This cross-sectional study recruited individuals of both sexes who were 40 to 60 years old. Volunteers were excluded if they were on vitamin D supplementation, received insulin or insulin-secretion–promoting drugs, had diseases that prevented them from performing an ergospirometry test, history of cardiovascular or cerebrovascular events, chronic respiratory diseases, systemic inflammatory conditions, diabetes mellitus, cancer, or were vegetarian or pregnant.

Participants were recruited between March 2017 and February 2019, in Medellin, Colombia. All of them underwent a complete medical history and physical examination, as well as biochemical tests. Based on that, the subjects were classified as having MS or not (controls), according to the internationally accepted criteria [24]. The waist circumference cutoff points were set according to the recommendation for the Colombian population [25]. All subjects with MS had IR, as defined by a homeostatic model assessment of insulin resistance (HOMA-IR) greater than 2.25 [25]. Most patients with MS also participated in the intraining-MET clinical trial [26].

After being assigned into study groups, the volunteers underwent cardiorespiratory fitness, body composition, and muscle composition studies, evenly scheduled at 3-day intervals over a period of 2 weeks (Fig. 1). All researchers performing the tests were blinded to the groups.

All interventions were approved by the Research Ethics Committee of the Faculty of Medicine at the University of Antioquia (minutes 013 of September 26, 2013) and the Research Ethics Committee at IPS-Universitaria (minutes 097 of June 22, 2016) in Medellin (Colombia), in accordance with the Ethical Principles for Medical Research Involving Human Subjects outlined in the Declaration of Helsinki in 1975 and Resolution no. 8430 by the National Ministry of Health of Colombia issued in 1993. All participants gave written informed consent.

We used G*Power v3.1.9.7 (University of Düsseldorf, Düsseldorf, Germany) to calculate the sample size. Assuming a difference in musclin of 20% between groups, with a standard deviation (SD) of 25% within each group, an allocation ratio of 2, an α of 0.05, and a power of 85%, we needed to enroll 27 subjects without and 55 with MS.

Data were analyzed using IBM SPSS software version 21.0 (IBM Corp., Armonk, NY, USA) and graphed in R packages (BoxPlotR and Corrplot). The Shapiro-Wilk test was performed to evaluate if data were normally distributed. Continuous variables were summarized using mean±SD for normally distributed data and median (interquartile range [IQR]) for variables with a skewed distribution. Percentages were used for categorical variables. The Mann-Whitney U or Student’s t test was used to compare groups with and without MS, according to the data distribution. A logarithmic transformation was performed for musclin and HOMA-IR, and multiple linear regression models evaluating the relationship between the natural logarithm (ln) of musclin and ln(HOMA-IR) or VO_2peak, adjusted for age, sex, fat percentage, total lean mass, and physical activity, were carried out. Pearson correlation coefficients were used to evaluate the relationship among ln(musclin) and other variables. In all analyses, a two-sided P<0.05 was considered to indicate statistical significance.

We included a total of 90 subjects, of whom 29 were controls and 61 had MS (Table 1). The groups were comparable across age, sex, indexes of lean mass, blood pressure, VO_2peak, smoking, alcohol consumption, and weekly physical activity levels (P>0.05). Although the participants with MS had higher carnosine content, and therefore a higher estimated area of type II fibers, the difference was not significant. The group with MS had higher BMI, VAT, waist circumference, triglyceridemia, glycemia, insulinemia, and HOMA-IR values (P<0.05).

The subjects with MS had higher median serum musclin values than the controls (206.7 pg/mL [IQR, 122.7 to 387.8] vs. 111.1 pg/mL [IQR, 63.2 to 218.5], P<0.05), as shown in Fig. 2.

This figure also shows that both groups partially overlapped at values around 50 to 300 pg/mL, but there was a subgroup of individuals with MS who showed serum musclin concentrations over 300 pg/mL. In order to better characterize this high-musclin subgroup, the whole sample was divided into tertiles according to the musclin concentrations and several variables were compared (Table 2). The subjects with higher musclin were those with higher HOMA-IR, visceral adiposity, and lean mass, but a lower body fat percentage. There were no differences regarding BMI, VO_2peak, or level of physical activity.

A significant correlation was found between ln(musclin) and ln(HOMA-IR) in a multiple linear regression model (β=0.228; 95% confidence interval [CI], 0.121 to 0.334; P<0.01) (Table 3) adjusted for age, sex, adiposity, lean mass, and level of physical activity.

There was no correlation between ln(musclin) and VO_2peak in a multiple linear regression model (β=−0.287; 95% CI, −1.337 to 0.762; P>0.05) (Table 4) adjusted for age, sex, adiposity, lean mass, and level of physical activity. The association was not significant even when using the VO_2peak corrected for total lean mass (VO_2peak/kg).

We constructed two Pearson correlation matrices to further explore the relationship between musclin and metabolic (Fig. 3) or cardiorespiratory fitness (Fig. 4) variables. In agreement with the results shown in Tables 2, 3, ln(musclin) showed significant positive correlations with ln(HOMA-IR), ln(insulin), glycemia, waist circumference, VAT, and thigh lean mass/total body mass (P>0.05). Ln(musclin) did not correlate with the BMI or the area of type II fibers in the VLM. There were significant negative correlations between ln(musclin) and fat percentage and fat mass/stature squared index (P>0.05) (Fig. 3).

Since the lack of correlation between musclin and BMI was noteworthy, we further explored the relationship between tertiles of BMI with cardiorespiratory capacity variables and serum musclin. The results confirmed that the tertile with the most obese subjects showed the lowest VO_2peak and the lowest maximum time of work, but did not show musclin concentrations different from those of the other tertiles (P>0.05) (not shown).

In accordance with the above results, there were no correlations between ln(musclin) and VO_2peak, VO_2peak corrected for body mass or lean mass, VCO_2peak, maximum HR, maximum velocity, maximum time of work, RER, or recovery HR (P>0.05) (Fig. 4). There was also no correlation between ln(musclin) and weekly physical activity levels (not shown).

Muscle mass, muscle composition, and endocrine function have clear implications for the understanding of the pathophysiology of chronic metabolic diseases such as MS [13,14]. In terms of endocrine function, the physiological role of musclin in mammals has been the subject of controversy. Moreover, the study of the function of musclin in humans has lagged significantly when compared to studies in murine models.

Our results showed that subjects with MS had higher values of serum musclin. Furthermore, musclin showed positive correlations with IR and VAT, but negative correlations with indicators of total body fat. Total adiposity, as indicated by fat mass and fat percentage measurements, mainly (approximately 80%) reflects SAT. Although higher total adiposity is related to metabolic impairment, VAT is the adipose compartment that is more pathogenic and more strongly related to increased risks of MS and IR [45–49]. This observation was confirmed in our population, in which we observed a positive association between HOMA-IR and VAT (r=0.58, P<0.05). Then, not only are VAT and SAT anatomically separated, but most importantly, they are developmentally and functionally distinct, which gives them different implications for the development of metabolic diseases [45–49]. For instance, the removal of VAT in murine models of obesity reversed IR and modified the profile of serum adipokines [50]. Moreover, mice that received a transplant of SAT into the VAT area improved insulin sensitivity [51], clearly demonstrating a different metabolic profile between VAT and SAT. The differences are further confirmed because VAT has more potential for inflammation (mainly by producing more tumor necrosis factor alpha, interleukin 6, and monocyte chemoattractant protein-1), which reduces insulin sensitivity, and is also the main source of resistin, an adipokine proposed to favor IR and boost inflammation [46]. VAT has also long been shown to elicit a higher turnover of palmitate [52], which is an important marker and inducer of IR in cellular models [53]. Subjects with MS and IR have greater total free fatty acids and palmitate than obese subjects without IR [54,55]. Instead, SAT is the main source of adiponectin, an insulin-sensitizing, anti-inflammatory adipokine [46]. In subjects with high trunk fat (mainly VAT) and IR, those who also had large SAT values had higher adiponectin and lower inflammatory markers than those with less SAT [48]. This confirms that, in contrast to VAT, large amounts of SAT have favorable effects on adiponectin, inflammatory markers, IR, and MS risk.

Therefore, the subjects in our study with high SAT (i.e., high total fat mass and fat percentage), irrespective of their VAT, were expected to have higher adiponectin, as well as lower palmitate and inflammatory markers, a profile associated with less IR. In this manner, the inverse association between musclin and total fat can be explained if adiponectin inhibits the expression and secretion of musclin. Conversely, subjects with high musclin and high VAT, which in our study were those with MS, IR, and low SAT, were expected to have lower adiponectin levels but higher levels of resistin, palmitate and inflammatory markers. Thus, the fact that musclin was positively associated with VAT and IR can be explained if resistin, palmitate, and inflammatory markers (either in isolation or through the concerted action of all these molecules) favor musclin expression and secretion.

In agreement with this hypothesis, it was recently shown that palmitate activates the phospho-protein kinase RNA-like ER kinase (PERK)/activating transcription factor 4 (ATF4)/CCAAT/enhancer-binding protein (C/EBP) signaling pathway in myotubes to increase the transcription of musclin [18], which may then appear in serum, as serum musclin concentrations seem to follow musclin intramuscular expression [19,20]. The specific relationship between musclin and adipokines, other free fatty acids, and inflammatory markers is an avenue that should be addressed in future basic and clinical studies.

In agreement with the above discussion, we found that musclin correlated with IR after adjusting for age, sex, fat percentage, and lean mass. This observation agrees with a recent study in obese people [15]. However, in that paper, the authors did not assess body composition, lean mass, or muscle composition in detail. Moreover, the lean patients in that study had a HOMA-IR over 3, which in our population corresponded to IR.

In the present study, we found that musclin positively correlated with HOMA-IR and visceral adiposity in participants who were clearly separated as with or without MS and IR (HOMA-IR over or below the cutoff value of 2.25, respectively), and this relationship remained consistent after adjusting for variables associated with body composition.

The lack of a correlation between musclin and muscle composition and the low magnitude of the correlation between musclin and muscle mass suggest that several circulating musclin isoforms might exist. Although skeletal muscle is the main source of musclin [11], this myokine is also produced by the bones [56] and the heart [23]. Even though it is not currently possible to distinguish the specific contribution of any of the isoforms, if they actually exist, we propose that our results mainly reflect muscle-derived musclin.

The results showed a positive correlation between musclin and insulinemia, in agreement with earlier experiments conducted in myotubes [11], which suggested that insulin is required for musclin to induce IR. Thus, it seems that in humans, the increased insulinemia associated with the presence of IR drives positive feedback that further favors the IR-inducing role of musclin.

In contrast to what could be extrapolated from some recent studies in mice, we did not find any association between musclin and variables related to cardiorespiratory capacity in untrained humans. Some data suggesting that musclin is a physical endurance enhancer were obtained in untrained musclin knock-out (KO) mice [21]. Although those mice showed major bone and muscle parameters similar to the wild type, we speculate that musclin KO may have unnoticedly affected one or more cardiovascular, respiratory, musculoskeletal, or neurological molecular and epigenetic determinants of physical performance in mammals (including mice). Those determinants include, for instance, adrenergic signaling, calcium signaling, hemoglobin, myoglobin, lung reserve, alpha-actinin, mitochondrial function, cardiac contractile function, and capillary density [57]. Their single or combined alterations may explain possible differences in physical performance between wild-type and KO mice with healthy musculoskeletal development, and may become evident only under exercise or stress conditions, or through more in-depth molecular studies.

In agreement with this, the musclin KO mice showed exacerbated muscle atrophy only when bearing cancer, indicating that several other regulators may be affected in this mouse model [22]. The existence of as yet unknown musclin isoforms, short forms, or several receptors for one or more isoforms, would also explain the dual function of this protein. In any case, the results obtained from musclin KO mice should be interpreted with caution until more knowledge about musclin-activated signaling pathways is gained and those animals are better characterized.

We calculated a sample size with an allocation ratio of MS to control subjects of 2 instead of 1, because control subjects have narrower ranges of musclin and HOMA-IR, which limits the power of the multiple and simple linear regression models. The subjects of both study groups were enrolled at the same time by the same researchers, and were evaluated by blinded researchers. Since the sample size was larger than required and the difference in musclin concentrations between groups was higher in magnitude than expected, we could estimate an actual study power of 96%. There are also at least three reasons to argue against the possibility that the lack of correlations between musclin and cardiorespiratory capacity variables arose due to insufficient power: (1) the multiple regression model had a very high P value, and even after adjustments the 95% confidence interval remained centered on zero; (2) none of the correlations showed a trend, which further suggests that type I errors were not present; and (3) several known correlations reported in Fig. 4 were significant, and the most obese subjects showed the worst cardiorespiratory capacity as expected, serving as quality controls and demonstrating that the sample was adequate. Thus, our good methodological design allowed us to test our hypotheses robustly.

About 3% of participants with MS were on metformin. Since metformin reduces insulin, HOMA-IR, and IR [58], we speculate that metformin might have prevented the serum musclin in the MS group from being even higher and attenuated the correlations of musclin with insulin and HOMA-IR. Thus, the increased musclin values found in the subjects with MS are not explained by any known effect of metformin.

Since aging reduces the number, size and satellite cells of type II fibers [59], which are the main producers of musclin in mammals [16], we speculate that musclin may be reduced in older people. However, the effect of aging on musclin is not expected to affect our results because we: (1) enrolled only adults, not elderly subjects; (2) adjusted our regression analyses for age; (3) and matched groups with and without MS for age.

We provide evidence to show that the role of musclin in humans is in line with that shown in cell cultures and wild-type mice—that is, musclin is likely involved in the regulation of glucose metabolism. Our results indicate that insulin, visceral and total adiposity, and muscle mass, but not BMI, fiber type, or physical fitness, are probably novel regulators of musclin in humans. The data supports a recent model proposed to integrate information on the role of skeletal muscle mass and its endocrine function in the pathophysiology of MS [13].

The evidence available up to now also allows us to propose musclin as a target to be regulated in patients with MS and IR. New public health policies should always consider and account for the role of the skeletal muscle as an important player in metabolic health issues.

In humans, musclin is positively associated with IR and with a body composition profile characterized by high visceral adiposity and lean mass, but low body fat percentage. This myokine is not related to BMI, cardiorespiratory fitness, or physical activity levels. The fact that most data obtained with cultured cells, wild-type murine models, and now in humans supports this conclusion suggests that the results obtained in KO mice should be interpreted carefully.