Subcutaneous Adipose Tissue and Systemic Inflammation Are Associated With Peripheral but Not Hepatic Insulin Resistance in Humans
Abstract
Obesity-related insulin resistance (IR) may develop in multiple organs, representing various etiologies for cardiometabolic diseases. We identified abdominal subcutaneous adipose tissue (ScAT) transcriptome profiles in liver or muscle IR by means of RNA sequencing in overweight or obese participants of the Diet, Obesity, and Genes (DiOGenes) (NCT00390637, ClinicalTrials.gov) cohort (n = 368). Tissue-specific IR phenotypes were derived from a 5-point oral glucose tolerance test. Hepatic and muscle IR were characterized by distinct abdominal ScAT transcriptome profiles. Genes related to extracellular remodeling were upregulated in individuals with primarily hepatic IR, while genes related to inflammation were upregulated in individuals with primarily muscle IR. In line with this, in two independent cohorts, the Cohort on Diabetes and Atherosclerosis Maastricht (CODAM) (n = 325) and the Maastricht Study (n = 685), an increased systemic low-grade inflammation profile was specifically related to muscle IR but not to liver IR. We propose that increased ScAT inflammatory gene expression may translate into an increased systemic inflammatory profile, linking ScAT inflammation to the muscle IR phenotype. These distinct IR phenotypes may provide leads for more personalized prevention of cardiometabolic diseases.
Introduction
Worldwide, over 2.1 billion people were overweight or obese in 2013 (1). Overweight and obesity are major risk factors for type 2 diabetes (2,3). Adipose tissue dysfunction, rather than excess fat mass per se, is frequently associated with progression toward skeletal muscle, liver, and whole-body insulin resistance (IR) (4,5).
Adipose tissue dysfunction is characterized by increased adipocyte size (hypertrophy) rather than increased adipocyte number (hyperplasia) (6), and this seems to be associated with impairments in adipose tissue lipid buffering capacity (7). In addition, expanded adipose tissue mass is accompanied by an increased production and secretion of proinflammatory cytokines, chemokines, and adipokines (8,9). Both systemic lipid overflow and low-grade inflammation are thought to contribute to the development of IR in skeletal muscle and liver (10).
Whole-body IR reflects defective insulin action in major metabolic organs, such as skeletal muscle, liver, brain, and adipose tissue, and has long been known to precede the development of cardiometabolic diseases (10). IR can develop simultaneously in multiple organs, but the severity may vary between organs. For instance, impaired fasting glucose and impaired glucose tolerance may represent distinct prediabetic phenotypes, which are characterized by more pronounced hepatic or muscle IR, respectively (11). In line with this, altered skeletal muscle fatty acid handling as well as more pronounced peripheral IR have been observed in individuals with impaired glucose tolerance compared with individuals with impaired fasting glucose (12,13). Moreover, the severity of IR at the tissue level may contribute to the differential response to lifestyle and pharmacological interventions. Indeed, physical activity has been shown to mainly target skeletal muscle insulin sensitivity (14), while metformin treatment might have more pronounced effects on hepatic insulin sensitivity (15). Moreover, a low-fat, high–complex carbohydrate diet may result in a more pronounced increase in the disposition index in individuals with prediabetes with hepatic IR, while a Mediterranean diet may have more beneficial effects in individuals with muscle IR (16).
Analysis of gene expression in subcutaneous adipose tissue (ScAT) may help to elucidate the pathways and mechanisms that link adipose tissue function to tissue-specific IR. Previous studies have shown significant dysregulation in the ScAT transcriptome in obesity and IR (17). For instance, Elbein et al. (18) found that ScAT inflammatory and cell cycle–regulatory pathways were upregulated in individuals with whole-body IR compared with insulin-sensitive BMI-matched individuals, while lipid metabolism pathways were downregulated. In addition, Soronen et al. (19) observed upregulation of inflammatory pathways and downregulation of mitochondrial respiratory and lipid metabolism pathways in obese women with whole-body IR. In addition, recent transcriptome analyses of abdominal ScAT during weight loss in individuals from the Diet, Obesity, and Genes (DiOGenes) study highlighted the link between lipid metabolism pathways and glycemic improvement (20).
Studies that focus on detailed characterization of tissue-specific IR phenotypes are scarce, and it is currently not known whether differential ScAT transcriptome profiles are related to more pronounced muscle or hepatic IR. Identification and quantification of metabolic anomalies in various IR phenotypes may provide direction for more personalized lifestyle or pharmacological interventions in the prevention and control of cardiometabolic diseases. Here, we studied the ScAT transcriptome of overweight and obese participants without diabetes from the DiOGenes study (21), a large European multicenter dietary intervention study. We analyzed baseline RNA sequencing data in abdominal ScAT from 368 individuals and identified gene clusters that were associated with muscle or hepatic IR. Subsequently, we investigated whether we could confirm mechanisms suggested by ScAT transcriptome findings by performing complementary analyses on systemic inflammatory profiles in two independent human cohorts, namely, CODAM (Cohort on Diabetes and Atherosclerosis Maastricht) (22) and the Maastricht Study (23).
Research Design and Methods
Study Design
DiOGenes is a multicenter, randomized, controlled dietary intervention study that involved eight European countries. Briefly, 938 overweight or obese adults without diabetes (age 18–65 years, BMI 27–45 kg/m2, and blood fasting glucose <6.1 mmol/L) were included. More details about this study have been described previously (21). For current analyses, baseline data were used of 368 participants for whom abdominal ScAT RNA sequencing and oral glucose tolerance test (OGTT) data were available.
CODAM is a prospective, observational study on, among other things, the natural progression of IR and glucose tolerance. A total of 574 individuals (aged >40 years) were included from a large population-based cohort, as previously described in detail (22). The Maastricht Study is a large population-based cohort (participants aged 40–75 years) that is enriched with participants with type 2 diabetes and that focuses on the etiology of type 2 diabetes, its classic complications, and its emerging comorbidities (23). The analyses described here include the baseline data of 325 CODAM (BMI >25 kg/m2) and 685 Maastricht Study (BMI >27 kg/m2) overweight/obese participants without diabetes for whom data on OGTT and systemic low-grade inflammation were available.
For the multicenter DiOGenes study, the protocol was approved by the ethics committee of each center/country. The DiOGenes study in Maastricht as well as CODAM and the Maastricht Study were approved by the local institutional medical ethics committee (Maastricht University Medical Center+), and the Maastricht Study was also approved by the Minister of Health, Welfare and Sports of the Netherlands (permit 131088-105234-PG). All participants gave written informed consent. These studies were carried out in accordance with the principles of the Declaration of Helsinki.
Estimates of Tissue-Specific IR and Classification of IR Subgroups
DiOGenes participants underwent a 5-point OGTT at baseline. In short, after an overnight fast, venous blood was sampled at baseline and after a 75-g flavored glucose load was ingested. Blood samples were taken at 0, 30, 60, 90, and 120 min, and plasma was stored at −80°C until analysis to determine glucose and insulin concentrations. CODAM participants underwent a 4-point OGTT with blood samples taken at 0, 30, 60, and 120 min. The Maastricht Study participants underwent a 7-point OGTT with blood samples taken at 0, 15, 30, 45, 60, 90, and 120 min. Muscle insulin sensitivity and hepatic IR were estimated using the methods of Abdul-Ghani et al. (24).
The muscle insulin sensitivity index (MISI) was calculated according to the following formula: MISI = (dG/dt)/mean plasma insulin concentration during OGTT. Here, dG/dt is the rate of decay of plasma glucose concentration during the OGTT, calculated as the slope of the least square fit to the decline in plasma glucose concentration from peak to nadir (24). The decline in plasma glucose concentration in the second part of the OGTT primarily reflects glucose uptake by peripheral tissues, mainly skeletal muscle. This index has been developed and validated against measures of peripheral insulin sensitivity as assessed during a hyperinsulinemic-euglycemic clamp using a stable isotope glucose tracer (24).
The hepatic IR index (HIRI) was calculated using the square root of the product of the areas under the curve (AUCs) for glucose and insulin during the first 30 min of the OGTT, i.e., sqrt(glucose 0–30 [AUC in mg/dL∗h] ∗ insulin 0–30 [AUC in µU/mL∗h]). This index has been developed and validated against the product of fasting plasma insulin and endogenous glucose production during a hyperinsulinemic-euglycemic clamp using a stable isotope glucose tracer (24).
DiOGenes, CODAM, and Maastricht Study participants were each divided into four groups based on the tertiles of HIRI and MISI scores for each cohort (16). The lowest tertile of MISI represented individuals with muscle IR; the highest tertile of HIRI represented individuals with hepatic IR. Accordingly, participants were categorized into one of four groups: 1) no IR, 2) IR primarily in muscle (muscle IR), 3) IR primarily in liver (liver IR), and 4) IR in both muscle and liver (muscle/liver IR).
Adipose Tissue Biopsy, RNA Preparation, and RNA Sequencing
In DiOGenes, needle biopsies of abdominal ScAT were performed 6–8 cm lateral to the umbilicus under local anesthesia; the biopsy specimens were snap-frozen in liquid nitrogen and stored at −80°C until analysis. Total RNA was extracted from these abdominal ScAT specimens as previously described (25). For each sample, we used GenomicAlignments to retrieve the number of reads mapping onto 53,343 genes (GRCh37 assembly). Only reads with both ends mapping onto a single gene were considered.
Statistical Analyses
Participant Characteristics
In DiOGenes, CODAM, and the Maastricht Study, descriptive variables with a skewed distribution were loge-transformed before further analyses. The phenotypic differences between tissue-specific IR groups were assessed using one-way ANOVA for continuous variables with Bonferroni post hoc tests, adjusted for sex, and χ2 test for categorical variables. The threshold for statistical significance was set at P < 0.05.
Differential RNA Expression Analyses
In DiOGenes, after alignment and determination of transcript abundance, we analyzed estimated raw count data for 53,343 genes and 368 samples in RStudio software (version 3.2). Differential expression analysis was performed in the DESeq2 package (version 1.12.4) using default settings (26). Prefiltering of genes with low counts was applied by removing genes with a total of zero reads or one read. The standard differential expression analysis steps are wrapped into a single function, DESeq (26). The differential expression analysis in DESeq2 implements a negative binomial generalized linear model with the tissue-specific IR phenotypes (muscle IR, liver IR, and no IR), study center, sex, BMI, and waist-to-hip ratio as covariates. Due to a small number of men (n = 12) in the muscle IR group, we were not able to stratify our results by sex. The DESeq2 model internally corrects for library size; no additional data normalization was applied. The result tables were generated using the function results comparing the tissue-specific IR phenotypes (liver IR vs. no IR, muscle IR vs. no IR). Furthermore, the method fpkm was used from DESeq2 to calculate relative expression of transcripts. We obtained the transcript lengths using the Biostrings R package. The GRange transcript lengths were added to the DESeq object through the use of the rowRanges method. Then the fpkm method extracted the values for fragments per kilobase of transcript per million mapped reads.
In an additional analysis, including all participants, MISI and HIRI were simultaneously included as continuous independent variables in the model with gene expression as the dependent variable to assess their independent effects. To allow for direct comparison of the effect sizes, we also standardized β coefficients by calculating their z scores. Study center, sex, BMI, and waist-to-hip ratio were included as other covariates in the model.
Gene Ontology Analysis
Gene ontology (GO) analysis was performed using Gorilla software (27). Applied settings were 1) organism, Homo sapiens; 2) running mode, two unranked lists of genes (target and background lists); 3) ontology, process; and 4) P < 0.05. All significant genes (nominal P < 0.05) were included (Supplementary Tables 1 and 2) and divided into up- and downregulated genes to provide direction for the involved biological processes, and this gave four different sets of results. For each result, based on Benjamini and Hochberg false discovery rate P values, the top five biological processes were selected.
Pathway Analysis
Pathway analysis was performed with PathVisio 3.2.1. The curated human pathway collection was obtained from WikiPathways (version 20171116) (28), containing 344 pathways. An overrepresentation analysis was performed with the RNA sequencing data set. The pathways were then ranked based on a standardized difference score (z score). Pathways were considered significantly changed when 1) z score was >1.96, 2) permuted P was <0.05, and 3) the pathway contained three or more significantly different genes (nominal P < 0.05).
Association With Low-Grade Inflammation Score
In CODAM and the Maastricht Study, linear regression analyses were performed with a low-grade inflammation score as the independent variable and HIRI or MISI (continuous variables) as the dependent variable (model 1). Skewed variables were loge-transformed. The low-grade inflammation score was calculated by averaging the z scores of eight inflammatory markers in CODAM (C-reactive protein [CRP], interleukin [IL]-6, IL-8, serum amyloid A [SAA], soluble intercellular adhesion molecule 1 [sICAM-1], tumor necrosis factor-α [TNF-α], ceruloplasmin, and haptoglobin) and six markers in the Maastricht Study (CRP, IL-6, IL-8, SAA, sICAM-1, TNF-α). Averaging these markers of low-grade inflammation produces a robust estimate of the overall inflammatory state with minimization of random error (29). To allow for direct comparison of the effect sizes, we also standardized HIRI and MISI by calculating their z scores. In model 2a, HIRI was included as a covariate in analyses on MISI, and vice versa, to assess the independent effects of tissue-specific IR, and in model 2b, age, sex, BMI, and waist-to-hip ratio were included as covariates. Finally, in model 3, the covariates of model 2a and 2b were combined. All data were analyzed using SPSS for Mac version 24.0 (SPSS Inc., Chicago, IL) and R statistical programming language (version 3.3.3).
Data and Resource Availability
RNA expression data are available from the Gene Expression Omnibus under accession no. GSE95640 (20). Other data are unsuitable for public deposition due to ethical restrictions and privacy of participant data. Data are available from the corresponding author for any interested researcher who meets the criteria for access to confidential data and upon reasonable request.
Results
Characteristics of the DiOGenes Population
Table 1 shows the demographic and metabolic characteristics of the DiOGenes participants according to their tissue-specific IR phenotype. In the no-IR group, 68% of participants were women. Participants in the liver-IR group were less often women (47%; P = 0.006), while participants with muscle IR were more often women (82%; P = 0.027). Comparisons of other characteristics between groups were adjusted for sex. Participants in the muscle/liver-IR group had significantly larger waist circumference and waist-to-hip ratio than those in the no-IR group. The muscle/liver-IR group also had significantly larger waist circumference, larger waist-to-hip ratio, and higher systolic blood pressure than the muscle-IR group. Plasma triacylglycerol concentrations were significantly higher in the liver-IR and muscle/liver-IR groups than in those with no IR. Cholesterol concentrations were highest in participants in the liver-IR group, while HDL cholesterol was lowest in the muscle/liver-IR group. Concentrations of the inflammatory marker CRP were significantly higher in the participants in the muscle/liver-IR group than in those with no IR. Insulin concentrations were significantly higher in all groups than in the no IR group. Previous studies have shown similar characteristics for the various tissue-specific IR groups (16).
Characteristics of DiOGenes participants
Extracellular Structure Organization Is Differentially Regulated in the Liver-IR and No-IR Groups
We performed RNA sequencing on the abdominal ScAT to assess transcriptome differences in adipose tissue in individuals with liver IR or muscle IR. The no-IR group was used as reference, and the analyses were adjusted for study center, sex, BMI, and waist-to-hip ratio. To gain insight into the differentially expressed transcriptome in the tissue-specific IR phenotypes, we performed GO enrichment analyses focusing on biological processes. We identified 2,416 genes that were differentially expressed in the liver-IR versus no-IR comparison (nominal P < 0.05), with in total 1,221 upregulated and 1,195 downregulated genes (Fig. 1 and Supplementary Table 1). The GO classifications with the strongest statistical significance for the upregulated genes was “extracellular structure organization” (including genes related to collagens, integrins, laminins, and thrombospondins) and “mitotic cell cycle process” (including genes related to cell division cycle and kinesins) (Table 2 and Supplementary Table 5). In additional analyses in which adipose tissue gene expression was the dependent variable, continuous HIRI was the independent variable, and adjustments were made for study center, sex, BMI, waist-to-hip ratio, and MISI, similar significant associated genes (Supplementary Table 9) and GO terms (Supplementary Table 11) were found.
Venn diagrams of number of significantly upregulated (A) and downregulated (B) genes (nominal P < 0.05) in comparisons of the liver-IR and no-IR groups and of the muscle-IR and no-IR groups.
GO analyses of differentially expressed genes in the liver-IR and muscle-IR groups compared with the no-IR groups in DiOGenes
Subsequently, we included a pathway enrichment analysis using the WikiPathways human collection of curated pathways. Eight significantly different pathways were overrepresented in the liver-IR versus no-IR comparison, with at least one gene with absolute fold change >1.2 and genes with nominal P < 0.05 (Supplementary Table 7). Pathways related to extracellular matrix (ECM) organization (e.g., focal adhesion pathways) and mitotic cell cycle (e.g., cell cycle) were among the most significantly differing pathways in the comparison of the liver-IR and no-IR groups (Figs. 2 and 3).
A heatmap representing the pathway enrichment analysis that applied the WikiPathways curated collection of human pathways. The pathways are ranked on the basis of a standardized difference score (z score). z scores >0 indicate enrichment for significantly different genes (blue gradients), and z scores <0 indicate an absence of enrichment for significantly different genes (white). Pathways were considered significantly different when 1) z score was >1.96, 2) permuted P was <0.05, and 3) three or more significantly different genes (nominal P < 0.05) existed in the pathway. ECM, extracellular matrix; ID, inhibitor of DNA binding; NF, nuclear factor; TGF, transforming growth factor.
Four large pathways from the WikiPathways curated collection of human pathways, with comprehensive information on inflammation and extracellular matrix (ECM) remodeling, were selected and integrated into two networks. Visualizations were created for each network: two diagrams representing the ECM remodeling processes in the comparison between the liver-IR and no-IR groups (A) and the muscle-IR and no-IR groups (B), and two diagrams representing inflammatory processes in the comparison between the liver-IR and no-IR groups (C) and the muscle-IR and no-IR groups (D). Only genes with absolute fold change >1.1 and a nominal P < 0.05 for the liver-IR vs. no-IR comparison, the muscle-IR vs. no-IR comparison, or both were included. Pathway nodes are visualized as rounded rectangles and genes as circles; the colors are based on their fold change. Gradients of blue indicate downregulation, gradients of red indicate upregulation, and white represents unchanged genes. Genes that changed significantly (nominal P < 0.05) are indicated with a gray outline around the circle representing the particular gene.
In the comparisons of both the liver-IR and muscle-IR groups with the no-IR group, the GO classification with the highest enrichment in downregulated genes was “regulation RNA metabolic process” (Supplementary Tables 3 and 4), which was accompanied by enrichment of upregulated genes in the “mitotic cell cycle process” (Table 2). These data might indicate gene-specific translational repression as a means of controlling the mitotic proteome, which may complement posttranslational mechanisms for inactivating protein function (30).
Inflammatory Pathways Are Differentially Regulated in the Muscle-IR and No-IR Groups
For the muscle-IR versus no-IR comparison, 3,716 genes were differentially expressed (nominal P < 0.05), of which 1,977 were upregulated and 1,739 were downregulated (Fig. 1 and Supplementary Table 2). ScAT of individuals with muscle IR was characterized by higher expression of inflammatory genes. The GO terms with the highest enrichment with upregulated genes in ScAT were “mitotic cell cycle process” (including genes related to cell division cycle and kinesins) (Table 2) and “immune system response” (including genes related to several complement factors and chemokine) (Table 2 and Supplementary Table 4). “Antigen processing and presentation” was highly statistically significant as well (including genes related to proteasome degradation and lysosomal cathepsins) (Table 2 and Supplementary Table 4). Like the GO enrichment analyses, WikiPathways-related inflammatory pathways (e.g., microglia pathogen phagocytosis, IL-1, and megakaryocytes in the obesity pathway and complement activation pathway) were predominant in the comparison of the muscle-IR and no-IR groups (Figs. 2 and 3 and Supplementary Table 8). In additional analyses in which the adipose tissue gene expression was the dependent variable and the continuous MISI was the independent variable, after adjusting for study center, sex, BMI, waist-to-hip ratio, and HIRI we found similar significantly associated genes (Supplementary Table 10) and GO terms (Supplementary Table 11). Thus, ScAT of individuals with muscle IR is characterized by increased expression of genes involved in chemotaxis, complement activation, and immune cell function related to lysosomal (e.g., cathepsins) and ubiquitin-proteasomal degradation.
Systemic Low-Grade Inflammation Is Associated With Muscle IR in CODAM and the Maastricht Study
Following up on the transcriptome analyses, we hypothesized that increased ScAT inflammatory gene expression, as observed in the muscle-IR group, may lead to the secretion of proinflammatory adipokines in the circulation. Subsequently, a systemic proinflammatory profile may induce peripheral insulin sensitivity. As systemic inflammation markers were not available in DiOGenes, we studied the relationship between systemic inflammation and either muscle or liver IR in two independent cohorts, namely, CODAM and the Maastricht Study (22,23). In these cohorts, data on systemic low-grade inflammation and tissue-specific IR are available for 325 and 685 overweight/obese individuals without diabetes, respectively (Supplementary Tables 12 and 13). In general, participants in CODAM and the Maastricht Study had risk profiles for cardiometabolic diseases similar to those of DiOGenes participants but were less often women, slightly older, and less obese (Supplementary Table 14).
A combined score of several plasma markers of low-grade inflammation (which include adipose tissue–derived factors including, among others, IL-6, IL-8, TNF-α, SAA, sICAM, and CRP) was inversely associated with MISI in a linear regression analysis that was adjusted for HIRI (CODAM: standardized β [stdβ], −0.107 [P = 0.028]; Maastricht Study: stdß, −0.131 [P < 0.001]) (Table 3, model 2a). Notably, the low-grade inflammation score was not significantly associated with HIRI in a linear regression analysis that was adjusted for MISI (CODAM: stdß, 0.065 [P = 0.184]; Maastricht Study: stdß, 0.000 [P = 0.995]) (Table 3, model 2a). After additional adjustments for age, sex, BMI, and waist-to-hip ratio, the association between low-grade inflammation and MISI was maintained (CODAM: stdß, −0.051 [P = 0.317]; Maastricht Study: stdß, −0.091 [P = 0.008]) (Table 3, model 3). No significant sex interactions were observed for the associations between the low-grade inflammation score and MISI (Pinteraction = 0.639) or HIRI (Pinteraction = 0.982). Therefore, associations between low-grade inflammation and MISI or HIRI were assessed in men and women combined.
Associations of plasma low-grade inflammation with MISI and HIRI
Finally, additional analyses in DiOGenes showed that in adipose tissue the RNA expression levels of seven out of the eight inflammatory genes that are included in the inflammation score were indeed higher in those with the muscle-IR than in those with no-IR (Supplementary Table 15). For three individual genes (IL-6, sICAM, haptoglobin), these differences were statistically significant. Such consistent upregulation of the inflammatory marker genes was not observed in adipose tissue of those with liver IR.
Discussion
Here, we demonstrate distinct adipose tissue transcriptome profiles in tissue-specific IR. We show that an altered ECM gene expression profile in abdominal ScAT was present in overweight and obese individuals with pronounced hepatic IR. Furthermore, an inflammatory gene expression profile was particularly present in individuals with pronounced muscle IR and not with hepatic IR. We propose that an increased systemic inflammatory profile is a mechanism linking the increased expression of inflammatory genes in abdominal ScAT to muscle IR. In line with this, in two independent human cohorts, we show a relation between the combined score of plasma markers of low-grade inflammation and muscle IR, but not liver IR. Although the relationship between adipose tissue inflammation and IR has previously been posed (8), the findings in the present study extend these previous observations by indicating the tissue-specific nature of this relationship.
Increased abdominal ScAT inflammation appeared to be a predominant phenotype in muscle-IR individuals. This was mainly determined by upregulation of genes involved in chemotaxis, complement activation, and immune cell function related to lysosomes (e.g., cathepsins) and ubiquitin-proteasome. In recent years, several chemokines have been reported to be of importance in the etiology of obesity-associated inflammation and IR (31). Chemokines are crucial for the attraction of leukocytes from the circulation into tissues, including macrophages and T cells. For instance, it is known that higher total leukocyte counts precede and predict the incident risk of type 2 diabetes (32). Additionally, overrepresentation of T-cell–recruiting and –activating genes is of interest, given the recent implication of T-cell recruitment in obesity-related adipose tissue inflammation and IR in animal models (33) and humans (34), although T-cell infiltration in humans is much less characterized (35).
Furthermore, the upregulation of lysosomal genes such as cathepsins, and genes in the ubiquitin-proteasome pathway by which many intracellular proteins are degraded, is in line with the findings of previous studies and may point toward increased removal of apoptotic adipocytes (36). Moreover, complement factors (e.g., C1Q) are known to enhance phagocytosis by linking apoptotic cells and phagocytes, and contribute to the rapid clearance of dead cells (37). Notably, cell death is associated with increased immune cell infiltration (e.g., macrophages) and local tissue inflammation (38). Strong upregulation of ScAT C1Q is consistent with previous reports on ScAT gene expression in humans (19,39–42). C1Qa knockout mice were protected from high-fat diet–induced adipose tissue inflammation, systemic glucose intolerance, and hepatic IR (43). In addition, obesity-related adipocyte death was slightly increased in C1Qa knockout mice, implying that a reduction in the efficacy of clearing dead adipocytes may reduce macrophage inflammatory activation (43).
Although ScAT inflammation has been posed as a central factor in the development of IR, our data show for the first time that a ScAT inflammatory gene expression profile is most pronounced in the muscle-IR phenotype. Secretion media derived from ScAT explants of obese individuals have been shown to impair insulin signaling in both human myotubes (44,45) and hepatocytes (44). Furthermore, intracellular organelles like inflammasomes and autophagosomes in ScAT have been recognized in organ crosstalk and IR, albeit in indirect and incompletely understood ways (46). The fact that we only found a relationship between ScAT inflammatory gene expression and muscle IR, not liver IR, might be related to a differential in vivo blood supply to both tissues. The liver may be more strongly affected by portal vein supply and factors derived from the visceral adipose tissue and the gut, while peripheral tissues like skeletal muscle might be more affected by the peripheral circulation. Indeed, as ScAT drains more systemically, this could affect skeletal muscle IR more than hepatic IR.
We hypothesized that increased ScAT inflammatory gene expression, as observed in the muscle-IR group, may lead to the secretion of proinflammatory adipokines in the circulation and a systemic proinflammatory profile inducing subsequently peripheral insulin sensitivity. In line with the adipose tissue transcriptome data, in two comparable cohorts, CODAM and the Maastricht Study, we showed that low-grade inflammation scores of plasma inflammatory markers were inversely associated with muscle insulin sensitivity (MISI), while we did not observe an association between the systemic low-grade inflammation score and hepatic IR (HIRI). These data provide support for the hypothesis that the link between an inflammatory ScAT gene expression profile, as found in DiOGenes, may translate into an increased systemic inflammatory profile, thereby explaining the link with peripheral insulin sensitivity.
In individuals with liver IR, increased ECM remodeling in abdominal ScAT was a predominant phenotype in the present study. The upregulation of genes related to ECM organization in ScAT is in line with the previously reported increased accumulation of ECM components in dysfunctional adipose tissue in obesity, which may decrease ECM flexibility and reduce adipose tissue plasticity by triggering adipocyte necrosis (47). In general, abnormal collagen deposition is tightly associated with increased local inflammation, characterized by infiltration by macrophages and other immune cells (48). Nevertheless, we did not observe a significant upregulation of ScAT inflammatory genes in individuals with liver IR compared with no-IR individuals. One explanation may be that our ScAT transcriptome data might reflect simultaneous processes related to ECM remodeling and fibrosis in the liver, which could in turn contribute to hepatic IR. Indeed, it has been shown that hepatic IR is also closely associated with increased ECM remodeling and fibrosis in the liver (49).
In this study, we were able to include large independent human data sets from DiOGenes and cohort studies (CODAM, the Maastricht Study). This included adipose tissue transcriptome analysis by RNA sequencing as well as detailed human phenotyping. The data from the multiple 5- to 7-point OGTTs, with glucose and insulin concentrations available, made it feasible to derive tissue-specific IR phenotypes (16,24,50). The estimated MISI and HIRI from OGTT results have been validated against gold standard hyperinsulinemic-euglycemic clamp studies (24) and have been used previously in large cohort and intervention studies (16,50). Nevertheless, contrary to standardized clamp-derived insulin sensitivity measures, MISI and HIRI may to some extent be determined by other biological processes such as the rate of glucose absorption as well as the incretin response. In addition, our data are adjusted for sex and body composition and they show robust and consistent associations between tissue-specific IR, adipose tissue transcriptome profiles, and systemic low-grade inflammation. A limitation of this study is that we do not have data regarding gene expression in visceral adipose tissue, as previous studies reported strong associations between hepatic IR and visceral adipose tissue (5).
To our knowledge, this study is the first to link the adipose tissue transcriptome with tissue-specific IR in a large population of overweight and obese individuals. These data open new and exciting avenues showing distinct tissue-specific IR phenotypes related to the development of type 2 diabetes and cardiovascular disease in overweight and obese individuals. Our findings of differential adipose tissue transcriptome and systemic inflammatory profiles in tissue-specific IR may provide targets and biomarkers for more tailored nutritional or pharmacological interventions in the prevention or treatment of cardiometabolic disease. Indeed, there is evidence from post hoc analyses that the response to nutritional intervention may depend on IR phenotype, e.g., being more insulin resistant at the level of the liver or skeletal muscle (16). Future studies are urgently warranted in order to obtain detailed insight into these differential metabolic phenotypes as well as related biomarkers.
Article Information
Funding. This study was supported through a grant from the Maastricht University Medical Center+ and the Nestlé Institute of Health Sciences, Lausanne, Switzerland. DiOGenes was supported by the European Commission, Food Quality and Safety Priority of the Sixth Framework Program (FP6-2005-513946). The Maastricht Study was supported in part by the European Regional Development Fund via OP-Zuid, the Province of Limburg, the Dutch Ministry of Economic Affairs (grant 31O.041), Stichting De Weijerhorst (Maastricht, the Netherlands), and the Pearl String Initiative Diabetes (Amsterdam, the Netherlands). The Maastricht Study was supported by the Cardiovascular Center (Maastricht, the Netherlands), the School for Cardiovascular Diseases (CARIM), the Care and Public Health Research Institute, School for Nutrition and Translational Research in Metabolism (Maastricht, the Netherlands), Stichting Annadal (Maastricht, the Netherlands), and Health Foundation Limburg (Maastricht, the Netherlands). The initiation of CODAM was supported by grants of the Netherlands Organisation for Scientific Research (940-35-034) and the Dutch Diabetes Research Foundation (98.901).
Duality of Interest. W.H.M.S. has received research support from several food companies, including Nestlé, DSM, Unilever, Nutrition et Santé, and Danone, and pharmaceutical companies such as GlaxoSmithKline, Novartis, and Novo Nordisk, and is an unpaid scientific advisor for the International Life Science Institute (ILSI) and ILSI Europe. A.A. has received grants and personal fees from Gelesis; personal fees from Acino International, BioCare Copenhagen, the Dutch Beer Institute, Groupe Éthique et Santé, IKEA Food Scientific Health Advisory Board, McCain Foods Ltd, Navamedic, Novo Nordisk, Denmark, Pfizer, Saniona, W International, Zaluvida; and grants from DC-Ingredients, outside the submitted work. A.V. is a full-time employee at the Nestlé Institute of Health Sciences SA. E.E.B. receives grant support from food industry companies such as DSM, Danone, Friesland Campina, Avebe, and Sensus, partly within the context of public-private consortia, has received funding from pharmaceutical companies such as Novartis, and is involved in several task forces and expert groups related to the ILSI and ILSI Europe. No other potential conflicts of interest relevant to this article were reported.
Author Contributions. B.W.v.d.K. wrote the manuscript. B.W.v.d.K. and M.K. performed the transcriptomics and statistical analyses. B.W.v.d.K., M.K., M.A., M.M.J.v.G., N.V., W.H.M.S., A.A., A.V., D.L., C.J.H.v.d.K., S.J.P.M.E., C.G.S., C.D.A.S., G.H.G., I.C.W.A., J.W.E.J., C.T.E., and E.E.B. revised the manuscript and read and approved the manuscript for publication. B.W.v.d.K., M.K., M.A., M.M.J.v.G., N.V., C.J.H.v.d.K., C.D.A.S., G.H.G., I.C.W.A., J.W.E.J., C.T.E., and E.E.B. interpreted data. M.A., M.M.J.v.G., and N.V. supervised the statistical analyses. M.A. and J.W.E.J. supervised the transcriptomics analyses. M.M.J.v.G., C.J.H.v.d.K., S.J.P.M.E., C.G.S., and C.D.A.S. designed or contributed to data acquisition in CODAM and the Maastricht Study. W.H.M.S. and A.A. designed the DiOGenes clinical study. A.V. and D.L. performed RNA sequencing analysis. I.C.W.A., C.T.E., and E.E.B. designed and led the current study. C.T.E. and E.E.B. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Footnotes
Clinical trial reg. no. NCT00390637, clinicaltrials.gov
B.W.v.d.K. and M.K. share first authorship.
C.T.E. and E.E.B. share last authorship.
This article contains Supplementary Data online at http://diabetes.diabetesjournals.org/lookup/suppl/doi:10.2337/db19-0560/-/DC1.
- Received June 6, 2019.
- Accepted August 27, 2019.
- © 2019 by the American Diabetes Association.
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 http://www.diabetesjournals.org/content/license.
References
In this Issue
