Serotonin signals through a gut-liver axis to regulate hepatic steatosis

Nonalcoholic fatty liver disease (NAFLD) is increasing in worldwide prevalence, closely tracking the obesity epidemic, but specific pharmaceutical treatments for NAFLD are lacking. Defining the key molecular pathways underlying the pathogenesis of NAFLD is essential for developing new drugs. Here we demonstrate that inhibition of gut-derived serotonin synthesis ameliorates hepatic steatosis through a reduction in liver serotonin receptor 2A (HTR2A) signaling. Local serotonin concentrations in the portal blood, which can directly travel to and affect the liver, are selectively increased by high-fat diet (HFD) feeding in mice. Both gut-specific Tph1 knockout mice and liver-specific Htr2a knockout mice are resistant to HFD-induced hepatic steatosis, without affecting systemic energy homeostasis. Moreover, selective HTR2A antagonist treatment prevents HFD-induced hepatic steatosis. Thus, the gut TPH1-liver HTR2A axis shows promise as a drug target to ameliorate NAFLD with minimal systemic metabolic effects.

Although the functions of central nervous system serotonin (5-HT) have tended to dominate the research focus, there is now a renewed emphasis on peripheral serotonin as a therapeutic target. This includes the role of serotonin in disorders such as diabetes and obesity as well as in disorders of the gastrointestinal tract, the location where the majority of mammalian 5-HT is produced. In the current study, the authors focus on the role of 5-HT in Nonalcoholic fatty liver disease (NAFLD). Using high-fat diet feeding in mice in combination with a number of experimental approaches (gut-specific Tph1 knockout mice and liver-specific Htr2a knockout mice, selective serotonin receptor 2A  antagonist treatment), the study convincingly demonstrates a role for gut derived serotonin synthesis, via liver 5-HT2RA signalling, in hepatic steatosis.
There is a lot to like about this interesting and topical study. The authors have taken a very comprehensive approach to verify their hypothesis with the data from both knockout animals and pharmacological studies supporting their conclusions. I have the following queries and recommendations.
(1) The initial characterisation of the effect of the high fact diet on the liver serotonergic system relies almost exclusively on hepatic gene expression analysis by qRT-PCR. This would benefit from verification at the protein level.
(2) Although the findings are very interesting conceptually, from a practical and translational perspective there may be an issue with therapeutic targeting of hepatic 5-HT2RA since that receptor also has a role in liver regeneration. This point requires inclusion in a revised discussion.
(3) In the introduction, the authors state that 5-HT has exclusive tissue expression patterns, and is produced by TPH1 in peripheral non-neuronal tissues and TPH2 in central neuronal tissues. However, enteric neurons also produce 5-HT and this should also be noted.
(4) The authors indicate that neither Tph1 nor Tph2 were expressed in the liver suggesting that the hepatocyte is not likely to produce 5-HT. This contradicts previous reports of hepatic 5-HT production including evidence of tph gene and protein expression as well as enzyme activity (Valdés-Fuentes et al 2015 https://physoc.onlinelibrary.wiley.com/doi/abs/10.14814/phy2.1238 9).
(5) The study includes measures of 5-HT levels in portal blood and peripheral blood of both mice and human subjects. Can the authors provide more information on the sample processing for this plasma 5-HT analysis since the use of platelet poor plasma, generated by higher than normal centrifugation, is usually required for such studies.
(6) In fig 1b, portal blood 5-HT levels are set at 100%. Why are human portal blood serotonin levels expressed as % when the rodent ones expressed as absolute levels? For consistency, please provide absolute levels of 5-HT in all figures.
(7) There is a lot of variation in the group n numbers for the various measures (e.g. 3 or 4 for the 5-HT measurements in figure 1 but 7-10 for the hepatic triclyceride levels). Why are 5-HT levels not provided for all animals?
Reviewer #2 (Remarks to the Author): The authors have recently shown that short term treatment of TPH inhibitors prevents development of hepatic steatosis in mice fed a high carbohydrate diet. In this manuscript, the authors extend these studies and explore the mechanism and function of 5-HT signals regarding liver energy metabolism in vivo. Results show link gut-derived 5-HT, through direct actions on liver, to the pathogenesis of hepatic steatosis. In order to investigate the functional role for gutderived 5-HT the authors generated gut-specific Tph1 knockout (KO) mice and induced hepatic steatosis with 8 weeks of HFD diet. Hepatic lipid droplet accumulation, NAFLD activity score (NAS), and hepatic TG levels were dramatically reduced in the liver of HFD-fed Tph1 GKO mice. In order to identify the relevant metabolic pathways that prevent hepatic steatosis in Tph1 GKO mice, the authors performed gene expression analyses for an extensive array of metabolic markers. In addition, hepatic Htr2a loss-of-function mice were protected against HFD-induced hepatic steatosis. Therefore, the authors evaluated selective HTR2A antagonism as a novel therapeutic strategy for NAFLD. Results presented indicate that selective HTR2A antagonist may prevent hepatic steatosis in HFD-fed mice likely through common mechanisms shared with Tph1 GKO and Htr2a LKO mice. Results highlight the functional importance of hepatic HTR2A signalling in the progression of NAFLD. Inactivation of HTR2A signalling in the liver using Htr2a LKO, attenuated hepatic steatosis independent of systemic energy homeostasis. 2) Serotonin levels are higher in portal as compared to peripheral venous blood.
3) High-fat diet increases 1) and 2), as expected 4) Tph1G KO results in improved NAS and hepatic triglyceride levels. The authors should include the data showing (as indicated in the text) that circulating serotonin is reduced. This seems to be an important finding.    Sarpogrelate acts as expected as a HTR2A agonist. This compound has been tested vs. aspirin in secondary prevention of cerebral infarction. A published subgroup analysis did not find any differences in individuals without T2DM, dyslipidaemia which indirectly would have supported the role of HTR2A agonists. Verification of the results in humans is therefore necessary.
Additional comments: 1. The data presented seems rather preliminary and it is necessary to further identify detailed molecular mechanisms on how GDS regulates hepatic lipid disposal in the liver and how inhibition of HTR2A signalling reduces hepatic TG accumulation. 2. As the authors discuss, there may be more various functions of 5-HT in different tissues. Therefore, a more detailed study of the different roles of 5-HT using tissue specific KO strategy is necessary to further our understanding of the function of the neurotransmitter. 3. Do the mice display differences in eating and behaviour? Studies in metabolic cages are necessary. 4. Hepatic lipids should be characterized using lipidomics technology 5. Suppl figure 1 c; significance seems to be driven by one single data point Reviewer #3 (Remarks to the Author): Reviewer's comment: This paper by Choi et al presents studies identifying a role for gut-derived serotonin synthesis in de novo lipogenesis and hepatic steatosis via reduction in liver serotonin receptor 2A (HTR2A) signalling. They showed increased serotonin concentrations in the portal blood of six liver living donor subjects as well as in mice on a high-fat diet. Both gut-specific Tph1 knockout mice and liver-specific Htr2a knockout mice were resistant to short-term HFD-induced hepatic steatosis without having any effects on systemic lipid, glucose, and energy metabolism. Finally, the use of a selective HTR2A antagonist treatment mimicked the protective effects found on the KO on HFDinduced hepatic steatosis. They concluded that gut TPH1-liver HTR2A axis is a promising novel drug target for NAFLD. The manuscript is interesting and of potential significance. There are number of concerns with the current dataset.
Major issues: 1. The human data presented in figures 1 is very hard to interpret. These are living donors were the presence of steatosis/NAFLD is strictly ruled out by extensive work up including liver biopsy that portal blood was obtained from at the time of transplantation. The authors showed correlation with serotonin levels in portal blood with various markers of steatosis and NAFLD progression?? 2. In both humans with NAFLD and mice on HFD the main mechanism of steatosis is increased uptake of free fatty acids from circulation mainly via increased lipolysis in the context of systemic and adipose tissue insulin resistance. The current study showed that gut-specific Tph1 knockout mice and liver-specific Htr2a knockout mice on a short-term HFD for eight weeks have no effects on systemic lipid, glucose, and energy metabolism. These results strongly suggest that the effects on steatosis noted are likely to be transient and disappeared with more prolonged feeding. In fact, these diets are typically given 16 to 24 weeks in NAFLD related research. Thus at least a later time point (16 or 24 wks) should be included to better understand the impact of GDS on NAFLD. 3. The suggested effects on the novo lipogenesis in the liver are not fully supported with only limited data on gene expression of various enzymes involved in lipid metabolism provided 4. The studies using sarpogrelate as a therapeutic strategy for NAFLD does not take into account the effects on platelet aggregation and on soluble adhesion molecules induce by this drug that may have a significant impact on the liver phenotype observed. Studies using RNA-based therapy to selectively disrupt the GDS would be warranted.

Response to Reviewers
We thank the reviewers for their careful review of our manuscript. We found the comments and suggestions very helpful. We have tried to comply with all of the comments of the reviewers and highlighted in red color for changes in the revised manuscript. Overall, reviewers raised concerns on the lack of mechanistic insight in the liver and the specificity of 5-HT signaling through gut-liver axis. To address these concerns, we performed RNA-seq analysis in the liver of Htr2a LKO mice and analyzed liver specific Htr2b KO mice. We also responded to the reviewer's other concerns and outlined those changes below.

Reviewer #1
Although the functions of central nervous system serotonin (5-HT) have tended to dominate the research focus, there is now a renewed emphasis on peripheral serotonin as a therapeutic target. This includes the role of serotonin in disorders such as diabetes and obesity as well as in disorders of the gastrointestinal tract, the location where the majority of mammalian 5-HT is produced. In the current study, the authors focus on the role of 5-HT in Nonalcoholic fatty liver disease (NAFLD). Using high-fat diet feeding in mice in combination with a number of experimental approaches (gut-specific Tph1 knockout mice and liver-specific Htr2a knockout mice, selective serotonin receptor 2A  antagonist treatment), the study convincingly demonstrates a role for gut derived serotonin synthesis, via liver 5-HT2RA signalling, in hepatic steatosis.
There is a lot to like about this interesting and topical study. The authors have taken a very comprehensive approach to verify their hypothesis with the data from both knockout animals and pharmacological studies supporting their conclusions. I have the following queries and recommendations.

Comment 1) The initial characterisation of the effect of the high fat diet on the liver serotonergic system relies almost exclusively on hepatic gene expression analysis by qRT-PCR. This would benefit from verification at the protein level
We agree that verification at protein level is beneficial. However, it is difficult to detect 5-HT receptors (HTRs) at protein level because their expression levels are generally very low, usually less than 1 RPKM in RNA-seq. In addition, we haven't had a good antibody for HTR. HTRs are structurally similar with each other and they all have 7 transmembrane domains, which make it difficult to make specific antibody for each HTR. We have been trying to detect HTR by western blot or immunostaining for more than 10 years without any success. This is why lots of papers could not show HTR expression by western blot. Although we did not show HTR expression at protein level, hepatic Htr2a expression was well documented by us and others (Cell Metabolism 16, 588-600, 2012). We also showed the disappearance of Htr2a expression in Htr2a LKO mice along with clear phenotype.

Comment 2) Although the findings are very interesting conceptually, from a practical and translational perspective there may be an issue with therapeutic targeting of hepatic 5-HT2RA since that receptor also has a role in liver regeneration. This point requires inclusion in a revised discussion.
We thank reviewer's suggestion. 5-HT is known to play an important role in liver regeneration after acute liver injury, such as hepatectomy, through hepatic HTR2A and HTR2B (Science 312, 104-107, 2006). Thus, HTR2A antagonists need to be carefully used in case patients have acute or chronic liver injury.
As reviewer's suggestion, we revised DISCUSSION as follows (Page 11, Line 4-7): A possible concern regarding the use of sarpogrelate for anti-NAFLD treatment is its potential adverse effects on hepatic regeneration after acute liver injury. However, as sarpogrelate shows rare hepatotoxicity in real world practice, this concern might not be a problem to use it as an anti-NAFLD drug.

Comment 3) In the introduction, the authors state that 5-HT has exclusive tissue expression patterns, and is produced by TPH1 in peripheral nonneuronal tissues and TPH2 in central neuronal tissues. However, enteric neurons also produce 5-HT and this should also be noted.
Yes, enteric neurons express TPH2 and produce 5-HT. We revised in manuscript as follows (Page 3, Line 24 ~ Page 4, Line 1-2): The two distinct isoforms of TPH show mutually exclusive tissue expression patterns, TPH1 in peripheral non-neuronal tissues and TPH2 in neurons of the central and enteric nervous system.

Comment 4) The authors indicate that neither Tph1 nor Tph2 were expressed in the liver suggesting that the hepatocyte is not likely to produce 5-HT. This contradicts previous reports of hepatic 5-HT production including evidence of tph gene and protein expression as well as enzyme activity (Valdés-Fuentes et al 2015).
As the reviewer indicated, Valdés-Fuentes et al showed that Tph1 is expressed in rat liver (Physiol Rep 3, e12389). However, we could not detect Tph1 expression in the liver of C57BL6/J mice. Izikki et. al. also reported that Tph1 is not expressed in mouse liver (Am J Physiol Lung Cell Mol Physiol 293, L1045-105, 2007). Furthermore, NCBI gene expression database also shows no expression of Tph1 in the mouse and human liver. We agree that this is somewhat controversial but it could be due to the difference between species.

Comment 5) The study includes measures of 5-HT levels in portal blood and peripheral blood of both mice and human subjects. Can the authors provide more information on the sample processing for this plasma 5-HT analysis since the use of platelet poor plasma, generated by higher than normal centrifugation, is usually required for such studies.
We measured 5-HT levels in platelet poor plasma (PPP) both in mice and human subjects. To obtain PPP, we centrifuged twice to ensure no platelet contamination of supernatant; initially at 2500 rpm for 10 minutes and then at 4000 rpm for a further 10 minutes.
We revised METHODS as follows (Page 14, Line 6-9): To measure mouse platelet poor plasma (PPP) serotonin, blood samples were collected by retro-orbital bleeding or sampling from the portal vein. To acquire a PPP, we centrifuged twice to ensure no platelet contamination of supernatant; initially at 2500 rpm for 10 minutes and then 4000 rpm for further 10 minutes. (Page 16, Line 15-16): Human PPP serotonin levels were measured using ClinRep high-performance liquid chromatography kit (Recipe, Munich, Germany) at GreenCross LabCell (Yongin, Korea).

Comment 6) In fig 1b, portal blood 5-HT levels are set at 100%. Why are human portal blood serotonin levels expressed as % when the rodent ones expressed as absolute levels? For consistency, please provide absolute levels of 5-HT in all figures.
We showed absolute levels in Supplementary Fig. 1a and converted this value to % in Supplementary Fig. 1b and Fig. 1b.

Comment 7) There is a lot of variation in the group n numbers for the various measures (e.g. 3 or 4 for the 5-HT measurements in figure 1 but 7-10 for the hepatic triclyceride levels). Why are 5-HT levels not provided for all animals?
These are different sets of experiment. Figure 1c, 1d, 1e are results from standard chow diet (SCD) or high fat diet (HFD) fed C57BL6/J mice. However, Figure. 1f, 1g, 1h are results from SCD or HFD fed wild type littermates (Tph1 flox/flox ) and Tph1 GKO (Villin-Cre +/-; Tph1 flox/flox ) mice.

Reviewer #2
The authors have recently shown that short term treatment of TPH inhibitors prevents development of hepatic steatosis in mice fed a high carbohydrate diet. In this manuscript, the authors extend these studies and explore the mechanism and function of 5-HT signals regarding liver energy metabolism in vivo. Results show link gut-derived 5-HT, through direct actions on liver, to the pathogenesis of hepatic steatosis. In order to investigate the functional role for gut-derived 5-HT the authors generated gut-specific Tph1 knockout (KO) mice and induced hepatic steatosis with 8 weeks of HFD diet. Hepatic lipid droplet accumulation, NAFLD activity score (NAS), and hepatic TG levels were dramatically reduced in the liver of HFD-fed Tph1 GKO mice. In order to identify the relevant metabolic pathways that prevent hepatic steatosis in Tph1 GKO mice, the authors performed gene expression analyses for an extensive array of metabolic markers. In addition, hepatic Htr2a loss-of-function mice were protected against HFD-induced hepatic steatosis. Therefore, the authors evaluated selective HTR2A antagonism as a novel therapeutic strategy for NAFLD. Results presented indicate that selective HTR2A antagonist may prevent hepatic steatosis in HFD-fed mice likely through common mechanisms shared with Tph1 GKO and Htr2a LKO mice. Results highlight the functional importance of hepatic HTR2A signalling in the progression of NAFLD. Inactivation of HTR2A signalling in the liver using Htr2a LKO, attenuated hepatic steatosis independent of systemic energy homeostasis. 2) Serotonin levels are higher in portal as compared to peripheral venous blood.

3) High-fat diet increases 1) and 2), as expected 4) Tph1G KO results in improved NAS and hepatic triglyceride levels. The authors should include the data showing (as indicated in the text) that circulating serotonin is
reduced. This seems to be an important finding. Fig. 2  5) Tph1G KO does not affect systemic energy homeostasis. Fig. 3 6) Tph1 GKO in general express less genes involved in hepatic lipogenesis. This seems to be expected if HTR2A signalling is reduced. Fig. 4  7) Confirms the role of HTR2A. Fig. 5 Sarpogrelate acts as expected as a HTR2A agonist. This compound has been tested vs. aspirin in secondary prevention of cerebral infarction. A published subgroup analysis did not find any differences in individuals without T2DM, dyslipidaemia which indirectly would have supported the role of HTR2A agonists. Verification of the results in humans is therefore necessary. Fig. 1. 4

) Tph1G KO results in improved NAS and hepatic triglyceride levels. The authors should include the data showing (as indicated in the text) that circulating serotonin is reduced. This seems to be an important finding.
Circulating 5-HT levels in Tph1 GKO mice was already reported that plasma 5-HT level was about 50% decreased in Tph1 GKO mice compared to WT littermates (Cell Metab 16, 588-600, 2012).

Comment on Fig. 5) Sarpogrelate acts as expected as a HTR2A agonist. This compound has been tested vs. aspirin in secondary prevention of cerebral infarction. A published subgroup analysis did not find any differences in individuals without T2DM, dyslipidaemia which indirectly would have supported the role of HTR2A agonists. Verification of the results in humans is therefore necessary.
We agree that further verification of our results in human is necessary. The published human data show that sarpogrelate does not affect the metabolic profiles in human subjects. However, those papers did not intend to test the effect of sarporgrelate on fatty liver. Those papers were to test its anti-platelet effects in human. Actually, sarpogrelate or other HTR2A antagonist has never been used to treat metabolic diseases such as diabetes, obesity, and NAFLD. As such, those data need to be carefully interpreted and do not mean that sarpogrelate does not have efficacy in NAFLD. Rather, sarpogrelate is not likely to affect the metabolic profiles in standard condition. We also showed that Htr2a LKO and Tph1 GKO mice are metabolically comparable to wild type littermates when they were fed with standard chow diet (SCD). These data match well with the published human data. Thus, we need to further confirm the effects of sarpogrelate on human NAFLD, as we have confirmed in mice. In this manuscript, we just provided some human data to suggest the possible importance of our study in human NAFLD. Since human study will take a lot more time and efforts, we decided to share our data with people in the field and motivate to test this model in human.

Comment 1) The data presented seems rather preliminary and it is necessary to further identify detailed molecular mechanisms on how GDS regulates hepatic lipid disposal in the liver and how inhibition of HTR2A signaling reduces hepatic TG accumulation.
We appreciate reviewer's constructive suggestion. Since we did not show detailed downstream molecular mechanism of HTR2A in the liver, our data presented in this manuscript may look a little preliminary. However, we showed the role of gut derived 5-HT on hepatic lipid accumulation using two different tissue specific KO mouse models. We furthermore confirmed this phenotype using pharmacological inhibition of HTR2A. Thus, we have shown comprehensively that inhibition of 5-HT signaling through gut-liver axis can inhibit the initiation/progression of NAFLD. We also provide molecular mechanism showing the changes in hepatic gene expressions. Our gene expression data indicate that de novo lipogenesis and TG synthesis pathways are downregulated in Htr2a LKO and Tph1 GKO mice. In addition, we showed decreased expression of SREBP-1c and PPARγ which can explain the downregulation of gene expressions of de novo lipogenesis and TG synthesis. Taken together, we already put large amount of data in this manuscript and showed our results comprehensively at organismal and cellular levels. We think that more detailed downstream molecular mechanism of HTR2A is beyond the focus of this manuscript.
Instead, we performed RNA-seq analysis in HFD-fed Htr2a LKO mice and WT littermates to provide more comprehensive data on the changes in global gene expressions. The results showed that gene sets which contribute to inflammation and fibrosis in nonalcoholic steatohepatitis (NASH) pathogenesis as well as gene sets that contribute to steatosis were significantly downregulated in Htr2a LKO mice.
We added these data and revised RESULTS as follows (Page 8, Line 18 ~ Page 9, Line 4): To further explore the molecular pathways underlying the decreased hepatic lipid accumulation in HFD-fed Htr2a LKO mice, we performed RNA-seq and profiled the liver transcriptomes of HFD-fed Htr2a LKO mice and WT littermates. We analyzed gene ontology (GO) gene sets by gene set enrichment analysis (GSEA). Among 5917 GO gene sets, 3614 gene sets with sufficient number of matched genes were analyzed and 617 gene sets and 27 gene sets were identified to be significantly enriched in WT littermates and Htr2a LKO mice, respectively (Supplementary Table 1). Interestingly, gene sets which contribute to inflammation and fibrosis in nonalcoholic steatohepatitis (NASH) pathogenesis as well as gene sets that contribute to steatosis were significantly less enriched in Htr2a LKO mice (Fig. 5a-c). qRT-PCR analysis further confirmed that proinflammatory and profibrogenic genes were downregulated in the liver of HFD-fed Htr2a LKO mice (Fig.  5d). These results suggest that inhibiting 5-HT signaling through hepatic HTR2A can inhibit the progression of NAFLD.

Comment 2) As the authors discuss, there may be more various functions of 5-HT in different tissues. Therefore, a more detailed study of the different roles of 5-HT using tissue specific KO strategy is necessary to further our understanding of the function of the neurotransmitter.
We agree the reviewer's opinion. Since we showed that gut derived 5-HT does not affect systemic energy homeostasis, we need to study more on the roles of 5-HT in other tissues with tissue specific KO strategy. Actually, we have been analyzing the phenotypes of adipocyte-specific Tph1 KO (Adiponectin-Cre +/-; Tph1 flox/flox , Tph1 FKO) mice. The key phenotypes of Tph1 FKO mice, decreased body weight gain by increased energy expenditure, are similar with the phenotypes of adipocyte-specific inducible Tph1 KO (aP2-CreERT2 +/-; Tph1 flox/flox , Tph1 AFKO) mice, as we published previously (Nat Commun 6, 6794, 2015). Since adipocytederived 5-HT has distinct physiological roles compared to gut derived 5-HT, we are preparing a separate manuscript for these results. As such, we think studies on Tph1 KO in other tissues will provide different functions of peripheral 5-HT in different tissues and those are beyond the scope of this manuscript.

Comment 3) Do the mice display differences in eating and behaviour? Studies in metabolic cages are necessary.
We share this concern with the reviewer. However, it is already documented that inhibiting peripheral 5-HT synthesis genetically or chemically does not alter the eating behavior (Nat Med 21, 166-172, 2015;Nat Commun 6, 6794, 2015). All of the mice we used in this paper showed similar weight gain upon high fat diet indicating that Tph1 GKO mice and Htr2a LKO mice are not likely to affect eating behavior.

Comment 4) Hepatic lipids should be characterized using lipidomics technology
We thank reviewer's suggestion. We performed lipidomic analysis with the liver of HFD-fed Htr2a LKO mice and wild type littermates. Htr2a LKO mice exhibited significantly reduced levels of a variety of triglyceride and diglyceride species compared to the wild type littermates (Additional Fig 1 with legend), which is consistent with the liver histology and total triglyceride measurement data ( Fig. 4d-f). On the other hand, polar lipid species abundant in membrane structures such as phospholipids and ceramides were similar or higher in the Htr2a LKO mice, indicating specific reduction of triglycerides and diglycerides, the hallmark of fatty liver. Thus, the lipidomics data support the notion that liver-specific deletion of Htr2a is protective against HFD-induced hepatic steatosis.