Introduction

Prevalence of obesity has been rapidly increasing in the United States as well as other countries in the last 20 years, with 48.2% of adults now classified as either overweight or obese [1]. Bariatric surgery has the highest effect for substantial weight loss compared to behavioral modifications. However, the recurrence of weight gain has been shown to be as high as 50% in patients who underwent a gastric bypass [2]. As such, extensive research and public health initiatives have been aimed to identify new strategies in preventing post-operative recurrence. The altered gastrointestinal anatomy and altered absorptive processes following surgery may positively influence the gut microbiome, which is significantly involved in fat and energy regulation [3, 4].

Milkfat likely reduces gut dysbiosis and facilitates many of the essential metabolic processes associated with weight loss [5]. While support for dairy consumption in controlling weight has been controversial, there is a paucity of research on the effects of milkfat and sustained weight loss via changes in gut microbiome [6]. We hypothesized that dairy consumption may have an important role in post-operative outcomes. With the rates of performed bariatric surgical procedures increasing each year, understanding the specific gut flora changes before and after surgery and its relationship to milkfat consumption may yield important strategies for managing sustained weight-loss after bariatric surgery. Herein, we profile microbiome changes in patients who underwent bariatric surgery and had varying degrees of milkfat consumption. We also analyzed patient samples to see if bariatric surgery and more specifically consumption of high milk-fat, leads to changes in saturated, monounsaturated and polyunsaturated fatty acids. Our results will also be useful for the management of obesity in general population.

Materials and Methods

Study Design

Patients who were at least 18 years of age and scheduled for bariatric surgery at Cooper University hospital, Camden, New Jersey, from April to December 2019 were recruited. Surveys about demographics, medical history including antibiotic use, and baseline milkfat consumption were collected at the time of recruitment. Patients were asked to provide stool samples and dairy intake surveys (Supplemental file) 2 weeks before surgery, at time of the surgery, and at 1, 3, and 6 months postop. Baseline patient characteristics at enrollment 3–6 weeks prior to surgery are shown in Table 1. Of the 9 patients studied, 1 was diabetic and 3 were pre-diabetic. Of these 4 patients, 1 patient was receiving treatment for impaired glucose tolerance (BS139—prediabetic) prior to surgery and treatment was discontinued following surgery. The remaining patients were not on any diabetes medications.

Table 1 Baseline patient characteristics at enrollment 3–6 weeks prior to surgery
Full size table

Sample Collection

Patients were given detailed instructions about sample collection and were provided with a kit that included all the necessary materials such as collection tube, gloves, toilet hat, labels, zip lock bags etc. Stool samples were collected at home in Stool Nucleic Acid Collection and Preservation Tubes (Norgen, Ontario, Canada). The stabilizer allows easy storage and transport of samples to laboratory. Upon receipt, samples were stored at − 80 °C for long term storage and subsequent analysis.

16S rRNA Gene Sequencing

Total genomic bacterial DNA was extracted from fecal samples using the Qiagen DNeasy PowerSoil HTP Kit (Qiagen, California, USA). For high throughput sequencing, 515F–806R Golay barcodes were used to target the hypervariable V4 region of the 16S rRNA gene, which is highly conserved and ideal for gut microbiome analysis. Marker genes in isolated DNA were PCR-amplified using GoTaq Master Mix (Promega, Wisconsin, USA) [7]. 16S rRNA gene sequencing was performed on normalized pooled amplicons using the Illumina MiSeq System using a MiSeq reagent kit v2.

Analysis of Data from 16S rRNA Gene Sequencing

Analysis was performed using Python-based packages (Python 3.6.12), Quantitative Insights Into Microbial Ecology (QIIME) 2 2020.11, and Linear Discriminant Analysis using LEfSe10 [8, 9]. Sequences were de-multiplexed, then filtered and clustered into sub-operational taxonomic units (sOTUs) using QIIME 2 DADA2 [10]. Phylogenetic tree was created using SATé-enabled phylogenetic placement (SEPP) via Qiime 2 [11]. Taxonomy was assigned using a naïve-Bayes classifier based on the latest SILVA 16S rRNA gene database (March 2021) via the QIIME 2 interface. Additional Python packages (SciPy.stats, Scikit-bio) were used for statistical tests on QIIME 2-generated data. Inter-class comparisons were made based on attributes of our population. Alpha analyses were performed in QIIME 2, rarefied to an even sampling depth of 22,400 [12].

Alpha diversity was assessed by evenness analysis for factors, time-points, age, number of people at home, initial BMI, frequency of smoking, drinks/week, dairy/week (milk, yogurt, cottage cheese, sour cream, kefir, cheese, butter and margarine, ice cream and similar foods, using butter for cooking, other milk products and total milk consumption). To identify variables on which later analyses could focus, differences in alpha diversity were assessed by Kruskal–Wallis non-parametric rank test for categorical variables for Spearman for continuous or discrete variables across the dataset. To address non-independence of samples, we identified the variables that were most strongly associated with alpha diversity from the exploratory analysis and then performed more conservative tests that took non-independence into account. For categorical variables that did not change across time-points within a participant (such as use of butter while cooking during the study), a Kruskal–Wallis test was performed on the average alpha diversity values for each participant. Likewise, for discrete and continuous variables that did not change over time, Spearman correlation tests were performed on each participant’s average alpha diversity value. For discrete and continuous variables that changed within each participant, such as time-point, a repeated measures correlation test was performed on the ranks of Faith’s phylogenetic diversity, since assumptions of normality were not met. Kruskal–Wallis and Spearman correlation tests were performed via the Python package SciPy.stats, and the repeated measures correlation was performed via the package Pingouin.

Comparisons of relative abundances of microbial taxa between groups were performed using Linear Discriminant Analysis (LDA) Effect Size (LEfSe) with the Huttenhower group online interface [13]. Taxa with an LDA value of 2.0 or greater and a two-tailed p value < 0.05 with Kruskal–Wallis and pairwise Wilcoxon analyses were considered significantly enriched. Volatility plot was created in Qiime2 via longitudinal plug in.

Non-targeted fatty acid analysis

Analysis of fecal samples was performed using gas chromatography–mass spectrometry (GC–MS) by Metabolon, Inc. (Durham, NC, USA). Metabolon TAM112-002 measures the total fecal content of 30 fatty acids after conversion into their corresponding fatty acid methyl esters (FAMEs). The measured concentrations are provided in weight corrected μg/g of sample. Values below the lower limit of quantification were treated as a concentration of 0.001 μg/mL. In this method, homogenized stool samples were weighed out in 100 mg aliquots in test tubes. Liquid/liquid extraction was performed to extract fatty acids and remove the nucleic acid preservative. A 250 μL aliquot of each extract was transferred to a clean analysis tube. The solvent was removed by evaporation under a stream of nitrogen. Internal standard solution was added to the dried sample extracts, quality controls (QCs), and calibration standards. The solvent was again removed by evaporation under a stream of nitrogen. The dried samples and QCs were subjected to methylation/transmethylation with methanol/sulfuric acid, resulting in the formation of the corresponding methyl esters (FAME) of free fatty acids and conjugated fatty acids. The reaction mixture was neutralized and extracted with hexanes. An aliquot of the hexanes layer was injected onto a 7890A/5975C GC/MS system (Agilent Technologies, Santa Clara, CA, USA). Mass spectrometric analysis was performed in the single ion monitoring (SIM) positive mode with electron ionization. Quantitation was performed using both linear and quadratic regression analysis generated from fortified calibration standards prepared immediately prior to each run. Raw data were collected and processed using Agilent MassHunter GC/MS Acquisition B.07.04.2260 and Agilent MassHunter Workstation Software Quantitative Analysis for GC/MS B.09.00/ Build 9.0.647.0. Analysis of data based on T-test/ANOVA and creation of a hierarchical clustering heat map was performed using MetaboAnalyst 5.0.

Calculation of Milkfat Units

Milkfat consumption was calculated for each subject at each time-point and in total in milkfat units (MFU) using the formula: Σ[(milkfat % of product)(frequency of product consumed)]. The formula was based on the type of dairy product, the specific milkfat percentage of the product, and frequency of consumption. The specific milkfat percentage of each product was obtained from the U.S. Food and Drug Administration’s website. A cutoff of 250 MFU was used to separate low and high milkfat consumers. Our calculation of the milkfat units thus takes into account the different percentages of milkfat in various dairy products (fat-free, skimmed etc.) as well as the frequency of consumption per week.

Results

Gut Microbiome Changes with Respect to Bariatric Surgery and Low/High Milkfat Consumption

The median age of the study cohort (n = 9) was 45.5 years old (range: 30–66). Four patients were male. Five patients identified as White/Non-Hispanic, one as African American, and three as Other. Six patients (BS124, BS135, BS138, BS139, BS142, BS143) were low milkfat consumers and three (BS131, BS140, and BS145) were high milkfat consumers. Table 2 shows clinical markers at the time of surgery, 6 months after surgery, and one year after surgery. These clinical markers include changes in weight, BMI, Hemoglobin A1C, systolic blood pressure, diastolic blood pressure, AST, ALT, alkaline phosphatase, and total bilirubin. The average percent change in weight 1-year after surgery for low milkfat consumers (n = 6) was 28.1% while high milkfat consumers (n = 3) demonstrated an average percent change of 28.05%. The median difference in BMI was 11.6 kg/m2 for low milkfat consumers and 11.5 kg/m2 for high milkfat consumers. The HbA1C status at surgery and 1 year later did not differ between the two groups.

Table 2 Laboratory work-up of patients before and after surgery
Full size table

Sequencing analysis and subsequent linear discriminant analysis effect size showed an overall post-surgery enrichment of Roseburia and Christensenellaceae (Fig. 1A). Similar analysis showed that the genus Lachnospira was abundant in low milkfat consumers. The genera Lachnospiraceae, Candidatus, Sellimonas, Faecalitalea, and Blautia along with the species Bacteroides were all enriched in high milkfat consumers (Fig. 1B). Additionally, there was an overall increase in the Bacteroides species in all high milkfat consumers. Volatility plot (Fig. 2) shows how taxa changed over time. Faecalibacterium significantly increased more than any other bacteria after bariatric surgery in patients 135, 139 and 142, over the duration of the study. Interestingly, as mentioned above, these three patients were low milkfat consumers.

Fig. 1
figure 1

Linear discriminant analysis effect size (LEfSe). LEfSe scores via Kruskal–Wallis and Wilcoxon tests with two-tailed α = 0.05, for taxa that were differentially distributed in (A) bariatric patients before (n = 9) versus after (n = 9) surgery (6 months) and (B) bariatric patients that had low (n = 6) versus high (n = 3) milkfat intake. Negative values represent taxa that were abundant before surgery, whereas positive values represent taxa that were abundant after surgery (A). Negative values represent taxa that were abundant in low milkfat intake group, whereas positive values represent taxa that were abundant in high milkfat intake group (B). Phyla corresponding to different colors are noted in the respective panels

Full size image
Fig. 2
figure 2

Volatility plot. Volatility plot was created in Qiime2 via longitudinal plug in. Lines represent abundance of Faecalibacterium at different time points indicated (2 weeks before surgery; at the time of the surgery; 1 month post-surgery, 3 months post-surgery, 6 months post-surgery) in nine participating patients

Full size image

Out of the dairy products tested, the frequency of use of butter for cooking was associated with decreased alpha diversity in patients (correlation =  − 0.68, p-value = 0.042) (Fig. 3A). Usage of butter was associated with decreased alpha diversity in patients (p-value = 0.038) (Fig. 3B).

Fig. 3
figure 3

Alpha diversity assessment by (A) the frequency of use of butter for cooking by evenness-vector-correlation analysis and (B) the usage of butter by evenness-vector-significance analysis

Full size image

Fecal Fatty Acid Analysis

To evaluate the influence of intake of dairy fat on fatty acid composition of the patients who underwent bariatric surgery, we performed non-targeted metabolomics analysis of a panel of 30 fatty acids including long chain fatty acids (LCFAs), monounsaturated fatty acids (MUFAs) and polyunsaturated fatty acids (PUFAs). Analysis of data based on T-test/ANOVA was performed using MetaboAnalyst 5.0. Hierarchical clustering heatmap presented in Fig. 4 provides visualization of the metabolome data. The heat map shows that myristic acid (SFA: saturated fatty acid) and/or myristoleic acid (MUFA) was observed in high amounts in patients BS124, BS135, BS138 and BS139. Patient BS124 also showed high amount of cis-13–16-docosadienoic acid, cis-11,14-eicosadienoic acid (20:2n6) and cis-11-eicosaenoic acid (20:1n9) and high amount of two more SFAs: arachidic and stearic acid. On the other hand, omega-6 PUFAs such as linoleic acid, gamma-linolenic acid and osbond acid were in high amount in patients BS124, BS131, BS139 and BS145. High concentration of an omega-3-PUFA, alpha-linolenic acid, was observed in patient BS131.

Fig. 4
figure 4

Hierarchical clustering heatmap of fecal fatty acids. Analysis is based on T-test/ANOVA. Each colored cell on the map corresponds to a concentration value in the data table, with patient samples in columns and analyzed fatty acids in rows. Class values (0–4) correspond to the five time-points at which samples were collected (2 weeks before surgery, at the time of the surgery, and at 1, 3, and 6 months postop, respectively). A diverging color palette is used with hierarchy of colors from the dark brown (highest values) to dark blue (lowest values). Fatty acids analyzed: myristic acid (14:0), pentadecanoic acid (15:0), palmitic acid (16:0), heptadecanoic acid (17:0), stearic acid (18:0), arachidic acid (20:0), behenic acid (22:0), lignoceric acid (24:0), myristoleic acid (14:1n5), palmitoleic acid (16:1n7), vaccenic acid (18:1n7), oleic acid (18:1n9), cis-11-eicosaenoic acid (20:1n9), erucic acid (22:1n9), nervonic acid (24:1n9), mead acid (20:3n9), linoleic acid (18:2n6), gamma-linolenic acid (18:3n6), dihomo-gamma-linolenic acid (20:3n6), arachidonic acid (20:4n6), adrenic acid (22:4n6), osbond acid (22:5n6), cis-11,14-eicosadienoic acid (20:2n6), cis-13–16-docosadienoic acid (22:2n6), alpha-linolenic acid (18:3n3), stearidonic acid (18:4n3), eicosatetraenoic acid (ETA) (20:4n3), eicosapentaenoic acid (EPA) (20:5n3), docosapentaenoic acid (22:5n3), docosahexaenoic acid (DHA) (22:6n3)

Full size image

Discussion

Obesity and type 2 diabetes are diseases that need a combined treatment such as bariatric/metabolic surgery[14]. In this study, we assessed the association milkfat may have with microbiome changes. It is important to note that other dietary components and lifestyle changes likely contributed to differences observed in the microbiome. Additionally, hormonal changes such as differences in the leptin-adiponectin ratio axis and serum amyloid A concentrations may contribute to the changes in microbiome profiles [15, 16]. Our pilot study however, introduces new insight on the association of milkfat with postoperative bariatric microbiome profiles.

The sample size of our pilot study although small was comparable to other studies focused on post-bariatric surgery changes in human gut microbiome. For example, one study included before and 3 months post-bariatric surgery samples from total of six patients [17], while another study included eight gastric surgery patients and four medical weight loss patients [18]. The small sample size of our study is a limitation, but we did find several important observations that will help further direct research on this topic. The post-bariatric surgery increase seen in our study for Roseburia, Christensenellaceae, and Blautia is encouraging. Roseburia is a bacterium associated with weight loss and diabetes remission [19]. Christensenellaceae is inversely related to host body mass index [20]. The genus Blautia was observed as the only gut microbe that was inversely associated with visceral fat, independent of sex in a weight loss study in Japan [21]. Blautia is proposed to decrease obesity by producing butyric acid and acetic acid, which regulate G-coupled receptors (GPR) 41 and 43. While the function of these receptors is debated, studies on knockout mice implicate GPR41 and GPR43 in obesity [22]. Additionally, we observed an overall increase in the Bacteroides species in all high milkfat consumers. An overall increase in Bacteroides is negatively associated with obesity and promotes health and long-term weight loss [23, 24]. This suggests that the gut microbiome changes observed in our study in bariatric surgery patients with higher milkfat consumption are beneficial. Although high milkfat was seen to have a beneficial association, we observed reduced gut microbiome diversity when butter was used. Reduced diversity of gut microbiome is associated with obesity and diseases such as inflammatory bowel disease, cardiovascular disease, and asthma. Our results thus suggest that the use of butter did not have a beneficial association. Unlike most other dairy products, butter does not contain any probiotics, which may explain the lack of diversity associated with its use. Thus, this dairy product may not be associated with promoting a healthy microbiome even though it is generally associated with high milkfat content.

We observed that Faecalibacterium significantly increased more than any other bacteria after bariatric surgery in patients 135, 139 and 142, over the duration of the study. These were low-milk-fat consumers. A study showed that bariatric surgery leads to an increase in F. prausnitzii [18]. The authors discussed this as a beneficial bacterium. In contrast, another study reported significantly higher levels of F. prausnitzii in the obese compared with the non-obese participants [25]. The role of Faecalibacterium in obesity thus remains to be fully elucidated. It was suggested that, these discrepancies emphasize that population characteristics, age and diets should be taken into consideration before drawing conclusion regarding the role of this bacterium in obesity [26].

Role of short chain fatty acids in health and disease is still controversial. One study showed that bariatric surgery can lead to reduction in short chain fatty acids such as acetate, butyrate and propionate in patient fecal samples [27]. These are perceived to be linked to obesity. Another study reported that whole milk supplementation can alter the intestinal gut microbiome composition, but does not lead to significant changes to fecal short chain fatty acids [28]. A study using rat model showed that incorporation of dairy lipids increased tissue levels of long chain omega-3-fatty acids [29]. In the present study, we thus explored if bariatric surgery and more specifically consumption of high milk-fat, leads to changes in the fecal patterns of SFA, MFA and PUFA. It is interesting to note that the three patients BS124, BS138, and BS139 showing higher concentration of saturated fatty acids such as myristic acid were low-milk-fat consumers. On the other hand, BS131, a high-milk-fat consumer, showed very high concentration of beneficial omega-3-PUFA. Direct relevance of these observations with respect to dairy intake remains to be seen.

It would have been informative to carry out gut microbiome analysis of samples and collection of dietary surveys one year post-surgery as well. However, our area was one of the hardest hit areas by the COVID-19 pandemic. One year post-surgery time-points of all the patients who participated in this study were after the onset of pandemic in our area. Pandemic led to severe restrictions on research-related interactions with patients and restrictions on non-emergency in-person patient visits, which in turn posed challenges for collection of one year post-surgery samples and surveys. This will be considered for future studies. Corroborating clinical data with milkfat usage did not show clear clinical relationships in the course of one year; however, this should be interpreted with caution due limiting factors such as (i) small sample size of the pilot study, (ii) lack of information if patients had continued these dietary practices for one year, (iii) lack of information about compounding factors such as stress and/or changes if any, imposed on diet and other lifestyle aspects due to the pandemic.

Conclusion

While bariatric surgery is considered an efficient intervention to achieve major and rapid weight reduction, a high percentage of patients experience weight relapse post-operatively. Dietary interventions aimed at manipulating microbiome composition may complement the long-term efficacy of bariatric procedures. The findings in the present study suggest that incorporating dairy products in post-operative dietary plans may help cultivate a gut microbiome that is effective in regulating fat storage as well as digesting beneficial metabolites.