Adipose tissue dendritic cell signals are required to maintain T cell homeostasis and obesity-induced expansion

Abstract

Adipose tissue derived chronic inflammation is a critical component of obesity induced type II diabetes. Major histocompatibility complex II (MHCII) mediated T cell activation within adipose tissue is one mechanism that contributes to this phenotype. However, the contribution of dendritic cells as professional antigen presenting cells in adipose issue has not previously been explored. Using Itgax^Cre x MHCII^fl/fl (M11cKO) mice we observed adipose tissue specific changes in adipose tissue leukocytes. While there was a complete knockout of MHCII in dendritic cells, MHCII was also absent on the majority of macrophages. This resulted in reduction of TCR expression in CD4⁺ T cells in obese adipose tissue, and an increase in CD8⁺ and CD4⁺ CD8⁺ double positive T cells with decreased CD4⁺ T cells independent of diet type. Increased CD8⁺ cells were not observed in the spleen, suggesting adipose tissue T cell regulation is tissue specific. In vitro studies demonstrated more potent antigen presentation function in adipose tissue dendritic cells compared to macrophages. Obese M11cKO mice had decreased CD11c⁺ adipose tissue macrophages. Despite the changes of immune cellularity in adipose tissue, M11cKO largely did not change inflammatory gene expression in adipose tissue and did not demonstrate differences in glucose and insulin intolerance. Overall MHCII expression on CD11c⁺ cells is important for maintaining CD4⁺ and CD8⁺ adipose tissue T cells, but these cellular changes fail to alter inflammatory output and systemic metabolism.

1. Introduction

Type II diabetes has quickly become a widely prevalent disease in the US. The latest CDC statistics indicate that 9.4% of the entire population has diabetes, and that 87.5% of this group is overweight or obese (Centers for Disease Control and Prevention, 2017). Obesity induced diabetes is characterized by insulin resistance, which is mechanistically linked to adipose tissue dysfunction and chronic inflammation (Kahn, Hull and Utzschneider, 2006,Lee and Pratley, 2005,Despres and Lemieux, 2006,Gustafson, Hammarstedt, Andersson et al., 2007). Preclinical and clinical studies have shown chronic low-grade inflammation is a critical link between obesity and insulin resistance, and inhibiting adipose tissue inflammation improves insulin sensitivity and metabolic capacity (Hotamisligil, Shargill and Spiegelman, 1993,Lumeng and Saltiel, 2011,Shu, Benoist and Mathis, 2012,Xu, Barnes, Yang et al., 2003). Therefore understanding the mechanisms inducing adipose tissue inflammation and developing methods to reverse this phenotype are critical steps in developing diabetes treatments.

Components of both the innate and adaptive immune system reside within adipose tissue and their inflammatory output shifts with increased adiposity. Lean fat predominantly contains cells with anti-inflammatory phenotypes, such as arginase expressing adipose tissue macrophages (ATMs), group 2 innate lymphoid cells (ILC2) and regulatory T cells (Tregs), which maintain metabolic homeostasis (Lumeng, Deyoung, Bodzin et al., 2007,Brestoff and Artis, 2015,Feuerer, Herrero, Cipolletta et al., 2009). However with obesity, cross-talk between adipocytes and tissue resident leukocytes leads to adipose tissue inflammation characterized by qualitative and quantitative changes in adipose immune cells. With obesity, both CD4⁺ Th1 and CD8⁺ adipose tissue T cells (ATTs) are activated and contribute to insulin resistance in obese conditions (Winer, Chan, Paltser et al., 2009,Nishimura, Manabe, Nagasaki et al., 2009). CD4⁺ T conventional (Tconv) ATTs have restricted TCR repertoires in obese adipose tissue, indicating that they are being activated and proliferating by in response to specific antigens residing within the adipose tissue (Yang, Youm, Vandanmagsar et al., 2010). Adipose tissue Tregs (ATTregs) comprise ~40% of CD4⁺ T cells in lean mice, have a distinct transcriptional profile, and have a restricted TCR repertoire that contributes to metabolic homeostasis (Feuerer et al., 2009,Cipolletta, Feuerer, Li et al., 2012). The quantity of ATTregs is diminished with obesity-induced inflammation contributing to the pro-inflammatory environment.

ATT content is dependent on MHC class II and the antigen presenting cells (APCs) that controls ATT cell content and activation appears to be diverse (Cho, Zamarron, Muir et al., 2016,Deng, Lyon, Minze et al., 2013,Kolodin, van Panhuys, Li et al., 2015). Adipocytes have been implicated in antigen presentation, but professional phagocytes are also prominent features of adipose tissue (Deng et al., 2013). An extensive resident ATM population exists in lean mice, and with obesity activated CD9/CD11c⁺ macrophages accumulate in conjunction with pro-inflammatory CD4⁺ Th1 cells (Hill, Lim, Kim et al., 2018,Strissel, Defuria, Shaul et al., 2010). ATMs have potent APC capacity required for ATT activation in obesity. Using Lyz2^Cre x MHCII^fl/fl mice, MHCII expression in ATMs was shown to be required for the obesity-induced generation of CD4⁺ conventional Th1 ATTs, but not the maintenance of adipose tissue CD4⁺ cells on normal diets (Cho, Morris, DelProposto et al., 2014). This mouse model also targeted some adipose tissue dendritic cells (ATDC), which have been shown to contribute to the inflammatory response to obesity in visceral adipose tissue depots and may explain the ability to maintain normal CD4⁺ ATT numbers (Cho et al., 2016,Macdougall, Wood, Loschko et al., 2018,Bertola, Ciucci, Rousseau et al., 2012).

It is currently not well understood if ATDCs and ATMs have redundant functions in activating ATT cells, or whether they possess unique activation or polarization of ATTs in obese states. Since ATDC are the main CD11c⁺ cell in lean adipose tissue (Cho et al., 2016), we generated mice with knockout of MHCII (H2-Ab1) in CD11c expressing cells using CD11c(Itgax)^cre drivers crossed to H2-Ab1^fl/fl transgenic mouse (M11cKO mice). We hypothesized that inhibiting ATDC antigen presentation using CD11c^Cre would disrupt Treg populations and subsequently worsen glucose tolerance in lean mice. We observed that M11cKO mice had fewer CD4⁺ adipose tissue T cells along with increase in CD8⁺ ATT independent of diet type. Additionally M11cKO had impaired ATT TCR expression on CD4⁺ Treg and Tconv in obese settings. This phenotype was related to loss of MHCII in ATDC as well as most ATMs and contributed to a slight decrease in the accumulation of CD11c⁺ ATMs in obese mice. In vitro studies demonstrated that ATDC had stronger APC function than ATMs to explain the requirement for MHCII in ATDC for CD4+ cell maintenance. Overall our results suggest differential regulation of ATT cell maintenance and activation by ATDC and ATM.

2. Materials and Methods

2.1. Animal Studies

C57Bl/6J, CD11c^Cre (JAX stock 008068, B6.Cg-Tg(Itgax-cre)1–1Reiz/J (Caton, Smith-Raska and Reizis, 2007)) and MHCII^fl/fl (B6.129X1-H2-Ab1^tm1Koni/J) mice were obtained from the Jackson Laboratory. MHCII^fl/fl x CD11c^Cre (M11cKO) mice were generated by breeding Itgax/Cd11c^Cre with MHCII^fl/fl mice. Cre-negative and MHCII^fl/+ littermates were used as controls. Male mice were fed ad libitum either a normal diet (LabDiet PicoLab 5L0D 4.09kcal/gm 29.8% protein, 13.4% fat, 56.7% carbohydrate) or a high fat diet (HFD; Research Diets D12492, 5.24kcal/gm 20% protein, 60% fat, 20% carbohydrate) beginning at 6 weeks of age. All mouse procedures were approved by the University Committee on Use and Care of Animals at the University of Michigan and were conducted in compliance with the Institute of Laboratory Animal Research Guide fore the Care and Use of Laboratory Animals.

2.2. Metabolic Evaluation

Glucose tolerance tests (GTT) and Insulin Tolerance Tests (ITT) were performed after a 6 hour fast. For GTTs, mice were injected IP with D-glucose (0.7g/kg). For ITTs, mice were injected IP with human insulin (Humulin 1 U/kg). For both GTTs and ITTs, blood glucose concentrations (mg/dL) were measured at 0, 15, 30, 45, 60, 90, and 120 mins after injection from tail nick with a glucometer.

2.3. Isolation of Adipose Tissue SVF and Flow Cytometry Analysis

The stromal vascular fraction (SVF) was isolated from whole adipose tissue as previously described (Cho, Morris and Lumeng, 2014). Briefly, adipose tissue depots were dissected and weighed. Tissue was then mechanically disrupted by mincing, and chemically digested by rocking tissue in 1mg/ml collagenase IV (Sigma Aldrich) at 37C for 30 mins. Cells were then quenched with RPMI media and filtered through 100nm mesh prior to RBC lysis and subsequent filtering with 70nm mesh filters.

Cells were incubated in Fc Block for 5 minutes on ice and stained with indicated antibodies for 30 minutes at 4C: Anti-Mouse CD45 eFluor 450 [48–0451-82], Anti-Mouse CD8a FITC [11–0081-82], Anti-Mouse MHC Class II (I-A/I-E) PE-Cy7 [25–5321-82], Anti-Mouse CD4 APC [17–0041-82], Anti-Mouse CD11c APC-eFluor® 780 [47–0114-80], from eBioscience, anti-mouse CD3ε PerCP/Cy5.5 [145–2C11] from Biolegend, and anti-Mouse CD64 a and b Alloantigens PE [558455] from BD Pharmingen Stained cells were washed twice with FACS buffer and fixed for intracellular staining with Anti-Mouse Foxp3 PE [12–4771-82] from eBioscience, using a FoxP3 transcription kit (BD Biosciences). Analysis was performed on a FACSCanto II Flow Cytometer and analyzed with Flow Jo software (Treestar).

Gating Strategy

Representative flow gating of eWAT was performed as the following: cells were gated on Singlets (FSC-W x FSC-H) → Scatter → CD45⁺→ and then separated by CD3⁺ and CD3⁻. CD3⁺ cells were further sub-gated into specific T cell populations (CD4 by CD8). The CD4+ population was further divided into FoxP3+ (Treg) and FoxP3- (Tconv) groups. CD3⁻ cells were used to assess APCs. ATDCs are CD11c+ CD64−, and ATMs are CD64+.

2.4. Gene Expression Analysis

RNA was extracted from tissue by homogenizing epididymal white adipose tissue (eWAT) in Trizol, followed by phase separation with chloroform. The aqueous fraction was then processed using RNeasy Midi Kit purification (Qiagen) following standard protocol. cDNA was generated from 0.5ug total RNA using high-capacity cDNA reverse transcription kits (Applied Biosystems). Power SYBR Green PCR Master Mix (Applied Biosystems) and the StepOnePlus System (Applied Biosystems) were used for real-time quantification PCR. Acidic ribosomal phosphoprotein P0 (Arbp) expression was used as an internal control for data normalization. Samples were assayed in triplicate, and relative expression was determined using the 2-ΔΔCT method.

2.5. Adipocyte sizing

Adipose tissue was fixed with 10% formalin, and imaged with an immunofluorescence microscope at 10X. Adipocyte circumference was measured using ImageJ.

2.6. CFSE Staining and T cell Proliferation Assays

Adipose tissue depots were pooled to maximize collection of APCs. SVF was extracted and stained with CD45, CD11c, CD64, and CD3 flow antibodies (listed above). A FACS sort was performed to isolate ATMs, ATDCs, CD11c- CD64- CD45⁺, and CD45- cell populations using a Sony MA900 sorter. CD3+ cells were excluded. Each of the 4 cell populations was plated overnight, before being pulsed with 10ug/ml of OVA or BSA for 6 hours. CD3⁺ cells were isolated from the lymph nodes and spleen of an OTII mouse. These cells were then stained with CFSE and co-cultured with APCs at a 1:10 ratio. After 4 days cells were taken and stained for flow cytometry analysis. Gating strategy is as follows: Singlets (FSC-W x FSC-H) → Scatter → Live cells (Live/Dead Fixable Violet Dead Cell Stain Kit, Invitrogen) → CD4+ → FoxP3− → CD25+ CFSE−

2.7. Hepatic triglyceride content

Livers were weighed, snap frozen in liquid nitrogen, and stored at −80°C. 50–100mg of frozen liver was cut and homogenized in buffer containing NP-40, Tris-Hcl, and NaCl. Chloroform was added, vortexed, and samples dried overnight in a speed vacuum. Another round of chloroform extraction was performed on dried samples before residual lipid was dissolved in butanol. Triglycerides were measured using the Infinity Triglyceride Assay Kit (Sigma), and normalized to the initial mass of tissue homogenized.

2.8. Statistical Analysis

All values are reported as mean ± SEM. Differences between groups were determined using unpaired, two-tailed Students t Test or two-way ANOVA with Tukey post hoc tests with Graph Pad Prism 7 software. P values less that 0.05 were considered significant.

3. Results

3.1. MHCII^fl/fl x CD11c^Cre (M11cKO) deletes MHCII expression in ATDCs and most ATMs and decreases CD11c⁺ ATMs in obese mice.

In lean mice, the primary CD11c⁺ adipose tissue myeloid cells are ATDC (Cho et al., 2016). Therefore, we sought to eliminate MHCII expression in ATDC by generating conditional dendritic cell knockout mice by crossing MCHII^fl/fl x CD11c^Cre(Itgax^cre) mice (M11cKO). Although transgenic lines such as Zbtb46^Cre are more specific for targeting conventional DCs, off target effects like colitis that prevent weight gain and illicit off target gut inflammation limit their utility for obesity studies centered on adipose tissue biology (Satpathy, Kc, Albring et al., 2012,Loschko, Schreiber, Rieke et al., 2016). M11cKO and littermate controls were fed either normal diet (ND) chow or high fat diet (HFD) for 15 weeks. Body weight of control and M11cKO mice on a ND did not differ significantly. After HFD feeding, both genotypes gained significant weight, but HFD fed M11cKO weighed slightly less than littermate controls (Figure 1A). Organ weight at termination demonstrated no significant differences in non-adipose tissues such as liver and spleen. Epididymal (eWAT) and inguinal (iWAT) adipose tissue mass in HFD fed M11cKO mice were mildly, but significantly, decreased compared to littermate controls (Figure 1B). When normalized to total body weight, fat mass in M11cKO mice was not significantly different than controls (Supplemental Figure 1A).

Figure 1:

CD11c^Cre Reiz drives MHCII knockdown in both adipose tissue dendritic cells and macrophages.

(A) Terminal body weight of transgenic M11cKO mice and littermate controls on ND or HFD diet after 15 weeks of HFD feeding

(B) Tissue weights of Spleen, Liver, eWAT, and iWAT

(C) Representative flow plot illustrating ATDC, ATM, and CD11c- CD64- populations using CD64 and CD11c expression on CD45⁺ CD3⁻ cells. Frequencies of ATDCs, ATMs, and CD11c- CD64- of a percentage of the stromal vascular fraction (SVF) shown on the right

(D) Representative flow cytometry histograms and frequency of MHCII expression on ATDCs

(E) Representative flow cytometry histograms and frequency of MHCII expression on ATMs

(F) Representative flow cytometry histograms and frequencies of CD11c⁺ and CD11c⁻ ATMs as a percentage of SVF

(G) Representative histograms showing MHCII expression and frequencies of MHCII expression on CD11c⁻ and CD11c⁺ ATMs.

* p < 0.05, ** p < 0.01, *** p < 0.0001. N = 5–17 mice/group

Adipose tissue leukocytes in eWAT were assessed by flow cytometry (Strategy shown in Supplemental Figure 1B). In both lean and obese mice, the quantity of total ATDCs (CD45⁺CD64⁻CD11c⁺) and ATMs (CD45⁺CD64⁺) were not different between M11cKO and controls. The frequency of ATDCs as percentage of immune cells was not significantly changed by diet or genotype. As expected, the frequency of total ATMs from the HFD fed cohorts was significantly elevated compared to ND controls. Obese M11cKO mice had a trend towards less total ATM content. CD11c⁻ CD64⁻ cells were unchanged by genotype, but decreased in HFD- fed M11cKO compared to ND controls (Figure 1C). Using flow cytometry, we verified that M11cKO mice had absent MHCII expression on ATDCs. While nearly all ATDCs from controls expressed MHCII, almost all ATDC from Mc11KO mice had lost MHCII expression (2.27% ± 0.58 of M11cKO ATDC expressed MHCII) (Figure 1D). M11cKO mice also lacked MHCII expression in the majority of ATMs in ND and HFD fed mice (12.2%± 3.14 of the ATMs retained MHCII expression in M11cKO mice) (Figure 1E).

Since obesity is known to induce CD11c⁺ ATMs, we quantified CD11c⁺ and CD11c⁻ ATMs in control and M11cKO mice (Figure 1F). M11cKO and controls did not differ in the proportion of CD11c⁻ or CD11c⁺ ATMs in lean mice fed ND. M11cKO mice fed HFD had fewer CD11c⁺ ATMs compared to controls. There were no differences in CD11c⁻ ATMs in HFD fed M11cKO mice. Given the significant reduction in MHCII expressing CD64⁺ ATMs, we assessed MCHII expression on both CD11c⁺ and CD11c⁻ subtypes in the M11cKO and controls (Figure 1G). Surprisingly, both CD11c⁺ and CD11c⁻ ATMs had reductions in MHCII expression in M11cKO mice. This suggests that that there may be transient CD11c expression in the CD11c⁻ population or precursors that give rise to CD11c⁻ ATMs. Overall, M11cKO mice lack MHCII in all ATDCs and also have lost MHCII in the most, but not all, ATMs. This contrasts with Lyz2^Cre x MCHII^fl/fl, which targets all ATMs, and most, but not all, ATDC (Cho et al., 2014). A summary of similarities and differences between Lyz2Cre and CD11cre are summarized in (Table 1).

Table 1 -.

ATDC v ATM MHCII contribution to systemic metabolism and inflammatory properties in ND and HFD-fed mice

	ND		HFD
	M11cKO v WT Con (ATDC)	MMKO v WT Con (ATM)	M11cKO v WT Con (ATDC)	MMKO v WT Con (ATM)
Systemic Metabolism
GTT AUC	Unchanged	Unchanged	Unchanged	Improved
ITT AUC	Unchanged	Unchanged	Unchanged	Improved
Inflammatory Properties
ATM CD11c expression	Unchanged	Unchanged	Decreased	Decreased
ATDC MHCII expression	Complete KO	Majority KO	Complete KO	Majority KO
ATM MHCII expression	Majority KO	Complete KO	Majority KO	Complete KO
Splenic Tconv	Decreased	Unchanged	Decreased	Unchanged
Splenic Treg	Decreased	Unchanged	Decreased	Unchanged
Splenic CD8+	Unchanged	Unchanged	Increased	Unchanged
Splenic DPT	Unchanged	Unchanged	Increased	Unchanged
eWAT Tconv	Decreased	Unchanged	Decreased	Decreased
eWAT Treg	Unchanged	Unchanged	Decreased	Decreased
eWAT CD8+	Increased	Increased	Increased	Unchanged
eWAT DPT	Increased	Unchanged	Increased	Unchanged

3.2. M11cKO mice have increased adipose tissue CD8⁺ and decreased CD4⁺ T cells

Previous studies have shown that deletion of MHCII in ATMs using Lyz2^Cre results in normal CD4⁺ levels in lean mice and a reduction of conventional CD4⁺ ATT cells in obese mice with subsequent metabolic improvement (Cho et al., 2014). We phenotyped ATT cells by flow cytometry in M11cKO mice and controls fed ND and HFD (Figure 2A). In both, lean and obese mice M11cKO mice had fewer CD4⁺ ATTs compared to WT controls suggesting that ATDC MHCII is required for maintenance this population (Figure 2B). Analysis of Tconv and Treg within the CD4⁺ showed that Tconv were significantly decreased in lean and obese M11cKO mice while Treg were decreased only in the HFD fed M11cKO mice (Figure 2C-D). However, when Tregs are assessed as the frequency of total CD4+ T cells, neither diet nor genotype changed the frequency of this population (Supplemental Figure 2A). This decrease in CD4⁺ ATT was associated with a significant increase in CD8⁺ cells (Figure 2E) and an increase in CD4⁺ CD8⁺ double positive T (DPT) cells (Figure 2F), while CD4⁻ CD8⁻ double negative T cells remained unchanged (Figure 2G) in both ND and HFD mice. Since an increase in CD4⁺ CD8⁺ double positive cells are an aberration when found outside of the thymus, we assessed the thymus (Supplemental Figure 2B) and bone marrow T cells (Supplemental Figure 2C). M11cKO mice did not have increased DPT cells in either compartment and similar CD8 and CD4 single positive content, suggesting that these cells are not escaping thymic regulation, but may be forming in adipose tissue. When expressed as a percentage of the total SVF, there were no significant differences in total CD3⁺ T cell content in ND mice, but HFD fed M11cKO mice had a significant increase in total CD3⁺ T cells (Figure 2H). This increase was attributed to CD8⁺ ATT cells as CD8⁺ cells are significantly increased as a percentage of the SVF in both ND and HFD fed M11cKO compared to littermate controls (Figure 2I).

Figure 2:

M11cKO decreases CD4⁺ and increases CD8⁺ ATTs in eWAT.

(A) Flow plots illustrating differences in CD4 and CD8 expression on CD3⁺ cells..

(B) Frequency of total CD4⁺ of CD3⁺ T cells

(C) Frequency of FoxP3- CD4⁺ (Tconv) of CD3⁺ T cells

(D) Frequency of FoxP3⁺ CD4⁺ (Treg) of CD3⁺ T cells

(E) Frequency of CD8⁺ of CD3⁺ T cells

(F) Frequency of CD4⁺ CD8⁺ double positive T cells (DPT) as a percentage of total CD3⁺ T cells

(G) Frequency of CD4- CD8- double negative cells as a percentage of CD3⁺ T cells

(H) Frequency of CD3⁺ of SVF

(I) Frequency of CD8⁺ and CD4⁺ T cell populations as a percentage of total SVF

(J) RT-qPCR analysis of H2-kB (MHCI) gene expression in eWAT

(K) Median fluorescence intensity (MFI) of CD3 in the CD8⁺ population

(L) MFI of CD3 in Tconv cells

(M) MFI of CD3 in Treg cells.

* p < 0.05, ** p < 0.01, *** p < 0.0001. N = 5–17 mice/group

Because of the increase in CD8⁺ ATTs in an environment that has reduced MHCII expression, we assessed H2-kb(MHCI) gene expression to see if it was changed in M11cKO mice to compensate for decreased MHCII. However, MHCI expression was unchanged by diet or M11cKO in whole eWAT (Figure 2J). To examine any qualitative differences in CD4⁺ T cells in the M11cKO mice, CD3/TCR expression was quantified by flow cytometry. While CD8⁺ ATTs had similar CD3 expression in control and M11cKO mice (Figure 2K), both CD4⁺ Tconv (Figure 2L) and Treg (Figure 2M) had a reduction in CD3 expression as quantified by MFI in HFD fed mice. This suggests that lack of MHCII stimulus in M11cKO lead to decreased expression of the TCR on both CD4⁺ Tconv and Treg. In sum, MHCII expression on ATDC is required for the normal and high fat diet induced homeostasis of CD4⁺ ATTs and results in an increased in CD8⁺ ATT cells in lean and obese conditions.

3.3. Splenic T cell phenotype in M11cKO indicates ATT maintenance is adipose tissue specific.

We evaluated spleens of the M11cKO mice to determine if the T cell changes observed in adipose tissue were fat specific (Figure 3A). Similar to adipose tissue, splenic total CD4⁺, Tconv, and Treg (Figure 3B–D) cells were significantly decreased as a percentage of T cells in M11cKO independent of diet type – although the magnitude of the decrease was not as significant as seen in adipose tissue. When expressed as percentage of the CD3⁺ population, CD8⁺ T cells were unchanged in ND M11cKO mice, but HFD led to an increased frequency of CD8⁺ T cells in the M11cKO mice (Figure 3E). Splenic DPT cells were increased in frequency only in HFD fed M11cKO mice (Figure 3F). Overall, the quantity of DPT cells in the spleen were half of that observed in M11cKO eWAT. M11cKO mice had an increase in CD4⁻ CD8⁻ double negative T cells only in ND fed mice (Figure 3G). As a percentage of all splenocytes, lean ND fed M11cKO mice had significantly decreased CD3⁺ T cells compared to littermate controls, however no differences in splenic CD3⁺ cells were seen in HFD fed mice (Figure 3H). Unlike ATTs, splenic cells showed decreased CD4⁺ cells as calculated by frequency of splenocytes in the M11cKO mice, while there were no differences in CD8⁺ T cells (Figure 3I). The decreased in CD4⁺ were not attributable to specific decreases in splenic Treg or Tconv in HFD fed mice (Figure 3J). Unlike adipose tissue, CD3/TCR expression on splenic CD8⁺ (Figure 3K), Tconv (Figure 3L), and Tregs (Figure 3M) are all unchanged by either diet or genotype. Overall, our data demonstrates ATT specific regulation in M11cKO mice compared to the spleen, thymus and bone marrow, and suggests there are environment specific signals in WAT that drive CD8⁺ accumulation.

Figure 3:

M11cKO decreases CD4⁺ T cells in the spleen.

(A) Flow plots illustrating differences in CD4 and CD8 expression on CD3+ cells

(B) Frequency of total CD4⁺ of CD3⁺ T cells

(C) Frequency of FoxP3^- CD4⁺ (Tconv) of CD3⁺ T cells

(D) Frequency of FoxP3⁺ CD4⁺ (Treg) of CD3⁺ T cells

(E) Frequency of CD8⁺ of CD3⁺ T cells

(F) Frequency of CD4⁺ CD8⁺ double positive T cells (DPT) of CD3⁺ T cells

(G) Frequency of CD4⁻ CD8⁻ double negative of CD3⁺ T cells

(H) Frequency of CD3⁺ as a percentage of Splenocytes

(I) Frequency of CD8⁺ and CD4⁺ T cells as a percentage of Splenocytes

(J) Frequency of Tconv and Treg CD4⁺ T cells as a percentage of Splenocytes

(K) Mean fluorescence intensity (MFI) of CD3 in CD8⁺cells

(L) MFI of CD3 in Tconv cells

(M) MFI of CD3 in Treg cells

* p < 0.05, ** p < 0.01, *** p < 0.0001. N = 5–17 mice/group

3.4. ATDCs are more efficient antigen presenting cells than ATMs.

Despite the ability of Lyz2^Cre and CD11c^Cre drivers to delete MHCII in both ATMs and ATDC, there are striking differences between T cell regulation in these mice as Lyz2^Cre MHCII have normal number of CD4⁺ on a normal diet. Lyz2^Cre MHCII knockout mice have residual expression of MHCII on ATDC, while M11cKO mice have residual expression of MHCII in ATMs (~12%). We wanted to assess whether these differences could be due to different antigen presentation capabilities of ATDCs and ATMs. We hypothesized that ATDCs are more potent APCs than ATMs to explain the significant loss of CD4⁺ ATT cells in M11cKO mice. To test this, we assessed FACS sorted ATDCs, ATMs, CD45⁺ CD11c⁻ CD64⁻ cells (containing B cells), and CD45⁻ non-immune cells (used as a negative control). Sorted populations were used in an antigen specific reaction to assess their ability to induce splenic OTII T cell proliferation after OVA or a BSA treatment. After 4 days of co-culture, CD4⁺ cells were significantly increased in OVA pulsed ATDCs, but not other adipose tissue cell types (Figure 4A). ATDC and ATMs were able to induce Tregs when pulsed with OVA, however this was not seen with other leukocytes or CD45⁻ cells ((Figure 4B). In contrast, only ATDC were able to induce Tconv in response to OVA (Figure 4C). Overall, this suggests that differences in Lyz2^Cre v CD11c^Cre mediated MHCII KO may be due to the efficiency of the residual population of ATDC or ATM expressing MHCII in these models.

Figure 4:

ATDCs are more efficient antigen presenting cells than ATMs.

4 adipose tissue cell populations were FACS sorted from pooled WAT to assess their antigen presenting capacity by pulsing them with OVA and co-culturing with CFSE stained CD3⁺ OTII splenic T cells. CD45⁻ non-immune cells, CD45⁺ CD64⁺ (ATM), CD45⁺ CD11c⁺ CD64⁻ (ATDC) and CD45⁺ CD11c⁻ CD64⁻ populations were compared side-by-side.

(A) Number of Live CD4⁺ cells co-cultured with the APCs pulsed with BSA or OVA

(B) Number of CFSE^- CD25⁺ FoxP3+ (Treg) cells

(C) Number of CFSE^- CD25⁺ FoxP3- (Tconv) cells.

* p < 0.05, ** p < 0.01, *** p < 0.0001. N = 3 biological replicates

3.5. Adipose tissue inflammatory gene expression is unchanged in M11cKO mice.

We assessed pro-inflammatory cytokine gene expression in M11cKO mice. Despite reduced quantity of pro-inflammatory CD11c⁺ ATMs in M11cKO mice, Tnfa expression trends towards an increase in control HFD fed eWAT, and gene expression was significantly increased in HFD fed M11cKO mice compared to ND controls (Figure 5A). Ccl2/Mcp1 expression was significantly increased in obese eWAT and did not differ between M11cKO and WT mice (Figure 5B). T cell related cytokine genes were assessed. In lean M11cKO mice, there was increased expression of Ifng compared to controls. In obese mice, there was no significant difference in expression in M11cKO mice (Figure 5C). Il2 expression was significantly decreased between ND and HFD fed M11cKO groups. (Figure 5D). Collectively this data suggests that M11cKO do not have significant alterations in inflammatory gene expression compared diet matched littermate controls.

Figure 5:

M11cKO does not change inflammatory gene expression in eWAT.

Gene expression of whole snap frozen epididymal white adipose tissue, as assessed by RT-qPCR. Calculations were performed by normalizing to housekeeping gene Arbp, using 2ΔΔ^CT.

(A) TNFa

(B) MCP1/CCL2

(C) IFNy

(D) IL2

* p < 0.05, ** p < 0.01, *** p < 0.0001. N= 5–11 mice/group

3.6. M11cKO mice have normal glucose and insulin tolerance.

To determine whether antigen presentation by ATDCs influences systemic metabolism, we assessed glucose tolerance in lean (ND) and obese (HFD) fed mice. The ND fed controls and M11cKO mice showed no significant differences in GTT. With HFD feeding, both genotypes showed similar degrees of fasting hyperglycemia and glucose intolerance with similar area under the curve (AUC) measures (Figure 6A). Similar to the GTT, while HFD feeding induced insulin resistance based on ITT, there were no significant differences between M11cKO mice and controls (Figure 6B). Although metabolic parameters were not significantly different in M11cKO mice, frequency of CD11c+ ATMs positively correlates with insulin resistance (Supplement 3A), but not glucose intolerance (Supplement 3B), of both control and M11cKO mice. Histologic analysis of eWAT suggested architectural differences between HFD fed M11cKO and controls. Analysis of adipocyte size distribution showed an increase in small adipocytes in M11cKO mice (Figure 6C). Although M11cKO mice had decreased lipid content in adipose tissue, triglyceride storage was not preferentially shuttled to the liver. HFD feeding increased liver triglyceride, but M11cKO did not significantly change triglyceride content per gram of tissue (Figure 6D). Collectively this data shows a mild protection from HFD induced obesity in M11cKO mice associated with decreased fat mass that does not alter glucose or insulin tolerance.

Figure 6:

M11cKO does not alter glucose or insulin tolerance.

(A) Glucose tolerance test of mice at 12 weeks of age and area under the curve (AUC) quantifications.

(B) Insulin tolerance test of mice at 13 weeks of age and AUC quantifications.

(C) Distribution of adipocyte circumferences, with smaller bin size indicating smaller adipocyte size.

(D) Triglyceride content measured from snap frozen liver samples.

* p < 0.05, ** p < 0.01, *** p < 0.0001. N = 3–9 mice/group

4. Discussion

This study has several primary findings relevant to the study of immunometabolism. ATDC are more potent APCs compared to ATMs and other non-leukocyte cells. ATDC derived MHCII signals were required for CD4⁺ ATT cell maintenance as M11cKO mice had decreased CD4⁺ cells, and more CD8⁺ ATT cells in lean and obese settings. The dependency of CD4⁺ cells on DC derived MHCII was much more prominent in adipose tissue compared to other lymphoid organs such as the spleen.

Similar to other studies, we observed that targeting myeloid cells with CD11c^Cre mice expands beyond DC (Abram, Roberge, Hu et al., 2014). CD11c^Cre deleted MHCII expression on all ATDC and ~88% of CD11c⁺ and CD11c⁻ ATMs leaving a residual MHCII⁺ population of ATMs. This suggests that Itgax is transiently expressed in ATMs or ATM precursors regardless of their CD11c surface expression at the time of harvest. This observation may also be related to “leaky” expression of the transgene. We note that use of another Itgax-cre-EGFP mouse (Caton et al., 2007) provided similar results of deletion of MHCII in both ATMs and ATDC (data not shown). In contrast to the M11cKO mice, Lyz2^Cre targets all ATMs such that Lyz2^Cre MHCII^fl/fl have a residual MHCII⁺ population of ATDCs and result in significant phenotypic differences (Cho et al., 2014).

While Lyz2^Cre driven MHCII-KO resulted in no changes in T cells in lean mice and decreased Tconv in HFD fed mice, M11cKO led to different results (summarized in Table 1). Most notably ATT populations were dysregulated in M11cKO mice independent of diet type. The absence of ATDC MHCII expression resulted in a reduction in CD4⁺ Tconv and Tregs and expansion of CD8⁺ cytotoxic T cells and CD4⁺ CD8⁺ DPT cells. An increase in CD8⁺ and DPT cells was adipose tissue specific in M11cKO mice. In this setting, the small number of MHCII⁺ ATMs was not sufficient to maintain normal CD4⁺ ATT cell numbers. In the spleen, we observed what would be considered canonical T cell dysregulation in the absence of DC mediated antigen presentation, as there was a decrease in Tconv CD4⁺ cells. Since Tconv cell maintenance requires low affinity “tonic” TCR and MHCII interactions, loss of antigen presentation by dendritic cells may have driven this decrease in the CD4⁺ T helper cell population (Fischer, Jacovetty, Medeiros et al., 2007). In addition, there was a significantly reduced expression of the TCR in CD4⁺ ATT cells in HFD-fed M11cKO mice. TCR expression is dynamically regulated to tune inflammatory capacity of T cells. Since TCR expression regulates T cell functionality, it is possible that although the Tconv and Treg cells in HFD M11cKO are still present, that they are functionally impaired and unable to mount a typical inflammatory response (Labrecque, Whitfield, Obst et al., 2001). CD3/TCR downregulation is observed in T cells after cognate pMHCII activation in acute/subacute models of antigen dependent inflammation (Lamb, Skidmore, Green et al., 1983,Zanders, Lamb, Feldmann et al., 1983,Valitutti, Muller, Cella et al., 1995,Valitutti, Muller, Dessing et al., 1996,Cai, Kishimoto, Brunmark et al., 1997). Our observation that obesity decreases CD3/TCR expression in ATT cells in KO mice suggest that either there is an increase in pMHCII signals from other cells in AT or that our chronic model results in a downregulation of the TCR due to a lack of tonic signaling. We have examined residual MHCII+ cells in the KO mice and have not observed an increase in surface MHCII expression in any cell type. Therefore, we hypothesize that in the absence of MHCII, CD3/TCR downregulation is due to a loss of tonic signals from APCs in adipose.

Expansion of CD8⁺ ATT cells with MHCII manipulation is consistent with other reports. MHCII-KO leads to an influx of CD8⁺ T cells into tumors of cancer laden mice, leading to increased inflammation and smaller tumor burden (Chaoul, Tang, Desrues et al., 2018). Although it is unknown whether the increase in CD8⁺ T cells in M11cKO adipose tissue is due to recruitment due to adipose tissue secreted chemoattractants, the increased skewing of ATT towards CD8⁺ production compared to spleen suggest that local ATDC dependent signals generate this observation. In addition to work showing the majority of ATTs are of the T resident memory phenotype, work has also shown that tissue resident lymphocytes often have unique characteristics (Han, Glatman Zaretsky, Andrade-Oliveira et al., 2017,Fan and Rudensky, 2016).

The appearance of ATT DPT cells was also an unusual and intriguing finding. DPTs are usually only found in the thymus during T cell development, but DPT cells appear in several dysregulated autoimmune and chronic inflammatory diseases (Parel and Chizzolini, 2004). The significant increase in this population seems to be an ATT specific adaptation to loss of MHCII expression by ATDCs. Since the presence of DPT was not significantly increased globally in M11cKO mice, we speculate that CD8⁺ T cells may have some plasticity allowing them to assume a Tconv phenotype. This could be a compensatory mechanism that occurred due to decreased TCR expression on Tconv and Treg ATTs.

The dependence of CD4⁺ ATT cells on MHCII expression in all ATDC (M11cKO) but not all ATMs (Lyz2^Cre x MHCIIfl/fl) suggested that ATDC are stronger APC than ATMs or other potential APCs in adipose tissue. Our studies in FACS sorted populations show that ATDCs are superior at inducing activation and proliferation of T cells compared to ATMs and other CD11c⁻ CD64⁻ leukocytes containing B cells. It is unclear if MHCII expression on B cells in adipose tissue may play a role in ATT maintenance or activation and should be considered in future studies. Some studies have suggested that preadipocytes express MHCII and have downstream effects on ATTs (Deng et al., 2013), however we did not see evidence for this in our functional assay using CD45⁻ cells. Since these studies were performed with splenic OTII T cells, further studies need to be performed assessing ATT specific characteristics in lean and obese adipose tissue.

Like Lyz2^Cre x MHCII^fl/fl mice, M11cKO mice a reduction in pro-inflammatory CD11c⁺ ATMs in HFD fed mice compared to the controls, which would be suggestive of improved metabolic inflammation in adipose tissue. Despite this, glucose and insulin tolerance was not altered in M11cKO mice, while Lyz2^Cre x MHCIIfl/fl mice had improved metabolism. This may be related to the sustained increase in CD8⁺ ATT in M11cKO mice, which have been shown to participate in obesity-induced adipose tissue inflammation (Nishimura et al., 2009). Indeed, analysis of inflammatory gene expression in eWAT showed that M11cKO mice had similar levels of Tnfa, Ccl2, and Il2 expression. In fact, Tnfa and Ifng were higher in M11cKO mice than controls, which could be mediated by the increased CD8⁺ ATT cells.

The body weights of HFD-fed M11cKO mice were significantly lower than the diet-matched littermate controls. Body weight discrepancy results in smaller fat pad mass in the M11cKO mice, but not spleen or liver weights. Liver triglyceride content is not significantly different in M11cKO mice compared to their diet matched controls, indicating that lipid storage capacity of adipocytes is not dysregulated. Similar to whole body MHCIIKO mice (Cho et al., 2014), lean mass of M11cKO mice may be decreased compared to controls. Regardless of the net adipose tissue weight, the fat mass is not significantly smaller when it is normalized to body weight. This suggests that net adipose tissue weight may not be a clear predictor of metabolism or adipose tissue inflammation. There are several possible explanations for the different metabolic phenotypes observed in the Lyz2^Cre v CD11c^Cre driven MHCII knockouts. It has been shown that inhibiting dendritic cell mediated antigen presentation in the gut induces changes in the microbiome and associated inflammation (Loschko et al., 2016). Consequently, even if there is attenuated inflammation in the adipose tissue, chronic inflammation from the gut could lead to sustained cytokine release and systemic inflammation, and subsequent metabolic disorder. MHCII may contribute to processes like beta-oxidation, lipolysis, and mitochondrial production, however it is likely that impairment of nutrient absorption via gut inflammation may underlie the weight differences.

Overall diet independent and adipose tissue specific changes in M11cKO mice give insight into the importance of ATDC mediated antigen presentation signals. CD8⁺ and CD4⁺ CD8⁺ ATT cells increase in M11cKO in ND and HFD fed mice. However, CD11c⁺ ATMs are decreased and TCR expression is decreased CD4⁺ ATTs in HFD fed obese M11cKO mice. These cellular changes did not significantly improve the inflammatory output of obese adipose tissue and also did not improve glucose or insulin tolerance. M11cKO shows that ATM and ATDCs have cell type specific regulation of ATTs and resultant inflammatory and metabolic phenotype of obese mice.

Supplementary Material

2

Supplement 1:

(A) Tissue weights normalized to body weights.

* p < 0.05, ** p < 0.01, *** p < 0.0001

(B) Representative flow gating of eWAT. Cells were gated on Singlets → Scatter → CD45⁺ → and then separated by CD3⁺ and CD3⁻. CD3⁺ cells were further sub-gated into specific T cell populations, and CD3⁻ cells were used to assess APCs.

3

Supplement 2:

(A) Frequency of FoxP3⁺ (Treg) of CD4+ T cells

(B) CD4⁺ and CD8⁺ thymic T cells gated on CD3⁺.

(C) CD4⁺ and CD8⁺ bone marrow T cells gated on CD3⁺.

4

Supplement 3:

(A) Area under the curve (AUC) of ITT v CD11c+ ATM frequency in Control and M11cKO

(B) AUC of GTT v CD11c+ ATM frequency in Control and M11cKO.

R² and p-values were calculated in linear regression analysis

Highlights.

• Adipose tissue dendritic cells are more potent activators of T cells than adipose tissue macrophages

• MHCII mediated ATT activation is not required for induction of chronic low grade inflammation in adipose tissue with obesity

• MHCII^fl/fl x CD11c^Cre results in decreased CD4 adipose tissue T cells and increased CD8+ and CD4+ CD8+ double positive T cells

Abbreviations

MHCII

Majorhistocompatability Complex II

M11cKO

CD11c^Cre x MHCII^fl/fl

TCR

T cell Receptor

ATDC

Adipose Tissue Dendritic Cell

ATM

Adipose Tissue Macrophage

ATT

Adipose Tissue T cell

GTT

Glucose Tolerance Test

ITT

Insulin Tolerance Test

eWAT

Epididymal White Adipose Tissue

iWAT

Inguinal White Adipose Tissue

ND

Normal Diet

HFD

High Fat Diet

SVF

Stromal Vascular Fraction

MFI

Median Fluorescence Intensity

DPT

Double Positive T cells (CD4⁺ CD8⁺)