LRRC8/VRAC anion channels enhance β-cell glucose sensing and insulin secretion

Glucose homeostasis depends critically on insulin that is secreted by pancreatic β-cells. Serum glucose, which is directly sensed by β-cells, stimulates depolarization- and Ca2+-dependent exocytosis of insulin granules. Here we show that pancreatic islets prominently express LRRC8A and LRRC8D, subunits of volume-regulated VRAC anion channels. Hypotonicity- or glucose-induced β-cell swelling elicits canonical LRRC8A-dependent VRAC currents that depolarize β-cells to an extent that causes electrical excitation. Glucose-induced excitation and Ca2+ responses are delayed in onset, but not abolished, in β-cells lacking the essential VRAC subunit LRRC8A. Whereas Lrrc8a disruption does not affect tolbutamide- or high-K+-induced insulin secretion from pancreatic islets, it reduces first-phase glucose-induced insulin secretion. Mice lacking VRAC in β-cells have normal resting serum glucose levels but impaired glucose tolerance. We propose that opening of LRRC8/VRAC channels increases glucose sensitivity and insulin secretion of β-cells synergistically with KATP closure. Neurotransmitter-permeable LRRC8D-containing VRACs might have additional roles in autocrine/paracrine signaling within islets.

1. Primary β-cells lacking LRRC8A have minor effects on high glucose-stimulated electric activity and Ca2+ influx. Consistently, upon high glucose stimulation, LRRC8A-deficient islets secrete slightly less or similar insulins compared to control islets. Although there is a statistically significant difference in insulin release after 8 min, it's not clear whether it's physiologically significant because the amounts of insulin secreted in the initial phase are very low (less than 0.1%). Thus, among many other cation and chloride channels implicated in KATP-independent β-cell depolarization, VRAC might not be an important/significant modulator of glucose induced insulin secretion in β-cells. The authors didn't observe any gross phenotype in islet morphology and KATP channels are fully functional, however, the delayed Ca2+ response could still be due to other general β-cell defects caused by loss of VRAC function.
2. In contrast to the minor insulin secretion defects in vitro, the conditional knockout mice did exhibit an impressive phenotype in glucose tolerance shown in Fig 6c. The GTT assay is the only data supporting a role of LRRC8A in glucose metabolism in vivo. However, it has to be interpreted with caution due to a number of critical issues raised in the field in the last decade on RIP-Cre lines such as Tg(Ins2-cre)23Herr used in this study (See review Magnuson MA et al 2013). For example, RIP-Cre activities were detected across the central nervous system (Wicksteed B et al 2010). More concerning than the leakage of Cre expression, RIP-Cre was constructed with the human growth hormone (hGH) minigene, which was recently reported to affect β-cell functions by itself (Brouwers B et al 2014). It would be more convincing that alternative β-cell specific Cre lines (such as MIP-CreER) are used to examine the physiological role of LRRC8A in insulin-secreting β-cells. Besides GTT, measurements of plasma insulin levels may provide more direct evidence of the involvement of VRAC in glucose-stimulated insulin secretion in vivo.
Minor point: It's assumed and concluded in the abstract that high extracellular glucose causes β-cell volume increase, which in turn activates VRAC. It would be more convincing to show data on glucose-stimulated β-cell volume changes for control and KO β-cells in Figure 3 to support this conclusion.
Reviewer #2 (Remarks to the Author): Pancreatic beta-cells respond to glucose with calcium influx, a modest swelling and insulin secretion. The glucose-induced beta-cell swelling activates a volume-regulated VRAC anion channel, which has been predicted to play an important role in tuning beta-cell function. The manuscript by Till Stuhlmann et al. details for the first time that LRRC8A is a subunit of the beta-cell VRAC channel. The manuscript determines that glucose increases LRRC8/VRAC activity and that this provides a depolarizing influence on the beta-cell membrane potential that enhances glucose-stimulated calcium entry and insulin secretion. Moreover, the LRRC8/VRAC channel is shown to play an important role in regulating glucose tolerance, thus, beta-cell deficient LRRC8A mice show glucose intolerance. The manuscript includes a significant amount of high quality data relevant to understanding the importance of LRRC8/VRAC channels to beta-cell function. There appears to be a few issues that require attention. These are specified below.

Major:
As the activation of VRAC has important influences on beta cell functions in response to swelling by allowing regulatory volume decrease (RVD), cellular function may become altered over time due to loss of VRAC mediated RVD. This could influence many aspects of beta cell function such as the activities of other ion channels that are sensitive to cell swelling (e.g. KATP and voltage-dependent calcium channels) or it may have negative influences on beta cell health (e.g. changes in inflammasomes activation). Thus, some experiments to ensure that the loss of VRAC is only influencing calcium based on its acute influence on the membrane potential would be important.
The Glucose-activated VRAC current is important; this is the first major current besides KATP that has been identified. As the size of the beta-cell typically adjusts fairly quickly after glucose treatment in part due to VRAC activation, this current would be predicted to be transient as the authors indicate in the discussion. Therefore, it would be good to add some temporal analysis of this VRAC current activation and how long it remains active for following glucose simulation to help increase the understanding of its control of beta cell membrane potential. Moreover, it may be that the current follows the oscillations in islet swelling thus turning on and off during the oscillatory phases of islet calcium influx. Thus, some analysis of calcium oscillations would add important insights into the importance of this channel for 2nd phase insulin secretion.

Minor:
The following sentence needs references, ". Glucose-sensing by β-cells involves Glut2-mediated cellular uptake of glucose and its conversion to ATP and other metabolites." While Glut2 is the primary glucose transporter in mouse beta-cells, Glut1 is also present in human beta-cells.
The images in figure 1f show some yellow in the merge images of glucagon and insulin staining. This would indicate that some of the islet cells may be expressing both glucagon and inulin. The amount of these bihormonal cells shown in the βc-Δ8a mice also seems to be greater. As glucagon and insulin colocalization is not typical in mouse pancreas, either a different imaging technique such as confocal is needed or some more representative sections are required for this figure.
In the Figure 3 legend the following line is presumably a mistake, "2cError! Reference source not found." In the discussion the authors mention that, ". Accordingly a marked effect of VRAC on insulin secretion was only observed during the first ~15 minutes of glucose addition to isolated islets." However, the manuscript shows that insulin secretion differed when looked at for 8 minutes. It is not clear why the authors state 15 minutes when they only tested 8 minutes for the short duration?
Reviewer #3 (Remarks to the Author): This is an interesting and provocative new paper on the role of VRAC Cl ion channels in beta cell stimulus-secretion coupling. The importance of the work is that it may well represent the molecular identification of previously described volume sensitive Cl currents described in beta cells and provide insights into their possible physiological role.
While the efforts to selectively delete LRRC8 in beta cells in mouse are pretty convincing, the physiological impact is more important to establish, especially given that overall, the mouse lacks a clear phenotype. There are some specific concerns that require attention regarding the manuscript that must be addressed.

MAJOR
1. The data shown in Fig. 1 are generally convincing but the westerns shown in Fig. 1C would benefit from quantitation to firm up the finding of selective knockout and statistics of the result. Fig. 2 support the conclusion that hypotonic solution activates a Cl current in WT mouse beta cells but not the deletion mutants. However, as the authors decry the use of drugs suchg as DCPIB by the field one wonders why they did not also show that siRNA against VRAC in WT blocks the rectifying Cl current?

The data shown in
3. While the authors confirmed that their Cl channel current exhibited the expected anion selectivity expected for canonical VRAC, two other groups did not find this to be the case. Do the authors have an explanation for this discrepancy? Also, might testing a complete series of potentially permeant anions shed additional light?
4. Fig. 3A shows that 20 mM glucose causes activation of the current in WT but not VRAC knockouts. Was cell volume measured in this case? And did application of DCPIB alter volume? As cyclic AMP was also reported to activate islet Cl current, was this also tested? 5. Fig. 5 is especially important to the paper since here, physiological rather than swelling-related stimuli are chosen for study. One can argue that hypotonic solutions are not especially relevant to the study of beta cells as they are not likely to be subjected to external swelling stress in patients with normal kidney function. However, in Fig. 5 glucose is used which is relevant. The issues with this figure, however, are that little is documented short of a time lag between glucose introduction and the activation of spiking or free Ca concentration. What about longer term recordings of spiking or bursting cells or islets? How do the results depend on the concentration of glucose being applied to the islet? Is the dose response right shifted in the knockouts? And for the Ca data in Fig. 5f, the raw tracings should be shown in addition to the plot of the fura2 responses.
6. Fig. 6 has its own issues. First, the choice of glucose challenges is somewhat off as only 3.3 and 25 mM glucose are studied and the latter is a hammer dose. Much nicer if the authors also showed 11-secretion, it would behoove them to do a proper glucose perifusion experiment where flowing solution is continuously collected as it travels over a small number of islets. This would be helpful for documenting the true actions of channel deletion on both phase 1 and phase 2 of secretion. 7. With regards to the impressive data in Fig. 6C, which shows clear glucose intolerance after genetic deletion of LRRC8a, I have two concerns. First, it has been shown that the RIP2 promoter is not specific for only beta cells but can alter metabolism through actions in the hypothalamus. Do the authors have data ruling out expression in hypothalamus? Second, it was not clear whether islets from control animals lacking cre but having floxed LRRC8a or having cre but lacking floxed LRRC8a were all analyzed? 8. Do the authors have data proving that the Cl channel under study is indeed localized only to the beta cell plasma membrane and not to intracellular organelles? MINOR 1. The comments about % of islet containing beta vs. other cell types included on P. 3 depend on the species being discussed. About 80% of mouse islets are beta cells, while a smaller percentage is present in human. Also, it is more proper to speak not of changes in ATP but changes in ATP/ADP in response to glucose stimulation of beta cells.
2. The sentence on line 75 of P. 4 is incomplete.
Reviewer #4 (Remarks to the Author): In this paper, Stuhlmann and colleagues show that mouse pancreatic beta cells express the volumeregulated anion channel subunits LRRC8A and LRRC8D. They generated a mouse line deficient of the essential LRRC8A channel subunit and show that beta cells undergo cell swelling and have increased LRRC8A-dependent Cl-current in response to glucose stimulation. The ability of glucose to induce membrane depolarization, electrical activity and Ca2+ influx was delayed in the LRRC8A-deficient beta cells. This effect was specific to glucose and was not observed when beta cells were depolarized using the K-ATP channel blocker tolbutamide or increasing the extracellular K concentration. The delay in membrane depolarization also resulted in reduced initial insulin secretion and glucose tolerance in vivo. The study is well conducted and written and addresses an important aspect of the pancreatic beta cell stimulus-secretion coupling. I have a few concerns with the study that needs to be addressed:

1.
The authors state that the Lrrc8a deficient mice have a normal lifespan but do not provide any data to support it.

2.
I am concerned that the mice were only allowed 48 hr recovery between the GTT and ITT. The animals need a longer recovery period (1 week) between tests. The lack of effect in the ITT is likely correct but I would encourage the authors to repeat the ITT in rested animals.

3.
The conclusions that insulin secretion in the Lrrc8a deficient mice is reduced in response to glucose, but not tolbutamide and KCl is not supported by the data since all agents were only measured at 30 min and not 8 min where the difference in secretion was observed. Please measure insulin secretion in response to tolbutamide and KCL at 8 min.

4.
With the availability of published single human beta cell RNAseq data, it would be interesting if the authors could briefly discuss the translatability of their findings in mice to the human situation.

Detailed responses to the reviewers
We thank all reviewers for appreciating our work and for their insightful comments.

Reviewer #1 (Remarks to the Author):
Volume-Regulated Anion Channel (VRAC) may constitute additional KATP-independent glucose-sensing e ha is i pa reati β-cells. Building on the recent VRAC molecular cloning by Jentsch and Patapoutian groups, Stuhlmann et al tested this important hypothesis and presented a thorough study on the role of LRRC8/VRAC chloride channel in regulating glucose-sti ulated pa reati β-cell excitability and insulin secretion. They found that LRRC8 proteins are expressed in pancreas and islets. By generating a conditional knockout mouse allele, they showed that the essential VRAC subunit LRRC8A is required for hypotonicity-and high glucose-i du ed VRAC urre ts i ouse β-cells. They further demonstrated that loss of LRRC8A modestly affects glucose-induced electrical activity, intracellular Ca2+ response, and insulin secretion (at 8 min, but not 30 min). The paper is well written and results presented clearly. However, there are two main issues limiting the overall impact and significance of the manuscript.
We thank the reviewer for appreciating the importance of our paper and stating that it is clearly presented.
1. Pri ar β-cells lacking LRRC8A have minor effects on high glucose-stimulated electric activity and Ca2+ influx. Consistently, upon high glucose stimulation, LRRC8A-deficient islets secrete slightly less or similar insulins compared to control islets. Although there is a statistically significant difference in insulin release after 8 i , it's ot lear hether it's ph siologi all sig ifi a t e ause the a ou ts of i suli secreted in the initial phase are very low (less than 0.1%). Thus, among many other cation and chloride channels implicated in KATP-i depe de t β-cell depolarization, VRAC might not be an i porta t/sig ifi a t odulator of glu ose i du ed i suli se retio i β-cells. The authors did 't observe any gross phenotype in islet morphology and KATP channels are fully functional, however, the dela ed Ca + respo se ould still e due to other ge eral β-cell defects caused by loss of VRAC function.

We agree with the reviewer that the effe t of VRAC deletio o β-cell excitability and insulin secretion is less than that observed with deletion/inhibition of KATP channels. Clearly VRAC is not required for β-cell excitability or Ca 2+ tra sie ts si e the a o i al KATP pathway remains
intact. An only modulatory effect is therefore expected, and the onset of VRAC-dependent cell volume regulation should temporally limit the effect of VRAC as stated in the manuscript.
We agree with the reviewer that one can never totally exclude secondary effects of a gene KO Although we are convinced that our conclusion on the role of VRAC in insulin secretion is very well supported by the experimental evidence, we inserted a half sentence as a caveat in the last sentence of Results: Although, like i e er ge e disruptio e peri e t, e a ot totall exclude secondary effects, e o lude that LRRC8/VRAC ha els of β-cells modulate the secretion of insulin and the regulation of blood glucose.

2.
In contrast to the minor insulin secretion defects in vitro, the conditional knockout mice did exhibit an impressive phenotype in glucose tolerance shown in Fig 6c. The GTT assay is the only data supporting a role of LRRC8A in glucose metabolism in vivo. However, it has to be interpreted with caution due to a number of critical issues raised in the field in the last decade on RIP-Cre lines such as Tg(Ins2-cre)23Herr used in this study (See review Magnuson MA et al 2013). For example, RIP-Cre activities were detected across the central nervous system (Wicksteed B et al 2010). More concerning than the leakage of Cre expression, RIP-Cre was constructed with the human growth hormone (hGH) minigene, which was re e tl reported to affe t β-cell functions by itself (Brouwers B et al 2014). It would be more convincing that alter ati e β-cell specific Cre lines (such as MIP-CreER) are used to examine the physiological role of LRRC8A in insulin-se reti g β-cells Indeed, the RIP-Cre line, which we and many other have used to efficiently delete floxed genes in β-cells, has been described to express Cre-recombinase also in a small number of CNS neurons.
We have re-assessed this issue and have crossed the RIP-Cre mice to tdTomato reporter mice to examine the expression of Cre in the CNS (Fig. R2 for the reviewers). We indeed see a sparse population of Cre-positive cells scattered over the brain, including the hypothalamus, as described previously. However, we believe that this expression in the brain is not relevant for the present stud . I parti ular, our e peri e ts o isolated β-cells, which provide very strong evidence for the proposed signal transduction cascade, would be very difficult to explain by an effect of the CNS. While it might have been better to use another Cre-line from the beginning, we cannot repeat nearly all experiments with another Cre-line. This would not only require more than a year of work (and a lot of research money), but would also be impossible to justify to the authorities to whom we would have to apply for a new animal experiment permit.
While the sparse brain expression of our Cre-line seems to be a rather minor concern of the reviewer, he/she is more concerned about the possible effect of Cre-e pressio per se i β-cells (as described by others). We have performed the appropriate control in our in vivo glucose-and insulin tolerance experiments. There is no discernible difference between the Lrrc8a +/+ ; Cre + and Lrrc8a lox/lox ; Cre-groups. These data, which indicate that expression of the Cre-re o i ase i βcells does not cause glucose intolerance, have now been added to Fig. 6c, d.
Besides GTT, measurements of plasma insulin levels may provide more direct evidence of the involvement of VRAC in glucose-stimulated insulin secretion in vivo.
We agree with the reviewer that the demonstration of an involvement of VRAC in glucosestimulated insulin secretion in vivo would further support our conclusions. Unfortunately, it is impossible to get the necessary permit for these new animal experiments from the Berlin authorities in less than 6-8 months. We therefore contacted the Mouse Metabolic Evaluation Facility (MEF) of the Université de Lausanne, Switzerland, who have extensive experience in these assays, and asked them to perform these experiments on a paid service basis.
Indeed, the new experiments, which are shown in the new Fig. 6e, suggest that VRAC is important for insulin secretion in vivo. Whereas we observe a nearly two-fold increase in insulin plasma levels of WT mice 2 min after glucose injection, such an increase is absent in beta-cell specific LRRC8A KO mice, with insulin levels being similar again at 15 min. Although with the number of mice tested (13 -14 per genotype), t-tests give a marginal level of significance (p=0.07), we think that these data are a valuable addition to our paper and support the role of VRAC in early insulin secretion.
Mi or poi t: It's assu ed a d o luded i the a stra t that high e tra ellular glu ose auses β-cell volume increase, which in turn activates VRAC. It would be more convincing to show data on glucosesti ulated β-cell volume changes for control a d KO β-cells in Figure 3 to support this conclusion.
As e tio ed i our paper, β-cell swelling by high extracellular glucose has been published by another group previously (Miley et al., J. Physiol. 1997). Nonetheless, following the suggestion of the re ie er, e ha e o easured s elli g of β-cells in response to glucose which we show as new panel a in Fig. 3. The extent of this swelling is minor compared to swelling induced by the very drastic, non-physiological decrease in extracellular osmolarity, which is used by almost everybody in the field to induce VRAC activity. For comparison, we now also show hypotonicityi du ed β-cell swelling in Fig. 2a. See also our response to point 4 of reviewer 3.

Reviewer #2 (Remarks to the Author):
Pancreatic beta-cells respond to glucose with calcium influx, a modest swelling and insulin secretion. The glucose-induced beta-cell swelling activates a volume-regulated VRAC anion channel, which has been predicted to play an important role in tuning beta-cell function. The manuscript by Till Stuhlmann et al. details for the first time that LRRC8A is a subunit of the beta-cell VRAC channel. The manuscript determines that glucose increases LRRC8/VRAC activity and that this provides a depolarizing influence on the beta-cell membrane potential that enhances glucose-stimulated calcium entry and insulin secretion. Moreover, the LRRC8/VRAC channel is shown to play an important role in regulating glucose tolerance, thus, beta-cell deficient LRRC8A mice show glucose intolerance. The manuscript includes a significant amount of high quality data relevant to understanding the importance of LRRC8/VRAC channels to beta-cell function. There appears to be a few issues that require attention. These are specified below.
We thank the reviewer for appreciating the high quality and importance of our paper.

Major:
As the activation of VRAC has important influences on beta cell functions in response to swelling by allowing regulatory volume decrease (RVD), cellular function may become altered over time due to loss of VRAC mediated RVD. This could influence many aspects of beta cell function such as the activities of other ion channels that are sensitive to cell swelling (e.g. KATP and voltage-dependent calcium channels) or it may have negative influences on beta cell health (e.g. changes in inflammasomes activation). Thus, some experiments to ensure that the loss of VRAC is only influencing calcium based on its acute influence on the membrane potential would be important. The Glucose-activated VRAC current is important; this is the first major current besides KATP that has been identified. As the size of the beta-cell typically adjusts fairly quickly after glucose treatment in part due to VRAC activation, this current would be predicted to be transient as the authors indicate in the discussion. Therefore, it would be good to add some temporal analysis of this VRAC current activation and how long it remains active for following glucose simulation to help increase the understanding of its control of beta cell membrane potential. Moreover, it may be that the current follows the oscillations in islet swelling thus turning on and off during the oscillatory phases of islet calcium influx. Thus, some analysis of calcium oscillations would add important insights into the importance of this channel for 2nd phase insulin secretion.

We agree with the reviewer that the VRAC current should be transient, as we had already discussed in the manuscript. We have quite extensive experience with VRAC currents measured in HEK cells for our structure-function studies. In these whole-cell experiments, in which we stimulate VRAC with commonly used rather brutal decreases in extracellular osmolarity, VRAC currents reach a maximum after about 8-17 minutes and then begin to decline with variable kinetics. This decline could be to various factors, including some intrinsic channel inactivation and changes in cytosolic ion composition. While we agree with the reviewer that it may be interesting to know the time-course of the decline of glucose-induced VRAC urre ts i β-cells, these experiments are exceedingly difficult and prone to artifacts. In our hands, it is very rare to get stable perforated
pat h re ordi gs of β-cells for more than 15-20 minutes, and we assume that even perforated patch measureme ts ill affe t β-cells and VRAC inactivation over that time period. Therefore, we were unable to perform the suggested experiments. Supplementary Fig. 4 and in Fig. R3 for the reviewers. We refer to these results on pages 7-8 in the main text.

Minor:
The follo i g se te e eeds refere es, . Glu ose-se si g β-cells involves Glut2-mediated cellular uptake of glucose a d its o ersio to ATP a d other eta olites. While Glut is the pri ar glu ose transporter in mouse beta-cells, Glut1 is also present in human beta-cells.
We had indeed focused here on Glut2 because our study is concerned with mice. In order to not discuss the presence of different Glut isoforms in different species, which may disturb the flow of the te t, e ha e ha ged the ordi g to a ore ge eri : Glu ose-se si g β-cells involves glucose transporter-mediated cellular uptake of glucose.
The images in figure 1f show some yellow in the merge images of glucagon and insulin staining. This would indicate that some of the islet cells may be expressing both glucagon and inulin. The amount of these ihor o al ells sho i the β -Δ8a i e also see s to be greater. As glucagon and insulin colocalization is not typical in mouse pancreas, either a different imaging technique such as confocal is needed or some more representative sections are required for this figure. Fig. R4. We see no significant overlap between insulin and glucagon labeling in either genotype and have replaced the image in Fig. 1f

by a new one to avoid confusion. (Please note that printing on a color printer at this resolution can still artificially lead to yellow color).
I the Figure lege d the follo i g li e is presu a l a istake, Error! Refere e sour e ot fou d.
Thank you, yes, that was a mistake introduced by EndNote, which has now been corrected.
I the dis ussio the authors e tio that, . A ordi gl a arked effe t of VRAC o i suli se retio as o l o ser ed duri g the first ~ 5 i utes of glu ose additio to isolated islets. Ho e er, the manuscript shows that insulin secretion differed when looked at for 8 minutes. It is not clear why the authors state 15 minutes when they only tested 8 minutes for the short duration?
Thank you for pointing this out. This statement (~15 min) was based on our belief that a marked difference would last longer than the eight minutes, the time at which we did the measurement, since there was still a (non-significant) tendency of lower insulin secretion at 30 min (Fig. 6a)

Reviewer #3 (Remarks to the Author):
This is an interesting and provocative new paper on the role of VRAC Cl ion channels in beta cell stimulus-secretion coupling. The importance of the work is that it may well represent the molecular identification of previously described volume sensitive Cl currents described in beta cells and provide insights into their possible physiological role.
We thank the reviewer for appreciating the importance of our paper.
While the efforts to selectively delete LRRC8 in beta cells in mouse are pretty convincing, the physiological impact is more important to establish, especially given that overall, the mouse lacks a clear phenotype. There are some specific concerns that require attention regarding the manuscript that must be addressed.

MAJOR
1. The data shown in Fig. 1 are generally convincing but the westerns shown in Fig. 1C would benefit from quantitation to firm up the finding of selective knockout and statistics of the result.
We have now quantified the data from three Western blots of lysates form three different mouse pairs. We display the results in a new bar diagram in Fig. 1c which shows a marked, highly significant reduction in Lrrc8a expression in beta cell KO islets. Fig. 2 support the conclusion that hypotonic solution activates a Cl current in WT mouse beta cells but not the deletion mutants. However, as the authors decry the use of drugs such as DCPIB by the field one wonders why they did not also show that siRNA against VRAC in WT blocks the rectifying Cl current? We a o l spe ulate that the Cl>I sele ti it reported for β-cell lines resulted from a contamination with background currents, or other experimental problems.

The data shown in
4. Fig. 3A shows that 20 mM glucose causes activation of the current in WT but not VRAC knockouts. Was cell volume measured in this case? And did application of DCPIB alter volume? As cyclic AMP was also reported to activate islet Cl current, was this also tested?
We have now followed this reviewer (and reviewer 1 (minor point)) and have measured the s elli g of β-cells in response to extracellular glucose (25 mM) using the calcein method (shown in new panel a of Fig. 3). Even with 25 mM glucose, this cell swelling is significantly less than elicited with a strong decrease in extracellular osmolarity, which we now show for comparison in Fig. 2a. Importantly, as we now also state i the te t, oth WT a d Lrr 8a KO β-cells swell to similar degrees, excluding impaired glucose-i du ed ell s elli g as e ha is for redu ed β-cell excitability and insulin secretion with Lrrc8a KO. Interestingly, while regulatory volume decrease (RVD) is clearly visible in the continued presence of hypotonicity (Fig. 2a), there is no clear sign of RVD during the same time span with high glucose. This may suggest that the continued metabolism of glucose compensates during that time for the release of osmolytes trough VRAC, as now hypothesized in Discussion (page 10-11).
We did not test the effect of DCPIB on cell volume, because the beta-cell specific Lrrc8a KO is the most specific way to test for an involvement of VRAC, and because DCPIB is notoriously nonspecific. The latter point is bolstered by our new experiments in which we tested DCPIB on glucoseinduced Ca 2+ oscillations (Fig. R1 for reviewers), where DCPIB also abolished the response of KATP. The effect of cAMP has previously been attributed to an effect on CFTR (Guo et al, Nat Comm. 2014) and was not investigated further in our study. Fig. 5 is especially important to the paper since here, physiological rather than swelling-related stimuli are chosen for study. One can argue that hypotonic solutions are not especially relevant to the study of beta cells as they are not likely to be subjected to external swelling stress in patients with normal kidney function. However, in Fig. 5 glucose is used which is relevant. The issues with this figure, however, are that little is documented short of a time lag between glucose introduction and the activation of spiking or free Ca concentration.

5.
What about longer term recordings of spiking or bursting cells or islets?
As stated above (response to reviewer 2) it is difficult to keep stable perforated patches over a long period of time, and even this technique, though more gentle than whole-cell recordings, is likely to affect the cell (e.g. by cytosolic changes in the concentration of cations, which permeate through gramicidin). We therefore preferred to address this issue with measurements of intracellular calcium, as requested also by reviewer 2. We now show our evaluation of cytosolic Ca 2+ oscillations in WT and beta-cell specific Lrrc8a KO islets in the new Supplementary Fig. 4 and in Figure R3 for the reviewers. No significant difference between the genotypes could be detected.
How do the results depend on the concentration of glucose being applied to the islet? Is the dose response right shifted in the knockouts?
This is indeed an interesting question. We now performed measurements of cytosolic Caconcentrations of WT and KO β-cells in response to 6 and 8 mM glucose (see Suppl. Figure 3).  Fig. 3) And for the Ca data in Fig. 5f, the raw tracings should be shown in addition to the plot of the fura2 responses.
We agree that it is useful to show the primary raw traces. We have now added these both to Fig.  5f and to the new Suppl. Fig. 3 described above.
6. Fig. 6 has its own issues. First, the choice of glucose challenges is somewhat off as only 3.3 and 25 mM glucose are studied and the latter is a hammer dose. Much nicer if the authors also showed 11-15 mM doses as in Fig. 5. Also, as they claim an important factor is the timing of the bout of secretion, it would behoove them to do a proper glucose perifusion experiment where flowing solution is continuously collected as it travels over a small number of islets. This would be helpful for documenting the true actions of channel deletion on both phase 1 and phase 2 of secretion.
We agree, of course, that 11-M e ter al glu ose is ore ph siologi al tha the ha er dose of M e ha e used pre iousl although this concentration can be nearly reached in glucose tolerance measurements, see Fig. 6). We have now performed new experiments using only 10 mM glucose, and indeed also observe a significant, roughly 50% decrease of insulin secretion (with sampling for 8 min), very similar to the results obtained previously for 25 mM. We have now shown the response to 10 mM glucose in the main text (Fig. 6b) and have moved the data for 25 mM glucose into the Supplement (Suppl. Fig. 5).

7.
With regards to the impressive data in Fig. 6C, which shows clear glucose intolerance after genetic deletion of LRRC8a, I have two concerns. First, it has been shown that the RIP2 promoter is not specific for only beta cells but can alter metabolism through actions in the hypothalamus. Do the authors have data ruling out expression in hypothalamus? Second, it was not clear whether islets from control animals lacking cre but having floxed LRRC8a or having cre but lacking floxed LRRC8a were all analyzed?
Please see our detailed response to reviewer 1 who had the same concern in his/her point 2. In short, we have confirmed that the used Cre line is expressed in scarce, scattered cells of the brain, including the hypothalamus (Figure for reviewers R2). However, the insulin tolerance test shows that there is no effect on peripheral glucose metabolism, it seems unlikely that it influences the acute glucose tolerance experiment, and an influence on our in vitro experiments with isolated islets or β-cells is almost excluded. Moreover, we have now included the Cre+, loxP negative controls that yield glucose-or insulin-tolerance results indistinguishable from Cre-, loxP controls (see Fig. 6c, d).

Do the authors have data proving that the Cl channel under study is indeed localized only to the beta cell plasma membrane and not to intracellular organelles?
This is a good pointindeed, immunohistochemistry for our Lrrc8d-tdTomato mice (Fig. 1b)  Reviewer #4 (Remarks to the Author): In this paper, Stuhlmann and colleagues show that mouse pancreatic beta cells express the volumeregulated anion channel subunits LRRC8A and LRRC8D. They generated a mouse line deficient of the essential LRRC8A channel subunit and show that beta cells undergo cell swelling and have increased LRRC8A-dependent Cl-current in response to glucose stimulation. The ability of glucose to induce membrane depolarization, electrical activity and Ca2+ influx was delayed in the LRRC8A-deficient beta cells. This effect was specific to glucose and was not observed when beta cells were depolarized using the K-ATP channel blocker tolbutamide or increasing the extracellular K concentration. The delay in membrane depolarization also resulted in reduced initial insulin secretion and glucose tolerance in vivo.
The study is well conducted and written and addresses an important aspect of the pancreatic beta cell stimulus-secretion coupling.
We thank the reviewer for appreciating the importance and quality of our work.
I have a few concerns with the study that needs to be addressed: 1. The authors state that the Lrrc8a deficient mice have a normal lifespan but do not provide any data to support it.
Thank you for pointing this out. Our statement was imprecise. We did not perform a systematic study of lifespan, waiting for a year or more for mice to die of natural causes. We rather meant that there was no increased lethality of beta-cell specific KO mice until 30 weeks of age, an age after which all mice had been sacrificed for experiments. We have now added new data (in Fig.  1j) showing that there was no difference in weight development during this time between the genotypes.
We now state at the end of the third paragraph of results: β -Δ8a i e showed no increased lethality during our observation period of 30 weeks and their weight developed normally during that time (Fig. 1j). The wording (at the end of the first paragraph of results) has been replaced by: … ere ia le a d la ked a o ert phe ot pe.  Figure R2. Ins2-Cre mediated recombinase activity is scarce and not restricted to specific regions of the brain hypothalamus.
Ins2-Cre mice were bred to TdTomato reporter mouse line (B6.Cg-Gt(ROSA)26Sor tm9(CAG-tdTomato)Hze /J). Images of two individual coronal brain slices show an anterior (A) and a posterior (B) view of the hypothalamus. 50 µm thick free floating slices were stained with RFP antibody and nuclei counterstained with DAPI. Tomato staining is shown in gray while merged images show nuclei co-staining. Zoom images in left column display detailed staining in different brain areas. 3V, third ventricle; PVA= Paraventricular thalamic nucleus; DM= Dorsomedial hypothalamic nuclei Figure R3. Calcium oscillations in intact islets. Different oscillation pattern were observed for both genotypes after stimulation of intact islets with 10 mM glucose. 10 mM glucose was already added 30 min before measering. N=18 for Lrrc8a lox/lox and βc-Δ8a islets from 4 different animals each.  Figure R4. Normal number and distribution of β-and α-cells in pancreatic islets upon β-cell-specific Lrrc8a disruption. Immunofluorescent staining of pancreatic islet sections of (a) Lrrc8a lox/lox and (b) βc-Δ8a mice using β-and α-cell marker proteins insulin (green) and glucagon (red), respectively. Representative islets from two separate mice per genotype; mouse ID number followed by genotype is depicted in the left.

Insulin
Till Stuhlmann et al. provided some discussion and additional data on the two major limitations raised last time. While some questions were addressed, there are still remaining concerns on those two major points.
1. The modest modulatory role of LRRC8A in regulating glucose-stimulated insulin secretion, Ca2+ influx and electrical activity ( Figure 5-6) is in direct contrast with a recent paper published in Nature Communications by Rajan Sah and colleagues (SWELL1 is a glucose sensor regulating β-cell excitability and systemic glycaemia). Data presented in that study showed that SWELL1/LRRC8A is absolutely required for glucose-stimulated insulin secretion, Ca2+ influx and electrical activity in both primary mouse β-cells and primary human β-cells. The authors should cite and discuss the differences. Because they are the main findings of two otherwise very similar studies, it's thus important to understand and resolve where the major differences came from. Rajan Sah et al. acutely knocked out SWELL1/LRRC8A through Cre expression in flox/flox mouse β-cells. They also acutely knocked down SWELL1 through shRNA expression in human β-cells. However, Till Stuhlmann et al. isolated β-cells from control and conditional knockout mice, where LRRC8A was deleted early on during development. It is known that many other ion channels, both cation (TRP) and chloride channels (TMEM16, CFTR etc), can contribute to β-cell excitability. Is it possible that other ionic mechanisms partially compensate the loss of LRRC8A during development? This can be readily tested with Cre expression in isolated flox/flox mouse β-cells. Any other potential explanation for this major discrepancy and confusion? 2. The authors added Ins-Cre mice as an important control and showed they behave similarly as flox/flox mice, which excluded the effect of transgene expression. However, consistent to the literature, the authors did find scattered yet widespread Ins-Cre mediated recombinase activities all cross the brain (I encourage the authors show the brain cre expression data together with those of pancreas in Figure S6). I disagree with the authors' believe that leaky Cre expression "is not relevant for the present study". Deletion of VRAC in those neuronal populations may partially contribute to the impaired glucose tolerance through well-established central regulation of energy metabolism, as it is revealed for some Ins-Cre conditional ko mice in recent years. This is the reason that β-cellspecific inducible Cre (like MIP-CreER) is a preferred tool in order to draw a convincing conclusion. The point is well taken on the constraint with new cre line breeding. However, the in vitro data