Restoration of Euglycemia After Duodenal Bypass Surgery Is Reliant on Central and Peripheral Inputs in Zucker fa/fa Rats

Goto-Kakizaki (GK), diet-induced obesity Sprague-Dawley and Wistar rats, and the obese, insulin-resistant ZF rats have previously been used to examine the impact of bariatric surgery on glucose homeostasis (5–8). Our data (Supplementary Fig. 1A and B) indicate that while both Sprague-Dawley and Wistar rats become obese and develop fatty liver after 30 weeks of high-fat diet, rats remain euglycemic with no apparent defect in glycemic control. While the reasons for lack of diet-induced insulin resistance are unclear, it is plausible that compensatory glucose disposal by peripheral tissues or increased glucose excretion provides a protective mechanism. As such, it is difficult to evaluate modalities that modulate insulin sensitivity in these animal models. The GK rat has stable hyperglycemia; however, the lack of obesity and peripheral insulin resistance limits its use for elucidating mechanisms of obesity-induced insulin resistance. Considering these limitations, we elected to characterize the impact of DJB in the obese, hyperlipidemic, hyperinsulinemic, insulin-resistant ZF rat.

All surgical procedures were performed on 11- to 13-week-old male ZF rats and metabolic studies 1 month post-surgery. DJBs were carried out as previously described (9). Sham-operated animals were exposed to similar surgical conditions and pair-fed when appropriate. Oral glucose tolerance tests (oGTTs) (1 g/kg) and mixed-meal tolerance tests (mmTTs) (15 g normal rat chow) were carried out after an overnight fast. For intravenous glucose tolerance test (ivGTTs) (1 g/kg), animals were cannulated in the jugular vein and glucose was injected in bolus. For assessment of hepatic gluconeogenic capacity, intraperitoneal pyruvate tolerance test (ipPTT) (1 g/kg) was conducted in overnight-fasted animals. For assessment of tissue-specific insulin sensitivity, hyperinsulinemic-euglycemic clamp procedures were conducted as previously described with minor modifications (10). Briefly, after an overnight fast, at −130 min a priming dose of 7.5 µCi D-[3-³H]glucose (PerkinElmer) was administered into the carotid artery and a constant-rate (0.167 µCi/min; 16.7 μL/min) infusion of D-[3-³H]glucose was started at –120 min. Blood samples (25 μL in duplicate) were taken at −30, −20, −10, and 0 min for measurement of tracer-specific activity for assessment of basal HGP. At time 0 min (i.e., after 120 min of tracer equilibration), glucose (variable infusion, 50% dextrose; Hospira) and tracer plus insulin (25 mU/kg/min; Humulin, Eli Lilly) infusions were started simultaneously. Steady-state glucose concentrations were maintained for at least 40 min for assessment of tracer-specific activity. For assessment of β-cell function, hyperglycemic clamps were conducted as previously described (11). All rats were individually housed on a 12:12-h light:dark cycle. Animals were fed a standard rodent chow (2018; Tekland Global) and water ad libitum. All animal care procedures and experimental procedures were approved by the institutional animal care and use committee at NGM Biopharmaceuticals, Inc. All animals were obtained from Harlan Laboratories.

Studies were conducted using the TSE PhenoMaster Systems 1 month post-DJB. Animals were acclimated to the metabolic cages for a period of 36 h prior to a 72-h study period.

TSV was performed 1 month post-DJB. The timing for TSV was selected to confirm the positive effect of DJB on glucose tolerance prior to examining the contribution of vagal nerves. The surgical procedures for TSV have previously been described in detail (12). Control animals underwent identical surgical procedures including isolation of the hepatic and subdiaphragmatic trunks, but the nerves were left intact.

Quadriceps, liver, and visceral fat tissues were harvested for measuring the mRNA and protein content of insulin-signaling molecules 1 month after surgery. For insulin-stimulated assays, animals were intraportally infused with Humulin at 0.5 units/kg for 6 min and tissues were excised in a timely manner and later homogenized in Tris lysis buffer (50 mmol/L Tris, pH 8.0; 150 mmol/L NaCl; 1% Triton X-100; and 0.1% SDS) containing protease inhibitor and phosphatase inhibitor cocktails (Sigma). Protein concentration was measured and adjusted to the same concentration across the samples. Samples were diluted with assay buffer provided by Bio-Rad. Triacylglyceride (TAG) and diacylglyceride (DAG) measurements were carried out as previously described (13). Ceramide content was analyzed using recombinant ceramide kinase (BPS Biosciences, San Diego, CA; catalog number 40617) following the same kinase assay protocol used by the manufacturer. All reagents required to assay lipid contents were purchased from WAKO (Richmond, VA). Protein content was assayed in an aliquot by bicinchoninic acid method. Fatty acid oxidation was determined using previously established procedures (14). Glycogen was assayed as previously described (15). Western blots were performed using 4–12% SDS-PAGE acrylamide gels (Invitrogen, CA). Antibodies were purchased from Cell Signaling Technology (Danvers, MA). Total RNA was isolated using TRIzol (Invitrogen). First-strand cDNA was synthesized using a high-capacity cDNA reverse-transcription kit (Applied Biosystems, Foster City, CA). The samples were run in 20-μl reactions using an MJ Research PTC-200 96-well thermocycler coupled with the Chromo4 four-color real-time system (GMI, Ramsey, MN). Gene expression levels were calculated as previously described (16). Primer information is available in Supplementary Table 2.

DJB, sham-operated and their Zucker lean littermates were fasted overnight followed by a 2-h refeeding with 15 g normal chow. Intestines were collected and divided into duodenum, jejunum, ileum, and colon segments. The intestinal segments were cut open longitudinally, rinsed with PBS twice, and immediately frozen in liquid nitrogen. The segments were then weighed and immersed in extraction buffer (methanol with 0.1% trifluoroacetic acid), 5 mL buffer/g tissue, in 50-mL conical tubes rotating for 1 h at room temperature. The tissues were centrifuged at 3,500 rpm for 10 min, and the supernatants were filtered, vacuum dried, and reconstituted in PBS. Total protein concentration was measured using BCA Protein Assay Reagent (Thermo Scientific). The peptides were measured using ELISA kits: glucagon-like peptide-1 (GLP-1), (active 7-36; Alpco and Millipore), glucose-dependent insulinotropic peptide (GIP) (rat/mouse total; Millipore), and peptide YY (PYY) (mouse/rat; Alpco).

Experimental results are expressed as means ± SEM. Results were tested statistically by an unpaired two-tailed Student t test or one-way ANOVA with the Bonferroni correction using GraphPad Prism 5.0 (San Diego, CA). The criterion for statistical significance was set at P < 0.05.

To determine whether reduction in body weight was a prerequisite for restoration of euglycemia, we performed DJB, a nonrestrictive surgery. DJB resolved hyperglycemia compared with sham-operated, pair-fed animals (sham) independent of body weight (Fig. 1A and B). The ideal time for detailed metabolic studies, defined by full recovery from surgery and metabolic stability, was determined based on data gathered from time course oGTT studies (Fig. 1C and D). DJB rapidly improved glucose tolerance, reaching a plateau prior to week 4. Therefore, we elected to examine the impact of DJB on insulin sensitivity at this time point. To assess nutrient-induced insulin secretion, we conducted oGTT and mmTT in separate sets of animals (Fig. 1E and F). Peak glucose and insulin excursion was reduced by ∼40 and ∼55%, respectively, indicating that whole-body glucose disposal was increased independent of augmented insulin secretion. To account for potential metabolic responses attributable to delayed gastric emptying or malabsorption, ivGTT was performed (Fig. 1G). Similarly, glucose and insulin excursion was reduced by ∼30 and ∼75%, respectively. When the peak insulin concentration of DJB during oGTT and mmTT was compared with that from ivGTT, a modest incretin effect was observed; however, it remained significantly lower compared with respective controls.

Increased hepatic glucose output is an important cause of fasting hyperglycemia (17). Elevated HGP is attributed to increased gluconeogenesis (18). To assess hepatic gluconeogenic capacity, we performed ipPTT and found that the glucose excursion in DJB was reduced by ∼50% compared with sham (Fig. 2A). For accurate assessment and dissociation of basal HGP from hepatic and peripheral insulin sensitivity, hyperinsulinemic-euglycemic clamp studies were performed. Body weights were similar between DJB and sham, and both groups were markedly obese compared with the lean group (Supplementary Fig. 1C). However, postprandial and fasted insulin concentrations were approximately twofold lower in DJB compared with sham (Supplementary Fig. 1D). The exogenous glucose infusion rate (GIR) required to maintain euglycemia (∼100 mg/dL) was markedly higher in DJB compared with sham, confirming enhanced systemic insulin sensitivity (Fig. 2B). The improvement in GIR was due to restoration of basal HGP; hepatic insulin sensitivity, as indicated by enhanced insulin-mediated suppression of HGP; and an increase in insulin-stimulated glucose disposal rate (IS-GDR), a marker for skeletal muscle insulin sensitivity (Fig. 2C–E). Insulin-mediated suppression of adipose tissue–derived free fatty acids was also greater in DJB, signifying increased adipocyte insulin sensitivity (Fig. 2F). Intriguingly, steady-state plasma insulin concentration was lower in DJB compared with sham (Supplementary Fig. 1E). This reduction was attributed to an increased metabolic clearance rate of insulin (Supplementary Fig. 1F).

For determination of whether DJB altered insulin secretion, hyperglycemic clamp studies were conducted (Supplementary Fig. 1G). Glucose-stimulated insulin secretion was lower in DJB during physiologically relevant bouts of hyperglycemia (Supplementary Fig. 1H). However, insulin concentration was not statistically different between DJB and sham at supraphysiological glucose levels (i.e., >350 mg/dL), suggesting that the maximal secretory capacity was not compromised. Insulin sensitivity after DJB was further confirmed, as GIR was increased by ∼80% (Supplementary Fig. 1I).

To understand the mechanism of improved insulin sensitivity, we measured key aspects of the insulin-signaling pathway. We observed an increase in total content and tyrosine phosphorylation of insulin receptor substrate (IRS)-2 in DJB liver in both the basal and insulin-stimulated states (Fig. 3A). Surprisingly, insulin-stimulated Akt phosphorylation was not different among the groups (Supplementary Fig. 2A). Phosphorylation of cAMP response element–binding protein (CREB), a key regulator of hepatic gluconeogenesis (19), was reduced in DJB liver (Fig. 3B) without any changes in hepatic glycogen content (Supplementary Fig. 2B). The expression of the key gluconeogenic genes phosphoenolpyruvate carboxykinase (PEPCK) and glucose 6-phosphatase (G6Pase) was also decreased (Fig. 3C).

To ascertain whether enhanced systematic insulin sensitivity was attributable to enhanced lipid oxidation, we performed indirect calorimetry (Supplementary Table 1). Although body weight, food intake, and water consumption and activity levels were not different between groups, we observed a small increase in body temperature. Remarkably, lipid use was increased in DJB and normalized to lean, as indicated by a lower respiratory exchange ratio (RER) (Fig. 4A). To evaluate whether DJB rats remained metabolically flexible, we performed similar tests in separate sets of animals exposed to an overnight fast. DJB and sham responded similarly when fasted, relying exclusively on lipid oxidation, but returned to prefasting conditions when reexposed to food (Supplementary Fig. 3A). Consistent with the RER values, plasma TAG was reduced without a change in TAG secretion capacity in DJB, while fasted total ketones and β-hydroxybutyrate, indirect markers of lipid oxidation, were increased by ∼40 and ∼50% in DJB compared with sham, respectively (Supplementary Fig. 3B–E).

To more accurately assess lipid use in liver, we measured hepatic palmitate oxidation. Palmitate oxidation was increased by ∼90% in DJB liver compared with sham (Table 1). This increase was paralleled by phosphorylation of AMP-activated protein kinase (AMPK) (Fig. 4B). The expression of key regulatory genes in lipid metabolism was also examined (Fig. 4C). Expression of liver-X receptor α (LXRα), a key regulator of genes involved in lipid metabolism, was dramatically reduced. Expression of sterol regulatory element-binding protein-1c (SREBP-1c) along with its transcription targets, fatty acid synthase (FAS) and stearoyl-CoA desaturase 1 (SCD-1), was reduced, while expression of peroxisome proliferator–activated receptor (PPAR) γ and α remained unchanged. The expression of acetyl-CoA carboxylase (ACC), a rate-limiting enzyme in fatty acid synthesis, was decreased and its phosphorylation increased after DJB (Figs. 4D and E). Collectively, these findings provide a mechanism for the reduction of hepatic lipid synthesis and enhanced lipid oxidation associated with DJB.

We also measured specific hepatic lipid levels and observed a reduction in DAG, ceramides, and fatty acyl-CoAs (FA-CoAs), while hepatic TAG content was not altered (Table 1). Consistently, histological examination of lipid content did not reveal a difference between DJB and sham (Supplementary Fig. 3F).

To further characterize the impact of DJB on insulin-sensitive peripheral tissues, we measured markers of lipid metabolism and insulin signaling in muscle and visceral fat. Similar to liver, skeletal muscle palmitate oxidation was enhanced and was associated with activation of AMPK (Table 1 and Supplementary Fig. 4A). Unlike the liver, skeletal muscle Akt phosphorylation was increased and was paralleled by a reduction in glycogen stores (Supplementary Fig. 4B and C). Moreover, skeletal muscle concentrations of TAG, FA-CoA, DAG, and ceramides were equally reduced in DJB (Table 1). In visceral fat, DAGs, FA-CoAs, and ceramides were equally reduced without a change in TAG content (Table 1). This improvement was associated with an increase in insulin-stimulated Akt phosphorylation (Supplementary Fig. 4D).

Gastrointestinal bypass surgery is accompanied by an array of physical and neurohormonal changes within the gastrointestinal system. While the circulating concentration of several intestinally derived factors, including GLP-1, GIP, and PYY, have previously been documented, it is not clear whether local tissue concentration of these peptides is equally altered. Therefore, we developed a method to extract peptides from the intestines in a segment-specific manner and evaluated the tissue content of GLP-1, GIP, and PYY. The duodenal and ileal mass was selectively increased in DJB (Fig. 5A). Although the plasma concentration of GIP and PYY was not altered after DJB (data not shown), their tissue concentration was dramatically increased in jejunum and ileum, respectively (Fig. 5B and C). Conversely, while tissue concentrations of GLP-1 were unaltered in DJB, GLP-1 concentration was increased in plasma postprandially and in response to refeeding (Fig. 5D and E). The divergence between the serum and tissue concentration of these peptides suggests that intestinally derived hormones, in addition to their endocrine mode of action, might exert effects through an autocrine and/or paracrine manner, perhaps by modulating the enteric nervous system. In support of this, both GLP-1 and PYY receptors are expressed on these glucomodulatory nerves innervating the gastrointestinal track (20–22).

On the basis of well-established effects of vagal inputs into glucose homeostasis (23), we examined the potential contribution of vagal signaling to the enhanced glucose metabolism after DJB. To account for potential contribution of all subbranches of the vagus nerve, we performed TSV in control ZF rats (TSV) (Supplementary Fig. 5A) and DJB-operated rats (DJB^TSV). Control animals were exposed to both sham DJB and TSV surgeries and were pair-fed to TSV animals (sham). At the time of metabolic studies, body weight was not different among groups (Supplementary Fig. 5B). Denervation was confirmed during surgery by monitoring of the heart rate response and postmortem by visual inspection of severed nerves (Supplementary Fig. 5C–E). For assessment of the immediate effect of vagotomy in DJB^TSV, ipPTT was performed 1 week post-surgery. While vagotomy did not further exacerbate glucose production in control ZF rats, TSV abolished DJB-induced normalization of gluconeogenic capacity (Fig. 6A). For determination of whether CNS-derived vagal nerves are required for maintenance of hepatic and peripheral insulin sensitivity, euglycemic-hyperinsulinemic clamps were performed. Consistent with the glucoregulatory role of vagal innervation, TSV abolished DJB-induced improvement in basal HGP (Fig. 6B). However, GIR during clamp studies remained markedly higher in DJB^TSV animals, indicating that DJB-induced improvement in insulin sensitivity was preserved independent of vagal innervation (Fig. 6C). In accord with the higher GIR, insulin-stimulated (steady-state) suppression of HGP, a marker of hepatic insulin sensitivity, and IS-GDR, a marker for skeletal muscle insulin sensitivity, remained markedly improved in the DJB^TSV (Fig. 6D–F). Adipocyte insulin sensitivity was also maintained in DJB^TSV (Fig. 6G). In fact, restoration of hepatic and peripheral insulin sensitivity was the same in DJB compared with DJB^TSV (compare Figs. 2 and 6), demonstrating that vagotomy failed to compromise the insulin-sensitizing effects of DJB.

Bariatric surgery is an increasingly popular and effective means to achieve substantial weight loss in obese patients, resulting in a wide range of improved health outcomes (24). In many post-surgical diabetic patients, antidiabetes medications can be stopped altogether, and oftentimes the underlying diabetic state is normalized within days and sustained for years. Of interest, nonrestrictive surgeries that simply bypass the duodenum similarly improve glucose homeostasis in the absence of weight loss. Collectively, these findings have generated great interest in understanding the mechanisms regulating metabolic balance after duodenal bypass.

The mechanisms underlying the metabolic effects of bariatric surgery are largely unknown, but several possibilities exist. 1) Decreased caloric intake can have an immediate beneficial effect on glucose homeostasis. 2) Gastrointestinal-CNS communication via the vagus nerve may contribute to enhanced glucose tolerance, HGP, and insulin sensitivity (25–28). 3) Duodenal bypass could lead to the amplification as well as generation of novel gut hormone peptides(s) that improve insulin sensitivity or, alternatively, could suppress the secretion of a gut-derived diabetogenic factor(s). 4) Finally, changes in bile acids and alteration in the intestinal microflora can favorably modulate obesity and glucose homeostasis (29,30).

In this study, DJB markedly improved glucose tolerance in obese insulin-resistant rats, and this was accompanied by a striking reduction in insulinemia. Hepatic, skeletal muscle, and adipose tissue insulin sensitivity were substantially improved in DJB animals compared with weight-matched, sham-operated controls. While improved glucose metabolism after bariatric surgery has been widely reported (31–34), improved insulin sensitivity has not always been observed (8,35). For example, while duodenal bypass was shown to improve glucose tolerance, it was not effective at improving insulin sensitivity in high-fat–fed Long-Evans and Wistar rats (8,35). While the underlying reasons for this discrepancy are not clear, it is likely owing to the animal model studied, as discussed above. Since weight gain, food intake, and increasing adiposity proceeded normally in DJB, we conclude that the effects of proximal intestine bypass surgery are due to a mechanism or mechanisms independent of body weight and food intake. Our results from total vagal ligation demonstrate that the rapid restoration of hepatic gluconeogenic capacity and basal HGP in ZF rats is largely regulated by vagal pathways. This finding agrees with recent reports highlighting the key role of upper-intestinal vagal nerves in the regulation of basal HGP (28,36). Importantly, however, the insulin sensitization effect of DJB was independent of vagal innervation. Since reduced food intake, decreased adiposity, or vagal inputs do not appear to be contributing mechanisms to the restoration of insulin sensitivity, an alternative explanation relates to the gastrointestinal tract.

The gastrointestinal tract contains a variety of secretory cells, including enterocytes and enteroendocrine cells, which are specialized neuroendocrine cells responsible for production of a large number of polypeptide hormones. To date, a handful of enterocyte- and enteroendocrine cell–enriched factors have been implicated in regulation of glucose and lipid homeostasis. However, it is quite plausible that the observed changes in the internal milieu of the intestinal tract after DJB, which include an expansion of the intestinal mass, particularly in the ileum, lead to alterations in secreted factors from resident enterocytes and enteroendocrine cells with antidiabetes effects. In support of this hypothesis, we have demonstrated that DJB increases the concentration of GIP and PYY in the jejunum and ileum but not in circulation. In contrast, tissue concentration of GLP-1 was unaltered in DJB, while the circulating concentration was increased.

Several studies have reported increased luminal and circulating concentration of bile acids after bariatric surgery (37–39). Consistent with these studies, we have equally observed an increase in total bile acids in circulation after DJB (Supplementary Fig. 6). In humans, the postprandial rise in serum bile acids is paralleled by an increase in serum concentration of ileum-derived enterokine, fibroblast growth factor (FGF)-19 (FGF-15 in rodents) (40) and has been associated with improved insulin sensitivity after Roux-en-Y gastric bypass (41). Although serum FGF-15 data are lacking in the current study, owing to lack of detection antibodies, we have observed that FGF-15 mRNA is upregulated by >150% after DJB (42). Therefore, it is plausible that DJB-induced insulin sensitization is in part due to amplified activation of the bile acid–FGF-15/19 axis. In support of this notion and in agreement with the current data, improved glycemic control after FGF-15/19 has been reported to be independent of the Akt/Foxo1 pathway but reliant on the CREB–peroxisome proliferator–activated receptor γ coactivator-1α pathway (43). Lastly, but equally intriguing, are the observations that in obesity, gut-derived factors such as lipopolysaccharides, or even bacteria themselves, can enter the circulation, exerting systemic proinflammatory effects and profound insulin resistance (44–46). Perhaps reengineering of the gastrointestinal tract through changes in bile acids alters gut microflora, or the “leakiness” of bacterial products into the systemic circulation, in such a way that limits these proinflammatory factors.

The improvement in insulin sensitivity and lipid metabolism, particularly in the liver, suggests the possibility of endocrine gastrointestinal factors with potent hepatic effect. In support of this, we found that the hepatic expression of key lipogenic and gluconeogenic genes were markedly reduced, while those involved in β-oxidation were enhanced. It is likely that the reduction in lipogenesis was caused by the decrease in LXR and SREBP-1c expression and decreased ACC expression/activity, while the increased β-oxidation rates were the result of the enhanced AMPK activity. It is of interest that despite an improvement in all parameters of lipid metabolism, hepatic TAG content did not decline in the DJB. This indicates that hepatic TAG content is not directly connected to enhanced hepatic insulin sensitivity; alternatively, our data suggest that DAGs and ceramides may play a more prominent role in the ZF rat. This finding is consistent with previous reports highlighting the association of DAG accumulation with insulin resistance (47,48).

Another surprising observation was the large increase in IRS-2 expression in the DJB. To this end, SREBP-1c has been shown to inhibit IRS-2 transcription by binding to the serum response element site of its promoter (49). Thus, an observed reduction in SREBP-1c expression in DJB may contribute to the increased IRS-2 transcription and subsequent hepatic insulin sensitivity.

In summary, we have demonstrated that selective proximal intestinal bypass ameliorates systemic insulin resistance in ZF rats independent of changes in body weight. Restoration of euglycemia was independent of augmented insulin secretion but was associated with a dramatic improvement in systemic insulin sensitivity, namely, hepatic insulin sensitivity, key insulin signaling molecules, and enhanced tissue-specific substrate metabolism. Importantly, vagal-mediated CNS inputs were essential for the rapid improvement in gluconeogenic capacity and regulation of basal HGP but were not required for the insulin-sensitization effect of DJB in ZF rats. Collectively, the observed morphological changes coupled with altered patterning of the intestinally derived hormone peptides and a robust improvement in glucose homeostasis after DJB support a gastrointestinal-driven mechanism with a potent antidiabetes effect. This work raises the possibility that gut-derived factor(s) responsible for observed beneficial effects could lead the way to the development of new therapeutic approaches for metabolic disease, bypassing the need for bariatric surgery.

This work was supported in part by grants from the National Institutes of Health (DK-033651, DK-074868, and DK-063491).

J.J., J.O., C.M., M.C., J.-Y.H., Y.C., H.T., and M.S. are current and/or past employees of NGM Biopharmaceuticals, Inc. J.M.O. is a member of the scientific advisory board at NGM Biopharmaceuticals, Inc. No other potential conflicts of interest relevant to this article were reported.

J.J. designed and performed the gastrointestinal peptide extraction and measurement, helped write the manuscript, and reviewed and analyzed all datasets. E.J.B. performed real-time PCR and immunoblotting, contributed to discussion, and reviewed and analyzed all datasets. G.B. performed tissue substrate assays, contributed to discussion, and reviewed and analyzed all datasets. J.O. performed glucose tolerance tests, pyruvate tolerance tests, and serum analysis; assisted with clamp procedures; contributed to discussion; and reviewed and analyzed all datasets. C.M. performed substrate assays, real-time PCR, histology, and immunoblotting; designed and wrote the manuscript; contributed to discussion; and reviewed and analyzed all datasets. M.C. performed glucose tolerance tests, pyruvate tolerance tests, and serum analysis; assisted with clamp procedures; contributed to discussion; and reviewed and analyzed all datasets. J.-Y.H. established and evaluated the peptide extraction method and reviewed and analyzed all datasets. Y.C. contributed to the experimental design of the gastrointestinal peptide measurement and reviewed and analyzed all datasets. H.T. contributed to discussion, reviewed the manuscript, and reviewed and analyzed all datasets. J.M.O. designed and wrote the manuscript and reviewed and analyzed all datasets. M.S. designed and wrote the manuscript; conducted all surgeries, in vivo experiments, and tissue analyses; and reviewed and analyzed all datasets. M.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Drs. Alex DePaoli and Daniel Kaplan of NGM Biopharmaceutical, Inc., who assisted with editing the manuscript. They also thank Drs. Michael M. Meguid and Carolina G. Goncalves (State University of New York Upstate Medical University) for Roux-en-Y gastric bypass demonstrations.