Introduction

In humans, obesity increases the risk of developing type 2 diabetes. Inflammatory cells, particularly macrophages in visceral adipose tissue, have been implicated in the progression of obesity to type 2 diabetes [1]. Skeletal muscle is the primary site of dietary glucose disposal in humans, and muscle insulin resistance is essential for whole-body insulin resistance [2]. Numerous studies support the concept that skeletal muscle cells (fibres) develop insulin resistance as a result of the lipotoxic environment on high fat-feeding and obesity [3]. In vitro, muscle cells exposed to saturated fatty acids develop insulin resistance and mount a cell-autonomous inflammatory response [4]. On the other hand, there is evidence that, in vivo, muscle insulin resistance is reactive to factors emanating from adipose and hepatic tissues, in particular adipose tissue macrophage infiltration [4].

However, a third possible scenario remains untested, namely that macrophages with inflammatory signatures increase within skeletal muscle beds and contribute locally to muscle insulin resistance. Muscle-infiltrating macrophages in obesity might affect muscle glucose metabolism via secreted factors that impair glucose homeostasis. Although the rise in human muscle inflammatory genes in obesity and type 2 diabetes correlates with insulin resistance [5], evidence for elevated immune cells in human muscle in obesity and type 2 diabetes is scant and controversial [6, 7]. Moreover, the phenotypic polarisation of muscle macrophages in obesity and type 2 diabetes is unknown, as is any possible correlation to metabolic status. Here, we assess the expression of genes reflecting different macrophage phenotypes in human muscle biopsies and establish correlations with whole-body insulin action.

Methods

Study participants and muscle biopsies

All studies were approved by the Ethics Committee of the University of Leipzig, Germany, and all participants gave written informed consent. Individuals in cohorts I and II were divided into those with normal glucose tolerance (NGT) and those with type 2 diabetes on the basis of a 75 g oral glucose tolerance test according to American Diabetes Association criteria: participants with NGT had fasting plasma glucose (FPG) <6.0 mmol/l and 120 min plasma glucose <7.8 mmol/l; those with type 2 diabetes had FPG ≥6.0 mmol/l and/or 120 min plasma glucose >11.1 mmol/l.

Cohort III comprised individuals with or without previous type 2 diabetes diagnosis and metformin requirement, who underwent a documented 12 month exercise programme: 60 min aerobic training twice weekly (20 min warming and cool down, 20 min running or cycling and 20 min resistance training) plus 60 min swimming once weekly. Insulin sensitivity and glucose disposal were assessed by euglycaemic–hyperinsulinaemic clamps. Skeletal muscle and blood samples were collected 8–10 h after an overnight fast. Plasma was analysed for glucose, insulin, HbA1c and C-reactive protein. Vastus lateralis muscle biopsies (<100 mg) were obtained under local anaesthesia and immediately snap frozen in liquid nitrogen.

Gene expression analysis

Frozen muscle tissue was homogenised in Trizol, RNA isolated (RNeasy Mini kit, Qiagen, Mississauga, ON, Canada) and cDNA generated (SuperScript VILO cDNA Synthesis Kit, Life Technologies, Burlington, ON, Canada). Quantitative PCR was performed with Gene Expression master mix (Life Technologies) using a 7900HT Fast Real-Time PCR System (Life Technologies). Specific gene expression was measured with the following Taqman assays in all samples (Life Technologies): ITGAM, encoding cluster of differentiation (CD)11b (Hs00355885_m1); ITGAX, encoding CD11c (Hs01015064_m1); CD163 (Hs00174705_m1); CLEC10A, encoding macrophage galactose-binding lectin (MGL) (Hs00197107_m1); MRC1, encoding mannose receptor (MR) (Hs00267207_m1); TNF encoding TNF-α (Hs00174128_m1); and TGFB1, encoding TGF-β1 (Hs00998133_m1). The cDNA input was 8 ng/reaction. Comparative cycle threshold results were expressed relative to the reference gene Ras-like without CAAX 1 (RIT1, Hs00608424_m1).

Gene expression data were analysed by Kruskal–Wallis test and Spearman rank correlation analysis. Pre- and post-exercise data for BMI and glucose-disposal rate (R d) were compared by Wilcoxon matched-pairs signed-rank test. Values of p < 0.05 were considered significant. Only results with significant p values are shown.

Results and Discussion

ITGAX expression in muscle correlates positively with glycaemia

In cohort I (Table 1), we tested the hypothesis that markers of pro- and anti-inflammatory macrophages in muscle biopsies correlate with metabolic variables in lean and obese NGT individuals, and in type 2 diabetes patients. The expression of the pro-inflammatory macrophage gene ITGAX (encoding CD11c) correlated with FPG (Fig. 1a, p = 0.0284) and HbA1c (Fig. 1b, p = 0.0139). ITGAX was not expressed by cultured human myotubes (data not shown). ITGAX expression tended to correlate inversely with R d (not shown, p = 0.0568) and also to increase with age (not shown, p = 0.0552). When adjusted for age, ITGAX expression no longer correlated significantly with HbA1c (Fig. 1c, p = 0.0933) or FPG (not shown, p = 0.1819). Expression of the anti-inflammatory cytokine TGFB1 (primarily expressed by leucocytes, including macrophages) correlated negatively with FPG (Fig. 1a, p = 0.0095), HbA1c (Fig. 1b, p = 0.0471) and fasting plasma insulin (FPI) (Fig. 1d, p = 0.0310).

Table 1 Patient characteristics per cohort
Fig. 1
figure 1

(ad) Correlations between vastus lateralis muscle gene expression and metabolic variables in individuals from cohort I (n = 26). (a) Expression of ITGAX, encoding CD11c (black squares; p = 0.0284; ρ = 0.43) correlates positively with FPG, and TGFB1 expression (white squares; p = 0.0471; ρ = −0.39) correlates negatively with FPG by Spearman rank correlation analysis. (b) ITGAX expression (black squares; p = 0.0139; ρ = 0.48) correlates positively and TGFB1 expression (white squares; p = 0.0095; ρ = −0.50) correlates negatively with HbA1c. (c) After correction for age, expression of ITGAX (black squares) tends to correlate with HbA1c (p = 0.0933; ρ = 0.34). (d) Expression of TGFB1 correlates negatively with FPI (white squares; p = 0.0310; ρ = −0.42). To convert values for HbA1c in % into mmol/mol, subtract 2.15 and multiply by 10.929. (e) Relative gene expression in muscle in groups of individuals from cohort I (mean ± SEM; white bars, NGT; BMI <25 kg/m2; grey bars, NGT, BMI >30 kg/m2; black bars, type 2 diabetes, BMI >25 kg/m2). Expression of ITGAM (p = 0.0094) and ITGAX (p = 0.0069) differed significantly between obese NGT patients and overweight/obese type 2 diabetes patients (Kruskal–Wallis test). (f) ITGAX expression tends to differ between obese NGT and type 2 diabetes patients, when corrected for age (p = 0.0542). (gh) Correlations between muscle gene expression and metabolic variables in cohort II (n = 16). (g) TGFB1 (white squares; p = 0.0374; ρ = −0.52) correlates negatively with FPG. (h) Expression of TGFB1 (white squares; p = 0.0484; ρ = −0.50) and CLEC10A (white circles; p = 0.0425; ρ = −0.51) correlates negatively with HbA1c. (i) Type 2 diabetes patients have lower TGFB1 expression in muscle compared with overweight/obese NGT individuals in cohort II (mean ± SEM shown; p = 0.0303). Expression of all genes was normalised to the expression of RIT1 for each sample: ΔCt = Ct(test gene) − Ct(reference gene). Comparative gene expression was calculated as ΔΔCt = ΔCt(disease group) − ΔCt(control group). Fold-change values were calculated as \( {2^{{-\varDelta \varDelta {{\mathrm{C}}_{\mathrm{t}}}}}} \). Correlations were based on −ΔCt values, as a low ΔCt value indicates high expression. BMI is given in kg/m2. * p < 0.05; ** p < 0.01. T2D, type 2 diabetes

The association of ITGAX with metabolic disease was substantiated by elevated ITGAX expression in ten patients with type 2 diabetes and BMI >25 kg/m2 compared with eight individuals with NGT and BMI >30 kg/m2 (Fig. 1e, p = 0.0069). However, this difference was not significant after correcting for age (Fig. 1f, p = 0.0542). Expression of ITGAM , encoding the pan-macrophage marker CD11b, was also higher in the type 2 diabetes group (Fig. 1e, p = 0.0094) and did not correlate with age.

Hence, we demonstrate that expression of the pro-inflammatory macrophage ‘M1’ marker CD11c correlates positively with the metabolic disturbances in type 2 diabetes. However, the increase in muscle ITGAX expression in type 2 diabetes may be an age-dependent phenomenon (see below).

TGFB1 and MGL expression in muscle correlates negatively with glycaemia

A second, age-matched cohort of individuals was recruited to re-test the hypothesis (cohort II, Table 1). Importantly, ITGAX expression did not differ between NGT and type 2 diabetes; however, similar to cohort I, TGFB1 expression was lower with higher FPG (Fig. 1g, p = 0.0374) and HbA1c (Fig. 1h, p = 0.0484), and tended to correlate positively with R d (not shown, p = 0.0609).

To detect ‘M2’-polarised (i.e. non-inflammatory) macrophage markers, we analysed the marker genes CLEC10A (MGL), MRC1 (MR) and CD163. CLEC10A expression was inversely correlated with HbA1c (Fig. 1h, p = 0.0425) and tended to correlate negatively with FPG (not shown, p = 0.0639). MRC1 and CD163 expression did not correlate significantly with metabolic variables. TGFB1 expression was significantly reduced in type 2 diabetes patients compared with overweight/obese NGT individuals (Fig. 1i). In this cohort, individuals with NGT and type 2 diabetes differed markedly in FPI levels. Thus, high TGFB1 expression clusters with low FPI.

Taken together, TGFB1 expression differed between NGT and type 2 diabetes patients in one cohort and correlated with low HbA1c and FPG in two cohorts, independent of age. This raises the interesting possibility that muscle TGF-β1, although implicated in adipose tissue remodelling and fibrosis in obesity [8], may also influence blood glucose homeostasis.

MGL, MRC1 and CD163 expression correlates with insulin sensitivity in trained overweight humans

Finally, muscle biopsies were obtained from 15 individuals with BMI >25 kg/m2 after 1 year of prescribed physical activity (cohort III, Table 1). Exercise had a pronounced beneficial effect on BMI (Fig. 2a, p = 0.0002 for pre-exercise vs post-exercise) and, particularly, R d (Fig. 2b, p = 0.0010). After the exercise regimen, R d in both NGT and type 2 diabetes groups (68.0 and 66.8 μmol kg−1 min−1, respectively) was higher than R d in type 2 diabetes in both other cohorts (41.7–44.2 μmol kg−1 min−1 range) (Table 1). Expression of the anti-inflammatory markers CLEC10A (p = 0.0031), MRC1 (p = 0.0125) and CD163 (p = 0.0268) correlated strongly with R d (Fig. 2c). There were no significant differences in gene expression between individuals with NGT and type 2 diabetes at this endpoint. Hence, expression of the anti-inflammatory ‘M2’ markers MRC1, CLEC10A and CD163 correlated with better insulin sensitivity in this group of overweight/obese people. This might suggest that the phenotype of muscle-resident macrophages and/or of a recruited population may be impacted by exercise intervention. Further studies are required to explore these possibilities.

Fig. 2
figure 2

(a, b) Significant effects of a 1 year exercise intervention programme on (a) BMI (p = 0.0002) and (b) glucose-disposal rate (p = 0.0010) in individuals with a BMI >25 kg/m2 (cohort III; n = 15; mean ± SEM; Wilcoxon matched-pairs signed-rank test). (c) Correlations between muscle gene expression of anti-inflammatory macrophage genes and insulin sensitivity in cohort III. Expression of CLEC10A (white circles; p = 0.0031; ρ = 0.71), CD163 (black triangles; p = 0.0268; ρ = 0.57) and MRC1 (white triangles; p = 0.0125; ρ = 0.63) correlates with glucose-disposal rate during clamping (R d). ** p < 0.01; *** p < 0.001. Post-ex., after exercise; pre-ex., before exercise

Exercise and weight loss have marked beneficial effects in lowering systemic inflammation markers in the plasma of humans; however, this is not necessarily reflected in the macrophage content within skeletal muscle [6, 9]. Muscle levels of CD68 did not change with a 15 week weight-loss programme in 27 people with a mean initial BMI of 45.8 kg/m2 [9]. Similarly, no differences in net macrophage number or gene expression were observed before and after exercise in patients with type 2 diabetes and BMI ≥30 kg/m2 [6]. Since another study failed to detect weight-loss-dependent alterations in adipose tissue [10], weight loss may not act oppositely to weight gain or greater weight loss may be necessary to modify macrophage numbers. Although we did not have pre-exercise samples, we found that the anti-inflammatory ‘M2’ macrophage markers CLEC10A, MRC1 and CD163 strongly correlated with muscle insulin sensitivity in exercised people with BMI >25 kg/m2.

Conclusion

In summary, we show that pro- and anti-inflammatory macrophage marker genes vary inversely in human muscle, depending on age, glycaemic status and insulin sensitivity. As muscle insulin sensitivity can be affected by some of the same factors as adipocyte insulin action, it is conceivable that the phenotype of the muscle-residing macrophages affects muscle insulin sensitivity. We further show that low expression of CLEC10A and TGFB1 is associated with hyperglycaemia and that low CLEC10A is also associated with low glucose-disposal rate. Optimal polarisation of macrophages may contribute to the amelioration of insulin resistance with routine exercise.