Distinct patterns in the gut microbiota after surgical or medical therapy in obese patients

Patient inclusion and clinical parameters

This study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Faculty of Medicine, Pontificia Universidad Catolica de Chile. All study participants provided written informed consent. Participants of this study were recruited from candidates of the Obesity Program from Red de Salud UC-Christus. Eligible subjects were men or women, 18–60 years old with a body mass index (BMI) 30–50 kg/m². Women who were pregnant or with the intention to get pregnant were excluded. Other exclusion criteria include chronic antibiotic use, record of small intestine and/or colon resection, intestinal inflammatory diseases and probiotic consumption. A total of 19 patients were recruited. Nine patients following medical dietary treatment (MT), based on a hypocaloric diet combined with moderate physical activity three times per week, in addition to a monthly doctor visit for 12 months. In addition, five recruited patients underwent Roux-en-Y gastric bypass (RYGB) and five sleeve gastrectomy (SG) with a 5-trocar technique as described previously (Escalona et al., 2007; Boza et al., 2012). Patients undergoing either RYBG or SG received nutrient supplementation during the follow-up period such as multivitamin supplements, iron, vitamin B12 and calcium, and were instructed to follow a hypocaloric diet.

Anthropometric and clinical parameters were obtained in a clinical assessment. The patients were evaluated at three opportunities (baseline evaluation before medical intervertion, 6 and 12 months post treatment), which consisted of anthropometry measurements (weight and size), laboratory studies taking blood sample (lipids, HOMA and HbA1c). Patients were instructed to bring a homogenized fecal sample in a sterile container the day of the evaluation, which was transported in ice and immediately stored at −80 °C.

Descriptive statistical parameters (mean and standard deviation), and significant differences between clinical data were estimated by using a non-parametric unpaired Wilcoxon rank-sum test at 0.05 significance level. Mean differences between phylum changes were contrasted using a parametric unpaired t-test at 0.05 significance level. Pairwise Spearman Rank correlations between clinical parameters changes and bacterial variation were done using R statistical environment (R Core Team, 2013).

DNA isolation

Fecal samples were thawed and 150 mg were used for total DNA isolation using the ZR Fecal DNA MiniPrep kit (Zymo Research, USA) following manufacturer instructions and using a Disruptor Genie device (Scientific Industries, USA). Total DNA concentration was measured in a NanoDrop 2000c device (Thermo Fisher Scientific, USA).

Analysis of gut microbiota by 16S rRNA gene sequencing

Fecal DNA samples were diluted to 20 ng/µl in Nuclease-free water (IDT, USA) and submitted for Illumina MiSeq sequencing to Molecular Research DNA sequence services (MR-DNA, USA). The 16S rRNA gene V3–V4 variable region was amplified using the 341F and 785R primers (Klindworth et al., 2013), adding a barcode on the forward primer. The reaction was performed in 30 cycles using the HotStarTaq Plus Master Mix Kit (Qiagen, USA). After amplification, PCR products were checked in a 2% agarose gel. Multiple samples were pooled together and purified using calibrated Ampure XP beads (Agencourt Bioscience Corporation, USA). The pooled and purified combined PCR products were used to prepare a DNA library using TruSeq DNA LT Sample Prep Kit (Illumina, USA) following manufacturer instructions. Sequencing was performed using MiSeq platform (Illumina, USA) by paired-end sequencing.

The Quantitative Insights Into Microbial Ecology (QIIME, v1.9.1) software was used to analyze the 16S rRNA sequences (Caporaso et al., 2010; Navas-Molina et al., 2013). Briefly, chimera sequences were removed, then paired sequences joined and barcode was depleted. Operational taxonomic units (OTUs) were picked by closed reference command and defined by clustering at 1% divergence (99% similarity) using as reference the GreenGenes database (DeSantis et al., 2006; McDonald et al., 2012) release 05-2013. Low sequence counts were filtered from BIOM table using the minimum value of count/sample between all samples. To compare phylum level changes, OTUs belonging to phyla with less than 1% representation were removed. Alpha and Beta diversity were calculated using QIIME. BIOM OTU table and weighted Unifrac tables were exported from QIIME to R environment (R Core Team, 2013) for statistical analysis and figure representation. The raw data reads obtained from the MiSeq platform were stored in the SRA NCBI online public database with accession number SRP076859 (http://www.ncbi.nlm.nih.gov/sra/SRP076859).

qPCR amplification

In order to validate 16S rRNA gene sequencing results, quantitative PCR analysis was performed in fecal DNA samples using specific primers (Table S1) that were previously described (Rinttilä et al., 2004; Fierer & Jackson, 2005; Frank et al., 2007; Bacchetti De Gregoris et al., 2011). Amplification and detection were carried out in a StepOnePlus equipment (Applied Biosystems, USA), using 96-well optical plates MicroAmp Fast Optical (ThermoFisher, USA), filled with a mixture containing for each well 5 µl of PowerUp SYBR Master Mix or Fast SYBR Green Master Mix (Applied Biosystems, USA), 0.3 µM of each primer (0.3 µl each), 4.4 µl nuclease-free water (IDT, USA) and 1 µl of DNA previously diluted to 10 ng/µl. DNA samples were amplified with an initial hold of 50 °C for 2 min and a polymerase activation step of 95 °C for 2 min for PowerUp SYBR Master Mix, or 95 °C for 20 s for Fast SYBR Green Master Mix, followed by 40 cycles of a denaturation at 95 °C for 3 s and 62 °C for 30 s annealing and elongation. To verify a single amplification peak, a melting curve was performed by incrementing the temperature from 62 °C to 95 °C. All the samples were amplified in triplicate, and to correct primer efficiency, each plate contained a standard curve with ten-fold dilutions of genomic DNA of one species of the corresponding phylum, starting from 10 ng of DNA of the following microorganisms: Firmicutes: Lactobacillus acidophilus ATCC 4356; Bacteroides: B. dorei CL03T12C01, HM-718; NIAID, NIH; Actinobacteria: Bifidobacterium longum subsp. infantis ATCC 15697; Proteobacteria: Escherichia coli K12. The 16S rRNA gene copy number for each phylum in each sample was estimated from the corresponding standard curve and adjusted by the average genome 16S rRNA gene copy number of bacteria. To convert bacterial DNA amounts into copy number, the following equation was applied: ${\displaystyle Copynumber16SrRNAgene=\frac{Avogadro\mathrm{No}\left({\mathrm{mol}}^{-1}\right)\ast DNAquantity\left(\mathrm{g}\right)\ast Genome16Scopynumber}{Genome\left(\mathrm{pb}\right)\ast 660\left(\frac{\mathrm{g}}{\mathrm{mol}}\right)}.}$

Effect of obesity treatments on clinical parameters

In this study we compared the impact of three treatments for obesity on the gut microbiota, which included 19 patients who have undergone SG or RYGB, or received MT. Clinical data and fecal samples were collected before the medical intervention, and after six and twelve months post-treatment.

Six months after surgery, both RYGB and SG patients showed a marked weight loss, accompanied by significantly lower BMI values (28% and 29% decrease) and significant waist and hip perimeter reduction (Table 1 & Table S2). In contrast, patients on MT showed similar weight after six months of treatment with no major improvements in their anthropometric parameters (2.8% reduction in BMI).

Other specific signatures were found in the clinical markers of this cohort (Table 1). In RYGB and SG treatments, we observed a decrease in insulin levels and glycemia, however these levels maintained in a normal range. Importantly, surgery led to a significant improvement in insulin sensitivity (HOMA < 3) after treatment. No statistical significant changes were observed in lipid metabolism markers, nevertheless cholesterol and triglycerides tended to decrease in RYGB after surgery (Table S2). In contrast to the above observations, none of these clinical parameters improved in MT patients. The general clinical observations presented here appeared stable after 12 months of treatment (Table S2).

Obesity treatments induce global microbiota changes

Next, we determined the composition of the gut microbiota of these patients before and six months after each treatment. The hypervariable regions V3–V4 of the 16S rRNA gene in each fecal DNA sample was amplified by PCR and the products were sequenced by MiSeq Illumina platform. We obtained over 2 million reads in 38 samples, which includes DNA from before and after 6 months of obesity treatment. Rarefaction curves of the number of OTUs at different sequencing depths were obtained for each DNA sample (Fig. 1A), and they indicated saturation near 25,000 sequences. To evaluate microbiota composition of each patient, beta diversity was calculated, which estimates the degree of similarity of each sample in terms of their microorganism composition. Weighted Unifrac metric indicated that microbiota composition of MT patients was similar between 0 and 6 months (Fig. 1B). In contrast, the microbiota of patients undergoing RYGB cluster together at the beginning of the study, but after treatment its composition was divergent (Fig. 1C). Conversely, microbiota composition in SG group was divergent before the medical intervention but after treatment the compositions of three patients clustered closely (Fig. 1D).

On average, more than 99% of sequences aligned to the phyla Firmicutes, Bacteroidetes, Actinobacteria and Proteobacteria, with the first two being dominant in these samples (Fig. 2A). Changes in the representation of each phylum in patients in each treatment were then expressed as the ratio of the relative abundance after the intervention compared to pre-treatment (Figs. 2B–2D). In MT patients, no major changes were observed in any of these bacterial groups after six months, consistent with their poor response to the treatment. In contrast, important changes were observed in RYGB and SG patients. In the RYGB group, both Bacteroidetes and Proteobacteria increased in abundance (Fig. 2D), while in SG patients we observed an increase in Proteobacteria, but a decrease in Bacteroidetes (Fig. 2C). Firmicutes abundance was mostly unaffected in these patients. These different changes in microbiota also caused an increase in the Bacteroides/Firmicutes ratio in SG patients (Fig. 2E), and conversely a strong decrease in the RYGB group, consistent with previous reports (Li et al., 2011; Kong et al., 2013; Walters, Xu & Knight, 2014). Abundance and microbial composition obtained from 16S rRNA gene sequencing at genus-species level is summarized in Table S3 for each group.

In order to validate the 16S rRNA gene sequencing results, the composition of the microbiota of these patients was also evaluated by qPCR, quantifying the abundance of the phyla Actinobacteria, Firmicutes and Bacteroides, and the order Enterobacteriales. These results showed a good correlation with data obtained by MiSeq sequencing. MT patients did not show major microbiota changes after six months of treatment (Fig. 3A). In RYGB patients, we observed an increase in Bacteroidetes and a decrease in Firmicutes abundance (Fig. 3B). Actinobacteria also appeared in higher amounts in these patients. In contrast, in SG patients the abundance of Bacteroidetes decreased, while the Firmicutes and Enterobacteriales proportions increased (Fig. 3C). This in turn caused a reduction in the ratio Bacteroidetes/Firmicutes (Fig. 3D), similar to our previous observations by 16S rRNA gene sequencing (Fig. 2D). qPCR raw data and gene copy number for each group are included in Table S4, while phylum relative abundance measured by qPCR and 16S rRNA sequencing is presented in Table S5.

A correlation analysis comparing the abundance of each microbiota across all patients is represented in Fig. 4. This analysis shows that in general the initial microbiota abundance of most patients was uniformly correlated (R Spearman > 7). However, after six months of each treatment, microbiota abundance of RYGB subjects was very heterogeneous, in contrast to MT and SG subjects which grouped according to their respective treatment (Fig. 4B).

Microbial signatures for each obesity treatment

The most abundant genera or species that displayed the highest change in abundance after each obesity treatment (determined by 16S rRNA sequencing) is shown in Table 2. These bacteria mostly belong to the Firmicutes phylum. Changes in the microbiota of MT patients were less pronounced compared to the surgical treatments. In the RYGB group, Succiniclastum sp. displayed a significant increase, as well as a few Bacteroides and Citrobacter species (Table 2). A Bulleidia OTU displayed the highest fold change in the SGgroup, followed by Escherichia coli and Akkermansia muciniphila OTUs. Other interesting signatures observed include an increase of Streptococcus luteciae in both RYGB and SG patients. Species such as Bacteroides eggerthii, Bacteroides coprophilus and Lactobacillales sp. showed an important increase in RYGB subjects, but a marked down-representation after six months in the SG group.

Associations between microbiota changes and clinical markers

Finally, we evaluated whether changes in microbiota composition in each group correlated with clinical and anthropometric data collected. Certain positive and negative associations between microbial changes and clinical markers after six months of each treatment were found, using Pairwise Spearman Rank correlation analysis (Fig. 5). Only a few of them were statistically significant (p-value < 0.05). In MT patients, no significant correlations were observed. Changes in Proteobacteria positively correlated with weight loss and bilirubin levels in RYGB patients. In addition, a negative association between Actinobacteria and liver markers such as Gamma-glutamyl transpeptidase and alkaline phosphatase was obtained. In the SG group, hip and waist perimeter reduction correlated negatively with Proteobacteria and Actinobacteria, and positively with Firmicutes abundance. In this group most significant clinical changes were associated with the reduction in Bacteroidetes, such as liver markers and HbA1c.