The inactivation of DHHC7 in mouse liver promotes diet-induced obesity through a hepatic Prg4–GPR146 axis

Reviewer #1 (Public review):

Summary:

In this study, the authors' aim was to determine whether hepatic palmitoylation is a physiologically relevant regulator of systemic metabolism. The data demonstrate that loss of DHHC7 in hepatocytes disrupts Gαi palmitoylation, enhances cAMP-PKA-CREB signaling, and drives transcriptional upregulation and secretion of Prg4. The KO mice display increased body weight, fat mass, and plasma cholesterol, but at 12 weeks on HFD, do not exhibit insulin resistance. The potential mechanism underlying the metabolic phenotype was examined by assessing adipocyte signaling and by exploring whether Prg4 acts through GPR146. Through this pathway, the authors intend to link DHHC7-dependent palmitoylation to the regulation of hepatokines that exert systemic metabolic effects.

Strengths:

(1) Hepatic palmitoylation in systemic metabolic regulation is largely unexplored. The authors demonstrate the role of DHHC7 in vivo using a successful liver-specific knockout mouse model that causes HFD-dependent obesity without insulin resistance.

(2) Several studies were performed on chow and HFD, as well as male and female mice.

(3) Plasma proteomics identified Prg4 as a circulating factor elevated in KO mice. Prg4 overexpression phenocopied the KO mice.

(4) There is solid mechanistic data supporting the hypothesis that hepatic DHHC7 loss selectively increases Prg4 secretion as a hepatokine.

(5) There is convincing evidence for the DHHC7 mechanism in liver: DHHC7 controls cAMP-PKA-CREB via Gαi palmitoylation. The authors recognize that the palmitoylation change is causative rather than correlated, and this needs to be more fully explored in the future.

(6) Strong in vitro data support that Prg4 acts through adipocyte GPR146 via its SMB domain

Weaknesses:

(1) The assessment of liver and adipose tissue responses to DHH7 loss is insufficient to support claims that it alters systemic lipolysis. In this new mouse model, liver histology is necessary, especially given the cholesterol increase in the KO. As this is a newly established mouse line, common assessments of the liver during HFD feeding would be important for interpreting the phenotype.

(2) The data show DHH7 loss causes adipose tissue dysfunction and alterations in lipid metabolism. Beyond that, I suggest not stating more regarding the phenotype of the DHH7 mice for this work. A thorough analysis would be needed to determine which factor drives the obesity and changes in energy balance in the mice. For example, the KO mice had lower oxygen consumption (but no change in CO2 production, which is also usually similarly altered), suggesting a CNS component could drive obesity. However, since the data are not normalized for lean mass and there is no information about locomotor activity, this analysis is incomplete. RER may be informative if available. A broad conservative description of the KO phenotype would be more accurate since Pgr4 has many paracrine targets and likely has autocrine signaling in the liver.

(3) Most references to lipolysis or lipolysis flux systemically would be inaccurate. To suggest a suppression of lipolysis, serum NEFA would need to be measured, and in vivo or in vitro lipolysis assays performed to test the effect of DHH7 loss or the specificity of PGR4 action on adipocytes in vivo. To demonstrate adipose tissue dysfunction, analysis of lipogenesis markers, canonical markers for insulin sensitivity, and mitochondrial dysfunction should be performed/measured.

(4) Line 179: The experiment was performed in brown adipocytes to show that Prg4 does not affect p-CREB Figure S8 under the heading: "DHHC7 controls hepatic PKA-CREB activity through Gαi palmitoylation to regulate Prg4 transcription." Unless repeated using liver lysate, the conclusions stated in the text throughout the paper should be revised.

(5) It appears that the serum and liver proteomics were only assessed for factors that increased in KO mice? Were proteins that were significantly decreased analyzed?

(6) The beige adipocyte culture method is unclear. The methods do not describe the fat pad used, and the protocol suggests the cells would be differentiated into mature white adipocytes. If they are beige cells, a reference for the method, gene expression, and cell images could support that claim.

(7) The use of tamoxifen can confound adipocyte studies, as it increases beigeing and weight gain even after a brief initiation period. Both groups were treated with Tam, but another way to induce Cre would be ideal.

(8) Evidence for the lack of the glucose phenotype is incomplete. One reason could be due to the IP route of glucose administration, which has a large impact on glucose handling during a GTT. To confirm the absence of a glucose tolerance phenotype, an OGTT should be performed, as it is more physiological. In addition, the mice should be fed for 16 weeks. Prg4 affects immune cells, changing how adipose tissue expands, and 12 weeks of HFD feeding is often not long enough to see the effects of adipose tissue inflammation spilling over into the system.

(9) There may be liver-adipose tissue crosstalk in KO mice, but this was not fully assessed in this study and would be difficult to determine in any setting, given the diverse cell types that are targets of Pdg4. The crosstalk claim is unnecessary to share the basic premises; there is the DHH7 mechanism/phenotype and the Pgr4 mechanism/phenotype, and while there is no Pgr4 adipose direct mechanism, the paper can be successfully reframed.

(10) Although the DHH7 loss on the chow diet did not result in a phenotype, did the Pgr4 increase in the KO mice on chow? This would determine whether either i) the expression of Pgr4 is dependent on HFD/obesity, or ii) circulating Pgr4 has effects only in an HFD condition. The receptors may also change on HFD, especially in adipocytes.

Impact:

This work would significantly contribute to the study of liver metabolism, provided it includes data describing the liver. The role of Pgr4 in adipocytes and other cell types is of substantial value to the field of metabolism. By reframing the paper and conducting some key experiments, its quality and impact can be increased.

Reviewer #2 (Public review):

In the current report, Sun and Colleagues sought to determine the liver-specific role that DHHC7, a DHHC palmitoyltransferase protein, plays in regulating whole-body energy balance and hepatic crosstalk with adipose tissues. The authors generated an inducible, liver-specific DHHC7 knockout mouse to determine how altered palmitoylation in hepatocytes alters hepatokine production/secretion, and in turn, systemic metabolism. The ablation of DHHC7 was found to alter the production of proteoglycan 4 (Prg4), a hepatokine previously linked to metabolic regulation. The authors propose that the change in Prg4 production is mediated by the loss of Gαi palmitoylation, due to DHHC7 ablation, thereby augmenting cAMP-PKA-CREB signaling in hepatocytes, which alleviates the 'brake' on Prg4 production. The authors further propose that Prg4 overexpression leads to excessive binding to GPR146 on adipocytes, which in turn suppresses PKA-mediated HSL activation, promoting impairments in lipolysis, leading to obesity. The report is interesting and generally well-written, but it appears to have some clear gaps in additional data that would aid in interpretation. The addition of confirmatory culture studies would be incredibly helpful for testing the hypotheses being explored. My comments, concerns, and/or suggestions are outlined below in no particular order.

(1) Figures: All data should be presented in dot-boxplot format so the reader knows how many samples were analyzed for each assay and group. n=3 for some assays/experiments is incredibly low, particularly when considering the heterogeneity in responsiveness to HFD, food intake, etc....

(2) Figure 1E-F: It is unclear when the food intake measure was performed. Mice can alter their feeding behavior based on a myriad of environmental and biological cues. It would also be interesting to show food intake data normalized to body mass over time. Mice can counterregulate anorexigenic cues by altering neuropeptide production over time. It is not clear if this is occurring in these mice, but the timing of measuring food intake is important. Additionally, the VO2 measure appears to be presented as being normalized to total body mass, when in fact, it would probably be more accurate to normalize this to lean body mass. Normalizing to total body mass provides a denominator effect due to excessive adiposity, but white fat is not as metabolically active as other high-glucose-consuming tissues. If my memory serves me right, several reports have discussed appropriate normalizations in circumstances such as this.

(3) Figure 1J-N: It is not all that surprising that fasting glucose and/or TGs were found to be similar between groups. It is well-established that mice have an incredible ability to become hyperinsulinemic in an effort to maintain euglycemia and lipid metabolism dynamics. A few relatively easy assays can be performed to glean better insights into the metabolic status of the authors' model. First, fasting insulin concentrations will be incredibly helpful. Secondly, if the authors want to tease out which adipose depot is most adversely affected by ablation, they could take an additional set of CON and KO mice, fast them for 5-6 hours, provide a bolus injection of insulin (similar to that provided during an insulin tolerance test), and then quickly harvest the animals ~15 minutes after insulin injections; followed by evaluating AKT phosphorylation. This will really tell them if these issues have impairments in insulin signaling. The gold-standard approach would be to perform a hyperinsulinemic-euglyemic clamp in the CON and KO mice. I now see GTT and ITT data, but the aforementioned assays could help provide insight.

(4) Figure 3A: This looks overexposed to me.

(5) Figures 3-4: It appears that several of these assays could be complemented with culture-based models, which would almost certainly be cleaner. The conditioned media could then be used from hepatocyte cultures to treat differentiated adipocytes.

(6) Figure 4: It is unclear how to interpret the phospho-HSL data because the fasting state can affect this readout. It needs to be made clear how the harvest was done. Moreover, insulin and glucagon were never measured, and these hormones have a significant influence over HSL activity. I suspect the KO mice have established hyperinsulinemia, which would likely affect HSL activity. This provides an example of why performing some of these experiments in a dish would make for cleaner outcomes that are easier to interpret.

Reviewer #3 (Public review):

Summary:

In the current manuscript, Sun et al aimed to determine the metabolic function of hepatocyte DHHC7, one of the key enzymes in protein palmitoylation. They generated inducible liver-specific Dhhc7 knockout mice and discovered that Dhhc7-LKO mice are more prone to gain weight and develop adipose expansion and obesity. Via unbiased proteomic analysis, they identified PRG4 as one of the top secreted factors in the liver of Dhhc7-LKO mice. Hepatic overexpression of PRG4 recapitulates the obesity phenotype observed in Dhh7-LKO mice. At the mechanistic level, PRG4, once secreted from the liver, can bind to GPR146 on adipocytes and inhibit PKA-HSL signaling and lipolysis. Taken together, their findings suggest a novel pathway by which the liver communicates with adipose tissue and impacts systemic metabolism.

Strengths:

(1) The systemic metabolic homeostasis depends on coordination among metabolically active tissues. Thus, active communication between the liver and adipose tissue when facing nutritional challenges (such as high-fat diet feeding) is crucial for achieving metabolic health. The concept that the liver can communicate with adipose tissue and impact the lipolysis process via secreted hepatokines is quite significant but remains poorly understood.

(2) Hepatocyte Dhhc7 knockout mice developed a significant obesity phenotype, which is associated with adipose expansion.

(3) Unbiased proteomic analysis identified PRG4 as one of the top secreted factors in the liver of Dhh7-LKO mice. Hepatic overexpression of PRG4 recapitulates the obesity phenotype observed in Dhh7-LKO mice.

(4) In vitro cell-based assay showed that PRG4 can bind to adipocyte GPR146, inhibit PKA-mediated HSL phosphorylation, and subsequently, the lipolysis process.

Weaknesses:

(1) Lack of a causal-effect study to generate evidence directly linking hepatocyte DHH7 and PRG4 in driving adipose expansion and obesity upon HFD feeding.

(2) Lack of direct evidence to support that PRG4 inhibits adipocyte lipolysis via GPR146. A functional assay demonstrating adipocyte lipolysis is required.

(3) The conclusion is largely based on the correlation evidence.

Author response:

Public reviews:

Reviewer #1 (Public review):

Weaknesses:

(1) The assessment of liver and adipose tissue responses to DHH7 loss is insufficient to support claims that it alters systemic lipolysis. In this new mouse model, liver histology is necessary, especially given the cholesterol increase in the KO. As this is a newly established mouse line, common assessments of the liver during HFD feeding would be important for interpreting the phenotype.

We will add the data of the liver histology in the revised version.

(2) The data show DHH7 loss causes adipose tissue dysfunction and alterations in lipid metabolism. Beyond that, I suggest not stating more regarding the phenotype of the DHH7 mice for this work. A thorough analysis would be needed to determine which factor drives the obesity and changes in energy balance in the mice. For example, the KO mice had lower oxygen consumption (but no change in CO2 production, which is also usually similarly altered), suggesting a CNS component could drive obesity. However, since the data are not normalized for lean mass and there is no information about locomotor activity, this analysis is incomplete. RER may be informative if available. A broad conservative description of the KO phenotype would be more accurate since Pgr4 has many paracrine targets and likely has autocrine signaling in the liver.

We will add the data of CO2 production, locomotor activity and RER in the revised version.

(3) Most references to lipolysis or lipolysis flux systemically would be inaccurate. To suggest a suppression of lipolysis, serum NEFA would need to be measured, and in vivo or in vitro lipolysis assays performed to test the effect of DHH7 loss or the specificity of PGR4 action on adipocytes in vivo. To demonstrate adipose tissue dysfunction, analysis of lipogenesis markers, canonical markers for insulin sensitivity, and mitochondrial dysfunction should be performed/measured.

We will measure the serum NEFA to test the effect of DHHC7. We will analyze the lipogenesis markers, canonical markers for insulin sensitivity, and mitochondrial dysfunction.

(4) Line 179: The experiment was performed in brown adipocytes to show that Prg4 does not affect p-CREB Figure S8 under the heading: "DHHC7 controls hepatic PKA-CREB activity through Gαi palmitoylation to regulate Prg4 transcription." Unless repeated using liver lysate, the conclusions stated in the text throughout the paper should be revised.

The figure S8 is to demonstrate that Prg4 has no impact on forskolin induced CREB phosphorylation at Ser133, and provide the evidence that the prg4 acts on the upstream of adenylyl cyclase. We will revise the description.

(5) It appears that the serum and liver proteomics were only assessed for factors that increased in KO mice? Were proteins that were significantly decreased analyzed?

We are analyzing the decreased proteins in the following project.

(6) The beige adipocyte culture method is unclear. The methods do not describe the fat pad used, and the protocol suggests the cells would be differentiated into mature white adipocytes. If they are beige cells, a reference for the method, gene expression, and cell images could support that claim.

We will add a reference for the method, gene expression, asn cell images.

(7) The use of tamoxifen can confound adipocyte studies, as it increases beigeing and weight gain even after a brief initiation period. Both groups were treated with Tam, but another way to induce Cre would be ideal.

We will use the Doxycycline-inducible systems in the future.

(8) Evidence for the lack of the glucose phenotype is incomplete. One reason could be due to the IP route of glucose administration, which has a large impact on glucose handling during a GTT. To confirm the absence of a glucose tolerance phenotype, an OGTT should be performed, as it is more physiological. In addition, the mice should be fed for 16 weeks. Prg4 affects immune cells, changing how adipose tissue expands, and 12 weeks of HFD feeding is often not long enough to see the effects of adipose tissue inflammation spilling over into the system.

We will perform the OGTT and feed the mice for 16 weeks in the future.

(9) There may be liver-adipose tissue crosstalk in KO mice, but this was not fully assessed in this study and would be difficult to determine in any setting, given the diverse cell types that are targets of Pdg4. The crosstalk claim is unnecessary to share the basic premises; there is the DHH7 mechanism/phenotype and the Pgr4 mechanism/phenotype, and while there is no Pgr4 adipose direct mechanism, the paper can be successfully reframed.

We will reframe the paper.

(10) Although the DHH7 loss on the chow diet did not result in a phenotype, did the Pgr4 increase in the KO mice on chow? This would determine whether either i) the expression of Pgr4 is dependent on HFD/obesity, or ii) circulating Pgr4 has effects only in an HFD condition. The receptors may also change on HFD, especially in adipocytes.

We will test the Prg4 in the KO mice on chow diet.

Reviewer #2 (Public review):

(1) Figures: All data should be presented in dot-boxplot format so the reader knows how many samples were analyzed for each assay and group. n=3 for some assays/experiments is incredibly low, particularly when considering the heterogeneity in responsiveness to HFD, food intake, etc.

We will present the data in dot-boxplot format.

(2) Figure 1E-F: It is unclear when the food intake measure was performed. Mice can alter their feeding behavior based on a myriad of environmental and biological cues. It would also be interesting to show food intake data normalized to body mass over time. Mice can counterregulate anorexigenic cues by altering neuropeptide production over time. It is not clear if this is occurring in these mice, but the timing of measuring food intake is important. Additionally, the VO2 measure appears to be presented as being normalized to total body mass, when in fact, it would probably be more accurate to normalize this to lean body mass. Normalizing to total body mass provides a denominator effect due to excessive adiposity, but white fat is not as metabolically active as other high-glucose-consuming tissues. If my memory serves me right, several reports have discussed appropriate normalizations in circumstances such as this.

We will see how to be more accurate to normalize.

(3) Figure 1J-N: It is not all that surprising that fasting glucose and/or TGs were found to be similar between groups. It is well-established that mice have an incredible ability to become hyperinsulinemic in an effort to maintain euglycemia and lipid metabolism dynamics. A few relatively easy assays can be performed to glean better insights into the metabolic status of the authors' model. First, fasting insulin concentrations will be incredibly helpful. Secondly, if the authors want to tease out which adipose depot is most adversely affected by ablation, they could take an additional set of CON and KO mice, fast them for 5-6 hours, provide a bolus injection of insulin (similar to that provided during an insulin tolerance test), and then quickly harvest the animals ~15 minutes after insulin injections; followed by evaluating AKT phosphorylation. This will really tell them if these issues have impairments in insulin signaling. The gold-standard approach would be to perform a hyperinsulinemic-euglyemic clamp in the CON and KO mice. I now see GTT and ITT data, but the aforementioned assays could help provide insight.

We have the data for evaluating AKT phosphorylation and will add it in the revised version.

(4) Figure 3A: This looks overexposed to me.

We will replace it with short exposed one.

(5) Figures 3-4: It appears that several of these assays could be complemented with culture-based models, which would almost certainly be cleaner. The conditioned media could then be used from hepatocyte cultures to treat differentiated adipocytes.

We will perform the cell culture experiments for Figures 3-4

(6) Figure 4: It is unclear how to interpret the phospho-HSL data because the fasting state can affect this readout. It needs to be made clear how the harvest was done. Moreover, insulin and glucagon were never measured, and these hormones have a significant influence over HSL activity. I suspect the KO mice have established hyperinsulinemia, which would likely affect HSL activity. This provides an example of why performing some of these experiments in a dish would make for cleaner outcomes that are easier to interpret.

We will perform some experiments in cell culture dish.

Reviewer #3 (Public review):

Weaknesses:

(1) Lack of a causal-effect study to generate evidence directly linking hepatocyte DHH7 and PRG4 in driving adipose expansion and obesity upon HFD feeding.

We will perform the causal-effect study to demonstrate the hypothesis.

(2) Lack of direct evidence to support that PRG4 inhibits adipocyte lipolysis via GPR146. A functional assay demonstrating adipocyte lipolysis is required.

We will add the direct evidence in the revised version.

(3) The conclusion is largely based on the correlation evidence.

We will perform the experiment to strengthen the conclusion base on the a causal-effect study.