The gut microbiota promotes hepatic fatty acid desaturation and elongation in mice

Interactions between the gut microbial ecosystem and host lipid homeostasis are highly relevant to host physiology and metabolic diseases. We present a comprehensive multi-omics view of the effect of intestinal microbial colonization on hepatic lipid metabolism, integrating transcriptomic, proteomic, phosphoproteomic, and lipidomic analyses of liver and plasma samples from germfree and specific pathogen-free mice. Microbes induce monounsaturated fatty acid generation by stearoyl-CoA desaturase 1 and polyunsaturated fatty acid elongation by fatty acid elongase 5, leading to significant alterations in glycerophospholipid acyl-chain profiles. A composite classification score calculated from the observed alterations in fatty acid profiles in germfree mice clearly differentiates antibiotic-treated mice from untreated controls with high sensitivity. Mechanistic investigations reveal that acetate originating from gut microbial degradation of dietary fiber serves as precursor for hepatic synthesis of C16 and C18 fatty acids and their related glycerophospholipid species that are also released into the circulation.

1 Point-to-point response to the referees (NCOMMS-17-28020) As some of the points raised by the Reviewers were similar, some answers are replicated. Most work done for the revision of the manuscript was related to Comment 1, which basically answers to most of the reviewer`s questions.
Our new investigations were added and discussed including a new figure (Fig. 5) and two supplementary figures (Fig. SI3 and SI4). All changes made to the original manuscript are marked in red.
We would really like to thank the reviewers for their helpful comments, which substantially improved this manuscript. We hope that we could provide the additional information and clarifications requested.
Editor/Reviewer I and II: Comment 1: "In particular, it is our editorial view that the revised manuscript should include new experiments providing additional mechanistic insight, as recommended by the reviewers." Now, we provide additional mechanistic insights by showing for the first time that SCFA/acetate (FA 2:0) originating from gut microbial degradation of dietary fiber is a precursor for hepatic synthesis of C16 and C18 fatty acids finally entering the circulation.
Experimental evidence was collected from the following new animal experiments: i.
To show that acetate produced in the gut lumen is used for hepatic FA synthesis, the following stable isotope labelling strategy was applied: Mice were supplied with different concentrations of 13 C-labeled acetate via oral gavage→ The enrichment of 13 C in FA 16:0 of liver and plasma was investigated after mass spec-based analysis. ii.
To confirm that synthesis of FA/MUFA depends on gut microbial SCFA producers and portal vein FA 2:0 levels, first, Bacteroidetes and Firmicutes as major SCFA producers were eliminated with antibiotics. Secondly, the antibiotics were displaced to recover the initial conditions/SCFA levels. FA and PC species were analyzed as for previous experiments, portal vein SCFA levels were quantified by LC-MS/MS. iii.
To demonstrate that the hepatic synthesis of MUFA and PUFA metabolism depend on dietary fiber that is degradable and fermentable by gut microbiota (a) GF and SPF mice were fed a diet containing a non-degradable fiber source/cellulose and (b) SPF mice were fed diets with different contents of degradable fiber.
To ask whether FA 2:0 originating from the gut lumen is a precursor for synthesis of long chain

13
C-Acetate also dose-dependently elevated the fraction of newly synthesized FA 16:0 in the liver 4 h after oral gavage from ~4 to 30 % demonstrating a stimulation of the hepatic FA de novo synthesis by gut-derived acetate ( Figure 5D).
To demonstrate that differences in the lipid profiles of GF and colonized mice ( Figure 2, Figure   SI3), particularly altered MUFA contents, can be associated with microbial FA 2:0 production in the gut, we next manipulated Bacteroidetes and Firmicutes as major SCFA producers. SPF mice received a combination of vancomycin and metronidazole (VM; as describe before) for two days (time-point 2, TP2), before they obtained a regular chow diet without antibiotics for additional two (TP4) or 10 days (TP 14). At TP2+VM gut microbial composition and diversity dropped ( and -diversity; Figure 5F, G, H). Bacteroidetes and Firmicutes were almost completely eliminated ( Figure 5I) and portal vein FA 2:0 concentrations were reduced 2-fold ( Figure 5J). After removal of the antibiotics (TP4 and TP14), the gut microbial ecosystem recovered and baseline FA 2:0 levels (TP0) were reached. Importantly, alterations of liver and plasma MUFA (16:1 n-7, FA 18:1 n-9, 18:1 n-7), MUPC (PC 34:1, PC 36:1) and PUFA (20:3 n-6, 22:6 n-3) levels clearly followed this trend ( Figure 5K, L). These results indicate a direct relation of liver and plasma FA levels to gut microbial SCFA producers and portal blood FA 2:0 levels.
Chow diet used in these experiments comprises polysaccharides from a grain-soybean-based crude fiber extract (5 %) that can be depolymerized and subsequently fermented by gut microbiota to SCFA. To demonstrate that hepatic synthesis of MUFA depends on gut microbial degradation of dietary fiber, GF and SPF mice were fed with an experimental control diet with carbohydrate, fat and protein content comparable to chow, but with 5 % purified cellulose instead of crude fiber. Refined cellulose is practically non-degradable andfermentable by gut microbiota leading to markedly reduced SCFA levels in portal blood (Flint et al., 2012;Slavin et al., 1981;Weitkunat et al., 2015). In contrast to mice fed a chow diet ( Figure 2A, C), plasma and liver FA profiles were not significantly different between GF and SPF mice fed the experimental control diet containing cellulose ( Figure SI4A, B). These results confirm that the observed lipid metabolic differences, particularly the generation of MUFA, in SPF mice depends on a degradable fiber source. In agreement, SPF mice fed with control diet containing 14% of fiber had higher levels of MUFA, but lower contents of PUFA compared to SPF mice fed a standard chow diet with 5% fiber ( Figure SI4C, D).
In summary, these data provide strong evidence that the gut microbiota promotes hepatic FA metabolism by providing a substantial amount of FA 2:0 as precursor for synthesis of C16 and C18 FA." (Results, p.10-11)  "Hepatic ELOVL5 activity is repressed by dietary PUFA, but induced by dietary oleic acid (Ducheix et al., 2017;Wang et al., 2005). Thus, it can be speculated that MUFA generated by SCD1 together with decreased fractions of PUFA lead to increased ELOVL5 activity and enhanced FA 20:3 n-6 levels observed in SPF animals, although this has to be clarified in further studies." (Discussion p.12)  "Application of an in vivo stable isotope labelling strategy in combination with mass spectrometric analysis revealed that FA 2:0 originating from the gut lumen is used as precursor for hepatic synthesis of FA, which are released into circulation, also as complex lipids. This data are supported by significantly elevated serum lipid levels found in human subjects after rectal infusion of SCFA (Wolever et al., 1989). Modulation of Bacteroidetes and Firmicutes contents and potentially the total gut bacterial load (discussed before) by a short-term antibiotic treatment demonstrated a correlation of FA 2:0 levels to liver MUFA and the PUFA fraction. We chose VM, since this antibiotic combination had most pronounced effects on the composite classification score, SCD1 expression, MUFA and Bacteroidetes contents ( Figure 4). For the production of SCFA, it is important that the gut microbiota work as a community (den Besten et al., 2013;el-Khoury et al., 1994). Thus, we conclude that alterations in gut microbial diversity significantly contribute to SCFA levels in our study. At TP2+VM specifically the fraction of Proteobacteria massively rose ( Figure 5I).
Previous studies suggested that Proteobacteria can be associated with FA 3:0 (but not FA 2:0) fitting very well to the observed changes in portal vein FA 3:0 levels ( Figure 5J) (Islam et al., 2011;Liou et al., 2013). At TP4, MUFA and MUPC fractions reached even higher concentrations than at TP0 ( Figure 5K, L) suggesting a massively induced FA synthesis after antibiotics displacement to restore the initial FA and PC profile.
When mice were fed with a diet containing purified cellulose as fiber source, the FA profiles of GF and SPF mice were similar ( Figure S4A, B). Purified cellulose is rarely degradable, fecal and portal vein concentrations of SCFA from mice fed cellulose are ~50% lower than in mice fed a degradable fiber source (den Besten et al., 2013;Weitkunat et al., 2015). When SPF mice were fed a diet containing elevated fractions of degradable fiber, gut microbiotamediated effects were boosted compared to SPF mice fed a regular diet confirming the importance of a degradable fiber source for gut microbiota driven FA synthesis in the liver. It was reported previously that colonic formation of SCFA depends on the amount of dietary fiber ingested (den Besten et al., 2013;Topping and Clifton, 2001)." (Discussion, p.13/14)  "An important precursor for hepatic long-chain FA is FA 2:0 generated from degradable dietary fiber sources by the gut microbiota." (Discussion, p.14)  " Figure

 "Dietary intervention experiments
To study the effects of non-degradable dietary fiber, SPF and GF C57BL/6N mice were fed at the age of 6 weeks a purified control diet with comparable carbohydrate, fat and protein content as chow but with purified cellulose (5%) as fiber source (autoclaved, S5745-E702, Ssniff) for 2 weeks. To study the effects of an altered dietary fiber content, SPF C57BL/6N mice were fed at the age of 6 weeks either a standard chow diet (5% grain-soybean-based crude fiber extract; V1534, Ssniff) or control diet comparable to chow, but with elevated content of degradable fiber (14% grain-soybean-based crude fiber extract; V1574, Ssniff) for 2 weeks." (Methods, p.33/34)" increase to 81 % A at 0.3 min follow by an increase to 78 % A at 2.5 min. Column was cleaned with 100 % B from 2.6 to 3.0 min and re-equilibrated from 3.1 to 4 min with 90 % A.
The column flow was 500 µl at 60°C and 10 µl sample were injected. The method included acetic acid, propionic acid, butyric acid and iso-butyric acid. Butyric acid and iso-butyric acid were separated by LC. For quantification a calibration lines were generated by addition of SCFA to human plasma." (Methods, p.37/38)" Reviewer I Comment 2: "The experimental setting is straightforward, but with limited information e.g. on the mouse study setting, e.g.
(1) was sterility of GF mice regularly checked and how?" The following was added to the manuscript:  "Sterility of GF mice was routinely confirmed by culturing and microscopic observation of feces after Gram-staining. In addition, 16S rRNA gene-targeted PCR of GF cecal content was performed at the end of the study." (Methods, p.33) Comment 3: " (2) if non-identical housing conditions (isolators) were used after weaning, could that also be one factor affecting the results?" The antibiotic experiments confirm that our results are independent of mouse housing ( Figure 4, Figure 5). To further support this we analyzed the hepatic and plasma fatty acid profiles of GF and Oligo-MM 12 mice housed under identical conditions (isolators).
The following was added to the manuscript: "The major issue in the study is that too much is concluded based on cross-sectional comparison of two groups of mice in a relatively small study. The observation that fatty acid desaturation and elongation are upregulated in the livers of SPF mice is well supported by the data, but the study does not provide an insight into the potential mechanism behind it." A couple of new in vivo experiments and analysis have been performed to provide mechanistic insights, please see Comment 1.
Comment 5: "The antibiotic experiments do suggest that different treatments all seem to produce the GF-like profile, but it would be important to see the effect on lipid metabolism after conventionalisation or colonisation with specific strains, and perhaps with respect to genetic/pharmacological targeting of the pathway of interest. The authors did attempt to analyse the microbial composition after antibiotic treatments, but the results did not offer any insight into their effect on lipid metabolism." Data on mice colonized with a defined consortium of 12 strains (Oligo-MM 12 ) were included in our manuscript, please see Comment 3.
Moreover, we elaborated that particularly SCFA producers were affected by the antibiotic treatment ( Figure 4). The current knowledge is that particularly strains from Bacteroidetes and Firmicutes generate SCFA including acetate (Bacteroidetes >Firmicutes). Thus, we eliminated and subsequently recovered Bacteroidetes and Firmicutes contents in a timeline experiment using a combination of vancomycin and metronidazole (VM). We quantified SCFA, FA and PC species at all time points. Our data show that these phyla can be associated with our lipid metabolic findings identified in the comparison between GF and SPF mice. We think that also the gut microbiota richness ( Figure 5F-H) and the total gut bacterial load (Discussion) might play an important role.
The following was added to the manuscript (also listed in Comment 1):  Figure  To the best of our knowledge energy balance parameters were not yet monitored in GF mice adequately/precisely. It would be necessary to design custom build metabolic cages fitting into isolators. As suggested, we focused on energy harvest by degradation and fermentation of dietary fiber to SCFA, we quantified SCFA in portal blood of antibiotic treated mice.
The following was added to the manuscript (also listed in Comment 1):  Figure Figure 5D).
To demonstrate that differences in the lipid profiles of GF and colonized mice (Figure 2, Figure   SI3), particularly altered MUFA contents, can be associated with microbial FA 2:0 production in the gut, we next manipulated Bacteroidetes and Firmicutes as major SCFA producers. SPF mice received a combination of vancomycin and metronidazole (VM; as describe before) for two days (time-point 2, TP2), before they obtained a regular chow diet without antibiotics for additional two (TP4) or 10 days (TP 14). At TP2+VM gut microbial composition and diversity dropped ( and -diversity; Figure 5F, G, H). Bacteroidetes and Firmicutes were almost completely eliminated ( Figure 5I) and portal vein FA 2:0 concentrations were reduced 2-fold ( Figure 5J). After removal of the antibiotics (TP4 and TP14), the gut microbial ecosystem recovered and baseline FA 2:0 levels (TP0) were reached. Importantly, alterations of liver and plasma MUFA (16:1 n-7, FA 18:1 n-9, 18:1 n-7), MUPC (PC 34:1, PC 36:1) and PUFA (20:3 n-6, 22:6 n-3) levels clearly followed this trend ( Figure 5K, L). These results indicate a direct relation of liver and plasma FA levels to gut microbial SCFA producers and portal blood FA 2:0 levels." (Results, p.10-11)  "Application of an in vivo stable isotope labelling strategy in combination with mass spectrometric analysis revealed that FA 2:0 originating from the gut lumen is used as precursor for hepatic synthesis of FA, which are released into circulation, also as complex lipids. This data are supported by significantly elevated serum lipid levels found in human subjects after rectal infusion of SCFA (Wolever et al., 1989). and Bacteroidetes contents (Figure 4). For the production of SCFA, it is important that the gut microbiota work as a community (den Besten et al., 2013;el-Khoury et al., 1994). Thus, we conclude that alterations in gut microbial diversity significantly contribute to SCFA levels in our study. At TP2+VM specifically the fraction of Proteobacteria massively rose ( Figure 5I).
Previous studies suggested that Proteobacteria can be associated with FA 3:0 (but not FA 2:0) fitting very well to the observed changes in portal vein FA 3:0 levels ( Figure  The column flow was 500 µl at 60°C and 10 µl sample were injected. The method included acetic acid, propionic acid, butyric acid and iso-butyric acid. Butyric acid and iso-butyric acid were separated by LC. For quantification a calibration lines were generated by addition of SCFA to human plasma." (Methods, p.37/38)" "Moreover, to clarify direct influence of gut microbes to host hepatic lipid metabolism, authors should examine relationship between host hepatic lipid metabolism and change of gut microbial composition and metabolites on dietary difference (for example; low or high fiber, normal diet or high fat diet) but not between colonized and germ free." We fed SPF mice high fiber diet and a chow diet as control, and quantified total FA. It was already previously shown that the dietary fiber contents correlate with gut microbial fiber metabolizers and SCFA levels (Discussion). Because high fat diet might lead to pathophysiologic metabolic consequences, i.e. obesity, this experimental strategy was not applied.
The following was added to the manuscript (also listed in Comment 1):  Figure SI4 (see last pages of this document)  "In agreement, SPF mice fed with control diet containing 14% of fiber had higher levels of MUFA, but lower contents of PUFA compared to SPF mice fed a standard chow diet with 5% fiber ( Figure SI4C, D)." (Results, p.11)  "When SPF mice were fed a diet containing elevated fractions of degradable fiber, gut microbiota-mediated effects were boosted compared to SPF mice fed a regular diet confirming the importance of a degradable fiber source for gut microbiota driven FA synthesis in the liver.
It was reported previously that colonic formation of SCFA depends on the amount of dietary fiber ingested (den Besten et al., 2013;Topping and Clifton, 2001)." (Discussion, p.14)

 "Dietary intervention experiments
To study the effects of non-degradable dietary fiber, SPF and GF C57BL/6N mice were fed at the age of 6 weeks a purified control diet with comparable carbohydrate, fat and protein content as chow but with purified cellulose (5%) as fiber source (autoclaved, S5745-E702, Ssniff) for 2 weeks. To study the effects of an altered dietary fiber content, SPF C57BL/6N mice were fed at the age of 6 weeks either a standard chow diet (5% grain-soybean-based crude fiber extract; V1534, Ssniff) or control diet comparable to chow, but with elevated content of degradable fiber (14% grain-soybean-based crude fiber extract; V1574, Ssniff) for 2 weeks." (Methods, p.33)" Comment 8: "I think that by integrating these data, authors can find key factors influenced to host hepatic lipid metabolism in gut microbes and metabolites." By integrating the suggested data SCFA/acetate was identified as key factor relevant to explain our lipid metabolic findings, see also Comment 1. The manuscript has substantially improved and all of this reviewer's concerns have been adequately addressed and responded to.
Reviewer #2 (Remarks to the Author): The Authors have adequately responded to Reviewer's comments and concerns. I have no further comments at this point.