High-throughput-compatible assays using a genetically-encoded calcium indicator

Measurement of intracellular calcium in live cells is a key component of a wide range of basic life science research, and crucial for many high-throughput assays used in modern drug discovery. Synthetic calcium indicators have become the industry standard, due their ease of use, high reliability, wide dynamic range, and availability of a large variety of spectral and chemical properties. Genetically-encoded calcium indicators (GECIs) have been optimized to the point where their performance rivals that of synthetic calcium indicators in many applications. Stable expression of a GECI has distinct advantages over synthetic calcium indicators in terms of reagent cost and simplification of the assay process. We generated a clonal cell line constitutively expressing GCaMP6s; high expression of the GECI was driven by coupling to a blasticidin resistance gene with a self-cleaving cis-acting hydrolase element (CHYSEL) 2A peptide. Here, we compared the performance of the GECI GCaMP6s to the synthetic calcium indicator fluo-4 in a variety of assay formats. We demonstrate that the pharmacology of ion channel and GPCR ligands as determined using the two indicators is highly similar, and that GCaMP6s is viable as a direct replacement for a synthetic calcium indicator.


Results
Transient transfection of GCaMP6s. GCaMP6s and fluo-4 have similar excitation and emission spectra (see Table 1), allowing direct comparison using fluorescein filter sets. For our initial characterization of GCaMP6s, we used flow cytometry to compare the fluorescence intensity of 293-F cells labelled with fluo-4 to cells 24 hours after transfecting with the pGP-CMV-GCaMP6s 28 construct (Figure 1a-c). Identical settings for  28 for GCaMP6s and by Gee et al. 57 for fluo-4, in in vitro (cell-free) environments under saturating calcium conditions at neutral pH. Parameters are λ max (ex): wavelength of fluorescence excitation maximum; λ max (em): wavelength of emission maximum; ε max : absorption coefficient at the excitation maximum; QY: fluorescence quantum yield; EC 50 : midpoint of calcium binding curve; Hill slope for the calcium binding curve; k off : off-rate for calcium unbinding; DR: fluorescence dynamic range from zero to saturating calcium concentration. a measured for fluo-3 58 . www.nature.com/scientificreports www.nature.com/scientificreports/ cells tested in buffer was well-fitted by a single gaussian distribution with a center at 6.2 AFU. Fluo-4-labelled cells treated with ionomycin, a calcium ionophore which allows free entry of calcium, showed a single peak at 410.0 AFU. 293-F cells transiently transfected with pGP-CMV-GCaMP6s and tested in buffer showed a peak centered at 44.5 AFU. When treated with ionomycin, the histogram of the GCaMP6s-transfected cell population showed a peak at 133.3 AFU.
With the goal of ensuring long-term expression stability of GCaMP6s in a clonal cell line, we generated an expression plasmid in which the drug selection gene is directly coupled to the GCaMP6s sequence, using a spontaneously-cleaving 2A linker. Figure 1d shows a schematic representation of the GCaMP6s-P2A-Bsr construct. GCaMP6s is a circularly-permuted GFP variant with an M13 domain on the N-terminus, and a modified calmodulin (CaM) domain on the C-terminus. As the protein is being expressed, residues from the P2A linker are added to the C-terminus of GCaMP6s. At the end of the P2A sequence, interactions between the emerging protein interact with the ribosome, preventing the formation of the terminal gly-pro peptide bond 32 . Translation continues into the Bsr sequence despite the failure of this bond to form. The end result is the expression of two completely separate proteins, with a 19-amino acid fragment appended to the C-terminus of the GCaMP6s and an extra proline at the N-terminus of the Bsr protein. Bsr functions as a tetramer; this system drives the expression of four GCaMP6s subunits for each functional Bsr complex.
We tested this construct using a calcium flux assay based on a clonal cell line stably expressing a fusion of human GluA1o with human TARP-γ4 in 293-F cells. The α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionic acid (AMPA) subtype of ionotropic glutamate receptors are ligand-gated ion channels that mediate the majority of fast synaptic transmission within the mammalian brain. AMPA receptors (AMPARs) comprise tetramers of pore-forming GluA subunits in complex with a variety of accessory proteins. Although AMPARs lacking RNA-edited GluA2 subunits are calcium-permeant, homotetramers of GluA1o desensitize almost completely after a few milliseconds of exposure to agonist. Since FLIPR assays typically operate on the timescale of 1-100 seconds, cells expressing GluA1o alone do not produce detectable responses. Co-expression of members of the transmembrane AMPAR regulatory protein (TARP) family, including TARP-γ4, reduces desensitization sufficiently to allow measurable calcium flux. Figure 1f shows fluorescence as a function of time as measured in FLIPR for control wells, using the synthetic calcium indicator fluo-4. Wells stimulated with 15 µM glutamate alone showed a steady-state response of F 2 /F 0 = 1.53 ± 0.05 (mean ± SD, n = 128 wells). To estimate the full effective dynamic range of the indicator, we measured the fluorescence in near-saturating calcium concentrations using an addition of 15 µM glutamate plus 10 µM LY-395153, a positive allosteric modulator (PAM) which prevents desensitization of the AMPA receptors. Under these conditions the fluorescence rose to a maximum of F ∞ /F 0 = 4.04 ± 0.10, followed by a decline to a steady-state response of F/F 0 = 3.47 ± 0.08 (mean ± SD, n = 128 wells). The fluorescence under full inhibition of the AMPA receptor was measured using a simultaneous addition of 15 µM glutamate plus 50 µM CP-465022. After a small, transient rise, the fluorescence dropped to a steady-state response of F 1 / F 0 = 0.89 ± 0.01 (mean ± SD, n = 128 wells).
We performed a transient transfection of GCaMP6s-P2A-Bsr in the GluA1o-γ4 cell line. Figure 1e shows the time course of fluorescence for control wells using these cells, using the same format as for GluA1o-γ4 cells labelled with fluo-4. Wells stimulated with 15 µM glutamate alone had a steady-state response of F 2 / F 0 = 1.87 ± 0.05 (mean ± SD, n = 128 wells). The working dynamic range of the indicator, using an addition of 15 µM glutamate plus 10 µM LY-395153, was F ∞ /F 0 = 3.41 ± 0.34 (mean ± SD, n = 128 wells). The fluorescence under full inhibition of the AMPA receptor was F 1 /F 0 = 0.70 ± 0.07 (mean ± SD, n = 128 wells).
Stable expression of GCaMP6s. We placed under blasticidin selection an aliquot of the GluA1o-γ4 cells that had been transfected with GCaMP6s-P2A-Bsr, and isolated single cells to select for a clonal cell line stably expressing both GluA1o-γ4 and GCaMP6s (designated 293F.GluA1o-γ4.GCaMP6s.blast). We also generated a clonal cell line in wild-type 293-F by transfecting with GCaMP6s-P2A-Bsr, followed by selection with blasticidin and isolation of single cells; this cell line is designated 293F.GCaMP6s.blast.
AMPA receptors. We compared the cellular and subcellular distributions of stably-expressed GCaMP6s in 293F.GluA1o-γ4.GCaMP6s.blast cells to the parental 293F.GluA1o-γ4 cells labelled with fluo-4, using confocal microscopy. Figure 2a shows representative images of 293F.GluA1o-γ4.GCaMP6s.blast cells. These cells were also labelled with the nuclear stain Hoechst 33342 (blue). Laser power and detector settings were identical for all of the images in Fig. 2a, so these images are directly comparable. In the presence of 10 μM CP-465022 (used to fully inhibit calcium flux due to residual glutamate), GCaMP6s fluorescence was uniformly dim. Dead cells, identified by morphology, were relatively sparse and showed no GCaMP6s labelling. In the presence of 15 µM glutamate plus 10 µM LY-395153 or 10 μM ionomycin, most cells show bright, uniform cytoplasmic fluorescence with minimal punctate staining. Cell nuclei appear dimmer than the cytoplasm, suggesting nuclear exclusion of GCaMP6s. This phenomenon is more clearly seen in the expanded views and vertical projections of individual cells in Fig. 2b. There is very little overlap of green and blue staining in the vertical projections. The cells used for these images had been in continuous culture for six months and had therefore undergone ~180 cell divisions. Figure 2d shows representative images of 293F.GluA1o-γ4 cells labelled with fluo-4. Laser power and detector settings were reduced relative to the cells labelled with GCaMP6s, as the fluorescence intensity for the fluo-4 labelled cells was substantially brighter (see cytometry results for quantification). Similar to the cells expressing GCaMP6s, cells labelled with fluo-4 showed very dim green fluorescence in the presence of CP-465022 and bright cytoplasmic fluorescence in the presence of glutamate/LY-395153 or ionomycin. Unlike GCaMP6s, fluo-4 is present in the cell nuclei and shows increased intensity in other organelles. While some vacuoles in the fluo-4-labelled cells appear to lack green fluorescence, the nuclei appear to be co-labelled with fluo-4 and Hoechst 33342. In the expanded views and vertical projections in Fig. 2e, green fluorescence clearly overlaps with the nuclear staining.
Histograms of the logarithmic fluorescence intensity of individual cells as determined by flow cytometry are shown in Fig. 2c for GCaMP6s and Fig. 2f for fluo-4. Flow cytometer settings were the same as for the measurements shown in Figure 1a-c, allowing for direct comparison of the intensities. GCaMP6s cells treated with ionomycin show a single peak at 139 AFU, indicating that essentially the entire population is uniformly labelled. For treated with CP-465022, two peaks are visible; based on gaussian fits, these have centers at 7.9 AFU (75% of the population) and 77.2 AFU (25% of the population). Two peaks are also visible for cells treated with glutamate/ LY-395103, with centers at 18.4 AFU (18% of the population) and 127.6 AFU (82% of the population). Because all cells were uniformly labelled, the pharmacological treatments indicate that when all cells are averaged in a FLIPR-type experiment, 18% of the cells lack sufficient receptor to contribute to a response; an additional 25% of the cells will contribute to background fluorescence with reduced dynamic range available. Thus, only 57% of the cells from this clone can contribute fully to an agonist response. www.nature.com/scientificreports www.nature.com/scientificreports/ Fluo-4-labelled cells treated with ionomycin showed a single peak at 1247 AFU containing 90% of the population; 10% of the cells had fluorescence intensities less than 300 AFU in a broad featureless distribution. Fluo-4-labelled cells treated with CP-465022 show two peaks, with centers at 71 AFU (87% of the population) and 257 AFU (13% of the population). Two peaks are also visible for cells treated with glutamate/LY-395103, with centers at 119 AFU (25% of the population) and 552 AFU (75% of the population). Similar to the situation for the cells labelled with GCaMP6s, only approximately 62% of the cells from this clone can contribute fully to an agonist response.
The cytometry experiments allow direct comparison of the fluorescence intensities between the two indicators. FLIPR-type experiments measure the total fluorescence intensity of 10,000-20,000 cells, which can be emulated by the arithmetic mean in a cytometry run of ~10,000 cells. A summary of multiple cytometry measurements is shown in Table 2. The ionomycin treatment indicates that the total fluorescence output of GCaMP6s-labelled cells was 6.8X lower than cells labelled with fluo-4; the fluorescence in resting calcium conditions (cells treated with CP-465022) was 3.0X lower in GCaMP6s-labelled cells than in cells labelled with fluo-4. Although the fluorescence output of GCaMP6s-labelled cells is substantially dimmer compared to fluo-4, they are sufficient to give robust responses in the FLIPR experiments (compare Fig. 3a,b).
We measured the apparent potency of a panel of AMPA receptor agonists, antagonists, and PAMs to compare the performance of stably-expressed GCaMP6s in 293F.GluA1o-γ4.GCaMP6s.blast cells, to the parental 293F. GluA1o-γ4 cells labelled with fluo-4. Using single-addition assays, we monitored the fluorescence in a 384-well plate format as cells were exposed to a range of concentrations of test compounds generated by serial dilution. The assay buffer contained 4 mM calcium. For antagonists and PAMs, test compounds were pre-incubated for 60 minutes prior to the assay, and the cells were challenged with 15 µM glutamate. All compounds were tested in duplicate. Time-courses for the control wells are shown in Fig. 3a for cells stably expressing GCaMP6s, and Fig. 3b for cells labelled with fluo-4. The time-course and magnitude of fluorescence response to agonist were similar for the two calcium indicators. The responses in wells treated with positive antagonist controls (red), positive PAM controls (blue), and negative control wells (black) showed modest scatter and excellent separation for both indicators. The PAM positive control wells were treated with 10 μM LY-395153, which prevents desensitization of AMPA receptors; the intracellular calcium in these wells is expected to approach saturation for the calcium indicators.
The normalized responses as functions of concentration for representative agonists, antagonists, and PAMs are shown in Fig. 3c-h. Three separate concentration-response experiments were performed using cells labelled with fluo-4, or with stably-expressed GCaMP6s. The mean potency of n = 3 measurements of the compounds in each assay format are shown in Table 3. Statistics describing the assay performance for individual plates are summarized in Table 4.
Muscarinic acetylcholine receptors. Muscarinic acetylcholine receptors (mAChRs) are G-protein coupled receptors (GPCRs) with widespread tissue distribution and diverse physiological functions. G q/11 -coupled mAChRs are expressed endogenously in HEK293 cells 33,34 , and serve as a convenient test system for intracellular calcium indicators. We measured the potencies of a panel of mAChR ligands in the 293F.GCaMP6s.blast cell line stably expressing GCaMP6s, compared to wild-type 293-F cells labelled with the synthetic indicator fluo-4. The assays were performed in 384-well plates in a single-addition assay format in FLIPR Tetra ® ; assay buffer contained 2 mM calcium. Test compounds were prepared in a serial dilution format, and each compound was tested in duplicate. For agonists, fluorescence was monitored as the test compounds were added. For antagonists, test compounds were incubated with the cells for 30 minutes prior to the assay, and fluorescence was monitored as the cells were challenged with acetylcholine.
Time-courses for the control wells and normalized responses as functions of concentration for representative agonists and antagonists are shown in Fig. 4a,b. The time-course and magnitude of fluorescence response to agonist were similar for the two calcium indicators. The responses in wells treated with positive antagonist controls (red) and negative control wells (black) show modest scatter and excellent separation for both indicators. Because the muscarinic receptors are GPCRs and respond via calcium release from internal stores, the degree of saturation of the calcium indicator could not be determined directly through activation of the receptor. Therefore, we also included an additional set of control wells in which the calcium ionophore ionomycin was added during the first addition to fully saturate the calcium indicators (blue traces).
Three separate concentration-response experiments were performed using cells labelled with fluo-4, or with stably-expressed GCaMP6s. Normalized responses as functions of concentration for representative agonists and antagonists are shown in Fig. 4d-f. The mean potency of n = 3 measurements of the compounds in each assay www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ format are shown in Table 5. Statistics describing the assay performance for individual plates are summarized in Table 4.
TRPV1. TRPV1 is a ligand-gated cation channel, whose activation by exogenous ligands such as capsaicin produce the burning sensation associated with hot peppers. TRPV1 is expressed in the central and peripheral nervous systems and serves a variety of physiological functions including pain perception and body temperature regulation. We used transient transfections of DNA encoding the guinea pig TRPV1 receptor in the 293F. GCaMP6s.blast cell line and in wild-type 293-F cells subsequently labelled with fluo-4, to compare the pharmacology of a panel of TRPV1 ligands. The assays were performed in 384-well plates in a single-addition assay format in FLIPR Tetra ® . For antagonists, a range of concentrations of test compounds generated by serial dilution were pre-incubated on the cells for 60 minutes prior to the assay, and the cells were challenged with 10 µM capsaicin. For agonists, the test compounds were added on-line in place of the 10 µM capsaicin stimulus.
Full activation of TRPV1 results in a large steady-state current, which can lead to saturation of a calcium indicator in physiological saline 35 and subsequent shifts in apparent potency of test compounds 36 . These assays were performed in low calcium (20 µM) to avoid dye saturation. Time-courses for the control wells are shown in Fig. 5a,b. The time-course and magnitude of fluorescence response to agonist are similar for the two calcium indicators. The responses in wells treated with positive antagonist controls (red) and negative control wells (black) show modest scatter and excellent separation for both indicators. To estimate the degree of saturation of the calcium dye, we included control wells of 10 µM capsaicin with 2 mM calcium. Normalized responses as functions of concentration for representative agonists and antagonists are shown in Fig. 5a-f. The mean potency of n = 3 measurements of the compounds in each assay format are shown in Table 6. Statistics describing the assay performance for individual plates are summarized in Table 4.

Discussion
Based on flow cytometry, transient transfection of the genetically-encoded calcium indicator construct pGP-CMV-GCaMP6s into 293-F cells yielded a population of cells with a robust increase in fluorescence in the presence of ionomycin. This plasmid also contains a neomycin selection marker on a separate promoter. Upon selection of stably-expressing cells using G418, the expression levels of GCaMP6s were too low for consideration of selecting a clonal cell line bright enough for further use. Therefore, we coupled the DNA encoding GCaMP6s   www.nature.com/scientificreports www.nature.com/scientificreports/ to a blasticidin resistance gene 29,37 using a spontaneously-cleavable 2A linker 30 . This strategy forces the cell to fully express the target gene before being able to express the resistance marker 38 , enabling the generation of clonal cell lines with sufficient brightness and expression stability to support a high-throughput assay. Blasticidin deaminase is a tetramer of subunits, theoretically requiring the generation of four GCaMP6s proteins per functional enzyme. Further, we expect better long-term expression stability, since a cell cannot gain a competitive advantage by disabling expression of GCaMP6s.
After expression and spontaneous cleavage, fragments of the 2A linker remain on the C-terminus of the first gene (in this case, GCaMP6s) and the N-terminus of the second (the blasticidin resistance gene) 39 . Transient transfection of the GCaMP6s-P2A-Bsr into 293F cells expressing an AMPA receptor construct (GluA1o-γ4) yielded cells with strong, highly-stereotyped responses to glutamate addition and to ionomycin. This www.nature.com/scientificreports www.nature.com/scientificreports/ demonstrates that the 2A linker fragment on the GCaMP6s has a minimal effect on the function of the indicator. The noise level and effective dynamic range of cells labelled with GCaMP6s were similar to those labelled with fluo-4. For both indicators, the fluorescence of the antagonist positive control wells dropped slightly following the compound addition. In this case, agonist and antagonist were added at the same time. Residual glutamate in the bath is sufficient to partially activate the GluA1o-γ4 receptors prior to full agonist challenge, so that the intracellular calcium is already slightly elevated prior to agonist addition. This effect is more pronounced for GCaMP6s, since this indicator has a higher affinity for calcium compared to fluo-4.
While performing transient transfections is relatively simple and straightforward, one of the advantages of a genetically-encoded system is the ability to create stably-expressing clonal cell lines. Such a cell line would continuously manufacture its own indicator, requiring minimal intervention prior to the assay. Clonal cell lines based upon the GCaMP6s-P2A-Bsr construct showed bright, uniform cytoplasmic staining. Functional Blasticidin resistance indicates that the extra N-terminal proline on Bsr from the 2A linker had no major negative impact.
One of the problems experienced with the use of GECIs in in vivo applications is the tendency for accumulation of the indicator in the nucleus of neurons over a weeks-to-months timescale, leading to aberrant behavior of those cells 40 . Treatment of cells with the calcium ionophore ionomycin floods every compartment of the cell, showing the subcellular compartmentalization of the indicator. The 293F.GluA1o-γ4.GCaMP6s.blast cells used for the images in Fig. 2a,c had been in continuous culture for over six months, and showed no indication of nuclear accumulation. This is consistent with the cytoplasmic targeting of the indicator. A likely explanation for the difference between neurons and the 293-F cells is that any nuclear entry of GCaMP6s will be diluted amongst the progeny of the rapidly dividing 293-F cells, whereas terminally differentiated neurons do not undergo nuclear division. In contrast, all synthetic calcium indicators show labelling of intracellular compartments, including the nucleus, to varying degrees 41 .
High throughput-compatible assays, defined here as parallelized semi-automated experiments with the capacity for >10,000 data points per day, can be configured for multiple purposes that include screening libraries of compounds, determination of mechanism-of-action of test compounds, and measurement of functional potency at individual targets. Here, we focused on potency measurements, as this tends to be the most stringent assessment of the accuracy and reproducibility of an assay. We compared the performance of stably-expressed GCaMP6s to exogenously-applied fluo-4 in FLIPR Tetra ® assays, using cells expressing calcium-permeant ion channels or a G q/11 -coupled receptor. The results were compared using multiple criteria: (1) reproducibility of the control responses, quantified by the assay quality parameter z′; (2) assay dynamic range, quantified by the ratio of the responses of the negative controls to the fully-inhibited controls F 2 /F 1 ; (3) effective dynamic range of the dye, quantified by the calcium-saturated fluorescence divided by the fully-inhibited response F ∞ /F 1 ; and (4) comparison of potencies of standard compounds measured by fits to the response as a function of compound concentration.
For each receptor type, the assay quality parameter z′ > 0.5 for all plates tested for both indicators; this suggests high confidence in screening results. The effective dynamic range (maximal divided by minimal fluorescence intensity) ranged from 3.6 to 5.5 for GCaMP6s, similar to the range for fluo-4 (range of 2.7 to 8.0). The assay dynamic range (negative control fluorescence divided by the fluorescence for fully-inhibited controls) was also very similar between the two indicators.
The potencies of a panel of pharmacological agents targeting TRPV1 and the muscarinic receptor were comparable for the two indicators. Taken as a group, agonist potencies for AMPA receptors were 0.6 log units (3.7X) more potent, on average, using fluo-4 compared to GCaMP6s; antagonists were 0.3 log units (2X) less potent. Application of the Black-Leff operational model of agonism 42,43 , including the non-linear response of the calcium indicators as signal transduction elements, implies that the substantially different calcium sensitivities of the two indicators is involved (see Table 1). GCaMP6s has a calcium binding constant in buffer approximately 2.4X more potent than fluo-4, with a relatively steep Hill slope due to the tetravalent nature of the calmodulin. Although a full treatment is beyond the scope of this work, preliminary modeling suggests that the apparent potency measured by the two indicators under similar conditions can differ by 1-to 5-fold, and that agonists and antagonists will have shifts in opposite directions. Improved accuracy could potentially be achieved by careful calibration of the dyes, converting fluorescence measurements to intracellular calcium concentrations prior to the fitting process 36 . www.nature.com/scientificreports www.nature.com/scientificreports/ The working dynamic range (measured as the maximal fluorescence achieved with near-saturating calcium divided by the level of fluorescence for a fully-inhibited receptor) was similar between cells labelled with GCaMP6s compared to those labelled with the synthetic calcium indicator, fluo-4 (Table 4). These dynamic ranges are substantially smaller than values reported under ideal conditions (see Table 1). Working dynamic range can be degraded from ideal by several factors, including background fluorescence of the cells and the multiwell plates, and the non-zero intracellular calcium levels at the beginning of the assay. With intracellular calcium levels measured at rest in HEK293 cells near 100 nM (see, for example, Patel et al. 44 ), both indicators would be expected to have a non-zero basal fluorescence. While there are no established criteria for the minimal dynamic www.nature.com/scientificreports www.nature.com/scientificreports/ range required for a successful assay, these effective dynamic ranges were sufficiently large to enable acceptable assay quality for both indicators.
GECIs have numerous advantages relative to synthetic calcium indicators with respect to their use in high-throughput pharmacology assays, including the following. (1) Simplification of the assay procedure. Because the cells manufacture the indicator, exogenous application is unnecessary. Examples of typical workflow for the two types of indicators are shown in Fig. 6. In an HTS environment, every additional step decreases throughput and adds noise to the final result. (2) Reduced reagent cost. Having made the initial investment in creating a stable GECI cell line, no additional costs are involved with labelling cells. (3) While synthetic indicators must be created and purified on a large scale by multi-step chemical processes, GECIs can be modified using standard molecular biology techniques to add or modify many of their properties. (4) The GECI sequence can be fused to other proteins with breakable or permanent linkers. In this work, we used a self-cleaving linker to ensure near-stoichiometric expression of GCaMP6s to an expression marker. Many other linker types are available, which could prove useful in certain assay types. For example, we could potentially increase the sensitivity of the probe by linking the GCaMP6s directly to a target protein, thereby concentrating the probe in close proximity to the site of action. (5) Targeting sequences can be added to direct the probe to specific subcellular compartments. In this work, we had no targeting sequence so the probe was directed to the cytoplasm. In contrast, synthetic chemical indicators label subcellular compartments to varying degrees. This means that the fluorescence output is generally a mixture of responses from these different compartments. (6) GECIs are highly modular in nature, with well-defined and separable motifs for calcium binding, fluorescence properties, cellular targeting, and attachment to other proteins. This enables a near-infinite capacity for directed tuning of the properties to suit various needs.  Table 6. Comparison of pharmacology for test compounds in the TRPV1 intracellular calcium assays. Data are expressed as the mean ± SEM of n = 3 individual concentration-response experiments. Steps marked 'optional' may or may not be required, depending upon the specific requirements for the target of interest.
A large set of spectral properties and calcium affinities have already been discovered and characterized 8,26,45 . One example of the power of these genetic tools was the development of the calcium-measuring organelle-entrapped protein indicators (CEPIA) 23 . GCaMP2 was systematically modified to (a) add signals for ER targeting and retention, (b) alter the fluorophore to generate variations in emission/excitation spectra, and (c) reduce calcium affinity to match the higher range of calcium concentrations inside the ER. One of these indicators (R-CEPIAer) was subsequently employed in a high-throughput screening assay 24 . (7) This modularity also allows for new discoveries to be readily incorporated into existing constructs. For example, the fluorophore for GCaMP6s could easily be replaced with a red-or blue-shifted fluorophore with a reasonable expectation that the rest of the functionality of the probe would remain intact. (8) Massive libraries of mutations of an existing probe are easily made using directed or random mutagenesis to adjust functional or expression properties. The GCaMP6s platform as described here has the potential for introduction of complications that may need to be addressed, including the following. (1) All GECIs are calcium chelators, and for fluorescence applications, need to be expressed at high levels. In some cases, constitutive expression of GECIs has been observed to have deleterious effects on cell metabolism and viability in neurons and cardiomyocytes in vivo 26,46 ; this appears to be correlated with high expression levels and the loss of nuclear exclusion of the indicator (reviewed by Rose et al. 26 ). Our GCaMP6s lines have been in continuous culture for 6 months with no apparent deterioration in cell health, rate of cell division, or fluorescence of the indicator. While we did not observe any issues with cell viability in these experiments, certain cell types may be more vulnerable. Such issues could be mitigated using an inducible promoter on the GECI, allowing on-demand expression of the indicator. (2) The calcium sensor of GCaMP6s is based upon calmodulin, which interacts with a wide variety of cellular targets. Overexpression of the modified CaM could potentially alter function or pharmacology of certain targets 47 . Fortunately, the importance of intracellular calcium signaling has driven the evolution of a large variety of calcium-binding proteins 48 . To date, all GECIs described in the literature are based on variants of either CaM or troponin C; development of GECIs using alternate calcium-binding motifs may provide additional advantages. (3) CaM comprises four calcium binding sites that each contribute to folding the protein into its activated state. This results in a steeper dependence of the fluorescence output over a narrower range of calcium concentrations compared to synthetic calcium indicators, which generally comprise a single calcium binding site. The Hill slope of calcium binding to CaM itself is approximately 3; mutagenesis has produced CaM-based GECIs with slopes ranging from 0.7-3.8 7 . Pharmacological concentration-response curves tend to be steeper for GCaMP6s compared to fluo-4. The relatively narrow range of useful calcium concentrations requires careful tuning of the assay conditions to avoid saturation of the indicator, and subsequent loss of pharmacological sensitivity. Future improvements to the platform could include linearizing the calcium-fluorescence curves by careful adjustment of the binding site affinities 6 or by reducing the number of calcium binding sites 19,49 . (4) The maximal fluorescence output of 293-F cells stably expressing GCaMP6s was approximately 7X lower than 293-F cells stained with fluo-4. While the reduced signal intensity did not affect the performance of the assays described here, stable expression of GCaMP6s may provide insufficient signal in certain applications that require maximal light output. However, part of the increased fluorescence for cells labelled with fluo-4 was due to labelling of the nucleus and other intracellular organelles, which was absent in the GCaMP6s-expressing cells.
In summary, these results demonstrate that GECIs can be used as a replacement for synthetic calcium indicators in assay formats compatible with high-throughput screening. The dynamic range, assay reproducibility, and accuracy of potency measurements are comparable between GCaMP6s and fluo-4. GECIs have significant advantages over chemical indicators, including reduced cost of reagents and simplification of the assay procedure. By linking the expression of the GECI to the blasticidin selection marker using the cleavable P2A linker, we have ensured a high level of stable expression in 293-F clonal cell lines, enabling the use of these cell lines as platforms for any desired intracellular calcium assay.

Methods
GCaMP6s expression construct. The construct designated pGP-CMV-GCaMP6s 28 was obtained from Douglas Kim (Addgene plasmid # 40753). The GCaMP6s open reading frame (ORF) in pGP-CMV-GCaMP6s contained an N-terminal sequence: MGSHHHHHHGMASMTGGQQMGRDLYDDDDKDLAT. This peptide contained a 6His tag, a T7 bacteriophage gene10 tag, and an XPRESS tag (tags underlined). We used polymerase chain reaction (PCR) with KOD DNA Polymerase (EMD Millipore, Darmstadt, Germany) to isolate the GCaMP6s ORF, remove the N-terminal tags, and append a BamHI site on to the C-terminus that codes for a gly-ser linker. A Kozak was added to the GCaMP6s ORF N-terminus, and the PCR product inserted into a proprietary vector driven by the CMV IE-1 promoter. The GCaMP6s ORF was fused in-frame with the BamHI gly-ser linker to the Porcine Teschovirus-1 2A (P2A) CHYSEL peptide 30 (GATNFSLLKQAGDVEENPGP) which was further fused in frame to the Bacillus cereus blasticidin-resistance gene 29 (Bsr, InvivoGen, San Diego, CA). The complete GCaMP6s-P2A-Bsr ORF was sequence-confirmed in full across the ORF (data not shown). The construct was expanded in TOP10 E. coli (Thermo Fisher, Carlsbad, CA) and purified with the Qiagen Mega Prep Kit (Qiagen, Valencia, CA).
GluA1o-CACNG4 fusion protein expression construct. cDNAs for human GluA1o (Uniprot accession number P42261-1) and human TARP-γ4 (Uniprot accession number Q9UBN1) were PCR-amplified from a human brain cDNA library. To ensure a 1:1 stoichiometry of GluA1o and TARP-γ4 in the expressed channel, we fused the cDNA encoding the C-terminus of GluA1o to the cDNA encoding the N-terminus of TARP-γ4 by inserting a linker sequence encoding QQQQQQQQQQEFAT between the two full-length cDNAs 50,51 . The channels expressed with this construct appear to have similar properties to channels formed by co-expression of GluA1o with an excess of TARP 50 . Human GluA1o-γ4 fusion protein expression constructs were generated by overlapping PCR followed by cloning into pCIneo between EcoR1 and Not1 sites. The resulting plasmid was sequence-confirmed. For assays based on the synthetic indicator, the cells were loaded with fluo-4 Direct (Invitrogen) at 20X dilution from stock, with 1.25 mM probenecid (Setareh Biotech, Eugene, OR). In this kit, fluo-4 is applied as the acetoxymethyl ester. Cells were incubated with dye at 37 °C for 30 minutes, followed by 30 minutes at room temperature. Cells were then washed with assay buffer immediately prior to the FLIPR assay. The composition for the assay buffer was the same as the wash buffer, with variations in the calcium concentration (specified in the Results section).
Fluorescence was monitored during the addition of reagents using a FLIPR Tetra ® (Molecular Devices, Sunnyvale, CA) with a 470-495 nm excitation source and a 515-575 nm bandpass emission filter. Assays were performed in single-addition mode. Test compounds known to be antagonists or positive allosteric modulators (PAMs) were either added to the wells 60 minutes prior to the assay, or added at the same time as the agonist (see details for each experiment). During the assay, fluorescence was monitored for ten seconds prior to the addition of agonist to ensure stable baselines. For time course measurements, the fluorescence was normalized to the baseline fluorescence F 0 (fluorescence measurement at the first time point).
The negative control wells had no added compounds, and the positive control wells were treated with a blocking concentration of a full antagonist. All wells contained 0.1% DMSO. For the AMPA receptor assays, the positive control was 50 µM CP-465022 53 . For the muscarinic receptor assays, the positive control was 3 µM scopolamine. For the TRPV1 assays, the positive control was 10 µM capsazepine 54 .
The response R in each well was the fluorescence F in the measurement window, normalized to the mean fluorescence of negative ( σ ± N N ) and positive ( σ ± P P ) control wells (mean ± standard deviation (SD)) in the same measurement window: Assay quality was assessed for each assay plate by calculating the z′ factor 55 . Under ideal conditions, the assay quality parameter z′ = 1, while z′ < 0 indicates excessive scatter in the controls:

P N
For potency determinations, test compounds were prepared at a range of concentrations by serial dilution. The normalized responses (R) as functions of the test compound concentrations (x) were fitted to a four-parameter Hill function: www.nature.com/scientificreports www.nature.com/scientificreports/ The fitted parameter corresponding to the midpoint (x 0 ) was taken to be the potency of the compound: for antagonists (IC 50 ; 50% inhibitory concentration); for PAMs and agonists (EC 50 ; concentration achieving 50% maximal effect). Potency is expressed as = − pIC log (IC [M]) 50 10 50 , or = − pEC log (EC [M]) 50 10 50 . The dynamic range for a fluorescent indicator is typically defined as = DR F F / max min , where F max and F min are the signals measured in calcium-saturated and calcium-depleted conditions, respectively. In an HTS setting, zero-calcium conditions are rarely used, and resting levels of intracellular calcium can be within the sensitivity range of the indicator. Therefore, a more-relevant measure is the effective dynamic range: Here, F ∞ is the maximal fluorescence signal when the cells of interest are challenged with a treatment expected to achieve near-saturating calcium conditions, and F 1 is the fluorescence when the cells are unchallenged or treated with a fully-inhibited control. Preferably, F ∞ should be measured using a receptor-mediated pathway. This enables measurement only of those cells that comprise both the target protein and the indicator. To account for potential dye bleaching or addition artifacts, F 1 is measured within the same time window as the normal assay response. In many assays, F 1 may also be used as either the negative or positive control response. Both measures incorporated in Equation 4 are sensitive to instrumental features (imperfect optics, background fluorescence of the multiwell plates, etc.) and assay-specific features (background fluorescence of the cells or other reagents, cell-to-cell variation of the indicator concentration, etc.).
The assay dynamic range is calculated as: Here, F 2 is the fluorescence for an uninhibited control. In many assays, F 2 may also be used as a positive or negative control.
imaging. Cells were plated onto 384-well poly-D-lysine-coated plates at 10-15 K cells/well 24 hours prior to imaging. Nuclei were labelled with Hoechst 33342 (Thermo Fisher, Waltham, MA) at a final dilution of 16.7 μM. Cells were treated as described in the Results section 5 minutes prior to imaging. Confocal imaging of live cells was performed using a Zeiss LSM 700 microscope. GFP and fluo-4 were excited by a 488 nm laser (emission filter 492LP) and Hoechst 33342 by a 405 nm laser (emission filter 590SP) with a dwell time of 3.15 μsec and a line averaging setting of 2. Laser power and detector gain were set to levels just below saturation of the detector for each dye based upon the images of cells treated with ionomycin. Tracks were acquired sequentially to avoid bleed-through. 1024 × 1024 pixel (159.89 × 159.89 mm) images or 512 × 512 pixel (159.73 × 159.73 × 16.25 mm) z-stacks were captured at a depth of 12 bits. Z-stacks were processed as orthogonal projections to allow determination of calcium indicator/nuclear label colocalization. All images were exported as 2-channel 8-bit RGB TIF files, then cropped for final presentation in Adobe Photoshop. flow cytometry. Assay buffer was wash buffer adjusted to 4 mM CaCl 2 . Cells were suspended at a density of 1-3 million cells/mL in assay buffer. For those cells labelled with fluo-4, Fluo-4 Direct (Invitrogen) was added to the suspension at 20X dilution from stock, with 1.25 mM probenecid (Setareh Biotech). Cells were incubated 60 minutes at room temperature on a rocker, and then washed with assay buffer. Pharmacological treatments were added 2-3 minutes prior to cytometry in a Moxi GO II (ORFLO Technologies, Ketchum, ID). Fluorescence was detected using 488 nm laser excitation, 525/45 nm emission filter, and Very Low Gain for the detector. Particle size was detected by the Coulter method; particles smaller than 9 µm diameter were considered debris and gated out. Data analysis was performed using FlowJo V10 (Ashland, OR), with fluorescence intensities reported in Arbitrary Fluorescence Units (AFU). Non-linear least squares fitting of histograms of the log-transformed fluorescence intensities was performed using Origin Pro 2017.