Serglycin Is Involved in Adipose Tissue Inflammation in Obesity

Obesity is one of the main factors with large impact on human health globally (1). It is the strongest causal factor for the development of common clinical conditions such as metabolic syndrome, hypertension, insulin resistance, and type 2 diabetes mellitus (2). Many of the complications caused by obesity are manifested with the appearance of low-grade inflammation in white adipose tissue (WAT) (3). Adipose tissue consists of specialized lipid-storing adipocytes, which serve as the major energy depot in the body (4), but also contains several other cell types, including immune cells and endothelial cells. WAT mass is constantly changing and may expand by increasing the size of existing adipocytes (hypertrophy) or by activating differentiation of preadipocyte stem cells into new adipocytes (hyperplasia) (3, 4). These processes are complex, involving strong genetic regulations (5). Studies in mice and humans suggest that there is a considerable increase in the expression of proinflammatory genes in obese compared with lean subjects (3, 6, 7). Notably, the excessive fat deposition seems to have an effect on immune cell abundance of WAT and the production of cytokines, which are important mediators in the development of metabolic syndrome and other complications. The mechanisms initiating obesity-driven inflammation in WAT are not completely understood, but activation of macrophages, mast cells, CD4- and CD8-positive lymphocytes, B cells, and NK cells has been linked to WAT inflammation (8–11).

Proteoglycans comprise a large group of glycoconjugates with a broad range of cellular functions. All proteoglycans consist of a core protein to which one or several glycosaminoglycan chains are covalently attached. The glycosaminoglycan chains are sulfated and thereby highly negatively charged (12, 13). Proteoglycans are especially abundant in the extracellular matrix but are also found on cell surfaces and in intracellular compartments. Proteoglycans have been implicated in a wide range of physiological and pathophysiological conditions, including embryogenesis cancer, metastasis, and inflammation (14–16). Their role in obesity and inflammation of WAT, however, has not been extensively addressed.

Out of the various proteoglycans, serglycin (Srgn) has a special role as an intracellular proteoglycan predominantly located in secretory vesicles (13, 17). Srgn is further a dominant proteoglycan in inflammatory cell types that infiltrate WAT in the context of obesity. For example, Srgn is the dominating proteoglycan in macrophages, T cells, mast cells, and platelets (15), suggesting that Srgn expression in WAT may originate from infiltrated immune cells. However, a recent study suggested that Srgn is expressed in adipocytes and is induced in expression with adipocyte differentiation (18). Further, Srgn expression has been linked to the regulation of plasma low-density lipoprotein levels (19), which are often altered in acute as well as chronic inflammation (20). Based on these observations, we hypothesized that Srgn might have an impact on processes related to development of adipose inflammation. To address this possibility, we studied development of adipose inflammation in response to a high-fat and high-sucrose diet in male mice with serglycin deficiency (Srgn^−/− mice). Intriguingly, our data suggest that Srgn has a protective role in adipose tissue and dampens the immune profile otherwise associated with obesity.

Animal experiments conformed to the ARRIVE guidelines and ethical guidelines in Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes and the guidelines of the Swedish National Board for Laboratory Animals and the Norwegian Animal Research Authority. The animal study involving Srgn^−/− mice was approved by the local animal ethics committee (Uppsala, Sweden), whereas studies in the C57BL6/N strain was approved and registered by the Norwegian Animal Research Authority (Mattilsynet, approvals FOTS no. 10901 and no. 10902). All mice were kept at 22–24°C with a strict 12-h light/dark cycle.

Male Srgn^+/+ and Srgn^−/− mice on C57BL/6J genetic background were bred and housed at the National Veterinary Institute (Uppsala, Sweden) as described previously (21). Age-matched male mice were housed with ad libitum access to a diet with high-fat and high-sucrose content (no. D12451; Research Diets, New Brunswick, NJ) for 8 wk. At the end of the intervention, animals were sacrificed in the morning (9–11 am) by cervical dislocation. Epididymal WAT (eWAT) was rapidly frozen in liquid nitrogen or fixed in 4% paraformaldehyde solution. A total of 20 mice were included in the study. Analyses were performed with all individuals included, with the exception of certain analyses in which a subgroup of animals was selected based on age and weight to represent the whole group.

Female and male C57BL/6N mice (obtained from Janvier Labs, Le Genest-Saint-Isle, France) were kept on a standard chow diet (2018 Teklad global 18% protein rodent diets; Envigo, Madison, WI) until 8 wk of age. Next, the animals were stratified into two intervention groups given two types of diet containing either 60 or 10% fat (no. D12492 and no. D12450J; Research Diets, New Brunswick, NJ) for 10 wk. The male mice increased their body weight significantly, from 33.5 ± 3.2 SD to 43.8 ± 5.1 SD, and the female mice from 26.2 ± 2.2 SD to 31.2 ± 6.1 SD. At the end of the experiment, all mice from the same breeding colony were euthanized by cervical dislocation in a refed state.

A total of 15 human s.c. WAT (sWAT) and visceral WAT (vWAT) samples were obtained from patients who underwent laparoscopic abdominal surgery for weight reduction (gastric bypass, n = 4; sleeve gastrectomy, n = 5; or a single anastomosis gastric bypass, n = 1) or gallbladder removal (cholecystectomy, n = 5). Adipose tissue samples from 10 subjects with morbid obesity (mean age, 45.8 ± 15 y SD; mean body mass index (BMI), 42.7 ± 4.3 kg/m² SD), and five lean individuals (mean age, 45.6 ± 11.8 y SD; mean BMI, 24.6 ± 1.6 kg/m² SD) were taken from abdominal regions, directly frozen in dry ice, and stored at −80°C. The anthropometric and biochemical traits of the sample population are presented in Table IV. No history of cancer was recorded and two of the persons with obesity had type 2 diabetes mellitus. All study protocols have been approved by the Norwegian regional ethics committee. All participants gave written informed consent before taking part in the study.

To study SRGN mRNA expression during weight loss, adipose tissue samples obtained from patients undergoing bariatric surgery were analyzed (22). Samples from sWAT were collected from 16 subjects (mean age, 39.3 ± 10.9 y SD; mean BMI, 53.3 ± 4.3 kg/m² SD, 12 of which were women) during the bariatric surgery and 1 y after surgery (mean BMI, 33.1 ± 5.0 kg/m² SD). Control sWAT samples were also taken from 13 lean subjects undergoing inguinal hernia repair (mean age, 47.6 ± 17.1 y SD; mean BMI, 23.0 ± 2.5 kg/m² SD, 6 of which were women). Microarray analysis was performed on total RNA from those adipose tissue samples using Illumina iScan system. More information about materials and methods and anthropometric and biochemical data are available in Dankel et al. (2010) (22).

Total RNA was isolated from eWAT from a subpopulation of animals (n = 5) using the column-based Direct-zol RNA MiniPrep kit (The Epigenetics Company, Irvine, CA). cDNA libraries were prepared and amplified with the Ion AmpliSeq Transcriptome Mouse Gene Expression Kit (Life Technologies, Carlsbad, CA), following the manufacturer’s instructions. Sequencing was performed on an Ion S5 XL Sequencer (Thermo Fisher Scientific, Waltham, MA). Data had read lengths with average between 98 and 110 bp and high mapping (95% of aligned bases). Expression values were calculated with Torrent Suite software (version 5.10.1). Raw AmpliSeq data were used for downstream differential expression analyses using the edgeR package (23) implemented in R (version 3.6.0) (https://www.R-project.org/). Gene expression values were normalized using the trimmed mean of m-values normalization method. Genes with sufficient expression values (trimmed mean of m-values normalization method >10) were processed to enable downstream statistical analyses. The Benjamini–Hochberg method was used to correct the p values for multiple comparisons (false discovery rate [FDR] or p-adjusted value [padj]). For clustering and heatmap generation, heatmap.2 from the Rstudio gplots3.0.1.1 was used. AmpliSeq data are accessible on Gene Expression Omnibus (GSE166019).

Gene Ontology (GO) analyses were performed using the Molecular Signature Database available at the Gene Set Enrichment Analysis from the University of California, San Diego and the Broad Institute (https://www.gsea-msigdb.org/gsea/msigdb/). After examination of hierarchical tree on the heatmap, significantly affected genes (based on p value and log₂ fold change [log₂FC]) were loaded into the database, and the enrichment of Biological Process gene GO terms were examined. The threshold for significant enrichment was p value <0.05 and log₂FC ≥1 or ≤−1.

Genes from the AmpliSeq data were selected for validation with reverse transcription quantitative real-time PCR (RT-qPCR) with eWAT samples of the five animals used for transcriptome analyses and additional animals (n = 9 and n = 11 for Srgn^+/+ and Srgn^−/− mice, respectively). Tissues were placed in tubes containing glass beads and QIAzol (QIAGEN, Hilden, Germany) and were rapidly homogenized in a Precellys 24 homogenizer at 6000 rpm for 3 × 20 s (Bertin Technologies, Montigny-le-Bretonneux, France). After adding chloroform, samples were centrifuged at 14,000 × g, at 4°C for 15 min. The water-soluble phase was loaded onto the Nucleospin RNA purification kit (Macherey-Nagel, Düren, Germany). RNA from human samples was extracted using (RNeasy Plus Universal Kit; QIAGEN). The quality of the RNA was assessed with NanoDrop ND-1000 (NanoDrop Technologies, Wilmington, DE). Total RNA was reverse transcribed using random hexamers and the MultiScribe Reverse Transcriptase kit (Thermo Fisher Scientific). Gene-specific regions were amplified from cDNA (5–10 ng/µl) with the Bio-Rad SsoAdvanced Universal SYBR Green Supermix. Assay primers were designed with Primer-BLAST software. Expression analysis was performed with CXF Software (Bio-Rad Laboratories, Hercules, CA). The 60S ribosomal protein L32 (Rpl32) and Tata-binding protein (TBP) mRNAs were verified as stably expressed in eWAT and fibroblast cell cultures, respectively, and used to normalize gene expression data. TBP mRNAs was used for normalization of human sWAT samples. Relative gene expression was calculated pairwise for every tissue or cell type by the relative quantification method (2^−ΔΔCq).

Adipose tissue was fixed in 4% paraformaldehyde, washed in PBS, and subsequently embedded in paraffin. Embedded tissue was sectioned at 5 µm using an automated microtome Microm HM 355s (Thermo Fisher Scientific). Heat-induced epitopes were retrieved in 10 mM sodium citrate buffer by heating in water bath for 10 min. Sections were blocked using 5% normal goat serum for 1 h.

To visualize M1 and M2 macrophages, sections were incubated with rat anti-mouse monoclonal F4/80 Ab (1:250, no. ab6640; Abcam, Cambridge, U.K.) at 4°C overnight. Next, sections were incubated with goat anti-rat Alexa Fluor 546 Ab (1:400, no. A-11035; Thermo Fisher Scientific). Sections were then washed six times with PBS for 5 min and incubated again with rabbit Arg1 Ab (1:100, no. PA5-85267; Thermo Fisher Scientific) at room temperature (RT) for 1 h. Subsequently, the sections were washed and incubated with goat anti-rabbit Alexa Fluor 488 (1:400, no. ab150077; Abcam) at RT for 1 h. Hoechst 33342 (5 µmol/l) was used to counterstain the nuclei at RT for 5 min.

To visualize crown-like structures (CLS), sections were incubated with rat anti-mouse monoclonal F4/80 Ab and goat anti-rat Alexa Fluor 350 (1:400, no. A-21093; Thermo Fisher Scientific) at RT for 1 h.

Fifteen representative pictures per animal were taken at 20× and 40× magnification using an Olympus BX61 microscope (Olympus, Tokyo, Japan). For the M1/M2 macrophage quantification, the overlay of F4/80 and Hoechst was used to calculate the total presence of macrophages, and the overlay of the F4/80, Arg1, and Hoechst was used to calculate M2 macrophages using Olympus CellSens software (Olympus). CLS were counted manually. All cells were expressed per surface area.

To determine adipocyte size, sections separated with 20 sequential cuts were used for quantification to ensure that no repetitive adipocytes were imaged. Tissue sections were rehydrated and stained with Gill’s hematoxylin and 0.1% eosin. Subsequently, sections were dehydrated, mounted, and then dried overnight at 37°C. Representative pictures were taken using an Olympus BX61 microscope. The adipocyte size was measured using CellProfiler software v2.1.1 (The Carpenter Lab, Cambridge, MA) and expressed in surface area (in square micrometers) per adipocyte. The adipocyte surface area was distributed in 100-μm² clusters in Excel. Subsequently, adipocytes were divided in four groups: very small (>100–1000 μm²), small (>1000–2500 μm²), medium (>2500–7000 μm²), and large (>7000 μm²), so that each group contains ∼25% cells.

A publicly available single-cell RNA analysis mouse dataset (GSE133486) (24) and human dataset GSE129363 (25) were used to identify adipose tissue cell types with high expression of Srgn. Seurat R software package was used for analysis of the single-cell RNA sequencing (scRNAseq) data. Cells with unique gene count of maximum of 2500 genes per cell and minimum 700 for the stromal vascular fraction (SVF) and 200 for the mature adipocyte fraction were filtered out and used for further analysis for the GSE133486 dataset. For the human adipose tissue dataset GSE129363, cells with unique gene count of maximum of 2500 genes per cell and minimum 200 were filtered out. After filtering, data were normalized using the global-scaling method LogNormalize. The data matrix ScaleData function was used for centering and scaling. Top 12 dimensions were deployed in the principal component analysis that was used for graph clustering. We examined the top 100 mostly expressed genes in all cells in each cluster using the function FindAllMarkers in Seurat with avg.logFC >0.25 and padj <0.01. Clusters were manually assigned using cell-specific markers as recommended in Refs. 24, 25. Genes with differential expression between the clusters were discovered with the function FindMarkers.

Human Simpson–Golabi–Behmel syndrome (SGBS) preadipocyte cells [a kind gift from Dr. M. Wabitsch from Ulm University, Ulm, Germany (26)] were cultured in DMEM/Nutrient Mix F12 Ham with 10% noninactivated FBS, 4 mM l-glutamine, 50 U/ml penicillin, and 50 µg/ml streptomycin. The medium was supplemented with 8 mg/ml biotin and 4 mg/l d-pantothenate. For differentiation experiments, cells were seeded in a 12-well dish. Differentiation was initiated after cells reached 90% confluence and by changing to a serum-free medium supplemented with 0.01 mg/ml human apo-transferrin, 20 nM human insulin, 0.2 nM triiodothyronine, and 100 nM hydrocortisol and a differentiation mix consisting of 500 μM 3-isobutyl-1-methylxanthine, 25 nM dexamethasone and 2 µM rosiglitazone for 4 d. The differentiation mix was then removed, and medium was changed every fourth day until day 12 after initiation of differentiation.

Mouse 3T3-L1 preadipocytes (no. CL-173; American Type Culture Collection, Manassas, VA) were cultured in DMEM growth medium (no. D6546; Sigma-Aldrich, Darmstadt, Germany) supplemented with 10% noninactivated FBS, 4 mM l-glutamine, and an antibiotic mix containing 50 U/ml penicillin/50 µg/ml streptomycin, as described previously (27). For differentiation studies, cells were seeded in 12-well plates (8 × 10⁴ cells per well) and grown 1 d postconfluence before initiation of adipocyte differentiation. Differentiation was stimulated by adding 20 nM insulin, 500 μM 3-isobutyl-1-methylxanthine, 25 nM dexamethasone, and 2 µM rosiglitazone to the grown medium. After 3 d of incubation, the medium was changed to normal growth medium supplemented with 20 nM insulin. After another 3 d of incubation, the medium was replaced with normal growth medium. The cells were differentiated for 9 d after initiation of differentiation.

Data from RT-qPCR analyses are presented as the mean ± SEM or SD. Error bars for all results were acquired from biological, not technical, replicates. Significant differences in the groups were estimated with two-tailed Student t tests for the human intervention and Srgn^−/− mice experiment. Two-way ANOVA was used to investigate possible significance in obese male and female C57BL/6N mice experiments, as well as in the human sWAT and vWAT samples from obese and lean individuals. One-way ANOVA was used for cell culture experiments by taking day 0 as a control, as well as for comparison of the pre- and postbariatric surgery, and lean control individuals (*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001). Obtained data were visualized in GraphPad Prism version 6.04 (La Jolla, CA).

To study if Srgn influences development of obesity-induced adipose tissue inflammation, male wild-type (Srgn^+/+) mice and male mice lacking serglycin expression (Srgn^−/−) were fed a high-fat and high-sucrose diet for 8 wk. Only male mice were used in these experiments to avoid the influence of hormonal fluctuation during the feeding (28–30). Mice in both groups had comparable body weight gain (Table I) but with a trend of lower body weight of Srgn^−/− mice at the end of the diet intervention. The weight of eWAT and sWAT was similar in Srgn^+/+ and Srgn^−/− mice after the diet intervention. Hence, the absence of serglycin did not affect total weight gain or the weight of adipose tissues in male mice receiving an energy-dense diet for 8 wk.

We next investigated if Srgn could have an effect on adipose tissue gene expression in the diet-induced obese male mice. For our transcriptomic analysis of WAT, we selected five mice that were the most representative for each group (based on body weight). We focused our analyses on eWAT. Obtained sequencing reads were analyzed with a common statistical workflow for RNA sequencing analysis (see Materials and Methods). A significant differential expression of numerous genes was observed in Srgn^+/+ versus Srgn^−/− eWAT after the 8-wk high-fat and high-sucrose diet intervention (Supplemental Table I). Gene expression patterns were visualized by examining the normalized gene values with altered (increased and reduced) expression (n = 1330, p < 0.05). These data were used for constructing a heatmap to inspect if there was a clear segregation of transcriptome clusters and patterns of altered gene expression in Srgn^+/+ versus Srgn^−/− eWAT (Fig. 1A). A clear segregation of the two transcriptomic profiles of each genotype were also evident with principal component analyses (Fig. 1B) in which the first component describes 77% of the variance of the initial data. Analysis of differential gene expression revealed 58 genes that were considered significantly different (FDR < 5%) in Srgn^+/+ versus Srgn^−/− eWAT as visualized in the volcano plot (Fig. 1C). After setting a stricter threshold for significance (FDR < 5% and log₂FC above 0.75 or below −0.75), we identified 13 genes with decreased (Table II) and 22 genes with increased expression (Supplemental Table II) in Srgn^−/− versus Srgn^+/+ eWAT. In further analyses, we noted that the genes that were less regulated in Srgn^−/− versus Srgn^+/+ eWAT mostly corresponded to genes implicated in immune responses. Among these were genes associated with B cells Jchain, Mzb1, Pou2af1 (also denoted OCAB and OBF1), and Tbc1d10c (also denoted carabin) (Table II). We also found that Vwa1, Zpf874, and Serpinb8 mRNAs were significantly repressed. S100a9 and CD83 expressed in dendritic cells and in a broad range of other immune cells (Supplemental Table I) showed a tendency for reduced expression in Srgn^−/− compared with Srgn^+/+ eWAT (Supplemental Table I), but the FDR was not below 5%. It was also noted that Tpsb2 (also denoted mast cell tryptase) was significantly reduced in Srgn^−/− eWAT.

To investigate if the heatmap clustering has biological relevance and was not an artifact of the clustering procedure in the statistical platform, we performed GO analysis. For this, we chose the most differentially expressed genes to search for overrepresented functional categories (log₂FC > ±1). The GO analysis revealed that Gene Set Enrichment Analysis Biological Process categories involved in the immune response and defense mechanisms were strongly overrepresented among the reduced genes in the Srgn^−/− eWAT (Table III). The categories included processes such as lymphocyte activation, lymphocyte/leukocyte differentiation, regulation of immune response, and defense responses. This included genes such as Lck, CD83, Tbcd10c, Mzb1, S100a9, Serpinb9, and Cxcl13. Furthermore, the GO analysis linked thrombospondin 1 (Thbs1), which was one of the most affected genes by the lack of serglycin, to immune defense response pathways and regulation of hydrolase activity.

Similar assessment of genes that were increased in Srgn^−/− eWAT identified functional classes connected to cell organization and ion transport (Supplemental Table III). Furthermore, it is notable that many of the increased genes in the Srgn^−/− group are associated with cilium and microtubule, as represented by the categories: cilium and flagellum motility and microtubule formation and movement.

Our initial transcriptomic and GO analysis indicated that the serglycin deficiency resulted in an alteration of the immune profile of the adipose tissue. To substantiate this, we performed RT-qPCR to validate and assess the expression of markers specific for distinct immune cell populations. As seen in (Fig. 2A, the trend for lower expression of the macrophage marker CD68 in the AmpliSeq data were confirmed with RT-qPCR. It was significantly decreased in Srgn^−/− eWAT. Also, the macrophage marker Adgre1 was expressed at significantly lower levels in Srgn^−/− eWAT. In agreement with the global transcriptome assessment, the RT-qPCR analysis revealed a profound reduction in the expression of six markers for dendritic cells (CD83, S100a9, Lyz1, Itgax (CD11c), Itgam (CD11b), and H2-Eb1), accompanied by a trend of reduced expression (p < 0.09) of two T cell markers (CD8a and CD8b1) similar to the AmpliSeq results. CD4 was not affected in Srgn^−/− eWAT, in consensus with the whole transcriptome data. In further agreement with the transcriptome approach, the RT-qPCR analysis revealed a reduction in the B cell marker Jchain, but not in CD79a. Thbs1 was also significantly downregulated in Srgn^−/− eWAT. To further characterize the macrophage population, we evaluated expression of genes enriched in M1 and M2 activated macrophages. These analyses revealed that established M1 macrophage mRNA markers Il1b, Infg, and Il12b were expressed at significantly lower levels in eWAT from Srgn^−/− versus Srgn^+/+ mice (Fig. 2B). In contrast, no significant differences in the expression of the M2 macrophage markers [Mrc1 (CD206) and Arg1] were seen.

To determine if the observed altered expression of immune cell marker genes were translated to alterations of macrophage populations in eWAT, we performed double immunostaining for F4/80 and Arg1 to differentiate between M1 (F4/80⁺ and Arg1⁻) and M2 (F4/80⁺ and Arg1⁺) macrophages (Fig. 3A). This revealed a decreased total number of F4/80⁺ macrophages in Srgn^−/− versus Srgn^+/+ eWAT (Fig. 3A, 3C), whereas no difference was observed for the M2 marker (F4/80⁺ and Arg1⁺). This suggests that the reduced total number of macrophages is due to a reduction in M1 macrophages (F4/80⁺ and Arg1⁻). Clearly, this is in agreement with our observed reduced mRNA expression of the M1 macrophage markers, as judged by RT-qPCR analysis (Fig. 2B).

We also assessed the tissue for the presence of CLS. CLS represent dying adipocytes surrounded by macrophages, and the appearance of CLS is established as a hallmark event occurring in adipose tissue inflammation (31). The absence of serglycin was associated with a profound reduction in the number of CLS in the adipose tissue of mice fed a high-fat and high-sucrose diet (Fig. 3B, 3D). Immunohistochemistry staining against the B cell marker CD19 revealed no difference between the groups and did not confirm the reduced mRNA expression of B cell markers (data not shown).

The data above indicate that the absence of serglycin does not affect the overall weight gain and overall weight of adipose tissues. However, this does not exclude the possibility that serglycin might affect the fine structure of the adipose tissue, possibly by effects related to altered presence of immune cell populations as a consequence of serglycin deficiency. To evaluate this possibility, we assessed Srgn^+/+ and Srgn^−/− eWAT with regard to adipocyte size using H&E staining (Fig. 4A). As seen in (Fig. 4A and 4B, this analysis showed a marked reduction of adipocyte mean size in Srgn^−/− versus Srgn^+/+ mice. Notably, eWAT from Srgn^+/+ and Srgn^−/− mice had similar distribution of adipocytes up to 2500 μm² (Fig. 4C–E), whereas the absence of serglycin was accompanied by a significant increase in medium-sized (2500–7000 μm²) adipocytes (Fig. 4F) and a decrease of the large-sized (>7000 μm²) adipocytes (Fig. 4G).

To analyze the cellular origin of Srgn expression in adipose tissue, we performed analysis of scRNAseq data from mouse and human adipose tissue. Data for analysis were retrieved from the GSE133486 and GSE129363 datasets (24, 25). Cells from the mouse scRNAseq data were segregated into 16 clusters of purified adipocytes (Fig. 5A) and 12 types of immune cells (Fig. 5C) were present in the remaining SVF of WAT. The t-distributed Uniform Manifold Approximation and Projection (t-UMAP) plot shows the normalized expression of Srgn mRNA in these clusters (Fig. 5B, 5D), revealing that Srgn mRNA was predominantly expressed by immune cells (Fig. 5D), with particularly high expression in dendritic cells. High expression was also seen in monocytes, macrophages, basophils, neutrophils, NK cells, CD8⁺ T cells, CD4⁺ T cells, and B cells. The scRNAseq analysis revealed high expression of B cell markers (Bank1 and Ms4a1) and dendritic cell markers (CD83, H2-Ob, H2-Ab1, and H2-Eb1) in the “mixed” cluster in the SVF. Low Srgn mRNA expression was observed in the adipocyte progenitors (stem cells) of the SVF fraction. A corresponding t-UMAP analysis of mature adipocyte populations (Fig. 5B) revealed that Srgn mRNA was considerably lower than in the different immune cell populations (Fig. 5D). Srgn mRNA expression showed no correlation with any of the mature adipocyte clusters.

To further validate these observations, we compared the expression of SRGN mRNA within different cell types of human s.c. SVF. Data for analysis were retrieved from GSE129363 dataset (25), which revealed 16 different cell type populations within SVF (Fig. 5E). Our manual or unsupervised clustering found three clusters of endothelial cells and six clusters of progenitor cells. The latter were expressing CFD and other adipose cell markers (MGP and FABP5), whereas the progenitor 6 cluster expressed proinflammatory markers (CCL5, CD3D, and IL7R), indicating that this cluster includes hematopoietic progenitor cells (result not shown). The t-UMAP graph showed that SRGN mRNA was predominantly expressed by clusters enriched with immune cell and proinflammatory markers (Fig. 5F). In line with the mouse scRNAseq results, SRGN mRNA in human WAT was predominantly high in dendritic cells, macrophages, NK cells, and other immune cell clusters compared with the progenitor and endothelial cells clusters.

We next evaluated if Srgn mRNA expression increases under the process of adipocyte differentiation. Experiments performed in the human SGBS preadipocyte cell line revealed that SRGN expression was downregulated upon culture of these cells with adipogenic stimuli (Fig. 6A). Notably, SRGN mRNA expression followed the opposite trend versus the expression of the classical adipocyte differentiation markers PPARG and SLC2A4 (GLUT4) (Fig. 6A). However, IL6 showed a similar expression pattern as Srgn during adipocyte differentiation, which is consistent with previous research showing a negative correlation between this inflammatory marker and adipocyte differentiation (32). In addition, we performed experiments in 3T3-L1 murine cell line. At day 6, the 3T3-L1 cells were fully differentiated to adipocytes, as shown by the increased expression of the adipocyte markers Pparg and Slc2a4 (Glut4) (Fig. 6B). However, the expression of Srgn did not statistically change for the duration of the entire differentiation experiment, and it was notable that the expression of Srgn was overall low at all stages of adipocyte differentiation (C_t values >34). We also noted a decrease in Il6 expression during the maturation of the murine adipocytes, in agreement with the observed downregulation of this gene during adipocyte differentiation of the human SGBS cells. Altogether, these results suggest that expression of Srgn mRNA is low in mature human or mouse adipocytes.

Our findings above indicated that SRGN mRNA is predominantly expressed by immune cells populating WAT, whereas adipocytes show low SRGN mRNA expression. It is well established that the number of immune cells in WAT increases with obesity (4), and we therefore considered the possibility that SRGN mRNA expression might be increased during the development of obesity in humans. To address this possibility, we compared SRGN mRNA between humans with obesity versus lean controls. The anthropometric traits of the sample population are presented in Table IV. As shown in (Fig. 7A, expression of SRGN mRNA in sWAT was nearly 3-fold higher in humans with obesity (BMI, 42.7 ± 4.3 kg/m² SD) compared with lean controls (mean BMI, 24.6 ± 1.6 kg/m² SD). Likewise, SRGN mRNA in vWAT showed an apparent increased expression compared with the corresponding WAT in lean individuals, but the trend did not reach statistical significance after two-way ANOVA test (Fig. 7A). A similar trend was also seen for the macrophage marker CD68 (data not shown). Consistent with these human data, Srgn mRNA expression was ∼50% increased in gonadal WAT (eWAT and ovarian WAT) and sWAT of both male and female obese compared with lean mice (Fig. 7B, 7C). In parallel with induction of SRGN/Srgn mRNA expression, several immune cell markers were elevated in humans and mice with obesity (Fig. 7D, 7E), compared with the corresponding lean controls. Altogether, these data suggest that adipose SRGN/Srgn mRNA expression is linked to immune cell abundance in WAT.

To further substantiate the link between Srgn and inflammatory status in vivo, we examined changes in SRGN mRNA expression in bariatric patients at the peak of their obesity (mean BMI, 53 kg/m²) and 1 y after surgery (postsurgery BMI, 33 kg/m²). The patients were not asked to lose weight prior to surgery and showed a marked reduction in the expression of inflammatory genes 1 y after surgery (see Ref. 22). As depicted in (Fig. 7F, SRGN mRNA in sWAT was ∼58% reduced after a profound fat loss, returning to the level seen in lean subjects. This reduction was tightly correlated with reduced expression of genes encoding inflammatory factors such as S100A9 (Pearson r = 0.77, p = 0.0003), CCL2 (Pearson r = 0.78, p = 0.0004), and IL6 (Pearson r = 0.55, p = 0.028) (Fig. 7G). These findings provide further support for a role of Srgn in adipose tissue inflammation in obesity.

We show that the absence of serglycin markedly affects the adipose tissue transcriptome after a high-fat and high-sucrose diet intervention for 8 wk, suggesting that Srgn is involved in diet-induced adipose inflammation. In particular, we note that the absence of Srgn mRNA caused differential expression of genes attributed to the functional categories: immune system processes, leukocyte activation, and differentiation. Examples of such genes include Pou2af1, Jchain, Mzb1, CD3e, CD83, and S100a9. These findings are in agreement with a role for Srgn in protease secretion and chemokine signaling of various immune cell populations (15). For example, it is known that Srgn^−/− mice have a low ability to clear Klebsiella pneumonia infection due to defective elastase storage in neutrophils (33). In addition, Srgn has been shown to inhibit complement activity by binding to C1q and mannose binding lectin (MBL) (34). It has also been demonstrated that Srgn can modulate inflammation by interacting with CD44 on normal mouse chondrocytes (35). In HUVEC cells, the storage of Cxcl1 was found to be dependent on Srgn, hence supporting a role for Srgn in regulating inflammatory processes (36). We also noted an effect of Srgn deficiency in the expression of Thbs1 in Srgn^−/− WAT. Notably, Thbs1 was one of highest expressed genes in the WAT in a rat model for visceral adiposity (37), and previous studies in mice have reported that Thbs1-null mice are protected against obesity and adipose tissue inflammation (38). However, future studies will be required to outline the possible functional relationship between serglycin and thrombospondin.

A major finding in this study is that the absence of Srgn was associated with a decrease in the number of M1 macrophages in eWAT, as well as with a reduction in expression of several M1 macrophage markers (Il1b, Infg, and Il12b). A reduction in the expression of several general macrophage/dendritic cell markers [CD83, CD68, Adgre1, Lyz1, Itgam (CD11c), S100a9, and Itgax (CD11b)] was also seen in Srgn^−/− WAT. There is now substantial evidence supporting that high-caloric intake in mice increases the number of proinflammatory M1 macrophages in WAT (39, 40) and that total macrophages may account for more than 50% of all cells in WAT under such conditions (31, 41). Our data thus introduce the notion that the switch of macrophages toward M1 phenotype under obese conditions at least partly depends on Srgn. It is thought that the accumulation of macrophages under obese conditions is caused by an autocrine mechanism in which the macrophages, rather that adipocytes, secrete macrophage-attracting compounds (42). Intriguingly, many of the known macrophage chemoattractants, including various chemokines, are known to interact with serglycin and/or other proteoglycans (43). A plausible explanation for decreased accumulation of macrophages in WAT from obese Srgn^−/− mice could therefore be decreased activity of macrophage-attracting chemotactic factors and a lack of functional interaction of these with serglycin expressed by the macrophages. It has been shown that serglycin is the major proteoglycan secreted by macrophages and monocytes, and it is also known that Srgn mRNA expression is induced in macrophages activated by proinflammatory stimuli (44–46). Moreover, serglycin has been shown to regulate the secretion of TNF-α (44), suggesting that serglycin is involved in the regulation of cytokine output from macrophages. Hence, a plausible scenario could be that the effects of serglycin deficiency on the immune profile of the WAT could be related to serglycin-mediated effects on the cytokine milieu in the tissue. In agreement with an impact of serglycin on features of WAT inflammation, we also demonstrate that serglycin expression is closely associated with development of obesity or fat loss. Both in humans and mice, SRGN/Srgn mRNA expression is accompanied by increased expression of several established inflammatory markers (including macrophage markers), and we also demonstrated that the profound fat loss following bariatric surgery involved lowered adipose expression of SRGN and inflammatory genes mRNA.

Interestingly, we found a decrease in the presence of CLS in WAT of obese Srgn^−/− mice. CLS are composed of macrophages surrounding dying adipocytes, and the formation of CLS constitutes a hallmark histologic feature of adipose inflammation. Our findings thus provide strong support for a role of Srgn in establishment of adipose inflammation that develops with of obesity. We also show that Srgn^−/− mice have fewer large adipocytes compared with Srgn^+/+ mice. Possibly, this might be related to an impact of Srgn on adipocyte turnover during the obese conditions. In line with this notion, previous findings have revealed that Srgn can have an impact on cell death by promoting apoptotic versus necrotic cell death (47). Further investigations are needed to investigate the exact mechanism underlying the effects of Srgn deficiency on the adipose morphology and function during obese conditions.

Serglycin plays a critical role in mediating the storage of granzyme B in cytotoxic T cells and in promoting their cytotoxic activity (48, 49). It is also notable that CD8⁺ T cells are known accumulate in both humans and mice WAT during obese conditions (50). In this study, we did not observe any effect of Srgn on the expression of CD8a and CD8b mRNA in WAT. We found that the absence of Srgn was accompanied by reduced expression of a number of B cell markers, including Jchain, Pou2af, and Mzb1. However, we did not see a corresponding decrease in the number of B cells infiltrating WAT. Hence, these data suggest that the B cell population of the adipose tissue may become functionally compromised, although this is not translated to an effect on B and T cell numbers. In support of an impact on B cells during obesity, previous studies have revealed an increase in B cell infiltration in WAT after high-caloric feeding in mice (25).

An important question is whether the observed effects of Srgn deficiency are related to effects within the adipocyte populations. The scRNAseq analysis of adipose tissue revealed that Srgn mRNA expression was considerably lower in adipocytes than in the different immune cell population. In particular, strong expression of Srgn mRNA was seen in dendritic cells and macrophages. High expression was also seen in neutrophils and NK, T, and B cells. However, previous investigations have suggested that Srgn mRNA is considerably expressed by mature mouse adipocytes (18). To provide more insight into this issue, we investigated in this study whether serglycin expression is increased in culture mouse or human adipocytes. We found that Srgn mRNA expression was either unaffected or downregulated during adipocyte differentiation. Hence, these findings support the notion that Srgn is predominantly associated with the immune cell populations that accumulate adipose tissue under obese conditions.

The authors wish to express sincere appreciation to Linda Dorg for preparing eWAT for histological staining and allowing the authors to make use of the facilities in the Department of Pathophysiology, Oslo University Hospital, Oslo, Norway. We thank Dr. Martin Wabitsch for kindly providing SGBS cell line. The authors are also grateful to Nikolay Stoychev for the technical assistance given with the machine learning analysis of the scRNAseq data. We thank Pratibha Kolan for the technical assistance with the gene expression assays. We also thank Drs. Jon Kristinsson, Tom Mala (Oslo University Hospital, Aker), Marius Svanevik (Vestfold Hospital Trust), Villy Våge, and Inge Glambæk for the help in obtaining human tissue samples.

The authors have no financial conflicts of interest.