Anion-Sensitive Fluorophore Identifies the Drosophila Swell-Activated Chloride Channel in a Genome-Wide RNA Interference Screen

When cells swell in hypo-osmotic solutions, chloride-selective ion channels (Clswell) activate to reduce intracellular osmolality and prevent catastrophic cell rupture. Despite intensive efforts to assign a molecular identity to the mammalian Clswell channel, it remains unknown. In an unbiased genome-wide RNA interference (RNAi) screen of Drosophila cells stably expressing an anion-sensitive fluorescent indicator, we identify Bestrophin 1 (dBest1) as the Drosophila Clswell channel. Of the 23 screen hits with mammalian homologs and predicted transmembrane domains, only RNAi specifically targeting dBest1 eliminated the Clswell current (IClswell). We further demonstrate the essential contribution of dBest1 to Drosophila IClswell with the introduction of a human Bestrophin disease-associated mutation (W94C). Overexpression of the W94C construct in Drosophila cells significantly reduced the endogenous IClswell. We confirm that exogenous expression of dBest1 alone in human embryonic kidney (HEK293) cells creates a clearly identifiable Drosophila–like IClswell. In contrast, activation of mouse Bestrophin 2 (mBest2), the closest mammalian ortholog of dBest1, is swell-insensitive. The first 64 residues of dBest1 conferred swell activation to mBest2. The chimera, however, maintains mBest2-like pore properties, strongly indicating that the Bestrophin protein forms the Clswell channel itself rather than functioning as an essential auxiliary subunit. dBest1 is an anion channel clearly responsive to swell; this activation depends upon its N-terminus.

In Drosophila, however, accumulating evidence indicates that dBest1 encodes for a Cl swell channel. RNAi targeting dBest1 eliminates Drosophila Schneider (S2) cell I Clswell , an effect rescued by re-introduction of dBest1 [15]. Further, swell activated dBest1 mutants have altered biophysical properties and reactivity to sulfhydryl reagents [16]. dBest1 likely forms the chloride conducting pore, but it may be an obligate auxillary subunit of Drosophila I Clswell that modifies channel properties akin to CaV b subunits [17].
Assigning chloride channel function to any protein is difficult. The known chloride channel families (e.g., ClC, Anoctamin/ TMEM16, CFTR, and ionotropic GABA A and GlyR) lack structural pore or gating motifs that might form the basis for in silico identification. Expression cloning approaches have also failed due to widespread Cl swell channel expression that precludes the separation of endogenous and over-expressed protein activities. Moreover, known chloride channels blockers are non-specific and their affinities are far too low to encourage affinity purification. Finally, previous chloride indicators are poor tools for screening due to loading and retention issues, inconsistent results, and poor reproducibility [18].
Here we present an unbiased genome-wide, high-throughput RNAi screen designed to identify the Drosophila Cl swell channel and its regulators. Our screen employed H148Q-YFP, a genetically encoded anion-sensitive yellow fluorescent protein [19], to report Cl swell activity in Drosophila S2R+ cells. Of our 595 initial hits that altered chloride handling, we concentrated on characterizing proteins with mammalian homology and at least one transmembrane domain as potential Cl swell channels. dBest1 emerged from our screen as the lead candidate for Drosophila Cl swell . Both RNAi knockdown of dBest1 and overexpression of a dominant-negative dBest1 eliminated the Cl swell current in Drosophila S2R+ cells. Conversely, dBest1 overexpression in a mammalian system (HEK cells) produced a Drosophila-like I Clswell . To identify domains necessary for swell activation we characterized chimeras between the swell-sensitive dBest1 and the swell-insensitive mBest2. Swell sensitivity is only apparent in mBest2, the closest mammalian homolog of dBest1, when the protein contains the dBest1 amino (N)-terminus. This chimera maintains the pore properties of mBest2, providing additional evidence that the protein itself forms a channel rather than functioning as a necessary auxiliary subunit. We conclude that dBest1 is the channel underlying the Drosophila I Clswell .

Drosophila S2R+ Cells have Robust I Clswell
Drosophila S2 cells are used extensively in genome-wide RNAi screens to dissect signaling pathways, determine protein functions, and assign protein molecular identity [20]. These macrophage-like cells, derived from primary culture of late stage Drosophila melanogaster embryos [21] readily take up RNAi from serum-free media. The subsequent process of targeted mRNA ablation is efficient and highly reproducible [22,23]. For our screen we used S2R+ cells [24], an adherent S2 variant well suited for assays that require multiple solution changes. Importantly, S2R+ cells have a consistent, large I Clswell that activates slowly upon a drop in extracellular osmolality ( Figure 1A). I Clswell starts to activate within 2 min exposure to hypo-osmotic media, reaching steady state activation by 5 min. The fully activated Cl swell conductance is anion selective ( Figure 1B). The relative permeability sequence of S2R+ Cl swell is I = SCN.Cl.MES.aspartate (ASP) while the slope conductance sequence is I = SCN = Cl.MES.ASP (Table 1 & 2). The Cl swell I-V relationship has a slight ''S'' shape, revealing rectification. An extended step protocol further illustrates features of S2R+ Cl swell ( Figure 1C). I Clswell exhibits an initial instantaneous activation followed by a second slow activation phase, suggesting that more than one type of Cl 2 conductance is turned on in S2R+ cells with cell swelling. In contrast, Chien & Hartzell [15] reported a single phase, time-independent I Clswell activation in their S2 cells, perhaps indicating Cl 2 channel expression differences in the two cell lines. Tail currents, normally indicating time dependence of deactivation of the channel, are evident in our recordings. However, since tail currents were not observed under symmetrical recording conditions [15], we attribute these currents to the exit of intracellular Cl 2 accumulated during the prolonged steps. S2R+ I Clswell has an interesting pharmacological profile (Figure 2A). Even at a high concentration (100 mM), the non-specific chloride channel blocker 4,49-diisothiocyano-2,29-stilbenedisulfonic acid (DIDS) [25] blocks less than 25% of the S2R+ I Clswell (Figure 2A). 4-2(2-butyl-6,7-dichloro-2-cyclopentyl-indan-1-on-5yl (DCPIB), a mammalian selective I Clswell blocker [26], fails to completely block S2R+ I Clswell at 30 mM (Figure 2A & E). DCPIB blocks in a voltage-dependent manner ( Figure 2E); at 0 mV 52% of S2R+ I Clswell is blocked, while 90% is blocked at 80 mV. Surprisingly, furosemide, a Na-K-2Cl cotransporter (NKCC; SLC12A2) blocker, almost completely inhibits S2R+ I Clswell at 1 mM ( H148Q-YFP Reliably Reports the Activity of S2R+ Cell Cl swell Channels Cl swell conducts iodide better than chloride, favoring the use of the H148Q-YFP indicator as a reporter of its activity (I 2 K D = 20 mM and Cl 2 K D = 100 mM [19,27,28]). Several anion-sensitive YFP variants accurately quantify intracellular Cl 2 concentration or changes [27,[29][30][31]; anion binding near the YFP chromophore suppresses fluorescence emission by altering chromophore resonance [19]. H148Q-YFP was chosen for Cl swell detection because it is bright and potently suppressed by I 2 ; these properties are critical for good signal-to-noise ratios during screening. H148Q-YFP (pK a = 6.7) is also sensitive to intracellular pH changes [19,28]. S2R+ cells stably expressing H148Q-YFP maintain their fluorescence in 240 mOSM NaCl however ( Figure 3A), indicating that cell swelling does not appreciably alter intracellular pH. Subsequent replacement of bath Cl 2 with I 2 rapidly suppresses indicator fluorescence by 50% as I 2 enters the cells through open channels and interacts with the probe. The large fluorescence change and low intrinsic assay variability favor clear separation of potential hits. In the absence of hypo-osmotic solution, I 2 is unable to enter the S2R+ cells and fluorescence is maintained ( Figure 3B), indicating that S2R+ cells lack alternative constitutively active I 2 entry pathways that could confound our ability to identify the Cl swell channel. Further, furosemide block of open Cl swell channels prevents appreciable fluorescence suppression ( Figure 3C & D), suggesting that RNAi effectively targeting the Cl swell channel will be readily identifiable as hits.
Genome-wide RNAi Screening of H148Q-YFP S2R+ Cells Identifies dBest1 as the Cl swell Channel The primary screen was conducted at the Harvard/HHMI Drosophila RNAi Screening Center using our stable H148Q-YFPexpressing S2R+ cell line. Each well of sixty-six 384-well assay plates contained a dsRNA targeting 1 of 13,900 genes encoding proteins or non-coding RNAs (DRSC 2.0; Figure 4A). Five days after S2R+ cells were treated with RNAi, we assessed cellular fluorescence under swell conditions in the presence of Cl 2 and I 2 . Wells with fluorescence or ratio (I 2 fluorescence/Cl 2 fluorescence) changes greater than 1.5 times the standard deviation of the plate mean were initially considered as hits (Cl swell channel candidates or regulators of its activation pathway). We pared the list of 595 hits to genes with mammalian homologs and those with predicted transmembrane domains ( Figure 4B, Table S1). In a secondary screen, we confirmed that each RNAi significantly reduced swelling-induced fluorescence and targeted only the mRNA from the identified gene (qPCR). We then directly measured I Clswell via whole-cell voltage clamp. Candidates genes, whose RNAi significantly reduced the S2R+ cell I Clswell , were cloned and expressed in HEK293 or CHO-K1 cells. I Clswell was then measured via wholecell recording and compared with currents from untransfected cells. The only candidate of our screen to satisfy all the criteria for a Cl swell channel was dBest1 (Table S1).
DRSC26457 RNAi Targeting dBest1 Eliminates I Clswell dBest1 is a protein of 769 amino acids containing 4 transmembrane domains [32,33] (Figure 4C). It is one of four Bestrophin family members in Drosophila, with highest homology to mBest2/hBest2 (51% identity and 67% similarity; BLAST). Hartzell and colleagues first proposed that dBest1 was a chloride channel activated by high intracellular Ca 2+ and cell swelling [15,16]. In our H148Q-YFP fluorescence assay dBest1 RNAi DRSC26457 abrogated the fluorescence change normally observed when I 2 enters the S2R+ cells through activated Cl swell conductances ( Figure 5A, B). Interestingly, DRSC26457 also decreased the baseline fluorescence variability of S2R+ cells ( Figure 5B), suggesting that I dBest1 contributes to resting intracellular Cl 2 concentrations. S2R+ I Clswell was essentially eliminated by dBest1 RNAi DRSC26457 treatment ( Figure 5C). This RNAi specifically and effectively reduced dBest1 mRNA by 91.5% 60.5 (n = 3; qPCR); mRNA levels for the 3 remaining Bestrophin members and other Cl swell candidates were unaffected. A second RNAi targeting dBest1 (DRSC16909; corresponds with dB1S [15]) was part of our initial screen. It was less effective at knocking down dBest1 mRNA (85% reduction, n = 3; significantly less than DRSC26457; p,0.001, Student's t-test) and had two predicted off-target hits: CG4623 (20/20) and CG16711 (18/18). DRSC16909 did not significantly alter H148Q-YFP Iinduced  fluorescence suppression ( Figure 5A), and was not a hit in our initial screen. It is possible that the 15% remaining mRNA translated sufficient amounts of functional dBest1/Cl swell channels to exclude it as a hit in our screen. This prospect emphasizes the importance of validated, effective RNAi for accurate screening.

Mutant dBest1 W94C Significantly Reduces S2R+ I Clswell
To substantiate the conclusion that dBest1 is an essential component of the Cl swell channel, we tested whether a mutant dBest1 would act as a dominant negative regulator of I Clswell . In humans, Bestrophin 1 is mutant in vitelliform macular dystrophy    (Best's disease [34,35]). One mutation, W93C, occurs in a conserved sequence of the channel's putative pore [36,37] ( Figure 4C). When we expressed the homolog dBest1 W94C-gfp in S2R+ cells, I Clswell was significantly reduced ( Figure 5D). Interestingly, the late activating component of I Clswell remained clearly evident at depolarized potentials ( Figure S1). We could not study this current in more detail as a loss of cell membrane integrity rapidly ensued. We conclude that dBest1 is responsible for the early activating S2R+ cell I Clswell . W94C might interact with WT dBest1 to disrupt the Cl swell channel pore or it may prevent proper protein trafficking [38]. In S2R+ cells, dBest1 W94C-gfp has a distinct intracellular expression pattern unaltered by osmotic changes (Figure 5E), suggesting the latter explanation over the former. Regardless, dBest1 W94C has a dominant negative impact on I Clswell , further evidence that dBest1 is integral to the Cl swell channel. Another disease-associated Bestrophin mutation, D308A, occurs in a putative Ca 2+ -binding bowl located in the channel's Cterminus ( Figure 4C; blue star). D308A is proposed to eliminate Bestrophin activation by disruption of calcium binding [39]. We introduced this mutation into dBest1 to determine if activation by calcium and cell swelling could be separated. Unfortunately dBest1 D308A-gfp was not functional in HEK cells (data not shown). Three possible explanations may underlie this result: 1) activation by multi-modal stimuli is simultaneously disrupted by the mutation; 2) the mutation causes protein misfolding and the channel function has been eliminated for reasons unrelated to activation; 3) the mutant channel is mislocalized. Our GFP-tagged protein was expressed (data not shown), but we cannot exclude the possibility that it mislocalizes or fails to interact appropriately with other proteins necessary for I Clswell activation or channel function [34].

Exogenous dBest1 Expression Creates a Drosophila-like I Clswell
Exogenous expression of a candidate protein substantiates whether the protein is necessary and/or sufficient in a given process. Our secondary screen assessed whether candidate protein expression resulted in a novel I Clswell or augmented the endogenous HEK I Clswell ( Table S1). The HEK cell line chosen for candidate over-expression lacked constitutive I Cl and I SCN (potentially contaminating conductances attributable to SLC1A family member expression [40]; data not shown). The endogenous HEK I Clswell develops very slowly ( Figure 6A & B); a two fold increase was noted within the first 2 min of swell. Once the HEK I Clswell reaches steady state, however, it has increased more than forty fold (44.4610.7 fold, n = 29; Student's t-test, p,0.000005). Tail currents are absent ( Figure 6B & C). Characteristic voltagedependent inactivation develops during steps to positive potentials ( Figure 6C). HEK I Clswell is anion selective; its permeability and conductance sequences match closely to those of S2R+ I Clswell ( Figure 6D; Table 1 & 2). The HEK I Clswell pharmacological profile ( Figure 2C) correlates well with the literature. 100 mM DIDS, slightly above the reported IC 50 [41], blocks 78% of the HEK I Clswell at +80 mV ( Figure 2C). DCPIB has an IC 50 of 4 mM [26]; at 30 mM 100% of HEK I Clswell is blocked ( Figure 2C). 1 mM furosemide barely inhibits HEK I Clswell ( Figure 2C). The endogenous HEK I Clswell recapitulates the key features noted for the mammalian I Clswell [1].
Bestrophin proteins are not universally accepted as bona fide chloride channels; alternatively they are intracellular ion channel regulators [33,42,43]. dBest1-gfp is clearly observed on or near the surface of HEK-293 cells ( Figure 7A). Its expression results in a Drosophila-like I Clswell (Figure 7B-E). Constitutively active I dBest1 is apparent in iso-osmotic 320 mOSM solution and is significantly increased 1664.5 fold (Student's t-test, p,0.005) during the first 2 min of hypo-osmotic stimulation ( Figure 7B). I dBest1 has the same ''S''-shaped rectification as Drosophila I Clswell during ramps ( Figure 7C); tail currents and time-dependent activation are both apparent in the step protocol (Figure 7d). I dBest1 is anion selective; it has the same permeability and conductance sequences as S2R+ I Clswell and HEK I Clswell ( Figure 7E; Table 1 & 2). Strikingly, I dBest1 and the endogenous S2R+ I Clswell share a similar pharmacological profile that differs significantly from HEK I Clswell (Figure 2A-C). 100 mM DIDS inhibits 35% of I dBest , while 30 mM DCPIB blocks 45%. 1 mM furosemide blocks nearly 100% of the I dBest1 . We conclude that dBest1 expression results in a Drosophila-like I Clswell ; it cannot be attributed to endogenous HEK I Clswell upregulation.

dBest1 Swell Activation can be Conferred on the Swellinsensitive mBest2
The structural domains necessary for swell-induced channel activation are unknown. Although dBest1 has a long poorly conserved C-terminus ( Figure 4C), it is not necessary for swell activation. dBest1 remains swell-sensitive despite the removal of up to 338 of its C-terminal amino acid residues (Stop 383, Figure 4C; Figure 8A). Next we examined whether chimeras might reveal the domains underlying swell activation. dBest1's closest mammalian homolog, mBest2, is not activated by hypo-osmotic solutions ( Figure 8B). Chimera d64m (the first 64 residues are dBest1; the remaining residues are identical to those of mBest2; Figure 4C) expression resulted in a constitutively active current that more than doubled with swelling (2.3 fold 60.3 increase; Student's t-test, p,0.05; Figure 8B-D). The d64m chimera maintained the relative permeability and slope conductance of mBest2 (Figure 8E  & F; Table 1 & 2), suggesting that the channel's pore domain is downstream of residue 64. Two other groups have assessed mBest2 selectivity [44,45] and found greater permeability for SCN than we report here. Both groups used high intracellular calcium to activate I mBest2 ; we report constitutive I mBest2 measured with high internal calcium buffering (i.e. ,10 nM free calcium). Our HEK cell line was also screened for potentially contaminating I SCN (data not shown) attributable to SLC1A family member expression [40]. The pharmacological profile of I d64m noticeably diverged from both that of HEK I Clswell and I dBest1 (Figure 2). Furosemide blocked 75% of I d64m , while DIDS and DCPIB both blocked I d64m to near completion ( Figure 2D). We conclude that the dBest1 Nterminal domain is required for swell activation of the mBest2 channel. The reverse chimera (m64d) was nonfunctional; exogenous currents were not observed with swelling or in the presence of high intracellular Ca 2+ (data not shown). We cannot conclude with this data however, that the N-terminus is a ''swelling'' domain as it lacks any predictive motifs. We hypothesize that it works in concert with domains present both in dBest1 and mBest2 to facilitate swell activation. The strong correlation between S2R+ I Clswell and I dBest1 , combined with the unique selectivity and pharmacology of the d64m chimera, support the conclusion that the Bestrophin protein itself forms the Cl swell channel rather than functioning as an auxiliary subunit.

Discussion
Our study validates the H148Q-YFP fluorophore as a reliable reporter of Cl swell channel activity in genome-wide RNAi screening studies. H148Q-YFP has been employed very effectively in the identification of novel chloride channel activators, modulators, and blockers of CFTR and Ca 2+ -activated Cl 2 channels [18]. This is the first reported RNAi screen using an anion-sensitive fluorescent protein to assign molecular identity to a chloride channel. Our screen supports the findings of the Hartzell lab [15]: dBest1 RNAi eliminates Drosophila I Clswell .
We found that the RNAi effectiveness was essential for Cl swell candidate identification. Two separate dBest1-targeting RNAi's were part of our initial screen: DRSC26457 and DRSC16909 (which corresponds to dB1S [15]), but only DRSC26457 was a hit. qPCR reported a 95% reduction in dBest1 mRNA with DRSC26457 treatment versus an 85% reduction with DRSC16909. Hartzell and colleagues found that S2 cell I Clswell was significantly reduced following treatment with 0.4 mg of DRSC16909 [15], while in our screen each assay well had a standardized 0.25 mg of RNAi. Using more RNAi may have effected greater target knockdown and resulted in the detection of DRSC16909 as a hit in our screen. This result emphasizes the importance of RNAi effectiveness in hit identification.
Exogenously expressed I dBest1 and endogenous S2R+ I Clswell share similar characteristics, including time-dependent activation, tail currents, relative permeability sequences, slope conductance sequences, and pharmacological profiles. The shared properties of Drosophila I Clswell and I dBest1 suggest that the same protein forms the channel responsible for both. Bestrophin is a known ion channel modulator, altering voltage-gated calcium channel activity [46]. If dBest1 expression simply modulated or upregulated the endogenous HEK Cl swell channel expression, we would have expected the resulting I Clswell to maintain the properties of HEK I Clswell . Instead we observed that dBest1 introduced an exogenous Drosophila-like I Clswell whose development preceeded that of the endogenous HEK I Clswell . I dBest1 matched the pharmacological profile of the S2R+ I Clswell . Moreover, we found that the exogenous Drosophilalike I Clswell permeability and conductance sequence could be transformed into that of mBest2 with the d64m chimera. The pharmacological profile of I d64m was again significantly different from the endogenous HEK I Clswell . We conclude that dBest1 is the Drosophila Cl swell channel.
Several Bestrophin mutations are associated with vitelliform macular dystrophy [36,37]. How these mutations are causally linked to the disease is not clear. Here we found that overexpression of the disease-linked W94C dBest1 mutant in S2R+ cells significantly suppressed the endogenous Drosophila I Clswell . The W94C mutation occurs in the putative pore of dBest1 and thus may disrupt Cl swell conductance. However, the fluorescently tagged W94C dBest1 protein appears to localize to intracellular compartments, consistent with mislocalization. Milenkovic et al, have recently proposed that disease-associated Bestrophin mutations cause defects in intracellular trafficking [38]. Both scenarios may explain the dominant negative effect of dBest1 W94C on Drosophila I Clswell : non-functional, pore-disrupting, mutant Bestrophin proteins complexing with wild-type dBest1 may be largely retained within the endoplasmic reticulum. The end result would be the elimination of endogenous Drosophila I Clswell . Our experiments support the hypothesis that mutant Bestrophin W93C expression could significantly disrupt chloride flux and homeostasis in the human macula, contributing to the disease state.
The distinction of Bestrophin function in Drosophila versus mammalian cells is most clearly illustrated by Hartzell and colleagues [16]. I Clswell measured in peritoneal mast cells isolated from mBest1 2/2 , mBest2 2/2 , and mBest1/2 double knockout mice was identical to wild-type I Clswell . hBest1 and mBest2 are swell sensitive in that their currents are inhibited by hyperosmotic solutions. However, their activity does not increase with swell [12]. We confirm here that mBest2 activation does not increase when cells swell. Our d64m chimera contained only a small portion of dBest1, yet it responded to cellular swelling. The crucial Nterminal region contains no distinct association domains or predictive structures that might explain its coupling to changes in cell stretch, tension, or osmolality. We speculate that the Nterminus contributes to a required tertiary structure that enables swell signaling events to activate the dBest1 channel.
Our genome-wide RNAi screen of S2R+ cells and follow-up study firmly establishes that the dBest1 protein forms the Drosophila Cl swell channel. It further validates a live cell genetically engineered fluorescent screening platform to identify other mammalian chloride channels.

Generation of the S2R+ YFP-H148Q Stable Cell Line
The stable S2R+ cell line expressing a halide-sensitive YFP (H148Q-YFP; kindly provided by Dr. Alan Verkman, UCSF) was generated with a selection vector (pCoBlast). H148Q-YFP was subcloned into the pAc5.1/V5-HisA vector (Invitrogen, CA). The S2R+ cells were transfected by electroporation (Amaxa cell line nucleofector kit V; Lonza). Cells were placed under selective pressure with 25 mg/ml blasticidin for 2 weeks. Two rounds of fluorescence-activated cell sorting (FACS; DFCI Flow Cytometry Core Facility) normalized YFP fluorescence intensities. These S2R+ cells exhibited a robust I Clswell as described in Figure 1. S2R+ cells were maintained in Schneider's Drosophila medium (Invitrogen), with 10% heat inactivated fetal bovine serum (Invitrogen), and 1% penicillin/streptomycin (Sigma-Aldrich). For the primary screen, cells were spun down and resuspended in serum-free medium at a density of 9610 4 cells/ml. 10 ml of the cell suspension was added to each of the 384 wells using the Matrix Wellmate 8-channel microplate dispenser (ThermoScientific; DRSC). Cells were incubated for 30 min, then 30 ml of serum containing medium were added to each well. Cells were cultured for 5 d before the swell assay was performed (5 d RNAi treatment is necessary to sufficiently knock down proteins with slow turnover rates). For the secondary screen, cells were plated at a density of 40% in a 6 well dish. Once cells were adherent, the medium was replaced with 1 ml of serum-free medium containing 0.015 mg/ml dsRNA. Cells were incubated for 30 min at room temperature followed by addition of 3 ml of serum-containing medium to each well. Transfections were performed in duplicate; 1 well was used for functional studies and the other for qPCR analysis of knockdown at day 5.

The Primary Screen
Our genome-wide screen was conducted at the Harvard/ HHMI Drosophila RNAi Screening Center using the DRSC 2.0 Genomewide RNAi Library. DRSC 2.0 is a collection of dsRNAs for genome-wide RNAi knockdown covering , 13,900 genes encoding proteins and non-coding RNAs while minimizing offtarget effects due to sequence similarity to other genes. Each gene is targeted by 1.3 dsRNA/gene. The screen consisted of 66 384well assay plates in duplicate. Each well of the 384-well plate contained 5 ml of 0.05 mg/ml dsRNA in water (0.25 mg dsRNA/ well). Each plate contained control RNAi specific for Thread (Drosophila inhibitor of apoptosis protein), Rho (a small GTPase activator of the EGFR signaling pathway), and GFP. On day 5 of RNAi treatment, the cellular fluorescence of the H148Q-YFP probe was measured under several treatment conditions using the Analyst GT plate reader (Molecular Devices; DRSC). The probe was excited at 485 nm and emissions collected at 530 nm. Before fluorescence measurements were taken, the media was aspirated (384-well aspirator; VP Scientific) and the cells were equilibrated in 80 ml of 320 mOSM NaCl solution. After 10 min this solution was removed by aspiration and cells incubated in 240 mOSM NaCl for 5 min. Fluorescence was then measured, the NaCl solution removed and cells were incubated in 240 mOSM NaI for 5 min. Fluorescence was again measured; the change in fluorescence was determined by dividing the fluorescence in 240 mOSM NaI by that in 240 mOSM NaCl. Wells with fluorescence or ratio changes (240 mOSM I 2 fluorescence/240 mOSM Cl 2 fluorescence) greater than 1.5 times the standard deviation (1.56S.D.) of the plate mean were initially considered as hits (candidates of Cl swell channel or regulators of its activation pathway). False positives could potentially result if the RNAi treatment caused a high internal pH as H148Q-YFP has a pKa of 6.7. Cell death was detected in control wells indicating effective RNAi treatment.
Generation of dsRNA cDNA templates were generated by PCR amplification of genomic DNA using primers designed by the DRSC (SnapDragon tool). These primers had the T7 promoter sequence (TAATAC-GACTCACTATAGGG) added to the 5' end of both primers. The templates generally corresponded to exons but occasionally sequences with two or more exons interrupted by introns were used. The PCR fragments were ,150-600 base pairs in length, and any complete 19-mer homology to other genes that could lead to non-specific dsRNA are reported. Individual RNAi sequences used here are found in the DRSC website (www.flyrnai.org). dsRNAs against Drosophila were synthesized with the MEGAscript in vitro transcription kit (Ambion). RNA was purified with the RNeasyPlus mini kit (Qiagen) and stored at 280uC.

qPCR Analysis of RNAi Efficiency
After 5 d RNAi treatment, RNA was prepared from the S2R+ cells using the RNeasy Plus mini kit (Qiagen). 2.5 mg RNA was used for each first-strand cDNA synthesis reaction (SuperScript Vilo cDNA Synthesis kit, Invitrogen). Primers for qPCR were designed on the NCBI/PrimerBlast site (http://www.ncbi.nlm. nih.gov/ tools/primer-blast/) with the following restrictions: PCR product size was between 70 and 300 bp, primer melting temperatures were between 57 and 63uC, primers spanned an exon-exon junction, and primers were specific to the intended PCR template as determined by BLAST analysis of the Drosophila melanogaster Refseq mRNA database. Primer sets were only used if the melting curve had a single peak. The RT 2 Real-Time SYBR Green/Rox PCR master mix (SABiosciences) was used for qPCR. qPCR reactions were set up in quadruplicate to minimize pipetting errors, and run on the Mastercycler ep Realplex real-time PCR system (Eppendorf). Average cycle numbers for each primer set were normalized to either dTaf8 or dAct79b average cycle numbers.

Secondary Screening of Candidates
Comprehensive bioinformatics analysis of the hit list was performed to identify potential candidates for Cl swell . Hits were limited to those with human homologs and at least a single transmembrane domain. Potential regulators of the Cl swell activation pathway were left for future consideration. The effects of RNAi on fluorescence changes were confirmed by plate reader or imaging experiments. The specificity and effectiveness of the RNAi was assessed by qPCR. S2R+ cells treated with RNAi were patch clamped and I Clswell was directly measured. Candidates were cloned and expressed in HEK293 or CHO-K1 cells. I Clswell was measured by whole-cell patch clamp recording.

Electrophysiology
Whole-cell patch clamp recordings were made at room temperature. Recordings were obtained using an Axopatch 200B amplifier, Digidata 1322A analog-to-digital converter, and pClamp 8.01 software (Molecular Devices, Union City, CA). Data were low-pass filtered at 2 kHz and digitized at 5 kHz. Fire-polished thin or thick wall borosilicate glass pipettes of 3-4 MV resistances were used for recordings; access resistance was compensated to .80%. Cells were held at 270 mV to clearly eliminate cells with leaky seals and voltage ramps (400 ms in duration) from 2100 to +100 mV were applied every 2-5 s. Liquid junction potentials were corrected during analysis, and ramp data were plotted between -100 and +80 mV.

Recording Solutions
Internal pipette solution contained (in mM): 160 CsASP, 10 Cs4BAPTA, 4 MgATP, 2 MgCl 2 , 8 NaCl, and 10 HEPES (pH 7.4 with CsOH). 10 mM BAPTA was used to prevent activation of channels by calcium and to reduce the endogenous HEK cell I Clswell , which is optimally activated with 100 nM Ca i [47]. 240 mOSM solution composition is detailed in 'Screening solutions'. 200 mOSM solutions contained in mM: 82.5 NaCl, NaI, NaSCN, NaMES, or NaASP, 2 CaCl 2 , 1 MgCl 2 , 5 KCl, 10 HEPES, and 10 Glucose (pH 7.4 with NaOH). 90 mM mannitol was added to bring osmolality to 320 mOSM. The relative permeabilities were estimated from the Goldman-Hodgkin-Katz equation. For our calculations, the [Cl] i was set to 0 mM. Cells were held at -70 mV during all recordings, rapidly depleting Cl i . Cation permeability was essentially nil, as replacement of 200 mOSM NaCl solution with 200 mOSM NMDG-Cl solution did not change the reversal potential (E rev ; data not shown). Slope conductances were calculated for each anionic substitution between the E rev and +80 mV. For the S2R+ and HEK cells E rev and I +80 mV were measured in 200 mOSM solutions. For exogenously expressed dBest1, mBest2, and d64m E rev and I +80 mV were measured in 320 mOSM solutions to prevent contamination with the endogenous HEK cell I Clswell . The HEK cell line chosen for over-expression studies had no constitutive I Cl (I 320 mOSM did not change when switching between Cl and ASP), the endogenous I Clswell developed very slowly, and the cells did not have an endogenous I SCN (attributable to SLC1A family member expression [40].

Molecular Biology
dBest1 was a kind gift from Dr. Criss Hartzell (Emory University). All other constructs were either ordered from Open Biosystems or cloned from a Drosophila cDNA library or Human Brain (whole Marathon ready cDNA library; BD Biosciences). Candidate cDNAs were subcloned into pEGFP-N3 (C-terminal tag; BD Biosciences) and an engineered Red pTracer vector (untagged). We found that an N-terminal EGFP tag rendered dBest1 nonfunctional (data not shown). Using site-directed mutagenesis we introduced the W94C mutation into dBest1 pEGFP-N3 (GeneArt site-directed mutagenesis, Invitrogen, CA). dBest1 W94C-gfp was then subcloned into the pAc5.1 V5-HisA vector (Invitrogen, CA). Table S1 Secondary screening identifies Best1 as the Drosophila Cl swell channel. Candidates with transmembrane domains and human homologs were further studied to determine if they formed the Cl swell channel. 3 indicates a positive secondary screening result; X indicates a negative result. B indicates that several qPCR primer sets consistently had more than 1 melting point peak suggesting nonspecific primer binding. The effectiveness of RNAi knockdown, therefore, could not be determined by qPCR. ?? indicates that two cells overexpressing SLC1A2 had substantial I SCN-currents but small I Clswell . Thus, SLC1A2 overexpression may upregulate endogenous HEK cell I Clswell in the majority of the population but does not form the channel itself. HeLa cells treated with SLC1A3 siRNA (which reduced SLC1A2 and SLC1A3 mRNA by 90% and 92% respectively) had unaltered I Clswell (data not shown). (TIF)