Label-free cell phenotypic profiling decodes the composition and signaling of an endogenous ATP-sensitive potassium channel

Current technologies for studying ion channels are fundamentally limited because of their inability to functionally link ion channel activity to cellular pathways. Herein, we report the use of label-free cell phenotypic profiling to decode the composition and signaling of an endogenous ATP-sensitive potassium ion channel (KATP) in HepG2C3A, a hepatocellular carcinoma cell line. Label-free cell phenotypic agonist profiling showed that pinacidil triggered characteristically similar dynamic mass redistribution (DMR) signals in A431, A549, HT29 and HepG2C3A, but not in HepG2 cells. Reverse transcriptase PCR, RNAi knockdown, and KATP blocker profiling showed that the pinacidil DMR is due to the activation of SUR2/Kir6.2 KATP channels in HepG2C3A cells. Kinase inhibition and RNAi knockdown showed that the pinacidil activated KATP channels trigger signaling through Rho kinase and Janus kinase-3, and cause actin remodeling. The results are the first demonstration of a label-free methodology to characterize the composition and signaling of an endogenous ATP-sensitive potassium ion channel.

induced receptor activation in native cells 20,21 . The DMR signal obtained offers a holistic view of receptor signaling, so it is possible to deconvolute the systems cell biology [21][22][23][24][25][26] and signaling waves 27 of different classes of receptors. Taking advantages of the label-free biosensor to be highly sensitive and non-invasive, we used it to identify the DMR signature of pinacidil, a K ATP channel opener 28,29 , across multiple cell lines, and deconvolute its origin and signaling pathways in HepG2C3A cells. We demonstrated that HepG2C3A cells express a functional K ATP channel, although its location remains unknown.

Results
Label-free cell phenotypic profiling of ion channel ligands. To identify functional receptors as well as appropriate cell line(s) for studying endogenous K ATP channels, we first profiled a commercially available library consisting of seventy-two ion channel ligands, each at 10 mM, in the five distinct cell lines using DMR agonist assay in microplate 27 . The negative controls (that is, the cell response to the buffer solution containing 0.1% DMSO, equal amount to those for all ligands) were also included. Using the 3s of the negative control response, we limited our analysis to active ligands that led to a DMR greater than 50 pm or smaller than 250 pm in at least one cell line. Results showed that out of seventy-two ligands twenty-two triggered robust DMR in at least one cell line (Fig. 1). Similarity analysis with the Ward hierarchical clustering algorithm and Euclidean distance metrics 26 showed that these active ligands are divergent in their label-free cell phenotypic agonistic activity in these cell lines.
Detailed analysis revealed several interesting findings. First, thapsigargin, A23187, and cyclopiazonic acid shared similarity in their cell phenotypic agonistic activity. All three ligands triggered robust positive DMR in A431, A549, and HT29. However, only thapsigargin and A23187 were active in HepG2 and C3A cells. The non-selective agonistic activity of these ligands across the five cell lines, together with their similarity in DMR characteristics to those arising from the activation of G q -coupled receptors in respective cell lines 30 , suggests that their DMR is due to the increasing cytosolic calcium concentrations induced by these ligands. Thapsigargin is a non-competitive inhibitor of sarco/endoplasmic reticulum Ca 21 ATPase (SERCA) 31 , while cyclopiazonic acid is a specific inhibitor of SERCA 32 . Both inhibitors are known to raise cytosolic calcium concentration by blocking the ability of the cell to pump calcium into the sarcoplasmic and endoplasmic reticula, which leads to depletion of these stores and activates plasma membrane calcium channels, allowing an influx of calcium into the cytosol 31 . A23187 is a Ca 21 ionophore used extensively to mimic the effect of many physiological cell stimuli related to calcium ion 33 . The less robust response induced by cyclopiazonic acid is consistent with its relatively low potency, comparing to thapsigargin 32 . Due to its low potency, cyclopiazonic acid at 10 mM may not trigger the maximal response.
Second, NPPB (5-nitro-2-(3-phenylpropylamino)benzoic acid), niflumic acid, IAA-94 (R(1)-methylindazone) and flufenamic acid shared similar cell phenotypic pharmacology; all were active in HT29 and C3A cells, and to less extent in HepG2 cells. Recently, we had showed that the positive DMR of niflumic acid in HT29 is due to the activation of GPR35 34 . Consistent with its known agonist activity at the GPR35 35 , NPPB triggered a dose-dependent DMR in HT29 (Fig. 2a), yielding an apparent logEC 50 of 25.95 6 0.07 (n 5 3) (Fig. 2b). It also desensitized the response to 1 mM zaprinast, a known GPR35 agonist, with a logIC 50 of 25.79 6 0.05 (n 5 3) (Fig. 2b). The known GPR35 antagonist CID2745687 dose-dependently and completely blocked the DMR of 4 mM NPPB (n 5 3) (Fig. 2c), with an apparent logIC 50 of 25.51 6 0.06 (n 5 3) (Fig. 2d). Although flufenamic acid was reported to be inactive in human GPR35-b-arrestin-2 interaction assays 36 , we found that flufenamic acid is a DMR biased partial agonist at the GPR35 (Fig. 2e-h). Flufenamic acid triggered a dose-dependent DMR in HT29 (Fig. 2e), yielding an apparent logEC 50 of 25.18 6 0.03 (n 5 3) (Fig. 2f). It also desensitized the cells responding to the second stimulation with 1 mM zaprinast with a logIC 50 of 24.95 6 0.06 (n 5 3) (Fig. 2f). CID274568 dose-dependently and partially blocked the DMR of 32 mM flufenamic acid with a logIC 50 of 25.28 6 0.05 Figure 1 | DMR heat map of ion channel ligands active in five cell lines, A431, A549, HT29, HepG2, and C3A. This heat map was obtained using clustering analysis of the DMR profiles of the ion channel ligands in the five cell lines. For each ligand, the real responses at six discrete time points poststimulation (3, 5, 9, 15, 30, 45 min) in each cell line were used for the cluster analysis. Only ligands that led to a DMR greater than 50 pm, or less than 250 pm in at least one cell line were included in this analysis.  Fig. 2g & h). Using GPR35 Tango b-arrestin gene reporter assay, we found that NPPB triggered a dose response with a logEC 50 of 24.55 6 0.04 and a maximal response that is 45 6 3% (n 5 3) of the full agonist zaprinast, but flufenamic acid was inactive (Fig. 2i). These results suggest that similar to several other GPR35 agonists including benserazide 37 , tolcapone 38 and rosmarinic acid 39 , flufenamic acid not only activates the endogenous GPR35 in HT29, but also activates an additional unknown receptor. Of note, we did not determine the mechanism accounted for the DMR of IAA-94 in HT29, given that the IAA-94 DMR is much smaller than those induced by GPR35 agonists.
Third, pimozide, SDZ-201106 and ZM226600 all triggered characteristically similar DMR specifically in A431. ZM226600 is a Kir6 K ATP channel opener with an EC 50 of 0.5 mM 40 . Pimozide is a dopamine D 2 antagonist and has high affinity to 5-HT 7 receptor 41 . SDZ-201106 is a sodium channel opener 42 . Given that no common mode of action is known among these ligands, it is possible that they activ-ate an unknown receptor in A431 cells; this possibility is currently under investigation.
Fourth, pinacidil was found to trigger a robust negative DMR in A431, A549, HT29 and C3A, but not in HepG2 cells. However, several other K ATP openers including diazoxide, minoxidil, minoxidil sulphate, and PCO-400 in the library, each at 10 mM, were inactive in the DMR agonist profiling across the five cell lines. Given that K ATP openers are classified based on their ability to open the channel per se, but the DMR of a ligand reflects the functional consequence of receptor-ligand interaction in cells, we postulated that similar to GPCR ligands 43 , K ATP openers may also display functional selectivity and the DMR of pinacidil is related to the signaling of endogenous K ATP channels. Of note, it was reported that C3A expresses K ATP channels that play a role in cell proliferation 44 , and HepG2 has little K ATP channel current on the cell surface but the K ATP channel expression can be induced by transfection of insulin and glucose transporter GLUT2 45 .  Together, these analyses suggest that the label-free cell phenotypic profiling approach is powerful for determining the on-target and offtarget activity of ion channel ligands. Given the unique signature of pinacidil, we herein were primarily focused on the deconvolution of the target and pathways activated by pinacidil.
The pinacidil DMR in C3A cells is due to the activation of endogenous K ATP channels. To determine the composition of endogenous K ATP channels, we first characterized the dose responses of known K ATP openers using DMR agonist assay. Results showed that pinacidil triggered a dose response in C3A cells (Fig. 3a), yielding a logEC 50 of 24.77 6 0.05 (n 5 6) (Fig. 3b). In comparison, both PCO-400 and diazoxide triggered much smaller negative DMR with lower potency; their logEC 50 s were 23.99 6 0.07 and 23.47 6 0.09 (n 5 3), respectively (Fig. 3b). In contrast, minoxidil up to 64 mM was inactive. Similarly, pinacidil also triggered a dose response in A549 cells (Fig. 3c), yielding a logEC 50 of 24.72 6 0.05 (n 5 6) (Fig. 3d). In comparison, PCO-400 triggered a much smaller negative DMR with a logEC 50 of 24.16 6 0.03 (n 5 3), while both diazoxide and monoxidil were inactive ( Fig. 3d; data not shown). The relatively low potency of active K ATP openers obtained may be related to the whole cell measurements, wherein the intracellular ATP/ADP ratio has direct effects. Another possibility is that the K ATP channels are located inside cells, and the effective intracellular concentration of these openers is a function of cell uptake and efflux and may be lower than those added extracellularly. Of note, in order to better visualize the DMR characteristics we used real DMR responses, instead of their absolute values, for all dose response analysis. Nonetheless, these results suggest that these openers display distinct potency and efficacy to trigger DMR in C3A and A549 cells.
Third, we applied RNAi knockdown to determine the functional K ATP channels in C3A cells. Results showed that the mock or scrambled RNAi transfection had little impact on the pinacidil DMR (Fig. 5a). The treatment with three RNAi against SUR1 had little impact on the pinacidil DMR (Fig. 5b), consistent with the absence of SUR1 mRNA in C3A cells. However, knockdown of SUR2 with three RNAi all markedly suppressed the pinacidil DMR (Fig. 5c). Similar to the mock or scrambled RNAi transfection (Fig. 5d), three RNAi knockdown of Kir6.1 had little impact on the pinacidil DMR (Fig. 5e). In contrast, the knockdown of Kir6.2 with three RNAi all markedly suppressed the pinacidil DMR (Fig. 5f). One-way ANOVA analysis suggests that RNAi for SUR2 and Kir6.2, but neither SUR1 or Kir6.1, significantly altered the pinacidil DMR ( Fig. 5g and h). Given the low expression level of these proteins and the moderate efficiency of typical RNAi knockdown in our laboratory 26 (also see below), we didnot attempt to use Western blot to quantify the knockdown efficiency of respective RNAi.
Fourth, the effects of different K ATP inhibitors on the pinacidil DMR were examined and used to further strengthen RNAi knockdown results. Results showed that three sulfonylurea blockers including tolazamide, glipizide and tolbutamine all dose-dependently blocked the pinacidil DMR, leading to a single logIC 50 of 23.85 6 0.09, 3.94 6 0.10, and 3.91 6 0.10 (n 5 3), respectively. In contrast, glibenclamide up to 250 mM only partially attenuated the pinacidil DMR, and U-37883A had little impact on the pinacidil DMR ( Fig. 5i & j). U-37883A is a non-sulfonylurea blocker that has been reported to selectively inhibit Kir6.1 containing K ATP channels 46 , while glibenclamide is a relatively selective SUR1 inhibitor 47,48 . Furthermore, the co-existence of 40 mM tolazamide caused the right shift of pinacidil's potency (Fig. 5k). Together, these results suggest that the pinacidil DMR in C3A is due to the activation of SUR2/Kir6.2 K ATP channels. Of note, we attempted, but failed to conclusively determine the exact location of K ATP channels in C3A cells, due to very low levels of K ATP channel proteins.
Collectively, these pharmacological profiling, RNAi knockdown, and RT-PCR results suggest that C3A endogenously expresses a functional SUR2/Kir6.2 K ATP channel, whose activation by pinacidil results in a robust DMR signal.
Kinase inhibition profiling. The regulation of K ATP channels has been extensively investigated; in particular, its upstream regulation mediated by protein kinase A (PKA) has been well documented [49][50][51][52] . The phosphorylation of ion channels by protein kinases is an important mechanism by which membrane excitability is regulated by signaling pathways. However, a clear understanding about the signaling of K ATP is still lacking. Thus, we investigated the impact of a library of known kinase inhibitors on the pinacidil DMR in C3A cells. Results showed that the majority of kinase inhibitors had little impact on the pinacidil DMR (Fig. 6a). However, a subset of inhibitors markedly suppressed the DMR of pinacidil. Among them, H-89, HA-1077, H-7, and H-8 are known PKA inhibitors, while AG490 is a putative Janus activated kinase (JAK) inhibitor, Y27632 is a ROCK inhibitor, and the alkaloid staurosporine is a broad spectrum, high affinity kinase inhibitor with highest affinity for protein kinase C (0.7 nM). DMR inhibition assay showed that H-89 and H-7 dose-dependently and completely blocked the DMR of 32 mM pinacidil, leading to a logIC 50 of 25.32 6 0.09 and 25.17 6 0.10, respectively (Fig. 6b). The sensitivity of the pinacidil DMR to these PKA inhibitors were consistent with the previous findings, which suggest that PKA is anchored in proximity to K ATP channels  in the caveolae 49 , and more than one site in Kir6.2 (e.g., S372 or T224 in Kir6.2) 51 and SUR (e.g., T633 and S1465 in SUR2B) 49 have been implicated in PKA phosphorylation. The positive regulation of K ATP by PKA was evidenced by the observed attenuation of K ATP channel currents in the presence of PKA inhibitors [49][50][51][52] . Of note, compounds containing 5-isoquinolinsulfonyl moieties such as H-89 also bind directly to the SUR subunit but with relatively low potency 53 . At the concentration range tested it is not possible to distinguish between direct blocking of the K ATP channel and indirect inhibition through abrogation of PKA activity. The inhibition of the pinacidil DMR by staurosporine may also be linked to PKA (Ki , 7 nM), but the prevalence of other mechanisms cannot be ruled out 54,55 . Nonetheless, these results suggest that the pinacidil DMR can be modulated by kinase activity.
JAK3 and JAK2 are involved in K ATP channel signaling. Janus kinases (JAK1, 2, 3) are protein tyrosine kinases involved in cytokine mediated cellular signaling and are crucial for a variety of cellular functions including cellular survival, proliferation, differentiation and apoptosis 56,57 . Given the attenuation of the pinacidil DMR by AG490, we examined the role of JAKs in K ATP channel signaling using multiple assays. First, DMR inhibition assay showed that AG490 dose-dependently suppressed the pinacidil DMR, yielding a logIC 50 of 24.77 6 0.10 (n 5 3) (Fig. 7a). Second, siRNA knockdown studies showed that the treatment with two siRNAs for JAK1 had little impact on the pinacidil DMR, but two siRNAs for JAK2 and two for JAK3 all markedly suppressed the pinacidil DMR ( Fig. 7b and c). Third, western blotting showed that C3A cells primarily express JAK3, to less extend JAK2, but not JAK1. Moreover, the treatment of cells with 100 mM pinacidil markedly increased the level of JAK2 (Fig. 7d) and JAK3 (Fig. 7e). Importantly, the Kir6.2 siRNA treatment impaired the pinacidil induced increase of JAK3 protein level ( Fig. 7f and g), suggesting that the pinacidil induced increase in JAK3 protein is due to the activation of Kir6.2 K ATP . We also attempted, but failed, to detect p-JAK levels in the absence and presence of pinacidil. Several possible mechanisms are accounted for this; one may be due to the poor activity of anti-p-JAK antibodies used; another is that the pinacidil activated K ATP did not cause phosphorylation of JAKs, but only increased the expression level of JAKs. Of note, we used moderately high dose (40 mM) of pinacidil for DMR profiling of the effect of kinase inhibitor and RNAi knockdown in order to improve sensitivity, but a higher dose (100 mM) for JAK expression analysis to achieve maximal effect. Nonetheless, these results suggest the important role of JAK3, to less extent JAK2, in the K ATP channel signaling in C3A cells.
K ATP channel signaling is linked to ROCK and actin remodeling. Rho kinases (ROCK1 and ROCK2) play important roles in the small GTPase RhoA initiated signaling pathways. Rho kinases are known to be involved in a variety of cellular functions including cytoskeleton organization, cell proliferation and apoptosis 58 . Given the ability of Y27632 to attenuate the pinacidil response, we next examined the role of ROCK in the K ATP signaling in C3A cells. DMR inhibition assay showed that Y27632 dose-dependently inhibited the pinacidil  (Fig. 8a). RNAi knockdown of ROCK1 or ROCK2 was found to markedly attenuate the pinacidil DMR (Fig. 8b and c). Western blotting confirmed that the efficiency of ROCK RNAi knockdown was about 40-50% ( Fig. 8d-f), consistent with the level of suppression of the pinacidil DMR by these siRNAs (comparing Fig. 8f with Fig. 8c). RhoA activity results showed that the treatment of C3A cells with pinacidil had little impact on the RhoA activity (Fig. 8g), suggesting that the pinacidil activated K ATP channels had no effect on the activity of RhoA, the upstream effector of ROCK.
Given the important roles of ROCK signaling in actin remodeling, we finally examined the impact of microfilament modulators on the pinacidil response in C3A cells. Results showed that both cytochalasin B and latrunculin A dose-dependently blocked the pinacidil In contrast, the microtubule disrupting agent nocodazole up to 32 mM had little effect on the pinacidil DMR (Fig. 8h). These results suggest that the K ATP channel initiated signaling in liver cells is linked to ROCK activity, leading to actin remodeling.

Discussion
We here present a strategy centered on label-free cell phenotypic profiling to discover the composition and signaling of an endogenous K ATP channel in C3A cells. Label-free cell phenotypic profiling of a compound library consisting of seventy-two ion channel ligands across the five different cell lines led to identification of twentytwo ligands that displayed agonistic activity in at least one cell lines. Several classes of active ligands can be assigned based on their known mechanism of action (e.g., Ca 21 mobilizing agents such as A23187, thapsigargin and cyclopiazonic acid), or using orthogonal assays (e.g., GPR35 agonists flufenamic acid, niflumic acid and NPPB).
Combining DMR RNAi knockdown, ligand pharmacology profiling, and RT-PCR has allowed us to ascertain the presence, composition and signaling of a functional SUR2/Kir6.2 K ATP ion channel in C3A cells. K ATP openers are a structurally diverse group of compounds, and also diverse in the modes of action. Diazoxide is a benzothiadiazine dioxide, pinacidil a cyanoguanidine, PCO-400 a benzopran, minoxidil a pyrimidine N-oxide sulfate, and   ZM226600 an anilide tertiary carbinol. The Kir6.2/SUR1 channel can be activated by diazoxide but is relatively insensitive to pinacidil, while the Kir6.2/SUR2A channel is only weakly responsive to diazoxide but can be activated by pinacidil 59 . On the other hand, the Kir6.2/SUR2B channel can be activated by diazoxide 60 ; ZM226600 can activate a channel consisting of SUR2B co-associated with Kir6.1 or Kir6.2 61 ; minoxidil can specifically activate Kir6.1/SUR2 channels 62 . Our DMR profiling showed that among all KCOs tested in C3A cells pinacidil triggered a robust DMR, while PCO-400 and diazoxide are less active, ZM226600 and minoxidil were inactive. Considering gene expression, RNAi knockdown and K ATP inhibitor profiling data, our DMR results are better explained by the presence of a functional Kir6.2/SUR2A channel in C3A cells. Of note, due to the low expression of K ATP channels, in particular Kir6.2 subunits at both mRNA and protein levels in C3A cells, we have difficulty to determine the exact location of these channels whose activation contributes to the DMR of pinacidil. We also did not determine the location and compositions of K ATP channels in other pinacidil responsive cell lines. Nonetheless, these results demonstrate the use of label-free cell phenotypic profiling for broad pharmacological profiling of K ATP channels.
The molecular events underlying the modulation of K ATP activity by protein kinases is poorly understood. Both JAK2 and ROCK inhibitors have been reported to be physiologically linked to K ATP channels in cardioprotection. The JAK/STAT pathway mediates opioid induced cardioprotection via glycogen synthase kinase-3b; AG490 abrogates this cardioprotective effect 63 . The proposed mechanism for reduced myocardial no-reflow during ischemic preconditioning, a strategy also used for reperfusion therapy following acute myocardial infarction, involves the opening of K ATP channels via inhibition of ROCKs 64 . ROCK inhibitors decrease the area of noreflow; K ATP inhibitors abolish the reduced no-reflow. Although the exact cellular pathways undetermined, we here showed that both JAK3 (to less extent JAK2) and ROCK are involved in the K ATP channel signaling in C3A cells, a liver cell line.
Information regarding the role and pharmacology of K ATP channels in hepatocytes is limited. While the physiological outcomes of K ATP channel activity in hepatocytes and cardiomyocytes are  expected to be different, it is likely that their modulation by kinases occur through similar molecular pathways. The ability to capture complex physiological phenomena surrounding ion channel activity in experimentally convenient cellular screening assays, as described herein, is powerful and unprecedented. Label-free DMR assays offer great flexibility in assay formats 27 , rich information content with real-time kinetics, high throughput, and more importantly, holistic view of receptor signaling and drug pharmacology with wide coverage of pathway/targets 65 , including ion channels as demonstrated here. DMR assays not only permit pharmacological profiling of distinct types of ligands (openers and inhibitors) for the channel proteins, but also allow for investigating the signaling mediated through ion channels, later of which is difficult to be assessed using electrophysiology. However, DMR assays also come with several limitations. First, the label-free assays cannot detect directly the ion flux related to the open and close of ion channels. Second, the pharmacological profiles of ligands for a specific ion channel may not correlate with those obtained using electrophysiology, given that the DMR is a whole cell phenotypic response linking to the functional responses of cells upon the activation of ion channels. Third, without validation using conventional approaches, label-free is difficult to differentiate the upstream (e.g., PKA phosphorylation of the K ATP channels), and downstream (e.g., the K ATP activation increased JAK expression) signaling events of the channels. Thus, label-free biosensor enabled cell phenotypic profiling for investigating ion channel pharmacology is an attractive complimentary technology to electrophysiology and other conventional techniques. Cell culture. All five cell lines were obtained from American Type Cell Culture (ATCC) (Manassas, VA, USA). They are human epidermoid carcinoma A431, adenocarcinomic human alveolar basal epithelial cell A549, human colorectal adenocarcinoma HT-29, human hepatocellular carcinoma HepG2 and its clonal derivative C3A (HepG2C3A). C3A is a clonal derivative of HepG2 that was selected for strong contact inhibition of growth, high albumin production, high production of alpha fetoprotein and ability to grow in glucose deficient medium. All cells were subcultured 1-2 times per week according to ATCC's instruction. Cell passage less than 15 was used for all experiments. All cells were passaged at 37uC with 5% CO 2 using the following complete medium: McCoy's 5A medium for HT-29, Eagle's Minimum Essential Medium (MEM) for HepG2 and HepG2C3A, F-12K Medium for A549, and Dulbecco's Modified Eagle's Medium (DMEM) for A431. All media were supplemented with 10% fatal bovine serum, 4.5 g/L glucose, 2 mM glutamine, 100 mg/ml penicillin and streptomycin. All cells were passed with trypsin/ethylenediaminetetraacetic acid when approaching 90% confluence to provide new maintenance culture on T-75 flasks and experimental culture on the biosensor microplates.

Materials. BioMol kinase inhibitor library and latrunculin A was obtained from Enzo
For DMR assays, the cell seeding density was optimized for each cell line. The optimal seeding density was found to be 32 K, 25 K, 25 K, 21 K, and 21 K per well for HT-29, A431, A549, HepG2 and C3A cells, respectively. After overnight culture (,20 hrs), all cells were directly washed using the HBSS buffer before assay, except for A431 which was subject to starvation for an extra day using the serum-free DMEM medium. The cell confluency was ,95% at the time of assaying for all cells.
Reverse transcriptase PCR. Total RNA was extracted from one T-75 flask with a confluent monolayer of cells (15-30 million cells) for each cell line using an RNeasy mini kit (Qiagen, Cat#74104, Valencia, CA). On-column DNase digestion was performed using RNase-free DNase set (Qiagen, Cat#79254) to eliminate genomic DNA contamination. The concentration and quality of total RNA were determined using a Nanodrop 8000 (Thermo Scientific). The primer sequences used were based on previous reports 66, 67 and custom synthesized through Sigma-Aldrich. Reverse transcriptase PCR was performed using One-Step RT-PCR kit from Qiagen Inc. (Cat# 210212). The PCR conditions were as follows: 50uC for 30 min, 95uC for 15 min, followed by 60 cycles of 1 min at 94uC, 1.5 min at 57uC and 2 min at 72uC, with a final extension of 10 min at 72uC. For all gel electrophoresis analysis equal amount of DNA was loaded. The DNA of actin obtained was used the control. The comparative level was estimated across different genes in a specific cell line, not across different cell lines.
siRNA transfection was performed using the N-TER nanoparticle siRNA Transfection system (Cat# N2913) from Sigma. Specifically, 5000 cells were first plated into each well of an EpicH 384well microplate, and cultured for 20 hrs using the complete medium. The cells were then transfected with 50 nM siRNA and incubated in siRNA containing medium for 24 hrs before replaced with fresh cell culture medium. Label-free DMR assays were performed 48 hrs after transfection. The cells treated with the transfection vehicle were used as the mock control.
Immunoprecipitation and Western blotting. C3A cells were plated in 6-well tissue culture treated plate with 3 3 10 5 cells per well. Cells were transfected with respective top-rank siRNA at 50 nM final concentration 20 hrs after plating. 24 hrs after transfection the transfection reagent containing media were removed and replaced with the respective complete media. 48 hrs after transfection cells were lysed using 1% NP40 lysis buffer (150 mM NaCl, 25 mM Tris-HCl, pH 7.6, 1% NP-40) with the protease inhibitor cocktail (Roche Applied Science, Indianapolis, IN, USA). The cell lysate was then centrifuged at 14,000 rpm for 20 min. The supernatant was then transferred to a new tube before western blotting.
DMR assays. DMR assays in microplate were performed using high throughput screening compatible EpicH system (Corning) 68 . This system consists of a temperature-control unit (26uC), an optical detection unit, and an on-board liquid handling unit operated by robotics. This system scans entire biosensor microplate every 6 sec, and two scans are averaged to reduce the signal noise, thus leading to a kinetic response with a temporal resolution of ,15 sec. The cultured cells were first washed with the assay buffer HBSS using a plate washer (Bio-Tek Microplate Washers ELx405 TM , Bio-Tek, Winooski, VT), and incubated inside the EpicH system for ,1 hr. After a steady baseline was established, the ligand solutions were introduced using the on-board liquid handling device, and the cell responses were then recorded over time. For inhibitor profiling, the cells were pretreated with an inhibitor for 1 hr before pinacidil stimulation. For RNAi, the cells were transfected with siRNA for 48 hrs before pinacidil stimulation. All real-time DMR signals were reported as a 2 min baseline, followed by ligand stimulation. All DMR signals were background corrected. At least two independent sets of experiments, each with at least duplicate, were performed.
Tango b-arrestin translocation gene reporter assay. Tango assays were performed in engineered Tango TM U2OS-GPR35-bla cell line (Life Technologies). This cell line allows an endpoint measure of the activity of agonists specific to the GPR35 activation-induced b-arrestin translocation 69,70 . The cells were passed using McCoy's 5A medium supplemented with 10% dialyzed fetal bovine serum, 0.1 mM NEAA, 25 mM Hepes (pH 7.3), 1 mM sodium pyruvate, 100 U/ml penicillin, 100 mg/ml streptomycin, 200 mg/ml zeocin, 50 mg/ml hygromycin, and 100 mg/ml geneticin in a humidified 37uC/5% CO 2 incubator. For Tango assays, 10000 U2OS-GPR35-bla cells per well were seeded in 384-well, black-wall, clear bottom assay plates with low fluorescence background (Corning). After overnight culture, the cells were stimulated with ligands for 5 hrs at 37uC under 5% CO 2 , and then loaded with the cell permeable LiveBLAzer TM FRET (fluorescence resonance energy transfer) B/G substrate. After www.nature.com/scientificreports SCIENTIFIC REPORTS | 4 : 4934 | DOI: 10.1038/srep04934 2 hr incubation the coumarin to fluorescein ratio was measured using Victor 4 plate reader (PerkinElmer, Waltham, MA, USA). Results obtained were normalized to the maximal response of zaprinast (set to be 100%).
RhoA activation assay. RhoA activity was determined from protein isolated from C3A cells without and with treatment with 100 mM pinacidil using the luminescence based G-LISATM RhoA activation assay kit (Cytoskeleton, Denver, CO, USA) according to the manufacturer's instructions. Protein was isolated using the provided cell lysis buffer, and cells were processed rapidly on ice and snap-frozen until the time of assaying. Lysates were clarified by centrifugation at 10,000 rpm at 4uC for 2 min. Protein concentration was determined according to the manufacturer's protocol, and cell extracts were equalized to contain protein concentrations of 2 mg/ml for the assay. Luminescence was detected as suggested by the manufacturer.
Data analysis. All dose-dependent responses were analyzed by using GraphPad Prism 5.0 (GraphPad Software Inc., San Diego, CA, USA). The EC 50 and IC 50 values were obtained by fitting the DMR dose response curves with nonlinear regression. For cluster analysis, we profiled the ion channel ligand library across the five cell lines. For effective clustering, we extracted the real responses at the six distinct time points (3, 5, 9, 15, 30, 45 min) for each DMR to form a numerical description of the label-free cell phenotypic characteristics [71][72][73] . All the six time points refer to the stimulation duration after renormalizing the responses starting from the time when the compound was added. The real responses at these discrete time points were color coded to illustrate relative differences in the direction and strength of a DMR signal. The red color indicates a positive value, while the green color refers to a negative value, the black a value at or near zero. Differences in color intensity illustrate differences in signal strength. In the ligand heatmap matrix (Fig. 1) each column represents one DMR value at a specific time in a specific cell line, and each row represents one ligand. The Ward hierarchical clustering algorithm and Euclidean distance metrics (http://www.eisenlab.org/eisen/) were used for similarity analysis, and every row and column carries equal weight. DMSO in the assay buffer at a concentration that equals to those for all ligands was also included as a negative control. Each DMR represents an average of four replicates. To assist with direct visualization of DMR characteristics of each ligand in a cell line, we did not carry out similarity analysis among distinct columns.
Statistical analysis. DMR data were analyzed by using GraphPad Prism 5.0 (GraphPad Software Inc., San Diego, CA, USA). The EC 50 values were obtained by fitting the dose DMR response curves with nonlinear regression. The RNAi effect was analyzed using one-way ANOVA Tukey analysis with the Prism 5.0.