“Dysfunctions” induced by Roux-en-Y gastric bypass surgery are concomitant with metabolic improvement independent of weight loss

Metabolic improvement is greater in the long-RL (30% of total intestine) group of nonobese GK rats

Significant changes in absorption and metabolism in the RL play roles in diabetes remission by RYGB for obese diabetic subjects3,9,12. Accordingly, we evaluated beneficial effects in diabetes remission by RYGB and relationships between such effects and RL length in the surgical procedure for diabetic but nonmorbidly obese subjects. We performed RYGB procedures with various RL lengths in nonobese GK rats (Fig. 1a; see Materials and Methods, “Surgical procedures” for details). In human clinical studies, a normal RL is ~12â€"15% of total intestinal length, which corresponds to the GK-S-12 group, while a super-long RL is ~24â€"30% of total intestinal length17, which corresponds to that in our GK-S-30 group. Next, we measured several physiological indices related to diabetes (i.e., food intake, body weight, oral glucose tolerance test (OGTT), and intraperitoneal (i.p.) glucose tolerance test (IPGTT)). In comparison with the control groups (i.e., Wistar and GK rats witho ut surgery), the five surgical groups showed remarkably reduced food intake at 72 h postoperation after adapting to normal chow. These differences persisted during weeks (wk) 1â€"6 postoperation. After wk 6, food intake did not differ significantly between the surgical and control groups. The sham operated surgical control pair-fed GK group (GK-Sham-PF) was given an amount of food equivalent to the smallest amount among the three RYGB groups (see Materials and Methods), usually that of GK-S-12 (Fig. 1b).

Body weight in the five surgical groups declined greatly during wk 1 and then gradually increased, whereas body weight in the two non-surgical control groups showed a steady, gradual increase (Fig. 1c). At 8 wk postoperation, body weight of the GK controls differed significantly from those of the GK-S-12 (regular RL), GK-S-30 (long RL), and GK-S-30R (RL excised) groups, but not from those of the GK-Sham-PF and GK-S-3 (short RL) groups. These findings indicate that the GK-S-12, GK-S-30, and GK-S-30R surgeries, but not the GK-S-3 surgery, were effective in maintaining low body weight. Similarly, a human study showed a positive correlation of RL length with weight loss18. OGTTs and IPGTTs were performed 12 wk postoperatively to assess glucose metabolism. In the OGTTs, all six GK groups showed reduced glucose tolerance in comparison with the age-controlled Wistar group (Fig. 1d). Among the five surgical groups, the area under the curve (AUC)-glucose values of the GK-S-12 and GK-S-30 groups were significantly lower than those of the GK group. The AUC-glucose value for GK-S-30 was significantly lower than those of GK-S-12, GK-S-3, and GK-S-30R (Fig. 1d). For IPGTT-AUC values, there were no significant differences in GK-S-30 rats, but remarkable increases in both GK and GK-Sham-PF rats in comparison with Wistar rats (Fig. 1e). These results indicated that nonobese diabetic GK rats benefited from long-RL RYGB surgeries in comparison with other types, and restricted food intake had no positive effects in glycemic controls. In comparison with BL and CC segments, the RL appeared to play a key role in glucose homeostasis following RYGB in nonobese diabetic GK rats; the metabolic benefits disappeared in the GK-S-30R group, which had an identical BL and CC but excised RL in comparison with those in GK-S-30 rats.

Amelioration of diabetes state following RYGB differs from both normal and diabetic states in RL in terms of gene expression

Generally, a whole-genome expression profile is analyzed to characterize a living state as well as to understand the underlying molecular mechanisms to maintain this state19. Improvements in glucose metabolism were clearly observed in GK-S-30 but not in GK-S-30R (RL excised) rats. Therefore, we measured whole-genome expression profiles of RL in GK-S-30 rats and the corresponding intestinal segment of GK-Sham-PF and Wistar rats to evaluate molecular changes in the RL resulting from RYGB. We identified 4942 differentially expressed genes (DEGs) by multiple t-tests (Fig. 2a and Supplementary Table S1) to characterize three physiological states (i.e., normal, diabetes, and amelioration of diabetes by RYGB). We performed unsupervised hierarchical clustering (Fig. 2a) and principal component analysis (PCA) (Fig. 2b) based on the 4942 identified DEGs. We observed, unexpectedly, that the three groups of rats representing different pathological and physiological states were distinctly clus tered into three independent groups, indicating that the state of diabetic amelioration following RYGB was remarkably different from both the normal and diabetic states. Next, we counted overlapping DEGs between two pairwise states and estimated their overlapping significances using a hypergeometric test (Fig. 2c). RYGB-associated DEGs (i.e., GK-S-30 vs GK-Sham-PF rats) had relatively few overlaps with diabetes-associated DEGs (i.e., GK-Sham-PF vs Wistar rats), indicating that RYGB may not recover all molecular “dysfunctions” associated with diabetes but still improves glycemic control.

Fig. 2: Analyses of transcriptomic profiles among three physiological states.

a Heatmap illustrating dynamics of 4942 DEGs among the three state groups (normal Wistar, diabetes-GK-PF, diabetes remission-GK-S (short for GK-S-30)). Unsupervised hierarchical clustering was performed to distinguish among physiological states. The three types of rats were distinctly clustered into three independent groups, indicating that physiological state of diabetes remission by RYGB differed strongly from both normal and diabetic states. b PCA results showing visually that the three types of rats were clustered distinctly into three independent groups, confirming similar results from hierarchical clustering. c Few overlapping DEGs in comparison with pairwise groups, with no statistical significance by the hypergeometric test (all P values are 1). d Schematic diagram illustrates four transition groups (positive, unchanged, opposite, and newly changed) of DEGs to fractionize functional roles of DEGs and clarify relationships between RYGB-associated and diabetes-associated cha nges at the gene expression level. S: surgery. e DEGs in the four transition groups were counted. Most of the DEGs belonged to the unchanged and newly changed groups. Positive group was out of proportion to unchanged group.

DEGs can be categorized into four transition groups for representing different directions of change induced by RYGB

To further clarify the molecular mechanisms of diabetes remission by RYGB, we defined four transition groups (positive, unchanged, opposite, and newly changed groups) (Fig. 2d and Supplementary Table S1) against surgery to fractionize the functional roles of DEGs and interpret relationships with RYGB-associated changes at the gene expression level. The positive group represented diabetes-associated gene expression recovered by RYGB, including DEGs without significant differences between GK-S-30 and Wistar rats, but with significant differences between GK-Sham-PF and Wistar rats or between GK-Sham-PF and GK-S-30 rats. The unchanged group represented diabetes-associated gene expression not improved by RYGB, including DEGs without significant differences between GK-S-30 and GK-Sham-PF rats, but with significant differences between GK-Sham-PF and Wistar rats or between GK-S-30 and Wistar rats. The opposite group represented diabetes-associated gene expression relatively worsened by RY GB, including DEGs with significant differences among GK-Sham-PF, GK-S-30, and Wistar rats. Opposite changing directions of GK-S-30 and Wistar rats were in comparison to corresponding differences between GK-Sham-PF and Wistar rats. The newly changed group represented diabetes-irrelevant genes abnormally expressed after RYGB, including DEGs without significant differences between Wistar and GK-Sham-PF rats, but with significant differences between GK-S-30 and Wistar rats or between GK-S-30 and GK-Sham-PF rats. Thus, gene expression of these newly changed genes in GK-Sham-PF rats was originally at the same level as that in normal (Wistar) rats, but did not remain “normal” after RYGB and actually underwent significant changes to “abnormal” despite amelioration of diabetes.

Next, we individually separated DEGs into these four transition groups. We found, unexpectedly, that most of the DEGs belonged to the unchanged and newly changed groups, while the positive group representing recovered genes was relatively small (Fig. 2e). The big unchanged gene group demonstrated that functional recovery of most diabetes-associated genes is difficult. We therefore hypothesized that newly emerging genes belonging to the newly changed group may play a compensatory roles in amelioration of diabetes following RYGB.

A rebalance strategy following RYGB: the newly changed group contributes to metabolic rebalance

Diabetes (particularly type 2 diabetes (T2D)) remission by RYGB has generally been considered to result from functional changes in a complex and systemic interplay of multiple molecules and pathways. We further examined the functional relationships of genes in the four transition groups from molecular interactions and metabolic pathways to more comprehensively understand the underlying mechanisms of diabetes remission by RYGB. After mapping all DEGs into the knowledge-based molecular interactions of rats (Supplementary Fig. S1) and counting individual nodes (i.e., DEGs) neighboring those of the unchanged group, we found that neighboring nodes of the unchanged genes belonged mainly to the newly changed and unchanged groups (Fig. 3a and Supplementary Table S2).

Fig. 3: Rebalance model to explain compensatory strategy of diabetes remission by RYGB.

a Histogram showing distributions of neighboring nodes of unchanged genes in four transition groups. b Overlapping functional analysis in a total of 37 metabolism-associated processes significantly enriched by each of four transition groups. c Enriched metabolism-associated pathways by four transition groups graphically shown in detail. C carbohydrate metabolism, E energy metabolism, L lipid metabolism, N nucleotide metabolism, A amino acid metabolism, OA metabolism of other amino acids, CF metabolism of cofactors and vitamins, X xenobiotic biodegradation and metabolism. d Schematic diagram illustrating rebalance strategy of diabetes remission by RYGB. Blue block: diabetes-related genes/functions not recovered by RYGB (unchanged group). Red block: newly changed genes, i.e., diabetes-irrelevant genes unusually expressed after RYGB. Not all diabetes-related genes could be restored following RYGB, but the physical system rearranges other normal metabolic pathways, newly changed � �dysfunctions”, which unexpectedly improve metabolic parameters and rebalance the abnormalities caused by unchanged diabetes-associated genes. In a and c, positive transition group is indicated by green color, unchanged group by blue, newly changed group by red, and opposite group by yellow.

To clarify the functional roles of the newly changed group for amelioration of metabolic parameters by RYGB, our analysis focused on a total of 37 metabolism-associated pathways significantly enriched by at least one of the four transition groups. Although each of the four groups was enriched in its preferred and specific metabolism pathways, there were few overlapping enriched pathways of metabolism among them. Gene members in the four groups could participate or be involved in many other metabolic processes that were enriched by other transition groups, and some of the preferred and specific pathways corresponding to one group had some functional associations (Fig. 3b; Supplementary Fig. S2 and Table S3).

Interestingly, we found after performing functional analyses with members of each transition group (Supplementary Table S3) that the functions enriched by the positive group were involved mainly with carbohydrate metabolism and the tricarboxylic acid (TCA) cycle (Fig. 3c, pathways starting with C), and amino acid metabolism (pathways starting with A or OA). The unchanged group was enriched in nucleic acid and drug metabolism (pathways starting with N and X). The opposite group was more widely spread. Particularly, metabolic pathways involved with carbohydrate metabolism (e.g., glycolysis, gluconeogenesis), amino sugar and nucleotide sugar metabolism, and lipid metabolism (pathways starting with L) were enriched in the newly changed group. These findings support our hypothesis that the newly changed group plays a compensatory role to unchanged dysfunctions in rebalancing glucose homeostasis.

We proposed a rebalance model to explain this strategy of diabetes remission by RYGB (Fig. 3d). Generally, the body's physical system is in equilibrium and balance with organized, synergic, and regulatory metabolic pathways under normal or healthy conditions, whereas if diabetes-associated genes or functions shift and become overwhelmed, the physical system enters an unbalanced state and symptoms of diabetes (e.g., hyperglycemia and hyperlipidemia) occur. After treatment with RYGB, amelioration of diabetes may result from not only functional recovery of diabetes-associated genes but also newly emerging or RYGB-induced “dysfunctions” as a compensatory role for rebalance, as demonstrated in healthy rats and mice after RYGB by the cross-tissue study16. In our schematic diagram (Fig. 3d), the blue block is shifted to the right, indicating diabetes-associated genes, which remain unchanged following RYGB. The red block, which previously did not differ between normal and diabet es, is shifted to the left, indicating a newly changed “dysfunction” following RYGB. Different from a traditional recovering of the disease-associated genes, the new strategy of making “normal” functions “abnormal” unexpectedly rebalances and contributes to diabetes remission.

An unusual increase in newly synthesized CHOL as an example of RYGB-induced “dysfunctions”

Functional analysis indicated that the newly changed genes associated with metabolic pathways were enriched in the carbohydrate and lipid metabolism categories, which may play a compensatory role in rebalancing diabetes parameters. To further map the major RYGB-induced changes in carbohydrate and lipid functional pathways, we obtained signatures of 74 genes that differed (false discovery rate (FDR) <0.05) between GK-S-30 and GK-Sham-PF. Our enrichment analysis revealed nine significant pathways (Supplementary Table S4; FDR <0.05) based on gene sets in MSigDB, of which four were observed in CHOL metabolism. After RYGB, lack of bile may lead to alterations of CHOL uptake and synthesis in the small intestine. Bile has a direct inhibitory effect on cholesterogenesis in the intestinal mucosa20. CHOL synthesis may be enhanced in the bile-deprived RL, and this process requires costly glucose consumption to produce very large amounts of basic materials (e.g., acetyl-CoA, ATP, NADPH) 21, which may have a direct beneficial effect on glycemic control. The intestine, like the liver, is an important site for high-density lipoprotein (HDL) synthesis and secretion and provides ~30% of plasma HDL6. We therefore focused on CHOL metabolism enriched by the newly changed group to elucidate its association with metabolic improvement by RYGB.

To further test our hypothesis, we mapped the rate-limiting enzymes involved mainly in glucose, fatty acid, and CHOL metabolism to show their mRNA expression changes following RYGB (Fig. 4a). Red-colored gene symbols, representing the newly changed group, indicate that such gene expression does not differ significantly between GK-Sham-PF and Wistar rats, but is notably activated or inhibited in the RL following RYGB. In glycolysis, most of the key enzymes were upregulated after RYGB. In particular, the rate-limiting enzyme hexokinase (Hk) belonged to the newly changed group. Gluconeogenesis was downregulated (e.g., fructose-1,6-bisphosphatase (Fbp), phosphoenolpyruvate carboxykinase (Pck), and glucose-6-phosphatase (G6pc)). After glycolysis, pyruvate comes into mitochondria and is converted to acetyl-CoA, which enters the TCA cycle. In the TCA cycle, mRNA level is significantly elevated for citrate synthase (CS), but not for isocitrate dehydrogenase (Idh3g) or alpha-ketoglutarate dehydrogenase (Ogdh), indicating that citrate may accumulate and shuttle out of mitochondria to the cytoplasm for either fatty acid or CHOL biosynthesis. Acetyl-CoA carboxylases (Acaca), key enzymes catalyzing the irreversible carboxylation of acetyl-CoA to produce malonyl-CoA for fatty acid synthesis, were not transcriptionally upregulated. These findings suggest that conversion of glucose to CHOL synthesis may become “dysfunctionally” active in RL after RYGB, and contribute to metabolic rebalance (Fig. 4b).

Fig. 4: Increase in newly synthesized HDL-CHOL contributes to glycometabolism rebalance.

a Changes in expression of key enzymes of glucose and CHOL metabolism in three groups (Wistar, GK-PF, GK-S-30; each n = 5) in terms of relative mRNA level. Gene symbols are colored green for positive transition group; blue: unchanged; red: newly changed; yellow: opposite; black: non-DEGs. b Schematic diagram showing that increased de novo CHOL synthesis in RL has direct beneficial effect on glycemic control after RYGB. c Plasma total CHOL, d plasma LDL-CHOL, e plasma HDL-CHOL, and f plasma TG were measured in overnight-fasted blood samples from GK and Wistar groups. Data are expressed as mean ± SE. Statistical terms (n ≥ 6) and symbol colors as in Fig. 1.

We found that the relative mRNA levels of most enzymes related to CHOL synthesis were significantly increased in comparison with those in both Wistar and GK-Sham-PF rats (new changes) (Fig. 4a). For HMG-CoA reductase (Hmgcr), the rate-limiting enzyme in CHOL synthesis, relative mRNA level in RL was >2-fold higher in GK-S-30 rats than in either Wistar or GK-Sham-PF rats. Intestinal low-density lipoprotein (LDL) receptor (Ldlr) mRNA levels were significantly higher following RYGB, indicating increased internalization of LDL-CHOL from the blood into intestinal cells (Fig. 4a). ABCG5 and ABCG8 at the mRNA level were significantly reduced following RYGB (Fig. 4a), consistent with the concept that RL promotes extracellular CHOL transport across the brush border membrane to prevent excessive biliary CHOL loss22 (Fig. 4b).

Precursors (e.g., ATP, acetyl-CoA) of CHOL synthesis may arise from beta-oxidation of fatty acids; however, this mechanism is not supported by mRNA expression of carnitine palmitoyl transferase (Cpt) and acyl-coenzyme A dehydrogenase (Acad) (Fig. 4a). These raw materials are usually derived from glucose (Fig. 4b). Dietary glucose absorbed from their luminal side is strongly reduced in bile-deprived RL9; blood glucose taken up from the basolateral side must therefore be increased to meet the requirement for unusually high CHOL biosynthesis in RL.

To validate our hypothesis regarding unusually increased CHOL synthesis (mainly HDL-CHOL) in RL following RYGB, we measured total CHOL, LDL-CHOL, HDL-CHOL, and triglyceride (TG) levels in blood samples from the seven groups after overnight fasting. Total CHOL and LDL-CHOL were much lower in GK-S-30 than in GK-Sham-PF and GK (Fig. 4c, d). HDL-CHOL was much higher in GK-S-30 and GK-S-12 than in GK-Sham-PF and GK (Fig. 4e). LDL-CHOL profiles of GK-S-30R (RL excised) rats were significantly higher (P < 0.001) than those of GK-S-30 rats, and were not notably lower than those of GK rats. These findings indicate that RL contributes to HDL-CHOL, which decreases total CHOL and LDL-CHOL levels. In contrast, fatty acid biosynthesis is not increased in RL; TG levels in GK-S-30 and GK-S-30R were similar to each other and significantly lower than those in GK (Fig. 4f). The extremely low TG levels were due to reduced absorption without bile in rats with long RL (GK-S-30) or without RL (GK-S-30R). The greater amelioration of glucose and CHOL profiles in GK-S-30 than in GK-S-30R suggested that metabolic improvement was not associated with reduced TG absorption in RL, but rather with unusually high HDL-CHOL biosynthesis in RL.

Confirming amelioration of diabetes in STZ rats with long-RL RYGB

We examined the possible beneficial effects of long-RL RYGB on metabolic parameters in another nonobese rat model (streptozotocin (STZ)-treated diabetic rats) to rule out limitations in data interpretation associated with use of a single-animal model. The same RYGB procedure performed in GK-S-30 rats was performed in STZ-treated diabetic rats (STZ-S-30), with nonoperated age-matched Wistar and STZ rats as controls. Three months after RYGB, body weight was significantly higher in STZ-S-30 rats than in STZ controls (Fig. 5a), and glycemic control in the STZ-S-30 rats was greatly improved (Fig. 5bâ€"d). Lipid profiles showed that total CHOL, LDL-CHOL, and TG were significantly reduced in the STZ-S-30 group compared with the STZ group, whereas HDL-CHOL levels did not differ significantly among the Wistar, STZ-S-30, and STZ groups (Fig. 5eâ€"h). To address these findings differently from those in GK rats with surgery, we used oral loading of [14C] glucose to confirm the increase in newly synthesized HDL-CHOL labeled by an isotope in STZ-S-30 rats following RYGB (Supplementary Fig. S3). Our results were consistent with the beneficial effects of long-RL RYGB in STZ rats, and confirmed that long RL is associated with amelioration of metabolic parameters.

Fig. 5: Amelioration of diabetes and changes in [14C] glucose metabolism in STZ-treated diabetic rats after long-RL RYGB.

aâ€"j Amelioration of diabetes in 30-cm RL RYGB STZ rats (STZ-S-30) was compared with both nonoperated STZ rats (STZ) and normal Wistar rats by measurement of physiological parameters 3 months postoperation. a Body weight. b Blood glucose. c IPGTT (1 g/kg glucose was injected i.p.). d Insulin tolerance test (ITT) (1 U/kg insulin was injected i.p.). e Plasma total CHOL. f Plasma LDL-CHOL. g Plasma HDL-CHOL. h Plasma TG. Rats were in fasted-overnight and fed states. i [14C] glucose was administered orally according to body weight. Three hours later, [14C] CHOL was precipitated and biodistribution was measured in the three groups. Newly synthesized CHOL from [14C] glucose was significantly increased in intestine 2 (RL) but remained low in liver. j [14C] glucose incorporation rate into CHOL. Data are expressed as µmol [14C] glucose incorporated into digitonin-precipitable sterols (DPS) per gram of tissue per hour. Data are expressed as mean ± SE. Statistic ally significant differences of means between groups were determined by ANOVA with Tukey’s multiple comparison test (in aâ€"h, n ≥ 4; in i and j, n = 3 for each group), where *P < 0.05, **P < 0.01 for multiple comparisons among the three groups by ANOVA.

Unusually high CHOL biosynthesis in RL by diversion of ingested [14C] glucose after RYGB

To further evaluate high CHOL biosynthesis as a new change in RL following RYGB, we analyzed isotope levels in CHOL following oral loading of [14C] glucose in long-RL RYGB (STZ-S-30), STZ, and Wistar groups. Three hours after oral loading of [14C] glucose, glucose biodistribution per gram (g) tissue was similar in the three groups, i.e., levels were highest in liver, brain, pancreas, and colon, and lowest in adipose tissue and muscle (Supplementary Fig. S4). Glucose uptake was 0.34 ± 0.12% dose per g tissue in RL of RYGB-treated rats vs. 0.17 ± 0.03% in corresponding intestine of STZ-treated rats, which was nearly twice as high but did not achieve significance by ANOVA test. We precipitated and quantified [14C] CHOL, which was newly synthesized from orally administered [14C] glucose (Fig. 5i). [14C] CHOL biodistribution was highest in liver for all three groups, and remained significantly lower in liver after RYGB in comparison with Wistar. Labeled CHOL d istribution in intestine 2 did not differ significantly between STZ and Wistar. However, this labeled signal in RL (intestine 2) of STZ-S-30 was almost fourfold higher than that in STZ and twofold higher than that in Wistar (P = 0.0057 and 0.0419, respectively; Fig. 5i), and was considered an abnormal change. We calculated the ratio of CHOL biosynthesis to glucose disposal (Fig. 5j). Incorporation of [14C] glucose into CHOL was not recovered in liver after RYGB. However, in comparison with STZ, STZ-S-30 showed a significantly upregulated ratio of C2 flux into digitonin-precipitable sterols (DPS) in RL. These findings indicate an enhanced rate of glucose incorporation into CHOL synthesis per g tissue per hour in RL. The intestine contributes 30% of HDL-CHOL; we therefore measured newly synthesized plasma HDL-CHOL in RL of STZ-S-30 and STZ using a [14C] tracer, and observed a significant increase in newly synthesized plasma HDL-CHOL in STZ-S-30 (Supplementary Fig. S3).

Tweet