Participants and recruitment
The Glasgow Visceral and Ectopic Fat With Weight Gain in South Asians (GlasVEGAS) study was registered on ClinicalTrials.gov (NCT02399423). Ethical approval was obtained from the University of Glasgow College of Medical, Veterinary and Life Sciences Ethics Committee, and all participants provided written informed consent. The ethics application and study protocol are included in the Supplementary Information. Participants were men (assessed by self-report), aged 18–45 years, with BMI < 25 kg m2, who had been weight stable (±2 kg) for >6 months, and were either of WE (self-report of both parents of WE ethnic origin) or SA (self-report of both parents of Indian, Pakistani, Bangladeshi or Sri Lankan ethnic origin) ethnic origin. Participants were recruited via personal contacts and local advertising. Exclusion criteria included diabetes (physician-diagnosed or haemoglobin A1c ≥ 6.5% (48 mmol mol−1) on screening), history of cardiovascular disease, regular participation in vigorous physical activity, current smoking, taking drugs or supplements thought to affect carbohydrate or lipid metabolism, or other significant illness that would prevent full participation in the study. The study was undertaken in Glasgow, United Kingdom and participant recruitment occurred between 2015 and 2017. Write-up was delayed by the first author returning to a clinical role after completion of PhD and the effects of the coronavirus disease 2019 pandemic delaying adipocyte gene expression analyses.
Study design
All participants underwent an overfeeding protocol to induce ~7% (minimum of 5%) gain of body mass over 4–6 weeks. This was achieved by asking participants to eat until they felt more full than usual and providing them with high-energy snacks (ice-cream, chocolate bars, potato crisps, cheese, dried fruit and nuts, sugary drinks) to supplement their usual food intake by ~6.2–8.4 MJ per day (1,500–2,000 kcal per day). Once participants gained the required weight, they were placed on a weight-neutral diet for 3 days before the post-weight-gain assessments were made. Assessments of body fat and fat distribution, adipose tissue morphology, metabolic responses to eating, energy expenditure, dietary intake, physical activity and fitness were undertaken at baseline and after weight gain. Participants were then supported to lose the excess weight gained over the following 12 weeks. All but one participant who completed the weight loss phase were followed up until their weight returned to baseline (Fig. 1). However, a large number of participants, particularly in the SA group, declined to undertake post-weight-loss assessments, which meant that power to robustly detect any ethnic differences in responses in the weight loss phase of the study were limited (32% power to detect a 1-s.d. difference between groups). Thus, data are presented for the weight gain phase only.
Measurement of body composition
Height, weight and waist circumference were measured at baseline and following weight gain. Magnetic resonance imaging (MRI) was performed using 3.0-Tesla MRI scanner (Magnetom, Siemens) using a dual-echo 50-min Dixon Vibe protocol from head to feet, which provided water- and fat-separated volumetric data34. Body composition analysed included whole-body lean tissue; upper-body, lower-body and whole-body adipose tissue; and abdominal subcutaneous adipose tissue and VAT. Arms were not included in the whole-body or upper-body MRI body composition measurements. 1H magnetic resonance spectroscopy was undertaken to measure liver fat and quadriceps intramuscular triglyceride levels. Spectra were analysed using jMRUI software and the AMARES algorithm35,36.
Metabolic testing
For 3 days before baseline and post-weight-gain metabolic assessments, all participants’ food was provided, with the amount calculated to ensure a stable body weight during this period. Resting metabolic rate was calculated via indirect calorimetry using a ventilated hood (Quark CPET metabolic cart, COSMED)37 and energy intake was calculated as 1.55 times resting metabolic rate, in line with estimated daily energy requirements for a lightly active man38. ‘Westernized’ foods were provided in these standardized diets with macronutrient proportions of 35–40% fat, 45–50% carbohydrate and 12–17% protein. During these 3-day periods, participants were advised not to exercise or to eat any additional food. Participants monitored their body weight daily to ensure weight remained stable (within 0.5 kg) during these 3-day periods. Physical activity was monitored using hip-worn accelerometers (Actigraph 3TX+, Actigraph) worn during all waking hours for 7 days before each metabolic test day.
On metabolic test days, participants arrived after an overnight fast. Resting metabolic rate and substrate utilization were measured by indirect calorimetry. Then a ~0.5 g subcutaneous abdominal adipose tissue biopsy sample was taken, under local anaesthetic (10 ml 1% lignocaine), lateral to the umbilicus by liposuction with a 14-gauge biopsy needle (Sterican Single Use Deep Intramuscular Needle, B Braun Medical) attached to a 50 ml Luer Lock sterile syringe pre-filled with 5 ml of sterile saline. An ice pack was applied for 20 min to minimize swelling and bruising. A cannula was inserted into an arm vein and a fasting blood sample taken. To provide a physiologically relevant metabolic challenge representative of daily life, participants were given a standard mixed test meal, comprising a plain bagel (New York Bakery Company) with polyunsaturated fat margarine (Flora original; Unilever) and a meal-replacement drink (Strawberry-flavoured Complan, Complan Foods Limited) made with whole milk, containing 3.34 MJ (~800 kcal), with 37% of energy from fat, 47% from carbohydrate and 17% from protein, which was consumed over 10–15 min. Further blood samples were taken 30, 45, 60, 90, 120, 180, 240 and 300 min postprandially. Plasma samples were analysed for glucose, insulin and triglycerides. Time-averaged areas under the postprandial concentration-versus-time curves (that is, AUC, calculated using the trapezium rule, divided by the duration of the postprandial observation period, to give the average postprandial concentration) were used as summary measures of postprandial metabolic responses. The Matsuda insulin sensitivity index was calculated as a measure of whole-body insulin sensitivity16. HOMAIR was calculated from glucose and insulin concentrations in the fasted state15. Insulin secretion was estimated as insulin AUC/glucose AUC39. The ISSI-2, defined as insulin AUC/glucose AUC multiplied by the Matsuda insulin sensitivity index, was used as an estimate of beta cell function40.
Preparation of adipocytes and adipocyte sizing
Adipose tissue samples were placed in collection buffer (Hank’s Medium 199, Thermo Fisher Scientific, 22350029), and tissue temperature was maintained at 37 °C throughout. Adipocyte cell suspensions were then prepared as described by Rodbell41. Five microlitres of adipocyte 90% cytocrit suspension was split into two 5 μl aliquots of Krebs Ringer Hepes buffer on a glass slide, and digital pictures of adipocytes were captured using a ×10 lens under a B×50 microscope by Image-Pro Plus 4.0 (Image-Pro Plus version 4, Media Cybernetics). The intracellular diameters of 150 adipocytes in each sample were measured using ImageJ (version 1.52a, W. Rasband at the National Institutes of Health). Each individual adipocyte diameter was placed in a 10-μm-size band from 0–10 μm to >220 μm. Mid-point values for each 10-μm band were used to calculate mean adipocyte diameter and adipocyte volume (expressed as microlitres per 100 cells). There is no consensus in the literature about thresholds for adipocyte size categories. Based on the size distribution of adipocytes in our cohort, we designated adipocytes of ≤30 μm as very small, 31–70 μm as small, 71–90 μm as medium and >90 μm as large, and calculated the proportion and volume of adipocytes within each size category for each participant at baseline and in response to weight gain.
Adipocyte gene expression
Total RNA was extracted from 50–100 mg adipocyte cells using RNeasy Lipid Tissue Mini Kit (QIAGEN, 74804). The integrity and purification of isolated RNA was confirmed using a NanoDrop 1000 (Thermo Fisher Scientific). The optical density as a 260/280 nm ratio (1.8–2.0 acceptable purity) and the concentration of RNA was recorded. A DNA-free kit (Thermo Fisher Scientific, AM1906) was used to remove any DNA contamination in the RNA samples. Single-stranded complementary DNA (cDNA) used for quantitative PCR was reverse transcribed from purified RNA using a high-capacity cDNA reverse transcription kit (Thermo Fisher Scientific, 4368813). The quantity of specific cDNA targets was pre-amplified using TaqMan PreAmp master mix (2×; Thermo Fisher Scientific, 4384266). Uniformity of pre-amplification was checked using CDKN1 according to the manufacturer’s instructions. Ingenuity pathway analysis was used to identify potential adipocyte genes that may differ between SAs and WEs at baseline or in response to weight gain using the terms ‘insulin resistance’, ‘weight gain’, ‘weight loss’, ‘size of adipocytes’ and ‘adipocyte differentiation’ from the Ingenuity Pathway Analysis database42. Twenty-five genes related to lipid metabolism (ADIPOQ, APOE, CYP19A1, KLF14, LDLR, LEP, LPL, SREBF1), insulin signalling (CASP1, CIDEA, ESR1, GHR, INSR, PIK3R1, PLIN2, SIRT1), adipocyte differentiation (BSCL2, EPAS1, HOXC13, PPARG, TGFB1), and tissue stress response or inflammation (HIF1A, TCF7L2, TNF, TLR2) were selected for study (see Supplementary Table 1 for details of gene functions and assay IDs). The RT–PCR reaction mix was made up for each target gene using 1.25 µl target TaqMan Gene Expression Assay and 12.5 µl TaqMan universal PCR master mix (Thermo Fisher Scientific, 4304437). Diluted pre-amplified cDNA (6.25 µl) was mixed with each PCR reaction mix in duplicate in a MicroAmp fast optical 96-well reaction plate (Thermo Fisher Scientific) and underwent PCR according to the following programme: 50 °C for 2 min, 95 °C for 10 min, then 40× of 95 °C for 15 s and 60 °C for 1 min (StepOnePlus, Real-Time PCR system). PPIA was used as the reference gene43, and relative expression to PPIA of each gene was calculated as: Relative expression to PPIA (%) = 2−(Ct target gene – Ct PPIA) × 100%.
Assessment of cardiorespiratory fitness
Maximal oxygen uptake was assessed using a modified Taylor incremental uphill walking protocol with a treadmill speed of 5.5 km h−1 and gradient increasing by 3% every 2 min44. Oxygen uptake was determined from 1-min expired air samples taken continuously and heart rate was monitored by short-range telemetry. Achievement of maximal oxygen uptake was verified by volitional exhaustion together with a rate of perceived exertion of 19–20, a respiratory exchange ratio of >1.15 and heart rate within ten beats of the age-predicted maximum.
Power calculation and statistical analysis
The sample size was based on the power to detect a 1-s.d. difference for change in outcomes of interest with weight gain between groups, based on the typically ≥1-s.d. difference in postprandial insulin response, fat mass and adipocyte size observed in previous comparisons between normoglycaemic SA and WE adults5,11. Twenty-three participants per group provided 90% power to detect this magnitude of difference, with 80% power provided with 17 per group. To achieve 23 completers per group, allowing for drop out, we aimed to recruit 30 participants per group (that is, 30% over-recruitment). Our final achieved power to detect a 1-s.d. difference with 21 WE and 14 SA participants completing the study was 80.3%.
Statistical analysis was undertaken using Statistica (version 10.0, StatSoft) and Minitab (version 20.0, Minitab, State College). Data distribution was assumed to be normal, but this was not formally tested. Differences between ethnic groups at baseline were analysed using unpaired t-tests. The effects of weight gain (main effect of weight gain) and differences between ethnic groups in changes due to weight gain (weight gain × ethnicity interaction) were analysed by two-way ANOVA (ethnicity × time), with repeated-measures analysis on the time factor (baseline, after intervention). To assess whether ethnic differences in responses to weight gain were independent of any ethnic differences in baseline values, two-way ANCOVAs were performed, with baseline values included as a covariate in the models. To help understand factors contributing to the ethnic differences in the changes in fasting insulin concentrations, the postprandial insulin response, HOMAIR, the Matsuda insulin sensitivity index, insulin AUC/glucose AUC and ISSI-2 with weight gain (change in glucose and triglyceride responses with weight gain did not differ between ethnic groups), univariable and multivariable linear regressions between the change in these variables with weight gain, with body composition variables, fitness and physical activity variables and adipocyte size variables were explored. Univariable associations between adipocyte gene expression and adipocyte size variables were also investigated in exploratory analyses. All tests were two-sided, and statistical significance was accepted at a nominal value of P < 0.05. In this basic human experimental study, our primary aims were around ethnic differences in responses to weight gain in three ‘families’ of data—anthropometric/body composition (10 outcomes, excluding change in weight and BMI, which is the same in both groups, by design), metabolic (12 outcomes) and adipocyte morphology (10 outcomes). We also reported data on changes in cardiorespiratory fitness and physical activity (8 outcomes) and undertook an exploratory analysis of ethnic differences in expression of adipocyte genes related to lipid metabolism, adipocyte differentiation, tissue stress response and inflammation (25 outcomes). To reduce the familywise risk of type I statistical error, we used the method described by Tukey et al.45 (Bonferroni adjustment is considered too conservative when there are multiple high-correlated outcomes as is the case in the present study45), to adjust the P value for statistical significance from 0.05 to 0.02 for body composition, metabolic, adipocyte morphology and fitness/physical activity outcomes (assuming 12 outcomes in each family) and to 0.01 for adipocyte gene expression outcomes (25 outcomes). Biochemical analyses and adipocyte sizing and gene expression analyses were blinded. Data collection and statistical analysis were not performed blind to the conditions of the experiments. All available data points were included in the analysis.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
