Development of phenotypic assays for identifying novel blockers of L-type calcium channels in neurons

L-type calcium channels (LTCCs) are highly expressed in the heart and brain and are critical for cardiac and neuronal functions. LTCC-blocking drugs have a long and successful record in the clinic for treating cardiovascular disorders. In contrast, establishment of their efficacy for indications of the central nervous system remains challenging given the tendency of existing LTCC drugs being functionally and mechanistically more selective for peripheral tissues. LTCCs in vivo are large macromolecular complexes consisting of a pore-forming subunit and other modulatory proteins, some of which may be neuro-specific and potentially harbor mechanisms for neuronal selectivity. To exploit the possibility of identifying mechanistically novel and/or neuro-selective blockers, we developed two phenotypic assays—a calcium flux-based primary screening assay and a patch clamp secondary assay, using rat primary cortical cultures. We screened a library comprised of 1278 known bioactive agents and successfully identified a majority of the potent LTCC-blocking drugs in the library. Significantly, we identified a previously unrecognized LTCC blocker with a novel mechanism, which was corroborated by patch clamp and binding studies. As such, these phenotypic assays are robust and represent an important step towards identifying mechanistically novel and neuro-selective LTCC blockers.

www.nature.com/scientificreports/ various spliced variants of Ca V 1.2 have also been identified 23 , some of which may be more neuronally enriched and confer distinct physiological and pharmacological properties 5,24 . Variant isoforms of LTCC genes can lead to changes in channel properties [24][25][26] . Therefore, it is conceivable that LTCCs in neurons may confer novel and/ or neuro-selective pharmacological properties through neuro-favoring/specific interactions. In this study, we took a phenotypic approach to screening for novel, neuro-selective LTCC blockers. Using primary cultures of rat cortical neurons, we developed two phenotypic assays, a fluorescence-based mediumthroughput screening assay and a secondary patch clamp confirmation assay. We conducted a primary screen of 1278 commercially available small molecule drugs and other bioactives, which successfully identified majority of the LTCC drugs, lending validation to these assays. Importantly, we identified a novel, potent LTCC blocker, suggesting the utility of this approach towards identifying novel and neuro-selective LTCC modulators.

Isolation of LTCC-mediated responses in rat primary cortical cultures. Cortical neurons express
LTCCs as well as other VGCCs. As shown in Fig. 1a, application of 100 mM K + evoked a large calcium fluorescent response in a rat primary cortical culture that was partially blocked by nimodipine (3 µM), a potent and selective LTCC antagonist, indicating that LTCCs represented a substantial fraction of VGCCs expressed in these cultures. Elevated concentrations of K + indiscriminately activate all VGCCs. To selectively activate LTCCs, we used Bay K8644, a LTCC-selective agonist, to stimulate responses in these cultures. In contrast to 100 mM K + , the Bay K8644 (1.1 µM)-induced response was completely blocked by 3 µM nimodipine (Fig. 1a), indicating that it was LTCC mediated. To better understand whether the Bay K-induced response was a result of direct calcium influx through LTCC channels or due to secondary activation of other VGCCs, we included two VGCC-(but not LTCC-) blocking peptides, ω-conotoxin GVIA (N-type calcium channel blocker) and ω-conotoxin MVIIC (Nand P/Q-type calcium channel blocker). As shown in Fig. 1b-d, the peptide cocktail partially blocked the high  Optimization of the phenotypic fluorescence assay. Bay K8644 potentiates LTCCs by shifting the voltage dependence of channel activation towards hyperpolarized membrane potentials 27 . Consequently, the amplitude of Bay K-evoked response is dependent on the resting membrane potential. Significant responses appear only when the resting membrane potential is depolarized sufficiently close to the threshold for LTCC activation. To maximize Bay K-evoked responses, we varied the resting membrane potential by preincubating cells with a range of holding K + concentrations. As shown in Fig. 2, Bay K (300 nM)-induced responses as a function of holding [K + ]s were bell-shaped, peaking at 20 mM holding [K + ], which corresponds to a resting membrane potential of roughly − 50 mV. At this membrane potential, T-type calcium channels are inactivated, preventing potential contamination by non-specific responses from this class of VGCCs. Together with the finding that the Bay K-induced responses at 20 mM holding [K + ] were LTCC specific ( Fig. 1), we chose to preincubate cells in a buffer containing 20 mM K + in our assay. We also studied the effects of several other parameters on Bay K-evoked responses, including cell density, tolerance for DMSO and culture days in vitro (DIV). The results are either shown in Supplementary Fig. S1 and Pharmacology of reference compounds. Next, we evaluated the LTCC pharmacology in the calcium flux assay using reference compounds that are well described in the literature. Bay K potently activated LTCCs with an EC 50 of 203.9 nM (Fig. 3a,b). The three major classes of LTCC antagonists, nimodipine (a DHP), verapamil (a phenylalkylamine) and diltiazem (a benzothiazepine) concentration-dependently inhibited 300 nM Bay K-induced responses with IC 50 values of 142.4 nM, 18.5 µM and 27.1 µM, respectively (Fig. 3c,d), in general agreement with the literature [28][29][30] . Taken together, our phenotypic calcium flux assay showed convincing hallmarks of LTCC specificity.

Prestwick library screen.
To further validate and test the robustness of the assay, we performed a pilot screen of 1278 known bioactive small molecules from the Prestwick (PW) library. Compounds were screened at a single concentration (3 µM) in duplicates and percent inhibition values were calculated for each compound. The Z' value for the screen was 0.52 ± 0.11 (mean ± SD). The correlation between two independent runs was 0.62 with a slope of 0.71 (Fig. 4a), suggesting reasonable reproducibility. The histogram of the mean % inhibition showed a major peak centered near zero (mean ± SD = -7.8 ± 22.5%) and a second, smaller population with greater % inhibition values (Fig. 4b). We set the cutoff value at 70% based on these data, resulting in the selection of 75 candidate hits (5.9% of the compounds screened) for confirmation in a concentration-response format. Sixty of them (80%) were confirmed with IC 50 values < 5 µM (n = 3). The profile of the confirmed hits is summarized Fig. 4c.
DHPs are a major class of LTCC-blocking drugs and are well represented in the PW library (12 DHPs). Ten  Table 1, eight are DHPs and two (niclosamide and cyproheptadine) are non-DHPs. To our knowledge, niclosamide (an anthelmintic; 80 nM) is the only compound in Table 1 for which no published information is available about the effect on LTCCs. Using a separately validated patch clamp assay, we confirmed the LTCC-blocking activity of niclosamide in rat cortical neurons (see below). www.nature.com/scientificreports/ In conclusion, the phenotypic calcium flux assay was robust and validated by its ability to identify most of the known LTCC blockers as well as a novel LTCC blocker in the PW library.
Development of a secondary patch clamp assay. We also developed and validated a secondary patch clamp assay using rat cortical neurons to confirm hits identified from the calcium flux assay. The buffers used in patch clamp were intended for isolating/maximizing calcium channel responses (substituting extracellular Ba 2+ for Ca 2+ ) and minimizing/eliminating sodium and potassium channel activity (extracellular choline replacing Na + and intracellular Cs + replacing K + ). They simultaneously served to abolish network activities normally present in neuronal cultures without the need to introduce cocktail inhibitors of synaptic activity. To obtain conditions optimized for LTCC currents, we varied the holding (− 90 mV or − 50 mV) and testing (− 20 mV or 0 mV) potentials. As shown in Fig. 5a, the fraction of nimodipine-sensitive current was lower when holding/ testing the cell at − 90/0 mV (75%) or at − 50/0 mV (85%) compared to the − 50/− 20 mV protocol (96%). Importantly, inclusion of 300 nM Bay K (a near-saturating concentration; Fig. 5b) in the extracellular solution greatly amplified the LTCC currents (Fig. 5c,d) and effectively reduced the fraction of nimodipine-insensitive (i.e., non-LTCC) currents (as seen in Fig. 5a from the 99% block by nimodipine in the presence of Bay K). As expected, Bay K decreased the potency of nimodipine inhibition by 16.7 fold (IC 50 was increased from 24.8 nM in the absence of Bay K to 414.6 nM in the presence of Bay K; Fig. 5a), consistent with the two DHPs competing for the same binding site 37 . In addition, the potency of nimodipine block was also decreased at the holding potential of − 90 mV (IC 50 = 4.8 µM; Fig. 5a), in agreement with the well-documented voltage dependence of DHP blockers 8 . Under the optimized recording conditions (i.e., with the -50/-20 mV protocol in the presence of Bay K), a cocktail of VGCC-(but not LTCC-) blocking peptides had little effect (11% inhibition) on Ba 2+ currents (Fig. 5e). By contrast, the same cocktail blocked a significantly higher fraction (56%) of Ba 2+ currents under conditions in which other VGCCs were allowed to contribute to the current (i.e., with the − 90/0 mV protocol in the absence of Bay K; Fig. 5e). Taken together, results in Fig. 5 demonstrated that the chosen recording solutions, inclusion   Confirmation of primary hits in patch clamp. Using the patch clamp assay validated above, we tested niclosamide and cyproheptadine, the two non-DHPs hits in Table 1. In rat cortical neurons, 3 µM cyproheptadine blocked 75.2 ± 9.1% (n = 4) of the LTCC currents, in line with its binding affinities to the DHP, phenylalkylamine and benzothiazepine sites (Table 1). In contrast, while niclosamide potently blocked neuronal LTCC currents (IC 50 = 183.1 ± 28.2 nM, n = 6; Fig. 6a,c), it showed no binding at any of these sites in rat cortical preparations (K i > 30 µM for all three sites; n = 4; also see Table 1), suggesting that niclosamide binds to a novel site in the LTCC channel complex in neurons. Similar to control, the I-V curve in the presence of niclosamide (200 nM) peaked near − 20 mV, although niclosamide block was more pronounced at depolarized potentials (Fig. 6d). The activation and inactivation kinetics in the presence of niclosamide were not significantly different from control. The 20-80% rise time (at − 20 mV) was 17.4 ± 2.9 ms (n = 5) and 16.1 ± 3.5 ms (n = 4) in the absence and presence . Surprisingly, niclosamide block of Cav1.2 (100 nM) was more pronounced in experiments in which niclosamide was applied for significantly longer durations (average of ~ 6.5 min) than those used for Ca V 1.2 concentration-response experiments (average ~ 3.5 min) (Fig. 6c). Though not corrected for current rundown, this extra block was not due to compoundindependent current rundown, which was only 8.4 ± 3.7% and 12.4 ± 7.6%, respectively, after 3.5 and 6.5 min in compound-free buffer (n = 4).

Discussion
In an effort to identify novel, neuro-selective LTCC blockers, we took a phenotypic approach and developed two assays using rat primary cortical cultures: a medium-throughput, calcium flux-based fluorescence assay and a patch clamp, secondary confirmation assay. We optimized the assays by using Bay K8644, a LTCC-selective agonist, along with a moderately elevated [K + ] in the extracellular incubating buffer (20 mM; the calcium flux assay) or moderately depolarized holding . At 100 nM niclosamide, inhibition of Ca V 1.2 was significantly greater when CHO cells were exposed to the compound for longer periods (open circle, 80.6 ± 4.9%, n = 4) than that of LTCCs in neurons (solid squares; p < 0.001, one-way ANOVA) or that of Ca V 1.2 when CHO cells were exposed to the compound for shorter periods (solid circles; p < 0.001, one-way ANOVA). www.nature.com/scientificreports/ potential (− 50 mV; the patch clamp assay). This combination selectively amplified LTCC responses with minimal activation of other VGCCs. We further showed that these protocols resulted in Ca 2+ responses or Ba 2+ currents specifically through LTCCs rather than non-specific or LTCC-mediated secondary responses. Moreover, the pharmacology of LTCCs using well-documented reference compounds was also in line with the literature. Brain LTCCs comprise Ca V 1.2 and Ca V 1.3, with the former being the dominant isoform 38 . While it is tempting to speculate that responses in our assays primarily reflect Ca 2+ /Ba 2+ influx through Ca V 1.2, it is not possible to estimate the relative contributions of these isoforms based on our data.
We conducted a pilot screen of 1278 known bioactive agents from the PW library using the calcium flux assay. The results of the screen validated our assay in several ways. First, the assay was robust as suggested by (1) a good correlation (r 2 = 0.62) between the two independent runs of the screen, (2) a hit confirmation rate of 80% in a concentration-dependent manner, and (3) a Z' value of ~ 0.5. Second, most of the DHP drugs in the library (~ 83%), which are all potent LTCC blockers, were correctly identified in the screen. They were all among and represented 50% of the top 20 most potent hits. Third, six of the 10 non-DHPs in the top 20 hits have been reported as LTCC blockers in the literature. The only two non-DHPs in the top 10 hits were both confirmed in the patch clamp assay. One of them, niclosamide, is a novel LTCC blocker identified from our screen for the first time. Independent confirmation of these compounds by patch clamp lent further support for the strength and validity of our primary screening assay.
Combining the literature information and our patch clamp results for the top 20 most potent hits, an upper limit can be placed for the false positive rate associated with the screen. Since no literature information on LTCC effects is available for 3 of the 20 hits (the 4th, niclosamide, was positively confirmed by our patch clamp assay), the false positive rate could be up to 15% (3/20) assuming all three are false positives. It should be cautioned, however, that this estimate could be less reliable when extrapolated to a larger-scale screen because of (1) the relatively small data set in this study and (2) the nature of the extrapolation, i.e., largely from one known mechanism, DHP block of LTCCs, to potentially different and/or novel mechanisms by which many hits cause apparent inhibition.
It should be noted that it is intrinsically a higher hurdle to demonstrate that a hit in our primary assay (or any phenotypic screen) is a false positive since negative outcomes (i.e., failure to confirm a hit) in confirmation assays using recombinant Ca V 1.2-expressing systems do not necessarily prove that a hit is a false positive. After all, our phenotypic assay is intended for identifying neuro-selective or otherwise mechanistically novel LTCC antagonists, which may well be inactive at recombinant Ca V 1.2-expressing cells if, for instance, the target protein that a hit binds to is not expressed in the recombinant system. Nor is lack of displacement binding, even in the same neuronal preps, indicative of false positivity as the compound may bind to a different site than the tracer used (as illustrated for niclosamide in this study). Therefore, it is critical to verify primary hits from a phenotypic screen by an independent, preferably functional assay (e.g., patch clamp) in a native environment.
Using cortical neurons prepared the same way as for the calcium flux assay, we developed and validated a manual patch clamp assay for hit confirmation. The patch clamp assay confirmed both non-DHPs in the top 10 primary hits, niclosamide and cyproheptadine. Of particular interest, niclosamide has not been noted in the literature for its actions at LTCCs. Combining patch clamp with binding studies, we showed that niclosamide bound to a novel site in neuronal LTCCs with at least two orders of magnitude higher affinity than that at the DHP, phenylalkylamine and benzothiazepine sites. Future studies will be necessary to characterize the novel binding site and molecular mechanism of niclosamide.
In summary, we have developed and validated two phenotypic assays that have promising potential for identifying mechanistically novel and neuro-selective LTCC blockers in large-scale screening campaigns.

Methods
Neuronal cultures. Rat cortical neurons were isolated from E-18 Sprague-Dawley rats as previously described 39 (BrainBits, Springfield, IL). Briefly, cortices were isolated, rapidly removed from decapitated pups, and stored in ice-cold Hibernate-E media. Cortices were then transferred into a 15 mL conical tube and incubated at 37 °C for 10 min in 2 mL papain dilution buffer containing 5 mM L-cysteine, 1 mM EDTA, 10 mM HEPES, 100 µg/mL BSA, 1% papain and 0.1% DNase I. Heat-inactivated FBS (2 mL) and additional DNase I (5 µL) were added at the end of the incubation period after which cells were mechanically dissociated until homogeneous using a 10 mL pipette and centrifuged (5 min at 1000 rpm at room temperature). The cell pellet was then resuspended in culture media containing neurobasal medium supplemented with B27 and 0.5 mM Glutamax. Resuspended cells were centrifuged and the pellet was resuspended again in culture media. Finally, cells were filtered using a 70 µm cell strainer, centrifuged, resuspended in culture media, and plated in Falcon 384-well black wall, clear bottom plates (for the calcium flux assay) or on poly-D-Lysine coated glass coverslips (for the patch clamp assay). Cortical cultures were placed in a 37 °C humidified incubator maintained at 5% CO 2 . One half of the culture media was replaced with fresh media 5 days after plating. Animal protocols were approved by the Institutional Animal Care and Use Committee of Janssen Research & Development, L.L.C. All experiments were performed in accordance with the relevant guidelines and regulations.
Chemicals. TTX was obtained from Alamone (Israel). The PW chemical library was from Prestwick Chemical (France) and contained 1278 small molecular-weight drugs or otherwise bioactive agents. All other reference compounds were from Tocris.

Data analysis.
For the calcium flux assay, the peak Bay K-induced change in fluorescence from baseline, R, was used to calculate the % inhibition value as follows: R = 100*(P-R)/(P-N), where P and N are the peak responses in Bay K alone and Bay K + nimodipine, respectively. Compounds that substantially altered the baseline fluorescence in either direction (≥ ± 30%) were excluded before hit confirmation. For IC 50 determination, concentration-response data were fitted to a logistic function of the form: R = (A 1 −A 2 )/(1 + (C/IC 50 ) h ) + A 2 , where R was the normalized response, C was the compound concentration, IC 50 was the concentration at which half-maximal response occurred, h was the Hill coefficient, and A1 and A2 were constants. To evaluate screen robustness, the Z' value was calculated as follows: Z' = 1-(3*SD positive + 3*SD negative )/(mean positive -mean negative ), where SD is the standard deviation and positive and negative refer to Bay K alone and Bay K + nimodipine, respectively. Statistical analyses were performed using two-tailed Student's t-test or one-or two-way ANOVA with post-hoc Tukey test as indicated.
The percent inhibition data from the PW screen were fitted to a Gaussian function of the form: y = y 0 + A/ (w*sqrt(π))*exp(-2*(x-x c ) 2 /w 2 ), with fitting parameters y 0 as the offset (fixed to zero), A as the area under the curve, x c as center of the peak, and w (= 2.35 SD) as the full width at half maximum (FWHM).
For patch clamp, leak currents were subtracted using a p/4 protocol. For I-V curves, test voltage-evoked peak currents were normalized for each cell before averaging. Concentration-response data were fitted to the same logistic function as for the calcium flux data. Data recorded from n different cells were reported as mean ± SEM. Methods of statistical tests were indicated in figure legends. Data fitting and statistical analysis above were performed using Origin v.9 (Originlab, Northampton, MA). www.nature.com/scientificreports/ For binding experiments, results were expressed (1) as percent of control specific binding: 100 × (measured specific binding/control specific binding) or (2) as percent inhibition of control specific binding in the presence of the test compounds:100-100 × (measured specific binding/control specific binding). The IC 50 values (the concentration that causes half-maximal inhibition of control specific binding) were determined by non-linear regression analysis of the competition curves using Hill equation: Y = D + (A−D)/[1 + (C/IC 50 )] nH , where Y = specific binding, A = left asymptote of the curve, D = right asymptote of the curve, C = compound concentration, and nH = slope factor. Analyses were performed using software developed at Eurofins (Hill software) and validated by comparison with data generated by the commercial software SigmaPlot 4.0. The inhibition constants (K i ) were calculated using the Cheng Prusoff equation K i = IC 50 × (1 + L/K D ), where L = concentration of radioligand in the assay and K D = affinity of the radioligand for the receptor. K D values were determined from Scatchard plots.

Data availability
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.