Fatty Acid Metabolic Remodeling During Type 2 Diabetes Remission After Bariatric Surgery

White adipose tissue (WAT) expands through adipocyte hypertrophy and hyperplasia (1). Increased fat cell size is linked to WAT dysfunction, characterized by excessive intracellular lipolysis, delayed chylomicron clearance, and impaired storage of nonesterified fatty acids (NEFAs) produced by intravascular lipolysis of triglyceride (TG)-rich lipoproteins. NEFA spillover from adipose tissues toward lean organs contributes to lipotoxicity and insulin resistance, in particular in the heart (2–5). Muscle insulin resistance is also characterized by metabolic inflexibility, described as the incapacity to switch efficiently from lipid to glucose oxidation in the postprandial state (6). Medium- and long-chain acylcarnitines (ACs) are metabolites derived from fatty acid oxidation that cross mitochondrial and plasma membranes to finally reach the systemic circulation. Their plasma levels during fasting and during insulin stimulation have been used as circulating biomarkers of metabolic flexibility in humans (7).

Bariatric surgery procedures have been demonstrated to be superior to conventional medical treatment in terms of weight management and glycemic control (8–10). Weight loss induced by bariatric surgeries substantially improves peripheral insulin sensitivity (11), but glucose homeostasis is normalized within the first days after the intervention, before any significant weight loss (8,9,12,13). Several mechanisms of type 2 diabetes (T2D) remission have been proposed, including the effects of caloric restriction (12,14), modifications of gastrointestinal hormone secretion (15,16), changes in bile acid dynamics (17,18), and adaptations of the gut microbiota (19). One hypothesis is that bariatric surgery improves metabolic function in adipose tissue, leading to improved fatty acid metabolism and reduced lean organ lipotoxicity (20). However, the contribution of this mechanism has, to our knowledge, never been examined longitudinally in response to bariatric surgery. To assess the contribution of WAT metabolic function to T2D remission, we measured glucose tolerance, β-cell function, insulin sensitivity, and NEFA and glycerol metabolism in a fasted state and during a euglycemic-hyperinsulinemic clamp (EHC) without and with i.v. fat loading (HI). Plasma AC concentrations were quantified as surrogates of metabolic flexibility and incomplete fatty acid oxidation. Subcutaneous adipocyte size distribution was determined.

We selected 17 severely obese participants scheduled for laparoscopic biliopancreatic diversion with duodenal switch (DS) at the Institut Universitaire de Cardiologie et de Pneumologie de Québec (IUCPQ). Twelve patients diagnosed with T2D according to American Diabetes Association guidelines (21) and treated with diet, metformin, or a sulfonylurea, as monotherapy or in combination, and five metabolically healthy participants (normal fasting glucose, normal glucose tolerance during a 75-g oral glucose tolerance test, and TG <1.7 mmol/L) were included in this study. The protocol was approved by the IUCPQ ethics review board. All participants provided written informed consent. Each subject completed a 6-h stable isotopic tracer infusion study (2) and a stepwise i.v. glucose tolerance test (22) on two consecutive days before DS (45 ± 8 days), 3–4 days after DS, and 3 and 12 months after DS (Supplementary Fig. 1). Partial remission of diabetes was defined as HbA_1c <6.5% and fasting plasma glucose (FPG) between 5.6 and 6.9 mmol/L. Complete remission was defined as FPG <5.6 mmol/L in the absence of active pharmacological treatment, according to American Diabetes Association criteria (23). Two patients did not complete the metabolic studies scheduled 12 months after DS because of pregnancy (n =1) and diagnosis of low-grade breast cancer (n =1) after the 3-month follow-up visit.

After a 12-h fasting period, body composition was measured by DEXA and bioimpedance; a percutaneous biopsy of superficial subcutaneous WAT was taken (see adipocyte cell sizing). A primed, continuous i.v. perfusion of [1,1,2,3,3-d⁵]-glycerol (priming dose, 1.28 μmol/kg fat-free mass [FFM]; infusion rate, 0.1 μmol/kg FFM/min) and a continuous perfusion of [U-¹³C]-palmitate in 25% human albumin (infusion rate, 0.0125 μmol/kg FFM/min) were started and maintained over 6 h. A high dose of insulin (1.2 mIU/kg/min) was perfused 120 min later, and glucose levels were monitored every 5 min and a 20% dextrose perfusion adjusted to achieve stable euglycemia (5 mmol/L). The EHC was maintained over 4 h, and 20% Intralipid (40 mL/h) mixed with [9,10-³H(N)]-triolein (200 μCi/h) was coinfused with a low dose of unfractionated heparin over the last 2 h of the EHC (referred to as EHC+HI). Blood samples were obtained during the last 30 min of each experimental condition in order to measure hormone and metabolite levels and tracer enrichment, and to isolate chylomicron-like particles by ultracentrifugation (24).

The morning after a 12-h fast, and after a 30-min baseline period, the glucose perfusion rate was gradually increased to 2, 4, and 8 mg/kg/min and maintained for 40 min at each step. After these intervals, the 20% dextrose perfusion was adjusted to maintain a steady hyperglycemic state at 10 mmol/L for 2 h. Serial blood samples were obtained during the last 30 min of each glucose perfusion period.

Plasma levels of glucose, insulin, C-peptide, leptin, glucose-dependent insulinotropic peptide, total glucagon-like peptide 1, and total and high–molecular weight adiponectin were quantified, as previously described (12). A colorimetric assay (Cayman Chemical, Ann Arbor, MI) was used to quantify β-hydroxybutyrate (β-OHB) plasma levels. Plasma concentrations and isotopic enrichment of glycerol, palmitate, oleate, and linoleate were determined as previously described (3,24). Chylomicron-like particles were isolated from plasma samples obtained during the EHC+HI by ultracentrifugation (40,000g, 40 min) (3,24). Total TGs, oleate content, and tritium activity were quantified in order to calculate [9,10-³H(N)]-triolein-specific activity within the TG fraction. NEFAs were isolated from plasma using Oasis solid-phase extraction columns. Tritium activity in the NEFA fraction was quantified by liquid scintillation (25). Long-chain ACs and medium-chain ACs (MCACs) were quantified using a liquid chromatograph coupled to a tandem mass spectrometer (Quattro micro; Waters).

Percutaneous biopsies of adipose tissue (∼50 mg) were obtained under local anesthesia, fixed in collidine-hydrochloride osmium tetroxide solution, and digested in 8 M urea solution. Adipocyte diameter was measured with a Beckman Coulter Multisizer 4 (26). The frequency of each cell diameters was calculated and locally weighted scatterplot smoothing regression was applied to generate a cell size distribution.

A standard laparoscopic biliopancreatic diversion with DS was performed in all participants, as previously described (27).

Whole-body rates of appearance of palmitate (Ra_palmitate), total NEFA flux (Ra_NEFA), and glycerol (Ra_glycerol) were calculated as described (24). Fractional extraction (FE) rates were obtained by dividing the rate of appearance by the plasma concentration of the tracee. NEFA plasma appearance from intravascular lipolysis of exogenous TG-rich particles (i.e., Ra_{NEFA-spillover}) was calculated as the product of the whole-body Ra_NEFA and the ratio of oleate-specific activity in the NEFA and TG pool (28). HOMA of insulin resistance (HOMA-IR) index, peripheral insulin sensitivity index (S_i), and insulin secretion rate (ISR) were calculated as previously described (22). Adipose tissue insulin resistance index was calculated by multiplying fasting Ra_glycerol by fasting insulin levels (nmol/L). Insulin clearance was calculated as the ISR divided by the plasma insulin level during the steady-state hyperglycemic clamp.

Normality of distribution was assessed according to Shapiro-Wilks criteria, and non–normally distributed data were log-transformed before statistical analyses were performed. The effects of group, time, and interaction were analyzed by two-way ANOVA. Multiple comparisons within the same group were performed with the Dunnett post hoc test using the presurgical values as the control. The effect of insulin (fasting vs. clamp) and lipid perfusion (clamp vs. clamp + HI) were analyzed by one-way ANOVA.

Anthropometric and metabolic measurements are shown in Table 1. Three months after DS, patients with T2D and normoglycemic (NG) patients achieved a similar mean excess weight loss of 37.1% and 35.6%, respectively (Table 1). A reduction in fat mass and FFM was observed during this first phase of weight loss. Both groups continued to lose fat mass between the 3rd and the 12th months after DS, whereas FFM remained stable (Table 1). A similar excess weight loss of 83.2% and 84.9% was finally observed in patients with T2D and NG patients 12 months after DS (Table 1).

Antidiabetes medication was discontinued at surgery in all patients with T2D. Complete diabetes remission (23) was achieved in 7 of 11 (63.4%) and 8 of 9 (88.9%) patients with T2D by 3 and 12 months after DS, respectively (Table 1). All other patients achieved partial remission. In patients with T2D, FPG, insulin, C-peptide, and HOMA-IR decreased 3 days after the surgery, followed by further decline to NG levels 12 months after DS (Table 1). The adipose tissue insulin resistance index rapidly decreased 3 days after DS in both groups and remained lower than the baseline value 3 and 12 months after DS (Table 1). Peripheral insulin sensitivity (S_i) did not significantly change 3 days after DS but increased gradually in both groups up to 12 months after DS (Table 1). Tolerance of i.v. glucose and pancreatic β-cell sensitivity to glucose was assessed during a stepwise i.v. glucose perfusion. Glucose excursion remained similar to baseline values 3 days after DS but gradually declined in both groups up to 12 months after DS (Fig. 1A–C). Incremental area under the glucose curve during i.v. glucose perfusion also improved in patients with T2D 3 and 12 months, but not 3 days, after DS (Fig. 1C). Plasma insulin levels were gradually reduced in both groups, but we found no significant change in incremental area under the curve for insulin levels (Fig. 1D–F). β-Cell sensitivity to i.v. glucose perfusion, assessed by the relationship between ISR and glucose levels, did not change in both groups over the first year after DS (Fig. 1G and H). Disposition index (DI) remained lower in participants with T2D than in NG participants but increased significantly in the former group 3 and 12 months after DS compared with before surgery (Table 1). Insulin clearance, measured during the steady state of the hyperglycemic clamp, increased in T2D, reaching a plateau 3 months after DS (Table 1).

Before and 3 days, 3 months, and 12 months after DS, no difference was observed between participants with T2D and NG participants with regard to NEFA and glycerol plasma levels (Fig. 2A and B), rates of appearance (Fig. 2C and D), and FE (Fig. 2E and F) when fasted and during the EHC. Plasma levels of β-OHB, the most abundant ketone body, were also similar between groups (Fig. 2G). However, Ra_glycerol and FE of glycerol (FE_glycerol) during EHC+HI were higher in patients with T2D than in NG patients before the DS (Fig. 2D and F). This was associated with higher plasma TGs (Fig. 3A), chylomicron-like particle TG content (Fig. 3B), and TG radiotracer concentration (Fig. 3C) during EHC+HI in patients with T2D than in NG patients. Ra_{NEFA-spillover} was similar between groups (Fig. 3D).

The high glycerol fluxes observed during the EHC+HI in T2D significantly dropped 3 days after DS and tended to stabilize at these levels 3 and 12 months after DS (Fig. 2D and E). Surprisingly, Ra_NEFA (Fig. 2C) did not significantly drop concomitantly 3 days after DS. However, FE of NEFAs (FE_NEFA) (Fig. 2E) decreased 3 days after DS in patients with T2D, but during the EHC+HI FE_NEFA gradually increased over 12 months toward levels higher than those before DS in both patients with T2D and NG patients. Fasting β-OHB rapidly increased 3 days after DS in both groups (Fig. 2G) and remained increased 3 months after DS in patients with T2D (Fig. 2G). However, β-OHB plasma levels returned to preoperative levels 12 months after DS (Fig. 2G). Plasma TG concentrations declined in all conditions studied and in both groups after DS (Fig. 3A). Twelve months after DS, the high chylomicron-like particle TG levels observed in patients with T2D during the EHC+HI before DS decreased to levels found in NG patients (Fig. 3B and C). Ra_{NEFA-spillover} decreased significantly 3 days after DS in patients with T2D but increased to preoperative levels 3 and 12 months after DS (Fig. 3D).

Long-chain AC species (Fig. 4B–D) and saturated MCAC species (Fig. 4E) were reduced by insulin perfusion. However, C18:0 and unsaturated MCACs remained unchanged during the EHC with and without HI (Fig. 4D and F). Lipid perfusion significantly increased C18:2 plasma concentrations and decreased those of C16:0 (Fig. 4B and C).

A significant group effect was observed for C18:2 plasma levels when fasted (Fig. 4B). Before and 3 days, 3 months, and 12 months after DS, no difference in any other AC species was observed between patients with T2D and NG participants when fasted and during the EHC with and without HI (Fig. 4). Fasting plasma levels of C18:1 increased 3 months after DS in both groups (Fig. 4D). Saturated long-chain AC concentrations and the sum of saturated and unsaturated MCAC concentrations were reduced during the EHC with and without HI 12 months after DS in both groups (Fig. 4A, C, E, and F).

Subcutaneous adipocyte cell size revealed a bimodal distribution in NG patients and those with T2D (Fig. 5A and B). Both groups displayed a similar mode of the adipocyte population over nadir before DS (Fig. 5C). A reduction of the mode of the larger adipocyte population was observed in both groups 3 and 12 months after DS (Fig. 5C), with a larger proportion of small adipocytes (51 and 100 μm) and a smaller proportion of very large cells (101–200 μm) in both groups (Fig. 5D).

Leptin levels rapidly decreased in both groups 3 days after DS and declined further with weight loss in both groups up to 12 months after DS (Table 1). Fasting total and high–molecular weight adiponectin levels increased gradually, reaching a significant increase only 12 months after DS in both groups (Table 1).

HOMA-IR improved rapidly 3 days after DS. Reductions in Ra_glycerol and FE_glycerol during the EHC+HI were significantly associated with the reduction in HOMA-IR (Table 2).

Three months after DS, decrease in fasting glycerol turnover was associated with reductions in HOMA-IR and HbA_1c and with increases in S_i and DI (Table 2). The reduction in Ra_glycerol during the EHC was significantly associated with increased S_i and DI. The reduction in glycerol turnover during EHC+HI correlated with the decrease in HOMA-IR (Table 2). The increase in fasting plasma β-OHB correlated with the reduction in HOMA-IR (Table 2). Moreover, a decrease in TG levels during EHC+HI was associated with an increase in DI (Table 2). Reduction in FE_NEFA during the EHC without and with HI was significantly associated with decreased HbA_1c (Table 2).

A decrease in BMI was associated with increased HOMA-IR 12 months after DS. A decrease in the adipocyte mode was significantly associated with a reduction in HbA_1c (Table 3). Reduction in plasma TG levels when fasted and during the EHC without and with HI was associated with an increase in DI (Table 2). Surprisingly, the increase in Ra_NEFA during the EHC+HI was significantly associated with a reduction in HbA_1c (Table 2). Moreover, the decrease in adipocyte mode was significantly associated with an increase in Ra_{NEFA-spillover} (Table 2).

We examined—to our knowledge for the first time—the temporal relationship between bariatric surgery–induced changes in adipocyte morphology, WAT metabolic function, and glucose homeostasis in morbidly obese individuals without and with T2D. We report three novel key findings in morbidly obese participants undergoing DS: 1) the very early reduction in the HOMA-IR index is best predicted by a decrease in systemic glycerol turnover and lipolysis during i.v. fat loading; 2) improvement in HbA_1c is associated with decreased fat cell size over time, without major effects on systemic NEFA spillover; and 3) improvement in peripheral insulin sensitivity with weight loss occurs concomitantly with a marked reduction in plasma TG levels, improved FE_NEFA, and reduced plasma AC levels during an i.v. lipid load.

DS is the most effective treatment for weight loss, leading to T2D remission in >90% of patients 2 years after surgery (13,27). Patients included in our study lost almost all of their excess weight over the first year after DS, and all patients with T2D were in partial or complete remission (23) 3 months after the intervention. In accordance with a previous study performed after Roux-en-Y gastric bypass (RYGB) (11), we found that T2D remission was achieved after DS by early (within the first 4 days) improvement of hepatic insulin sensitivity, then by progressive improvement of muscle insulin sensitivity and DI, concomitant with marked weight loss, 3 and 12 months after DS. Our results are in accordance with a previous demonstration that weight loss is the major mechanism leading to improved peripheral insulin sensitivity after bariatric surgery (29). However, DS did not significantly change β-cell sensitivity to i.v. glucose stimulation; this finding is in accordance with that of Dutia et al. (15) in RYGB but in contrast to studies that examined β-cell function using an oral glucose challenge (12,30–32).

Our hypothesis stipulated that metabolic function in adipose tissue would improve after DS, with a decrease in cell size leading to greater NEFA storage during lipid overload, less NEFA spillover, and improved TG clearance. We also hypothesized that exposure of lean tissue to fewer NEFAs would alleviate lipotoxicity and improve insulin sensitivity in muscle and liver. We found a significant reduction in NEFA spillover 3 days after DS, but NEFA spillover increased steadily thereafter, going back to baseline by 12 months, despite improvement in all insulin sensitivity measures, glucose homeostasis, and adipose tissue cell size. Furthermore, Ra_NEFA when fasted and during the EHC with and without the lipid load did not significantly change after DS. We found a significant relationship between reduced HbA_1c level and increased NEFA spillover at 12 months. We therefore reject our hypothesis that reduced NEFA spillover is a major mechanism leading to reduced lean organ lipotoxicity and improved insulin sensitivity after DS. Furthermore, reduced plasma NEFA spillover cannot explain the significant association observed between reduced HbA_1c and reduced subcutaneous adipose tissue cell size after DS. It should be noted, however, that lack of change in systemic NEFA appearance occurred despite a marked reduction of plasma insulin level, demonstrating improvement in insulin-mediated suppression of intracellular lipolysis and/or stimulated fatty acid storage in adipose tissue. Weight loss leads to increased rate of abdominal subcutaneous storage of meal fatty acids per volume of tissue (33,34). Our findings suggest that increased NEFA storage capacity per volume of tissue exactly compensates for the decrease in adipose tissue volume during weight loss, resulting in conserved systemic fatty acid availability (35).

The very rapid decrease of systemic glycerol turnover 3 days after DS was closely linked to the improvement in HOMA-IR. Endogenous glycerol fluxes when fasted and during the EHC were reduced in patients with T2D, suggesting that DS induces short-term inhibition of intracellular lipolysis in WAT. Our group (12) and others (14) previously showed that caloric restriction plays an important role in the marked increase of hepatic insulin sensitivity in the first few days after DS and RYGB. However, these earlier studies did not report plasma glycerol kinetics. A previous study reported unchanged Ra_glycerol 2 weeks after RYGB but a significant decrease 1 year after surgery in NG patients and patients with T2D (36). Moreover, glycerol fluxes remained unchanged 1 week after RYGB, gastric banding, or caloric restriction (37). However, RYGB rapidly improved insulin sensitivity in adipose tissue (37). These results underscore the importance of impaired glycerol metabolic regulation for fasting glucose levels and the HOMA-IR index in T2D. In healthy men, the conversion of glycerol to glucose contributes from 10 to 20% of hepatic glucose production after an overnight fast, and this proportion increases after a longer period of starvation (38). Hepatic TG accumulation and hepatic insulin resistance have been associated with impaired inhibition of gluconeogenesis (39). Jin et al. (40) recently showed that an increase in gluconeogenesis from the tricarboxylic acid cycle intermediates and glycerol equally contribute to this phenomenon during an EHC.

We observed higher circulating β-OHB plasma levels in a fasted state during the first 3 months after DS, suggesting increased ketogenesis from enhanced hepatic lipid oxidation. This study was not designed to determine the source of carbon that was used to produce ketones after DS. However, it is unlikely that increased NEFA mobilization from WAT was the main source of this increased ketogenesis, because the NEFA turnover rate remained stable after compared with before DS. This increase in fasting β-OHB was associated with a reduction in HOMA-IR (41). Interestingly, experimental elevation of plasma β-OHB using i.v. infusion was shown to reduce plasma glucose and glycerol appearance rates despite a reduction of insulin level during fasting. The intriguing suggestion that ketone levels may mediate the early DS-mediated reduction of plasma glucose level and systemic lipolysis needs further investigation.

We found a bimodal adipocyte distribution in morbidly obese individuals. Cardiometabolic risk factors and insulin resistance have been associated with a higher proportion of very small adipocytes and with hypertrophy of larger adipocytes (26,42). These reports concluded that WAT hypertrophic remodeling relates to adipose tissue dysfunction and metabolic diseases (26,42). Large adipocyte size is associated with reduced adipose tissue acyl-CoA synthetase and diacylglycerol acyltransferase activities, suggesting lower capacity to store fatty acids (43). In this study we observed a gradual reduction of the proportion of large adipocytes, without major modifications to the proportion of very small adipocytes, in accordance with results from other groups (44,45). The magnitude of weight loss and the reduction of fat cell size were associated with further improvement of HOMA-IR and muscle insulin resistance and with better glycemic control, suggesting that WAT remodeling may be linked to T2D remission. However, we found an increased in vivo NEFA spillover rate with reduced subcutaneous adipocyte size after weight loss in our participants. This suggests that normalization of the higher plasma NEFA spillover observed in participants with prediabetes or T2D (2,3) does not mediate the relationship between improved glycemic control and reduced adipocyte hypertrophy.

Fasting plasma NEFA and VLDL appearance rates were reduced in severely obese participants 6 and/or 12 months after RYGB (46,47). This can explain the reduction in plasma TG levels 3 and 12 months after DS when fasted and during the EHC in the current study. TG clearance was improved, as shown by lower TG and radiotracer plasma levels from the chylomicron-like fraction perfused intravenously. Diet-induced weight loss does not change plasma NEFA levels (34,48–50), but a reduced NEFA appearance rate during fasting was also reported in some (49,50) but not all studies (48). Diet-induced weight loss enhanced NEFA mobilization in adipose tissue and systemic availability during acute exercise in obese women (48). We could not find any study reporting the effect bariatric surgery on i.v. lipid tolerance. However, a hypocaloric diet with a 12-week exercise intervention did not further reduce NEFA turnover during EHC+HI compared with exercise alone in mildly obese individuals without diabetes (50). The latter study also showed that an increase in total-body fatty acid oxidation with weight loss was fully accounted for by fatty acid sources other than plasma NEFAs. This suggests that improved peripheral insulin resistance with weight loss is more closely associated with increased fatty acid utilization than reduced systemic NEFA availability. In our study, FE_NEFA was increased during i.v. lipid perfusion 12 months after DS. Despite an unchanged plasma Ra_nefa, MCAC and long-chain AC plasma levels were reduced during the EHC with and without HI, suggesting that fatty acid handling by lean tissues was improved and that mitochondrial overload was alleviated. Altogether, these data suggest improved TG and NEFA clearance and reduced mitochondrial fatty acid overload despite unchanged plasma Ra_nefa over the long term after DS.

Our study included a small number of participants. However, given the trends observed, the repeated-measurement study design, and the relatively small variance of most metabolic data in our study, it is unlikely that we failed to detect a large change in NEFA turnover rate. We perfused heparin and Intralipid for lipid loading instead of performing a meal test in order to avoid changes in the tolerance of lipids consumed during a meal and their absorption occurring after DS (12). This may not fully recapitulate physiological intravascular TG lipolysis. Finally, we did not measure the energy substrate oxidative rate to assess the metabolic fate of fatty acids.

In conclusion, plasma Ra_{nefa-spillover} and Ra_glycerol were reduced 3 days after DS and were associated with reduced HOMA-IR index. With weight loss, muscle insulin sensitivity improved and plasma TG and NEFA clearance during the i.v. fat load increased. Subcutaneous adipocyte size gradually decreased with weight loss, in association with improved glycemic control. However, plasma NEFA spillover also gradually increased and systemic NEFA availability was maintained despite remission of T2D.

Acknowledgments. The authors acknowledge the important contributions of Suzy Laroche (IUCPQ), Dominique Caron-Dorval (IUCPQ), Marc Lapointe (IUCPQ), Lucie Bouffard (Centre de recherche du CHUS), Mélanie Fortin (Centre de recherche du CHUS), and Diane Lessard (Centre de recherche du CHUS).

Funding. The Canadian Institutes of Health Research (CIHR) (Institute of Nutrition, Metabolism and Diabetes grant MOP97947) and Canadian Diabetes Association (grant OG-3-14-4507-AC) funded this work. T.G.-L. holds a CIHR Doctoral Award. A.-M.C. received awards from the Fonds de recherche du Québec - Santé/Ministère de la santé et des services sociaux (FRQS/MSSS), Canadian Diabetes Association, and Endocrine Fellow Foundation.

Duality of Interest. L.B. and A.T. have received funding from Johnson & Johnson Medical Companies for research on bariatric surgery. D.R. has received the CIHR-Merck Frosst Research Chair on Obesity. A.C.C. has received the GSK Chair in Diabetes of the Université de Sherbrooke. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. T.G.-L. performed all the metabolic studies, collected and analyzed data, and wrote the manuscript. A.-M.C. performed metabolic studies and critically reviewed the manuscript. A.G. performed cell-sizing experiments and critically reviewed the manuscript. F.F. collected data, provided expertise on sample analysis, and critically reviewed the manuscript. L.B., S.M., S.L., and F.-S.H. managed medical issues; performed DS, percutaneous adipose tissues biopsies, and postoperative follow-up; and critically reviewed the manuscript. D.R., A.T., and A.C.C. designed the study and critically reviewed the manuscript. A.C.C. 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.