Sexually Dimorphic Diet-Induced Insulin Resistance in Obese Tissue Inhibitor of Metalloproteinase-2 (TIMP-2)-Deficient Mice

Animals

All procedures were in accordance with approved University of Vermont Animal Care and Use Committee protocols. TIMP-2 KO mice (22) (after 10 back-crosses with Charles Rivers' C57Bl/6 mice) and wild-type (WT) controls were maintained as congenic homozygous breeders. At 2 months of age, littermates were singly housed and either remained on the standard chow diet (ProLab RMH 3000, 4.1 kcal/g with 26% protein, 60% carbohydrate, 14% fat; Purina LabDiet, Scotts Distributing Inc., Hudson, NH) or switched to a HFD (5.24 kcal/g with 20% protein; 20% carbohydrate; and 60% fat, which is 37.1% saturated, 46.0% monounsaturated, and 16.9% polyunsaturated; Research Diets, Inc., New Brunswick, NJ) for 3 months. In both cases, mice had ad libitum access to food and water. Food consumption and animal weight were measured daily within 1 h of the light (sleep) cycle. At the termination of the study, the mice were humanely killed via exsanguination, tissues were rapidly removed and frozen at −80 C for quantitative real-time PCR and Western blot analyses or immersion fixed for immunohistochemical analysis.

Serological testing

Fed glucose and plasma insulin levels were monitored from a small nick in the tail vein at 1000 h. Fasting commenced at 1800 h at the beginning of the dark (feeding) cycle, the next day mice were weighed, and their fasting blood glucose and insulin levels were monitored at 1000 h. The mice were injected with 2 g/kg body weight of glucose [ip glucose tolerance test (IPGTT)] or 0.75 mU/kg of human insulin (Eli Lilly) [ip insulin tolerance test (IPITT)] in sterile saline after 55 and 78 d, respectively. Blood glucose values were monitored using a standard glucometer (LifeScan One Touch Ultra, Milpitas, CA) at 0, 5, 15, 30, 60, 90, and 120 min after injection. To monitor insulin levels, plasma was collected 0, 30, and 120 min after glucose injection.

Insulin and leptin enzyme immunoassay

Plasma insulin and leptin levels were determined by enzyme immunoassay according to manufacturer's instructions (Alpco Diagnostics, Salem, NH). Optical density readings were converted to ng/ml or pg/ml, respectively, using kit provided standards and Prism software (GraphPad, San Diego, CA). Correlation of leptin concentration to body weight was determined using Microsoft Excel.

β-Cell morphometry

β-Cell mass and proliferation were determined as described (23). For β-cell mass, 5-μm-thick paraffin sections were immunostained with guinea pig antiinsulin primary antibody (1:1000, AB3340; Millipore, Billerica, MA). Only islets more than five-cell diameters were included for islet β-cell surface area measurements. The proportion of islet β-cell surface area vs. whole pancreas surface area was determined planimetrically by digital imaging. β-Cell mass was determined by multiplying the average β-cell surface area per animal by their pancreatic weight. The relative size frequency of β-cell clusters, categorized into one of six classes, was analyzed with National Institutes of Health ImageJ using the same sections used for β-cell mass measurements. For β-cell proliferation, 5-μm-thick paraffin sections were immunostained with mouse antihuman Ki-67 (1:500, no. 610968; BD Biosciences, San Jose, CA), then biotin-conjugated donkey antimouse (Jackson ImmunoResearch, West Grove, PA), and incubation with streptavidin-horseradish peroxidase (1:300; Zymed, San Francisco, CA). After development with 3,3'-diaminobenzidine tetrahydrochloride/H₂O₂, sections were immunostained for insulin, as described above, and finally counterstained with hematoxylin. The number of Ki-67-positive nuclei per 1000–1500 islet β-cells was counted for each animal.

Islet isolation

Pancreatic islets were isolated by collagenase digestion, separated by histopaque density gradient centrifugation, and hand picked under a stereomicroscope as described (24).

Western blot analysis

Western blot analysis was performed as previously described (25). Protein expression in isolated islets was determined using rabbit antihuman TIMP-2 (1:1500, AB801; Millipore), rabbit antirat glucose transporter 2 (GLUT2) (1:1000, no. 07-1402; Millipore), mouse antirat pancreatic and duodenal homeobox 1 (Pdx1) (1:1000, Developmental Studies Hybridoma Bank clone F6A11), and normalization to mouse antirat β-tubulin (1:2000, T4026; Sigma, St. Louis, MO). Protein expression in sc adipose tissue was first assessed using rabbit antihuman MMP-14 (1:2500, Ab38971; Abcam, Cambridge, MA). Immunocomplexes were visualized by enhanced chemiluminescence (PerkinElmer Life Sciences, Boston, MA), stripped from the membrane by incubation with Restore Plus Western Blot Stripping buffer (Pierce, Rockford, IL) and stripping solution [0.2 m glycine and 0.5 m NaCl (pH 2.5)] for 10 min each followed by 15-min washes in Tris-buffered saline and Tris-buffered saline with Tween 20. Blots were then incubated with goat antimouse CD31/platelet/endothelial cell adhesion molecule 1 (PECAM-1) (1:250; sc-1506; Santa Cruz Biotechnology, Inc., Santa Cruz, CA) and immunocomplexes detected as before. Finally, after stripping as above, blots were reprobed with goat antihuman actin (1:1000, sc-1616; Santa Cruz Biotechnology, Inc.). Densitometry was performed using Quantity One software (Bio-Rad, Hercules, CA).

Microscopy

Confocal immunofluorescence imaging with guinea pig antiglucagon (1:1000, no. 4031-01F; Millipore), rabbit antirat GLUT2 (1:1000, no. 07-1402; Millipore), mouse antirat Pdx1 (1:1000, Developmental Studies Hybridoma Bank clone F6A11), rabbit antihuman TIMP-2 (1:200, no. 9013-2609; Biogenesis, Kingston, NH), rat antimouse Mac-2 (1:250; CedarLane Laboratories Ltd, Hornby, Ontario, CA), and rabbit antihuman MMP-14 (1:500, Ab38971; Abcam) was performed as described (23).

Adipose tissue morphology was revealed by hematoxylin and eosin staining using standard histological techniques. Differential collagen staining was revealed by Sirius Red staining and polarization microscopy (26).

Adipose weight and adipocyte size determination

Four main fat depots with known discrete anatomical locations were collected: 1) posterior sc depot, located at the base of the hind legs, indicated as sc; 2) perigonadal depots, enveloped and bound to the epididymis by the peritoneal leaflets in males or surrounding ovaries, uterus, and bladder in females, indicated as epi or gonad; 3) three visceral fat depots located in the abdominal cavities, including ip, mesenteric, and omental fat, combined as visceral fat, indicated as visc; and 4) brown adipose tissue from the interscapular area, indicated as brown. Tissues were weighed and snap frozen in liquid nitrogen for RNA and protein extractions.

Adipocyte size was determined on 8-μm-thick hematoxylin stained sections. Between 150 and 1000 adipocyte diameters were measured in 8–10 random high-power fields from at least two noncontiguous tissue sections per sample using MetaMorph software (Molecular Devices, Sunnyvale, CA).

Quantitative real-time PCR

Total RNA from adipose tissue was extracted using an RNeasy Mini kit (QIAGEN, Valencia, CA) and cDNA synthesized using the Advantage RT for PCR kit (CLONTECH; Palo Alto, CA) following manufacturer's instructions. Quantitative real-time PCR with cDNA corresponding to 20 ng total RNA, 10 μl of 2× TaqMan Universal PCR Master Mix, and 1 μl of TaqMan assay in a 20 μl reaction volume was performed on a 7300 Real-Time PCR System (Applied Biosystems, Foster City, CA) using standard cycling conditions. Relative gene expression was determined using standard curves. TaqMan Gene Expression Assays (Applied Biosystems) were used to measure the transcription levels of TIMP-2 (Mm00441825_m1), adiponectin (Mm01343606_m1), leptin (Mm00434759_m1), epidermal growth factor module-containing mucin-like hormone receptor 1 (EMR1) (Mm00802530_m1), monocyte chemotactic protein 1 (MCP-1) (Mm00441242_m1), IL-6 (Mm00446191_m1), and TNFα (Mm00443258_m1). Gene expression was normalized to hypoxanthine-guanine phosphoribosyltransferase 1 (Mm01545399_m1).

Statistical analysis

Data are presented as mean ± sem. Significant interactions were identified by two-way ANOVA with Bonferroni's multiple comparison tests or unpaired Student's t test using Prism software (GraphPad). Statistical significance was assigned at probability of P < 0.05.

TIMP-2 KO mice become obese independent of diet composition

TIMP-2 KO mice develop age-dependent obesity. At postnatal d 21, TIMP-2 KO mice of both sexes had weights comparable with WT mice. However, by 3 months of age, TIMP-2 KO mice exhibited significantly greater body weights than WT mice, and weight continued to increase unabated to 1 yr of age (Jaworski, D. M., unpublished observation). Because mice were maintained as homozygous breeders and perturbations in maternal nutritional status (e.g. undernutrition, obesity, and diabetes) can irreversibly alter structures responsible for the control of ingestive behavior and energy expenditure (27), weight gain in offspring from heterozygous pairings was determined. Even with the small sample size (n = 3), due to reduced frequency of sex-matched WT and TIMP-2 KO littermates, TIMP-2 KO mice weighed more than their WT littermates at 3 months of age (M: WT, 25.45 ± 0.33 g; KO, 29.50 ± 0.69 g; P = 0.006; and F: WT, 19.14 ± 0.25 g; KO, 23.01 ± 0.37 g; P = 0.001), suggesting that altered gestational metabolic programming is not the basis for TIMP-2 KO obesity. Thus, studies were undertaken to determine the basis for excessive weight gain in the absence of TIMP-2.

To determine the effect of diet on TIMP-2 KO obesity, 8-wk-old (i.e. before the onset of TIMP-2 KO obesity; M: WT, 22.48 ± 0.84 g; KO, 24.02 ± 0.38 g; P = 0.13; and F: WT, 18.75 ± 0.74 g; KO, 19.91 ± 0.48 g; P = 0.21) WT and TIMP-2 KO mice were either maintained on a standard chow diet (i.e. 4.1 kcal/g) or switched to a HFD containing 60% fat (i.e. 5.24 kcal/g), and their body weights were monitored daily for 3 months (Fig. 1A). Corroborating our previous observations, chow-fed TIMP-2 KO mice of both sexes gained more weight than WT mice over the 3-month study, and female TIMP-2 KO mice weighed as much as male WT mice at the study termination. Weight gain was exacerbated in HFD-fed TIMP-2 KO mice and was associated with greater adiposity (Fig. 1A and also see Fig. 5B). TIMP-2 KO mice maintained on the HFD for 5 months (e.g. ∼60 g male weight) failed to show plateau of weight gain (data not shown), suggesting a defect in central recognition of and compensation to adequate energy stores.

A parsimonious explanation for TIMP-2 KO mouse obesity is a failure to regulate caloric intake; thus, food consumption was measured (Fig. 1, B–D). To most accurately assess food intake, mice need to be individually housed (Fig. 1, B–D). Because social isolation is a stressor, the mean chow consumption of group housed mice was monitored daily for 5 d before individual housing. TIMP-2 KO mice of both sexes were hyperphagic (i.e. 8.37 kcal/d excess in males and 3.36 kcal/d excess in females) even before obesity onset. Two-way ANOVA revealed that regardless of diet, mean food consumption of individually housed TIMP-2 KO mice of both sexes was significantly greater than similarly housed WT mice over the 3-month study (Fig. 1B). Hyperphagia was more pronounced in chow-fed male than female TIMP-2 KO mice (Fig. 1C). HFD-fed WT mice of both sexes showed little daily variability in food intake (Fig. 1D). In contrast, HFD-fed TIMP-2 KO mice displayed dramatic increase in food intake upon cage change (males) or upon individual housing (females) (Fig. 1D). Taken together, our data demonstrate that, regardless of diet, TIMP-2 KO mice were hyperphagic and did not adjust caloric intake concomitant with increased adiposity or consumption of an energy-rich diet.

Sexually dimorphic insulin resistance in TIMP-2 KO mice

To assess the physiological effects of TIMP-2 KO obesity, plasma glucose, insulin, and leptin levels were measured at the study start and termination (Fig. 2). Although a significant sex effect was present at the study start (i.e. decreased fed and fasting glucose and increased insulin and leptin levels in female mice; P = 0.02, P = 0.039, P = 0.0005, and P < 0.0001, respectively), no genotype effect was present, suggesting that TIMP-2 KO mice were not overtly diabetic (Fig. 2A). Furthermore, although TIMP-2 KO mice were hyperphagic at the study start, this was not due to reduced leptin levels. Similar results were obtained in chow-fed mice at the study termination (Fig. 2B). Therefore, although chow-fed TIMP-2 KO mice were obese, they did not exhibit diabetes. In contrast, HFD-fed TIMP-2 KO mice at the study termination were hyperglycemic, hyperinsulinemic, and hyperleptinemic (Fig. 2C). Although hyperleptinemic, the strong correlation (M, R² = 0.85 and F, R² = 0.89) with body weight suggests that the increased plasma leptin in TIMP-2 KO mice is proportionate to their greater adiposity (see Fig. 5B). Similar results were obtained when normalized to net adipose weight. Inasmuch as chow-fed TIMP-2 KO mice appear to be protected from the adverse metabolic consequences of their obesity, diet quality, and not obesity per se, is a key factor in the development of their diabetes.

To further characterize altered energy homeostasis in HFD-fed TIMP-2 KO mice, we next measured glucose tolerance and insulin sensitivity by IPGTT and IPITT, respectively (Fig. 3). There was no significant difference in fasting glucose, and both genotypes and sexes of chow-fed mice displayed comparable responses to glucose administration (Fig. 3A). Although chow-fed male TIMP-2 KO mice did not exhibit insulin resistance (Fig. 3E) or enhanced β-cell mass (Fig. 4A), their elevated postprandial glucose levels (Fig. 2B) and fasting insulin levels (Fig. 3C) suggest a mild β-cell defect or a defect in hepatic glucose production but not severe enough to induce hepatic insulin resistance. These data also suggest that these mice may be in a progressive prediabetic phase. In contrast, HFD-fed TIMP-2 KO mice had increased fasting glucose levels and showed impaired glucose tolerance that was more pronounced in male than female TIMP-2 KO mice (Fig. 3B). In fact, the glucose values for male HFD-fed TIMP-2 KO mice are underestimations, because many readings exceeded the glucometer maximum of 600 mg/dl. Although displaying normal glucose tolerance, chow-fed male TIMP-2 KO mice exhibited increased fasting insulin levels that remained elevated during the IPGTT (Fig. 3C). However, values were at least 2-fold lower than male HFD-fed TIMP-2 KO mice (Fig. 3D). Both fasting and glucose-stimulated insulin levels were increased in HFD-fed TIMP-2 KO mice of both sexes, possibly due to hepatic insulin resistance and a compensatory increase in insulin secretion (Fig. 3D). Although chow-fed WT and TIMP-2 KO mice displayed comparable insulin sensitivity (Fig. 3E), male HFD-fed TIMP-2 KO mice were insulin resistant (Fig. 3F). The increased glucose levels in male and female TIMP-2 KO mice at 15 min after injection is likely due to stress-mediated glucose release, because TIMP-2 KO mice exhibit increased anxiety (28). Although female HFD-fed TIMP-2 KO mice were obese and moderately glucose intolerant, their peripheral tissues were insulin sensitive, suggesting a diet-dependent, male-specific defect in glucose utilization and insulin responsiveness.

Enhanced β-cell mass and proliferation in male HFD-fed TIMP-2 KO mice

Pancreatic β-cell mass dynamically changes to maintain euglycemia in response to altered metabolic demand and can be influenced by β-cell size (hypertrophy or atrophy), β-cell renewal (proliferation and differentiation), and β-cell death (29). Because only male TIMP-2 KO mice displayed insulin resistance (Fig. 3F), pancreatic physiology was only examined in male mice (Fig. 4). β-Cell mass was significantly increased in HFD-fed TIMP-2 KO mice (Fig. 4A). Nonetheless, relative islet size and number in both chow- and HFD-fed WT and TIMP-2 KO mice were comparable (P = 0.32) (data not shown), suggesting that the mass augmentation was not due to islet hypertrophy or enhanced neogenesis. In contrast, β-cell proliferation was significantly affected by diet and genotype (Fig. 4B) and most likely contributed to enhanced β-cell mass (Fig. 4A). Expression of the β-cell transcriptional regulator Pdx1 was decreased in TIMP-2 KO males (Fig. 4, C and E). Insulin and the glucose transporter GLUT2 were similarly decreased (Fig. 4, D and E). Taken together, these data suggest reduced insulin content and β-cell function in HFD-fed TIMP-2 KO mice. However, given that TIMP-2 expression was unaffected by HFD in isolated WT islets or islets in vivo (Fig. 4F), it suggests that the β-cell alterations observed in TIMP-2 KO mice may be secondary to diet-induced lipotoxicity and not directly to TIMP-2 deficiency.

Expression of inflammatory markers is enhanced in male TIMP-2 KO adipose tissue

Next, the effect of TIMP-2 deficiency on adipose tissue was examined (Fig. 5). In contrast to the pancreas, HFD consumption induced a significant reduction in TIMP-2 expression in adipose tissue but only in male WT mice (Fig. 5A). Furthermore, TIMP-2 displayed distinct sex-specific expression in adipose stores (i.e. most abundant in sc adipose tissue in males with comparable expression in all white adipose depots in females). Although body mass was increased in chow-fed TIMP-2 KO mice (Fig. 1A), when adipose tissue was meticulously removed and weighed, no difference in percentage body fat was present in chow-fed mice (Fig. 5B). In contrast, adipose tissue mass was significantly increased in HFD-fed TIMP-2 KO mice of both sexes. Moreover, gonadal adipose mass was increased in females, and it was decreased in males, suggesting a redistribution of adipose depots. In contrast to pancreatic β-cells, adipose mass augmentation was associated with hypertrophic sc adipocytes in both chow-fed (WT, 17.79 ± 6.73 μm; KO, 20.46 ± 8.31 μm; P < 0.0001) and HFD-fed (WT, 26.49 ± 12.63 μm; KO, 35.57 ± 14.72 μm; P < 0.0001) TIMP-2 KO mice (Fig. 5, C and D). Given that adipogenesis is tightly associated with angiogenesis (30) and TIMP-2 exerts antiangiogenic effects (31), we examined whether TIMP-2 KO mice exhibited increased vascularization (Fig. 5E). HFD increased vascularization (as revealed by CD31/PECAM-1 expression) in male, but not female, mice, but no genotype difference was observed. Finally, the transcriptional regulation of several adipose-derived bioactive molecules was examined (Fig. 5F). Although adipokine (e.g. adiponectin and leptin) expression was only regulated in sc adipose tissue, expression of macrophage markers (e.g. EMR1 and MCP-1) and proinflammatory cytokines (e.g. IL-6 and TNFα) were up-regulated in both sc and epididymal adipose tissue of HFD-fed TIMP-2 KO males. Immunohistochemical examination revealed increased macrophage infiltration in HFD-fed male sc adipose tissue, which was greater in TIMP-2 KO mice (Fig. 5G). Thus, these data support an enhanced inflammatory state in adipose tissue of HFD-fed TIMP-2 KO males.

Contribution of membrane type 1 (MT1)-MMP-mediated collagenolysis to the TIMP-2 KO hyperlipotrophic phenotype

Inasmuch as numerous MMPs are transcriptionally regulated by inflammatory cytokines (32), MT1-MMP (MMP-14) expression is increased in TIMP-2 KO heart (33), lung (22), and myoblasts (34), and MT1-MMP deficient mice (13) and adipocytes (12) display a lipodystrophic phenotype, we examined whether, by extension, increased MT1-MMP expression in adipose tissue could contribute to the hyperlipotrophic TIMP-2 KO phenotype (Fig. 6). MT1-MMP expression (i.e. pro-MT1-MMP) and activity (i.e. active MT1-MMP) were comparable in the sc adipose tissue of chow-fed WT and TIMP-2 KO mice of both sexes (Fig. 6, A and B). Interestingly, HFD increased MT1-MMP expression and activity in male, but not female, mice (Fig. 6, A and B). More strikingly, MT1-MMP activity was increased a greater extent in TIMP-2 KO mice (Fig. 6A) and was associated with decreased Sirius Red staining (Fig. 6C), indicative of increased collagenolysis. Collectively, our data support a contribution of MT1-MMP and its primary inhibitor, TIMP-2, to increased collagen turnover during diet-induced adipogenesis.