Data stratification
It is well documented that obesity and related health complications are affected by gender, and that sex-specific differences have a genetic basis and cannot be attributed only to differences in hormonal regulation [24]. Gender-specific differences in adiposity as well as fat distribution, in addition to the distinctive genetic basis and hormonal regulation of men and women, may result in sex-specific patterns.
In order to assess the presence of confounding factors, we performed a multi-dimensional scaling analysis, using the EdgeR R package. According to the MDS plot (Figure 1) the points separate into two groups, but not on the basis of the differences between the two conditions (normal vs obese). This finding led to the conclusion that the observed separation in the principal component analysis performed, was the result of gender-specific differences, meaning that a straightforward approach would be prone to confounding biases.
In order to partition the samples into non-overlapping groups, we performed a stratified sampling dividing the total sample population into four strata, considering the sex and diabetes status of the mice. Of the total 85 samples, 21 samples were categorised in the Male-Nondiabetic group, 21 samples were categorised in the Male-Diabetic group, 36 samples were categorised in the Female-Nondiabetic group and 5 samples in the Female-Diabetic group. One sample (DT 123) was not used in the analysis due to missing data.
QTL analysis
We performed haplotype association analysis, using the median value of the weight gained in the twelve weeks of the high-fat diet challenge. We mapped one QTL on chr3 for female mice and one QTL on chr5 for male mice, designated as ObFL and ObML for obesity female locus and obesity male locus, respectively. Details on the position and size of each QTL are given in Table 1. We performed functional analysis with BioInfoMiner on the genes extracted from each of the QTLs using the Gene Ontology and MGI Mammalian Phenotype Ontology vocabularies.
Female mice body weight gain: In female mice, prioritised genes included Sgms2 and Hadh that were found to be involved in various processes, such as increased energy expenditure (MP:0004889), decreased susceptibility to diet-induced obesity (MP:0005659) and increased circulating free fatty acid level (MP:0001554). Interestingly, Sgms2 deficiency in mice increases insulin sensitivity and ameliorates high-fat diet-induced obesity and Hadh-/- mice, while having a disrupted β-oxidation pathway, are also protected from diet-induced obesity [25, 26]. Other prioritised genes are Lef1, Dkk2 and Egf, which are all involved in the Wnt signaling pathway (GO:0016055). Noncanonical Wnt signaling has been shown to contribute to obesity-associated metabolic dysfunction by increasing adipose tissue inflammation [27].
Male mice body weight gain: In male mice the gene that is most highly prioritised is Ppargc1a, which is found to be involved in various enriched processes both in GO and in MGI Mammalian Phenotype Ontology, including decreased muscle weight (MP:0004232), regulation of muscle tissue development (GO:1901863) and lipid modification (GO:0030258). Ppargc1a is a transcriptional coactivator that regulates genes involved in energy metabolism through its interaction with Pparγ. In human GWAS analysis a single nucleotide variant of Ppargc1a (rs8192678) has been associated with susceptibility to obesity and insulin resistance [28]. Another important gene, prioritised by both vocabularies, was Cckar, involved in processes such as abnormal small intestinal transit time (MP:0006002) and abnormal intestinal cholesterol absorption (MP:0002645). Rats with a naturally occurring mutation in Cckar (Otsuka Long-Evans Tokushima Fatty (OLETF) rat) develop diabetes and obesity [29]. Prioritised gene Sod3 is involved along with Ppargc1a in: response to reactive oxygen species (GO:0000302), increased susceptibility to injury (MP:0005165), and abnormal cytokine secretion (MP:0003009). Over-expression of Sod3 in high-fat diet fed mice has been shown to block diet induced obesity [30]. Med28 is involved in the regulation of muscle cell differentiation (GO:0051147). Med28 is one of the subunits of the Mediator complex, which acts as a transcription factor coactivator and plays an important role in muscle metabolism by enhancing the transcriptional activity of Ppargc1a and Pparα [31]. Lastly, Slit2 is also prioritised and also shares enriched terms with Ppargc1a, such as regulation of smooth muscle cell migration (GO:0014910) and response to organonitrogen compound (GO:0010243). Slit2 has been shown to regulate metabolic function and thermogenic activity and improve glucose homeostasis in diet-induced obese mice, mainly through Ucp1 [32], which has been found highly under-expressed in male obese mice, in comparison to female obese mice, in our analysis.
RNA-seq analysis
DE analysis was performed on the Male-Nondiabetic and Female-Nondiabetic groups separately. In order to perform the statistical differential expression tests we divided the samples into two subgroups, 10 samples were categorised in the Male-Nondiabetic-Obese group (the male case group), 11 samples were categorised in the Male-Nondiabetic-Normal (the male control group) 12 samples were categorised in the Female-Nondiabetic-Obese group (the female case group) and 24 samples in the Female-Nondiabetic-Normal group (the female control group). Differences in library size were addressed through the TMM normalisation.
Female non-diabetic normal vs female non-diabetic obese: As a first step we tried to assess the DE genes within the two genders, for this reason we performed the DE tests between control and case groups that belonged to the same gender. The Female-Nondiabetic-Normal samples (control group) versus the Female-Nondiabetic-Obese samples (case group) DE genes list consisted of 1382 DE genes. Using BioInfoMiner for the enrichment analysis we obtained two prioritised gene list of 23 DE genes from GO and 21 DE genes from MGI Mammalian Phenotype. Seven genes were common in both gene lists. Processes that were highly enriched in DE genes in female mice include digestion (GO:0007586), regulation of insulin secretion (GO:0050796), response to lipid (GO:0033993), fatty acid biosynthetic process (GO:0006633), increased energy expenditure (MP:0004889), decreased susceptibility to diet-induced obesity (MP:0005659), abnormal glucose tolerance (MP:0005291) and decreased circulating leptin level (MP:0005668). Top prioritised, linker genes in female mice include Lepr and Ppargc1a that were under-expressed in obese females and Pnlip, Pyy, and Ins2 that were over-expressed in obese females. Lepr functions as a receptor for leptin, an adipose secreted hormone that regulates energy expenditure, satiety, lipid and glucose metabolism and immune system activation. Pancreatic lipase is normally secreted from the pancreas and is efficient in the digestion of dietary fats. Inhibition of Pnlip may prevent high-fat diet-induced obesity in mice. Orlistat, an inhibitor of Pnlip was the first FDA-approved anti-obesity therapeutic target for the treatment of diet-induced obesity [33]. Extra pancreatic expression of Pnlip has been observed in mice induced by fasting via the PPARα-FGF21 signalling pathway [34]. PYY is synthesised and released from specialised cells found predominantly within the distal gastrointestinal tract and regulates appetite. Transgenic mice with increased circulating PYY are resistant to diet-induced obesity [35]. Ins2 ectopic expression in the liver has been observed before in mice subjected to high-fat diet [36]. We hypothesise that Pyy may be expressed ectopically in the liver in the same way as Ins2 in response to the high-fat diet.
Male non-diabetic normal vs male non-diabetic obese: The DE genes list for the Male-Nondiabetic-Normal samples (control group) versus the Male-Nondiabetic-Obese samples (case group) consisted of 1589 DE genes. We performed functional analysis on this list and we obtained two prioritised gene lists of 24 DE genes from GO and 22 genes from MGI Mammalian Phenotype. Ten of the genes from the two prioritised gene lists were common in both. In the case of male mice enriched pathways are overwhelmingly related to muscle and cardiac muscle processes: impaired muscle contractility (MP:0000738), abnormal skeletal muscle mass (MP:0004817), myopathy (MP:0000751), cardiac muscle hypertrophy (GO:0003300). A second category of enriched processes involves ion transport (GO:0006811). Top prioritised linker genes in male mice include Nos1, Ryr1, Des, Ttn all under-expressed in obese males while Ryr2 is over-expressed. Ryr1 is mainly expressed in skeletal muscle. The encoded protein functions as a calcium release channel in the sarcoplasmic reticulum. However, there is a number of studies suggesting that RyRs are widely expressed and have been linked with inositol 1,4,5-trisphosphate receptors in hepatocytes [37]. Down-regulation of Ryr1 could lead to reduced release of Ca2+ from the sarcoplasmic (muscle cells) and the endoplasmic reticulum (hepatic cells) into the cytoplasm and therefore hinder the triggering of muscle contraction as well as the regulation of mitochondrial metabolism and glycogen degradation [38, 39]. Des encodes a muscle-specific class III intermediate filament, which is important to help maintain the structure of sarcomeres and has been linked with hepatic fibrosis due to obesity [40]. Titin is an essential component of skeletal and cardiac muscles, but it has been suggested in recent studies that titin isoforms are expressed in non-muscle tissues including the liver, with an essential role in maintaining cellular organisation and contributing to signal transduction [41]. Ryr2 is primarily expressed in cardiac muscle. Ryr2 channels are associated with mitochondrial metabolism, gene expression and cell survival, in addition to their role in cardiomyocyte contraction. A recent study links Ryr2 to insulin release and glucose homeostasis, suggesting that the upregulation of Ryr2 might be a coping mechanism in response to stress [42]. Nos1 produces nitric oxide (NO), which has multiple biological functions. Downregulation of Nos1 in obesity and diabetes is largely attributed to insulin resistance. [43].
Male vs female comparisons: Finally, in order to assess the DE genes between the two genders we performed DE tests between groups that belonged to different genders. The DE genes list for the Male-Nondiabetic-Normal samples versus the Female-Nondiabetic-Normal samples comparison, consisted of 1923 genes. Functional analysis produced two prioritised gene lists consisting of 28 genes from GO and 25 genes from MGI Mammalian Phenotype, with 9 common genes. As for the DE genes for the Male-Nondiabetic-Obese samples versus the Female-Nondiabetic-Obese samples, the genes list consisted of 1940 DE genes. Functional analysis produced two prioritised gene lists consisting of 40 genes from GO and 22 genes from MGI, with 8 common genes. Between the two groups 482 DE genes are shared while 1441 genes are uniquely DE in the Male-Nondiabetic-Normal versus the Female-Nondiabetic-Normal samples and 1458 genes are uniquely DE in the Male-Nondiabetic-Obese versus the Female-Nondiabetic-Obese samples.
The 482 DE genes that are common in the two comparisons are enriched in genes that are involved in processes such as the epoxygenase P450 pathway (GO:0019373), long-chain fatty acid metabolic process (GO:0001676), negative regulation of gluconeogenesis (GO:0045721), increased circulating gonadotropin level (MP:0003362), and regulation of NF-κB import into nucleus (GO:0042345). Functional analysis of these 482 DE expressed genes produced 21 prioritised genes from GO and 20 prioritised genes from MGI Mammalian Phenotype. Of these 8 genes were common in both lists. Cav1, Egfr, Gstp1, Gcg, and Nox4 are consistently over-expressed in male versus female mice both in the non-obese and obese comparisons. Caveolin–1 regulates hepatic lipid accumulation and glucose metabolism and plays an important role in metabolic adaptation [44]. Hepatic glucagon action has been associated with elevated fatty acid oxidation and might act protectively against non-alcoholic fatty liver disease (NAFLD) [45]. Esr1, Il1b and Ptgs2 are consistently over-expressed in female versus male mice in both comparisons. Ptgs2 is involved in inflammation in fat and drives obesity-linked insulin resistance and fatty liver [46]. Tph1 is under-expressed in non-obese males and over-expressed in obese males in comparison to females. In contrast, Ryr2 is over-expressed in non-obese males and under-expressed in obese males in comparison to females. Tph1 is involved in the synthesis of serotonin, which is known to modulate appetite, energy expenditure and thermogenesis. Tph1-deficient mice fed a high-fat diet are protected from obesity, insulin resistance and NAFLD [47].
Functional analysis of the 1441 genes that were uniquely DE in the Male-Nondiabetic-Normal versus the Female-Nondiabetic-Normal comparison produced enriched processes like abnormal bone volume (MP:0010874), increased mesenteric fat pad weight (MP:0009298), unsaturated fatty acid and lipid metabolic process (GO:0033559, GO:0006629), regulation of hormone levels (GO:0010817), and a number of terms related to inflammatory processes, such as abnormal IgG2a level (MP:0020176), increased susceptibility to autoimmune diabetes (MP:0004803) and small intestinal inflammation (MP:0003306). The list of prioritised genes consisted of 21 genes from GO and 23 genes from MGI, with 9 overlapping genes. Genes involved in inflammatory processes, such as Il2, Il4, Il10, and Il1r1, were predominantly over-expressed in female mice. Vdr and Lepr were also over-expressed in females. Males had over-expressed Mrap2 and Oprm1. Mrap2 encodes a protein that modulates melanocortin receptor signalling. Mice deficient in Mrap2 exhibit severe obesity and a mutation in Mrap2 may be associated with severe obesity in human patients [48].
Finally, functional analysis of the 1458 unique DE genes from the Male-Nondiabetic-Obese versus the Female-Nondiabetic-Obese comparison resulted again in enriched immune response related processes (GO:0006955) and positive regulation of inflammatory response (GO:0050729) with related genes also predominantly over-expressed in obese females. The list of prioritised genes consisted of 34 genes from GO and 20 genes from MGI Mammalian Phenotype, with 4 overlapping genes. Genes related to immunological processes include Ifng, Il6, and Tnf. Other enriched processes were muscle system processes (GO:0003012) and abnormal muscle contractility (MP:0005620), abnormal adrenaline level (MP:0003962), abnormal blood pH regulation (MP:0003027), digestion (GO:0007586) and decreased glucagon secretion (MP:0002711). Genes involved in these processes were predominantly under-expressed in male mice, for example, the Mc2r and Pyy. Notably, both Ins1 and Ins2 are also highly over-expressed in obese female mice compared to almost zero levels of expression in male mice. Mc2r belongs to the MCR family that plays an important role in appetite and energy regulation via leptin signalling. MCRs expression has been found to increase in the liver in response to muscle tissue damage and potentially exert a protective effect through metabolic regulation [49]. Prioritised genes of male versus female comparisons are presented in heatmaps in Figure 2.