OBJECTIVE

Roux-en-Y gastric bypass (RYGB) and sleeve gastrectomy (SG) are effective procedures to treat and manage type 2 diabetes (T2D). However, the underlying metabolic adaptations that mediate improvements in glucose homeostasis remain largely elusive. The purpose of this study was to identify metabolic signatures associated with biochemical resolution of T2D after medical therapy (MT) or bariatric surgery.

RESEARCH DESIGN AND METHODS

Plasma samples from 90 patients (age 49.9 ± 7.6 years; 57.7% female) randomly assigned to MT (n = 30), RYGB (n = 30), or SG (n = 30) were retrospectively subjected to untargeted metabolomic analysis using ultra performance liquid chromatography with tandem mass spectrometry at baseline and 24 months of treatment. Phenotypic importance was determined by supervised machine learning. Associations between change in glucose homeostasis and circulating metabolites were assessed using a linear mixed effects model.

RESULTS

The circulating metabolome was dramatically remodeled after SG and RYGB, with largely overlapping signatures after MT. Compared with MT, SG and RYGB profoundly enhanced the concentration of metabolites associated with lipid and amino acid signaling, while limiting xenobiotic metabolites, a function of decreased medication use. Random forest analysis revealed 2-hydroxydecanoate as having selective importance to RYGB and as the most distinguishing feature between MT, SG, and RYGB. To this end, change in 2-hydroxydecanoate correlated with reductions in fasting glucose after RYGB but not SG or MT.

CONCLUSIONS

We identified a novel metabolomic fingerprint characterizing the longer-term adaptations to MT, RYGB, and SG. Notably, the metabolomic profiles of RYGB and SG procedures were distinct, indicating equivalent weight loss may be achieved by divergent effects on metabolism.

Obesity and type 2 diabetes (T2D) are two of the most prevalent diseases contributing to diminished quality of life globally. In the U.S. alone, it is estimated that more than one-third (∼42%) of adults have obesity (1), and 38.4 million (∼12%) have T2D (2). Metabolic surgery is now widely accepted as a gold-standard treatment for moderate to severe obesity and T2D (3). However, there remains immense dispute as to whether the metabolic and glycemic benefits are primarily due to surgery or secondary effects of weight loss. For example, fixed weight loss by means of surgery and by means of lifestyle intervention similarly improved parameters of whole-body metabolic function, such as postprandial glucose kinetics, insulin action, and free fatty acid (FFA) turnover, suggestive of weight-dependent effects (4). In contrast, larger clinical trials with prospective randomization have shown that the metabolic benefits induced by bariatric surgery are synergistic to and potentially independent of weight loss (5). There is also evidence of procedure-dependent changes in nutrient absorption and metabolic parameters, observed most robustly after intestinal bypass operations, such as Roux-en-Y gastric bypass (RYGB) (4,6,7). Nevertheless, if the mechanisms of surgery are divergent from lifestyle and/or medical therapy (MT), there is enormous opportunity to use combinatorial strategies to enhance long-term patient outcomes.

Metabolites are small bioactive molecules that are substrates, intermediates, or products of complex biologic processes. Comprehensive profiling of metabolites, termed metabolomics, enables the investigation of complex interactive metabolic disorders, such as obesity and T2D. Furthermore, determination of system-wide metabolic function allows for the identification of biomarkers and algorithms to predict the efficacy of surgical interventions in weight loss and diabetes outcomes. Several previous studies have identified potential bioactive compounds that are associated with improvements in insulin sensitivity and bile acid turnover after bariatric surgery (8–11). However, most of the studies to date are cohort based with short duration of follow-up.

The purpose of this study was to determine system-wide changes in metabolism after MT or bariatric surgery. We conducted an untargeted metabolomic analysis of blood samples from patients with obesity and T2D who had been randomly assigned to receive either MT or one of the two most widely used bariatric surgery procedures, sleeve gastrectomy (SG) or RYGB (12). We hypothesized that RYGB and SG would both dramatically alter the metabolome compared with advanced MT. Furthermore, we anticipated that the phenotypic signatures of the metabolome would differ between RYGB and SG, indicating operation-specific and weight-independent effects on health outcomes.

Study Population

The Surgical Therapy and Medications Potentially Eradicate Diabetes Efficiently (STAMPEDE) trial was a parallel-arm randomized controlled trial comparing the effects of advanced MT, RYGB, and SG on diabetes remission in patients with obesity (12–14). In brief, patients with diagnosed T2D were randomly assigned to receive either MT, SG, or RYGB surgery and were observed for 2 years postintervention. The study inclusion criteria allowed patients between ages 20 and 60 years, with an HbA1c >7.0% and BMI 27–45 kg/m2. This study was approved by the Cleveland Clinic Institutional Review Board. A subset of 30 patients per group were randomly selected based on sample availability for the metabolomic analyses.

Treatment Allocation

MT

Patients underwent nutritional and exercise counseling along with pharmacotherapy according to the American Diabetes Association standards of medical care in diabetes guidelines (15). Briefly, patients had clinical visits every 3 months for the first year, where medications were titrated with a goal of achieving an HbA1c <6.0%. In addition, all patients were treated with lipid-lowering and antihypertensive medications with target systolic blood pressure of <130 mmHg, diastolic blood pressure of <80 mmHg, and LDL of <100 mg/dL.

Metabolic Surgery

SG and RYGB surgeries were performed laparoscopically by an experienced surgeon as described previously (16). SG involved a reduction in gastric volume by 75–80% beginning 3 cm from the pylorus toward the esophagogastric angle. RYGB involved creation of a 15- to 20-mL gastric pouch, 150-cm Roux limb, and 50-cm biliopancreatic limb.

Sample Collection

Fasting whole-blood samples were collected at baseline and 2 years after intervention. HbA1c was measured in whole blood using turbidimetric inhibition immunoassay (Cleveland Clinic Laboratories, Cleveland, OH). The remaining blood was centrifuged at 4°C for 10 min and stored frozen at −70°C for subsequent analysis.

Metabolomics

Untargeted metabolomic profiling was conducted on samples at baseline and 2 years after intervention by ultra performance liquid chromatography (UPLC) with tandem mass spectrometry (MS/MS). Samples were prepared using the automated MicroLab STAR. After extraction with methanol to remove the protein fraction, metabolites were analyzed by reverse-phase UPLC and MS/MS in both positive and negative ion modes and hydrophilic interaction chromatography/UPLC-MS/MS in the negative ion mode. Raw data were extracted, peak identified, and checked for data quality, and compounds were identified by comparison with library entries of purified standards based on retention time/index, mass-to-charge ratio, and chromatographic data (including MS/MS spectral data). Peaks were quantified using area under the curve. For studies spanning multiple days, a data normalization step was performed to correct variation resulting from instrument interday tuning differences. For studies that did not require >1 day of analysis, no normalization was necessary other than for purposes of data visualization.

Statistical Analyses

All the analyses were performed using R studio (version 1.4.1717). All continuous variables with a normal distribution are reported as means and SD. Variables with a nonnormal distribution are reported as medians and interquartile ranges. Categorical variables are summarized with the use of frequencies. We used an ANOVA model adjusted for BMI to assess change in continuous laboratory measurements and compare change in metabolite abundance among the three study groups. Main effects of treatment or time were adjusted for the false discovery rate. A q value of <0.1 was established a priori as the accepted level of false positivity. All metabolomic analyses were performed using metaboAnalystR 3.0 and POMA packages in R (17,18). The metabolomic data were normalized using log transformation and autoscaling, and missing metabolite data were imputed using probabilistic principal component analysis. The sparse partial least squares–discriminant analysis (sPLS-DA) algorithm was used to effectively reduce the number of variables (metabolites) in high-dimensional metabolomic data to produce robust and easy-to-interpret models. Random forest, a supervised machine learning algorithm suitable for high-dimensional data analysis, was performed using the randomForest package in R. Secondary visualization of data was performed using GraphPad Prism 10.

Baseline characteristics are listed in Table 1. In brief, the age of all participants was approximately 50 ± 8 years, with 57% being women. No baseline differences were observed for age, sex, BMI, duration of T2D, fasting plasma glucose, or HbA1c, with a mean duration of T2D of 8.1 ± 5.4 years and HbA1c average of 9.2 ± 1.5%. At baseline, patients required three (±1) medications for the management of T2D. BMI decreased significantly from baseline to 2 years after SG (−24.6%) and RYGB (−24.6%) compared with MT (−1.6%). Similarly, HbA1c decreased significantly from baseline to 2 years after SG (−29.8%) and RYGB (−30.4%) compared with MT (−15.4%). No significant differences in BMI, HbA1c, or markers of glycemic control or cardiometabolic health were observed between SG and RYGB after 2 years of treatment. However, use of medications to treat T2D was lower after RYGB (26.7%) relative to both SG (40.0%) and MT (100%) (Supplementary Fig. 1).

Table 1

Patient demographics and markers of metabolic function at baseline and 2 years after treatment allocation

CharacteristicMT
(n = 30)
SG
(n = 30)
RYGB
(n = 30)
P
Baseline24 Months
Baseline24 MonthsBaseline24 MonthsBaseline24 MonthsMT vs. SGMT vs. RYGBSG vs. RYGBMT vs. SGMT vs. RYGBSG vs. RYGB
Age, years 50.8 (±7.0) — 48.6 (±9.0) — 49.5 (±6.9) — 0.512 0.781 0.901 — — — 
Sex, % female 56.7 — 66.7 — 50 — 0.717 0.862 0.4 — — — 
Duration of T2D, years 9.5 (±5.7) — 8.1 (±4.6) — 6.8 (±5.5) — 0.578 0.139 0.63 — — — 
BMI, kg/m2 36.1 (±3.0) 35.5 (±3.3) 35.8 (±4.2) 27 (±3.2) 36.2 (±3.3) 27.3 (±3.3) 0.977 0.958 <0.0001 <0.0001 0.979 
HbA1c, % 9.1 (±1.5) 7.7 (±2.0) 9.4 (±1.7) 6.6 (±0.9) 9.2 (±1.3) 6.4 (±1.0) 0.764 0.98 0.934 0.011 0.002 0.925 
Glucose, mg/dL 176.7 (±62.3) 140.8 (±66.1) 182.1 (±72.9) 101.7 (±27.7) 200.9 (±69.7) 99.6 (±32.7) 0.977 0.322 0.547 0.042 0.029 0.999 
Insulin, μU/mL 25.4 (±24.6) 34.05 (±35.9) 14.8 (±11.1) 8.3 (±10.1) 25.4 (±19.1) 16.5 (±20.3) 0.162 >0.9999 0.172 <0.0001 0.01 0.378 
HOMA-IR 10.8 (±11.8) 10.3 (±9.8) 6.1 (±4.2) 2.2 (±3.3) 12.7 (±10.2) 4.4 (±5.8) 0.082 0.75 0.007 0.001 0.03 0.705 
SBP, mmHg 136.5 (±18.8) 133.8 (±19.8) 138.4 (±19.1) 132.5 (±19.7) 137.7 (±17.1) 132.5 (±24.5) 0.979 0.986 >0.9999 0.993 0.993 >0.9999 
DBP, mmHg 83.6 (±12.1) 77.8 (±10.4) 82.8 (±11.9) 74.4 (±10.1) 84.5 (±8.2) 77.2 (±9.7) 0.986 0.967 0.855 0.504 0.994 0.664 
Cholesterol, mg/dL 182.6 (±46.7) 172.3 (±34.3) 184.6 (±45.7) 183.4 (±40.3) 187.6 (±46.5) 178.3 (±28.6) 0.996 0.955 0.992 0.834 0.996 0.922 
Triglycerides, mg/dL 186.5 (±46.7) 119.5 (±63.1) 200.9 (±158.8) 107.2 (±52.9) 272.1 (±268.6) 115.3 (±56.6) 0.97 0.071 0.167 0.978 0.996 0.998 
CharacteristicMT
(n = 30)
SG
(n = 30)
RYGB
(n = 30)
P
Baseline24 Months
Baseline24 MonthsBaseline24 MonthsBaseline24 MonthsMT vs. SGMT vs. RYGBSG vs. RYGBMT vs. SGMT vs. RYGBSG vs. RYGB
Age, years 50.8 (±7.0) — 48.6 (±9.0) — 49.5 (±6.9) — 0.512 0.781 0.901 — — — 
Sex, % female 56.7 — 66.7 — 50 — 0.717 0.862 0.4 — — — 
Duration of T2D, years 9.5 (±5.7) — 8.1 (±4.6) — 6.8 (±5.5) — 0.578 0.139 0.63 — — — 
BMI, kg/m2 36.1 (±3.0) 35.5 (±3.3) 35.8 (±4.2) 27 (±3.2) 36.2 (±3.3) 27.3 (±3.3) 0.977 0.958 <0.0001 <0.0001 0.979 
HbA1c, % 9.1 (±1.5) 7.7 (±2.0) 9.4 (±1.7) 6.6 (±0.9) 9.2 (±1.3) 6.4 (±1.0) 0.764 0.98 0.934 0.011 0.002 0.925 
Glucose, mg/dL 176.7 (±62.3) 140.8 (±66.1) 182.1 (±72.9) 101.7 (±27.7) 200.9 (±69.7) 99.6 (±32.7) 0.977 0.322 0.547 0.042 0.029 0.999 
Insulin, μU/mL 25.4 (±24.6) 34.05 (±35.9) 14.8 (±11.1) 8.3 (±10.1) 25.4 (±19.1) 16.5 (±20.3) 0.162 >0.9999 0.172 <0.0001 0.01 0.378 
HOMA-IR 10.8 (±11.8) 10.3 (±9.8) 6.1 (±4.2) 2.2 (±3.3) 12.7 (±10.2) 4.4 (±5.8) 0.082 0.75 0.007 0.001 0.03 0.705 
SBP, mmHg 136.5 (±18.8) 133.8 (±19.8) 138.4 (±19.1) 132.5 (±19.7) 137.7 (±17.1) 132.5 (±24.5) 0.979 0.986 >0.9999 0.993 0.993 >0.9999 
DBP, mmHg 83.6 (±12.1) 77.8 (±10.4) 82.8 (±11.9) 74.4 (±10.1) 84.5 (±8.2) 77.2 (±9.7) 0.986 0.967 0.855 0.504 0.994 0.664 
Cholesterol, mg/dL 182.6 (±46.7) 172.3 (±34.3) 184.6 (±45.7) 183.4 (±40.3) 187.6 (±46.5) 178.3 (±28.6) 0.996 0.955 0.992 0.834 0.996 0.922 
Triglycerides, mg/dL 186.5 (±46.7) 119.5 (±63.1) 200.9 (±158.8) 107.2 (±52.9) 272.1 (±268.6) 115.3 (±56.6) 0.97 0.071 0.167 0.978 0.996 0.998 

Bold font indicates significance.

DBP, diastolic blood pressure; HOMA-IR, HOMA for insulin resistance; SBP, systolic blood pressure.

A total of 1,037 known metabolites were identified in the study participants’ serum samples at baseline and 2 years after intervention. Principal component analysis revealed a discrete shift of the metabolome after SG and RYGB compared with baseline, whereas there was minimal separation after MT (Fig. 1A). A summary of the number of biochemicals that achieved statistical significance (P ≤ 0.05), as well as those approaching significance (0.05 < P < 0.10), is visualized via volcano plot (Fig. 1B and C). Briefly, there were 115 metabolites that were significantly altered in the MT group compared with baseline, of which 52 were increased and 63 were decreased. In the SG group, 159 metabolites were significantly altered compared with baseline, of which 36 were increased and 123 were decreased. After 2 years of RYGB, 186 were significantly altered, of which 55 were increased and 131 were decreased. The metabolites that were significantly changed after post hoc analysis between MT, SG, and RYGB were 1-carboxyethylphenylalanine and 1-carboxyethyltyrosine. The metabolites that were significantly changed in the SG and RYGB groups compared with MT were 1-carboxyethylvaline, 1-carboxyethylleucine, 1-carboxyethylphenylalanine, 1-carboxyethylisoleucine, 1-carboxyethyltyrosine, 1-oleoyl-GPC (18:1), 1-(1-enyl-palmitoyl)-2-oleoyl-GPC (P–16:0/18:1), 12,13-dihydroxy-9Z-octadecenoic acid (12,13-DiHOME), 1-lignoceroyl-GPC (24:0), 1-(1-enyl-palmitoyl)-2-palmitoleoyl-GPC (P–16:0/16:1), 1-stearoyl-2-oleoyl-GPE (18:0/18:1), 1,5-anhydroglucitol (1,5-AG), and (S)-3-hydroxybutyrylcarnitine. Taken together, the pathway analyses of SG and RYGB in relation to MT yielded a divergent set of metabolic pathways. The most enriched pathways in SG compared with MT were myriad and pertained to long-chain fatty acid metabolism and bile acid synthesis, in addition to amino acid metabolism. In contrast, RYGB elicited enrichments in glucose utilization pathways, along with pathways associated with amino acid metabolism. Several metabolites were differentially enriched from baseline to 2 years after all three interventions, including glucose, 1,5-AG, mannose, and isoleucine, among others (Supplementary Fig. 2).

Figure 1

Differential plasma metabolomic signatures after MT, SG, and RYGB surgery. A: Principal component analysis (PCA) at baseline and 24 months of treatment. B: Volcano plot visualization of differentially expressed metabolites (false discovery rate q < 0.1) from baseline to 24 months of treatment within groups. C: Visualization of differentially expressed metabolites by treatment allocation. D: Differential pathway enrichment analysis from baseline to 24 months of treatment. E: Unsupervised machine learning prediction of pathway significance to metabolome phenotype.

Figure 1

Differential plasma metabolomic signatures after MT, SG, and RYGB surgery. A: Principal component analysis (PCA) at baseline and 24 months of treatment. B: Volcano plot visualization of differentially expressed metabolites (false discovery rate q < 0.1) from baseline to 24 months of treatment within groups. C: Visualization of differentially expressed metabolites by treatment allocation. D: Differential pathway enrichment analysis from baseline to 24 months of treatment. E: Unsupervised machine learning prediction of pathway significance to metabolome phenotype.

Close modal

To estimate the patterns of change in metabolomic features between the study groups, hierarchic clustering was performed based on Euclidean distance (Supplementary Fig. 1). The group of metabolites correctly identified the study group (i.e., both the bariatric surgery groups were clustered into one, which hierarchically was different from the MT group). Several metabolites were noted to have significantly changed across the study groups. Using Ward’s method of clustering, changes in relative abundances of the metabolites were obtained and visualized (Supplementary Fig. 3). The top metabolites that changed the most across the groups were 2-O-methyluridine, 2-deoxyuridine, 1-(1-enyl-palmitoyl)-2-linoleoyl-GPC (P–16:0/18:2), 1,2-dilinoleoyl-GPC (18:2/18:2), and 12,13-DiHOME. The above mentioned metabolites within each study group were enriched in a similar fashion (i.e., moderately increased in RYGB and mildly increased in SG), whereas they were decreased with MT alone. sPLS-DA validated metabolites identified by logistic regression with interactions to predict patients who underwent MT, SG, or RYGB. To determine the relative abundance of which metabolites made the largest contribution to the group differentiation between the 2-year metabolome of postintervention groups, a random forest analysis was performed. The overall predictive accuracy of the model was 74%. The top 30 metabolites that showed the greatest mean decrease in accuracy of the model are visualized in a dot plot (Fig. 2). When the metabolites were classified according to their biochemical class, various metabolites of lipids were noted to be more frequently found, followed by amino acids and bile acids. 1,5-AG displayed significant predictive power in MT, SG, and RYGB alike (Fig. 2A–E). Mannose was not predictive for MT but was highly predictive similarly for SG and RYGB (Fig. 2A–E). 2-hydroxydecanoate was revealed as having high selectivity for RYGB and, in a comparative model, was the most discriminate variable between treatments (Fig. 2C–E). To identify the significance of the metabolites in relation to glycemic control, we compared metabolites that significantly correlated with changes in HbA1c and plasma glucose after RYGB (Fig. 2F). Of the 10 metabolites with overlapping relationships, 2-hydroxydecanoate was the most enriched (Fig. 2G). Furthermore, there was no significant relationship between 2-hydroxydecanoate and MT (r = 0.140; P = 0.480) or SG (r = 0.123; P = 0.516); this relationship was solely observed with RYGB (r = 0.424; P = 0.012) (Fig. 2H–J).

Figure 2

Identification of metabolic regulators of glucose homeostasis after MT, SG, and RYGB surgery. AD: Random forest analysis of MT (A), SG (B), RYGB (C), and MT vs. SG vs. RYGB (D). E: Venn diagram illustrating overlapping vs. discriminate metabolites in each intervention. F: Venn diagram illustrating overlapping vs. discriminate metabolites correlated with change in glucose and HbA1c from baseline to 2 years. G: Expression intensity of 2-hydroxydecanoate before and 2 years after MT, SG, or RYGB. HJ: Regression modeling of relationship between 2-hydroxydecanoate expression (baseline expression intensity) and plasma glucose (change in mg/dL from baseline to 2 years) with MT (H), SG (I), and RYGB (J).

Figure 2

Identification of metabolic regulators of glucose homeostasis after MT, SG, and RYGB surgery. AD: Random forest analysis of MT (A), SG (B), RYGB (C), and MT vs. SG vs. RYGB (D). E: Venn diagram illustrating overlapping vs. discriminate metabolites in each intervention. F: Venn diagram illustrating overlapping vs. discriminate metabolites correlated with change in glucose and HbA1c from baseline to 2 years. G: Expression intensity of 2-hydroxydecanoate before and 2 years after MT, SG, or RYGB. HJ: Regression modeling of relationship between 2-hydroxydecanoate expression (baseline expression intensity) and plasma glucose (change in mg/dL from baseline to 2 years) with MT (H), SG (I), and RYGB (J).

Close modal

Table 2 summarizes pathway enrichment analyses illustrating differences between SG versus RYGB. Robust changes were observed in aromatic amino acid metabolism, particularly phenylalanine metabolism, which differed in RYGB relative to SG and had the highest overall impact on topology. Supplementary Table 1 shows the metabolites that were differentially changed in RYGB with respect to the SG group. The initial analysis revealed that the significantly different metabolites between RYGB and SG were 1-carboxyethylphenylalanine, 1-carboxyethyltyrosine, 12,13-DiHOME, and 1-lignoceroyl-GPC (24:0). To determine the optimal number of metabolomic principal components to effectively discriminate patients in the RYBG versus SG group, sPLS-DA was performed. Lowest error rate, calculated by random sampling and cross validation, was four components (error rate 10%). The set of metabolites with the highest impact were carotene diols, perfluoroctan, 2-hydroxysebacate, and phenyllactate in the first component; sphingomyelins, lignoceroylcarnitine, 17-α-hydroxypregnanolone glucuronide, and glycoursodeoxycholate in the second component; histidine, saccharin, octadecanedioate (C18-DC), glycoursodeoxycholate, and glycochenodeoxycholate in the third component; and 2-hydroxynervonate, bilirubin, galactonate, 3-hydroxybutyrate (BHBA), and (S)-3-hydroxybutyrylcarnitine in the fourth component. Pathway analysis revealed enriched metabolites mainly pertaining to amino acid metabolism (Supplementary Fig. 4). Phenylalanine, tyrosine, and tryptophan metabolism had the highest impact (P = 0.007), driven mainly by phenylalanine metabolism (P < 0.001), ubiquinone biosynthesis (P = 0.04), and taurine metabolism (P = 0.02). The pathways that were significantly changed, but with slightly lesser impact on the overall model, were glycine, serine, and threonine metabolism (P < 0.001) and alanine, aspartate, and glutamate metabolism (P = 0.05).

Table 2

Top 10 enriched pathways between RYGB and SG sorted by impact value

Total compoundsHitsPFDRImpact
RawInverse logHolm adjusted
Phenylalanine, tyrosine, and tryptophan biosynthesis 4.92E−04 3.31E+00 3.25E−02 2.92E−02 
Phenylalanine metabolism 10 8.85E−04 3.05E+00 5.76E−02 2.92E−02 0.62 
Tyrosine metabolism 42 3.49E−03 2.46E+00 2.24E−01 7.08E−02 0.25 
Glycerolipid metabolism 16 4.76E−03 2.32E+00 3.00E−01 7.08E−02 0.37 
Pentose phosphate pathway 22 5.36E−03 2.27E+00 3.32E−01 7.08E−02 0.05 
Arginine and proline metabolism 38 11 8.01E−03 2.10E+00 4.89E−01 8.27E−02 0.42 
Glycine, serine, and threonine metabolism 33 14 8.77E−03 2.06E+00 5.26E−01 8.27E−02 0.73 
Ubiquinone and other terpenoid-quinone biosynthesis 1.73E−02 1.76E+00 1.00E+00 1.43E−01 
Aminoacyl-tRNA biosynthesis 48 19 2.29E−02 1.64E+00 1.00E+00 1.51E−01 0.17 
Glycolysis/gluconeogenesis 26 2.27E−02 1.64E+00 1.00E+00 1.51E−01 0.1 
Total compoundsHitsPFDRImpact
RawInverse logHolm adjusted
Phenylalanine, tyrosine, and tryptophan biosynthesis 4.92E−04 3.31E+00 3.25E−02 2.92E−02 
Phenylalanine metabolism 10 8.85E−04 3.05E+00 5.76E−02 2.92E−02 0.62 
Tyrosine metabolism 42 3.49E−03 2.46E+00 2.24E−01 7.08E−02 0.25 
Glycerolipid metabolism 16 4.76E−03 2.32E+00 3.00E−01 7.08E−02 0.37 
Pentose phosphate pathway 22 5.36E−03 2.27E+00 3.32E−01 7.08E−02 0.05 
Arginine and proline metabolism 38 11 8.01E−03 2.10E+00 4.89E−01 8.27E−02 0.42 
Glycine, serine, and threonine metabolism 33 14 8.77E−03 2.06E+00 5.26E−01 8.27E−02 0.73 
Ubiquinone and other terpenoid-quinone biosynthesis 1.73E−02 1.76E+00 1.00E+00 1.43E−01 
Aminoacyl-tRNA biosynthesis 48 19 2.29E−02 1.64E+00 1.00E+00 1.51E−01 0.17 
Glycolysis/gluconeogenesis 26 2.27E−02 1.64E+00 1.00E+00 1.51E−01 0.1 

FDR, false discovery rate.

Bariatric surgery has been established over time as a safe and effective treatment for obesity and related diseases, such as T2D (19). However, response to surgical interventions with intention to treat T2D is variable (20), which has historically been attributed to alterations in calorie intake, physical activity, and/or alterations in gut hormone responses (21). We carried out untargeted metabolomic analysis to identify pathways and metabolites associated with changes in glucose homeostasis using a randomized control sample after RYGB, SG, and MT. We observed that RYGB and SG profoundly altered the circulating metabolome and specifically identified 2-hydroxydecanoate as having selective importance related to improvements in glucose homeostasis after RYGB.

As part of our analysis, we identified changes in metabolites previously associated with improvements in glycemic control. For example, all three interventions increased 1,5-AG. Plasma 1,5-AG reflects short-term glucose status, unlike HbA1c, which is a marker of longer-term glycemic control (22). Nonetheless, 1,5-AG is a validated biomarker of diabetes status and is clinically useful for monitoring strict control of glycemic status. Increased 1,5-AG with lower glucose also indicates improved insulin sensitivity, because 1,5-AG competes with glucose for reabsorption in the kidney (23,24). In the context of this study, the greatest changes were observed with RYGB and SG (∼6.5-fold), reflective of the vastly greater improvements in fasting glucose and HbA1c. Higher changes in 1,5-AG and lower pyruvate and lactate also denote increased pyruvaldehyde degradation in the RYGB compared with SG group, with increased glucose utilization via glycolysis.

High levels of plasma branched-chain amino acids (BCAAs) are correlated with insulin resistance (25). Leucine, isoleucine, and valine were all significantly lower after SG and RYGB compared with their baseline values. MT resulted in significantly lower levels of isoleucine and trends toward lower leucine and valine. Alterations in plasma levels of BCAAs may reflect dietary intake or muscle protein catabolism (26). Accompanying these BCAA changes were decreased levels of the catabolic products of BCAAs, including the C3 and C5 acyl carnitines, propionylcarnitine, 2-methylbutyrylcarnitine (trend in RYGB), and isovalerylcarnitine (RYGB only). In human studies, increased C3 and C5 acyl carnitines in plasma and muscle are associated with insulin resistance (25). This was not observed in the MT group. Previous studies of bariatric surgery have identified decreases in plasma propionylcarnitine (C3), 2-methylbutyrylcarnitine (C5), and isovalerylcarnitine (C5), which were not observed after dietary interventions with similar weight loss (5,27). Similarly, patients with obesity and T2D exhibited reduced BCAA metabolites after significant weight loss (28). The decrease in these acyl carnitines in SG and RYGB groups may be part of the mechanism that leads to greater changes in markers of increased insulin sensitivity. In addition to BCAA metabolism, aromatic amino acid (phenylalanine, tyrosine, and tryptophan) metabolism has been implicated in adiposity and insulin resistance (29,30). Our study noted higher enrichment of downstream metabolites related to phenylalanine and tryptophan metabolism in the RYGB compared with SG group. These data shed light on the possible role of phenylalanine and tryptophan metabolism in insulin secretion and disposition index, as previously described (31). In comparison with SG, RYGB activated the glycine, serine, and threonine pathway, raising the possible role of 1-carbon metabolism in mediating its effects (32).

Changes in plasma FFAs can be a result of higher dietary intake, de novo fatty acid synthesis in the liver, phospholipid breakdown, lipolysis, or triacylglycerol hydrolysis. In the current study, there were decreased levels of circulating long-chain saturated (e.g., stearate [18:0]), long-chain monounsaturated (e.g., 10-nonadecenoate [19:1n9]), and polyunsaturated (e.g., docosahexaenoate [22:6n3]) fatty acids after SG intervention. Consistent with lower FFAs, markers of fatty acid metabolism via β-oxidation, such as acetylcarnitine and BHBA, were also decreased after SG intervention. Despite few changes in FFA levels, BHBA was also decreased in the MT group at 24 months compared with baseline. Also, we noted a marginal decrease in 2-hydroxysebacate in the SG compared with RYGB group, which is a marker of dicarboxylate metabolism (33). 12,13-DiHOME, a lipokine associated with increased skeletal muscle fatty acid uptake, browning of adipose tissue, and improved cardiac function, was noted to be significantly increased after SG (34,35). Taken together, these data suggest that SG had a greater impact on metabolite processes associated with lipid metabolism than RYGB or MT.

Increased 2-hydroxydecanoate after RYGB was the most selectively important phenotypic trait revealed by supervised machine learning. 2-hydroxydecanoate is a weakly acidic, anionic medium-chain fatty acid (MCFA). MCFA's are unique in that they are rapidly released into circulation after ingestion through absorption from small-intestine mucosal cells. MCFAs are modified and catabolized by duodenal pancreatic lipases, which aid in the hydrolysis and digestion of lipid-soluble esters and metabolites. In this regard, one may speculate that 2-hydroxydecanoate becomes selectively enriched after RYGB because of near-complete bypassing of the duodenum and, subsequently, duodenal lipases. This notion is supported in part by prior studies demonstrating dramatic augmentation of acylcarnitine metabolites after bariatric surgery (4). Nonetheless, further research is required to establish the role and regulation of 2-hydroxydecanoate in the context of RYGB.

Primary bile acids are synthesized in the liver by cytochrome P450–mediated oxidation of cholesterol and are stored in the gall bladder and released into the duodenum. Bile acids are important for the digestion of lipids and especially fat-soluble vitamins, including vitamins A, D, E, and K. Most bile acids released into the intestinal tract are recycled to the liver. Bile acids in the gut are subject to modification by gut microbiota, which make the secondary bile acids (36). In the current study, MT resulted in a decrease in several primary bile acids (including but not limited to glycocholate, glycochenodeoxycholate, and taurochenodeoxycholate), with little impact on secondary bile acids. Despite this, metabolites associated with vitamins A and E were higher in the plasma and may reflect an increase in intake because of dietary changes. Increased levels of glycolithocholate sulfate, trends toward increased taurolithocholate 3-sulfate 24 months after SG and RYGB compared with baseline, and increased glycohyocholate after RYGB intervention were observed. Changes in plasma bile acids after bariatric surgery may be attributed to two different phenomena: 1) bile acids have less time to mix with food, leaving them free for ileal uptake and entry into circulation, and 2) secondary bile acids are the products of bacterial catabolism of the primary bile acids (36). Because there were few changes in primary bile acids, it is likely that changes in secondary bile acids reflect microbiome alterations. Microbiome remodeling is a feature of bariatric surgery and is independent of weight loss (37). Interestingly, with MT, there were some alterations in benzoate, phenylalanine, and tyrosine metabolites (e.g., catechol sulfate and p-cresol sulfate), which are the products of bacterial cometabolism; however, we did not identify strong patterns of change in microbiome markers in the SG and RYGB groups. This may indicate a greater influence of dietary changes on the microbiome compared with bariatric surgery or normalization of the microbiome long term postbariatric surgery. Further analysis by correlating dietary intake data, as well as metagenomic analysis, would reveal the effect modification of the dietary changes on the microbiome after bariatric surgery.

Our approach was strengthened by leveraging patients from a randomized controlled trial with a longer follow-up period and high compliance. Furthermore, our study is one of the first to report the long-term adaptations of the metabolome in a procedure-dependent manner including RYGB and SG. Potential limitations of our study include the lack of an interim metabolomic data set and the differential use of medications that may affect whole-body metabolism. We did not observe associations between 2-hydroxydecanoate and changes in HbA1c, indicating that the relationship may be limited to fasting conditions. Future studies would benefit from short-term metabolomic analysis coupled with longer-term phenotype observations to aid in the prediction of remission and/or relapse of T2D.

In conclusion, SG and RYGB profoundly affected plasma metabolic profiles 2 years after surgery. Shared mechanisms included metabolites and pathways associated with glucose utilization, insulin sensitivity, and amino acid catabolism. SG produced more profound alterations in FFA metabolites, whereas RYGB most dramatically altered primary and secondary bile acid and aromatic amino acid metabolism. Furthermore, we identified 2-hydroxydecanoate as the most important distinguishing signature among all interventions, which was selectively increased after RYGB. Taken together, these data support the notion that SG and RYGB improve glucose homeostasis and promote remission of T2D by shared and discrete mechanisms of action. Further consideration of 2-hydroxydecanoate as a candidate biomarker in RYGB as it pertains to remission of T2D is warranted.

Clinical trial reg. no. NCT00432809, clinicaltrials.gov

This article contains supplementary material online at https://doi.org/10.2337/figshare.26870587.

S.R.K. is currently affiliated with Weill Cornell Medicine.

Acknowledgments. The authors thank prior and current members of the Integrated Physiology and Molecular Medicine Laboratory for their technical support and thoughtful insights that supported this work.

Funding. This work was supported by Ethicon Endo-Surgery, LifeScan, Inc., the Cleveland Clinic, and the National Institutes of Health (DK089547).

Duality of Interest. The STAMPEDE trial recieved support from Ethicon Endo-Surgery, a company which develops minimally invasive devices for metabolic surgery, and LifeScan, Inc., which develops blood glucose monitoring solutions for patients with diabetes. Metabolon, Inc., a company which provides bulk metabolomic and bioinformatic services, provided support with sample processing and data analysis. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. C.L.A., A.H., and W.S.D. performed the formal analysis and wrote the original draft of the manuscript. C.L.A., A.H., W.S.D., S.R.K., P.R.S., and J.P.K. performed investigations. A.H., P.R.S., and J.P.K. conceptualized the study. P.R.S. and J.P.K. provided resources and acquired funding. J.P.K. supervised the study. All authors reviewed and edited the manuscript. J.P.K. 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.

Prior Presentation. These data were presented in part at the American Diabetes Association 80th Scientific Sessions, 12–16 June 2020 (https://doi.org/10.2337/db20-1904-P).

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Elizabeth Selvin and Casey M. Rebholz.

1.
Hales
CM
,
Carroll
MD
,
Fryar
CD
,
Ogden
CL
.
Prevalence of obesity and severe obesity among adults: United States, 2017–2018
.
NCHS Data Brief
2020
;
360
:
1
8
2.
Centers for Disease Control and Prevention
.
Diabetes Report Card 2019.
Accessed 11 March 2024. Available from https://stacks.cdc.gov/view/cdc/103877
3.
American Diabetes Association Professional Practice Committee
.
8. Obesity and weight management for the prevention and treatment of type 2 diabetes: Standards of Care in Diabetes—2024
.
Diabetes Care
2024
;
47
:
S145
S157
4.
Yoshino
M
,
Kayser
BD
,
Yoshino
J
, et al
.
Effects of diet versus gastric bypass on metabolic function in diabetes
.
N Engl J Med
2020
;
383
:
721
732
5.
Laferrère
B
,
Reilly
D
,
Arias
S
, et al
.
Differential metabolic impact of gastric bypass surgery versus dietary intervention in obese diabetic subjects despite identical weight loss
.
Sci Transl Med
2011
;
3
:
80re2
6.
Hofsø
D
,
Fatima
F
,
Borgeraas
H
, et al
.
Gastric bypass versus sleeve gastrectomy in patients with type 2 diabetes (Oseberg): a single-centre, triple-blind, randomised controlled trial
.
Lancet Diabetes Endocrinol
2019
;
7
:
912
924
7.
Saeidi
N
,
Nestoridi
E
,
Kucharczyk
J
,
Uygun
MK
,
Yarmush
ML
,
Stylopoulos
N
.
Sleeve gastrectomy and Roux-en-Y gastric bypass exhibit differential effects on food preferences, nutrient absorption and energy expenditure in obese rats
.
Int J Obes (Lond)
2012
;
36
:
1396
1402
8.
Jüllig
M
,
Yip
S
,
Xu
A
, et al
.
Lower fetuin-A, retinol binding protein 4 and several metabolites after gastric bypass compared to sleeve gastrectomy in patients with type 2 diabetes
.
PLoS One
2014
;
9
:
e96489
9.
Guo
K
,
Yu
H
,
Lu
J
,
Bao
Y
,
Chen
H
,
Jia
W
.
Decreased serum betatrophin levels correlate with improved fasting plasma glucose and insulin secretion capacity after Roux-en-Y gastric bypass in obese Chinese patients with type 2 diabetes: a 1-year follow-up
.
Surg Obes Relat Dis
2016
;
12
:
1343
1348
10.
Vaz
M
,
Pereira
SS
,
Monteiro
MP
.
Metabolomic signatures after bariatric surgery—a systematic review
.
Rev Endocr Metab Disord
2022
;
23
:
503
519
11.
Staels
B
,
Fonseca
VA
.
Bile acids and metabolic regulation: mechanisms and clinical responses to bile acid sequestration
.
Diabetes Care
2009
;
32
(
Suppl. 2
):
S237
245
12.
Schauer
PR
,
Bhatt
DL
,
Kirwan
JP
, et al.;
STAMPEDE Investigators
.
Bariatric surgery versus intensive medical therapy for diabetes—3-year outcomes
.
N Engl J Med
2014
;
370
:
2002
2013
13.
Schauer
PR
,
Bhatt
DL
,
Kirwan
JP
, et al.;
STAMPEDE Investigators
.
Bariatric surgery versus intensive medical therapy for diabetes—5-year outcomes
.
N Engl J Med
2017
;
376
:
641
651
14.
Schauer
PR
,
Kashyap
SR
,
Wolski
K
, et al
.
Bariatric surgery versus intensive medical therapy in obese patients with diabetes
.
N Engl J Med
2012
;
366
:
1567
1576
15.
ElSayed
NA
,
Aleppo
G
,
Aroda
VR
, et al
.
9. Pharmacologic approaches to glycemic treatment: Standards of Care in Diabetes—2023
.
Diabetes Care
2023
;
46
(
Suppl. 1
):
S140
S157
16.
Kashyap
SR
,
Bhatt
DL
,
Schauer
PR
;
STAMPEDE Investigators
.
Bariatric surgery vs. advanced practice medical management in the treatment of type 2 diabetes mellitus: rationale and design of the Surgical Therapy And Medications Potentially Eradicate Diabetes Efficiently trial (STAMPEDE)
[
published correction appears in Diabetes Obes Metab 2010;12:833
].
Diabetes Obes Metab
2010
;
12
:
452
454
17.
Castellano-Escuder
P
,
González-Domínguez
R
,
Carmona-Pontaque
F
,
Andrés-Lacueva
C
,
Sánchez-Pla
A
.
POMAShiny: a user-friendly web-based workflow for metabolomics and proteomics data analysis
.
PLoS Comput Biol
2021
;
17
:
e1009148
18.
Pang
Z
,
Chong
J
,
Li
S
,
Xia
J
.
MetaboAnalystR 3.0: toward an optimized workflow for global metabolomics
.
Metabolites
2020
;
10
:
186
19.
Chang
S-H
,
Stoll
CRT
,
Song
J
,
Varela
JE
,
Eagon
CJ
,
Colditz
GA
.
The effectiveness and risks of bariatric surgery: an updated systematic review and meta-analysis, 2003-2012
.
JAMA Surg
2014
;
149
:
275
287
20.
Magro
DO
,
Geloneze
B
,
Delfini
R
,
Pareja
BC
,
Callejas
F
,
Pareja
JC
.
Long-term weight regain after gastric bypass: a 5-year prospective study
.
Obes Surg
2008
;
18
:
648
651
21.
Shah
M
,
Simha
V
,
Garg
A
.
Review: long-term impact of bariatric surgery on body weight, comorbidities, and nutritional status
.
J Clin Endocrinol Metab
2006
;
91
:
4223
4231
22.
Dungan
KM
,
Buse
JB
,
Largay
J
, et al
.
1,5-anhydroglucitol and postprandial hyperglycemia as measured by continuous glucose monitoring system in moderately controlled patients with diabetes
.
Diabetes Care
2006
;
29
:
1214
1219
23.
Hassanein
M
,
Shafi
T
.
Assessment of glycemia in chronic kidney disease
.
BMC Med
2022
;
20
:
117
24.
Selvin
E
,
Wang
D
,
McEvoy
JW
, et al
.
Response of 1,5-anhydroglucitol level to intensive glucose- and blood-pressure lowering interventions, and its associations with clinical outcomes in the ADVANCE trial
.
Diabetes Obes Metab
2019
;
21
:
2017
2023
25.
Newgard
CB
,
An
J
,
Bain
JR
, et al
.
A branched-chain amino acid-related metabolic signature that differentiates obese and lean humans and contributes to insulin resistance
.
Cell Metab
2009
;
9
:
311
326
26.
Mann
G
,
Mora
S
,
Madu
G
,
Adegoke
OAJ
.
Branched-chain amino acids: catabolism in skeletal muscle and implications for muscle and whole-body metabolism
.
Front Physiol
2021
;
12
:
702826
27.
Magkos
F
,
Bradley
D
,
Schweitzer
GG
, et al
.
Effect of Roux-en-Y gastric bypass and laparoscopic adjustable gastric banding on branched-chain amino acid metabolism
.
Diabetes
2013
;
62
:
2757
2761
28.
Corbin
LJ
,
Hughes
DA
,
Bull
CJ
, et al
.
The metabolomic signature of weight loss and remission in the Diabetes Remission Clinical Trial (DiRECT)
.
Diabetologia
2024
;
67
:
74
87
29.
Mikkola
TM
,
Salonen
MK
,
Kajantie
E
,
Kautiainen
H
,
Eriksson
JG
.
Associations of fat and lean body mass with circulating amino acids in older men and women
.
J Gerontol A Biol Sci Med Sci
2020
;
75
:
885
891
30.
Geidenstam
N
,
Magnusson
M
,
Danielsson
APH
, et al
.
Amino acid signatures to evaluate the beneficial effects of weight loss
.
Int J Endocrinol
2017
;
2017
:
6490473
31.
Vangipurapu
J
,
Stancáková
A
,
Smith
U
,
Kuusisto
J
,
Laakso
M
.
Nine amino acids are associated with decreased insulin secretion and elevated glucose levels in a 7.4-year follow-up study of 5,181 Finnish men
.
Diabetes
2019
;
68
:
1353
1358
32.
Arnoriaga-Rodríguez
M
,
Mayneris-Perxachs
J
,
Contreras-Rodríguez
O
, et al
.
Obesity-associated deficits in inhibitory control are phenocopied to mice through gut microbiota changes in one-carbon and aromatic amino acids metabolic pathways
.
Gut
2021
;
70
:
2283
2296
33.
Henninger
J
,
Eliasson
B
,
Smith
U
,
Rawshani
A
.
Identification of markers that distinguish adipose tissue and glucose and insulin metabolism using a multi-modal machine learning approach
.
Sci Rep
2021
;
11
:
17050
34.
Pinckard
KM
,
Shettigar
VK
,
Wright
KR
, et al
.
A novel endocrine role for the BAT-released lipokine 12,13-diHOME to mediate cardiac function
.
Circulation
2021
;
143
:
145
159
35.
Stanford
KI
,
Lynes
MD
,
Takahashi
H
, et al
.
12,13-diHOME: an exercise-induced lipokine that increases skeletal muscle fatty acid uptake
[
published correction appears in Cell Metab 2018;27:1357
].
Cell Metab
2018
;
27
:
1111
1120.3
36.
Larabi
AB
,
Masson
HLP
,
Bäumler
AJ
.
Bile acids as modulators of gut microbiota composition and function
.
Gut Microbes
2023
;
15
:
2172671
37.
Ocaña-Wilhelmi
L
,
Martín-Núñez
GM
,
Ruiz-Limón
P
, et al
.
Gut microbiota metabolism of bile acids could contribute to the bariatric surgery improvements in extreme obesity
.
Metabolites
2021
;
11
:
733
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.