2.1. Human participants and clinical measurements
Results from two separate clinical studies (studies A, and B) are reported here.
2.1.1. Study A
To identify genes correlated to insulin sensitivity in skeletal muscle, we studied 39 non-diabetic men from Malmö, Sweden [17, 18]. Briefly, the Malmö Exercise Intervention cohort consists of 50 sedentary but otherwise healthy male subjects from southern Sweden. They all have European ancestry and 24 of them have a first-degree family member with T2D. Muscle biopsies were collected from 39 of the subjects. The mean age and body mass index (BMI) were 37.71 ± 0.71 years and 28.47 ± 0.48 kg/m2, respectively, and the mean 1/the homeostatic model assessment-insulin resistance (HOMA-IR) was 0.69 ± 0.04 (Supplementary Table 1).
2.1.2. Study B
To replicate the findings from study A, we studied an additional 10 healthy young non-diabetic men without any family history of diabetes, from a previously described study [19]. The mean age and BMI were 25.33 ± 0.33 years and 24.57 ± 0.62 kg/m2, respectively, and the mean 1/HOMA-IR was 1.17 ± 0.12 (Supplementary Table 2). Here, we included baseline gene expression profile data (i.e. only before bed rest) from part of a larger study on the influence of physical inactivity in healthy and prediabetic individuals [19].
None of the study participants were engaged in vigorous exercise on a routine basis and they were directed to avoid extreme physical exercise and alcohol intake for at least 2 days before the studies. The participants were asked to fast for 10–12 hours before examination days. Fasting blood samples and anthropometric data were obtained from all participants. All participants underwent an oral glucose tolerance test (OGTT; 75 g) and glucose tolerance was classified in accordance with World Health Organization criteria [20]. Homeostasis model assessment -insulin resistance (1/HOMA-IR = 22.5 / (fasting plasma insulin (µU/ml) x fasting plasma glucose (mmol/l))) was calculated for all participants in both studies and used as a surrogate measure of insulin sensitivity [21, 22]. The muscle biopsies were obtained from the vastus lateralis muscle under local anesthesia in individuals participating in all studies using a modified Bergström needle [23].
We excluded data from two participants (one from each studies A and B) with extreme values of insulin sensitivity (more than 1.5 * interquartile range) for further analysis. Both studies were approved by local ethics committees and all participants gave their informed consent for participation.
2.2. RNA extraction and hybridization
Muscle biopsies were taken from the right vastus lateralis muscle under local anesthesia (Lidocaine 1%), using a 6 mm Bergström needle (Stille AB, Sweden). In both studies, biopsies were immediately stored in RNAlater (Ambion, Austin, TX) and after overnight incubation at 4 °C snap frozen at -80 °C until further processing. The double staining method was used for capillary staining. Myofibrillar ATPase histochemistry was performed by preincubation at pH 4.4, 4.6, and 10.3 to identify muscle fiber types [17]. Computer image analysis was performed using BioPix IQ 2.0.16 software (BioPix AB, Sweden). RNA was extracted using Tri reagent (Sigma-Aldrich, St. Louis, MO) followed by RNeasy Midi kit (Qiagen, Düsseldorf, Germany). The RNA was further concentrated by RNeasy MiniElute (Qiagen, Düsseldorf, Germany) and SpeedVac (DNA 120 SpeedVac, Thermo Savant, Waltham, MA).
For study A, synthesis of biotin-labeled cRNA and hybridization to the Affymetrix Custom Array NuGO-Hs1a520180 GeneChip (http://www.nugo.org) were performed according to the manufacturer’s recommendation. This GeneChip contains 23,941 probesets for interrogation, including known genes and expressed sequenced tags. Images were analyzed using the GeneChip Operating System (GCOS; Affymetrix) software. For each array, the percentage present call was greater than 40.
For study B, targets were hybridized to the one-color (Cy3, green) Agilent Whole Human Genome Oligo Microarray (G4112A) which contains 44,000 60-mer oligonucleotide probes representing 41,000 unique genes and transcripts. Probe labeling and hybridization were performed according to manufacturer’s recommendation. Images were analyzed using the Agilent Feature Extraction Software (version 9.5).
2.3. Quantitative real-time PCR (QPCR)
A technical replication of the key findings from the microarray data, as well as expression analysis of key genes to be correlated with insulin stimulated glucose update, was conducted using QPCR. Reverse transcription was performed on 250 ng RNA (from 36 subjects in study A) or 200 ng RNA (from 7 subjects in the Muscle SATellite cell (MSAT) cohort) using the QuantiTect Reverse Transcription kit (Qiagen). QPCR was performed on a ViiA 7 real-time PCR system (Thermo Fisher Scientific) with 2 ng cDNA in 10 µl reactions and TaqMan Expression PCR Master Mix with duplex assays according to the manufacturer’s instructions. Samples were analyzed in triplicates on the same 384 well plate with 3 endogenous controls (POL2A, HPRT1 and PPIA). The expression levels were calculated and normalized by geometric averaging of the endogenous controls as previously described [24]. Assays: SIRT2 (Hs00247263_m1), FBXW5 (Hs00382591_g1) and CPT1B (Hs00189258_m1). Endogenous control assays: POLR2A (Hs00172187_m1), HPRT1 (4326321E, VIC-MGB) and PPIA (4326316E, VIC-MGB).
2.4. Isolation and cultivation of human muscle satellite cells
Muscle satellite cells were isolated from 7 subjects from an ongoing unpublished MSAT study. Subjects were male with a mean age of 35.6 ± 10.6 years, a mean BMI of 25.1 ± 3.6 kg/m2 and a mean fasting plasma glucose value of 5.2 ± 0.2 mmol/L. Muscle biopsies were obtained from the vastus lateralis muscle under local anesthesia in individuals participating in all studies using a modified Bergström needle. Biopsies were minced into small pieces with scissors and digested in a digestion solution (Ham's F-10 Nutrient mix (Gibco®, #31550015), Trypsin-EDTA (0.25 %) (HyClone, SV30031.01), Collagenase IV (1 mg/ml) (Sigma, C5138), BSA (5 mg/ml) (Sigma, A2153)) at 37 oC for a total of 15-20 minutes. After this, cells were passed through a 70 µm cell strainer and centrifuged at 800 g for 7 minutes. The pellet was washed and resuspended in growth medium (Ham's F-10 Nutrient Mix, GlutaMAX™ Supplement (Gibco®, #41550021), FBS (20 %) (Sigma, F7524), Antibiotic/Antimycotic Solution (Gibco®, #15240062)) and cells were pre-plated on a culture dish and incubated for 3 hours at 37 oC and 5% CO2 to allow fibroblast to attach to the plate. After this, the suspended cells were transferred to a flask pre-coated with matrigel (Corning #356234) and were incubated for 4 days at 37 oC and 5% CO2 in growth medium. Medium was then changed every other day. After about a week, cells were detached using TrypLE (TrypLE™ Express, no phenol red (Gibco®, #15090046)) and re-plated on the same flask to allow even distribution of cells over the surface.
At 70-80% confluence medium was changed first to an intermediate medium (DMEM, low glucose, GlutaMAXTM Supplement, pyruvate, No HEPES (Gibco® #21885025), FBS (10 %) (Sigma, F7524), Antibiotics) for 24 hours, and then to a differentiation medium (DMEM, low glucose, GlutaMAXTM Supplement, pyruvate, No HEPES (Gibco® #21885025), Horse serum (2 %) (Invitrogen, #16050-130), Antibiotics) for 8 days, where glucose uptake experiments were performed. After 3 days of differentiation, Cytarabine (Ara-C) (10 µg/ml) (Sigma, C1768) was added to the differentiation medium, for 2 days, to prevent excessive growth of proliferating cells, e.g. fibroblasts [25].
2.5. Measurement of glucose uptake in cultured muscle cells
Measurement of glucose uptake in cultured muscle cells was performed using an enzymatic fluorometric assay as previously described [26]. Briefly, cells differentiated for 8 days grown in 12-well plates, were starved for 3 hours in FBS-free DMHG low glucose medium (Gibco® #21885025) at 37 °C and 5% CO2. The cells were then washed in warm PBS and treated with either Cytochalasin B (10 µM) (Sigma, C6762) (for non-specific glucose uptake), Krebs-Ringer-HEPES (KRH) buffer only (basal glucose uptake) or with 100 nM insulin (Actrapid 100 IE/ml, Novo Nordisk) (stimulated glucose uptake) in a KRH buffer containing 0.1% BSA (pH 7.4) for 1 hour at 37 °C and 5% CO2. After this, cells were incubated in a KRH buffer containing 2-Deoxy-D-glucose (2DG) (1 mM) (Sigma, D6134) for 15 minutes at room temperature, after which the cells were washed in ice-cold PBS and then frozen and stored at -80 oC (for less than a week). Lysis was done by adding 0.1 M ice cold NaOH to the cells and incubate at 70 °C for 60 minutes, after which HCL and triethanolamine (TEA) buffer (pH 8.1) (Sigma, T1502) at final concentrations of 0.1 M and 50 mM respectively, were added to neutralize the lysate. Lysates and prepared series 2-Deoxy-D-glucose 6-phosphate (DG6P) (Santa Cruz, SC-220734) dilution standards (30, 15, 7.5, 3.75, 1.875, 0 μM) (dissolved in “lysate buffer” (0.1 M NaOH / 0.1 M HCl / 50 mM TEA buffer, pH 8.1; same proportion as samples), were transferred to a black 96-well assay plate (Greiner Bio-one International, 655076), 250 µl of assay solution (TEA buffer (50 mM) with KCl (50 mM) (pH 8), BSA (0.02 %), NADP (0.1 mM) (Sigma, N8035), Diaphorase (0.2 U/ml) (Sigma, D2197), Resazurin (6 µM) (Sigma, R7017), G6PDH (15.4 U/ml) (Sigma, G8404)) was added to each well, and the plates were incubated for 60 minutes at 37 °C. Fluorescence was measured using the microplate reader (Infinite M200 Pro, Tecan) at wavelengths λex = 545 nm and λem = 590 nm. DG6P was then quantified by comparing the fluorescence intensity from the experimental samples to the DG6P standard curve. Value were adjusted for protein concentration determined with the PierceTM Coomassie (Bradford) Protein Assay Kit (Thermo Fisher Scientific, 23200).
2.6. Quantification of mtDNA content
DNA was isolated from the muscle biopsies by phenol/chloroform/isoamyl alcohol extraction according to the manufacturer’s recommendation (Diagenode, Belgium). Concentration and purity were measured using a NanoDrop ND-1000 spectrophotometer (A260/A280 > 1.6 and A260/A230 > 1.0) (NanoDrop Technologies, Wilmington, DE, USA). QPCR was carried out using an Applied Biosystems 7900HT sequence detection system with 5 ng genomic DNA in 10 µl reactions and TaqMan Expression PCR Master Mix according to the manufacturer’s recommendations. All samples were analyzed in triplicates on the same 384 well plate (maximum accepted standard deviation in Ct-value of 0.1 cycles). Two assays (16S and ND6) were used to analyze mitochondrial DNA content (mtDNA) targeting the heavy and light strand, respectively. To analyze nuclear DNA (nDNA) content RNaseP was used as a target. The mtDNA content is calculated as the mean value of ND6 and 16S divided by 2 x RNaseP. Assays used: ND6 (Hs02596879_g1), 16S (Hs02596860_s1) and RNaseP (4316838).
2.7. Statistical analysis
2.7.1. Study A
We used ENTREZ custom chip definition files (http://brainarray.mbni.med.umich.edu) to regroup the individual probes into consistent probesets and remap to the correct sets of genes for Affymetrix Custom Array NuGO-Hs1a520180 array which resulted in a total of 16,313 genes from study A. Methods used for calculating gene expression from Affymetrix array data can have a major impact on the results [27-29]. Hence, we used three different procedures for normalization and summarization which combines the multiple probe intensities for each gene to produce an expression value: (1) by the GC-content robust multi-array average (GC-RMA) method with additional background adjustment using sequence information to estimate probe affinity for nonspecific binding, quantile based normalization, and summarization based on a multi-array model fit using median polish algorithm, (2) by probe logarithmic intensity error (PLIER) method (Affymetrix) utilizing both perfect match and mismatch signaling with quantile based normalization, and (3) by robust multi-array average (RMA) method [30, 31] which implements model-based background adjustment, quantile based normalization and summarization based on a multi-array model fit using median polish algorithm. We conducted filtering based on the MAS5.0 present/absent calls which classified each gene as expressed above background (present call) or not (absent or marginal call). We included genes, which have detection call as present call in at least 25% of arrays [32], which left 7,947 genes out of 16,313 for further analysis in study A.
To identify a reliable list of genes regulating insulin sensitivity, Spearman partial correlation analysis was performed to determine the individual effects of each gene expression on a surrogate measure of insulin sensitivity (1/HOMA-IR) after adjusting for BMI, age and family history of T2D for each of three normalization methods namely GC-RMA, PLIER and RMA separately. We considered only those genes that were significantly correlated with insulin sensitivity with a P < 0.05 in all three different normalization methods.
To technically validate the microarray findings, real time quantitative PCR (QPCR) was used to measure the mRNA expression of FBXW5 and SIRT2 in human skeletal muscle from study A. Correlation between the microarray and QPCR experiments was determined using Spearman's rank correlation coefficient test.
In the study A cohort, correlation between the QPCR expression values of SIRT2, FBXW5, CPT1B, FABP3, MLYCD, PPARG1A and ESRRA with % fiber type and mitochondrial DNA was determined using Spearman’s rank correlation coefficient test. All data except that of SIRT2 and FBXW5 was collected and reanalyzed from a previously described study [17, 18].
Enrichment analyses were performed on the genes whose expression levels in skeletal muscle were significantly correlated with insulin sensitivity in study A using the WEB-based GEne SeT AnaLysis Toolkit (WebGestalt) which implements the hypergeometric test [33].
2.7.2. Study B
The median intensities of each spot on the array were calculated using the GenePix Pro software (version 6). We performed quantile-based normalization between arrays without background subtraction using linear models for microarray data (limma) package in R [34, 35]. We removed poor quality probes that were either saturated (i.e. > 50% of the pixels in a feature are above the saturation threshold) or flagged as non-uniformity outlier (i.e. the pixel noise of feature exceeds a threshold for a uniform feature) in at least one array, which left 29,297 probes for further analysis [36].
Spearman partial correlation analysis was performed to determine the individual effects of each gene expression on a surrogate measure of insulin sensitivity (1/HOMA-IR) after adjusting for BMI, age and family history of T2D. Due to the exploratory nature of the study, no correction for multiple testing was performed. Instead, we considered only those genes that were significantly, positively or inversely, correlated with insulin sensitivity in both studies A and B with a significance level set to 0.05. Paired Wilcoxon signed-rank test was conducted to assess for the change before and after insulin-stimulated glucose uptake. Spearman correlation analysis was between basal- and insulin-stimulated glucose uptake and mRNA expression of FBXW5, SIRT2 and CPT1B. All statistical analyses were performed using IBM® SPSS® Statistics, MATLAB® and R statistical software.