ATP activates bestrophin ion channels through direct interaction

Human Bestrophin1 (hBest1) is a Ca2+-activated Cl− channel in retinal pigment epithelium (RPE) essential for retina physiology, and its mutation results in retinal degenerative diseases that have no available treatments. Here, we discover that hBest1’s channel activity in human RPE is significantly enhanced by adenosine triphosphate (ATP) in a dose-dependent manner. We further demonstrate a direct interaction between ATP and bestrophins, and map the ATP-binding motif on hBest1 to an intracellular loop adjacent to the channel activation gate. Importantly, a disease-causing mutation of hBest1 located within the ATP-binding motif, p.I201T, diminishes ATP-dependent activation of the channel in patient-derived RPE, while the corresponding mutants in bestrophin homologs display defective ATP binding and a conformational change in the ATP-binding motif. Taken together, our results identify ATP as a critical activator of bestrophins, and reveal the molecular mechanism of an hBest1 patient-specific mutation.

that "no other protein" was present.
The data suggest that the channel is gated by the free acid of ATP (without complexed Mg). It would be interesting to know whether Mg.ATP also works.   4C. states n=3. I presume this means 3 bilayers for each construct, or does it mean 3 different protein purifications? I think a larger n is required from at least 2 different protein purifications. Also, Po was calculated from how many events? All-points histograms for the entire experiment should be shown accompanying the sample traces.
Reviewer #2 (Remarks to the Author): NCOMMS-18-08461-T Zhang et al. report the identification of ATP as a co-regulator of the calcium activated chloride channel Bestrophin (BEST) using a combination of binding assays and functional studies. The data support this main conclusion.
The authors claim on p. 8 that the A1 and A3 mutants of KpBest lost interaction with ATP because of non-specific channel disruption. However, there are no data presented to support this claim. The lack of data is surprising, as it seems that these proteins were purified for the experiments. Thus, the authors must have size exclusion data that could be presented and that would establish whether the gross properties of these mutants are similar to or different from the wild-type channel. Such data would help establish whether disruption of the putative ATP also destroys the integrity of the channel and should be shown.
The authors perform functional experiments in the background of cells having a non-functional mutation, P274R. Because Best channels are multimers, this background raises the possibility that the measurements they make are do not purely represent the introduced channels but may include mixed heteromultimers containing the introduced channel and some number of P274R mutants. Even though P274R is non-functional on its own, it is unclear whether one or a few of these mutant subunits could form a functional channel when co-assembled with other functional subunits. The authors need to address this point as otherwise the exact nature of the measured channels in Fig. 5 is unclear making it difficult to drawn any conclusions from these experiments.
The authors determine the structure of KpBEST mutant, I180A and claim (p.12) that this mutant dramatically increases the activation gate opening, but does not affect the neck. No evidence for the latter claim is presented. Please show this point.
The authors claim that KpBEST and cBest1 have a 'very similar' all atom RMDS. The value is 12.3Å!!! and is reduced after 5 cycles (of what?) to 4.5Å Neither value can be used to claim similarity. Some explanation is need. Inspection of Fig. 7A indicates that at least two of the transmembrane helices appear to have a register shift. This may be the source of the mismatch. Further, given that the sequences of the two proteins are not identical, the authors might be better served focusing on the Calpha superposition. The figure clearly indicates some similarity, and I suspect that the issues with the large numbers are due to sidechain mismatches and the two (or maybe three, it is hard to see from the figure, especially as the blue and green are similar hues) helix mismatches.
Reference to Fig. S3 should occur immediately after the claim that the authors obtained the first purified hBest1 protein (Line 34 of last full paragraph on p. 6).

Fig. 7C and d would be improved if the two structures each had labeling indicating which is KpBest and which is cBEST1.
Reviewer #3 (Remarks to the Author): The manuscript by Zhang et al reported activation of a bacterial homolog of human bestrophin channel by ATP and its analogs. The authors went on to show that L177T can abolish ATP binding and also renders the bacterial channel insensitive to ATP. The simplest explanation is that ATP binds to KpBest through L177, although an allosteric mechanism cannot be ruled out.
The authors then expanded the study to human and bovine bestrophin channels, and eventually concluded that the mammalian channel can also bind to and be activated by ATP and that one of the intracellular loop is likely the binding site for ATP.
While I appreciate the combination of structural and functional approaches, and recognize that the use of induced human RPE cells for functional studies could add significance and physiological relevance to the study, I feel that several key experiments are missing and therefore the conclusions on the mammalian channels are not well justified. In the current format, the manuscript only demonstrated that KpBest is activated by ATP, and that ATP activates KpBest through changes at regions near L177.
Here are the questions I have: 1. Studies of ATP activation of bovine or human bestrophin channel should be done on either purified protein reconstituted into liposomes (for the bovine channel) or on a heterologous expression system such as HEK cells (for either bovine or human channels). That way, bestropin channel currents can be rigorously validated. This step is necessary to establish that ATP activates mammalian bestrophin channels. 2. Once step 1 is complete, mutations can then be tested to identify regions that are sensitive to ATP activation. At this point, ATP binding to either purified bovine or chicken bestrophin (both the wild type and mutations at the homologous position of L177) should be measured, and functions of the wild type and mutant channels recorded and compared. 3. Once step 1 and 2 are complete, the recordings on induced human RPE cells would then become impactful. And even at this stage, validation of the recorded currents on RPE cells is necessary because there are other channels that could produce chloride current on a native cell. 4. Independent of issues with the mammalian bestrophin channels, the logic of presenting the structure of the bacterial I180A channel is not clear. This is a channel that has a high Popen without the presence of ATP. The opening of the channel is almost entirely due to the truncation of the Ile side chain. In which way does it represent an ATP activated channel? Does it still bind to ATP? If so, how does channel activity change in the presence of ATP? 5. If ATP affinity is at the micromolar range, a structure of ATP in complex with KpBest should be attainable and that would address the question of whether ATP activates the channel by directly interacting with L177 or through an allosteric effect. 6. Related to #5, the loop mutations that produced functionally null channels should be examined for ATP binding. This should be done for both the BpBest and mammalian bestrophins. The reason given for not following up on these mutations is not compelling.

"The authors claim on p. 8 that the A1 and A3 mutants of KpBest lost interaction with ATP
because of non-specific channel disruption. However, there are no data presented to support this claim. The lack of data is surprising, as it seems that these proteins were purified for the experiments. Thus, the authors must have size exclusion data that could be presented and that would establish whether the gross properties of these mutants are similar to or different from the wild-type channel. Such data would help establish whether disruption of the putative ATP also destroys the integrity of the channel and should be shown".
Response: As the reviewer suggested, we now have added the MST data for A1/A3 and size exclusion results for A1-A4 in Supplementary Fig. 4. The size exclusion profiles of A1-A4 are indistinguishable from that of the WT KpBest ( Supplementary Fig. 1), suggesting that the integrity of the channel is retained in all these mutants. We agree with the reviewer that the involvement of loops 1 and 3 in ATP binding cannot be ruled out. However, it is difficult to further explore this possibility when conserved residues in these motifs are irreplaceable for constituting a functional channel. It should be noted that in the literature mutagenesis of hBest1 often results in loss of channel function, stressing the structural rigidness of the channel. Therefore, it is more informative to focus on loop 2 which is specifically involved in ATP binding and ATP-dependent activation. We have revised the text on Page 8 accordingly.
2. "The authors perform functional experiments in the background of cells having a non-functional mutation, P274R. Because Best channels are multimers, this background raises the possibility that the measurements they make do not purely represent the introduced channels but may include mixed heteromultimers containing the introduced channel and some number of P274R mutants. Even though P274R is non-functional on its own, it is unclear whether one or a few of these mutant subunits could form a functional channel when co-assembled with other functional subunits. The authors need to address this point as otherwise the exact nature of the measured channels in Fig. 5 is unclear making it difficult to drawn any conclusions from these experiments."

Response:
We have added co-IP results in Supplementary Fig. 5, showing that the P274R mutant does not interact with WT or any of the A1-A4 mutants. Therefore, P274R cannot interfere with the function of the introduced WT or A1-A4 channels by forming heteromultimers.
3. "The authors determine the structure of KpBEST mutant,I180A and claim (p.12 Fig. 7A indicates that at least two of the transmembrane helices appear to have a register shift. This may be the source of the mismatch. Further, given that the sequences of the two proteins are not identical, the authors might be better served focusing on the Calpha superposition. The figure clearly indicates some similarity, and I suspect that the issues with the large numbers are due to sidechain mismatches and the two (or maybe three, it is hard to see from the figure, especially as the blue and green are similar hues) helix mismatches."

Response:
The reviewer raised an important point, and we agree that the Pymol alignment results were not clear. In order to re-evaluate the structure similarity between KpBest and cBest1, we performed CCP4 superpose using the Secondary Structure Matching mode, and the resulting RMSD between KpBest (4WD8) and cBest1 (4RDQ, without antibody molecules) is 2.4216 Å (1077 residues alignment). Thus, our new alignment data support that the overall structures of KpBest and cBest are very similar. We have revised the manuscript accordingly (Page 15, paragraph 2). Fig. S3 should occur immediately after the claim that the authors obtained the first purified bBest2 protein (Line 34 of last full paragraph on p. 6)."