Gene Expression Profiling Uncovers Molecular Classifiers for the

Transcription

Gene Expression Profiling Uncovers Molecular Classifiers for the
VOLUME
28
䡠
NUMBER
9
䡠
MARCH
20
2010
JOURNAL OF CLINICAL ONCOLOGY
From the Department of Pathology, Center
for Experimental Research and Medical
Studies, Laboratory of Functional Genomics, and Institute for Cancer Research and
Treatment, University of Torino, Torino;
Department of Medical Sciences, Leukemia Study Center, University of Milan,
Hematology 1, Fondazione Istituto di
Ricovero e Cura a Carattere Scientifico Policlinico and Pathology & Lymphoid Malignancies Units, San Raffaele Scientific
Institute, Milan; Department of Pathology,
University of Verona, Verona; Department
of Surgical Pathology, University of Brescia,
Brescia; Institute of Hematology and Medical Oncology L. and A. Seràgnoli, S. OrsolaMalpighi Hospital, University of Bologna,
Bologna; Department of Pathology,
National Cancer Institute, Naples, Italy;
Department of Pathology and New York
University Cancer Center, New York
University School of Medicine, New York,
NY; Institute of Pathology, University of
Wuerzburg, Wuerzburg, Germany; and
Department of Pathology, University of
Leuven, Leuven, Belgium.
Submitted November 11, 2008; accepted
October 1, 2009; published online ahead of
print at www.jco.org on February 16, 2010.
Supported by Associazione Italiana per la
Ricerca sul Cancro; Fondazione Guido
Berlucchi; Ministero dell’Università e
Ricerca Scientifica; Regione Piemonte;
Compagnia di San Paolo, Torino (Progetto
Oncologia); Sixth Research Framework
Program of the European Union, Project
RIGHT (Grant No. LSHB-CT-2004-005276),
and fellowships from Fondazione Italiana
Ricerca sul Cancro (L.A. and K.T.).
Both L.A. and E.P. contributed equally to
this work.
Authors’ disclosures of potential conflicts
of interest and author contributions are
found at the end of this article.
© 2010 by American Society of Clinical
Oncology
0732-183X/10/2809-1583/$20.00
DOI: 10.1200/JCO.2008.20.9759
R E P O R T
Gene Expression Profiling Uncovers Molecular Classifiers
for the Recognition of Anaplastic Large-Cell Lymphoma
Within Peripheral T-Cell Neoplasms
Roberto Piva, Luca Agnelli, Elisa Pellegrino, Katia Todoerti, Valentina Grosso, Ilaria Tamagno,
Alessandro Fornari, Barbara Martinoglio, Enzo Medico, Alberto Zamò, Fabio Facchetti, Maurilio Ponzoni,
Eva Geissinger, Andreas Rosenwald, Hans Konrad Müller-Hermelink, Christiane De Wolf-Peeters,
Pier Paolo Piccaluga, Stefano Pileri, Antonino Neri, and Giorgio Inghirami
A
B
S
T
R
A
C
T
Purpose
To unravel the regulatory network underlying nucleophosmin-anaplastic lymphoma kinase (NPMALK) –mediated lymphomagenesis of anaplastic large-cell lymphoma (ALCL) and to discover
diagnostic genomic classifiers for the recognition of patients with ALK-positive and ALK-negative
ALCL among T-cell non-Hodgkin’s lymphoma (T-NHL).
Patients and Methods
The transcriptome of NPM-ALK–positive ALCL cell lines was characterized by silencing the
expression of ALK or STAT3, a major effector of ALK oncogenic activity. Gene expression profiling
(GEP) was performed in a series of systemic primary T-NHL (n ⫽ 70), including a set of
ALK-positive and ALK-negative ALCL (n ⫽ 36). Genomic classifiers for ALK-positive and ALKnegative ALCL were generated by prediction analyses and validated by quantitative reversetranscriptase polymerase chain reaction and/or immunohistochemistry.
Results
In ALCL cell lines, two thirds of ALK-regulated genes were concordantly dependent on STAT3
expression. GEP of systemic primary T-NHL significantly clustered ALK-positive ALCL samples in
a separate subgroup, underscoring the relevance of in vitro ALK/STAT3 signatures. A set of
genomic classifiers for ALK-positive ALCL and for ALCL were identified by prediction analyses.
These gene clusters were instrumental for the distinction of ALK-negative ALCL from peripheral
T-cell lymphomas not otherwise specified (PTCLs-NOS) and angioimmunoblastic lymphomas.
Conclusion
We proved that experimentally controlled GEP in ALCL cell lines represents a powerful tool to
identify meaningful signaling networks for the recognition of systemic primary T-NHL. The
identification of a molecular signature specific for ALCL suggests that these T-NHLs may
represent a unique entity discernible from other PTCLs, and that a restricted number of genes can
be instrumental for clinical stratification and, possibly, therapy of T-NHL.
J Clin Oncol 28:1583-1590. © 2010 by American Society of Clinical Oncology
This article was written on behalf of the
European T-Cell Lymphoma Study
Group.
Corresponding authors: Giorgio Inghirami,
MD, and Roberto Piva, PhD, Department
of Pathology and CeRMS, University of
Torino, Via Santena 7, Torino 10126 Italy;
e-mail: [email protected];
[email protected].
O R I G I N A L
INTRODUCTION
Non-Hodgkin’s lymphoma (NHL) is a heterogeneous group of malignancies corresponding to the
neoplastic, clonal expansion of B or T lymphocytes
possibly transformed at different stages of differentiation and maturation. Immunophenotypic and
molecular genetic studies have demonstrated that
several pathogenetic events are acquired during the
development and progression of NHL and that molecular fingerprints allow more objective diagnoses
and/or precise tumor stratifications.1,2 Within the
T-cell lymphoproliferative disorders, the classifica-
tion of the World Health Organization (WHO)
has recognized several specified and unspecified
entities.3 The T-cell NHLs (T-NHLs) account for
approximately 10% to 15% of all lymphoid neoplasms.4,5 They include, among several entities,
peripheral T-cell lymphomas not otherwise specified (PTCLs-NOS), angioimmunoblastic lymphoma (AILT), and anaplastic large-cell lymphoma
(ALCL). ALCLs, which comprise approximately
12% of all T-NHLs, are a heterogeneous group
whose definition, origin, and relationship with other
T-NHLs have frequently raised considerable questions and often controversies.6,7 The discovery that
© 2010 by American Society of Clinical Oncology
Downloaded from ascopubs.org by 78.47.27.170 on November 18, 2016 from 078.047.027.170
Copyright © 2016 American Society of Clinical Oncology. All rights reserved.
1583
Piva et al
ALCLs strongly express CD308 and display recurrent chromosomal
translocations involving the anaplastic lymphoma kinase (ALK)
gene9,10 led to a better recognition of these tumors and to their distinction
in two different subsets, according to ALK expression. Depending on the
pathologic criteria and the median patient age, ALK-positive ALCLs correspond to 50% to 85% of all nodal ALCLs. Gene expression profiling
(GEP) and comparative genomic hybridization (CGH) studies, although
performed in a limited number of cases, have shown that ALK-positive
andALK-negativeALCLshaverestrictedgenomicsignaturesand/orpreferential genomic aberrations.11-15 These findings, in association with
the unique epidemiologic and clinical features of ALK-positive
ALCL,16,17 led to the consideration of ALK-positive and ALKnegative systemic ALCLs as separate entities by the expert panel of
the WHO.3 Although ALK-positive ALCL can be readily diagnosed,
the distinction of ALK-negative ALCL from PTCL-NOS can be in
some instances excessively subjective. In fact, immunophenotypic or
genetic features to precisely define these T-NHLs are missing, suggesting that ALK-negative ALCLs may represent a morphologic variant
within the otherwise heterogeneous category of PTCL-NOS.18,19
Here, we undertook a systematic approach to profile the expression signatures of ALK-positive ALCL cell lines and of primary
T-NHL, including a subset of ALCL samples. This approach defined
small gene-cluster classifiers, capable of distinguishing ALK-positive
and ALK-negative ALCL from other PTCLs-NOS, and strengthened
the hypothesis that ALCLs correspond to a distinctive pathologic
subgroup within T-NHL.
PATIENTS AND METHODS
Patients and Case Selection
Cryopreserved samples of 16 PCTLs-NOS and 48 ALCLs (22 ALKpositive and 26 ALK-negative ALCLs) were provided from the Universities of
Leuven, Wuerzburg, Torino, Bologna, Verona, Brescia, and Napoli and by the
San Raffaele Scientific Institute of Milan. T-NHLs were selected on the basis of
stringent criteria: (1) lymph node biopsy site; (2) presence of ⱖ 50% neoplastic
cells; (3) RNA preservation; (4) high CD30 expression, T-cell associated/
restricted markers, GranzymeB, and TIA-1 positivity, and PAX-5 negativity
(for ALCL cases). ALCL samples (16 ALK-positive and 20 ALK-negative
ALCLs) displaying the best RNA quality were selected for GEP; 16 PTCL-NOS
and 12 ALCL samples were used for quantitative reverse-transcriptase polymerase chain reaction (Q-RT-PCR) validation. All samples were obtained at
the time of diagnosis, before treatment. ALCL cases were submitted to central
pathologic review by a panel of four expert hematopathologists (S.P.,
C.D.W.P., H.K.M.H., and G.I.). Final diagnoses were assigned according to the
criteria of the WHO classification. Unclassifiable cases (n ⫽ 2) were excluded
from the study. PTCL-NOS, AILT samples, and normal purified T-cell preparationswerepreviouslydescribed.20 Representativeformalin-fixedtumorcoreswere
processed to tissue microarrays for immunohistochemical analyses.21 Informed
consent was obtained from all enrolled patients following the procedures
approved by the local ethical committees of each participating Institution.
GEP
Total RNA was extracted using the TRIZOL reagent (Invitrogen, Carlsbad, CA) and purified using the RNeasy total RNA Isolation Kit (Qiagen, Santa
Clarita, CA). For the experiments using ALCL cell lines, the hybridization was
performed on HumanWG-6 BeadChips v2.0 (Illumina, San Diego, CA), using
biologic triplicates for each condition. cDNA and biotinylated cRNA were
generated by Illumina TotalPrep RNA Amplification Kit (Ambion, Austin,
TX). Data were processed with the Illumina Beadstudio software using the
following thresholds for significant detection: P value less than .001, detection
more than 0.99, and fold change more than 1.5. Gene expression data were
1584
clustered and visualized with the Gene Expression Data Analysis Suite
(GEDAS) software (http://gedas.bizhat.com/gedas.htm).
ALCL samples were hybridized on HG-U133 Plus 2.0 arrays (Affymetrix,
Santa Clara, CA), in accordance with previously published data.14,20 Hierarchical
agglomerative clustering and dendrogram were generated as described.23,24 Fisher’s exact test was used to evaluate cluster significance. The Significant Analysis of
Microarrays (SAM) software v3.02 was used for supervised analyses (http://www.
stat.stanford.edu/⬃tibs/SAM/).25 The search of classifier genes and the validation
of the identified signature on an independent data set were performed by prediction analysis of microarrays (PAM), setting the optimal value of ⌬ to obtain the
minimum cross-validation error using a leave-one-out cross-validation process.26
The selected probe lists were visualized by DNA-Chip Analyzer software.
shRNA Sequences, Generation of Inducible Cell Lines, and
Lentiviral Preparations
Expression plasmid for inducible STAT3 silencing was produced by
subcloning into pLVTHM vector,27 an shRNA sequence directed to human
STAT3.28 Additional human STAT3 shRNAs were purchased from the The RNA
Consortium library (Open Biosystems, Huntsville, AL).29 Self-inactivating lentiviral particles and inducible cell lines were produced as described.30,31
Q-RT-PCR
cDNA was transcribed using SuperSCRIPT III following the manufacturer’s instructions (Invitrogen). Semiquantitative PCR reactions were carried
out in triplicate for 25, 30, and 35 cycles. Q-RT-PCR was performed in
triplicate on ABI PRISM 7900HT thermal cycler (Applied Biosystems, Foster
City, CA) with SYBR green dye. The results were expressed using the comparative Ct method, according to the manufacturer’s manual. The predictive
power of the investigated genes was tested using the conventional procedure of
Quadratic Discriminant Analysis.32 The oligonucleotide primer pairs and
polymerase chain reaction conditions are available on request.
Cell Culture
Human ALK-positive (TS [a subclone of Sup-M2],30 Karpas 299, SuDHL-1, and JB6), and ALK-negative (Mac-1) ALCL cells were grown in
RPMI-1640 medium supplemented with 10% fetal calf serum (Lonza, Rockland, ME). Cell cycle and apoptosis analyses were performed by flow cytometry.30
Antibodies and Western Blotting
The following primary antibodies were used for Western blotting: mouse
anti-ALK and anti-STAT3 from Zymed (Carlsbad, CA), mouse anti–␣tubulin from Sigma-Aldrich (St Louis, MO), rabbit antiphospho STAT3-Y705
and STAT5-Y694 from Cell Signaling Technology (Danvers, MA), rabbit
anti-Survivin and antiphospho eIF2␣ from Oncogene Research (San Diego,
CA), and rabbit anti-GFP from Molecular Probes (Carlsbad, CA).30
Immunohistochemistry
Immunohistochemical stains were performed on formalin-fixed,
paraffin-embedded tissue microarrays of ALCL and PTCL-NOS samples.
Sections were incubated with antibodies anti-ALK (Zymed), CD30
(DAKO, Fremont, CA), phospho-STAT3-Y705 (Cell Signaling Technology), C/EBP␤ (Santa Cruz Biotechnology, Santa Cruz, CA), GAS1 (provided by Dr. Schneider, Trieste, Italy), and NFATC2 (Sigma-Aldrich).
Bound complexes were revealed on a semiautomated immunostainer.30
RESULTS
Expression Signature of ALK-Positive ALCL Cells
Depends on STAT3 Activity
The transcription factor STAT3 is a major substrate of ALK
chimera in human ALCL and is required for growth and survival of
nucleophosmin (NPM) -ALK–transformed cells.10,33,34 To determine
whether STAT3 is essential in mediating NPM-ALK–regulated genes,
we generated ALCL cell lines expressing a doxycycline-inducible
shRNA to knockdown (KD) STAT3 expression. Inducible loss of STAT3
led to G1 cell cycle arrest, followed by cell death. These findings were
© 2010 by American Society of Clinical Oncology
Downloaded from ascopubs.org by 78.47.27.170 on November 18, 2016 from 078.047.027.170
Copyright © 2016 American Society of Clinical Oncology. All rights reserved.
JOURNAL OF CLINICAL ONCOLOGY
Molecular Classifiers for ALCL
B
A5
895
TS -TTA
dox
1,116
shSTAT3
TS
- + + - +
NI
L
S3
AS
S3
S
39
B
39
D
C
337
genes
162
S3
S
shALK
A5
M
DO
X9
6h
S3
S+
39
D
39
B
S3
S
DO
X
S3
AS
S3
S–
Co
n
tro
l
96
h
A
STAT3
ALK
TNFRSF8
SCCA2
GAS1
ICOS
IL2RA
548
Fig 1. (A) Identification of a STAT3 signature in anaplastic lymphoma kinase
(ALK) –positive cells. Supervised analysis
was performed in untreated Sup-M2-TS
cells (control; n ⫽ 6), transduced with
mock shRNA (S3AS; n ⫽ 3), three different STAT3 shRNA sequences (39B, 39D,
S3S; n ⫽ 9), or inducible STAT3 shRNA
(TS-TTA) in the absence of doxycycline
(S3S – DOX; n ⫽ 3) or presence of doxycycline (S3S ⫹ DOX; n ⫽ 3). The color
scale bar represents relative gene expression changes, normalized by the standard
deviation. (B) A restricted STAT3 signature
in ALK-positive cells identified by overlapping gene expression profiling analyses of
STAT3 and ALK knock-down (KD) in SupM2-TS cells. (C) Validation of STAT3 signature
after ALK (A5) or STAT3 KD (S3S, 39B, 39D)
by a reverse transcriptase polymerase chain
reaction approach. A5M and S3AS correspond to mock shRNA sequences.
HSD17B7
SGK
RGS16
β-2M
-1.0 0 +1.0
confirmed by additional and unrelated shRNA sequences (Data Supplement Fig S1). On the basis of the kinetic of STAT3 protein level reduction,
wecarriedoutGEPinALCLcells(Sup-M2-TS),expressingthreedifferent
STAT3 shRNA sequences, in conditions of acute or inducible KD. Supervised analysis identified a selected number of genes (1,453), specifically
modulated after STAT3 silencing (Fig 1A); these included known
STAT3 targets and putative STAT3-regulated genes. Notably,
more than 60% of differentially expressed genes were upregulated
after STAT3 ablation, indicating that STAT3 acts largely as a transcriptionalrepressor.Thetop30hitsincludedtranscriptsalsoregulatedby
NPM-ALK,31 such as IL2RA, LEF1, ICOS, RGS16, GAS1, and SGK (Data
Supplement Table S1). Comparison of NPM-ALK and STAT3 signatures
showed that 67% (337 of 499) of NPM-ALK target signals overlapped to
STAT3 KD genes (Fig 1B). To validate this integrated signature, we arbitrarily selected eight targets and verified that their expression was specifically downregulated by either ALK or STAT3 KD (Fig 1C). Thus in vitro
gene silencing experiments underscored the relevance of STAT3 transcriptional activity to the NPM-ALK signaling in ALCL cells.
Gene Expression Profiling Analysis Clusters Patients
With ALCL According to ALK Expression
We performed a global transcriptional analysis of systemic ALCL
tumor samples including 16 ALK-positive and 20 ALK-negative cases,
chosen on strict selection criteria (see Patients and Methods). To determine whether the generated expression profiles could identify distinct
clinical entities, we first performed an unsupervised analysis using a hierarchical agglomerative clustering algorithm.23 ALCL samples, described
by 4,676 most highly variable probe sets (ie, at least two-fold average ratio
www.jco.org
to the mean across all the values for each probe), generated a dendrogram
with two main branches in which ALK-positive and ALK-negative cases
were not distinguishable (Data Supplement Fig S2). To exclude possible
misinterpretation due to relatively low biologic heterogeneity of closely
related tumor cells within a heterogeneous non-neoplastic environment,35 weimplementedanunsupervisedanalysisusinganextendeddata
set including 28 PTCL-NOSs, six AILTs, and 20 purified normal T-cell
samples.20 Samples were clustered on the basis of the expression of 6,083
most-variable probe sets (specific for 2,964 genes). Normal T cells displayed a profile distinct from all T-NHLs, whereas complete separation
could not be achieved among PTCL-NOS, AILT, and ALK-negative
ALCL cases. Notably, ALK-positive ALCL cases gathered within a single
branch (Data Supplement Fig S3), indicating that these tissues could be
clustered according to ALK expression alone (P ⬍ 1 ⫻ 10⫺6).
ALK/STAT3 Signature Predicts ALK Status in Patients
With T-NHL
We then asked whether the STAT3 signature of ALK-positive
ALCL cell line (Sup-M2-TS) could disclose diagnostic significance in
primary T-NHL. Patients’ expression profiles were clustered accordingly to the 337 genes modulated by either ALK or STAT3 KD (Fig
1B). Unsupervised analysis of the corresponding 716 probe sets found
ALK-positive samples clustered in a single subgroup (P ⬍ 1 ⫻ 10⫺6),
revealing a smaller set of genes strongly correlated to ALK status (Fig 2,
between green lines). We therefore searched for genes characterizing
ALK-positive fingerprint using PAM software, a statistical method for
ranking genes and performing multiclass classification on the basis of
gene expression data.26 The classifier led to the identification of 34
© 2010 by American Society of Clinical Oncology
Downloaded from ascopubs.org by 78.47.27.170 on November 18, 2016 from 078.047.027.170
Copyright © 2016 American Society of Clinical Oncology. All rights reserved.
1585
Piva et al
+ ++++++
-3.0
+ +++++++
0
T-cell
ALCL
PTCL-NOS
+
ALK
+3.0
ANXA2
C13orf18
C14orf101
CCDC46
CD63
ETS2
EVI1
G0S2
GAS1
IL1RAP
IL2RA
IMPA2
IQCG
ITG81
LEPROT
LGALS1
MCAM
MYO10
NQO1
PRF1
PTPRG
RHOC
S100A6
SLC12A8
SLC20A1
SNFT
TCEA2
TEAD4
TMEM158
TNFRSF8
TST
UGCG
AILT
Fig 2. Hierarchical clustering of peripheral T-cell lymphomas not otherwise specified (PTCL-NOS), angioimmunoblastic lymphoma (AILT), normal T cells, and anaplastic
large-cell lymphoma (ALCL) samples according to the expression of 716 probe sets (specific for 337 genes) commonly modified in TS cell line by ALK and STAT3 KD.
The in vitro ALK/STAT3 signature clusters 15 out of 16 ALK ⫹ ALCL cases in a separate subgroup (P ⬍ .001). A smaller group of genes more strongly correlated with
ALK status is depicted between the green lines and indicated in the inset. ALK, anaplastic lymphoma kinase.
probes specific for 24 genes (Fig 3A), whose reliability was first confirmed by semiquantitative RT-PCR of selected genes (Fig 3B). Supplementary analyses determined that the five best-ranked genes
(IL1RAP, GAS1, PRF1, TMEM158, and IL2RA) were highly informative in the discrimination of patients with ALK-positive ALCL (sensitivity, 92.86%; specificity, 97.30%). Q-RT-PCR performed on
samples not included in the microarray profiling (eight PTCL-NOS,
eight ALK-positive and eight ALK-negative ALCLs) confirmed higher
expression of the three best classifier genes of the five identified
(IL1RAP, GAS1, and PRF1; Fig 3C). The predictive power of their
combined expression levels reached a classification score of 85% using
a quadratic discriminant analysis for classification of multivariate observations (data not shown). The performance of this classifier was
further challenged on a publicly available microarray ALCL set (15
expressing high level of ALK and eight ALK negative).13 The model
correctly recognized all ALK-positive specimens, with three ALKnegative samples misclassified (Data Supplement Fig S4). The high
1586
predictive classification power (sensitivity, 100%; specificity, 62.5%)
suggests that the identified ALK signature is independent from institutional bias and is a conserved feature of ALK-positive ALCL.
Finally, to validate the prediction analysis in a larger data set,
immunohistochemistry for CD30, ALK, GAS1, and STAT3 phosphorylation was applied to 91 ALCL cases (49 ALK-positive and 42
ALK-negative cases). GAS1 reactivity was detected in 84% of ALKpositive (34 expressing high and 12 expressing low levels of GAS1) and
in 14% of ALK-negative cases (five expressing high and one expressing
low levels of GAS1; Figs 3D and 3E). Notably, GAS1 was strictly correlated
with ALK expression with higher significance compared with STAT3
phosphorylation and C/EBP␤ (Data Supplement Fig S5).13,31,34,36,37
A Restricted Number of Genes Are Upregulated in
ALK-Negative ALCLs
WHO classification has recently defined ALK-negative ALCL as
a provisional pathologic entity.3 However, the diagnosis of ALK-negative
© 2010 by American Society of Clinical Oncology
Downloaded from ascopubs.org by 78.47.27.170 on November 18, 2016 from 078.047.027.170
Copyright © 2016 American Society of Clinical Oncology. All rights reserved.
JOURNAL OF CLINICAL ONCOLOGY
Molecular Classifiers for ALCL
ALCL ALK+
D
O
B
ALCL ALK
NPMALK
GAS1
IL2RA
C13ORF18
IL1RAP
PRF1
SCCA2
β2-MICRO
ALK
C
ALK P = .00116
PRF1 P = .22993
GAS1
-2
4
-4
-6
2
-8
0
-10
-12
-2
-14
PTCL
ALCL
ALCL
ALK neg ALK pos
PTCL
ILRAP P = .04275
0
0
-2
-2
-4
-4
-6
ALCL
ALCL
ALK neg ALK pos
ALCL
ALCL
ALK neg ALK pos
GAS1 P = .10834
2
PTCL
E
GAS1 Expression (%)
IL1RAP
GAS1
PRF1
TMEM158
IL2RA
MCAM
C14orf101
IMPA2
TEAD4
PTPRG
G0S2
C13orf18
AGT
SNFT
TNFRSF8
FUT7
NQO1
MYO10
SLC12A8
SPP1
CLEC3B
RHOC
S100A9
FBN1
H2
A
100
GAS1 expression
80
Negative
60
Low
High
40
20
0
ALK neg
PTCL
ALK pos
ALCL
ALCL
ALK neg ALK pos
Fig 3. (A) A genomic classifier of anaplastic lymphoma kinase (ALK) –positive systemic anaplastic large-cell lymphoma (ALCL) was identified by prediction analysis
of microarrays method. A restricted classifier with average sensitivity and specificity more than 95% is highlighted in yellow. (B) mRNA expression for PRF1, IL1RAP,
GAS1, and other selected ALK/STAT3 targets was determined by semiquantitative reverse-transcriptase polymerase chain reaction (RT-PCR) in 18 systemic ALCLs.
(C) Box plot of ALK, PRF1, IL1RAP, and GAS1 expression levels obtained by quantitative RT-PCR analysis in eight peripheral T-cell lymphomas (PTCLs) not otherwise
specified, eight ALK-negative, and eight ALK-positive ALCL samples. (D) Representative immunohistochemical staining for ALK and GAS1 of single ALK-positive or
ALK-negative ALCL samples. (E) Quantitative analysis of GAS1 staining relative to ALK status in patients with ALCL. Immunohistochemistry was performed in 49
ALK-positive and 42 ALK-negative systemic ALCL samples. neg, negative; pos, positive; ⌬Ct, difference in cycle threshold between reference and target gene.
ALCL is often problematic, because these T-NHLs lack a unique immunophenotype or restricted genetic markers. Therefore, their classification frequently relies on morphologic grounds and/or on the
expression of a limited number of antigens.6,7 Thus we investigated
whether a distinct gene-expression pattern was associated with
ALK-negative ALCLs. Among genes differentially expressed between ALK-positive and ALK-negative samples, only a minority
(3%) were upregulated in ALK-negative samples, irrespective of
the stringency of supervised analysis. Using a Q-RT-PCR approach
in an independent set of PTCL-NOS and ALCL samples, the predicted trend of overexpression for CD86 and ZNF267 was confirmed, even though differences were not statistically significant
(Data Supplement Fig S6). In agreement with previous analyses, PRF1,
IL1RAP, and GAS1 scored within the top upregulated transcripts of
ALK-positive specimens (Data Supplement Table S2); the most significantly over-represented functional categories included genes regulating development, neurogenesis, apoptosis, cell communication,
proteolysis, adhesion, motility, and transcription (data not shown).
www.jco.org
A Genomic Classifier Discriminates ALCL From
Other T-NHLs
To discover new predictors of ALK-negative ALCL, we then
compared the profiles of ALK-negative ALCL with those of other
T-NHLs and uncovered 14 genes capable of distinguishing ALKnegative ALCL from PTCL-NOS and AILT samples (sensitivity, 95%;
specificity, 100%). Unexpectedly, all 14 ALK-negative predictors were
similarly expressed by ALK-positive ALCL, suggesting the existence of
a common ALCL signature (data not shown). This hypothesis was
subsequently confirmed, comparing all ALCL samples with T-NHL.
PAM analysis led to the identification of a overlapping list of genes that
included 34 probes. The new classifier clearly separated ALCL from
PTCL-NOS, AILT, and normal T-cells (sensitivity, 97.22%; specificity, 100%; Fig 4A). The identified fingerprint was confirmed by Q-RTPCR in independent cases using four targets (TNFRSF8, SNFT,
NFATC2, and PERP), which were differentially regulated in patients
with ALCL (Fig 4B). As predicted, the immunostaining also revealed
weak/rare expression of NFATC2 in the anaplastic cells of patients with
© 2010 by American Society of Clinical Oncology
Downloaded from ascopubs.org by 78.47.27.170 on November 18, 2016 from 078.047.027.170
Copyright © 2016 American Society of Clinical Oncology. All rights reserved.
1587
Piva et al
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
A
ALK
NFATC2
LGALS3
ARNTL2
ARNTL2
PERP
SNFT
TNFRSF8
M6PRBP1
FLOT1
FLOT1
LGALS1
ANXA2
ANXA2
ANXA2
CLIC1
NDUFB7
COMT
ENO1
TPI1
GPATCH4
TMEM5
TNPO1
C1ORF21
KIAA0152
NUPL1
KPNB1
SART3
TMED4
PRF40A
ZC3H1S
IFI16
KIAA1033
PCYOX1
CHST11
-3.0
B
T-cell
ALCL
PTCL-NOS
TNFRSF8 P = .00171
AILT
0
+3.0
SNFT P = .00752
4
4
PERP P = .02019
4
1
3
2
2
2
0
2
-1
-2
1
0
0
-2
-1
0
-3
PTCL
ALCL
ALCL
ALK neg ALK pos
-4
-2
-2
-2
NFATC2 P = .00388
-5
PTCL
ALCL
ALCL
ALK neg ALK pos
PTCL
ALCL
ALCL
ALK neg ALK pos
PTCL
ALCL
ALCL
ALK neg ALK pos
Fig 4. (A) Hierarchical clustering of peripheral T-cell lymphoma not otherwise specified (PTCL-NOS), angioimmunoblastic lymphoma (AILT), normal T cells, and
anaplastic large-cell lymphoma (ALCL) samples according to the expression of 34 probe sets (specific for 30 genes) identified by prediction analysis of microarrays as
diagnostic predictors of ALCL. (B) Box plot of TNFRSF8, SNFT, PERP, and NFATC2 expression levels obtained by quantitative reverse-transcriptase polymerase chain
reaction analysis in eight patients with PTCL-NOS, eight patients with anaplastic lymphoma kinase (ALK)–negative anaplastic large-cell lymphoma (ALCL), and eight
patients with ALK-positive ALCL. ⌬Ct, difference in cycle threshold between reference and target gene.
ALCL, whereas it was consistently expressed in the neoplastic compartments of PTCL-NOS samples (Data Supplement Fig S7). Therefore, the
identification of a gene cluster for ALCL represents a valuable diagnostic
toolfortherecognitionofALK-negativeALCLandsupportsthenotionof
ALCL as a pathologic entity, molecularly discernible from other T-NHLs.
DISCUSSION
It is recognized that the transformation of mammalian cells requires
the sequential acquisition of genetic defects and that distinct neoplas1588
tic entities accumulate defined/recurrent aberrations. A molecular
characterization of human tumors is now mandatory to achieve precise diagnoses and to stratify subsets of patients sharing similar clinical
progression and response to therapy. This knowledge served as a
platform to open a new era for “targeted” protocols.38-40 However, a
major undertaking is the identification of patient-specific molecular signatures.41
Here, using ALK or STAT3 GEP signatures obtained from
ALK-positive ALCL cell lines, we demonstrated that primary systemic
ALK-positive ALCLs express a distinct profile, mainly dictated by
© 2010 by American Society of Clinical Oncology
Downloaded from ascopubs.org by 78.47.27.170 on November 18, 2016 from 078.047.027.170
Copyright © 2016 American Society of Clinical Oncology. All rights reserved.
JOURNAL OF CLINICAL ONCOLOGY
Molecular Classifiers for ALCL
STAT3 signaling. Importantly, the preferential expression of a limited
number of genes is sufficient to recognize ALK-positive ALCLs from
other T-NHLs, independent from ALK expression. On the contrary,
no significant markers specifically expressed in ALK-negative ALCLs
were identified. However, we recognized that ALCLs share a cluster of
transcripts, which allow their stratification and distinction from other
T-NHLs, suggesting a common ALCL signature and possibly unique
origin. Therefore, the ALCL classifier could represent a successful tool
for distinguishing ALK-negative ALCL from CD30⫹ PTCL-NOS.6
We have previously demonstrated that the transcriptional profile
of ALCL cell lines depends on NPM-ALK activity.31 Because STAT3 is
an essential target of ALK oncogenic signaling,33,34 we asked whether it
was also the main effector of ALK-mediated transcriptional regulation. Comparison of profiles obtained using either ALK or STAT3
silencing strategies demonstrated that the majority of ALK transcriptional targets are regulated by STAT3-mediated activity. Importantly,
a class-prediction analysis identified a restricted ALK/STAT3 signature, sufficient to distinguish systemic ALK-positive ALCL sample,
and demonstrating that RNAi-based GEP is a powerful tool to dissect
the molecular fingerprint of primary T-NHL. These findings were
further supported by meta-analyses on an independent ALCL database13 and by Q-RT-PCR validation, confirming that ALK-positive
ALCL can be successfully stratified. Notably, in Lamant’s analysis,
distinction between ALK-positive and ALK-negative ALCL could not
be obtained using an unsupervised approach. We faced a similar issue
when we limited our studies to the ALCL cases. This finding could be
due to several matters, including a limited number of samples and small
percentage of ALCL cells, which reduced the variability and heterogeneity
of the whole signature. Remarkably, when we selected ALCL with a high
number of tumor cells and incorporated in the analysis normal (resting
and stimulated T cells) and other pathologic (PTCL-NOS and AILT)
lesions, our findings led to the dissection of distinct subtypes.
We have also discovered 30 predictor genes differentially expressed both in ALK-positive and ALK-negative ALCLs, suggesting a
commonality among all ALCLs. Only a few of them are regulated/
associated via ALK signaling, indicating that ALK-independent genes
may be part of a common signature of the ALCL precursor or alternatively that their expression is due to genetic aberrations/defects that
regulate similar/identical pathways in all ALCLs. In fact, gene
ontology and gene-set-enrichment-analysis approaches (data not
shown) established that ALCLs share common pathways (ie, loss of
T-cell signaling, hypoxia, and mitochondrial signature). This hypothesis is further supported by the fact that most ALK-positive and ALKnegative ALCLs lack expression of TCR␤ protein.42 Moreover, a
subset of ALK-negative ALCLs express “bona fide” ALK-positive associated proteins (ie, phospho-STAT3 and/or C/EBP␤), which may
be positively upregulated via unknown activator(s) other then
REFERENCES
1. Harris NL, Jaffe ES, Stein H, et al: A revised
European-American classification of lymphoid neoplasms: A proposal from the International Lymphoma Study Group. Blood 84:1361-1392, 1994
2. Rizvi MA, Evens AM, Tallman MS, et al:
T-cell non-Hodgkin lymphoma. Blood 107:1255-1264,
2006
3. Swerdlow SH, Campo E, Harris NL, et al:
WHO Classification of Toumors of Haematopoietic
www.jco.org
ALK. Notably, when we test whether morphologic/cytologic features could further stratify ALCL samples, no significant correlations were found, further supporting a shared relationship of
these neoplasms.
In conclusion, an ALCL signature may be very useful to dissect
cases in which a definitive diagnosis of ALK-negative ALCL or PCTLNOS cannot be reached via morphology and/or the immunohistochemistry. This is not a trivial issue, because ALK-negative ALCLs
have a different clinical outcome as compared with PTCL-NOS.6 The
erroneous diagnosis of PTCL-NOS may lead to more toxic chemotherapeutic protocols in patients with ALK-negative ALCL; conversely, patients with PTCL-NOS with an incorrect ALK-negative
ALCL diagnosis may be treated with suboptimal therapies, leading
to clinical failure. If our findings, generated in a relatively small
group of tumors, are confirmed in multi-institutional studies, we
foresee the introduction in routine clinical settings of Q-RT-PCR
cards and/or antibody panels for a more precise diagnosis of
T-NHL. Thus a gene expression ALCL classifier may provide a new
approach to precisely define T-NHL and to a select more appropriate therapeutic protocols.
AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS
OF INTEREST
The author(s) indicated no potential conflicts of interest.
AUTHOR CONTRIBUTIONS
Conception and design: Roberto Piva, Elisa Pellegrino, Enzo Medico,
Stefano Pileri, Antonino Neri, Giorgio Inghirami
Financial support: Roberto Piva, Giorgio Inghirami
Administrative support: Giorgio Inghirami
Provision of study materials or patients: Alessandro Fornari, Alberto
Zamò, Fabio Facchetti, Maurilio Ponzoni, Eva Geissinger, Andreas
Rosenwald, Hans Konrad Müller-Hermelink, Christiane De
Wolf-Peeters, Pier Paolo Piccaluga, Stefano Pileri, Giorgio Inghirami
Collection and assembly of data: Roberto Piva, Elisa Pellegrino, Luca
Agnelli, Katia Todoerti, Valentina Grosso, Ilaria Tamagno, Alessandro
Fornari, Barbara Martinoglio, Enzo Medico, Eva Geissinger, Andreas
Rosenwald, Hans Konrad Müller-Hermelink, Antonino Neri
Data analysis and interpretation: Roberto Piva, Elisa Pellegrino, Luca
Agnelli, Enzo Medico, Pier Paolo Piccaluga, Stefano Pileri, Antonino
Neri, Giorgio Inghirami
Manuscript writing: Roberto Piva, Giorgio Inghirami
Final approval of manuscript: Roberto Piva, Luca Agnelli, Enzo Medico,
Fabio Facchetti, Maurilio Ponzoni, Andreas Rosenwald, Hans Konrad
Müller-Hermelink, Christiane De Wolf-Peeters, Pier Paolo Piccaluga,
Stefano Pileri, Antonino Neri, Giorgio Inghirami
and Lymphoid Tissues. Lyon, France, International
Agency for Research on Cancer, 2008
4. Morton LM, Wang SS, Devesa SS, et al:
Lymphoma incidence patterns by WHO subtype in
the United States, 1992-2001. Blood 107:265-276,
2006
5. Vose J, Armitage J, Weisenburger D: International
peripheral T-cell and natural killer/T-cell lymphoma study:
Pathology findings and clinical outcomes. J Clin Oncol
26:4124-4130, 2008
6. Savage KJ, Harris NL, Vose JM, et al: ALKanaplastic large-cell lymphoma is clinically and
immunophenotypically different from both ALK⫹
ALCL and peripheral T-cell lymphoma, not otherwise specified: Report from the International
Peripheral T-Cell Lymphoma Project. Blood 111:
5496-5504, 2008
7. Fornari A, Piva R, Chiarle R, et al: Anaplastic
large cell lymphoma: One or more entities among
T-cell lymphoma? Hematol Oncol [epub ahead of
print on April 8, 2009]
8. Stein H, Mason DY, Gerdes J, et al: The
expression of the Hodgkin’s disease associated
antigen Ki-1 in reactive and neoplastic lymphoid
© 2010 by American Society of Clinical Oncology
Downloaded from ascopubs.org by 78.47.27.170 on November 18, 2016 from 078.047.027.170
Copyright © 2016 American Society of Clinical Oncology. All rights reserved.
1589
Piva et al
tissue: Evidence that Reed-Sternberg cells and histiocytic malignancies are derived from activated
lymphoid cells. Blood 66:848-858, 1985
9. Morris SW, Kirstein MN, Valentine MB, et al:
Fusion of a kinase gene, ALK, to a nucleolar protein
gene, NPM, in non-Hodgkin’s lymphoma. Science
263:1281-1284, 1994
10. Chiarle R, Voena C, Ambrogio C, et al: The
anaplastic lymphoma kinase in the pathogenesis of
cancer. Nat Rev Cancer 8:11-23, 2008
11. Thompson MA, Stumph J, Henrickson SE, et
al: Differential gene expression in anaplastic lymphoma kinase-positive and anaplastic lymphoma
kinase-negative anaplastic large cell lymphomas.
Hum Pathol 36:494-504, 2005
12. Trempat P, Villalva C, Xerri L, et al: Gene
expression profiling in anaplastic large cell lymphoma and Hodgkin’s disease. Leuk Lymphoma
45:2001-2006, 2004
13. Lamant L, de Reynies A, Duplantier MM, et al:
Gene-expression profiling of systemic anaplastic
large-cell lymphoma reveals differences based on
ALK status and two distinct morphologic ALK⫹
subtypes. Blood 109:2156-2164, 2007
14. Piccaluga PP, Agostinelli C, Califano A, et al:
Gene expression analysis of angioimmunoblastic lymphoma indicates derivation from T follicular helper cells
and vascular endothelial growth factor deregulation.
Cancer Res 67:10703-10710, 2007
15. Salaverria I, Bea S, Lopez-Guillermo A, et al:
Genomic profiling reveals different genetic aberrations in systemic ALK-positive and ALK-negative
anaplastic large cell lymphomas. Br J Haematol
140:516-526, 2008
16. Falini B, Pileri S, Zinzani PL, et al: ALK⫹
lymphoma: Clinico-pathological findings and outcome. Blood 93:2697-2706, 1999
17. Gascoyne RD, Aoun P, Wu D, et al: Prognostic
significance of anaplastic lymphoma kinase (ALK)
protein expression in adults with anaplastic large cell
lymphoma. Blood 93:3913-3921, 1999
18. ten Berge RL, Oudejans JJ, Ossenkoppele
GJ, et al: ALK-negative systemic anaplastic large cell
lymphoma: Differential diagnostic and prognostic
aspects–a review. J Pathol 200:4-15, 2003
19. ten Berge RL, de Bruin PC, Oudejans JJ, et al:
ALK-negative anaplastic large-cell lymphoma demonstrates similar poor prognosis to peripheral T-cell
lymphoma, unspecified. Histopathology 43:462-469,
2003
20. Piccaluga PP, Agostinelli C, Califano A, et al:
Gene expression analysis of peripheral T cell lymphoma, unspecified, reveals distinct profiles and
new potential therapeutic targets. J Clin Invest
117:823-834, 2007
21. Zinzani PL, Dirnhofer S, Sabattini E, et al:
Identification of outcome predictors in diffuse large
B-cell lymphoma: Immunohistochemical profiling of
homogeneously treated de novo tumors with nodal
presentation on tissue micro-arrays. Haematologica
90:341-347, 2005
22. Fu L, Medico E: FLAME, a novel fuzzy clustering method for the analysis of DNA microarray
data. BMC Bioinformatics 8:3, 2007
23. Eisen MB, Spellman PT, Brown PO, et al:
Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 95:1486314868, 1998
24. Mattioli M, Agnelli L, Fabris S, et al: Gene
expression profiling of plasma cell dyscrasias reveals molecular patterns associated with distinct
IGH translocations in multiple myeloma. Oncogene
24:2461-2473, 2005
25. Tusher VG, Tibshirani R, Chu G: Significance
analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 98:51165121, 2001
26. Tibshirani R, Hastie T, Narasimhan B, et al:
Diagnosis of multiple cancer types by shrunken
centroids of gene expression. Proc Natl Acad Sci U
S A 99:6567-6572, 2002
27. Wiznerowicz M, Trono D: Conditional suppression of cellular genes: Lentivirus vector-mediated druginducible RNA interference. J Virol 77:8957-8961,
2003
28. Chatterjee M, Stuhmer T, Herrmann P, et al:
Combined disruption of both the MEK/ERK and the
IL-6R/STAT3 pathways is required to induce apoptosis of multiple myeloma cells in the presence of
bone marrow stromal cells. Blood 104:3712-3721,
2004
29. Moffat J, Grueneberg DA, Yang X, et al: A
lentiviral RNAi library for human and mouse genes
applied to an arrayed viral high-content screen. Cell
124:1283-1298, 2006
30. Piva R, Chiarle R, Manazza AD, et al: Ablation
of oncogenic ALK is a viable therapeutic approach
for anaplastic large-cell lymphomas. Blood 107:689697, 2006
31. Piva R, Pellegrino E, Mattioli M, et al: Functional validation of the anaplastic lymphoma kinase
signature identifies CEBPB and BCL2A1 as critical
target genes. J Clin Invest 116:3171-3182, 2006
32. Abruzzo LV, Barron LL, Anderson K, et al:
Identification and validation of biomarkers of IgV(H)
mutation status in chronic lymphocytic leukemia
using microfluidics quantitative real-time polymerase chain reaction technology. J Mol Diagn 9:546555, 2007
33. Chiarle R, Simmons WJ, Cai H, et al: Stat3 is
required for ALK-mediated lymphomagenesis and
provides a possible therapeutic target. Nat Med
11:623-629, 2005
34. Zamo A, Chiarle R, Piva R, et al: Anaplastic
lymphoma kinase (ALK) activates Stat3 and protects
hematopoietic cells from cell death. Oncogene 21:
1038-1047, 2002
35. Klein U, Tu Y, Stolovitzky GA, et al: Gene
expression profiling of B cell chronic lymphocytic
leukemia reveals a homogeneous phenotype related
to memory B cells. J Exp Med 194:1625-1638, 2001
36. Quintanilla-Martinez L, Pittaluga S, Miething
C, et al: NPM-ALK-dependent expression of the
transcription factor CCAAT/enhancer binding protein
beta in ALK-positive anaplastic large cell lymphoma.
Blood 108:2029-2036, 2006
37. Jundt F, Raetzel N, Muller C, et al: A rapamycin derivative (everolimus) controls proliferation
through down-regulation of truncated CCAAT enhancer binding protein beta and NF-kappaB activity
in Hodgkin and anaplastic large cell lymphomas.
Blood 106:1801-1807, 2005
38. Chin L, Gray JW: Translating insights from the
cancer genome into clinical practice. Nature 452:
553-563, 2008
39. Strausberg RL, Simpson AJ, Old LJ, et al:
Oncogenomics and the development of new cancer
therapies. Nature 429:469-474, 2004
40. Sawyers C: Targeted cancer therapy. Nature
432:294-297, 2004
41. van’t Veer LJ, Bernards R: Enabling personalized cancer medicine through analysis of geneexpression patterns. Nature 452:564-570, 2008
42. Bonzheim I, Geissinger E, Roth S, et al:
Anaplastic large cell lymphomas lack the expression of T-cell receptor molecules or molecules of
proximal T-cell receptor signaling. Blood 104:33583360, 2004
■ ■ ■
1590
© 2010 by American Society of Clinical Oncology
Downloaded from ascopubs.org by 78.47.27.170 on November 18, 2016 from 078.047.027.170
Copyright © 2016 American Society of Clinical Oncology. All rights reserved.
JOURNAL OF CLINICAL ONCOLOGY