C57BL/6J mice and hGFAP-CreERT2 were used at 8~10 weeks of age. The hGFAP-CreERT2 mice are inducible transgenic mice under control of human GFAP promoter and estrogen. All mice were kept on a 12 hours light-dark cycle in a specific-pathogen-free facility with controlled temperature and humidity and had free access to food and water. All experimental procedures were conducted according to protocols approved by the directives of the Institutional Animal Care and Use Committee of KIST (Seoul, Republic of Korea).

Primary cultured astrocytes were prepared from cortex of C57BL/6 mouse pups (P0-P2) as described [37]. The cerebral cortex was dissected from the brain and adherent meninges were removed, minced and dissociated into single cell suspension by trituration. Cells were cultured in Dulbecco's modified Eagle's medium (DMEM, #10-013, Corning) supplemented with (in mM) 25 glucose, 4 L-glutamine, 1 sodium pyruvate, 10% heat-inactivated horse serum (#260500-088, Gibco), 10% heat-inactivated fetal bovine serum (#10082-147, Gibco) and 10,000 units/ml penicillin-streptomycin (#15140-122, Gibco). Cultures were maintained at 37℃ in a humidified 5% CO₂ incubator. On DIV3, cells were vigorously washed with repeated pipetting and the media was replaced to get rid of debris and other floating cell types. During maintaining the culture before use, the media was replaced every 3~4 days. For gene-silencing experiments with primary cultured astrocytes, various shRNA vectors were electroporated (Neon Transfection system kit; #MPK10096, Invitrogen) into trypsinized cultured astrocytes 4 days before the experimental day and replated onto culture dish. One day before the experimental day (for recording I_Cl,swell, DIV21 primary astrocyte culture were used), cells were replated onto cover-glass coated with 0.1 mg/ml Poly D-Lysine (PDL, #P6407, Sigma-Aldrich).

Human embryonic kidney 293T (HEK293T) and Chinese hamster ovary-K1 (CHO-K1, referred as CHO) cells were purchased from ATCC (#CRL-3216, ATCC) and the Korean Cell Line Bank (Seoul National University, Republic of Korea), respectively. All cell lines have been tested for mycoplasma contamination. HEK293T cells and CHO cells were cultured in DMEM (#10-013, Corning) and F-12 (#21127-022, Gibco), respectively. Both media were supplemented with (in mM) 25 glucose, 4 L-glutamine, 1 sodium pyruvate, 10% heat-inactivated fetal bovine serum (#10082-147, Gibco) and 10,000 units/ml penicillin-streptomycin (#15140-122, Gibco). Cell lines were maintained at 37℃ in a humidified atmosphere of 95% air and 5% CO₂. 18 hours before the experimental day, cells were transfected with DNA clone by transfection reagent (Effectene, #301425, Qiagen). On experiment day, the transfected cells were replated onto cover-glass for electrophysiological recordings. For immunocytochemistry, HEK293T cells were plated onto PDL coated cover-glass, and transfected with DNA clone 20 hours before fixation.

Various kinds of shRNA were synthesized as for following: Two kinds of oligos were purchased (Sequence of oligos; [phos]5′-t [sense sequence of target] ttcaagaga [reverse complement sequence of target] ttttttc-3′ and [phos]5′-tcgagaaaaaa [sense sequence of target] tctcttgaa [reverse complement sequence of target] a-3′). The sequence information for control shRNA and various candidates of shRNA are listed in Table 1.Two oligos were annealed with annealing buffer (in mM; 200 potassium acetate, 60 HEPES-KOH, 4 Mg-acetate, pH 7.3 was adjusted by KOH) and incubated at 95℃ for 5min and 70℃ for 10 min. The annealed double-stranded oligo was inserted into HpaI-XhoI restriction enzyme sites of pSicoR lentiviral vector [44] (#11579, Addgene) and verified by sequencing. The shRNA containing pSicoR vectors were electroporated into cultured astrocytes. In experiments using triple combination of TTYH shRNA, we used either (pSicoR-Ttyh1-shRNA-GFP, pSicoR-Ttyh2-shRNA-mcherry and pSicoR-Ttyh3 shRNA-mcherry) or (pSicoR-scrambled-shRNA-GFP and pSicoR-scrambled-shRNA-mcherry). The sequence of each shRNAs are listed in the Table 1.

The 7~9 weeks old male mice were anesthetized with 2% tribromoethanol (Avertin, 20 µl/g) and placed in a stereotaxic frame. The scalp was incised and exposed the skull. The connective tissue was gently scraped away. Stereotaxic coordinates were: AP: −1.7 mm; L: 1.7 mm; and DVL:−1.85 mm, according to Allen Mouse Brain Atlas (http://www.brain-map.org). The cocktail of lentiviruses carrying either (pSicoR-Ttyh1-shRNA-GFP, pSicoR-Ttyh2-shRNA-mcherry, and pSicoR-Ttyh3-shRNA-mcherry) or (pSicoR-scrambled-shRNA-GFP and pSicoR-scrambled-shRNA-mcherry) were bilaterally injected into hippocampal CA1 stratum radiatum of naïve C57BL/6J mice for slice patch-clamp recording or tamoxifen-treated hGFAP-creERT2 mice for intrinsic optical signal (IOS) imaging. The adeno-associated virus (AAV) of Lentivirus carrying pSicoR-Ttyh1-shRNA-GFP or scrambled-shRNA-GFP was also injected into the CA1 stratum radiatum of naïve C57BL/6J mice for confirming the knock-down efficiency with immunohistochemistry. 10~14 days after virus delivery by stereotaxic injection, the mice were sacrificed for next experiments at 9~10 weeks of age.

The tamoxifen-inducible Cre-mediated recombination is expected to result in deletion of the floxed sequences. To activate Cre recombinase, 7~9 weeks-old male mice were administrated with tamoxifen (#T5648, Sigma-Aldrich) once a day (200 mg per kg of body weight, intraperitoneal) for 7 days. Tamoxifen was dissolved in sunflower oil (#1642347, Sigma-Aldrich) with 10% ethanol at a final concentration of 20 mg per milliliter at room temperature, and stored at 4℃ in the dark. The shRNA is floxed with loxP sites in the pSicoR construct, which is cleaved by Cre-expression [44]. The cocktail of lentiviruses carrying pSicoR-Ttyh1-shRNA-GFP, pSicoR-Ttyh2-shRNA-mcherry and pSicoR-Ttyh3-shRNA-mcherry were injected into CA1 stratum radiatum of the hippocampus in tamoxifen-treated hGFAP-creERT2 transgenic mice, which allowed us to achieve astrocyte-specific recovery of TTYH family expression in the hippocampal CA1 stratum radiatum.

The design for Ttyh1-shRNA-insensitive-GFP clone construction was done by changing every third single nucleotide of Ttyh1 shRNA target site to another nucleotide that makes same amino acid due to the codon redundancy (Ttyh1-shRNA target sequence- 5′- GCACAGAAGCCCTTGTTATCC -3′; shRNA-insensitive sequence- 5′-GCGCAAAAACCTTTATTGTCA-3′ included 7 mismatches, also described in Fig. 2B). The site-directed mutagenesis kit was used for the construction of Ttyh1-shRNA-insensitive-GFP vector (Ezchange™ site-directed mutagenesis kit, EZ400, Enzynomics). The sequence information of oligomers for cloning of mutants clone are listed in Supplemental Table 2.

All truncation mutant and single amino acid mutants of TTYH1 were constructed in shRNA-insensitive TTYH1 tagged with GFAP clone (Ttyh1 shRNA-insensitive-EGFP). Also, a single amino acid mutant of TTYH2 or TTYH3 were constructed in TTYH2 or TTYH3-IRES2-EGFP clone. All truncations, single amino acid mutants, and shRNA insensitive clones were made by site-directed mutagenesis kit (Ezchange™ site-directed mutagenesis kit, EZ400, Enzynomics).

4-(2-butyl-6,7-dichloro-2-cyclopentylindan-1-on-5-yl)oxybutyric acid (DCPIB), 2-(2-Amino-3-methoxyphenyl)-4H-1-benzopyran-4-one (PD98059) were purchased from Tocris (#1540, 1213). 4′,5,7-Trihydroxyisoflavone, 5,7-Dihydroxy-3-(4-hydroxyphenyl)-4H-1-benzopyran-4-one (Genistein) was purchased from Sigma-Aldrich (#G6649). Other chemicals used for recording solution of electrophysiology and intrinsic optical signal imaging were purchased from Sigma-Aldrich.

For VRAC current (herein defined as Cl⁻ current induced by hypo-osmotic stimulation, I_Cl,swell) recording, primary cultured cortical astrocyte or transfected HEK292T or CHO cells, and acute hippocampal slices from 8~10 weeks mice were used for in vitro and ex vivo (slice) whole cell patch clamp recording, respectively. The preparation procedures with cultured coverslip and hippocampal slices followed a previous report [43]. In both in vitro and ex vivo whole cell patch clamp recording, we use the same internal and external solution; isotonic (ISO), hypo-osmotic solution (HOS). In case of ex vivo whole-cell patch clamp recording, the slices were left to recover for at least 1 hour before recording in oxygenated (95% O₂ and 5% CO₂) artificial cerebrospinal fluid (aCSF) containing (in mM) 130 NaCl, 24 NaHCO₃, 3.5 KCl, 1.25 NaH₂PO₄, 1 CaCl₂, 3 MgCl₂ and 10 glucose (pH 7.4) at room temperature. After 1 hour, the slices were placed on a chamber with an aCSF solution. After complete rupture of the cell membrane, the external aCSF was changed to standard ISO solution for baseline recording.

Standard external bath solutions of both ISO and HOS for VRAC_swell were composed of (in mM): 70 Tris-HCl, 1.5 CaCl₂, 10 HEPES, and 10 glucose, 5 TEA-Cl, 5 BaCl₂ and adjusted to pH 7.3 with CsOH, as previously described [6]. The osmolality of each solution was adjusted with sucrose: 100 mM sucrose for 280~290 mOsm (for ISO) and 30 mM sucrose for 220~230 mOsm (for HOS), Solution osmolality was confirmed by a vapor pressure osmometer (Vapro osmometer #5600, Wescor). The external solution of both ISO(NaCl) and HOS(NaCl) for recording of LRRC8-mediated VRAC_Γ was composed of (in mM) : 150 NaCl for ISO(NaCl), 105 NaCl for HOS(NaCl), respectively, 6 KCl, 1 MgCl₂, 1.5 CaCl₂, 10 Glucose, 10 HEPES, adjusted to pH 7.4 with NaOH. The solution osmolality is 320mOm for ISO(NaCl) and 240mOsm for HOS(NaCl). For Ca²⁺ dependency with TTYH mediated I_Cl,swell, 1.5 mM CaCl₂ is changed to 0.15 mM CaCl₂ for all external bath solution. TEA-Cl and BaCl₂ were added to eliminate the potassium currents enabling selective recording of Cl⁻ current. For glutamate permeability experiments, 70 mM Tris-Cl containing HOS external solution was substituted for 70 mM Tris-base and 70 mM glutamate (glutamate containing HOS, gHOS) to omit external Cl⁻. When the maximal current was generated after standard HOS treatment, the standard HOS (70 mM Tris-Cl) was changed to gHOS (70 mM Tris-glutamate). The reversal potentials from both standard HOS (70mM Tris-Cl) and gHOS (70mM Tris-glutamate) were measured to calculate the ratio of glutamate permeability (detailed description in the analysis of electrophysiological data).

For the recording of various anion (SCN⁻, I⁻, Br⁻, and F⁻) permeability, 105mM of NaCl in HOS (NaCl) was changed to the same concentration, 105mM of NaSCN, NaI, NaBr, and NaF at the maximal conductance of VRAC_swell or VRAC_Γ recording. The reversal potentials from each HOS(NaCl) (105mM NaCl) and HOS(NaX⁻) (105mM NaSCN, NaI, NaBr, and NaF) were measured to calculate ratio of ion permeability (The calculating equation for of relative anion permeability is the same with glutamate permeability, detailed description in analysis of electrophysiological data).

Standard internal solution for VRAC_swell was composed of (in mM) : 60 Trizma-HCl, 70 Trizma-base, 70 Aspartic acid, 15 HEPES, 0.4 CaCl₂, 1 MgCl₂, 4 Mg-ATP, 0.5 Na-GTP, and 1 EGTA, adjusted to pH 7.25 with CsOH, as previously described [6].

The internal solution for recording of LRRC8-mediated VRAC_Γ was composed of (in mM) : 40 CsCl, 100 CsMeS, 1 MgCl₂, 1.9 CaCl₂, 5 EGTA, 4 Na2ATP, 10 HEPES, adjusted to pH 7.2 with CsOH, as previously described(Voss et al., 2014). For Ca²⁺ free experiment, 1mM EGTA was replaced into 20mM BAPTA which leads 100nM free Ca²⁺ to be reduced by 10nM. Patch pipettes had tip resistances of 5~8MΩ when filled with internal solution. Pipettes were pulled from borosilicate thin-wall glass capillaries (TW150F-4, World Precision Instruments) with the micropipette puller (P-97, Sutter Instrument). Unlike other previous studies [25,26], we used Tris·Cl-based external solution without Na⁺ and K⁺ and minimal level of Cs⁺ (15mM of CsOH for pH adjustment of external solution), and Tris·Cl-based internal solution to eliminate any contribution of Na⁺, K⁺ and to minimize Cs⁺ to I_Cl,swell. This is particularly important for isolating the anionic current (See Table 2).

For substituted-cysteine accessibility method (SCAM), 10 amino acids in TTYH1 (from V161 to A170) were each substituted with cysteine residue. The cysteine mutant of TTYH1 or TTYH1-WT were co-transfected with AQP4 in HEK29T cells. The standard bath solution of HOS containing MTSES (100mM) was applied at the maximal I_Cl,swell. The percent block was determined.

To test the dominant negative function of TTYH1-R165A, pEGFP-N1-TTYH1-WT and pEGFP-N1-TTYH1-R165A were co-expressed with mAQP4-M1-mcherry in HEK293Tcells. The total DNA amount of TTYH1 (WT or R165A) is 2 mg and AQP4 is 750ng. The relative amount of R165A to WT is varied with 1:0, 0.75:0.25, 0.5:0.5, 0.25:0.75, and 0:1. The maximal current of I_Cl,swell was recorded in each condition by treatment of a standard solution of HOS.

To identify shRNA-transfected or infected astrocytes from primary cultures or hippocampal slices, GFP or mcherry fluorescence was used (Fig. 10D and 11J), which were driven under CMV promoter in pSicoR vector construct (Fig. 11I). In the case of hippocampal slices, astrocytes were distinguished by electrophysiological properties such as low membrane resistance (Rm), low resting membrane potential after rupturing. The whole cell configuration was achieved by rupturing the cell membrane with suction after achieving a giga-seal. Holding voltage was −60 mV. Stability and quality of the patch were determined by monitoring the cell parameters: Rm, cell capacitance (Cm), and series resistance (Ra).

The episodic ramp protocol (+100 mV to −100 mV, 1000 ms, 15 s interval) for measurement of the maximal conductance of VRAC_swell and VRAC_Γ and the current step protocol (from−100 mV to +140 mV with 20 mV step) for measurement of the voltage-dependent inactivation of VRAC_swell and VRAC_Γ were given and recorded with a Digidata 1322A interface and pClamp10 software (Molecular Devices). A continuous gap-free recording was simultaneously conducted with Minidigi digitizer and Axoscope10 software (Axon Instruments).

For drawing the I~V curve for I_Cl,swell, the averaged mean value and standard error of mean (SEM) at every 5 mV are plotted. The current density (pA/pF) of I_Cl,swell was measured by subtraction of maximal Cl⁻ current (average of maximal 4weeps in HOS treatment) from basal Cl⁻ current (average 4weeps of ISO treatment, 1 min before HOS treatment) divided by the capacitance of the membrane (Cm), and measured the amplitude of current density (pA/pF, from −100 mV to 100 mV) by I_{Cl,swell (at 100 mV)} – I_{Cl,swell (at −100 mV)}. Rectification index was defined here as the following equation.

Rectification index=Icl−, swell−100mVIcl−, swell+100mV

To examine the glutamate permeability (P_Glu) of VRACs, we measured the reversal potential shift of I_Cl,swell caused by external solution change from standard HOS solution (70 mM Tris-Cl) to gHOS (70 mM Tris-glutamate) solution, which was corrected for junction potential (9.4mV). The detailed composition of the gHOS solution is described above. The reversal potential of I_Cl,swell was calculated with following Goldman-Hodgkin-Katz (GHK) equation.

Erev=RTF*lnPclCl−i+PglugluiPclCl−o+Pglugluo

With this equation, reversal potential difference (ΔE_rev) was calculated as (previously also shown in [35,45]).

ΔErev=RTF*lnPclCl−i+PglugluiPclCl−HOS+PglugluHOS

−RTF*lnPclCl−i+PglugluiPclCl−gHOS+PgluglugHOS

The ‘i’ and ‘o’ indicates internal and external solution respectively. HOS and gHOS represent standard hypo-osmotic stimulation condition and glutamate containing HOS condition respectively.

For IOS imaging devices, Infrared (IR) light source with optical filter (775 nm wavelength, Omega Filters) was used for transillumination of brain slices and these optical signals were obtained as IOS images using a microscope (BX50WI, Olympus) equipped with a CCD camera (ORCA-R2, Hamamatsu). Imaging Workbench software (INDEC Biosystems) was used for image acquisition and image analysis.

For the detailed procedure of experiments, we first prepared mouse brain hippocampal slices (described in the methods for electrophysiology). Next, we fixed hippocampal slice into the recording chamber and positioned electrical stimulator in the CA1 stratum radiatum region. We determined the region of interest (ROI) by calculating 1.5~2% increase in IOS by single pulse electrical stimulation (20Hz, 1 sec, 200~300mA). Images were acquired every 1 second and after stabilization of the IOS signal, we acquired 10 min baseline IOS transmittance (T_base). Next, we applied 1Hz-30min electrical stimulation (which has been previously shown to induce activity-dependent cell swelling [2]) and acquired two IOS indexes required for analysis: maximum IOS transmittance (T_max) and final IOS transmittance (T_end). With these defined indicators, we calculated the induced maximal volume increase (b) and regulated volume decrease (a) as T_max−T_base and T_max−T_end respectively (Fig. 11C, E, and J). Finally, the percentage of regulated volume decrease (RVD%) was calculated by

RVD%=ab*100

Gene silencing test with shRNAs that target VRAC candidates including Best1, Ttyh1,2 and 3, LRRC8A and C were tested by quantitative real-time PCR (qRT-PCR). All shRNAs were electroporated into cultured cortical astrocytes. Approximately 3 days after electroporation, total RNA was extracted by using RNA isolation kit (RNeasy Mini Kit, #74104, Qiagen). cDNA was synthesized by using reverse transcriptase (SuperScript III reverse transcriptase, #18080-044, Invitrogen). For qRT-PCR, SYBR-green (SYBR Green PCR master mix, #4309155, Applied Biosystems) were used. Primers for qRT-PCR for each gene were listed in Supplemental Table 2.

The knock-down efficiency of mLrrc8a shRNA or m/h Lrrc8a shRNA was confirmed by performing western blot. mLRRC8A-GFP with control or mLrrc8a shRNA or m/h Lrrc8a shRNA-mcherry were co-transfected with HEK293T cells. After the 30hr, the transfected HEK293T cells were lysed with RIPA buffer (#R4100, GenDEPOT) containing protease inhibitor cocktail (#P3100, GenDEPOT). 30 µg of protein lysates were separated by protein electrophoresis using protein gels (#4561084, BIORAD) and blotted onto PVDF membranes (#1704156, BIORAD). The blots were incubated overnight at 4℃ with chicken anti-GFP (1:5000, #AB13970, Abcam) and rabbit anti-b-actin (1:5000, #AB133626, Abcam). Blots were then washed and incubated with horseradish peroxidase-conjugated rabbit anti-chicken IgY (1:3000, #AP162P, Millipore), goat anti-rabbit IgG (1:3000, #AB6721, Abcam), respectively. The immunoreactivity with enhanced chemiluminescence (GE Healthcare Life Sciences). The band intensity was acquired by ImageQuant LAS 4000 (GE Healthcare) and quantified using ImageJ software (NIH). For surface biotinylation, wild-type, several truncated mutant or single amino acid mutants of mTTYH1-EGFP with AQP4 overexpressing HEK293T cells were incubated at 4℃ and washed three times with PBS. Surface-expressed proteins were then biotinylated in PBS containing Ez-link sulfo-NHS-LC-Biotin (Thermo 21335) for 30 min. After biotinylation, cells were washed with quenching buffer (100 mM glycine in PBS) to remove excess biotin and then washed three times with PBS. The cells were then lysed and incubated with high capacity NeutrAvidin-Agarose Resin (Thermo 29204). After three washes with lysis buffer, bound proteins were eluted by SDS sample buffer and subjected to western blot analysis. Primary antibody was rabbit anti-GFP (Millipore, AB3080). The secondary antibody was Donkey anti-rabbit HRP (Amersham, NA9340).

To determine whether the direction of N-, C-terminus, and Loop1 of TTYH1 are located in the cytosolic or extracellular side, HEK293T cells were transfected with TTYH1-EGFP or TTYH1-GFP (Loop1-FLAG), grown on coverslips for 24 hours. The cells were fixed in 4% paraformaldehyde for 30 min at room temperature. The permeabilized group are treated PBS with 3% Triton X-100 for 5 min and the non-permeabilized group is treated PBS for 5min. Non-specific binding was prevented with 1-hour incubation in 2% donkey serums. Cells were incubated the rabbit anti-Ttyh1 (for N-terminus of TTYH1, #NBP1-59909, Novus Biologicals), anti-GFP (for C-terminus of TTYH1, #ab13970, Abcam), and anti-FLAG (for Loop1 of TTTYH1, #F1804, Sigma-Aldrich) primary antibodies for overnight at 4℃. After washing, DyLight 594-conjugated secondary antibody (Jackson lab, 1:400) was added and incubated for 2 hours at room temperature. The cells were washed and mounted, and then observed under a Nikon A1 confocal microscope.

Mice under anesthesia with 2% tribromoethanol (avertine, 20 µl/g) and were trans-cardially perfused with 0.9% saline followed by 4% PFA, and then the brains were removed. The dissected brains were post-fixed with 4% PFA at 4℃ for 18~24 hours and dehydrated with 30% sucrose at 4℃ for 48 hours. For immunostaining, brain sections with a thickness of 30 µm were cut using a cryostat, blocked with the serum (donkey serum 2%, goat serum 2%, the solution including 0.3% Triton X-100) of the appropriate species for 1 hour and then treated with primary antibodies, including rabbit anti-TTYH1 (1:200, #NBP1-59909, Novus Biologicals), chicken anti-GFAP (1:500, #ab5541, Millipore), mouse anit-NeuN (1:500, #MAB377, Millipore), DAPI (1:5000, #46190, Chemicon). During over-night. The next day, tissues were rinsed 3 times with 0.1M PBS and subsequently incubated with secondary antibodies conjugated with Alexa Fluor 488, 594 and 647 for 1 hour 30 min. After treating with secondary antibodies, the tissues were washed out with PBS including DAPI staining in the second washing step. After three rinses in PBS, the sections were mounted on slide glasses. Images were acquired on a Nikon A1 confocal microscope.

Statistical analyses were performed using ImageJ software (NIH Image). For analyzing the knock-down efficiency of AAV-virus containing control or Ttyh1/2/3 shRNA, we measured the TTYH1 intensity in the binary image of GFAP positive (+) cells. The statistical analysis of TTYH1 intensity in GFAP (+) cells between control and Ttyh1/2/3 shRNA groups were analyzed with the two-tailed Mann-Whitney t-test.

Publically available brain RNA-seq data sets [42] of VRAC candidates such as TTYH, LRRC8, BEST family were used in Fig. 1D (https://web.stanford.edu/group/barres_lab/brain_rnaseq.html). Multiple alignments of amino acids were processed from CLUSTAL OMEGA (http://www.ebi.ac.uk/Tools/msa/clustalo/). The dendrogram was drawn by the PHYLIP program. The prediction of the transmembrane domain of TTYH family was processed by TMHMM (http://www.cbs.dtu.dk/services/TMHMM/). The potential phosphorylation site and kinases were predicted by GPS 3.0 (http://gps.biocuckoo.org/online.php). Statistical analysis was performed using GraphPad Prism 7.01 software. p<0.05 was considered statistically significant. No statistical methods were used to predetermine sample size. All statistical analysis results and methods were described in the Supplemental Table 1 for main figures.

Statistical parameters including the exact value of n, the definition of center, dispersion and precision measures (mean ± SEM) and statistical significance are reported in the Supplemental Table 1. All data points are tested if the values come from a Gaussian distribution by D'Agostino-Pearson omnibus normality test, and then appropriate statistical methods are applied. In figures, asterisks denote statistical significance as ^*, p<0.05; ^**, p<0.01; ^***, p<0.001; ^****, p<0.000, as well as non-significance with NS, p>0.05. Statistical analysis was performed in GraphPad PRISM 7.

Previous studies suggested that water movement via AQP4 is critical for an initial volume increase which then somehow triggers an opening of VRAC_swell in astrocytes [46]. To test the requirement of AQP4 for activation of VRAC_swell, we recorded I_Cl,swell under voltage-clamp configuration from cultured mouse astrocytes expressing either control or Aqp4 shRNA. To isolate genuine I_Cl,swell without any current contaminated by cations, we used a special recipe of intracellular and extracellular solutions as previously described (see Table 2 in Materials and Methods) [6]. HOS-treatment caused a slow onset of Cl⁻ conductance increase as a series of vertical lines representing the current trace in response to a periodic voltage ramp from +100 mV to −100 mV (Fig. 1A inset). We found that the I_Cl,swell was mostly abolished by Aqp4 shRNA (Fig. 1A~C, see Table 1 in Materials and Methods). Consistent with previous reports [46], these results indicate that astrocytes regulate their volume with water movement through AQP4.

Previously, two groups concurrently reported that LRRC8A is an essential component of VRAC_Γ in various cell lines [25,26,27]. Thus, we tested whether LRRC8A is the major component of VRAC_swell in astrocytes. We developed a specific shRNA for Lrrc8a and Lrrc8c, which shows relatively higher mRNA expression than other Lrrc8 isoforms in cultured astrocytes (Fig. 1D~F, see Table 1 in Materials and Methods). We observed that the maximal amplitude of I_Cl,swell was not affected by Lrrc8a shRNA or Lrrc8c shRNA or both (Fig. 1G~I). Taken together, these results indicate that LRRC8A and C do not contribute majorly to the astrocytic VRAC_swell.

It has been subsequently contended that BEST1, not LRRC8A, is indispensable for volume regulation in hiPSC-RPE cells [36]. Over the years, we have extensively characterized BEST1 as a Ca²⁺-activated anion channel, localized at peri-synaptic sites of the hippocampal astrocytes, functioning as the glutamate- and GABA-permeable channel in cerebellum and hippocampus, respectively [47]. To test whether there is any contribution of BEST1 in the astrocytic VRAC_swell, we utilized previously characterized BEST1 knockout mouse [48] and Best1 shRNA [45]. We observed that I_Cl,swell was normally generated in either cultured astrocytes from BEST1 knockout mice or in Best1 shRNA expressing astrocytes (Fig. 1J~O), eliminating the possibility that BEST1 contributes to the astrocytic VRAC_swell. Taken together, the astrocytic VRAC_swell is not encoded by Lrrc8a or Best1.

We next focused on Ttyh family as the candidate genes for VRAC_swell. Firstly, we confirmed that cultured astrocytes show much higher mRNA expression of Ttyh1/2/3, compared to other VRAC_swell candidates such as Lrrc8 and Bestrophin family (Fig. 1D), as previously reported [42]. Thus, we developed and tested the Ttyh isoform-specific shRNAs (Fig. 2A) on the astrocytic I_Cl,swell. We found that the astrocytic I_Cl,swell was almost completely abolished by a combination of all three Ttyh1/2/3 shRNAs, but not by a single Ttyh shRNA or any combination of double Ttyh shRNAs (Fig. 2E~H). To exclude any possibility of off-target effect by shRNAs, we developed a fully functional shRNA-insensitive clone of TTYH1 (Fig. 2B~D). We found that in astrocytes co-expressing all Ttyh1/2/3 shRNAs with shRNA-insensitive form of TTYH1, the I_Cl,swell was fully recovered to the control level (Fig. 2G and 2H). Taken together, these results indicate that TTYH family are necessary for the astrocytic VRAC_swell with functional redundancy.

The reduction of I_Cl,swell by Ttyh1/2/3 shRNAs might be due to two possibilities: a reduction of the pore-forming subunits of VRAC_swell to reduce the amplitude of I_Cl,swell or a reduction of a modulatory protein of VRAC_swell to reduce the activation kinetics of I_Cl,swell. To distinguish these two possibilities, we analyzed the current kinetics of I_Cl,swell by plotting the average current density of I_Cl,swell versus time after HOS treatment (Fig. 2I and 2K). To compare each shRNA condition, we normalized current density by the maximum to calculate the time to reach 50% of peak (Fig. 2J and 2L). We found that the maximal current density at both +100mV and −100mV was dramatically decreased in Ttyh1/2/3 shRNAs expressing astrocytes compared to control shRNA (Fig. 2I). However, the time to reach 50% of peak was not significantly different (Fig. 2J, inset), suggesting that the current reduction by Ttyh1/2/3 shRNAs was due to a reduction of the pore-forming VRAC_swell subunits. In contrast, we found that the maximal current density at both +100mV and −100mV was not significantly different in Lrrc8a shRNA condition compared to control shRNA (Fig. 2K). However, the time to reach 50% of peak in Lrrc8a shRNA expressing astrocytes was significantly slower compared to control shRNA condition (Fig. 2L, inset), indicating that Lrrc8a shRNA slows the activation of VRAC_swell without affecting the maximal amplitude of I_Cl,swell. Taken together, these results suggest that Ttyh1/2/3 are the essential component of VRAC_swell, whereas LRRC8A might act as a modulatory channel of VRAC_swell.

To test if each TTYH isoform confers VRAC_swell activity, we transiently expressed each TTYH isoform clone in HEK293T cells together with AQP4 to maximize HOS-induced volume increase and recorded I_Cl,swell (Fig. 3A~C). We found that each TTYH isoform co-expressed with AQP4 showed significantly increased I_Cl,swell compared to the condition of AQP4 alone (Fig. 3A~C), indicating that all three isoforms of TTYH are sufficient for activation of the I_Cl,swell in the presence of AQP4 and that each isoform of TTYH family can mediate VRAC_swell possibly as a homomeric channel. In marked contrast, LRRC8A has been repeatedly shown to be insufficient for VRAC_swell activity in heterologous expression systems [25,26,28,31]. Consistently, we did not observe I_Cl,swell in CHO-K1 cells co-expressing LRRC8A and AQP4 (Fig. 3D~F), indicating that LRRC8A is not sufficient for VRAC_swell.

The unexpected lack of HOS-induced I_Cl,swell in naïve HEK293T cells (Fig. 3A~C) was a surprising observation which called for further investigation. In the past, LRRC8A-mediated VRAC_Γ in naïve HEK293T cells has been recorded with external solution containing cations such as Na⁺, Cs⁺, or K⁺ (see Table 2 in Materials and Methods). In contrast, the external solutions that we have utilized contain minimal cations as previously described for the recording of astrocytic VRAC [6]. Therefore, we directly tested the two different solutions in native HEK293T cells. By surprise, we found that I_Cl,swell was not observed when the native HEK293T cells with endogenous LRRC8A-mediated VRAC_Γ was treated with the HOS which contained minimal cations, whereas a robust I_Cl,swell was induced by the Na⁺-containing HOS (Fig. 3G~I), suggesting that an activation of LRRC8A either requires external cation or possesses a sensitivity to Tris. To further confirm whether this I_Cl,swell induced by Na⁺-containing HOS was mediated by LRRC8A, we engineered previously reported Lrrc8a shRNA [49] which targets both human and mouse LRRC8A (Fig. 1F). We found that the I_Cl,swell induced by Na⁺-containing HOS was almost completely eliminated by the m/h Lrrc8a shRNA and overexpression of TTYH1 and AQP4 along with Lrrc8a shRNA caused a robust I_Cl,swell (Fig. 3J~L), indicating that TTYH1 is sufficient for VRAC_swell even in the absence of LRRC8A. Furthermore, the same m/h Lrrc8a shRNA did not have any effect on the maximal amplitude of I_Cl,swell induced by the HOS which contained minimal cations in astrocytes (Fig. 3M~O) as well as in TTYH1 and AQP4 over-expressing HEK293T cells (Fig. 3P~R). These results indicate that LRRC8A-mediated VRAC_Γ is profoundly different from TTYH-mediated VRAC_swell in its external-cation-dependence and that the astrocytic VRAC_swell is distinct from LRRC8A. This conclusion was further strengthened by the observations that TTYH-mediated VRAC_swell and LRRC8A-mediated VRAC_Γ showed profoundly different voltage-dependent inactivation (Fig. 4A and 4B) and relative permeability to various anions (Fig. 4D), although there was no difference in rectification index (Fig. 4C).

The astrocytic I_Cl,swell was reported to display an outwardly rectifying conductance [6] with a calculated rectification index to be less than one. The rectification index of each Ttyh1/2/3-mediated I_Cl,swell was also identical to that of I_Cl,swell recorded from cultured astrocytes with a strong outward rectification (Fig. 4E). Crepel et al. also showed that the Ca²⁺ is not essential for the activation of astrocytic VRAC_swell [6]. We observed that each of the Ttyh1/2/3-mediated I_Cl,swell is also independent of Ca²⁺ as evidenced by the similar amplitude of I_Cl,swell in low external Ca²⁺ (0.15mM) and high internal Ca²⁺-chelator BAPTA (20mM) (Fig. 4F~H). We next examined the pharmacological profiles of astrocytic VRAC_swell compared to Ttyh1/2/3-mediated I_Cl,swell in HEK293T cells. DCPIB is a known inhibitor for both VRAC_swell and VRAC_Γ recorded from various cell lines including cultured astrocytes [50,51]. We found that DCPIB rapidly blocked I_Cl,swell in cultured astrocytes (Fig. 4I~K) and Ttyh1/2/3-mediated I_Cl,swell in HEK293T cells (Fig. 4L and 4M). Taken together, these results indicate that Ttyh1/2/3 share the same biophysical (outward rectification), biochemical (Ca²⁺-independency) and pharmacological (sensitivity to DCPIB) properties as the astrocytic VRAC_swell.

As previously reported, tyrosine kinase (TK) and mitogen-activated protein kinase (MAPK) are essential for activation of the astrocytic VRAC_swell [6]. We also confirmed that pre-treatment of TK inhibitor genistein (Fig. 5A~C) and MAPK inhibitor PD98059 (Fig. 5E~G) completely abolished I_Cl,swell in cultured astrocytes. Each of the Ttyh1/2/3-mediated I_Cl,swell was similarly blocked by pre-treated genistein (Fig. 5A and 5D) or PD98059 (Fig. 5H~J). In fact, the surface expression of TTYH1 was enhanced by HOS treatment, which was prevented by either TK or MAPK inhibitor (Fig. 5K). These results indicate that each TTYH isoform shares the identical kinase-dependency as the astrocytic VRAC_swell. In contrast, LRRC8A-mediated VRAC_Γ showed a profoundly different sensitivity to MAPK inhibitor (Fig. 5L~N).

Another unique property of the astrocytic VRAC_swell is its relatively high P_Glu/P_Cl [8,32,52]. To measure the P_Glu/P_Cl of I_Cl,swell, we exchanged the external solution from HOS (83 mM Cl⁻) to glutamate-containing HOS (gHOS: 70 mM glutamate and 13 mM Cl⁻) after reaching a maximal current amplitude of I_Cl,swell (Fig. 5O). We found that in both cultured astrocyte (Fig. 5P and 5Q) and each Ttyh1/2/3-expressing HEK293T cells (Fig. 5R), the reversal potential of I_Cl,swell shifted to a positive potential in gHOS treatment. Based on the reversal potential shift, we calculated the P_Glu/P_Cl according to Goldman-Hodgkin-Katz equation and found that I_Cl,swell in cultured astrocytes and each Ttyh1/2/3-mediated I_Cl,swell shared the similar high P_Glu/P_Cl at around 0.26 (Fig. 5S). Taken together, these results indicate that kinase-dependency and glutamate-permeability of each TTYH-mediated I_Cl,swell resemble those of the astrocytic VRAC_swell.

If TTYH isoforms are the true channel-forming subunit of VRAC_swell, we should be able to determine the channel-pore as a single amino acid residue. To identify the exact pore-forming amino acid residue, we first utilized bioinformatical approaches to predict the membrane topology of TTYH isoforms. Based on the dendrogram analysis (Fig. 6A) and sequence alignment (Fig. 7) of Ttyh1/2/3 isoforms across different species, we found that TTYH1, which is the most abundantly expressed isoform in astrocytes (Fig. 1D), showed the order of sequence homology: TTYH1>TTYH2>TTYH3 >> Drosophila Tweety homologs. The hydrophobicity-based sequence analysis predicted a membrane topology with four highly-conserved trans-membrane domains (TM) and additional intra-membrane domain (IM) in TTYH1 or TM domain in TTYH2 and TTYH3 near the C-terminus (Fig. 6B). To validate the predicted topology, we over-expressed the chimeric clone of TTYH1-GFP (GFP attached to the C-terminus of TTYH1) in HEK293T cells. We performed immunocytochemistry with either the antibody recognizing the N-terminus of TTYH1 (Ab_N-term) or the GFP-antibody (Ab_GFP) which should bind to the C-terminus of TTYH1-GFP (Fig. 8A). The intracellular or extracellular topology of each terminus was determined by the presence of the membrane detergent Triton X-100, which allowed the access of antibodies to the cytosolic compartment. We found that both Ab_N-term and Ab_GFP showed positive immunoreactivity only when Triton X-100 was present (Fig. 8A). These results indicate that contrary to the previous suggestion [53], TTYH1 has four TM domains and one IM domain with cytosolic N- and C-termini (Fig. 8C). In support of this topology, we found that the location of Loop1 between TM1 and TM2 (Fig. 8C) is in the extracellular region (Fig. 8B).

Because channel-pores usually reside within a loop between two transmembrane domains, we next closely examined the loop-sequences of Loop1-4 of TTYH1 (Fig. 8C) and found that Loop2 and Loop4 were highly conserved across different species but not Loop1 (Fig. 7). Loop3 appeared to be too short for a pore. Therefore, we focused on Loop2 and Loop4 to search for the channel-pore. We generated a series of truncation mutants (Δ) with a 20 amino-acid-deletion for each loop2 and 4 and also recorded I_Cl,swell. We identified potential candidates, Δ2A (S116-T135) and Δ2C (T156-A175) in Loop2 and Δ4B (L289-N304) in Loop4, which all showed an elimination of I_Cl,swell (Fig. 8D). Then, we performed surface biotinylation assay for each candidate to eliminate the possibility that the disappearance of I_Cl,swell is due to an impaired surface expression of the channel protein. We eliminated Δ4B in Loop4 and ΔYY (ΔY299, Y300) in Loop4B as a potential pore because of the lack of surface expression of Δ4B and ΔYY (Fig. 8E). We made additional truncation mutants with Δ2A and Δ2C into 5-amino-acid-deletions and narrowed down to Δ2A4, Δ2C2, and Δ2C3, which showed significantly reduced I_Cl,swell (Fig. 8D). After examining the surface biotinylation results (Fig. 8E), we found that the positively charged arginine residue at 165 position (R165) satisfied the definition of putative single amino acid pore residue, displaying a signficant reduction of I_Cl,swell (Fig. 8F and 8G) and a positive shift in the reversal potential of about +7 mV (Fig. 8G) without affecting the surface expression after substitution with alanine (R165A) (Fig. 8E). This result suggests that the TTYH1-R165A is less permeable to Cl⁻ compared to TTYH1-WT. The substitution at the homologous residue, R164A in TTYH2 similarly showed significantly reduced I_Cl,swell (Fig. 8H and 8I) indicating that the arginine residue (R165 for TTYH1 and R164 for TTYH2) is the key pore-forming residue. We further found that R165A mutant of shRNA-insensitive TTYH1, when co-expressed with all Ttyh1/2/3 shRNA, failed to rescue the I_Cl,swell in astrocytes (Fig. 8J and 8K), confirming that R165 is the putative pore residue. Taken together, R165 in TTYH1 and R164 in TTYH2 are the pore-forming amino acids located between TM2 and TM3.

To determine whether R165 and neighboring residues in TTYH1 are located at the pore lining vestibule, we performed the SCAM with cysteine substituted mutants of each of 10 amino acids from V161 to A170 (Fig. 9A~D). The control TTYH1-WT mediated I_Cl,swell was reduced by 9.01% with 100mM of MTSES which has no permeability through the plasma membrane, suggesting that there is a basal level sensitivity of TTYH1 for 100mM of MTSES (Fig. 9A~C). Among 10 amino acids, R165C was most significantly blocked by 100mM of MTSES with the highest block percentage (36.58%) (Fig. 9A~C), suggesting that R165 is the bona fide pore-forming amino acid residue, but not an amino acid involved in channel gating. The V164C was also substantially blocked by 100mM of MTSES (28.85%, Fig. 9C). These results suggest that R165 might be located at the apex of the outer vestibule of the pore in TTYH1, whereas V164 might be partially buried in the outer vestibule (Fig. 9D). Finally, to investigate whether one R165 functions as one pore in a multi-subunit complex or a collection of multiple R165 functions as an assembly of one pore, we recorded I_Cl,swell when the relative abundance of TTYH1-R165A and TTYH1-WT was varied in co-expression system (Fig. 9E). Intriguingly, the I_Cl,swell was linearly reduced according to the relative abundance of TTYH1-R165A (Fig. 9E), which fitted well with a straight line (R²=0942, p=0.006) but not with a dominance model of tetramer or pentamer [54]. These results suggest that, instead of an assembly of several R165's to form one functional pore, each R165 in a multi-subunit complex of TTYH1 homomeric channel constitutes an individual functional conducting pore as a part of a large conducting pore.

Due to the lack of molecular identity, the brain distribution and in-vivo function of VRAC_swell have remained unknown. Using the antibody against TTYH1, which is the most abundant isoform, we found that the protein expression pattern of TTYH1 was exclusively in GFAP-expressing astrocytes in the hippocampus and cortex (Fig. 10A). The astrocyte-specific expression and antibody specificity were verified by the use of adeno-associatedvirus carrying Ttyh1-specific shRNA (Fig. 10B and 10C). Then, we recorded I_Cl,swell from ex-vivo astrocytes in acute hippocampal slices infected with either all Ttyh1/2/3 shRNAs or control shRNA. After making a whole-cell patch configuration under the conventional artificial cerebrospinal fluid, we exchanged the external bath solution to the Tris-Cl based isotonic solution which contained 5mM TEA and 5mM BaCl₂ to eliminate the K⁺ conductance of astrocytes and found that the reversal potential of the membrane current shifted from −80mV to near −10mV, which was the equilibrium potential for Cl⁻. Under these recording conditions, the HOS-induced astrocytic I_Cl,swell was significantly reduced by all Ttyh1/2/3 shRNAs compared to control shRNA (Fig. 10D~F), indicating that I_Cl,swell of hippocampal astrocyte is mediated by Ttyh1/2/3. Taken together, these results indicate that TTYH family function as the astrocytic VRAC_swellex vivo.

The primary role of VRAC_swell in volume regulation is its participation during RVD. To recapitulate RVD in hippocampal slices, we performed intrinsic optical signal (IOS) imaging which represents the astrocytic volume change through water movement via AQP4 upon repetitive neuronal activity [2,55]. Among various stimulation protocols, we found that low-frequency stimulation (LFS; 1 Hz) for 30 min at Schaffer-collateral pathway caused an RVD with a continuous increase in IOS, which peaked at around 15 min, followed by a steady decrease in IOS until the end of 30 min stimulation period (Fig. 11A~C, and 11E~G). In contrast, a brief 20 Hz or 100 Hz stimulation only casued a transient IOS, but not DCPIBsensitive RVD (Fig. 11D). The LFS-induced RVD was completely blocked by DCPIB and genistein treatments (Fig. 11E~G), raising a possibility that the LFS-induced RVD might be mediated by Ttyh1/2/3.

To test whether LFS-induced RVD is mediated by TTYH isoforms, we injected lentivirus carrying all Ttyh1/2/3 shRNAs into CA1 hippocampus (Fig. 11J) and found that the RVD was eliminated by all Ttyh1/2/3 shRNAs (Fig. 11J~L). The eliminated RVD by all Ttyh1/2/3 shRNAs was fully rescued by tamoxifen-dependent flanking of shRNAs, which are under the cre/loxP control (pSicoR system [44]) in hGFAP-CreERT2 mice (Fig. 11H~L). These results demonstrate that astrocytic Ttyh1/2/3 are necessary and sufficient for RVD in the hippocampus.

One of the unique properties of an ion channel family is its ability to assemble as a homomeric or heteromeric complex to form a functional channel. For TTYH family there are three possible scenarios: (1) each TTYH subunits can assemble as a homomer; (2) TTYH subunits can assemble as a heteromer within TTYH family; and (3) TTYH subunits can assemble as a heteromer with other unknown proteins. Although we cannot completely rule out the possibilities of TTYH subunits assembling as heteromers, our results support the possibility that TTYH subunits can assemble and function as a homomer. If TTYH subunits were assembled only as a heteromer within TTYH family, gene-silencing by the combination of two shRNAs should have been effective to reduce the I_Cl,swell in astrocytes. However, this was not the case: Only the combination of all three shRNAs (Fig. 2E, 2G and 2H), but not the combinations of two shRNAs (Fig. 2E and 2F), was able to significantly and fully silence the I_Cl,swell in astrocytes. These strongly suggest that TTYH subunits can assemble and function at least as a homomer. It is also unlikely that TTYH subunits assemble with other unknown protein to form a functional channel because heterologous overexpression of each TTYH subunits produced an equal magnitude of I_Cl,swell in HEK293T and CHO-K1 cells, which are originated from human kidney and hamster ovary, respectively. The chance of such unknown protein to be expressed at high level in those two seemingly different cell lines and astrocytes is very low. More importantly, the strongest evidence for homomeric assembly of TTYH in astrocytes comes from the observation that co-expression of shRNA-insensitive TTYH1-WT with all three Ttyh1/2/3 shRNAs fully rescued I_Cl,swell (Fig. 2E, 2G and 2H), whereas co-expression of the pore-mutant TTYH1-R165A (shRNA-insensitive) and all three Ttyh1/2/3 shRNAs caused a complete impairment of I_Cl,swell (Fig. 8J and 8K). Therefore, we propose that astrocytic VRAC_swell is composed of mostly homomers of all three Ttyh1/2/3 subunits. Nevertheless, further studies are needed to clarify the exact subunit composition of the TTYH channel.

In this study, we provide unprecedented evidence that each Ttyh1/2/3 isoform acts as a pore-forming subunit of VRAC_swell, while LRRC8A acts as a pore-forming subunit of VRAC_Γ, which modulates the speed of cell swelling and subsequent activation kinetics of VRAC_swell. These ideas are derived from the detailed analysis of activation kinetics of I_Cl,swell recorded from both Lrrc8a and Ttyh1/2/3 shRNAs expressing cultured astrocytes (Fig. 2I~L). The observations that the peak amplitude of I_Cl,swell was almost completely eliminated by Ttyh1/2/3 shRNAs without affecting the activation kinetics (Fig. 2I and 2J), while there was no change in peak amplitude of I_Cl,swell with significantly slower activation kinetics by Lrrc8a shRNA (Fig. 2K and 2L), implying that TYH1/2/3 and LRRC8A might directly or indirectly interact with each other and have distinct roles on activation of I_Cl,swell in astrocytes. In support of the possibility of a direct interaction, a recent report demonstrated that LRRC8A directly binds to other ion channels such as TMEM16A (ANO1) in response to swelling by coimmunoprecipitation assay [56], raising a possibility that LRRC8A might also interact with TTYH family.

In support of the possibility of an indirect interaction between TTYH-mediated VRAC_swell and LRRC8-mediated VRAC_Γ, a recent report provided a crystal structure of the pore-forming subunits of LRRC8 family [29]. In this report, the authors clearly demonstrated that activation of LRRC8 requires reduced cytosolic ionic strength, but not swelling or membrane stretch [27,29]. Based on this report and our findings, we now propose a mechanistic model (Fig. 12A) to give a more comprehensive view and to explain the observed difference in activation kinetics of I_Cl,swell after Lrrc8a shRNA and the maximal current density of I_Cl,swell by Ttyh shRNA in astrocytes. When HOS(Tris-Cl) is applied in cultured astrocytes, water permeates the astrocytic plasma-membrane through the abundant AQP4 water channel, down the gradient of osmotic pressure. At the intracellular vicinity of AQP4, the ionic strength is instantaneously lowered by the influx of water (due to a dilution of ionic concentration by water). This lowering of ionic strength at the local region activates and opens nearby LRRC8A-mediated VRAC_Γ, allowing the Cl⁻ ions to initially influx through VRAC_Γ. The influx of Cl⁻ ions is ensured and potentiated particularly during the current recording by the ramp protocol (a periodic voltage step to +100mV and ramping down to −100 mV). This influx of Cl⁻ ions through VRAC_Γ further accelerates the water influx via AQP4 and subsequently induces a cell-swelling. This cell swelling eventually activates and open TTYH-mediated VRAC_swell. By this time the intracellular Cl⁻ concentration would have been significantly accumulated, and the opening of VRAC_swell allows efflux of Cl⁻ ions. And the water efflux through AQP4 follows the Cl⁻ efflux via VRAC_swell, causing a regulatory volume decrease (RVD). Based on this mechanistic model, we propose that TTYH family constitute the pore-forming subunits of astrocytic VRAC_swell which senses the cell swelling, and LRRC8 family constitutes a modulatory channel, VRAC_Γ which senses the low cytosolic ionic strength to boost anionic influx, water influx, and cell-swelling, facilitating the activation of VRAC_swell. This model now explains how the gene-silencing of LRRC8A by Lrrc8a shRNA slows the activation kinetics of I_Cl,swell without affecting the maximal current density of I_Cl,swell. Taken together, the precise relationship between Ttyh1/2/3 and LRRC8A and the molecular mechanisms of activation and modulation of I_Cl,swell by the association of both TTYH-mediated VRAC_swell and LRRC8-mediated VRAC_Γ remain undetermined. Future investigations are needed to test those exciting possibilities.

The precise molecular mechanisms of sensing the cell swelling induced by hypo-osmotic shock, leading to an opening of VRAC_swell are poorly understood. Several hypothetical models for sensing the cell-swelling have been proposed [57,58,59]; (1) direct sensing of membrane stretch, (2) direct or indirect tethering to the cytoskeleton, (3) alterations in membrane curvature such as caveolae complex by mechanical or membrane-compositional changes, and (4) via interaction with integrins activated by the cell volume perturbation, and (5) via swelling-induced interaction between actin-binding protein (ACTN4) and a cytosolic member of ABC transporter superfamily (ABCF2) prevents ABCF2 from suppressing VRAC_swell. Previous studies supporting these models have identified caveolin-1 and caveolin-related proteins that form a membrane curvature called caveolae in astrocytes [60]. This caveolin-1 is tethered to actin microfilaments [61] and positively modulates the activity of VRAC via c-Src tyrosine kinase [62]. Meanwhile, integrins are known to be associated with not only the extracellular matrix but also caveolin and positively regulate VRAC activation via various kinases such as FAK tyrosine kinase [63]. It has been previously demonstrated that a5-integrin is co-localized with TTYH1 in TTYH1-GFP over-expressing HEK293T cells [64]. We have demonstrated that Ttyh1/2/3-mediated I_Cl,swell is fully blocked by genistein and PD98059, which are the inhibitors of tyrosine kinase and MAP kinase, respectively (Fig. 5A, 5D and Fig. 5H~J). Based on these results, we can propose a mechanistic insight that there exists a complex of caveolin, integrin, cytoskeleton, and TTYH channels at astrocytic membrane curvature called caveolae, and hypo-osmotic shock somehow transduces through the integrin-caveolin-cytoskeleton complex to activate MAP kinase and tyrosine kinase to phosphorylate TTYH channels, which is required for either activation or surface expression of TTYH channels (Fig. 12B). In support of this possibility, we have observed that the surface expression of TTYH1 is enhanced by HOS treatment and prevented by either TK or MAPK inhibitor (Fig. 5K). These exciting possibilities await future investigations.

In addition to the fundamental role of VRAC_swell in cellular volume regulation, VRAC_swell is also implicated in many other functions including regulation of cell proliferation, apoptosis and motility, which occur during an early developmental stage in the embryonic neural stem cells [17,18]. Consistently, TTYH1 has been shown to be prominently expressed in embryonic stem cells during developmental stages and various brain structures in developing and adult brain, while limited expression was detected in non-neural tissues and cell types [65]. In the brain, TTYH1 appears to be associated with the general functions of VRAC_swell such as proliferation, apoptosis, and motility. For example, TTYH1 expression was shown to be increased at the edge of extending processes of the migrating cultured astrocytes after a scratch injury [66], suggesting its role in migration. TTYH1 expression is also increased in hypertrophied reactive astrocytes following status epilepticus [66], alluding to its role in morphogenesis. In glioma cells, TTYH1 was implicated as a potent regulator of tumor microtubule morphology, tumor-cell invasion, and proliferation [41], which are the putative functions of VRAC_swell.

We have demonstrated that TTYH channels are highly permeable to glutamate (P_Glu/P_Cl = 0.2~0.3) (Fig. 5O~S), as VRAC_swell has been already shown to release glutamate and other osmolytes [8,9,10,15,51]. We have also demonstrated that 1 Hz stimulation in the Schaffer-collateral pathway in the hippocampus causes a profound volume increase and eventual TTYH1-mediated RVD (Fig. 11J~L). It is possible that glutamate released through TTYH channels during RVD might regulate synaptic transmission as well as synaptic plasticity. Consistent with this idea, it has been previously shown that taurine, released through DCPIB-sensitive VRAC_swell, can mediate glycinergic neurotransmission in the hypothalamus [11]. We show that there is no contribution of DCPIB-sensitive RVD after a brief 20 Hz or 100 Hz stimulation at Schaffer-collateral pathway in the hippocampus (Fig. 11D). In contrast, a prolonged 1 Hz stimulation, which is a conventional protocol for long-term depression (LTD) at this synapse, causes a profound increase in volume and initiation of RVD (Fig. 11C, Fig. 11E~G, J~L). During such RVD, glutamate and other osmolytes are expected to be released through TTYH channels and affect synaptic transmission as well as NMDA receptor- or mGluR-dependent synaptic plasticity, in particular, LTD. These exciting possibilities should be explored in the future.

In conclusion, with the advent of molecular identification of TTYH family as the VRAC_swell of the brain, the putative functions of VRAC_swell can be systematically addressed at the molecular, cellular and systemic levels in the future. Furthermore, our study should provide novel therapeutic targets for cerebral edema following ischemic stroke, brain trauma, and brain cancer, which are known to be associated with VRAC_swell [14,67].