Adipocyte CD1d Gene Transfer Induces T Cell Expansion and Adipocyte Inflammation in CD1d Knockout Mice

Adipose tissues play an important role in energy homeostasis, functioning as both energy storage and an endocrine organ secreting multiple hormones, adipokines, and cytokines. Adipose inflammation is implicated in obesity-associated metabolic syndromes, and anti-inflammatory treatments are proposed to combat obesity-associated metabolic diseases (1). Adipose inflammation results from the cross-talk between adipocytes that comprise the bulk of adipose tissue, and immune cells such as macrophages and T cells in the stromal vascular fraction (SVF). Adipose T cells are among the first responders to a high-fat diet challenge. Both Th1 CD4⁺ and CD8⁺ T cells are implicated in obesity-associated adipose inflammation (2–5), but the specific signals that activate proinflammatory T cells are not fully understood. There is evidence in obesity that adipocytes express MHC class II molecules that determine the CD4⁺ T cell subset activation and IFN-γ production as well as insulin resistance (6).

Together with MHC class I and class II molecules, CD1 proteins constitute a distinct, third lineage of lipid Ag-presenting molecules. The crystal structure shows that CD1 adopts an MHC fold that is more closely related to that of MHC class I than to that of MHC class II (7). Five CD1 genes in humans (CD1a, CD1b, CD1c, CD1d, and CD1e) and only one in the rodents (CD1d) have been found (8), and CD1d is adapted to bind lipid-based Ags rather than peptides. CD1d-presented lipid Ags activate a special class of T cells known as NKT cells with two broad categories. Type I NKT cells, also known as invariant NKT (iNKT) cells, are typically characterized by the expression of a semi-invariant TCR, defined by their invariant TCR α-chain (Vα14-Jα18 in mice, Vα24-Jα18 in humans) and limited TCRβ repertoire (Vβ8.2, Vβ7, and Vβ2 in mice, Vβ11 in humans) and their reactivity with the glycosphingolipid Ag α-galactosylceramide (α-GalCer) (9, 10). Murine type I NKT cells are either CD4⁺ or CD4⁻CD8⁻ double negative (DN) and exhibit an activated/memory phenotype. In contrast, TCRs expressed by type II NKT cells are more diverse and do not bind α-GalCer-CD1d tetramers, and their phenotype can be quite heterogeneous, including CD4⁺, CD8⁺, or DN NKT cells (11). iNKT cells are enriched in both mouse and human adipose tissue (12, 13). CD1d is expressed on immune cells such as monocytes, macrophages, dendritic cells, B lymphocytes (14), and non-immune cells, including epithelial and vascular smooth muscle cells and adipocytes (15). In murine visceral adipose tissue (VAT), CD1d is abundantly expressed in adipocytes where the mRNA level is ∼30-fold higher than that in endothelial cells or macrophages. In contrast, other Ag-presenting molecules, for example, MHC class I and MHC class II, are highly enriched in macrophages and endothelial cells (16).

The role of adipose iNKT cells in insulin resistance and obesity is controversial, with both protective (12, 17, 18) and pathological (19, 20) effects having been reported. There is evidence that adipocyte CD1d plays an important role in adipose tissue homeostasis thought to be through regulation of iNKT cells. However, the results from different reports are also inconsistent. One study demonstrates that deletion of CD1d specifically from adipocytes alleviates diet-induced obesity (DIO) and insulin resistance (21). On the contrary, other studies report that DIO is associated with downregulation of CD1d expression in adipocytes, and adipocyte CD1d deletion by Cre-loxP recombination reduces adipose iNKT numbers and aggravates adipose tissue inflammation and insulin resistance in obesity (16, 22).

To genetically manipulate individual adipose depots in vivo, we have characterized an engineered hybrid serotype of adeno-associated virus (AAV), Rec2 (23–25). This hybrid serotype achieves superior transduction of fat by direct injection when compared with the naturally occurring AAV serotypes. Interestingly, oral administration of the Rec2 vector leads to selective transduction of the intrascapular brown adipose tissue. Several laboratories have used the Rec2 serotype vectors to manipulate adipose tissues in various mouse models (26–28). To further improve selectivity of adipose tissue gene transfer, we recently developed a dual-cassette AAV vector system: one cassette using the nonselective CBA promoter to drive transgene expression, and the other cassette using a liver-specific albumin promoter to drive a microRNA targeting the woodchuck posttranscriptional regulatory element (WPRE) sequence, which exists only in the same rAAV vector. This dual-cassette Rec2 vector by a single i.p. injection achieves highly selective transduction of VAT with minimal off-target transgene expression in the liver and no detectable transgene expression in other tissues (kidney, intestine, testis) (25, 29, 30). In this study, we employed this unique delivery system to rescue CD1d gene expression in VAT adipocytes of CD1d knockout (KO) mice to investigate the interactions between adipocytes and immune cells within the adipose tissue. The results show that introducing selective expression of CD1d to VAT adipocytes in CD1d KO mice leads to a massive expansion and activation of CD8⁺ T cells. Furthermore, the ensuing activation of CD8⁺ T cells plays a major role in the subsequent induction of inflammatory responses within adipocytes.

CD1d KO mouse breeders were purchased from The Jackson Laboratory (CD1d^−/−, catalog no. 008881) and bred at our facility. Male C57BL/6 mice (6 wk old) were purchased from Charles River Laboratories (Wilmington, MA). All mice were housed in temperature (22–23°C)- and humidity-controlled rooms with food (chow diet, 11% fat, caloric density 3.4 kcal/g, Teklad) and water ad libitum. All animal experiments were performed in accordance with the guidelines approved by the Ohio State University Institutional Laboratory Animal Care and Use Committee.

Mouse CD1d1 cDNA (GenBank: NM_007639.3) was synthesized by Integrated DNA Technologies. The synthesized DNA fragment contains XhoI (CTCGAG) and SacI (GAGCTC) sequences at the 5′ and 3′ termini, respectively. The cDNA was cloned into a novel AAV plasmid of dual cassettes that restricts off-target transduction in liver (25). The transgene expression cassette consists of the CMV enhancer and CBA (chicken β-actin) promoter, WPRE enhancing transgene expression, and bovine growth hormone poly(A). The liver-restricting cassette consists of albumin promoter and a microRNA targeting the WPRE. The dual cassettes are flanked by AAV2 inverted terminal repeats. Engineered hybrid serotype Rec2 vectors were packaged and purified as described previously (23). The control vector contained the identical backbone with no transgene and was termed Rec2-empty.

Six-week-old male CD1d KO mice or wild-type mice were randomized to receive the Rec2-CD1d or Rec2-empty by i.p. injection at the dose of 2 × 10¹⁰ viral genomes (vg) per mouse diluted in AAV buffer to a volume of 100 µl. Mice were monitored for body weight and food intake until termination of the experiment.

CD1d KO mice were randomized to receive Rec2-CD1d or Rec2-empty as described above. Starting at day 7 after AAV injection, each vector group was randomly assigned to receive MCC950 (Sigma-Aldrich, catalog no. PZ0280) or vehicle (Vedco veterinary 0.9% sodium chloride injection, USP). MCC950 (10 mg/kg body weight) was i.p. injected every other day until the end of the study 4 wk after AAV injection.

CD1d KO mice were randomized to receive Rec2-CD1d or Rec2-empty as described above. Starting at day 11 after AAV injection, each vector group was randomly assigned to receive anti-CD8 (Bio X Cell, clone YTS 169.4.2, catalog no. BE0117) or IgG (Bio X Cell, catalog no. BE0090). The Abs (0.2 mg/mouse) were i.p. injected once weekly until the end of the study 4 wk after AAV treatment.

Mice were anesthetized with 2.5% isoflurane followed by decapitation, and then truncal blood was collected. The VAT and spleen were dissected and weighed. The dissected VAT was minced into small pieces in Krebs–Ringer HEPES buffer (pH 7.4). Collagenase (1 mg/ml, Sigma-Aldrich, catalog no. C6885) was added to all tissues and incubated for 40 min at 37°C with shaking. The mixture was centrifuged to separate the floating adipocytes from the adipose SVF. The SVF pellet was treated with ammonium chloride solution to lyse the RBCs, then washed and resuspended in FACS buffer. The spleens were mechanically dissociated through a 70-µm strainer to obtain a single-cell suspension. RBCs were lysed with ammonium chloride solution, then washed and resuspended in FACS buffer. For surface staining, cells were stained with fluorescent dye–conjugated Abs with the appropriate surface markers for 20 min. Conjugated Abs NK1.1 (catalog no. 108728, PK136), CD3 (catalog no. 100220, 17A2), CD19 (catalog no. 115528, 6D5), CD8a (catalog no. 100762, 53-6.7), CD4 (catalog no. 100526, RM4-5), CD69 (catalog no. 563290, H1.2F3), and CD8a (catalog no. 100738, 53-6.7) were purchase from BioLegend. Conjugated Abs CD44 (catalog no. 553133, IM7), CD62L (catalog no. 560507, MEL-14), and CD69 (catalog no. 563290, H1.2F3) were purchased from BD Biosciences. Conjugated Abs CD62L (catalog no. 45-0621-082, MEL-14) and CD44 (catalog no. 56-0441-82) were purchased from eBioscience. The CD1d tetramer (PBS-57) was provided by the National Institutes of Health. Cell events were acquired using an LSR II flow cytometer (BD Biosciences), and the results were analyzed using FlowJo v10 software (Tree Star).

Mice were i.p. injected with glucose solution (1 mg glucose/g body weight) after an overnight fast. Blood was drawn from the tail at various time points, and the blood glucose concentrations were measured with a portable glucose meter (Bayer Contour Next).

Trunk blood was collected at euthanasia. Serum was prepared by allowing the blood to clot for 30 min on ice followed by centrifugation. The amount of leptin and adiponectin in serum was measured using the following DuoSet ELISA development system (R&D Systems): mouse leptin (catalog no. DY498) and adiponectin/Acrp30 (catalog no. DY1119).

Adipose tissues or adipocytes were homogenized in radioimmunoprecipitation assay (RIPA) buffer (Pierce, catalog no. 89901) containing protease inhibitor cocktail set III (Calbiochem, catalog no. 539134). Tissue lysates were separated by gradient gel (4–20%, Mini-PROTEAN TGX, Bio-Rad, catalog no. 4561096) and transferred to a nitrocellulose membrane (Bio-Rad, catalog no. 1620115). Blots were incubated overnight at 4°C with the following primary Abs: tubulin (1:1000, Cell Signaling Technology, catalog no. 2144), CD1d (1:500, Santa Cruz, catalog no. sc-373858), NLRP3 (1:1000, Cell Signaling Technology, catalog no. 15101), caspase-1(1:500, eBioscience, catalog no. 14-9832-82), IL-1β (1:1000, Cell Signaling Technology, catalog no. 12242), PARP (1:1000, Cell Signaling Technology, catalog no. 9542), caspase-9 (1:1000, Cell Signaling Technology, catalog no. 9508), cleaved caspase-9 (1:1000, Cell Signaling Technology, catalog no. 52873), and caspase-3 (1:1000, Cell Signaling Technology, catalog no. 14220).

Total RNA was isolated using the RNeasy kit plus RNase-free DNase treatment (Qiagen, catalog no. 74804). cDNA was reverse transcribed using a TaqMan reverse transcription reagent (Applied Biosystems, catalog no. N8080234). Quantitative PCR (qPCR) for adipocytes was carried out on a StepOnePlus real-time PCR system (Applied Biosystems) using Power SYBR Green PCR master mix (Applied Biosystems, catalog no. A25742). Primers were designed to detect the mouse mRNAs (Table I). Data were calibrated to endogenous control Actb quantified using the 2^−ΔΔCt method (31). For TCR Vβ repertoire analysis the PCR amplifications were performed using 24 TCR spectratyping primers as previously reported (32, 33). The sequences of qPCR primers are presented in Table I.

Adipose tissue was fixed in 10% formalin. Paraffin-embedded sections (4 μm) were processed and stained with H&E by the Comparative Pathology and Mouse Phenotyping and Histology/Immunohistochemistry core of The Ohio State University Comprehensive Cancer Center. Adipose tissue sections were imaged using a Nikon Eclipse 50i microscope with an Axiocam 506 color camera attachment through ZEN 2 software.

Values are expressed as mean ± SEM. Means between two groups were compared with a two-tailed Student t test on Microsoft Excel. For multiple comparisons, two-way ANOVAs were used to determine statistical significance on GraphPad Prism 8 (GraphPad Software). A p value <0.05 was considered significant, and the level of significance was indicated as *p < 0.05, **p < 0.01, and ***p < 0.001.

To investigate the consequences of reintroducing CD1d to adipocytes, we used a dual-cassette AAV serotype Rec2 vector system (25) to deliver mouse CD1d to the VAT in CD1d KO mice (vector design shown in (Fig. 1A). In a pilot study to test vector efficacy, male CD1d KO mice were randomized to receive a single i.p. injection of either Rec2-CD1d or a Rec2 vector carrying the same expression cassettes but no transgene (Rec2-empty), at a dose of 2 × 10¹⁰ vg per mouse. The mice were sacrificed at 4 wk after AAV injection. Surprisingly, we observed an increase in the absolute number of SVF cells from the VAT of CD1d-treated mice by ∼5-fold (data not shown). Hence, we were interested in evaluating which immune populations were affected in the adipose tissue by the restoration and overexpression of adipocyte CD1d. Overall, the absolute number of lymphocytes in the adipose tissue was increased by ∼11-fold compared with mice treated with the empty vector (Supplemental Fig. 1A, 1B). Flow cytometric quantification of lymphocyte subpopulations in the adipose tissue revealed that T cells had a 20-fold increase in absolute number, with CD8⁺ T cells experiencing a 60-fold increase, CD4⁺ T cells experiencing a 15-fold increase, and DN cells experiencing a 10-fold increase. B cells and NK cells also showed significant increases in absolute numbers but less so compared with T cell subsets (Supplemental Fig. 1A, 1B). These changes suggest that the restoration of adipocyte CD1d in CD1d KO mice resulted in an increase in number of all lymphocyte subsets in the adipose tissue, but particularly in the CD8⁺ T cell subset.

To confirm the T cell phenotypic changes, we repeated this experiment but for a longer duration of 7 wk to assess potential metabolic outcomes. The body weight (Fig. 1B) and food intake (data not shown) were monitored and showed no difference between the mice treated with CD1d and the mice treated with empty vector. A glucose tolerance test (GTT) was performed at 4 wk postinjection of AAV with no significant difference observed between the two groups (Fig. 1C). Mice sacrificed at 7 wk postinjection of AAV showed no differences in thymic or splenic mass (Fig. 1D). Among the fat pads examined, only VAT showed an ∼25% reduction of mass in CD1d-treated mice versus control mice (225.2 ± 19.5 mg versus 331.7 ± 22.9 mg, respectively), whereas there was no mass difference for s.c. adipose tissue, retroperitoneal adipose tissue, and brown adipose tissue (Fig. 1D). Leptin and adiponectin are critical mediators in the maintenance of metabolic homeostasis through adipose tissue cross-talk (34), yet we found no differences between experimental and control groups (Fig. 1E). The VAT adipocytes were isolated to perform quantitative RT-PCR (data not shown) and an immunoblot to verify AAV-mediated CD1d expression (Fig. 1F). Moreover, the CD1d mRNA expression in CD1d-treated mice was predominantly in adipocytes whereas its expression in the SVFs was only ∼2% of that found in adipocytes (Supplemental Fig. 1C). To assess whether the Rec2 vector transduces immune populations residing in the adipose tissue, a dual-cassette Rec2 vector carrying GFP was injected i.p. to mice at the dose of 2 × 10¹⁰ vg per mouse that resulted in robust GFP expression in adipose tissue (25). Flow cytometry detected minimal or no GFP fluorescence among B cells, T cells, NK cells, or macrophages from the VAT (Supplemental Fig. 2).

Massive increases of T cells, particularly CD8⁺ T cells, was reproduced (Fig. 2A, 2B). The CD4/CD8 ratio was reduced by ∼50% in CD1d-treated VAT (Fig. 2C, 2D), confirming a large increase in the absolute number of CD8⁺ cytotoxic T lymphocytes over the increase in the absolute number of CD4⁺ T cells. Compared with empty vector–treated mice, CD8⁺ T cells from CD1d-treated mice exhibited a significantly higher absolute number of CD62L^hiCD44^lo naive cells (CD1d, 6,433 ± 1,076; empty, 2,790 ± 394), CD62L^loCD44^hi effector cells (CD1d, 173,531 ± 37,276; empty, 4,433 ± 607), and CD62L^hiCD44^hi central memory cells (CD1d, 37,774 ± 6,168; empty, 2,794 ± 272) (Fig. 2E). In addition, CD8⁺ T cells from CD1d-treated mice showed a significantly higher absolute number of cells expressing CD69, a lymphoid activation Ag (CD1d, 188,612 ± 39,246; empty, 5,888 ± 612). CD4⁺ T cells showed a similar pattern of increase of these subsets (Fig. 2E).

CD1d KO mice are iNKT cell deficient (35, 36). To examine whether adipocyte gene transfer of CD1d could rescue the iNKT in VAT, T cells were subjected to ligand staining with α-GalCer analog PBS-57/CD1d tetramer selection, and the results showed no increase in the number or percentage of iNKT cells in adipose tissue (Supplemental Fig. 3). Furthermore, we did not observe a change in the percentage or absolute numbers of immune cell populations in the thymus (data not shown), blood (data not shown), or spleen (Supplemental Fig. 3), indicating that the immune profile changes in the CD1d-treated mice were restricted to the adipose tissue microenvironment.

Although systemic metabolic outcomes were limited, AAV-CD1d treatment resulted in a significant reduction of VAT mass (Fig. 1B–E). Thus, we profiled the expression of genes known to be major regulators of adipose tissue homeostasis and function, including important adipokines such as adiponectin (Adipoq), leptin (Lep), lipolysis marker hormone-sensitive lipase (Hsl), lipogenic marker sterol regulatory element binding transcription factor 1 (Srebp1c), and adipogenesis marker peroxisome proliferator-activated receptor isoform γ (Pparg) (37–40). All of the adipocyte functional genes were downregulated in the VAT adipocytes from CD1d-treated mice (Fig. 3A), suggesting that the homeostasis and function of adipose tissue might be compromised. Meanwhile, we tested the expression of apoptosis-related proteins by immunoblot and found that the expression levels of caspase-9 and caspase-3 were upregulated, suggesting an activation of the apoptotic signaling pathway in adipose tissue (Fig. 3B). Next, we examined the expression of the immune-related cytokines and inflammation-related genes. CCL2 (Ccl2), IL-10 (Il10), and TNF-α (Tnfa) were significantly upregulated in the adipocytes from CD1d-treated mice (Fig. 3C). In addition, CD1d treatment led to a massive induction of the NLRP3 inflammasome-related transcripts (Nlrp3, Casp1, and Il1b) (Fig. 3C), and protein levels were verified by immunoblotting (Fig. 3D).

H&E staining showed rare leukocyte infiltration in the VATs of mice treated with empty vector. In contrast, widespread leukocyte extravasation and infiltration were found in the VATs of the CD1d-treated mice, often adjacent to blood vessels (Fig. 3E).

Furthermore, we profiled gene expression in SVF cells and found significant upregulation of the T cell marker Cd3e and the proliferation marker Ki67 (Fig. 3F). Expression of CXCR3 (Cxcr3) is tightly linked to Th1 CD4⁺ and CD8⁺ effector cells and permits T cells to enter into the inflammatory sites (41). Src-family kinase Lck, a key regulator of T cell activation, is involved in the initiation of TCR signaling (42). Both Cxcr3 and Lck were significantly upregulated in CD1d-treated VAT SVF cells (Fig. 3F).

Regarding the link between adipocyte CD1d-induced T cell expansion/activation and adipocyte inflammation, we hypothesized two possibilities: 1) NLRP3 inflammasome activation in adipocytes leads to IL-1β stimulation of T cells; 2) CD1d-expressing adipocytes function as APCs for a perceived alloantigen (i.e., CD1d) to CD8⁺ T cells and thus induce T cell expansion and activation, subsequently causing adipocyte inflammation.

To test the hypothesis of adipocyte inflammasome activation resulting in T cell activation, we used a NLPR3 inhibitor to suppress IL-1β production in vivo. MCC950 inhibits both canonical and noncanonical NLRP3 inflammasome activation and IL-1β secretion (43, 44). One week after AAV injection of either Rec2-CD1d or Rec2-empty, mice were treated with an i.p. injection of MCC950 or vehicle control every other day until the end of the experiment at 4 wk after AAV injection.

We verified a significant reduction in Il1b by ∼50% in the CD1d/MCC950 group versus the CD1d/vehicle group (Fig. 4C). The CD1d-treated mice that received the i.p. injection of vehicle control induced T cell changes that were similar to those observed in previous experiments (compare (Fig. 1 and Supplemental Fig. 1 to (Fig. 4A, 4B). CD1d-treated mice that received the i.p. injection of MCC950 had virtually identical T cell changes when compared with mice treated with the i.p. injection of vehicle control (Fig. 4A, 4B), including CD4 and CD8 differentiation from naive to effector cells (data not shown). VAT adipocyte gene expression analysis showed that MCC950 treatment significantly inhibited Il1b gene expression whereas it had no effect on the upregulation of Nlrp3 and Casp1 in CD1d-treated mice (Fig. 4C). These results suggest that inhibiting NLRP3/IL-1β has no impact on adipocyte CD1d-induced T cell expansion and activation, further suggesting that IL-1β released by NLRP3 inflammasome is not essential for the activation of CD8⁺ T cells.

In addition to the classical function of presenting lipid Ag, nonclassical biological functions of CD1d are reported in various cell types presumably through intracellular signaling of CD1d (45–48). To test whether CD1d expression could directly induce adipocyte inflammatory responses, we repeated the AAV-CD1d experiment in CD1d KO mice and examined an early time point when transgene was expressed but T cell activation did not occur. At 10 d after AAV-CD1d injection, CD1d transgene expression was detected in VAT adipocytes at ∼4-fold of the CD1d mRNA level in naive wild-type mice (Supplemental Fig. 4A), confirming a successful gene transfer of CD1d in VAT. At this early time point, immune changes, particularly the expansion of T cells, were not observed in CD1d-treated mice (Supplemental Fig. 4B). Moreover, in the absence of T cell expansion, adipocyte overexpression of CD1d had no effect on the expression of NLRP3 inflammasome–related genes or adipocyte functional genes (Supplemental Fig. 4C). These data suggest that CD1d expression per se is insufficient to induce adipocyte inflammasome, and the adipocyte inflammatory responses do not occur prior to the activation of T cells or other immune cells.

To investigate the association of CD8⁺ T cells with the NLRP3 inflammasome that we observed in CD1d-transduced adipocytes within CD1d KO mice, we performed a CD8 T cell depletion study. CD1d KO mice were subjected to the same CD1d treatment as previously described. At 11 d after AAV injection, mice from each vector group were randomly assigned to receive a weekly injection of anti-CD8 Ab or control IgG until the end of the study at 4 wk after AAV injection. In mice treated with empty vector, anti-CD8 Ab treatment sufficiently depleted CD8⁺ T cells in the VAT (Fig. 5B) and spleen (data not shown). In mice treated with CD1d, infusion of control IgG displayed the same T cell changes observed in previous experiments (compare (Fig. 2 with (Fig. 5A–C). Anti-CD8 treatment did not prevent the proportional increase of SVF lymphocytes in CD1d KO mice treated with CD1d but did significantly decrease both the percentage and absolute number of total T cells compared with CD1d KO mice treated with CD1d followed by IgG control (Fig. 5A), due specifically to significant reductions in CD8⁺ T cells but not CD4⁺ T cells in VAT (Fig. 5C). Adipocyte gene expression profiling showed that CD8⁺ T cell depletion significantly attenuated the upregulation of inflammasome-related genes (Fig. 5D) and largely prevented the downregulation of a cluster of adipocyte functional genes associated with CD1d-induced T cell activation (Fig. 5E).

Accumulating evidence suggests the existence of iNKT cell autoreactivity, whereas most iNKT cells react minimally to self-antigens continually presented by CD1d molecules in the steady state (49). To examine whether T cell activation could be induced by adipocyte CD1d expression in the presence of iNKT cells, we performed the same adipose-specific gene transfer experiments in wild-type mice. At week 7 after AAV-CD1d injection, neither T cell activation (Fig. 6A, 6B) nor dysregulation of adipocyte functional genes (Fig. 6C) was observed in wild-type VAT, although AAV-CD1d resulted in >12-fold higher CD1d expression in VAT adipocytes.

We repeated this experiment to examine an earlier time point, that is, week 4 after AAV injection, in wild-type mice. Overall, we did not observe any major immune change, but rather a slight yet statistically significant decrease in the absolute number of CD4⁺ T cells was observed in VAT (Fig. 7B). These data suggest that the T cell activation and expansion is caused by the introduction of CD1d in VAT adipocytes of CD1d KO mice, with CD1d behaving as a perceived alloantigen in the absence of adipocyte iNKT cells. To further determine whether the recognition of CD1d expression in VAT adipocytes of CD1d KO mice as a strong T cell alloantigen in this context is specific, we performed selective GFP gene transfer to VAT adipocytes in CD1d KO mice and did not observe T cell activation (Fig. 8A).

To further determine whether the expanded CD8⁺ T cells are derived from iNKT cells, we isolated VAT CD8⁺ T cells from CD1d KO mice by a flow cytometry cell sorter. We also purified CD8⁺ T cells, excluding CD1d-tetramer⁺ cells, from the spleen of and iNKT cells (CD3⁺CD1d-tetramer⁺) from the liver of untreated wild-type C57BL/6 mice by a flow cytometry cell sorter. PCR amplification of 24 TCR Vβs using spectratyping primers (32, 33) showed that iNKT cells typically express an invariant Va14-Ja18 TCR α-chain paired with either Vβ8, Vβ7, or Vβ2 (50, 51). We found that VAT CD8⁺ T cells from CD1d KO mice did not express Vβ7 and Vβ8, different from control iNKT cells, suggesting that the expanded CD8⁺ T cells unlikely are iNKT cells. Compared to the CD8 T⁺ cell control, VAT CD8⁺ T cells have less Vβ variation. This can be due to adipose tissue specificity or some VAT CD8⁺ T cells are under clonal expansion (Fig. 8B). This result suggests that the introduction of the CD1d molecule into the VAT adipocytes of CD1d KO mice causes allogeneic responses based on at least in part its particular molecular structure.

We did not observe changes in the absolute number of iNKT cells in the wild-type mouse VAT at either week 4 or week 7 after AAV-CD1d injection (Figs. 6B, 7B). We sought to determine whether the presence of iNKT cells specifically resulted in the absence of a T cell response to expression of CD1d in VAT adipose tissue in wild-type mice. We therefore selected another iNKT cell–deficient mouse strain with an otherwise normal repertoire of lymphocytes, that is, the invariant TCR α-chain Jα18 KO mouse. Following our selective introduction of CD1d into VAT adipocytes of the invariant TCR α-chain Jα18 KO mice, we did not see any T cell activation (Fig. 9), suggesting that the absence of the iNKT cells is not itself responsible for the T cell activation observed when CD1d is selectively expressed in VAT adipocytes of CD1d KO mice.

In this study, we observed that rAAV-mediated gene transfer of CD1d to VAT adipocytes in CD1d KO mice resulted in a robust autoimmune response, including a massive increase in SVF cellularity, T cell expansion and activation, and adipocyte inflammation and apoptosis. Depletion of CD8⁺ T cells significantly attenuated inflammasome activation and abolished the dysregulation of adipocyte functional genes induced by the introduction of adipocyte CD1d expression. Our studies reveal an adipocyte CD1d → CD8⁺ T cell → adipocyte inflammasome loop as the mechanism responsible for these observations (Fig. 10).

Most of the studies on CD1d to date have investigated its role in regulating iNKT cells (13, 52, 53). The CD1d KO mice are iNKT cell deficient, and adipocyte-specific CD1d KO mice display decreased numbers of iNKT cells in the adipose tissues and impaired responses to α-GalCer–induced iNKT cell activation (16, 54). Our results showed that restoring CD1d expression only in VAT adipocytes failed to rescue adipose iNKT cells in CD1d KO mice, which is in support of the notion that iNKT cells develop within the thymus and are thymus CD1d-dependent (13, 52, 53). Moreover, CD1d overexpression in VAT of wild-type mice had no effect on the percentage or absolute numbers of iNKT cells (Figs. 6A, 6B, 7), suggesting that CD1d alone is insufficient to activate iNKT cells and likely requires additional signals. These results are consistent with prior reports demonstrating that CpG-induced activation of iNKT cells is more likely dependent on IL-12 with a minimal role for CD1d (55–57).

In the absence of iNKT cells, selective CD1d expression in adipocytes resulted in a massive increase of lymphocytes within the transduced VAT without changes in blood, thymus, or spleen cell numbers at 4 and 7 wk postinjection of AAV-CD1d. Although T cells, B cells, and NK cells all increased in number, the greatest increase in both percentage and absolute numbers was in T cells, with a decrease in naive T cells and an increase in effector T cells for both CD4 and CD8 T cell subsets, resulting in a lower CD4/CD8 ratio (Fig. 2). SVF gene expression profiling showed upregulation of CD3ε and various molecules associated with proliferation, T cell activation, and trafficking (Fig. 3D).

One key finding of the study is that CD1d gene transfer resulted in activation of NLRP3 inflammasome in the adipocytes associated with downregulation of genes critically involved in adipogenesis, lipolysis, and lipogenesis (Fig. 3), suggesting impaired adipose function, which was consistent with the noted elevated level of proteins involved in the apoptotic pathway. The inflammatory response therefore likely explains the significantly reduced mass of the CD1d-transduced VAT. Regarding the cause of the adipocyte inflammation, several lines of evidence support CD8⁺ T cells as a key mediator. First, the NLRP3 inhibitor MCC950 significantly inhibited the adipocyte level of IL-1β but had no effect on the activation and proliferation of T cells induced by adipocyte gene transfer of CD1d (Fig. 4), suggesting that T cell expansion is independent of adipocyte inflammation. Second, CD1d expression per se was unable to induce adipocyte inflammatory responses at an early time point when the expansion and activation of T cells did not occur, ruling out the intracellular signaling of CD1d as a cause of inflammasome activation (Supplemental Fig. 4). Third, the depletion of CD8⁺ T cells significantly attenuated the upregulation of adipocyte inflammasome genes and completely abolished the downregulation of adipocyte functional genes associated with CD1d gene transfer (Fig. 5). Of note, CD8 depletion reduced the total T cell numbers by >50% in the CD1d-treated VAT, confirming that CD8⁺ T cells account for a major population induced by adipocyte CD1d. Meanwhile, the expansion of CD4⁺ T cells remained, which might underlie the incomplete blockade of inflammasome induction. Nevertheless, attenuating the induction of inflammatory genes by ∼50% was sufficient to fully normalize the levels of adipokines and adipocyte functional genes, indicating that inflammasome activation is upstream of the dysregulation of adipocyte functional genes.

Adipose-resident T cells have been increasingly recognized as an important player in obesity-associated adipose inflammation and systemic insulin resistance (2, 58, 59). In diet-induced obesity, CD8⁺ T cells contribute to adipose inflammation through the promotion of proinflammatory M1 polarization of adipose macrophages (59, 60). Future studies will assess whether adipocyte CD1d-induced CD8⁺ T cell activation affects adipose macrophages and whether these macrophages serve as a mediator of adipocyte inflammation. Furthermore, Th1 CD4⁺ T cells are also implicated in adipose inflammation through direct interactions with both adipocytes and indirect interactions via macrophages (3, 4). Investigations on the subtypes of CD4⁺ T cells and their roles in CD1d-induced adipocyte inflammation are therefore also warranted. In addition to the massive T cell expansion, expansion of NK cells and B cells, often considered proinflammatory, was observed in CD1d-expressing VAT, although at a much lesser extent compared with T cells (Fig. 2). It will be of interest to investigate whether T cell activation leads to NK and B cell expansion in future studies.

Another interesting question is how CD1d expressed in adipocytes triggers T cell activation and expansion. In previous studies using the Rec2 serotype AAV vectors for adipose gene delivery, no overt changes of T cells numbers or ratios were observed in adipose tissues, including viral vectors carrying no transgene, various endogenous genes (e.g., VEGF, IL-15, connexin 43, SWELL1) (26, 27, 29, 61), or foreign genes (e.g., GFP, Cre recombinase) (23, 30). Thus, the robust activation of T cells in AAV-CD1d–transduced fat could not be attributed to the response to any foreign Ags, but rather a unique immune response against CD1d protein or CD1d-presented self-antigens. Actually, we confirmed the absence of T cell activation after gene transfer of the foreign Ag GFP to CD1d KO mice (Fig. 8). Several more lines of evidence support the notion that this observation is a specific alloreactive immune response to CD1d expression in VAT adipocytes by CD8⁺ T cells in CD1d KO mice. First, CD1d is structurally similar to MHC class I molecules, and allogeneic MHC molecules have been demonstrated to be potent immunogens for T cells due to their polygenesis and polymorphism for a variety of MHC-restricted T cells (62). Second, the same AAV-CD1d treatment failed to induce T cell activation in wild-type mice (Figs. 6, 7). Finally, the activated T cells in CD1d KO mice were CD1d tetramer negative, indicating that they were not thymus CD1d-dependent iNKT cells (52, 53). One study by Park et al. (8) clearly demonstrated that classical MHC-restricted CD8⁺ T cells do not cross-react with CD1d. This suggests that T cell activation and expansion in our model is not likely caused by direct interaction of the TCR with CD1d as an Ag-presenting molecule. Instead, the activation of T cells induced by CD1d rescue in CD1d KO mice is likely due to the autoimmune response against peptides derived from the CD1d protein or CD1d-presented self-antigens. The gene transfer of CD1d to VAT adipocytes in invariant TCR Jα18 KO mice failed to induce T cell activation as observed in CD1d KO mice, suggesting that the absence of iNKT cells in the CD1d KO mouse is not singularly responsible for induction of the inflammation that we observed when in the CD1d KO VAT adipocytes are selectively expressing CD1d (Fig. 9). TCR Vβ repertoire analysis showed that the expanded CD8⁺ T cells unlikely are iNKT cells because they expressed different Vβ gene products without Vβ7 and Vβ8. It is possible that non-invariant CD1d-restricted T cells induced the inflammation. Further investigations are underway to address this question.

The inflammasomes are a multi‐protein complex activating caspase‐1 that mediates the activation and secretion of the proinflammatory cytokines IL‐1β and IL-18 (63, 64). Obesity is associated with activation of NLRP3/caspase-1/IL‐1β in adipose tissues. Inhibition of caspase-1 has been shown to alleviate obesity-associated systemic insulin resistance (65). In the present study, we found a robust activation of the NLRP3/caspase-1/IL‐1β pathway in the CD1d-treated VAT, but systemic changes (weight, food intake, GTT, circulating leptin, and adiponectin) were not observed. The lack of notable systemic metabolic consequences may be explained by multiple factors. In this study, only VAT was transduced and the inflammation of one fat depot might not be sufficient to cause systemic metabolic disturbance, particularly in experiments with relatively short duration. In addition, imposing a local metabolic challenge may help to reveal the systemic metabolic phenotypes. In fact, both global CD1d KO mice and adipocyte-specific CD1d KO mice display impaired insulin sensitivity under a high-fat diet but not under normal diets (12, 16). In future studies, we will repeat the AAV-CD1d experiment in the DIO model and examine the long-term metabolic outcomes.

In summary, our studies uncover an adipocyte CD1d → CD8⁺ T cell → adipocyte inflammasome cascade in CD1d KO mice treated with AAV-mediated gene transfer of CD1d specifically to VAT adipocytes. We observed that this leads to massive expansion and activation of T cells with skewing to CD8⁺ T cells within the transduced VAT caused by an autoimmune response to an alloantigen. The CD8⁺ T cells in turn induce an adipocyte inflammasome pathway and subsequent dysregulation of adipocyte functions, resulting in a significant loss of adipose tissue.

We thank the National Institutes of Health Tetramer Core Facility for providing CD1d tetramers (α-GalCer analog PBS-57).

L.C. and W.H. are inventors of a provisional patent application related to the liver-restricting AAV vector. The other authors have no financial conflicts of interest.