Loss of periostin occurs in aging adipose tissue of mice and its genetic ablation impairs adipose tissue lipid metabolism

Abstract Remodeling of the extracellular matrix is a key component of the metabolic adaptations of adipose tissue in response to dietary and physiological challenges. Disruption of its integrity is a well‐known aspect of adipose tissue dysfunction, for instance, during aging and obesity. Adipocyte regeneration from a tissue‐resident pool of mesenchymal stem cells is part of normal tissue homeostasis. Among the pathophysiological consequences of adipogenic stem cell aging, characteristic changes in the secretory phenotype, which includes matrix‐modifying proteins, have been described. Here, we show that the expression of the matricellular protein periostin, a component of the extracellular matrix produced and secreted by adipose tissue‐resident interstitial cells, is markedly decreased in aged brown and white adipose tissue depots. Using a mouse model, we demonstrate that the adaptation of adipose tissue to adrenergic stimulation and high‐fat diet feeding is impaired in animals with systemic ablation of the gene encoding for periostin. Our data suggest that loss of periostin attenuates lipid metabolism in adipose tissue, thus recapitulating one aspect of age‐related metabolic dysfunction. In human white adipose tissue, periostin expression showed an unexpected positive correlation with age of study participants. This correlation, however, was no longer evident after adjusting for BMI or plasma lipid and liver function biomarkers. These findings taken together suggest that age‐related alterations of the adipose tissue extracellular matrix may contribute to the development of metabolic disease by negatively affecting nutrient homeostasis.


| INTRODUCTION
Aging is associated with increased body weight gain (Chumlea et al., 2002). Overweight is a risk factor for the metabolic syndrome and promotes progression of different pathologies (Rosen & Spiegelman, 2014). White adipose tissue (WAT) stores energy as triglycerides, whereas brown adipose tissue (BAT) dissipates energy in the form of heat, and may thus protect against hypothermia. The extracellular matrix (ECM) provides a structural scaffold for adipose depots. It functions as a gatekeeper that regulates the exchange of adipokines and growth factors between fat cells and the circulation (Lu, Takai, Weaver, & Werb, 2011;Mariman & Wang, 2010). Aside from adipocytes, adipose tissue-resident mesenchymal stem cells are a main source of secreted signals. Their secretory profile, which includes matrix-modifying enzymes, changes significantly with increased age (Tchkonia et al., 2010). ECM remodeling during adipose tissue expansion is essential to allow sufficient de novo vascularization and control of adipocyte size (Lin & Kang, 2016). For instance, loss of collagen 6 increases expansion of adipocytes and improves metabolic features suggesting that excessive matrix accumulation is a pathological feature of obesity (Khan et al., 2009). The 90 kDa matricellular protein periostin is encoded by the Postn gene. It is secreted into the matrix space and can activate integrins (Itg), specifically heterodimers of Itgαv with Itgβ1, Itgβ3, or Itgβ5, which in turn regulate processes such as proliferation and differentiation (Idolazzi et al., 2017).
Here, we investigated the role of periostin in BAT and WAT function and its potential involvement in lipid metabolism during aging. While aging results in a marked downregulation of Postn expression, metabolic interventions that require adipose tissue remodeling induce Postn gene expression. In line with these observations, systemic deletion of Postn in mice resulted in loss of adipose tissue (AT) mass after cold exposure, reduced AT expansion and adipocyte size during high-fat diet (HFD) feeding, and impaired lipolytic enzyme activity after acute β3-adrenergic stimulation. Unexpectedly, expression of periostin mRNA in human WAT was positively associated with aging but inversely correlated with BMI. The age-related correlation could partially be explained by changes to BMI, as well as plasma lipid and liver biomarkers.

| Aging and metabolic interventions regulate the expression of periostin in adipose tissue
To test whether aging negatively affects the function of adipose tissue-resident stem/progenitor cells, we isolated adipogenic progenitors cells (APCs) from BAT and inguinal WAT (iWAT) by flow cytometry for microarray-based analysis (Schulz et al., 2011).
Aside from a number of genes belonging to an imprinted gene network that has been linked to adiposity (Morita et al., 2016), we found the gene encoding for periostin (Postn) to be the most strongly downregulated candidate gene (Figure 1a,b). The age-related reduction in Postn mRNA could be verified by real-time quantitative PCR (Figure 1c). Analysis of periostin protein levels confirmed the negative effect of aging in iWAT and BAT, but not in gonadal WAT (gWAT; Figure 1d, Supporting Information . To determine the distribution of periostin expression in adipose tissues, a transgenic mouse strain with β-galactosidase (lacZ) knock-in was used for histological analysis. The reporter is expressed from the Postn gene locus in lieu of the native protein.
These analyses revealed a marked reporter activity in cells adherent to blood vessels and in interstitial cells (Figure 1e). The predominant expression in nonadipocytes was confirmed by increased Postn levels in the stromal-vascular fraction (SVF) compared to mature adipocytes in iWAT and gWAT ( Figure 1f). To test whether Postn expression was regulated by physiological stimuli in AT, mRNA levels were assessed after 6 weeks of HFD, short-term cold exposure, and following acute beta-adrenergic stimulation with the β3adrenergic receptor agonist, CL316,243 (CL). HFD and CL injections resulted in upregulation of Postn in iWAT and gWAT, and a similar trend was found after cold, while only beta-adrenergic stimulation increased Postn expression in BAT (Figure 1g). Human periostin protein was found to be higher in iWAT and BAT compared to gWAT in males and but comparable between depots in female mice (Figure 1h,i,Supporting Information Figure S1d,e). In summary, these findings suggest that periostin may be involved in AT remodeling following physiological and dietary cues, whereas this effect is attenuated with increased age.

| Genetic ablation of Postn results in reduced body size but normal AT distribution
To directly address the potential role of periostin in AT biology, we generated homozygous LacZ reporter mice, thereby creating a strain with whole-body deletion of Postn gene expression (Figure 2a,b). Recapitulating previous observations (Oshima et al., 2002;Rios et al., 2005), Postn-KO mice displayed significantly reduced body weight as well as body and tibia lengths (Figure 2ce). AT depots displayed no size differences implying that a reduction in periostin does not affect AT development (Figure 2f,g).
Brown and white fat depots displayed normal morphology and ECM distribution, and unchanged gene expression patterns of ECM components, white and brown adipogenesis, or senescence (Supporting Information Figure S2). Similarly, pre-adipocytes isolated from iWAT and BAT of knockout mice and wild-type littermates displayed normal differentiation and adipogenic gene expression patterns suggesting intact adipogenic differentiation capacities (Supporting Information Figure S3).  . Further analysis revealed that the weight loss during cold exposure was due to a significant reduction in WAT mass, but also liver mass, rather than weight loss in other tissues (Figure 3c,d). Plasma lipid analysis revealed a marked decrease in circulating lipid levels during cold but no consistent genotype-dependent changes were observed in either male or female knockout mice (Supporting Information Figure S4a). These data suggest that systemic lipid homeostasis was regulated in a gender-specific manner in knockout animals but was unlikely to contribute to the observed phenotype.
As brown adipocyte nonshivering thermogenesis controls body temperature, we assessed expression of the brown fat-defining marker uncoupling protein 1 (Ucp1) but found it to be unchanged on . (f, g) AT depot weights of male and female WT and Postn-KO animals prior to normalization (f) or normalized to body weight (g; n = 3-6). Mean ± SEM; *p < 0.05, **p < 0.01, ***p < 0.001, assessed by unpaired t test (Glut4; Slc2a4) was induced in WAT and could potentially explain the reduced glucose levels observed in cold-exposed knockouts compared to wild-type littermates ( Figure 3h, Supporting Information Figure S4d). These data taken together suggest a potential but limited defect of brown/beige adipocyte function that may contribute the lower thermogenic potential and cold tolerance of  To test whether the HFD-related loss of fat mass could be explained by changes to thermogenesis or ECM function, gene expression was assessed but found to be unchanged in either BAT or WAT depots (Supporting Information Figure S8). Together, these data suggest that the deletion of periostin leads to impaired lipid metabolism in adipose tissues during HFD feeding which is consistent with the defective metabolic response observed during cold exposure.

| Periostin expression in human white adipose tissue is regulated by aging and obesity
To determine a potential relationship of periostin expression and AT lipid homeostasis in humans, we analyzed the human periostin mRNA expression in WAT biopsies (n = 471) of lean (BMI < 25 kg/ m 2 ), overweight (BMI 25-29.9 kg/m 2 ), and obese subjects (BMI > 29.9 kg/m 2 ). A significant inverse correlation between human periostin mRNA levels and obesity was observed in subcutaneous WAT (sWAT) as well as in visceral WAT (vWAT; Figure 6a,b). In healthy subjects with normal glucose tolerance, expression of human periostin is significantly higher in sWAT compared to vWAT, and this difference was blunted in patients with impaired glucose tolerance or overt type 2 diabetes (Figure 6c). In contrast to murine AT, a positive correlation between participant age and human periostin mRNA was found in both human AT depots ( inverse correlation was observed in normal weight individuals (BMI < 25 g/m 2 ), although it should be noted that these differences were not significant due to the comparably low sample size in the normal weight and overweight groups (Supporting Information   Table S1). In addition to BMI, the correlations with age were further explained by plasma lipids and, to a more limited extent, also by liver parameters (i.e., ASAT, ALAT; Table 1). Collectively, these data suggest that expression of human periostin mRNA in human WAT interacts with age and age-associated phenotypes depending on individual levels of adiposity and metabolic health.

| DISCUSSION
Aging is associated with defective ECM function leading to tissue fibrosis, that is, an excessive accumulation of fibrous connective tissue and ECM components (Wynn, 2007). Similar alterations have also been described in adipose tissue (Sun, Tordjman, Clement, & Scherer, 2013). Interestingly, undifferentiated adipogenic progenitor cells are a major source of ECM components in AT (Kubo, Kaidzu, Nakajima, Takenouchi, & Nakamura, 2000;Nakajima, Yamaguchi, Ozutsumi, & Aso, 1998), suggesting that alterations in progenitor cell function with increased age may contribute to homeostatic changes in the tissue matrix (Tchkonia et al., 2010). Periostin is a secreted factor and its binding integrin heterodimers has been shown to acti- , and PLIN1 in BAT, iWAT, and gWAT. Western blot signals were normalized to glyceraldehyde 3phosphate dehydrogenase (GAPDH). § pHSL and Plin1 antibodies were probed on the same membrane and normalized to GAPDH in the upper panel; # HSL, ATGL, and CD36 were probed on a separate membrane and normalized to GAPDH in the lower panel. Data are shown as mean ± SEM. n = 5. *p < 0.05, **p < 0.01 compared with WT animals assessed by Mann-Whitney U test of impaired cold tolerance, yet our analysis suggests that reduced substrate availability due to impaired lipolysis rates could be a partial explanation. The concept of catecholamine resistance has also recently been introduced in obesity and could also contribute to impaired lipid metabolism in our model (Reilly & Saltiel, 2017  important to consider other mechanisms of cold intolerance which could be related to impaired insulation due to loss of AT, yet a poor insulation could lead to an increase in metabolic rates which we did not observe in knockout animals (Cannon & Nedergaard, 2011). Periostin plays a vital role in skin health and its ablation could translate into increased heat loss due to a skin defect that only becomes apparent under challenging conditions (Murota, Lingli, & Katayama, 2017). In summary, the impaired thermoregulation in periostin-deficient mice is likely a multifactorial process that is accompanied by impaired AT lipolysis and lipid metabolism gene expression during cold. A similar process may also partially contribute to an age-related defect in AT lipid metabolism (Benjamin, Gellhorn, Wagner, & Kundel, 1961;Gellhorn & Benjamin, 1965).
Chronic overfeeding leads to AT expansion due to adipocyte hypertrophy and is accompanied by enhanced vascularization to assure adequate oxygen supply which requires substantial ECM remodeling (Lemoine et al., 2012). In our model, deletion of Postn resulted in a reduced ability for AT expansion during HFD feeding.
Given the expression of periostin in cells of the blood vessels, deletion of Postn may affect nutrient supply during AT expansion as permeability of the endothelium of blood vessels is pivotal in the release of nutrients to the tissue. Moreover, the substantial reductions in adipose tissue mass in the apparently opposing interventions of cold exposure and HFD suggest that the deletion of Postn confers a defect in AT function that is mainly characterized by impaired control of adipocyte size and could be a sign of defective lipid handling.
For instance, it was previously shown that changes in adipocyte function due to altered cell size are mediated by activation of integrin/ERK signaling (Farnier et al., 2003). Similarly, genetic ablation of laminin α4, a component of the ECM directly surrounding adipocytes, results in reduced ability to expand in response to diet-induced and age-related obesity (Vaicik et al., 2014). This is further  * ** ** * ** * * * * * * * ** * ** ** ** * * Age, corrected for Age, corrected for F I G U R E 6 Human periostin mRNA expression in different fat depots is negatively associated with obesity and parameters of glucose metabolism in humans. (a, b) Gene expression analysis of periostin mRNA in human sWAT (a) and human vWAT (b) biopsies collected from normal weight (body mass index (BMI) <25 kg/m 2 ), overweight (BMI 25-29.9 kg/m 2 ), and obese (BMI > 29.9 kg/m 2 ) subjects. (c) Human periostin mRNA levels in sWAT (white bars) and vWAT (gray bars) of humans with normal glucose tolerance, impaired glucose tolerance and type 2 diabetes. Data are shown as mean ± SEM. n = 12-408. *p < 0.05, **p < 0.01, ***p < 0.001 assessed by one-way (a, b) or two-way (c) ANOVA with Bonferroni post hoc test. (d, e) Graphical depictions of Spearman partial correlation coefficients (ρ) and 95% confidence intervals (CI) of associations between participant age and human periostin mRNA expression in sWAT (panel d) and vWAT (panel e) following adjustment for body mass index (BMI), or lipid and liver function parameters (data shown in Table 1). *p < 0.05, **p < 0.01 indicate significant correlations parameters of metabolic health as covariates, suggesting that the regulation of periostin gene expression in human WAT responds to different metabolic cues including aging. The discrepant effects of age on periostin gene expression in mice and human white fat remain to be further elucidated but our observations suggest a potential relevance in adipose tissue function. High-fat feeding in mice induced Postn mRNA after a relatively short time course of 6 weeks, while the majority of our study participants had chronically established metabolic dysfunction including insulin resistance or type 2 diabetes, which could help explain this apparent discrepancy. In livers, Postn expression is elevated in obese mice and humans, and overexpression of periostin in mouse livers promoted hepatic steatosis and hypertriglyceridemia (Lu et al., 2014). These counteracting processes in liver and AT may explain the limited effects on plasma lipids and glucose metabolism and imply a tissue-specific regulation and function of periostin. Interestingly, increased circulating periostin was described as a biomarker of increased risk to develop nonalcoholic fatty liver disease and insulin resistance during obesity (Yang et al., 2016). The respective contribution of different tissues to circulating periostin and their involvement in the clinical manifestation of metabolic diseases remains to be clarified in further detail. In our animal studies, periostin expression was induced during HFD feeding. Consequently, the predictive value of periostin as a biomarker may also depend on acute challenges such as nutrition and lifestyle.
In summary, these analyses reveal a potential involvement of periostin in metabolic dysfunction of adipose tissue, a novel crosstalk mechanism between the extracellular matrix and adipose tissue lipid metabolism which relays metabolic signals via periostin. T A B L E 1 Spearman partial correlation coefficients (ρ) and 95% confidence intervals (CI) to depict associations between participant age and human periostin mRNA expression in sWAT and vWAT following adjustment for body mass index (BMI), or lipid and liver function parameters 4.2 | Isolation of adipogenic progenitors and mRNA analysis AT-derived progenitor cells were isolated as described (Schulz et al., 2011;Steinbring, Graja, Jank, & Schulz, 2017

| Human studies
Our study includes data from 471 participants, who have been recruited at the University Hospital in Leipzig (Germany) as previ-

| Statistical analysis
Statistical differences between groups were evaluated using either an unpaired two-tailed Student's t test, Mann-Whitney U test, or analysis of variance (ANOVA) with Bonferroni post hoc test.