Selective Inhibition of Human Brain Tumor Cells through
Transcription
Selective Inhibition of Human Brain Tumor Cells through
Angewandte Chemie DOI: 10.1002/anie.200905126 Bionanotechnology Selective Inhibition of Human Brain Tumor Cells through Multifunctional Quantum-Dot-Based siRNA Delivery** Jongjin Jung, Aniruddh Solanki, Kevin A. Memoli, Ken-ichiro Kamei, Hiyun Kim, Michael A. Drahl, Lawrence J. Williams, Hsian-Rong Tseng, and KiBum Lee* One of the most promising new chemotherapeutic strategies is the RNA interference (RNAi)-based approach, wherein small double-stranded RNA molecules can sequence-specifically inhibit the expression of targeted oncogenes.[1] In principle, this method has high specificity and broad applicability for chemotherapy. For example, the strategy of small interfering RNA (siRNA) enables manipulation of key oncogenes that modulate signaling pathways and thereby regulate the behavior of malignant tumor cells. To harness the full potential of this approach, the prime requirements are to deliver the siRNA molecules with high selectivity and efficiency into tumor cells and to monitor both siRNA delivery and the resulting knockdown effects at the single-cell level. Although several approaches such as polymer- and nanomaterial-based methods[2] have been attempted, limited success has been achieved for delivering siRNA into the target tumor cells. Moreover, these types of approaches mainly focus on the enhancement of transfection efficiency, knockdown of non-oncogenes (e.g. the gene coding for green fluorescent protein (GFP)), and the use of different nanomaterials such as quantum dots (QDs), iron oxide nanoparticles, and gold nanoparticles.[3, 4] Therefore, to narrow the gap between current nanomaterial-based siRNA delivery and chemotherapies, there is a clear need to develop methods for target-oriented delivery of siRNA,[5] for further monitoring the effects of siRNA-mediated target-gene silencing by means of molecular imaging probes,[4] and for investigating the corresponding up/down-regulation of signaling cascades.[6] Perhaps most importantly, to begin the development of the necessary treatment modalities, the strategies for nanomate- rial-based siRNA delivery must be demonstrated on oncogenes involved in cancer pathogenesis. Herein, we describe the synthesis and target-specific delivery of multifunctional siRNA–QD constructs for selectively inhibiting the expression of epidermal growth factor receptor variant III (EGFRvIII) in target human U87 glioblastoma cells, and subsequently monitoring the resulting down-regulated signaling pathway with high efficiency.[7] Glioblastoma multiforme (GBM) is the most malignant, invasive, and difficult-to-treat primary brain tumor. Successful treatment of GBM is rare with a mean survival of only 10– 12 months.[8] EGFRvIII, the key growth factor receptor triggering cancer cell proliferation in many cancer diseases such as brain tumors and breast cancer, is a constitutively active mutant of EGFR which is expressed in only human GBM and several other malignant cancers, but not in normal healthy cells (Figure 1 A).[9] We targeted EGFRvIII, since it is known that knockdown of this gene is one of the most effective ways to down-regulate the PI3K/Akt signaling pathway, a key signal cascade for cancer cell proliferation and apoptosis.[6, 10] Hence by targeting EGFRvIII, our siRNA delivery strategy based on multifunctional nanoparticles [*] J. Jung, A. Solanki, K. A. Memoli, H. Kim, M. A. Drahl, Prof. L. J. Williams, Prof. K.-B. Lee Department of Chemistry and Chemical Biology, Rutgers University Piscataway, NJ 08854 (USA) Fax: (+ 1) 732-445-5312 E-mail: [email protected] Homepage: http://rutchem.rutgers.edu/ ~ kbleeweb/ Dr. K. Kamei, Prof. H.-R. Tseng Department of Molecular and Medical Pharmacology University of California, Los Angeles Los Angeles, CA 90095 (USA) [**] We thank V. Starovoytov and Dr. Y. Horibe for helping us with TEM, and the New Jersey Biomaterial Center (Prof. Kohn) for allowing us to use the cell culture facilities. K.-B.L. acknowledges the NIH Directors’ Innovator Award (1DP20D006462-01) and is also grateful to the NJ commission on Spinal Cord grant (09-3085-SCR-E-0). Supporting information for this article is available on the WWW under http://dx.doi.org/10.1002/anie.200905126. Angew. Chem. Int. Ed. 2010, 49, 103 –107 Figure 1. A) Quantum dots as a multifunctional nanoplatform to deliver siRNA and to elucidate the EGFRvIII-knockdown effect of PI3K signaling pathway in U87-EGFRvIII B) Detailed structural information of multifunctional siRNA–QDs. C) Two different strategies for the siRNA–QD conjugate. L1 shows the linker for attaching siRNA to QDs through a disulfide linkage which is easily reduced within the cells to release the siRNA. L2 shows the linker for covalently conjugating siRNA to QDs which enables the tracking of siRNA–QDs within the cells. 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim 103 Communications could potentially minimize the side effects caused by conventional chemotherapies, specifically immune suppression, while significantly improving the efficacy of chemotherapy against GBM. We prepared two types of siRNA–QD conjugates, one for siRNA delivery and the other for siRNA tracking (Figure 1 B,C). Core–shell CdSe/CdS/ZnS QDs with a diameter of 7 nm were synthesized[11] and coated with either trioctylphosphine oxide (TOPO) or hexadecylamine (HDA). In order to make the QD constructs water-soluble and suitable for conjugating with siRNA, we displaced these hydrophobic ligands with a dihydrolipoic acid (DHLA) derivatized with an amine-terminated poly(ethylene glycol) (PEG) spacer. The expectation was that the dithiol moiety would provide strong coordination to the QD surface and increase stability in aqueous media, the PEG spacer would increase water solubility and reduce nonspecific binding, and the amine group would enable conjugation to the siRNA element.[12] Two bifunctional linkers were synthesized and evaluated for siRNA conjugation. The linker shown in L1, PTPPf [3-(2pyridyl)-dithiopropionic acid pentafluorophenyl ester], was designed to release siRNA upon entering the cell by cleavage of the disulfide linkage, through enzymatic reduction or ligand exchange (e.g. glutathione).[13] The linker in L2, MPPF (3-maleimidopropionic acid pentafluorophenyl ester), was designed to be more robust, thereby enabling evaluation of cellular uptake and localization of the siRNA construct within the cellular compartments.[14] Details of the synthesis, characterization and conjugation protocols are given in the Supporting Information. The final design component was functionalizing the construct for tumor-cell-selective transfection. For this purpose two functional peptides, thiol-modified RGD peptide and thiol-modified HIV-Tat derived peptide, were attached to the siRNA–QDs by the conjugation methods described above. Brain tumor cells (U87 and U87-EGFRvIII) overexpress the integrin receptor protein avb3, which strongly binds to the RGD binding domain.[15] RGD-functionalized siRNA–QDs selectively accumulate in brain tumor cells in vitro, and can be tracked by fluorescence microscopy.[16] In addition, the HIV-Tat peptide enables efficient transfection of siRNA–QDs in cells when it is directly attached to the QD surface.[17] The density of siRNA on the QDs and the ratio between siRNA strands and peptides were optimized for gene knockdown. It was found that the density of 10 siRNAs per nanoparticle and the ratio of 1:10 (siRNA for each peptide), which was in close agreement with literature values,[4] was optimal for knocking down the target genes (EGFP and EGFRvIII) overexpressed in our U87 cell lines. To optimize gene silencing with our siRNA–QD constructs and to assess the transfection efficiency and RNA interference (RNAi) activity, we examined the suppression of EGFP expressed in U87 cell lines that were genetically modified to express EGFP. The cytotoxicity of the constructs was determined by serial dilution studies. The range of concentration causing minimal or negligible cytotoxicity was identified, and subsequent experiments employed the concentrations within this range (see Figure S1 in the Supporting Information).[18] Importantly, the EGFP cell line has been 104 www.angewandte.org widely used to investigate siRNA-based silencing of EGFP, since the suppression of EGFP expression does not compromise cell viability. The transfection efficiency of three different kinds of constructs were evaluated; constructs modified with the RGD peptide only, those modified with the HIV-Tat peptide only, and those with both HIV-Tat and RGD peptide. Although the siRNA–QDs modified with only RGD showed considerable selective internalization within U87-EGFP cells, siRNA–QDs modified with a combination of RGD and HIVTat peptides (the ratio of siRNA/RGD/HIV-Tat being 1:10:10 per QD) showed maximum internalization within U87-EGFP cells, in close agreement with previous studies.[4] This optimal condition was used for subsequent siRNA–QD experiments. The U87-EGFP cell line was then treated with siRNA– QDs (siRNA/QDs = 0.12 mm :0.011 mm), modified with HIVTat [ 1.2 mm] and RGD [ 1.2 mm], and simultaneously imaged using fluorescence microscopy (Figure 2). Cationic lipids (X-tremeGENE, Roche) were used to further enhance cellular uptake and prevent degradation of the siRNA within the endosomal compartment of the cells. The siRNA–QDs showed significant internalization into the cells. Knockdown of the EGFP signal was observed after 48–72 h (Figure 2 B). Fluorescence intensity was influenced by other factors such as exposure time, media conditions, and cell shrinkage. To minimize the influence from these external factors, the Figure 2. Knockdown of EGFP in U87 cells using siRNA–QDs modified with RGD and HIV-Tat peptides. (Note that yellow arrows mark U87EGFP cells transfected with the siRNA–QDs and the blue arrows indicate PC-12 cells.) A) Control U87-EGFP cells without siRNA–QDs; phase-contrast image (A1) and the corresponding fluorescence image (A2). B) EGFP knockdown using multifunctional siRNA–QDs; B1) Phase-contrast image shows that the morphology of U87-EGFP cells has not changed relative to the control cells in (A). B2) Fluorescence image clearly shows the knockdown of EGFP in cells (marked by yellow arrows) which have internalized the siRNA–QDs (red) after 48 h. C) U87-EGFP control cells (without siRNA–QDs) and U87-EGFP cells transfected with siRNA–QDs were cocultured so as to investigate them under the same conditions; C1) Phase-contrast image clearly shows no difference in the morphology of the U87-EGFP control cells and the siRNA–QDs transfected cells. C2) Fluorescence image clearly shows the decrease in the EGFP signal in the U87-EGFP cells transfected with siRNA–QDs as compared to the surrounding U87-EGFP control cells. D) Phase-contrast image showing the target-oriented delivery of siRNA–QDs in cocultures of the malignant U87-EGFP cells, overexpressing the avb3 integrin receptors, and the less tumorigenic PC-12 cells (blue arrows) incubated with the siRNA–QDs. It can be clearly seen that most of the siRNA–QDs, owing to the presence of RGD and HIV-Tat peptides, were taken up by the U87-EGFP cells and not by the PC-12 cells. Scale bars: 100 mm. 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2010, 49, 103 –107 Angewandte Chemie control U87-EGFP cells (without siRNA) were trypsinized and co-cultured with U87-EGFP cells transfected with siRNA–QDs in the same well. The U87 cells containing siRNA–QDs were easily distinguishable from the control cells owing to the bright fluorescence of the QDs (Figure 2 C2). Cells with internalized siRNA–QDs showed considerable knockdown of the EGFP protein relative to the surrounding control U87-EGFP cells (Figure 2 C). To further demonstrate the target-specific delivery of the siRNA–QDs, we incubated the siRNA–QDs modified with Tat and RGD against EGFP in co-cultures of the U87-EGFP cell line with other less-tumorigenic cell lines, such as PC-12 and SK-N-BE(2)C (see Figure S2 in the Supporting Information), which have a considerably small number of integrin receptors.[19] The presence of RGD tripeptide molecules on the surface of the siRNA–QDs led to specific binding with integrin receptors overexpressed in the U87 cells, resulting in higher cellular uptake by the malignant U87 cells than by the less tumorigenic PC-12 cells as seen by the selective accumulation of the QDs within the U87-EGFP cells (Figure 2 D). These results confirmed our hypothesis that the target-specific delivery of the siRNA–QDs into brain cancer cells can be significantly enhanced by functionalizing the QDs with targeting moieties like RGD tripeptide. The intracellular delivery of the siRNA–QDs within the U87-EGFP cells was also confirmed by transmission electron microscopy (TEM), which clearly shows the presence of QDs in the cytoplasm of the cells (Figure 3 A). The knockdown efficiency of the siRNA–QDs was similar to or slightly better than that of the positive control consisting of U87-EGFP cells transfected with only siRNA using X-tremeGENE (see Figure S3 in the Supporting Information). This high transfection efficiency appears to result from synergistic effects of the two transfection peptides. Decrease in fluorescence intensities (EGFP signal, green fluorescence) within cells treated with the above-mentioned systems were then compared with the intensity of U87-EGFP without siRNA. As shown in (Figure 3 B), the decrease in fluorescence intensity of U87-EGFP incubated with siRNA–QDs and siRNA alone was comparable, but drastically lower than that observed for the control without siRNA. Cells containing siRNA–QDs show a weaker green fluorescence (EGFP signal) than the control. This data strongly suggests that siRNA–QDs can be used simultaneously as delivery and imaging probes. Having demonstrated the selective manipulation of the U87-EGFP cell line, we then focused on the knockdown of EGFRvIII with our siRNA–QD constructs. U87-EGFRvIII cells were genetically modified to overexpress EGFRvIII, a mutant-type epidermal growth factor receptor (EGFR) only expressed within cancer cells.[20] This cell type was incubated with our siRNA–QDs modified with Tat and RGD peptides and armed with EGFRvIII-targeting siRNA. The cells were simultaneously imaged for the internalization of siRNA–QDs using fluorescence microscopy. Significant cell death was observed in the wells loaded with siRNA–QDs against EGFRvIII after 48 h (Figure 4 A). Quantitative analysis revealed that the number of viable U87-EGFRvIII cells, as observed by fluorescence microscopy, decreased with increasing incubation time. Relative to the control (U87-EGFRvIII Angew. Chem. Int. Ed. 2010, 49, 103 –107 Figure 3. Knockdown efficiency of EGFP within U87-EGFP cells and internalization of multifunctional siRNA–QDs. A) TEM analysis of the internalization of the multifunctional siRNA–QDs into the U87-EGFP cells; A1) Presence of multifunctional siRNA–QDs (yellow arrows) within the cytoplasm and the endosome (scale bar: 5 mm). A2) Enlarged image showing individual siRNA–QDs within the cytoplasm (scale bar: 2.5 mm). B) The bar graph represents the knockdown of EGFP over 24 h, 48 h, and 96 h in U87-EGFP cells treated with siRNA [0.12 mm] only (dark gray), and siRNA–QD [siRNA:QD = 0.12 mm:0.011 mm] (light gray). The EGFP knockdown data was normalized with the expression levels of EGFP in the control U87-EGFP cells (black). without siRNA–QDs), there was a significant decrease in the number of viable cells, thus demonstrating the effectiveness of our nanoparticle-based siRNA delivery to knockdown the oncogene. The result was confirmed using the MTT assay which showed a decrease in the number of viable cells in the well incubated with siRNA–QDs against EGFRvIII (Figure 4 B). This assay further confirmed that the QDs themselves were noncytotoxic when used alone as they did not result in any appreciable cell death (see Figure S1 in the Supporting Information). The knockdown of EGFRvIII and the inhibition of the downstream proteins in the PI3K signaling pathway were confirmed using Western immunoblotting. The results (Figure 4 C) confirm a considerable decrease in the expression of EGFRvIII, and down-regulation of phospho-Akt and phospho-S6 relative to the control. Thus, these results demonstrate the specificity of the siRNA against EGFRvIII, the inherent noncytotoxicity of the QDs, and the facile evaluation and manipulation of cancer cell proliferation with multifunctional QD constructs. In summary, we have demonstrated an application of multifunctional siRNA–QDs focusing on targeted delivery, high transfection efficiency, and multimodal imaging/track- 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 105 Communications egies but also for dissecting signaling cascades triggered by inhibiting specific proteins. Collectively, our strategy for siRNA delivery based on multifunctional QDs has significant potential for simultaneous prognosis, diagnosis, and therapy. Received: September 13, 2009 Published online: November 30, 2009 . Keywords: antitumor agents · gene knockdown · nanoparticles · nonviral siRNA delivery · target-specific delivery Figure 4. Knockdown of EGFRvIII in U87-EGFRvIII using multifunctional siRNA–QDs. A) Phase-contrast images showing the internalization of siRNA–QDs into the U87-EGFRvIII cells. A1) Morphology of U87-EGFRvIII cells before incubation with siRNA–QDs on Day 0. A2) U87-EGFRvIII cells after incubation with siRNA–QDs (red) on Day 0. A3) Morphology of U87-EGFRvIII cells 48 h after incubation with siRNA–QDs. Note that effect of the EGFRvIII knockdown by the siRNA–QDs can be clearly seen as the cells have clearly shrunk (yellow arrows) and appear to have collapsed (cf. Day 0), marking the onset of apoptosis; scale bar: 100 mm. B) Cell viability assay using MTT assay. B1) Optical image of cell viability (MTT) assay in a well plate. Dark blue color indicates a high number of viable cells and pale blue indicates a low population of viable cells. B2) MTT-assayed wells were quantified with UV absorbance and the data was converted to cell viability data. Untreated control C1 and C2 represent control cell population and viable cell population, respectively, in the presence of a cationic lipid based transfection reagent. siRNA–QD-transfected cells in experiment E1 and siRNA treated cells in E2 show low numbers of the viable cells due to knockdown of EGFRvIII gene. C) Western immunoblotting shows the silencing effect of the EGFRvIII gene. Protein expression level of EGFRvIII is dramatically decreased, and phosphorylation levels of key proteins in PI3K signaling pathway are reduced significantly. The upstream protein (AKT) and the downstream protein (S6), which play an important role in cell proliferation, are selected to investigate the gene-knockdown effect on the PI3K signaling pathway. ing. Our siRNA–QDs could be used for the development of novel chemotherapies and diagnostics relevant to brain cancer research. These novel methods and applications complement recent advances in nanomaterial-based siRNA delivery, nanomaterial-based molecular imaging, and siRNAbased chemotherapeutic strategies reported recently. While the ability to functionalize as well as control the surface of quantum dots with specific linkers and multifunctional molecules (siRNA and peptides) is critical for nanoparticlebased drug delivery, this method could also provide highly useful information regarding biological surface chemistry of nanomaterials. In addition, the application of multifunctional siRNA–QDs to modulate the key cancer signaling pathways is important not only for selective chemotherapeutic strat- 106 www.angewandte.org [1] a) Z. Medarova, W. Pham, C. Farrar, V. Petkova, A. Moore, Nat. Med. 2007, 13, 372 – 377; b) D. M. Dykxhoorn, D. Palliser, J. Lieberman, Gene Ther. 2006, 13, 541 – 552. [2] a) L. Wasungu, D. Hoekstra, J. Controlled Release 2006, 116, 255 – 264; b) J. H. Jeong, H. Mok, Y. K. Oh, T. G. Park, Bioconjugate Chem. 2009, 20, 5 – 14; c) M. L. Patil, M. Zhang, S. Betigeri, O. Taratula, H. He, T. Minko, Bioconjugate Chem. 2008, 19, 1396 – 1403; d) A. Zintchenko, A. Philipp, A. Dehshahri, E. Wagner, Bioconjugate Chem. 2008, 19, 1448 – 1455. [3] a) J. H. Lee, K. Lee, S. H. Moon, Y. Lee, T. G. Park, J. Cheon, Angew. Chem. 2009, 121, 4238 – 4243; Angew. Chem. Int. Ed. 2009, 48, 4174 – 4179; b) L. F. Qi, X. H. Gao, Acs Nano 2008, 2, 1403 – 1410; c) J. S. Lee, J. J. Green, K. T. Love, J. Sunshine, R. Langer, D. G. Anderson, Nano Lett. 2009, 9, 2402 – 2406; d) M. V. Yezhelyev, L. F. Qi, R. M. ORegan, S. Nie, X. H. Gao, J. Am. Chem. Soc. 2008, 130, 9006 – 9012; e) D. A. Giljohann, D. S. Seferos, A. E. Prigodich, P. C. Patel, C. A. Mirkin, J. Am. Chem. Soc. 2009, 131, 2072. [4] A. M. Derfus, A. A. Chen, D. H. Min, E. Ruoslahti, S. N. Bhatia, Bioconjugate Chem. 2007, 18, 1391 – 1396. [5] a) R. Kikkeri, B. Lepenies, A. Adibekian, P. Laurino, P. H. Seeberger, J. Am. Chem. Soc. 2009, 131, 2110; b) A. A. Bhirde, V. Patel, J. Gavard, G. F. Zhang, A. A. Sousa, A. Masedunskas, R. D. Leapman, R. Weigert, J. S. Gutkind, J. F. Rusling, Acs Nano 2009, 3, 307 – 316. [6] Q. W. Fan, W. A. Weiss, Oncogene 2005, 24, 829 – 837. [7] E. Arwert, S. Hingtgen, J. L. Figueiredo, H. Bergquist, U. Mahmood, R. Weissleder, K. Shah, Cancer Res. 2007, 67, 7335 – 7342. [8] a) A. Veeravagu, Z. F. Liu, G. Niu, K. Chen, B. Jia, W. B. Cai, C. J. Jin, A. R. Hsu, A. J. Connolly, V. Tse, F. Wang, X. Y. Chen, Clin. Cancer Res. 2008, 14, 7330 – 7339; b) M. S. Lesniak, H. Brem, Nat. Rev. Drug Discovery 2004, 3, 499 – 508. [9] a) A. J. Ekstrand, N. Sugawa, C. D. James, V. P. Collins, Proc. Natl. Acad. Sci. USA 1992, 89, 4309 – 4313; b) P. A. Humphrey, A. J. Wong, B. Vogelstein, M. R. Zalutsky, G. N. Fuller, G. E. Archer, H. S. Friedman, M. M. Kwatra, S. H. Bigner, D. D. Bigner, Proc. Natl. Acad. Sci. USA 1990, 87, 4207 – 4211. [10] a) I. Vivanco, C. L. Sawyers, Nat. Rev. Cancer 2002, 2, 489 – 501; b) F. Yamoutpour, V. Bodernpudi, S. E. Park, W. H. Pan, M. J. Mauzy, R. A. Kratzke, A. Dudek, D. A. Potter, R. A. Woo, D. M. ORourke, D. J. Tindall, F. Farassati, Mol. Cancer Ther. 2008, 7, 3586 – 3597. [11] a) J. J. Li, Y. A. Wang, W. Z. Guo, J. C. Keay, T. D. Mishima, M. B. Johnson, X. G. Peng, J. Am. Chem. Soc. 2003, 125, 12567 – 12575; b) R. G. Xie, U. Kolb, J. X. Li, T. Basche, A. Mews, J. Am. Chem. Soc. 2005, 127, 7480 – 7488. [12] a) W. Liu, M. Howarth, A. B. Greytak, Y. Zheng, D. G. Nocera, A. Y. Ting, M. G. Bawendi, J. Am. Chem. Soc. 2008, 130, 1274 – 1284; b) K. Susumu, H. T. Uyeda, I. L. Medintz, T. Pons, J. B. Delehanty, H. Mattoussi, J. Am. Chem. Soc. 2007, 129, 13987 – 13996. [13] a) S. Kida, M. Maeda, K. Hojo, Y. Eto, S. Nakagawa, K. Kawasaki, Chem. Pharm. Bull. 2007, 55, 685 – 687; b) A. S. M. 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Angew. Chem. Int. Ed. 2010, 49, 103 –107 Angewandte Chemie [14] [15] [16] [17] Kamruzzahan, A. Ebner, L. Wildling, F. Kienberger, C. K. Riener, C. D. Hahn, P. D. Pollheimer, P. Winklehner, M. Holzl, B. Lackner, D. M. Schorkl, P. Hinterdorfer, H. J. Gruber, Bioconjugate Chem. 2006, 17, 1473 – 1481; c) S. Bauhuber, C. Hozsa, M. Breunig, A. Gpferich, Adv. Mater. 2009, 21, 3286 – 3306. Y. L. Chiu, A. Ali, C. Y. Chu, H. Cao, T. M. Rana, Chem. Biol. 2004, 11, 1165 – 1175. J. Xie, K. Chen, H. Y. Lee, C. J. Xu, A. R. Hsu, S. Peng, X. Y. Chen, S. H. Sun, J. Am. Chem. Soc. 2008, 130, 7542. W. B. Cai, D. W. Shin, K. Chen, O. Gheysens, Q. Z. Cao, S. X. Wang, S. S. Gambhir, X. Y. Chen, Nano Lett. 2006, 6, 669 – 676. a) C. C. Berry, Nanomedicine 2008, 3, 357 – 365; b) G. Ruan, A. Agrawal, A. I. Marcus, S. Nie, J. Am. Chem. Soc. 2007, 129, 14759 – 14766. Angew. Chem. Int. Ed. 2010, 49, 103 –107 [18] a) A. M. Derfus, W. C. W. Chan, S. N. Bhatia, Nano Lett. 2004, 4, 11 – 18; b) C. Kirchner, T. Liedl, S. Kudera, T. Pellegrino, A. M. Javier, H. E. Gaub, S. Stolzle, N. Fertig, W. J. Parak, Nano Lett. 2005, 5, 331 – 338. [19] a) A. Meyer, C. M. van Golen, B. Kim, K. L. van Golen, E. L. Feldman, Neoplasia 2004, 6, 332 – 342; b) L. Martiniova, E. W. Lai, A. G. Elkahloun, M. Abu-Asab, A. Wickremasinghe, D. C. Solis, S. M. Perera, T. T. Huynh, I. A. Lubensky, A. S. Tischler, R. Kvetnansky, S. Alesci, J. C. Morris, K. Pacak, Clin. Exp. Metastasis 2009, 26, 239 – 250; c) Z. B. Li, W. B. Cai, Q. Z. Cao, K. Chen, Z. H. Wu, L. N. He, X. Y. Chen, J. Nucl. Med. 2007, 48, 1162 – 1171. [20] M. Y. Wang, K. V. Lu, S. J. Zhu, E. Q. Dia, I. Vivanco, G. M. Shackleford, W. K. Cavenee, I. K. Mellinghoff, T. F. Cloughesy, C. L. Sawyers, P. S. Mischel, Cancer Res. 2006, 66, 7864 – 7869. 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org 107 PAPER www.rsc.org/loc | Lab on a Chip An integrated microfluidic culture device for quantitative analysis of human embryonic stem cells† Ken-ichiro Kamei,‡ab Shuling Guo,‡cd Zeta Tak For Yu,‡abe Hiroko Takahashi,ab Eric Gschweng,cd Carol Suh,ab Xiaopu Wang,ab Jinghua Tang,d Jami McLaughlin,c Owen N. Witte,*acdg Ki-Bum Lee*abf and Hsian-Rong Tseng*ab Received 29th May 2008, Accepted 14th October 2008 First published as an Advance Article on the web 20th November 2008 DOI: 10.1039/b809105f We have successfully designed and fabricated an integrated microfluidic platform, the hESC-mChip, which is capable of reproducible and quantitative culture and analysis of individual hESC colonies in a semi-automated fashion. In this device, a serpentine microchannel allows pre-screening of dissociated hESC clusters, and six individually addressable cell culture chambers enable parallel hESC culture, as well as multiparameter analyses in sequence. In order to quantitatively monitor hESC proliferation and pluripotency status in real time, knock-in hESC lines with EGFP driven by the endogenous OCT4 promoter were constructed. On-chip immunoassays of several pluripotency markers were carried out to confirm that the hESC colonies maintained their pluripotency. For the first time, our studies demonstrated well characterized hESC culture and analysis in a microfluidic setting, as well as a proofof-concept demonstration of parallel/multiparameter/real-time/automated examination of self-renewal and differentiation in the same device. Introduction Human embryonic stem cells (hESCs),1–3 derived from the inner cell mass of blastocyst-stage embryos, hold great potential for the treatment of many devastating diseases and injuries. This is mainly due to two distinct properties: (i) they can self-renew indefinitely and (ii) they can potentially generate all cell types in the human body. Intrinsic regulators (e.g., growth factors and signaling molecules) and cellular microenvironments (e.g., extracellular matrices, ECMs) play critical roles in the regulation of self-renewal and differentiation of hESCs. Conventionally, hESCs are passaged in clusters (containing approximately a Department of Molecular & Medical Pharmacology, University of California, Los Angeles, CA 90095, USA. E-mail: hrtseng@mednet. ucla.edu b Crump Institute for Molecular Imaging, David Geffen School of Medicine, University of California, Los Angeles, CA 90095, USA c Department of Microbiology, Immunology and Molecular Genetics, University of California, Los Angeles, CA 90095, USA d The Eli and Edythe Broad Center of Regenerative Medicine and Stem Cell Research, University of California, Los Angeles, CA 90095, USA. E-mail: [email protected] e Department of Mechanical and Aerospace Engineering, University of California, Los Angeles, CA 90095, USA f Department of Chemistry & Chemical Biology, Institute for Advanced Materials, Devices and Nanotechnology, The Rutgers Stem Cell Research Center, Rutgers, The State University of New Jersey, Piscataway, NJ 08854, USA. E-mail: [email protected] g The Howard Hughes Medical Institute, David Geffen School of Medicine, University of California, Los Angeles, CA 90095, USA † Electronic supplementary information (ESI) available: Fabrication of hESC-mChips, generation of genetically modified hESCs (i.e., HSF1-LG, HSF1-OCT4-EGFP and H1-OCT4-EGFP), microscopy settings and induction of hESC differentiation. See DOI: 10.1039/b809105f ‡ These three authors contributed equally to this work. This journal is ª The Royal Society of Chemistry 2009 20–200 cells) using well plates or culture dishes. Growth-arrested mouse embryonic fibroblast (mEF) feeder layers are co-cultured in serum replacement-containing medium to supply the essential intrinsic regulators and environmental cues. However, there have been concerns associated with xenogenic contamination that would restrict potential therapeutic applications of hESCs in clinical settings.4,5 In order to harness the unique potential of hESCs and to improve self-renewal6–12 and controlled differentiation of hESCs,13–16 systematic approaches have been adopted to screen a broad range of serum- and feeder-free culture conditions to obtain a better understanding of the roles of intrinsic regulators and cellular microenvironments. The cost to perform these screening experiments is high, since they consume a considerable quantity of hESCs, ECM materials and culture media containing expensive growth factors. There is a clear need for developing a miniaturized platform on which to carry out large-scale screening in a cost-efficient fashion. There is a growing interest to develop microfluidics-based technologies17 for performing cell culture and analysis. Microfluidic systems offer intrinsic advantages over conventional macroscopic culture such as reduced sample/reagent consumption and precise control over the delivery of culture fluids and soluble factors. A continuous-flow microfluidic system composed of the simplest device configuration (i.e., individual microchannels and the respective inlets/outlets) has been utilized to implement miniaturized cell culture and analysis.18 In this case, bovine adrenal capillary endothelial cells were seeded in protein-coated microchannels, where culture media and assay reagents were introduced and withdrawn from the microchannels through inlets and outlets, respectively. Several challenges remain, however, to explore these continuous-flow cell culture/ assay chips for systematic screenings where combinations of multiple parameters should be tested to obtain the desired Lab Chip, 2009, 9, 555–563 | 555 outcomes. For example, when many cell culture conditions are screened in a microchannel network, it is inevitable that the individual conditions would be cross-contaminated through diffusion. Moreover, this multiparameter screening necessitates a delicate microfluidic delivery/mixing system for handling small amounts of culture components coordinated by an automated operation system. To overcome these challenges, different miniaturized functional modules, including isolation valves and mechanical pumps, have been developed to prevent cross contamination and to attain precise fluidic delivery and mixing. Most importantly, these miniaturized valves and pumps can be digitally controlled, thus allowing automated cell culture in a microfluidic chip.19 Among the exciting automated microfluidic systems, the poly(dimethylsiloxane) (PDMS)-based integrated microfluidic system represents a large-scale architecture of microchannel networks that enables the execution of sequential and parallel processes in individual devices.20 Particularly, the biocompatible and gas-permeable properties of PDMS matrices help to retain proper physiological conditions for a wide range of mammalian cells suitable for different screening applications. The cooperation of integrated hydraulic valves confines distinct regions for testing specific screening combinations/conditions without the concern of cross contamination,21,22 and a peristaltic pump (composed of three consecutive isolation valves) is capable of delivering, metering, and mixing of nanoliter (nL)-level fluids with great precision.22 Over the past seven years, different PDMS-based integrated microfluidic devices have been developed for complicated chemical23,24 and biological operations,25–27 including recent demonstrations on culturing human mesenchymal stem cells.28 Obviously, the characteristics of the PDMS-based integrated microfluidic system meet the needs of conducting systematic screenings of optimal hESC culture conditions. Although there are examples of the culture of human neural stem cells29,30 and mouse embryonic stem cells31 in different microfluidic systems, there are few reports to demonstrate the culture and manipulation of hESCs in a microfluidic platform.32–35 Here, we demonstrate an integrated microfluidic platform (hESC-microChip, hESC-mChip), which allows reproducible and quantitative culture and analysis of individual hESC colonies in a semi-automated fashion. Initially, several challenges were envisioned to conduct this study in a hESC-mChip. For example, hESCs are extremely sensitive to changes of intrinsic regulators, cellular microenvironments and ambient pressure/temperature. The effects of culturing hESCs into a hESC-mChip on hESCs should be addressed in these contexts. Further, hESCs have to be passaged in clusters and co-cultured in the presence of growtharrested mEF feeder layers. Experience in handling hESC clusters in the chip and co-culturing of hESC clusters with the adherent mEF cells should be acquired. Moreover, to confirm that the chip-cultured hESC colonies maintain their pluripotency over a certain culture period, immunoassays for a number of pluripotency markers have to be carried out in sequence. Each immunoassay for chip-based operation will be optimized and some of them will be compiled in sequences. The goal of our study was not meant to unveil novel insights in hESC biology or develop a new type of microfluidic technology, but to acquire solid experience and practical knowledge of performing 556 | Lab Chip, 2009, 9, 555–563 microfluidic hESC culture, which will constitute a useful foundation for exploring further application of microfluidic platforms in hESC research. Experimental hESC culture in a hESC-mChip All hESC research described here was approved by the UCLA Embryonic Stem Cell Research Oversight Committee. A newly fabricated hESC-mChip was sterilized under UV light for 15 min prior to on-chip hESC culture. Based on a two-layer coating approach, a bovine fibronectin solution (FN, 1 mg mL1 in PBS, Sigma) and a gelatin solution (0.2% in PBS) were sequentially introduced into the hESC-mChip from ‘‘Inlet 2’’ using Teflon tubing (Fig. S1†). g-Irradiated mEFs (1 107 cells mL1) were loaded into the cell culture chambers from ‘‘Inlet 3’’. mEFs were cultured for 12 hr in a humidified incubator (37 C, 5% CO2, Thermo Fisher Scientific) before loading cells. hESCs cultured in a 6-well plate were passaged with 1 mg mL1 of collagenase IV in DMEM/F12 (see the ESI†). The freshly dissociated hESC clusters were introduced into the cell culture chambers through ‘‘Inlet 1’’ connected to a serpentine microchannel, where every hESC colony was visually inspected (Fig. 1c). Gravity flow36 was adopted in order to introduce hESC clusters into each cell culture chamber. To ensure the quality and uniformity of hESC colonies in our studies, only hESC clusters with desired sizes (100 20 mm) and disc-shaped morphologies were selected for seeding. In general, four to six hESC colonies were accommodated in each cell culture chamber. The locations of individual hESC colonies were registered according to the ruler, allowing continuous fate mapping by an inverted microscope. The hESC-mChip-based hESC culture was carried out in a humidified incubator (37 C, 5% CO2). By programming the cooperation of isolation valves and peristaltic pumps, media stored in Teflon tubing was introduced into each cell culture chamber every 12 hr. Immunocytochemistry and histology hESC colonies were fixed by introducing paraformaldehyde (4%, Electron Microscope Science) into the cell culture chambers in the hESC-mChip. After permeabilization with Triton X-100 (0.5%, Fluka) in PBS for 30 min, a blocking solution containing normal goat serum (5%, Vector Laboratory), normal donkey serum (5%, Jackson Laboratory), bovine serum albumin (3%, Fraction V, Sigma) and N-dodecyl-b-D-maltoside (0.1%, Pierce)37 was loaded into the device from ‘‘Inlet 1’’, and the device was incubated at room temperature for 1 hr. After rinsing with PBS containing 0.1% Tween 20 (PBS-T), the hESC colonies were incubated with human specific antibodies for OCT4 (2 mg mL1, mouse monoclonal IgG, Santa Cruz Biotechnology), NANOG (2 mg mL1, rabbit polyclonal IgG, Abcam), SSEA1 (2 mg mL1, mouse monoclonal IgM, Santa Cruz Biotechnology), SSEA4 (2 mg mL1, mouse monoclonal IgG, Santa Cruz Biotechnology), TRA-1-60 (2 mg mL1, mouse monoclonal IgM, Santa Cruz Biotechnology) or TRA-1-81 (2 mg mL1, mouse monoclonal IgM, Santa Cruz Biotechnology) for 24 hr at 4 C. After rinsing the cell culture chambers with a blocking solution, the respective secondary antibody: Alexa Fluor 514-conjugated goat antimouse IgG (H + L) (10 mg mL1, Invitrogen), R-Phycoerythrin This journal is ª The Royal Society of Chemistry 2009 (R-PE)-conjugated goat anti-mouse IgM (10 mg mL1, BD Pharmigen), Cy5-conjugated goat anti-rabbit IgG (H + L) (7.5 mg mL1, Jackson ImmunoResearch), or Alexa Fluor 750conjugated goat anti-mouse IgG (H + L) (20 mg mL1, Invitrogen) was loaded into the cell culture chambers to detect the bound primary antibodies. After incubating at room temperature for 1 hr, the chambers were rinsed with PBS-T. Finally, 10 mg mL1 of DAPI solution was loaded for nuclear staining. For alkaline phosphatase (AP) staining, the hESC colonies were fixed with paraformaldehyde (4%) for 30 min at room temperature. After fixation, a freshly prepared AP staining solution (1 mg mL1 Fast Red TR salt in water with 0.01% AS-MX alkaline phosphate solution, Sigma) was loaded into the cell culture chambers and incubated for 30 min in the dark. Fluorescence and phase contrast images were taken with an inverted microscope (TE2000S, Nikon), and quantitatively analyzed with MetaMorph software (version 7.1.3.0; Molecular Devices) (Fig. S2†). Results Design and operation of hESC-mChips Fig. 1 Design of the hESC-mChip. (a) Schematic illustration of a hESCmChip capable of semi-automated operation for hESC culture and analysis. The functions of different hydraulic valves are illustrated by their colors: Red for pneumatic valve operation and yellow for fluidic delivery and metering. The 6 1 array of cell culture chambers (with dimensions of 3000 mm (l) 500 mm (w) 100 mm (h) and total volume of 150 nl) are numbered 1 to 6. Each cell culture chamber is separated by hydraulic valves to achieve individual addressability. There are four inlets and two outlet channels in each device, providing accesses to hESC colonies, culture media and immunostaining reagents. (b) Optical micrograph of the actual device. Food dyes were introduced into the various microchannels to help visualize the functional components of the hESC-mChip: Red and yellow as illustrated in (a); blue indicates the fluidic channels. A ruler was fabricated alongside of each cell culture chamber to serve as a landmark that directs continuous fate mapping of individual hESC colonies by an optical microscope. For hESC culture in the hESC-mChip, freshly prepared hESC clusters were introduced into cell culture chambers through the inlet connected to a serpentine microchannel as shown in (c), where every hESC cluster was visually inspected (i,iv). To ensure the uniformity of hESC clusters used in our studies, only hESC clusters with the desired size and morphology were introduced into cell culture chambers (i–iii). Undesirable hESC clusters were removed as waste (iv–vi). This journal is ª The Royal Society of Chemistry 2009 A typical hESC-mChip (Fig. 1a and b) is composed of a 6 1 array of cell culture chambers (with dimensions of 3000 mm (l) 500 mm (w) 100 mm (h) and total volume of 150 nL) for accommodating individual hESC colonies. A ruler was fabricated alongside each cell culture chamber as a landmark, so that individual hESC colonies were registered for continuous fate mapping using an inverted microscope. There are four inlets and two outlet channels in each device, providing accesses to culture media and immunostaining reagents. For hESC culture in the hESC-mChip, freshly dissociated hESC clusters (obtained by digesting conventionally cultured hESC colonies with collagenase IV) were introduced into cell culture chambers through the inlet via a serpentine microchannel where every hESC cluster was visually inspected (Fig. 1c). To ensure the uniformity of hESC clusters in our studies, only disc-shaped clusters with diameters within 100 20 mm were introduced to the cell culture chamber. In general, four to six hESC clusters were selected and seeded per chamber. To allow parallel examination of multiple variables over time, six pairs of hydraulic valves (Fig. 1a and b) conferred individual addressability to the six cell culture chambers in the device. A laptop computer was utilized to control the valves and pumps to achieve automated operation of the hESC-mChip. To ensure general applicability of the hESC-mChips, we conducted our studies using a collection of hESC lines, including two parental hESC lines (i.e., HSF1 and H1) and three genetically modified hESC lines–(i) HSF1-LG which expresses firefly luciferase and enhanced green fluorescent protein (EGFP) as a fusion protein driven by the ubiquitin promoter, (ii) HSF1-OCT4EGFP and (iii) H1-OCT4-EGFP which express EGFP under the endogenous OCT4 promoter. In our proof-of-concept studies, hESC-mChip-based culture experiments were carried out in the presence of g-irradiated mEFs, using serum replacement-containing media with either 10 or 100 ng mL1 of bFGF. The g-irradiated mEFs were seeded in the protein-coated cell culture chambers for 12 hr prior to the introduction of the dissociated hESC clusters. Throughout the experiment, hESC-mChips Lab Chip, 2009, 9, 555–563 | 557 were stored in a humidified incubator (5% CO2, 37 C). The gas-permeability of PDMS allowed rapid gas exchange between the atmosphere around the hESC-mChips and the media in the cell culture chambers. The results revealed that medium with a concentration of 100 ng mL1 bFGF gave better reproducibility of hESC self renewal in the device. Due to the higher surface area-to-volume ratio of the microfluidic environment, a significant amount of bFGF was absorbed on the microchannels surfaces. The use of a higher concentration of bFGF was sufficient to maintain the chip-cultured hESC colonies. Since the hESC-mChip consumes only 150 nL of medium in each culture chamber, the use of 100 ng mL1 bFGF has very limited impact on experimental cost. Optimization of hESC culture conditions Since this digitally controlled hESC-mChip is capable of smallscale screening, we were able to utilize these devices to progressively define an optimal surface coating protocol and a cell feeding schedule which are optimized for the hESC colonies. Initially, several protein coating combinations and approaches were examined in the device in search of a recipe (Fig. S3†) which led to efficient plating of the g-irradiated mEF layer and reproducible self-renewal of hESC colonies. We identified a layer-by-layer coating method: a layer of fibronectin (FN) was first coated onto the PDMS surfaces (by introducing 1 mg mL1 FN solution into the cell culture chambers and incubated at 37 C for 30 min), followed by sequential deposition of a gelatin layer (0.2% gelatin solution at 37 C for 15 hr). This coating method resulted in a uniform and long-lasting FN/gelatin layer on the PDMS surface for maintaining hESC colonies. Using a hESC-mChip with six FN/gelatin-coated cell culture chambers, we then carried out a parallel examination of different cell feeding schedules. By programming the cooperation of hydraulic valves and peristaltic pumps, the medium stored in Teflon tubing was periodically introduced into each cell culture chamber at different feeding intervals (i.e., 3, 6, 12, 18, 24 and 36 hr). As a result of monitoring morphology and survival rate of hESC colonies, we identified a 12-hr feeding cycle which allowed the reproducible self renewal of hESC colonies in our hESC-mChip for 6 days. By using the optimized hESC culture condition (i.e., in the presence of serum replacement-containing medium, g-irradiated mEFs and FN/gelatin coated cell culture chambers, as well as using a cell feeding cycle of 12 hr), we were able to culture HSF1, H1, HSF1-LG, HSF1-OCT4-EGFP and H1OCT4-EGFP in the hESC-mChips for 6 days (Fig. S4†). In addition, HSF1 cells could be cultured in our mChips up to 12 days for the longest culturing periods (Fig. S5†). By chance, a single hESC (HSF1) colony was cultured in a cell culture chamber (Fig. S6†). There was no significant difference observed in contrast with the multi-colonies culture. Chip-based immunocytochemistry to confirm hESC pluripotency To confirm pluripotency of hESC-mChip-cultured hESCs, immunocytochemistry for a number of pluripotency markers, including alkaline phosphatase (AP), stage-specific embryonic antigen 4 (SSEA4), OCT4 (also known as POU5F1), NANOG, tumor-related antigen (TRA)-1-60 and TRA-1-81, was carried 558 | Lab Chip, 2009, 9, 555–563 out in the same device. The digitally controlled interface allowed automated execution (Supplementary Methods) of the immunostaining processes, where multiple reagents, including paraformaldehyde (4%) for fixation, Triton X-100 (0.5%) in PBS for permeabilization of the cell membrane, and antibodies for fluorescent immunocytochemical analyses, were introduced into the cell culture chambers in sequence. It is noteworthy that mixtures of four different pluripotency markers could be introduced in individual culture chambers, allowing four fluorescence immunocytochemical analyses at the same time. Finally, the resulting hESC-mChip was mounted on either a fluorescence microscope or a confocal microscope to collect immunofluorescence micrographs. Fig. 2a and b show immunofluorescence images of hESC-mChip-cultured HSF1 and H1 colonies, respectively. These cells retained characteristic hESC morphology, and strong fluorescence signals of pluripotency markers, indicating that they maintained their stemness over the six-day culture period. Threedimensional (3D) confocal micrographs of hESC-mChipcultured hESCs (Fig. 2c–e, and the visualization of its 3-D structure in a movie clip in Supplemental Information) revealed 3D structures of the hESC colonies, and merged 3D confocal micrographs indicate the co-localization of different pluripotency markers. Quantification of hESC growth in the hESC-mChip To monitor hESCs in vitro and in vivo, HSF1-LG cells were generated by infecting HSF1 cells with lentivirus containing a mutated thermostable firefly luciferase (mtfl)38 and EGFP as a fusion protein (LG) driven by a ubiquitin promoter (Fig. S7a– h†). The EGFP signal allows the quantification of cell growth in real time.39 To test this idea, freshly dissociated HSF1-LG clusters were cultured in the hESC-mChip for 6 days,40 and their EGFP signals were measured every other day (Fig. 3a). In parallel, these dissociated clusters were cultured in conventional culture dishes under similar conditions. The growth rates of hESCs were quantified by measuring the increased surface area or integrated EGFP intensities of individual hESC colonies at different time points. As shown in Fig. 3b and c, both quantification approaches gave similar results. Although inhibition of cell proliferation has been reported in other microfluidic cell culture settings,41 possibly due to the constrained accumulation of soluble factors in the diffusion dominant microfluidic environment, the growth rates of hESC-mChip cultured colonies were not significantly different from those observed for hESCs in conventional dishes (p ¼ 0.21). Multiparameter monitoring of hESC pluripotency status In order to monitor the pluripotent status of hESCs in realtime, we constructed OCT4-EGFP knock-in reporter lines in HSF1 and H1 cells (i.e., HSF1-OCT4-EGFP and H1-OCT4EGFP) (Fig. S8a–d†). In both cases, the linearized OCT4EGFP knock-in construct42 was introduced into hESCs via Nucleofector (Amaxa Biosystems). These genetically modified hESCs could be passaged as their parental HSF1 or H1 cells. To ensure that EGFP expression in these hESCs faithfully represents pluripotency, both HSF1-OCT4-EGFP and H1OCT4-EGFP were induced to differentiate in the presence of This journal is ª The Royal Society of Chemistry 2009 Fig. 2 On-chip immunocytochemistry to confirm hESC pluripotency. Bright-field and fluorescence micrographs of hESC-mChip-cultured hESC colonies stained with a collection of pluripotency markers: (a) Three HSF1 colonies were stained by DAPI and alkaline phosphatase (AP), as well as immunostained for OCT4, NANOG, TRA-1-60 and TRA-1-81. (b) Two H1 colonies were stained with DAPI, SSEA4, NANOG, TRA-1-60 and TRA-1-81. The characteristic morphologies and strong fluorescent signals of pluripotency markers indicate that the hESCs cultured in hESC-mChips retained their pluripotency over the six-day culture period. (c–f) Three-dimensional (3D) confocal micrographs of a genetically modified hESC colony (HSF1-LG). (c) DAPI nuclear staining, (d) EGFP expression, (e) OCT4 immunostaining and (f) the merged image. These images revealed information on the 3D structure of the hESC colonies. fetal bovine serum (FBS, 15%) and the absence of mEFs. After about 10 days in culture, over 90% of the cells lost EGFP expression, correlating with their differentiated morphology (Fig. S8e†). Additionally, if the EGFP signal truly correlates with the endogenous OCT4 expression, this marker could be used to rescue pluripotent cells from a differentiated population. To show this, OCT-EGFP-knock-in cells were differentiated as embryoid bodies in serum containing medium. After 21 days, the EGFP positive population (approximately 3%) was This journal is ª The Royal Society of Chemistry 2009 sorted from the non-expressing cells (Fig. S8g†) and re-plated into conventional culture conditions. Indeed, these cells re-grew into typical ES colonies and maintained pluripotency markers (Fig. S8h–j†). Either HSF1-OCT4-EGFP or H1-OCT4-EGFP hESCs were utilized for the demonstration of parallel examinations of controlled differentiation and proliferation in individual hESCmChips. In a given study, differentiation of hESCs was carried out in cell culture chambers No. 1, 3 and 5, where only a layer of Lab Chip, 2009, 9, 555–563 | 559 Fig. 3 Real-time quantitative monitoring of growth of hESC-mChip-cultured hESC colonies. (a) Fate mapping of hESCs cultured in a hESC-mChip with bright-field microscopy. As we show in Fig. 2, the hESC colonies still had pluripotency, even in hESC colonies attached onto the channel. And, since there are PDMS walls in the EGFP images of HSF1-LG at day 6, there is no EGFP signal from those areas. In addition, since EGFP expression in HSF1-LG is under the regulation of a ubiquitin promoter which constitutively active in any kinds of cells, EGFP intensity doesn’t reflect their pluripotency. (b) Quantitative comparison of growth rate of the size of hESC colonies in conventional culture dishes and hESC-mChips. (c) Quantitative comparison of growth rate of EGFP intensity of hESC colonies in conventional culture dishes and hESC-mChips. Each bar represents the standard deviation (n > 7). FN was coated and no feeder cells were applied. In parallel, proliferation of hESCs was carried out in culture chambers No. 2, 4 and 6, where FN/gelatin coating was applied and g-irradiated mEFs were cultured. After 24 hr, the genetically modified hESCs clusters were introduced into the 6 cell culture chambers. After 3 hr, differentiation medium (containing 5 mM retinoic acid (RA) and 15% FBS) and self-renewal medium were separately introduced into the respective sets of chambers with a 12-hr feeding schedule. The EGFP signals in the differentiating or selfrenewing cells were monitored every other day to record the status of their pluripotency. hESC colonies in differentiation medium gradually lost their compact morphologies and spread 560 | Lab Chip, 2009, 9, 555–563 out. Concurrently, the EGFP signals started to diminish after 2 days, whereas the hESC colonies in the self-renewal medium grew larger accompanied by increased EGFP signal (Fig. 4a). After 4 days of culture in a hESC-mChip, immunocytochemistry for SSEA1 was performed to confirm differentiated or pluripotent status. In general, SSEA1 is the marker for pluripotency for murine ESCs, whereas only differentiated hESCs show expression.42 As shown in Fig. 4b and 4c, hESCs in differentiation medium showed strong staining for SSEA1, correlating with the loss of EGFP signal. In contrast, hESCs in the self-renewal medium maintained strong expression of EGFP, but no expression of SSEA1 was detected. This demonstrated that a single This journal is ª The Royal Society of Chemistry 2009 Fig. 4 A single hESC-mChip serves as a platform for parallel examination of controlled differentiation and self-renewal for hESCs. Either differentiation medium (containing 5 mM retinoic acid and 15% FBS) or self-renewal medium (100 ng mL1 bFGF) was introduced into cell culture chambers No. 1/3/5 or No. 2/4/6, respectively. (a) Fate mapping of HSF1-OCT4-EGFP colonies under the differentiation and self-renewal conditions using an inverted fluorescent microscope. After 2 days, hESC colonies under the differentiation condition gradually lost their hESC morphology and EGFP signal, whereas the hESC colonies under the self-renewal condition grew larger accompanied by an increased EGFP signal. (b) HSF1-OCT4-EGFP cells were immunostained for SSEA1 (a differentiation marker) at Day 4. (c) Quantitative comparison of EGFP intensity of hESC colonies in differentiation medium and self-renewal medium. Each bar represents the standard deviation (n > 7). hESC-mChip could carry out controlled self-renewal and differentiation in parallel without cross contamination. Discussion We have successfully demonstrated reproducible and quantitative culture and analysis of individual hESC colonies in an integrated microfluidic platform, the hESC-mChip. The six individually addressable cell culture chambers in the hESCThis journal is ª The Royal Society of Chemistry 2009 mChip allowed parallel examination of combinations of variables over time to obtain optimal culture conditions for self-renewal and controlled differentiation of hESCs. In addition to the intrinsic advantages of microfluidic systems, the hESC-mChip provides an opportunity to culture hESCs in different conditions in parallel as well as to run sequential phenotypical and functional analyses. Several small-scale screenings were performed to identify the optimal chip-based culture conditions that are widely applicable for a collection of hESCs, including two parental Lab Chip, 2009, 9, 555–563 | 561 hESC lines and three genetically modified hESC lines. Semiautomated immunoassays for a number of pluripotency markers were carried out in sequence to confirm that the chip-cultured hESC colonies maintained their pluripotency over a culture period of at least 6 days.43 Two more hESC lines, HSF6 and H9, could also be cultured in the hESC-mChip, and maintained their pluripotency (data not shown). Finally, we were able to demonstrate parallel examination of proliferation or controlled differentiation in a single hESC-mChip. Three genetically modified hESC lines allowed quantitative monitoring of hESC proliferation and pluripotency of the hESC-mChip-cultured hESC colonies in a real-time manner. Conventional hESC research is conducted in a collective fashion which overlooks a great deal of information on individual hESC colonies and their microenvironments over time. Lack of precise control of experimental and analytical conditions makes it difficult to interpret the results obtained from different experiments. In the hESC-mChip, there are six identical cell culture chambers providing a closely related microenvironment for multiparameter analysis. In each cell culture chamber, there is a built-in landmark to register individual hESC colonies for continuous fate mapping. The hESC-mChip is controlled by a laptop PC, allowing reproducible culture and analysis of individual hESC colonies in a semi-automated fashion. Although there were microfluidic devices developed for the culture of hESCs,33,34,44 no quantitative and integrated culture and analysis has been reported. In conjunction with a fluorescent microscope and three genetically modified hESC lines, we demonstrated, for the first time, that the hESC-mChip is capable of integrated and quantitative culture and analysis of hESCs. We also realized that we have a limited number of samples per chip in the hESC-mChip, and it cannot be operated in a fully automated fashion. Currently, a new generation of fully automated hESC-mChip incorporating hundreds of individual cell culture chambers is under development. We envision that the new generation hESC-mChip will be applied for high-throughput screening of feeder-free and chemically defined conditions which better regulate self-renewal and differentiation of hESCs. Furthermore, by using HSF1-LG and OCT4-EGFP knock-in cell lines, the integrated microfluidic hESC culture platform can provide a new screening system of the condition for single hESC expansion and fate mapping for individual hESCs. Acknowledgements This work was partially supported by the Eli and Edythe Broad Center of Regenerative Medicine and Stem Cell Research at the Institute of Molecular Medicine at University of California, Los Angeles and the DOE-UCLA Institute of Molecular Medicine. O.N.W. is an investigator of the Howard Hughes Medical Institute. We thank Dr. Michael Teitell at UCLA for his expert opinions on teratoma histology. We thank Dr. James Thomson at WiCell for kindly providing the OCT4-EGFP knock-in DNA construct. References 1 M. Amit and J. Itskovitz-Eldor, Meth. Mol. Biol. (Clifton, N. J.), 2006, 331, 43–53. 2 I. Singec, R. Jandial, A. Crain, G. Nikkhah and E. Y. Snyder, Annu. Rev. Med., 2007, 58, 313–328. 562 | Lab Chip, 2009, 9, 555–563 3 J. A. Thomson, J. Itskovitz-Eldor, S. S. Shapiro, M. A. Waknitz, J. J. Swiergiel, V. S. Marshall and J. M. Jones, Science, 1998, 282, 1145–1147. 4 B. S. Mallon, K. Y. Park, K. G. Chen, R. S. Hamilton and R. D. McKay, Int. J. Biochem. Cell Biol., 2006, 38, 1063–1075. 5 M. J. Martin, A. Muotri, F. Gage and A. Varki, Nat. Med., 2005, 11, 228–232. 6 M. Amit, C. Shariki, V. Margulets and J. Itskovitz-Eldor, Biol. Reprod., 2004, 70, 837–845. 7 J. Lu, R. Hou, C. J. Booth, S. H. Yang and M. Snyder, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 5688–5693. 8 T. E. Ludwig, V. Bergendahl, M. E. Levenstein, J. Y. Yu, M. D. Probasco and J. A. Thomson, Nat. Meth., 2006, 3, 637–646. 9 T. E. Ludwig, M. E. Levenstein, J. M. Jones, W. T. Berggren, E. R. Mitchen, J. L. Frane, L. J. Crandall, C. A. Daigh, K. R. Conard, M. S. Piekarczyk, R. A. Llanas and J. A. Thomson, Nat. Biotechnol., 2006, 24, 185–187. 10 C. Xu, E. Rosler, J. Jiang, J. S. Lebkowski, J. D. Gold, C. O’Sullivan, K. Delavan-Boorsma, M. Mok, A. Bronstein and M. K. Carpenter, Stem Cells, 2005, 23, 315–323. 11 R.-H. Xu, R. M. Peck, D. S. Li, X. Feng, T. Ludwig and J. A. Thomson, Nat. Meth., 2005, 2, 185–190. 12 S. Yao, S. Chen, J. Clark, E. Hao, G. M. Beattie, A. Hayek and S. Ding, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 6907–6912. 13 H. Lee, G. Al Shamy, Y. Elkabetz, C. M. Schofield, N. L. Harrsion, G. Panagiotakos, N. D. Socci, V. Tabar and L. Studer, Stem Cells, 2007, 25, 1931–1939. 14 S. Levenberg, N. F. Huang, E. Lavik, A. B. Rogers, J. Itskovitz-Eldor and R. Langer, Proc. Natl. Acad. Sci. U. S. A., 2003, 100, 12741– 12746. 15 M. Pick, L. Azzola, A. Mossman, E. G. Stanley and A. G. Elefanty, Stem Cells, 2007, 25, 2206–2214. 16 M. Schuldiner, O. Yanuka, J. Itskovitz-Eldor, D. A. Melton and N. Benvenisty, Proc. Natl. Acad. Sci. U. S. A., 2000, 97, 11307–11312. 17 J. El-Ali, P. K. Sorger and K. F. Jensen, Nature, 2006, 442, 403–411. 18 S. Takayama, J. C. McDonald, E. Ostuni, M. N. Liang, P. J. Kenis, R. F. Ismagilov and G. M. Whitesides, Proc. Natl. Acad. Sci.U. S. A., 1999, 96, 5545–5548. 19 W. Gu, X. Zhu, N. Futai, B. S. Cho and S. Takayama, Proc. Natl. Acad. Sci. U. S. A., 2004, 101, 15861–15866. 20 T. Thorsen, S. J. Maerkl and S. R. Quake, Science, 2002, 298, 580– 584. 21 B. J. Kirby, T. J. Shepodd and E. F. Hasselbrink, J. Chromatogr., A, 2002, 979, 147–154. 22 M. A. Unger, H. P. Chou, T. Thorsen, A. Scherer and S. R. Quake, Science, 2000, 288, 113–116. 23 C. C. Lee, G. Sui, A. Elizarov, C. J. Shu, Y. S. Shin, A. N. Dooley, J. Huang, A. Daridon, P. Wyatt, D. Stout, H. C. Kolb, O. N. Witte, N. Satyamurthy, J. R. Heath, M. E. Phelps, S. R. Quake and H. R. Tseng, Science, 2005, 310, 1793–1796. 24 J. Wang, G. Sui, V. P. Mocharla, R. J. Lin, M. E. Phelps, H. C. Kolb and H. R. Tseng, Angew. Chem. Int. Ed. Engl., 2006, 45, 5276–5281. 25 F. K. Balagadde, L. You, C. L. Hansen, F. H. Arnold and S. R. Quake, Science, 2005, 309, 137–140. 26 E. A. Ottesen, J. W. Hong, S. R. Quake and J. R. Leadbetter, Science, 2006, 314, 1464–1467. 27 Z. T. F. Yu, K. Kamei, H. Takahashi, C. J. Shu, X. Wang, G. W. He, R. Silverman, G. G. Radu, O. N. Witte, K. B. Lee and H. R. Tseng, Biomed. Microdevices, in press. 28 R. Gomez-Sjoberg, A. A. Leyrat, D. M. Pirone, C. S. Chen and S. R. Quake, Anal. Chem., 2007, 79, 8557–8563. 29 V. I. Chin, P. Taupin, S. Sanga, J. Scheel, F. H. Gage and S. N. Bhatia, Biotechnol. Bioeng., 2004, 88, 399–415. 30 B. G. Chung, L. A. Flanagan, S. W. Rhee, P. H. Schwartz, A. P. Lee, E. S. Monuki and N. L. Jeon, Lab Chip, 2005, 5, 401–406. 31 L. Kim, M. D. Vahey, H. Y. Lee and J. Voldman, Lab Chip, 2006, 6, 394–406. 32 V. V. Abhyankar and D. J. Beebe, Anal. Chem., 2007, 79, 4066–4073. 33 V. V. Abhyankar, G. N. Bittner, J. A. Causey and T. J. B. Kamp, 7th lnternational Conference on Miniaturized Chemical and Biochemical Analysts Systems, 2003, p. 17. 34 N. Korin, A. Bransky, U. Dinnar and S. Levenberg, in Biomedical Applications of Micro- and Nanoengineering III, SPIE, Adelaide, Australia, 2006, pp. 64160N-64168. This journal is ª The Royal Society of Chemistry 2009 35 E. Figallo, C. Cannizzaro, S. Gerecht, J. A. Burdick, R. Langer, N. Elvassore and G. Vunjak-Novakovic, Lab Chip, 2007, 7, 710–719. 36 B. Yao, G. A. Luo, X. Feng, W. Wang, L. X. Chen and Y. M. Wang, Lab Chip, 2004, 4, 603–607. 37 B. Huang, H. Wu, S. Kim and R. N. Zare, Lab Chip, 2005, 5, 1005– 1007. 38 P. Ray, R. Tsien and S. S. Gambhir, Cancer Res., 2007, 67, 3085–3093. 39 Z. Li, Y. Suzuki, M. Huang, F. Cao, X. Xie, A. J. Connolly, P. C. Yang and J. C. Wu, Stem Cells, 2008, 26, 864–873. 40 P. Morier, C. Vollet, P. E. Michel, F. Reymond and J. S. Rossier, Electrophoresis, 2004, 25, 3761–3768. 41 H. Yu, I. Meyvantsson, I. A. Shkel and D. J. Beebe, Lab Chip, 2005, 5, 1089–1095. This journal is ª The Royal Society of Chemistry 2009 42 T. P. Zwaka and J. A. Thomson, Nat. Biotechnol., 2003, 21, 319– 321. 43 In routine hESC culture, hESCs are passaged in clusters using well plates or culture dishes. Freshly dissociated hESC colonies contain approximately 20–200 cells. Due to the fast-growing nature of hESCs, freshly dissociated hESC colonies grow to critical mass at about Day 6. To prevent loss of pluripotency of large hESC conies, hESC passage has to be carried out within 6 days. For the same reason, we normally carry out hESC culture no more than 6 days. When small hESC conies were loaded into the chips, we were able to culture hESCs in the microfluidic devices for more than six days (maximum 12 days). 44 V. V. Abhyankar and D. J. Beebe, Anal. Chem., 2007, 79, 4066–4073. Lab Chip, 2009, 9, 555–563 | 563 Cell Stem Cell Resource Phosphoproteomic Analysis of Human Embryonic Stem Cells Laurence M. Brill,1,2,5,* Wen Xiong,3,5 Ki-Bum Lee,3,5,6 Scott B. Ficarro,1,7 Andrew Crain,2,4 Yue Xu,3 Alexey Terskikh,2 Evan Y. Snyder,2,* and Sheng Ding3,* 1Genomics Institute of the Novartis Research Foundation, 10675 John Jay Hopkins Drive, San Diego, CA 92109, USA Institute for Medical Research, 10901 North Torrey Pines Road, La Jolla, CA 92037, USA 3Department of Chemistry, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037, USA 4Biomedical Sciences Graduate Program, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093, USA 5These authors contributed equally to this work 6Present address: Department of Chemistry and Chemical Biology, Institute for Advanced Materials, Devices, and Nanotechnology, The Rutgers Stem Cell Research Center, Rutgers University, 610 Taylor Road, Piscataway, NJ 08854, USA 7Present address: Dana-Farber Cancer Institute, 44 Binney Street, Boston, MA 02115, USA *Correspondence: [email protected] (L.M.B.), [email protected] (E.Y.S.), [email protected] (S.D.) DOI 10.1016/j.stem.2009.06.002 2Burnham SUMMARY Protein phosphorylation, while critical to cellular behavior, has been undercharacterized in pluripotent cells. Therefore, we performed phosphoproteomic analyses of human embryonic stem cells (hESCs) and their differentiated derivatives. A total of 2546 phosphorylation sites were identified on 1602 phosphoproteins; 389 proteins contained more phosphorylation site identifications in undifferentiated hESCs, whereas 540 contained more such identifications in differentiated derivatives. Phosphoproteins in receptor tyrosine kinase (RTK) signaling pathways were numerous in undifferentiated hESCs. Cellular assays corroborated this observation by showing that multiple RTKs cooperatively supported undifferentiated hESCs. In addition to bFGF, EGFR, VEGFR, and PDGFR activation was critical to the undifferentiated state of hESCs. PDGF-AA complemented a subthreshold bFGF concentration to maintain undifferentiated hESCs. Also consistent with phosphoproteomics, JNK activity participated in maintenance of undifferentiated hESCs. These results support the utility of phosphoproteomic data, provide guidance for investigating protein function in hESCs, and complement transcriptomics/epigenetics for broadening our understanding of hESC fate determination. INTRODUCTION Human embryonic stem cells (hESCs) are a model developmental system that may have potential clinical value for mitigating diseases. Mechanisms of hESC fate determination are not well defined, although there has been progress in elucidating molecular circuitry of self-renewing ESCs. Transcriptional profiles of hESCs (Brandenberger et al., 2004; Sato et al., 2003; Sperger et al., 2003) and more limited ChIP-on-chip (Boyer et al., 2005) and proteomic (Bendall et al., 2007; Van Hoof et al., 2006) analyses 204 Cell Stem Cell 5, 204–213, August 7, 2009 ª2009 Elsevier Inc. suggest mechanisms underlying hESC self-renewal and differentiation. In addition to transcriptional and translational regulation, cell-fate determination is controlled by protein phosphorylation, a critical determinant of cell signaling (Mann et al., 2002; Schlessinger, 2000). Recent phosphoproteomic analyses of human mesenchymal stem cells identified 716 and 703 protein phosphorylation sites (Thingholm et al., 2008a, 2008b). However, protein phosphorylation has not been well characterized in pluripotent cells. Therefore, we performed a large-scale multidimensional liquid chromatography (MDLC)- tandem mass spectrometry (MS/MS)-based phosphoproteomic analysis of undifferentiated hESCs and their differentiated derivatives for identification of protein phosphorylation sites in these cells. Undifferentiated hESCs were cultured under feeder-free conditions with bFGF. Comparable differentiated derivatives were obtained by removal of bFGF and treatment with retinoic acid (RA), which induces nearly complete, albeit nonspecific, differentiation to a heterogeneous population of cells. Removal of bFGF alone does not result in complete differentiation, whereas concurrent RA treatment causes virtually complete loss of the undifferentiated population in 4 days (required for this type of analysis). Our data provide a freely available resource of protein phosphorylation sites in hESCs and differentiated derivatives (http://www. ebi.ac.uk/pride/). These data have begun to prove informative and predictive. For example, as proof of concept, pathway analyses of the phosphoproteins suggested potential responses of hESCs to perturbations of receptor tyrosine kinase (RTK) signaling pathways. To test some RTK pathways for a role in the maintenance of undifferentiated hESCs, we treated hESC cultures with selected agonists or antagonists of these pathways. Their effects were consistent with predictions of the phosphoproteomic analyses. Furthermore, the data suggested previously unidentified protein roles in hESC self-renewal or differentiation, thus providing extensive guidance for future research. RESULTS Phosphoproteomic Analysis of hESCs Because phosphoproteomic analysis is challenging (Mann et al., 2002) and has not been reported in hESCs, we chose to analyze Cell Stem Cell Human ESC Phosphoproteomics Figure 1. Undifferentiated hESCs Expressed Markers of Pluripotency, whereas the Markers Were Downregulated upon Differentiation Cells were cultured to yield undifferentiated hESCs (hESCs), or differentiated hESC derivatives (derivs) under feeder-free conditions by withdrawing bFGF and including 5 mM RA in the media for the final 4 days of culture. Nuclei were stained with DAPI (A and B; left column). (A) Cells were stained with antibodies detecting OCT4 (center column), and OCT4 and DAPI images were merged (right column). (B) Cells were stained with antibodies detecting SSEA-4 (center column), and SSEA-4 and DAPI images were merged (right column). All photomicrographs were at the same magnification. The scale bar represents 50 mM. the well-characterized hESC line H1 (WiCell; WA01) (Thomson et al., 1998), which has been used in molecular studies of hESCs (Bendall et al., 2007; Brandenberger et al., 2004; Wang et al., 2007). Fifty-nine hESC lines, including H1, showed remarkable conservation of hESC markers (Adewumi et al., 2007), which provided confidence that our findings would be representative. Before analyzing protein phosphorylation, the undifferentiated hESC markers OCT4 (Thomson et al., 1998) and SSEA-4 (Reubinoff et al., 2000) were examined to assess whether the hESCs were truly undifferentiated under our culture conditions and whether differentiation was complete. Undifferentiated hESCs were cultured on Matrigel-coated plates in feeder-free cultures using conditioned media (CM) that contained 8 ng/ml of added bFGF. A heterogeneous population of differentiated derivatives of the hESCs was obtained by removal of bFGF and treatment with 5 mM RA for 4 days. OCT4 was detected in 97% of the hESCs under the feeder-free conditions, whereas it was nearly undetectable in differentiated derivatives (Figure 1). Similarly, SSEA-4 was positive in the undifferentiated hESCs and nearly absent in differentiated derivatives. Moreover, the nucleus-to-cytoplasm ratio, also monitored as an indicator of whether hESCs are undifferentiated or differentiated, was consistent with OCT4 and SSEA-4 expression (Figure 1). These observations suggested that our cells represented two distinct populations—‘‘undifferen- tiated’’ or ‘‘differentiated’’ hESC derivatives—that might then be reliably subjected to phosphoproteomic analysis, using MDLC-MS/MS technology, that can result in unbiased discovery of protein phosphorylation sites (Kruger et al., 2008). Phosphoproteomic analyses of hESCs and their differentiated derivatives were performed using automated MDLC, a linear ion trap mass spectrometer, and readily available bioinformatics algorithms. Phosphorylated peptides from total proteins from undifferentiated hESCs or their differentiated derivatives were separated, enriched, and analyzed using MDLC comprised of strong cation exchange chromatography (SCX), reversed-phase (RP) desaltFe3+-immobilized metal affinity chromatography (desalt-IMAC), and RP HPLC coupled to nano-electrospray ionization-tandem mass spectrometry (ESI-MS/MS; see a schematic diagram in Figure S1, available online). IMAC, for phosphopeptide enrichment, coupled to RP HPLC-ESI-MS/MS is a robust technique for phosphoproteomic analyses (Bodenmiller et al., 2007; Brill et al., 2004; Gruhler et al., 2005), and automation improves reliability and reproducibility (Ficarro et al., 2005). Because phosphorylated proteins are frequently at low abundance, substoichiometrically phosphorylated, and difficult to identify (Mann et al., 2002), replicate analyses were performed to increase phosphoproteome coverage. Replicates increase proteome coverage, especially of lower abundance proteins (Liu et al., 2004), and the impact of experimental variation in LC-MS/MS can be minimized by replicates (Washburn et al., 2003). Phosphopeptides were identified with high confidence (see Supplemental Experimental Procedures). Examples of typical MS/MS spectra used to identify phosphopeptides are in Figure S2. To complement identification, extracted ion chromatograms (XICs) were used to quantify the relative abundance of phosphopeptides. The normalized abundance of randomly selected phosphopeptides identified in all four phosphoproteomic analyses (two biological replicates, i.e., phosphopeptides from two Cell Stem Cell 5, 204–213, August 7, 2009 ª2009 Elsevier Inc. 205 Cell Stem Cell Human ESC Phosphoproteomics pairs of independent cultures of undifferentiated hESCs or their differentiated derivatives) demonstrated relatively low variability (Table S1). This degree of consistency agrees with previous findings in which proteomic data can be reliably compared among experiments (Washburn et al., 2003). In contrast, differential phosphopeptide identification implies differential phosphopeptide abundance. We used data-dependent MS/MS, and peptide abundance and identification correlate in data-dependent MS/MS (Liu et al., 2004). Selected phosphopeptides identified in undifferentiated hESC or differentiated derivative cell populations were also quantified using XICs. Furthermore, signal from each of the selected phosphopeptides was manually sought in the MS/MS data from analyses in which it had not been identified by SEQUEST searches, in order to test whether the phosphopeptide was detectable and, if so, its relative abundance among the phosphoproteomic analyses. Only a fraction of the phosphopeptides not identified in SEQUEST searches were detectable (via a poor quality MS/MS spectrum) when searching the raw data (Table S2). However, every phosphopeptide that was examined demonstrated a higher normalized abundance in analyses in which it was identified than in analyses in which it was not identified by SEQUEST searches. Although lack of identification of a phosphopeptide is not evidence for its absence, identification versus lack of identification implies that the phosphopeptide is likely to be more abundant in the cell population in which it was identified, consistent with our results (Table S2) and those of others (Liu et al., 2004). Western blots were performed on proteins from undifferentiated hESCs and differentiated derivatives, using antibodies recognizing phosphorylation sites previously identified by MDLC-MS/MS. All nine antibodies that were used recognized bands with the expected mobility on western blots, providing confidence in phosphorylation site identifications. Representative western blots, including normalized integrated intensities of phosphoprotein bands, are shown in Figure S3. Phosphorylation of mTOR on Ser2448 was apparently more abundant in undifferentiated than differentiated cells (Figure S3A), and mTOR Ser2448 phosphorylation was identified in undifferentiated, but not differentiated cells, using MDLC-MS/MS (Table S3A). PAK1 phosphorylation on Ser144 was identified twice in undifferentiated cells and once in differentiated cells by MDLC-MS/MS (Table S5A), and western blots suggested that PAK1 phosphoserine 144 was more abundant in undifferentiated than in differentiated cells (Figure S3B). Antibodies recognizing PTK2 phosphotyrosine 576/577 suggested that phosphorylation of this site was more abundant in differentiated derivatives than undifferentiated hESCs (Figure S3C), consistent with identification of PTK2 phosphorylated on Tyr576, using MDLC-MS/MS, only in differentiated derivatives (Table S4A). Phosphorylation of CDK1/2/3/5 on Thr14 and Tyr15 (two conserved residues in all four CDK proteins) was more abundant in undifferentiated cells (Figure S3D), and XIC peak areas suggested that phosphorylation of CDK1/2/3 on Thr14 and Tyr15 was more abundant in undifferentiated cells (Table S1). CDK1, -2, -3, and -5 phosphorylated on Thr14 and Tyr15 are recognized in western blots (Supplemental Experimental Procedures), and the corresponding phosphopeptides identified by MDLC-MS/MS (IGEGT*YGVVY and IGEGTY*GVVY; for brevity, designated as originating from CDK2 in Tables S1 and S5) are identical among CDK1/2/3, whereas the 206 Cell Stem Cell 5, 204–213, August 7, 2009 ª2009 Elsevier Inc. corresponding peptide from CDK5 differs at two amino acid residues (IGEGT*Y*GTVF), which is easily distinguishable by MS/MS. The relative abundance of JUN phosphorylated on Ser63, and HSP27 phosphorylated on Ser82, was similar in undifferentiated and differentiated cells on western blots (data not shown), and phosphorylated JUN Ser63 and phosphorylated HSP27 Ser82 were both identified the same number of times in undifferentiated and differentiated cells (Table S5A), demonstrating further agreement between western blots and MDLC-MS/MS. If subsequent studies focus on one or a few especially critical sites of protein phosphorylation, it is advisable to examine the phosphorylation site using an independent technique. However, MDLC-MS/MS is reliable for phosphoproteome analysis and can yield unbiased, large-scale discovery of protein phosphorylation (Bodenmiller et al., 2007; Brill et al., 2004; Ficarro et al., 2005; Gruhler et al., 2005; Kruger et al., 2008; Thingholm et al., 2008a), and our findings support its accuracy. Together, these results suggest that application of MDLC-MS/MS for identification of phosphopeptides was suitable for phosphoproteomic analysis of undifferentiated hESCs and their differentiated derivatives. Phosphopeptide identifications are in Tables S3A–S5B. Each phosphoprotein, from which phosphopeptides were derived, was classified as either (1) containing more phosphorylation site identifications in undifferentiated hESCs, (2) containing more phosphorylation site identifications in differentiated hESC derivatives, or (3) containing a similar number of phosphorylation site identifications in both cell populations. A protein is conservatively defined to contain more phosphorylation site identifications in a cell population if its phosphorylation was identified exclusively in this population or at least 3-fold more frequently than in the other population; otherwise, the protein is considered to contain a similar number of phosphorylation site identifications in populations from both cell states. Although identification of protein phosphorylation sites was unlikely to be comprehensive, as implied by studies using different cell types (Bodenmiller et al., 2007; Mann et al., 2002), among the 2546 nonredundant phosphorylation sites, 472 were on proteins containing more phosphorylation site identifications in undifferentiated hESCs, whereas 726 were on proteins containing more phosphorylation site identifications in differentiated hESC derivatives (Figure 2A). Of the peptides, 94% were singly phosphorylated, whereas the rest were doubly phosphorylated, similar to other studies using IMAC for phosphopeptide enrichment (Bodenmiller et al., 2007; Thingholm et al., 2008a). Serine, threonine, and tyrosine phosphorylation comprised 82%, 14%, and 4% of the sites, respectively (Tables S3A–S5B), and tyrosine phosphorylation was relatively prominent in undifferentiated hESCs (Figure 2C). Among the 1602 proteins, 389 contained more phosphorylation site identifications in undifferentiated hESCs, whereas 540 contained more phosphorylation site identifications in differentiated hESC derivatives (Figure 2B). Transcription factors can reprogram differentiated cell types to ESC-like cells when ectopically expressed (Takahashi et al., 2007; Takahashi and Yamanaka, 2006; Yu et al., 2007) and were the most abundant known phosphoprotein category (Figures 2G–2I). This observation, not typical of proteomic analyses, could reflect the growing consensus that many transcription regulators are important in control of ESC state. Among the Cell Stem Cell Human ESC Phosphoproteomics Figure 2. Number of Protein Phosphorylation Sites and Phosphoproteins Identified in hESCs and Their Differentiated Derivatives, Prominence of Tyrosine Phosphorylation, Predicted Subcellular Location of the Phosphoproteins, and Phosphoprotein Categories (A and B) (A) Total number of nonredundant phosphorylation sites and (B) number of proteins with more phosphorylation site identifications in undifferentiated hESCs (line H1/WA01) (represented in red), RA differentiated, H1-hESC derivatives (represented in gold), or with a similar number of phosphorylation site identifications in the two cell populations (represented in gray). The percentage of the phosphorylation sites and phosphoproteins in each of the three groups of proteins is shown in parentheses. (C) Percentage of nonredundant tyrosine phosphorylation sites, among the sites for which the phosphorylated residue could be defined as serine, threonine, or tyrosine (94% of all sites), that were on proteins containing more identified sites in undifferentiated hESCs, differentiated hESC derivatives, or that were on proteins with a similar number of identified sites between undifferentiated and differentiated cells. (D–F) The subcellular localization of the phosphoproteins is shown; those widely associated with more than one subcellular location are designated as variable. (G–I) Phosphoprotein categories, among those whose functions are known, are shown. The percentage of proteins with known functions are 45.8%, 55.7%, and 57.2% for proteins with more phosphorylation site identifications in undifferentiated hESCs, differentiated hESC derivatives, or a similar number of phosphorylation site identifications between the two cell populations, respectively. Each chart progresses from the protein category containing the most to the fewest entries. Abbreviations and definitions include the following: transcript. reg., transcription regulator; enzyme, protein with enzymatic activity outside of the other categories; RNA meta., RNA-binding proteins and proteins participating in metabolic processes involving RNA; prot. degr., protein degradation; transport., transporter; apop. reg., apoptosis regulator; transmem. recep., transmembrane receptor; GEF and GAP, guanine nucleotide exchange factor and GTPase-activating protein; cytoskel., proteins that are components of, closely associated with, or regulate cytoskeletal function; cell prolif., proteins participating in regulation of cellular proliferation and/or cell-cycle progression; tum. sup., tumor suppressor; translat. reg., translation regulator; phosphoinos. sig., proteins participating in phosphoinositide signaling; gen. assem., genome assembly; GF, growth factor; cell adhes., proteins functioning in cell adhesion; telomere mainten., protein functioning in telomere maintenance; prom. differ., proteins promoting cellular differentiation; GF buffer, proteins regulating the availability of growth factors; comp. casc., complement cascade; nuc. receptor, ligand-dependent nuclear receptor; and hormone biosyn., hormone biosynthesis. 158 phosphorylated transcription regulators, 41 contained more phosphorylation site identifications in undifferentiated hESCs, 46 contained more phosphorylation site identifications in differentiated hESC-derivatives, and 71 contained a similar number of phosphorylation site identifications in both cell populations. Most of the transmembrane receptors and predicted extracellular proteins contained more phosphorylation site identifications in either undifferentiated or differentiated hESCs, whereas fewer of these proteins contained a similar number of phosphorylation site identifications in both cell populations (Figures 2D–2I, Table S6), implying that growth factors, cytokines, their receptors, and corresponding signaling pathways could participate in controlling hESC fate. Furthermore, kinases, which are key players in cell signaling, represented the second-largest cate- gory of known phosphoproteins (Figures 2G–2I). Phosphorylation of cytoplasmic, cytoskeletal, and cell-adhesion proteins was identified relatively frequently in differentiated derivatives (Figures 2D–2I). Phosphorylated Transcription Regulators in Undifferentiated hESCs The transcription regulator ESG1 (official symbol TLE1; Table S7) is expressed only in preimplantation embryos, ESCs, and primordial germ cells (Western et al., 2005). ESG1 is coexpressed with OCT4 and SOX2 in both mouse and human ESCs, suggesting it is a potential pluripotency marker (Western et al., 2005). In addition, SUPT16H and SSRP1 (Tables S7 and S8) were phosphorylated in undifferentiated hESCs and are the Cell Stem Cell 5, 204–213, August 7, 2009 ª2009 Elsevier Inc. 207 Cell Stem Cell Human ESC Phosphoproteomics Table 1. Signaling Proteins with More Phosphorylation Site Identifications in Undifferentiated hESCs than in hESC Derivatives Proteins Participating in Receptor Tyrosine Kinase Signaling Receptors, Growth Factors AREGa, KDR, IGF2R, EPHA1 Kinases Phospholipases Adaptors Other LCK, NEK4, MAPK6, MAPK7, FRAP1 (mTOR), PIK3C3, PIK3R4, DBF4, CDC42BPA, CRKL, MINK1, KIAA1804 (MLK4), CDKL5, EIF2AK1, CRKRS PLCG1, PLCG2, PLCH1 SHC1, GAB1, NCK2, KIAA1303 (RAPTOR), CNKSR1, CNKSR2, ABI2, CDC37L1, PLEKHA1 PPAP2B, EPS15L1, TRAF4, APC, CDH17, IGFBP2, RAPGEF1, TRIP10, TSC1, WDR62, NUMB Signal Transduction Pathways and Member Proteins MAPK: JNK, ERK CRKLb, MINK1, KIAA1804 (MLK4), TRAF4, TRIP10, WDR62, CNKSR1, DBF4, CDC42BPA, RAPGEF1, PLCG1, SHC1, PLCG2, GAB1, LCK, MAPK6, MAPK7, NEK4, NCK2 PI3K/AKT/mTOR FRAP1 (mTOR), TSC1, GAB1, PIK3C3, PLCG2, PIK3R4, KIAA1303 (RAPTOR), ANRT (HIF-1b) a Official symbols of the proteins, some of which are followed by synonyms in parentheses, are used in this table. b Symbols in bold text represent proteins that are relatively specific to JNK signaling. two subunits of FACT (facilitates chromatin transcription). FACT destabilizes nucleosomes to allow transcription without disruption of the epigenetic state (Belotserkovskaya et al., 2003) and promotes initiation of DNA replication in the S phase of the cell cycle (Tan et al., 2006). CREBBP (Table S7) has histone acetyltransferase activity. Its mRNA is enriched in undifferentiated hESCs (Brandenberger et al., 2004) (Table S8). AKT (Table S5A) phosphorylates CREBBP, increasing CREBBP acetyltransferase activity and promoting NF-kB-mediated transcription and enhanced cell survival (Liu et al., 2006). Furthermore, CREBBP increases ERK1 expression (Chu et al., 2005). ERK1 activity contributes to hESC self-renewal in the presence of bFGF (Li et al., 2007). At least 18 phosphorylated transcription regulators identified in undifferentiated hESCs can modify chromatin structure via histone methylation or acetylation (Table S7) and may contribute to the epigenetic pattern that is likely to be important to hESCs (Bernstein et al., 2006; Lee et al., 2006; McCool et al., 2007). We identified phosphorylation of DNMT3B, MBD3 (Table S3A) and EZH2 (Table S5A) in undifferentiated hESCs. DNMT3B encodes a DNA methyltransferase (Table S7), which was expressed in all 59 hESC lines tested (Adewumi et al., 2007), was enriched in undifferentiated hESCs (Brandenberger et al., 2004), and was phosphorylated in undifferentiated hESCs (Table S8). Differential phosphorylation could modulate EZH2 activity. Phosphorylation at S21 by AKT inhibits the histone H3 Lys27 methyltransferase activity of EZH2 (Cha et al., 2005), and we identified a phosphorylation site of EZH2 in undifferentiated hESCs (S371 or T372; Table S5A), a site whose phosphorylation was also identified in undifferentiated mouse ESCs (L.M.B., K.-B.L., W.X., and S.D., unpublished data). Phosphorylated transcription regulators in undifferentiated hESCs can participate in transcriptional activation or repression, histone modification, and more (Table S7). These and other functions may be integrated to favor the undifferentiated state of hESCs, as implied by the complexity of the phosphoproteome (Figure 2). Although some of these transcriptional and epigenetic regulators were previously reported to influence hESCs, the mechanisms are unclear. The identified phosphorylation sites provide focused information for future studies of the function of these factors in hESCs. Furthermore, we also identified 208 Cell Stem Cell 5, 204–213, August 7, 2009 ª2009 Elsevier Inc. hundreds of phosphoproteins whose presence in hESCs was unknown, providing a rich resource for further investigation. For instance, TNRC6A, a factor for gene silencing via RNA interference (Liu et al., 2005), was phosphorylated in undifferentiated hESCs (Table S3A). Growth Factor-Mediated Signaling Pathways in Undifferentiated hESCs Tyrosine phosphorylation, which plays a dominant role in growth factor/RTK signaling pathways (Schlessinger, 2000), was relatively prominent in undifferentiated hESCs (Figure 2C). Signaling pathways participating in self-renewal of hESCs include bFGF, TGF-b/activin, insulin/IGF, EGFR family, PDGF, Wnt, neurotrophin, integrin, and Notch pathways (Beattie et al., 2005; Bendall et al., 2007; James et al., 2005; Pebay et al., 2005; Wang et al., 2007; Xu et al., 2005; Yao et al., 2006). However, detailed understanding of the action of these pathways is lacking. The phosphoproteins were grouped into signaling pathways, as described in the Supplemental Experimental Procedures, to further explore their functional potential. Forty-one canonical and metabolic pathways were suggested using the phosphoproteins as input for pathway analysis (data not shown). Proteins in RTK pathways were phosphorylated in undifferentiated hESCs, including the adaptors GAB1, SHC1, and NCK2; the kinases LCK, NEK4, MAPK6, MAPK7, mTOR, PIK3C3, and PIK3R4; phospholipases PLC-g1 and PLC-g2; and the phosphatase PPAP2B (Table 1). Some phosphoproteins are shared among pathways, and some are more pathway specific, such as APC in Wnt signaling and NUMB in Notch signaling. Table 1 and Figure 2C imply that a variety of signaling pathways are important in undifferentiated hESCs. For example, EGF pathway members ErbB2, AREG, and EPS15L1 were phosphorylated in undifferentiated hESCs (Table 1 and Table S5), complementing a report showing that the ErbB2/ErbB3 ligand heregulin-1b helps support undifferentiated hESCs (Wang et al., 2007). KDR (VEGFR2, FLK1) was phosphorylated in undifferentiated hESCs (Table 1), and stimulation of hESCs with CM elicits tyrosine phosphorylation (site[s] undefined) of PDGFRA (Wang et al., 2007). Components of the VEGF and PDGF pathways were phosphorylated in undifferentiated hESCs, including some proteins in Table 1. We also identified phosphoproteins from signaling pathways whose Cell Stem Cell Human ESC Phosphoproteomics presence in hESCs has not been reported, and a large number of proteins not previously known to be phosphorylated (Tables S3A–S5B). Molecular profiling studies typically lack biological follow-up (e.g., Bodenmiller et al., 2007; Boyer et al., 2005; Brandenberger et al., 2004; Brill et al., 2004; Ficarro et al., 2005; Gruhler et al., 2005; Lee et al., 2006; McCool et al., 2007; Sperger et al., 2003; Thingholm et al., 2008a, 2008b; Van Hoof et al., 2006). However, a few, including transcriptomic (Armstrong et al., 2006) and proteomic (Bendall et al., 2007; Kratchmarova et al., 2005; Mukherji et al., 2006; Wang et al., 2006; Wang et al., 2007) studies, demonstrated that cells responded to stimulation in manners consistent with molecular profiles. To test the cellular relevance of the phosphoproteomic and pathway analyses, we began by targeting EGF, VEGF, and PDGF pathways in undifferentiated hESCs using inhibitors of their receptors. Although specificity of RTK inhibitors is imperfect, we used some of the widely accepted ones (see the Supplemental Experimental Procedures). Treatment of undifferentiated hESC cultures with an EGFR inhibitor at 10 mM resulted in extensive apoptosis (data not shown), similar to another report (Wang et al., 2007). The hESCs were also treated with 10 mM KDR inhibitor II or 10 mM Gleevec, a PDGFRA inhibitor (Zhang et al., 2003). Undifferentiated control colonies were compact and expressed OCT4 and SSEA-4 (Figure 3B and data not shown). In contrast, most cells differentiated in the presence of KDR or PDGFR inhibitor, shown by flattening of the colonies, altered cellular morphology and nearly undetectable OCT4 and SSEA-4 (Figure 3C and data not shown). Vehicle-only controls lacked any noticeable effect on the cells (Figure 3B). The results were similar under feeder-free conditions in CM and feeder-free conditions in chemically defined media (CDM; Yao et al., 2006). Furthermore, KDR or PDGFR inhibitor, at 10 mM, resulted in decreased expression of NANOG and OCT4 (Figure 3A). To further investigate the effect of RTK signaling pathways, we decreased bFGF to a subthreshold 4 ng/ml (at least 20 ng/ml is required under feeder-free conditions in CDM [Yao et al., 2006]) and systematically supplemented cultures with EGF, PDGF-AA, or VEGF-AA at different concentrations to determine which trophic factor could complement bFGF deficiency. Although PDGF-AA without bFGF was unable to maintain longterm cultures of undifferentiated hESCs, PDGF-AA at 10 ng/ml and the subthreshold concentration of 4 ng/ml of bFGF (subsequently abbreviated PDGF/bFGF) stably maintained undifferentiated hESCs under feeder-free conditions in CDM for >15 passages, and the hESCs remained undifferentiated throughout all four experiments (Figure 4D). The cells displayed undifferentiated morphology and robust expression of OCT4. In contrast, when undifferentiated hESCs, which had been stably maintained in CDM containing PDGF/bFGF for >15 passages, were subsequently cultured for 4 days in CDM containing 4 ng/ml of bFGF but no PDGF, the cells differentiated (Figure 4B). FACS analyses demonstrated that 89% of the hESCs in CDM containing PDGF/bFGF were positive for SSEA-4, comparable to cultures in CDM containing 20 ng/ml of bFGF (86%; Figure 4). Similar FACS results were obtained when cells were stained and sorted for the pluripotency marker Tra-1-60 (data not shown). Moreover, PDGF/bFGF in CDM resulted in sustained expression of NANOG and OCT4 transcripts, whereas their abundance Figure 3. Protein Kinase Inhibitors Resulted in Differentiation of hESCs (A) Expression of NANOG (Chambers et al., 2003) and OCT4 mRNAs was assessed by RT-PCR, in the presence of protein kinase inhibitors that resulted in differentiation of hESCs. Cells were cultured with 20 ng/ml of bFGF, and inhibitors (10 mM) were included in the cultures for the final 4 days. Inhibitor identities are indicated in the figure. Slower decline of OCT4 than NANOG was typically observed during hESC differentiation. GAPDH was an internal control. (B and C) Undifferentiated, vehicle-only control (B) and differentiated, KDR inhibitor-treated (C) cells are shown under imaging conditions indicated above the columns. All photomicrographs were at the same magnification, and the scale bar (bottom right) represents 50 mM. Abbreviations include the following: i, inhibitor; uhESCs, undifferentiated hESCs. declined within 4 days in the absence of PDGF-AA or the presence of the PDGFR inhibitor Gleevec (Figures 3A and 4A), further supporting the proposal that PDGF-AA facilitates maintenance of undifferentiated hESCs. Together, phosphoproteomic and pathway analyses suggested that PDGF should favor maintenance of undifferentiated hESCs. PDGFR inhibitor, and separate use of PDGF-AA, provided clear evidence that PDGF, when bFGF is at a subthreshold concentration, can promote the undifferentiated state of hESCs in CDM under feeder-free conditions, insights that derived directly from the phosphoproteomic analysis. Our data further suggested that ErbB and VEGFR activation participate in maintenance of undifferentiated hESCs, because disruption of these pathways caused apoptosis (data not shown) and/or differentiation (Figure 3) (although EGF and VEGF-AA demonstrated limited efficacy at complementing the deficiencies of 4 ng/ml bFGF). The ErbB2/ErbB3 ligand heregulin-1b contributes to maintenance of undifferentiated hESCs (Wang et al., 2007). In addition, insulin/IGF pathway members (Bendall et al., 2007) were phosphorylated in hESCs (including proteins in the PI3K/AKT/mTOR pathway; Table 1). Phosphoproteomics, cellular assays, and other reports (Bendall et al., 2007; Wang et al., 2007; Yao et al., 2006) suggest that multiple RTK pathways are required, although none of them alone is sufficient to support self-renewal in the absence of Cell Stem Cell 5, 204–213, August 7, 2009 ª2009 Elsevier Inc. 209 Cell Stem Cell Human ESC Phosphoproteomics and pathway analyses also imply that additional pathways could favor undifferentiated hESCs. Figure 4. PDGF and a Subthreshold Concentration of bFGF Sustained Long-Term Culture of hESCs (A) RT-PCR to amplify NANOG and OCT4 transcripts in long-term hESC cultures (>15 passages) in CDM containing 10 ng/ml of PDGF-AA and 4 ng/ ml of bFGF (lane PDGF, bFGF4). Lanes bFGF20 or bFGF4 refer to 20 or 4 ng/ml of bFGF in the CDM for 4 days, respectively, in the absence of PDGF, following culture in 10 ng/ml of PDGF-AA and 4 ng/ml of bFGF for >15 passages. (B–D) Colony morphology, OCT4 staining, and fluorescence-activated cell sorting (FACS) demonstrated that PDGF/bFGF in CDM maintained undifferentiated hESCs passaged >15 times. Imaging conditions or FACS analyses of SSEA-4 expression, detected via Cy3-conjugated secondary antibodies, is indicated above the columns, and the culture additives that were varied are indicated beside the rows. In FACS plots, dotted lines delineate boundaries of fluorescence intensity approximately indicative of cellular identity as undifferentiated hESCs (uhESC) and differentiated hESC derivatives (deriv). Decline of SSEA-4 is incomplete in differentiated hESCs after 4 days (Figure 1). Following maintenance of the hESCs in CDM containing bFGF at 4 ng/ml and PDGF-AA at 10 ng/ml for >15 passages, cells were cultured for 4 days in CDM lacking PDGF and containing bFGF at 4 ng/ml (B) or 20 ng/ml (C), or in the continued presence of bFGF at 4 ng/ml and PDGF-AA at 10 ng/ml (D). All photomicrographs were at the same magnification, and the scale bar (bottom center panel) represents 100 mM (B–D). bFGF. Also consistent with our results, although less clear due to the undefined media that was used, Sphingosine-1-phosphate plus PDGF contributes to maintenance of undifferentiated hESCs in the presence of mouse embryonic fibroblasts (MEFs) or MEF-conditioned media (Pebay et al., 2005). It previously appeared that bFGF alone might sustain self-renewal of hESCs. However, as predicted by our phosphoproteomic analysis, several other factors that exist in serum and/or are secreted by feeders, acting through autocrine or paracrine effects or as culture additives, are also important for hESC self-renewal (Bendall et al., 2007; Wang et al., 2007). Our phosphoproteomic 210 Cell Stem Cell 5, 204–213, August 7, 2009 ª2009 Elsevier Inc. Phosphorylated Signal Transduction Proteins in Undifferentiated hESCs PI3K signaling facilitates ESC self-renewal (Armstrong et al., 2006), and the PI3K pathway is activated by PDGF in mesenchymal stem cells (Kratchmarova et al., 2005), but the mechanism of action of the PI3K pathway has been unclear. PI3K/ AKT/mTOR pathway members were phosphorylated in undifferentiated hESCs. For example, PIK3C3 is enriched in undifferentiated hESCs (Brandenberger et al., 2004), and PIK3C3 was phosphorylated in undifferentiated hESCs (Table 1). mTOR (Table 1) plays a role in proliferation of undifferentiated hESCs (Wang et al., 2007) and is phosphorylated at Ser2448 during mitogenic stimulation (Chiang and Abraham, 2005). mTOR, phosphorylated at Ser2448 and Ser2454 in undifferentiated hESCs (Figure S3A, Table S3A) is a protein that enhances cell survival (Peponi et al., 2006). TSC1 was also phosphorylated in undifferentiated hESCs (Table 1). TSC1 can limit cell size (Rosner et al., 2003), and its overexpression caused cells to form compact clusters with increased reaggregation in vitro (Li et al., 2003), similar to the small size of undifferentiated hESCs and compact morphology of hESC colonies. Phosphorylated PI3K/AKT/mTOR pathway members in undifferentiated hESCs (Table 1) suggest which pathway members may regulate undifferentiated hESCs. Phosphoproteins participating in MAPK signaling were identified (Table 1). The ERK pathway contributes to hESC selfrenewal under conditions that include bFGF (Li et al., 2007), whereas JNK signaling in hESCs has not been reported. Some phosphoproteins downstream of RTK pathways are relatively specific to JNK signaling, such as TRAF4, MLK4, CRKL, and MINK1 (Table 1). To test for JNK signaling in undifferentiated hESCs, we tested two JNK inhibitors in hESC cultures under feeder-free conditions in CM. JNK inhibitor II, a small molecule (SP600125) widely used in JNK studies (Bennett et al., 2001; Han et al., 2001; Shin et al., 2002), and JNK inhibitor III, a polypeptide (Holzberg et al., 2003), were used. Each inhibitor alone resulted in cellular differentiation, demonstrated by colony morphology and decreased OCT4 expression (Figure S4). In contrast, controls lacking JNK inhibitors, including vehicle-only controls, remained undifferentiated (Figure S4 and data not shown). Induction of differentiation by JNK inhibitors was similar under feeder-free conditions in CDM (data not shown). Furthermore, OCT4 and NANOG mRNA was depleted in the presence of JNK inhibitor II (Figure 3A). Thus, this phosphoproteomic analysis provides the first suggestion that JNK, an important signal transduction protein downstream of many RTKs, may facilitate maintenance of undifferentiated hESCs. Moreover, these experiments further demonstrate agreement between phosphoproteomic and cellular analyses in hESCs. DISCUSSION Analysis of molecular mechanisms underlying hESC properties is essential for optimal use of these cells. Complementing previous analyses of promoters, transcripts, and protein expression, our phosphoproteomic analysis suggests that multiple protein phosphorylation events participate in control of hESC Cell Stem Cell Human ESC Phosphoproteomics fate. Application of MDLC-MS/MS-based phosphoproteomics to pluripotent cells may represent an important tool for stem cell biologists. While this study focused on its use for hESCs, one can envision its application to induced pluripotent somatic cells and other somatic stem cells. Our phosphoproteomic analyses identified proteins potentially participating in self-renewal or differentiation of hESCs and focused attention on pathways heretofore underappreciated and underexplored. Transcription regulators, including epigenetic and transcription factors, and kinases contained many phosphorylated members, suggesting that these proteins may be key determinants of hESC fate decisions. Although a variety of proteins have been implicated in hESC self-renewal, some of their functions have been unclear. The identified phosphorylation sites, some on central signaling proteins, expand the knowledge of protein phosphorylation in hESCs. We also identified many proteins whose potential functions in hESCs had not been identified previously. In other words, phosphoproteomic analyses may provide guidance for systematic, rather than solely serendipitous or overly broad-based, approaches in future studies of self-renewal and differentiation of pluripotent cells. Phosphoproteomic analyses identified proteins favoring an undifferentiated or differentiated state of hESCs. For example, phosphorylation of proteins in the JNK pathway was identified, and our cellular follow-up experiments, which are atypical of molecular profiling studies, suggested that inhibition of JNK leads to differentiation of hESCs. A role of JNK in undifferentiated hESCs has not been reported. The VEGF and PDGF pathways are candidates to favor maintenance of undifferentiated hESCs because inhibitors of their receptors resulted in hESC differentiation. However, the growth factors that were added singly could not replace bFGF. Together, these results suggested that activation of these pathways is necessary but not sufficient to sustain self-renewal of hESCs, consistent with increasing evidence that multiple growth factor-driven pathways act together to maintain undifferentiated hESCs. For example, PDGF-AA complemented a subthreshold concentration of bFGF, shown by long-term maintenance of undifferentiated cultures under feeder-free conditions in CDM. Use of CDM allowed improved knowledge of the composition of the media, rather than use of undefined media in the presence of, or conditioned by, feeder fibroblasts (Yao et al., 2006), so the pathways that were targeted in our cellular assays were more clearly defined. Together, our results expanded the repertoire of pathways that facilitate hESC culture and support the suggestion that multiple signaling inputs are needed to maintain undifferentiated hESCs (Wang et al., 2007). Moreover, phosphoproteomic analyses complement epigenetics, gene expression profiles, and total protein MS to facilitate an improved understanding of hESC fate determination. The functions of most of the phosphorylated proteins in pluripotent cells are unknown and should be evaluated for their influence on stem cell behavior. Application of further advances in proteomic and allied technologies should enhance future studies through improved analysis of protein phosphorylation. As phosphoproteins controlling pluripotent behavior are understood better, methods for developing model systems with stem cells, and potential therapeutic applications may become increasingly clear. EXPERIMENTAL PROCEDURES Cell Culture, Phosphoproteomic Analysis Feeder-free cultures were in Matrigel-coated plates in CM containing 8 ng/ml bFGF (Xu et al., 2001). Differentiation was with 5 mM RA and no added bFGF. In CDM, hESCs were cultured in Matrigel-coated plates in N2/B27-CDM (Yao et al., 2006). Phosphoproteomic analyses used cells from CM. Cells were rinsed with PBS, lysed, and centrifuged, and proteins were precipitated with (NH4)2SO4 and pelleted by centrifugation. Proteins were resuspended in 100 mM NH4HCO3, 8 M urea containing phosphatase inhibitors, reduced, alkylated, digested with trypsin, and peptides desalted. Peptides were separated by SCX, phosphopeptides enriched by desalt-IMAC (Brill et al., 2004; Ficarro et al., 2005), separated by nanoflow HPLC, and analyzed by ESI-MS/MS. MS/MS spectra were matched to amino acid sequences using SEQUEST (Eng et al., 1994). All reported phosphopeptide identifications were manually verified (Bernstein et al., 2008; Brill et al., 2004; Ficarro et al., 2005). Normalized XIC peak areas of some phosphopeptides were quantified. For analyses lacking the identification, MS/MS data were exhaustively searched for the phosphopeptide, which was rarely found via a poor quality MS/MS spectrum, and its XIC peak area was quantified. Phosphoproteins were classified as containing more phosphorylation site identifications in undifferentiated hESCs or differentiated derivatives, or as containing a similar number of phosphorylation site identifications in the two cell populations, as described in the Results. Western Blot Analysis Proteins were run on Bis-Tris gels, transferred to PVDF membranes, blocked, and incubated with antibodies recognizing phosphorylation sites identified by MDLC-MS/MS. Anti-GAPDH was the loading control. Membranes were washed, incubated with fluorophore-conjugated secondary antibodies, washed, imaged, and bands quantified according to the manufacturer (LI-COR). Phosphoprotein Category, Subcellular Location, and Pathway Analysis Ingenuity Pathway Analysis, Metacore, NCBI, Gene Ontology, and peerreviewed literature were used to identify phosphoprotein subcellular location, category, and signaling pathways. Cellular Assays, RT-PCR EGFR, JNK, or PDGFR inhibitors were used. Untreated and vehicle-only controls were included for each experiment. PDGF-AA/bFGF was used in cultures for >15 passages. For immunostaining and DAPI staining, monoclonal mouse anti-OCT4 and anti-SSEA-4 were used. Secondary antibodies were Cy2-conjugated rabbit anti-mouse IgM and Cy3-conjugated rabbit anti-mouse IgG. For RT-PCR, mRNA was isolated and cDNA was synthesized; OCT4, NANOG, and GAPDH were amplified. For FACS, cells were incubated with mouse monoclonal antiSSEA-4 or anti-TRA-1-60 antibodies, washed with PBS, and incubated with Cy3-conjugated rabbit anti-mouse IgG. Details are in the Supplemental Experimental Procedures. ACCESSION NUMBERS All supplemental data are deposited in the PRIDE database (http://www.ebi. ac.uk/pride/) under accession numbers 9253–9257 and 9259–9264. SUPPLEMENTAL DATA Supplemental Data include four figures, Supplemental Experimental Procedures, Supplemental References, and 11 tables and can be found with this article online at http://www.cell.com/cell-stem-cell/supplemental/S19345909(09)00286-0. ACKNOWLEDGMENTS We thank Fang C. Kuan, Andrew Su, Jeff Janes, and Ali Iranli for help with bioinformatics; and Michelle Stettler-Gill, Anthony Boitano, Jacqueline Cell Stem Cell 5, 204–213, August 7, 2009 ª2009 Elsevier Inc. 211 Cell Stem Cell Human ESC Phosphoproteomics Lesperance, and Brandon Nelson for technical assistance. Support was from a postdoctoral fellowship from the California Institute for Regenerative Medicine (CIRM) (K.-B.L.), the Genomics Institute of the Novartis Research Foundation (GNF), and the 1 P20 GM 075059-01. Received: October 23, 2008 Revised: May 7, 2009 Accepted: June 9, 2009 Published: August 6, 2009 REFERENCES Adewumi, O., Aflatoonian, B., Ahrlund-Richter, L., Amit, M., Andrews, P.W., Beighton, G., Bello, P.A., Benvenisty, N., Berry, L.S., Bevan, S., et al. (2007). Characterization of human embryonic stem cell lines by the International Stem Cell Initiative. Nat. Biotechnol. 25, 803–816. Armstrong, L., Hughes, O., Yung, S., Hyslop, L., Stewart, R., Wappler, I., Peters, H., Walter, T., Stojkovic, P., Evans, J., et al. (2006). The role of PI3K/ AKT, MAPK/ERK and NFkappabeta signalling in the maintenance of human embryonic stem cell pluripotency and viability highlighted by transcriptional profiling and functional analysis. Hum. Mol. Genet. 15, 1894–1913. Beattie, G.M., Lopez, A.D., Bucay, N., Hinton, A., Firpo, M.T., King, C.C., and Hayek, A. (2005). Activin A maintains pluripotency of human embryonic stem cells in the absence of feeder layers. Stem Cells 23, 489–495. Belotserkovskaya, R., Oh, S., Bondarenko, V.A., Orphanides, G., Studitsky, V.M., and Reinberg, D. (2003). FACT facilitates transcription-dependent nucleosome alteration. Science 301, 1090–1093. Bendall, S.C., Stewart, M.H., Menendez, P., George, D., Vijayaragavan, K., Werbowetski-Ogilvie, T., Ramos-Mejia, V., Rouleau, A., Yang, J., Bosse, M., et al. (2007). IGF and FGF cooperatively establish the regulatory stem cell niche of pluripotent human cells in vitro. Nature 448, 1015–1021. Bennett, B.L., Sasaki, D.T., Murray, B.W., O’Leary, E.C., Sakata, S.T., Xu, W., Leisten, J.C., Motiwala, A., Pierce, S., Satoh, Y., et al. (2001). SP600125, an anthrapyrazolone inhibitor of Jun N-terminal kinase. Proc. Natl. Acad. Sci. USA 98, 13681–13686. Bernstein, B.E., Mikkelsen, T.S., Xie, X., Kamal, M., Huebert, D.J., Cuff, J., Fry, B., Meissner, A., Wernig, M., Plath, K., et al. (2006). A bivalent chromatin structure marks key developmental genes in embryonic stem cells. Cell 125, 315–326. Bernstein, E., Muratore-Schroeder, T.L., Diaz, R.L., Chow, J.C., Changolkar, L.N., Shabanowitz, J., Heard, E., Pehrson, J.R., Hunt, D.F., and Allis, C.D. (2008). A phosphorylated subpopulation of the histone variant macroH2A1 is excluded from the inactive X chromosome and enriched during mitosis. Proc. Natl. Acad. Sci. USA 105, 1533–1538. Chiang, G.G., and Abraham, R.T. (2005). Phosphorylation of mammalian target of rapamycin (mTOR) at Ser-2448 is mediated by p70S6 kinase. J. Biol. Chem. 280, 25485–25490. Chu, B.Y., Tran, K., Ku, T.K., and Crowe, D.L. (2005). Regulation of ERK1 gene expression by coactivator proteins. Biochem. J. 392, 589–599. Eng, J., McCormack, A., and Yates, J.R., 3rd. (1994). An approach to correlate tandem mass spectral data of peptides with amino acid sequences in a protein database. J Am Soc Mass 5 (Spec), 976–989. Ficarro, S.B., Salomon, A.R., Brill, L.M., Mason, D.E., Stettler-Gill, M., Brock, A., and Peters, E.C. (2005). Automated immobilized metal affinity chromatography/nano-liquid chromatography/electrospray ionization mass spectrometry platform for profiling protein phosphorylation sites. Rapid Commun. Mass Spectrom. 19, 57–71. Gruhler, A., Olsen, J.V., Mohammed, S., Mortensen, P., Faergeman, N.J., Mann, M., and Jensen, O.N. (2005). Quantitative phosphoproteomics applied to the yeast pheromone signaling pathway. Mol. Cell. Proteomics 4, 310–327. Han, Z., Boyle, D.L., Chang, L., Bennett, B., Karin, M., Yang, L., Manning, A.M., and Firestein, G.S. (2001). c-Jun N-terminal kinase is required for metalloproteinase expression and joint destruction in inflammatory arthritis. J. Clin. Invest. 108, 73–81. Holzberg, D., Knight, C.G., Dittrich-Breiholz, O., Schneider, H., Dorrie, A., Hoffmann, E., Resch, K., and Kracht, M. (2003). Disruption of the c-JUN-JNK complex by a cell-permeable peptide containing the c-JUN delta domain induces apoptosis and affects a distinct set of interleukin-1-induced inflammatory genes. J. Biol. Chem. 278, 40213–40223. James, D., Levine, A.J., Besser, D., and Hemmati-Brivanlou, A. (2005). TGFbeta/activin/nodal signaling is necessary for the maintenance of pluripotency in human embryonic stem cells. Development 132, 1273–1282. Kratchmarova, I., Blagoev, B., Haack-Sorensen, M., Kassem, M., and Mann, M. (2005). Mechanism of divergent growth factor effects in mesenchymal stem cell differentiation. Science 308, 1472–1477. Kruger, M., Kratchmarova, I., Blagoev, B., Tseng, Y.H., Kahn, C.R., and Mann, M. (2008). Dissection of the insulin signaling pathway via quantitative phosphoproteomics. Proc. Natl. Acad. Sci. USA 105, 2451–2456. Lee, T.I., Jenner, R.G., Boyer, L.A., Guenther, M.G., Levine, S.S., Kumar, R.M., Chevalier, B., Johnstone, S.E., Cole, M.F., Isono, K., et al. (2006). Control of developmental regulators by Polycomb in human embryonic stem cells. Cell 125, 301–313. Li, S., Braverman, R., Li, H., Vass, W.C., Lowy, D.R., and DeClue, J.E. (2003). Regulation of cell morphology and adhesion by the tuberous sclerosis complex (TSC1/2) gene products in human kidney epithelial cells through increased E-cadherin/beta-catenin activity. Mol. Carcinog. 37, 98–109. Bodenmiller, B., Mueller, L.N., Mueller, M., Domon, B., and Aebersold, R. (2007). Reproducible isolation of distinct, overlapping segments of the phosphoproteome. Nat. Methods 4, 231–237. Li, J., Wang, G., Wang, C., Zhao, Y., Zhang, H., Tan, Z., Song, Z., Ding, M., and Deng, H. (2007). MEK/ERK signaling contributes to the maintenance of human embryonic stem cell self-renewal. Differentiation 75, 299–307. Boyer, L.A., Lee, T.I., Cole, M.F., Johnstone, S.E., Levine, S.S., Zucker, J.P., Guenther, M.G., Kumar, R.M., Murray, H.L., Jenner, R.G., et al. (2005). Core transcriptional regulatory circuitry in human embryonic stem cells. Cell 122, 947–956. Liu, H., Sadygov, R.G., and Yates, J.R., 3rd. (2004). A model for random sampling and estimation of relative protein abundance in shotgun proteomics. Anal. Chem. 76, 4193–4201. Brandenberger, R., Wei, H., Zhang, S., Lei, S., Murage, J., Fisk, G.J., Li, Y., Xu, C., Fang, R., Guegler, K., et al. (2004). Transcriptome characterization elucidates signaling networks that control human ES cell growth and differentiation. Nat. Biotechnol. 22, 707–716. Brill, L.M., Salomon, A.R., Ficarro, S.B., Mukherji, M., Stettler-Gill, M., and Peters, E.C. (2004). Robust phosphoproteomic profiling of tyrosine phosphorylation sites from human T cells using immobilized metal affinity chromatography and tandem mass spectrometry. Anal. Chem. 76, 2763–2772. Cha, T.L., Zhou, B.P., Xia, W., Wu, Y., Yang, C.C., Chen, C.T., Ping, B., Otte, A.P., and Hung, M.C. (2005). Akt-mediated phosphorylation of EZH2 suppresses methylation of lysine 27 in histone H3. Science 310, 306–310. Chambers, I., Colby, D., Robertson, M., Nichols, J., Lee, S., Tweedie, S., and Smith, A. (2003). Functional expression cloning of Nanog, a pluripotency sustaining factor in embryonic stem cells. Cell 113, 643–655. 212 Cell Stem Cell 5, 204–213, August 7, 2009 ª2009 Elsevier Inc. Liu, J., Rivas, F.V., Wohlschlegel, J., Yates, J.R., 3rd, Parker, R., and Hannon, G.J. (2005). A role for the P-body component GW182 in microRNA function. Nat. Cell Biol. 7, 1261–1266. Liu, Y., Denlinger, C.E., Rundall, B.K., Smith, P.W., and Jones, D.R. (2006). Suberoylanilide hydroxamic acid induces Akt-mediated phosphorylation of p300, which promotes acetylation and transcriptional activation of RelA/p65. J. Biol. Chem. 281, 31359–31368. Mann, M., Ong, S.E., Gronborg, M., Steen, H., Jensen, O.N., and Pandey, A. (2002). Analysis of protein phosphorylation using mass spectrometry: deciphering the phosphoproteome. Trends Biotechnol. 20, 261–268. McCool, K.W., Xu, X., Singer, D.B., Murdoch, F.E., and Fritsch, M.K. (2007). The role of histone acetylation in regulating early gene expression patterns during early embryonic stem cell differentiation. J. Biol. Chem. 282, 6696– 6706. Cell Stem Cell Human ESC Phosphoproteomics Mukherji, M., Brill, L.M., Ficarro, S.B., Hampton, G.M., and Schultz, P.G. (2006). A phosphoproteomic analysis of the ErbB2 receptor tyrosine kinase signaling pathways. Biochemistry 45, 15529–15540. Pebay, A., Wong, R.C., Pitson, S.M., Wolvetang, E.J., Peh, G.S., Filipczyk, A., Koh, K.L., Tellis, I., Nguyen, L.T., and Pera, M.F. (2005). Essential roles of sphingosine-1-phosphate and platelet-derived growth factor in the maintenance of human embryonic stem cells. Stem Cells 23, 1541–1548. Peponi, E., Drakos, E., Reyes, G., Leventaki, V., Rassidakis, G.Z., and Medeiros, L.J. (2006). Activation of mammalian target of rapamycin signaling promotes cell cycle progression and protects cells from apoptosis in mantle cell lymphoma. Am. J. Pathol. 169, 2171–2180. Reubinoff, B.E., Pera, M.F., Fong, C.Y., Trounson, A., and Bongso, A. (2000). Embryonic stem cell lines from human blastocysts: somatic differentiation in vitro. Nat. Biotechnol. 18, 399–404. Rosner, M., Hofer, K., Kubista, M., and Hengstschlager, M. (2003). Cell size regulation by the human TSC tumor suppressor proteins depends on PI3K and FKBP38. Oncogene 22, 4786–4798. Sato, N., Sanjuan, I.M., Heke, M., Uchida, M., Naef, F., and Brivanlou, A.H. (2003). Molecular signature of human embryonic stem cells and its comparison with the mouse. Dev. Biol. 260, 404–413. Schlessinger, J. (2000). Cell signaling by receptor tyrosine kinases. Cell 103, 211–225. Shin, M., Yan, C., and Boyd, D. (2002). An inhibitor of c-jun aminoterminal kinase (SP600125) represses c-Jun activation, DNA-binding and PMA-inducible 92-kDa type IV collagenase expression. Biochim. Biophys. Acta 1589, 311–316. Sperger, J.M., Chen, X., Draper, J.S., Antosiewicz, J.E., Chon, C.H., Jones, S.B., Brooks, J.D., Andrews, P.W., Brown, P.O., and Thomson, J.A. (2003). Gene expression patterns in human embryonic stem cells and human pluripotent germ cell tumors. Proc. Natl. Acad. Sci. USA 100, 13350–13355. Takahashi, K., and Yamanaka, S. (2006). Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell 126, 663–676. Takahashi, K., Tanabe, K., Ohnuki, M., Narita, M., Ichisaka, T., Tomoda, K., and Yamanaka, S. (2007). Induction of pluripotent stem cells from adult human fibroblasts by defined factors. Cell 131, 861–872. Tan, B.C., Chien, C.T., Hirose, S., and Lee, S.C. (2006). Functional cooperation between FACT and MCM helicase facilitates initiation of chromatin DNA replication. EMBO J. 25, 3975–3985. Thingholm, T.E., Jensen, O.N., Robinson, P.J., and Larsen, M.R. (2008a). SIMAC (sequential elution from IMAC), a phosphoproteomics strategy for the rapid separation of monophosphorylated from multiply phosphorylated peptides. Mol. Cell. Proteomics 7, 661–671. Thingholm, T.E., Larsen, M.R., Ingrell, C.R., Kassem, M., and Jensen, O.N. (2008b). TiO(2)-based phosphoproteomic analysis of the plasma membrane and the effects of phosphatase inhibitor treatment. J. Proteome Res. 7, 3304–3313. Thomson, J.A., Itskovitz-Eldor, J., Shapiro, S.S., Waknitz, M.A., Swiergiel, J.J., Marshall, V.S., and Jones, J.M. (1998). Embryonic stem cell lines derived from human blastocysts. Science 282, 1145–1147. Van Hoof, D., Passier, R., Ward-Van Oostwaard, D., Pinkse, M.W., Heck, A.J., Mummery, C.L., and Krijgsveld, J. (2006). A quest for human and mouse embryonic stem cell-specific proteins. Mol. Cell. Proteomics 5, 1261–1273. Wang, J., Rao, S., Chu, J., Shen, X., Levasseur, D.N., Theunissen, T.W., and Orkin, S.H. (2006). A protein interaction network for pluripotency of embryonic stem cells. Nature 444, 364–368. Wang, L., Schulz, T.C., Sherrer, E.S., Dauphin, D.S., Shin, S., Nelson, A.M., Ware, C.B., Zhan, M., Song, C.Z., Chen, X., et al. (2007). Self-renewal of human embryonic stem cells requires insulin-like growth factor-1 receptor and ERBB2 receptor signaling. Blood 110, 4111–4119. Washburn, M.P., Ulaszek, R.R., and Yates, J.R., 3rd. (2003). Reproducibility of quantitative proteomic analyses of complex biological mixtures by multidimensional protein identification technology. Anal. Chem. 75, 5054–5061. Western, P., Maldonado-Saldivia, J., van den Bergen, J., Hajkova, P., Saitou, M., Barton, S., and Surani, M.A. (2005). Analysis of Esg1 expression in pluripotent cells and the germline reveals similarities with Oct4 and Sox2 and differences between human pluripotent cell lines. Stem Cells 23, 1436– 1442. Xu, C., Inokuma, M.S., Denham, J., Golds, K., Kundu, P., Gold, J.D., and Carpenter, M.K. (2001). Feeder-free growth of undifferentiated human embryonic stem cells. Nat. Biotechnol. 19, 971–974. Xu, R.H., Peck, R.M., Li, D.S., Feng, X., Ludwig, T., and Thomson, J.A. (2005). Basic FGF and suppression of BMP signaling sustain undifferentiated proliferation of human ES cells. Nat. Methods 2, 185–190. Yao, S., Chen, S., Clark, J., Hao, E., Beattie, G.M., Hayek, A., and Ding, S. (2006). Long-term self-renewal and directed differentiation of human embryonic stem cells in chemically defined conditions. Proc. Natl. Acad. Sci. USA 103, 6907–6912. Yu, J., Vodyanik, M.A., Smuga-Otto, K., Antosiewicz-Bourget, J., Frane, J.L., Tian, S., Nie, J., Jonsdottir, G.A., Ruotti, V., Stewart, R., et al. (2007). Induced pluripotent stem cell lines derived from human somatic cells. Science 318, 1917–1920. Zhang, P., Gao, W.Y., Turner, S., and Ducatman, B.S. (2003). Gleevec (STI-571) inhibits lung cancer cell growth (A549) and potentiates the cisplatin effect in vitro. Mol. Cancer 2, 1. Cell Stem Cell 5, 204–213, August 7, 2009 ª2009 Elsevier Inc. 213 Biomed Microdevices (2009) 11:547–555 DOI 10.1007/s10544-008-9260-x Integrated microfluidic devices for combinatorial cell-based assays Zeta Tak For Yu & Ken-ichiro Kamei & Hiroko Takahashi & Chengyi Jenny Shu & Xiaopu Wang & George Wenfu He & Robert Silverman & Caius G. Radu & Owen N. Witte & Ki-Bum Lee & Hsian-Rong Tseng Published online: 9 January 2009 # Springer Science + Business Media, LLC 2008 Abstract The development of miniaturized cell culture platforms for performing parallel cultures and combinatorial assays is important in cell biology from the single-cell level to the system level. In this paper we developed an integrated microfluidic cell-culture platform, Cell-microChip (CellμChip), for parallel analyses of the effects of microenvironmental cues (i.e., culture scaffolds) on different mammalian cells and their cellular responses to external stimuli. As a model study, we demonstrated the ability of culturing and assaying several mammalian cells, such as NIH 3T3 fibroblast, B16 melanoma and HeLa cell lines, in a parallel way. For functional assays, first we tested drug-induced apoptotic responses from different cell lines. As a second functional assay, we performed “on-chip” transfection of a reporter gene encoding an enhanced green fluorescent protein (EGFP) followed by live-cell imaging of transcriptional activation of cyclooxygenase 2 (Cox-2) expression. Collectively, our CellμChip approach demonstrated the capability to carry out parallel operations and the potential to further integrate advanced functions and applications in the broader space of combinatorial chemistry and biology. Keywords Microfluidic devices . Cell-based assay . Apoptosis . Transfection . Cell culture Electronic supplementary material The online version of this article (doi:10.1007/s10544-008-9260-x) contains supplementary material, which is available to authorized users. Zeta Tak For Yu and Ken-ichiro Kamei contributed equally to this work. Z. T. F. Yu Department of Mechanical and Aerospace Engineering, University of California, Los Angeles, CA 90095, USA C. J. Shu : O. N. Witte Department of Microbiology, Immunology, and Molecular Genetics, University of California, Los Angeles, CA 90095, USA Kamei Takahashi·: X. X. Wang Wang :· G. G.W. W.He He: · Z. T. F. Yu : K.-i. Kamei · :H.H.Takahashi C. G. Radu (*) : H.-R. Tseng (*) Crump Institute for Molecular Imaging, University of California, Los Angeles, CA 90095, USA e-mail: [email protected] e-mail: [email protected] O. N. Witte The Howard Hughes Medical Institute, University of California, Los Angeles, CA 90095, USA R.Silverman Silverman: · Kamei Kamei Wang·: G. G. W. W. He He :· R. Z. T. F. Yu : K.-i. · :X.X.Wang C. G. Radu : O. N. Witte (*) : H.-R. Tseng Department of Molecular and Medical Pharmacology, University of California, Los Angeles, CA 90095, USA e-mail: [email protected] K.-B. Lee (*) Department of Chemistry and Chemical Biology, Institute for Advanced Materials, Devices and Nanotechnology, The Rutgers Stem Cell Research Center, Rutgers, The State University of New Jersey, Piscataway, NJ 08854, USA e-mail: [email protected] DO09260; No of Pages 548 1 Introduction Conventional cell-based experiments are typically performed on a large cell population. Researchers have begun to appreciate that there are local variables associated with the heterogeneous microenvironment in the macroscopic culture setting, which often leads to experimental inconsistency. On the other hand, differences among individual cells are often ignored since the conventional assays (e.g., Weston blots and microarray analysis) are conducted at a collective fashion. As a result, it is challenging to elucidate complex cellular systems and analyze dynamic signaling pathways (Irish et al. 2006) using the conventional experiment systems. To overcome these challenges, it is essential to develop a new technology platform to enable (1) improved control on the cell culture microenvironment, (2) precise cell assays with the single-cell resolution, and (3) sequential and parallel operation by combining those mentioned in (1) and (2). We envision such a technology can be applied for screening drug candidates (Dittrich and Manz 2006; Padron et al. 2000), evaluating biological pathways (Minor 2003), and understanding pharmacological effects (Hill et al. 1998; Umezawa 2005), thus constituting critical technological foundations for a broad spectrum of biomedical research. Microfluidic devices (Auroux et al. 2002; Dittrich and Manz 2006; Dittrich et al. 2006; Reyes et al. 2002) offer a robust analytical tool that allows rapid analysis of cellular responses to external stimuli in a parallel way. Moreover, microfluidics, with its intrinsic advantages of sample/reagent economy, precise control over physical and chemical microenvironments, high throughput, scalability and digital controllability, whose features allow microfluidics to investigate complex and dynamic biological processes at the single-cell level. In addition, microscale cell culture using microfluidics allows investigating the function of microenvironmental cues at the single-cell level. Especially, compared to static microfluidic cell-culture systems, integrated microfluidics allows for the control of adding and/or removing of biochemical cues at specific temporal as well as spatial points. This unique advantage makes it possible to do novel microenvironmental experiments, such as cell–cell interactions and extracellular matrix (ECM)–cell interactions. However, there are several issues associated with the fabrication and control of integrated microfluidic cell culture systems in order to achieve routine cellular assays of mammalian cells. Even though microfluidic cell culture systems have been developed extensively, they can mainly be described into two ways: what kinds of cell culture platforms were used, and what types of cells and applications were tested. Many different cell culture platforms have been developed, including two-dimensional, three-dimensional (Cartmell et al. 2003; Toh et al. 2007) and co-culture platforms Biomed Microdevices (2009) 11:547–555 (Sin et al. 2004). Those microfluidic approaches enabled to culture and assay many different cells, such as liver (Kane et al. 2006; Sin et al. 2004; Zhang et al. 2008), muscle (Tourovskaia et al. 2005; Tourovskaia et al. 2006), neural (Millet et al. 2007; Park et al. 2006), and stem cells (Chung et al. 2005; Gomez-Sjoberg et al. 2007; Kim et al. 2006). Despite these recent advances, several critical questions and several challenges should be addressed more explicitly, and optimized to fully achieve the potential of microfluidic cell culture and assays. It requires the abilities: (1) to test a robust and flexible fluidic configuration (e.g., flow through vs. circulation) for cell culture and (2) to fabricate microfluidic network capable of performing sequential and parallel operations such as parallel culture of multiple cell types and subsequent phenotypic and functional assays in the same microfluidic chip. In this context, incorporation of isolation valves (Unger et al. 2000) and peristaltic pumps (Chou et al. 2001) should allow individual addressability and digital controllability (Lee et al. 2005; Wang et al. 2006) of each cell culture chamber embedded on a microfluidic device, which in turn should enable complex manipulations of the culture microenvironments as well as multiple analytical measurements. Over the years, there have been a variety of microfluidic chips directed for functional biological assays. These include differentiation of cell through different flow rates (Gu et al. 2004), fully automated cell culture system by two-layer PDMS chips (Gomez-Sjoberg et al. 2007), optimization of drug cocktail to regulate cell activities through closed-loop control algorithm and microfluidic platform (Wong et al. 2008), and modelling of galactose pathway in an alternating culture environment (Bennett et al. 2008). All these devices or systems have provided additional modules and thus are superior to the conventional setting on conducting biological research or routine operations. In this paper, we describe the design and operation of polydimethylsiloxane (PDMS)-based Cell-microChip (CellμChip, Fig. 1) envisioned as a digitally controlled platform for performing parallel cell culture and sequential cellular assays. The potential of the Cell-μChip to support the optimal culture and assays of human and murine cell lines was demonstrated. To determine the optimal culture conditions, these cells were cultured in six cell culture chambers embedded on a Cell-μChip under two different medium supply modes in parallel (i.e., circulation and direct feeding). The growth and viability of the microcultures were monitored and quantified in real time in the Cell-μChip using an integrated CCD camera. Sequential staining with acridine orange (AO) and propidium iodide (PI) (Hudson et al. 1969; Traganos et al. 1977) allowed the identification of viable and dead cells, respectively. To Biomed Microdevices (2009) 11:547–555 Fig. 1 (a) Schematic representation of an integrated CellmicroChip (Cell-μChip) for performing multiple cell culture and assays under a digitally controlled interface. Three pairs of parallel-oriented cell culture chambers are incorporated in a Cell-μChip, where multiple cell types can be cultured under two different modes of medium supply, i.e., circulatory (channels i, iii and v) and direct feeding (channels ii, iv and vi). The operation of this microchip is controlled by pressure driven valves with their delegated functions indicated by their colors: red for regular valve (for isolation and gating) and yellow for pumping valve (for fluid transport and circulation). (b) Optical image of the actual device. The microchip was loaded with various colors of food dyes to enhance the visualization of different parts in the entire system: red and yellow as in (a); blue indicates the flow channel and the medium reservoir 549 (a) Outlet port Inlet port i Medium tubing ii iii iv v vi (1) Cell culture chambers i vi: 100– 150 nL (2) Metering pump Medium reservoir 5–15µL Collection Regular valve Valve for pump (b) demonstrate the ability to perform “on chip” functional assays, we analyzed drug-induced apoptotic responses. Furthermore, we showed that the Cell-μChip is amenable to complex sequential operations such as genetic manipulation and monitoring of transcriptional activation of gene expression. 2 mm 2 Experimental outside the culture chambers were removed by washing with fresh DMEM. The Cell-μChip was placed in an incubator for 6 h. The pump was turned on to introduce the DMEM in a circulating or feed through fashion in the respective culture chambers. The flow rate of the medium was controlled in the range of 0.1–4 nL s−1. Cell growth was monitored by collecting bright field micrographs of cells inside the Cell-μChip at 12-h intervals. 2.1 Microfluidic cell culture 2.2 Immunoassay for fibronectin Using the integrated valves and pumps, bovine fibronectin (FN) (Sigma) solution (1.0 mg mL−1) filled in Teflon tubing was introduced into the six cell culture chambers of the Cell-μChip. The Cell-μChip was kept at 37°C for 30 min for FN coating. DMEM was then introduced into the cell culture chambers to extrude the FN solution. The medium reservoirs and medium tubings were then dead-end filled with DMEM at 10 psi for 60 min (Song et al. 2008). Then, individual cell suspensions (NIH 3T3, HeLa and B16) with 2× 106 cells mL−1 were sequentially introduced by gravitation into the cell culture chambers. After cell loading, cells located To confirm FN coating efficiency in cell culture chambers, immunoassay for FN was performed. After FN coating in a Cell-μChip, blocking solution containing 5% BSA and 0.1% N-dodecyl-β-D-maltoside (DDM) (Pierce) was introduced in a Cell-μChip, and then kept at room temperature for 1 h. Mouse anti-FN (BD Biosciences) was loaded into cell culture chambers, and incubated at room temperature for 2 h. Excess antibody in cell culture chambers were rinsed with PBS with 0.1% Tween 20 (PBS-T) twice. Then secondary goat anti-mouse IgG conjugated with Alexa555 (Invitrogen) were introduced in 550 cell culture chambers and incubated at room temperature for 1 h. After washing with PBS-T twice, fluorescent intensity was measured with a fluorescent microscope, and quantified with MetaMorph. 2.3 AO/PI fluorescence staining After culturing the cells in the Cell-μChip for 4 days, a solution composed of DMEM, AO and PI in a ratio of 10:1:1 was introduced into the cell culture chambers. After 5 min incubation at 37°C, the cells were imaged under a fluorescence microscope. Biomed Microdevices (2009) 11:547–555 3 Results and discussion 3.1 Design of the Cell-μChip The PDMS-based Cell-μChips (Fig. 1) were fabricated by multilayer soft lithography approach (see supplementary information) (Unger et al. 2000; Xia and Whitesides 1998). It is important to note that the biocompatible and gaspermeable properties (Kim et al. 2007; Korin et al. 2007) of PDMS matrices help to retain proper physiological (a) 30 min 2h 1 mg mL-1 NIH 3T3, HeLa and B16 cells were cultured in the cell culture chambers of the Cell-μChip for 24 h. Staurosporine (0, 0.1, 1 and 10 μM as final concentrations) or actinomycin D (0, 0.1, 1 and 10 μM as final concentrations) in DMEM culture medium were loaded into the cell culture chambers. After 2 h incubation, the medium in all cell culture chambers was replaced by the MitoTracker Red dye (Invitrogen) for staining of viable cells. The CellμChip was incubated for 30 min at 5% CO2, 37°C. Sequentially, the MitoTracker Red solution was replaced by a solution containing 100 μL of Annexin binding buffer and 5 μL of Annexin V-Alexa488 for staining the apoptotic cells. Following 15 min incubation at RT, cell culture chambers were flushed with Annexin binding buffer and the cells were imaged using a fluorescence microscope. 0.01 mg mL-1 2.4 Apoptosis assay (b) NIH 3T3 cells were loaded into the six culture chambers of the Cell-μChip and were allowed to settle for 24 h. Cells were transfected with the pCox2-EGFP plasmid (a kind gift from Prof. Harvey R. Herschman, UCLA), encoding an enhanced green fluorescent protein (EGFP) under a murine Cox-2 promoter (Liang et al. 2004). The transfection mixture containing plasmid (0.5 μg), medium (30 μL) and transfection reagent (2.5 μL, Superfect reagent, Qiagen) was incubated at RT for 10 min. The mixture was further diluted with 150 μL of DMEM and loaded into all culture chambers of the CellμChip. After 3-h incubation at 5% CO2, 37°C, the mixture was replaced with serum-free DMEM and cells were incubated for an additional day. To activate the Cox-2 promoter the culture media in three out of six chambers was replaced with media containing the induction agent TPA (50 ng mL−1). Following 7 h incubation at 5% CO2, 37°C, EGFP expression was imaged using a fluorescence microscope. Fluorescence Intensity (A.U.) 2.5 On-chip transfection and reporter gene imaging 4000 3500 30 min 2h 3000 2500 2000 1500 1000 500 0 0.01 mg mL-1 1 mg mL-1 Fibronectin Concentration Fig. 2 Fibronectin coating efficiency on the PDMS surface in a CellμChip determined by immunofluorescence assay. (a) Fluorescence images of immunostained FN on the PDMS surface. (b) Quantitative analysis of FN coating efficiency determined with fluorescence images shown in (a) Biomed Microdevices (2009) 11:547–555 551 (a) Day 0 Day 1 Day 2 Day 7 Day 8 200 µm Day 3 matic valves and peristaltic pumps. This design enabled to digitally control sequential operations. The chip consisted of three identical pairs of parallel-oriented culture chambers with identical dimensions (3×0.5×0.1 mm3, corresponding to a volume of 150 nL). To allow synchronized pumping, six internally connected peristaltic pumps were incorporated at the ends of the six cell culture chambers. Each pair of culture chambers was configured to have two types of medium supplies: one allowing media recirculation through the culture chamber for cellular auto-conditioning and the other enabling direct feeding of cells with fresh media. The size of medium reservoir could accommodate about 10 μL of culture media, a volume sufficient to sustain continuous on-chip cell culture for 8 days. In contrast, supply Teflon tubings were utilized to (a) (b) (b) (c) (c) 200 µm NIH 3T3 HeLa B16 (d) 20 (e) 100 µm Fig. 3 Long term culture of NIH 3T3 cells in a Cell-μChip. (a) Time lapse images of the NIH 3T3 cell proliferation in the microchip in a duration of 8 days. (b)–(e) Dead (PI)/live (AO) staining of NIH 3T3 cells cultured in a Cell-μChip for 4 days. (b) A bright field micrograph of NIH 3T3 cell morphologies. (c) A green fluorescence micrograph of the live-stained cells. (d) A red fluorescence micrograph of the dead-stained cells. (e) A merged fluorescence image of (b) and (c) conditions for a wide range of mammalian cells suitable for different screening applications. A fluidic network for individually addressable cell culture chambers and solution/reagent transport was integrated with embedded pneu- Fold Changes of Cell Number NIH 3T3 (d) HeLa 15 with medium re-circulation B16 NIH 3T3 with direct medium feeding 10 5 0 0 2 4 6 Days Fig. 4 Demonstration of six parallel cell cultures in a closely related microenvironment. After 3 days culture, the cell morphologies were shown in (a) NIH 3T3, (b) HeLa and (c) B16. (d) Growth curves of chip-cultured NIH 3T3, HeLa and B16 cells were quantified by monitoring the number of cells inside the cell culture chambers over time. After 5 days culture, we could not count cell number precisely due to cell confluence and multiple cell layers in the cell culture chambers 552 Biomed Microdevices (2009) 11:547–555 Staurosporine (µ µM) 1 10 0.1 1 10 Apoptosis NIH3T3 Living Apoptosis HeLa Living Apoptosis B16 Living 0.1 ActinomycinD (µM) Fig. 5 Multiparametric apoptosis assays performed in the CellμChips. NIH 3T3, HeLa and B16 cells were treated with either staurosporine or actinomycin D to induce apoptosis. Apoptotic cells were stained with Annexin V conjugated with Alexa488 (green), and living cells were stained with MitoTracker (red) store and deliver fresh culture media. This design allowed us to perform six cell culture experiments in a closely related microenvironment under two different culture media supply modes. searching optimal ECMs, FN is effectively coated on the PDMS surface in our Cell-μChip (Fig. 2). Several different conditions, such as different FN concentrations and incubation time, were tested to optimize the FN coating condition. We confirm the efficiency and homogeneity of FN coating on PDMS by immunoassay. As a result, 1 mg mL−1 of FN for 30 min incubation at 37°C is the optimal condition in a Cell-μChip. 3.2 Surface modification with fibronectin We initially used bare PDMS surface to seed cells, however, we found that cells either could not attach to the surface well or they detached so easily when new fresh media were supplied. We reasoned that this problem was due to the inherent hydrophobicity of PDMS materials. Thus, we tested to use ECMs in order to make the surface hydrophilic and biocompatible for cell adhesion. In our 3.3 Cell culture in the Cell-μChip We initially demonstrated to culture NIH 3T3 cells in all culture chambers of a Cell-μChip (Fig. 3, Supplementary Information and Fig. S1). Generally, following loading on Biomed Microdevices (2009) 11:547–555 (a) (b) 20µm (c) (d) 553 numbers increased significantly for all three cell lines. To examine cell viability in the three reservoir-attached cell culture chambers, we performed sequential PI/AO staining. As a result, most of the cells are viable, and few dead cells were observed (data not shown). To quantify the culture parameters in the Cell-μChip we sought to determine growth rate for the three cell types (Fig. 4(d)). We observed that the curves reflected a linear growth phase until the cells reached confluence followed by a stationary phase. Comparing the cell growth with medium-recirculation or direct feeding setting, first we used NIH 3T3 cells and continuously monitored them. There was no clear difference in NIH 3T3 cell growth between the two settings (Fig. 4(d)). In the case of HeLa and B16 cells, we obtained similar results as NIH 3T3 cells (data not shown). This result indicates that a Cell-μChip serves a platform to perform multiple cell culture in a single device. 3.5 On-chip apoptosis assay Fig. 6 On-chip transfection and EGFP induction in NIH 3T3 cells. The plasmid vector which encodes EGFP driven by a Cox-2 promoter was transfected with NIH 3T3 cells. (a) Bright field and (b) fluorescence images of NIH 3T3 cells stimulated with TPA for 7 h. (c) Bright field and (d) fluorescence images of NIH 3T3 cells without stimulation the chip, it took about 1 h for cells to attach and spread onto the FN-coated culture chamber surfaces. NIH 3T3 cells grew to confluence, and the whole microchannel was fully occupied at day 8. During this period, we continuously monitored cell growth inside the Cell-μChip using a CCD camera (see ESM Movie 1). Interestingly, cells in a CellμChip grew on the bottom as well as ceiling. This phenomenon is not allowed under conventional culture conditions using Petri dishes. This observation indicated that a Cell-μChip could provide unique and intrinsic characteristics of cell culture manipulations. To determine cell viability we performed AO/PI fluorescence staining (Fig. 3(b)–(e)). AO staining (Fig. 3(c)) indicated that the majority of cells in the chamber were viable. PI staining (Fig. 3(d)) showed a small percentage of dead cells. 3.4 Parallel culture of multiple cell lines in a single Cell-μChip To demonstrate this concept, we cultured NIH 3T3, HeLa and B16 cells in a single Cell-μChip (Fig. 4). NIH 3T3, HeLa and B16 cells were sequentially loaded into the pairs of culture chambers; i–ii, iii–iv and v–vi accordingly. After cell adhesion, the six peristaltic pumps were turned on to feed cells with the medium, in either recirculation or direct feeding setting. The results were consistent with those observed for the parallel culture of NIH 3T3 cells. Cell Apoptosis is not only fundamentally involved in developing cells and maintaining tissue homeostasis, but also can be closely related to several diseases including cancer, autoimmune, and neurodegeneration. Even though, extensive studies have been reported to dissect apoptosis’s molecular basis, more system level as well as single-cell level analysis by using integrated microfluidics would bring new insights for the underlying mechanisms of these biological processes. To demonstrate the capability of “on chip” functional analyses with the Cell-μChip, we performed a drug-induced apoptosis assay (Fig. 5). We used two kinds of apoptosis inducers, staurosporine (ST) (Rajotte et al. 1992; Tafani et al. 2001; Wang et al. 1996) and actinomycin D (AD) (Martin et al. 1990). NIH 3T3, HeLa and B16 cells were treated with apoptosis inducers at four concentrations (0, 0.1, 1 and 10 μM) in the CellμChips. ST or AD treated cells increased apoptotic cell population with a dose-dependent manner of apoptosis inducers. Although cell viability among untreated and treated cells appeared to be the same based on the MitoTracker staining, the cells underwent the apoptosis process at various speeds according to the drug concentration, as demonstrated by the Annexin V staining. We conclude that the Cell-μChip served as a platform to perform multiparametric functional assays. 3.6 On-chip transfection and monitoring of reporter gene expression Plasmid DNA transfection is one of the common methods to manipulate gene expression in mammalian cells. However, the detailed conditions of optimal transfection for different cell lines are variable. Combinatorial approach 554 from parallel microfluidic cellular assays can help to identify the optimal condition. To demonstrate the feasibility of performing the other assay in the Cell-μChip, gene transfection experiments were carried out (Fig. 6). For proof-of-concept, we used a plasmid vector with an enhanced green fluorescent protein (EGFP) driven by a cyclooxygenase-2 (Cox-2) promoter. Therefore, EGFP expression serves as a reporter of Cox-2 transcription. The basal expression level of Cox-2 in NIH 3T3 cells is very low. Tetradecanoylphorbol acetate (TPA) can activate Cox-2 transcription. EGFP expression in these cells was monitored under a fluorescence microscope. As shown in Fig. 6, even though weak EGFP signals were detected in the negative control chambers, strong EGFP signals were observed in TPA-induced transfected cells in the corresponding cell culture chambers. 4 Conclusion In summary, we developed a fully digitally controlled microfluidic cell culture and assay platform that could support parallel cell culture and sequential cell assays. Through the integration of isolation valves, murine and human cells lines could be cultured in different cell culture chambers and tested for different conditions of cell culture and assays in a single Cell-μChip. Real-time monitoring of cell morphology and numbers, viability assay, apoptosis assay and transfection to monitor expression of a reporter gene vector were also performed using the same platform. Our results indicate that intrinsic advantages of microfluidic devices enable the execution of complicated and integrated biological operations in stand-alone devices such as the Cell-μChip. We envision that this platform will be further integrated with advanced functions and utilities for more sophisticated cell culture applications. Acknowledgements This research was supported by the NIH NanoSystems Biology Cancer Center, the DOE-UCLA Institute of Molecular Medicine and the NIH-UCLA Center for In Vivo Imaging in Cancer Biology and Siemens Medical Solutions USA Inc. We thank Stephanie M. Shelly, Dan Rohle, Shirley Quan and Mireille Riedinger for the outstanding technical support with conventional cell culture conditions. ONW is an Investigator of the Howard Hughes Medical Institute. CGR was supported by a Developmental Project Award (ICMIC, NIH/ NCI grant no. CA08630). C.J.S. was supported by a National Institutes of Health (NIH) Research Training in Pharmacological Sciences training grant PHS T32 CM008652. References P.A. Auroux, D. Iossifidis, D.R. Reyes, A. Manz, Micro total analysis systems. 2. Analytical standard operations and applications Anal. Chem. 74, 2637–2652 (2002). doi:10.1021/ac020239t Biomed Microdevices (2009) 11:547–555 M.R. Bennett, W.L. Pang, N.A. Ostroff, B.L. Baumgartner, S. Nayak, L.S. Tsimring, J. Hasty, Metabolic gene regulation in a dynamically changing environment Nature 454, 1119–1122 (2008). doi:10.1038/nature07211 S.H. Cartmell, B.D. Porter, A.J. Garcia, R.E. Guldberg, Effects of medium perfusion rate on cell-seeded three-dimensional bone constructs in vitro Tissue Eng. 9, 1197–1203 (2003). doi:10.1089/10763270360728107 H.P. Chou, M.A. Unger, S.R. Quake, A microfabricated rotary pump Biomed. Microdevices 3(4), 323–330 (2001). doi:10.1023/ A:1012412916446 B.G. Chung, L.A. Flanagan, S.W. Rhee, P.H. Schwartz, A.P. Lee, E.S. Monuki, N.L. Jeon, Human neural stem cell growth and differentiation in a gradient-generating microfluidic device Lab Chip 5, 401–406 (2005). doi:10.1039/b417651k P.S. Dittrich, A. Manz, Lab-on-a-chip: microfluidics in drug discovery Nat. Rev. Drug Discov. 5, 210–218 (2006). doi:10.1038/nrd1985 P.S. Dittrich, K. Tachikawa, A. Manz, Micro total analysis systems. Latest advancements and trends Anal. Chem. 78, 3887–3907 (2006). doi:10.1021/ac0605602 R. Gomez-Sjoberg, A.A. Leyrat, D.M. Pirone, C.S. Chen, S.R. Quake, Versatile, fully automated, microfluidic cell culture system Anal. Chem. 79, 8557–8563 (2007). doi:10.1021/ac071311w W. Gu, X. Zhu, N. Futai, B.S. Cho, S. Takayama, Computerized microfluidic cell culture using elastomeric channels and Braille displays Proc. Natl. Acad. Sci. U. S. A. 101, 15861–15866 (2004). doi:10.1073/pnas.0404353101 D.C. Hill, S.K. Wrigley, L.J. Nisbet, Novel screen methodologies for identification of new microbial metabolites with pharmacological activity Adv. Biochem. Eng. Biotechnol. 59, 73–121 (1998). doi:10.1007/BFb0102297 B. Hudson, W.B. Upholt, J. Devinny, J. Vinograd, The use of an ethidium analogue in the dye-buoyant density procedure for the isolation of closed circular DNA: the variation of the superhelix density of mitochondrial DNA Proc. Natl. Acad. Sci. U. S. A. 62, 813–820 (1969). doi:10.1073/pnas.62.3.813 J.M. Irish, N. Kotecha, G.P. Nolan, Mapping normal and cancer cell signalling networks: towards single-cell proteomics Nat. Rev. Cancer 6, 146–155 (2006). doi:10.1038/nrc1804 B.J. Kane, M.J. Zinner, M.L. Yarmush, M. Toner, Liver-specific functional studies in a microfluidic array of primary mammalian hepatocytes Anal. Chem. 78, 4291–4298 (2006). doi:10.1021/ ac051856v L. Kim, M.D. Vahey, H.Y. Lee, J. Voldman, Microfluidic arrays for logarithmically perfused embryonic stem cell culture Lab Chip 6, 394–406 (2006). doi:10.1039/b511718f L. Kim, Y.C. Toh, J. Voldman, H. Yu, A practical guide to microfluidic perfusion culture of adherent mammalian cells Lab Chip 7, 681–694 (2007). doi:10.1039/b704602b N. Korin, A. Bransky, U. Dinnar, S. Levenberg, A parametric study of human fibroblasts culture in a microchannel bioreactor Lab Chip 7, 611–617 (2007). doi:10.1039/b702392h C.C. Lee, G. Sui, A. Elizarov, C.J. Shu, Y.S. Shin, A.N. Dooley, J. Huang, A. Daridon, P. Wyatt, D. Stout, H.C. Kolb, O.N. Witte, N. Satyamurthy, J.R. Heath, M.E. Phelps, S.R. Quake, H.R. Tseng, Multistep synthesis of a radiolabeled imaging probe using integrated microfluidics Science 310, 1793–1796 (2005). doi:10.1126/science.1118919 Q. Liang, M. Yamamoto, D.T. Curiel, H.R. Herschman, Noninvasive imaging of transcriptionally restricted transgene expression following intratumoral injection of an adenovirus in which the COX-2 promoter drives a reporter gene Mol. Imaging Biol. 6, 395–404 (2004). doi:10.1016/j.mibio.2004.09.002 S.J. Martin, S.V. Lennon, A.M. Bonham, T.G. Cotter, Induction of apoptosis (programmed cell death) in human leukemic HL-60 Biomed Microdevices (2009) 11:547–555 cells by inhibition of RNA or protein synthesis J. Immunol. 145, 1859–1867 (1990) L.J. Millet, M.E. Stewart, J.V. Sweedler, R.G. Nuzzo, M.U. Gillette, Microfluidic devices for culturing primary mammalian neurons at low densities Lab Chip 7, 987–994 (2007). doi:10.1039/b705266a L.K. Minor, Assays to measure the activation of membrane tyrosine kinase receptors: focus on cellular methods Curr. Opin. Drug Discov. Devel. 6, 760–765 (2003) J.M. Padron, C.L. van der Wilt, K. Smid, E. Smitskamp-Wilms, H.H. Backus, P.E. Pizao, G. Giaccone, G.J. Peters, The multilayered postconfluent cell culture as a model for drug screening Crit. Rev. Oncol. Hematol. 36, 141–157 (2000). doi:10.1016/S10408428(00)00083-4 J.W. Park, B. Vahidi, A.M. Taylor, S.W. Rhee, N.L. Jeon, Microfluidic culture platform for neuroscience research Nat. Protoc. 1, 2128– 2136 (2006). doi:10.1038/nprot.2006.316 D. Rajotte, P. Haddad, A. Haman, E.J. Cragoe Jr., T. Hoang, Role of protein kinase C and the Na+/H+ antiporter in suppression of apoptosis by granulocyte macrophage colony-stimulating factor and interleukin-3 J. Biol. Chem. 267, 9980–9987 (1992) D.R. Reyes, D. Iossifidis, P.A. Auroux, A. Manz, Micro total analysis systems. 1. Introduction, theory, and technology Anal. Chem. 74, 2623–2636 (2002). doi:10.1021/ac0202435 A. Sin, K.C. Chin, M.F. Jamil, Y. Kostov, G. Rao, M.L. Shuler, The design and fabrication of three-chamber microscale cell culture analog devices with integrated dissolved oxygen sensors Biotechnol. Prog. 20, 338–345 (2004). doi:10.1021/bp034077d W.H. Song, J. Kwan, G.V. Kaigala, V.N. Hoang, C.J. Backhouse, Readily integrated, electrically controlled microvalves J. Micromechanics Microengineering 8, 1071–1078 (2008) M. Tafani, D.A. Minchenko, A. Serroni, J.L. Farber, Induction of the mitochondrial permeability transition mediates the killing of HeLa cells by staurosporine Cancer Res. 61, 2459–2466 (2001) Y.C. Toh, C. Zhang, J. Zhang, Y.M. Khong, S. Chang, V.D. Samper, D. van Noort, D.W. Hutmacher, H. Yu, A novel 3D mammalian cell perfusion-culture system in microfluidic channels Lab Chip 7, 302–309 (2007). doi:10.1039/b614872g 555 A. Tourovskaia, X. Figueroa-Masot, A. Folch, Differentiation-on-achip: a microfluidic platform for long-term cell culture studies Lab Chip 5, 14–19 (2005). doi:10.1039/b405719h A. Tourovskaia, X. Figueroa-Masot, A. Folch, Long-term microfluidic cultures of myotube microarrays for high-throughput focal stimulation Nat. Protoc. 1, 1092–1104 (2006). doi:10.1038/ nprot.2006.123 F. Traganos, Z. Darzynkiewicz, T. Sharpless, M.R. Melamed, Simultaneous staining of ribonucleic and deoxyribonucleic acids in unfixed cells using acridine orange in a flow cytofluorometric system J. Histochem. Cytochem. 25, 46–56 (1977) Y. Umezawa, Genetically encoded optical probes for imaging cellular signaling pathways Biosens. Bioelectron. 20, 2504–2511 (2005). doi:10.1016/j.bios.2004.10.015 M.A. Unger, H.P. Chou, T. Thorsen, A. Scherer, S.R. Quake, Monolithic microfabricated valves and pumps by multilayer soft lithography Science 288, 113–116 (2000). doi:10.1126/science. 288.5463.113 X. Wang, N.G. Zelenski, J. Yang, J. Sakai, M.S. Brown, J.L. Goldstein, Cleavage of sterol regulatory element binding proteins (SREBPs) by CPP32 during apoptosis EMBO J. 15, 1012–1020 (1996) J. Wang, G. Sui, V.P. Mocharla, R.J. Lin, M.E. Phelps, H.C. Kolb, H. R. Tseng, Integrated microfluidics for parallel screening of an in situ click chemistry library Angew. Chem. Int. Ed. Engl. 45, 5276–5281 (2006). doi:10.1002/anie.200601677 P.K. Wong, F. Yu, A. Shahangian, G. Cheng, R. Sun, C.M. Ho, Closed-loop control of cellular functions using combinatory drugs guided by a stochastic search algorithm Proc. Natl. Acad. Sci. U. S. A. 105, 5105–5110 (2008). doi:10.1073/pnas. 0800823105 Y.N. Xia, G.M. Whitesides, Soft lithography Annu. Rev. Mater. Sci. 28, 153–184 (1998). doi:10.1146/annurev.matsci.28.1.153 M.Y. Zhang, P.J. Lee, P.J. Hung, T. Johnson, L.P. Lee, M.R. Mofrad, Microfluidic environment for high density hepatocyte culture Biomed. Microdevices 10, 117–121 (2008). doi:10.1007/s10544007-9116-9 COMMUNICATION www.rsc.org/loc | Lab on a Chip Microfluidic image cytometry for quantitative single-cell profiling of human pluripotent stem cells in chemically defined conditions† Ken-ichiro Kamei,‡*abcde Minori Ohashi,‡abcde Eric Gschweng,ef Quinn Ho,abcdeg Jane Suh,abcde Jinghua Tang,e Zeta Tak For Yu,abcde Amander T. Clark,eh April D. Pyle,ef Michael A. Teitell,ei Ki-Bum. Lee,j Owen N. Witteacefk and Hsian-Rong Tseng*abcde Received 4th November 2009, Accepted 11th February 2010 First published as an Advance Article on the web 16th March 2010 DOI: 10.1039/b922884e Microfluidic image cytometry (MIC) has been developed to study phenotypes of various hPSC lines by screening several chemically defined serum/feeder-free conditions. A chemically defined hPSC culture was established using 20 ng mL1 of bFGF on 20 mg mL1 of Matrigel to grow hPSCs over a week in an undifferentiated state. Following hPSC culture, we conducted quantitative MIC to perform a single cell profiling of simultaneously detected protein expression (OCT4 and SSEA1). Using clustering analysis, we were able to systematically compare the characteristics of various hPSC lines in different conditions. Human pluripotent stem cells (hPSCs) such as human embryonic stem cells (hESCs)1 and human induced pluripotent stem cells (hiPSCs)2–6 exhibit unique characteristics and may provide great opportunities for cell-based therapy and regenerative medicine. These characteristics include unlimited propagation capacity in the undifferentiated stage with a normal euploid karyotype and the ability to differentiate into all cell types in the human body. a Department of Molecular & Medical Pharmacology, University of California, Los Angeles, CA, 90095, USA. E-mail: kkamei@mednet. ucla.edu; [email protected] b Crump Institute for Molecular Imaging, David Geffen School of Medicine, University of California, Los Angeles, CA, 90095, USA c California NanoSystems Institute, University of California, Los Angeles, CA, 90095, USA d Institute for Molecular Medicine, University of California, Los Angeles, CA, 90095, USA e Eli and Edythe Broad Center of Regenerative Medicine and Stem Cell Research, University of California, Los Angeles, CA, 90095, USA f Department of Microbiology, Immunology and Molecular Genetics, University of California, Los Angeles, CA, 90095, USA g Department of Biological Chemistry, University of California, Los Angeles, CA, 90095, USA h Department of Molecular Cell and Developmental Biology, Los Angeles, CA, 90095, USA i Department of Pathology and Laboratory Medicine, Los Angeles, CA, 90095, USA j Department of Chemistry & Chemical Biology, Institute for Advanced Materials, Devices and Nanotechnology, Rutgers Stem Cell Research Center, Rutgers, The State University of New Jersey, Piscataway, NJ, 08854, USA k Howard Hughes Medical Institute, David Geffen School of Medicine, University of California, Los Angeles, CA, 90095, USA † Electronic supplementary information (ESI) available: Fabrication of microfluidic hPSC array, generation of hiPSC (i.e., hiPSA1 and hiPSB2), microscopy settings and image processing are available. See DOI: 10.1039/b922884e ‡ These authors contributed equally to this work. This journal is ª The Royal Society of Chemistry 2010 Typically, hPSC culture conditions contain serum such as KnockOut serum replacement (KSR) and feeders such as mouse embryonic fibroblasts (MEFs). Although these conditions can successfully maintain pluripotency of hPSCs, these animal products could cause xenogenic contamination and immunorejection in patients after transplantation of hPSCs, posing a major challenge to the use of hPSCs in cell-based therapy applications. Additionally, these factors are undefined and some are proprietarily formulated, forming an obstacle in being able to systematically study the regulation of stem cell biology. Therefore, it is essential to develop serum/ feeder-free culture methods for hPSCs in order to define culture elements and later apply them to effective therapeutic use. Currently, there is an ongoing trend towards establishing chemically defined conditions for hPSC culture. Several chemically defined culture systems have been introduced to maintain hESCs in combination with (i) growth factors/cytokines (e.g., basic fibroblast growth factor (bFGF), nodal, transforming growth factor-b1 (TGF-b1), activin A and insulin-like growth factor-1 (IGF-1) analog (heregulin1b)) and (ii) supplements (e.g., GABA, pipecolic acid and lithium chloride,7,8 and N2/B279) on extracellular matrices (ECM) such as Matrigel or other ECM components. A chemically defined culture system with serum/feeder-free conditions is ideal since it excludes the unknown factors and enhances the reproducibility and robustness of hPSC propagation. Thus, to facilitate practical applications involving hPSCs, optimal chemically defined culture conditions must be established that will not only maintain phenotypically and karyotypically stable cells for extended periods but will also retain the ability for directed and reproducible differentiation. Until now, even with the conventional culturing methods, controlling hPSC fate (e.g., self-renewal, differentiation, apoptosis and quiescence) has been challenging and underlying mechanisms are mostly unidentified. However, recent studies have uncovered some extrinsic factors that can influence state stability of hPSCs and contribute to fate decisions.10 These extrinsic factors include various soluble factors, cell-cell interactions, and ECM, which are key components of the hPSC microenvironment by definition (Fig. 1a).11 Additionally, soluble factors such as growth factors added to the culture or secreted by stem cells are often potent in their effects on cell fate.12 Indeed, undifferentiated hPSCs are highly sensitive to the soluble growth factors that are usually contained in these media. However, the effects of various defined media for maintaining selfrenewal states over extended periods have not been fully studied and optimally defined culture conditions have yet to be further refined. Therefore, screening chemically defined media (CDM) to evaluate the influence of these factors will also be essential for acquiring more Lab Chip, 2010, 10, 1113–1119 | 1113 Fig. 1 (a) The extrinsic factors such as soluble growth factors, cell-cell interactions, and ECM play an important role in controlling stem cell fate in the microenvironment. (b) Schematic illustration of a microfluidic hPSC array for hPSC culture and phenotype assay. (c) Microfluidic image cytometry (MIC) was conducted followed by segmentation and quantification analysis. 1114 | Lab Chip, 2010, 10, 1113–1119 This journal is ª The Royal Society of Chemistry 2010 qualified and defined culture methods that support self-renewal of hPSCs. Still, there are several other parameters that must be addressed in the study of hPSCs. First, although hPSCs can self-renew indefinitely, it is known that there is enormous variation between different PSC lines with regard to expression of pluripotency and differentiation markers.13 This is likely due to that fact that hES cell lines have been derived from embryos with different characteristics and further isolated by different procedures.14 In the case of hiPSCs, there is also variation due to (i) the factors used for reprogramming, (ii) the methods to deliver these factors, (iii) the source of the original cell lines, (iv) the expression levels of delivered factors, (v) the culture conditions for obtained hiPSCs and (vi) the methods to identify obtained hPSCs.15 Second, various commercially available hPSC defined culture media and ECM7–9,16–18 contain different components that may cause variable effects depending on the cell line and culture periods. Thus, taking into consideration all of these parameters as influential factors, there is also a need to systematically compare the differences between hPSC lines in order to comprehend their fundamental biology. However, there are some disadvantages in conventional experimental settings for hPSCs. Especially, when screening the characteristics among various hPSC lines, conventional analyses such as flow cytometry, microarray or RT-PCR require large amounts of cells, resulting in high costs in maintenance.19 On the other hand, the introduction of microfluidics can allow major advances in stem cell research. While there are tremendous efforts to compare the similarities and differences of various hPSC characteristics worldwide,13 it is highly important to establish a standardized hPSC culture condition, which causes less deviation and uncertainty. In microfluidics, miniaturization of cell culture platforms not only allows us to observe cellular behavior on the scale found in living systems but also provides a means to engineer miniaturized cell culture platforms that are more in vivo-like than conventional dish cultures,20,21 thereby fostering robust, reproducible and uniform culture conditions. Additionally, with the ability to manipulate the fluid flow precisely, microfluidics can make excellent perfusion cell-culture devices, which are powerful tools to control the soluble and mechanical parameters of the cell culture environment.22 These aspects are extremely essential since hPSCs interact strongly with their microenvironmental factors, which can directly influence the fate decisions. Furthermore, microfluidic technology can be integrated with a variety of biological assays and is compatible with Micro Electro Mechanical System (MEMS) technology for further applications including electrophoresis and cell sorting.19 A microfluidic device is made out of polydimethylsiloxane (PDMS), which is an elastomeric material utilizing the process of soft lithography for fabrication. Its beneficial features for stem cell biology include biocompatibility, gas permeability and durability. It is also safe and easy to handle within general laboratories performing biological research. Additionally, with its scalability and automation, it has more potential for clinical applications. Ultimately, since microfluidic devices can perform standard tissue culture in a more rapid, controllable and reproducible fashion with considerably low costs in a high-throughput fashion,23,24 microfluidic technology is well-suited for evaluating multiple hPSC culture conditions and simultaneously observing their responses. Previously, Villa-Diaz et al. and we have reported maintaining hESCs in conventional KSR/MEF conditions inside a hESCmChip.25,26 However, to date, it has not yet been reported that hPSCs, This journal is ª The Royal Society of Chemistry 2010 especially hiPSCs, can be cultured in chemically defined culture conditions and quantitatively studied in a microfluidic device. More importantly, there has not been a systematic comparison of the similarities or differences of each of these hPSCs cultured in various chemically defined culture conditions. Therefore, we have developed a microfluidic hPSC array (Fig. 1b) to perform hPSC culture and phenotype assay. Subsequently, microfluidic image cytometry (MIC) was conducted (Fig. 1c) followed by segmentation and quantification analysis. In this study, we have reported (i) a microfluidic platform to optimize ECM and various CDM for evaluating optimal culture conditions in hPSCs, combined with (ii) a systematic and quantitative analysis and smallscale screening of the hPSCs cultured in various CDM using multiparallel detected protein expressions. Using this array, we have also performed (iii) a side-by-side comparison of the hPSC phenotypic responses across available stem cell lines and CDM. This analysis allowed for examination of the cell fate of a single hPSC in a hPSC colony in each condition and demonstrated the sensitivity and effectiveness of our microfluidic hPSC array for use in quantification of multiple stem cell culture parameters. For fabrication of a PDMS-based microfluidic hPSC array, we used the process of soft lithography (ESI Fig. S1a †). The PDMS was mounted and assembled on a glass slide. During the hPSC culture assays, this array was set on an inverted microscope stage for routine monitoring of hPSCs. Our PDMS-based microfluidic hPSC array was comprised of 24 cell culture chambers (700 mm (W) 900 mm (L) 100 mm (H), Total volume 630 nL). For on-chip cell culture, each chamber was used for the static culture conditions. Using an electrical pipette (0.5–12.5 mL, Thermo Fisher Scientific) capable of handling precise volume and flow rates, four mL of solution containing hPSCs or reagents were filled into the tip. The tip was gently inserted into the inlet of a microfluidic hPSC culture array and solution was dispensed at 6 mL sec1 with accurate piston movement (ESI Fig. S1b†). A few hours after hPSC loading, the medium was changed every 12 h (see also in ESI Methods and Fig. S2†). In this array, each chamber can perform immunocytochemical analysis under discrete hPSC culture conditions to determine the levels of protein expression (see also in ESI Methods†). For cell line study, we examined 5 lines including (i) OCT4-enhanced green fluorescent protein (EGFP) knock-in HSF1 cell line (HSF1-OCT4-EGFP),25,27 (ii) hESC lines (HSF1 and H1) and (iii) hiPSC lines (iPSA1 and iPSB2, ESI Fig. S3†). The OCT4-EGFP cell line is unique in that it allows live cell monitoring of its pluripotency status in real-time. We therefore used it to optimize the defined culture conditions. Other cell lines were used to further make comparisons between their protein expressions. For the purpose of establishing optimal culture conditions in a microfluidic hPSC array, we began with examining the optimal concentration of ECM by using MEF-conditioned medium (CM). We chose to use hESC qualified Matrigel, (see also in ESI Methods†) since this is commonly used for feeder-free hPSC culture in current stem cell research. As mentioned, we used HSF1-OCT4-EGFP cell lines to monitor the morphology of hPSC colonies and EGFP expression levels during culturing periods. The results showed that HSF1-OCT4-EGFP colonies were unable to attach, spread out and grow well on the substrate coated with 100 mg mL1 (Fig. 2a). On the other hand, the HSF1-OCT4-EGFP colonies extended well and maintained their growth in an undifferentiated state for 7 days with 20 mg mL1. Thus, we determined that 20 mg mL1 of Matrigel was an optimal ECM condition for hPSC culture. Lab Chip, 2010, 10, 1113–1119 | 1115 Next, using an optimized ECM condition, we then screened CDM for optimal culture conditions. We tested three CDM (StemPro,16 mTeSR7,8,32 and N2B279), which had been published to support undifferentiated growth of hPSCs with defined components. Each medium was also supplemented with varying bFGF concentrations and the morphology of hPSC colonies and EGFP expression levels were then monitored over 5 days (Fig. 2b). After five days in culture, HSF1-OCT4-EGFP cells were able to form colonies and express EGFP driven by an OCT4 promoter in all three CDM conditions. We also found that bFGF concentrations did not influence cell viability and pluripotency of hPSCs. In a previous study, we used 100 ng mL1 of bFGF with KSR/MEF conditions.15 However, at this time, we observed that 20 ng mL1 bFGF in feeder-free chemically defined hPSC culture conditions was sufficient to grow undifferentiated hPSCs. Here, we confirmed that all three chemically defined conditions were able to sustain the growth of hPSCs with undifferentiated states using optimized ECMs and accordingly established optimal defined culture conditions in a microfluidic hPSC array. Interestingly, within optimal conditions we now found a variation in physical and biochemical characteristics in hPSCs cultured with different media. We then compared the effects of these media on various phenotypes across the cell lines including morphology, growth rates and expression level of pluripotency protein markers. A recent study showed that colony morphology was an important parameter to determine characteristics of hPSCs and molecular phenotype and differentiation potential could vary within morphologically different hPSC colonies.28 According to the results of DAPI nuclear staining, we found that HSF1-OCT4-EGFP cells cultured in StemPro formed colonies with sharp pointed edges (Fig. 2c). These cells also had a tendency to form a relatively larger nuclear size than those cultured in the other CDM. Additionally, HSF1-OCT4-EGFP cells cultured in mTeSR represented more dense and tight colonies. We then conducted growth assays to examine the average growth rate of colonies cultured in each medium by measuring the surface area of HSF1-OCT4-EGFP colonies (Fig. 2d). We found that although all the conditions were able to support self-renewal of hPSCs and maintain pluripotency marker protein expression over four days, the growth rate of colonies differed depending on culturing media. Among three CDM, HSF1-OCT4-EGFP cultured in StemPro showed the fastest growth rate. Compared to N2B27, StemPro and mTeSR conditions showed 2.65 and 1.85-fold changes in their average colony size, respectively. We speculated that the components heregulin-1b and activin A were responsible for promoting proliferation of HSF1-OCT4-EGFP cells in StemPro.16 To further characterize the effects of these media on a collection of hPSC lines, we performed immunocytochemistry to evaluate expression of pluripotency markers in hES and hiPS cells quantitatively (Fig. 3 and ESI Fig. S2†). The pluripotent markers we used were OCT4, NANOG, SSEA4, TRA-1-60 and TRA-1-80. Here, we introduced one more condition where we induced differentiation by Fig. 2 The establishment of serum/feeder-free chemically defined hPSC culture conditions in a microfluidic hPSC array. Bright field (BF) and fluorescence images of hPSCs are shown on the top and bottom, respectively. (a) Optimization of Matrigel coating conditions. HSF1OCT4-EGFP cells were cultured on ECM with two concentrations (20 or 100 mg mL1) using MEF-CM. Scale bar represents 50 mm. (b) Screening of CDM (StemPro, mTeSR and N2B27) with different concentrations of 1116 | Lab Chip, 2010, 10, 1113–1119 bFGF using HSF1-OCT4-EGFP cells. HSF1-OCT4-EGFP cells were cultured on the glass slide coated with an optimal Matrigel concentration (20 mg mL1). Scale bar represents 100 mm. (c) Morphologically different HSF1-OCT4-EGFP colonies cultured in three CDM. (d) Quantitative comparison of the growth curves of HSF1-OCT4-EGFP cells cultured in three CDM. Each dot represents mean S.D. (*p < 0.05, ***p < 0.001). Scale bar represents 50 mm. This journal is ª The Royal Society of Chemistry 2010 Fig. 3 Evaluation of pluripotency/differentiation marker expression in a microfluidic hPSC array using MIC. (a) Bright-field (BF) images, DAPI nuclear fluorescence images and other fluorescence images of HSF1 cells cultured in StemPro immunostained against pluripotent markers (OCT4, NANOG, SSEA4, TRA-1-60 and TRA-1-80). Scale bar represents 50 mm. (b) BF images, DAPI nuclear fluorescence images, OCT4 and SSEA1 fluorescence images of HSF1 cells cultured in StemPro or Differentiation condition (Diff). Scale bar represents 50 mm. (c,d) Singlecell based immunofluorescent histograms of (c) OCT4 expression and (d) SSEA1 expression in individual HSF1, H1, iPSA1 and iPSB2 cells cultured in KSR/MEF, StemPro, mTeSR, N2B27 and Differentiation condition. (e,f) Heat maps based on the quantified (e) OCT4 expression and (f) SSEA1 expression. The protein expression level was normalized among samples of H1, HSF1, iPSA1 and iPSB2 cultured in KSR/MEF, StemPro, mTeSR, N2B27 and Differentiation condition and analyzed by Euclidean distance hierarchical clustering to categorize similar groups This journal is ª The Royal Society of Chemistry 2010 adding 10% fetal bovine serum to DMEM medium as a negative control. The differentiation marker we used was SSEA1. As the results indicated, all of the hPSCs cultured in chemically defined conditions uniformly expressed these pluripotency markers (Fig. 3a). After confirming their pluripotency, we further quantified OCT4 and SSEA1 expression at the single-cell level based on immunofluorescence imaging (Fig. 3b). Image cytometry is an image-based measurement that allows quantitative analysis of these marker expressions at the single cell level by using software such as CellProfiler, which can generate flow-cytometry-like data. Single-cell based immunofluorescent histograms presented a variable distribution in OCT4 and SSEA1 expression (Fig. 3c,d, respectively) after each cell line in each condition was co-stained with OCT4 and SSEA1 and analyzed in a single cell. In general, this visually expressed how heterogeneous/homogeneous each colony was within the undifferentiated and differentiated conditions. As an illustration, when a histogram of protein expression showed the broad distribution, populations were more heterogeneous and vice versa. The histograms of OCT4 expression in HSF1, H1 and iPSB2 cultured in KSR/MEF and three CDM conditions had a homogenous distribution with the higher level of OCT4 expression compared to Differentiation condition. The iPSA1 cells cultured in mTeSR and N2B27 had two subpopulations with both high and low OCT4 expression. The histograms for SSEA1 expression in HSF1, iPSA1 and iPSB2 cultured in KSR/MEF and three CDM conditions exhibited a homogeneous distribution with the low SSEA1 expression level, indicating that most of the hPSC populations remained undifferentiated. On the other hand, all the cell lines in Differentiation condition exhibited the broad distribution of the histogram for SSEA1 expression, revealing various responsiveness and sensitivity of the highly heterogeneous cells upon the serum inducement. This could also be attributed to the weak and/ or short-time period of treatment, but still enabled visual dynamics of the protein expression during the differentiation process. Additionally, while the H1 cells cultured in mTeSR and KSR/MEF conditions had low SSEA1 expression, some populations in StemPro and N2B27 conditions showed relatively high SSEA1 regardless of the fact that they simultaneously expressed high OCT4. For systematic analysis, we further conducted Euclidean distance hierarchical clustering29 based on the mean values of OCT4 and SSEA1 expression and generated the heat maps, which represented values of data in two-dimensional maps as colors (Fig. 3e,f, respectively. ESI Fig. S2), to compare the distinct protein expression resulting from the various cell lines and CDM. Hierarchical clustering generates a hierarchy of sample groups represented by a dendrogram (a tree-like diagram). To determine the similarities of two groups, we used Euclidean distance calculated with the eqn (1), sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi n X d ¼ jp-qj ¼ ðpi qi Þ2 (1) i¼1 if p and q are the two points (here, samples) in Euclidean n-space for finding nearest neighbors of sample groups. The heat maps together. Each row represents marker expression in each cell line. Each column represents a particular condition. (g,h) Evaluation of MIC in terms of the reproducibility. The quantification of (g) OCT4 fluorescence intensity and (h) SSEA1 fluorescence intensity between the two experiments is shown. Lab Chip, 2010, 10, 1113–1119 | 1117 generated based on Euclidean distance hierarchical clustering are commonly used for, for instance, microarray analysis to reduce data dimension, categorize samples, and show a multidimensional data set in 2-D maps in colors. In terms of the different media treatments for 4 days, according to the heat map, hPSCs cultured in StemPro and mTeSR expressed high OCT4 and low SSEA1 across all the cell lines. The heat map also showed a similar expression pattern across all the cell lines in the OCT4 level between hPSCs cultured in StemPro and mTeSR conditions. In contrast, hPSCs cultured in N2B27 condition rendered relatively lower OCT4 expression across the cell lines, exhibiting its tendency to direct differentiation during the culturing periods. Therefore, N2B27 condition was categorized as similar to the hPSCs cultured in Differentiation condition based on the clustering. In the case of SSEA1 expression, all cell lines cultured in Differentiation condition showed strong SSEA1 expression. Additionally, we observed that some populations of H1 and iPSB2 cells cultured in N2B27 condition also expressed relatively high SSEA1. In terms of cell lines within the same condition, each cell line responded differently and resulted in various phenotypes. The iPSB2 line especially appeared to have different OCT4 and SSEA1 expression compared to the other three lines. However, there seems to be no clear trend in cell lines concluded based on the level of pluripotent marker expression. Finally, we evaluated the robustness of MIC to confirm the fidelity of our study (Fig. 3g,h). Two microfluidic chips that cultured H1 cells in Stem Pro (Exp1 and Exp2) were randomly chosen and both OCT4 (Fig. 3g) and SSEA1 (Fig. 3h) expression were quantified. Between the two chips, there were no significant differences in OCT4 or SSEA1 fluorescence intensity value therefore we concluded that our microfluidic hPSC array in conjunction with MIC was precise and reproducible. conventional 6 well plates. According to the results, we found that culturing in various CDM resulted in different phenotypes in each hPSC line including morphology, growth rate and pluripotent marker expression. We speculated that over culturing periods, heterogeneous cell populations within a single colony showed varied growth factor responsiveness and protein expressions by intricately interplaying with the microenvironmental factors at the single cell level. In general, the final phenotype in a single cell relies on the current state of the cell and the microenvironment that is composed of the extrinsic factors such as soluble factors in media and output signals of hPSCs.11,31 Here, by presenting the detail phenotypic analysis, we have also demonstrated the ability of our device to study the heterogeneity of hPSCs and the interaction of different hPSC lines with the microenvironment, which will have an overall effect in governing stem cell fate. With a carefully selected set of markers (e.g. pluripotency, apoptosis, differentiation and cell cycle), this tool can be applied to conduct more phenotype studies when combined with signaling cascades transduced by extrinsic factors using its multiplexity to determine the hPSC molecular signatures. Because of these unique features, we envision that this microfluidic platform will be beneficial to investigate stem cell biology in a wide range of biomedical settings and applications in regenerative medicine. Conclusions Notes and references We developed a simple microfluidic platform to optimize ECM, screen CDM and establish the optimal chemically defined culture system for both human ESCs and iPSCs. By using this microfluidic platform, we were also able to study hPSC phenotypic response by comparing the effects of various CDM and hPSC lines. Although we cannot ignore the fact that PDMS may absorb molecules from solution due to their characteristics (e.g., highly porous and hydrophobic material)30 and release them during culturing periods, according to the results, not only this microfluidic platform can effectively maintain pluripotency of hPSCs over a week in CDM with 20 ng mL1 of bFGF but all the results were consistent and reproducible across the hPSC lines. Also, our concentration of bFGF was the original concentration7–9,16 found in other studies with the conventional macro-scale settings. Thus, we considered this PDMS effect was negligible. Additionally, we found that the condition with StemPro medium on the ECM of 20 mg mL1 of Matrigel for culturing hPSCs generally provides high OCT4 and low SSEA1 expression across the cell lines including hiPSCs. In this work, we have demonstrated that a microfluidic hPSC array can achieve robust and reproducible hPSC cultures on a simple setting when combined with highly quantitative single-cell profiling methods. Furthermore, not only can this array be utilized for real-time live cell monitoring of hPSCs, but this platform can also perform small-scale screening with multi-parallel detection system, using a small amount of samples and reagents that are roughly 3 orders of magnitude less than the 1 J. A. Thomson, Science, 1998, 282, 1145–1147. 2 K. Takahashi and S. Yamanaka, Cell, 2006, 126, 663–676. 3 K. Takahashi, K. Tanabe, M. Ohnuki, M. Narita, T. Ichisaka, K. Tomoda and S. Yamanaka, Cell, 2007, 131, 861–872. 4 J. Yu, M. A. Vodyanik, K. Smuga-Otto, J. Antosiewicz-Bourget, J. Frane, S. Tian, J. Nie, G. A. Jonsdottir, V. Ruotti, R. Stewart, I. I. Slukvin and J. Thomson, Science, 2007, 318, 1917–1920. 5 W. E. Lowry, L. Richter, R. Yachechko, A. D. Pyle, J. Tchieu, R. Sridharan, A. T. Clark and K. Plath, Proc. Natl. Acad. Sci. U. S. A., 2008, 105, 2883–2888. 6 I. Park, R. Zhao, J. West, A. Yabuuchi, H. Huo, T. Ince, P. Lerou, M. W. Lensch and G. Daley, Nature, 2008, 451, 141–146. 7 T. Ludwig, M. Levenstein, J. Jones, W. Berggren, E. Mitchen, J. Frane, L. Crandall, C. Daigh, K. Conard, M. Piekarczyk, R. Llanas and J. Thomson, Nat. Biotechnol., 2006, 24, 185–187. 8 T. Ludwig, V. Bergendahl, M. Levenstein, J. Yu, M. D. Probasco and J. Thomson, Nat. Methods, 2006, 3, 637–646. 9 S. Yao, S. Chen, J. Clark, E. Hao, G. M. Beattie, A. Hayek and S. Ding, Proc. Natl. Acad. Sci. U. S. A., 2006, 103, 6907–6912. 10 T. Enver, M. Pera, C. Peterson and P. W. Andrews, Cell Stem Cell, 2009, 4, 387–397. 11 D. E. Discher, D. J. Mooney and P. W. Zandstra, Science, 2009, 324, 1673–1677. 12 C. E. Murry and G. Keller, Cell, 2008, 132, 661–680. 13 O. Adewumi, B. Aflatoonian, L. Ahrlund-Richter, M. Amit, P. W. Andrews, G. Beighton, P. A. Bello, N. Benvenisty, L. S. Berry, S. Bevan, B. Blum, J. Brooking, K. G. Chen, A. B. Choo, G. A. Churchill, M. Corbel, I. Damjanov, J. S. Draper, P. Dvorak, K. Emanuelsson, R. A. Fleck, A. Ford, K. Gertow, M. Gertsenstein, P. J. Gokhale, R. S. Hamilton, A. Hampl, L. E. Healy, O. Hovatta, J. Hyllner, M. P. Imreh, J. Itskovitz-Eldor, J. Jackson, J. L. Johnson, M. Jones, K. Kee, 1118 | Lab Chip, 2010, 10, 1113–1119 Acknowledgements This work was supported by the Eli and Edythe Broad Center of Regenerative Medicine and Stem Cell Research at the Institute of Molecular Medicine at University of California, Los Angeles. H.R.T., M.A.T., A.D.P., A.T.C., and O.N.W. were supported by California Institute of Regenerative Medicine. O.N.W. is an investigator of the Howard Hughes Medical Institute. This journal is ª The Royal Society of Chemistry 2010 14 15 16 17 18 19 B. L. King, B. B. Knowles, M. Lako, F. Lebrin, B. S. Mallon, D. Manning, Y. Mayshar, R. D. McKay, A. E. Michalska, M. Mikkola, M. Mileikovsky, S. L. Minger, H. D. Moore, C. L. Mummery, A. Nagy, N. Nakatsuji, C. M. O’Brien, S. K. Oh, C. Olsson, T. Otonkoski, K. Y. Park, R. Passier, H. Patel, M. Patel, R. Pedersen, M. F. Pera, M. S. Piekarczyk, R. A. Pera, B. E. Reubinoff, A. J. Robins, J. Rossant, P. Rugg-Gunn, T. C. Schulz, H. Semb, E. S. Sherrer, H. Siemen, G. N. Stacey, M. Stojkovic, H. Suemori, J. Szatkiewicz, T. Turetsky, T. Tuuri, S. van den Brink, K. Vintersten, S. Vuoristo, D. Ward, T. A. Weaver, L. A. Young and W. Zhang, Nat. Biotechnol., 2007, 25, 803–816. L. M. Hoffman and M. K. Carpenter, Nat. Biotechnol., 2005, 23, 699– 708. N. Maherali and K. Hochedlinger, Cell Stem Cell, 2008, 3, 595–605. L. Wang, T. C. Schulz, E. S. Sherrer, D. S. Dauphin, S. Shin, A. M. Nelson, C. B. Ware, M. Zhan, C. Song, X. Chen, S. N. Brimble, A. Mclean, M. J. Galeano, E. W. Uhl, K. A. D’Amour, J. D. Chesnut, M. S. Rao, C. A. Blau and A. J. Robins, Blood, 2007, 110, 4111–4119. M. K. Furue, J. Na, J. P. Jackson, T. Okamoto, M. Jones, D. Baker, R. Hata, H. D. Moore, J. D. Sato and P. W. Andrews, Proc. Natl. Acad. Sci. U. S. A., 2008, 105, 13409–13414. R. Xu, R. Peck, D. Li, X. Feng, T. Ludwig and J. Thomson, Nat. Methods, 2005, 2, 185–190. J. El-Ali, P. K. Sorger and K. F. Jensen, Nature, 2006, 442, 403–411. This journal is ª The Royal Society of Chemistry 2010 20 D. van Noort, S. M. Ong, C. Zhang, S. Zhang, T. Arooz and H. Yu, Biotechnol. Prog., 2009, 25, 52–60. 21 S. N. Bhatia, U. J. Balis, M. L. Yarmush and M. Toner, FASEB J, 1999, 13, 1883–1900. 22 S. Takayama, E. Ostuni, P. LeDuc, K. Naruse, D. E. Ingber and G. M. Whitesides, Chem. Biol., 2003, 10, 123–130. 23 I. Meyvantsson, J. W. Warrick, S. Hayes, A. Skoien and D. Beebe, Lab Chip, 2008, 8, 717–724. 24 J. Warrick, I. Meyvantsson, J. Ju and D. J. Beebe, Lab Chip, 2007, 7, 316–321. 25 K. Kamei, S. Guo, Z. Yu, H. Takahashi, E. Gschweng, C. Suh, X. Wang, J. Tang, J. Mclaughlin, O. N. Witte, K. Lee and H. Tseng, Lab Chip, 2009, 9, 555–563. 26 L. G. Villa-Diaz, Y. Torisawa, T. Uchida, J. Ding, N. C. Nogueirade-Souza, K. S. O’Shea, S. Takayama and G. D. Smith, Lab Chip, 2009, 9, 1749–1755. 27 T. P. Zwaka and J. A. Thomson, Nat. Biotechnol., 2003, 21, 319–321. 28 E. M. Chan, S. Ratanasirintrawoot, I. H. Park, P. D. Manos, Y. H. Loh, H. Huo, J. D. Miller, O. Hartung, J. Rho, T. A. Ince, G. Q. Daley and T. M. Schlaeger, Nat Biotechnol, 2009, 27, 1033–1037. 29 J. Quackenbush, Nat. Rev. Genet., 2001, 2, 418–427. 30 M. W. Toepke and D. J. Beebe, Lab Chip, 2006, 6, 1484–1486. 31 A. Solanki, J. D. Kim and K. B. Lee, Nanomedicine, 2008, 3, 567–578. 32 The mTeSR medium is commercially available in the original condition with 100 ng mL1 of zebrafish bFGF (zbFGF) to begin with.8 Lab Chip, 2010, 10, 1113–1119 | 1119 Stem cell differentiation Controlling Differentiation of Neural Stem Cells Using Extracellular Matrix Protein Patterns Aniruddh Solanki, Shreyas Shah, Kevin A. Memoli, Sung Young Park, Seunghun Hong, and Ki-Bum Lee* The ability of stem cells to differentiate into specialized lineages within a specific microenvironment is vital for regenerative medicine. For harnessing the full potential of stem cells for regenerative therapies, it is important to investigate and understand the function of three types of microenvironmental cues—soluble signals, cell–cell interactions, and insoluble (physical) signals—that dynamically regulate stem cell differentiation.[1] Neural stem cells (NSCs) are multipotent and differentiate into neurons and glial cells,[2] which can provide essential sources of engraftable neural cells for devastating diseases such as Alzheimer’s disease,[3] Parkinson’s disease[4] and spinal cord injury.[5] One of the major challenges involved in the differentiation of NSCs is to identify and optimize factors which result in an increased proportion of NSCs differentiating into neurons as opposed to glial cells. To this end, soluble cues such as brain-derived neurotrophic factor (BDNF),[6] sonic hedgehog (Shh),[7] retinoic acid (RA),[6c] and neuropathiazol[8] have been shown to significantly increase neuronal differentiation of NSCs in vitro. However, the research toward studying the function of the other two microenvironmental cues (cell– cell interactions and insoluble cues) during the neurodifferentiation of NSCs is limited, mainly due to the lack of availability of methods for the investigation.[9] While various aspects such as cell–cell interactions,[10] combinations of extracellular matrix (ECM) proteins,[1a,11] and physical properties of substrates have been shown to play a vital role in determining the fate of other adult stem cells such as mesenchymal stem cells (MSCs),[12] cardiac stem A. Solanki, S. Shah, K. A. Memoli, Prof. K.-B. Lee Department of Chemistry and Chemical Biology, Rutgers The State University of New Jersey Piscataway, NJ 08854, USA Email: [email protected], http://rutchem.rutgers.edu/∼kbleeweb/ S. Y. Park Interdisciplinary Program in Nano-Science and Technology Seoul National University Seoul, South Korea Prof. S. Hong Department of Physics Department of Biophysics and Chemical Biology Interdisciplinary Program in Nano-Science and Technology Seoul National University Seoul, South Korea DOI: 10.1002/smll.201001341 small 2010, 6, No. 22, 2509–2513 cells,[13] and hematopoetic stem cells,[14] little is known about the influence of such factors on the neuronal differentiation of NSCs. Therefore, there is a pressing need to develop methods for investigating the role of cell–cell interactions and insoluble signals in selectively inducing the differentiation of NSCs into specific neural cell lineages. Herein, we demonstrate how ECM protein patterns can be used to investigate the effect of physical cues combined with cell–cell interactions on the differentiation of NSCs. Bio-surface chemistry combined with soft lithography was used to generate combinatorial patterns with varying geometries and dimensions of ECM proteins (e.g., laminin, fibronectin, and collagens) to study the influence of surface features and ECM compositions on the differentiation of NSCs. We hypothesized that the ECM protein patterns with variant geometries and dimensions would provide physical cues (e.g., mechanical or topographical cues), as well as guide cell–cell and cell–ECM interactions in a controlled manner, both of which would ultimately lead to a pattern geometry-dependent and pattern dimensiondependent neuronal and glial differentiation (Figure 1). Our data confirmed that the difference in the extent of neuronal and glial differentiation of NSCs on the ECM protein patterns was entirely due to the pattern geometry and dimension, as all the experiments were carried out in the absence of exogenous factors that promote neurogenesis; this suggests that NSCs can undergo differentiation by purely sensing the difference in ECM pattern geometries and dimensions. Extracellular matrix protein patterns with variant geometries and dimensions were fabricated by initially patterning octadecanethiol (ODT, 5 mm in ethanol), a hydrophobic alkanethiol, which formed self-assembled monolayers (SAMs) of squares, stripes, and grids on glass substrates coated with a thin film (12 nm) of gold. In order to minimize the nonspecific attachment of laminin, the background of the substrates was passivated by incubating in a solution (5 mm in ethanol) of tetraethylene glycol terminated alkanethiol [EG4-(CH2)11-SH, 12 h] (See Supporting Information for synthesis and characterization). After passivating the background, a solution of ECM protein [e.g., laminin (10 μg mL−1) in phosphate buffered saline (PBS) buffer, pH = 7.4] was added onto the substrates (3 h) and was preferentially adsorbed onto the hydrophobic regions (ODT patterns). The selective adsorption of laminin on hydrophobic regions was consistent with the results of other groups[15] and was also confirmed by immunostaining using anti-laminin IgG © 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim wileyonlinelibrary.com 2509 communications A. Solanki et al. interactions of NSCs with the passivated areas, and then incubated in fresh basal medium. The media was exchanged with pattern geometry fresh media every other day. During our screening approach to investigate the squares stripes grids function of physical cues on neuronal difH H H H H H O O O O O O O O O O O O ferentiation of NSCs, we monitored the O O O O O O differentiation on ECM protein patterns O O O O O O O O O O O O by using two orthogonal assays, namely immunocytochemical and morphological assays. To assess the differentiation of S S S S S S S S S S S S NSCs, the down-regulation of the NSC Au Thin Film marker (Nestin) and the geometryECM Protein ECM Protein Patterns dependent expression of the neuronal marker (β-III Tubulin, TuJ1) and glial marker (glial fibrillary acidic protein, GFAP) were monitored. In addition, the development of branches or spindle-like morphologies, and neurite outgrowths were observed by using an inverted phase contrast microscope (Zeiss Axiovert squares stripes grids 200M equipped with AxioCam CCD). Patterns of ECM proteins with different geometries contributing to adheFigure 1. A schematic diagram of our approaches. A) The fabrication of ECM protein patterns sion, proliferation, growth and migration and their application for NSC differentiation. B) The selective attachment of NSCs on the protein of various cells (including stem cells) have patterns and differentation into two different kinds of neural cells. C) The differentiation of been reported.[16] In addition, reports from NSCs into either neurons (red) or astrocytes (green) on the protein patterns. D) Increased the literature have shown cell–cell interacneuronal differentiation on the grid patterns, as compared to the stripes and squares. tions to play a critical role in the differentiation of adult stem cells. For instance, (See Supporting Information, Figure S1). Only the patterned it was recently shown that cell–cell interactions played an regions, coated with ECM proteins, promoted cell adhesion important role in the osteogenic (bone) differentiation of and growth whereas the rest of the substrate remained inert MSCs.[10] To study the influences pattern geometries and cell– (Figure 1). We similarly patterned several different ECM pro- cell interactions on the differentiation of NSCs, we initially teins including fibronectin and collagen, but found that lam- patterned the NSCs on stripes of laminin, which promoted inin provided the optimum microenviromental cues for NSC one-way interactions in a controlled manner (Figure 2.A1). adhesion and growth. Hence, all our differentiation studies We found that after six days, 36% of NSCs on the isolated were carried out using laminin patterns. stripes differentiated into neurons (Figure 2.A2 and Figure 3). To examine the effect of the ECM protein patterns on At the same time we observed that 64.3% of NSCs on stem cell differentiation, primary rat hippocampal neural these stripes differentiated into astrocytes (Figure 2.A3 and stem cells (Millipore) were first expanded and maintained Figure 3). in an undifferentiated state in a homogeneous monolayer on To further confirm the influence of such interactions a polyornithine and laminin-coated Petri dish in a defined on the differentiation of NSCs, we used square patterns of serum-free growth medium [DMEM/F12 supplemented with laminin to isolate NSCs and restrict their growth within the B27 and basic fibroblast growth factor (bFGF, 20 ng mL−1)]. square patterns (Figure 2.B1). We hypothesized that the difFor obtaining reproducible and consistent results, all ferentiation behaviour of NSCs can be considerably influexperiments were carried out using NSCs from passages enced by limiting cell–cell interactions. We observed that 2–5 at a constant cell density of 150 000 cells per substrate NSCs patterned on squares, having the same dimensions and (1.5 cm × 1.5 cm), which was optimum for cell growth spaces as the stripes, differentiated into neurons to a considwithout clustering. Arresting the proliferation of NSCs and erably lesser extent (28.1%, Figure 2.B2 and 3) as compared initiating their spontaneous differentiation was achieved to the NSCs involved in one-way interactions on the striped by withdrawing bFGF from the culture medium (resulting laminin patterns. At the same time, the number of NSCs that in basal medium), without the additional treatment with differentiated into astrocytes increased considerably on exogenous factors (proteins and small molecules). The basal squares –76.9% on squares as compared to 64.3% on stripes medium (2 mL) containing the NSCs (75 000 cells mL−1) (Figure 2.B3 and 3). Thus, the reduced cell–cell interactions was put in a single well of a 6-well plate containing a sub- with the NSCs on the surrounding patterns may have led to strate with laminin patterns. After the NSCs attached onto reduced neuronal differentiation and increased glial differenthe laminin patterns (1 h), the substrates were rinsed with tiation of the NSCs. Based on the observed differentiation of copious amounts of media in order to minimize nonspecific NSCs on stripes and squares, we further hypothesized that using neurons B pa tte rn di m en sio n A astrocytes C 2510 www.small-journal.com D © 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim small 2010, 6, No. 22, 2509–2513 Controlling Differentiation of Neural Stem Cells Figure 2. Growth and differentiation of NSCs on the laminin patterns. Phase contrast images show NSC attachment and growth on stripes (A1), squares (B1), and grids (C1) on Day 2 after seeding. Fluorescent images of cells stained for the neuronal marker TuJ1 (red) and nucleus (blue) show the extent of neuronal differentiation of NSCs on stripes (A2), squares (B2), and grids (C2) on Day 6 after seeding. Similarly, cells stained for astrocyte marker GFAP (green) and nucleus (blue) show the extent of glial differentiation on stripes (A3), squares (B3), and grids (C3) on Day 6 after seeding. Scale bars: 50 μm. specific pattern geometries promoting cell–cell interactions could lead to higher neuronal differentiation. For this purpose, we used grid patterns of laminin, having the same dimensions as the stripe and square patterns, for NSC growth and differentiation. The grid patterns were specifically designed to increase cell–cell interactions in a controlled manner (by promoting two-way interactions, Figure 2.C1). After six days in basal medium, as compared to the NSCs patterned on stripes and squares of laminin, we observed a remarkable increase in the number of NSCs that underwent neuronal differentiation (45.6%, Figure 2.C2 and 3) and a decrease in the number of cells that underwent glial differentiation on grid patterns of laminin (49.6%, Figure 2.C3 and 3). All the experiments were repeated several times under the same conditions. To maintain consistency and minimize the effects from other variables, we fabricated and used PDMS stamps to generate ECM protein patterns of all the three geometries (having the same dimensions and spacing) on the same substrate. Using this method, we could reproduce and confirm our results with relative ease. Neuronal and glial differentiation of NSCs was also monitored on control substrates which included substrates coated with laminin (unpatterned) and substrates without laminin. The NSCs on substrates without laminin did not attach and failed to survive, whereas 32.5% of the NSCs on the unpatterned substrates coated with laminin differentiated into neurons and 71.2% of the NSCs differentiated into astrocytes six days after seeding. In addition to investigating the effect of pattern-geometry, we also studied the effect of dimensions on NSC differentiation. To this end, we generated ten different dimensions for each of the geometries, ranging from sizes as small as 10 μm and as large as 250 μm (Figure 4B). Interestingly, for the three different geometries above 50 μm, we observed little difference in the percentage of NSCs undergoing neuronal % of Cells Expressing Neural Markers 90 TuJ1 (Neurons) 80 70 60 50 40 30 20 10 0 Squares Stripes Grids Pattern Geometry (10-50 µm) Figure 3. Quantitative comparison of the percentage of cells expressing the neuronal marker TuJ1 and astrocyte marker GFAP on laminin patterns of squares, stripes and grids. Six days after seeding, the differentiated cells were counted and plotted as a ratio of TuJ1-positive cells or GFAPpositive cells to the total number of cells (n = 3). Student’s unpaired t-test was used for evaluating the statistical significance for cells stained for TuJ1 on stripes and squares, compared to those on grids. (∗ = P < 0.01, ∗∗ = P < 0.001). small 2010, 6, No. 22, 2509–2513 Figure 4. NSC alignment and differentiation on combinatorial ECM patterns. A) NSCs on grids of laminin express the neural stem cell marker, nestin (purple) on Day 2 after seeding, thus confirming that the NSCs are undifferentiated. B) NSCs stained for actin (green) show extensive spreading and cell–cell interactions on grid patterns of laminin on Day 2 after seeding, confirming that the NSCs, while still in the undifferentiated state, extensively interact with each other. C) SEM image of NSCs on Day 2 after seeding, showing the early alignment and extension of processes on grid patterns of laminin. D) NSCs previously shown to extend and grow on the grid patterns of laminin undergo neuronal differentiation and express the neuronal marker synapsin (pseudocolored yellow) on Day 6 after seeding. Scale bars: 20 μm. © 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.small-journal.com 2511 communications A. Solanki et al. and glial differentiation. The result observed for pattern dimensions above 50 μm was similar to that observed with unpatterned substrates. We believe the cells may not be able to sense the difference in pattern geometries above 50 μm and thus show similar behaviour to the cells on unpatterned substrates. Since the NSCs showed remarkable difference in differentiation on patterns ranging from 10–50 μm, all of our statistical analysis and investigation was done using pattern features within this range. We observed that the laminin patterns of all three geometries enabled the NSCs to attach and grow within a day or two day after seeding. By staining for actin using phalloidin and using field emission scanning electron microscopy (FESEM, Zeiss Gemini), we further observed that the cytoskeleton of the NSCs aligned well within the laminin patterns, guiding cellular morphology and interactions (Figure 4B,C). To confirm that the laminin patterns influenced morphological changes before differentiation (as opposed to an early differentiation of NSCs which might have caused a change in alignment and morphology), the NSCs were immunostained for the neural stem cell marker nestin two days after seeding in basal medium. We observed that most of the NSCs that aligned along the patterns, stained positive for nestin (Figure 4A), confirming that cells were in an undifferentiated (multipotent) state when they aligned along the patterns (See Supporting information, Figure S2 for NSCs on squares and stripes stained for actin and nestin). We further confirmed neuronal differentiation of NSCs on the laminin patterns using synapsin as another neuronal marker in addition to TuJ1. After six days in basal medium, a remarkably high number of the NSCs growing along the grid patterns of laminin expressed synapsin (Figure 4D). In addition, colocalization of TuJ1 and synapsin was observed within the NSCs differentiated on the grid patterns, confirming that the neurons expressed both neuronal markers (Supporting information, Figure S3). In summary, we fabricated and utilized patterns of ECM proteins for modulating the extent of neuronal and glial differentiation of NSCs in the absence of soluble cues such as small molecules and exogenous proteins. Potentially, our approach and methodology can be helpful for deconvoluting physical cues and cell–cell interactions from complex microenvironmental cues. More detailed mechanistic studies on how physical cues modulate the signaling cascades and the signaling pathways that are primarily involved in stem cell differentiation induced by such factors are currently under investigation. The implications of our results could also potentially be significant for tissue engineering for brain and spinal cord injuries, where NSCs or NSC-based differentiated cells can be transplanted into the damaged regions with scaffolds. For example, scaffolds having patterns promoting cell–cell interactions in a controlled manner could potentially lead to increased neuronal differentiation in vivo. Even though we have explored only proof-of-concept experiments focusing on differentiation of NSCs, a similar strategy could be extended to study and control the fate of other stem cells, such as MSCs and embryonic stem cells (work in progress). Our results substantiate the importance of pattern 2512 www.small-journal.com dimensions, pattern geometries, and cell–cell interactions in controlling stem cell fate. Supporting Information Supporting Information is available from the Wiley Online Library or from the author. Acknowledgements A. S. and S. S. have contributed equally to the authorship of this paper. We acknowledge John Kim and Neal Patel for their help in the experimental procedures. This work was supported by the NIH Director’s Innovator Award [(1DP20D006462–01), K.-B. L.] and the N.J. Commission on Spinal Cord grant [(09–3085-SCR-E-0), K.-B. L.]. S. H. acknowledges the support from the NRF grant (2009– 0079103) and the System 2010 program. K.-B. L. acknowledges partial support from Bioforce Nanosciences. [1] a) F. Guilak, D. M. Cohen, B. T. Estes, J. M. Gimble, W. Liedtke, C. S. Chen, Cell Stem Cell 2009, 5, 17; b) C. J. Flaim, S. Chien, S. N. Bhatia, Nat. Methods 2005, 2, 119; c) C. J. Flaim, D. Teng, S. Chien, S. N. Bhatia, Stem Cells and Dev. 2008, 17, 29. [2] a) F. H. Gage, J. Ray, L. J. Fisher, Annu. Rev. Neurosci. 1995, 18, 159; b) F. H. Gage, Science 2000, 287, 1433. [3] E. D. Roberson, L. Mucke, Science 2006, 314, 781. [4] a) A. Bjorklund, O. Lindvall, Nat. Neurosci. 2000, 3, 537; b) J. H. Kim, J. M. Auerbach, J. A. Rodriguez-Gomez, I. Velasco, D. Gavin, N. Lumelsky, S. H. Lee, J. Nguyen, R. Sanchez-Pernaute, K. Bankiewicz, R. McKay, Nature 2002, 418, 50; c) O. Lindvall, Z. Kokaia, Nature 2006, 441, 1094. [5] a) C. B. Johansson, S. Momma, D. L. Clarke, M. Risling, U. Lendahl, J. Frisen, Cell 1999, 96, 25; b) G. Martino, S. Pluchino, Nat. Rev. Neurosci. 2006, 7, 395; c) F. Barnabe-Heider, J. Frisen, Cell Stem Cell 2008, 3, 16. [6] a) K. G. Bath, F. S. Lee, Dev. Neurobiol. 2010, 70, 339; b) N. Matsuda, H. Lu, Y. Fukata, J. Noritake, H. F. Gao, S. Mukherjee, T. Nemoto, M. Fukata, M. M. Poo, J. Neurosci. 2009, 29, 14185; c) H. J. Song, C. F. Stevens, F. H. Gage, Nat. Neurosci. 2002, 5, 438. [7] a) X. J. Li, X. Q. Zhang, M. A. Johnson, Z. B. Wang, T. LaVaute, S. C. Zhang, Development 2009, 136, 4055; b) N. E. Szabo, T. Y. Zhao, X. L. Zhou, G. Alvarez-Bolado, J. Neurosci. 2009, 29, 2453. [8] M. Warashina, K. H. Min, T. Kuwabara, A. Huynh, F. H. Gage, P. G. Schultz, S. Ding, Angew. Chem., Int. Ed. 2006, 45, 591. [9] a) A. Banerjee, M. Arha, S. Choudhary, R. S. Ashton, S. R. Bhatia, D. V. Schaffer, R. S. Kane, Biomaterials 2009, 30, 4695; b) N. D. Leipzig, M. S. Shoichet, Biomaterials 2009, 30, 6867; c) K. Saha, A. J. Keung, E. F. Irwin, Y. Li, L. Little, D. V. Schaffer, K. E. Healy, Biophys. J. 2008, 95, 4426; d) A. K. Magnusson, P. Linderholm, C. Vieider, M. Ulfendahl, A. Erlandsson, J. Neurosci. Res. 2008, 86, 2363. [10] J. Tang, R. Peng, J. D. Ding, Biomaterials 2010, 31, 2470. [11] M. M. Harper, E. A. Ye, C. C. Blong, M. L. Jacobson, D. S. Sakaguchi, J. Mol. Neurosci. 2010, 40, 269. © 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim small 2010, 6, No. 22, 2509–2513 Controlling Differentiation of Neural Stem Cells [12] a) S. Y. Park, S. Namgung, B. Kim, J. Im, Y. Kim, K. Sun, K. B. Lee, J. M. Nam, Y. Park, S. Hong, Adv. Mater. 2007, 19, 2530; b) C. M. Nelson, R. P. Jean, J. L. Tan, W. F. Liu, N. J. Sniadecki, A. A. Spector, C. S. Chen, Proc. Natl. Acad. Sci. USA 2005, 102, 11594; c) S. A. Ruiz, C. S. Chen, Stem Cells 2008, 26, 2921. [13] a) G. Forte, F. Carotenuro, F. Pagliari, S. Pagliari, P. Cossa, R. Fiaccavento, A. Ahiuwalia, G. Vozzi, B. Vinci, A. Serafino, A. Rinaldi, E. Tizaversa, L. Carosella, M. Minieri, P. Di Nardo, Stem Cells 2008, 26, 2093; b) T. P. Kraehenbuehl, P. Zammaretti, A. J. Van der Vlies, R. G. Schoenmakers, M. P. Lutolf, M. E. Jaconi, J. A. Hubbell, Biomaterials 2008, 29, 2757; c) T. P. Kraehenbuehl, L. S. Ferreira, P. Zammaretti, J. A. Hubbell, R. Langer, Biomaterials 2009, 30, 4318. small 2010, 6, No. 22, 2509–2513 [14] a) R. S. Taichman, Blood 2005, 105, 2631; b) M. J. Kiel, S. J. Morrison, Nat. Rev. Immunol. 2008, 8, 290; c) G. B. Adams, D. T. Scadden, Nat. Immunol. 2006, 7, 333. [15] H. Agheli, J. Malmstrom, E. M. Larsson, M. Textor, D. S. Sutherland, Nano Lett. 2006, 6, 1165. [16] a) M. Nakajima, T. Ishimuro, K. Kato, I. K. Ko, I. Hirata, Y. Arima, H. Iwata, Biomaterials 2007, 28, 1048; b) A. Ruiz, L. Buzanska, D. Gilliland, H. Rauscher, L. Sirghi, T. Sobanski, M. Zychowicz, L. Ceriotti, F. Bretagnol, S. Coecke, P. Colpo, F. Ross, Biomaterials 2008, 29, 4766; c) B. Knoll, C. Weinl, A. Nordheim, F. Bonhoeffer, Nat. Protoc. 2007, 2, 1216. Received: August 4, 2010 Published online: September 21, 2010 © 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.small-journal.com 2513 Biomedical Applications Carbon Nanotube Monolayer Cues for Osteogenesis of Mesenchymal Stem Cells Ku Youn Baik, Sung Young Park, Kwang Heo, Ki-Bum Lee, and Seunghun Hong* Recent advances in nanotechnology present synthetic bioinspired materials to create new controllable microenvironments for stem cell growth, which have allowed directed differentiation into specific lineages.[1,2] Carbon nanotubes (CNTs), one of the most extensively studied nanomaterials, can provide a favorable extracellular environment for intimate cell adhesion due to their similar dimension to collagen. It has been shown that CNTs support the attachment and growth of adult stem cells[3–6] and progenitor cells including osteoblasts and myoblasts.[7,8] In addition, surface-functionalized CNTs provide new opportunities in controlling cell growth. Surface functionalization improves the attachment of biomolecules, such as proteins, DNA, and aptamers, to CNTs.[9] Zanello et al. cultured osteoblasts on CNTs with various functional groups and showed reduced cell growth on positively charged CNTs.[10] Recent reports have shown that human mesenchymal stem cells (hMSCs) formed focal adhesions and grew well on single-walled CNTs (swCNTs).[5,6] However, the effect of naïve swCNT substrates on the differentiation of stem cells has not been reported before. Herein, we report the osteogenic differentiation of hMSCs induced by swCNT monolayer cues without any chemical treatments. Interestingly, the surface treatment of swCNTs via oxygen plasma showed synergistic effects on the differentiation as well as the adhesion of hMSCs. The stress due to the enhanced cell spreading on swCNT layers was proposed as a possible explanation for the Dr. K. Y. Baik, Prof. S. Hong Department of Physics and Astronomy Seoul National University Seoul, 151–747, Korea Dr. K. Y. Baik Plasma Bioscience Research Center Kwangwoon University Seoul, 139–701, Korea S. Y. Park, K. Heo, Prof. S. Hong Interdisciplinary Program in Nano-Science and Technology Seoul National University Seoul, 151–747, Korea Prof. K.-B. Lee Department of Chemistry and Chemical Biology, Rutgers The State University of New Jersey NJ 08854, USA Prof. S. Hong Department of Biophysics and Chemical Biology (WCU Program) Seoul National University Seoul, 151–747, Korea E-mail: [email protected] DOI: 10.1002/smll.201001930 small 2011, 7, No. 6, 741–745 enhanced osteogenesis of hMSCs on the swCNT monolayers. Previous reports showed that the stress to stretch stem cells on microscale molecular patterns generated the tension on actin filaments, which eventually enhanced the osteogenesis.[11,12] Since our method relies on monolayer coating of swCNTs, it can be applied to a wide range of substrates including conventional scaffolds without any complicated fabrication processes. Figure 1 shows a schematic diagram depicting our experimental procedure. In this experiment, three substrates were used: glass as a control, a pristine swCNT monolayer adsorbed on a glass surface, and an oxygen-plasma-treated swCNT (O-swCNT) monolayer on a glass surface (Figure S1a in the Supporting Information (SI)).[13] swCNTs had the average diameter of approximately 1–2 nm, and their average length was 1.5 μm. When the cleaned cover slips were dipped in a dispersed swCNT solution, swCNTs were adsorbed onto the glass surface to form a monolayer. Oxygen plasma treatment was performed to modify the surface chemistry of swCNTs, which is known to generate hydroxyl or carboxyl groups on the surface.[14,15] Atomic force microscopy (AFM) analysis revealed that the swCNT layer maintained its surface roughness even after 40 s of the oxygen plasma treatment, while its contact angle decreased abruptly after 20 s of exposure to oxygen plasma (Figure S1b in the SI). Our control experiments show that, as the time for plasma treatment increased, the abrupt change of contact angle or surface roughness occurred at around 20 or 40 s, respectively. Due to such an abrupt transition, the properties of swCNT layers treated with oxygen plasma for 20 or 40 s usually exhibited rather large variations, which reduced the reproducibility of the following stem cell growth experiments. On the other hand, we could reproducibly obtain similar surface properties for the swCNT layers after 30 s of plasma treatment, thus enabling reliable stem cell growth experiments. Consequently, in our experiment, we utilized swCNT monolayers treated with oxygen plasma for 30 s, which reproducibly exhibited hydrophilic properties while maintaining the roughness of pristine swCNT monolayers. The hMSCs from bone marrow were seeded on these substrates, and then their adhesion, proliferation, and differentiation were examined. Figure 2 shows the adhesion and proliferation of hMSCs cultured on glass (Figure 2a), a swCNT monolayer (Figure 2b), and an O-swCNT monolayer (Figure 2c). Their actin filaments and nuclei were visualized after 24 h from seeding. It is notable that hMSCs spread wider, and their actin fibers look thicker on swCNT monolayers than on glass substrates. For quantitative analysis of cell adhesion, the area of 200 individual cells on each substrate was estimated from that of stained actin fibers (see Figure S2 in the SI) and utilized to © 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim wileyonlinelibrary.com 741 communications K. Y. Baik et al. in hMSCs cultured on an innate swCNT film.[6] Other researchers also have observed wider adhesion of various cells on swCNT substrates than on conventional substrates such as glass. Osteoblast-like cells and hMSCs spread better on a swCNT film than on glass substrates, resulting in the larger cell area and higher occurrence of filopodia at the cell boundaries.[6,17] Besides the effects of nanoscale surface roughness,[16] our results also indicate that surface chemistry plays a role in cellular interaction between cells and swCNTs. Note that even though our swCNT monolayers with or without oxygen plasma treatment had similar surface roughness Figure 1. Schematic diagram depicting hMSC growth on swCNT (Figure S1 in the SI), hMSCs on O-swCNT substrates exhibmonolayers. a) Glass substrate was used as a control. b) swCNTs were adsorbed onto the glass substrate to form a swCNT monolayer. ited an enhanced cell area and proliferation compared with c) Oxygen plasma treatment was applied to modulate the swCNT surface those on pristine swCNT substrates (Figure 2). Presumably, the chemical changes of the O-swCNT layer, such as enhanced properties. hydrophilicity and surface oxygen content, increased the procalculate the average area per cell (Figure 2d). The results liferation and adhesion of hMSCs on it.[18] The morphological change of hMSCs has been reported to show that the adherent area of individual hMSCs is higher on swCNT or O-swCNT monolayers compared with that on the be related to their capacity for multipotentiality. For example, glass substrate. At the same time, the ratio of long and short spindle-shaped hMSCs have high potential for adipogenesis, axial lengths of the hMSCs is smaller on swCNT monolayers while flat hMSCs have high potential for osteogenesis.[19,20] and much smaller on O-swCNT monolayers than on glass Similarly, the actin cytoskeleton changed from thin and parsubstrates (Figure 2e). The MTS (3-(4,5-dimethylthiazol-2- allel microfilament bundles to thick and crisscrossed bundles yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetra- under the osteogenic differentiation process, which is conzolium) assay results show that the proliferation of hMSCs sistent with Figure 2b,c.[21,22] This implies that the swCNT substrates could enhance the osteogenesis of hMSCs. was enhanced on the O-swCNT substrate (Figure 2f). In order to check the osteogenic induction by the swCNT One possible explanation regarding the enhanced cell area on the swCNT substrate can be the nanoscale rough- monolayer, the osteogenic proteins and corresponding genes ness of the swCNT monolayers. Zhang et al. reported that were tested. Core binding factor alpha1 (CBFA1), osteocalcin nanoscale surface roughness of swCNTs can deform the cell (OCN), and alkaline phosphatase (ALP) were used as ostemembrane and hinder the motion of vesicles inside cells.[16] ogenic markers. CBFA1 is the main transcription factor for This membrane deformation may affect the distribution committing hMSCs to the osteoblastic lineage; OCN is a difand diffusion of membrane proteins including focal adhe- ferentiated osteoblast-specific gene for mineralizing the bone sion proteins, which are critical factors in cell adhesion. Tay matrix, and ALP is an early osteoblastic marker.[23,24] hMSCs et al. showed increased number of focal adhesion proteins were cultured on three different substrates with and without osteogenic induction media for 17 days. Immunostaining was performed at day 12, and the quantitative gene analysis was performed at day 7 and 14 from seeding. Figure 3a shows that the hMSCs filled the whole area on all substrates at day 12. Also, note that the cells cultured in osteogenic induction media were slightly detached from the glass due to strong cell–cell interactions. The detachment was retarded on swCNT substrates (data not shown). The osteogenic protein OCN was detected and visualized with fluorescent dyes (Figure 3b). The hMSCs grown on swCNT substrates exhibited brighter OCN immunofluorescence than those on glass substrates, but it was not statistically Figure 2. Adhesion and proliferation of hMSCs on various substrates. Fluorescence images significant (Figure S3 in the SI). However, of actin filaments show the morphology of hMSCs on a) a glass substrate, b) a swCNT the hMSCs grown on O-swCNT substrates monolayer, and c) an oxygen-plasma-treated swCNT monolayer (O-swCNT). Scale bars are definitely exhibited brighter immunofluo100 μm. Quantitative analysis was visualized with d) the averaged value of area per cell (number of cells, n = 200), e) the averaged value of the ratio of long and short axial lengths rescence than those on glass substrates, (a/b in Figure 2b; n = 200), and f) the averaged MTS assay value at day 6 (n = 3). In all and quantitative statistical analysis showed analyses, the student’s t-test was utilized for the significance calculation (∗: p < 0.05). much more brightness on O-swCNT 742 wileyonlinelibrary.com © 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim small 2011, 7, No. 6, 741–745 Carbon Nanotube Monolayer Cues for Osteogenesis of Mesenchymal Stem Cells Figure 3. Immunohistochemistry of an osteogenic protein and mRNA analysis of hMSCs on glass substrates, swCNT substrates, and O-swCNT substrates with normal culture media; and hMSCs cultured on glass substrates with osteogenic induction media. Scale bars are 100 μm. a) Brightfield images, b) osteocalcin (OCN) protein immunostaining at day 12, c) quantitative polymerase chain reaction (qPCR) data of mRNA extracted from hMSCs on each substrate at day 7 and 14. An early marker of osteogenic commitment (CBFA1), a late marker of developing osteoblasts (OCN), and an early osteoblastic marker (ALP) were tested (n = 5). The results indicate the enhanced osteogenic differentiation of hMSCs on swCNT monolayer and on an O-swCNT monolayer without any differentiation-inducing chemicals. substrates. This high protein expression was confirmed in mRNA expression. The mRNA of beta-actin was incorporated as an endogenous housekeeping gene for all the test substrates, and the relative transcription level to that of the hMSCs cultured on glass is shown in Figure 3c. After 14 days of culture, the osteogenic genes such as CBFA1 and OCN were upregulated on both swCNT and O-swCNT substrates.[24] In the case of ALP, the expression was enhanced only on the O-swCNT substrates in the early days of culture (Figure 3c; Figure S4 in the SI). The brighter fluorescence and enhanced mRNA expression indicate the enhanced commitment of hMSCs on O-swCNT monolayers for osteogenesis without any differentiation-inducing chemicals. Although the enhanced osteogenic function of osteoblast on CNTs have been reported previously,[25,26] this is the first report regarding the enhanced commitment of hMSCs to osteoblast lineage on a CNT monolayer. A possible explanation for the enhanced osteogenesis can be the stress on cells due to the enhanced cell spreading on the swCNT monolayers. It was reported that the stress to stretch a stem cell generated the tension on actin filaments and eventually enhanced the osteogenesis.[11,12,27] In order to verify this hypothesis, we fabricated differently sized swCNT square patterns. The size of the patterns was determined considering a previous report utilizing 12 μm extracellular matrix (ECM) molecular patterns to enhance adipogenesis and 100 μm small 2011, 7, No. 6, 741–745 patterns to osteogenesis.[11] We fabricated patterns larger than 100 μm to see only the effect of enhanced spreading on osteogenesis. Figure 4a shows the bright field images of hMSCs cultured for 12 days. The hydrophobic molecules efficiently blocked the hMSC adhesion, resulting in the confined growth of hMSCs only in the swCNT regions.[3] A closer look at cell morphologies showed that the nuclei were located at the center of the pattern, and cytoplasm was extended to the boundaries. Actin fibers were anchored and stretched to the focal adhesions located at the boundaries (Figure S5 in the SI). This indicates that the hMSCs tended to spread out to the boundaries of swCNT patterns using focal adhesion, and thus actin fibers were stretched as the square size increased. The DAPI stained nuclei count 2, 13, 17, and 17 for 100, 200, 300, and 400 μm square patterns, respectively (Figure 4b). The spreading area per cell was calculated by dividing the actin stained area by the cell number. Data from 20 square swCNT patterns were plotted in Figure 4d. As the pattern size increased, the cell density decreased and the averaged area per cell increased. This implies that the hMSCs in larger swCNT patterns were spread wider. The OCN immunofluorescence value per single cell was averaged from nine of each swCNT pattern size (Figure 4c), and the values were plotted in Figure 4e. The results indicate that hMSCs on larger square patterns exhibited easier commitment to the osteogenic lineage. This is consistent with our hypothesis that larger cell spreading in swCNT patterns enhanced the osteogenesis. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim wileyonlinelibrary.com 743 communications K. Y. Baik et al. differentiation-inducing media. In addition, we found that simple oxygen plasma treatment amplifies the adhesion, proliferation, and even osteogenic differentiation of hMSCs by adding chemical effects on the main topographical effects. The stress on hMSCs due to the enhanced cell spreading on swCNT monolayers was proposed as a possible explanation. However, a more detailed mechanism, involving such details as the signal pathways in the stem cells, is not yet clear and requires further study. In any case, this work suggests that the swCNTs can be a powerful scaffold in mesenchymal stem cell engineering. Experimental Section Preparation of CNT Substrates: swCNTs in powdered form (Carbon Nanotechnologies, Inc.) were sonicated in dichlorobenzene (0.05 mg mL−1), and cleaned glass cover slips (piranha solution; H2SO4:H2O2 = 3:1) were dipped in dispersed swCNT solution so that swCNTs were adsorbed onto the glass surface. In this case, the first adsorbed swCNTs blocked the additional adsorption of swCNTs, resulting in monolayer coverage with a certain maximum density.[28] After 2 min, the swCNT-coated glass Figure 4. Immunohistochemistry and fluorescence quantification of confined hMSCs in was rinsed with dichlorobenzene vigorously differently sized square patterns of swCNT monolayers. Scale bars are 100 μm. a) Bright-field to remove any weakly adsorbed swCNTs, and images of hMSCs on the square patterns of a swCNT monolayer after 12 days. b) Fluorescence images of nuclei stained with 4’,6-diamidino-2-phenylindole (DAPI). The panels show 2, 13, then dried with N2 gas. Oxygen plasma treat17, or 17 cells in 100, 200, 300, or 400 μm sized square patterns, respectively. c) Fluorescence ments were performed to modify the surface images of OCN immunostaining, indicating osteogenic differentiation. d) Area per cell in chemistry of the swCNTs (Expanded Plasma differently sized square patterns. Measurements were averaged over 20 squares for each Cleaner (PDC-002) from Harrick Plasma, size. e) Intensity of OCN immunostaining per cell in differently sized square patterns. Values radio frequency (RF) power ≈ 30 W, pressure were averaged over 9 squares for each size. The results indicate that enhanced osteogenesis ≈ 120 mTorr; 1 mTorr = 0.133 Pa). is accompanied with enlarged cell area. Cell Culture and Reagents: hMSCs from human bone marrow (purchased from As previously reported, the enhanced adhesion might be Lonza, Walkersville, USA) were expanded in MSC growth medium linked to Rho-family GTPase (guanine triphosphatase) sign- (MSCGM) and used for our experiments at passage 4–6 in culaling and nonmuscle myosin contraction within the cell. They ture medium (high-glucose Dulbecco’s Modified Eagle Medium showed that the overexpression of ras homolog gene family (Gibco) + 10% fetal bovine serum (FBS; Gibco) + 1% penicillinmember A (RhoA) or Rho-associated coiled-coil-containing streptomycin (Gibco)). All the substrates were cleaned with 70% protein kinase 1 (Rock1) stimulated myosin contraction and ethanol and phosphate buffered saline (PBS) to remove residual promoted osteogenesis.[11] Microarray analysis and pathway toxic solvents. The cells were seeded with a density of about inhibition studies should be followed to elucidate a more 3000 cells cm−2 on the prepared substrates, and culture media detailed mechanism. Furthermore, the effect of plasma was changed every 2–3 days. For osteogenic differentiation, treatment on swCNT should be studied systemically. As the hMSCs were cultured in osteogenic differentiation media (100 nM plasma treatment was known to enhance protein adsorption dexamethasone, 50 μM ascorbic acid, and 10 mM glycerol 2onto swCNT surfaces, it might have an influence on the cel- phosphate in culture medium).[11] Immunohistochemistry: To stain actin fibers, cells were fixed in lular interaction with the nanostructured surface. Interestingly, surface chemical modulation with –OH and –COOH 4% formaldehyde solution, permeabilized with 0.1% Triton X-100, groups itself was reported to downregulate the osteogen- and then stained by tetramethylrhodamine isothiocyanate (TRITC)esis.[28] This implies that there is a synergistic effect of both conjugated phalloidin (1:100, Molecular Probes). For osteocalcin staining, cells were fixed with 4% formaldehyde solution in PBS, persurface roughness and chemistry on O-swCNT substrates. In summary, we showed that the osteogenic differentiation meabilized with 0.1% Triton X-100 in PBS, blocked with 10% normal of hMSC was promoted from swCNT monolayers without goat serum for 1 h at room temperature, and incubated with mouse 744 wileyonlinelibrary.com © 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim small 2011, 7, No. 6, 741–745 Carbon Nanotube Monolayer Cues for Osteogenesis of Mesenchymal Stem Cells anti-human osteocalcin IgG (1:100 dilution, 50114, QED Bioscience Inc.) for 1 h at room temperature. The second antibody (1:500 dilution, Alexa-Fluor 488 conjugated anti-mouse IgG, Sigma) was then adhered for immunofluorescence. For vinculin staining, monoclonal anti-vinculin antibody produced in mouse (1:100, Sigma) was used. After counterstaining the nuclei with DAPI (prolong gold antifade reagent with DAPI, Invitrogen), fluorescence images were obtained using a Nikon Eclipse TE2000-U microscope and a complementary metal-oxide semiconductor camera (INFINITY1–1C, Lumenera Corp.). Cell Proliferation Test: The cells were seeded at the same density, and the number of cells on each substrate was determined using CellTiter 96 Aqueous One Solution Cell Proliferation Assay (G3580, Promega) and spectrometer (HP845, Hewlett-Packard). Cell Area Measurement: Actin filaments of cells were stained with phalloidin. The fluorescence signal was converted to the values of ‘0’ or ‘1’ (black and white) by subtracting the averaged background value. The number of pixels whose value is ‘1’ was counted to calculate the area of each cell. The area of 200 cells for each case was measured. qPCR Analysis: Total RNA was extracted from hMSCs using RNeasy Mini Kit (74104, Qiagen), and converted to cDNA using reverse transcriptase and random primers (ImProm-II Reverse Transcription System, Promega). The same amount of total RNA was used in cDNA synthesis. Resulting cDNAs was used in qPCR (7300 Real Time PCR system, Applied Biosystems) with the primers for β-Actin (NM_001101.3), CBFA1 (NM_001015051), OCN (NM_199173.3), and ALP (NM_000478.3). Preparation of Micropatterned Substrates: Micropatterned substrates were prepared by photolithography. Photoresist (AZ 5214) patterns were first prepared via photolithography, and the substrate was immersed in octadecyltrichlorosilane (OTS, Aldrich) solution (1:250 v/v in anhydrous hexane) for 5 min to cover bare SiO2 regions with OTS molecules. The substrate was then sonicated in acetone and methanol solution to remove the photoresist, resulting in hydrophobic OTS self-assembled monolayer (SAM) patterns. When the patterned substrate was immersed in swCNT (Carbon Nanotechnologies, Inc.) solution (0.05 mg mL−1 in 1,2-dichlorobenzene) for 1 min, swCNTs were adsorbed only on the bare surface regions, while the CH3-terminated SAM prevented their adsorption. After that, the substrate was immersed in the same SAM solution to passivate remaining bare surface regions in the CNT patterned region. Supporting Information Supporting Information is available from the Wiley Online Library or from the author. Acknowledgements This work was supported by the National Research Foundation grant (No. 2010–0000799) and the International Research & Development Program from the MEST (No. 2010–00293). SH acknowledges support from the Converging Research Center Program (No. 2010k001138) and the Happy tech. program (No. 20100020821) from the MEST. KBL acknowledges the NIH Directors’ Innovator small 2011, 7, No. 6, 741–745 Award (1DP20D006462–01) and is also grateful to the NJ commission on Spinal Cord grant (09–3085-SCR-E-0). KYB acknowledges the support from the National Research Foundation of Korea Grant funded by the Korean Government (No.2010–0029418). We thank Jaehyuk Choi, Aniruddh Solanki, Birju Shah, and Shreyas Shah for helpful discussions. [1] M. Dalby, N. Gadegaard, R. Tare, A. Andar, M. Riehle, P. Herzyk, C. Wilkinson, R. Oreffo, Nat. Mater. 2007, 6, 997–1003. [2] S. Oh, K. Brammer, Y. Li, D. Teng, A. Engler, S. Chien, S. Jin, Proc. Natl. Acad. Sci. USA 2009, 106, 2130–2135. [3] S. Y. Park, S. Park, S. Namgung, B. Kim, J. Im, J. Kim, K. Sun, K. Lee, J.-M. Nam, Y. Park, S. Hong, Adv. Mater. 2007, 19, 2530–2534. [4] E. Jan, N. A. Kotov, Nano Lett. 2007, 7, 1123–2248. [5] M. Kalbacova, M. K., L. Dunsch, H. Kataura, and U. Hempel. Phys. Status Solidi B 2006, 243, 3514–3518. [6] C. Tay, H. Gu, W. Leong, H. Yu, H. Li, B. Heng, H. Tantang, S. Loo, L. Li, L. Tan, Carbon 2010, 48, 1095–1104. [7] T. Akasaka, A. Yokoyama, M. Matsuoka, T. Hashimoto, S. Abe, M. Uo, F. Watari, Bio-Med. Mater. Eng. 2009, 19, 147–153. [8] X. Li, H. Gao, M. Uo, Y. Sato, T. Akasaka, Q. Feng, F. Cui, X. Liu, F. Watari, J. Biomed. Mater. Res. Part A 2008, 91, 132–139. [9] H. So, K. Won, Y. Kim, B. Kim, B. Ryu, P. Na, H. Kim, J. Lee, J. Am. Chem. Soc. 2005, 127, 11906–11907. [10] L. Zanello, B. Zhao, H. Hu, R. Haddon, Nano Lett. 2006, 6, 562–567. [11] R. McBeath, D. Pirone, C. Nelson, K. Bhadriraju, C. Chen, Dev. Cell 2004, 6, 483–495. [12] J. Settleman, Mol. Cell 2004, 14, 148–150. [13] M. Lee, J. Im, B. Y. Lee, S. Myung, J. Kang, L. Huang, Y.-K. Kwon, S. Hong, Nat. Nanotechnol. 2006, 1, 66. [14] V. Brüser, M. Heintze, W. Brandl, G. Marginean, H. Bubert, Diam. Relat. Mat. 2004, 13, 1177–1181. [15] A. Felten, C. Bittencourt, J. Pireaux, G. Van Lier, J. Charlier, J. Appl. Phys. 2005, 98, 074308. [16] J. Zhang, D. Fu, M. B. Chan-Park, L. Li, P. Chen, Adv. Mater. 2009, 21, 790–793. [17] N. Aoki, A. Yokoyama, Y. Nodasaka, T. Akasaka, M. Uo, Y. Sato, K. Tohji, F. Watari, J. Biomed. Nanotechnol. 2005, 1, 402–405. [18] A. Martins, E. Pinho, S. Faria, I. Pashkuleva, A. Marques, R. Reis, N. Neves, Small 2009, 5, 1195–1206. [19] B. Neuhuber, S. Swanger, L. Howard, A. Mackay, I. Fischer, Exp. Hematol. 2008, 36, 1176–1185. [20] I. Sekiya, B. Larson, J. Smith, R. Pochampally, J. Cui, D. Prockop, Stem Cells 2002, 20, 530–541. [21] J. Rodriguez, M. Gonzalez, S. Rios, V. Cambiazo, J. Cell. Biochem. 2004, 93, 721–731. [22] G. Yourek, M. Hussain, J. Mao, Asaio J. 2007, 53, 219–228. [23] T. Komori, H. Yagi, S. Nomura, A. Yamaguchi, K. Sasaki, K. Deguchi, Y. Shimizu, R. Bronson, Y. Gao, M. Inada, Cell 1997, 89, 755–764. [24] H. Qi, D. Aguiar, S. Williams, A. La Pean, W. Pan, C. Verfaillie, Proc. Natl. Acad. Sci. USA 2003, 100, 3305–3310. [25] S. Sirivisoot, C. Yao, X. Xiao, B. Sheldon, T. Webster, Nanotechnology 2007, 18, 365102–365200. [26] Y. Usui, K. Aoki, N. Narita, N. Murakami, I. Nakamura, K. Nakamura, N. Ishigaki, H. Yamazaki, H. Horiuchi, H. Kato, Small 2008, 4, 240–246. [27] K. Kilian, B. Bugarija, B. Lahn, M. Mrksich, Proc. Natl. Acad. Sci. USA 2010, 107, 4872–4877. [28] H. Curran, R. Chen, J. Hunt, Biomaterials 2006, 27, 4783–4793. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Received: October 29, 2010 Revised: December 12, 2010 Published online: February 7, 2011 wileyonlinelibrary.com 745 ARTICLE Polarization-Controlled Differentiation of Human Neural Stem Cells Using Synergistic Cues from the Patterns of Carbon Nanotube Monolayer Coating Sung Young Park,†,‡ Dong Shin Choi,† Hye Jun Jin,‡ Juhun Park,‡ Kyung-Eun Byun,‡ Ki-Bum Lee,§,* and Seunghun Hong†,‡,^,* † Interdisciplinary Program in Nano-Science and Technology and ‡Department of Physics and Astronomy, Seoul National University, Seoul 151-747, Korea, Department of Chemistry and Chemical Biology, Rutgers, The State University of New Jersey, Piscataway, New Jersey 08854, United States, and ^Deparment of Biophysics and Chemical Biology, Seoul National University, Seoul 151-747, Korea § D ue to the ability to generate the main phenotypes in the nervous systems, neural stem cells (NSCs) offer great potential in regenerative medicine.1 For therapeutic applications, such as rebuilding damaged nerves, one should be able to precisely control the direction and structural polarization of individual axonal growth.2 Previously, attempts have been made to control the structural polarization of cultured neurons by using several key strategies such as molecular cues of diffusible gradient or substrate-bound chemical/ extracellular matrix (ECM) protein patterns and topographical cues.3 12 However, they have several disadvantages. For example, a diffusible gradient is not suitable for the membrane/matrix proteins due to the difficulty of maintaining it over time. In the case of printed protein patterns, the protein molecules may undergo conformational changes during the process of stamping, which often leads to the denaturation and the loss of biological activities.13 Besides, to create optimal nanotopographical cues which can help cell growth, various microor nanofabrication techniques are required, such as electron-beam lithography or chemical/reactive-ion etching.14 On the other hand, various synthetic nanomaterials such as biocompatible nanofibers and carbon nanomaterials have been recently proposed for novel nanostructured scaffolds.14 19 However, neuronal polarization control of NSCs, especially at the level of individual axons or dendrites, has not been demonstrated using these nanomaterials. Herein, we report a method for the structural-polarization-controlled neuronal PARK ET AL. ABSTRACT We report a method for selective growth and structural-polarization-controlled neuronal differentiation of human neural stem cells (hNSCs) into neurons using carbon nanotube network patterns. The CNT patterns provide synergistic cues for the differentiation of hNSCs in physiological solution and an optimal nanotopography at the same time with good biocompatibility. We demonstrated a polarization-controlled neuronal differentiation at the level of individual NSCs. This result should provide a stable and versatile platform for controlling the hNSC growth because CNT patterns are known to be stable in time unlike commonly used organic molecular patterns. KEYWORDS: neural stem cells . carbon nanotubes . polarization . nanotopography . micropattern differentiation of human NSCs (hNSCs) using the patterns of CNT network structures with good biocompatibility. In this strategy, the CNT network patterns provided synergistic cues of selective laminin adsorption and optimal nanotopography, which resulted in selective adhesion and growth of hNSCs on them. CNT network structures were found to induce the enhanced adhesion and growth of hNSCs even better than conventional cell-culture substrates, such as glass. CNT patterns with various geometries were utilized to explore their effect on the outgrowth of hNSC during the growth and differentiation process. As a proof of concept, a structural-polarization-controlled neuronal differentiation using CNT network patterns was demonstrated at the level of individual axons and neurites. Furthermore, we applied our strategy for the controlled hNSC growth on flexible and biocompatible polymer substrates such as polyimide. Since CNT monolayer coatings can be applied to versatile substrates and provide stable microenvironments for hNSC growth control even better than commonly used organic VOL. 5 ’ NO. 6 ’ * Address correspondence to [email protected], [email protected]. Received for review February 14, 2011 and accepted May 9, 2011. Published online May 13, 2011 10.1021/nn2006128 C 2011 American Chemical Society 4704–4711 ’ 2011 4704 www.acsnano.org ARTICLE Figure 1. CNT network patterns for selective hNSC growth and polarization. (a) Schematic diagram showing structural-polarization-controlled neuronal differentiation using CNT patterns. CNT monolayer patterns were fabricated on a substrate using a previously reported method,22 and laminin was absorbed selectively on the CNT-coated regions. This structure induced preferential adhesion of hNSCs, finally achieving structural-polarization-controlled neuronal differentiation. (b) SEM image of CNT patterns (dark spots). Scale bar represents 40 μm. (c) Immunofluorescence image of anti-laminin (green) bound to the laminin which was selectively adsorbed on the CNT patterns. The scale bar represents 200 μm. It confirms the selective adsorption of laminin on the CNT. The inset shows the AFM topography image of the laminin-coated CNT monolayer in phosphate buffered saline (PBS). The scale bar in the inset represents 2 μm. (d) Cell viability assay of hNSCs on CNT patterns for 3 day proliferation. The viability was measured by flow cytometry. The obtained data in the graph clearly indicate that 98% of hNSCs grown on the CNT layer were alive (red). molecular patterns, our work should provide a simple but efficient way to control the structural polarization of NSCs and may open up various applications in neural engineering and regenerative medicine. RESULTS AND DISCUSSION Figure 1a shows a schematic diagram illustrating our basic experimental procedure. CNT patterns were prepared according to previously reported methods.20,21 Briefly, a self-assembled monolayer (SAM) of methylterminated 1-octadecanethiol (ODT) was first patterned on thin Au films on cover glass substrates by microcontact printing, while leaving some bare Au surface regions unaltered (see method in Supporting Information). When the patterned substrate was placed in CNT suspensions (0.05 mg/mL in 1,2-dichlorobenznene), CNTs were selectively adsorbed onto bare Au regions, forming CNT monolayer patterns. The CNT patterns were then placed in laminin solution (10 20 μg/mL) for 10 30 min so that laminin molecules were selectively adsorbed onto the CNT patterns. Laminin is one of the ECM components that is helpful for hNSC adhesion and growth. After cell seeding, the hNSCs grew preferentially along these laminin-coated CNT patterns in the culture media with growth factors, such PARK ET AL. as basic fibroblast growth factor (bFGF) and epidermal growth factor (EGF). For hNSCs, the growth factors (bFGF and EGF) are known to enhance hNSC growth and proliferation, while blocking the differentiation process. Afterward, the substrate was placed in culture media without bFGF and EGF for 2 weeks to study the differentiation of hNSCs on laminin-coated CNT patterns. Figure 1b shows the scanning electron micrograph (SEM) image of the prepared CNT patterns. It shows the well-defined CNT regions (darker square regions) as well as ODT-coated area (lighter region). The highresolution atomic force microscopy (AFM) image also confirmed the highly selective adsorption of CNTs on bare Au regions (Figure S1a in Supporting Information). When placed in laminin solution, CNT patterns selectively adsorb laminin molecules from solution. This was verified by immunochemistry (Figure 1c). For this purpose, after the laminin adsorption, the substrate was placed in the fluorescent-labeled anti-laminin solution so that the anti-laminin molecules would bind to the laminin molecules on the substrate. The fluorescence image shows much stronger fluorescence intensity in the CNT regions (brighter green regions in Figure 1c) than on ODT regions, confirming the high-density adsorption VOL. 5 ’ NO. 6 ’ 4704–4711 ’ 4705 2011 www.acsnano.org ARTICLE Figure 2. hNSC growth and differentiation depending on the size of CNT patterns. The phase contrast images of hNSCs grown for 1 day (a,d,g) and those of the differentiated cells for 2 weeks (b,e,h), and the immunofluorescence images of the differentiated cells (c,f,i) are shown. The immunofluorescene markers are Hoechst for nuclei, glial fibrillary acidic protein (GFAP) for astroglial cells, and TUJ1 and neurofilament light (NF-L) for neuronal cells. All scale bars represent 200 μm, unless otherwise noted. The dotted black squares (a,b,d,e) indicate some of the CNT-coated regions. (a c) hNSC growth and differentiation on rather large square-shape CNT patterns (300 μm 300 μm, 200 μm spacing). Note that neural networks were constructed in arbitrary manner after differentiation. The immunofluorescence image (c) shows the differentiated cells positive for the astroglial marker, GFAP (green). (d f) Restrictive neurite growth of hNSCs in individual CNT square patterns (50 μm 50 μm, 50 μm spacing). We did not observe any indication of neurite outgrowth of hNSC after the growth and differentiation from the immunostaining image of NF-L (red). (g i) Outgrowths of hNSCs directed by rather small square-shape CNT patterns (5 μm 5 μm, 5 μm spacing). The inset figure (g) shows that a single hNSC was attached on seven individual CNT square patterns. The immunofluorescence image (i) indicates that the differentiated cells are positive for neuronal cell marker, TUJ1 (red). The scale bar in the phase contrast image (g) represents 100 μm. of laminin molecules on the CNT patterns. It is also consistent with previous reports regarding the preferential adsorption of protein molecules to CNT sidewalls22,23 and the resistance of alkyl chains of the ODT SAM to laminin adsorption.24 The CNT patterns with laminin coating were also investigated via an AFM topography image (inset in Figure 1c and Figure S1b in Supporting Information). It exhibited the average roughness of 26 nm, which is in the optimal range of surface roughness (20 50 nm) promoting the adhesion and longevity of primary neurons.25,26 This result indicates that the nanotopographic cues of CNT network structures as well as laminin molecules adsorbed on the CNT patterns can synergistically induce the selective growth of hNSCs. The biocompatibility of the CNT network structure as a substrate for hNSC growth was investigated via cell PARK ET AL. viability assay using flow cytometry. For the assay, the adherent hNSCs were detached from the CNT patterns after 3 day growth and 3 day differentiation, respectively. After the 3 day growth period, 98% of the cells were found to be viable (Figure 1d). The assay result of a 3 day differentiation also exhibited nearly 97% cell viability (Figure S2 in Supporting Information). This suggests the good biocompatibility of CNT patterns for hNSC growth and differentiation. Furthermore, we utilized the Western Blot method to confirm the protein expression of hNSCs before and after the differentiation (Figure S3 in Supporting Information). The results show that the hNSCs grown with the growth factors (EGF and bFGF) were positive for neural stem cell markers (nestin and SOX2), which shows that they just proliferated and undifferentiated. Meanwhile, those VOL. 5 ’ NO. 6 ’ 4704–4711 ’ 4706 2011 www.acsnano.org ARTICLE Figure 3. Control of hNSC orientation using line shape CNT patterns. (a c) hNSC growth on CNT line shape patterns (30 μm width). The hNSCs inside individual 30 μm wide line patterns were observed to grow, extending their neurites in the same direction along the predefined CNT line patterns. (d f) Individual hNSC growth on each CNT line pattern (5 μm width). The hNSCs were aligned to form a bipolar shape on the CNT line patterns during the growth and differentiation. (g i) Neural network formed on narrow line shape CNT patterns combined with large square-shape ones. Note that highly oriented hNSC growth was induced by the predefined CNT patterns, and eventually well-organized neural networks were formed after differentiation. (a,d,g) Phase contrast images of hNSC grown for 1 day, and the scale bars in the phase contrast images are 200 μm. (b,e,h) Phase contrast images of the differentiated cell. (c,f,i) Immunofluorescence images of the differentiated cells. The scale bars in the phase contrast images are 50 μm. The dotted black squares indicate some of the CNT-coated regions. grown without these growth factors were positive for glial fibrillary acidic protein (GFAP) and neuron-specific class III β-tubulin (TUJ1), which indicates that they differentiated. When the hNSCs were seeded on the laminin-coated CNT patterns in the culture media with the growth factors, they selectively adhered onto the CNT pattern regions and grew along the patterns (Figure 2a,d,g). In this stage, the growth factors blocked the differentiation of hNSCs. When the substrate was placed in the culture media without the growth factors, the hNSCs started to differentiate (Figure 2b,e,h). The differentiation was confirmed by immunocytochemistry (Figure 2c,f,i). Here, we used three different markers to look at cytoskeletal distributions on the CNT patterns after the differentiation: GFAP as an astroglial cell marker (Figure 2c), neurofilament light (NF-L, Figure 2f), and TUJ1 (Figure 2i) as neuronal cell markers. We also performed an experiment to investigate the hNSC PARK ET AL. growth on CNT networks compared with that on conventional substrates such as coverglass (Figure S5 in Supporting Information). After the hNSC seeding on the laminin-coated CNT patterns that were prepared on coverglass (Figure S5A), we observed that the hNSCs grew selectively in the CNT regions (Figure S5B D). This result clearly indicates that the CNT network can provide a better extracellular environment for hNSC growth than conventional cell-culture substrates such as coverglass. Depending on the geometries of CNT patterns, the hNSCs exhibited significantly different outgrowing behaviors during growth and differentiation (Figure 2a c). When the size of the CNT square patterns was large enough (300 μm 300 μm, 200 μm spacing) to hold multiple cells, the hNSCs in the CNT patterns could maintain their cell cell interactions and proliferated very well (Figure 2a). Eventually, they outgrew over the 200 μm wide ODT SAM regions toward the adjacent VOL. 5 ’ NO. 6 ’ 4704–4711 ’ 4707 2011 www.acsnano.org ARTICLE Figure 4. Control of hNSC growth and differentiation on biocompatible and flexible polyimide (PI) substrate. (a) Optical image of a polyimide membrane with CNT patterns, which is flexible and transparent. (b) Immunofluorescence image of antilaminin (green). It confirms that the laminin was selectively adsorbed onto the CNT patterns on PI substrate. Scale bar represents 200 μm. (c) Phase contrast image of selective hNSC adhesion on CNT patterns on PI after cell seeding. Scale bar represents 200 μm. (d) Immunofluorescence image of the differentiated hNSCs on CNT patterns on PI (TUJ1 for neural cells and GFAP for astroglial cells). The inset shows the magnified image of the region marked by the white solid square. Scale bar represents 200 μm, and that of the inset represents 50 μm. It should be noted that the orientation-controlled neural networks were constructed along the CNT patterns on the PI membrane. CNT square patterns and formed the neural networks, where the cells grown on the distanced CNT square patterns were connected (Figure 2b). The fluorescence image clearly shows that the outgrowing astrocytes (green regions marked as GFAP) were connecting the hNSC population on the distanced patterns after differentiation (Figure 2c). We then reduced the size of CNT square patterns (50 μm 50 μm, 50 μm spacing) such that each square could hold only a single hNSC (Figure 2d f). In this case, the hNSC outgrowth was extremely restricted during the growth and differentiation process (Figure 2d). Even after we removed the growth factors to induce differentiation, the hNSCs did not exhibit any indication of major outgrowth over the ODT regions (Figure 2e). The fluorescence image of neuronal cytoskeletons (NF-L, red) does not show any outgrowing hNSCs from the patterns (Figure 2f). This result clearly shows that the cell cell interaction can be controlled by the geometries of CNT patterns, which can be critical for hNSC growth and differentiation. We also tested hNSC behaviors on CNT square patterns smaller (5 μm 5 μm, 5 μm spacing) than individual hNSCs (Figure 2g i). Here, the hNSCs first adhered and outgrew over several CNT square patterns (Figure 2g). Note that each cell was bound strongly on the small CNT pattern regions and outgrew and extended over the ODT regions. In this case, the spacing of the CNT square patterns should significantly affect PARK ET AL. the cytoskeletal tensions of the individual hNSCs, which probably should affect the differentiation of stem cells.27 After the differentiation, we could observe that the neuronal outgrowths extended and bound on the nearby CNT patterns (Figure 2h). The fluorescence image clearly shows the neuronal cytoskeletal marker (TUJ1, green) indicating the connected neural networks bound on the small CNT patterns. Overall, the results in Figure 2 clearly show that the size and spacing of CNT patterns can play a critical role in controlling the hNSC outgrowths during the growth and differentiation process, which can possibly affect cell cell interactions or cytoskeletal tensions. Furthermore, line shape CNT patterns can be utilized to control the neuronal orientation with high precision (Figure 3). In the line shape CNT patterns with a line width (30 μm width, 60 μm spacing) that can hold two or three cells, the hNSCs adhered (Figure 3a) along the line pattern. We observed that they differentiated to form neural networks along the inside of the line patterns (Figure 3b). Here, the differentiation was also confirmed by immunocytochemistry with the TUJ1 marker. When the CNT line width was narrowed down to about 5 μm, which can hold only a single hNSC, the hNSCs grew and differentiated into bipolar shapes along the individual CNT line patterns (Figure 3f). Significantly, this result indicates that we can control the orientation of hNSCs with single-celllevel precision. VOL. 5 ’ NO. 6 ’ 4704–4711 ’ 4708 2011 www.acsnano.org ARTICLE Figure 5. Structural-polarization-controlled neuronal differentiation of individual hNSCs using CNT patterns. (a) SEM image of a CNT pattern with a single narrow strip as shown in dark gray. (b) Phase contrast images of hNSC adhesion on CNT patterns. The dotted square and line (red) represent the CNT patterns. Scale bar represents 50 μm. After cell seeding, the cell bodies of hNSCs were attached within the CNT square region. (c) Phase contrast images of the differentiated cells on the CNT patterns. Note that the growing parts in the hNSCs were observed along the CNT single narrow strip regions during the differentiation. (d) Immunofluorescence images of growth-associated protein 43 (GAP 43, green) and Hoechst (blue, for nucleus). Scale bar represents 50 μm. It should be noted that the GAP 43 (green dots) was distributed along the narrow strip region. (e) Immunofluorescence image of GFAP (green), TUJ1 (red), and Hoechst (blue). Scale bar represents 50 μm. It should be noticed that the differentiated neuronal cells (TUJ1, red) were surrounded by astroglial cells (GFAP, green) on the structural-polarizationcontrolled CNT pattern, where the neuronal polarization was also directed by the CNT narrow strip region. When circle-shape patterns are connected with narrow line shape ones, we observed quite an interesting hNSC behavior during the growth and differentiation (Figure 3g i). After being seeded in the culture media with growth factors, the cell bodies of hNSCs tended to adhere and proliferated on the circle-shape pattern regions (Figure 3g). After withdrawal of the growth factors in the culture media, they started to differentiate and the outgrowing neurites were observed mostly along the narrow line shape CNT pattern regions (Figure 3h). The differentiation was also confirmed via immunostaining (Figure 3i). Since the hNSCs first adhered and grew on the circle-shape patterns, their nuclei (blue regions) were mostly located on the circle regions, while the long neurites (red regions) extended along the line shape regions (Figure 3i). This result indicates that the CNT patterns can be utilized to control both of the locations of cell nuclei and the direction of neurite growth, thus allowing us to control the structural polarization of the neuronal differentiation of hNSCs. Furthermore, the synapse formation of the neurons was checked by a neuronal presynaptic vesicle marker, synaptophysin (Figure S6 in Supporting Information). The results clearly show that the neurons differentiated from the hNSCs grown on the CNT patterns can also form the synapses, which are important for a neuron to pass a chemical/ electrical signal to another neuron. PARK ET AL. For future therapeutic applications, such as regenerative medicine, it would be crucial to apply our strategy to a flexible and biocompatible substrate such as polyimide (PI) (Figure 4), which has been widely utilized for implantable neural devices such as threedimensional artificial nerve conduits28 and stimulating electrodes.29,30 We prepared CNT patterns on thin Aufilm-coated PI substrates and performed the experiments of hNSC growth and differentiation on them (Figure 4a). We could achieve high-quality CNT patterns on the Au-coated PI substrates, as shown in the SEM images (Figure S7a in Supporting Information). The immunofluorescence image indicates the highly selective adsorption of laminin onto the CNT patterns on the PI substrate (green regions in Figure 4b). After being seeded on it, the hNSCs adhered selectively onto the CNT pattern regions on PI substrates and proliferated (Figure 4c). Eventually, we achieved the orientation-controlled growth and differentiation of hNSCs along the CNT patterns on the flexible PI substrate (Figure 4d and Figure S7b in Supporting Information). Finally, the structural-polarization-controlled differentiation of individual hNSCs can be achieved by CNT patterns composed of one square and one line shape (Figure 5a). Here, the width of the line shape region is much smaller than the size of an individual hNSC. After cell seeding, we were able to observe the selective VOL. 5 ’ NO. 6 ’ 4704–4711 ’ 4709 2011 www.acsnano.org EXPERIMENTAL METHODS Fabrication of CNT Monolayer Patterns. CNT (multiwalled CNT, 98% purified, NanoLab, MA, USA) patterns were fabricated on Au-coated glass substrate according to the methods described previously (Supporting Information).20,21 To prepare a polymer substrate, polyimide (PI, VTEC Polyimide 1388, Richard Blaine International, Inc., PA, USA) in solution was coated on a cover glass by spin coating at 1000 rpm for 1 min and then cured on a hot plate (Supporting Information). CNT patterns on Au-coated PI were generated by the same method as before.20,21 hNSC Culture. Immortalized human NSCs (ReNcell VM, Millipore, Temecula, CA, USA) were purchased and maintained according to the manufacturer's protocol.32 Differentiation was initiated by removal of growth factors such as basic fibroblast growth factor (bFGF) and epidermal growth factor (EGF) from the culture media, and the cells were allowed to differentiate usually for 2 weeks. For the hNSC culture, the prepared CNT patterns were incubated in laminin solution (20 μg/mL, Sigma, MO, USA) for 30 min. The laminin-coated CNT patterns were washed with PBS several times and subsequently seeded with suspensions of hNSC at a cell density of 105/mL. All of the hNSC experiments were carried out between passages 3 and 10. Cell Viability Assay. The hNSCs were either grown for 3 days or subsequently differentiated for 3 days, and then they were used for cell viability assay. The NSCs were first detached and made into 106/mL cell suspensions, of which only a fraction was used for counting cell viability. The cells were incubated with a reagent composed of a mixture of a cell permeant and a noncell permeant dye (ViaCount Reagent, Millipore, Heyward, CA, USA) according to the manufacturer's protocol, and the viability was determined using a single-laser four-color flow cytometry detection system (EasyCyte Plus, Millipore, Heyward, CA, USA) at 500 cells per one flow rate with predefined gating. PARK ET AL. CONCLUSION In summary, we demonstrated a structural-polarization-controlled neuronal differentiation of hNSCs using the patterns of CNT monolayer coating. Due to the synergistic effect of CNT network structures for selective laminin adsorption and optimal nanotopography, we could effectively promote the selective growth of hNSCs on the CNT patterns. The result of the cell viability assay (>97%) suggested the good biocompatibility of CNT patterns for hNSC growth and differentiation. We also confirmed that CNTs could induce the adhesion and growth of hNSCs even better than conventional cell-culture substrates such as bare glass. Importantly, the structural-polarization-controlled neuronal differentiation was demonstrated at the level of an individual axon or neurite. Furthermore, we also applied it to flexible and biocompatible PI substrates, which should significantly expand the possible therapeutic applications of our method. Since CNT monolayer coatings can be applied to versatile substrates including flexible ones and provide a better cell-growth environment than conventional cell-culture substrates such as glass, our strategy should provide many new opportunities in various areas such as neural engineering, stem cell therapy, and regenerative medicine. ARTICLE hNSC adhesion inside the square regions (Figure 5b). Then, the hNSCs on the square region outgrew along the narrow line shape regions during the growth and differentiation stages (Figure 5c). The neuronal differentiation was confirmed by growth associated protein 43 (GAP 43, green in Figure 5d), which is known to be expressed in the growth cone regions of neural cells. We observed that GAP 43 was also highly expressed on the line shape CNT regions, indicating that the neurites outgrew along the line shape regions (Figure 5d and Figure S8 in Supporting Information). We carried out immunocytochemistry to check the neural lineages of the differentiated cells on these CNT patterns (Figure 5e). Here, GFAP and TUJ1 indicate astroglial and neural cells, respectively. To confirm their lineages, the relative fluorescence intensities of GFAP and TUJ1 from the cell nuclei on the square pattern regions were quantified using a method similar to that reported previously (Figure S9 in Supporting Information).31 The result shows that 20% of them were TUJ1-positive, whereas another 20% were GFAP-positive. It should be noted that the hNSCs were differentiated with controlled structural polarity on the CNT patterns, while maintaining their capabilities to differentiate into the main phenotypes in the nervous system, such as neuronal or astroglial cells. Immunocytochemistry. The hNSCs were fixed for 15 min in 4% paraformaldehyde in PBS and permeabilized with 0.1% Triton X-100 in PBS for 15 min, followed by overnight incubation at 4 C in the following primary antibodies: TUJ1 (1:500; clone SDL.3D10, Sigma, MO, USA), GFAP (1:1000; Dako, Glostrup, Denmark), NF-L (1:200; Millipore, Temecula, CA, USA), GAP 43 (1:200; Millipore, Temecula, CA, USA), and synaptophysin (Millipore, Temecula, CA, USA). Cells were washed with PBS, incubated with either goat anti-mouse FITC (1:200; Sigma, MO, USA) or goat anti-rabbit TRITC (1:500; Sigma, MO, USA), then counterstained with 10 mM Hoechst 33342 (Sigma, MO, USA). The mounted samples were imaged using an inverted fluorescence microscope (Nikon, TE2000, Tokyo, Japan) with an EMCCD monochrome digital camera (DQC-FS, Nikon, Tokyo, Japan). ImageJ software (freely downloadable from National Institutes of Health Web site, http://rsbweb.nih.gov/ij/) was used for subsequent processing of the fluorescence images. Acknowledgment. We appreciate J.H. Yi and E. Miljan for the fruitful discussion to evaluate NSC culture and differentiation. We also thank A. Solanki, B. Shah, and S. Shah for helpful discussions. This project has been supported by the NRF Grant (No. 2011-0000390), and partial support from the Happy Tech Program (No. 20100020821). S.H. acknowledges the support from the Converging Research Center program (No. 2010K001138) and the System 2010 program of the MKE. K.-B.L. acknowledges the NIH Directors' Innovator Award (1DP20D006462-01) and is also grateful to the NJ commission on Spinal Cord Grant (09-3085-SCR-E-0). Supporting Information Available: Supplementary methods, additional details on fabrication method, and supplementary figures. This material is available free of charge via the Internet at http://pubs.acs.org. VOL. 5 ’ NO. 6 ’ 4704–4711 ’ 4710 2011 www.acsnano.org 1. Gage, F. H. Mammalian Neural Stem Cells. Science 2000, 287, 1433–1438. 2. Reh, T. A. Neural Stem Cells: Form and Function. Nat. Neurosci. 2002, 5, 392–394. 3. Stenger, D. A.; Hickman, J. J.; Bateman, K. E.; Ravenscroft, M. S.; Ma, W.; Pancrazio, J. J.; Shaffer, K.; Schaffner, A. E.; Cribbs, D. H.; Cotman, C. W. Microlithographic Determination of Axonal/Dendritic Polarity in Cultured Hippocampal Neurons. J. Neurosci. Methods 1998, 82, 167–173. 4. Lauer, L.; Klein, C.; Offenhausser, A. Spot Compliant Neuronal Networks by Structure Optimized Micro-Contact Printing. Biomaterials 2001, 22, 1925–1932. 5. von Philipsborn, A. C.; Lang, S.; Loeschinger, J.; Bernard, A.; David, C.; Lehnert, D.; Bonhoeffer, F.; Bastmeyer, M. Growth Cone Navigation in Substrate-Bound Ephrin Gradients. Development 2006, 133, 2487–2495. 6. Shi, P.; Shen, K.; Kam, L. C. Local Presentation of L1 and N-Cadherin in Multicomponent, Microscale Patterns Differentially Direct Neuron Function In Vitro. Dev. Neurobiol. 2007, 67, 1765–1776. 7. Suter, D. M.; Forscher, P. Substrate-Cytoskeletal Coupling as a Mechanism for the Regulation of Growth Cone Motility and Guidance. J. Neurobiol. 2000, 44, 97–113. 8. Dertinger, S. K. W.; Jiang, X. Y.; Li, Z. Y.; Murthy, V. N.; Whitesides, G. M. Gradients of Substrate-Bound Laminin Orient Axonal Specification of Neurons. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 12542–12547. 9. Mahoney, M. J.; Chen, R. R.; Tan, J.; Saltzman, W. M. The Influence of Microchannels on Neurite Growth and Architecture. Biomaterials 2005, 26, 771–778. 10. Cai, J.; Peng, X.; Nelson, K. D.; Eberhart, R.; Smith, G. M. Permeable Guidance Channels Containing Microfilament Scaffolds Enhance Axon Growth and Maturation. J. Biomed. Mater. Res., Part A 2005, 75, 374–386. 11. Dowell-Mesfin, N. M.; Abdul-Karim, M. A.; Turner, A. M.; Schanz, S.; Craighead, H. G.; Roysam, B.; Turner, J. N.; Shain, W. Topographically Modified Surfaces Affect Orientation and Growth of Hippocampal Neurons. J. Neural Eng. 2004, 1, 78–90. 12. Goldner, J. S.; Bruder, J. M.; Li, G.; Gazzola, D.; Hoffman-Kim, D. Neurite Bridging across Micropatterned Grooves. Biomaterials 2006, 27, 460–472. 13. von Philipsborn, A. C.; Lang, S.; Bernard, A.; Loeschinger, J.; David, C.; Lehnert, D.; Bastmeyer, M.; Bonhoeffer, F. Microcontact Printing of Axon Guidance Molecules for Generation of Graded Patterns. Nat. Protoc. 2006, 1, 1322–1328. 14. Seidlits, S. K.; Lee, J. Y.; Schmidt, C. E. Nanostructured Scaffolds for Neural Applications. Nanomedicine 2008, 3, 183–199. 15. Ferreira, L.; Karp, J. M.; Nobre, L.; Langer, R. New Opportunities: The Use of Nanotechnologies To Manipulate and Track Stem Cells. Cell Stem Cell 2008, 3, 136–146. 16. Kostarelos, K.; Bianco, A.; Prato, M. Promises, Facts and Challenges for Carbon Nanotubes in Imaging and Therapeutics. Nat. Nanotechnol. 2009, 4, 627–633. 17. Hu, H.; Ni, Y. C.; Montana, V.; Haddon, R. C.; Parpura, V. Chemically Functionalized Carbon Nanotubes as Substrates for Neuronal Growth. Nano Lett. 2004, 4, 507–511. 18. Kam, N. W. S.; Jan, E.; Kotov, N. A. Electrical Stimulation of Neural Stem Cells Mediated by Humanized Carbon Nanotube Composite Made with Extracellular Matrix Protein. Nano Lett. 2009, 9, 273–278. 19. Cellot, G.; Cilia, E.; Cipollone, S.; Rancic, V.; Sucapane, A.; Giordani, S.; Gambazzi, L.; Markram, H.; Grandolfo, M.; Scaini, D.; et al. Carbon Nanotubes Might Improve Neuronal Performance by Favouring Electrical Shortcuts. Nat. Nanotechnol. 2009, 4, 126–133. 20. Rao, S. G.; Huang, L.; Setyawan, W.; Hong, S. Nanotube Electronics: Large-Scale Assembly of Carbon Nanotubes. Nature 2003, 425, 36–37. 21. Park, S. Y.; Park, S. Y.; Namgung, S.; Kim, B.; Im, J.; Kim, Y.; Sun, K.; Lee, K. B.; Nam, J. M.; Park, Y.; et al. Carbon Nanotube Monolayer Patterns for Directed Growth of Mesenchymal Stem Cells. Adv. Mater. 2007, 19, 2530–2534. PARK ET AL. 22. Chen, R. J.; Zhang, Y. G.; Wang, D. W.; Dai, H. J. Noncovalent Sidewall Functionalization of Single-Walled Carbon Nanotubes for Protein Immobilization. J. Am. Chem. Soc. 2001, 123, 3838–3839. 23. Wang, S. Q.; Humphreys, E. S.; Chung, S. Y.; Delduco, D. F.; Lustig, S. R.; Wang, H.; Parker, K. N.; Rizzo, N. W.; Subramoney, S.; Chiang, Y. M.; et al. Peptides with Selective Affinity for Carbon Nanotubes. Nat. Mater. 2003, 2, 196–200. 24. Chang, C. H.; Liao, J. D.; Chen, J. J. J.; Ju, M. S.; Lin, C. C. K. Alkanethiolate Self-Assembled Monolayers as Functional Spacers To Resist Protein Adsorption upon Au-Coated Nerve Microelectrode. Langmuir 2004, 20, 11656–11663. 25. Fan, Y. W.; Cui, F. Z.; Hou, S. P.; Xu, Q. Y.; Chen, L. N.; Lee, I. S. Culture of Neural Cells on Silicon Wafers with Nano-Scale Surface Topograph. J. Neurosci. Methods 2002, 120, 17–23. 26. Khan, S. P.; Auner, G. G.; Newaz, G. M. Influence of Nanoscale Surface Roughness on Neural Cell Attachment on Silicon. Nanomedicine 2005, 1, 125–129. 27. Bray, D. Mechanical Tension Produced by Nerve Cells in Tissue Culture. J. Cell Sci. 1979, 37, 391–410. 28. Schmidt, C. E.; Leach, J. B. Neural Tissue Engineering: Strategies for Repair and Regeneration. Annu. Rev. Biomed. Eng. 2003, 5, 293–347. 29. Stieglitz, T.; Beutel, H.; Schuettler, M.; Meyer, J.-U. Micromachined, Polyimide-Based Devices for Flexible Neural Interfaces. Biomed. Microdevices 2000, 2, 283–294. 30. Lago, N.; Ceballos, D.; Rodriguez, F. J.; Stieglitz, T.; Navarro, X. Long Term Assessment of Axonal Regeneration through Polyimide Regenerative Electrodes To Interface the Peripheral Nerve. Biomaterials 2005, 26, 2021–2031. 31. Soen, Y.; Mori, A.; Palmer, T. D.; Brown, P. O. Exploring the Regulation of Human Neural Precursor Cell Differentiation Using Arrays of Signaling Microenvironments. Mol. Syst. Biol. 2006, 2, 37. 32. Donato, R.; Miljan, E. A.; Hines, S. J.; Aouabdi, S.; Pollock, K.; Patel, S.; Edwards, F. A.; Sinden, J. D. Differential Development of Neuronal Physiological Responsiveness in Two Human Neural Stem Cell Lines. BMC Neurosci. 2007, 8, 36. VOL. 5 ’ NO. 6 ’ 4704–4711 ’ ARTICLE REFERENCES AND NOTES 4711 2011 www.acsnano.org BRIEF ARTICLE pubs.acs.org/molecularpharmaceutics Synergistic Induction of Apoptosis in Brain Cancer Cells by Targeted Codelivery of siRNA and Anticancer Drugs Cheoljin Kim,† Birju P. Shah,† Prasad Subramaniam, and Ki-Bum Lee* Department of Chemistry and Chemical Biology, Rutgers, The State University of New Jersey, Piscataway, New Jersey 08854, United States bS Supporting Information ABSTRACT: Multiple dysregulated pathways in tumors necessitate targeting multiple oncogenic elements by combining orthogonal therapeutic moieties like short-interfering RNAs (siRNA) and drug molecules in order to achieve a synergistic therapeutic effect. In this manuscript, we describe the synthesis of cyclodextrin-modified dendritic polyamines (DexAMs) and their application as a multicomponent delivery vehicle for translocating siRNA and anticancer drugs. The presence of βcyclodextrins in our DexAMs facilitated complexation and intracellular uptake of hydrophobic anticancer drugs, suberoylanilide hydroxamic acid (SAHA) and erlotinib, whereas the cationic polyamine backbone allowed for electrostatic interaction with the negatively charged siRNA. The DexAM complexes were found to have minimal cytotoxicity over a wide range of concentrations and were found to efficiently deliver siRNA, thereby silencing the expression of targeted genes. As a proof of concept, we demonstrated that upon appropriate modification with targeting ligands, we were able to simultaneously deliver multiple payloads —siRNA against oncogenic receptor, EGFRvIII and anticancer drugs (SAHA or erlotinib)—efficiently and selectively to glioblastoma cells. Codelivery of siRNA-EGFRvIII and SAHA/erlotinib in glioblastoma cells was found to significantly inhibit cell proliferation and induce apoptosis, as compared to the individual treatments. KEYWORDS: RNA interference, codelivery, cyclodextrins, SAHA, brain tumor cells, targeted delivery ’ INTRODUCTION Advances in the field of chemical genetics and molecular cell biology have triggered a surge in development of genetic manipulation based therapies for cancer.1,2 Such genetic manipulation methods typically rely on either the traditional smallmolecule/protein modalities3 or the newly discovered RNA interference (RNAi) based modalities,4 each having their own advantages and disadvantages. For example, RNAi therapeutics can provide attractive solutions to the major shortcomings of the conventional therapeutics, including difficulty in lead identification and complex synthesis of small organic molecules and proteins, and potentially can be applicable to all molecular targets for cancer therapy.5 However, RNAi-based therapeutics, such as small interfering RNA (siRNA) and micro RNA (miRNA), are inherently antagonistic and their downstream effects (i.e., genesilencing) are delayed, compared to those of conventional smallmolecule/protein-based therapeutics.6 Additionally, owing to their short serum half-life and poor cellular uptake, successful clinical application of siRNA requires appropriate chemical modifications and better delivery vehicles to overcome the numerous cellular barriers.4 On the other hand, small organic molecules can act as both antagonists and agonists for molecular targets and their drug effects can be much faster than siRNA with minimal problems during their intracellular uptake.5 Hence, from a biological perspective, it would be beneficial to combine the r 2011 American Chemical Society advantages of these therapeutic modalities to potentially enhance their individual efficacy. For example, it was recently demonstrated that simultaneous delivery of siRNA against multidrug resistance genes in cancer cells led to the enhanced efficacy of the codelivered anticancer drugs.7,8 These studies show that it would be desirable to target multiple oncogenic signaling elements using different therapeutic modalities for cooperative effect, especially considering the molecular heterogeneity of tumors. However, to achieve this goal, the primary requirement is to develop noncytotoxic codelivery platforms capable of efficient translocation of siRNA and small molecules with specificity as well as identify the right combination of siRNA and small molecules for a cooperative therapeutic effect. To address the aforementioned need for cooperative chemotherapeutics, herein we describe the synthesis of a multifunctional delivery platform consisting of a dendritic polyamine backbone conjugated with β-cyclodextrin (β-CD) moieties [henceforth referred to as DexAMs] and its application for target-specific codelivery of two orthogonal chemotherapeutic molecules (siRNA and anticancer drug). We hypothesize that Received: December 27, 2010 Accepted: July 27, 2011 Revised: July 11, 2011 Published: July 27, 2011 1955 dx.doi.org/10.1021/mp100460h | Mol. Pharmaceutics 2011, 8, 1955–1961 Molecular Pharmaceutics BRIEF ARTICLE Scheme 1. (A) General Scheme Showing Codelivery of Small Molecules Like Anticancer Drugs and siRNA to Cancer Cells Using Cyclodextrin Modified Polyamines (DeXAMs). (B) Chemical Structure of the Delivery Vehiclea a See Supporting Information for other DexAM generations. codelivery of siRNA and anticancer drugs will have a cooperative therapeutic effect against the target oncogenic signaling pathway (EGFRvIII-PI3K/AKT), resulting in the selective induction of apoptosis in brain tumor cells (Scheme 1). Additionally, conjugation of targeting ligands against receptors overexpressed in brain cancer cells (EGFR) would allow for selective uptake of our complexes into glioblastoma cells, thereby minimizing toxic side effects on normal cells. Additionally, our delivery platform and synthetic methods have several advantages, as compared to conventional carrier molecules (e.g., polyethyleneimine (PEI) and polyamidoamine (PAMAM)). These include (i) minimal cytotoxicity and high transfection efficiency of siRNA/drugDexAM constructs, (ii) significantly higher yields and purity of DexAMs and increased aqueous solubility of DexAM constructs, and (iii) capability of simultaneously delivering nucleic acids, small organic molecules and proteins, thereby achieving cooperative therapeutic effects. ’ MATERIALS AND METHODS Starting materials, reagents, and solvents were purchased from commercial suppliers (Sigma-Aldrich, Acros, and Fisher) and used as received unless otherwise noted. All reactions were conducted in flame-dried glassware with magnetic stirring under an atmosphere of dry nitrogen. Reaction progress was monitored by analytical thin layer chromatography (TLC) using 250 μm silica gel plates (Dynamic Absorbents F-254). Visualization was accomplished with UV light and potassium permanganate stain, followed by heating. Proton nuclear magnetic resonance (1H NMR) spectra were recorded on either a Varian-300 instrument (300 MHz), a Varian-400 instrument (400 MHz) or a Varian-500 instrument (500 MHz). Chemical shifts of the compounds are reported in ppm relative to tetramethylsilane (TMS) as the internal standard. Data are reported as follows: chemical shift, integration, multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, br = broad, m = multiplet), and coupling constants (Hz). Quantification of siRNA Loading Efficiency. The complexes were prepared at various charge ratios by mixing equal volumes of DexAM with siRNA in PBS. Charge ratios (N/P) were calculated as a ratio of the number of primary amines in the polymer, determined from 1H NMR spectra, to the number of anionic phosphate groups in the siRNA. The samples were then incubated at room temperature for 30 min to ensure complex formation. The complexes were prepared at a final siRNA concentration of 0.2 μg of siRNA/100 μL of solution. 100 μL of each complex was transferred to a 96-well (black-walled, clear-bottom, nonadsorbing) plate (Corning, NY, USA). A total of 100 μL of diluted PicoGreen dye (1:200 dilution in Tris- EDTA (TE) buffer) was added to each sample. Fluorescence measurements were made after 10 min of incubation at room temperature using a M200 Pro Multimode Detector (Tecan USA Inc., Durham, NC, USA), at excitation and emission wavelengths of 485 and 535 nm, respectively. All measurements were corrected for background fluorescence from a solution containing only buffer and PicoGreen dye. Particle Size and Zeta Potential Analysis. Dynamic light scattering (DLS) and zeta potential analyses were performed using a Malvern Instruments Zetasizer Nano ZS-90 instrument (Southboro, MA) with reproducibility being verified by collection 1956 dx.doi.org/10.1021/mp100460h |Mol. Pharmaceutics 2011, 8, 1955–1961 Molecular Pharmaceutics and comparison of sequential measurements. Polymer/siRNA complexes (siRNA concentration = 330 nM) at different polymer concentrations were prepared using purified water (resistivity = 18.5 MΩ-cm). DLS measurements were performed at a 90° scattering angle at 25 °C. Z-average sizes of three sequential measurements were collected and analyzed. Zeta potential measurements were collected at 25 °C, and the Z-average potentials following three sequential measurements were collected and analyzed. Cell Culture. Cells were cultured in the following growth media: DMEM (Dulbecco’s modified Eagle’s medium) with high glucose (Invitrogen), 10% fetal bovine serum (FBS), 1% streptomycinpenicillin, 1% glutamax (Invitrogen), and selection markers, G418 (100 μg/mL) and hygromycin B (30 μg/mL) for U87-EGFP and U87-EGFRvIII respectively. PC-12 cells were cultured in DMEM with 10% horse serum, 5% FBS and 1% streptomycinpenicillin. For the knockdown experiment and targeted delivery, passaged cells were prepared to 4060% confluency in 24-well plates. For the knockdown experiment, targeted delivery and cell viability assay, medium was exchanged with serum-free basal medium (500 μL) and siRNADexAM solution (50 μL) was added after 2030 min. After incubation for 12 h, medium was exchanged with normal medium. Fluorescence measurement and cellular assays were performed after 4896 h from the starting point. Cytotoxicity Assays. The percentage of viable cells was determined by MTS assay following standard protocols described by the manufacturer. All experiments were conducted in triplicate and averaged. The quantification of polymer-mediated toxicity was done using MTS assay after incubating the glioblastoma cells in the presence of varying concentrations of only polymer vehicle for 4896 h. The data is represented as formazan absorbance at 490 nm, considering the control (untreated) cells as 100% viable. Quantification of Knockdown of EGFP Expression (ImageJ). Following siRNA treatment, cells were washed with DPBS and fixed with 24% paraformaldehyde solution prior to imaging. For the fluorescence, DIC and phase contrast images were obtained using the Zeiss Axio observer inverted epifluorescence microscope. Each image was captured with different channels and focus. Images were processed and overlapped using Image-Pro (Media Cybernetics) and ImageJ (NIH). Targeted Delivery. Highly tumorigenic U87-EGFP cells and low-tumorigenic PC-12 cells were cultured in 24-well plates, at a density of 5 104 cells per well. For PC-12 cells, the normal growth medium was DMEM (with high glucose, Invitrogen), 5% horse serum, 10% FBS, 1% Glutamax, and 1% penicillinstreptomycin. For the delivery of EGFR-Ab conjugated DexAM polyplexes, medium was exchanged with serum free DMEM medium. The cells were incubated in the Ab-conjugated polyplex medium for 68 h. Fluorescence images were taken after replacing the serum-free medium with regular medium. Apoptosis Assay. Cells were harvested by trypsinization and stained using an Annexin V FITC Apoptosis Detection kit (Roche, Cambridge, MA) according to the manufacturer’s protocol. The stained cells were immediately analyzed by flow cytometry (FACScan; Becton Dickinson, Franklin Lake, NJ). Early apoptotic cells with exposed phosphatidylserine but intact cell membranes bound to Annexin VFITC but excluded propidium iodide. Cells in necrotic or late apoptotic stages were labeled with both Annexin VFITC and propidium iodide. BRIEF ARTICLE ’ RESULTS AND DISCUSSION Using multistep solution-phase and solid-phase synthesis, we generated a series of highly water soluble dendritic polyamine compounds conjugated to one or more β-cyclodextrin (β-CD) molecules, referred to as DexAMs, with higher yield and purity as compared to reported syntheses (Scheme 2). The first step for synthesizing DexAM involved generating a dendritic polyamine backbone by Michael addition of tris(2-aminoethyl)amine and methyl acrylate, followed by amidation of the amino esters generated after Michael addition. The use of tris(2-aminoethyl)amine as the core initiator yielded higher surface amine groups and hence more compact dendrimers as compared to the reported synthetic methods (for, e.g., ethylenediamine, ammonia) for PAMAM dendrimers.9 The conjugation of β-cyclodextrin to the polyamine backbone involved tosylation of β-cyclodextrin, followed by nucleophilic addition with amine group. Compared to the previously reported protocol,10 where tosyl chloride was used for regioselective tosylation of β-cyclodextrin resulting in very low yields, we improved the synthetic yield (∼50%) and purity by using tosylimidazole, instead of tosyl chloride, under reflux conditions to generate 6-monotosylated β-cyclodextrin (see Supporting Information). In the final step, polyamine backbone was conjugated to tosylated CD via nucleophilic addition to generate cyclodextrin conjugated polyamines, resulting in a 25fold increase in the aqueous solubility of CD (>50 g/100 mL) as compared to that of CD alone (<1.8 g/100 mL), owing to generation of an aminium salt (see Supporting Information for the detailed synthesis). The first component of our delivery vehicle—β-CD—has been extensively used in pharmaceutical applications to improve solubility of hydrophobic moieties, such as anticancer drugs.11 Many anticancer drugs are known to have poor aqueous solubility, thereby necessitating the use of toxic organic solvents like dimethyl sulfoxide (DMSO), which can be detrimental in biological applications.12 The presence of β-CD in our DexAM moiety and the optimized drug loading would not only prevent the use of such toxic solvents but also improve the water solubility of CDdrug complex for the optimal cellular uptake and drug efficacy. In our study, two hydrophobic anticancer drugs [erlotinib and suberoylanilide hydroxamic acid (SAHA)] were synthesized and loaded into the β-CD cavity by using our optimized protocols (Figure 2b).13,14 For instance, by utilizing the pH-dependent solubility of erlotinib, we could load drug up to almost 50% of the molar ratio of β-CD, resulting in a significant increase in its aqueous solubility (178 mg/100 mL).15,16 Similarly, we complexed SAHA with β-CD under reflux conditions to obtain highly water soluble SAHACD complexes (solubility: 175 mg/100 mL) (see Supporting Information for the more detailed synthesis and experimental protocols).17 The second component of our DexAMs—dendritic polyamine backbone— provides a positive surface charge which can interact electrostatically with the negatively charged nucleic acids, condensing them into cationic complexes (known as polyplexes), thus facilitating their intracellular uptake and endosomal escape.1820 However, these primary/tertiary amines are also responsible for cytotoxicity by interacting with the cellular components and interfering in the cellular processes.21 Hence, there is a clear need to develop synthetic chemistry to control the ratio of electrostatic properties and the size of polymer structures. Our synthetic methods enabled us to precisely control the number of primary amine head groups from 4 to 48 leading to four different generations 1957 dx.doi.org/10.1021/mp100460h |Mol. Pharmaceutics 2011, 8, 1955–1961 Molecular Pharmaceutics BRIEF ARTICLE Scheme 2. (A) Schematic Representation of Synthesis of DexAMs. (B) Conjugation of Drugs and Antibodies to DexAMs of DexAMs molecules (D1D4), thereby allowing us to achieve an optimal balance between cytotoxicity and complexation ability. We assessed the capability of our four different generations of DexAMs (D1D4) to spontaneously form complexes with the negatively charged siRNA using a well-established dye exclusion assay. As the number of amine groups increased from DexAM-1 (D1, 4 primary amines) to DexAM-4 (D4, 48 primary amines), the amount of free/unbound siRNA decreased correspondingly at a given DexAM concentration (see Figure S1 in the Supporting Information). Since we found that the complexation ability of DexAM-4 is higher than that of the other generations with minimal cytotoxicity, we proceeded with using DexAM-4 for the subsequent experiments. Additionally, the hydrodynamic diameters of the resultant polyplexes could be controlled from 250 to 400 nm with polydispersity index of 0.81.0 by increasing the polymer concentration (see Figure S2a in the Supporting Information). The zeta potentials of the resulting polyplexes were in the range of 810 mV at pH 7.4 (see Figure S2b in the Supporting Information), demonstrating the cationic nature of the polyplexes. Cytotoxicity of the DexAM molecules was assessed using MTS assay. First we confirmed the effect of the β-CD moiety on the cytotoxicity of DexAMs by comparing the cytotoxicity of the DexAM (containing CD) to that of the DexAM without CD. Our cytotoxicity assay data clearly shows that the DexAM constructs with CD show significantly less cytotoxicity as compared to those without CD (Figure 1a). We believe this is due to the presence of CDs on a polycationic backbone in DexAM, which can potentially reduce nonspecific binding of the DexAM constructs with proteins or cellular structures.2023 We also compared the cytotoxicity of our DexAMs with the commercially available transfection agents, polyethyleneimine (PEI), Lipofectamine 2000 (LF) and X-tremeGENE (Xgene) at the recommended concentrations for transfection, and found that those agents were significantly more cytotoxic at those concentrations as compared to DexAMs (Figure 1b). The optimization of gene silencing with our siRNADexAM constructs and assessment of knockdown efficiency were performed by measuring the suppression of enhanced green fluorescent protein (EGFP) in glioblastoma cell lines (U87-EGFP), which were genetically modified to constitutively express EGFP. The decrease of green fluorescence intensity due to siRNAmediated EGFP silencing was monitored over a time period of 4896 h to quantify the knockdown efficiency of our DexAM/ siRNA constructs (see Figure S4 in the Supporting Information). Approximately 70% of the U87-EGFP cells showed no EGFP signal after 96 h of siRNA treatment as compared to the control cells at a polymer concentration of 100 μM (Figure 1c) with negligible cytotoxicity (∼95% cell viability). In parallel, we compared the transfection efficiency and the corresponding cytotoxicity of our delivery platform with that of the commercially available transfection agent (X-tremeGENE) under the same condition, in which X-tremeGENE-based transfection demonstrated similar levels of EGFP knockdown (∼70% knockdown efficiency), albeit with significant toxicity (∼30% cell viability) (see Figure S5 in the Supporting Information). In addition to efficient translocation of siRNA across the cell membrane with minimal cytotoxicity, successful therapeutic application of siRNA also requires the siRNA to interact with the RNAi machinery within a target cell, thereby minimizing off-target effects.24 Brain tumor cells, particularly glioblastoma cells, are known to present high levels of epidermal growth factor receptors (EGFRs) on their cell surface, thus making it a specific biomarker for cellspecific delivery toward brain tumor cells.25 For targeted delivery to glioblastoma cells, we modified our DexAM-4 with appropriate ratios of EGFR antibodies (DexAM-4:EGFR-Ab = 1:5) 1958 dx.doi.org/10.1021/mp100460h |Mol. Pharmaceutics 2011, 8, 1955–1961 Molecular Pharmaceutics BRIEF ARTICLE Figure 1. Cytotoxicity and transfection efficiency of DexAM-4 (D4). (A) Effect of cyclodextrin grafting on polymer-mediated toxicity. (B) Comparison of toxicities of DexAM-4 with commercially used transfection agents at optimized concentrations of delivery agent and siRNA. (LF 2000, Lipofectamine 2000; and Xgene, X-tremeGENE). (C) Phase contrast (C1, C2) and fluorescent (C3, C4) images showing siRNA-mediated decrease in green fluorescence in treated and control (untreated) U87-EGFP cells. and incubated them in U87 (glioblastoma cell line, target cells) and other less-tumorigenic PC-12 cells (control cells) which tend to have low levels of expression of EGFRs. The DexAM-4 constructs were also labeled with a fluorescent dye (Alexa Fluor 594) to monitor their intracellular uptake using fluorescence microscopy. From our data we could see that EGFR-antibody modified DexAM-4 were selectively translocated into U87 (target glioblastoma cells) with high efficiency as compared to the PC-12 (control cells) (Figure 2a). Having demonstrated the target-specific delivery and efficient gene silencing capability of the siRNADexAM constructs, we then focused on our main goal of codelivering siRNA and anticancer drugs for targeting key oncogenic signaling pathways (e.g., EGFRvIII-(phoshphatidylinositol-3-kinase)PI3K/AKT) to achieve a cooperative chemotherapeutic effect. Tumors harbor multiple dysregulated signaling pathways, thus limiting the clinical utility of single target agents.26 Hence, combining approaches targeting multiple oncogenic elements, using a single delivery platform, can not only increase the likelihood of blocking tumor survival and metastasis, as compared to individual treatments, but also simplify clinical applications. For this purpose, we focused on developing a combinatory therapeutic approach based on siRNA and anticancer cancer drugs targeting oncogenic pathways in glioblastoma multiforme (GBM), an extremely aggressive and difficult-to-treat form of primary brain tumor. We aimed at downregulating the EGFRvIII-PI3K/AKT pathway, implicated in the proliferation and apoptosis of brain tumor cells, by delivering siRNA against epidermal growth factor receptor variant III (EGFRvIII), which is known to enhance the tumorigenicity of GBM.2729 However, due to tumor molecular heterogeneity, only siRNA-based downregulation of a single oncogenic target (EGFRvIII) may not be efficacious. Histone deacetylase (HDAC) inhibitors like suberoylanilide hydroxamic acid (SAHA) and EGFR tyrosine kinase inhibitors like erlotinib have been reported to enhance the efficacy of other EGFR antagonists.26,30 To this end, we used either SAHA or erlotinib for codelivery with siRNA against EGFRvIII oncogene to deactivate the target signaling pathway in a selective and efficient manner. These drugs have already shown some promising results for GBM therapy, but have met with limited success since they require higher doses and longer exposures, which may lead to increased toxic side effects.31 Our hypothesis is that combination of anticancer drugs against complementary therapeutic targets with siRNA therapeutics against EGFRvIII would have a cooperative effect on induction of apoptosis in brain tumor cells. To test this hypothesis, we initially compared the antiproliferative capability of anticancer drugs (SAHA and erlotinib) and siRNA against EGFRvIII in glioblastoma cells, either individually or in combination by using cell viability assay (Figure 2b). From the data, we could clearly observe a cooperative inhibition of glioblastoma cell proliferation when SAHA (5 μM) was codelivered with the siRNA (200 nM; polymer concentration 100 μM), as compared to treating the cells with only SAHA at the same concentration (5 μM). This can be attributed to the fact that SAHA is known to significantly enhance the efficacy of agents targeting EGFR signaling pathway by modulating several indirect downstream targets, which in turn are key regulators of EGFR pathways. Similarly, codelivery of erlotinib (30 μM) and siRNA (200 nM) also inhibited tumor cell proliferation to a higher extent (Figure 2b). Additionally, we also monitored the effect of codelivery of both siRNA and anticancer drugs on inducing cell death in glioblastoma cells using the apoptosis assay (Annexin-V/propidium iodide assay). A significantly higher proportion of cell population treated with both siRNA and SAHA was Annexin-VFITC-positive as compared to the individual treatments as well as untreated cells. These results indicate greater induction of apoptosis in cells treated with both siRNA and SAHA, as compared to those with only SAHA and only siRNA treatment (Figure 2c). A similar trend in the cooperative induction of apoptosis was seen in the case of 1959 dx.doi.org/10.1021/mp100460h |Mol. Pharmaceutics 2011, 8, 1955–1961 Molecular Pharmaceutics BRIEF ARTICLE Figure 2. Targeted delivery of DexAMs and cooperative effect of anticancer drugs and siRNA on glioblastoma cells. (A) Targeted delivery of DexAMs modified with EGFR antibodies in highly tumorigenic U87-EGFP cells and less-tumorigenic PC-12 cells. (B) Viability of glioblastoma cells following individual treatments and codelivery of drugs and siRNA, based upon MTS assay. (C) Flow cytometry based Annexin-V/PI assay demonstrating the apoptotic effect of combined and individual siRNA and drug treatments. Percentages represent Annexin-V-positive (apoptotic cells). For all experiments, the polymer concentration was kept constant (100 μM), whereas the concentrations of SAHA, erlotinib and siRNA were 5 μM, 30 μM and 200 nM respectively. combined erlotinib/siRNA treatment (Figure 2c). We also found that complexation of SAHA and erlotinib within the CD cavity improved their aqueous solubility and hence increased their potency, measured as IC 50 values, by approximately 2-fold as compared to its DMSO solution (see Figure S6 in the Supporting Information). Thus, these results show the cooperative effect on selectively inducing the apoptosis of brain tumor cells by the right combination of siRNA and anticancer drugs and the capability of our delivery molecules (DexAMs) for target-specific delivery and improved chemotherapeutic efficacy. In conclusion, we synthesized a multimodal delivery platform to simultaneously deliver two orthogonal therapeutic modalities having cooperative therapeutic efficacy in an efficient and selective manner to the target brain tumor cells. As a proof-ofconcept experiment, we demonstrated that target-specific codelivery of siRNA and anticancer drugs having complementary therapeutic results would be a novel method to enhance the apoptotic signaling pathways and inhibit the proliferation signaling pathways in brain tumor cells. Potentially, our approach and methodology can be beneficial for introducing exogenous siRNA combined with small molecules into other mammalian cells, which can represent a powerful approach for the optimal manipulating signal transduction. Our synthetic techniques afforded facile manipulation of the polymer structure to achieve efficient transfection with minimal polymer-mediated cytotoxicity. The strategy of codelivering anticancer drug with therapeutic siRNA is particularly advantageous for in vivo applications, so that both the moieties are delivered to the target cells using a single delivery platform. Our versatile delivery platform can also be used to codeliver different kinds of small molecules and nucleic acids to regulate cancer cell fate such as proliferation, migration and apoptosis by targeting multiple signaling pathways. Collectively, our DexAM-based codelivery strategy has significant potential for cancer therapy as well as regulating cell fate by modulating key signaling cascades. ’ ASSOCIATED CONTENT bS Supporting Information. Detailed synthesis of DexAMs; NMR characterization of synthesized compounds; conjugation of targeted moieties to DexAMs, complexation of drugs, particle size and zeta potentials of DexAM polyplexes, siRNA loading efficiency. This material is available free of charge via the Internet at http://pubs.acs.org. ’ AUTHOR INFORMATION Corresponding Author *Department of Chemistry and Chemical Biology, The State University of New Jersey, Piscataway, NJ 08854, United States. Tel: (+1) 732-445-2081. Fax: (+1) 732-445-5312. E-mail: kblee@ rutgers.edu. Homepage: http://rutchem.rutgers.edu/∼kbleeweb/. Author Contributions † These authors contributed equally to this work. ’ ACKNOWLEDGMENT The authors thank Joan Dubois for assisting with the flow cytometry measurements and Kevin Memoli for providing us with Erlotinib and SAHA. We are also grateful to KBLEE group members for their valuable suggestions for the manuscript. This work was supported by NIH Director’s Innovator Award (1DP20D006462-01) and N.J. Commission on Spinal Cord Research grant (09-3085-SCR-E-0). ’ REFERENCES (1) Mac Gabhann, F.; Annex, B. H.; Popel, A. S. Gene therapy from the perspective of systems biology. Curr. Opin. Mol. Ther. 2010, 12 (5), 570–577. (2) Caldwell, J. S., Cancer cell-based genomic and small molecule screens. In Genomics in Cancer Drug Discovery and Development; 1960 dx.doi.org/10.1021/mp100460h |Mol. Pharmaceutics 2011, 8, 1955–1961 Molecular Pharmaceutics Hampton, G., Sikora, K., Eds.; Academic Press: New York, 2007; Vol. 96, pp 145173. (3) Collins, I.; Workman, P. New approaches to molecular cancer therapeutics. Nat. Chem. Biol. 2006, 2 (12), 689–700. (4) Kim, D. H.; Rossi, J. J. Strategies for silencing human disease using RNA interference. Nat. Rev. Genet. 2007, 8 (3), 173–184. (5) Bumcrot, D.; Manoharan, M.; Koteliansky, V.; Sah, D. W. Y. RNAi therapeutics: a potential new class of pharmaceutical drugs. Nat. Chem. Biol. 2006, 2 (12), 711–719. (6) Castanotto, D.; Rossi, J. J. The promises and pitfalls of RNAinterference-based therapeutics. Nature 2009, 457 (7228), 426–433. (7) Chen, A. M.; Zhang, M.; Wei, D. G.; Stueber, D.; Taratula, O.; Minko, T.; He, H. X. Co-delivery of Doxorubicin and Bcl-2 siRNA by Mesoporous Silica Nanoparticles Enhances the Efficacy of Chemotherapy in Multidrug-Resistant Cancer Cells. Small 2009, 5 (23), 2673–2677. (8) Zhu, C. H.; Jung, S.; Luo, S. B.; Meng, F. H.; Zhu, X. L.; Park, T. G.; Zhong, Z. Y. Co-delivery of siRNA and paclitaxel into cancer cells by biodegradable cationic micelles based on PDMAEMA-PCL-PDMAEMA triblock copolymers. Biomaterials 2010, 31 (8), 2408–2416. (9) Tomalia, D. A.; Baker, H.; Dewald, J.; Hall, M.; Kallos, G.; Martin, S.; Roeck, J.; Ryder, J.; Smith, P. A new class of polymer: starburst-dendritic macromolecules. Polym. J. 1985, 17, 117–132. (10) Tang, W.; Ng, S. C. Facile synthesis of mono-6-amino-6-deoxyalpha-, beta-, gamma-cyclodextrin hydrochlorides for molecular recognition, chiral separation and drug delivery. Nat. Protoc. 2008, 3 (4), 691–697. (11) Uekama, K.; Hirayama, F.; Irie, T. Cyclodextrin drug carrier systems. Chem. Rev. 1998, 98 (5), 2045–2076. (12) Brewster, M. E.; Loftsson, T. Cyclodextrins as pharmaceutical solubilizers. Adv. Drug Delivery Rev. 2007, 59 (7), 645–666. (13) Knesl, P.; Roseling, D.; Jordis, U. Improved synthesis of substituted 6,7-dihydroxy-4-quinazolineamines: Tandutinib, erlotinib and gefitinib. Molecules 2006, 11 (4), 286–297. (14) Stowell, J. C.; Huot, R. I.; Vanvoast, L. The Synthesis of N-Hydroxy-N0 -Phenyoctanediamide and its Inhibitory Effect on Proliferation of Axc Rat Prostate Cancer Cells. J. Med. Chem. 1995, 38 (8), 1411–1413. (15) Lee, Y. H. P.; Sathigari, S.; Lin, Y. J. J.; Ravis, W. R.; Chadha, G.; Parsons, D. L.; Rangari, V. K.; Wright, N.; Babu, R. J. Gefitinibcyclodextrin inclusion complexes: physico-chemical characterization and dissolution studies. Drug Dev. Ind. Pharm. 2009, 35 (9), 1113–1120. (16) Khramov, A. N.; Stenken, J. A. Enhanced microdialysis recovery of some tricyclic antidepressants and structurally related drugs by cyclodextrin-mediated transport. Analyst 1999, 124 (7), 1027–1033. (17) Hockly, E.; Richon, V. M.; Woodman, B.; Smith, D. L.; Zhou, X. B.; Rosa, E.; Sathasivam, K.; Ghazi-Noori, S.; Mahal, A.; Lowden, P. A. S.; Steffan, J. S.; Marsh, J. L.; Thompson, L. M.; Lewis, C. M.; Marks, P. A.; Bates, G. P. Suberoylanilide hydroxamic acid, a histone deacetylase inhibitor, ameliorates motor deficits in a mouse model of Huntington’s disease. Proc. Natl. Acad. Sci. U.S.A. 2003, 100 (4), 2041–2046. (18) Pack, D. W.; Hoffman, A. S.; Pun, S.; Stayton, P. S. Design and development of polymers for gene delivery. Nat. Rev. Drug Discovery 2005, 4 (7), 581–593. (19) Wang, H.; Chen, K. J.; Wang, S. T.; Ohashi, M.; Kamei, K. I.; Sun, J.; Ha, J. H.; Liu, K.; Tseng, H. R. A small library of DNAencapsulated supramolecular nanoparticles for targeted gene delivery. Chem. Commun. 2010, 46 (11), 1851–1853. (20) Wang, H.; Wang, S. T.; Su, H.; Chen, K. J.; Armijo, A. L.; Lin, W. Y.; Wang, Y. J.; Sun, J.; Kamei, K.; Czernin, J.; Radu, C. G.; Tseng, H. R. A Supramolecular Approach for Preparation of Size-Controlled Nanoparticles. Angew. Chem., Int. Ed. 2009, 48 (24), 4344–4348. (21) Pun, S. H.; Bellocq, N. C.; Liu, A. J.; Jensen, G.; Machemer, T.; Quijano, E.; Schluep, T.; Wen, S. F.; Engler, H.; Heidel, J.; Davis, M. E. Cyclodextrin-modified polyethylenimine polymers for gene delivery. Bioconjugate Chem. 2004, 15 (4), 831–840. (22) Heidel, J. D.; Hu, S. W.; Liu, X. F.; Triche, T.; Davis, M. E. Cyclodextrin-containing polycations (CDPs) for in vivo delivery of siRNA. Abstr. Paper. Am. Chem. Soc. 2004, 227, 208-BIOT. BRIEF ARTICLE (23) Davis, M. E.; Bellocq, N. C. Cyclodextrin-containing polymers for gene delivery. J. Inclusion Phenom. Macrocyclic Chem. 2002, 44 (14), 17–22. (24) Dykxhoorn, D. M.; Palliser, D.; Lieberman, J. The silent treatment: siRNAs as small molecule drugs. Gene Ther. 2006, 13 (6), 541–552. (25) Heimberger, A. B.; Suki, D.; Yang, D.; Shi, W. M.; Aldape, K. The natural history of EGFR and EGFRvIII in glioblastoma patients. J. Transl. Med. 2005, 3. (26) Lai, C. J.; Bao, R.; Tao, X.; Wang, J.; Atoyan, R.; Qu, H.; Wang, D. G.; Yin, L.; Samson, M.; Forrester, J.; Zifcak, B.; Xu, G. X.; DellaRocca, S.; Zhai, H. X.; Cai, X.; Munger, W. E.; Keegan, M.; Pepicelli, C. V.; Qian, C. G. CUDC-101, a Multitargeted Inhibitor of Histone Deacetylase, Epidermal Growth Factor Receptor, and Human Epidermal Growth Factor Receptor 2, Exerts Potent Anticancer Activity. Cancer Res. 2010, 70 (9), 3647–3656. (27) Engelman, J. A. Targeting PI3K signalling in cancer: opportunities, challenges and limitations. Nat. Rev. Cancer 2009, 9 (8), 550–562. (28) Liu, P. X.; Cheng, H. L.; Roberts, T. M.; Zhao, J. J. Targeting the phosphoinositide 3-kinase pathway in cancer. Nat. Rev. Drug Discovery 2009, 8 (8), 627–644. (29) Jung, J. J.; Solanki, A.; Memoli, K. A.; Kamei, K.; Kim, H.; Drahl, M. A.; Williams, L. J.; Tseng, H. R.; Lee, K. Selective Inhibition of Human Brain Tumor Cells through Multifunctional Quantum-DotBased siRNA Delivery. Angew. Chem., Int. Ed. 2010, 49 (1), 103–107. (30) Friess, T.; Scheuer, W.; Hasmann, M. Combination treatment with erlotinib and pertuzumab against human tumor xenografts is superior to monotherapy. Clin. Cancer Res. 2005, 11 (14), 5300–5309. (31) Learn, C. A.; Hartzell, T. L.; Wikstrand, C. J.; Archer, G. E.; Rich, J. N.; Friedman, A. H.; Friedman, H. S.; Bigner, D. D.; Sampson, J. H. Resistance to tyrosine kinase inhibition by mutant epidermal growth factor receptor variant III contributes to the neoplastic phenotype of glioblastoma multiforme. Clin. Cancer Res. 2004, 10 (9), 3216–3224. 1961 dx.doi.org/10.1021/mp100460h |Mol. Pharmaceutics 2011, 8, 1955–1961 APPLIED PHYSICS LETTERS 98, 173702 共2011兲 ZnO thin film transistor immunosensor with high sensitivity and selectivity Pavel Ivanoff Reyes,1 Chieh-Jen Ku,1 Ziqing Duan,1 Yicheng Lu,1,a兲 Aniruddh Solanki,2 and Ki-Bum Lee2,a兲 1 Department of Electrical and Computer Engineering, Rutgers University, 94 Brett Road, Piscataway, New Jersey 08854-8058, USA 2 Department of Chemistry and Chemical Biology, Rutgers University, 610 Taylor Road, Piscataway, New Jersey 08854-8058, USA 共Received 11 January 2011; accepted 3 April 2011; published online 29 April 2011兲 A zinc oxide thin film transistor-based immunosensor 共ZnO-bioTFT兲 is presented. The back-gate TFT has an on-off ratio of 108 and a threshold voltage of 4.25 V. The ZnO channel surface is biofunctionalized with primary monoclonal antibodies that selectively bind with epidermal growth factor receptor 共EGFR兲. Detection of the antibody-antigen reaction is achieved through channel carrier modulation via pseudo double-gating field effect caused by the biochemical reaction. The sensitivity of 10 fM detection of pure EGFR proteins is achieved. The ZnO-bioTFT immunosensor also enables selectively detecting 10 fM of EGFR in a 5 mg/ml goat serum solution containing various other proteins. © 2011 American Institute of Physics. 关doi:10.1063/1.3582555兴 The ion-selective field effect transistors 共ISFETs兲 has been used popularly as a sensitive pH sensor and various biochemical sensors.1–3 Recently the ISFET structure has been integrated with poly-Si thin film transistors 共TFTs兲 and GaN/AlGaN high electron mobility transistors for detection of DNA, penicillin, and cellular potentials.4,5 However, the sensing procedure using the ISFET can be invasive as its entire gate serves as the sensing area which contains both the analyte solution and the reference electrode. Another class of FET-type biosensors is based on organic field-effect transistors 共OFETs兲.6–9 The general structure of an OFET consists of a back-gate metal-oxide semiconductor field-effect transistor 共MOSFET兲 with the conducting channel made of organic semiconductors. The OFET has the advantage of being easily controlled through biasing due to the back-gate configuration. However, OFETs require high bias voltages, and suffer from low channel mobility. Currently, nanowire-based FET sensors are demonstrated with high sensitivity reaching the order of fM.10,11 However, these prototypes of sensors generally involve a complex fabrication process as they are constructed individually by manipulating and aligning a single strand of semiconducting nanowire such as TiO2 or Si as the FET channel between the source and drain patterns. It is difficult to achieve repeatability and manufacturability in fabrication and integration of these devices for larger sensor arrays. ZnO is emerging as a wide band gap semiconductor oxide with multifunctional properties that makes it an attractive sensor material. ZnO and its nanostructures are compatible with intracellular material and ZnO–based sensors have been demonstrated for detection of biochemicals such as enzymes, antibodies, DNA immobilization, and hybridization.12–15 In this letter, we report the highly sensitive and selective immunosensing ability of a ZnO based TFT biosensor 共ZnObioTFT兲. The epidermal growth factor receptor 共EGFR兲 is used as the example because the sensing of EGFR-antibodies reacting with EGFR proteins has its implications in cancer related studies and drug screening for cancer, as EGFR is a兲 Authors to whom correspondence should be addressed. Electronic addresses: [email protected] and [email protected]. 0003-6951/2011/98共17兲/173702/3/$30.00 well-known to be overexpressed in solid tumors, especially breast cancers. The ZnO-TFT devices possess excellent and repeatable characteristics. It can be fabricated using conventional microelectronic process and can be integrated into a large-scale at sensor arrays low cost, which will benefit further development of a device platform not only for diagnosing cancers, but also for monitoring a patient’s response to therapy in real-time. The device schematic is shown as the inset of Fig. 1共a兲. It follows a back-gate inverted-staggered configuration. A Si substrate was covered with 1 m layer of SiO2 through wet oxidation followed by e-beam deposition of a layer of Au 共50 nm兲/Cr 共100 nm兲 that serves as the gate electrode. A 70 nm layer of SiO2 serving as the gate oxide was then deposited through plasma enhanced chemical vapor deposition with substrate temperature of 250 ° C and using SiH4 and N2O as the source gases. A 50 nm ZnO thin film was grown using metalorganic chemical vapor deposition on the top of the FIG. 1. 共a兲 Transconductance curve of the ZnO-bioTFT and its vertical structure schematic 共inset兲; 共b兲 transistor characteristic curves for various gate bias, and the top view of the device 共inset兲. 98, 173702-1 © 2011 American Institute of Physics 173702-2 Reyes et al. Appl. Phys. Lett. 98, 173702 共2011兲 FIG. 3. Drain current vs gate bias for various Molar concentrations of pure EGFR proteins detected by the ZnO-bioTFT to demonstrate sensitivity. FIG. 2. 共Color online兲 共a兲 Drain current vs gate bias for fixed drain bias of 10 V. Step 1: bare device, step 2: EGFR-antibody immobilization, and step 3: EGFR protein detection; 关共b兲–共d兲兴 schematic of the carrier modulation mechanism for steps 1 to 3, respectively. SiO2 to serve as the n-type conduction channel, with substrate temperature at 350 ° C and using diethyl zinc 共DEZn兲 as the metal precursor and ultrahigh purity O2 as oxidizer. Au 共50 nm兲/Ti 共100 nm兲 was deposited through e-beam evaporation for the source and drain Ohmic contacts. The exposed ZnO channel acts as the sensing area and has a dimension of 200 m ⫻ 400 m, giving a W/L ratio of 2. Shown in the inset of Fig. 1共b兲 is the top view of the TFT device. The electrical characteristics of the ZnO-bioTFT are shown in Figs. 1共a兲 and 1共b兲. The transconductance curve 关drain current 共ID兲 versus gate voltage 共VGS兲兴 in Fig. 1共a兲 shows that the bioTFT is a normally-OFF enhancement mode transistor with a threshold voltage of 4.25 V and an ON-OFF ratio of ⬃108. The high ON-OFF ratio of the device provides the high sensitivity of the device to the charge modulation within the ZnO channel. Figure 2共b兲 shows the transistor characteristic curves with drain current versus drain voltage for various gate-biasing of the device. To realize the immunosensing ability of the ZnObioTFT, the exposed ZnO channel was functionalized using linkage chemistry, which involves three basic steps. First, the ZnO channel was functionalized with trimethoxysilane aldehyde 共having a reactive aldehyde end group兲 by incubating the device in 1% v/v solution of the silane-aldehyde in 95% ethanol for 30 min. The device was then cured at 120 ° C for 15 min. Second, the aldehyde groups were coupled to the amine groups of the monoclonal EGFR antibodies 共1:50兲 through reductive amination in the presence of 4 mM sodium cyanoborohydride in PBS 共pH 7.4兲 for two hours. Third, unreacted aldehyde groups were blocked using 100 mM ethanolamine in a similar manner to prevent nonspecific in- teractions of proteins. Finally, the device was rinsed in a continuous flow of PBS, pH 7.4 for 10 min. The biofunctionalization enables the exposed ZnO channel direct interaction with the biochemical species being detected. The mechanism of detection of antibody-antigen reaction is illustrated in Figs. 2共a兲–2共d兲. In the first step 关Fig. 2共b兲兴 the unfunctionalized ZnO-bioTFT is positively biased at the drain and gate electrode. The positive voltage at the gate causes the majority carriers of the n-type ZnO channel to accumulate near the base of the ZnO layer to facilitate a conduction path for the current flow from drain to source. The positive voltage at the drain causes some of the carriers to also accumulate near the side of the drain electrode forming a wedge-shaped conduction path. The bias at the drain also acts as the electron pump to drive the current to flow. For the second step 关Fig. 2共c兲兴, the exposed ZnO channel is functionalized with EGFR monoclonal antibodies 共mAbs兲 having free lysine groups. The immobilized antibody molecules caused significant decrease in conductivity of the ZnO surface layer, thus, reducing the drain current. In the third step 关Fig. 2共d兲兴, the EGFR protein captured by the EGFR mAbs forms a polarized molecule with a dominant partiallypositive charged tip16 which led to the accumulation of negative carriers within the ZnO channel to accumulate near the exposed surface where the antibody-protein pairs were present. This carrier accumulation was in addition to the conduction path created near the gate. The combined amount of accumulation layer caused an increase in the current flow. The top molecule layer 共reacted protein兲 acted as a virtual top gate and the antibody layer acted as a virtual insulator layer, thus forming a pseudodouble gated field-effect conduction scheme for the ZnO-bioTFT. The actual measured drain currents that confirmed each step of the detection process are shown in Fig. 2共a兲. The drain voltage is fixed to 10 V and the gate voltage is varied from ⫺5 to +15 V, and the drain current is measured using an HP4156C semiconductor parameter analyzer and Cascade Microtech probe station. To demonstrate the high sensitivity of the ZnO-bioTFT, solutions of pure EGFR 共in PBS兲 were prepared with four different Molar concentrations using serial dilutions, namely, 10 nM, 100 pM, 1 pM, and finally 10 fM. Each EGFR solution 共2 l兲 was introduced to a separate but similar ZnObioTFT fabricated on a single chip that were simultaneously functionalized with EGFR mAbs. The drain current was monitored as a function of gate voltage with a fixed drain voltage of 10 V, for each concentration. Figure 3 shows the measured drain current versus gate voltage of the bioTFT. An increase in drain current was measured as the EGFR concen- 173702-3 Appl. Phys. Lett. 98, 173702 共2011兲 Reyes et al. device. Moreover, the device was able to discern as low as 10 fM of EGFR protein concentration in the serum solution. The sensitivity plot of the device for both pure and in-serum detection is shown in Fig. 4共b兲 which exhibits linearity in the x-y logarithmic scale. In summary, we have demonstrated a ZnO bioTFT that has the ability to perform immunosensing with high sensitivity and selectivity. The channel of the bioTFT is functionalized with amine-terminated EGFR mAbs. EGFR proteins with the lowest concentration of 10 fM were detected by the device in both pure state and selectively in a concentration serum solution containing various other protein species. The ZnO-bioTFT enables bias-controlled operation though its bottom gate configuration. The high sensitivity of the device is attributed to its high on-off ratio, and the output current trend is explained by the pseudo-double gating electric field effect. The realization of the ZnO-bioTFT functionalized with EGFR mAbs reacting with EGFR proteins has potential applications in cancer diagnosis and treatment. FIG. 4. 共a兲 Drain current vs gate bias for various molar concentrations of EGFR-proteins in a serum solution containing many different proteins. 共b兲 Sensitivity plot of the device for pure protein and protein in serum detection. tration was increased and the graph also shows that the device was able to detect as low as 10 fM of EGFR concentration. The trend in the current readings agrees with the hypothesis provided by the pseudodouble gating effect discussed above. The highly selective sensing of EGFR using the ZnObioTFT was also demonstrated. In this experiment, a 5 mg/ml 共in PBS, pH 7.4兲 goat serum solution was prepared, which contains many different species of proteins. As mentioned above, different EGFR solutions were prepared, namely, 100, 1, and 10 fM, using this serum solution as the solvent and not pure PBS. For all the concentrations, the total amount of serum present remained approximately the same. Each of the different solutions 共2 l兲 was introduced onto a chip containing multiple similar bioTFT devices that were biofunctionalized with EGFR mAbs. The drain current of each device was measured as a function of gate voltage, with a fixed drain voltage of 10 V. As a control, we first introduced serum solution without the EGFR proteins to the ZnO-bioTFT. Figure 4共a兲 shows no change in the drain current for the pure serum confirming that there were no EGFR molecules in the solution. The drain current increased as a function of EGFR concentration. The bio-TFT detected only the EGFR proteins out of the many different proteins present in the serum solution introduced onto the sensing area of the This work has been supported in part by the AFOSR under Grant No. FA9550-08-01-0452, and by the NSF under Grant No. ECCS 1002178. P. Bergveld, Sens. Actuators B 88, 1 共2003兲. M. Asahi and T. Matsuo, Suppl. Jpn. Soc. Appl. Phys. 44, 339 共1975兲. P. Gimmel, K. D. Schierbaum, W. Gopel, H. H. Van den Vlekkert, and N. F. de Rooy, Sens. Actuators B 1, 345 共1990兲. 4 P. Estrela, A. G. Stewart, F. Yan, and P. Migliorato, Electrochim. Acta 50, 4995 共2005兲. 5 J. Yu, S. K. Jha, L. Xiao, Q. Liu, P. Wang, C. Surya, and M. Yang, Biosens. Bioelectron. 23, 513 共2007兲. 6 J. T. Mabeck and G. G. Malliaras, Anal. Bioanal. Chem. 384, 343 共2006兲. 7 L. Torsi, A. Dodabalapur, L. Sabbatini, and P. G. Zambonin, Sens. Actuators B 67, 312 共2000兲. 8 J. Liu, M. Agrawal, and K. Varahramyan, Sens. Actuators B 135, 195 共2008兲. 9 Q. Zhang and V. Subramanian, Biosens. Bioelectron. 22, 3182 共2007兲. 10 K. S. Chang, C. C. Chen, J. T. Sheu, and Y.-K. Li, Sens. Actuators B 138, 148 共2009兲. 11 Y.-M. Chu, C.-C. Linb, H.-C. Chang, C. Li, and C. Guoc, Biosens. Bioelectron. 26, 2334 共2011兲. 12 S. M. Al-Hilli, R. T. Al-Mofarji, and M. Willander, Appl. Phys. Lett. 89, 173119 共2006兲. 13 A. Wei, X. W. Sun, J. X. Wang, Y. Lei, X. P. Cai, C. M. Li, Z. L. Dong, and W. Huang, Appl. Phys. Lett. 89, 123902 共2006兲. 14 P. I. Reyes, Z. Zhang, H. Chen, Z. Duan, J. Zhong, G. Saraf, Y. Lu, O. Taratula, E. Galoppini, and N. N. Boustany, IEEE Sens. J. 9, 1302 共2009兲. 15 Z. Zhang, N. W. Emanetoglu, G. Saraf, Y. Chen, P. Wu, J. Zhong, Y. Lu, J. Chen, O. Mirochnitchenko, and M. Inouye, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 53, 1330 共2006兲. 16 R. Murali, P. Brennan, T. Kieber-Emmons, and M. I. Greene, Proc. Natl. Acad. Sci. U.S.A. 93, 6252 共1996兲. 1 2 3 www.advmat.de www.MaterialsViews.com Sung Myung, Aniruddh Solanki, Cheoljin Kim, Jaesung Park, Kwang S. Kim, and Ki-Bum Lee* Nanomaterials such as silicon nanowires (SiNWs),[1,2] carbon nanotubes (CNTs),[3–6] and graphene,[7,8] have gained much attention for use in electrical biosensors due to their nanoscopic and electrical properties. For instance, SiNWs and CNTs can be integrated into field-effect transistors (FETs) to detect small amounts of target biomolecules with high sensitivity and selectivity by measuring electrical disturbances induced by the binding of these biomolecules to the surface of the nanostructure.[9,10] The detection of biomarker proteins with high sensitivity and selectivity is vital for the early diagnosis of many diseases including cancer and HIV. For this purpose, carbonbased nanomaterials such as CNTs and graphene have attracted significant attention for fabricating highly sensitive FET-based biosensors.[6,8,9,11–15] In particular, the use of graphene in FETbased biosensors is becoming more and more appealing not only due to its unique properties, such as higher 2D electrical conductivity, superb mechanical flexibility, large surface area, and high chemical and thermal stability, but also due to its ability to overcome the limitations of CNTs, such as variations in electrical properties of CNT-based devices and the limited surface area of CNTs.[16–24] Nevertheless, there have been only a few reports on the development of graphene FET-based biosensors,[14,25] and their potential as biosensors has not been fully explored. It is therefore critical to develop nanoscopic graphenebased biosensors that are simple in device structure, small in size, and allow label-free detection and real-time monitoring of biomarkers, all of which are essential criteria for biosensors. A key challenge in the above requirements is the achievement of both, well-organized 2D or 3D graphene structures, in microscopic and nanoscopic biosensing devices and well-defined bioconjugation chemistry on graphene. Dr. S. Myung, A. Solanki, C. Kim, Prof. K.-B. Lee Department of Chemistry and Chemical Biology Rutgers, The State University of New Jersey Piscataway, NJ 08854, USA E-mail: [email protected] Prof. K.-B. Lee Institute for Advanced Materials Devices and Nanotechnology (IAMDN) Rutgers, The State University of New Jersey Piscataway, NJ 08854, USA J. Park, Prof. K. S. Kim Center for Superfunctional Materials Department of Chemistry Pohang University of Science and Technology Pohang, Korea DOI: 10.1002/adma.201100014 Adv. Mater. 2011, 23, 2221–2225 Here, we demonstrate a novel strategy for the fabrication and application of a reduced graphene oxide (rGO)[26,27] encapsulated nanoparticle (NP)-based FET biosensor for selective and sensitive detection of key biomarker proteins for breast cancer. It is important to note that we used Human Epidermal growth factor Receptor 2 (HER2) and epidermal growth factor receptor (EGFR), which are known to be over-expressed in breast cancers,[28–30] only as a proof-of-concept to demonstrate the high sensitivity and selectivity of the graphene-encapsulated NP biosensor. This biosensor could be used to detect any important cancer markers with relative ease. In the typical experiments for fabricating grapheneencapsulated NP-based biosensors, individual silicon oxide NPs (100 nm) functionalized with 3-aminopropyltriethoxysilane (APTES) were first coated with thin layers of graphene oxide (GO), which prevent aggregation and maintain high electrical conductivity (Figure 1). This was mainly achieved via the electrostatic interaction between the negatively charged GO (see Supporting Information, Figure S1) and the positively charged silicon oxide NPs. The GO solution (0.05 mg mL−1 in deionized water) was simply injected into the NP solution (5 mg mL−1), wherein the negatively charged GO assembled on the positively charged NP surface until equilibrium coverage was reached.[31–33] The transmission electron microscopy (TEM) image of the GO-coated NPs clearly shows the uniform assembly and saturation density of GO on the NP surface (Figure 1a). The GO thickness on the surface of the NPs was 5 nm, as measured from high-resolution TEM (HR-TEM). As seen in the image, the NPs were connected through a film of GO that was used as an electrical carrier after its reduction to rGO. For efficient use of NP junctions as electrical channels, it was imperative to assemble the NPs with high density on the device. The scanning electron microscopy (SEM) images show well-defined, dense rGO-NP patterns uniformly covering a large area of the silicon oxide substrate (Figure 2a,b). Furthermore, the modified self-assembly method (using centrifugation) allowed us to assemble NPs with high density in a short span and by using minimal amount and concentration of the NP solution. Importantly, the high surface-to-volume ratio of the GO-encapsulated NPs can generate 3D electrical surfaces that significantly enhance detection limits and enable label-free, highly reproducible detection of clinically important cancer markers. One of the most attractive and advantageous aspects of the graphene-encapsulated NP biosensor is the ease of fabrication and measurement (Figure 2). The device was fabricated using photolithography, followed by a lift-off process, both of which are well-established.[33,34] We first generated gold electrodes on a silicon oxide substrate using photolithography and lift-off. To generate arrays of GO-NPs, we patterned the photoresist (AZ 5214) on the substrates with gold electrodes using © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim wileyonlinelibrary.com COMMUNICATION Graphene-Encapsulated Nanoparticle-Based Biosensor for the Selective Detection of Cancer Biomarkers 2221 www.advmat.de COMMUNICATION www.MaterialsViews.com Figure 1. Fabrication process of biomolecular sensor based on graphene-coated NPs. a) Schematic diagram of GO assembly on amine-functionalized NPs and TEM image of NPs coated with GO. b) Fabrication of a metal electrode on the oxide substrate and surface modification for the assembly of GO-NP. c) Photoresist (PR) patterns on the metal electrodes. d) GO-NP assembly in the centrifuge tube. e) Removal of PR patterns and reduction of GO coated on the NP surface. Figure 2. Reduced GO-NP patterns and the electrical property of the rGO-NP device. a) SEM images of rGO-NPs assembled on a large area. b) The SEM images of biosensors consisting of a rGO-NP array with gold electrodes. c) The schematic diagram of measuring process of the gate effect utilizing ionic liquid (left) and the gate effect of ten rGO-NP junctions with a 50 mA channel length (right). 2222 wileyonlinelibrary.com © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2011, 23, 2221–2225 www.advmat.de www.MaterialsViews.com One of the key barriers to using graphene FET-based biosensors is to operate the device under physiological conditions (e.g., different pH and salt concentrations), in which different ionic environments affect the conductivity of graphene FET-based biosensors.[36] To study the working conditions of the graphene FETbased biosensors in aqueous solutions, we measured the gating effect of the rGO-NP-based biosensor using an ionic liquid gate (Figure 2c). A typical source–drain current versus gate potential plot was obtained in an ionic liquid, 1-butyl-3-methylimidazolium tetrafluoroborate (BmimPF6) (see Supporting Information for the synthesis of BmimPF6). In an ionic liquid, the high concentration of ions renders the thickness of the diffusion layer negligible, thus making it useful as a gate insulating layer. Silver (Ag) wire was used as the reference electrode for the measurements in the ionic liquid. The gating effect observed in the GO-NP devices was similar to that observed in rGO thin-film transistors that have ambipolar conduction and p-type behavior near zero gate voltage.[33,37] In the present case, the top-gate bias was swept with ≈0.05 V s−1 sweep speed under the source–drain bias of 0.5 V (Figure 2c). Once the device containing the rGO-NP array was optimized for biosensing, the selective detection of HER2 and EGFR was carried out by functionalizing the rGO-NPs with monoclonal antibodies (mAbs) against HER2 or EGFR. The bioconjugation chemistry is well-established and involved three basic steps.[2,10] First, the reduced GO surface was functionalized with 4-(pyren-1-yl)butanal via π–π interactions by incubating the device in a methanol solution (1:500) of 4-(pyren-1-yl) butanal (see Supporting Information for synthesis) for 30 min (Figure 3a). Second, the aldehyde groups were coupled to the COMMUNICATION photolithography. The exposed silicon oxide surface and gold surface were functionalized with self assembled monolayers (SAMs) of positively charged 3-aminopropyltriethoxysilane (APTES) and cysteamine, respectively. The SAM formation promoted the assembly of the negatively charged GO-NPs (through electrostatic interactions).[34,35] We then employed a relatively simple technique involving centrifugation for the uniform assembly of GO-NPs on the positively charged SAMs. In this technique, the substrate containing the patterned photoresist along with the SAMs was centrifuged in a solution of GO-NPs at 2000 rpm for 3 min in a centrifuge tube. Despite a low concentration of GO-NPs, we were able to achieve uniform films of NPs with a high density in a reproducible manner. This is in stark contrast to the standard methods used for assembling NPs on surfaces. Other methods generally rely on using larger volumes of the solution containing higher concentrations of NPs, where contact of the NPs with the surface is mainly made through infrequent Brownian motion, which eventually causes NP assembly. On the other hand, the centrifugation technique achieved uniform, very dense layers of graphene-encapsulated NPs over a large area in a short time span (Figure 2a,b). We then generated a uniform NP array by removing the patterned photoresist using acetone. The removal of photoresist did not disturb the assembly of GO-NPs. To render the insulating GO electrically conductive, we reduced the GO through an overnight exposure to hydrazine vapor. This method of fabricating the device is very powerful because it can be integrated with conventional microfabrication processes, which makes the device cost effective and relatively easy to produce on a large scale. Figure 3. Real-time detection of cancer marker, HER2. a) The preparation of rGO-NP device. b) Surface functionalization of rGO for immobilizing the antibody. c) Measuring conductance of the devices when the target protein is introduced. d) The sensitivity of the biosensor (relative conductance change,%) in response to the concentration of HER2 with VDS (voltage drain to source) = 1 V and Vg (gate voltage) = 0 V. e) The selectivity of the biosensor in response to PBS buffer, BSA with 50 μg mL−1 and HER2 (100 pM and 1 μM). f) Sensor sensitivity (relative conductance change,%) as a function of the HER2 concentration with VDS = 1 V and Vg = 0 V. All experiments were performed multiple times (sample number, n = 30) to collect statistical data (with error bars) and confirm the reproducibility and robustness of the biosensing system. Adv. Mater. 2011, 23, 2221–2225 © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim wileyonlinelibrary.com 2223 www.advmat.de COMMUNICATION www.MaterialsViews.com 2224 amine groups of the monoclonal HER2 or EGFR antibodies (1:50) through reductive amination in the presence of 4 mM sodium cyanoborohydride in phosphate buffered saline (PBS; pH 7.4) for 2 h (Figure 3b). Third, unreacted aldehyde groups were blocked using 100 mM ethanolamine in a similar manner to prevent non-specific interactions of proteins. Finally, we rinsed the device in a continuous flow of PBS (pH 7.4) for 10 min. We show that the surface chemistry used in the device plays a crucial role in achieving highly selective and sensitive detection of HER2 or EGFR protein (Figure 3c,d). Furthermore, due to the large surface-to-volume ratio of the rGO-NPs, the biosensors were highly efficient as compared to the thin-filmtransistor-based biosensors. The sensitivity of the rGO-NP devices, functionalized with HER2 mAbs, was determined by measuring the changes in conductance as the solution concentration of HER2 was varied from 10 fM to 1 μM (Figure 3d). In all experiments, only 1 μL of each solution was added onto the device. Representative timedependent data show that on the addition of a 10 fM solution of HER2, no change in conductance was observed. However on increasing the concentration to 1 pM, a decrease in conductance of the p-type rGO-NP device was observed due to the binding of HER2 to the mAbs. As the concentrations of the solutions were subsequently increased, a concentration-dependent decrease in the conductance of the rGO-NP device was observed. Thus, the detection limit of the biosensor was observed to be 1 pM in a solution containing only HER2 protein, which is a significant improvement over thin-film-transistor-type sensors based on graphene.[7,14,15] The observed change in electrical conductivity can be attributed to the p-type characteristics of the rGO-NP FET-based sensors because the amine groups on the protein surface are positively charged. Binding of these positively charged target biomolecules, such as HER2 or EGFR, to the rGO surface will induce positive potential gating effects that generate reduced hole density and electrical conductance. To test the selectivity of the graphene FET-based biosensors, we further investigated the selective detection of the device in competitive binding studies with bovine serum albumin (BSA) (Figure 3e). Time-dependent conductance measurements recorded on the rGO-NP devices functionalized with HER2 mAbs showed no change in conductance upon addition of PBS and 50 μg mL−1 BSA. However, upon addition of 1 μL of 100 pM solution of HER2 to the BSA solution on the device, a rapid and sharp change in conductance was observed, demonstrating the high selectivity of the device. Upon adding the 1 μM solution of HER2, the conductance further decreased rapidly and drastically. In spite of the presence of a solution with a very high concentration of BSA (50 μg mL−1), the detection limit of the target protein, HER2, was 100 pM, clearly demonstrating the remarkable sensitivity and selectivity of the rGO-NP biosensor. Figure 3f shows the sensitivity (relative conductance change) of the biosensor as a function of the HER2 concentration. The lowest HER2 concentration level that could be detected is 1 pM, which shows a decrease in conductance (3.9%). Similar to the non-linear behavior of CNT FET-based sensors,[38–40] the sensor responses increase non-linearly with the increase in the HER2 concentration from 1 pM to 1 μM, which clearly shows that the sensor response is due to the binding of HER2 to the HER2 mAbs. wileyonlinelibrary.com In addition to HER2, we similarly investigated the sensitivity and selectivity of the device for detecting EGFR. We functionalized the device with EGFR mAbs and observed the change in conductance upon addition of EGFR solution. The trend in conductance change was similar to that observed with HER2, with the detection limit being 100 pM for EGFR and 10 nM in the presence of BSA (50 μg mL−1; see Supporting Information, Figure S2). We believe the slight decrease in sensitivity for detecting EGFR (relative to HER2) might be due to the difference in binding affinities of the two mAbs to their respective proteins. However, the result demonstrates the capacity of the biosensor to detect different biomarkers in a sensitive and selective manner. In conclusion, we have demonstrated the application of a graphene-encapsulated NP-based biosensor for highly selective and sensitive detection of cancer biomarkers by using surface chemistry principles combined with nanomaterials and micro- and nanofabrication techniques. The novel 3D structure of graphene-encapsulated NPs significantly increases the surface-to-volume ratio in FET-type biosensors, thereby improving the detection limits (1 pM for HER2 and 100 pM for EGFR) for the target cancer biomarkers. In addition, we demonstrated the highly selective nature of the biosensor as we detected low concentrations of the target cancer biomarkers in the presence of a highly concentrated BSA solution. The ease of fabrication and biocompatibility, along with excellent electrochemical and electrical properties of graphene nanocomposites, makes the graphene-encapsulated NP-based biosensor an ideal candidate for future biosensing applications in a clinical setting. Experimental Section Preparation of Reduced Graphene Oxide and SiO2 NPs: GO was obtained from SP-1 graphite utilizing the modified Hummer method.[27] For the GO assembly on the surface of NPs, the GO suspension was injected into the nanoparticle solution for 10 min, and GO-NPs were separated from GO solution using a centrifuge. SiO2 NP (100 nm) solution was purchased from Corpuscular Inc. For GO-NP assembly, the photoresist-patterned substrate was placed in the NP solution and GO-NPs were assembled on the substrate by applying centrifugal force. After the deposition of GO-NPs on the substrate, the GO on the NP surface, having the low conductance, was reduced to graphene by exposure to hydrazine vapor overnight. Surface Molecular Pattering: 3-Aminopropyltriethoxysilane (APTES) and cysteamine molecules, used to form SAMs, and the solvents were purchased from Sigma-Aldrich. For the patterning of APTES SAM on SiO2, the photoresist (AZ5214) was first patterned by photolithography using a short baking time (<10 min at 95 °C). The patterned substrate was placed in the APTES solution (1:500 (v:v) in anhydrous hexane) for 7 min. For the patterning of APTES on the on SiO2 layer, the substrate with photoresist patterns was placed in an APTES solution (1:500 (v:v) in anhydrous hexane) for 10 min. The photoresist was then removed with acetone. Metal Deposition and Measurement of Graphene Devices: For the electrode fabrication, the photoresist was first patterned on the substrate. Ti/Au (10/30 nm) was then deposited on the substrate and the remaining photoresist was then removed with acetone for the lift-off process. A Keithley-4200 semiconductor parameter analyzer was used for measurement and data collection. Supporting Information Supporting Information is available from the Wiley Online Library or from the author. © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2011, 23, 2221–2225 www.advmat.de www.MaterialsViews.com This work was supported by the NIH Director’s Innovator Award [(1DP20D006462–01), K.-B. L.] and the N.J. Commission on Spinal Cord Injury grant [(09–3085-SCRE-0), K.-B. L.]. K.S.K. acknowledges support from NRF (National Honor Scientist Program: 2010-0020414). Received: January 4, 2011 Revised: February 16, 2011 Published online: April 5, 2011 [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] Y. Cui, Q. Q. Wei, H. K. Park, C. M. Lieber, Science 2001, 293, 1289. F. Patolsky, G. F. Zheng, C. M. Lieber, Nat. Protoc. 2006, 1, 1711. K. Bradley, M. Briman, A. Star, G. Gruner, Nano Lett. 2004, 4, 253. R. J. Chen, H. C. Choi, S. Bangsaruntip, E. Yenilmez, X. W. Tang, Q. Wang, Y. L. Chang, H. J. Dai, J. Am. Chem. Soc. 2004, 126, 1563. A. Star, Y. Lu, K. Bradley, G. Gruner, Nano Lett. 2004, 4, 1587. X. W. Tang, S. Bansaruntip, N. Nakayama, E. Yenilmez, Y. L. Chang, Q. Wang, Nano Lett. 2006, 6, 1632. J. Lu, I. Do, L. T. Drzal, R. M. Worden, I. Lee, ACS Nano 2008, 2, 1825. S. Mao, G. H. Lu, K. H. Yu, Z. Bo, J. H. Chen, Adv. Mater. 2010, 22, 3521. A. T. Woolley, C. Guillemette, C. L. Cheung, D. E. Housman, C. M. Lieber, Nat. Biotechnol. 2000, 18, 760. G. F. Zheng, F. Patolsky, Y. Cui, W. U. Wang, C. M. Lieber, Nat. Biotechnol. 2005, 23, 1294. D. Du, Z. X. Zou, Y. S. Shin, J. Wang, H. Wu, M. H. Engelhard, J. Liu, I. A. Aksay, Y. H. Lin, Anal. Chem. 2010, 82, 2989. J. H. Jung, D. S. Cheon, F. Liu, K. B. Lee, T. S. Seo, Angew. Chem. Int. Ed. 2010, 49, 5708. C. H. Lu, H. H. Yang, C. L. Zhu, X. Chen, G. N. Chen, Angew. Chem. Int. Ed. 2009, 48, 4785. Y. Ohno, K. Maehashi, Y. Yamashiro, K. Matsumoto, Nano Lett. 2009, 9, 3318. C. S. Shan, H. F. Yang, J. F. Song, D. X. Han, A. Ivaska, L. Niu, Anal. Chem. 2009, 81, 2378. P. Avouris, Nano Lett. 2010, 10, 4285. K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, A. A. Firsov, Science 2004, 306, 666. Adv. Mater. 2011, 23, 2221–2225 [18] K. S. Kim, Y. Zhao, H. Jang, S. Y. Lee, J. M. Kim, J. H. Ahn, P. Kim, J. Y. Choi, B. H. Hong, Nature 2009, 457, 706. [19] X. S. Li, W. W. Cai, J. H. An, S. Kim, J. Nah, D. X. Yang, R. Piner, A. Velamakanni, I. Jung, E. Tutuc, S. K. Banerjee, L. Colombo, R. S. Ruoff, Science 2009, 324, 1312. [20] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Katsnelson, I. V. Grigorieva, S. V. Dubonos, A. A. Firsov, Nature 2005, 438, 197. [21] Y. B. Zhang, Y. W. Tan, H. L. Stormer, P. Kim, Nature 2005, 438, 201. [22] A. K. Geim, K. S. Novoselov, Nat. Mater. 2007, 6, 183. [23] L. Y. Jiao, L. Zhang, X. R. Wang, G. Diankov, H. J. Dai, Nature 2009, 458, 877. [24] X. Du, I. Skachko, A. Barker, E. Y. Andrei, Nat. Nanotechnol. 2008, 3, 491. [25] X. C. Dong, Y. M. Shi, W. Huang, P. Chen, L. J. Li, Adv. Mater. 2010, 22, 1649. [26] D. A. Dikin, S. Stankovich, E. J. Zimney, R. D. Piner, G. H. B. Dommett, G. Evmenenko, S. T. Nguyen, R. S. Ruoff, Nature 2007, 448, 457. [27] G. Eda, G. Fanchini, M. Chhowalla, Nat. Nanotechnol. 2008, 3, 270. [28] A. Citri, Y. Yarden, Nat. Rev. Mol. Cell Biol. 2006, 7, 505. [29] N. E. Hynes, H. A. Lane, Nat. Rev. Cancer 2005, 5, 341. [30] Y. Yarden, M. X. Sliwkowski, Nat. Rev. Mol. Cell Biol. 2001, 2, 127. [31] I. Jung, D. A. Dikin, R. D. Piner, R. S. Ruoff, Nano Lett. 2008, 8, 4283. [32] T. Kobayashi, N. Kimura, J. B. Chi, S. Hirata, D. Hobara, Small 2010, 6, 1210. [33] S. Myung, J. Park, H. Lee, K. S. Kim, S. Hong, Adv. Mater. 2010, 22, 2045. [34] S. Myung, M. Lee, G. T. Kim, J. S. Ha, S. Hong, Adv. Mater. 2005, 17, 2361. [35] M. Lee, J. Im, B. Y. Lee, S. Myung, J. Kang, L. Huang, Y. K. Kwon, S. Hong, Nat. Nanotechnol. 2006, 1, 66. [36] W. R. Yang, K. R. Ratinac, S. P. Ringer, P. Thordarson, J. J. Gooding, F. Braet, Angew. Chem. Int. Ed. 2010, 49, 2114. [37] F. Chen, Q. Qing, J. L. Xia, J. H. Li, N. J. Tao, J. Am. Chem. Soc. 2009, 131, 9908. [38] A. Star, J. C. P. Gabriel, K. Bradley, G. Gruner, Nano Lett. 2003, 3, 459. [39] M. Abe, K. Murata, A. Kojima, Y. Ifuku, M. Shimizu, T. Ataka, K. Matsumoto, J. Phys. Chem. C 2007, 111, 8667. [40] B. L. Allen, P. D. Kichambare, A. Star, Adv. Mater. 2007, 19, 1439. © 2011 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim wileyonlinelibrary.com COMMUNICATION Acknowledgements 2225 Multifunctional Nanomaterials Graphite-Coated Magnetic Nanoparticles as Multimodal Imaging Probes and Cooperative Therapeutic Agents for Tumor Cells Joung Kyu Park, Jongjin Jung, Prasad Subramaniam, Birju P. Shah, Cheoljin Kim, Jong Kyo Lee, Jee-Hyun Cho, Chulhyun Lee, and Ki-Bum Lee* An effective therapeutic approach against cancer typically requires the combination of several modalities, such as chemotherapy, radiation, and hyperthermia. In this regard, the development of multifunctional nanomaterial-based systems with combined therapeutic and molecular imaging capabilities has shown great potential but has not been fully explored. In particular, magnetic nanomaterials have been at the forefront of cancer research as noninvasive imaging probes as well as multifunctional therapeutics.[1] For example, magnetic nanoparticles (MNPs) with appropriate surface modifications have been successfully applied to deliver therapeutic biomolecules, such as anticancer drugs, antibodies, and siRNAs, to target tumor cells or tissues.[2] Moreover, the unique physical and chemical properties of these magnetic nanostructures have enabled their wide applications in cancer imaging and therapy, including magnetic resonance imaging (MRI) and hyperthermia.[3] Promising advances have been made in synthesizing multifunctional MNPs from various materials, including metals,[4] metal oxides,[5] metal alloys,[6] and metal– graphitic-shell nanomaterials,[7] with different properties. However, current studies are mostly focused on the synthesis and characterization of materials with limited demonstration of their biomedical applications, like molecular imaging and therapy. As a result, research efforts towards developing MNP-based multimodal therapeutics to control the tumor microenvironment are highly limited and have not been fully explored. Therefore, in order to address the challenges Dr. J. K. Park, J. Jung, P. Subramaniam, B. P. Shah, Dr. C. Kim, Prof. K.-B. Lee Department of Chemistry and Chemical Biology Rutgers, The State University of New Jersey Piscataway, NJ 08854, USA E-mail: [email protected] Dr. J. K. Park, Dr. J. K. Lee Center for Nano-Biofusion Research Korea Research Institute of Chemical Technology Daejon 305–600, Korea Dr. J.-H. Cho, Dr. C. Lee Division of Magnetic Resonance Research Korea Basic Science Institute Ochang 363–883, Korea DOI: 10.1002/smll.201100012 small 2011, 7, No. 12, 1647–1652 of MNP-based therapeutics, as well as to narrow the gap between current nanoparticle-based multimodal imaging approaches and their clinical applications, there is a clear need to synthesize effective chemotherapeutic MNPs and to develop multimodal therapies for targeting specific oncogenes, thereby activating/deactivating corresponding key signaling pathways. In this Communication, we describe the novel synthesis and a systematic in vitro evaluation and application of multifunctional magnetic nanoparticles (MNPs) with an iron cobalt core and a graphitic carbon shell (FeCo/C) for the targeted delivery of small interfering RNA (siRNA) to tumor cells with a concomitant hyperthermia-based therapy, thereby cooperatively inhibiting proliferation of and inducing apoptosis in tumor cells (Figure 1). In parallel, we also demonstrate that our MNPs can be used as highly sensitive magnetic resonance and Raman imaging probes. As a model study, we used glioblastoma multiforme (GBM) cell lines, the most malignant and difficult-to-treat brain tumor cells. We hypothesized that the targeted delivery of our siRNA–MNP constructs against the oncogenic receptor (EGFRvIII) and subsequent hyperthermal treatment would selectively, as well as cooperatively, damage the tumor cells, resulting in the synergistic inhibition of tumor-cell proliferation and the induction of apoptosis via the deactivation of the PI3K/AKT signaling pathway. Hence, these MNP-based therapeutics could potentially be used for the simultaneous imaging and therapy of malignant tumors both in vitro and in vivo. Recent efforts in cancer therapy have demonstrated the application of hyperthermia, which involves localized heating of cancerous cells or tissues, as an adjuvant to chemotherapy and radiation to improve their efficacy.[8] Hyperthermia typically involves increasing the local temperature of the tumor region to 42–46 °C over a given time period, ultimately resulting in apoptosis of the heat-sensitized cancer cells.[9] One of the best methods of achieving a localized hyperthermal effect is to deliver MNPs to the target cells and subsequently apply electromagnetic fields following their cellular uptake/localization.[10] Furthermore, hyperthermia and its downstream effects can be significantly enhanced by the concomitant use of other cancer therapies, including radiation and drug/gene delivery and vice versa. To develop cooperative (hyperthermia and siRNA delivery) therapeutic systems based upon MNPs, we © 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim wileyonlinelibrary.com 1647 communications J. K. Park et al. 11-nm FeCo ≈235 emu g−1), has a high Curie temperature, and has high magnetic anisotropy energies, all of which are critical in order to enhance their potential for biomedical applications, such as MRI and hyperthermia; ii) the thickness of the outer graphite-shell layers can be controlled by our synthetic method, which would lead to an improved Raman signal intensity for detecting cancerous cells; and iii) the nanoparticles are chemically inert due to the presence of a graphitic carbon shell[11] and can be made biocompatible by appropriate surface modifications (e.g., dextran-ligand coating). Moreover, compared to conventional methods[1b,12] for the synthesis of core–shell metal-alloy magnetic nano[13] Figure 1. Magnetic FeCo–graphite nanoparticles for multimodal imaging and targeted materials, such as electric-arc discharge, tumor therapy. a) Detailed structure of the MNPs depicting the highly magnetic FeCo core, high-temperature thermal decomposiprotective Raman-active graphite shell, and the biocompatible dextran coating. b) Inhibition tion,[14] and chemical vapor deposition of proliferation and induction of apoptosis via combined siRNA delivery and hyperthermia (CVD),[11] our novel hydrothermal synusing siRNA–FeCo/C NP constructs. thetic approach has several advantages, such as relatively milder synthetic condisynthesized graphitic-carbon-protected iron cobalt (FeCo/C) tions, low environmental impact, cost effectiveness, ease of nanoparticles (7 and 11 nm in diameter) with a body-centered scalability (see Figure S3 in the SI), and the exclusion of toxic cubic (bcc) crystalline structure using hydrothermal synthetic solvents and size-separation techniques. methods followed by an annealing process at 1000 °C. HighAnother critical step to realize the full potential of our resolution transmission electron microscopy (HR-TEM) and multimodal FeCo/C NPs for in vitro/in vivo biomedical appliX-ray diffraction (XRD) confirmed the excellent chemical/ cations (e.g., targeted drug/gene delivery, MRI, or hyperphysical properties of our FeCo/C NPs, such as monodisper- thermal therapy) is to make the nanoparticles biocompatible sity, narrow size distribution of the nanoparticles, and the presence of a crystalline bcc FeCo core (Figure 2). The graphiticcarbon shells surrounding the FeCo core were confirmed by Raman spectroscopy analysis and HR-TEM. Furthermore, the thickness of the graphitic-carbon coating could be monitored by the intensity of the Raman signal (marked by arrows in Figure 2d). We also characterized the magnetic properties of our FeCo/C NPs using a superconducting quantum interference device (SQUID). Our FeCo/C NPs were found to display remarkable superparamagnetic properties at room temperature, as suggested by the significantly higher value of the saturation magnetization (Ms) for our FeCo/C NPs as compared to that of the commercially available Fe3O4 (Figure 2e). This was attributed to the higher magnetic moments of the FeCo/C NPs as a result of the high-temperature Figure 2. Structural and magnetic properties of the FeCo/C nanoparticles. a) TEM image annealing, as shown in Figure S2 in the of the 11-nm nanoparticles (scale bar = 20 nm). b) HR-TEM image of the nanoparticles showing the crystalline lattice structure of the FeCo core. c) Powder XRD for the 7- and 11-nm Supporting Information (SI). In this work, our FeCo/C MNPs have nanoparticles showing the presence of a bcc crystalline core and graphitic shell. d) Raman spectrum (excitation at 632.8 nm) of the 11-nm nanoparticles with single and multiple carbon several advantages over other conven- shells (marked by arrows) showing the D and G bands of graphitic carbon (scale bar = 2 nm). tional magnetic nanoparticles, such as e) Room-temperature magnetization versus applied magnetic field for 11-nm FeCo/C NPs FePt, Fe2O3, and Fe3O4: i) FeCo exhibits (red symbols) and Fe3O4 (black symbols). Note that no hysteresis loop exists owing to the an excellent magnetization value (e.g., superparamagnetism of the nanoparticles. 1648 www.small-journal.com © 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim small 2011, 7, No. 12, 1647–1652 Graphite-Coated Magnetic Nanoparticles as Multimodal Imaging Probes and modify the surface for achieving targetspecific intracellular delivery.[15] Several strategies, including coating with molecules such as dextran, have been successfully used for functionalizing magnetic nanoparticles and rendering them biocompatible.[16] Hence, we synthesized a series of dextran derivatives (see Figure S4a in the SI) and used them to coat our FeCo/C NPs (see Figure S4b in the SI). The dextrancoated FeCo/C NPs were found to have excellent colloidal stability and the hydrodynamic diameter was found to be ≈150 nm (see Figures S5 and S6 in the SI). These dextran-modified FeCo/C NPs were then conjugated to specific cancer-targeting biomolecules, such as epidermal growth factor receptor (EGFR) antibodies and cyclic-RGD (c-RGD) peptide, to target glioblastoma cells and to improve their intracellular uptake. The conjugation of biomolecules such as EGFR antibodies and c-RGD to our FeCo/C NPs not only increases transfection of our nanoparticles via receptor-mediated endocytosis but can also selectively target the glioblastoma cells by binding to receptors (EGFR and integrins) known to be overexpressed in glioblastoma cells.[17] Highly sensitive multimodal imaging tools like MRI, computed tomography (CT), positron-emission tomography (PET), and Raman imaging, which rely on nanomaterials as molecular probes, have gained much attention as diagnostic tools for specific cancers, such as breast and brain cancers.[18] In order to investigate the capability of our FeCo/C NPs as multimodal imaging nanoprobes, we focused on two important imaging methods for cancer, in vivo MRI and in vitro Raman imaging, both of which provide complementary information about tumor microenvironments at the tissue level and/or cellular Figure 3. MR measurements and imaging of FeCo/C NPs and Resovist. a) Concentrationdependent T2 measurements of FeCo/C NPs and Resovist solutions. b) T2-weighted MR images level.[3b,18b] In order to test the capability of various nanoparticle solutions. The FeCo/C NPs show higher MR image contrast (several of our MNPs as MRI contrast agents, the fold) as compared to Resovist, a traditional MRI contrast agent. c) T -weighted MR images of a 2 transverse relaxation times (T2) of the rat before (t = 0 min) and 30 min after (t = 30 min) injection of FeCo/C NPs (left) and Resovist water protons were measured in a specific (right) into a rat’s tail vein. The nanoparticles were seen to localize in the liver (green arrow), magnetic field and their values were com- spleen (blue arrow), and kidneys (red arrow) of the animal. Also, FeCo/C NPs show a higher pared to those of the commercially avail- imaging contrast at a lower concentration (0.25 mg of Fe) as compared to Resovist (2.5 mg able MRI contrast agent Resovist (Fe3O4, of Fe). d) Raman spectra of U87-EGFP cells (marked by arrows) treated with 11-nm FeCo/C NPs. The spectra shows the capability of the FeCo/C NPs to be used as imaging agents at the from Roche). Both the 7- and 11-nm single-cell level (see Figure S8 in the SI for detailed Raman imaging). FeCo/C NPs exhibited a higher relaxivity coefficient (r2, 252 and 392 mm−1 s−1, respectively) and enhanced T2-weighted MR contrast (Figure 3a,b) nanoparticles accumulated in the liver, spleen, and kidneys as compared to Resovist (r2 = 140 mm−1 s−1). Characteriza- and their imaging contrast was significantly higher (about tion of our biocompatible dextran-coated FeCo/C NPs as in 10 times) than that of Resovist (Figure 3c). We also monitored vivo MRI agents was performed by injecting the MNPs into the in vivo biodistribution and MRI contrast of these FeCo/C a rat’s tail vein. Preliminary MRI studies showed that our NPs over 10 days. It was found that the NPs were retained small 2011, 7, No. 12, 1647–1652 © 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.small-journal.com 1649 communications J. K. Park et al. in the liver and spleen and that there was no significant loss of imaging contrast over this time period (see Figure S7 in the SI). Furthermore, the analysis of in vitro Raman imaging results, wherein confocal microscopy was used to observe the NPinternalized U87 cell lines, confirmed that the intensity of the D and G bands of the graphitic-carbon shells was indeed proportional to the thickness of the carbon shells on the FeCo/C NPs (Figure 3d). This data strongly suggests that our FeCo/C NPs have a well defined correlation and sensitive response when used as Raman imaging probes at the single-cancer-cell level. These preliminary in vitro and in vivo studies showed no visible cytotoxicity due to the degradation of the FeCo/C NPs. It has been reported that FeCo/C NPs are safely excreted over time through the biliary system after being taken up by the liver and spleen.[19] Hence, it is highly Figure 4. In vitro hyperthermia and siRNA-delivery studies of the FeCo/C NPs. a,b) U87–EGFP unlikely that these FeCo/C NPs would cell death induced by hyperthermia in cocultures of the highly tumorigenic U87–EGFP cells pose any serious toxicity issues when used (marked by arrows) and the less-tumorigenic PC-12 cells (marked by arrows) via the targeted in vivo. A detailed toxicological evaluation delivery of FeCo/C NPs to the U87 cells. Fluorescence images (a) and quantitative analysis of these MNPs in animal models would be (b) show that significant hyperthermia-induced cell death is observed in U87 cells, while the PC-12 cells keep proliferating with time. (Note that the number of cells at 0 h was taken to forthcoming in the future. be 100% and the cell counts at other time points were normalized to this value.) Annexin V Having demonstrated the potential of assays for detection of early apoptosis proved that the cell death was caused by localized our FeCo/C NPs as multimodal imaging hyperthermia-induced apoptosis rather than necrosis (see Figure S12 in the SI). c) MTS assay probes, we then focused on evaluating demonstrating the synergistic inhibition of proliferation and induction of cell death by the their ability to be used as targeted hyper- combined siRNA and hyperthermia treatment using siRNA-FeCo/C NPs in U87–EGFRvIII cells thermia agents for tumor therapy in vitro. as compared to individual treatments and nontreated controls. The scale bar in all images We hypothesized that our FeCo/C NPs is 100 μm. would be more efficient hyperthermia agents than Fe3O4 NPs of similar size due to their high mag- incubated in cocultures of glioblastoma cells (U87–EGFP), netization and therefore would be more effective in inhibiting which had been genetically labeled with enhanced green fluproliferation and inducing apoptosis of target brain tumor orescent protein (EGFP) and present EGFRs on their memcells (bTCs). To test their efficacy as targeted hyperthermia branes and several other less tumorigenic cells, such as PC-12 agents, we first measured the minimum time required to and astrocytes, which tend to have low expression levels of attain the therapeutic temperature (≈43 °C) when placed in integrins and EGFRs. Our data clearly indicates that the sura homogeneous magnetic field. The specific absorption rates face modification of our FeCo/C NPs with c-RGD peptide or (SARs) of FeCo/C and Fe3O4 NPs (≈69 W g−1 for FeCo/C EGFR antibodies resulted in their selective cellular uptake and 13 W g−1 for Fe3O4), as derived from the plots of temper- by the target bTCs, as compared to PC-12 or astrocytes (see ature versus time in aqueous solutions under a 334 kHz mag- Figure S11 in the SI). Following intracellular uptake of the netic field (which is the optimum frequency range estimated aforementioned NP constructs, the cells were exposed to by the Neel and Brownian relaxation-time simulations[20]), an alternating current (AC) magnetic field for 15 min. Sigindicated that the time required for our FeCo/C NPs to reach nificant inhibition of proliferation and hyperthermia-induced the therapeutic temperature was ten times shorter than that cell death was observed mainly in the U87 cells while the of Fe3O4 NPs (see Figure S9 in the SI). We next evaluated less-tumorigenic PC-12 cells largely continued proliferating the concentration-dependent cytotoxicity of our FeCo/C NPs with time (Figure 4a,b). by serial-dilution investigations. From this study, the range of In the past decade, there has been considerable interest concentrations inducing negligible cytotoxic effects on cells in the development of nanoparticle-based siRNA-delivery was identified and the concentrations (95% cell viability at methods for cancer therapy. RNA interference (RNAi) 300 μg mL−1 after ≈24 h post transfection) within this range involves the use of siRNAs to selectively mediate the cleavage were used for our subsequent experiments (see Figure S10 in of complementary mRNA sequences and thus regulate target the SI). We then sought to precisely increase the temperature gene expression.[21] In combination with other chemotheraof tumor cells while minimizing the exposure of other cells peutic methods like hyperthermia, RNAi could prove to be to hyperthermic temperatures. For this purpose, FeCo/C NPs a powerful tool to manipulate the tumor microenvironment. functionalized with c-RGD peptide or EGFR antibodies were In order to deliver siRNA to the target cells, the FeCo/C 1650 www.small-journal.com © 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim small 2011, 7, No. 12, 1647–1652 Graphite-Coated Magnetic Nanoparticles as Multimodal Imaging Probes NPs were conjugated to the siRNA using polyethyleneimine (PEI) as a cationic polymer via a layer-by-layer approach[22] (see Figure S13 in the SI). We initially optimized the knockdown efficiency of our siRNA–FeCo/C NP construct by the suppression of EGFP in U87–EGFP cells. The decrease in green fluorescence intensity due to siRNA-mediated knockdown of EGFP was monitored to assess the knockdown efficiency (≈80% after 3 days of transfection) of our siRNA–NP constructs (see Figure S14 in the SI). Once the conditions for the siRNA delivery and knockdown were optimized, we focused on inhibiting the proliferation and inducing apoptosis of U87–EGFRvIII cell lines overexpressing the oncogenic EGFRvIII gene. EGFRvIII is a mutant type of EGFR as well as an oncogenic receptor that is highly expressed only in tumor cells, and not in normal cells.[23] The effect of the knockdown of the EGFRvIII oncogene using our siRNA– FeCo/C NP constructs was assessed using a microscope at 96 h post transfection (see Figure S14 in the SI). We hypothesized that the knockdown of the target oncogene, EGFRvIII, with our multifunctional siRNA–FeCo/C NP constructs, combined with hyperthermia, would lead to a cooperative inhibition of tumor-cell proliferation and increase in cell death. The U87–EGFRvIII cells were incubated with our FeCo/C NPs modified with siRNA against EGFRvIII, followed by hyperthermia treatment for 5 min at 72, 96, and 120 h post siRNA treatment. Quantitative analysis based upon the 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)2-(4-sulfophenyl)-2H-tetrazolium assay, commonly known as the MTS assay (Figure 4c) showed that treatment of cells with our siRNA–MNPs against EGFRvIII followed by hyperthermia induced significantly more cell death as compared to the controls. This could be attributed to the fact that silencing of the EGFRvIII oncogene results in a decrease in expression of the focal adhesion proteins that make the cells more susceptible to heat, thereby leading to a synergistic increase in cell death.[24] Our study clearly demonstrates that the appropriate combination of various therapeutic modalities using our FeCo/C NPs can significantly enhance the therapeutic efficacy relative to the individual components. In summary, this work provides an early demonstration of integrating multimodal imaging with combined hyperthermia and siRNA-based therapy in malignant tumor cells using highly efficient FeCo/C NPs. Our FeCo/C NPs have successfully been demonstrated to have a well defined correlation and a fast and sensitive thermal response to the strength of the applied magnetic field. At the same time, our FeCo/C NPs could be developed as novel therapeutic and diagnostic tools for cancer research. The ability to functionalize the graphitic surface of the FeCo/C NPs with targeting ligands and biomolecules would be critical to realize the potential of nanoparticle-based diagnosis and therapy of various cancers. It should be noted that our FeCo/C NPs showed excellent MRI contrast results compared to the conventional MRI contrast agents as well as enabling us to collect the Raman spectral information at the single-cell level. More importantly, the use of our FeCo/C NPs for site-specific and localized hyperthermia in conjunction with siRNA therapy against oncogenes would greatly complement and enhance the effects of other therapeutic modalities, including gene therapy and small 2011, 7, No. 12, 1647–1652 chemotherapy, thereby reducing the dose of anticancer drugs, mitigating their toxic side effects, and effectively circumventing drug resistance in cancers. Supporting Information Supporting Information is available from the Wiley Online Library or from the author. Acknowledgements We thank Prof. Huixin He for helping us with the hyperthermia studies, Prof. Sang-Wook Cheong and Prof. Martha Greenblatt for their support for the SQUID measurements and the IAMDN center for allowing us to use their high-resolution TEM facility. We are grateful to the KBLEE group members (Aniruddh Solanki, Shreyas P. Shah, and Michael Koucky) for their valuable suggestions for the manuscript. J.K.P. acknowledges the research program at KRICT and Dr. Young-Duk Suh and Dr. Kee-Suk Jun for the Raman imaging. K.-B.L. acknowledges the NIH Director’s Innovator Award (1DP20D006462–01) and is also grateful to the N.J. Commission on Spinal Cord Research grant (09–3085-SCR-E-0). [1] a) J. W. M. Bulte, T. Douglas, B. Witwer, S. C. Zhang, E. Strable, B. K. Lewis, H. Zywicke, B. Miller, P. van Gelderen, B. M. Moskowitz, I. D. Duncan, J. A. Frank, Nat. Biotechnol. 2001, 19, 1141–1147; b) A. H. Lu, E. L. Salabas, F. Schuth, Angew. Chem. Int. Ed. 2007, 46, 1222–1244; c) H. Zeng, S. H. Sun, Adv. Funct. Mater. 2008, 18, 391–400; d) A. Solanki, J. D. Kim, K. B. Lee, Nanomedicine 2008, 3, 567–578. [2] a) J. H. Gao, H. W. Gu, B. Xu, Acc. Chem. Res. 2009, 42, 1097–1107; b) S. H. Hu, K. T. Kuo, W. L. Tung, D. M. Liu, S. Y. Chen, Adv. Funct. Mater. 2009, 19, 3396–3403. [3] a) J. Kim, J. E. Lee, S. H. Lee, J. H. Yu, J. H. Lee, T. G. Park, T. Hyeon, Adv. Mater. 2008, 20, 478–483; b) W. B. Lin, T. Hyeon, G. M. Lanza, M. Q. Zhang, T. J. Meade, MRS Bulletin 2009, 34, 441–448; c) D. Hogemann, V. Ntziachristos, L. Josephson, R. Weissleder, Bioconj. Chem. 2002, 13, 116–121. [4] D. L. Huber, Small 2005, 1, 482–501. [5] H. Lee, M. K. Yu, S. Park, S. Moon, J. J. Min, Y. Y. Jeong, H. W. Kang, S. Jon, J. Am. Chem. Soc. 2007, 129, 12739–12745. [6] S. H. Sun, C. B. Murray, D. Weller, L. Folks, A. Moser, Science 2000, 287, 1989–1992. [7] M. Mahmood, Y. Xu, Z. R. Li, E. Dervishi, N. Ali, V. Saini, A. S. Biris, S. Trigwell, V. P. Zharov, A. R. Biris, IEEE Trans. Ind. Appl. 2009, 45, 2162–2169. [8] N. G. Huilgol, D. Gupta, R. Dixit, Int. J. Hyperther. 2010, 26, 21–25. [9] N. K. Prasad, K. Rathinasamy, D. Panda, D. Bahadur, J. Mater. Chem. 2007, 17, 5042–5051. [10] a) P. Cherukuri, E. S. Glazer, S. A. Curleya, Adv. Drug Deliv. Rev. 2010, 62, 339–345; b) A. Ito, F. Matsuoka, H. Honda, T. Kobayashi, Canc. Gene Ther. 2003, 10, 918–925. [11] a) W. S. Seo, S. M. Kim, Y. M. Kim, X. Sun, H. J. Dai, Small 2008, 4, 1968–1971; b) W. S. Seo, J. H. Lee, X. M. Sun, Y. Suzuki, D. Mann, Z. Liu, M. Terashima, P. C. Yang, M. V. McConnell, D. G. Nishimura, H. J. Dai, Nat. Mater. 2006, 5, 971–976. [12] a) J. Park, K. J. An, Y. S. Hwang, J. G. Park, H. J. Noh, J. Y. Kim, J. H. Park, N. M. Hwang, T. Hyeon, Nat. Mater. 2004, 3, 891–895; © 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.small-journal.com 1651 communications [13] [14] [15] [16] [17] [18] 1652 J. K. Park et al. b) M. Bystrzejewski, A. Huczko, H. Lange, P. Baranowski, G. Cota-Sanchez, G. Soucy, J. Szczytko, A. Twardowski, Nanotechnology 2007, 18, 145608; c) T. Hyeon, S. S. Lee, J. Park, Y. Chung, H. Bin Na, J. Am. Chem. Soc. 2001, 123, 12798–12801. X. L. Dong, Z. D. Zhang, S. R. Jin, B. K. Kim, J. Appl. Phys. 1999, 86, 6701–6706. a) M. Yamada, S. J. Okumura, K. Takahashi, J. Phys. Chem. Lett. 2010, 1, 2042–2045; b) S. Behrens, H. Bonnemann, N. Matoussevitch, A. Gorschinski, E. Dinjus, W. Habicht, J. Bolle, S. Zinoveva, N. Palina, J. Hormes, H. Modrow, S. Bahr, V. Kempter, J. Phys. Condens. Matter 2006, 18, S2543-S2561. Y. Zhang, N. Kohler, M. Q. Zhang, Biomaterials 2002, 23, 1553–1561. a) C. C. Berry, S. Wells, S. Charles, G. Aitchison, A. S. G. Curtis, Biomaterials 2004, 25, 5405–5413; b) H. Hifumi, S. Yamaoka, A. Tanimoto, T. Akatsu, Y. Shindo, A. Honda, D. Citterio, K. Oka, S. Kuribayashi, K. Suzuki, J. Mater. Chem. 2009, 19, 6393–6399. J. Xie, K. Chen, H. Y. Lee, C. J. Xu, A. R. Hsu, S. Peng, X. Y. Chen, S. H. Sun, J. Am. Chem. Soc. 2008, 130, 7542–7543. a) H. Lee, E. Lee, D. K. Kim, N. K. Jang, Y. Y. Jeong, S. Jon, J. Am. Chem. Soc. 2006, 128, 7383–7389; b) X. M. Qian, X. H. Peng, D. O. Ansari, Q. Yin-Goen, G. Z. Chen, D. M. Shin, L. Yang, www.small-journal.com [19] [20] [21] [22] [23] [24] A. N. Young, M. D. Wang, S. M. Nie, Nat. Biotechnol. 2008, 26, 83–90; c) A. S. Haka, K. E. Shafer-Peltier, M. Fitzmaurice, J. Crowe, R. R. Dasari, M. S. Feld, Cancer Res. 2002, 62, 5375–5380. J. H. Lee, S. P. Sherlock, M. Terashima, H. Kosuge, Y. Suzuki, A. Goodwin, J. Robinson, W. S. Seo, Z. Liu, R. Luong, M. V. McConnell, D. G. Nishimura, H. Dai, Magn. Reson. Med. 2009, 62, 1497–1509. S. Mornet, S. Vasseur, F. Grasset, E. Duguet, J. Mater. Chem. 2004, 14, 2161–2175. J. J. Jung, A. Solanki, K. A. Memoli, K. Kamei, H. Kim, M. A. Drahl, L. J. Williams, H. R. Tseng, K. Lee, Angew. Chem. Int. Ed. 2010, 49, 103–107. A. Elbakry, A. Zaky, R. Liebkl, R. Rachel, A. Goepferich, M. Breunig, Nano Lett. 2009, 9, 2059–2064. H. K. Gan, A. H. Kaye, R. B. Luwor, J. Clinical Neurosci. 2009, 16, 748–754. Y. Ning, R. Zeineldin, Y. Y. Liu, M. Rosenberg, M. S. Stack, L. G. Hudson, Canc. Res. 2005, 65, 9280–9286. © 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Received: January 4, 2011 Revised: March 1, 2011 Published online: May 11, 2011 small 2011, 7, No. 12, 1647–1652 www.advmat.de COMMUNICATION www.MaterialsViews.com Generation of a Library of Non-Toxic Quantum Dots for Cellular Imaging and siRNA Delivery Prasad Subramaniam, Seung Jae Lee, Shreyas Shah, Sahishnu Patel, Valentin Starovoytov, and Ki-Bum Lee* Dissecting the spatio-temporal interaction of biomolecules inside cells at the subcellular level is an important facet of molecular cell biology and chemical biology. Over the last few decades, a variety of fluorescent molecular probes have been developed to investigate these complex bio-interactions for both in vitro and in vivo cellular imaging and subcellular detection. In particular, semiconductor nanoparticles like quantum dots (QDs) have shown great potential as nanoparticle-based fluorescent probes due to their excellent physiochemical properties, which allow them to overcome the limitations of conventional fluorescent probes such as organic dyes and fluorescent proteins.[1] These unique attributes of QDs have proven to be crucial in elucidating the intricate interactions through which small molecules and biomolecules (e.g., proteins, peptides and nucleic acids) bind to their targets in specific signaling cascades.[2] While QDs are excellent molecular probes, they also can be used as effective delivery vehicles for several therapeutic biomolecules. For example, the use of QDs for the simultaneous imaging and delivery of small interfering RNA (siRNA) for selectively knocking-down target oncogenes in tumor cells has been successfully demonstrated.[2c,3] However, the major limiting factor in harnessing the maximum potential of QDs as multifunctional imaging probes and delivery systems is their inherent cytotoxicity; most of the wellestablished QDs are composed of highly toxic elements, such as cadmium (Cd), selenium (Se) or tellurium (Te).[2b,4] Owing P. Subramaniam, S. J. Lee, S. Shah, S. Patel, Prof. K.-B. Lee Department of Chemistry and Chemical Biology Institute for Advanced Materials Devices and Nanotechnology (IAMDN), Rutgers The State University of New Jersey Piscataway, NJ 08854, USA E-mail: [email protected]; http://rutchem.rutgers.edu/∼kbleeweb/ S. J. Lee Center for Nano-Bio Fusion Research Korea Research Institute of Chemical Technology Daejeon 305-600, Korea V. Starovoytov Department of Cell Biology and Neuroscience, Rutgers The State University of New Jersey, Piscataway, NJ 08854, USA Prof. K.-B. Lee School of Medicine Kyung Hee University Seoul, South Korea DOI: 10.1002/adma.201201019 4014 wileyonlinelibrary.com to this obstacle, the wide applications of QDs are currently delayed and the main focus of QD imaging has been limited to the cell and small animal studies. In response to the above issues, the recent development of I-III-VI2 type QDs[5] like AgInS2,[5c] CuInS2[5b,5d] and ZnS-AgInS2[5e] offers better control of band-gap energies and demonstrates the great potential of these QDs as non-toxic molecular probes. For instance, several research groups have successfully synthesized ZnS-AgInS2 (ZAIS) QDs through the decomposition of single source precursors using thermal,[5e] hydrothermal,[6] photothermal[7] and microwave-assisted methods.[8] Yet, these conventional synthetic methodologies for preparing these I-III-VI2 type QDs have several shortcomings such as high reaction temperatures, poor control of growth rates, long reaction times, difficulty of high throughput synthesis, and the need for complicated synthetic procedures to prepare QDs with different emissions profiles, all of which would be critical in investigating the diversity and dynamic processes of multiple biomarkers in cancer and stem cells. Thus, in order to harness the full potential of QDs as biological imaging probes as well as drug delivery platforms for clinical and translational research, there is an urgent need to develop a simple and straightforward methodology that affords both the synthesis of non-toxic QDs and the versatility of generating a library of QDs with tunable properties favoring their use as imaging agents in biology. To address the aforementioned issues, we developed a novel sonochemical approach for the high throughput synthesis of a library of biocompatible ZnxS-AgyIn1-yS2 (ZAIS) quantum dots with tunable physical (photoluminescence, PL) properties, thereby allowing them to be used as multifunctional nanoparticles for the simultaneous imaging and effective delivery of siRNA to brain tumor cells with negligible cytotoxicity. These ZAIS QDs also proved to be useful for imaging stem cells, which are otherwise quite sensitive towards nanomaterial-based molecular imaging probes (Figure 1). A sonochemical synthetic method uses ultrasound irradiation to drive the main synthetic reaction process. With the rapid growth of its use for applications in material science, the sonochemical synthetic approach is particularly attractive for the preparation of novel nanomaterials. Its advantages include a fast reaction rate (for e.g., its possible to generate a whole library of nanoparticles of several compositions in a span of few hours), controllable reaction conditions and the ability to form nanoparticles with uniform shapes, narrow size-distributions and high purity in relatively less time and at ambient conditions. Specifically, the synthetic methodology used for the preparation of our ZAIS QDs involved the sonochemical decomposition of the organometallic precursor (AgyIn1-y)Znx(S2CN(C2H5)2)4 at ambient conditions. Compared © 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2012, 24, 4014–4019 www.advmat.de www.MaterialsViews.com COMMUNICATION Figure 1. Synthesis of a library of ZnS-AgInS2 QDs for simultaneous imaging and delivery of siRNA. A) Synthetic procedure to obtain a library of dodecylamine-capped hydrophobic ZAIS QDs. B) Conversion of the the hydrophobic ZAIS QDs into water-soluble ones via ligand exchange with 3-mercaptopropionic acid (MPA) followed by attachment of siRNA using a layer-by-layer approach. C) Delivery of the siRNA against EGFP using ZAIS QD-siRNA conjugates into brain cancer cells overexpressing EGFP. QD) to 700 nm (Red QD) by just varying the mole ratios of zinc (Zn), silver (Ag), and indium (In) in the precursor solutions (Figure 2A). It is worthwhile to note that as compared to conventional CdSe or CdTe QDs, which show size-dependent fluorescence properties, our composition-dependent tunable fluorescence of the ZAIS QDs presents a unique advantage for obtaining QDs with emission in the near-UV range (blue or blue-green emission). This observation can be explained by the fact that a blue-colored CdSe QD needs to be less than 2.0 nm in size, which is practically impossible to obtain using traditional thermal decomposition techniques. After assessing the optimum conditions for obtaining the desired emission profiles, we observed that for a given concentration of precursor ions, ultrasonication for 5 minutes at 20 kHz and 200W output power, gave the highest emission intensity (data not shown). Figure 2B depicts a 3D heat map which summarizes the physical and chemical properties of the partial ZAIS QD library. For the different combinations of the precursor elements (i.e. Zn, Ag, and In), the 3D heat map shows the Figure 2. Photoluminescent properties of the ZAIS QD library. A) Representative fluorescent emission wavelength and the corresponding image of the entire library of ZAIS QDs with varying compositions (ZnxS-AgyIn1-yS2) synthesized photoluminescence (PL) peak intensities of via the sonochemical approach. B) Heat map depicting the PL intensity (z axis) vs. Zn and Ag/In concentrations (x and y axis) for select compositions of the ZAIS nanoparticle library. selected ZAIS QDs under UV irradiation (λex = 365 nm). This data analysis not only Column color indicates the maximum emission wavelength. to the traditional synthetic methods for preparing QDs, this approach does not require high-temperature and high-pressure conditions (See Experimental Section for the detailed synthetic information). The resulting ZAIS QDs exhibited intense emission at room temperature, regardless of the size of the particles (Figure 2). More interestingly, the energy gap of the ZAIS QDs and their emission wavelength can be controlled by varying the concentration of each precursor element. For instance, the emission of the ZAIS QDs could be tuned from 480 nm (Blue Adv. Mater. 2012, 24, 4014–4019 © 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim wileyonlinelibrary.com 4015 www.advmat.de COMMUNICATION www.MaterialsViews.com 4016 provides the key characteristic properties for the various compositions, but it can also facilitate the selection of ZAIS QDs with appropriate physicochemical properties for further studies and applications. For example, examining the 3D heat map, it is evident that ZAIS QDs without zinc (i.e. AgyIn1-yS2) had the longest emission wavelengths (Figure 2B). We also observed that the PL peak wavelength was blue-shifted on increasing the Zn concentration. Furthermore, QDs obtained with 0.5 mol Ag and 0.5 mol In, had the longest emission wavelength (λem = 697nm), with an observed blue-shift on increasing or decreasing the Ag concentration. In conclusion, the emission profiles of the ZAIS QDs could be easily tuned by varying the Zn or Ag (In = 1-Ag) concentrations and hence the reported method used to synthesize our QD library can allow scientists to investigate the composition-dependent physicochemical properties of QDs of interest. (See Supporting Information, Figure S1 for the detailed information about absorption and PL spectra). Comprehensive characterization of the resulting ZAIS QDs was performed using several different methods. The elemental composition of one representative ZAIS QD (Zn0S-Ag0.2In0.8S2) was analyzed using X-Ray fluorescence spectroscopy (XRF). It was observed that the relative atomic ratios of the constituent elements was consistent with the calculated mole ratios (Figure 3 A), thus confirming the efficiency of the sonochemical synthetic methodology in obtaining QDs with desired chemical compositions. We also analyzed the crystal structures of several compositions of the ZAIS QDs using powder X-ray diffraction (XRD). The particles prepared with varying mole ratios of Zn, Ag and In exhibited three broad peaks which lie in between the diffraction patterns of the tetragonal AgInS2 and bulk cubic ZnS crystals (Figure 3B). Furthermore, there was a clear peak shift towards a higher angle with an increase in the amount of Zn2+, thus indicating that the QDs obtained were a not a just mixture of bulk ZnS and AgInS2 but a ZnS-AgInS2 solid solution.[5e] Transmission electron microscopy (TEM) analysis clearly revealed the spherical shape and monodispersity of our ZAIS QDs (Figure 3C). The average size of the QDs as determined by TEM was found to be 12 ± 1.3 nm. This was further confirmed by measuring the hydrodynamic size and the polydispersity index (PDI) of the ZAIS QDs using dynamic light scattering (Figure S3). Additionally, the size of the QDs did not significantly change when their composition was altered (data not shown). We also determined the PL quantum yields of the ZAIS QDs and compared them to the quantum yield of the CdSe/ZnS QDs (∼0.4). It was found that the relative quantum yields of the ZAIS QDs depended on their composition and were comparable or in some cases higher than that of the CdSe/ZnS QDs (Figure S4). Finally, in terms of stability, the physical properties of the ZAIS QDs such as absorption and photoluminescence remained unchanged for two months when stored at ambient conditions (see Figure S2). Hence, these ZAIS QDs could be potentially used for long term cellular labeling without any loss of photoluminescence. The solubility and stability of QDs in an aqueous solution is essential for their wide application as molecular probes in molecular cell biology. The ZAIS QDs reported here were functionalized with 3-mercaptopropionic acid (MPA) to render them soluble in physiological conditions (See Experimental section for the details regarding the surface modification).[9] These water-soluble ZAIS QDs were found to be extremely stable at physiological conditions without any signs of aggregation, even when stored for several months (Figure S5). This was further confirmed by monitoring the PL intensity of one of the ZAIS QDs (x = 0, y = 0.2) in phosphate-buffered saline (PBS, pH = 7.4) at 37 ºC over a period of 6 days. The ZAIS QDs were found to be quite stable without any significant loss of photoluminescence over the test period (Figure S6). Another critical factor limiting the use of conventional QDs for varFigure 3. Physical characterization of the ZnxS-AgyIn(1-y)S2 QDs. A) X-ray Fluorescence analysis ious biological applications is their inherent [10] This could be partially of one of the ZAIS QD (x = 0, y = 0.2) showing the elemental composition and relative atomic cellular toxicity. mole ratios of the constituent elements. The composition agrees to that calculated theoreti- attributed to the lack of suitable methods to cally. B) X-ray diffraction patterns of the ZAIS QDs prepared using sonochemistry. The values of generate QDs with different compositions in x and y are indicated in the figure. Reference patterns of bulk ZnS and AgInS2 are also shown. a high throughput manner for subsequent Broad peak width is attributed to the amorphous nature of the ZAIS QDs. (C) TEM images of toxicological screening. Hence, the generaone of the ZAIS QDs (x = 0, y = 0.2) clearly shows the monodispersity and narrow size distrition of a QD library and assaying their cytobution (n = 100). Inset shows the high resolution image of a single nanoparticle (Scale bar = 10 nm). (D) Photoluminescence spectra of select water-soluble ZAIS QDs (a: x = 0.6, y = 0.5.; toxicity in a simple and quick way can facilitate the selection of appropriate QDs eliciting b: x = 0.3, y = 0.4; c: x = 0, y = 0.2; d: x = 0, y = 0.5) wileyonlinelibrary.com © 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2012, 24, 4014–4019 www.advmat.de www.MaterialsViews.com COMMUNICATION Figure 4. Biocompatibility and cellular imaging studies of ZAIS-QDs in mammalian cells. A) Comparison of the cellular cytotoxicity of the water soluble ZAIS QD (Zn0S-Ag0.2In0.8S2) and CdSe/ZnS QDs at different concentrations in U87 cells (A1) and hMSCs (A2). The results are presented as means ± SD from three independent experiments. Student’s unpaired t-test was used for evaluating the statistical significance of the cytotoxicity of ZAIS QDs (* = P < 0.001, ** = P < 0.01, *** = P < 0.05) as compared to the CdSe/ZnS QDs. B) Transmission electron microscopy of the ZAIS QD in hMSCs. The image clearly shows the presence of the QDs (marked by blue arrows) in the cytoplasm and the nucleus (marked in red). The inset depicting the magnified image of the QD cluster, confirms the monodisperse nature of the QDs inside the cell. C,D) Fluorescence microscopy imaging demonstrating the uptake of the water soluble ZAIS-QD (λem = 606 nm) in U87 cells (C) and hMSCs (D). minimum toxic effects and thus enabling the long term cellular imaging of cancer and stem cells in vitro and in vivo. To test the biocompatibility of our ZAIS QDs for use as imaging probes in vitro, a cytotoxicity assay was carried out in both cancer and stem cells (Figure 4A). The concentration-dependent cytotoxicity of the water-soluble ZAIS QDs was assessed in human bone marrow-derived mesenchymal stem cells (hMSCs) and human brain tumor cells (U87 glioblastoma cell line) for two days using a cell proliferation assay (MTS). The toxicity of the ZAIS QDs was also compared to the water-soluble CdSe/ZnS QDs (control sample). The ZAIS QDs showed significantly improved biocompatibility (less cytotoxic, 95% cell viability) in both U87 cells and hMSCs even at high concentrations (upto 100 μg/mL) in comparison to the control sample (CdSe/ZnS QDs), which were found to noticeably cytotoxic at even low concentrations. The cytotoxicity of other compositions of the ZAIS QDs is presented in the supporting information (Figures S7 and S8). In addition, we carried out the cytotoxicity assay of the ZAIS QDs in normal healthy cells (NIH-3T3 mouse fibroblasts). The ZAIS QDs were found to be extremely biocompatible as compared to the CdSe/ZnS QDs (Figure S9). The prolonged exposure of nanoparticles, especially QDs, to an oxidative environment (aerial oxidation or UV-induced oxidation) has been known to catalyze their decomposition, thereby leading to the leaching of the metal ions, which is known to induce cytotoxic effects in cells. In order to test the cytotoxicity of our ZAIS QDs under an oxidative environment, we subjected the ZAIS QDs to high-intensity UV light to catalyze the oxidation Adv. Mater. 2012, 24, 4014–4019 process. The water-soluble ZAIS QD solutions were exposed to a UV-light source (λem = 365 nm, power density of 12 mW/cm2) for 1–4 hours after which they were incubated with U87 cells. It was found that even after 4h of UV-induced photo-oxidation, the ZAIS QDs were not toxic to the cells (>85% viability, Figure S10) which is in stark contrast to the photo-toxicity of CdSe/ZnS QDs reported previously. For our cellular imaging experiment, Figure 4B shows the internalization of the water-soluble ZAIS QDs (with 606 nm emission) in human MSCs. The QDs were primarily found to be present in the cytoplasm of the hMSCs with a few QDs being found in the nucleus of the cells. Furthermore, fluorescent microscopy images of U87 cells (Figure 4C) and hMSCs (Figure 4D) incubated with the water soluble ZAIS-QDs proved that the QDs showed strong fluorescence even when inside the cells. Collectively, our ZAIS-QDs showed excellent biocompatibility (non-toxic) and these results can be extended to their wide use in cellular imaging and delivery applications, especially for stem cells, which are known to be extremely sensitive towards nanomaterials. Having demonstrated the potential of our ZAIS QDs as nontoxic fluorescent imaging probes, we evaluated their ability to be used as vehicles for the efficient delivery and tracking of siRNA in vitro. In the past decade, there has been a considerable interest in the development of nanomaterial-based siRNA and gene delivery methods for controlling cell fate and behavior.[11] RNA interference (RNAi) involves the use of small interfering RNAs (siRNAs) to selectively mediate the cleavage of complementary mRNA sequences and thus regulate target © 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim wileyonlinelibrary.com 4017 www.advmat.de COMMUNICATION www.MaterialsViews.com Figure 5. In vitro testing of the ZAIS-QD-siEGFP cell uptake and silencing efficiency in stably transfected U87-EGFP glioblastoma cells. A) Control U87EGFP cells with PEI-coated ZAIS-QD; A1) represents the phase contrast image, and A2) is the corresponding fluorescence image. B) EGFP knockdown using the ZAIS QD-siRNA constructs; B1) Phase contrast image showing the the viability of U87-EGFP cells has not changed appreciably after the transfection of the ZAIS QD-siRNA constructs as compared to the control cells in (A). B2) Fluorescence image clearly shows the knockdown of EGFP in cells which have internalized the siRNA-QDs (red) after 72 h. The red fluorescence from the ZAIS QDs correlates well with the loss of the green fluorescence in cells (indicated by yellow arrows). Scale bar is 50 μm. gene expression. In combination with other modalities like small molecules and peptides, RNAi could prove to be a powerful tool to manipulate the cellular microenvironments.[11b] In this context, the use of QDs as a multimodal delivery vehicle becomes a promising choice, thereby allowing for the efficient delivery of siRNA and real-time tracking of the siRNA-mediated gene knockdown.[2c] As a proof-of-concept experiment, we used a brain tumor cell line (U87) which was genetically labeled to express the green fluorescent protein (GFP). In order to deliver the siRNA to target brain tumor cells expressing EGFP (U87-EGFP), the MPA-coated ZAIS QDs were conjugated to the siRNA (against the EGFP gene) using polyethyleneimine (PEI) as a cationic polymer via a layer-by-layer approach (Figure 1B).[12] The efficiency of the PEI coating and siRNA conjugation was monitored using zeta potential (Figure S11). The decrease in the green fluorescence due to the siRNA-mediated knockdown of the EGFP gene using our ZAIS QD-siRNA constructs was then monitored (Figure 5) in order to to assess the transfection efficiency and RNA interference (RNAi) activity. It was found that the ZAIS QD-siRNA constructs were efficiently taken up by cells as evident by the intracellular red fluorescence of the QDs. In addition, the QD localization correlated well with the decrease of the green fluorescence (∼80% after 3 days of transfection) resulting from the siRNA-mediated knockdown of the EGFP mRNA. These results show the ability of our ZAIS QDs to efficiently translocate siRNA into the cells and achieve gene knockdown in vitro. With further surface modifications 4018 wileyonlinelibrary.com using targeting ligands (RGD or TAT peptides and antibodies) and bifunctional linkers (cleavable linkers or covalent linkers), the ZAIS QDs could be used for the targeted delivery of siRNA to cancer/stem cells and for the real-time monitoring of its delivery in an efficient manner. In addition, the presence of a library of multicolored QDs obtained via our sonochemical approach, would allow the for multiplexed imaging of transplanted cell populations in vivo (i.e., tracking different cell populations with different QDs using different emission wavelengths at the same time). In summary, we successfully demonstrated the preparation of ZnxS-AgyIn1-yS2 (ZAIS) QDs using a facile sonochemical synthetic method. The physicochemical and bio-relevant properties of the resulting QDs can be easily tuned over the entire visible spectrum by varying the chemical composition of the precursors. We also demonstrated that our ZAIS QDs can exhibit excellent biocompatibility for the efficient delivery of siRNA and simultaneous imaging/tracking of the same in cancer cells (and stem cells) with negligible QD-induced cytotoxicity. While the ZAIS QDs show great potential for the imaging and delivery of siRNA in vitro, a more thorough investigation of their long-term cytotoxicity is needed before they can be used in vivo. Efforts in this direction are underway. Overall, the ease of the synthesis of the ZAIS QDs, their excellent cyto-compatibility, and their versatility as multiplexed imaging agents provides an attractive alternative over conventional QD-based molecular imaging probes and siRNA delivery vehicles. The above methodology © 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Adv. Mater. 2012, 24, 4014–4019 www.advmat.de www.MaterialsViews.com and the IAMDN (Rutgers) for allowing us to use their high-resolution TEM facility. We are also grateful to Drs. Joung Kyu Park and Jeong Je Cho for their scientific input and the K.-B.L. group members for their valuable suggestions for the manuscript. K.-B.L. acknowledges the NIH Director’s Innovator Award (1DP20D006462-01) and is also grateful to the N.J. Commission on Spinal Cord grant (09-3085-SCR-E-0). This study is also funded in part by the US EPA (Grant# 83469302) and the UK NERC (Grant# NE/H012893). Experimental Section Synthesis of Precursor Complexes: An aqueous solution of sodium diethyldithiocarbamate (0.05 M, 5.0 mL) was mixed with an aqueous solution containing appropriate amounts of AgNO3, In(NO3)3·xH2O and Zn(NO3)2·6H2O in order to get the required mole ratios (total concentration of the metal ions was 0.025 M). The solution was allowed to stir for 5 minutes after which it was filtered using a buchhner funnel, washed several times with distilled water and MeOH and finally dried in a convection oven at 60 deg C overnight to obtain the precursor as a dried powder. Several precursor powders were synthesized by varying the mole ratios of the metal salts. Sonochemical Synthesis of Dodecylamine-Capped ZnS-AgInS2 Quantum Dots: The precursor complex (0.1 g) and dodecylamine (10.0 mL) were put into a 20 mL vial and sonicated using a tip probe-based high frequency sonicator (Branson) for 5 minutes in an air-atmosphere. The resulting suspension was allowed to sit at room temperature for two minutes after which 5.0 mL of chloroform (5.0 mL) and MeOH (5.0 mL) were added to it and centrifuged at 4000 rpm. The supernatant containing the ZAIS QDs was collected and equal amount of MeOH was added to it in order to isolate the nanoparticles. The obtained ZAIS QDs were then resuspended in chloroform for absorbance and photoluminescence measurements. Surface Modification of ZAIS-QDs with 3-Mercaptopropionic Acid (MPA): The dodecylamine-capped ZAIS QDs were subjected to a ligand exchange reaction using 3-mercaptopropionic acid (MPA) according to a previously reported protocol.[9] Briefly, a 3.0 mL ethanolic solution of MPA (0.2 M) and KOH (0.3 M) was added dropwise to an equal amount of the dodecylamine-capped ZAIS QD solution in chloroform. The turbid solution was stirred for 3h at room temperature followed by centrifugation at 4000 rpm. The wet precipitate of the MPA-coated ZAIS QDs was washed with EtOH and redissolved in phosphate-buffered saline (PBS). The water soluble MPA-coated ZAIS QDs were stable in buffer solution with no significant change in absorption and photoluminescence for upto 2 months, when stored at ambient conditions. Culture of Human U87 Glioblastoma Cells, NIH-3T3 Mouse Fibroblasts and Human Mesenchymal Stem Cells: The EGFRvIII overexpressed U87 glioblastoma cells (U87) and human mesenchymal stem cells (hMSCs) were cultured using previously reported methods. For U87-EGFRvIII cells, DMEM with high glucose, 10% fetal bovine serum (FBS, Gemini Bioproducts), 1% Streptomycin-penicillin and 1% Glutamax (Invitrogen, Carlsbad, CA) were used as basic components of growth media including Hygromycin (30 μg/mL, Invitrogen) as a selection marker. Human bone marrow-derived MSCs (Lonza, Walkerville) were cultured in the conditioned media (Lonza, Walkerville) according to manufacturer’s recommendations. For the NIH-3T3 mouse fibroblasts, DMEM (with high glucose) supplemented with 10% fetal calf serum (FCS, Gemini Bioproducts), 1% Streptomycin-penicillin and 1% Glutamax was used as the growth medium. All cells were maintained at 37 oC in humidified 5% CO2 atmosphere. Supporting Information Supporting Information is available from the Wiley Online Library or from the author. Acknowledgements We thank Prof. Gene Hall for help with the XRF measurements, Dr. Tom Emge, and Vaishali Thakral for their support with the XRD Adv. Mater. 2012, 24, 4014–4019 Received: March 10, 2012 Revised: April 5, 2012 Published online: June 29, 2012 COMMUNICATION could be potentially extended to synthesize libraries of various types of nanoparticles (magnetic nanoparticles and upconverting near-IR fluorescent nanoparticles), thereby allowing for rapid screening of the nanomaterials for biomedical applications such as drug delivery and cellular labeling. [1] a) T. Pons, H. Mattoussi, Ann. Biomed. Eng. 2009, 37, 1934– 1959; b) A. Solanki, J. D. Kim, K.-B. Lee, Nanotechnology 2008, 3, 567–578. [2] a) J. B. Delehanty, H. Mattoussi, I. L. Medintz, Anal. and Bioanal. Chem. 2009, 393, 1091–1105; b) A. M. Derfus, W. C. W. Chan, S. N. Bhatia, Nano Lett. 2003, 4, 11–18; c) J. Jung, A. Solanki, K. A. Memoli, K.-I. Kamei, H. Kim, M. A. Drahl, L. J. Williams, H.-R. Tseng, K.-B. Lee, Angew. Chem. Int. Ed. 2010, 49, 103–107; d) X. Wu, H. Liu, J. Liu, K. N. Haley, J. A. Treadway, J. P. Larson, N. Ge, F. Peale, M. P. Bruchez, Nat. Biotech. 2003, 21, 41–46. [3] a) A. M. Derfus, A. A. Chen, D.-H. Min, E. Ruoslahti, S. N. Bhatia, Bioconjugate Chem. 2007, 18, 1391–1396; b) A. M. Smith, H. Duan, A. M. Mohs, S. Nie, Adv. Drug Del. Rev. 2008, 60, 1226–1240; c) M. V. Yezhelyev, L. Qi, R. M. O’Regan, S. Nie, X. Gao, J. Am. Chem. Soc. 2008, 130, 9006–9012. [4] Y. Su, Y. He, H. Lu, L. Sai, Q. Li, W. Li, L. Wang, P. Shen, Q. Huang, C. Fan, Biomat. 2009, 30, 19–25. [5] a) P. M. Allen, M. G. Bawendi, J. Am. Chem. Soc. 2008, 130, 9240–9241; b) E. Cassette, T. Pons, C. Bouet, M. Helle, L. Bezdetnaya, F. Marchal, B. Dubertret, Chem. Mat. 2010, 22, 61176124; c) Y. Hamanaka, T. Ogawa, M. Tsuzuki, T. Kuzuya, J. Phys. Chem. C 2011, 115, 1786–1792; d) X. Tang, W. Cheng, E. S. G. Choo, J. Xue, Chem. Comm. 2011, 47, 5217–5219; e) T. Torimoto, T. Adachi, K.-I. Okazaki, M. Sakuraoka, T. Shibayama, B. Ohtani, A. Kudo, S. Kuwabata, J. Am. Chem. Soc. 2007, 129, 12388–12389; f) R. Xie, M. Rutherford, X. Peng, J. Am. Chem. Soc. 2009, 131, 5691–5697. [6] H. Zhong, Y. Zhou, M. Ye, Y. He, J. Ye, C. He, C. Yang, Y. Li, Chem. Mat. 2008, 20, 6434–6443. [7] J. J. Nairn, P. J. Shapiro, B. Twamley, T. Pounds, R. von Wandruszka, T. R. Fletcher, M. Williams, C. Wang, M. G. Norton, Nano Lett. 2006, 6, 1218–1223. [8] J. Gardner, E. Shrudha, L. Lau, C. Wang, R. Rodriguez, J. Pak, J. Nanopart.Res. 2008, 10, 633–641. [9] a) C.-C. Chen, C.-P. Yet, H.-N. Wang, C.-Y. Chao, Langmuir 1999, 15, 68456850. b) G. P. Mitchell, C. A. Mirkin, R. L. Letsinger, J. Am. Chem. Soc. 1999, 121, 8122–8123. [10] a) P. Mitchell, Nat. Biotech. 2001, 19, 1013–1017; b) X. Gao, L. Yang, J. A. Petros, F. F. Marshall, J. W. Simons, S. Nie, Curr.Opin. Biotechnol. 2005, 16, 63–72; c) W. Liu, M. Howarth, A. B. Greytak, Y. Zheng, D. G. Nocera, A. Y. Ting, M. G. Bawendi, J. Am. Chem. Soc. 2008, 130, 1274–1284; d) B. S. Shah, P. A. Clark, E. K. Moioli, M. A. Stroscio, J. J. Mao, Nano Lett. 2007, 7, 3071–3079. [11] a) J. K. Park, J. Jung, P. Subramaniam, B. P. Shah, C. Kim, J. K. Lee, J.-H. Cho, C. Lee, K.-B. Lee, Small 2011, 7, 1647–1652; b) B. P. Shah, C. Kim, P. Subramaniam, K.-B. Lee, Mol. Pharmaceutics 2011, 8, 1955–1961; c) H. Wang, K.-J. Chen, S. Wang, M. Ohashi, K.-I. Kamei, J. Sun, J. H. Ha, K. Liu, H.-R.Tseng, Chem. Comm. 2010, 46, 1851–1853; d) X. Xue, F. Wang, X. Liu, J. Mater. Chem. 2011, 21, 13107–13127. [12] A. Elbakry, A. Zaky, R. Liebl, R. Rachel, A. Goepferich, M. Breunig, Nano Lett. 2009, 9, 2059–2064. © 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim wileyonlinelibrary.com 4019