Time-Dependent Molecular Responses Differ Between Gastric Bypass and Dieting But Are Conserved Across Species

Summary

The effectiveness of Roux-en-Y gastric bypass (RYGB) against obesity and its comorbidities has generated excitement about developing new, less invasive treatments that use the same molecular mechanisms. Although controversial, RYGB-induced improvement of metabolic function may not depend entirely upon weight loss. To elucidate the differences between RYGB and dieting, we studied several individual organ molecular responses and we generated an integrative, inter-organ view of organismal physiology. We also compared murine and human molecular signatures. We show that although dieting and RYGB can bring about the same degree of weight loss, post-RYGB physiology is very different. RYGB induces distinct organ-specific adaptations in a temporal pattern, characterized by energetically demanding processes, which may be coordinated by HIF1α activation and the systemic repression of growth hormone receptor signaling. Many of these responses are conserved in rodents and humans, and may contribute to the remarkable ability of surgery to induce and sustain metabolic improvement.

Graphical Abstract

In Brief (eTOC)

Using an integrative, inter-organ view of organismal physiology, XXX et al compared RYGB gastric bypass and dieting in mouse and humans. Although dieting and RYGB can bring about the same degree of weight loss, the molecular signature of surgery highlights an essential role for metabolic regulators and the circadian clock.

Introduction

Despite a rapidly growing understanding of energy balance regulation, efforts to turn that knowledge into treatments for obesity and its comorbidities have met with minimal success. Weight loss surgery and especially Roux-en-Y gastric bypass (RYGB) remain the most effective treatment options (Arterburn and Courcoulas, 2014). The mechanisms underlying the remarkable RYGB-induced effects in body weight, glycemic and lipidemic control remain unclear. Many investigators have advocated that they do not depend entirely upon weight loss, based on clinical observations that metabolic improvement occurs early in the postoperative period before significant weight loss (Laferrere, 2011) and based on documented physiological permutations that body weight changes cannot account for (Evers et al., 2017). This theory, however, remains controversial and others have suggested that the postoperative calorie restriction is sufficient to explain the metabolic effects of RYGB (Isbell et al., 2010; Jackness et al., 2013). Unraveling the molecular underpinnings of the improvement of metabolic function after RYGB will be helpful in developing ways to improve surgery or produce these effects without surgery.

In this study, we used an established mouse model to characterize the response of multiple organs to RYGB focusing on metabolic pathways and examined the differential effects of RYGB and calorie restriction early and late after surgery. In addition, we performed a comparative analysis between murine and human responses to determine whether they are preserved in the two species. This strategy likely highlights the most important underlying mechanisms and molecular basis of the metabolic effects of RYGB. It also provides insight about the usefulness and limitations of the rodent model as an experimental system to study the effects of RYGB.

Results

We used C57BL/6 male mice that had developed diet-induced obesity and we performed either RYGB or a sham operation. A group of sham-operated animals was subsequently calorie restricted to match the body weight of RYGB-treated mice (Weight Matched Sham; WMS). To compare the acute and long-term effects of RYGB, we analyzed tissues 9 days and 9 weeks after surgery (Figure 1A,B).

Figure 1. RYGB leads to sustained weight loss and improved glucose metabolism in a mouse model of diet induced obesity (DIO).

A. Experimental design (see also section “Postoperative care” in Methods for details).

B. Intestinal anatomy following RYGB. Food (black arrow) passes from esophagus through a small gastric pouch directly to the jejunum. Duodenum drains bile, pancreatic and gastric secretions and connects to jejunum, forming a Y shaped intestinal configuration. We analyzed the 3 segments of the small intestine (duodenum, jejunum and ileum), the liver, gastrocnemius muscle and inguinal white adipose tissue.

C. Body weight of RYGB-treated and WMS mice over the study period (N=5 for both groups).

D. Blood glucose levels. Measurements were collected after overnight fasting on the day of surgery and on postoperative days 9, 28 and 63.

E–F. Respiratory exchange ratio (RER) during a 24-hour dark/light cycle (RYGB N=8; WMS N=7). Shadows indicate SEM (E). Boxplot of 24-hour measurements (F).

G–I. Heat production during a 24-hour dark/light cycle. Shadows indicate SEM (G). Boxplots of total (H) and resting (I) heat production. * RYGB vs WMS independent samples t-test, p<0.05.

RYGB-operated mice lost approximately 36% (18g) of their preoperative body weight and maintained the weight loss through the 9^th postoperative week (Figure 1C). These mice had lower fasting blood glucose levels than WMS, 9 days after surgery and onwards, despite the same degree of body weight loss (Figure 1D). In agreement with previous reports, RYGB led to an increase in respiratory exchange rate and heat production, particularly during the resting period (Figure 1E–I) (Hao et al., 2016; Nestoridi et al., 2012).

RYGB Induces A Late Sharp Metabolic Enhancement in White Adipose Tissue (WAT), but No Changes in Skeletal Muscle

We have previously shown that WAT increases its glucose uptake after RYGB, potentially contributing to improved metabolic function (Saeidi et al., 2013). To determine whether WAT responds differently to RYGB and dieting, we characterized the transcriptome of subcutaneous inguinal WAT (iWAT) by RNA Sequencing.

There was an upregulation of genes of the TCA cycle, β-oxidation, mitochondrial electron transport chain and fatty acid synthesis in iWAT at 9 weeks after RYGB (Figure 2A, S1A–D). Morphological evaluation of iWAT at 9 weeks after RYGB showed patches of cells that appear more similar to brown adipose tissue (Figure 2B). Consistently, we detected UCP1 protein expression in iWAT sections (Figure 2C). Several beige adipose tissue markers and components of β-adrenergic receptor (Wu et al., 2012) were upregulated in our dataset, including Ucp1, Elovl3, Dio2 and Cidea, all of which were virtually not expressed in calorie restricted mice (Figure 2D–E, S1E). Beiging of iWAT was not observed early after surgery. We then reanalyzed publicly available gene expression datasets from subcutaneous WAT samples of human patients collected before and 3 months after RYGB (Poitou et al., 2015). There was no sign of de-novo beiging of subcutaneous WAT, consistent with other clinical studies of post-surgical patients (Hoffstedt et al., 2016; Rachid et al., 2015). The human data were weakly (r=0.13) but significantly (p<0.01) correlated with our murine datasets at both time-points (Figure S1F–H).

Figure 2. RYGB induces beiging of subcutaneous white adipose tissue in DIO mice.

A. Mean log₂ fold change in expression levels of genes in metabolic pathways (GO/KEGG with FDR<0.05) in subcutaneous inguinal WAT (iWAT): RYGB-treated vs. WMS mice, 9 days (N=4 per group) and 9 weeks (wks; N= 4 for RYGB and N= 5 for WMS) after surgery; and in human patients before and 3 months (mos) after RYGB (reanalysis of dataset from (Poitou et al., 2015)). MM: mus musculus and HS: homo sapiens. Genes are sorted by fold change at the 9-week post-RYGB time-point. N.E: no expression was detected.

B–C. Hematoxylin & Eosin staining (B) and immunohistochemistry against UCP1 (C) for murine iWAT tissues, 9 weeks after surgery (N=5 per group). Scale bar=100 μm. Slides in (C) were counterstained with hematoxylin. Arrows show areas of browning.

D. Heat map showing mean log₂ fold change in gene expression of markers of brown and beige adipose tissue. Columns/comparisons/color code similar to panel A.

E. mRNA expression of selected beige/brown adipose tissue markers in individual mice 9 weeks after surgery. Each symbol represents a single animal.

** p<0.01, RYGB vs WMS independent samples t-test.

In contrast to WAT, RYGB in mice did not induce sustained long-term changes in skeletal muscle despite an early increase in oxidative energy harvest and an associated fast-to-slow twitch muscle fiber shift. This finding was also corroborated by 3 human datasets, which we reanalyzed (Barres et al., 2013; Campbell et al., 2016; Park et al., 2006) (Figure 3).

Figure 3. RYGB induces an increase in oxidative metabolism and in the number of slow twitch skeletal muscle fibers early after the surgery, but the changes are not sustained in the long-term.

A. Mean log₂ fold change in expression levels of genes of regulated metabolic pathways (entire gene sets of GO/KEGG pathways with FDR<0.05). RYGB-treated vs. WMS mice, early (9 days) and late (9 weeks; wks) after surgery (N=4 for all groups); and human patients 3 months (mos) after RYGB (reanalysis of dataset from (Campbell et al., 2016). β-ala: β-alanine metabolism, MM: mus musculus and HS: homo sapiens. Genes are ranked by fold change at the 9-day post-surgery timepoint. The same color map is used in the entire figure. N.E: no expression was detected.

B. Partial metabolic network for lipid and β-alanine metabolism showing genes with FDR<0.05, 9 days after RYGB. Differentially expressed genes in this subset are red or blue if they were up or down-regulated respectively. None of these genes were differentially expressed 9 weeks post-RYGB.

C. Heat map for mean log₂ fold change in expression of markers of slow twitch and fast twitch skeletal muscle fibers. Columns/comparisons/color code similar to panel A.

D. mRNA expression levels for markers of very fast, fast, medium and slow twitch muscle fibers of individual mice 9 days after surgery. Each individual symbol represents a single mouse. * p<0.05, ** p<0.01; both for RYGB vs. WMS independent samples t-test.

E. Immunohistochemical staining (and quantification) of skeletal muscle sections against MYH7, a marker of slow twitch fibers (N=5 per group). Slow twitch fibers (arrows) were rare in WMS (left panel) and more common 9 days after RYGB (right panel). Scale bar=100μm. Slides were counterstained with hematoxylin. * RYGB vs. WMS independent samples t-test, p<0.05.

F. Clustering of mice according to slow and fast twitch muscle fiber. Heat map of mean log₂ fold change in gene expression of individual animals over the log₂ average of each gene. Each column represents a single mouse. Each row is a single gene (gene names are omitted for clarity). R: RYGB, S: WMS.

G–H. Heat map of mean log₂ fold change in expression of secreted ligands (G) and metabolic transcription factors (H), which were found to significantly correlate with the induction of slow and fast twitch muscle fibers 9 days post-RYGB (see “Quantification and Statistical Analysis” in Methods for details). Comparison of RYGB-treated vs. WMS mice. Expression of genes regulated in endurance exercise such as Fndc5 and the transcription factors Esrrg and Esrrb paralleled the induction of slow twitch muscles.

I. Correlation of changes in gene expression for all differentially expressed genes (FDR<0.05) between murine and human datasets. ** p<0.01 for Pearson (r).

Hepatic Metabolic Responses Exhibit A Temporal Pattern after RYGB

We then sought to examine the effects of RYGB and dieting on the liver. RNA sequencing of liver samples revealed opposite early and late responses to RYGB compared with WMS. Early after surgery, RYGB primarily affected lipid metabolism. We detected an increase in triglyceride breakdown and β-oxidation with key genes such as Lpl, Pnpla2, Cpt1a upregulated by 2.7, 1.4 and 1.7-fold accordingly (Figure 4A–B). In contrast, fatty acid and triglyceride synthesis pathways were strongly repressed and the levels of Acly, Fasn and Scd1 decreased by 2.5, 3.9 and 46-fold accordingly. Histological analysis revealed reduced lipid accumulation in the liver (Figure 4C), suggesting that hepatic lipids were used to fuel lipid metabolism.

Figure 4. Hepatic metabolic responses exhibit a temporal pattern after RYGB.

A. Heat map of mean log₂ fold change in expression levels of genes of differentially regulated metabolic pathways. RYGB-treated vs. WMS mice 9 days (N=4 per group) and 9 weeks (wks; N= 5 for RYGB and N= 4 for WMS) after surgery; and human patients less than 1 year (yr) post-bariatric surgery vs. lean patients, who did not undergo surgery (reanalysis of dataset from (Ahrens et al., 2013)). MM: mus musculus and HS: homo sapiens. Here and in entire figure, genes are sorted by fold change at the 9-day post-RYGB time-point. N.E: no expression was detected.

B. Partial metabolic network for differentially expressed genes related to lipid, glucose and amino acid metabolism (FDR<0.05). RYGB-treated vs. WMS mice. Genes were mapped red or blue if they were up or down-regulated respectively. Regular fonts: 9 days after surgery, bold fonts and within brackets: 9 weeks after surgery.

C. Hematoxylin and Eosin staining of liver sections (N=5 per group). Large lipids droplets (black arrowhead) were prevalent in WMS mice and rare after RYGB. Scale bar=50μm. Quantification of lipid droplet area. * RYGB vs. WMS independent samples t-test, p<0.05.

D. Immunohistochemical staining for PPARα in liver sections (N=5 per group) displaying increased nuclear localization of PPARα after RYGB. Scale bar=50μm, 250μm in expanded panel. Slides were counterstained with hematoxylin.

E. Mean log₂ fold change in expression levels of PPARα target genes that were differentially expressed (FDR<0.05) in RYGB-treated compared to WMS mice at both 9-day and 9-week time-points.

F. Pparα activation score 9 days and 9 weeks after RYGB.

G. Pathway analysis of serum metabolites, measured by HILIC-POS in human patients. Comparison between RYGB-treated patients and controls with obesity (BWH cohort, RYGB N=25, Controls N=12). Many pathways involved in nitrogen metabolism were significantly enriched, suggesting that RYGB induces substantial changes in amino acid metabolism.

The early metabolic profile was nearly reversed at the late post-surgery time-point. The temporal pattern of respective activation and repression of PPARα target genes at early and late time-points further reflects the opposite trends of metabolic function (Figure 4D).

Amino acid catabolic enzymes were upregulated in RYGB compared to WMS 9 weeks after surgery (Figure S2A). Alanine, glutamine and proline degradation pathways were upregulated, likely activating urea cycle and feeding cataplerotic reactions of the TCA cycle (Figure 4A–B). Other metabolic pathways involving amino acid synthesis and degradation, and in particular 1-carbon metabolism were also upregulated (Figure S2A), suggesting an increase in hepatic nitrogen load.

Examining the differentially regulated transcription factors, we found changes in the expression of genes of the core circadian clock 9-weeks after RYGB (Figure S2B–C, 7C). Several studies have shown that meal timing and food composition can alter circadian phase (Kohsaka et al., 2007; Vollmers et al., 2009). Thus, the changes in circadian network may be an artifact resulting merely from the calorie restriction of the dieting group. However, circadian shifts may also be physiologically important consequences of RYGB itself, for example due to altered nutrient absorption or differences in feeding patterns (Laurenius et al., 2012; Shin et al., 2013). As described in detail in Data S1, we found that the changes among the differentially expressed genes following RYGB were best but not entirely explained by a phase shift of clock-controlled genes of 8 hours. Restricted feeding could explain approximately 60% of the changes in diurnally expressed genes, but less than 2.5% of the changes in all differentially expressed genes observed in the RYGB vs. WMS comparison.

Figure 7. Inter-organ integration of RYGB-induced metabolic responses 9 days and 9 weeks after surgery in mice.

A–B. Principal component analysis (PCA) of log₂ fold change in metabolic gene expression across the 6 studied tissues in mice. Orange circles denote genes with a Z-score smaller than −2.5 or greater than 2.5 (see Quantification and Statistical Analysis).

C–E. Mean log₂ fold change in gene expression levels across tissues in: (C) genes associated with transcription factors that were found to play a role in the coordination of the transcriptional response based on the PCA, as shown in panels A–B. See also supplementary file Data S1. RYGB-treated vs. WMS mice; duod: duodenum, jejun: jejunum, (D) Il33 (red) and Ghr (blue), (E) components of growth hormone receptor pathway.

F. Immunohistochemical staining for GHR in liver sections. Scale bars: 50μm. Slides were counterstained with hematoxylin.

G. Pearson correlation matrix (mean log₂ fold change in differentially expressed genes) across tissues and time-points in mice (RYGB-treated vs. WMS). Gray denotes a correlation of 1 to itself. Dashed gray line corresponds to correlation of a tissue to itself 9 days and 9 weeks after surgery.

* p<0.05, **p<0.01, both for RYGB vs. WMS independent samples t-test.

To determine whether the liver in human patients exhibits a similar gene expression profile in response to RYGB, we reanalyzed a dataset comparing the gene expression profile of liver biopsies obtained from patients less than one year post bariatric surgery and biopsies derived from the healthy part of the liver of patients undergoing procedures for liver cancer (Ahrens et al., 2013). There was a negative correlation between our murine, early post-RYGB and this long-term human dataset (r=−0.49, p<0.01), further supporting a temporal pattern in regulation. Core circadian network and circadian regulated genes, transcription factor expression and amino acid degradation pathways were correlated between the human and the late post-RYGB datasets (Figure 4A, S2D–F).

To further examine whether amino acid metabolism is affected by RYGB, we used Hydrophilic Interaction Liquid Chromatography (HILIC-POS) and analyzed the serum amino acid profile of a group of human patients, who underwent RYGB and well-matched controls with severe obesity. There was a significant decrease in several essential amino acids and an increase in several nonessential amino acids, which is consistent with increased amino acid utilization or protein deprivation (Figure S2G–H). An enrichment analysis showed the biochemical processes, in which the statistically different metabolites participate (Figure 4G). Overall, this serum amino acid metabolic profile was consistent with the hepatic gene expression signatures of enhanced amino acid metabolism.

RYGB Elicits Adaptive Functional and Metabolic Responses in Duodenum, Jejunum and Ileum

The hallmark of RYGB is a reconfiguration of intestinal anatomy and thus, all the physiological and molecular effects of the surgery originate from this anatomical change. Intestinal mucosa and especially the villi have mechanisms to sense the microenvironment and elicit a response according to local conditions. To interrogate these responses, we performed RNA sequencing on samples enriched in villi from the duodenum, jejunum and ileum in the acute and late postoperative period.

We observed an increase in HIF1α pathway activation with increased expression of Hif1α and its targets within the glycolytic metabolic pathway, in the jejunum, early and late post-RYGB (Figure 5A, S3A–B). Expression of genes involved in gluconeogenesis was downregulated in the jejunum (Figure 5A).

Figure 5. RYGB induces a distinct metabolic adaptation in the three segments of the small intestine.

A–C. Heat map of mean log₂ fold change in gene expression. Villi collected from duodenum (duod.), jejunum (jejun.) and ileum. RYGB-treated vs. WMS mice, 9 days and 9 weeks after surgery (N=5 for all groups); and biopsies we collected from the jejunum (villi and crypts) of patients before vs. one month after RYGB. MM: mus musculus and HS: homo sapiens. Genes are sorted by fold change in the jejunum at the 9-week post-surgery time-point.

D. Immunohistochemical staining for REG3b in jejunal sections (N=5 per group). In WMS mice, REG3b was located at the base of the villi, while after RYGB, REG3b was detected throughout the villi. Scale bar=200μm. Slides were counterstained in DAPI.

E. Heat map of mean log₂ fold change in expression of genes involved in intestinal response to bacteria (FDR<0.05). Columns/comparisons as in panel A.

Fatty acid but not triglyceride biosynthesis was induced in the jejunum in both time-points and repressed in the duodenum early after surgery (Figure 5B). Jejunal cholesterol synthesis was strongly upregulated at both time-points but repressed in the duodenum early after surgery (Figure 5C). Ldlr, which is key for basolateral cholesterol uptake was dramatically upregulated by 10 and 6.9-fold in the jejunum in the acute and long-term phase after surgery. Asns, which marks amino acid deprivation (Pan et al., 2003) was upregulated in the jejunum (Figure S3C). Altogether we detected an increase in the machinery responsible for basolateral nutrient uptake and utilization of glucose in the jejunum and to a lesser extent the ileum at both time-points.

In line with clinical reports (Stein et al., 2014), many processes related to micronutrient metabolism were affected (Figure S3D–G). Although pathways of enterohepatic bile acid regulation did not significantly change in our data, early after surgery, ileal bile acid transporters were upregulated as was the expression of Fgf15, the hormone regulating bile acid metabolism. In contrast, in the long-term postoperative phase, Fgf15 and Cyp7a1, the rate-limiting enzyme for hepatic bile acid synthesis, were downregulated in most mice (Figure S3H–I).

There was a robust induction of the oxidases Duox2, Duox2a and Cyba, the induced nitric oxide synthase Nos2, antibacterial genes such as Areg and Reg3b and the mucin Muc4 especially in the jejunum (Figure 5D–E, S4A). The response to reactive oxygen species (ROS) was activated in the jejunum and repressed in the duodenum shortly after surgery. Cytokines expressed by either immune or epithelial cells such as Il33 (Schiering et al., 2014) were upregulated in the jejunum (Figure S4B–C). These data suggest that RYGB elicits a strong protective response in intestinal mucosa, possibly in response to changes in bacterial luminal ecology, the chemical composition of chyme and other stressors such as mechanical forces.

RYGB Induces A Similar Metabolic Adaptive Response in the Jejunum of Human Patients

We then analyzed samples of jejunal mucosa collected from patients undergoing RYGB. These biopsies included both villi and crypts (Methods). Transcriptional analysis showed remarkable similarity in metabolic responses between patients and mice (Figures 5A–C, 5E, 6). As in mice, the results suggest that the jejunum of patients may increase utilization of systemic glucose through glycolysis and induction of HIF1α and may augment uptake of glutamine and cholesterol from circulation, while decreasing luminal uptake and export of nutrients (Figure 6B). HK2, G6PD and HMGCR were upregulated at the protein level (Figure 6C). 4EBP1 and PDK1 were hyper-phosphorylated after RYGB, suggesting that mTOR pathway activation may facilitate the upregulation of HIF1α and cholesterol synthesis. As in mice, oxidases, mucins and nitric oxide synthase were highly upregulated after surgery (Figure 5E, 6K). Furthermore, the number of goblet cells was increased (Figure 6H, S4E).

Figure 6. RYGB affects the metabolism and structure of the jejunal crypt-villus unit in human patients.

A. Correlation and clustering of mean log₂ fold change in metabolic gene expression between intestinal segments in mice 9 days (9d) and 9 weeks (9w) after surgery and human patients (HS; UPMC cohort). In both mice and patients, the profile of the jejunum at 9 days and 9 weeks post-surgery is highly correlated, with the profile of the ileum 9 days after surgery showing a similar expression pattern.

B. A model for the metabolic regulation in jejunal enterocytes 1 month after RYGB. Upregulated processes shown in red, downregulated in blue and unchanged in black.

C. Increase in protein levels of HK2, G6PD and HMGCR after RYGB (PUC cohort, N=13 per group). Samples were pooled and B2M was used as control. Increase in protein levels of phosphorylated 4EBP1 and PDK1.

D. Mean log₂ fold change in expression of intestinal epithelial cell markers in the jejunum after RYGB. Each column represents an individual array.

E. Mean log₂ fold change in expression of cMYC-MAD-MAX complex and induction of cell cycle regulators after RYGB.

F. Increase in protein levels of CYC2 and CYCE1 following RYGB.

G. Mean log₂ fold change in expression levels of all differentially expressed genes (FDR<0.05) following RYGB (left). Comparison with the known expression pattern of these genes along the crypt-villus axis (right), based on data of reference (Mariadason et al., 2005). Each row represents a gene; names have been omitted for clarity. Looking at the graph from top to bottom, most differentially up- or down- regulated genes following RYGB were similarly up- or down- regulated in crypts vs. villi (top) and only a small number of genes showed an opposite pattern (bottom).

H. Hematoxylin and eosin staining of jejunal sections (PUC cohort, N=13 per group). Black arrows point to crypts, yellow arrows to goblet cells. scale bar=500μm.

I–J. Quantification of number of cells per crypt (I) and crypt cross-sectional area (J) in patients after RYGB compared to controls. * RYGB vs. WMS independent samples t-test, p<0.05.

K. Validation of selected targets of glucose metabolism and cellular proliferation with RT-PCR. * one sample t-test comparing RYGB/WMS fold change to 1, p<0.05.

Markers of intestinal stem cells and Paneth cells were upregulated post-RYGB in patients (Figure 6D). Correspondingly, there was an increase in Ki67 staining, cell cycle regulators that are active in the proliferating cells of the crypts were induced at mRNA and protein level, and the c-MYC pathway was affected (Figure 6E–K, S4F–G). Analysis of all differentially regulated genes post-RYGB showed that most upregulated genes were predominantly expressed in crypts, while downregulated genes were expressed preferentially in villi (Figure 6G). Histological analysis confirmed that crypt size and number of cells per crypt were increased by over 25% after RYGB (Figure 6H–J).

Integrative Analysis of Gene Expression Patterns

After determining the molecular signatures of the acute and long-term response of individual organs to RYGB, we wanted to conduct an integrative analysis of the gene expression patterns of all the tissues that we analyzed. Therefore, we performed a principal component analysis (PCA) on the fold-change in metabolic gene expression signatures in the six tissues, at the 9-day and 9-week time-points (Figure 7A–B). The first principal component (PC1) accounted for over 40% of the variability in the data at both time-points. We then calculated the angle between PC1 and the fold-change in expression of all transcription factors expressed in these tissues. We interpret the transcription factors with a significantly small or large angle to PC1 as factors that could explain the metabolic gene expression changes we detect across tissues (Figure 7A).

Srebf1, Srebf2 and components of the Srebp complex, the core circadian rhythm network and both thyroid receptors were differentially expressed in the acute and long-term postoperative phase (Figure 7B–C). We were surprised to find that Nfkb1b aligned with PC1, possibly linking the immune with the metabolic responses. We also identified Hif1α in this independent analysis as a central systemic metabolic regulator. In the long-term postoperative phase, transcription factors most similar to PC1 included Atf2 and Atf4, which control the amino acid response pathways, the xenobiotic regulator Ahr, nuclear receptors Esrrg and Nr2f2, and signaling factors such as Stat3 (Figure 7B).

Growth Hormone Receptor Is Downregulated and Il33 Is Upregulated Systemically after RYGB

As a second approach for examining the systemic effects of surgery, we searched for genes that exhibit either up- or downregulation across tested tissues. Il33, an atypical cytokine (Miller et al., 2010) that has been implicated in many diseases including diabetes was upregulated in most tissues following RYGB (Figure 7D). In addition, Ghr, coding for growth hormone receptor was downregulated in most tissues following surgery. Accordingly, Igf1 was downregulated in the liver, while Socs3, which can inhibit growth hormone signaling, was upregulated (Figure 7D–E). Immunostaining in the jejunum, ileum and liver confirmed these results (Figure 7F, S5). To further explore this finding, we compared the hepatic signature of RYGB-treated mice to those from male mice lacking GHR signaling (ames mouse strain) vs. wild type animals (Amador-Noguez et al., 2004; Estep et al., 2009; Swindell, 2007; Yang, 2006). Consistent with a global decrease in GHR signaling, these gene expression signatures positively correlated in the acute post-RYGB phase (Figure S5I).

Similar and Distinct Metabolic Responses in The Acute and Long-term Phase after RYGB

Our results show that tissues respond differently to RYGB, and moreover, the same tissue may change the regulation of its metabolic processes between the acute and the long-term phase after surgery. To quantify these observations, we determined the correlations of significantly regulated metabolic genes across the tissues (Figure 7G). Notably, the jejunum exhibited very similar changes in metabolic expression profile in the acute and long-term postoperative phase (r=0.64, p<0.0001). The persistence of this response most likely reflects a tight link to the ongoing changes of the local microenvironment, which result from the anatomical reconfiguration of this intestinal section. Conversely, the liver displayed a negative correlation between the early and the long-term response (r=−0.21, p<0.0001), reflecting a transition from post-surgical fasting to a new metabolic steady state.

Discussion

In this study, the simultaneous analysis of many tissues provides an opportunity to generate an integrative organismal perspective of the differences in the mechanisms underlying the metabolic improvement in the acute and long-term phase after RYGB surgery and dieting.

Acute Response to RYGB vs. Dieting

The energetically costly intestinal remodeling, the postoperative immune response and the reduced metabolic efficiency can account for the an augmented acute, adaptive response to fasting and the glucose-sparing metabolic processes seen in the liver and muscle after RYGB. In the duodenum, most metabolic processes were downregulated, possibly due to the lack of nutrient exposure. Remarkably, the jejunum entered an anabolic state very early after surgery; and consistent with recent studies, the observed metabolic and morphological adaptations may effectively transform it to a fuel sink (Cavin et al., 2016; Ku et al., 2017; Saeidi et al., 2013).

Long-term Response to RYGB vs. Dieting

In the steady state, the predominant late metabolic change in the liver after RYGB appeared to be an enhancement of amino acid catabolic processes. This may reflect an increase in nitrogen load and may account for the reduction in circulating amino acid levels, which is a consistent finding in clinical blood metabolite profiling studies (Laferrere et al., 2011; Magkos et al., 2013).

In this phase, RYGB-treated animals appear to be in a new metabolic steady state that is driven primarily by changes in the gut and the white adipose tissue, but not the muscle. Murine and human data on the long-term effects of skeletal muscle showed a limited transcriptional response to surgery. RYGB induced beiging of white adipose tissue, a finding that is corroborated by previous studies in mice (Neinast et al., 2015), but remains controversial in human studies (Rachid et al., 2015). The small intestine and particularly the jejunum maintained the expression pattern of the acute phase. Strikingly, very similar metabolic and antimicrobial adaptations were observed in jejunal biopsies taken from patients shortly, just one month, after surgery.

The temporal pattern of the metabolic changes in various organs and especially in the liver may not be surprising. Notably, the curve describing body weight as a function of time after RYGB in animals (Figure 1) and human patients (Sjostrom et al., 2007)) consists of 3 separate parts with different rates of change (slopes), which may reflect the temporal pattern of the changes in metabolic pathways.

Inter-organ Integration of RYGB-induced Metabolic Responses

Adaptation to metabolic changes requires inter-organ integration of metabolic pathways. Our data highlighted the central role of metabolic regulators such as SREBP2 and HIF1α in the global gene expression response early after surgery. Intriguingly, genes of the circadian clock network exhibited a similar pattern of regulation across murine tissues. This synchronicity may be important for coordination of the anabolic processes of the intestine with the catabolic processes seen in the other tissues. Recent studies have now linked SREBP2, HIF1α, and the circadian clock, and it is possible that this transcriptional network is central in the metabolic response to RYGB (Adamovich et al., 2017) (see also Data S1).

Another striking finding in our study is the global downregulation of growth hormone receptor, in the acute and long-term postoperative phase. Although downregulation of GHR signaling is consistent with the strong fasting response we observed in the acute postoperative phase, it is surprising that it was maintained in the long-term, despite the ad libitum access to food. This phenotype has been observed in GHR deficient animals, which similarly to RYGB-treated mice, also exhibit increased energy expenditure and increased food intake (Bergad et al., 2000; Morrison et al., 2016). GHR and insulin signaling often oppose each other, and reduction in GHR signaling may effectively increase insulin sensitivity and glycemic control (Dreval et al., 2014). Consistent with our findings, circulating levels of IGF1, the downstream effector of GHR signaling are decreased immediately following surgery in rodents and human patients (Itariu et al., 2014; Rubino et al., 2004; Stemmer et al., 2013). Thus, a global reduction in GHR signaling may be another mechanism by which RYGB improves glycemic control.

Finally, it is important to note that Il33 expression increased in many tissues. IL33 induces a protective Th2 immune response, while higher IL33 serum levels are associated with a healthier metabolic profile and beiging of white adipose tissue (Miller et al., 2010). Although RYGB appeared to induce a stronger immune response in most of the tissues we studied, there was no histological evidence for inflammation, particularly in the intestine. This raises the possibility that the observed immune response is part of a positive remodeling process.

Limitations of Study

Our study centered on a mouse model for RYGB and a diet- and weight- matched sham-operated control. This animal model does not capture certain elements of the postoperative recovery of human patients. In humans, for example, surgery or fasting does not induce the substantial degree of weight loss that is seen in mouse models within days. For this reason, we validated and interpreted the results of the murine model by comparing them to human data. Nonetheless, human studies have inherent variability and do not control for factors such as diet and many clinical and metabolic parameters.

The primary outcome of the study was the transcriptional signatures of the tissues. This allows unbiased, high resolution comparison among different tissues, facilitates an integrative analysis of the results and enables comparative analysis between mouse and human data. Although transcriptional regulation correlates with the underlying activation or repression of metabolic and signaling pathways, the correlation is not perfect, especially in metabolic circuits, in which protein phosphorylation and substrate inhibition control the activity of many enzymes (Schwanhäusser et al., 2011). Importantly, although our findings highlight the role of HIF1α and SREBP2 in the coordination of the transcriptional response; both HIF1α and SREBP2 activity is regulated post-transcriptionally and through nuclear localization. Therefore mRNA levels of these factors may correlate with but are not indicative of their actual activity in vivo. As described in detail in Methods and in Data S1, there are several methodological issues associated with the interpretation of the data that are based on dieting and calorie restriction models. We followed published guidelines and took special care to standardize the metabolic status of the tissues, by collecting them following the same fasting protocol and duration. Furthermore, the central findings were validated at the protein level. Overall, the conserved molecular responses to RYGB in rodents and human patients may provide an additional level of confidence in the conclusions that can be reached about the mechanisms of action of RYGB.

Conclusion: RYGB Establishes A New Metabolic State

Our data provide a resource to map the many metabolic changes following RYGB. They stress that although dieting and RYGB can bring about the same degree of weight loss and metabolic improvement, the physiology post-RYGB is very different. Taken together with other studies characterizing metabolic, neuronal and endocrine regulation, it emerges that a new metabolic state is established following RYGB, characterized by the upregulation of costly and inefficient metabolic processes. In addition, the intestine responds to the anatomical changes within days, engaging energetically expensive processes. The numerous parallel mechanisms that surgery elicits may contribute to its remarkable ability to sustain the metabolic improvement compared with dieting; and may underlie the reason it has been the most successful treatment option across genders, age groups, ethnic and racial backgrounds, BMI range, comorbidities and genetic backgrounds. This new metabolic state could guide us to develop new means to achieve durable weight loss and metabolic improvement without the surgery.

STAR Methods

Contact for Reagent and Resource Sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Nicholas Stylopoulos (Nicholas.stylopoulos@childrens.harvard.edu).

Experimental Model and Subject Details

Animals

All experiments were approved by and performed in compliance with the Institutional Animal Care and Use Committee of Boston Children’s Hospital. Animals were individually housed and were maintained on 12-hour light-dark cycle (lights on at 7:00 am) in facilities with an ambient temperature of 19–22 °C and 40–60% humidity. C57BL/6 mice with diet-induced obesity (Jackson Laboratory, ME, strain DIO C57BL/6J DIO, Stock No 380050) were used for all studies. Obesity was induced in these mice by feeding the animals ad libitum with a high fat diet (HFD) that provides 60% of total energy as fat, 20% as carbohydrate and 20% as protein (D12492 diet, Research Diets Inc, New Brunswick, NJ). At the time of surgery all animals weighed 45 – 50 g. All mice in these studies were male, because a) male mice develop higher degree of obesity than female animals when exposed to high fat diets and b) most obesity and metabolic studies use male mice and this facilitates comparisons with other datasets. In Table 1, we present the sample size for described experiments and corresponding figures. For each RYGB-treated group (9-days, 9 weeks) the overall mortality was 2 out of 10 operated mice and this was consistent with previous reports using the same model and surgical technique (Liou et al., 2013). The same animals were used for all the analyses (at each timepoint) and we included only mice that were from the same litters. All animals were assigned to 3 groups (RYGB, sham operation with ad libitum feeding and sham operation with weight matching) randomly, before any intervention was performed.

Human Patients

There were no published datasets of the gene expression profile of the intestine after RYGB. We analyzed samples of a tissue bank that we have established and includes intestinal samples collected from three different sources. We performed the microarray study in patients before and 1 month after RYGB. Importantly, these samples were collected from research subjects that participated in research studies, which involve the analysis of intestinal samples at protocol specified intervals. Thus, they were not simply collected at convenient time points, when other surgical procedures were done for a clinical indication (e.g. for management of surgical complications). Specifically, the tissues have been collected under the registered clinical trial NCT02710370 at the University of Pittsburgh Medical Center (UPMC). To be able to directly compare the human signatures to the murine intestinal profile, we analyzed samples collected from the same patients, at the time of surgery and the earliest timepoint of that study, which is 1-month (15 – 45 days) post-RYGB. The mean age was 37.1 years and the BMI at the time of the biopsy was 41.2, while the preoperative BMI was 42.9. For a detailed description of the clinical trial, the study population and the inclusion/exclusion criteria, please see: https://clinicaltrials.gov/ct2/show/NCT02710370

Validation experiments were performed in two other cohorts. Specifically, for gene expression, histological and protein expression studies, we analyzed samples that have been collected from two separate cohorts of research subjects in two different Institutions (Brigham and Women’s Hospital - BWH, Boston MA and the Pontifical Catholic University of Chile - PUC, Santiago Chile). BWH cohort included subjects who had undergone RYGB at least 1 year prior to tissue collection (mean 52 months) and well-matched unoperated controls with obesity. The mean age was 56.3 (RYGB) and 42.5 (Controls) years. The BMI at the time of the samples’ collection was 37.7 (RYGB) and 34.4 (Controls), while the preoperative BMI of the RYGB-treated group was 48.4. PUC cohort included subjects who had undergone RYGB 1–2 years prior to tissue collection (mean 18.4 months) and well matched controls, who had undergone tissue collection at the time of their RYGB. The mean age was 46.8 (RYGB) and 47.1 (Controls) years. The BMI at the time of the samples’ collection was 28.9 (RYGB) and 38.9 (Controls), while the preoperative BMI of the RYGB-treated group was 39.5.

The inclusion and exclusion criteria were common in all studied cohorts. There was no difference in age, sex, preoperative BMI and comorbidities between the three cohorts. Written informed consent was obtained from all subjects, and all protocols were approved by the Institutional Review Board of the University of Pittsburgh Medical Center, Brigham and Women’s Hospital and the Ethics Committee of the Faculty of Medicine of the Pontifical Catholic University of Chile.

Method Details

Interventions in Mice

DIO mice were subjected to RYGB or a sham operation. After the surgery, mice were placed on an elemental liquid diet and then were gradually advanced to the same high-fat diet they consumed before the operation. This feeding protocol is necessary for the recovery of the animals and resembles the postoperative feeding protocol of human patients. Sham-operated controls consumed the same type of diet as the animals in the RYGB-treated group. One group of sham-operated animals was calorie restricted to match the body weight of RYGB-treated mice (Weight Matched Sham; WMS).

RYGB was performed as previously described in detail (Kucharczyk et al., 2013; Nestoridi et al., 2012). The operation in human patients usually includes a rearrangement of 20–30% of the small intestine (i.e., Roux limb and biliopancreatic limb each about 10%–15% of total intestinal length). The total length of small intestine was measured, the ligament of Treitz identified and the jejunum divided at the appropriate 3–5 cm from the ligament of Trietz. A gastric pouch was formed and was connected to the jejunum with a side to end anastomosis, forming the Roux limb. The distal stomach was then closed without compromising blood supply. The biliopancreatic limp, comprising of the duodenum and proximal jejunum was anastomosed to the jejunum 5 cm distal to the gastric pouch. 8–0 Vicryl sutures (Ethicon Inc.) were used for the anastomoses; 6–0 silk sutures (Ethicon Inc.) were used to close the abdominal muscular layer and 6–0 PDS for the skin incision.

In all murine experiments, we used as controls mice undergoing a sham operation. The sham operation consisted of laparotomy, jejunal transection and reanastomosis.

Postoperative Care in Mice

RYGB and sham operated mice received the same post-surgical care as in our previous studies (Kucharczyk et al., 2013; Nestoridi et al., 2012). Mice were fasted the day of the surgery and water was provided either the evening of the surgery or the morning after. Analgesia and fluids were also administered subcutaneously. Water and elemental liquid diet (Peptamen 1Cal, Nestle Inc.) were provided ad libitum as early as postoperative day 2 and up to postoperative day 9–14. RYGB-treated and a group of sham-operated mice were then placed back ad libitum on the same high fat diet they consumed before surgery (D12492 diet, Research Diets Inc, New Brunswick, NJ) until the end of the study.

Dieting Mouse Model (Calorie Restriction - Weight Matched Sham Model)

Sham-operated mice started to regain weight immediately after re-introduction of high fat diet. This gain was controlled through food restriction in the group of dieting (weight matched sham; WMS) mice. These mice consumed the same type of diet as the other groups, but the total amount of food intake was adjusted daily, so their body weight matched the mean body weight of the animals in the RYGB-treated group. Overall, these animals had to consume approximately 40% less food than the mice of the RYGB group.

Specifically, mice underwent first a sham operation. In the first 9–14 days, animals had ad libitum access to the same liquid diet (Peptamen 1Cal, Nestle Inc.) that was consumed by the RYGB-treated mice. At the end of this period, there was no difference in the body weight of the two groups of animals. Upon switching back to solid, high fat diet (D12492 diet, Research Diets Inc.), we measured food intake daily in the RYGB-treated animals. Animals in the WMS group were given the amount of food that was required to sustain the same body weight, on average, as their RYGB-treated counterparts. Food was provided at noon every weekday, while animals had ad libitum access in the weekends. Over the entire study period, the animals in the dieting group consumed approximately 40% less food than the RYGB-treated animals. Importantly, the animals in the WMS group weighed 32–37 gram, which is higher than the body weight of normal, chow-fed animals. Thus, the animals were not calorie restricted or cachectic. They had the ability to gain weight compared to regular, normal chow fed animals. Another important methodological detail is that all animals in both groups were handled the same time. Specifically, every day at noon, husbandry/food intake measurements were performed in the RYGB-treated group and husbandry and food was provided in the WMS mice. As shown in Figures 1E and 1G, there was no difference in RER and heat production between RYGB-treated and WMS animals at that time of the day. Overall, the feeding methods that we followed are the same as the ones used in other studies on calorie restriction of lean animals, including those that have performed gene expression studies in lean, calorie-restricted animals. Recent studies (Acosta-Rodríguez et al., 2017; Ellacott et al., 2010) have highlighted the behavior and the limitations of this feeding method, because the restricted animals eat quickly and get habituated to the clock, which may alter metabolic patterns. One of the essential recommendations of these studies to improve the accuracy of the experiment is to house the animals in single cages, which we followed in our studies. All our animals were singly housed for the entire duration of the study.

Metabolic Assessments in Mice

Blood glucose levels were measured using a blood glucose meter (LifeScan Inc., Milpitas, CA). Indirect calorimetry was performed 5 weeks postoperatively. The Oxymax Lab Animal Monitoring System (CLAMS) was used (Columbus Instruments, OH, USA). All mice were placed in the cages of the indirect calorimeter system for acclimation before experimental measurements. The CLAMS system is located in a room where ambient temperature can be controlled and the temperature inside the cages was documented to be always 29 °C. This temperature is in the thermoneutrality zone for mice, so all measurements were performed under thermoneutral conditions. Measurements were performed for 72 hours during which animals had access to HFD and water ad libitum. Oxygen consumption, carbon dioxide production and heat production were recorded for each animal. Resting energy expenditure was calculated by correlating oxygen consumption, food intake and activity based on the recordings of the metabolic system.

Tissue Collection in Mice

Two postoperative time-points were chosen for the analysis of the tissues. In our experience, approximately 9 days are required for mice to recover well, to ensure long-term and healthy survival and to reduce the effects of post-surgical trauma on data analysis. On the 9^th postoperative day, mice are ready to advance gradually from elemental diet to the high fat diet they consumed before the surgery. Thus, the 9^th postoperative day was chosen as a meaningful time-point that could allow study and interpretation of the acute effects of RYGB. After this day, the animals start consuming gradually (as a supplement to the elemental liquid diet), and by postoperative day 14 exclusively, the high fat diet they consumed before the surgery. By the 9^th postoperative week, there is no further change in body weight or other metabolic parameters and thus the 63^th postoperative day (the end of the 9^th postoperative week) was chosen as a time-point to study the long-term effects of RYGB.

For tissue collection, animals were fasted at 6pm and tissues were collected between 9AM-11AM the following day. Animal (WMS or RYGB) order alternated each day and tissues were always collected from pairs and not from a single group. During tissue collection, we observed that there was no food in the entire gastrointestinal tract in any of the animals in both groups. Liver, inguinal white adipose tissue, gastrocnemius muscle, duodenum, jejunum and ileum were analyzed. Intestinal samples were opened longitudinally and washed with ice cold PBS and then suspended in 5mM EDTA for 10 minutes. Epithelium was gently scraped and the crypts were removed by a 70um cell strainer (Falcon), leaving samples enriched in villi. This could be confirmed by examining the samples under a microscope. Intestinal villi, liver and skeletal muscle were suspended in RNA later (ThermoFisher Scientific), and inguinal white adipose tissue was frozen in liquid nitrogen.

RNA Sequencing and Microarray

RNA kits (Qiagen RNeasy, Qiagen) were used to extract RNA. RIN scores>8 were used for further analysis. RNA-Seq libraries were prepared using PrepX kits on Apollo 324 (Wafergen). For murine tissues, a HiSeq 2500 (Illumina) was used for sequencing, and STAR (Dobin et al., 2013) for alignment within the Bioconductor environment (Huber et al., 2015; Ritchie et al., 2015). Differential expression was determined using voom (Law et al., 2014). For the gene expression profiling of the intestine of human patients, we used the GeneChip™ Human Transcriptome Array 2.0 (Affymetrix), which covers more than 285,000 full-length transcripts, 245,000 coding transcripts and 40,000 non-coding transcripts. Benjamini-Hochberg correction was used to account for multiple hypothesis testing.

Immunohistochemistry and Western Blotting

Samples were placed in 10% formalin buffer for 24 hours and then transferred to ethanol and embedded in paraffin. Paraffin embedded tissues were de-paraffinized using HistoClear (National diagnostics). Slides were then boiled in pH 6.0 citric buffer for antigen retrieval. Morphological analysis was peformed using ImageJ (Research Services Branch, National Institutes of Health, Bethesda, MD).

Tissue lysate preparation, SDS-PAGE and Western blotting was preformed as previously described (Saeidi et al., 2013). The following antibodies were used: GHR (1:50, Abcam 202964), UCP1 (1:100, ThermoFisher Scientific, PA5–29575), PPARα (1:100, Abcam 8934), Il33 (1:50 R&D Systems af3626), Reg3b (1:100, ThermoFisher Scientific, AF5110), MYH7 (1:100, Abcam 11083), HMGCR (1:250, Biovision 3952–100), CYCE (1:1000, Cell Signaling 4129), CDC2 (1:1000, Cell Signaling 9112), HK2 (Cell Signaling 2867), phosphoPDK1 (1:1000, Cell Signaling C49H2), Ki67 (1:250, Abcam ab66155), phospho4EBP1 (1:1000, Cell Signaling, 236B4), B2M (1:1000, Abcam ab75853). Tissues were counterstained with DAPI or Hematoxylin.

Quantification and Statistical Analysis

We searched the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/gds/) for datasets that included gene expression profiles of samples collected from human patients who have undergone RYGB. We searched using the keywords “gastric bypass” and only results that included data from samples collected before and after the surgery were included in the study. Gene expression data were then extracted and analyzed further.

Pearson correlation was used to determine similarity between gene expression datasets. Correlation between human and mouse datasets was performed on genes that were differentially regulated in the human study (p<0.01) and their mouse homologues. For the comparison of liver expression profiles in RYGB-treated, calorie restricted, ames mice and female mice, each regulated gene was assigned a score of 1, −1 or 0 for being significantly up regulated, down regulated or not regulated. Within our data set, log₂ fold change in gene expression was used.

To determine secreted ligands and metabolic transcription factors that significantly correlate with the induction of slow and fast twitch muscle fibers, we correlated the degree of induction with the mean fold change of slow twitch genes. From the Bonferroni adjusted correlations with q<0.05, we analyzed the secreted factors and metabolic transcription factors to outline key genes that likely help induce this switch.

A Mann-Whitney test was used to determine differences in glucose levels and body weight, or differences in circadian rhythm regulation. Activation scores for specific pathways were the – log₁₀ p-value for the probability of finding differentially expressed genes within a specific subgroup, compared with the entire gene population, using the hypergeometric distribution.

Pathway analyses were performed using Metacore (Clarivate Analytics). KEGG database was used to curate metabolic gene lists. Other lists were based on GSEA curated lists (Subramanian et al., 2005) and literature data for muscle fiber type markers (Egan and Zierath, 2013), PPAR target genes (Rakhshandehroo et al., 2010), HIF1α targets (Benita et al., 2009), transcription factors (RIKEN database) and circadian regulated genes (Wu et al., 2012; Wu et al., 2017). The list of core circadian network genes (clock, per2, per3, cry2, rora, rorc, nr1d1, nr1d2, dbp, hlf, nfil3, bhlhe40) was compiled based on references (Brown et al., 2012; Dallmann et al., 2014; Koike et al., 2012) and is very similar to the one used in reference (Wang et al., 2017). Since murine intestinal samples contained villi only and the human biopsies included villi and crypts, we compared the expression pattern of genes along the crypt-villus axis in the two datasets using the dataset from reference (Mariadason et al., 2005).

A MATLAB script was used for PCA analysis of fold change in expression of genes in the central metabolic pathways between RYGB and WMS over the 6 tissues. The angle between each transcription factor expressed in all 6 tissues and the first two principal components was calculated. The angle distribution was found to be normal, with mean very close to 90°. An angle over 2.5 standard deviations from the mean (p<0.0125) was considered significantly small/large and close to the principal component.

Sample size for each experiment and the information about the statistical tests that were used are reported in each figure legend. Importantly, a post-hoc analysis showed that with N=5 mice/group, we had 91% power to detect an effect of the magnitude observed in blood glucose levels between RYGB-treated and WMS groups. Since the primary goal of our studies is to determine the mechanisms by which RYGB improves glucose levels and metabolic function, we used this sample size for all murine experiments.

Data and Software Availability

All datasets and the array files have been uploaded on GEO database. The reference series is GSE113823.

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE113823

Sub-series that are linked to GSE113823 include:

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE113819

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE113821

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE113822

GSE113819 includes expression data from human jejunum before and one month after RYGB.

GSE113821 includes mRNA Seq of mouse tissues 9 days after RYGB or sham surgery

GSE113822 includes mRNA Seq of mouse tissues 9 weeks after RYGB or sham surgery.

Supplementary Material

1

Data S1. Methods, data and discussion about the effects of RYGB on circadian network. Related to Figures 4 and 7.

Highlights.

Compared to dieting RYGB leads to sustained weight loss and metabolic improvement

RYGB dynamically modulates organ-specific signaling and metabolic pathways

Many of these effects are conserved in rodents and humans

RYGB induces a new metabolic state, which may explain its remarkable effectiveness