Desacetyl-α-MSH and α-MSH have sex specific interactions with diet to influence mouse gut morphology, metabolites and microbiota

The melanocortin peptides have an important role in regulating body weight and appetite. Mice that lack the desacetyl-α-MSH and α-MSH peptides (Pomctm1/tm1) develop obesity. This effect is exacerbated by a high fat diet (HFD). However, development of obesity in female Pomctm1/tm1 mice during chronic HFD conditions is not fully accounted for by the increased energy intake. We hypothesized that the protection against chronic HFD-induced obesity imparted by MSH peptides in females is mediated by sex-specific alterations in the gut structure and gut microbiota. We determined that female WT mice had reduced jejunum villus length and increased crypt depth in response to chronic HFD. WT males and Pomctm1/tm1 mice lacked this adaptation to a chronic HFD. Both Pomctm1/tm1 genotype and chronic HFD were significantly associated with gut microbiota composition. Sex-specific associations between Pomctm1/tm1 genotype and gut microbiota were observed in the presence of a chronic HFD. Pomctm1/tm1 females had significantly reduced fecal acetate and propionate concentrations when compared to WT females. We conclude that MSH peptides influence jejunum villus length, crypt depth and the structure of the gut microbiota. These effects favor reduced nutrient absorption and occur in addition to the recognized roles of desacetyl-α-MSH and α-MSH peptides in appetite control.

. Cartoon of the study design. Pomc tm1/tm1 and C57BL/6J mice were assigned to eight groups according to the genotype, diet and sex differences 6 . Diet was stratified according to kilocalories provided by the fat content (LF-fat = 10% of the kilocalories; HF-fat = 45% of kilocalories). Mice were culled at age 23 weeks for tissue collection. H&E: haematoxylin and eosin; PAS: periodic acid Schiff; SCFAs: short chain fatty acids; BCAAs: branched chain amino acids (BCAAs); GC-MS: gas chromatography mass spectrometry.

Scientific Reports
| (2020) 10:18957 | https://doi.org/10.1038/s41598-020-75786-z www.nature.com/scientificreports/ We measured villi length to determine if the trend towards increased small intestine length was associated with an increased absorption surface area. Diet was not associated with changes in villi length in male mice or female Pomc tm1/tm1 mice (Fig. 2C,D). By contrast, female WT mice had significantly shorter jejunum villi in response to a chronic HF diet (Fig. 2D, p = 0.0217).
Neither duodenum nor ileum showed genotype or diet specific differences in villus length or crypt depth, in either sex (Tables S1, S2). Furthermore, no group specific differences were observed in the numbers of goblet cells in the duodenum, jejunum or ileum (Tables S1, S2). These findings are consistent with a role for desacetylα-MSH and α-MSH in intestinal morphological adaptations that limit the nutrient absorptive surface in WT female mice chronically exposed to HF diets. Pomc tm1/tm1 genotype, chronic HF diet and sex were associated with gut microbiota composition. Cecal contents were collected to enable profiling of the microbial communities by 16S rRNA amplicon sequencing. The V3-V4 region of the 16S rRNA gene was amplified and paired-end reads (2 × 250 bp) were sequenced on an Illumina MiSeq. The numbers of quality-filtered sequences per sample ranged between 9899 and 61,808 (Table S4). In total, 5370 OTUs were identified using a 97% sequence similarity threshold. The impact of cage, maternal factor, genotype, diet or sex as single drivers of gut microbiota β-diversity was investigated using Permutational Analysis of Variance (PERMANOVA) (Fig. 3). We identified a significant cage effect in the mice microbiota (Bray-Curtis dissimilarity index, PERMANOVA: R 2 : 0.5284, p < 0.001). Therefore, cage-effect was corrected in all subsequent analyses. Maternal factor (i.e., litter) was not associated with differences in the Bray-Curtis dissimilarity index (PERMANOVA: p = 0.093).
Genotype and diet explained 6.3% and 4.1% of the observed variation in the gut microbiota, respectively (Fig. 3A,B). Sex explained 10.9% of the observed variation in the gut microbiota composition (Fig. 3C).
Differences in gut microbiota between Pomc tm1/tm1 and WT mice were intensified in females on a chronic HF diet. Mice were stratified by diet, sex and genotype, and sequence reads rarefied to enable pairwise comparisons (Table S1A). Diet (HF v LF) was associated with gut microbiota composition in WT (Table 1; Fig. S1A, p = 0.01 and S2A, p = 0.004) and Pomc tm1/tm1 mice (Table 1; Fig. S1B, p = 0.013 and S1B p = 0.024), in both sexes. The Pomc tm1/tm1 genotype was associated with gut microbiota composition in female mice that were fed either LF or HF diets (Table 1; Fig. S1C, p = 0.003 and D, p = 0.001). In male mice, the Pomc tm1/tm1 genotype was significantly associated with gut microbiota only when the mice were fed a LF diet (Table 1; Fig. S1C, p = 0.001). The genotype effect in male mice was of borderline significance on the HF diet (Table 1; Fig. S1D, p = 0.045).

Figure 2.
Small intestine and jejunum villus in male and female WT and Pomc tm1/tm1 mice fed on a LF or a HF diet. The small intestine and jejunum villus were measured in male (A, C) and female (B, D) mice fed on a LF or a HF diet for 23 weeks post weaning. Data are mean ± S.E.M. Significant differences were determined either using two-way ANOVA and Tukey's post-hoc test or student's t test (female jejunum villus length). *p < 0.05; WT: Wild-type; LF: Low-fat; HF: High-fat.

Scientific Reports
| (2020) 10:18957 | https://doi.org/10.1038/s41598-020-75786-z www.nature.com/scientificreports/ Interactions between diet and genotype explained less microbial variation in Pomc tm1/tm1 mice compared to WT mice (Bray-Curtis dissimilarity index, Table 1). The interaction between diet and genotype was sexdependent. The Pomc tm1/tm1 genotype was associated with 26% of the variation in the male gut microbiota on the LF diet. However, this reduced to 12% of the variation in the gut microbiota when male Pomc tm1/tm1 mice were fed the HF diet (Bray-Curtis dissimilarity index, Table 1). By contrast, the genotype association with the gut microbiota in female Pomc tm1/tm1 mice increased from 19% on the LF diet to 29% on the HF diet (Bray-Curtis  Exposure to chronic HF diet was only associated with reduced bacterial richness in WT mice. We investigated genotype and dietary associations with α-diversity (i.e., number of observed OTUs per sample and the Shannon's diversity index). Eight pairwise comparisons were performed to test either genotype effect within diet and sex on bacterial α-diversity, or diet effect within genotype and sex on bacterial α-diversity (Table 2). HF diet was significantly associated with reduced bacterial richness in male and female WT (p = 0.0086 and B, p = 0.0014) but not Pomc tm1/tm1 mice ( Table 2). When investigating the Pomc tm1/tm1 genotype effect, Pomc tm1/tm1 females on HF showed significantly higher bacterial richness than WT females on the chronic HF diet (Fig. 4C, p = 0.0021). However, no significant differences were observed in the other pairwise comparisons ( Table 2). Analyses of the Shannon diversity index identified no significant differences between diet, or genotype ( Table 2).
Sex-specific alterations of acetic and propionic acid concentrations were observed between genotypes. More than 90% of the gut microbiota in WT and Pomc tm1/tm1 mice was composed of Bacteroi-  S3). Linear modelling (MaAsLin) did not identify significant associations between genotype, diet, SCFAs, BCAAs and specific bacterial taxa at phylum or genus levels within sex. Linear mixed-effect models identified significantly gender-specific effect of genotypes on the fecal acetic acid (gender: genotype: p = 0.0314) and propionic acid (gender: genotype: p = 0.0129). Specifically, WT females had 2.78 times the concentration of acetic acid (p = 0.0239, Fig. 5A), and 3.34 times the concentration of propionic acid (p = 0.0099, Fig. 5B) observed in Pomc tm1/tm1 females. By contrast, there were no significant differences in acetic and propionic acid levels between male genotypes ( Fig. S4A and B). No significant interactions or main effects were observed for butyric acid (Figs. 5C, S1C) or BCAA concentrations (Fig. S5).

Discussion
We have identified anatomical changes (e.g., increased intestinal length) in male and female Pomc tm1/tm1 mice when they were fed a chronic HF diet. Deletion of desacetyl-α-MSH and α-MSH from the female mice resulted in loss of protection from diet induced obesity and a failure to reduce villus length on exposure to the chronic HF diet. These anatomical changes were accompanied by changes in the gut microbiota and fecal SCFA concentrations. We conclude that these structural changes to the gut contributed to the observed increase in female body weight that occurred in the absence of an appreciable increase in food intake or caloric consumption 6 .
It has previously been demonstrated that adult male mice fed a HF diet (45% kcal from fat) for 12 weeks had an enlarged villus area in the small intestine 28 . Similarly, short term exposure to HF diets (45% kcal from fat; 20 weeks) increase villus height in the jejunum of adult male and female Sox9-EGFP reporter mice 29 . We contend that the differences between these studies and ours, despite the use of the same diet, stem from exposure of our animals to a chronic HF diet from weaning. Therefore, the age of exposure might be critical for structural adaptations of the intestine that determine the absorptive surface area available for nutrient absorption. Human studies of total small bowel length show a remarkable variation from 2.5 to 13 m. It is generally longer in males than females 30 . A weak correlation with height has been reported although there is no correlation with degree of obesity in abdominal surgery cohorts 30 .  31 . For example, there were more phylum Actinobacteria and Tenericutes in males; while genus Dorea, Coprococcus and Ruminococcus were more abundant in females 31 . The gender specific gut microbiota composition Org et al. observed was genotypedependent when strains were analyzed individually 31 . The most apparent sex specific differences in gut microbiota among the 86 strains were observed in C57BL/6J mice; and the gender specific microbiota composition in C57BL/6J mice is greater in response to a HF diet 31 . Sexually dimorphic changes in the gut microbiome are likely mediated by sex-dependent immune differences 32 . POMC-derived peptides are key regulators of innate and adaptive immunity 33 . Consistent with this, our results identified that deletion of desacetyl-α-MSH and α-MSH is associated with sex-specific differences in the gut microbiota, in response to exposure to a chronic HF diet.
The association between reduced bacterial diversity and obesity has been described by some 16,34 , but not all studies 35 . Similarly, inconsistent results have been observed between bacterial richness and obesity in mouse studies 29,36 . In this study, we observed that HF compared with LF diet was associated with reduced bacterial richness in male and female WT mice, but not in male or female Pomc tm1/tm1 mice. This reduction in bacterial richness was, therefore, independent of sex and independent of obesity, since the HF diet-induced obesity in male but not female WT mice and the HF diet exacerbated both male and female Pomc tm1/tm1 obesity. Interestingly, an increased bacterial richness was observed in female Pomc tm1/tm1 mice compared to WT mice on a HF diet, while this Pomc tm1/tm1 genotype effect was absent in female mice on a LF diet or male Pomc tm1/tm1 mice regardless of diet. This is consistent with the results of the β-diversity analysis, which showed the effect of Pomc tm1/tm1 genotype on gut microbiota composition was intensified by a HF diet, but not in males. Taken together, our study indicates a sex-specific interaction between melanocortin peptides and diet in influencing gut microbiome. The most likely implication of these findings are that α-MSH acts to protect female mice from chronic HFD related obesity, and in the absence of these peptides, there are alterations to the gut microbiota which promote increased energy extraction from the HFD. However, there are other potential explanations that need to be considered.
We cannot conclude a causal relationship between the gut microbiota, Pomc tm1/tm1 genotype, HF diet, and fecal SCFAs. Pomc tm1/tm1 mice gained significantly more body weight than WT mice on either the LF or chronic HF diets 5,6 . Given that food consumption influences gut microbiota composition 37,38 , our experimental design does not allow us to determine if the significant difference in body weight between groups was causal in changing the gut microbiota composition and fecal SCFA concentrations. For example, HF diets are known to induce hyperphagia by reducing sensitivity to satiety signals (e.g., cholecystokinin) in mice 39 . HF diet also induces inflammation in the hypothalamus, which has impacts on appetite-stimulating NPY/AgRP neurons 40 . Therefore, it remains possible that the altered gut microbiota composition we observed is a consequence of: (1) hyperphagia induced by desacetyl-α-MSH and α-MSH deficiency; or (2) the chronic HF diet itself. Pair-fed mice would be a better control, than ad-libitum fed mice, to untangle this effect. Since the female Pomc tm1/tm1 mice gained more body weight than WT females when fed a LF diet, despite consuming the same amount of this food 6 , it is likely that the differences between these groups are due to the deficiency of the desacetyl-α-MSH and α-MSH impacting on increased energy extraction from the diet, although decreased energy expenditure could also be involved.
Female Pomc tm1/tm1 mice had lower fecal acetic acid and propionic acid levels, regardless of diet. The main source of SCFAs in this study was dietary fiber. The fiber concentration of LF diet (50 g/1055.05 g) was lower than the HF diet 50 g/858.15 g). However, the diet effect on SCFAs was not significantly different and could be masked by genotype and sex effects. Acetic acid, butyric acid and propionic acid account for > 95% of SCFAs, and each compound has different distributions and individual functions 41 . For example, butyric acids are mainly produced by members of the phylum Firmicutes (e.g., Eubacterium, and Roseburia species) and used by the colon epithelial cells 42 . By contrast, acetic acid and propionic acid can pass through the blood brain barrier and are known to contribute to the regulation of lipid metabolism, insulin sensitivity, anti-inflammation and feeding behavior 41,43 . The reduced acetic acid and propionic acid levels we measured are consistent with the obese phenotype in female Pomc tm1/tm1 mice, based on other studies (reviewed in 44 ). This study provides further evidence supporting the existence of sex-specific differences in gut communities and metabolites in response to HF diet.
Fecal SCFA concentrations need not mirror concentrations of SCFAs in the peripheral circulation. SCFAs are the end products of anaerobic metabolism in the gut and generally have the highest concentrations in the large intestine 45 . After local use by the colonocytes, SCFAs are released into the portal, hepatic and peripheral venous blood 45 . Therefore, we are unable to conclude whether the reduced fecal SCFAs in female Pomc tm1/tm1 mice were due to decreased production or an increase in circulating SCFAs in these mice. Future analyses involving plasma SPME headspace GCMS may help to shed light on this.
The results of our study are consistent with the POMC derived peptides having a sex-specific role in modulating energy balance beyond appetite alone 6 . We have shown that the sex-specific effect of POMC peptides is associated with altered gut morphology and gut microbiota, which could be either a consequence or a cause of the Pomc tm1/tm1 obesity, or both, given that there is bidirectional communication between the gut microbiome and brain. Cross-talk between the gut and brain is essential for integrating peripheral and central immune, metabolic and endocrine signals to maintain energy and metabolism homeostasis [46][47][48] . It remains possible that the microbes contribute to this gut-brain communication by: (1) the production and activation of appetiteregulating hormones (e.g., cholecystokinin) 49 ; (2) impacts on central appetite regulation through disruption of the blood-brain barrier (BBB) 50,51 , or (3) production of peptides that mimic appetite-regulating hormones (e.g., C1pB, as a mimetic of α-MSH) 52 . Resolving this will require fecal microbiome transfer experiments to establish a causal relationship for the microbiome in Pomc tm1/tm1 mediated obesity.
In conclusion, this study supports a role for POMC derived peptides in conferring protection from chronic HF diet related obesity in females through compensatory changes in the gut morphology and gut microbiota that collectively limit nutrient absorption. We also demonstrated a sex-specific effect of Pomc tm1/tm1 genotype in influencing gut microbiota in the presence of a HF diet.  6 . Totally, 63 mice were sacrificed for analysis at age 23 weeks (32 male mice; 31 female mice: one mouse died on the way). There were no more than three mice per group from the same mother (Table S4; 6 ).
Tissue collection. Mice were culled by CO 2 asphyxiation for tissue collection after 23 weeks on a LF or HF diet 6 . The intestinal tract, from the duodenum to colon, was collected immediately post cull. Intestinal length was measured using a ruler. Ice-cold phosphate buffered saline was flushed through the gut tissues using a BD PrecisionGlide Needle (18G 1 ½ TW 1.2 × 38 mm). The small intestine was dissected into the duodenum, jejunum, and ileum. Segments (2 cm) of the duodenum, jejunum and ileum were collected for histological analysis.
The caecum was collected for gut microbiota analysis. The contents of the caecum were transferred to RNAand DNA-free Eppendorf tubes, using sterile forceps. Samples were snap frozen on dry ice. Fecal pellets, for quantification of SCFAs and BCAAs, were collected between 9 am and 12 pm, 1-7 days before the mice were culled. Pellets were stored frozen (-80 °C).
Histology. All histological images were acquired using a Zeiss AxioCam HRc Axioskop 2 mot plus microscope. Morphometric analysis and cell counts were performed using Image J 1.46r 53 .
Intestinal segments from the duodenum, proximal jejunum, and distal ileum were fixed in 10% neutral buffered formalin for 24 h, before storage in 70% ethanol. Tissues were embedded longitudinally in paraffin using standard techniques. Tissues were sectioned (5 µm cross sections). Four serial cross sections, from each animal, were obtained and mounted on each slide (n = 8 or 7 mice per group). One slide per tissue per animal was used for villus length and crypt depth measurement. One slide was used for goblet cell analysis.
Haematoxylin and eosin (H&E) staining was used to evaluate villus length and crypt depth 54,55 . All welloriented villi and crypts from four serial sections per tissue, per mouse, were used to determine villus height and crypt depth. The villus length was defined as the length between the top to base of each villus, excluding the crypt, using 20 × magnification. The crypt depth was defined as the distance from the base of the villus to the crypt interface with the outer muscularis mucosae 56 .
Periodic Acid Schiff (PAS) stain was used to visualize goblet cells 57 . Goblet cell numbers in each section were counted for all well-oriented villi. The number of goblet cells was normalized to the area of each villus, measured using Image J 1.46r 53 . 16S gene amplicon sequencing libraries were quantified using a Qubit Fluorometer. Libraries were normalized, pooled, denatured (0.1 M NaOH) and diluted to 20 pM. PhiX (PhiX Control Kit v3; Illumina) was added (5% final concentration). PhiX spiked libraries were pooled (600 μl) and loaded onto an Illumina MiSeq according to the manufacturer's instructions (MiSeq reagent Kit v2). FastQ files (2 × 250 bp Paired-End reads) were generated, run quality checked using the PhiX Control and each paired-end sequence assigned to its sample using the multiplexing index. Paired-end reads were quality filtered and assembled using Paired-End read mergeR (version 0.9.10). The open-reference operational taxonomic units (OTUs) picking method was used to identify bacteria and compute their distribution across different taxonomic levels. Specifically, reads were clustered into OTUs using the UCLUST algorithm and a 97% sequence similarity threshold. Singleton OTUs were removed. Taxonomic annotation of the cluster centroids was performed using the Green genes 16S gene database (GG, version 13_8) 59,60 . Rarefaction curves based on the phylogenetic diversity (PD) metric, showed a trend towards saturation and confirmed that the sequence depth in each sample was sufficient to include most OTUs (Fig. S6).
β-diversity analyses were performed in MicrobiomeAnalyst 61 using the Jaccard and Bray-Curtis dissimilarities matrix and visualized by Principal Coordinates Analysis (PCoA). The Jaccard index compares microbiome profiles in terms of presence/absence of OTUs 62 . By contrast, the Bray-Curtis dissimilarity index compares OTU abundance profiles 63 .
PERMANOVA was used to test for significant differences in gut microbial communities and structure between groups. PERMANOVA, with 999 permutations, returns a P value for statistical significance, alongside an R 2 value, indicative of the proportion of variation explained by a specific grouping variable.
The Shannon diversity index (accounting for both abundance and evenness) and richness (the number of observed OTUs) was used to measure α-diversity of the bacterial communities. The Mann-Whitney U test was used to identify any significant differences between LF and HF diets, and between WT and Pomc tm1/tm1 genotypes in terms of bacterial richness and evenness. Cage effects were not accounted for in this test.
Fecal SCFA and BCAA quantification. SCFAs and Branched Chain Amino Acids (BCAAs) were analyzed using propyl chloroformate (PCF) derivatization followed by gas chromatography mass spectrometry (GC-MS) 64 . In brief, fecal samples were mixed with NaOH (0.005 M) containing stable-isotope labelled internal standards (acetic acid-d4, l-leucine-d10 and l-alanine-d4, all at a concentration of 25 μg/ ml). A mixture of 1-propanol: pyridine (3:2, v/v) and propyl chloroformate were added and the derivatives were extracted with hexane. The upper hexane layer was transferred to a GC-MS vial for analysis. The remaining solid fecal matter was placed in a centrifugal vacuum concentrator (Thermo SpeedVac) until visibly dry (~ 5 h). The dried fecal biomass (Tables S5, S6) was used for normalization.
SCFAs and BCAAs were separated and identified using an Agilent 7890A GC coupled to an Agilent 5975C MSD. Quantification was performed using a 10-point calibration curve (0.5-250 µg/ml) for each target analyte, normalizing by internal standard and using the regression function of the curve to predict the concentration for any given peak area. Raw GC-MS data were converted to NetCDF files using MSD ChemStation G1701FA F.01.01.2317. MassOmics (XCMS R packaged in GUI) was used to integrate peak area.
Multivariate association testing. Individual associations between bacterial genera or phyla and genotype, diet, SCFAs and BCAAs were analyzed by linear modelling in multivariate association with linear model (MaAsLin) 65 . Cage effect was accounted for when testing the differences of SCFAs and BCAAs between genotypes, diets and genders using linear mixed-effects models (nlme, version 3.1-137) 66 .
Linear mixed-effects models (nlme, version 3.1-137, R version 3.5.0) 66 were used to test for differences in SCFA and BCAA concentrations between genotypes, diets and gender. Zero values were removed, and the data were logarithm transformed to satisfy normality assumption, prior to analysis. Cage effect was a confounder in this study and was accounted for in the linear mixed-effect models. Six linear mixed-effects models were generated based on six response variables (acetic acid, butyric acid, propionic acid, l-leucine, l-isoleucine and l-valine) with diet, genotype and gender modelled as the fixed effects and cage modelled as random effects. Analysis of variance (ANOVA) was used to test for the three-way interactions, two-way interactions and main effects. Pairwise comparisons that passed the p < 0.05 threshold of significance in the ANOVA were subjected to post-hoc analysis.