Perforin Is a Novel Immune Regulator of Obesity-Related Insulin Resistance

Obesity-associated inflammation of visceral adipose tissue (VAT) and associated chronic release of proinflammatory cytokines and adipokines are major contributors to the development of insulin resistance (1–3). During obesity, increased activity of proinflammatory immune effector cells overwhelms anti-inflammatory regulators, leading to impaired insulin signaling and insulin resistance. Obesity-related insulin resistance is associated with a dramatic influx of macrophages into VAT (4,5) and a shift in their activation state from an anti-inflammatory or “M2-polarized” state toward a proinflammatory phenotype or “M1 polarized” (6,7). Recently, other immune cell types have also been implicated in obesity-associated inflammation of VAT, including mast cells (8), neutrophils (9,10), CD4⁺ Th1 (11), CD4⁺ Th17 (12), CD8⁺ T cells (13), and B cells (14).

The T-cell receptor repertoire in VAT becomes restricted during high-fat diet (HFD) feeding (11,15), suggesting a potential role for antigen-specific T-cell immunity in inflamed VAT. Furthermore, interferon γ (IFN-γ)–secreting CD8⁺ T cells increase in VAT starting at 2 weeks of HFD prior to the accumulation of adipose tissue macrophages, and CD8⁺ T cell–deficient mice fed HFD have improved insulin sensitivity and VAT inflammatory microenvironment (13). Adoptive transfer of CD8⁺ T cells into CD8-deficient mice reverted these effects and increased the numbers of M1 polarized in VAT (13). Activated CD8⁺ T cells in VAT produce increased amounts of proinflammatory cytokines such as IFN-γ and G-CSF, associated with HFD feeding (15). Despite evidence of increased numbers and functional activity of pathogenic CD8⁺ T cells, the mechanisms by which T-cell homeostasis is regulated in inflamed VAT are largely unknown. Moreover, it is unclear whether adipocyte cell death in VAT, manifested as a dying cell surrounded by macrophages (the so-called “crown-like structures [CLS]”), is dependent on CD8⁺ T cell–mediated killing mechanisms.

Perforin is a cytotoxic effector molecule primarily released by CD8⁺ T cells and natural killer (NK) cells to eliminate infected or dangerous cells via the perforin-granzyme cell death pathway (16). Perforin-mediated cytotoxicity is also involved in the homeostatic regulation of CD4⁺ and CD8⁺ T cells in vivo, in a physiologic function distinct from pathogen clearance (17). Interestingly, expression of perforin and associated granzymes is increased in VAT of HFD-fed mice, likely related to increased CD8⁺ T-cell activation (15). In humans, impaired perforin-dependent cytotoxic function leads to excessive T-cell activation, severe hyperinflammation, and the often fatal disorder hemophagocytic lymphohistiocytosis (18,19). Indeed, perforin-deficient mice are susceptible to increases in CD8⁺ T-cell activation and IFN-γ production in response to distinct stimuli (20). Notably, this heightened immune activation seems to be mediated via signaling of the T-cell receptor and results from increased antigen-specific stimulation (20). Perforin-mediated regulation of CD8⁺ T-cell activation and expansion may also include a regulatory feedback on dendritic cells, which prime or initiate T-cell responses during an immune response (21,22).

Since obesity-mediated insulin resistance is associated with increased perforin expression, increased inflammation, and antigen-specific immunity in VAT, we investigated the effects of perforin deficiency in diet-induced obesity. Here, we show that mice lacking perforin (Prf1^null) placed on HFD develop worsened obesity, glucose tolerance, and insulin resistance when compared with wild-type (WT) controls. These changes are accompanied by dysregulation of T-cell homeostasis in VAT with increases in T-cell number and cytokine production, local proliferation, and reduced T-cell apoptosis. Interestingly, the ability to form CLS in VAT, a hallmark of adipocyte cell death, was not reduced in Prf1^null mice, suggesting that perforin is not necessary for the formation of CLS in VAT. These findings suggest that a dominant role of perforin in obesity-associated insulin resistance is to restrict T-cell expansion and activation in inflamed VAT, providing further evidence of a critical role for adaptive immune cells in governing obesity and glucose homeostasis.

Age-matched C57BL/6 (Jax 664), C57BL/6 Prf1^null mice (Jax 2407, derived using a BL/6 embryonic stem cell line), and C57BL/6 CD8^null mice (Jax 2665, backcrossed to C57BL/6 for at least 13 generations) were purchased from The Jackson Laboratory. NK^null mice were a gift of Dr. Tak Mak (University Health Network), and analysis of these mice involved littermate controls. Mice were maintained in a pathogen-free, temperature-controlled, and 12-h light and dark cycle environment. Mice received either a normal control diet (NCD) or HFD (60 kcal % fat; Research Diets) beginning at 6 weeks of age. All mice were males and age matched between groups. All experiments were approved by the Institutional Animal Care and Use Committee of University Health Network.

Total RNA was isolated from epididymal fat pads using the RNeasy Plus Mini Kit (Qiagen) and reverse transcribed using random primers by M-MLV (Invitrogen). RT-PCR was performed using SYBR Green master mix (Applied Biosystems) to determine the amounts of fatty acid binding protein 4 (FABP4), CCAAT/enhancer-binding protein-α (CEBP-α), peroxisome proliferator-activated receptor-γ (PPAR-γ), and sterol regulatory element–binding protein (SREBP) mRNA. Each sample was run in triplicate, and the level of gene expression was normalized to the housekeeping gene 18s. Primer sets are listed in Supplementary Table 1. Changes in gene expression were calculated by the 2(−ΔΔC(T)) method using mean cycle threshold (CT) values. Fold changes in gene expression were assessed by calculating the difference between the CT values of 18s and that of the target gene (ΔCT). Expression of 18s did not differ between WT and Prf1^null mice (P > 0.20). The mean ΔCT were subtracted from individual ΔCT values to obtain ΔΔCT. Fold changes in gene expression were calculated using the equation 2^−ΔΔCT.

Mice were placed in automated metabolic cages for 48 h. Airflow was held at 0.5 L/min, and metabolic activity was assessed using indirect calorimetry recording maximal O₂ consumption (VO₂), CO₂ production (VCO₂), and heat production normalized to body weight. Respiratory exchange ratio was calculated as VCO₂/VO₂. Ambulatory activity was assessed by the breaking of infrared laser beams in the x-y plane. Energy expenditure was adjusted for body mass using ANCOVA (23). Data are the average between light and dark measurements.

Glucose tolerance tests (GTTs), insulin tolerance tests (ITTs), and measurements of serum insulin and fat cell diameter were performed as previously described (11). Mice were fasted overnight for GTTs and fasting insulin. For GTTs, we administered 1 g ⋅ kg⁻¹ of d-glucose intraperitoneally and measured blood glucose levels at indicated time points. For ITTs, mice were fasted for 6 h, insulin was administered intraperitoneally at 1.5 units ⋅ kg⁻¹, and blood glucose concentrations were determined at indicated time points. VAT was stained with hematoxylin-eosin, and CLS, defined as adipocytes enveloped by several macrophages (24), were counted by two blinded observers (D.A.W. and S.W.). Liver triglyceride concentrations were determined using a coupled enzyme assay according to the manufacturer’s recommendations (Sigma-Aldrich).

Mice were fasted overnight and injected intraperitoneally with 1.5 units ⋅ kg⁻¹ of insulin. VAT, muscle, and livers were harvested 15 min after acute insulin injection and snap frozen in liquid nitrogen. Tissues were lysed on ice using the complete radioimmunoprecipitation assay buffer (Santa Cruz Biotechnology), and their protein concentration was determined using the Pierce BCA Protein Assay (Thermo Scientific). Protein lysates were separated by SDS-PAGE and immunoblotted with antibodies (Cell Signaling Technology) against total and Ser473-phosphorylated Akt (pAkt), and GAPDH. The blotted bands were visualized with a MicroChemi chemiluminescent imaging system (FroggaBio) and quantitated using the GelQuant analysis software (FroggaBio).

VAT immune cells were isolated as previously described (11). For cytokine measurements, 3.0 × 10⁵ VAT stromal vascular cells (SVCs) were cultured for 12–16 h in RPMI with 10% FCS and penicillin-streptomycin antibiotics. Bone marrow (BM) cells were obtained by flushing one femur and one tibia with PBS using a 26-gauge needle, dissociated into a single-cell suspension, hemolyzed, and counted. To purify immune cells from spleens and axillary lymph nodes (LNs), we mechanically dissociated the organs on 70-μm nylon cell strainers followed by red blood cell lysis and cell washing.

Cytokines in the serum and supernatants of SVC VAT cultures were analyzed using the Bio-Plex Pro Mouse Cytokine Luminex assay (Bio-Rad). Serum adipokines and cytokines were also measured by ELISA (TNF-α; Biolegend) as well as through a multiplex ELISA screening kit (Signosis), according to the manufacturers' instructions. Serum cytokines are indicated in pg/mL, or shown as a ratio of Prf1^null to WT levels for the SVC VAT cultures.

Cells were stained for 30 min with fluorophore-conjugated antibodies to CD3 (145-2C11), CD4 (GK1.5), CD8 (53-6.7), CD45.2 (104), CD44 (1M7), CD62L (MEL-14), CD69 (H1.2F3), CD107a (1D4B), CD11b (M1/70), CD11c (N418), F4/80 (BM8), Gr-1 (RB6–8C5), NK1.1 (PK136), IFN-γ (XMG1.2), and interleukin 17 (IL-17) (TC11-18H10.1) using recommended dilutions (Biolegend). For intracellular staining, T and NK cells were incubated with a cell stimulation cocktail and Golgistop (eBioscience) for 5 h. Phycoerithrin-conjugated antiperforin and isotype control antibodies were purchased from eBioscience. Annexin V staining kit and propidium iodide (eBioscience) were used for the detection of early apoptotic CD8⁺ T cells harvested from the spleen and VAT. Splenocytes and VAT cells were left untreated whereas thymocytes were incubated with 1 μmol/L dexamethasone at 37°C for 4 h as a positive control. Data were acquired on a Fortessa flow cytometer (BD Biosciences) and analyzed with FlowJo software (Tree Star).

Splenic CD8⁺ T cells from WT and Prf1^null mice were purified using negative selection (>95% purity, EasySep; StemCell Technologies) and injected intraperitoneally (5 × 10⁶) in 16-week-old, HFD-fed CD8^null mice.

Mice were injected intraperitoneally with 4 μg/g body weight of EdU (Invitrogen) and their VAT, spleens, BM, and LNs collected 20 h later and processed into single-cell suspensions. Intracellular EdU was labeled with Alexa Fluor 488 in a Click-iT reaction (Invitrogen) according to the manufacturer’s protocol and measured by flow cytometry.

Statistical difference between two means was determined by two-sided unpaired Student t tests. In figure legends involving multiple experiments from pooled animal tissue, the number of experiments is listed, followed by the number of mice. Data are presented as means ± SEM. Statistical significance was set at <0.05.

To determine a potential role for perforin in the regulation of inflammation and insulin resistance during diet-induced obesity, we first assessed perforin expression by intracellular staining in immune cells in the VAT, spleen, and BM from mice fed either NCD or HFD. One of the dominant cells expressing perforin was CD8⁺ T cells in the VAT (Fig. 1A) and spleen (Supplementary Fig. 1A). Compared with NCD controls, HFD-fed mice showed an increased frequency of perforin⁺ CD8⁺ T cells in the VAT upon anti-CD3/CD28 stimulation (Fig. 1A) and in the spleen when left unstimulated (Supplementary Fig. 1A). Perforin expression was also detected in NK cells (NK1.1⁺ CD3⁻) and in CD4⁺ Foxp3⁺ T cells at low levels in all tissues (Supplementary Fig. 1B). In contrast, perforin was nondetectable in CD4⁺ Foxp3⁻ T, γδ T, CD19⁺, CD11b⁺, and CD11c⁺ cells (Supplementary Fig. 1B). No differences were detected in the frequency of NK or CD4⁺ Foxp3⁺ T cells expressing perforin between NCD- and HFD-fed mice in the VAT, spleen, and BM (Supplementary Fig. 1B). Next, to assess the function of perforin in diet-induced obesity, we compared Prf1^null with age-matched C57BL/6 WT mice fed HFD starting at 6 weeks of age. Relative to HFD-fed WT mice, Prf1^null mice receiving the same diet had slight but significant increases in body weight between 6 and 18 weeks of age; although by 20 weeks of age, this difference was not statistically significant (Fig. 1B). To determine the location of this increase in body weight, organs and adipose tissues from WT and Prf1^null mice were harvested and weighed after 10 and 20 weeks of HFD feeding. Prf1^null mice had increased subcutaneous fat and liver weights 10 weeks after initiation of HFD (Fig. 1C), but no significant differences in organ weights were detected at 20 weeks in HFD (Fig. 1D). Due to the increased body weight at 10 weeks of HFD, we focused on further characterizing adipose tissue at this time. Interestingly, after 10 weeks of HFD, VAT weight trended higher in Prf1^null mice, and this correlated with increased VAT volume by MRI (Fig. 1E and Supplementary Fig. 1C) and adipocyte size (Fig. 1F and G) in Prf1^null mice, compared with WT control mice. There was no difference in the number of CLS (Fig. 1G and H) and expression of multiple genes associated with lipogenesis (Fig. 1I). However, the livers of Prf1^null mice showed upregulated expression of lipogenic genes (Fig. 1J) and higher levels of triglycerides (Fig. 1K) compared with WT mice. Despite increased body weight and adiposity, HFD-fed Prf1^null mice had similar food intake (Supplementary Fig. 1D, left), core body temperature (Supplementary Fig. 1D, right), VO₂, VCO₂, respiratory exchange ratio, locomotor activity (Supplementary Fig. 1E), and energy expenditure adjusted for body mass (Supplementary Fig. 1F), compared with WT controls.

After 10 and 20 weeks of HFD feeding, we performed GTTs in Prf1^null and age-matched WT control mice on the same diet. Compared with WT mice, Prf1^null mice had higher fasting glucose (Fig. 2A) and generally worsened glucose tolerance after 10 (Fig. 2B) and 20 weeks of HFD feeding (Fig. 2C), even in the absence of body weight differences at 20 weeks. In agreement with these defects in glucose homeostasis, Prf1^null mice fed HFD for 10 and 20 weeks had markedly increased fasting insulin (Fig. 2D) and worsened insulin sensitivity upon insulin challenge after 13 weeks of HFD, compared with WT controls (Fig. 2E). To determine which tissues show lower insulin sensitivity in Prf1^null mice, we measured the protein levels of total and pAkt in VAT, muscle, and liver harvested from HFD-fed WT and Prf1^null mice after an acute insulin challenge. Compared with WT controls, all three metabolic tissues in Prf1^null mice showed lower pAkt/Akt ratios (Fig. 2F and Supplementary Fig. 1G), in agreement with our insulin tolerance findings. Interestingly, Prf1^null mice maintained on a standard NCD showed increased body weight between 6 and 26 weeks of age (Fig. 3A) and a trend for worsened fasting glucose (Fig. 3B) and worsened glucose tolerance (Fig. 3C) but similar fasting insulin (Fig. 3D), compared with WT controls. These results suggest that perforin may also have potential effects on regulating glucose metabolism under standard caloric intake.

Proinflammatory T-cell and macrophage infiltration of VAT directly contributes to obesity-related insulin resistance (3). Thus, to investigate mechanisms responsible for the worsening in insulin resistance observed in Prf1^null mice during HFD feeding, we determined the immune cell composition of epididymal VAT, spleens, BM, and LNs harvested from Prf1^null and WT control mice. Prf1^null mice had increased total T cells and CD8⁺ T-cell accumulation in VAT after 10–20 weeks of HFD (Fig. 4A), whereas no differences in T-cell subsets were detected in spleens and BM (Supplementary Fig. 2A). Compared with WT controls, HFD-fed Prf1^null mice had increased numbers of total T cells and CD4⁺ cells in the LNs, although no differences were observed in CD8⁺ cells (Supplementary Fig. 2A). There was no difference in the total number of macrophages in VAT; however, Prf1^null mice showed a higher percentage of VAT M1-polarized macrophages (CD11b⁺ F4/80⁺ CD11c⁺) after 10 weeks of HFD feeding, consistent with worsened metabolic parameters in these mice (Fig. 4B). Interestingly, whereas the T-cell subset distribution was similar in the spleen, BM, and LNs between HFD-fed Prf1^null and WT mice (Supplementary Fig. 2B), there was a trend toward a lower CD4-to-CD8 ratio in the VAT of Prf1^null mice 10 and 20 weeks after initiation of HFD (Fig. 4C).

Since no differences in body weights were observed after 20 weeks of HFD despite worsening metabolic control in the Prf1^null mice, we decided to further investigate immunological parameters at this time point. To determine the relative distribution of infiltrating immune cell subsets in VAT inflammation during HFD feeding, we assessed the proportion of regulatory forkhead box P3⁺ (Foxp3⁺) T cells in VAT, as well as the proportions of IFN-γ–producing CD4⁺, CD8⁺, and NK cells, and IL-17–producing CD4⁺ and γδ⁺ T cells. No difference was detected in the percentage of VAT-resident Foxp3⁺ T cells between HFD-fed Prf1^null and WT control mice (Fig. 4D). As well, the frequency of Foxp3⁺ regulatory T cells in LN and BM CD4⁺ T cells was similar between HFD-fed Prf1^null and WT control mice (Supplementary Fig. 2C). After HFD feeding, VAT of Prf1^null mice contained increased percentages of CD8⁺ and CD4⁺ T (Th1) cells expressing IFN-γ but no differences in the percentage of NK cells producing IFN-γ (Fig. 4E). Increased frequency of IFN-γ–producing CD8⁺ and CD4⁺ T cells was also observed in the BM (Supplementary Fig. 2D), but not in LNs (Supplementary Fig. 2E), of HFD-fed Prf1^null compared with WT control mice. Furthermore, the Th17 response was unchanged in Prf1^null mice, as the percentages of CD4⁺ and γδ⁺ T cells positive for IL-17 were similar between the two strains in VAT (Fig. 4F), BM (Supplementary Fig. 2F), and LNs (Supplementary Fig. 2G). Overall, these findings depict a state of heightened Th1-mediated inflammation in Prf1^null mice in the diet-induced obesity model that is localized predominantly to metabolic tissues such as VAT.

We measured the amounts of proinflammatory cytokines secreted by VAT SVCs from Prf1^null and WT mice cultured for 48 h. SVCs from Prf1^null produced increased levels of IL-6, IL-12 (p70), IFN-γ, GM-CSF, MCP-1, and TNF-α compared with SVCs from HFD-fed WT controls, consistent with increased inflammation in VAT from Prf1^null mice (Fig. 5A). We also determined changes in several serum adipokines using a multiplex screening kit. Only leptin levels were increased after 10 weeks of HFD feeding, consistent with increased fat mass (Supplementary Fig. 3A). However, by 20 weeks of HFD, increased amounts of some cytokines were identified. Serum leptin levels fell in Prf1^null mice at 20 weeks of HFD feeding (Supplementary Fig. 3B), in agreement with the trend to lower VAT weight observed at this time point. Using more sensitive detection methods, we were able to observe additional inflammatory cytokines increased in the serum of Prf1^null mice compared with WT controls, primarily after 20 weeks of HFD feeding (Fig. 5B). These changes are consistent with the increased inflammation and associated worsening of insulin resistance observed in the HFD-fed Prf1^null mice.

To determine whether the increased VAT T-cell numbers in Prf1^null mice, particularly in pathogenic CD8⁺ T cells, were associated with altered cell turnover, we assessed apoptotic changes and the proliferative potential of WT and Prf1^null T cells after 10 weeks of HFD feeding. Compared with WT controls, Prf1^null mice had a marked decrease in the percentage of early apoptotic VAT CD8⁺ T cells but not splenic CD8⁺ T cells (Fig. 6A), suggesting that Prf1^null mice have a decreased ability to regulate CD8⁺ T-cell numbers in VAT through a reduced induction of apoptosis. Next, we assessed the effect of perforin deficiency on the proliferative potential of T cells. Although CD3/CD28-stimulated WT and Prf1^null splenic CD8⁺ T cells proliferated similarly in vitro (Fig. 6B), Prf1^null mice showed a dramatic increase in the in vivo proliferation of CD8⁺ T cells in VAT as determined by EdU incorporation (Fig. 6C). Unrestrained CD8⁺ T-cell proliferation in Prf1^null mice was only observed in VAT as no differences were detected between WT and Prf1^null mice in the proliferation of CD8⁺ T cells in the spleen, LNs, and BM (Supplementary Fig. 4A). Enhanced in vitro proliferation of VAT-resident CD8⁺ T cells in response to VAT coculture was also observed in Prf1^null mice, compared with WT controls (Supplementary Fig. 4B). These findings suggest that perforin deficiency leads to CD8⁺ accumulation in VAT in part due to diminished apoptosis, as well as increased proliferation. Moreover, the increased proliferation of CD8⁺ T cells in Prf1^null mice is unlikely solely due to an intrinsic abnormality of the cells themselves (Fig. 6B) but is also related to the in vivo environment generated in the inflamed VAT of these mice (Fig. 6C).

To assess the activation status of T cells, we determined the expression of the degranulation (CD107a), activation (CD44), and effector memory (CD44 and CD62L) markers in CD4⁺ and CD8⁺ T cells in the VAT, spleens, and LNs from HFD-fed Prf1^null and WT controls. Compared with WT controls, CD8⁺ T cells from Prf1^null mice had increased expression of the degranulation marker CD107a in VAT (Fig. 6D) but not in the spleens, BM, and LNs (Supplementary Fig. 4C). No differences were detected in T-cell expression of the activation marker CD44 in VAT (Fig. 6E), spleen, BM, and LNs (Supplementary Fig. 4D). However, CD8⁺ T cells from Prf1^null mice are enriched for CD44^hi CD62L⁻ effector memory cells in the VAT (Fig. 6F) but not in the spleen (Supplementary Fig. 4E), suggesting that these cells are primed and poised to respond to inflammatory stimuli locally in the VAT.

To identify CD8⁺ T cell–derived perforin in VAT as a critical regulatory factor governing glucose sensitivity, we investigated the pathogenicity of perforin-deficient CD8⁺ T cells derived from Prf1^null mice compared with their perforin-sufficient counterparts via adoptive transfer. HFD-fed CD8^null mice were reconstituted intraperitoneally with 1 × 10⁷ CD8⁺ T cells either from HFD-fed WT or HFD-fed Prf1^null mice. After 2 weeks, CD8⁺ T cells from Prf1^null and WT mice had similar reconstitution in the VAT, which was greater than reconstitution rates in the spleen of recipient CD8^null mice (Supplementary Fig. 5A and B). Despite the lack of differences in body weight (Fig. 7A) after adoptive transfer, recipients of CD8⁺ T cells from Prf1^null mice developed worsened glucose tolerance (Fig. 7B) and hyperinsulinemia (Fig. 7C) compared with recipients of CD8⁺ T cells from WT controls, suggesting that perforin derived from CD8⁺ T cells is at least partially responsible for maintaining normal glucose homeostasis. Furthermore, recipients of CD8⁺ T cells from Prf1^null had a decreased pAkt-to-Akt protein ratio in the VAT but not in muscle and liver (Fig. 7D and Supplementary Fig. 5C) after insulin challenge, suggesting that the adoptive transfer promoted insulin resistance locally in the VAT. No difference in VAT lipogenesis gene expression was detected between recipients of CD8⁺ T cells from Prf1^null or WT mice (Fig. 7E). We also detected no differences in liver weights (Fig. 7F) or triglyceride levels (Fig. 7G) between recipients of CD8⁺ T cells from Prf1^null and recipients of CD8⁺ T cells from WT mice.

In addition to CD8⁺ T cells, perforin is highly expressed in NK cells (16). Thus, we next investigated the overall effect of global knockdown of NK cells on metabolic parameters in mice. If a strong essential role for NK cell–derived perforin existed in control glucose homeostasis, we would expect similar phenotype metabolic abnormalities as observed in the Prf1^null mice. Thus, we studied the metabolic response of NK^null and age-matched WT mice fed HFD for 12 weeks. However, contrary to HFD-fed Prf1^null mice, HFD-fed NK^null mice had decreased body weight (Fig. 7H) and improvements in glucose tolerance (Fig. 7I) and fasting insulin (Fig. 7J) compared with age-matched WT mice. These results highlight that NK cell–derived perforin likely does not have a nonredundant role in promoting glucose tolerance.

We have identified a novel role for perforin in regulating inflammatory changes during obesity-related insulin resistance. Previous work has identified critical roles for perforin in promoting inflammatory-mediated diseases, including type 1 diabetes (25), cerebral malaria (26), and viral myocarditis (27). In diet-induced obesity, a dominant overarching effect of perforin appears to be protective against metabolic abnormalities such as insulin resistance.

Compared with age-matched WT controls, Prf1^null mice already showed increased body weight at the start of HFD feeding, and this difference was sustained up to 18 weeks of age. The presence of increased body weight before HFD initiation in Prf1^null mice suggests that perforin may also regulate early body size and growth independent of diet. Indeed, perforin has been shown to be expressed in some nonimmune cells, such as chondrocytes of articular cartilage, and may be involved in tissue remodeling (28). Consistently, our findings indicate that Prf1^null mice fed NCD have increased body weight up to 26 weeks of age.

After 10 weeks of HFD, part of the increase in body weight was located in liver and adipose tissues. At this time point, Prf1^null mice also had increased VAT volume and fat cell diameter; however, no changes in expression of genes controlling lipogenesis were detected, suggesting that adipocyte hypertrophy is an early manifestation of perforin deficiency. In liver, on the other hand, perforin deletion resulted in upregulated expression of lipogenic genes and triglyceride accumulation, likely owing to hyperinsulinemia-induced increase in uptake and de novo synthesis of fatty acids, typical of the insulin-resistant state (29). The increase in body weights and adiposity were not accompanied by changes in food intake, metabolic rate, and activity. The mechanisms by which perforin deficiency affects early weight gain and adipocyte hypertrophy is unclear. However, there is evidence that the immune system can directly regulate these parameters. For example, certain cytokines, such as interleukin 15 (IL-15) can induce perforin expression in naïve CD8⁺ T cells (30) and NK cells (31) and can promote weight loss independent of food intake and lymphocyte function (32). We also cannot exclude a potential role for direct CD8⁺ T-cell interactions with adipocytes and perforin-mediated apoptosis of expanding adipocytes in this process.

Interestingly, the effects of perforin deficiency on adipose tissue weight gain may be transient because by around 20 weeks of HFD feeding, Prf1^null mice tended to have decreased VAT weight, confirmed by falling levels of leptin on longer durations of HFD. It is possible that these effects might be related to excessive T-cell expansion and proinflammatory cytokine secretion, including IFN-γ, leading to local insulin resistance and lipolysis of fat. Other organ weights at 20 weeks of HFD feeding were unaffected, confirming that perforin influences body weight early during the course of HFD feeding.

To assess the effect of perforin on the regulation of HFD-induced insulin resistance, we then investigated whole-body glucose homeostasis and insulin sensitivity of HFD-fed Prf1^null and age-matched WT controls. After HFD feeding, Prf1^null mice had worsened fasting hyperglycemia and hyperinsulinemia and showed deteriorated glucose tolerance and peripheral insulin sensitivity. Infiltration of immune cells into adipose tissue is a key event in the development of insulin resistance during HFD-induced obesity (2,3,33). In particular, IFN-γ–secreting CD8⁺ T cells and Th1 CD4⁺ T cells infiltrate VAT and enhance macrophage M1 polarization and proinflammatory functions (11,13). These classically activated or M1 macrophages in VAT produce IL-1β, IL-6, and TNF-α, factors, which can alter insulin receptor signaling in target tissues, directly contributing to insulin resistance (3). In the current study, all three metabolic tissues, liver, muscle, and VAT, showed reduced insulin sensitivity in Prf1^null mice.

HFD-fed Prf1^null mice displayed increased CD8⁺ T cells and IFN-γ–secreting Th1 CD4⁺ T cells in VAT, associated with VAT proinflammatory cytokine production and increased M1 macrophage polarization. This augmented presence of inflammatory T cells in inflamed VAT is likely a direct consequence of perforin’s critical ability to maintain T-cell homeostasis in inflamed tissues (17,20,21) through downregulation of peripheral T-cell expansion (34). Consistently, we observed a substantial decrease in the percentages of early apoptotic CD8⁺ T cells and an increase in CD8⁺ T-cell proliferation, degranulation, and cytokine production in the VAT of HFD-fed Prf1^null mice. Thus, perforin-mediated control of T-cell homeostasis in VAT may oppose previously described mechanisms of induced CD8⁺ T-cell proliferation in VAT, including local activation of CD8⁺ T cells by the proinflammatory cytokines IL-12 and IL-18 (35). Given the worsened insulin sensitivity observed in muscle and liver of Prf1^null mice, more studies are needed to characterize the role of perforin in regulating immune cell activation in these metabolic tissues during HFD.

In mouse models of hemophagocytic lymphohistiocytosis, Prf1^null mice injected with lymphocytic choriomeningitis virus develop a striking increase in CD8⁺ and CD4⁺ T-cell activation and increased antigen presentation by rare dendritic cells (20,21). Diet-induced obesity is also a state of increased inflammation, especially in metabolic tissues like VAT. Prf1^null mice show an increase in IFN-γ production by VAT T cells as well as antigen-presenting cell proinflammatory cytokines such as IL-12(p70), IL-6, and TNF-α. During viral clearance, such as in lymphocytic choriomeningitis virus models, CD8⁺ T cells function to kill dendritic cell populations in a perforin-dependent manner to limit T-cell activation (21). Collectively, our results indicate that perforin has a regulatory role in T-cell turnover, activation, and inflammatory cytokine production in inflamed VAT during HFD feeding. A question that remains unanswered is whether there is cross talk between CD8⁺ T cells and dendritic cells in VAT to limit pathological T-cell function during obesity-related insulin resistance.

Perforin plays a critical role in the autoimmune destruction of insulin-producing β-cells of the pancreas, leading to type 1 diabetes. Prf1^null mice crossed onto the nonobese diabetic (NOD) background show markedly reduced disease incidence (25,36). In these mice, islet inflammation is not diminished in protected mice but targeted β-cell killing is impaired (25,36). In diet-induced obesity, VAT contains a limited T-cell receptor repertoire of CD4⁺ and CD8⁺ T cells, a hallmark of antigen-specific immunity. Moreover, HFD feeding induces VAT adipocyte cell death, and dying fat cells become surrounded by macrophages and other immune cells, including prominent numbers of CD8⁺ T cells (13). To screen for perforin-dependent CD8⁺ T-cell killing of adipocytes as a potential mechanism of antigen-specific immunity in VAT, we counted CLS in HFD Prf1^null mice and WT controls. Interestingly, HFD Prf1^null mice showed no differences in the ability to form CLS in VAT, suggesting alternative mechanisms of VAT adipocyte cell death, which might include Fas (CD95)-mediated apoptosis (37) or activation of the key proapoptotic molecule Bid (38). More recently, mechanisms of adipocyte cell death have focused on pyroptosis, involving the NLRP3 inflammasome and caspase-1 activation (39).

We next investigated whether CD8⁺ T cells or NK cells are dominant sources of protective perforin (16). NK cells play a crucial role in limiting viral replication through perforin-dependent cytotoxic effects and, along with CD8⁺ T cells, are dominant producers of perforin in the immune system (40). Transfer of CD8⁺ cells from HFD-fed Prf1^null to CD8^null mice results in worsened glucose tolerance and hyperinsulinemia. These effects indicate that the abnormal glucose and insulin metabolism in Prf1^null mice were at least partially triggered by the perforin deficiency in CD8⁺ T cells. Consistently, a high frequency of perforin expression was detected in stimulated VAT CD8⁺ T cells. Perforin was also detected in low amounts in other immune cells in VAT, including CD4⁺ Foxp3⁺ T cells and NK cells. Thus, although we have shown that perforin derived from CD8⁺ T cells plays an important role in obesity-related VAT inflammation, we cannot entirely exclude that contributions from other immune cell sources, including CD4⁺ Foxp3⁺ T cells and NK cells, exist to regulate inflammation during insulin resistance.

To examine a potential nonredundant role for NK cell–derived perforin in limiting obesity-related insulin resistance, we examined glucose and insulin metabolism of HFD-fed NK^null mice, caused by deletion of the Nfil3 (E4bp4) gene, a critical transcription factor involved in NK cell development (41). HFD-fed NK^null mice had improvements in glucose tolerance and fasting insulin. Although NK cell deficiency may influence diet-induced obesity–induced glucose intolerance and insulin sensitivity via perforin-independent mechanisms, these results suggest that NK cells overall are not a nonredundant source of protective perforin in diet-induced obesity–induced insulin resistance. Glucose homeostasis in NK^null mice is likely either compensated by perforin from other sources or overshadowed by other pathogenic functions of the NK cell in promoting glucose intolerance. Furthermore, we acknowledge that some of the improvements observed in glucose homeostasis in NK^null mice are likely attributable to the reduction in body weight observed in age-matched mice. In type 2 diabetic patients, the frequency of blood NK cells expressing the activation markers NKG2D⁺ and CD107a⁺ is increased in correlation with larger body mass (42). Relative to subcutaneous depots, obese patients also have increased frequency of IFN-γ–producing CD56⁺ NK cells in VAT, suggesting that NK cells can influence the inflammatory tone of human VAT (43). In agreement, our results indicate that NK cells may have a dominant pathogenic effect on HFD-induced weight gain, glucose intolerance, and insulin resistance in mice. However, additional studies are required to elicit the exact roles and mechanisms of NK cell function in the low-grade inflammation during diet-induced obesity.

Collectively, our work supports a model whereby perforin limits expansion, activation, and cytokine production of pathogenic T cells in inflamed VAT during DIO. Because T-cell activation in VAT regulates macrophage polarization and function, perforin seems to have a key role in obesity-related chronic inflammation. Our findings provide further evidence for a role for adaptive immune cells in insulin resistance and highlight the potential benefit of therapies that target pathogenic T-cell homeostasis and activation in VAT.

Acknowledgments. The authors thank the Spatio-Temporal Targeting and Amplification of Radiation Response Program (STTARR) facility, in particular Warren Foltz, for performing the MRIs in this study.

Funding. This work was supported in part by Canadian Institutes of Health Research grant 119414 (D.A.W.) and Canadian Diabetes Association grants OG-3-12-3844 (D.A.W.) and CS-5-12-3886 (D.A.W.). X.S.R. is the recipient of a Banting & Best Diabetes Centre Fellowship (funded by Eli Lilly Canada).

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

Author Contributions. X.S.R., S.T., S.W., and D.A.W. designed and conducted experiments, analyzed data, and wrote the manuscript. H.Le., H.Lu., M.G., S.Y.S., S.S., C.T.L., and G.H.Y.L. performed experiments. H.T., T.W.M., and M.W. reviewed the manuscript. D.A.W. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.