Gut microbiome and bile acid changes after male rodent sleeve gastrectomy: what comes first?

Abstract

Background

Understanding how a sleeve gastrectomy (SG) achieves metabolic improvement is challenging due to the complex relationship between the liver, bile acid (BA) pool, and gut microbiome. We hypothesized that SG alters the gut microbiome which then increases the BA pool leading to metabolic efficacy.

Methods

We performed fecal material transfer (FMT) from SG or sham mice to surgically-naïve mice with an intact microbiome. We evaluated the effect of surgery and FMT on BA-related liver enzymes, BA concentrations, and gut microbiome composition via 16s and metagenomic analysis.

Results

SG significantly deflected weight gain compared to sham surgery, 5±2 g vs 10±3 g respectively (p= 0.004). SG significantly increased the BA pool and decreased liver transcription of slc10a1 (p=0.04) and cyp8b1 (p=0.03). Random forest analysis identified several features with significantly increased relative abundance in SG compared to sham mice including Lactobacillus. Examination of metabolic profiles with metagenomic analysis revealed a BA salt hydrolase produced by the Ligilactobacillus species.

FMT of SG stool to surgically-naïve mice significantly decreased the BA pool compared to sham FMT (p=0.034). Unlike SG surgery, we found no effect of SG or sham FMT on bile acid related enzymes in the liver after 14 weeks of treatment.

Conclusion

Overall, we propose that the metabolic benefits of SG surgery are related to decreased liver transcription of cyp8b1and slc10a1 with subsequent increases in the systemic and enterohepatic BA pool including LCA. The gut microbiome adapts to the altered BA pool with associated increases in Ligilactobacillus and bile salt hydrolase production.

Graphical Abstract

New and noteworthy:

We propose that the metabolic benefits of sleeve gastrectomy are initiated by decreased liver transcription of cyp8b1and slc10a1. A notable downstream effect includes changes in systemic bile acid composition and circulation including increased LCA. An altered gut microbiome after surgery includes increases in Ligilactobacillus that was shown to express a bile salt hydrolase which could be a contributor to the post-sleeve gastrectomy gut microbiome changes.

Introduction

Metabolic and bariatric surgery (MBS) is the most effective treatment of severe obesity and associated medical conditions [1,2]. The most commonly performed MBS is a sleeve gastrectomy (SG) where the stomach volume is reduced by approximately 75% as a greater curvature gastrectomy [3]. SG induces weight loss through multiple potential mechanisms including systemic and central alterations in endocrine physiology and metabolism [4,5]. While there are multiple mechanisms that induce weight loss and metabolic disease improvement still under exploration, there is supportive evidence from rodent studies that both changes in the bile acid pool and gut microbiome are implicated in the weight-loss and metabolic effects [6–8]. There are multiple mechanisms that may induce weight loss and metabolic disease. It is still under investigation if changes in the gut bile acid pool and microbiome are critical to the weight loss and metabolic effects of SG.

Documented shifts in the post-SG microbiome include increased diversity and gene richness, increased Akkermansia, and decreased Firmicutes [9]. These shifts are all associated with metabolic health [10,11]. Furthermore, previous studies have shown that the use of peri-operative antibiotics in SG disrupts its beneficial effect on obesity and glucose homeostasis [12]. We have previously published increased post-prandial bile acids which correlate to the weight loss response [13].

There is a complex relationship between the liver, the bile acid pool within the enterohepatic circulation, and the gut microbiome. Primary bile acids are cholesterol metabolites synthesized in the liver, conjugated with taurine or glycine, and secreted into the small intestine where they can be transformed to secondary bile acids by gut bacteria and reabsorbed through the portal circulation back to the liver [14]. A small fraction of the bile acid pool circulates and acts as critical metabolic regulators and nutrient sensors [14].

Elucidating a role for bile acids and gut bacteria following SG regarding weight loss and metabolic improvements has been challenging. Modification of one system through molecular techniques causes a compensatory change in another system outside the surgical intervention. In particular, the use of germ-free or antibiotic-depleted animal models to study the specific role of the gut microbiome with SG is limited as lack of a gut microbial system inherently alters the bile acid pool, obesity sensitivity, and glucose homeostasis [15–17].

To determine if SG alters the gut microbiome which then increases the bile acid pool and systemic bile acid concentrations, we performed fecal material transfer (FMT) from high-fat diet (HFD) fed, obese SG donor mice to HFD fed, obese surgically naïve recipient mice with an intact gut microbiome. With this methodology, any changes noted in the recipient mice bile acid pool would be mechanistically linked to the donor SG gut microbiome and not an artifact of a germ-free or gut microbial depleted animal model.

Methods

All methods presented here were reviewed and accepted by the institutional review board at the Medical College of Wisconsin.

Study design for donor mice.

Twenty-four male C57Bl/6J mice at 3 weeks of age were obtained from Jackson Laboratories. After acclimation, mice were housed individually on standard rodent chow. As shown in Figure 1A, at 4 weeks of age, individual stool collection was performed (time point A), and the mice were switched to a 45% high fat diet (HFD, D12451 Research Diets). This diet was continued for 8 weeks prior to surgery (12 weeks of age) and individual stool was collected (time point B). Mice were matched by body weight in each group and then underwent SG (n=12/sex) or sham (n=12/sex) surgeries at 12 weeks of age.

Figure 1.

A) Timeline of male C57Bl/6J surgical donor mice cohort. Red and green represent sham (n=11) and sleeve (n=6) surgery respectively. Week number in terms of age and experimental week included. The liver icon represents when liver enzyme expression measurements were taken. The bile acid icon represents when bile acid measurements were taken. The knife icon represents when the surgeries were performed. The food icon represents when the high fat diet was started. The stool icon represents weeks that stool for 16s sequencing was taken. Stool from timepoint C also underwent metagenomic sequencing. Stool was collected within the four-week gap between timepoint C and euthanasia for use in the FMT experiment. Experimental week 20 represents the end of the experiment in which serum bile acid, and hepatocyte transcription analyses were performed. B) Weight gain after sleeve gastrectomy (n=6) or sham surgery (n=11) on a 45% high fat diet. (*) is p<0.05 C) Absolute weight gain after sleeve gastrectomy (n=6) or sham surgery (n=11) on a 45% high fat diet. (*) is p<0.05 D) Bar graph depicting the cumulative food intake in grams of sham or sleeve gastrectomy mice throughout the experiment. (*) is p<0.05. Error bars represent standard error of the mean.

Post-operatively, mice were continued on a 45% HFD for 12 more weeks (24 weeks of age). Individual stool samples were collected at 8 weeks post-surgery (time point C) for 16S analysis. Stool was then collected from individual home cages every 3 days over the next 4 weeks and batched per surgical group to create SG or sham stool for future FMT treatment (post-op weeks 8–12). The mice were then euthanized. Prior to euthanasia mice were gavaged with 100 µL chocolate Ensure to create a post-prandial metabolic state for tissue and serum collection. At euthanasia, blood was collected from the inferior vena cava for future post-prandial bile acid analysis 15–20 minutes after oral gavage. A segment of the right lobe of the liver was flash frozen in liquid nitrogen and stored at −80°C. Body weight and food intake were monitored at least weekly throughout the entirety of the experimental timeline. This study and all the associated procedures were approved by the Medical College of Wisconsin’s Institutional Animal Care and Use Committee.

Surgical Procedures.

Mice were fasted from food overnight but had ad libitum access to water. A subcutaneous injection of long-acting buprenorphine was given for perioperative pain control and 0.9 NaCl to prevent post-operative dehydration. Perioperative antibiotics were not administered due to the previously shown negative effects of antibiotics on metabolic efficacy of rodent SG [18]. The abdomen was incised in the midline through the abdominal muscles from the xiphoid extending 1.5 cm inferiorly. The liver was retracted to the right exposing the stomach. The fibro-fatty tissue of the greater curvature was taken up to the gastroesophageal junction with scissors. For SG, a medium-large vascular Weck Hemoclip (Teleflex) was placed from the antrum to just lateral to the GEJ excluding the greater curvature of the stomach including all of the non-glandular lateral stomach. The excluded stomach was excised with scissors. For all mice, the abdominal muscles were closed with a running 4–0 Vicryl suture. The skin was closed with interrupted vertical mattress 4–0 permanent sutures. The animals were kept on a warmer until fully awake and then returned to their individual cages for the remainder of the study. Mice were given a liquid nutritional supplement (chocolate Ensure) for 48 hours post-operatively before transition back to the HFD. Sham surgery was performed similar to SG surgery, but no clip or division of the stomach was performed.

FMT Formulation.

Frozen batched stool from surgical group donors was ground into a powder. Two separate mortar and pestles were used to prevent cross contamination. PB2 (PB2- Powdered Peanut Butter Original, PB2 Foods) was used as the oral carrier of the FMT. PB2 only for the control group was made as 70 mg of PB2 Original powdered peanut butter mixed with 100 µL of PBS per animal per day. FMT was made as 70 mg of PB2, 10 mg of frozen powdered SG or sham stool, and 100 µL of PBS per animal per day. PB2 control or FMT was made as batches each week as individual pellets and stored at −20 until use.

Study design for FMT treatment of recipient mice.

As shown in Figure 5A, 30 male C57Bl/6J mice at 4 weeks of age were purchased from Jackson Laboratories. After acclimation, mice were housed individually and started on a 45% HFD (D12451 Research Diets). At 7 weeks of age, all mice were acclimated to the PB2 carrier of the FMT for 7 days. No antibiotics were used to diminish the host gut microbiome. The PB2 pellet was placed in the cage on the lick spot and left for 30 minutes at the same time each day. After 30 minutes, all cages were checked and documented if the PB2 pellet was eaten. 100% acclimation took a total of 4 days. Experimental treatment was started the following week (8 weeks of age) as one of three FMT treatment groups (n=10/group): PB2 control, PB2 + SG stool (SG FMT), or PB2 + sham stool (SH FMT). Treatment was continued for 14 weeks until 22 weeks of age. Individual stool was collected at baseline (pre-PB2 training), and at 2, 8, and 14 weeks of FMT for 16s analysis. Mice were euthanized at 14 weeks of treatment (22 weeks of age). Mice were again gavaged with 100 µL chocolate Ensure prior to euthanasia as in the surgical cohort and blood collected from the inferior vena cava 15–20 minutes after oral gavage. Tissue samples were collected as in the surgical cohort.

Figure 5.

A) Timeline of the male c57Bl/6J fecal material transplant recipient mice cohorts. Weeks of intervention listed at bottom. Orange, red, and green represent PB2 (n=9), PB2+Sham stool (n=10) and PB2+Sleeve stool (n=10) treatments respectively. The stool icon represents weeks that stool for 16s sequencing was taken. Week 18 represents the end of the experiment in which 16s, serum bile acid, and hepatocyte transcription analyses were performed. B) Weight gain experienced by all male c57Bl/6J fecal material transfer groups shown as grams gained per week of 45% HFD and stool treatment. ns= non-significant differences for each week between groups with a p-value <0.05. C) Absolute weight gain experienced by all male C57Bl/6J fecal material transfer groups shown as grams gained per week of 45% HFD and stool treatment. ns= non-significant differences for each week between groups with a p-value <0.05 D) Cumulative grams of 45% HFD food eaten by all male c57Bl/6J fecal material transfer groups throughout the experiment. Error bars represent standard error of the mean.

RNA isolation and qRT-PCR.

Total RNA was isolated from 30 mg of liver tissue by first homogenization using the red Eppendorf lysis kit (NextAdvance). RNA was then isolated using the RNEasy kit (Qiagen) into 50ul nuclease free distilled water. Total RNA concentration was determined by nano-drop measurement. Reverse transcription was performed with 500 ng RNA using the Superscript IV VILO with ezDNase kit (Thermofisher) according to the manufacturer protocol. qPCR was then performed using Ssofast EvaGreen Supermix (BioRad). Annealing was done at 55°C for 40 cycles and resulting data analyzed using CFX Maestro software (BioRad). Gene expression was normalized to GAPDH for each sample. Primers (5’ to 3’) utilized are shown below:

GAPDH aacgaccccttcattgacct (F); tggaagatggtgatgggctt (R)

cyp7a1 AGCAACTAAACAACCTGCCAGTACTA (F); GTCCGGATATTCAAGGATGCA (R)

cyp8b1 GCCTTCAAGTATGATCGGTTCCT (F); GATCTTCTTGCCCGACTTGTAGA (R)

fxr AGCTCTTTATTCTGTCAACCACA (F); GCCAGCTGCTTACTTGTCTCT (R)

slc10a1 TGGCTACCTCCTCCCTGATG (F); GCCAGGTTGTGTAGGAGGAT (R)

slco1a1 TGAGAAAGACAGCAGTAGGACTTT (F); GTGATTTGGCTAGGTATGCAC (R)

shp GATCCTCTTCAACCCAGATGTGC (F); CTACCAGAAGGGTGCCTGGA (R)

Bile Acid Mass Spectrometry and Analysis.

Creative Proteomics (Shirley, NY) performed bile acid analysis from post-prandial serum samples at study end. According to the provided protocol, samples were prepared as 20 µL of serum mixed with 80 µL of internal standard (IS) solution of 14 deuterium-labeled bile acids in mixed acetonitrile-methanol (1:1) solution. An Agilent 1290 UHPLC system coupled to an Agilent 6495B QQQ mass spectrometer was used. The MS instrument was operated in the multiple reaction monitoring (MRM) mode with negative ion detection. A Waters C18 column (2.1*150 mm, 1.7 µm) was used for LC separation and the mobile phase was 0.01% formic acid in water and in acetonitrile for binary-solvent gradient elution. Bile acids measured included: cholic (CA), chenodeoxycholic (CDCA), deoxycholic (DCA), lithocholic (LCA), omega-muricholic (OMCA), alpha-muricholic (αMCA), and beta-muricholic (βMCA) acids as well as their glycine and taurine conjugated metabolites.

DNA Extraction.

Bacterial DNA was extracted from stool pellets frozen at −80°C using the Qiagen Powerlyzer Powersoil Kit following manufacturers protocol with modifications including two heat step incubations, extensions of incubation and centrifugation times, as well as the inclusion of an ice incubation step [19]. DNA quality was assessed via Nanodrop and QuBit before being frozen at −80°C until sequencing.

16s rDNA Bacterial Gene Sequencing.

DNA was sent to the University of Wisconsin- Biotechnology Center (UWBC; Madison, WI) for processing. Composition of the gut microbiome was determined through sequencing the 16S rDNA gene V3–V4 region spanning nucleotides 341–806 [9]. PCR amplicons were produced and sequenced on the MiSeq platform (Illumina, San Diego, CA, USA) using the 2 × 300-bp protocol using a dual-indexing amplification strategy, optimized for the Illumina MiSeq platform.

16s Ribosomal Data Analysis.

16s ribosomal DNA analysis was performed as described previously [20]. Briefly, 16S rDNA sequences were analyzed using Quantitative Insights Into Microbial Ecology (QIIME2) [21]. Sequences were denoised, filtered and trimmed using the DADA2 plugin with a PHRED cut-off of 25. Chimeric sequences were also removed in this step. Alpha-diversity and beta-diversity were estimated using the core-metrics-phylogenetic command of QIIME2 [9,22,23,24]. The q2-diversity plugin was run after samples were rarefied. PCoA plots of beta diversity were examined using Emperor [25,26]. Abundance data was exported from QIIME2 for downstream analysis of discriminant features. Taxonomy was assigned to amplicon sequence variants (ASVs) using the SILVA reference database and used to create taxonomy bar plots and generate relative abundance tables [9, 27].

Linear discriminant analysis effect size (Lefse) of 16s data was performed to identify differential ASVs between treatment groups [28]. Random forest analysis of 16s data was also performed on R Studio using the Boruta filter to identify statistically significantly differential ASVs by iteratively removing features found to be less treatment specific than randomly assigned probes [29]. A cut-off Gini score of 0.09 was used to determine significance as described previously [9].

Metagenomic Analysis.

Metagenomic analysis of stool microbiota was performed on sterile collection samples from the surgical cohort at timepoint C, Figure 1A. This analysis was performed to further characterize the post-SG microbiome and identify potential bile acid altering metabolic functions and their associated metagenomes [30]. Samples were sequenced using the Iluminia NovaSeq 6000 technology at the UWBC. Raw sequences were filtered using Ilumina Utills and Trimmomatic to remove host sequences and low-quality reads. Reads from all samples were assembled with MEGAHIT to generate contigs of a minimal length of 1000 bp. The generated contigs file was reformatted using Anvi’o v7 (anvi-script-reformat-fasta) with Bowtie2 used to map short reads to the assembled contigs and Samtools to convert contigs into sorted BAM files. The script (anvi-run-scg-taxonomy) was run to assign taxonomy to single-copy core genes before adding taxonomic annotations via the KAIJU database and functional annotations. Via the NCBI’s. Cluster of Orthologous Groups (COGs). Anvi’o was then used to generate profile databases. For each sample with a minimum length cut off of 2500 bp. The individual profiles were merged and binned into Metagenomic Assembled Genomes (MAGs) using the CONCOCT feature (anvi-cluster-contigs). Anvi’o was also used for storage of open reading frames as well as provision of sample specific information such as mean coverage, standard deviation, and single nucleotide variant information. The program also allows for manual binning to further microbiome members and functions. MAGs were manually curated for high-quality. Genomes with a threshold of over 50% genome completion and less than a 10% redundancy rate based on bacterial core gene collections found in Anvi’o. Key differences between surgical groups were identified and explored for bile acid altering metabolomics using Anvi’o’s search feature and KEGG orthologs. Identified enzymes of significance were confirmed using the NCBI protein BLAST function [31,32].

Statistical Analysis.

Comparisons between groups for variables collected were analyzed using an unpaired, two-tailed T-test for the surgical experiments and by ANOVA for the FMT recipient experiments. All samples were assessed for normality prior to subsequent testing. If identified as not meeting normality requirements by a Shapiro-Wilk test, either a T-test with Welch’s correction or a Kruskal-Wallis with Dunn’s post comparison test were performed. Significance was determined as p<0.05. All plots were made in Prism GraphPad v10.

Results

Donor Mice

Body weight and food intake.

As shown in Figures 1B–1C, SG significantly deflected weight gain compared to sham surgery by the first week postoperatively. SG mice gained 5.33 ± 2.1g post-surgery, while sham mice gained 11.09 ± 3.1g, p= 0.004. SG significantly decreased cumulative food intake for the first 4 weeks after surgery compared to sham mice (58.3 ± 5.4 g vs. 65.9 ± 5.6 g respectively, p=0.02). No difference in cumulative food intake was detected over the last 8 weeks of study period (p=0.29) (Figure 1D).

Surgical changes in bile acids.

Total post-prandial, serum bile acids were not significantly different between SG, 4.87 ± 6.24 µM, and sham mice, 0.98 ± 0.72 µM, p=0.08. However, we found significant differences in multiple individual bile acids (Figure 2). SG significantly increased concentrations of LCA (p=0.01), and both glycine (G-, p<0.001) and taurine (T-, p=0.02) conjugated LCA compared to sham surgery. SG also significantly increased G-CDCA (p<0.001) and G-DCA (p=0.001).

Figure 2.

Bar graphs depicting serum bile acid levels in male c57Bl/6J undergoing sham surgery (n=11) or sleeve gastrectomy (n=6). A) Total bile acids B) Cholic acid C) Chenodeoxycholic acid D) Deoxycholic acid E) Lithocholic acid F) Alpha-muricholic acid G) Omega-muricholic acid H) Beta-muricholic acid I) Glyco-cholic acid J) Glyco-chenodeoxycholic acid K) Glyco-deoxycholic acid L) Glyco-lithocholic acid M) Tauro-cholic acid N) Tauro-chenodeoxycholic acid O) Tauro-deoxycholic acid P) Tauro-lithocholic acid Q) Tauro-omega-muricholic acid R) Tauro-alpha-muricholic acid S) Tauro-beta-muricholic acid. (*) is p<0.05. Error bars represent standard error of the mean.

Surgical changes in liver-specific bile acid enzymes.

To determine the effect of SG on bile acid synthesis and recycling in the liver, we measured the following enzymes from liver tissue: fxr, shp, slc10a1, slc1a1, cyp8b1, and cyp7a1 as shown in Figure 3. SG significantly decreased liver transcription of slc10a1 (p=0.04), a sodium-dependent taurocholate co-transporter (also known as ntcp) that mediates electrogenic uptake of bile acids into hepatocytes [33]. Compared to sham surgery, SG also decreased liver transcription of cyp8b1 (p=0.03), the enzyme responsible for the transformation of CDCA to CA. There were no significant differences in expression levels of the fxr bile acid sensing receptor shp, or cyp7a1 between the surgical cohorts.

Figure 3.

Bar graph depicting liver enzyme transcription levels in male c57Bl/6J undergoing sham surgery (n=11) or sleeve gastrectomy (n=6). Transcription levels were measured using qRT-PCR on mRNA isolated from hepatocytes. A) cyp7a1 B) cyp8b1 C) scl10a D) fxr E) shp F) slc01a1. (*) is p<0.05. Error bars represent standard error of the mean.

Donor Gut Microbiome Bioanalysis.

We found a significant difference in beta-diversity metrics of Weighted Unifrac (abundance and phylogenetic distance) p=0.001, Jaccard (presence or absence) p=0.001, and Bray-Curtis (abundance) p=0.001 between SG and sham surgery at 8 weeks post-surgery. There were no significant differences in unweighted Unifrac or in alpha-diversity metrics at 8 weeks post-operatively between groups.

Random forest analysis identified the following as features with significantly increased relative abundance in SG compared to sham mice: Akkermansia, Lactobacillus, Muribaculaceae, and Parasutterella (Figure 4A). Lefse analysis had similar results with significantly differential Lactobacillus, Akkermansia, and Verrucomicrobiota for SG (Figure 4b). Significantly elevated features in sham mice identified with random forest analysis included: Lactococcus, Clostridium, Peptostreptococcaeceae, and others. Lefse results were consistent with random forest analysis, identifying significantly differential Eubacterium, Clostridium, and Erysipelotrichales for sham mice.

Figure 4.

A) Bar graph of significantly differential features in male c57Bl/6J mice undergoing sham surgery (n=11) or sleeve gastrectomy (n=6) using Random Forest with Boruta analysis (Gini score > 0.09) of 16s data. B). Lefse plot of significant discriminant features between surgical cohorts. C.) Metagenomic analysis of stool collected from surgical cohorts at the final experimental timepoint (week 18). There is increased relative abundance of cholylglycine bile salt hydrolase containing Ligilactobacillus in sleeve vs sham groups.

Metagenomic analysis allowed for further investigation of microbiota members as well as their metabolic capabilities. The elevations in Lactobacillus in SG stool from 16S analysis were further resolved to be two different species: Lactobacillus apodemi and Ligilactobacillus. Examination of their metabolic profiles revealed a bile acid salt hydrolase produced by the Ligilactobacillus species (Figure 4C). KEGG ortholog search and BLAST analysis revealed the enzyme as cholylglycine hydrolase (CGH), a bile acid hydrolase which deconjugates glycine from conjugated bile acids.

FMT Recipient Mice

Body weight and food intake.

There was no significant difference in body weight or food intake between PB2, SH FMT, or SG FMT treatments after 14 weeks in male C57Bl/6J mice on a 45% HFD (Figures 5B,5C,5D). At the study end point, compared to PB2 only mice who gained 13.3 ± 2.6 g, SH FMT mice gained 14.9 ± 4.6 g (p=0.83) and SG FMT mice gained 12.1 ± 2.1 g (p=0.31).

FMT changes in bile acids.

Surprisingly as shown in Figure 6, we found SG FMT significantly altered the bile acid pool in an inverse fashion in comparison to the BA changes of SG donors. Total post-prandial, serum bile acids were significantly lower with SG FMT, 1.11 ± 0.54 µM, compared to SH FMT, 2.04 ± 0.37 µM, p=0.034. SG FMT also significantly decreased T-CA (p=0.014), T-DCA (p=0.013), T-βMCA (p=0.002), and T-αMCA (p=0.023) compared to SH FMT.

Figure 6.

Bar graphs depicting serum bile acid levels in male c57Bl/6J mice treated with fecal material transplant with PB2 (n=9), sham stool (n=10), and sleeve stool (n=10). A) Total bile acids, B) Cholic acid, C) Chenodeoxycholic acid, D) Deoxycholic acid, E) Lithocholic acid, F) Alpha-muricholic acid, G) Omega-muricholic acid, H) Beta-muricholic acid, I) Glyco-cholic acid, J) Glyco-chenodeoxycholic acid, K) Glyco-deoxycholic acid, L) Glyco-lithocholic acid, M) Tauro-cholic acid, N) Tauro-chenodeoxycholic acid, O) Tauro-deoxycholic acid, P) Tauro-lithocholic acid, Q) Tauro-omega-muricholic acid, R) Tauro-alpha-muricholic acid, S) Tauro-beta-muricholic acid. (*) is p<0.05. Error bars represent standard error of the mean.

FMT changes in liver-specific bile acid enzymes.

Unlike SG surgery, we found no effect of SG or SH FMT on bile acid related enzymes in the liver at 14 weeks of treatment (Figure 7).

Figure 7.

Bar graph depicting liver enzyme transcription levels in male c57Bl/6J mice treated with fecal material transplant with PB2 (n=9), sham stool (n=10), and sleeve stool (n=10). Transcription levels were measured using qRT-PCR on mRNA isolated from hepatocytes. A) cyp7a1, B) cyp8b1, C) scl10a1. ns= non-significant differences for each week between groups with a p-value <0.05. Error bars represent standard error of the mean.

Recipient Gut Microbiome Bioanalysis.

After 14 weeks of FMT treatment, the gut microbiome displayed significant differences in all beta-diversity metrics between treatment groups (Weighted Unifrac p=0.002, Jaccard p=0.002, and Bray-Curtis p=0.000) similar to the donor cohort. Likewise, there were no significant differences in Unweighted Unifrac or alpha-diversity metrics with FMT treatment.

Random forest with Boruta analysis of 16s data revealed significant elevations of Muribaculaceae and Lactobacillus in SG FMT groups (similar to SG donors), and increased Clostridium and Turicibacter in SH FMT groups (similar to sham donors). Interestingly, PB2 only treatment increased Akkermansia compared to both SG FMT and SH FMT (Figure 8A). Lefse analysis between SG FMT and SH FMT recipients was overall unremarkable, identifying only elevated Dubosiella and Negativibacillus in SH FMT recipients as significantly differential features (Figure 8B).

Figure 8.

A) Bar graph of significantly differential features between all three male c57Bl/6J mice fecal material transplant groups (PB2 (n=9), Sham stool (n=10), and Sleeve tool (n=10) using Random Forest with Boruta analysis (Gini score > 0.09) of 16s data. B) Lefse plot of significant discriminant features between fecal material transplant groups. A three-way analysis was unable to be performed due to Lefse program constraints.

Discussion

Our results support the well-documented metabolic improvement provided by SG including a significant reduction in weight-gain on HFD due to a post-surgical decrease in early post-operative caloric intake [34–38]. The effect of SG on the enterohepatic circulation and gut microbiome appears to be related to changes in the bile acid pool due to reductions in liver slc10a1 and cyp8b1expression resulting in enhanced systemic circulation of postprandial bile acids and increased CDCA concentrations. SG increased bile acid concentrations and altered the gut microbiome, while SG FMT reversed and lowered bile acid concentrations despite similar alterations in the gut microbiome. Rather than the gut microbiome acting as a causative mechanism for the high bile acid load after SG, this study supports that the gut microbiome after SG has adapted to handle and process a high bile acid load initiated by the liver, with increased presence of bile salt hydrolases from species like Ligilactobacillus.

Our finding of decreased cyp8b1 expression was initially reported by McGavigan et after SG in mice which was TGR5-dependent and contributed to shifts in bile acid subpopulations of 12α-hydroxylated to12α-non-hydroxylated ratios (primarily chenodeoxycholic : cholic acid) [39]. The effect of SG to reduce liver cyp8b1 expression has been verified in rat models as well [40]. Cyp8b1 knockdown in mice resulted in complete absence of cholic acid, and improved glucose tolerance, insulin sensitivity, and β-cell function [41]. Further, reduced cholic acid leads to increased free fatty acids reaching the distal small bowel and enhanced intestinal GLP-1 secretion, a universal finding after SG [42]. While in our study we did not find a decrease in circulating cholic acid levels after SG, we had notable increases in conjugated-CDCA and its corresponding secondary bile acid, LCA, an expected finding with decreased expression of hepatic cyp8b1. Most recently, Liu et al found that cyp8b1 overexpression diminished the metabolic efficacy of rodent SG including a reduced magnitude of SG weight loss and cyp8B1 downregulation was required for SG-related changes to the gut microbiota [7]. Using an antibiotic-suppression model for FMT, SG FMT reduced body weight, as did FMT from cyp8b1KO mice, but no assessment of the bile acid pool was made. Our results expand on these findings, suggesting that SG suppresses cyp8b1, alters the bile acid pool with increased conjugated-bile acids including CDCA, which in turn results in gut microbial adaptation with increased diversity and abundance of several species including Lactobacillus containing bile acid detoxification machinery [8].

Bile acids can be toxic to gut bacteria; thus, microbial adaptation after SG-induced increases to the bile acid pool is possibly survival driven [43]. The significant increase in Ligilactobacillus abundance could be due to expression of the bile acid modifying enzyme CGH which promotes bacterial survival, while other bacteria unable to adapt may decrease in abundance. This enzyme could be a protective factor for the Ligilactobacillus against the solubilizing nature of bile acids and has unintended beneficial consequences supporting weight loss, host fat digestion, and absorption [44]. CGH has also been associated with reduced serum cholesterol, and its increased abundance remains a potential additional mechanism for the favorable metabolism changes seen post-SG [45]. The decreased BA concentrations seen in our FMT cohort may be a result of the increased abundance of Ligilactobacillus and CGH. It is possible that once being transferred to a naïve gut, the CGH continues hydrolyzing BAs and consequently lowers systemic BAs [43–45].

Our analysis of changes in liver gene expression post-SG also identified a significant decrease in expression levels of slc10a1. The slc10a1 gene product, the sodium taurocholate co-transporting polypeptide (ntcp), is not just involved in bile acid handling but also has been implicated in the progression of nonalcoholic fatty liver disease [46]. Hepalatide, a ntcp antagonist, attenuates inflammation and apoptosis on liver histology which mirrors the remarkable effect of SG on hepatic inflammation and steatosis [47–49] and suggests a possible mediator for hepatic health. Similar to cyp8b, the reduction in slc10a1 gene expression was not replicated by our FMT cohort, suggesting that the changes in liver enzyme expression are more likely to be a result of the altered gastric anatomy in the post-SG state rather than a response to an altered gut microbiome.

We did not find a direct link between the gut microbiome and subsequent beneficial effects of SG on body weight and food intake in this study despite effectively altering the recipient gut microbiome. This could mean the weight-loss response after SG is bile-acid dependent. Quante et al have found the bile acid changes after SG directly impact body weight, with tauro-DCA able to induce a profound weight loss analogous to SG [50]. Ryan et al published a sentinel paper that the metabolic efficacy of SG was lost with farsenoid X receptor knockdown, a nuclear bile acid receptor [6]. However, it may be that a FMT in an intact microbiome such as employed in our study cannot induce weight loss. Using similar methodology, we have found SG FMT in HFD-fed rats prevents the development of hypertension with high-fat feeding without changes in body weight [8]. Contrastingly, Liu et al noted significant differences in body weight and glucose tolerance between SG FMT and Sham FMT in mice using antibiotic microbial suppression [7]. These results suggest that the downstream effect of SG on the gut microbiome provides beneficial contributions to weight loss and metabolic efficacy independent of systemic bile acid signaling.

Limitations.

The intact microbiome model has inherent limitations as described above which could reduce our ability to produce a microbial-mediated phenotype of SG. Further, larger doses of FMT, or more frequent dosing in an intact microbiome may more successfully uncover SG FMT supported weight loss and metabolic mechanisms. Rodent high-fat feeding does not recapitulate the complex polygenetic, environmental, and behavioral contributions to humans with obesity as the rodent surgical model of SG is limited due to differences in gastric anatomy between mice and humans. The effect of SG on cyp8b1 must be confirmed clinically to move forward as a mechanistic target for weight loss interventions. Limitations also include lack of functional testing of the identified bile salt hydrolase and inability to piece apart bile acid handling at each step in enterohepatic circulation. Further studies must be done to characterize these changes in the post SG state.

Conclusions.

Overall, we propose that the metabolic benefits of SG surgery may be initiated by decreased liver transcription of cyp8b1and slc10a1 altering systemic bile acid circulation including increasing LCA concentrations. An altered gut microbiome after SG is possibly a survival adaptation toward bile salt detoxification. We could not distinguish from this study the contribution of liver slc10a1 versus cyp8b1on the metabolic efficacy of SG and will need to be explored in future studies.

Data Availability

All sequences can be found on the NCBI GenBank database using the BioProject ID PRJNA1165775.