Genetically-Encoded Neuronal Indicator and Effector (GENIE) Project

Genetically-Encoded Neuronal Indicator and Effector (GENIE)
Project
Steering Committee (2015-2016)
Reed George, Senior Director, Scientific Services
Vivek Jayaraman, Group Leader
Douglas Kim, Program Scientist
Wyatt Korff, Associate Director and Program Scientist, Team Projects
Loren Looger, Group Leader
Nelson Spruston, Scientific Program Director and Laboratory Head
Karel Svoboda, Group Leader
Research Staff
Stephan Brenowitz, Senior Scientist (voltage sensor development)
Hod Dana, Research Specialist (mouse imaging)
Benjamin Fosque, Research Technician (protein biochemistry)
Jeremy Hasseman, Research Specialist (molecular cloning)
Bei-Jung Lin, Research Specialist (mouse imaging)
Yi Sun, Research Specialist (Drosophila imaging)
Getahun Tsegaye, Research Specialist (molecular cloning)
Collaborating Scientists
Adam Cohen, Associate Professor, Department of Chemistry and Chemical Biology, Harvard
University (voltage sensor development)
William Schafer, Group Leader, Medical Research Council, Laboratory of Molecular Biology,
Cambridge University (C. elegans imaging)
Eric Schreiter, Senior Scientist, Looger Lab (calcium integrator development)
1
1.0 Summary
For more than 15 years, fluorescent proteins have been engineered to report neural activity.
Neural circuits process information on spatial scales ranging from single synapses to large
assemblies of neurons at time scales from milliseconds to months. Imaging based on fluorescent
protein sensors spans all of these temporal and spatial scales. Although protein sensors are now
routinely used to image neuronal activity in vivo, they still have major limitations.
The GENIE project develops prototype protein sensors into the best-in-class tools for in vivo
neurophysiology. Our focus is on high-impact tools that will benefit from protein engineering
and large-scale screening technology.
The project will continue to optimize green and red calcium indicators. Red indicators facilitate
deeper imaging in vivo and multi-color imaging. We have also initiated a program to engineer
genetically-encoded voltage sensors and switches that detect the coincidence of light triggers and
calcium.
2.0 Background
The goal of the GENIE Project is to engineer fluorescent sensors for imaging of neuronal
activity in the intact nervous system.
Understanding the function and plasticity of neural circuits requires measurements of neural
activity over a large range of spatial and temporal scales. Calcium indicators have been
especially useful for neurobiology. In most neurons, action potentials (APs) are tightly coupled
to large (20-fold) and rapid (rise time <1 millisecond) increases in intracellular free calcium
concentration. These calcium transients can be measured as fluorescence changes in neuronal
somata (1, 2), dendrites (3, 4), and axons (5, 6) in vitro and in vivo as a proxy for the underlying
electrical activity. Calcium indicators are also useful in non-spiking, graded potential neurons (7,
8), although optimal indicator parameters may differ. Additionally, excitatory synaptic
transmission is typically associated with large-amplitude (20- to 100-fold) increases in calcium
concentration in small synaptic compartments (e.g. dendritic spines). These synaptic calcium
transients can be used to quantify synaptic transmission in vitro (9) and in vivo (10, 11). Calcium
2
imaging therefore provides a versatile tool to probe neural activity over timescales of
milliseconds to months and spatial scales of micrometers to millimeters.
Genetically-encoded calcium indicators (GECIs) occupy a niche complementary to single unit
electrophysiology. Single unit methods have better sensitivity and time resolution. A major goal
of the GENIE Project is to optimize GECIs to achieve signal-to-noise ratios (SNRs) that rival
electrophysiological methods. We have developed the general-purpose GFP-based GCaMP6
sensors (11), which under favorable conditions detect single spikes in neuronal populations in
vivo. GECIs with even higher sensitivity will allow larger populations of neurons to be imaged
more rapidly and under more challenging conditions (e.g. during behavior). High sensitivity
comes at the price of limited dynamic range; high-sensitivity indicators saturate at modest
activity levels, making them unsuitable for monitoring activity patterns in neurons with high
spike rates (~100 Hz). Existing GCaMP6 indicators are not sufficiently sensitive to measure
single APs in some cell types, such as parvalbumin-positive interneurons in rodents. Beyond
GCaMP6, we will develop a suite of green indicators tailored to specific applications, including
imaging of large populations of neurons, single-spike detection across neuron types,
measurement of spike rates over a wide dynamic range, and imaging spike times with high-speed
indicators.
We are also developing red fluorescent GECIs (12, 13). There are four major reasons why red
indicators are of particular interest. First, longer wavelength excitation light penetrates deeper
into tissue. In addition, red fluorescence is absorbed much less than green fluorescence,
especially in mammalian tissue, giving red fluorophores a competitive advantage for in vivo
imaging. Red GECIs (R-GECIs) therefore promise deeper and less invasive imaging compared
to green GECIs. Second, many transgenic animals with green fluorescence have been created; RGECIs can be used in these animals. Third, R-GECIs are more easily combined with
channelrhodopsin-2 (ChR2)-based manipulations of neural circuits. Fourth, combinations of red
and green GECIs promise simultaneous imaging of multiple distinct neural structures (14). We
will thus develop a suite of sensitive R-GECIs.
Calcium imaging provides only an indirect report of neural activity. Electrical measurements are
required to understand how neuronal input is converted to neuronal output, the heart of neuronal
3
computation (15). Calcium is not an accurate proxy for subthreshold electrical dynamics of
neurons. Electrode-based measurements of membrane potential have been central to
neuroscience research for many years, but microelectrodes are invasive and difficult to use. A
long-standing goal has been to develop imaging methods for membrane potential (16). However,
despite several attempts by multiple groups, the best available protein voltage sensors still suffer
from small fluorescence changes in response to APs, low brightness, poor targeting to the plasma
membrane, large capacitive load on the membrane, and/or slow response kinetics (17-20).
Existing GEVI scaffolds require substantial improvement before they can be used to study neural
coding in vivo. We have begun a program to develop genetically-encoded voltage indicators
(GEVIs) for subcellular imaging.
A major goal of neuroscience is to identify neural ensembles, groups of neurons that are coactive
during specific phases of behavior (21). Tools to mark these ensembles in vivo, and to target
them for later characterization or manipulation are lacking. Promoters from immediate early
genes such as c-fos have been harnessed to trigger transgene expression in active cells, but these
methods lack sensitivity, respond with slow kinetics, and are modulated by many physiological
parameters other than neural activity (22). The GENIE project has developed switches that detect
the coincidence of light and calcium to mark neurons that are active during experimenter-defined
periods. We propose to develop similar switches to activate gene expression, either through
transcriptional regulation of a target gene or through recombination.
In mammalian systems, stable long-term expression of protein sensors remains a major
challenge. High protein concentrations are required for in vivo imaging (23)(GECIs, 10-100
micromolar). Viral infection using AAV viral vectors is currently the gene delivery method of
choice (24). Although AAV produces sufficient expression, expression levels continue to ramp
up over months (6-fold from 1 to 6 months of expression) (11), eventually causing cytotoxicity
and changes in neuronal behavior (24). GECIs can be expressed at constant levels over 10
months in transgenic mice, without signs of cytotoxicity (25). However, current schemes (e.g.
CAG promoter-GCaMP3 targeted to the Rosa26 gene locus) still fail to provide sufficient
expression levels for many applications. The lack of suitable methods for controlled long-term
expression might be the single most important impediment to progress in using geneticallyencoded sensors in rodents. We will develop novel transgenic mice for Cre-dependent gene
4
expression, optimized for in vivo imaging. In addition, we will optimize sensors for protein
stability, which may provide several benefits: boosting expression levels, reducing cytomorbidity
and decreasing levels of inactive sensor, which contributes to fluorescence background but not
signal.
3.0 Progress Report (September 2012-present)
3.1 Assay development
Underlying the development of GCaMP6 and other advances in GECI development has been a
neuronal culture assay for characterizing GECI variants rapidly in the relevant cell type (Fig. 1AC). This was necessary because neurons have fast and small-amplitude calcium accumulations
that are difficult to model in non-neuronal systems. Primary neuronal cultures in 96-well plates
are transduced with GECI variants and a nuclear fluorescent protein to control for expression
level. APs are triggered using extracellular stimulation electrodes, while neurons are imaged
using fluorescence microscopy. The output of the assay is the fluorescence dynamics of multiple
individual neurons to a variety of trains of APs (26). This assay therefore allows characterization
of large numbers of GECI variants under relevant conditions.
Over the last three years we quadrupled the throughout of our assay (December 2010: 0.99
variants tested/day; February 2014: 4.5 variants tested/day). A major change is that gene
transduction is now accomplished by DNA electroporation into primary neurons instead of viral
infection. Transfections are performed using a liquid handling robot. Screening is now routinely
performed in 96-well format instead of 24-well format.
We mutagenize scaffolds using gene assembly cloning methods (27). Amino acid positions are
selected for mutagenesis based on the crystal structures of sensors (28-30). Substitutions are
made at single positions, tested in the culture assay, and the best point mutants are combined.
Individual, sequenced clones are assayed in multiple wells.
The sensitivity of the neuronal culture assay is limited by variability in neuronal physiology. A
100% improvement over R-GECO1.0 or GCaMP3 can be detected for a 1AP response with eight
replicates (p<0.01). The screening cost is $36 per variant (without labor). Throughput is limited
5
Figure 1. GCaMP6 engineering.
(A) Neuronal culture screening platform. (B) Schematic of stimulation geometry. (C) Expression vector with human synapsin-1 promoter (syn), GECI
variant, internal ribosome entry site (IRES), nuclear localization signal fused with fluorescent protein (nls-fluorescent protein), and woodchuck hepatitis
virus post-transcriptional regulatory element (WPRE). (D) Optimization of GCaMP for detecting neuronal activity. GCaMP SNR after 1AP was
estimated for GCaMP, GCaMP1.3, GCaMP1.6, GCaMP2, GCaMP3, G-GECO1, GCaMP5G, and GCaMP6s. Note that SNR was estimated across
different types of assays published by multiple groups and is thus a rough estimate. Data from (2, 11, 24, 47, 56-60). (E) GCaMP structure (29, 30) and
mutations (red) found in different GCaMP6 variants. CaM-binding peptide M13 from myosin light chain kinase (M13, yellow), linker 1 (gray),
circularly-permuted EGFP (cpEGFP, green), linker 2 (gray), CaM (blue), calcium ions (gray spheres). (F) Screening results for GCaMP6. Top,
fluorescence change in response to 1 AP (vertical bars, ΔF/F0; green bar, OGB1-AM, left; black bars, single GCaMP mutations; red bars, combinatorial
mutations; blue, GCaMP6 indicators) and significance values for different AP stimuli (color plot). Middle, half decay time after 10 APs. Bottom, resting
fluorescence, F0 normalized to nuclear mCherry fluorescence. Red line, GCaMP3 level; green line, GCaMP5G level; blue line, OGB1-AM level. (G)
Single AP Responses averaged across multiple neurons and wells for GCaMP3, 5G, 6f, 6m, 6s, and OGB1. (H) GCaMP6 performance in the mouse
visual cortex in response to a single spike. Left, schematic of the experiment. Right, median fluorescence change in response to 1 AP for different
calcium indicators. Shading corresponds to s.e.m., n = 9 (GCaMP5K, data from (47)), 11 (GCaMP6f), 10 (GCaMP6m), 9 (GCaMP6s) cells. GCaMP5K
and GCaMP5G have similar properties.
6
Figure 2. Fly visual system assay.
(A) Schematic of the fly visual assay.
Female flies (3-5 days old) are glued onto a
custom-built pyramid and 2-photon imaged
through a window in the cuticle on the head.
Visual stimulation is delivered through a
modified digital-mirror-device that reflects a
dispersed laser beam (532nm, DPSS) onto a
pattern on a screen in front of the fly. A
signal is also acquired from a silicon
photodiode that detects changes in light
intensity to synchronize visual stimulation
with imaging. Laser light is isolated from
GCaMP fluorescence through spectral
filtering and physical masking. Inset:
Schematic of the optic lobe, including lamina
monopolar cell L2, where the axons
terminating in the second layer in the
medulla are imaged (dashed rectangle); and
the lobula plate tangential cell HS, where
dendritic trees are imaged (dashed
rectangle). (B) GCaMP6f is faster and more
sensitive than GCaMP6s and GCaMP5G in
L2. L2 is a graded potential neuron that
depolarizes in response to light offset.
Whole-field flickering with different
durations and duty cycles are used as visual
stimulation. Gray bars denote light offset.
(C) GCaMP6f responses to whole-field
flickering visual stimulation flashed at
various frequencies. (D) Representative
traces of GCaMP6s and GCaMP6f in the HS
neuron to moving gratings tilted at different
angles, moving at different speeds and with
different contrast (single condition shown).
In these large non-spiking neurons,
GCaMP6s responses are larger than
GCaMP6f. GCaMP6f captures the stimulus
onset and offset and is more similar to the
envelope of the intracellular recording curves
of this neuron. (E) Direction tuning of the
imaged dendritic area of HS neuron
measured with GCaMP6f (blue) and
GCaMP6s (red). (F) Frequency tuning of the
imaged dendritic area of HS neuron
measured with GCaMP6s. (G) Contrast
tuning of the imaged dendritic area of HS
neuron measured with GCaMP6s.
by the cloning rate (7.7 constructs cloned/day for RCaMP and R-GECO; an improvement of a
factor of 6 over the GCaMP cloning rate, 1.3 constructs cloned/day). The neuronal culture assay
has excess capacity.
7
Fly visual system
The GENIE team has validated the performance of GECIs in adult Drosophila by comparing
odor responses in projection neurons (PNs) of the fly antennal lobe. Although recording the
activity of PNs in specific glomeruli in this early olfactory system assay has been useful for
characterizing GECI performance in vivo (11), it has some limitations. First, odor delivery is
notoriously difficult to control with precision and consistency. Second, although it is possible to
perform cell-attached/loose-patch recordings from PNs, each glomerulus is innervated by
multiple PNs, and a single GAL4 driver typically labels most or all of them, making it difficult to
rigorously relate glomerular calcium activity to electrical activity in a single PN (31). Thus, we
have begun to develop a visual system assay for testing GECIs. We have validated the assay by
recording optic lobe responses in fly lamina and lobula plate neurons (Fig. 2).
Going forward, we will increasingly focus on spiking neurons, and specifically interneurons
similar to H1 and H2 in the optic lobe (32). We have chosen this class of neurons for their large
size and tractability for electrophysiology, well-characterized visual responses, and the high
dynamic range of firing rates that can be induced with appropriately controlled visual
stimulation. We have identified candidate GAL4 lines that target such large neurons and are in
the process of characterizing their visual responses using 2-photon imaging and
electrophysiology.
3.2 Sensitive and fast GCaMPs
We developed the general-purpose GCaMP6 family of indicators with improved SNR for spike
detection in vivo (Fig. 1D-H) (11). We characterized close to 500 structure-guided mutations,
with as many as 11 combined mutations in GCaMP3. Many of these mutations were at the
interface between GFP and calmodulin (CaM), which was extensively remodeled. We identified
GCaMP variants with 3-fold faster rise times and 2-fold decreased decay times. The kinetic
improvements are based on a mutation in the M13/CaM interface. This CaM bump combines
modularly with sensitivity-increasing mutations. The GCaMP6 sensors show single AP
responses that equal or exceed those of Oregon Green Bapta-1 (OGB1), the most widely used
calcium indicator. GCaMP6s detects all active neurons in the visual cortex in vivo. GCaMP6s,
6m, and 6f (for slow, medium, and fast) differ in their response kinetics, with 6f having the
8
Figure 3. GCaMPs with varying affinity and cooperativity.
(A) Schematic and representative traces of the NMJ assay in Drosophila larvae. In a fillet preparation, motor
neuron axons are cut and electrically stimulated with a suction electrode, and type Ib/s boutons are wide-field
imaged with an EMCCD camera. Images are segmented and analyzed in response to stimulation. Scale bars, 5
µm. (B) Fluorescence changes (∆F/Fapo) from calcium titrations of purified GCaMP protein variants. (C)
Frequency tuning curves for low affinity variants (EF-hand mutants), GCaMP6, and GCaMP5G. The averaged
maximum ΔF/F0 at different stimulation frequencies are plotted. Different variants are color-coded according to
their rank order of dissociation constant (KD). Note that the low affinity variants are right-shifted and have
shallower slopes. (D) Decay kinetics for the low affinity variants, GCaMP6, and GCaMP5G. The low affinity
variants are faster than previous GCaMPs including GCaMP6f. (E) Rise kinetics for the low affinity variants,
GCaMP6, and GCaMP5G. The low affinity variants are faster than previous GCaMPs including GCaMP6f.
fastest kinetics of any useful GECI. In 2013, the GCaMP6 indicators were amongst the most
widely requested reagents at Addgene.org (481 GCaMP6s requests as of March 2014; 48th most
requested plasmid ever) and the University of Pennsylvania Viral Vector Core. GCaMP6expressing flies were amongst the most requested fly strains at the Bloomington Drosophila
Stock Center (148 GCaMP6m requests in 2013; most requested stock out of 37,125 stocks in
2013).
Beyond GCaMP6, we have begun to engineer GCaMP6 variants with low resting fluorescence,
to maximize SNR for large-scale in vivo imaging in rodents, higher resting fluorescence for
applications in flies, and GCaMP6 variants with varying calcium affinities and binding
9
cooperativity, to enable activity imaging in neurons with graded responses and fast-spiking
neurons (Fig. 3).
3.3. FRET-based GECIs
We have completed a collaboration with Oliver Griesbeck (MPI Martinsried), resulting in the
creation of the best-in-class troponin-based FRET GECIs (33). Ratiometric GECIs have
advantages for imaging moving preparations, such as C. elegans. We focused on the TN-XXL
ratiometric (CFP/YFP) indicator scaffold. TN-XXL uses a calcium-binding domain, troponin,
which is orthogonal to CaM in terms of structure and perhaps toxicity profile (34). We screened
121 TN-XXL-based variants from the Griesbeck lab using neuronal culture screening. The
resulting ‘Twitch’ sensors have 8-fold greater single AP responses compared with TN-XXL (33).
3.4 R-GECIs
We have focused on optimizing two scaffolds, RCaMP and R-GECO, which are based on
mRuby (35) and mApple (36), respectively. We first targeted our engineering to improve
sensitivity (including SNR and brightness) and kinetics. Overall, the strategy employed is similar
to GCaMP6 (11). Libraries of point mutations are being generated in a structure-guided manner
(Fig. 4) (12). Targets for mutagenesis include the calcium-binding domains, the binding pocket
in CaM for M13, and the interface between CaM and fluorescent protein. More than 700 RCaMP
and 500 R-GECO variants have been characterized in the neuronal culture assay thus far.
A novel R-GECO variant reliably detects single APs in the cell culture assay (Fig. 5G-L). Some
variants have kinetics in the range of GCaMP6f (~25 ms half rise time). Structure-guided
engineering of the RCaMP and R-GECO scaffolds has improved the 1AP SNR in cultured
neurons by 7- and 5-fold, respectively, thus far (Fig. 4 and 5).
Compared to GFP, red fluorescent proteins, and the derived red calcium sensors, suffer from
photostability and brightness problems. Under 2-photon excitation, RFPs tend to bleach rapidly
or show reversible dark states. The peak brightness per molecule (an experimental quantity
proportional to quantum efficiency (QE) and inversely related to bleaching rate) (37) is
substantially smaller for R-GECO1.0 and RCaMP1h compared to GCaMP (~33% and ~50%,
respectively) (12). In addition, R-GECO1.0 is transiently induced to become red fluorescent after
10
Figure 4. R-GECI variants.
(A) RCaMP structure (12). M13 peptide (orange), linker 1 (gray), circular permuted mRuby (red), linker 2 (yellow), and CaM (purple). (B)
RCaMP amino acid positions that were mutated (red). (C) Screening results, 700 RCaMP1h variants. Top, fluorescence change in response
to 1 AP (vertical bars, ∆F/F0; green bars at right, GCaMP5G, OGB1, GCaMP6s, respectively; black bars, RCaMP1h variants; and
significance values for different AP stimuli (color plot). Middle, half decay time after 10 APs. Bottom, resting fluorescence, F0 normalized
to nuclear GFP fluorescence. Red line, RCaMP1h level; green line, GCaMP5G level; blue line, GCaMP6s level. (D) R-GECO1.0 structure
(12). M13 peptide (orange), linker 1 (gray), circular permuted mRuby (red), linker 2 (yellow), and CaM (purple). (E) R-GECO1.0 amino
acid positions that were mutated (red). (F) Screening results, 500 R-GECO1.0 variants. Top, fluorescence change in response to 1 AP
(vertical bars, ∆F/F0; green bars at right, GCaMP5G, OGB1, GCaMP6s, respectively; black bars, R-GECO1.0 variants; and significance
values for different AP stimuli (color plot). Middle, half decay time after 10 APs. Bottom, resting fluorescence, F0 normalized to nuclear
GFP fluorescence. Red line, R-GECO1.0 level; green line, GCaMP5G level; blue line, GCaMP6s level.
11
Figure 5. Performance of selected R-GECI variants in neuronal culture.
(A) Responses averaged across multiple neurons and wells for RCaMP1h (black), OGB1 (green), RCaMP1h
variant 471 (blue), and variant 692 (red). Fluorescence changes in response to 1 AP. Error bars correspond to
s.e.m. (B) Responses to 10 APs. (C) Comparison of RCaMP1h, variants, and OGB1-AM as a function of AP
number at 83 Hz (colors as in A). (D) SNR. (E) Half decay time. (F) Half rise time (after stimulus offset). (G)
Responses averaged across multiple neurons and wells for R-GECO1.0 (black), OGB1 (green), R-GECO1.0
variant 783 (blue), variant 819 (red), and variant 1027 (magenta). Fluorescence changes in response to 1 AP. (H)
Responses to 10 APs. (I) Comparison of R-GECO1.0, variants, and OGB1-AM as a function of AP number at 83
Hz (colors as in G). (J) SNR. (K) Half decay time. (L) Half rise time (after stimulus offset).
exposure to blue light, which makes it incompatible with the action spectrum of ChR2. We hope
to address these issues over the next year.
Despite these challenges, one novel R-GECO variant performs well in mouse V1 neurons (Fig.
6). We expect to release next generation RCaMP and R-GECO variants by the end of 2014.
3.5 GEVIs
Over the last fifteen years several different types of fluorescent protein sensors of voltage have
been developed. One scaffold is based on archaerhodopsin-3 (Arch) (38). We modified our
neuronal culture assay for screening fast voltage sensors. This required illumination with a highintensity laser (638 nm; ~1 kW/cm2) and imaging with a high-speed camera (EM-CCD, 500 fps,
128x128 pixels). A limited set of Arch variants was tested in collaboration with Adam Cohen
12
(Harvard). Wild-type Arch-GFP exhibited a 1AP ΔF/F0 response of 15%, whereas the D95E, Y,
and N variants showed responses in the 3-7% range. More importantly, Arch quantum efficiency
is tiny, making it very challenging to use in thick specimens. The quantum efficiency is also
likely intrinsic to the retinal chromophore and unlikely to be substantially increased by protein
engineering. We have therefore abandoned the Arch scaffold. Instead we will focus on
fluorescent protein-based GEVIs (17) or perhaps in hybrid Arch-FP sensors (39).
3.6 Calcium-activated and light-dependent switch prototypes
We have developed a genetically-encoded, light-gated integrator of calcium concentration for
neurons (Calcium-modulated Photoconvertible Ratiometric Integrator, CaMPARI) (Fig. 7).
CaMPARI allows the tagging of neurons that are active during a user-specified epoch of
illumination (milliseconds to seconds). CaMPARI is a photoswitchable protein (based on EosFP
(40)) where the photoswitching has been engineered to be calcium-dependent. CaMPARI
survives tissue fixation and can thus be imaged post hoc in histological preparations of fly and
other brains (Fig. 7H-K). This allows post hoc, brain-wide mapping of neurons that were active
during defined epochs, defined by illumination in vivo.
We have made incremental improvements to light-activated Cre recombinase (41) (Fig. 8).
Light-activated Cre recombinase could be made dependent on neural activity (e.g. calcium) in
the future to provide a powerful tool to switch genes on/off in active neurons.
3.7 Transgenic/Targeted Mice
GCaMP6s and GCaMP6f transgenic mice were produced using Thy1-GCaMP6-WPRE (42) and
Camk2a-GCaMP6 (43) expression cassettes. Most founders (>25 lines) showed negligible
expression. Five Thy1 mouse lines were found suitable for in vivo imaging, with high expression
levels in the hippocampus, neocortex, and other structures (Fig. 9). Neurons did not show any
cytopathology even after long-term expression of GCaMP6. These mice have been made
available at The Jackson Laboratory. Unfortunately, Thy1 mice show frustratingly inconsistent
expression patterns across brain regions and neurons. Thus each line is suitable only for a small
set of brain areas. A stable, retargetable, and tunable strategy for mouse transgenesis is needed
for in vivo imaging.
13
Figure 6. Performance of novel R-GECO variant in mouse visual cortex.
(A) Top, schematic of the experiment. Bottom, field of view showing neurons expressing R-GECO1.0 (left) and RGECO variant 850 (right). (B) Example traces from neurons expressing R-GECO1.0. Single sweeps (gray) and
averages of sweeps (red) are overlaid. Directions of grating motion (8 directions) are shown above traces (arrows).
(C) Example traces from neurons expressing R-GECO variant 850. Single sweeps (gray) and averages of sweeps
(black) are overlaid. (D) The fraction of cells scored as responding to visual stimulation when loaded with different
R-GECO1.0 and R-GECO variant 850. Error bars correspond to s.e.m. (E) The distribution of fluorescence
changes across cells at the preferred orientation.
14
4.0 Research Plan
4.1 Specific Aims for the Next Project Period
4.1.1 GCaMPs
(FY2015 commitment: 21%, FY2016 commitment: 16%)
•
Engineer and characterize 500 additional GCaMP variants.
•
Screen for high SNR GCaMP variants.
•
Screen for GCaMP variants with lower affinity and Hill coefficient.
•
Screen for fast GCaMP variants.
•
Develop and test low-F0 variants for imaging large volumes in vertebrates.
•
Develop and test higher-F0 variants for imaging in the fly brain.
•
Develop a protein stability assay and screen for stable GCaMP variants.
•
Test the 10 best variants in vivo.
4.1.2 R-GECIs
(FY2015 commitment: 32%, FY2016 commitment: 32%)
•
Engineer and screen 1,000 RFP-based calcium indicator variants.
•
Screen for sensitive, fast and photostable variants.
•
Test the 10 best variants in vivo.
4.1.3 GEVIs
(FY2015 commitment: 37%, FY2016 commitment: 37%)
•
Benchmark existing scaffolds.
•
Establish a high-throughput platform for evaluating protein voltage sensors.
•
Engineer and screen 100,000 voltage sensors
•
Test most promising 200 variants in cultured neurons
•
Test 10 best variants in mouse, Drosophila, and zebrafish in vivo.
4.1.4 Calcium-Activated and Light-Dependent Switch Prototypes
(FY2015 commitment: 5%, FY2016 commitment: 5%)
•
Prototype two light and activity-dependent switches (light-sensitive Cre recombinase and
transcription factors).
15
•
Combine switches with reporter constructs to control neuronal ensembles (e.g. using
ChR2).
4.1.5 Transgenic/Targeted Mice
(FY2015 commitment: 5%, FY2016 commitment: 11%)
•
Engineer mice with high and stable neuronal expression of GCaMP6, RCaMP, and RGECO.
4.2 Experimental Plan
4.2.1 GCaMPs
Sensitive GCaMPs
The SNR of GECIs is largely determined by the fluorescence change per unit of activity (i.e. an
AP, or a synaptic input) (44-46). This fluorescence change, ∆F/F, in turn depends on the
brightness of the calcium-bound bright form, the fraction of indicator that is calcium-bound, and
the brightness of the calcium-free indicator. Increasing ∆F/F can be achieved by tweaking three
parameters:
1) Increase the affinity for calcium. This causes a larger ∆F/F per AP if spike rates are low.
Increasing affinity was a major part of the GCaMP6 design strategy (11). One drawback of
further pushing in this direction is that small stimuli will bring the indicator close to saturation,
reducing the effective dynamic range.
2) Maximize fluorescence of the calcium-bound indicator. Screening for higher QE was also part
of the GCaMP5 (47) and GCaMP6 (11) design strategies. However, the brightness of calciumbound GCaMP6 is already higher than GFP and it is therefore unlikely that this parameter will
yield additional improvements.
3) Minimize resting fluorescence. Our screen has already revealed linker mutations with reduced
resting fluorescence, without a large reduction of the peak fluorescence. In the past we have not
focused on these mutations because our strategy was to produce indicators with appreciable
resting fluorescence, which is critical for structural imaging. However, transgenic animals now
allow imaging in a background in which neurons, nuclei, or other structures of interest are
16
Figure 7. Performance of CaMPARI in vitro
and in fly neurons.
(A) Schematic of a fluorescent protein that
preferentially undergoes 400 nm light-induced
green-to-red photoconversion in the presence of
high calcium ion concentrations. (B) Schematic
of the primary structure of CaMPARI
illustrating the domain organization of nuclear
export signal (NES), CaM, linker 1 (L1),
mEos2 variant C-terminal portion (mEos2-C),
circular permutation linker (cpL), mEos2
variant N-terminal portion (mEos2-N), and
CaM-binding fragment from myosin light chain
kinase (M13). (C) Timecourse of red
fluorescence appearance during exposure of
CaMPARI to photoconversion light in the
presence of saturating calcium (closed circles)
or in the presence of calcium (open circles).
Lines represent single exponential fits to the
photoconversion timecourse. (D) CaMPARI
photoconversion rate (calculated as in (C)) as a
function of different concentrations of free
calcium ions. Black line represents a sigmoidal
fit to the data. (E) Green and red (purple false
color) composite epifluorescence micrograph
(bottom) of primary neurons expressing
CaMPARI after a 2 s photoconversion light
pulse (top). (F) Same as (D) except that 2 s of
80 Hz field electrode stimulation of the neurons
accompanied the 2 s photoconversion light
pulse. (G) Quantitation of the red/green
fluorescence ratio in neurons photoconverted
for 2 s concomitant with 2 s of field electrode
stimulation to induce AP firing at different
frequencies. The black line is the best linear fit
to the data. Values represent means across at
least 10 independent measurements, normalized
to the group that was not stimulated. (H)
Schematics of olfactory assay with CaMPARI
in Drosophila. Adult flies are tethered to a
holder contained within an air-tight chamber.
Odors are delivered through an olfactometer
into the chamber and rapidly evacuated . UV
photo-conversion light (405nm, ~4mW/mm2) is
delivered through the objective onto the head.
Odor delivery duration is 1 second, while UV
light is delivered in four 500 millisecond
windows with a 200 ms interstimulus interval.
After the experiment, brains are dissected and
imaged by confocal microscopy. (I) Control
experiments where UV light illuminates for up
to 300 s in total. (J)When fly is presented with
phenylacetic acid, VL2a glomerulus is
photoconverted into red. (K) When fly is
presented with geosmin, DA2 glomerulus is
photoconverted into red.
17
labeled with red fluorescence protein, obviating the need for substantial resting fluorescence.
Reducing resting fluorescence even by a factor of two will have a substantial impact on SNR,
increasing the yield in cellular imaging experiments by several-fold. We believe that this is the
most fruitful path forward. We will therefore create GCaMPs with low resting fluorescence for
large-scale, high SNR imaging. 200 linker variants of GCaMP6s and 6f will be screened. We aim
to engineer variants maintaining the peak fluorescence of GCaMP6s and 6f, while lowering F0
by 5-fold.
In flies, where it is common to use Gal4 drivers to target GECIs to sparse neural populations in
neuropil (rather than cell bodies), substantial resting fluorescence remains desirable to easily
locate and identify neural structures for imaging. Higher resting fluorescence is particularly
important in behaving flies, in which high illumination intensity can artificially excite neural
tissue and result in behavioral artifacts. In addition, the use of a co-labeling FP (e.g. RFP)
complicates the use of multiple GECIs for simultaneous imaging of disparate neural populations.
Thus, moderate-to-high resting fluorescence will continue to be a consideration when screening
GCaMPs and R-GECIs for use in Drosophila.
Stable GCaMPs
GCaMP6 sensors have to be used at high concentrations (approximately 10-150 micromolar) for
in vivo imaging. These protein concentrations are among the highest normally achieved in the
brain and are difficult to mimic in transgenic mice. Instead GCaMPs are typically delivered
using viral vectors such as AAV. AAV-mediated expression has undesired consequences.
Protein expression increases monotonically with time, ultimately leading to cytotoxic GCaMP
concentrations. The high expression levels required interfere with expression of endogenous
proteins. Protein concentrations depend on the rate of production and the rate of degradation. To
increase protein concentrations without increasing expression, we will engineer GCaMPs with
increased protein stability. We anticipate that more stable GCaMPs will allow higher protein
concentrations at modest expression rates. This will allow use of transgenic methods for
expression and imaging with reduced cytotoxicity.
We will test 200 more GCaMP6s and 6f variants to improve protein stability. This will include
altering potential ubiquitination sites, improving thermodynamic stability, and optimizing codon
18
usage for mice and flies. Variants will be screened by transient transfection of HEK293T cells,
application of a protein synthesis inhibitor (cycloheximide), and then epifluorescence imaging of
or fluorescence-activated cell sorting (FACS) to measure GCaMP variant fluorescence and
fluorescence of a co-expressed mCherry control. We will also directly measure extent of
degradation through the proteasomal pathway, through use of inhibitors such as lactacystin (24).
The goal will be to achieve a 5-fold increase in stability leading to a 5-fold increase in protein
concentration, while maintaining GCaMP6 SNR levels. The best variants will be tested in
cultured neurons and in vivo in mice and flies.
Linear GCaMPs
In fly larval neuromuscular junction (NMJ) assays, GCaMP6 responses were as much as 3-fold
increased over those of GCaMP5G with single AP stimulation. But responses were not improved
at higher frequencies (80-160 Hz, Fig. 3C). Although it may not be possible to reduce the Hill
coefficient to the level of the best synthetic indicators (e.g., OGB1, with Hill coefficient of 1)
and produce linear responses across low and high calcium concentrations, we hope to produce a
family of indicators with different KD's that span the dynamic range of typical neuronal firing
rate regimes and calcium concentrations. To enable improved calcium imaging over a wider
range of stimuli in fly neurons, we will undertake engineering of (1) a series of lower affinity
indicators that extend across the calcium range of various types of fly neurons, and (2) a single
indicator with much wider dynamic range.
This will be done by testing 50 more mutants with mutations at one or more of the 4 calciumbinding sites of CaM and the M13/CaM interface. Calcium-binding mutations have been
previously demonstrated to lower affinity and cooperativity (48). The reduction of cooperativity
of calcium binding could aid in widening the dynamic range of calcium sensing, which we have
begun to explore (Fig. 3). The M13/CaM interface can affect calcium-binding and unbinding
rates. The goals will be to create (1) a series of 5-10 indicators with calcium KD's from 500
nanomolar to 10 micromolar, while maintaining GCaMP6 SNR levels, and (2) an indicator with
a GCaMP6s-like calcium KD but a Hill slope of 1 to 1.5, while maintaining GCaMP6s SNR
levels. The kinetics of lower KD mutants could be improved compared to GCaMP6, which may
allow instantaneous firing rates to be tracked in high-firing-rate neurons.
19
Figure 8. Light-induced Cre recombinase.
(A) Schematic of split Cre halves interacting on DNA using light-sensitive dimerization domains from CRY2
and CIBN proteins. (B) Schematic of split Cre halves. Control (*) constructs with CRY2-Linker-Cre Nterminal domain ending at amino acid position 104 (CRY2 L CreN-104) and CIBN-Linker Cre C-terminal
domain (CIBN L CreC-106) beginning with amino acid position 106 as in (Kennedy M J 2010). Constructs
with various split sites are also shown. (C) Light-induced recombination of CAG-lox-stop-lox-GFP reporter
construct. Various combinations of split Cre constructs were co-transfected with a reporter construct into
HEK293T cells. Cells were exposed to pulsed blue light (2 s on, 180 s off) for 0, 1, 24 h. A red fluorescent
protein transfection control construct was also used. Percent recombination was the ratio of the number of green
to red cells.
Milestones
For FY2015, we will engineer low-F0 GCaMPs such that 5-fold more neurons can be imaged in
vivo with GCaMP6s-like SNR for mouse and limited fly applications. For FY2016, stable
GCaMPs will be produced that have 5 times greater stability of GCaMP6 when expressed in vivo
and are less toxic. GCaMPs with lower calcium affinity and cooperativity will be made that span
the calcium ranges of various fly high frequency and analog neurons for improved activity
detection.
4.2.3 R-GECIs
We propose to engineer improved RCaMPs and R-GECOs. After making considerable gains in
amplitude and speed (Fig. 4 and 5), we will next address the following design parameters:
Peak brightness is an easily measurable parameter that is highly relevant to imaging. It is
proportional to the QE and related inversely to the bleaching cross-section (37). We will
maintain ∆F/F0 and increase peak brightness. So far our protein engineering has focused on sites
20
away from the chromophore of RCaMP and R-GECO (Fig. 4A, D). We will first identify highest
SNR and highest SNR/highest speed variants and then mutate near the chromophore. We believe
that given the high level of sequence homology between red FPs it should be possible to graft
chromophore-forming residues and surrounding side-chains from recently improved FPs (e.g.
mCrimson, mCardinal) without having to start from scratch with the new scaffolds. It is likely
that subsequent optimization will be required after chromophore grafting, but such designs may
improve brightness or photostability in an efficient manner. Photophysical measurements will be
performed in collaboration with the JFRC Applied Physics and Instrumentation Group (APIG) in
crude protein preparations. The goal will be obtain improvements in photobleaching rates and
photoswitching. Spectral separation from green fluorescent calcium indicators will also be tested.
We will also explore long Stokes' shift R-GECIs that can be co-excited with GCaMP but yield
easily separable emission.
Lower F0 for improved imaging of neuronal populations. We seek to lower the background
fluorescence from inactive neurons and processes in order to increase SNR. A co-expressed GFP
could be used to identify sensor-expressing neurons in vivo. Re-engineering of the M13/FP and
FP/CaM linkers will be done to lower F0 while maintaining the peak fluorescence, similar to the
strategy for low-F0 GCaMPs. The goal will be to increase SNR so that thousands of neurons in
vivo can be imaged simultaneously at high speed in mouse brains.
Lower affinity for wider dynamic range in fly neurons. The gains in RCaMP and R-GECO
response amplitudes have been obtained through driving the apparent calcium KD to lower levels
than for the parent sensors. The response amplitudes are improved for low numbers of action of
potentials. But for high numbers of APs, responses are not much improved. By mutating
individual or several of the 4 calcium-binding domains in CaM, we will create an affinity series
of RCaMP and R-GECO sensors. Our goal is also to widen the dynamic range of calcium
detection by reducing the cooperativity of calcium binding.
Milestones
For FY2015, RCaMP and R-GECO variants will be produced with these properties: high fidelity
detection of single AP spikes in vivo, brightness sufficient to image from thousands of neurons,
at high speed, spectrally compatible with GCaMP and ChR2. For FY2016, RCaMP and R-GECO
21
variants will be produced with these properties: lower F0 to reduce background signal for
improved population imaging, and lower affinity and wider dynamic range for activity
monitoring of high frequency fly neurons.
4.2.5 GEVIs
Benchmarking existing voltage sensors. Several fluorescent protein voltage sensors have been
developed. These sensors differ in topology, voltage sensor moiety, contrast mechanism, etc. We
will begin by benchmarking the best versions of the existing scaffolds, including the singlewavelength sensors ArcLight (19), ASAP1 (18), VSFP3.1 (49), and FRET-based sensors
VSFP2.3 (49), VSFP-Butterfly (49) and Mermaid (50). We currently do not plan to pursue Archbased sensors because of their low fluorescence quantum yield (38) that is inherent to the retinal
and not likely to be significantly improved by protein engineering. Benchmarking will be based
on fluorescence microscopy and whole-cell recordings in HEK cells (Fig. 10) and in primary
neuronal cultures.
Screening assay for voltage sensors. In contrast to GECIs, the mechanism underlying
fluorescence changes are not understood in GEVIs. Large-scale mutagenesis and engineering,
coupled with a high-throughput assay for GEVI function, will be required to make substantial
advances.
We will thus develop a non-neuronal assay that will provide higher screening rates at reduced
cost. The assay requires robust dividing cells, rather than primary cultured neurons. Also needed
are cells with a stereotyped, timed voltage response that can be evoked synchronously across a
population in a well. This is critical because for a high-throughput assay we would like to avoid
single cell, microscopy-based assays and instead interrogate entire wells. Ideally, cells will be
efficiently transfectable.
One candidate system is a HEK cell line stably expressing NaV1.3 and Kir2.1 (‘HEK-NK’) (51).
These cells exhibit spontaneous activity of AP-like spikes when grown to confluency, and can be
induced to fire by field stimulation (Fig. 11).
Alternatively we will generate a HEK cell line expressing ChR2 or one of its red-shifted variants
(e.g. Chrimson, ReaChR). Such a line would permit simultaneous optical control and readout of
22
Figure 9. GCaMP6 transgenic mice.
(A) Epifluorescence image of coronal brain section (bregma -4.3 mm) from a Camk2a-GCaMP6f transgenic mouse (founder line GP7.7).
Unstained 50-µm section. (B) Thy1-GCaMP6f-WPRE (founder line GP5.17). (C) Thy1-GCaMP6f-WPRE (founder line GP5.5). Look up
table with fluorescence signal scale (arbitrary units). (D) Confocal image of somatosensory cortex and hippocampus from a Camk2aGCaMP6f transgenic mouse (founder line GP7.7). (E) Thy1-GCaMP6f-WPRE (founder line GP5.17). (F) GCaMP fluorescence level per
neuron measured from confocal optical sections of transgenic mice and AAV-transduced mice from motor (blue bars), barrel (black bars),
or visual cortex (red bars). Quantification of cortical layer 2/3 (left graph). Quantification of cortical layer 5 (right graph). (G) Functional
imaging of GCaMP6f in visual cortical neurons in Thy1-GCaMP6f-WPRE (founder line GP5.17) transgenic mice. Example traces from a
neuron expressing GCaMP6s. Single sweeps (gray) and averages of sweeps (red) are overlaid. Directions of grating motion (8 directions)
are shown above traces (arrows). Layer 2/3 neuron (left graph). Layer 5 neuron (right graph).
23
membrane potential. The choice between these systems will depend on the consistency of
voltage control.
A factor limiting the throughput of our current neuron-based imaging assay is that each well is
interrogated in series by high-resolution imaging. The high-throughput GEVI assay will measure
fluorescence in all 96 wells at once (52), increasing throughput by two orders of magnitude, at
the expense of single cell resolution. This will allow screening 100,000 variants per year.
A secondary component of screening will be measurement of surface expression of voltage
sensors using an antibody assay and varying permeabilization conditions. Because of the time
and expense involved in such an assay, this method will be applied to a subset of sensors
identified in the activity-based screen with improved voltage sensing. After identifying residues
in GEVIs that improve voltage signals, systematic changes in the signal sequence involved in
membrane targeting will be examined to maximize membrane insertion.
Screening libraries of sensor variants generated by large-scale site-directed mutagenesis. To
date, testing of GEVI scaffolds has been done only with modest numbers of variants (dozens).
Therefore it is not possible to predict which existing scaffold holds the greatest promise for
improvements with large-scale engineering and screening. Based on existing data we propose to
divide our screening resources roughly as follows: 40% ASAP1, 40% ArcLight, 20% other
scaffolds, including new scaffolds. Thus our main effort will be to improve existing scaffolds,
and a smaller but significant effort will be dedicated to demonstrating feasibility and function of
fundamentally new GEVI scaffolds that hold promise for vast performance improvements.
Selection of variants during screening will be based on several criteria to generate significant
performance improvements in optical imaging of neuronal membrane potential with improved
sensitivity as well as spatial and temporal resolution. Performance metrics that will be measured
are sensitivity (amplitude and SNR), linearity, and speed. Secondary screens will also probe
capacitative loading by the sensor gating charge.
Developing GEVIs based on novel scaffolds. Despite more than a decade of work, GEVIs are far
less mature than GECIs. The mechanisms coupling voltage change to fluorescence change are
unknown. There is little structural information that guides the paths for optimization. It is likely
24
that the best GEVI scaffolds in the long run may not currently exist. Use of FP chromophores (as
opposed to retinal in Arch, etc.) ensures that probes are bright. The voltage paddle of the
ArcLight/VSFP family may not be an ideal voltage sensor for several reasons including
capacitative loading. We will develop additional scaffolds, including direct placement of the
GFP chromophore into the membrane, through targeted modification of the surface to increase
lipophilicity and membrane intercalation. Potential designs include "stapling" FPs between
transmembrane helices alongside specific intra-cellular and extra-cellular targeting domains;
inserting FPs into the middle of pore-forming proteins, such as has been done inside designed
cavitated chaperonins (53); or by reengineering FP surface residues to be more hydrophobic
and/or be targeted for fatty acid attachment (the reverse of this approach has been used
successfully to reengineer G-protein coupled receptors (obligate membrane proteins) to become
soluble proteins for high-throughput drug screening and crystallization (14).
Our goal is to develop a set of voltage sensors that reliably detect a single AP and have a range
of decay kinetics (1, 10, 100 milliseconds) suitable for a variety of experimental situations. The
goals of the project will be met when we have indicators that can measure AP trains in
populations of neurons in the intact mouse and Drosophila brain, and that can be used to
measure subthreshold electrical activity in dendritic arbors of individual neurons.
Milestones
For FY2015, we will produce a standardized benchmarking and screening assay, and the
performance of current voltage indicators (Arch, ArcLight, ASAP, VSFP) will be compared. For
FY2016, optimized voltage indicator variants will be made that can detect subthreshold activity
of hundreds of neurons in vivo. Improved voltage sensor variants will be engineered for high
fidelity tracking of neuronal spikes at 100 Hz in vivo.
4.2.7. Transgenic/targeted mice
We will optimize the Rosa26-CAG-lox-STOP-lox-GCaMP6s-WPRE expression cassette to
increase expression levels. Linker sequences between the CAG promoter and sensor start codon
will be re-engineered to allow for more efficient transcription and/or translation. Thousands of
constructs with altered linker sequences will be tested for expression level in transient
25
Figure 10. Performance of GEVIs in
HEK cells.
(A) Topology of ArcLight-Q239 protein
(19) at membrane with intracellular FP. (B)
Topology of ASAP1 protein (18) at
membrane with extracellular FP. (C)
Voltage clamp of HEK293T cell
expressing ArcLight-Q239. Voltage steps
(-100 to 40 mV from -60 mV holding
potential; top) and transmembrane currents
(bottom). (D) Voltage clamp of HEK293T
cell expressing ASAP1, with same
conditions as C. (E) ArcLight-Q239
fluorescence responses to voltage steps
from -60 mV holding potential
(fluorescence traces, top; ∆F/F0 traces after
correction for bleaching, bottom). (F)
ASAP1 fluorescence responses to voltage
steps, under same conditions as E. (G)
ArcLight-Q239 fluorescence changes
versus voltage steps from -60 mV holding
potential. (H) ASAP1 fluorescence
changes versus voltage steps. (I) ∆F/F0
responses to voltage steps from -60 to 40
mV from HEK cells expressing either
ArcLight-Q239 (blue, n=4) or ASAP1
(green, n=5). (J) Kinetics for the fast
component of the rise (τ-on) and decay (τoff) time for GEVIs. (K) SNR calculated as
the ratio of the peak fluorescence at 40 mV
divided by the standard deviation of the
resting fluorescence at -60 mV after
background subtraction. (L) Bleaching
measured as the percentage loss of
fluorescence after illumination for 2 s as in
E and F.
26
transfection assays in HEK293T tissue culture cells by epifluorescence imaging and FACS. A
Cre recombinase-expressing construct and co-expressed RFP control (to normalize for
transfection level) will be transfected transiently. The Rosa26-CAG variants with the highest
fluorescence will be selected for use in gene targeting and mouse production for next generation
GCaMPs, RCaMPs, and R-GECOs. In the absence of an improved strategy, we will continue
using Thy1 transgenic expression cassettes and selecting for founder lines with suitable
expression patterns and levels. We will also continue using improved expression systems (e.g.
TIGRE insulator, Tet-regulatable, lox-gated system (54)) in collaboration with other laboratories.
Milestones
For FY2015, thousands of Rosa26-CAG-lox-STOP-lox-GCaMP6s-WPRE variants will be
screened in tissue culture cells. For FY2016, Rosa26-targeted mice with optimized cassettes will
be produced with high and stable GCaMP6s, GCaMP6f, RCaMP, and R-GECO expression
levels comparable to AAV-mediated expression in cortical neurons.
4.2.9 Calcium-activated and light-dependent switch prototypes
Our goal is to create proteins that transduce neural activity (i.e. calcium) and a user-defined
stimulus (i.e. light) into activation of a downstream transgene, either through significantly upregulated transcription of the target locus, or through recombination of the locus to become
compatible with transcriptional activation. Several promising scaffolds and approaches will be
tested, including Cre for recombination, tetracycline repressor (TetR) and TAL effectors for
transcriptional activation, and CRISPR/Cas9 for genome editing.
For Cre recombinase, we will pursue alternative topologies to the published split, wild-type
enzyme (41) that we have characterized to date (Fig. 8). Cre must bind to both DNA strands at
both ends of a target locus, so that a split Cre requires the simultaneous assembly of 8 molecular
fragments onto double-stranded DNA to potentially achieve a recombination event. All lightgated, split-Cre designs tested to date have suffered from very low levels of activation. For this
reason, further experimentation with Cre will involve full Cre domains and potentially tandem
dimers, drastically reducing the number of molecules, which should improve efficacy. Calciumdependence will be introduced into Cre recombinase by several strategies: computational
27
Figure 11. Spiking HEK cells.
(A) Schematic of HEK-NK cells that stably express voltage-gated
sodium channel NaV1.3 and the inwardly-rectifying potassium
channel KIR2.1 under drug selection and are electrically-coupled
by endogenous gap junctions (51). (B) Voltage steps (-60 to 0
mV) delivered from a -60mV holding potential activate a slowlyinactivating sodium current. (C) Current steps (-40 to 100 pA)
were applied to an HEK-NK cell. Hyperpolarizing and
subthreshold steps generated a passive response. Steps of 20 pA
or greater evoked spikes with threshold of -30 mV. (D)
Extracellular field stimulation evoked synchronous spikes in a
population of HEK-NK cells transfected with ArcLight-Q239
(Top). One cell was recorded in current clamp mode (Top left,
pipette attached). ROIs were selected from the epifluorescence
image (Top right). (Middle) Field stimuli of 1 ms duration were
delivered at 15 second intervals. Ten consecutive spikes are
shown (red traces). Inset shows synchrony of the spike onset
(black traces). (Bottom) ∆F/F0 responses for 5 cells selected
shown at top, obtained during a single field stimulus trial.
28
calcium-binding-site design into the Cre itself, or by grafting M13/CaM as in the GCaMP,
RCaMP, R-GECO, and CaMPARI scaffolds. Low background levels of enzyme activity will be
retained by both design and directed evolution. We have preliminary data with improved activity
several-fold over the starting molecules (Fig. 8).
Recently, light-gated versions of TAL effectors have been introduced, taking advantage of the
cryptochrome (CRY)/CIB interaction to bring together a TAL DNA-binding domain and a
transcriptional activation motif (55). We will expand on this scaffold by introducing calciumdependence into the gene activation step, similarly through either computational design or by
CaM/M13 grafting. A similar strategy to that used to generate the light-gated TALs may work
for TetR and its derivatives TetOn and TetOFF, wherein the DNA-binding domains and
activator/repressor modules are contained in the same protein. We will also explore the
modification of endogenous mouse/fly/etc. transcription factors to introduce these
characteristics.
Finally, recent advances in understanding of the structure/function of the CRISPR/Cas9
microbial genome-editing enzyme system creates the possibility of engineering in neural
activity-dependence to this tool as well. Such a reagent would be useful for targeted
manipulation of neurons in culture, for instance (the dependence on guide RNA delivery limits in
vivo utility).
Milestones
For FY2015, we will produce assays to test and screen calcium switch scaffolds based on light/calcium-sensitive Cre recombinase and transcription factors. For FY2016, we will demonstrate
prototype light-sensitive Cre recombinase and transcription factors that can switch on ChR2 gene
expression based on activity and light gating in vivo in mice and flies.
4.3 Publication and Reagent Distribution Plan
Our goal is to produce well-characterized and quality-controlled reagents and make them
immediately and widely available to the research community. For example, GCaMP5 was
available at Addgene.org for almost one year before publication of the relevant manuscript (47).
29
4.3.1 Timeline of Reagent Distribution
0-9 Months After Discovery of a New Sensor
Characterization in vitro and in vivo requires at least 9 months (including creation of viral
vectors, transgenic flies, and worms) after the initial discovery. During this time period, reagents
will be made available to key collaborators who are testing and calibrating the sensor using
previously agreed upon procedures and standards; one obvious example would be collaborators
who test in species/systems not represented on the project team. Additionally, laboratories at
JFRC will receive access to the reagents for use at JFRC. Collaborators including JFRC
laboratories will not distribute the reagents and will not publish results using the sensors prior to
publication of the sensors by the project. Our goal is to complete testing within 9 months of
initial characterization in cultured neurons.
Transgenic mouse production will be initiated during this period, but production and testing will
take considerably longer. Reagents will be distributed to other rodent labs if they plan to create
transgenic mice that are not the same as those created at JFRC by the project, mainly to avoid
duplication of effort, and are willing to widely share these mice on a pre-publication basis (i.e.
via deposition at The Jackson Laboratory).
9 Months After Discovery
•
DNA reagents will be deposited in Addgene, Inc.
•
Viruses will be deposited in the University of Pennsylvania Viral Vector Core.
•
Flies will be submitted to the Bloomington Drosophila Stock Center and/or
•
JFRC's shared Drosophila resource after validation.
Transgenic mice will be deposited at The Jackson Laboratory as soon as they have been
characterized, no later than 9 months after germline transmission.
30
Reagents that have been tested, but are not being pursued further, will be advertised on the JFRC
website and made available through Addgene, Inc., or the University of Pennsylvania Viral
Vector Core.
4.7.2 Publication Policies
The generation and characterization of new reagents will be published as a collaboration among
the relevant GENIE members and other contributors at JFRC and elsewhere. The GENIE Project
at JFRC will be listed as the first address, unless an alternative is mutually agreed upon. If the
manuscript is accepted less than 9 months after discovery, all materials will be distributed
without restrictions as described above.
31
5.0 References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
Stosiek C, Garaschuk O, Holthoff K, & Konnerth A (2003) In vivo two-photon calcium
imaging of neuronal networks. Proc Natl Acad Sci U S A 100(12):7319-7324.
Reiff DF, et al. (2005) In vivo performance of genetically encoded indicators of neural
activity in flies. J Neurosci 25(19):4766-4778.
Seelig JD & Jayaraman V (2013) Feature detection and orientation tuning in the
Drosophila central complex. Nature.
Svoboda K, Denk W, Kleinfeld D, & Tank DW (1997) In vivo dendritic calcium
dynamics in neocortical pyramidal neurons. Nature 385:161-165.
Petreanu L, et al. (2012) Activity in motor-sensory projections reveals distributed coding
in somatosensation. Nature 489(7415):299-303.
Cox CL, Denk W, Tank DW, & Svoboda K (2000) Action potentials reliably invade
axonal arbors of rat neocortical neurons. Proc Natl Acad Sci U S A 97(17):9724-9728.
Seelig JD, et al. (2010) Two-photon calcium imaging from head-fixed Drosophila during
optomotor walking behavior. Nat Methods 7(7):535-540.
Haag J & Borst A (2000) Spatial distribution and characteristics of voltage-gated calcium
signals within visual interneurons. Journal of neurophysiology 83(2):1039-1051.
Yuste R & Denk W (1995) Dendritic Spines as Basic Functional Units of Neuronal
Integration. Nature 375(6533):682-684.
Jia H, Rochefort NL, Chen X, & Konnerth A (2010) Dendritic organization of sensory
input to cortical neurons in vivo. Nature 464(7293):1307-1312.
Chen TW, et al. (2013) Ultrasensitive fluorescent proteins for imaging neuronal activity.
Nature 499(7458):295-300.
Akerboom J, et al. (2013) Genetically encoded calcium indicators for multi-color neural
activity imaging and combination with optogenetics. Frontiers in molecular neuroscience
6:2.
Zhao Y, et al. (2011) An expanded palette of genetically encoded Ca(2) indicators.
Science 333(6051):1888-1891.
Perez-Aguilar JM, et al. (2013) A computationally designed water-soluble variant of a Gprotein-coupled receptor: the human mu opioid receptor. PloS one 8(6):e66009.
London M & Hausser M (2005) Dendritic computation. Annu Rev Neurosci 28:503-532.
Grinvald A & Hildesheim R (2004) VSDI: a new era in functional imaging of cortical
dynamics. Nat Rev Neurosci 5(11):874-885.
Knopfel T (2012) Genetically encoded optical indicators for the analysis of neuronal
circuits. Nature reviews. Neuroscience 13(10):687-700.
St-Pierre F, et al. (2014) High-fidelity optical reporting of neuronal electrical activity
with an ultrafast fluorescent voltage sensor. Nature neuroscience.
Jin L, et al. (2012) Single action potentials and subthreshold electrical events imaged in
neurons with a fluorescent protein voltage probe. Neuron 75(5):779-785.
Gong Y, Li JZ, & Schnitzer MJ (2013) Enhanced Archaerhodopsin Fluorescent Protein
Voltage Indicators. PloS one 8(6):e66959.
Wallace DJ & Kerr JN (2010) Chasing the cell assembly. Current opinion in
neurobiology.
Cruz FC, et al. (2013) New technologies for examining the role of neuronal ensembles in
drug addiction and fear. Nature reviews. Neuroscience 14(11):743-754.
32
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
Huber D, et al. (2012) Multiple dynamic representations in the motor cortex during
sensorimotor learning. Nature 484(7395):473-478.
Tian L, et al. (2009) Imaging neural activity in worms, flies and mice with improved
GCaMP calcium indicators. Nat Methods 6(12):875-881.
Zariwala HA, et al. (2012) A Cre-Dependent GCaMP3 Reporter Mouse for Neuronal
Imaging In Vivo. The Journal of neuroscience : the official journal of the Society for
Neuroscience 32(9):3131-3141.
Wardill TJ, et al. (2013) A neuron-based screening platform for optimizing geneticallyencoded calcium indicators. PloS one 8(10):e77728.
Gibson DG, et al. (2009) Enzymatic assembly of DNA molecules up to several hundred
kilobases. Nat Methods 6(5):343-345.
Ding J, Luo AF, Hu L, Wang D, & Shao F (2014) Structural basis of the ultrasensitive
calcium indicator GCaMP6. Science China. Life sciences 57(3):269-274.
Akerboom J, et al. (2009) Crystal structures of the GCaMP calcium sensor reveal the
mechanism of fluorescence signal change and aid rational design. J Biol Chem
284(10):6455-6464.
Wang Q, Shui B, Kotlikoff MI, & Sondermann H (2008) Structural basis for calcium
sensing by GCaMP2. Structure 16(12):1817-1827.
Jayaraman V & Laurent G (2007) Evaluating a genetically encoded optical sensor of
neural activity using electrophysiology in intact adult fruit flies. Front Neural Circuits
1:3.
Egelhaaf M & Warzecha AK (1999) Encoding of motion in real time by the fly visual
system. Current opinion in neurobiology 9(4):454-460.
Thestrup T, et al. (2014) Optimized ratiometric calcium sensors for functional in vivo
imaging of neurons and T lymphocytes. Nat Methods 11(2):175-182.
Direnberger T, et al. (2012) Biocompatibility of a genetically encoded calcium indicator
in a transgenic mouse model. Nat Commun 3.
Kredel S, et al. (2009) mRuby, a Bright Monomeric Red Fluorescent Protein for Labeling
of Subcellular Structures. PloS one 4(2).
Shaner NC, et al. (2008) Improving the photostability of bright monomeric orange and
red fluorescent proteins. Nat Methods 5(6):545-551.
Mutze J, et al. (2012) Excitation spectra and brightness optimization of two-photon
excited probes. Biophysical journal 102(4):934-944.
Kralj JM, Douglass AD, Hochbaum DR, Maclaurin D, & Cohen AE (2012) Optical
recording of action potentials in mammalian neurons using a microbial rhodopsin. Nat
Methods 9(1):90-95.
Gong Y, Wagner MJ, Zhong Li J, & Schnitzer MJ (2014) Imaging neural spiking in brain
tissue using FRET-opsin protein voltage sensors. Nat Commun 5:3674.
McKinney SA, Murphy CS, Hazelwood KL, Davidson MW, & Looger LL (2009) A
bright and photostable photoconvertible fluorescent protein. Nat Methods 6(2):131-133.
Kennedy MJ, et al. (2010) Rapid blue-light-mediated induction of protein interactions in
living cells. Nat Methods.
Caroni P (1997) Overexpression of growth-associated proteins in the neurons of adult
transgenic mice. J Neurosci Methods 71(1):3-9.
Tsien JZ, et al. (1996) Subregion- and cell type-restricted gene knockout in mouse brain.
Cell 87(7):1317-1326.
33
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
Yasuda R, et al. (2004) Imaging calcium concentration dynamics in small neuronal
compartments. Science's STKE : signal transduction knowledge environment
2004(219):pl5.
Sankaranarayanan S, De Angelis D, Rothman JE, & Ryan TA (2000) The use of
pHluorins for optical measurements of presynaptic activity. Biophysical journal
79(4):2199-2208.
Wilt BA, Fitzgerald JE, & Schnitzer MJ (2013) Photon shot noise limits on optical
detection of neuronal spikes and estimation of spike timing. Biophysical journal
104(1):51-62.
Akerboom J, et al. (2012) Optimization of a GCaMP Calcium Indicator for Neural
Activity Imaging. The Journal of neuroscience : the official journal of the Society for
Neuroscience 32(40):13819-13840.
Sun XNR, et al. (2013) Fast GCaMPs for improved tracking of neuronal activity. Nat
Commun 4.
Akemann W, Lundby A, Mutoh H, & Knopfel T (2009) Effect of voltage sensitive
fluorescent proteins on neuronal excitability. Biophysical journal 96(10):3959-3976.
Tsutsui H, Karasawa S, Okamura Y, & Miyawaki A (2008) Improving membrane voltage
measurements using FRET with new fluorescent proteins. Nat Methods 5(8):683-685.
Park J, et al. (2013) Screening fluorescent voltage indicators with spontaneously spiking
HEK cells. PloS one 8(12):e85221.
Hempel CM, et al. (2011) A system for performing high throughput assays of synaptic
function. PloS one 6(10):e25999.
Paavola CD, et al. (2006) A versatile platform for nanotechnology based on circular
permutation of a chaperonin protein. Nanotechnology 17(5):1171-1176.
Zeng HK, et al. (2008) An inducible and reversible mouse genetic rescue system. Plos
Genet 4(5).
Liu JY, et al. (2012) Efficient and Specific Modifications of the Drosophila Genome by
Means of an Easy TALEN Strategy. J Genet Genomics 39(5):209-215.
Nakai J, Ohkura M, & Imoto K (2001) A high signal-to-noise Ca(2+) probe composed of
a single green fluorescent protein. Nat Biotechnol 19(2):137-141.
Pologruto TA, Yasuda R, & Svoboda K (2004) Monitoring neural activity and [Ca2+]
with genetically encoded Ca2+ indicators. J Neurosci 24(43):9572-9579.
Ohkura M, Matsuzaki M, Kasai H, Imoto K, & Nakai J (2005) Genetically encoded
bright Ca2+ probe applicable for dynamic Ca2+ imaging of dendritic spines. Anal Chem
77(18):5861-5869.
Tallini YN, et al. (2006) Imaging cellular signals in the heart in vivo: Cardiac expression
of the high-signal Ca2+ indicator GCaMP2. Proc Natl Acad Sci U S A 103(12):47534758.
Mao T, O'Connor DH, Scheuss V, Nakai J, & Svoboda K (2008) Characterization and
subcellular targeting of GCaMP-type genetically-encoded calcium indicators. PloS one
3(3):e1796.
34