The transcriptional corepressor CtBP2 serves as a metabolite sensor orchestrating hepatic glucose and lipid homeostasis

Biological systems to sense and respond to metabolic perturbations are critical for the maintenance of cellular homeostasis. Here we describe a hepatic system in this context orchestrated by the transcriptional corepressor C-terminal binding protein 2 (CtBP2) that harbors metabolite-sensing capabilities. The repressor activity of CtBP2 is reciprocally regulated by NADH and acyl-CoAs. CtBP2 represses Forkhead box O1 (FoxO1)-mediated hepatic gluconeogenesis directly as well as Sterol Regulatory Element-Binding Protein 1 (SREBP1)-mediated lipogenesis indirectly. The activity of CtBP2 is markedly defective in obese liver reflecting the metabolic perturbations. Thus, liver-specific CtBP2 deletion promotes hepatic gluconeogenesis and accelerates the progression of steatohepatitis. Conversely, activation of CtBP2 ameliorates diabetes and hepatic steatosis in obesity. The structure-function relationships revealed in this study identify a critical structural domain called Rossmann fold, a metabolite-sensing pocket, that is susceptible to metabolic liabilities and potentially targetable for developing therapeutic approaches.

1. Figure 1. CtBP2 is shown to inhibit FoxO1 activity via its interaction with the C-terminal domain of FoxO1. CtBP2 inhibition of FoxO1 is enhanced by insulin and diminished by glucose administration, without providing mechanistic insights. This observation seems at odd with the physiological state, in which glucose administration is coupled with acute insulin secretion from the pancreas in mice. 2. Likewise, CtBP2 acts to inhibit SREBP1 activity, without providing mechanistic insights. This mechanism seems important but appears counterintuitive. SREBP1 and FoxO1 are acting in opposite pathways, meaning that FoxO1 protein expression and activity are induced in fasted states, whereas SREBP1c protein expression and activity become activated in fed states. Physiologically, it is difficult to conceptualize the idea that CtBP2 functions to inhibit both SREBP1 and FoxO1 pathways, regardless of fasting or refeeding. 3. This also begs the question as to how CtBP2 expression is regulated in the liver and why CtBP2 is uniquely required for inhibiting two opposing pathways (FoxO1 in fasted phase and SREBP1 in fed state) in the liver? 4. Figure 3 panel g. Data from limited human liver autopsies (n=3-4) indicated that FoxO1-CtBP2 interaction was lower in obese subjects. Close examination of these data in Panel g does not strongly support the authors' conclusion. Neither FoxO1 nor CtBP2 protein levels were changed in lean vs. obese liver in human subjects. Fasting or fed states were not known either. 5. Figure 3 Panel I. Data in this panel appeared from a single mouse without statistics in each group. 6. Figure 4c. This panel shows that FoxO1 occupancy in the promoters of G6pc and Pck1 was similar in lean and HFD mice, contradicting the published data that FoxO1 activity, along with augmented gluconeogenic gene expression, is higher in the liver in obese mice. This panel's data are also at odd with the authors' idea that FoxO1 activity is higher in obese liver, due to the lack of CtBP2 inhibition on FoxO1 activity in obese mice ( Figure Fig. 3e-i and Extended Data Fig. 4f). Please double-check these experiments to resolve these apparent self-conflicting results and controversies with data in the literature. 7. Figure 4b. It was not understandable why the authors chose a heterozygous HEK293 cell line (instead of a liver-based cell line) for ChIP assays. All other studies were performed on the mouse liver or hepatocytes. 8. Figure 4g. Despite a significant increase in fasting blood sugar levels and postprandial blood glucose levels, CtBP2-KO mice and WT had a similar fasting plasma insulin levels. This observation deserves a discussion. 9. Figure 4J and Figure 4m. CtBP2-KO mice developed mild steatosis on regular chow, and this effect was exacerbated on MCD diet. These effects ensued without changes in lipogenesis in the liver. What is the underlying mechanism? What are blood triglyceride levels in CtBP2-KO mice? A significant induction of steatosis in the absence of altered lipogenesis could derived from decreased VLDL secretion or reduced fatty acid oxidation in the liver -a possibility that were not attempted by the authors in the manuscript. Indeed, MCD diet feeding is known to cause steatosis by curbing hepatic VLDL-triglyceride secretion. 10. Figure 4. The authors did not determine hepatic FoxO1 or SREBP1 mRNA or proteins in the liver of CtBP2-KO mice. It remained unknown and unanswered whether hepatic CtBP2 depletion caused a significant increase in FoxO1 and SREBP1 activity, giving rise to augmented gluconeogenesis and increased lipogenesis in the liver of CtBP2-KO mice. 11. Figure 5. The authors used adenoviruses to increase CtBP2 expression in the liver of HFD obese mice. This experiment was based on the idea that CtBP2 is inactivated in the liver of obese mice. However, data presented in Figure 2 panels e and f did not show any significant alterations in liver CtBP2 protein levels in both ob/ob mice and HFD obese mice. Please confirm these studies. 12. Figure 5c. Despite the lack of a clear rationale for replenishing CtBP2 in obese liver, the authors showed that adenovirus-mediated CtBP2 production in the liver improved glucose tolerance. However, a puzzling observation is that obese mice with CtBP2 overproduction had blunt responses to pyruvate injection, indicating that CtBP2-overexpressed liver was impaired to undergo gluconeogenesis. Liver TG content was lower and blood glucose levels were lower in Adv-CtBP2 vector-treated obese mice. Where did the energy go and were blood lipid levels higher in Adv-CtBP2 vector-treated obese mice? It is desirable that the authors provide some explanation or insightful discussion for these observations. Ideally, the authors need to show that Adv-CtBP2 vector-treated obese mice might have increased energy expenditure and/or with a concomitant reduction in body weight gain. But the authors did not provide a reasonable clue to these observations. 13. Figure 5e. Oil Red O staining is required to better show the changes in hepatic lipid content in the liver. Please show these data in addition to H&E staining. 14. Figure 5. Fasting blood sugars in both control vector and Adv-CtBP2 vector-treated obese mice are relatively lower within a normal range. This might derive from a short duration of HFD feeding. Please acknowledge this point in the manuscript. 15. Figure 5. Please show the protein level of Adv-CtBP2 vector-mediated CtBP2 production in the liver over the control group. This information will help gauge how many folds of CtBP2 proteins were needed to achieve the beneficial effect on glucose and lipid metabolism in Adv-CtBP2 vectortreated obese mice. 16. In general, there is a lack of clarity about the state of mice (fasting or fed), in which ex vivo studies were performed in liver tissues. 17. Figure 7b. Data were based on a single measurement and were relatively weak in supporting the authors' conclusion, without quantification. 18. Figure 7. Overexpression of a constitutively active CtBP2 mutant in the liver corrected diabetes and reversed steatosis in obese mice, without changes in body weight. This conclusion needs more experimental data to support. For example, where did the energy go in this model? CtBP2 is presumed to inhibit both gluconeogenesis and lipogenesis in this model. What about the blood fatty acids and triglyceride levels? Does this model, as opposed to its controls, had altered energy expenditure? 19. Please limit the use of vague words such as "robustly" inactivated or "profoundly" decreased, etc, in the manuscript. 20. Please convert all bar groups to dot-based graphs with individual dots representing individual mice for in vivo experiments or individual replicates for in vitro experiments. Dot-based graphs better illustrate the intra-subject variations and reproducibility. 21. Please indicate the molecular weight in all western blots. Reviewer #2: Remarks to the Author: The authors in this manuscript showed that Ctbp2 acts as a metabolite sensor in hepatic glucose and lipid homeostasis by supporting evidence that Ctbp2 cooperated with FoxO1 regulates metabolic genes in obese liver. There is no question about the quality and skills in this work, but, the concept that Ctbp is a metabolic sensor has been already accepted in this field.
Major Concerns 1) Page7 line 6-7: they mentioned that Ctbp1 is excluded in hepatic glucose metabolism through preliminary screening stage. Since Ctbp1, just like Ctbp2, acts as a transcriptional repressor in many cell types, they should provide the more reliable evidence that Ctbp1 is excluded in this experiment.
2) In Supplementary Figure 2e, they showed that FoxO1 mutant (deltaPSDL) did not alter the association with Ctbp1. As far as I know, Ctbp isoforms interact with substrate protein in a same manner, more sophisticated experiments may be needed. As shown in Figure 4f, mouse liver contains sufficient amount of both Ctbp1 and Ctbp2. Thus, without Ctbp1, all the data in this manuscript is incomplete to explain the FoxO1-Ctbp axis in hepatic glucose metabolism. 3) In Figure 2a, addition of 50 mM of NADH in cell lysates causes to increase the interaction of FoxO1 with Ctbp2. However, cell lysates have already contained the endogenous NADH (and NAD+), they should check the endogenous amount of NADH and calculate the ratio between endogenous NADH and exogenous NADH. NADH addition in cell lysates is different from the buffering system just like Figure 2b. 4) In Figure 3, they mentioned that Ctbp2 is markedly inactivated in the liver of obesity by showing that the dissociation between Ctbp2 and FoxO1 (other transcription factors) in several diet conditions. But these appears to be indirect. I'd like to point out that Ctbp2 itself does not have repressive activity. To strengthen their idea, they should provide the evidence of the dissociation of Ctbp (and FoxO1) with Ctbp2-associated repressors (such as Lsd1, Nurd complex) by IP experiment and ChIP experiments as well as the change of epigenetic marks (such as histone acetylation and so on) in the target genes. 5) In this manuscript, they generated the liver-specific Ctbp2 deficient mice (Figure 4f-m). I think this animal is good for analyzing the function of Ctbp2 in liver under various diet conditions. It would strengthen their idea.
Reviewer #3: Remarks to the Author: An important aspect of this manuscript is the structural model presented for the binding of palmitoyl CoA and acetyl CoA to CtBP2. The description of the structural modeling is quite superficial; if this reflects the actual structural modeling performed, that would be a cause for concern. The authors state that "palmitoyl-CoA and acetyl-CoA were docked into the NAD+ binding site of human CtBP2" and energy minimized. One would expect that the ADP substructure would fit in similarly for NAD+ and the CoA molecules, which provides a nice starting point. However, looking at the structure, it is not at all clear why the remainder of the CoA structures would continue down into the nicotinamide pocket, given the very different chemical structures, rather than exiting the interior of the protein as could easily occur near the phosphate groups. Based on examining the structures, it does not appear that palmitoyl-CoA would disrupt the dimeric assembly if it exited through this passage, whereas wiggling through the nicotinamide pocket to exit through the substrate pocket might. As the authors are pushing the idea that binding would lead to disruption of the dimeric assembly, how the docking was done and what choices were made becomes crucial. Unfortunately, based on the superficial description given, it is impossible to know if their model is likely or not, given the alternatives. Even if the authors are correct on this hypothesis, the presentation of the resultant models (Fig 2c and Fig. 6h) is not very illuminating. The authors also do not show a comparison of the packing of the palmitoyl-CoA with the location of the dimeric interface, which would be helpful for understanding the authors hypothesis that binding would disrupt the interface. It is worth noting that the CtBP dimer interface is very extensive (with buried surface area of over 2500Å2) so it would take substantial interactions from the acyl chain to disrupt it.
A central basis for considering that CtBP acts as a redox sensor is the conclusion from the Goodman group in the early 2000s that the affinity CtBP for NADH is 100 times greater than that of NAD+. (This work was done on CtBP1, so there is no evidence that CtBP2 binds NADH significantly tighter than NAD+ as the authors claim in their manuscript.) As far as I am aware, no lab has been able to repeat/verify the conclusions of the Goodman group. The reported Kd for NADH binding in that work is similar to that reported by other methods of around 100-500nM. However, the reported Kd for NAD+ binding by a more indirect competition measurement of 8-11uM is much higher than any values reported by other groups. In addition to direct measurements of affinity (see, for instance, Madison et al. 2013, JBC), there are measurements of the response of CtBP to NADH and NAD+ in the oligomerization (Bellesis et al. 2018), and ability to stimulate binding of partners (Kumar et al 2002, Balasubramanian et al. 2002. None of these studies found evidence for substantially different responses of CtBP to NADH vs. NAD+. The interpretation of this manuscript rests heavily on a much higher affinity for NADH compared with NAD+, which is not supported by the preponderance of the data available. The dissociation constant that the authors present using microscale thermophoresis between CtBP2 and NADH is two orders of magnitude higher (weaker binding) than reported by other methods. Although this is very briefly addressed in the appendix, I do not find their suggestions for the basis of this discrepancy very convincing, which adds additional concerns to the accuracy of their analysis.
In this study, Dr. Sekiya and colleagues reported that the transcriptional co-repressor C-terminal binding protein 2 (CtBP2) acts a NADH/NAD+ metabolite-sensor for modulating hepatic metabolism in mice. They showed that CtBP2 functions to inhibit FoxO1 activity and this effect contributes to the inhibition of gluconeogenesis in the liver.
Likewise, CtBP2 acts to inhibits SREBP1 activity and this action contributes to the inhibition of lipogenesis in the liver. They further showed that dietary obese mice had lower CtBP2 protein levels in the liver. Liver-conditional CtBP2 depletion promoted both hepatic gluconeogenesis and lipogenesis, causing steatosis in mice.
Adenovirus-mediated production of CtBP2 or its constitutively active mutant ameliorated diabetes and steatosis in dietary obese mice. Data from limited human liver autopsies (n=3-4) indicated that FoxO1-CtBP2 interaction was lower in obese subjects. Based on these data, the authors suggest that CtBP2 serves as a metabolite sensor for integrating intracellular NADH/NAD+ levels to hepatic glucose and lipid homeostasis.
Response: We appreciate the Reviewer 1's careful evaluation of our manuscript. The Reviewer 1 raised a series of important points and we were able to improve our manuscript by responding to them. Some of them were unexpected, and we could not achieve this advancement without the Reviewer 1's suggestions.
Among the comments above, we would like to make sure that CtBP2 protein levels are relatively maintained even in the obese animals but that the activity of CtBP2 to bind to metabolic transcription factors was dramatically reduced in obesity. The Reviewer 1 would be aware of this, but we would like to emphasize that the function of CtBP2 is regulated at the levels of allosteric conformational changes, not at the levels of protein expression.
Overall there are a lot of studies performed in vitro and in vivo, with data obtained from multiple animal models. However, the data are not quite coherent. There are many piece of data derived from single measurement, without quantification and statistics. In particular, mechanistic insights are lacking in supporting the authors' conclusion that CtBP2 inhibits both FoxO1 and SREBP1 -two hepatic pathways that are differentially regulated in opposite manners. There are a number of specific comments for the authors to improve the manuscript.
Response: We are afraid that our description was not sufficient for the Reviewer1 and readers to understand our complicated model. We would like to explain two major points before going to the point-by-point responses.
1) We would respectfully encourage the Reviewer 1 to understand our concept on the role of CtBP2 in physiological conditions. The activity of CtBP2 is relatively maintained stable in physiological conditions ( Fig.3a), therefore contributions of CtBP2 activity to fasting feeding cycle would be relatively small compared to pathological conditions. In addition, CtBP2-mediated SREBP1 suppression requires intermediary molecule(s), LRH1 in our study or other molecules to be identified. Therefore, even when CtBP2 is active, CtBP2 would not be able to suppress SREBP1 without proper LRH1/SREBP1 complex formation. Here we show some data not included in our manuscript (shown below). These sets of data would hopefully help the Reviewer 1 to understand what we mentioned above. However, too much data and emphasis on physiology would make our description too disorganized to lead readers to understand our major point that is potential benefits of CtBP2 activation in obesity, a pathological condition. Considering the structural organization of our paper and relative contribution of CtBP2 activity in physiology, we would like to discuss this aspect elsewhere, not in this paper.
2) The Reviewer1 might be understanding in a way that more FoxO1 would be recruited in the promoters of gluconeogenic genes upon liberation from CtBP2's repression.
However, what we propose is that CtBP2 forms a complex with FoxO1 bound to the promoters, and just dissociation of CtBP2 from the FoxO1 leads to upregulation of those genes while the amount of FoxO1 in those promoters are left unchanged upon liberation from CtBP2. All molecular events occur on the promoters. The Reviewer1's view would be that CtBP2/FoxO1 complex exists in cytosol or nuclei without binding to promoters and that the liberated FoxO1 from the complex would be recruited and bind to promoters upon liberation from CtBP2. We modified our schematic description (Fig. 8h) to clarify that CtBP2 dissociation from FoxO1 is occurring in the promoters and added a graphical description below to help the Reviewer1 to understand our proposed model.
1. Figure 1. CtBP2 is shown to inhibit FoxO1 activity via its interaction with the C-terminal domain of FoxO1. CtBP2 inhibition of FoxO1 is enhanced by insulin and diminished by glucose administration, without providing mechanistic insights. This observation seems at odd with the physiological state, in which glucose administration is coupled with acute insulin secretion from the pancreas in mice.
Response: The Reviewer is asking the role of CtBP2 in physiology and this comment may be related to Figure 3. Glucose administration may be accompanied with insulin secretion, however, administration of either glucagon, insulin or glucose would emphasize the effect of each despite the overall response may reflect metabolic changes evoked secondarily. We showed the effect of fast/fed to have a fair assessment of physiological state without emphasizing specific metabolic cues where we did not observe robust alterations of the CtBP2 activity. Our main claim here is that CtBP2 activity is relatively maintained in a physiological state whereas it is markedly inactivated in obesity. We also described possible mechanisms behind this in our supplementary discussion which may have been missed due to its presence in supplementary discussion. We included the discussion in the main text.
As the Reviewer 1 pointed out, what we expected was that glucose administration would increase the CtBP2/FoxO1 interaction. However, what we observed was opposite but consistent with the literature cited in our discussion that proposes FoxO1 activation in response to glucose to cope with oxidative stress (page 27, line 7-12).
Since glucose utilization inhibits fatty acid oxidation, increased fatty acyl-CoA may explain this observation as we discussed. CtBP2 regulates part of the FoxO1 activity and we would like to emphasize that CtBP2/FoxO1 interaction does not necessarily correlate with total FoxO1 activity especially in physiological conditions. 2. Likewise, CtBP2 acts to inhibit SREBP1 activity, without providing mechanistic insights. This mechanism seems important but appears counterintuitive. SREBP1 and FoxO1 are acting in opposite pathways, meaning that FoxO1 protein expression and activity are induced in fasted states, whereas SREBP1c protein expression and activity become activated in fed states. Physiologically, it is difficult to conceptualize the idea that CtBP2 functions to inhibit both SREBP1 and FoxO1 pathways, regardless of fasting or refeeding.
Response: Here again the activity of CtBP2 is relatively maintained stable in both fasted and fed states, indicating relative contributions of CtBP2 to alterations of gluconeogenic and lipogenic gene expression in physiology would be limited.
Inactivation of CtBP2 in obesity may in part explain the concomitant activation of gluconeogenesis and lipogenesis in obese liver. In addition, CtBP2 regulates SREBP1 indirectly. Thus, the intermediary molecule, LRH1 in our manuscript or unidentified molecule(s), may also play roles in lipogenic gene expression in this context. CtBP2 activation would directly repress gluconeogenesis irrespective of LRH1 status but CtBP2 would repress lipogenesis in a LRH1-dependent manner as we described in our discussion. To help the Reviewer 1 to understand this complexity, we provided data above mentioned. Interaction of CtBP2 with FoxO1 is not always in parallel with that with SREBP1 in physiology that may be explained by the presence of intermediary molecule(s) or other mechanisms. We can further investigate this aspect, but relative contributions of CtBP2 activities to physiological changes (i.e. fasting feeding cycle) are limited, therefore, we would like to focus on pathological conditions in this report. Too much data to clarify the details in physiology would not be friendly for broad readership in this journal, we believe.
3. This also begs the question as to how CtBP2 expression is regulated in the liver and why CtBP2 is uniquely required for inhibiting two opposing pathways (FoxO1 in fasted phase and SREBP1 in fed state) in the liver?
Response: Regarding the expression of CtBP2, we showed it in Extended Data Fig. 5c.
What we have observed is that protein expression of CtBP2 is relatively maintained compared to the drastic alterations of CtBP2 activity in obesity. We would emphasize that functional binding ability of CtBP2 is diminished in obesity and that protein expression of CtBP2 is relatively maintained.
Based on our observations, CtBP2 expression at mRNA and protein levels may respond to physiological and pathological conditions but the extent of alterations is relatively small and we cannot draw a consistent conclusion through multiple experiments. Therefore, we would like to focus on the CtBP2 activity in this study that is consistent throughout multiple experiments. We would like to avoid showing data with uncertainties and we believe we can propose our model without determining the relatively small fluctuation of CtBP2 expression since the CtBP2 activity predominantly determines the effects of CtBP2. 4. Figure 3 panel g. Data from limited human liver autopsies (n=3-4) indicated that FoxO1-CtBP2 interaction was lower in obese subjects. Close examination of these data in Panel g does not strongly support the authors' conclusion. Neither FoxO1 nor CtBP2 protein levels were changed in lean vs. obese liver in human subjects. Fasting or fed states were not known either.
Response: These are human autopsy samples and we cannot control fasted/fed state and quality of samples. Those samples were collected hours after donors died. We already tried to improve the quality of data, but we cannot expect high quality for these human samples compared to those from experimental animals. If this set of data can be an obstacle, we would withdraw the data, which would not influence rigidity of our proposal. Again, we would like to emphasize that CtBP2 activity shown in the co-immunoprecipitation experiments are reduced and that we don't necessarily need to observe reduced protein expression of CtBP2. 5. Figure 3 Panel I. Data in this panel appeared from a single mouse without statistics in each group.
Response: We increased the number of samples with statistics (new Fig.3l).
6. Figure 4c. This panel shows that FoxO1 occupancy in the promoters of G6pc and Pck1 was similar in lean and HFD mice, contradicting the published data that FoxO1 activity, along with augmented gluconeogenic gene expression, is higher in the liver in obese mice. This panel's data are also at odd with the authors' idea that FoxO1 activity is higher in obese liver, due to the lack of CtBP2 inhibition on FoxO1 activity in obese mice ( Figure Fig. 3e-i and Extended Data Fig. 4f). Please double-check these experiments to resolve these apparent self-conflicting results and controversies with data in the literature.
Response: The Reviewer1 raised quite an important point. The question whether or how much FoxO1 occupancy in the G6pc promoter is altered in response to physiological and pathological conditions is not completely settled although the insulin-dependent nuclear cytosolic shuttling has been predominantly accepted.
Although we don't need to argue against this shuttling view to explain our proposal, we can find several reports contradictory to this view. In this paper, they quantify nuclear FoxO1 contents in diet-induced obesity in Figure 3.
They claim that high fat diet feeding increased nuclear FoxO1 contents but we cannot recognize the difference as far as we see the actual blots. Although the shuttling system of FoxO1 has been accepted as an exclusive mechanism, we may need to be careful about the relative contribution of this system based on the literature and our unpublished data.
What we propose is that loss of CtBP2 binding liberates FoxO1 and promotes gluconeogenic gene transcription. In this system, we don't need to observe increased 8. Figure 4g. Despite a significant increase in fasting blood sugar levels and postprandial blood glucose levels, CtBP2-KO mice and WT had a similar fasting plasma insulin levels. This observation deserves a discussion.
Response: The Reviewer1 is again raising an important point. Although we have not identified possible systemic effects on insulin biology induced by liver-specific CtBP2 deletion, we don't deny that possibility based on our preliminary data. We added a statement in our Results section (page 17, line 2-5).
9. Figure 4J and Figure 4m. CtBP2-KO mice developed mild steatosis on regular chow, and this effect was exacerbated on MCD diet. These effects ensued without changes in lipogenesis in the liver. What is the underlying mechanism? What are blood triglyceride levels in CtBP2-KO mice? A significant induction of steatosis in the absence of altered lipogenesis could derived from decreased VLDL secretion or reduced fatty acid oxidation in the liver -a possibility that were not attempted by the authors in the manuscript. Indeed, MCD diet feeding is known to cause steatosis by curbing hepatic VLDL-triglyceride secretion.
Response: This is quite an important suggestion but we again omitted a set of data to keep readers on the main path. The extent of alterations of plasma lipids induced by CtBP2 deletion in liver was modest, but the lipid fluxes would at least in part contribute to phenotype. We added data in Extended Data Fig. 6h, k. In addition, plasma cholesterol levels are negatively correlated with CtBP2 activity in liver, that would be informative for pharmacologists looking for new druggable targets for dyslipidemia considering the broad readership of the journal.
10. Figure 4. The authors did not determine hepatic FoxO1 or SREBP1 mRNA or proteins in the liver of CtBP2-KO mice. It remained unknown and unanswered whether hepatic CtBP2 depletion caused a significant increase in FoxO1 and SREBP1 activity, giving rise to augmented gluconeogenesis and increased lipogenesis in the liver of CtBP2-KO mice.
Response: We again appreciate the Reviewer 1's suggestion. We assumed the expression levels of these two transcription factors were unchanged, but in fact both FoxO1 and SREBP1 expression levels were reduced in CtBP2 knockout mice (new Fig.   5g, Extended Data Fig. 6i). This would be an additional compensatory change and would explain modest phenotype observed in the knockout mice compared to that observed in the overexpression models.
11. Figure 5. The authors used adenoviruses to increase CtBP2 expression in the liver of HFD obese mice. This experiment was based on the idea that CtBP2 is inactivated in the liver of obese mice. However, data presented in Figure 2 panels e and f did not show any significant alterations in liver CtBP2 protein levels in both ob/ob mice and HFD obese mice. Please confirm these studies.
Response: We may have had to explain more carefully although we have a statement in our discussion. What we are proposing is that obese mice have decent amount of CtBP2 protein but that is functionally inactivated. CtBP2 in obesity cannot bind to FoxO1 due to the functional deterioration. We firstly overexpressed wild-type CtBP2 to activate it although it does not faithfully mimic metabolite-dependent CtBP2 activation. Therefore, we added a study using a CtBP2 mutant which mimic functional activation in Figure 8.
We need to identify small molecules to activate CtBP2 metabolite-sensing pocket to completely convince readers. This will be out of scope of this paper, but we indeed identified and have it in our hands with expected metabolic benefits. We hope the amended schematic description of our model in Fig.8h would facilitate proper understanding.
12. Figure 5c. Despite the lack of a clear rationale for replenishing CtBP2 in obese liver, the authors showed that adenovirus-mediated CtBP2 production in the liver improved glucose tolerance. However, a puzzling observation is that obese mice with CtBP2 overproduction had blunt responses to pyruvate injection, indicating that CtBP2-overexpressed liver was impaired to undergo gluconeogenesis. Liver TG content was lower and blood glucose levels were lower in Adv-CtBP2 vector-treated obese mice.
Where did the energy go and were blood lipid levels higher in Adv-CtBP2 vector-treated obese mice? It is desirable that the authors provide some explanation or insightful discussion for these observations. Ideally, the authors need to show that Adv-CtBP2 vector-treated obese mice might have increased energy expenditure and/or with a concomitant reduction in body weight gain. But the authors did not provide a reasonable clue to these observations. Response: We will discuss this in the response to Q18 that is same as this question.
13. Figure 5e. Oil Red O staining is required to better show the changes in hepatic lipid content in the liver. Please show these data in addition to H&E staining. Fig. 7g. Thanks to this comment, we were able to highlight a substantial difference induced by CtBP2 overexpression.

Response: We included Oil Red O staining in Extended Data
14. Figure 5. Fasting blood sugars in both control vector and Adv-CtBP2 vector-treated obese mice are relatively lower within a normal range. This might derive from a short duration of HFD feeding. Please acknowledge this point in the manuscript. 17. Figure 7b. Data were based on a single measurement and were relatively weak in supporting the authors' conclusion, without quantification.
Response: We increased the number of samples with statistics (new Fig. 8b).
18. Figure 7. Overexpression of a constitutively active CtBP2 mutant in the liver corrected diabetes and reversed steatosis in obese mice, without changes in body weight. This conclusion needs more experimental data to support. For example, where did the energy go in this model? CtBP2 is presumed to inhibit both gluconeogenesis and lipogenesis in this model. What about the blood fatty acids and triglyceride levels?
Does this model, as opposed to its controls, had altered energy expenditure?
Response: We provided plasma lipid profiles (new Extended Data Fig. 9f) as well as energy expenditures (new Extended Data Fig. 9g) in these animals. But if we roughly calculate mass turnover, liver lipid reduction would be around (50 mg/g tissue) x ( ~1.5 g liver/mouse) = ~75 mg. If we assume the amount of whole body plasma ~1 ml, triglyceride in blood would be (~100 mg/dl) x (1 ml) = 1 mg. Of course, we need to take the turnover of lipids into account, these are trivial compared to the body weights (~ 40g). This is also the case for blood glucose (we can roughly convert glucose/lipid based on the formula 4 kcal/g for glucose and 9 kcal/g for lipid). Although we agree that CtBP2 activation may have a potential to reduce body weights, we may need longer period of time to observe significant reduction of body weights, which would not be evaluated by the transient adenovirus-mediated overexpression.
However, the Reviewer 1 is raising an important question again as he/she asked a potential role of CtBP2 in insulin biology. We have preliminary data supporting this idea in another model that will be discussed in future publications.
19. Please limit the use of vague words such as "robustly" inactivated or "profoundly" Response: We added some statements in our Introduction section. One is the existence of controversy regarding the redox sensing capability of CtBP2 based on the differential affinities for NADH and NAD + . The other is non-transcriptional function of CtBP1.
Reviewer #2 (Remarks to the Author): The authors in this manuscript showed that Ctbp2 acts as a metabolite sensor in hepatic glucose and lipid homeostasis by supporting evidence that Ctbp2 cooperated with FoxO1 regulates metabolic genes in obese liver. There is no question about the quality and skills in this work, but, the concept that Ctbp is a metabolic sensor has been already accepted in this field.
Response: We sincerely appreciate the Reviewer 2's acknowledgement of our quality.
Regarding the novelty of our findings, CtBP2 as an NADH sensor is widely accepted, but no existing paper demonstrated that it is a fatty acyl-CoA sensor. Importantly, inactivation of CtBP2 in obesity seems to be induced by the increased fatty acyl-CoA, not by NADH/NAD + ratio, suggesting fatty acyl-CoA sensing is one of the major biological roles of CtBP2. While previous studies attempted to understand CtBP2 in the 'NADH vs NAD + ' context, our model in this study is rather that CtBP2 is a sensor working in the 'NADH vs fatty acyl-CoA' context. Indeed, the Reviewer 3 pointed out that our core finding in this manuscript would be the fatty acyl-CoA sensing by CtBP2.
This idea would provide new perspectives to understand our metabolic system and may solve some controversy regarding NADH/NAD + sensing function of CtBP2 that was asked by Reviewer 3. We amended our statements to make these things clearer.
Moreover, our findings indicate that the balance between NADH/NAD + and fatty acyl-CoA, two metabolites without any direct connections reported, can be a novel gauge for cellular metabolic homeostasis, that can be the beginning of a new paradigm.
In addition, the link between CtBP2 and metabolic transcription factors is also novel, which redefine CtBP2 as a central player in a new metabolic system. The metabolic system regulated by CtBP2 is composed of a lot of novel interactions and idea. We refrained from repeating the word 'novel' in our manuscript, but please let us list our major novel findings here. We believe our manuscript provide a conceptual change supported by a lot of novel findings although it is always hard to assure 'novelty'.

1) CtBP2 interacts with FoxO1 and FoxO1-mediated regulation of gluconeogenesis is
under the control of CtBP2.

4) Negative regulation of CtBP2 by fatty acyl-CoA. Previous papers reported that NADH
activates CtBP2 more efficiently than NAD + but that both NADH and NAD + activate CtBP2. Therefore, the identification of fatty acyl-CoA as a negative regulator of CtBP2 can offer a drastic conceptual change. 5) CtBP2 activity is dramatically altered in liver tissues in physiological and pathological conditions. 6) CtBP2 interacts with SREBP1 and SREBP1-mediated regulation of lipogenesis is under the control of CtBP2. In addition, we identified the intermediary molecule LRH1. 7) Generation of CtBP2 flox mice and the phenotype of liver specific CtBP2 knockout mice: diabetes and hepatic steatosis and liver damage in a specific condition. It is of note that concurrent activation of hepatic gluconeogenesis and lipogenesis in obesity has been a long-lasting mystery in this field. 8) Liver-specific overexpression of CtBP2 ameliorates diabetes and hepatic steatosis. 9) Identification of A201H mutation that facilitates CtBP2 dimer formation.
10) Based on these findings, we proposed that CtBP2 plays a critical role in liver glucose and lipid metabolism and could be targeted to develop future medicine.
11) The balance between NAD(H) and fatty acyl-CoA can be a novel indicator for metabolic homeostasis.
Major Concerns 1) Page7 line 6-7: they mentioned that Ctbp1 is excluded in hepatic glucose metabolism through preliminary screening stage. Since Ctbp1, just like Ctbp2, acts as a transcriptional repressor in many cell types, they should provide the more reliable evidence that Ctbp1 is excluded in this experiment.
Response: As far as we examined, CtBP1 seems to be involved in our metabolic system as well. However, it did not seem to be straightforward (shown below). For example, knockdown of CtBP1 and CtBP2 had synergistic effects on gluconeogenic gene expression in vitro whereas knockdown of CtBP1 and CtBP2 in vivo provided different expression profile. We believe that we need to understand the basis for this discrepancy and the metabolic roles of CtBP1 seems to be more complicated. If the roles of CtBP1 and CtBP2 were just synergistic, we may describe them in a single paper, but we think it would be too broad to have both CtBP isoforms in a single paper. We also added a statement regarding the possible role of CtBP1 in metabolism as a future avenue in our Discussion (page 29, line15-17). In addition, we clearly stated that CtBP2 isoform may have 'direct' involvement rather than CtBP1 isoform in our Results section (page 7, line 7) since our original statement may give an impression that CtBP1 does not have any metabolic effects.
2) In Supplementary Figure 2e, they showed that FoxO1 mutant (deltaPSDL) did not alter the association with Ctbp1. As far as I know, Ctbp isoforms interact with substrate protein in a same manner, more sophisticated experiments may be needed. As shown in Figure 4f, mouse liver contains sufficient amount of both Ctbp1 and Ctbp2. Thus, without Ctbp1, all the data in this manuscript is incomplete to explain the FoxO1-Ctbp axis in hepatic glucose metabolism.
Response: We think that CtBP1 also binds to FoxO1 through its PSDL motif directly and indirectly, and the contribution of indirect binding is strong enough to minimize the effect of deletion of the direct binding. Again, if we demonstrate the roles of CtBP1 as well as CtBP2 in this paper, the amount of data would be tremendous. We believe it would be friendly for readers if we focus on CtBP2 in this paper.
3) In Figure 2a, addition of 50 mM of NADH in cell lysates causes to increase the interaction of FoxO1 with Ctbp2. However, cell lysates have already contained the endogenous NADH (and NAD+), they should check the endogenous amount of NADH and calculate the ratio between endogenous NADH and exogenous NADH. NADH addition in cell lysates is different from the buffering system just like Figure 2b.
Response: At first, we need to discriminate between bound and free NADH/NAD + since CtBP2 would be under the influence of free NADH/NAD + . Since it is really hard to determine free NADH/NAD + concentrations in cell lysates, it is often estimated by indirect measurements. In addition, we are using an exaggerated system where CtBP2 We also showed a more simplified system using recombinant proteins and metabolites ( Fig.7g,i), which would sufficiently convince readers. Figure 3, they mentioned that Ctbp2 is markedly inactivated in the liver of obesity by showing that the dissociation between Ctbp2 and FoxO1 (other transcription factors) in several diet conditions. But these appears to be indirect. I'd like to point out that Response: We challenged those mice with an MCD diet in this manuscript. High fat diet and other obesogenic diet would inactivate CtBP2 and the difference between wild-type and CtBP2-KO would disappear. We also tried a streptozotocin-induced type 1 diabetes model. We did not include this data since it would not add new insights over existing data. If the reviewer requires additional model(s), we can add the data.

4) In
Reviewer #3 (Remarks to the Author): An important aspect of this manuscript is the structural model presented for the binding of palmitoyl CoA and acetyl CoA to CtBP2. The description of the structural modeling is quite superficial; if this reflects the actual structural modeling performed, that would be a cause for concern. The authors state that "palmitoyl-CoA and acetyl-CoA were docked into the NAD+ binding site of human CtBP2" and energy minimized. One would expect that the ADP substructure would fit in similarly for NAD+ and the CoA molecules, which provides a nice starting point. However, looking at the structure, it is not at all clear why the remainder of the CoA structures would continue down into the nicotinamide pocket, given the very different chemical structures, rather than exiting the interior of the protein as could easily occur near the phosphate groups. Based on examining the structures, it does not appear that palmitoyl-CoA would disrupt the dimeric assembly if it exited through this passage, whereas wiggling through the nicotinamide pocket to exit through the substrate pocket might. As the authors are pushing the idea that binding would lead to disruption of the dimeric assembly, how the docking was done and what choices were made becomes crucial. Unfortunately, based on the superficial description given, it is impossible to know if their model is likely or not, given the alternatives. Even if the authors are correct on this hypothesis, the presentation of the resultant models (Fig 2c   and Fig. 6h) is not very illuminating. The authors also do not show a comparison of the packing of the palmitoyl-CoA with the location of the dimeric interface, which would be helpful for understanding the authors hypothesis that binding would disrupt the interface.
It is worth noting that the CtBP dimer interface is very extensive (with buried surface area of over 2500Å2) so it would take substantial interactions from the acyl chain to disrupt it.
Response: We appreciate the Reviewer 3's comments since we would have readers with similar impression upon publication of our original manuscript. The acyl-chain moiety extends out of the pocket, pivots with fixation to the pocket through its CoA moiety and finally locates at the dimerization interface with hydrophobic and electrostatic interactions. We provided another image from a different angle to clarify the location of palmitoyl-CoA in CtBP2 dimeric assembly (Extended Data Fig. 8f). And of note, not only the acyl-chain moiety at the dimerization interface but also CoA moiety play a pivotal role by competing the pocket with NADH.
We further performed the first-principles calculation using fragment molecular orbital (FMO) method to quantify the interaction energies ( Fig.7j and Extended Data Fig. 8g) and this analysis was further validated by a mutation study (Extended Data Fig. 8h). If the Reviewer 3 requires more solid data such as X-ray crystallography or cryo-electron microscopy, that should be reported separately.
A central basis for considering that CtBP acts as a redox sensor is the conclusion from the Goodman group in the early 2000s that the affinity CtBP for NADH is 100 times greater than that of NAD+. (This work was done on CtBP1, so there is no evidence that CtBP2 binds NADH significantly tighter than NAD+ as the authors claim in their manuscript.) As far as I am aware, no lab has been able to repeat/verify the conclusions of the Goodman group. The reported Kd for NADH binding in that work is similar to that reported by other methods of around 100-500nM. However, the reported Kd for NAD+ binding by a more indirect competition measurement of 8-11uM is much higher than any values reported by other groups. In addition to direct measurements of affinity (see, for The interpretation of this manuscript rests heavily on a much higher affinity for NADH compared with NAD+, which is not supported by the preponderance of the data available. Response: We thank the Reviewer 3 raising this controversy. We evaluated the effect of NADH/NAD + ratio in Figure 2f and 2i indirectly using live cultured cells. In most of our studies in this manuscript, we evaluated the effects of NADH without any comparison to those of NAD + . And we also stated the binding of NADH seems to be relatively unstable compared to fatty acyl-CoA in Figure 2b. As mentioned above, one of our major points in this study is that CtBP2 is a fatty acyl-CoA sensor. In fact, we showed inactivation of CtBP2 in two different obese mouse models that cannot be explained by NADH/NAD + ratio (Extended Data Fig.4g,h). Our proposal is that CtBP2 is serving as a sensor in the context of NADH/fatty-acyl-CoA ratio rather than NADH/NAD + ratio, which would solve the controversy raised by the Reviewer 3. We attempted to clarify this by adding statements in our Introduction (page 5, line 18-20), Results (page 23, line12-14) and Discussion (page 24, line 14-18) sections. We completely agree with the existence of this controversy which needs to be solved. There are a lot of possibilities, for instance, purified CtBP protein alone may not be able to distinguish NADH and NAD + and some cellular environments may be required. Indeed, we can observe redox sensitivities in cultured cells (Fig. 2f,2i) that cannot be observed in purified proteins in agreement with the literature provided by the Reviewer 3 (please refer to the data below). Although this question is of great interest for scientists in a broad field as well as for us, we can construct our proposed model in this study without solving this issue.
The dissociation constant that the authors present using microscale thermophoresis between CtBP2 and NADH is two orders of magnitude higher (weaker binding) than reported by other methods. Although this is very briefly addressed in the appendix, I do not find their suggestions for the basis of this discrepancy very convincing, which adds additional concerns to the accuracy of their analysis.
Response: No existing reports evaluated binding kinetics of CtBP2 using microscale thermophoresis. Since there might be technical pitfalls in this experimental system, we additionally examined this aspect by different systems. We attempted to estimate the Kd using the differential sccaning fluorimetry (DSF) assay (Extended Data Fig. 3c in the revised manuscript), where we observed similar Kd values for both NADH and oleoyl-CoA. We also attempted to measure the kinetics using Biacore technology, a surface plasmon resonance (SPR) biosensor system, where Kd for NADH and oleoyl-CoA were 5-50 nM range and 10 mM-10 M range (please refer to the data shown above), respectively. We observed a discrepancy between MST/DSF and SPR in terms of the estimated Kd values. The major difference would be that SPR requires immobilization of recombinant proteins and that MST and DSF provide more natural conditions for molecules to interact. Although we showed data obtained through MST and DSF because of the relatively natural conditions of the experiments and consistency, we don't deny potential difficulties in determination of binding kinetics.
Although every experiment may have some possible limitations, we showed multiple layers of evidence at the levels of protein, cultured cells and in vivo models to support our conclusion. We moved our corresponding discussion into the main text as well.
We as well as other scientists in this field have not performed any unbiased screening experiments to identify the most biologically relevant ligand(s) for CtBP2. Thus, no one knows if NADH is exclusively regulating CtBP2 activity. We also added a statement to imply potential contribution of non-NADH/NAD + ligand(s) in our discussion (page 25, line 19). We believe we can only accumulate evidence through many experimental conditions that will eventually lead us to the truth. We hope our data could be part of that accumulated evidence.