METRNL decreases during adipogenesis and inhibits adipocyte differentiation leading to adipocyte hypertrophy in humans

Abstract

Background:

Meteorin-like (METRNL) is a recently described circulating protein shown to be highly expressed in white adipose tissue and to beneficially affect energy metabolism in mice.

Objective:

We systematically evaluated the role of METRNL for human adipogenesis and its association with obesity, browning and hyperinsulinemia in children. In addition, we assessed the functional relevance of METRNL for human adipogenesis.

Results:

METRNL expression decreased during human adipocyte differentiation in vitro. Coherently, METRNL expression was lower in isolated adipocytes compared with stromal vascular fraction (SVF) cells in human samples. Withdrawal of the peroxisome proliferator-activated receptor-γ (PPARγ) agonist rosiglitazone from adipogenic media partially preserved the METRNL downregulation during adipogenesis. METRNL expression was higher in adipocytes of obese compared with lean children and correlated with adipocyte size, whereas in SVF METRNL expression correlated with proliferation capacity. Concordantly, overexpression of METRNL inhibited human adipocyte differentiation as shown by decreased lipogenesis and lower expression of PPARγ and markers of adipogenesis, whereas experimental downregulation promoted adipogenesis. Proliferation, in contrast, was advanced by METRNL overexpression. These interactions with adipose tissue dynamics may contribute to the clinically observed body mass index-independent association of METRNL expression with hyperinsulinemia and adipose tissue inflammation in human samples. METRNL was not associated with UCP1 expression or induction of browning in white adipocytes.

Conclusions:

Taken together, the downregulation of METRNL during adipogenesis and functional induction of increased proliferation in SVF cells with concomitant inhibition of adipocyte differentiation may result in hypertrophic AT accumulation. This may also explain our observations of increased METRNL expression in adipocytes but not SVF cells in obese children compared with lean children and the subsequent hyperinsulinemia.

Introduction

Meteorin-like (METRNL), previously referred to as subfatin, Il38 or cometin,¹ is a recently described circulating protein, which is highly expressed in white adipose tissue (AT) and which beneficially affects energy metabolism in mice.^{2, 3}

The expression of METRNL was shown to (transiently) increase in response to metabolic challenges such as resistance exercise training and acute cold exposure² and to decrease after caloric restriction in mice.⁴ This raised the hypothesis that Metrnl may act as an adapting factor in the regulation of energy homeostasis.

Concordant with this theory, circulating Metrnl was shown to increase energy expenditure and to induce browning of white AT in mice indirectly through alternative activation of macrophages within the AT.² Furthermore, Metrnl improved glucose tolerance by enhancing insulin sensitivity^{2, 3} and suppressed chronic inflammation in mice.²

This 'overall lean phenotype' had at least partly been attributed to effects of Metrnl on AT biology in mice,^{2, 3} such as peroxisome proliferator-activated receptor-γ (PPARγ)-mediated induction of adipocyte differentiation³ or anti-inflammatory effects.^{2, 3, 4} On the other hand, Metrnl was shown to be upregulated in adipogenesis and high-fat diet induced obesity in rodents.^{2, 3}

Hence, Metrnl may act as an adaptive factor affecting AT composition under metabolic challenges in mice. For humans, the situation is less clear. METRNL was shown to be abundantly expressed in human white AT,⁴ although no correlation of the circulating protein was found with body mass index (BMI) in humans.³

Here we aimed to evaluate systematically METRNL expression during human adipogenesis and whether METRNL expression in white AT is associated with obesity, browning and metabolic parameters in lean and obese children. We finally studied the effects of knockdown and overexpression of METRNL on human adipogenesis to assess the relevance of METRNL for humans.

Materials and methods

Subjects and samples

We obtained 93 subcutaneous AT samples from lean and obese children with detailed anthropometric, clinical and metabolic assessments (Supplementary Table 1) of the previously described Leipzig Adipose Tissue Childhood cohort.⁵ Written informed consent was obtained from all parents. The study was approved by the local ethics committee (Reg. No.: 265-08, 265-08-ff; NCT02208141). AT samples were collected during elective surgery including orthopedic surgery, herniotomy, orchidopexy, abdominal surgery, thoracic surgery and back surgery, and weighed 0.04 to 16.4 g. Biological characterization of AT samples regarding macrophage infiltration, adipocyte cell size, proliferation capacity or UCP1 expression was described in detail previously.^{5, 6}

Briefly, adipocyte diameter and number was determined after osmium fixation using a Coulter counter (Multisizer III; Beckmann Coulter, Krefeld, Germany), and total adipocyte number was estimated by dividing adipocyte number per g sample by total body AT mass. Stromal vascular fraction (SVF) cells were directly seeded for proliferation or differentiation analyses. Proliferation was assessed by counting Hoechst-stained nuclei at days 2, 4, 6, 8 and 10 after seeding. Adipogenic differentiation was performed according to the Poietics Human Adipose-Derived Stem Cell Adipogenesis Protocol (Lonza, Cologne, Germany), and efficiency is given as % of Nile Red/Hoechst double-stained cells from the total number of Hoechst-positive cells.

Macrophage infiltration into AT was analyzed by immunohistochemical staining of AT sections with a monoclonal CD68 antibody (M0718; Dako, Glostrup, Denmark) using the Dako REAL APAAP Immunocomplex system, and is given as the number of macrophages per 100 adipocytes.

Cell culture, differentiation and stimulation of human preadipocytes and adipocytes

We applied the Simpson–Golabi–Behmel syndrome (SGBS) human adipocytes cell model (kindly provided by M Wabitsch (Ulm, Germany)),⁷ which represents a widely applied model of human adipogenesis of non-immortalized, non-neoplastic cells with spontaneously retained proliferation and differentiation capacity.⁸ Preadipocytes were cultured in DMEM/Ham F12 medium with 10% fetal calf serum (Life Technologies, Karlsruhe, Germany) supplemented with 33 μm biotin and 17 μm pantothenic acid. Cells were differentiated into adipocytes as described previously⁹ under serum-free conditions by supplementing basal medium with 20 nm insulin, 0.2 nm triiodothyronine, 100 nm hydrocortisone and 0.13 nm apotransferrin, 2 μm rosiglitazone, 25 nm dexamethasone and 500 μM 3-isobutyl-1-methylxanthine. After 4 days, differentiation was continued with differentiation medium without rosiglitazone, dexamethasone and 3-isobutyl-1-methylxanthine. Cells were harvested at day 0, day 4, day 8 and day 12 post induction.

For trans-differentiation, SGBS preadipocytes were cultured and treated as described above. In addition to supplements used for standard differentiation, 8.3 nm bone morphogenetic protein 7 (BMP7) or 2 μm rosiglitazone were added during complete course of differentiation according to Tseng et al.¹⁰

Stimulation experiments were performed in SGBS preadipocytes and in adipocytes at day 10 post induction. Confluent preadipocytes were serum starved for 24 h and then cultivated in SGBS basal medium supplemented with either 100 nm insulin, 100 nm dexamethasone, 100 nm insulin growth factor-1 or 10 μm isoproterenol for additional 24 h. For stimulation experiments in adipocytes, the adipocyte cell medium was directly replaced by basal SGBS medium containing respective stimulants (see above) for 24 h. Untreated cells incubated in basal SGBS medium for 24 h served as control.

To identify critical components for the METRNL expression, insulin, triiodothyronine, hydrocortisone, apotransferrin, rosiglitazone, dexamethasone or 3-isobutyl-1-methylxanthine were selectively withdrawn from the differentiation medium for the complete period of differentiation. Cells treated with complete differentiation medium served as control. Cells marked with d0 represent cells in the preadipocyte state.

For quantification of triglyceride accumulation, adipocytes at day 12 post induction were fixed in Roti-Histofix 4% (Carl Roth, Karlsruhe, Germany), washed with phosphate-buffered saline and stained with Oil Red-O working solution (0.3% in 60% isopropanol) for 15 min. After repeated washes with water, Oil Red-O was extracted by incubation with isopropanol, and quantified at 540 nm using the FLUORstar OPTIMA (BMG Labtech, Offenburg, Germany).

Cell proliferation of SGBS cells was measured using the commercial Cell Proliferation Reagent WST-1 (Roche, Grenzach-Wyhlen, Germany). According to the manufacturer’s protocol absorbance at 450 nm was detected.

Ectopic METRNL expression and siRNA-mediated knockdown

SGBS preadipocytes were transfected using the Neon Transfection System 100 μl Kit (Invitrogen, Carlsbad, CA) as described previously.¹¹ siRNAs and ON-TARGETplus control reagents (Dharmacon, Lafayette, CO, USA) were used at a final concentration of 500 nm. METRNL expression vector was purchased from OriGene (Rockville, MD, USA) and 3.5 μg were used per transfection. After electroporation, 2 50 000 cells per well were seeded into 6-well format and differentiated as described above.

mRNA expression studies

Quantification of PPARG, TBP, UCP1, PRDM16, PAT2, P2RX5 and TMEM26 in children AT samples has been described previously.⁶ Primer and probe sequences are listed in Supplementary Table 2.

Total RNA was extracted using TRIzol (Ambion, Life Technologies, Carlsbad, CA, USA) according to the manufacturer's instructions. Total RNA of 500 ng were reverse transcribed using the miScript II RT Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. METRNL was quantified by real-time PCR with Maxima SYBR Green Master Mix (Thermo Scientific, Carlsbad, CA, USA) on the ABI 7500 Sequence Detection System (Applied Biosystems, Darmstadt, Germany). For standardization of gene expression target genes were normalized to the housekeeping gene TATA-box-binding protein (T BP).

Immunoblotting

Cells were lysed in 50 mm HEPES (pH 7.5), 150 mm NaCl, 10 mm EDTA, 1% Triton X-100 and Roche complete. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis separated proteins were transferred to a polyvinylidene fluoride membrane and detected by chemiluminescence. Antibodies against METRNL (ab177740) and β-actin (A5316) were purchased from Abcam (Cambridge, UK) and Sigma (Taufkirchen, Germany), respectively.

Statistical analyses

Data are presented as means±s.e.m. of at least three independent cell culture experiments, each performed in triplicates. Normal distribution of the data was assessed using the Kolmogorov–Smirnov test and by quantile–quantile plots. Non-normally distributed data were log-transformed (log) before analysis. For comparison of quantitative traits between two groups, t-test was used. Correlation analyses were performed using Pearson's correlation analysis or partial correlation analysis as indicated. Expression during adipogenesis was analyzed applying one-way analysis of variance (ANOVA) with repeated measurements and Dunnett’s post-test.

For multiple regression analyses, the stepwise forward model was used. Independent variables were included in the regression model if a significant contribution to the model was confirmed by applying the F-test. Threshold for statistical significance was P<0.05. Statistical analyses were performed using the Statistica 7.1 software package (StatSoft, Tulsa, OK, USA) and GraphPadPrism5 (GraphPad Software, La Jolla, CA, USA).

Results

METRNL expression decreases during human adipogenesis

Metrnl was shown to be upregulated during differentiation of mouse 3T3-L1 preadipocytes.⁴ We first evaluated whether METRNL expression is modulated during human adipogenesis. During differentiation of human preadipocytes of the SGBS cell line, METRNL expression decreased on mRNA level (to 38.5±1.5%) as well as on protein level (Figure 1a). The 14-fold increase of PPARγ expression confirmed an effective adipocyte differentiation. This downregulation of METRNL was confirmed by lower expression of METRNL after differentiation of primary human stroma-vascular cells into adipocytes for 8 days (Figure 1b), which consistently occurred in seven of eight samples.

Figure 1

METRNL expression during human adipogenesis. METRNL expression decreases during in vitro differentiation of SGBS preadipocytes to adipocytes on mRNA and protein level (a). An expected increase in mRNA expression of PPARγ confirmed efficient adipogenesis. Immunoblot of SGBS cell lysates of the indicated time-points during adipogenesis show a distinct band at around 35 kDa representing METRNL (marked by arrow) and confirmed downregulation on protein level. Untreated cells served as control (ctr.) and an immunoblot of β-actin served as a loading control. Expression data are shown for at least three independent cell experiments. Expression in preadipocytes was set 1. (b) METRNL was downregulated at day 8 of adipogenesis in primary subcutaneous preadipocytes of children. Data are shown as mean of eight individual primary human preadipocyte samples differentiated into adipocytes. (c) METRNL expression is higher in the SVF fraction compared with isolated adipocytes in children. Data are shown as mean±s.e.m. of 93 individual biopsy samples. Statistical significance was assessed by t-test. (d) Withdrawal of adipogenic agents: SGBS preadipocytes were differentiated into mature adipocytes with adipogenic media lacking indicated adipogenic ingredients. METRNL expression was partly preserved in cells treated with differentiation medium lacking rosiglitazone. Expression in mature adipocytes differentiated with complete medium (c) was set=1. (e) Stimulation with metabolic regulators: METRNL expression is downregulated by dexamethasone in preadipocytes. Preadipocytes and adipocytes (day 10 of differentiation) were stimulated with 100 nm insulin (Ins), 100 nm dexamethasone (Dex), 100 nm insulin growth factor-1 (IGF-1) and 100 nm isoproterenol (Iso) for 24 h. Expression in untreated cells (C) was set=1. Data are shown as mean±s.e.m. for three independent cell experiments. Statistical significance was assessed by one-way ANOVA and post hoc Dunnett’s test and was compared with the control sample.

To further validate this finding, we compared the METRNL expression pattern between isolated adipocytes and the SVF purified from white adipose tissue samples of 93 children available from the Leipzig Childhood Adipose Tissue cohort (Supplementary Table 1). METRNL expression was threefold higher in the SVF fraction (Figure 1c) compared with adipocytes consistent with a downregulation of METRNL during human adipogenesis.

We subsequently assessed whether the downregulation of METRNL during experimental adipocyte differentiation was merely due to direct effects of the adipogenic inducers by withdrawing single components from the differentiation medium for the complete time of differentiation. Rosiglitazone depletion resulted in a significant preservation of METRNL mRNA levels (Figure 1d), which was also accompanied by impaired adipocyte differentiation (data not shown). Stimulation of preadipocytes or differentiated adipocytes with metabolic factors did not affect METRNL expression except for downregulation after dexamethasone treatment in preadipocytes (Figure 1e).

METRNL is associated with hypertrophic AT accumulation and related hyperinsulinemia and AT inflammation in humans

Considering this dynamic regulation of METRNL during human adipogenesis and previous findings of upregulation of Metrnl in obese mice,⁴ we hypothesized that METRNL expression in AT may be altered with obesity and insulin resistance in children. To test this, we investigated the association of METRNL expression in adipocytes (Table 1) and SVFs (Table 2) and clinical measures of obesity and metabolic parameters.

Table 1 Correlation of METRNL expression in adipocytes with anthropometric, metabolic and cardiovascular parameters in 93 children

Table 2 Correlation of METRNL expression in SVF with anthropometric, metabolic and cardiovascular parameters in 93 children

METRNL expression in adipocytes (Figure 2a), but not in SVF (Figure 2b), was higher in obese children compared with lean children and positively correlated with BMI-SDS (standard deviation score) and age (Tables 1 and 2).

Figure 2

Impact of METRNL on AT accumulation and inflammation. (a) METRNL expression in adipocytes is significantly increased in obese compared with lean children. Expression in SVF is not significantly affected by obesity (b). Data are given as mean±s.e.m. Statistical significance was assessed by t-test. The expression of METRNL in adipocytes and SVF is positively correlated with serum insulin levels (c, d). METRNL expression in adipocytes is positively associated with adipocyte diameter (e). METRNL expression in SVF is negatively associated with preadipocyte doubling time, thus indicating a positive association with proliferation (f). Pearson's correlation coefficient r and P-value for the whole cohort, as well as stratified for lean and obese are shown in each scatter plot.

Similarly, we observed a positive correlation between METRNL expression in both adipocytes and SVFs with fasting insulin (Figures 2c and d), and in adipocytes with CD68 expression (Table 1), which were independent from sex, age and BMI-SDS. In subsequent multiple regression analyses, we confirmed METRNL expression in adipocytes (Supplementary Table 3) and in SVF (Supplementary Table 4) to be independently associated with insulin. Hence, METRNL expression appears to be associated with obesity and furthermore with fasting hyperinsulinemia and AT inflammation independent of BMI-SDS in children.

AT accumulation in children results from hypertrophy and hyperplasia and particularly adipocyte hypertrophy was related to insulin resistance and AT inflammation.⁵ We were, therefore, interested whether proliferation and/or adipocyte size were related to METRNL expression. We analyzed potential correlations between METRNL expression in adipocytes (Table 1) and SVFs (Table 2) and parameters of adipose tissue proliferation, as well as differentiation capacity. Adipocyte diameter correlated positively with METRNL expression (Figure 2e), whereas we observed a negative correlation between METRNL expression and doubling time in SVF cells (Figure 2f), which withstood adjustment for age, BMI-SDS and pubertal stage.

These findings indicate that METRNL expression is positively associated with preadipocyte proliferation and adipocyte hypertrophy, as well as fasting insulinemia in humans.

METRNL inhibits human adipogenesis

To finally assess whether the association of METRNL with AT accumulation, hypertrophy and proliferation is just an epiphenomenon or whether METRNL may be functionally involved in obesity-associated AT alterations, we studied the effect of METRNL overexpression and knockdown in human adipocytes.

We verified overexpression of METRNL protein and mRNA after transfection of SGBS preadipocytes (Figure 3a). METRNL overexpression inhibited adipocyte differentiation as shown by decreased lipid accumulation at day 12 of adipogenesis (Figure 3b). Furthermore, overexpression of METRNL significantly decreased the expression of PPARγ, master-regulator of adipogenesis, and the expression of subsequent adipogenic genes such as FABP4 (Figures 3c and d).

Figure 3

Impact of METRNL on human adipogenesis. Transfection of SGBS preadipocytes with a METRNL expression vector significantly increased METRNL until d4 of differentiation procedure (a). Immunoblot of undifferentiated SGBS cell lysates transfected with a METRNL expression vector at indicated concentrations—20 μg cell lysate were analyzed—a distinct band at around 35 kDa representing METRNL. An immunoblot of β-actin serves as a loading control. Transfection of the METRNL expression vector significantly decreased lipid accumulation at d12 (b) and significantly inhibited PPARG (c) and FABP4 (d) expression during SGBS in vitro adipogenesis. Transfection of SGBS preadipocytes with siMETRNL oligos significantly decreased METRNL expression (f) and increased lipid accumulation at d12 (g). PPARG (h) and FABP4 (i) expression is increased after METRNL knockdown. PPARG knockdown decreased lipid accumulation as well as PPARG and FABP4 expression and served as a control. SGBS preadipocytes overexpressing METRNL exhibit a higher proliferation rate than control cells transfected with empty vector (e). Knockdown of METRNL in preadipocytes has no effect on proliferation rate (k). Data are given as mean±s.e.m. and are shown for at least two independent cell experiments. Result in preadipocytes (d0) was set 1 in (c, d, h and i), Oil Red-O staining was normalized to siControl- or empty vector-treated cells. For WST-1 assay, result in preadipocytes (d0) was set 1. Statistical significance was assessed by two-way ANOVA and Dunnett’s multiple comparisons test. Samples were compared with empty vector- or siControl-treated samples and marked with *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001. Data are given as mean±s.e.m. and are shown for at least two independent cell experiments.

Conversely, knockdown of METRNL expression using RNA interference (Figure 3f) slightly increased the adipogenic capacity of SGBS cells as shown by increased lipid accumulation (Figure 3g) and increased PPARγ and FABP4 expression (Figures 3h and i). The magnitude of these effects was, however, smaller compared with the downregulation of the master-regulator of adipogenesis PPARγ (Figures 3f–i). Furthermore, METRNL expression was less reduced during adipogenesis after PPARg knockdown, compared with control transfected cells (Figure 3f).

Hence, overexpression of METRNL inhibited human adipocyte differentiation, whereas downregulation promoted adipogenesis. Proliferation, in contrast, was advanced by METRNL overexpression, but was unaffected by attenuated METRNL (Figures 3e and k).

METRNL is not induced by browning of white adipocytes in humans

As Metrnl was shown to induce the browning of adipose tissue,² we evaluated the relationship of adipocyte METRNL with brown/beige-selective marker gene expression in human AT samples. We did not find any correlation with UCP1 (r=0.034, P>0.5), PRDM16 (r=0.032, P>0.5), PAT2 (r=0.201, P>0.5), P2RX5 (r=0.194, P>0.5) or TMEM26 (r=−0.099, P>0.5). In addition, we trans-differentiated SGBS cells into adipocytes with a brown-like phenotype by treatment with high rosiglitazone or BMP7. Rosiglitazone increased UCP1 expression 22.7-fold (±14.2) and BMP7 4.8-fold (±0.2), indicating successful induction of a brown-like phenotype (Supplementary Figure S1A). However, METRNL expression was not affected by rosiglitazone and was rather reduced by BMP7 to 71±0.002%, indicating an inhibitory role of BMP7 (Supplementary Figure S1B).

Discussion

We show that METRNL is downregulated during human adipocyte differentiation. Compared with lean children, METRNL expression is higher in adipocytes of obese children and correlates with the adipocyte size as an indicator of hypertrophy, whereas in the SVF higher METRNL expression correlates with the proliferation capacity. On the functional level, overexpression of METRNL inhibited human adipocyte differentiation, while downregulation promoted adipogenesis. Proliferation, in contrast, was advanced by METRNL overexpression. These interactions of METRNL with adipose tissue dynamics may contribute to the clinically observed association of METRNL with hypertrophic AT accumulation and ensuing hyperinsulinemia and AT inflammation (Figure 4).

Figure 4

Schematic overview of results and conclusions. METRNL decreases during human adipocyte differentiation. Even though it promotes the proliferation of SVF cells, the functional repression of human adipogenesis (with inhibited formation of new adipocytes) would lead to a hypertrophic state of adipose tissue with subsequent association with hyperinsulinemia as a sign of insulin resistance and adipose tissue inflammation.

Our finding of downregulation of METRNL expression during human adipogenesis is substantiated by (i) decreasing the expression of METRNL during adipocyte differentiation in the SGBS cell line, (ii) in primary human adipocytes and (iii) by lower expression in isolated adipocytes compared with SVF cells in human AT samples. These first findings in the human background contradict findings of a recent study performed in mouse 3T3-L1 cells⁴ showing a >20-fold increase in Metrnl expression at day 16 of differentiation after a first transient decrease at the beginning of differentiation. The authors argued that adipogenic compounds such as 3-isobutyl-1-methylxanthine, insulin and dexamethasone inhibited Metrnl expression, and their removal from differentiation medium at day 2 subsequently facilitated Metrnl expression.⁴ In our human cell models, dexamethasone slightly suppressed METRNL expression, but the cAMP agonist or insulin did not affect METRNL expression. However, when the PPARγ agonist rosiglitazone was absent in the adipogenic medium, the downregulation of METRNL was blunted, which may indicate a PPARγ-dependent effect.

In contrast to Li et al.,⁴ we found a clearly higher expression of METRNL in SVF compared with adipocytes in humans. Owing to the small sample volume of children’s biopsy samples, we were not able to separate the different SVF sub-populations, typically consisting of preadipocytes,¹² endothelial and immune cells such as macrophages and lymphocytes.¹³ In particular, the immune cells would be of interest as METRNL expression has been shown in activated monocytes and PBMCs.¹⁴ The link to inflammation is further supported by observations of increased METRNL in some immune diseases such as psoriasis and rheumatoid arthritis.¹⁴ Inflammation of adipose tissue is a pathological factor contributing to the remodeling of adipose tissue in obesity.¹⁵ One could hence speculate that higher amounts of macrophages within the SVF directly or indirectly stimulate METRNL expression in adipocytes or that high METRNL expression in adipocytes leads to an increased macrophage infiltration of adipose tissue. It also needs to be considered that macrophages may be an additional, although minor source of METRNL.⁴ Interestingly, we did not find a correlation between METRNL expression in adipocytes or SVF with macrophage infiltration seen histologically. However, CD68 expression in SVF, a marker for macrophages, was the strongest independent predictor for METRNL expression in adipocytes.

From the finding of downregulation of METRNL expression during adipogenesis one may expect lower METRNL expression in AT of obese subjects. We observed, however, a higher METRNL expression exclusively in adipocytes, but not in the SVF, of obese compared with lean subjects. This is in line with mice studies showing increased expression in obese mice. The higher expression in adipocytes of obese children is likely to be due to adipocyte hypertrophy.⁵ This direct relationship would further be supported by observations in mice where adipocyte overexpression of Metrnl led to increased adipocyte size.³ From this, one may conclude that METRNL expression is associated with hypertrophic adipose tissue accumulation. Interestingly, obese conditions and adipocyte hypertrophy have also been associated with a decrease in the activity and amount of PPARG, which is implicated in the pathogenesis of the metabolic syndrome.¹⁶ Considering that particularly the adipocyte hypertrophy is a major and independent predictor for hyperinsulinemia and insulin resistance in human and in children,⁵ this may also explain the association of METRNL expression with increasing fasting insulin levels seen in our cohort.

On the other hand, we found higher METRNL expression in the SVF to be associated with a faster proliferation of these cells. In addition, higher METRNL levels in SVF generated by overexpression resulted in facilitated proliferation of these cells. Hence, METRNL expression in SVFs may mediate an increased proliferation of preadipocytes. This would also fit to the finding of downregulation of METRNL once the adipogenic program is induced and proliferation no longer occurs.¹⁷

In this regard, it is furthermore of interest as to how METRNL functionally affects human adipogenesis. Our results from overexpression and knockdown studies coherently show an inhibiting effect of METRNL on human adipocyte differentiation, which may at least partially be mediated by affecting PPARγ expression and subsequent downstream targets. Again, there may be species-specific differences concerning the effect of METRNL on adipogenesis, as in rodents METRNL has been found to induce adipogenesis via a PPARγ-dependent pathway and to affect beneficially glucose metabolism.³ The species homology of the METRNL molecule is 77% and for example differs in the presence of N-glycosylation in mice, which does not exist in human.¹ Also, the experimental approach (murine vs human cell lines with distinct differentiation protocols, experimental animal models) may involve distinct cofactors that modulate adipogenesis in combination with METRNL.

Taken together, our findings of METRNL causing increased proliferation in SVF cells with concomitant inhibition of adipocyte differentiation in humans would result in hypertrophic AT accumulation through storage of lipids in existing adipocytes, while the formation of new adipocytes is disturbed in the presence of high METRNL. As obese children have larger adipocytes,⁵ this may also explain the higher levels of METRNL in adipocytes but not SVF cells in obese children compared with lean children, and also the subsequent hyperinsulinemia.

Previous studies suggested that METRNL has a role in thermogenesis and browning of AT. METRNL is induced by metabolic challenges such as exercise and cold exposure. The circulating Metrnl does then stimulate browning of WAT and thermogenesis in mice by inducing the release of certain immune cytokines from AT macrophages, which subsequently activate expression of thermogenic genes.² Brown (beige) adipose tissue constitutes an important player in energy metabolism and has been shown to be present in humans after the new-born period in adults¹⁸ and children.⁶ We, therefore assessed whether the expression of METRNL is associated with the expression of brown/beige-selective marker genes including UCP1, PRDM16 and TMEM26. In contrast to previous studies, we did not find evidence for an association of brown/beige-selective marker gene expression with METRNL in adipose tissue samples of children. In those previous mouse studies, the described browning effect² was transient (between day 5 and 7 after METRNL injection) and we may have missed such a dynamic regulation as we have only 'snapshot' information about the current METRNL and BAT gene levels. However, when we trans-differentiated SGBS cells using rosiglitazone or BMP7, the induction of a brown-like phenotype as verified by an increase in UCP1 expression, did not lead to an induction of METRNL expression, indicating that AT METRNL is not associated with adipocyte browning. These findings are further underlined by our results showing that METRNL is not regulated in response to dexamethasone and that it reduces PPARG expression rather than activating it, which would be required for brown adipocyte formation. An explanation for this might be that METRNL acts indirectly on browning of adipocytes via alternative activation of macrophages as shown by Rao et al.,² which are not present in our cell model. Moreover, our analyses were restricted to intracellular METRNL. We did not analyze the relationship of muscle- or adipose tissue-derived circulating METRNL levels and parameters of adipocyte browning, which might be interesting as a METRNL-mediated muscle–fat crosstalk was suggested by Rao et al.² In this regard, we can also not exclude that muscle-derived METRNL mediates different effects than adipose tissue-derived METRNL.

Hence, with our experimental approach we could not support browning of AT as one of the previously described major functions of METRNL for humans. Regarding a role of METRNL in glucose and insulin metabolism, we did see an association of METRNL expression in AT with circulating insulin levels. However, as outlined above, the interpretation of our findings is a role of METRNL in hypertrophic AT accumulation and secondary clinical hyperinsulinemia.

Finally, so far three functions have been described for METRNL. It has a role in the cold adaptation by increasing the beige fat thermogenesis and it improves white adipose tissue function and affects insulin sensitization.¹ In the context of the nervous system, it acts a neurotropic factor supporting neuroblast migration and neurite outgrowth.¹⁹ The detailed function and signaling (including a yet unknown receptor) remain largely unknown.

One limitation of our study is the use of human SGBS cells as a somewhat ‘artificial’ in vitro cell culture model for the analyses of effects of downregulation or overexpression of METRNL on adipocyte differentiation and the effect of adipocyte browning on METRNL expression. It would be intriguing to reproduce our results in isolated preadipocytes of children. However, we were limited by the often small sample volumes obtained during elective surgery, which did not allow separation of SVF cell into distinct cell subsets or more detailed functional analyses.

Taken together, our findings of downregulation of METRNL during adipogenesis and functional effects of METRNL on increased proliferation in SVF cells with concomitant inhibition of adipocyte differentiation result in hypertrophic AT accumulation through storage of lipids in existing adipocytes. As obese children have larger adipocytes, this may also explain our observations of higher METRNL expression in adipocytes but not SVF cells in obese children compared with lean children and the subsequent hyperinsulinemia.