Simulating the Post-gastric Bypass Intestinal Microenvironment Uncovers a Barrier-Stabilizing Role for FXR.

Regional changes to the intestinal microenvironment brought about by Roux-en-Y gastric bypass (RYGB) surgery may contribute to some of its potent systemic metabolic benefits through favorably regulating various local cellular processes. Here, we show that the intestinal contents of RYGB-operated compared with sham-operated rats region-dependently confer superior glycemic control to recipient germ-free mice in association with suppression of endotoxemia. Correspondingly, they had direct barrier-stabilizing effects on an intestinal epithelial cell line which, bile-exposed intestinal contents, were partly farnesoid X receptor (FXR)-dependent. Further, circulating fibroblast growth factor 19 levels, a readout of intestinal FXR activation, negatively correlated with endotoxemia severity in longitudinal cohort of RYGB patients. These findings suggest that various host- and/or microbiota-derived luminal factors region-specifically and synergistically stabilize the intestinal epithelial barrier following RYGB through FXR signaling, which could potentially be leveraged to better treat endotoxemia-induced insulin resistance in obesity in a non-invasive and more targeted manner.

Introduction

No currently available, non-surgical treatment intervention for morbid obesity approaches the clinical efficacy of metabolic surgeries such as Roux-en-Y gastric bypass (RYGB) (Ruban et al., 2019). By creating a small gastric pouch and adjoining it to the transected mid-jejunum, in combination with adjoining the freed upper to the lower jejunum, RYGB typically achieves 30–40% weight loss, which is sustained in the long-term (Adams et al., 2016; Sjostrom et al., 2004, 2007), and, in most cases, rapid remission of type 2 diabetes (Mingrone et al., 2012, 2015; Panunzi et al., 2016). Numerous surgery-specific mechanisms have been proposed for this remarkable restoration of glycemic control post-operatively, including loss of upper jejunal glucose absorption from ingested food by the apical sodium-glucose co-transporters 1/3 (Baud et al., 2016; Pal et al., 2019; Stearns et al., 2009), increased lower jejunal glucose extraction from the circulation by the basolateral glucose transporter 1 (Cavin et al., 2016; Makinen et al., 2015; Pal et al., 2019; Saeidi et al., 2013), as well as gut hormone-mediated improvements in pancreatic islet cell function (Guida et al., 2019; Jorgensen et al., 2013; Ramracheya et al., 2016; Salehi et al., 2014). In addition, circulating bile acid levels steeply rise (Ahlin et al., 2019; Bhutta et al., 2015; Chen et al., 2019; Ferrannini et al., 2015; Gerhard et al., 2013; Kohli et al., 2013; Pournaras et al., 2012; Spinelli et al., 2016; Yan et al., 2019; Zhai et al., 2018) and circulating branched chain amino acid levels decline (Laferrere et al., 2011; Lips et al., 2014; Yoshino et al., 2020) in association with favorable shifts in the intestinal microbiota (Arora et al., 2017; Liou et al., 2013; Tremaroli et al., 2015). Despite significant progress in the field, further understanding precisely how these changes add or contribute to the restored hepatic and peripheral insulin sensitivities that appear to develop sequentially following RYGB (Albers et al., 2015; Bikman et al., 2008; Bojsen-Moller et al., 2014; Campos et al., 2010; Chambers et al., 2011; Gancheva et al., 2019; He et al., 2014; Lima et al., 2010; Martinussen et al., 2015; Meirelles et al., 2009; Steven et al., 2016; Yoshino et al., 2020) will invariably inform the design of more effective pharmacotherapies against insulin resistance and type 2 diabetes.

The intestinal epithelial barrier (IEB) comprises a single continuous layer of cells lining the gastrointestinal tract and serves as the main line of defense against ingested toxins and potentially pathogenic microbes residing in the gastrointestinal wall and lumen (Vancamelbeke and Vermeire, 2017). Tight junctions, adherens junctions, and desmosomes are multi-protein complexes located to the cell membrane of enterocytes that together form a homophilic, tripartite seal from the apical to the basolateral sides, respectively (Garcia et al., 2018; Heinemann and Schuetz, 2019; Schlegel et al., 2020). Since the original findings in mice (Cani et al., 2007, 2008), evidence has mounted that obesity is associated with IEB breakdown primarily due to unfavorable shifts in the intestinal microbiota caused by frequent overconsumption of a high-fat diet (Chakaroun et al., 2020). Consequently, bacteria and their components such as lipopolysaccharide (LPS) leak into the circulation and accumulate in metabolic tissues, where they induce a chronic, low-grade state of inflammation leading to local and systemic insulin resistance (Chakaroun et al., 2020). Stabilizing the IEB in obesity either by small molecule drugs (Lin et al., 2019; Luck et al., 2015; Natividad et al., 2018) or by microbiota transplant (Depommier et al., 2019; Natividad et al., 2018; Wang et al., 2019b) thus shows promise in attenuating endotoxemia-induced insulin resistance.

Whether RYGB stabilizes the IEB to impact on glycemic control has been addressed in several clinical and pre-clinical studies. Early findings from pre-diabetic/diabetic patients revealed that RYGB reduces plasma LPS/LPS binding protein (LBP) levels to those of healthy lean individuals associated with a reduction in fasting plasma glucose, glycated hemoglobin (HbA1c), and insulin levels (Monte et al., 2012; Troseid et al., 2013; Yang et al., 2014). Similarly, RYGB has been shown to reduce intestinal permeability in diet-induced obese mice (Steensels et al., 2017) and rats (Wang et al., 2019a), as well as to normalize plasma LPS levels and glycemic status in streptozotocin-induced diabetic (Wu et al., 2019) and genetically obese and insulin-resistant Zucker diabetic fatty (ZDF) (Guo et al., 2019) rats. Collectively, these clinical and pre-clinical findings strongly suggest that at the later stages post-operatively (Guo et al., 2019; Monte et al., 2012; Steensels et al., 2017; Troseid et al., 2013; Wang et al., 2019a; Wu et al., 2019; Yang et al., 2014), RYGB improves glycemic control by stabilizing the IEB and attenuating endotoxemia-induced insulin resistance; however, the underlying molecular and cellular mechanisms remain poorly understood. We previously found that the improved oral glucose tolerance in genetically obese and glucose-intolerant Zucker fatty (ZF) rats following RYGB compared with sham surgery can be transmitted through their cecal, but not their ileal, contents to recipient germ-free mice (Arora et al., 2017). We therefore hypothesized that soluble factors within the intestinal contents of RYGB-operated rats region-specifically regulate IEB stability to exert distinct effects on endotoxemia and glycemic control.

Results

RYGB Improves Metabolic Parameters and Attenuates Endotoxemia in Diet-Induced Obese Rats

We first metabolically phenotyped a cohort of adult male Wistar rats made obese on a high-fat diet and then subjected to either RYGB or sham surgeries (Figure S1). Post-operatively, animals were given simultaneous free access to both high-fat and low-fat diets to model the real-world situation in humans and assess changes in food choice (Dischinger et al., 2019; Hankir et al., 2017) (Figure 1A). In line with previous findings (Dischinger et al., 2019; Hankir et al., 2017), RYGB-operated compared with sham-operated rats consumed less food over the 3-week recording period (1318 ± 90.6 kcal vs. 1889 ± 52.2 kcal, respectively, P < 0.0001; Figure 1B). They also progressively obtained more of their daily calories from the low-fat diet rather than from the high-fat diet, unlike sham-operated rats (Figure 1C). Accordingly, RYGB-operated rats weighed considerably less than their sham-operated counterparts at the time of sacrifice 6 weeks after surgeries (451.9 ± 12.5 g vs. 592.6 ± 13.9 g, respectively, P < 0.0001; Figure 1D), which equated to a 23.7 ± 2.1% difference (Figure 1E), and had lower fasting plasma insulin (Figure 1F) and LPS (Figure 1G) levels. Together, these findings provide confidence that our rat model closely matches the clinical features of RYGB by improving eating behavior and reducing body weight in association with attenuated insulinemia and endotoxemia.

Figure 1

RYGB Improves Metabolic Parameters and Attenuates Endotoxemia in Diet-Induced Obese Wistar Rats

(A) Schematic diagram illustrating experiments performed on sham-operated (Sham) and RYGB-operated (RYGB) rats.

(B and C) Cumulative total food intake (B) and average daily high-fat diet (HFD) preference (C) of Sham (n = 10) and RYGB (n = 15) rats over the 21-day recording period after surgeries. Dashed vertical line and gray panel in (B) indicate time of surgeries and 3-day recovery period when animals were placed on a liquid diet, respectively. Dashed horizontal line in (C) indicates equal preference for HFD and low-fat diet (LFD).

(D and E) Body weight of Sham (n = 10) and RYGB (n = 15) rats (D) and percentage (%) body weight change relative to Sham rats (E) over the 6-week recording period after surgeries. Dashed vertical lines in (D and E) indicate time of surgeries.

(F and G) Fasting plasma insulin (F) and lipopolysaccharide (LPS) (G) levels of a subset of Sham (n = 6) and RYGB (n = 5) rats at the time of sacrifice 6 weeks after surgeries.

Data in (B-G) are presented as mean ± SEM. Statistical significance was determined by two-way ANOVA (main effect of treatment) in (B-E) with Sidak's post-hoc test in (C) and two-tailed, unpaired t test in (F and G).

See also Figure S1.

The Jejunal and Colonic Contents of RYGB-Operated Rats Confer Superior Glycemic Control to Recipient Germ-free Mice in Association with Suppression of Endotoxemia

Next, to test our hypothesis that the intestinal contents region-specifically regulate glycemic control and endotoxemia following RYGB, we transferred the jejunal and colonic contents of RYGB-operated and sham-operated rats to germ-free mice and performed oral glucose tolerance tests (OGTTs) 2 weeks later in conjunction with serum LBP measurements, a reliable indicator of endotoxemia (Guerville et al., 2017; Moreno-Navarrete et al., 2012) (Figure 2A). We have previously shown that such transfer leads to colonization of the gastrointestinal tract of recipient germ-free mice with the microbiota present in the intestinal contents of rat donors (Arora et al., 2017). We found that the jejunal contents of RYGB-operated compared with sham-operated rats conferred marginally better blood glucose clearing capability to recipient germ-free mice (Figure 2B). This was largely driven by the lower peak blood glucose levels at 15 min after the oral glucose challenge and the greater drop in blood glucose levels 15 min later (Figure 2B). On the other hand, the colonic contents of RYGB-operated compared with sham-operated rats had a more pronounced beneficial effect on glycemic control in recipient germ-free mice as their blood glucose levels were lower at all time-points after the oral glucose challenge (Figure 2D). Notably, germ-free mice that had received the contents from both intestinal regions of RYGB-operated compared with sham-operated rats had lower serum LBP levels (Figures 2C and 2D). When merging the data from both surgical treatment groups (Li et al., 2019), regression analysis revealed a positive correlation between serum LBP levels in recipient germ-free mice and the area under the curve (AUC) from their OGTTs (Figure 2F). In contrast, the duodenal contents of RYGB-operated compared with sham-operated, ZF rats had no effect on oral glucose tolerance in recipient germ-free mice (Figure S2). These findings provide evidence that the altered jejunal and colonic, but not duodenal, contents following RYGB improve glycemic control in recipient germ-free mice in association with suppression of endotoxemia.

Figure 2

The Jejunal and Colonic Contents of RYGB-Operated Rats Confer Superior Glycemic Control to Recipient Germ-Free Mice in Association with Suppression of Endotoxemia

(A) Schematic diagram illustrating the jejunal and colonic contents transfer experiments and oral glucose tolerance tests (OGTTs) performed on recipient germ-free mice.

(B and C) Tail vein blood glucose concentrations following an oral glucose challenge with the associated area under the curve (AUC) (B) and serum LPS binding protein (LBP) concentrations (C) of germ-free mice that had received the pooled jejunal contents of Sham or RYGB rats. n = 4–6 mice/group.

(D and E) Tail vein blood glucose concentrations following an oral glucose challenge with the associated AUC (D) and serum LBP concentrations (E) of germ-free mice that had received the pooled colonic contents of Sham or RYGB rats. n = 3–4 mice/group.

(F) Pearson correlation between AUC from OGTTs and serum LBP of germ-free mice from (B, C, E, and F). Solid regression lines indicate least squares fit of data and dashed lines indicate 95% confidence intervals.

Data in (B-E) are presented as mean ± SEM. Statistical significance was determined by two-way ANOVA (main effect of treatment) in (B and D) and two-tailed, unpaired t test in (B–E and F).

See also Figure S2.

The Intestinal Contents of RYGB-Operated Rats Directly Stabilize Barrier Function and Structure in Caco-2 Cells

Having established that the intestinal contents of RYGB-operated compared with sham-operated rats favor the suppression of endotoxemia in recipient germ-free mice, we then proceeded to ask in more detail if they contain soluble factors that directly regulate IEB stability. To do so, we partially simulated the regional intestinal microenvironments of RYGB-operated and sham-operated rats by applying their duodenal, jejunal, and colonic contents onto the apical side of confluent Caco-2 cell monolayers, a human colonic cancer cell line differentiated into polarized enterocytes routinely used to test factors that regulate IEB function and structure (Lea, 2015) (Figure 3A). We note that because cell culture media was supplemented with antibiotics for these experiments (see Transparent Methods), a static simulation was achieved; that is, only microbiota-derived soluble factors already present in the intestinal contents upon collection could potentially exert effects in this setting.

Figure 3

The Intestinal Contents of RYGB-Operated Rats Directly Stabilize Barrier Function and Structure in Caco-2 Cells

(A) Schematic diagram illustrating intestinal microenvironment simulation experiments performed on Caco-2 cells.

(B–D) Permeability coefficients (PE) obtained from trans-well assays performed on confluent Caco-2 cell monolayers treated for 24 hr with the pooled duodenal (B), jejunal (C), and colonic (D) contents of Sham and RYGB rats. n = 5 cultures/group from 1 independent experiment.

(E–G) Trans-epithelial electrical resistance (TER) values obtained from low-frequency electrical impedance assays performed on confluent Caco-2 cell monolayers treated for 24 hr with the pooled duodenal (E), jejunal (F), and colonic (G) contents of Sham and RYGB rats. n = 16 cultures/group from 2 independent experiments.

(H–J) Immunoblot analysis of claudin-1 (Cldn-1), occludin (Ocldn), and desmoglein-2 (Dsg-2) protein expression in confluent Caco-2 cell monolayers treated for 24 hr with the pooled duodenal (H), jejunal (I), and colonic (J) contents of Sham and RYGB rats. n = 5–6 cultures/group from 1 independent experiment.

(K–M) Immunofluorescent analysis of Cldn-1, Ocldn, and Dsg-2 protein expression and distribution in confluent Caco-2 cell monolayers treated for 24 hr with the pooled duodenal (K), jejunal (L), and colonic (M) contents of Sham and RYGB rats. Scale bar: 25μm.

Data in (B-K) are presented as mean ± SEM. Statistical significance was determined by two-tailed, unpaired t test with Welch's correction in (B-D), two-way ANOVA (main effect of treatment) in (E-G), and two-tailed, unpaired t test in (H–M).

See also Figure S3.

For our functional experiments, we took two complementary approaches. We first performed measurements of 4 kDa fluorescein isothiocyanate (FITC)-dextran paracellular flux from the apical to the basolateral sides of Caco-2 cells in trans-well assays, followed by transepithelial electrical resistance (TER) measurements between Caco-2 cells in low-frequency electrical impedance assays. This revealed that the duodenal, jejunal, and colonic contents of RYGB-operated compared with sham-operated rats markedly decreased 4 kDa FITC-dextran passage across Caco-2 cells (reflected by decreased epithelial permeability coefficients - PE) (Figures 3B–3D). Accordingly, they all markedly increased TER values (Figures 3E–3G).

For our structural analyses, we also took two complementary approaches. We first performed immunoblot analysis to determine overall barrier protein expression in Caco-2 cells treated with the duodenal, jejunal, and colonic contents of RYGB-operated and sham-operated rats, followed by high-magnification immunofluorescent analysis to determine their cellular distribution in response to the same treatments (Tables S1 and S2 for antibody details). This revealed that the duodenal and colonic contents of RYGB-operated compared with sham-operated rats increased expression of the barrier-stabilizing tight junction protein claudin-1 (Garcia-Hernandez et al., 2017) and the barrier-stabilizing desmosomal protein desmoglein-2 (Schlegel et al., 2020) (Figures 3H and 3J). On the other hand, the jejunal contents of RYGB-operated compared with sham-operated rats increased expression of the barrier-stabilizing tight junction protein occludin (Garcia-Hernandez et al., 2017), as well as desmoglein-2 (Figure 3I). High-magnification immunofluorescent analysis was largely consistent with the immunoblot findings (Figures 3K-M), with the addition that the jejunal contents of RYGB-operated compared with sham-operated rats increased claudin-1 protein at the cell border of Caco-2 cells (Figure 3L).

In order to assess potential differences in the transcription of barrier proteins in Caco-2 cells in response to the duodenal, jejunal, and colonic contents of RYGB-operated compared with sham-operated rats, we performed real-time quantitative polymerase chain reaction (RT-qPCR) analysis (See Table S3 for primer sequences). This revealed no significant effect on CLDN1, OCLDN, and DSG2 mRNA expressions (Figure S3), suggesting that post-transcriptional processes mainly drive the observed regulation of barrier proteins under our in vitro experimental conditions. Such a dissociation between mRNA and protein expression levels of various barrier proteins has also been shown for intestinal enteroids derived from patients with inflammatory bowel disease compared with healthy patients (Meir et al., 2020). Together, these findings reveal that the duodenal, jejunal, and colonic contents following RYGB contain soluble factors that directly stabilize IEB function and structure. In conjunction with the germ-free mice findings, they further suggest dissociation between regulation of IEB stability and endotoxemia by the intestinal contents of RYGB-operated compared with sham-operated rats.

The Duodenal and Colonic, but not the Jejunal, Contents of RYGB-Operated Rats Stabilize Barrier Function Partly through FXR in Caco-2 Cells

Genetic inactivation of the bile acid receptors Takeda G-protein–coupled receptor 5 (TGR5) and farnesoid X receptor (FXR) leads to IEB breakdown in mice (Cipriani et al., 2011, Inagaki et al., 2005b), whereas TGR5 and FXR agonists, including the intestinally restricted FXR agonist fexaramine, have barrier-stabilizing properties in multiple settings (Cipriani et al., 2011; Fang et al., 2015; Hartmann et al., 2018; Sorribas et al., 2019; Verbeke et al., 2015). Furthermore, while TGR5 and/or FXR are required for the full beneficial effects on glycemic control of various metabolic surgeries including RYGB (Li et al., 2020) and vertical sleeve gastrectomy (VSG) (Ding et al., 2016; McGavigan et al., 2017; Ryan et al., 2014) in diet-induced obese mice, it is unknown whether this is through regulating IEB stability. We therefore tested the effects of the duodenal, jejunal, and colonic contents of RYGB-operated compared with sham-operated rats in electrical impedance assays performed on Caco-2 cell monolayers in the presence of selective TGR5/FXR antagonists/agonists compared with vehicle control (Figure 4).

Figure 4

The Duodenal and Colonic, but not the Jejunal, Contents of RYGB-Operated Rats Stabilize Barrier Function Partly through FXR in Caco-2 Cells

(A–F) TER measurements and associated AUCs obtained from low-frequency electrical impedance assays performed on confluent Caco-2 cell monolayers treated for 24 hr with the pooled duodenal (A and B), jejunal (C and D), and colonic (D and E) contents of Sham and RYGB rats in the presence of 10μM of the TGR5 agonist 3-(2-chlorophenyl)-N-(4-chlorophenyl)-N,5-dimethyl-4-isoxazolecarboxamide (TGR5 agonist), 10μM of the TGR5 antagonist triamterene, 10uM of the FXR agonist WAY-362450 (WAY), 10 μM of the FXR antagonist DY-268 (DY), or vehicle control (0.1% DMSO). n = 7–16 cultures/group from 1 to 2 independent experiments.

Data are presented as mean ± SEM. Statistical significance was determined by two-way ANOVA (main effect of treatment) in (B and F) and one-way ANOVA with Tukey's post-hoc test in (A-F).

In order to gain insight into whether intestinal bile acid receptors region-specifically contribute to stabilization of the IEB following RYGB, the duodenal, jejunal, and colonic contents of RYGB-operated rats were supplemented with the TGR5 antagonist triamterene (10μM) (Cheng et al., 2019; Li et al., 2017; Lo et al., 2017; Wang et al., 2017a, 2017b) or the selective FXR antagonist DY-268 (10μM) (Yu et al., 2014). This revealed that only the latter treatment partially negated the effects on TER values of the duodenal and the colonic (Figures 4B and 4F), but not the jejunal (Figure 4D), contents of RYGB-operated rats. These findings suggest that duodenal and colonic luminal factors following RYGB improve IEB function partly through FXR, whereas jejunal luminal factors do so independently of FXR.

In order to gain insight into whether intestinally restricted bile acid receptor agonists could region-specifically stabilize the IEB in obesity, the duodenal, jejunal, and colonic contents of sham-operated rats were supplemented with the selective TGR5 agonist 3-(2-chlorophenyl)-N-(4-chlorophenyl)-N, 5-dimethyl-4-isoxazolecarboxamide (10μM) (Evans et al., 2009) or the selective FXR agonist WAY-362450 (10μM) (Flatt et al., 2009). This revealed that only the latter treatment modestly increased TER values in response to the duodenal and the colonic (Figures 4B and 4F), but not the jejunal (Figure 4D), contents of sham-operated rats. These findings suggest that soluble factors specifically in the duodenal and colonic contents following RYGB act in synergy with FXR agonism to stabilize IEB function.

Plasma Cholic acid, Chenodeoxycholic acid, and FGF19 Levels Negatively Correlate with Endotoxemia Severity in Pre-diabetic/Diabetic RYGB Patients

In order to translate our animal and cell-based findings, we revisited phenotypic data and re-analyzed fasting plasma samples previously obtained from a mixed longitudinal cohort of 38 pre-diabetic/diabetic patients at baseline and 12 months after RYGB (Bankoglu et al., 2018). As expected, RYGB markedly reduced body weight, BMI, fasting plasma leptin, and LPS levels (Table 1). This was associated with a reduction in various circulating pro-inflammatory cytokines/markers such as tumor necrosis factor-alpha (TNF-α), interleukin-6 (Il-6), and C-reactive protein, which is in-line with previous studies (Illan-Gomez et al., 2012; Viana et al., 2013), as well as an improvement in several markers of glycemic status including fasting plasma HbA1c, glucose, and insulin levels, which drove homeostatic model of insulin resistance (HOMA-IR) indices down (Table 1).

Table 1

Fasting Metabolic Parameters of Pre-diabetic/diabetic RYGB Patients at Baseline and 12 Months after Surgery

Metabolic Parameter	Baseline	12 months Post-RYGB	p Value
Body weight (kg)	146.5 ± 3.1 (n = 38)	104.3 ± 3.7 (n = 34)	≤0.0001
Body mass index (kg/m2)	51.2 ± 1.0 (n = 38)	36.4 ± 1.1 (n = 34)	≤0.0001
Plasma leptin (ng/mL)	29.58 ± 2.17 (n = 37)	9.47 ± 1.51 (n = 23)	≤0.0001
Plasma LPS (EU/mL)	0.42 ± 0.02 (n = 34)	0.32 ± 0.04 (n = 26)	≤0.0001
Plasma Tnf-α (pg/mL)	4.10 ± 0.40 (n = 37)	3.11 ± 0.33 (n = 23)	0.0035
Plasma Il-6 (pg/mL)	10.42 ± 2.42 (n = 37)	5.94 ± 1.84 (n = 23)	0.0081
Plasma CrP (μg/mL)	93.76 ± 9.84 (n = 37)	28.89 ± 6.06 (n = 35)	≤0.0001
Plasma HbA1c (%)	5.77 ± 0.11 (n = 31)	5.26 ± 0.07 (n = 33)	≤0.0001
Plasma glucose (mg/mL)	9.4 ± 0.19 (n = 37)	8.6 ± 0.15 (n = 33)	0.0004
Plasma insulin (ng/mL)	0.64 ± 0.09 (n = 37)	0.23 ± 0.03 (n = 23)	≤0.0001
HOMA-IR	3.10 ± 0.38 (n = 34)	1.15 ± 0.18 (n = 20)	≤0.0001

Data are presented as mean ± SEM. p Values in bold are statistically significant as determined by two-tailed, paired t test or Wilcoxan matched-pairs signed rank test.

We next measured fasting plasma bile acid levels by electrospray ionization liquid chromatography–mass spectrometry (ESI-LC-MS/MS). In-line with previous studies (Ahlin et al., 2019; Chen et al., 2019; Gerhard et al., 2013; Kohli et al., 2013; Pournaras et al., 2012; Spinelli et al., 2016), this revealed an approximately order of magnitude increase for the unconjugated primary bile acids cholic acid (CA) and chenodeoxycholic acid (CDCA), as well as comparatively smaller (approximately 50%) increases in their glycine-conjugated derivatives glycocholic acid (GCA) and glycochenodeoxycholic acid (GCDCA) 12 months after RYGB compared with baseline (Table 2). Similarly, for secondary bile acids, there was a 4-fold increase in deoxycholic acid (DCA) and a 2-fold increase in ursodeoxycholic acid (UDCA) along with a corresponding increase in glycodeoxycholic acid (GDCA), whereas glycoursodeoxycholic acid (GUDCA) did not change (Table 2). Interestingly, while fasting plasma lithocholic acid (LCA) levels also did not change 12 months after RYGB compared with baseline, there was a significant increase in glycolithocholic acid (GLCA) (Table 2).

Table 2

Fasting Plasma Bile Acid and FGF19 Concentrations of Pre-diabetic/Diabetic RYGB Patients at Baseline and 12 Months after Surgery

Plasma BIle Acid or Related Factor	Baseline	12 months Post-RYGB	p Value
CA (μmol/L)	0.146 ± 0.05 (n = 37)	1.65 ± 0.57 (n = 27)	0.0047
CDCA (μmol/L)	0.22 ± 0.06 (n = 37)	1.60 ± 0.38 (n = 27)	0.0061
GCA (μmol/L)	0.28 ± 0.06 (n = 37)	0.37 ± 0.07 (n = 27)	0.071
GCDCA (μmol/L)	0.81 ± 0.13 (n = 37)	1.27 ± 0.19 (n = 27)	0.031
Total primary bile acids (μmol/L)	1.47 ± 0.26 (n = 37)	4.90 ± 0.93 (n = 27)	0.0024
DCA (μmol/L)	0.23 ± 0.03 (n = 37)	0.96 ± 0.20 (n = 27)	0.0051
LCA (μmol/L)	0.023 ± 0.002 (n = 37)	0.021 ± 0.002 (n = 27)	0.64
UDCA (μmol/L)	0.06 ± 0.01 (n = 37)	0.12 ± 0.03 (n = 27)	0.17
GDCA (μmol/L)	0.32 ± 0.08 (n = 37)	0.52 ± 0.09 (n = 27)	0.26
GLCA (μmol/L)	0.009 ± 0.001 (n = 37)	0.013 ± 0.002 (n = 27)	0.037
GUDCA (μmol/L)	0.09 ± 0.01 (n = 37)	0.09 ± 0.02 (n = 27)	0.58
Total secondary bile acids (μmol/L)	0.70 ± 0.09 (n = 37)	1.74 ± 0.26 (n = 27)	0.0079
FGF 19 (pg/mL)	22.76 ± 2.95 (n = 33)	63.72 ± 11.79 (n = 20)	0.0006

Data are presented as mean ± SEM. p values in bold are statistically significant as determined by Wilcoxan matched-pairs signed rank test.

Because CA, CDCA, GCA, GCDCA, DCA, and GLCA are all FXR agonists (Parks et al., 1999; Wang et al., 1999), and fibroblast growth factor 15/19 (FGF15/19) is released from enterocytes in response to activation of FXR (Inagaki et al., 2005a), we measured fasting plasma FGF19 levels in patients at baseline and 12 months after RYGB. In-line with previous studies (Chen et al., 2019; DePaoli et al., 2019; Gerhard et al., 2013; Pournaras et al., 2012), RYGB almost tripled fasting plasma FGF19 levels (Table 2) indicating intestinal FXR activation. When merging the data from baseline and 12 months after RYGB (Mani et al., 2019), regression analysis additionally revealed that fasting plasma CA (Figure 5A), CDCA (Figure 5B), and FGF19 (Figure 5D) levels weakly, but significantly, negatively correlated with fasting plasma LPS levels, while DCA trended in the same direction (Figure 5C). Further, fasting plasma LPS levels positively correlated with fasting plasma insulin levels and HOMA-IR indices (Figures 5E and 5F).

Figure 5

Plasma CA, CDCA, and FGF19 Levels Negatively Correlate with Endotoxemia Severity in Pre-Diabetic/Diabetic RYGB Patients

(A–D) Pearson correlations between fasting plasma CA (A), CDCA (B), DCA (C), and FGF19 (D) concentrations with fasting plasma LPS concentrations in RYGB patients at baseline (black circles; n = 31–34) and 12 months after surgery (white circles; n = 20–26).

(E and F) Pearson correlations between fasting plasma LPS concentrations with fasting plasma insulin concentrations (E) and HOMA-IR indices (F) in RYGB patients at baseline (black circles; n = 31–34) and 12 months after surgery (white circles; n = 20–26).

Solid regression lines indicate least squares fit of data and dashed lines indicate 95% confidence intervals. Statistical significance was determined by two-tailed, unpaired t test.

Discussion

The main objective of the present study was to gain mechanistic insight into how RYGB consistently stabilizes the IEB/attenuates endotoxemia across species (Guo et al., 2019; Monte et al., 2012; Steensels et al., 2017; Troseid et al., 2013; Wang et al., 2019a; Wu et al., 2019; Yang et al., 2014) in order to inform the design of more effective pharmacotherapies against insulin resistance and type 2 diabetes. We administered the intestinal contents of RYGB-operated and sham-operated rats to germ-free mice and Caco-2 cells. This unique approach led us to identify a close interaction between intestinal luminal factors with intestinal FXR signaling as a potential contributor to the stabilized IEB/attenuated endotoxemia following RYGB (Figure 6), which was corroborated in a longitudinal cohort of 38 pre-diabetic/diabetic patients.

Figure 6

Region-Specific Mechanisms for Intestinal Luminal Factors in Stabilizing the Intestinal Epithelial Barrier following RYGB

Schematic diagram illustrating a proposed model based on the findings of the present study. While duodenal luminal factors following RYGB may locally stabilize the IEB partly through FXR, this would not be expected to have an overall impact on endotoxemia-induced insulin resistance and glycemic control due to the low bacterial load in the duodenal contents. However, colonic luminal factors following RYGB may locally stabilize the IEB partly through FXR to attenuate endotoxemia-induced insulin resistance and improve glycemic control due to the high bacterial load in the colonic contents. Similarly, jejunal luminal factors following RYGB may locally stabilize the IEB to attenuate endotoxemia-induced insulin resistance and improve glycemic control due to the moderate bacterial load in the jejunal contents, although this would be independent of FXR signaling possibly due to the redirection of bile flow away from the upper jejunum post-operatively.

We previously found that the cecal contents of RYGB-operated compared with sham-operated ZF rats conferred superior oral glucose tolerance to recipient germ-free mice (Arora et al., 2017). We have now extended these findings to include the jejunal and colonic contents of RYGB-operated compared with sham-operated diet-induced obese rats and have further provided an association with suppression of endotoxemia. Interestingly, the blood glucose excursion curves of recipient germ-free mice in the present study suggest that the jejunal contents of RYGB-operated compared with sham-operated rats may mainly enhance insulin release, while the colonic contents may mainly enhance peripheral insulin sensitivity. Future glucose-stimulated insulin secretion assays and hyperinsulinaemic-euglycaemic clamp studies performed on recipient germ-free mice are needed to confirm this. In contrast, the lack of effect of the duodenal contents of RYGB-operated compared with sham-operated ZF rats on oral glucose tolerance in recipient germ-free mice is consistent with the low (gram-negative) bacterial load in this small intestinal region (Arora et al., 2017) and highlights the region-specific contribution of the gastrointestinal tract to regulating systemic endotoxemia (Chakaroun et al., 2020).

By applying the intestinal contents of RYGB-operated and sham-operated rats directly onto the apical side of Caco-2 cells, we partially simulated the regional intestinal microenvironments of the respective surgical groups. This revealed region-specific, barrier-stabilizing effects for the intestinal contents of RYGB-operated compared with sham-operated rats, possibly mediated by the action of various host- and/or microbiota-generated soluble factors. These potentially include, but are not limited to, hepatocyte-derived conjugated primary bile acids, that may become concentrated in the duodenum following RYGB due to redirection of ingested food passage to the mid-jejunum (Ise et al., 2019), as well as microbiota-derived metabolites such as indole-3-propionic acid (Jennis et al., 2018; Natividad et al., 2018), secondary bile acids (Haange et al., 2020; Lajczak-McGinley et al., 2020), taurine (Ahmadi et al., 2020; Wang et al., 2019a), and the short chain fatty acid propionate (Liou et al., 2013; Tong et al., 2016). Future metabolomics studies are required to confirm this, and to potentially discover novel soluble factors that stabilize the IEB.

Murine models of metabolic surgeries were originally developed to identify the molecular mediators behind their unique metabolic benefits using specific knockout lines (Stevenson et al., 2019). A landmark achievement using this approach is the identification of FXR and/or TGR5 as important molecular targets of VSG (Ding et al., 2016; McGavigan et al., 2017; Ryan et al., 2014), bile diversion to the ileum (BDI) (Albaugh et al., 2019; Pierre et al., 2019), and RYGB (Li et al., 2020). Remarkably, while mice deficient for FXR show normal changes in whole-body energy balance following RYGB, they do not present with fully improved oral glucose tolerance, insulin tolerance, and HOMA-IR indices (Li et al., 2020). The findings of the present study suggest that this could be due to persistently elevated endotoxemia post-operatively, possibly from the loss of intestinal FXR-mediated stabilization of the IEB, which can be formally assessed in mice lacking FXR specifically in enterocytes (Albaugh et al., 2019). In contrast, mice deficient for TGR5 respond normally to RYGB (Hao et al., 2018), which is in line with the lack of effect of TGR5 blockade on barrier function in Caco-2 cells in response to the intestinal contents of RYGB-operated rats reported here.

We could confirm the widely documented rise in various primary and secondary plasma bile acids in patients following RYGB (Ding et al., 2019). This may reflect increased intestinal luminal bile acid levels and organic anion-transporting peptide (OATP)-mediated re-uptake by enterocytes, as has been shown in rats following duodenal-jejunal bypass (Ueno et al., 2020), an experimental surgical procedure anatomically similar to RYGB. By extension, FXR could be activated in enterocytes to stabilize the IEB following RYGB in patients as proposed for rodents above. Indeed, we could link the post-operative rise in plasma FGF19 levels, a readout of intestinal FXR activation, to attenuated endotoxemia, which itself positively associated with HOMA-IR indices. However, it should be stressed that correlation does not prove causation. Future studies are required to confirm if the intestinal contents or at least the stool of patients following RYGB compared with baseline have FXR-mediated, barrier-stabilizing effects in enterocytes.

Limitations of the Study

We did not analyze intestinal microbiota in the present study so cannot exclude the possibility that the attenuated endotoxemia following RYGB is due to an overall reduction of intestinal LPS-producing, gram-negative bacteria. We think that this is unlikely however as we have previously shown that such intestinal bacteria, including Bacteroides species, are markedly increased throughout the gastrointestinal tract in RYGB-operated compared with sham-operated rats (Arora et al., 2017). Our in vitro system did not include intestinal immune cells, which could contribute to the stabilized IEB following RYGB (Luck et al., 2019; Wang et al., 2019a). Nevertheless, our approach using Caco-2 cells alone allowed us to establish direct, cell-autonomous roles for intestinal contents in regulating IEB function and structure. We could not employ pair-fed or body weight-matched groups in our rat experiments, so it is possible that the stabilized IEB/attenuated endotoxemia reported here are not specific to RYGB. However, previous findings from diet-induced obese mice and rats, as well as genetically obese and insulin-resistant ZDF rats, have established that reduced intestinal permeability/attenuated endotoxemia following RYGB occur independently of food/high-fat diet restriction and weight loss (Guo et al., 2019; Steensels et al., 2017; Wang et al., 2019a). Additionally, we previously found that colonic CA and CDCA concentrations are twice as high in RYGB-operated, diet-induced obese rats compared with body weight-matched, sham-operated counterparts (Haange et al., 2020), further supporting the view that gastrointestinal reconfiguration-induced changes in bile flow and bile acid metabolizing intestinal microbiota per se, contribute to increased colonic FXR activation and stabilization of the IEB (Figure 6). This is also supported by the finding that the jejunal contents of RYGB-operated rats stabilized barrier function in Caco-2 cells independently of FXR, since bile flow is redirected away from the upper/mid jejunum post-operatively. Nevertheless, future bile acid measurements of the duodenal and colonic contents of RYGB-operated rats by nuclear magnetic resonance spectroscopy are needed to identify the precise molecular mechanisms responsible for their barrier-stabilizing effects, akin to that performed for mouse cecal contents in the context of enhanced enteroendocrine cell function following VSG (Chaudhari et al., 2020).

In summary, we have presented a combination of pre-clinical and clinical evidence suggesting that RYGB stabilizes the IEB potentially through a synergistic interaction between intestinal luminal factors and intestinal FXR signaling. Our findings also provide an experimental framework for identifying other signaling pathways in enterocytes responsible for attenuating endotoxemia following RYGB, which could be exploited for the more effective, non-invasive treatment of insulin resistance in obesity and possibly also other metabolic disorders.

Resource Availability

Lead Contact

Further information and requests for resources should be directed to and, where possible, will be fulfilled by the Lead Contact, Mohammed Hankir (hankir_m@ukw.de).

Materials Availability

All reagents used in this study will be made available on reasonable request to the Lead Contact.

Data and Code Availability

The original/source data are available from the Lead Contact on reasonable request.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.