Regulation of the Fibrosis and Angiogenesis Promoter SPARC/Osteonectin in Human Adipose Tissue by Weight Change, Leptin, Insulin, and Glucose

Volunteers were recruited from adults undergoing bariatric surgery (obese group) and routine abdominal surgery (lean group) at the University Hospital Aintree, Liverpool, after ethical permission from the local research ethics committee. Subjects with diabetes, inflammatory disorders, or infectious or malignant diseases and subjects on drug treatments that would likely affect body weight or any study variables, including endocrine diseases and treatments with systemic glucocorticoids, were excluded. Lean and obese participants were of similar age and had the following patient characteristics: obese group—13 men and 26 women, age 44.5 ± 1.4 years (mean ± SD), BMI 46.8 ± 1.9 kg/m², n = 39; lean group—10 men and 8 women, age 42.3 ± 4.1 years, BMI 23.6 ± 0.8 kg/m², n = 18. Volunteers attended the clinical investigation unit after an overnight fast from midnight for a physical examination and venesection before admission for surgery. Blood pressure, weight, and height were recorded by the same observer, and BMI was calculated. Waist circumferences were recorded by measurement at the midpoint between iliac crest and lowest point of the costal margin. Body composition was estimated by electrical bio-impedance measurement using a four-pole system (Bodystat, Isle of Man). Patient characteristics are shown in Table 1.

Subjects for the VLCD study that aims to identify gene expression changes in adipose tissue of obese subjects undergoing weight loss from caloric restriction were recruited after approval by the ethics review board at University of Gothenburg, Göteborg Sweden. As previously described (9), a total of 40 obese (BMI >30) men and women age 25–61 years participated in the study. Smokers and those with pharmacological treatment of diabetes or lipid-lowering medication were excluded. Of the 40 subjects recruited, 21 met the criteria for metabolic syndrome based on the modified World Health Organization criteria (10): type 2 diabetes or impaired glucose tolerance as measured by oral glucose tolerance test, lipid disturbance [triglycerides >1.7 mmol/l or HDL <0.9 mmol/l in men and HDL <1.0 mmol/l in women], and hypertension (blood pressure >140/90 mmHg). All subjects were treated with three VLCD meals daily from Cambridge Manufacturing (Northants, U.K.); the daily energy intake was 1,883 kJ (450 kcal) for 16 weeks followed by a 2-week period when regular food was gradually reintroduced. Study assessments were performed at the start of VLCD treatment (week 0), twice during the VLCD phase (weeks 8 and 16), and 2 weeks after the end of VLCD treatment (week 18). Anthropometrical measurements, blood pressure recording, blood sampling, oral glucose tolerance test, and a SCAT biopsy were performed at each of the four time points. Computed tomography performed at weeks 0 and 16 was used for adipose tissue area calculations. SCAT biopsy samples were obtained with a syringe with manually applied vacuum under local anesthesia and immediately frozen in liquid nitrogen and stored at −80°C until analysis. Microarray analysis of SCAT biopsies was performed on 24 patients of which 12 subjects (three women and nine men) were obese but healthy and 12 subjects (three women and nine men) were dysmetabolic according to the above criteria. Patient characteristics are shown in Table 2. Biochemical and anthropometrical measurements and examinations were performed as described (9). Computed tomography was used to determine body composition as previously described (11).

The fast-food study included a total of 18 lean subjects who accepted a fast-food diet–induced increase in body weight of 5–15% during a period of 4 weeks as described previously (12). The goal for each individual was to double his or her calorie intake by a diet rich in protein and saturated animal fat combined with a sedentary lifestyle. The participants were free from current diseases as judged by medical checkup and history. Subcutaneous abdominal fat biopsies from six of the subjects were collected at baseline and after 4 weeks (four men, two women; age 24.1 ± 3 years; BMI 21.4 ± 2.5 kg/m²) for microarray analysis. Adipocytes were isolated from the adipose tissue samples by collagenase digestion (collagenase type 1, Worthington, NJ) as described previously (13). Total RNA was extracted using TRIzol (Invitrogen, Carlsbad, CA) and then further purified with RNeasy mini columns (Qiagen, Hilden, Germany). High-quality RNA was confirmed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). Labeled cRNA from reverse transcription of total RNA was fragmented and hybridized to HG-U133 Plus 2.0 arrays according to manufacturers' instructions (Affymetrix, Santa Clara, CA). Fat mass was assessed by dual-energy X-ray absorptiometry at baseline and after 4 weeks. Patient characteristics are shown in Table 3.

VAT biopsies were obtained (0800–1000 h) from six women undergoing elective abdominal surgery [age (mean ± SD): 27.5 ± 7 years, BMI: 23.8 ± 2.8 kg/m²] for adipose tissue explant studies. Exclusion criteria included known cardiovascular disease, thyroid disease, neoplasms, current smoking, diabetes, hypertension (blood pressure >140/90 mmHg), and renal impairment (serum creatinine >120 μmol/l). None of the subjects were on any medications for at least 6 months before the study. The local research ethics committee approved the study, and all patients involved gave their informed consent, in accordance with the guidelines in the Declaration of Helsinki 2000.

The subjects recruited at Aintree Hospital for the adipose tissue depot study had blood taken for analysis of adipokines, glucose (glucose oxidase method, Yellow Springs), insulin (insulin ELISA, BioSource Europe S.A.; sensitivity 0.15 μIU/ml, intra-assay coefficients of variation 2.5–5.3%; human specific and with 0% crossreactivity with proinsulin), and high-sensitivity C-reactive protein (hsCRP, ELISA ImmunDiagnostik AG, Bensheim, Germany; sensitivity, 0.124 ng/ml, intra-assay coefficients of variation 5.0–6.0%). The computer-solved homeostasis model assessment method (14) was used to determine insulin sensitivity and β-cell function. Serum levels were measured for adiponectin with the Quantikine ELISA (R&D Systems, Abingdon, Oxon, U.K.) that has a sensitivity of 0.25 ng/ml and intra-assay coefficients of variation of 2.5–4.7%. Serum leptin was measured by ELISA from DRG Diagnostics (Marburg, Germany); sensitivity 1.0 ng/ml, intra-assay coefficients of variation 3.5–5.0%. All ELISA-based measurements were performed according to manufacturers' instructions.

Adipose tissue was obtained during surgery from subcutaneous and visceral compartments from each patient (paired samples). Tissue was collected immediately and frozen at −70° for further analysis. Total RNA was isolated from frozen adipose tissue samples with the RNeasy lipid tissue kit (Qiagen, Crawley, U.K.). The isolated total RNA was quantified by measurement of absorbency at 260 and 280 nm, and its integrity was verified using agarose gels (1%) stained with ethidium bromide. Reverse transcription to first-strand cDNA was carried out using the ImProm-II Reverse Transcription System (Promega, Madison, WI), according to the manufacturer's instructions. Real-time quantification of adipokines and SPARC mRNA was performed using QuantiTect Multiplex PCR kits (Qiagen), with commercially available prevalidated and assay-on demand primers and Quantiprobes, on a ROTORGENE 2000 analyzer (Corbett Research, Cambridge, U.K.) and compared with the housekeeping gene β-actin. SPARC primers were forward 5′ATTAGGCTGTTGGTTCAAA 3′ and reverse 5′ AGCCCTGGTTCTCCAAAA 3′. The use of a duplex configuration allowed simultaneous amplification of the adipokine gene of interest with the β-actin housekeeping gene. In brief, 1 μl cDNA (50 μg) was added to a mixture of 2× QuantiTec Multiplex PCR Master Mix, 20× Primer Mix for target gene, 20× Quantiprobe for target gene, 10× Assay mix for β-actin gene, and RNAase-free water to a final volume of 25 μl, in different 100-μl PCR tubes and in duplicates. Thermocycling conditions were: 15 min initial activation step followed by 50 cycles of 45 s of denaturation at 94°C, 45 s of annealing at 56°C, and 45 s of extension at 76°C. Data were obtained as threshold cycle (computed tomography) values, which is defined as the fractional cycle number at which the fluorescence reached 10 times the standard deviation of the baseline. The increase in fluorescence was measured in real time and analyzed using the Rotorgene Application Software v6.0 (Build 38). The relative gene expression of adipokines was calculated according to the method of Pfaffl as a relation to β-actin that takes into account the differing amplification efficiencies of each gene during the PCR (15).

Total RNA from human tissues was prepared with Qiagen Lipid tissue kit (Qiagen). Total RNA from adipocytes was prepared with the phenol-chloroform extraction method of Chomczynski and Sacchi (16). After further purification with RNeasy (Qiagen), the RNA concentration was measured spectrophotometrically. The A260/A280 ratio was 1.8–2.0 and the quality of the RNA was verified by agarose gel electrophoresis before reverse transcription into cDNA. Preparation of cRNA and hybridization in the different studies was performed according to standard Affymetrix protocols as previously described (9).

Gene expression was measured using the Human Genome U133A DNA microarray (Affymetrix), composed of 22,283 probe sets representing almost 14,000 expressed genes. Arrays were scanned with a confocal laser scanner (GeneArray scanner G2500A Hewlett Packard, Palo Alto, CA). The Affymetrix probesets 200665_at and for control 212667_at were used for SPARC analyses, and the data from probeset 200665_at were reported.

SPARC gene expression in different human tissues was assessed using DNA microarrays. Duplicate GeneChip HG U133A expression profiles from 17 different tissues were downloaded from the SymAtlas dataset (http://symatlas.gnf.org/SymAtlas/), as previously described (17). In addition, our own expression profiles, originating from small and large adipocytes, were included and normalized, as previously described (17).

Surgical adipose tissue samples were placed into sterile containers containing Medium 19 (Sigma Aldrich, Gillingham, U.K.) for primary adipose tissue culture following a protocol that was a modification of the method described by Fried and Moustaid-Moussa (18). Briefly, 1–3 g of adipose tissue was minced into 5–10 mg (∼1 mm³) fragments, washed with a 230 μm mesh (Filter no. 60; Sigma Aldrich, Gillingham, U.K.) and rinsed with sterile PBS warmed to 37°C. Samples were then transferred to six-well plates (∼50 mg/well) containing 3 ml of Media 199 (Invitrogen, Paisley, U.K.) supplemented with 50 μg/ml gentamicin and 1% FCS (containing insulin at a concentration of 10⁻¹⁴ M) and cultured for 24 h with or without the addition of insulin or d-glucose in a 37°C incubator under an atmosphere of 5% CO₂/95% air.

Protein lysates were prepared by homogenizing adipose tissue in radioimmunoprecipitation lysis buffer (Upstate, Lake Placid, NY) according to manufacturer's instructions. Protein samples (40 μg/lane) containing SDS sample buffer (5 mol/l urea, 0.17 mol/l SDS, 0.4 mol/l dithiothreitol, and 50 mmol/l Tris-HCl, pH 8.0) were subjected to SDS-PAGE (10% resolving gel) and transferred to polyvinylidene difluoride (PVDF) membranes. The PVDF membranes were incubated with primary rabbit–anti-human antibody for SPARC (Abcam, Cambridge, U.K.) [1:500 dilution] or primary rabbit–anti-human antibody for β-actin (Cell Signaling Technology, Beverly, MA) [1:1,000 dilution] overnight at 4°C. The membranes were washed thoroughly for 60 min with TBS-0.1% Tween before incubation with the secondary anti-rabbit horseradish peroxidase–conjugated Ig (Dako, Ely, U.K.) [1:2,000] for 1 h at room temperature. Antibody complexes were visualized using chemiluminescence (ECL+; GE Healthcare, Little Chalfont, U.K.). Band densities were measured using a scanning densitometer coupled to scanning software Scion Image (Scion Corporation, Frederick, MD), and the membranes were reprobed with the β-actin antibody (Cell Signaling Technology; 1:10,000 dilution) to determine equal protein loading.

Data were analyzed using the package SPSS version 15 (SPSS, Chicago, IL). Positively skewed data were log transformed where possible, and then analyzed parametrically, or else nonparametric methods were used. Student's paired t test and Pearson or Spearman correlation coefficients were used as appropriate, and multiple regression analysis was used for correction for dependent variables. For the cell culture comparison, we used group comparison by Friedman's ANOVA and post hoc Dunn's test. Statistical significance was regarded as P < 0.05 (two tailed).

Our data showed that SPARC was expressed in various human tissues but especially in adipocytes where the expression was higher in larger cells (Fig. 1,A). The expression in adipocytes was greater than the expression in stroma cells of adipose tissue (data not shown). When comparing adipose tissue depot expression, SPARC expression was higher in SCAT than VAT (0.81 ± 0.06 vs. 0.50 ± 0.03 signal units, P < 0.001, n = 47, Fig. 1,B) similar to the expression of the adipokine leptin in SCAT (1.53 ± 0.1 vs. 0.6 ± 0.03 signal units, P < 0.001, n = 47, Fig. 1 A).

We did not find a sex difference in expression of BMI-matched men and women. Both SCAT- and VAT-derived SPARC showed a positive correlation with fat mass (Fig. 2) and with waist circumference (Table 4) in the depot study. SPARC was strongly correlated with high-sensitivity C-reactive protein but not with circulating tumor necrosis factor-α (TNF-α) or interleukin-6 (IL-6). SPARC was correlated with local adipose tissue expression of the macrophage-migration inhibitory factor (MMIF; r = 0.3, P < 0.05), but only SCAT expression of SPARC correlated with local IL-6 expression (r = −0.3, P < 0.05). TNF-α, macrophage inflammatory protein-1 (MIP-1), monocyte chemoattractant protein-1 (MCP-1), or regulated upon activation normal T-cell expressed and secreted expression showed no association with SPARC expression (Table 4).

SPARC was associated with fasting insulin levels (SCAT: r = 0.31, P = 0.04; VAT: r = 0.34, P = 0.02) and homeostasis model assessment–insulin resistance (HOMA-IR; SCAT: r = 0.32, P = 0.03; VAT: r = 0.36, P = 0.016). In addition, VAT-SPARC expression but not SCAT-SPARC was correlated with fasting glucose (r = 0.29, P = 0.046, Table 4).

VAT-SPARC correlated with VAT leptin (r = 0.55, P < 0.001) and SCAT-SPARC with SCAT leptin (r = 0.68, P < 0.001); this remained significant after correction for fat mass, waist-to-hip ratio, and HOMA-IR. VAT-SPARC was negatively correlated with serum adiponectin (r = −0.37, P < 0.01) and SCAT-SPARC but not VAT-SPAC negatively correlated with SCAT adiponectin (r = −0.3, P < 0.05; Fig. 2,C, Table 4).

Obese subjects in the VLCD study lost 18 kg in the first 8 weeks and an additional 10 kg by week 16. There was no further weight loss from week 16 to 18. The average SPARC expression in the adipose tissue of these subjects decreased by 33% in the first 8 weeks (P < 0.0001, paired t test). However, despite further weight loss, there was no change in SPARC expression between 8 and 16 weeks (P = 0.74, paired t test). Between 16 and 18 weeks, during which period the subjects went from VLCD to regular food while remaining weight stable, the SPARC expression increased significantly (20% rise, P < 0.0001, paired t test) and remained significantly correlated with insulin, glucose, and leptin levels but not adiponectin levels (data not shown). There was no difference in SCAT-SPARC expression in the group with the metabolic syndrome, but there was a tendency to a further decline of SPARC at week 16 (Fig. 3).

The fast food study subjects gained on average of 7.2 ± 1.6 kg with resulting BMI of 23.7 ± 3.3 kg/m² and fat mass increment of 4.2 ± 5.8 kg. This resulted in a significant increase in fasting insulin by 52%, and HOMA-IR and serum leptin were more than doubled. SPARC-expression in SCAT increased with hyperalimentation in 4 weeks from 3.5 ± 0.8 to 6.2 ± 0.4 signal units (P < 0.05; Table 3), and the increment in SPARC correlated with the amount of weight gain (r = 0.77, P < 0.01).

Cell culture studies were performed to determine whether the observed correlations of SPARC with metabolic parameters are because of direct regulation by glucose, leptin, or insulin. Increasing doses of glucose (5, 10, 20, and 40 mmol/l) decreased SPARC protein expression (5 mmol/l: 100 ± 8 optical density units; 10 mmol/l: 93 ± 6 optical density units; 20 mmol/l: 24 ± 6 optical density units; 40 mmol/l: 19 ± 4 optical density units [SPARC/β-actin], respectively, n = 6; P < 0.01, Fig. 4,A). In contrast, culture of VAT with insulin (0, 0.01, 1, and 100 nmol/l) increased SPARC expression (C: 100 ± 20 optical density units; 0.01 nmol/l: 105 ± 35 optical density units; 1 nmol/l: 156 ± 27 optical density units; 100 nmol/l: 292 ± 30 optical density units, n = 6; P < 0.001; Fig. 4 B).

Similar to insulin, leptin treatment of visceral fat explants showed a dose-dependent increment of SPARC expression (C: 100 ± 20 optical density units; 0.1 nmol/l: 271 ± 30 optical density units; 10 nmol/l: 465 ± 52 optical density units; P < 0.001; Fig. 4 C).

Adipose tissue plays a key role in energy homeostasis both as an energy store and as an endocrine organ. Dysfunction of adipose tissue fat storage leads to fat deposition in other organs as ectopic fat and results in disadvantageous metabolic consequences as most apparent in subjects with familial lipodystrophy (19). The cause of impaired adipose tissue storage capacity is unclear and SPARC that influences matricellular composition could contribute to its development by inhibition of adipocyte differentiation. Disturbance of the 3D extracellular matrix (ECM), for example, increased rigidity by increased collagen 1 content and has previously been shown to compromise in vitro adipocyte differentiation (20), and collagen dysregulation has recently been attributed to metabolic dysregulation (3).

The following attributes make SPARC a likely candidate to limit fat deposition in adipose tissue itself. Although it is an evolutionarily conserved collagen-binding glycoprotein, it does not contribute to the structure of the ECM, but it was recently shown to be a ligand of the integrin receptor α5-β1–integrin (21) and inhibit adipogenesis through stimulation of β-catenin signaling (8), which is part of the Wnt pathway that enhances osteoblastogenesis alongside the inhibition of adipogenesis (22). SPARC null mice have a phenotype marked by an increased subcutaneous fat deposition, reduction in collagen 1 in SPARC-null fat, adipocytes of higher diameters, and fat pads with an increase in adipocyte number (23).

The results of our study showed an increase of SPARC expression with increased fat mass that is consistent with previous reports showing a higher expression of SPARC in rodent obesity (24) and elevated SPARC levels in obese subjects (25). In addition, this is the first study to highlight the depot-specific expression of SPARC in humans. SPARC/osteonectin is expressed in various human tissues with particularly high expression in subcutaneous abdominal fat where is appears to be secreted primarily by adipocytes in comparison to the adipose tissue stromal fraction that is consistent with previous findings (2). Subjects with familial lipodystrophy lack SCAT, and the metabolic consequences could be attributed to the limitation of subcutaneous tissue expansion to which SPARC as an inhibitor of adipogenesis (8) with higher expression in SCAT may contribute.

This is the first study to assess metabolic parameters in combination with adipose tissue SPARC expression. We found a positive correlation with fasting insulin, fasting glucose, and HOMA-IR and waist circumference as well as hsCRP for subcutaneous and visceral depots. However, we did not find a correlation of SPARC with fasting lipids or blood pressure, which may in part be masked by treatment of subjects with hypercholesterolemia and hypertension (not shown).

Higami et al. (26) examined genes downregulated by energy restriction in epididymal fat in mice, and SPARC was one of the prominently affected genes lowered by long-term energy restriction. Consistent with this we have found a strong downregulation of SPARC with weight loss after 8 weeks of a VLCD and an increment of SPARC with weight gain. Although HOMA-IR improved with weight loss in our VLCD study population, and SCAT-SPARC expression correlated with fasting insulin and HOMA-IR, the SPARC expression was no different in the BMI-matched groups with and without the metabolic syndrome, which is explained and consistent with the lack of correlation of SPARC with blood pressure and lipids. Leptin is a strong predictor of SPARC in the depot study as well as the VLCD study in which it is independent of the BMI at baseline. SPARC remains correlated with leptin expression but loses its association with weight loss after 8 weeks, after which subjects continued to lose weight. The strong correlation with insulin, glucose, and leptin in the depot and the VLCD study prompted us to study and confirm their regulation by further in vitro studies, while adiponectin showed an inverse correlation with SPARC expression in SCAT but not VAT and thus appeared less likely to be because of a direct effect.

Culture of visceral explants showed that glucose lowered and insulin and leptin increased SPARC expression that apart from the glucose correlation is consistent with the findings in the clinical studies. However, circulating glucose in vivo showed a positive correlation with SPARC-AT expression that contradicts the in vitro study in which supraphysiological glucose levels lower SPARC. The lack of influence of circulating glucose levels may be explained by the insulin resistance of adipocytes for glucose transport in obese individuals. In diabetic subjects, we would expect the glucose levels required to lower SPARC to be supraphysiological as found in marked hyperglycemia that is usually avoided by antidiabetic treatment. Although we show that SPARC is regulated by leptin, SPARC in turn suppresses leptin gene expression in mouse preadipocytes (21), which may explain the high leptin levels observed in SPARC knockout mice (23).

It was suggested that SPARC may be involved in inflammatory processes (27,28), but data from our study show only a correlation of adipose tissue SPARC expression with the local expression of the MMIF-1—a proinflammatory adipocytokine (29). An association of SPARC with MMIF was to our knowledge only reported in connection with malignant melanoma cells that showed overexpression of both (30). Thus, the interaction of SPARC and MMIF remains to be confirmed but may explain a potential association of SPARC with diabetes with MMIF being associated with the pathogenesis of type 1 and type 2 diabetes and latter among others through a decrease in insulin signal transduction (31).

Much pathology related to diabetes and obesity has been linked with SPARC. An increased SPARC expression was found in rodent models of diabetic nephropathy (32), and SPARC null mice were shown to be protected from renal fibrosis (27). Increased circulating levels of SPARC have been observed in subjects with cardiovascular disease (25). SPARC is also expressed in human retinal endothelial cells (33), and with its close interaction with VEGF and PAI-1 (34,35) it may further the progression of diabetic retinopathy. Furthermore, SPARC is linked with tumorigenesis and appears to favor certain tumors (36), although its exact role in tumor development is controversial (37) and it remains to be shown whether SPARC could contribute to the association of obesity and some cancers (38).

Apart from focusing on the potential advantages of lowering SPARC levels, the inducers of SPARC excess such as hyperleptinemia in subjects with leptin resistance and hyperinsulinemia in type 2 diabetes require similar attention. Leptin is a proinflammatory molecule, and it has been suggested that leptin is permissive in the pathogenesis of liver fibrosis as shown by protection of leptin-deficient mice from fibrosis during steatohepatitis or in response to chronic toxic liver injury (39). As such, hyperleptinemia in adipose tissue may through upregulation of SPARC also induce adipose tissue fibrosis that requires further study.

In summary, our study showed that SPARC-AT expression is related to human metabolism and SPARC expression that is predominant in SCAT is upregulated by insulin and leptin. Together with the previous reported consequences of SPARC in other tissues, a role for SPARC in the development of obesity- and diabetes-related complications is likely. Further research is required to show whether increased adipose tissue SPARC limits the expansion of normal adipose tissue in response to energy excess and promotes ectopic fat deposition and associated metabolic dysfunction.

This work was supported by Diabetes UK, Swedish Research Council (11285), University Hospital of Linkoping Research Funds; Diabetes Research Centre of Linkoping University; and the Gamla Tjaenarinnor Foundation.

No potential conflicts of interest relevant to this article were reported.

Parts of this study were presented in abstract form at the 69th Scientific Sessions of the American Diabetes Association, New Orleans, Louisiana, 5–9 June 2009.