CD4+ T cells regulate glucose homeostasis independent of adipose tissue dysfunction in mice

Obesity is frequently associated with a chronic low‐grade inflammation in the adipose tissue (AT) and impaired glucose homeostasis. Adipose tissue macrophages (ATMs) have been shown to accumulate in the inflamed AT either by means of recruitment from the blood or local proliferation. ATM proliferation and activation can be stimulated by TH2 cytokines, such as IL‐4 and IL‐13, suggesting involvement of CD4‐positive T cells in ATM proliferation and activation. Furthermore, several studies have associated T cells to alterations in glucose metabolism. Therefore, we sought to examine a direct impact of CD4‐positive T cells on ATM activation, ATM proliferation and glucose homeostasis using an in vivo depletion model. Surprisingly, CD4 depletion did not affect ATM activation, ATM proliferation, or insulin sensitivity. However, CD4 depletion led to a significant improvement of glucose tolerance. In line with this, we found moderate disturbances in pancreatic endocrine function following CD4 depletion. Hence, our data suggest that the effect on glucose metabolism observed after CD4 depletion might be mediated by organs other than AT and independent of AT inflammation.


Introduction
Obesity has become an increasing burden for health systems around the world and has led amongst others to an unrivaled upsurge of type 2 diabetes [1]. The socioeconomic and personnel efforts to tackle the consequences of the obesity epidemic have become enormous [2]. It is therefore pivotal to elucidate the molecular foundation of obesity in order to develop targeted therapies.
Correspondence: Prof. Martin Gericke e-mail: martin.gericke@medizin.uni-halle.de On a cellular level, obesity is associated with a chronic lowgrade inflammation in the adipose tissue (AT) that is characterized by an accumulation of various immune cells, namely T cells, B cells, neutrophils, mast cells, and macrophages [3][4][5][6][7]. This inflammation is most pronounced in the visceral adipose tissue (VAT), which is believed to have a more detrimental effect on metabolism than subcutaneous adipose tissue (SAT) [8,9]. Macrophages constitute the largest proportion of immune cells in the AT of obese rodents and humans respectively [3], and the extent of the accumulation of macrophages in the AT seems to be closely correlated to the degree of insulin resistance [10]. Macrophages occur in a continuum of different phenotypes from which two types are generally discerned: M1 macrophages are induced by pro-inflammatory cytokines such as IFN-γ and express CD11c as a surface marker [11,12]. In contrast, activation of M2 macrophages occurs in a TH2-related microenvironment, particularly due to IL-4 and IL-13, and leads to expression of CD206 and CD301 [11,12]. Both populations seem to increase in number in the AT during obesity, but there is a marked increase in the proportion of M1 macrophages [13], which are located preferentially around dead adipocytes, in so-called crown-like structures (CLS). Further, it has been shown that the accumulation of adipose tissue macrophages (ATMs) during obesity occurs either by means of monocyte recruitment from the blood or proliferation of resident ATMs [14]. We and others have recently shown that the number of proliferating ATMs increases during obesity [15][16][17] and that this proliferation is at least partially mediated by cytokines such as IL4, IL13, and osteopontin (OPN), which are related to CD4-positive T helper cells (TH2 cells) [18][19][20]. This observation might suggest a link between macrophage activation and proliferation and CD4-positive TH2 cells.
CD4-positive T cells belong to the network of adaptive immune cells, which seem to play an important role in obesity-induced inflammation and insulin resistance [21]. To what extent the effect of lymphocytes on metabolism is immediate or mediated by macrophages and other immune cells is not known. Studies on mice with a RAG1 knockout, which abrogates the rodents capacity to produce any type of lymphocyte, have shown worsening of metabolic parameters, suggesting a protective role of lymphocytes [5]. It is assumed that the phenotype and function of lymphocytes alter during the course of obesity, leading to a loss of their protective properties. In line with this, depletion of CD3-positive T cells, CD20-positive B cells, and CD8-positive T cells in HFD-fed mice have shown attenuation of AT inflammation and improvement of metabolic parameters [4,5,22]. More accurately, depletion of CD3-positive T cells led to a marked upregulation of CD206-expressing ATMs while leaving adipocyte diameter unchanged. Likewise, depletion of CD8-positive T cells resulted in a decrease of CD11c-positive ATMs and a significant decrease of CLS formation. Both studies hence bring to attention the association between T-cell-mediated changes in glucose metabolism on one side and T-cell-mediated changes in VAT microenvironment on the other [5,22]. The effect of an in vivo CD4 cell depletion on AT inflammation and glucose homeostasis during obesity is as yet undescribed, although it has been suggested that certain distinct CD4-positive cell populations might contribute to immune-mediated alterations in AT inflammation and glucose metabolism [20,23]. Among these, VAT regulatory T cells (Tregs) have recently gained considerable attention for their potential to regulate AT and metabolic homeostasis [24]. In line with this, numerous studies have demonstrated that in situ expansion of Treg population in VAT led to attenuation of AT inflammation and improvement of glucose metabolism [23][24][25][26][27]. Importantly, in vivo loss-of-function models of VAT Tregs failed to show an effect on glucose metabolism in HFD-fed mice [24,[28][29][30].
CD4 cells, analogous to other immune cells in the AT, seem to increase in absolute numbers during obesity, yet inconsistencies exist as for the proportional changes: Some authors argue for a relatively stable CD4 population [31], whereas others suggest a proportional diet-dependent decrease of CD4 cells in the course of obesity [5,22]. Evidence further points towards a phenotypic switch among CD4 cells in the inflamed AT with a growing IFN-γ-positive population [5,31] and a loss of forkhead box P3-(FOXP3) and suppression of tumorigenicity 2-(ST2) positive cells during obesity [30,32,33]. Both FOXP3 and ST2 are generally believed to have anti-inflammatory properties in the context of obesity-induced inflammation and various experiments suggest a beneficial effect on glucose metabolism [23,25,[33][34][35]. The loss of these cells hence suggests the notion of a CD4 population that acquires a detrimental effect on glucose metabolism during the development of obesity.
The present work sought to elaborate further on the modus operandi of CD4-positive cells in the AT and aimed to elicit a possible impact on ATM activation and proliferation. We therefore established 3-day and 2-wk CD4 depletion protocols to evaluate short-term and potentially more direct changes as well as longterm effects on changes to the microenvironment. We show that short-term and long-term CD4 depletion have no apparent influence on either macrophage activation or macrophage proliferation in the AT, yet seem to have an impact on glucose homeostasis as measured by glucose tolerance testing (GTT). Furthermore, we show that long-term depletion of CD4-positive T cells causes dysregulation of the pancreatic endocrine axes, possibly contributing to the altered glucose metabolism following CD4 depletion. We conclude that in our study systemic CD4 depletion was shown to have no apparent effect on either CLS formation, ATM proliferation, or activation, yet entailed beneficial glucose tolerance potentially due to reasons beyond effects on VAT inflammation.

CD4-positive T cells show time-and diet-dependent changes in the inflamed AT
First, we performed immunofluorescence staining of CD3- (Fig. 1B) and CD4- (Fig. 1C) positive T cells to establish their presence and evaluate their distribution in the murine AT. We observed that CD3-and CD4-positive cells were mainly present in the CLS. Using flow cytometry, we then aimed to phenotype the CD3-/CD45-positive T cell population into their CD4-and CD8-expressing subsets. Therefore, we analyzed VAT of mice fed either an HFD or an NCD for 4, 12, or 24 wks. We detected a significant proportional increase of CD4-positive cells over time and independent of diet ( Fig. 1A). At 24 wks, the proportion of CD8-positive cells was significantly larger in HFD-fed compared to control animals. This diet-dependent difference was absent in the overall proportion of CD4 cells (Fig. 1A). However, further phenotyping of the CD4-positive subset revealed significant diet-dependent changes of CD4 subsets. IFN-γ-positive CD4 cells significantly dropped at 12 wks and then seemed to increase after 24 wks in the HFD-fed animals (Fig. 1D). In contrast, expression of both Treg-associated FOXP3 and TH2-associated ST2 among , and FOXP3 among CD4-positive ATLs of mice fed either an NCD (white; n = 5-10, cumulative data from more than five independent experiments) or an HFD (black; n = 5-10; cumulative data from more than five independent experiments) for 4, 12, or 24 wks. Data are presented as column bar graphs and as means ± SEM and were tested for statistical significance by Mann-Whitney U-test. *p < 0.05. **p < 0.01. ***p < 0.001. Scale bar represents 50 μm.
CD4-expressing lymphocytes showed a marked increase until 12 wks in both groups followed by a further increase in NCD-fed mice and a decrease in HFD-fed mice ( Fig. 1E and F). At 24 wks, expression of both ST2 and FOXP3 were significantly lower in the HFD-fed group compared to NCD-fed controls ( Fig. 1E and F). This reduced ST2 expression in obese mice at 24 wks was observed in both FOXP3-positive and FOXP3-negative cells, but was more pronounced in the FOXP3-negative population ( Fig. 1G and H).

Three-day depletion of CD4-positive cells does not alter ATM activation or proliferation
To study the direct impact of CD4 cells on AT inflammation, we established a 3-day depletion protocol. CD4-specific antibodies were shown to efficiently deplete CD4-positive T cells by ∼99%. We further showed that depletion of CD4-positive T cells was consistent in both adipose and splenic tissue ( Fig. 2C and D). Further, depletion did not result in significant weight loss averaging ∼1 g in both groups after 3 days of depletion (data not shown).
After HFD-feeding for 24 weeks and antibody treatment according to protocol, we analyzed the AT phenotype. Fat cell expansion and formation of CLS are signs of tissue remodeling under high-caloric diet and could be observed in all animals ( Fig. 3A and B). However, AT morphology as measured by adipocyte diameter did not alter between T-cell-depleted mice and mice treated with an appropriate isotype control (Fig. 3E). In order to evaluate AT inflammation, we counted the number of CLS in relation to adipocytes, which also revealed no significant difference (Fig. 3F). Another marker for the progression of AT inflammation is the proportion of ATMs among all stromal vascular cells. Again, no difference was observed (Fig. 3H). ATM activation as measured by the ratio of CD11c-positive ATMs (classically activated or M1 macrophages) in relation to CD206expressing ATMs (alternatively activated or M2 macrophages) was quantified using flow cytometry and showed no effect under CD4 cell depletion (Fig. 3I). We then went on to assess a possible impact of CD4 depletion on ATM proliferation. First, we stained VAT for the proliferation marker proliferating cell nuclear antigen (PCNA; Fig. 3C and D). Since CLS have been shown to be the primary site of ATM proliferation, we focused our analysis on cells Figure 2. Flow cytometric analysis of T cell depletion in VAT and spleen. Efficiency of T cell depletion was assessed using flow cytometry. Cells were stained for the lymphocyte markers CD45, CD3, CD4, and CD8 prior to analysis. (A and B) Representative flow cytometry plots for the relative distribution of T cell subsets following depletion of CD4-positive cells or injection of an appropriate isotype control (IgG2b). (C and D) CD4-specific antibodies efficiently depleted CD4-positive T cells (white) by ∼99% (n = 4-6 in both groups; cumulative data from 3 independent experiments). Depletion of CD4-positive T cells was consistent in VAT and spleen. Data are presented as means ± SEM and as column bar graphs and were tested for statistical significance by Mann-Whitney U-test. **p < 0.01. residing within these structures. Quantification of immunohistochemistry showed no significant changes under CD4 cell depletion (Fig. 3G). Second, we performed flow cytometric analysis of Bromodeoxyuridine (BrdU) staining ( Fig. 3J) that in accordance with the immunohistochemistry showed no significant difference between the groups. Likewise, mRNA levels of genes associated with ATM proliferation or activation showed no significant changes under CD4 cell depletion (Fig. 3K).
In the next step, we intended to compare the impact of CD4 cell depletion, as described above, with the effect of total T cell depletion. For that reason, we designed a similar 3-day depletion protocol for CD3 cells. However, we became aware that some mice in the CD3-depleted cohort showed signs of diarrhea and peritoneal inflammation in conjunction with a significant weight loss when compared with their IgG-injected control littermates (Fig.  S2N). The depletion itself showed a moderate depletion efficiency of ∼66% in both spleen and VAT (Fig. S1). Importantly, ATM proliferation and activation remained unaffected by CD3 depletion (Fig. S2). Due to the observed weight loss and inflammatory changes, we concluded that the interpretation of our data was severely biased and decided to show these results in the supplementary section only.

Two-week depletion of CD4-positive cells does not alter ATM activation or proliferation
In order to rule out any indirect long-term effect of CD4-positive lymphocytes on ATM proliferation and activation, we established a 14-day depletion protocol. We then assessed AT integrity by means of immunohistochemistry as described in the previous section. Similar to the 3-day depletion experiment, AT morphology Whole AT gene expression analysis of Ki67 and various ATM proliferation-and ATM activation-associated cytokines was performed in IgG2b (n = 3-4) and anti-CD4 (n = 5-6) treated mice using quantitative real-time PCR (cumulative data from 3 experiments). Gene of interest mRNA levels were measured in duplicate and normalized to TBP. (L and M) Representative flow cytometry plots for M1/M2 and BrdU staining after treatment with IgG2b or CD4 antibody respectively (data analysis is shown in I and J). Data are presented as means ± SEM and as column bar graphs and were tested for statistical significance by Mann-Whitney U-test. All scale bars represent 25 μm.
showed no marked difference after 14 days of CD4 depletion compared to the isotype control ( Fig. 4A and B). Likewise, adipocyte diameter and CLS density did not differ significantly between CD4-depleted and control mice ( Fig. 4E and F). Quantification of proliferation by means of PCNA staining (Fig. 4C, D, and G) and by means of flow cytometry (Fig. 4J) also showed no alteration between the two groups, as did the proportion of ATMs and M1 to M2 ratio ( Fig. 4H and I). Finally, the study of mRNA associated with ATM proliferation and activation revealed no significant changes in T-cell-depleted murine AT (Fig. 4K-N).

Glucose tolerance is improved under CD4-specific antibody treatment
In the next step, we aimed to determine any effect of CD4specific antibody treatment on glucose metabolism. We therefore performed GTT and ITT as described above. Interestingly, during GTT we observed significantly lower glucose levels at 30 and 120 min after glucose injection among the T-cell-depleted mice compared to mice treated with an appropriate isotype control (Fig. 5A). To evaluate the metabolic relevance, we performed area-under-the-curve analysis which also showed a significant difference between the two groups ( Fig. 5B), suggesting a beneficial effect of CD4 cell depletion on glucose tolerance. During insulin tolerance testing we observed a mild, yet not significant, trend towards improvement of insulin tolerance under CD4 depletion ( Fig. 5C and D).

Pancreatic endocrine function is altered following CD4 depletion
Finally, we sought to elaborate on a possible contribution of the pancreas to the observed difference in glucose metabolism. Morphometric analysis of pancreatic tissue as measured by islet quantity and size ( Fig. 6B and C, representative images in Fig. 6A) revealed no significant alterations in mice injected with CD4-depleting antibody. However, frequency distribution analysis revealed a significantly higher proportion of very small islets (40-60 μm) in pancreata of CD4-depleted mice (Fig. 6D). We then aimed at assessing the pancreatic endocrine function under CD4 depletion. Interestingly, fasting plasma insulin appeared to be similar between the two groups (Fig. 6E). Likewise, pancreatic insulin content was unchanged following CD4 depletion (Fig. 6G). In contrast, we observed a significant decrease in whole pancreatic content of glucagon (Fig. 6H). Further, protein analysis normalized to the islet protein content, as measured by GLUT2, also revealed a relative increase of somatostatin (Fig. 6I).

Discussion
The role of T cells in AT inflammation has been extensively studied over the past two decades. Two major in vivo depletion studies have suggested a substantial effect of adipose tissue lymphocytes (ATLs) on AT integrity and glucose homeostasis: CD3 depletion was shown to reduce M1 and increase M2 ATMs in VAT and positively influence glucose homeostasis [5]. Likewise, depletion of CD8-positive T cells resulted in an increased M2/M1 ratio in VAT, a reduction in CLS density, and beneficial effects on insulin sensitivity and glucose tolerance [22]. These two studies both suggest that ATLs acquire a pro-inflammatory phenotype during highfat feeding, depletion of which partially restores the disrupted AT integrity and attenuates the impaired glucose metabolism in HFDfed mice. In this study, we aimed to substantiate this claim by There are conflicting propositions in the literature, first, as to the sequence of events regarding T cell involvement and, second, as to the extent of quantitative and qualitative changes in the VAT T cell population. Studies using mRNA as a target have suggested a relatively early expansion of CD3-positive cells in HFD-induced inflammation that coincides with changes in glucose homeostasis and precedes macrophage expansion [36,37]. It has been argued that this early expansion of CD3-positive cells in VAT is limited to the CD8-positive subset whereas CD4-positive T cells decrease [5,22]. However, other studies make the case for an early CD4 cell expansion [20] or a relatively stable CD4 population [31]. In our study, we demonstrate an age-but not diet-dependent proportional increase of CD4 cells in VAT and a numerically but not phenotypically stable CD4 population with regard to the impact of an HFD. Equally unclear is the role of IFN-γ-releasing CD4 cells with studies arguing for either an increase [5] or a reduction [38] of this population under the influence of HFD-feeding. Our results seem to concur with the latter, although we also detected an increase to lean levels at a later time point. Our data thus argue against a major impact of IFNγ released by CD4 cells in the context of late AT inflammation. However, the highly significant drop of IFN-γ-releasing CD4 cells at 12 wks of HFD also described by Zamarron et al. [38] poses the question of different AT responses to high-caloric stimuli during different time points and could instigate further research. Relative consensus exists as to the FOXP3-positive Treg population that seems to increase with age and decrease under high-caloric feeding [5,22,23,30,32] selectively in VAT but not splenic tissue or SAT [23]. A similar dynamic has been reported for ST2 expression on FOXP3-positive and FOXP3-negative T cells respectively [33]. Our data agree with this observation, leading us to argue for an obesity-induced decrease of FOXP3 and ST2 expression that is inverse to the age-related increase in chow-fed animals. ST2 and FOXP3 are generally considered to have anti-inflammatory properties in the context of obesity-induced inflammation [23,25,[33][34][35].
Our results hence led us to believe that depletion of the CD4 population that has tilted towards a more pro-inflammatory phenotype would propel similar changes as seen with CD8 or CD3 depletion respectively. We therefore established a highly efficient 3-day and 14-day depletion protocol to account for short-and long-term effects on AT integrity. It should be noted that the mice cohorts used in both experiments differed quite significantly regarding their total body weight. However, the two experiments were not designed to be compared directly and both groups express a degree of AT inflammation that is distinct from AT morphology in lean animals. Interestingly, both depletion experiments showed no effect of depletion on adipocyte size, number of CLS, macrophage activation, or proliferation. We were able to reproduce these results using a short-term CD3 depletion protocol, Figure 6. Impact of 14-day CD4 depletion on pancreatic tissue integrity and secretory function. (A) Hematoxylin and eosin staining of pancreatic tissue (representative images of islets, data analysis shown in B and C). (B and C) Morphometric analysis of pancreatic tissue as measured by number of islets per mm 2 pancreatic tissue (B; n = 6 per group; cumulative data from three independent experiments) and islet diameter (C; n = 6 per group; cumulative data from 3 independent experiments). (D) Frequency distribution analysis of islet size (shown as histogram with 20 μm bin width (n = 6 per group; cumulative data from three independent experiments). (E) Fasting plasma insulin in mice treated with either CD4depleting or control antibody was measured using ELISA (n = 8 per group; cumulative data from four independent experiments). (F) Representative Western blots of insulin, somatostatin, glucagon, GLUT2 and GAPDH following depletion of CD4-positive cells or injection of an appropriate isotype control. Uncropped Western blots can be found in Supporting Information Fig. 4. (G-I) The abundance of insulin (G; both groups n = 8), glucagon (H; both groups n = 8), and somatostatin (I; both groups n = 8) is normalized to both GAPDH to account for whole pancreatic cell mass and GLUT2 to account for islet cell mass (panel G, H, and I represent cumulative data from four independent experiments with four mice per experiment). Data are presented as means ± SEM and as column bar graphs and were tested for statistical significance by unpaired t-testing, *p < 0.05. **p < 0.01. Scale bar represents 50 μm. however interpretation of the data was confounded by weight loss in depleted mice.
It should be acknowledged that weight loss has been reported in numerous studies using the CD3-depleting hamster IgG 145 2C11 [5,[39][40][41][42][43]. This has been attributed to a cytokine release syndrome triggered by activation of Fc receptors. However, in this study, we used the rat IgG 17A2 that has been explicitly demonstrated not to induce the same significant morbidity associated with the 145 2C11 antibody [42,43]. The severe reaction of some of the mice in our anti-CD3 cohort was therefore unexpected. Caution in using and interpreting data obtained from CD3 depletion with both, 17A2 and 145 2C11, should therefore be advised.
However, for appropriate comparison between data after CD3 and CD4 depletion, a similar IgG-based approach has been used, which did not show any adverse effects in CD4 depleted mice.
We and others have recently described ATM proliferation as a characteristic of AT inflammation [16,17], which seems to be dependent on cytokine stimulation namely by TH2-related cytokines IL-4, IL-13, and OPN [18,19]. Furthermore, CD4 cells have been shown to be the main source of OPN in VAT [20]. However, various techniques used in this study to evaluate ATM proliferation have been consistent in showing no effect of CD4 depletion, leading us to believe that CD4 cells play a minor role in prompting ATM proliferation under high-caloric feeding.
CD4 depletion however seems to have some effect on glucose metabolism, since we showed significant improvement of GTT, but not ITT following CD4 depletion. This suggests some degree of impaired insulin production in reaction to glucose challenge but not insulin resistance per se. This discrepancy led us to hypothesize a potential pancreatic involvement. Both groups showed signs of islet hypertrophy, which have been described as characteristic of diet-induced obesity [44]. Yet, we failed to observe a significant difference in trophic conditions under CD4 depletion. Furthermore, we did neither detect an increase of insulin expression nor enhanced fasting insulin levels under CD4 depletion. Interestingly however, we observed a significant reduction of glucagon in the pancreatic tissue and a significant increase of somatostatin in proportion to islet cell mass in CD4-depleted mice, suggesting an impact of CD4 cells on alpha and delta cell homeostasis. Alpha-cell dysfunction has been described as a hallmark of type 2 diabetes [45,46]. Indeed, hyperglucagonism was shown to be present in obese subjects in the fed and fasted state in numerous studies [47][48][49][50]. Furthermore, alpha cell dysfunction seems to result in an inability to suppress glucagon in reaction to glucose challenge [51][52][53]. Novel therapeutics rely partly on the antagonism of glucagon action [54]. The inadequate alpha cell output, in turn, appears to be counter-regulated by somatostatin [50,55]. It has been suggested that this effect is mediated by an absolute decrease in delta cell output and a loss of somatostatin sensitivity of alpha cells in the state of impaired glucose tolerance. [50,56] Therefore, it might be reasonable to speculate that glucagon and somatostatin disturbance, as observed in our study, partly contribute to the beneficial metabolic phenotype exhibited by CD4depleted mice. However, studying the pancreatic involvement was not the primary focus of this study and should therefore warrant further research.
Likewise, the beneficial effect of CD4 depletion on whole-body glucose tolerance is in agreement with the notion that CD4 cells acquire a phenotype that is detrimental to insulin sensitivity during the course of obesity. However, whereas CD3 and CD8 depletion seem to influence both glucose homeostasis and ATM activation, effects of CD4 depletion seem to be independent of such changes. Potentially these effects are mediated by cytokines that do not have a major impact on either ATM phenotype or proliferation.
That said, in the absence of an established method to selectively deplete organ-specific cells in vivo without prior genetic modifications, we currently rely on systemic depletion protocols, making it difficult to attribute an observed systemic effect to sitespecific changes in the microenvironment. Such inherent limitations apply to all depletion studies and therefore it has to be highlighted that this study was not designed to provide a mechanistic insight into the origins of CD4-mediated effects on glucose metabolism and that the above-mentioned models postulating a pancreatic or VAT involvement in CD4-mediated changes in glucose tolerance are speculative in nature and do not rule out the potential involvement of other organs.
In conclusion, our results show that the CD4 cell population in VAT shifts towards a more pro-inflammatory phenotype and that systemic depletion of CD4 cells improves glucose tolerance while leaving insulin sensitivity, AT morphology, ATM activation, and ATM proliferation as signs of AT dysfunction unaffected.

Experimental animals
Mice strains were kept in our local animal facility in a temperature-controlled, 12 h-light/dark cycle environment with ad libitum access to food and water. In order to induce obesity C57BL/6 mice were fed an HFD (60% kcal fat; Ssniff Spezialdiäten, Soest, Germany) or a normal chow diet (9% kcal fat; Ssniff Spezialdiäten) up to 24 weeks, starting at 6 weeks of age. For some experiments, CSF1R-eGFP +/-(MacGreen; [57]) or LysMCre x TDTO flx/flx [58] on a C57BL/6 background were used, depending on the spectral requirements. Of note, reporter mice develop an AT inflammation similar to wild-type C57BL/6 mice [59]. All studies used males and were approved by the local ethics committee (42502-2-1554 MLU).

Depletion
For depletion of CD4-positive cells, mice were injected with 120 μl of anti-CD4 (GK 1.5; BioLegend, San Diego, CA) or an isotype control (RTK4530; BioLegend) using a 3-day protocol and a 2-wk protocol. With the 3-day depletion protocol, mice received either injection for 3 consecutive days. With the 2-weeks protocol, mice received 3 injections each week for 2 wks as described by Nishimura et al. [22]. ITT and GTT were performed at the end of the 2-wk depletion protocol only. Experiments were performed under specific pathogen free (SPF) conditions, following FELASA criteria. Importantly, mice after CD4 depletion were indistinguishable from control mice, without any signs of diarrhea, conjunctivitis, or a significant drop in body weight.
For depletion of CD3-positive cells, mice were injected with 120 μl of anti-CD3 (17A2; BioLegend, San Diego, CA) or an isotype control (RTK4530; BioLegend) using a 3-day protocol as described above. Strikingly however, some mice showed signs of diarrhea and peritoneal inflammation in association with a considerable weight loss (Fig. S2N) following CD3 cell depletion, for which reason, we have decided to show the acquired data in the supplementary section only (Fig. S1 and 2).

ITT/GTT
For ITT, baseline glucose level was measured before mice were injected with insulin (Insuman Rapid, 100 IU/ml). Blood glucose was measured again at 15, 30, and 60 min after injection. Mice received 1.5 U/kg insulin solution intraperitoneally. ITT was performed 3 days prior to GTT. For GTT, mice were fasted for 12 h prior to testing. Baseline sugar was measured prior to injection of glucose (20%; B. Braun, Melsungen, Germany). Glucose level was then measured at 15, 30, 60 and 120 min after injection. Mice received 1 g/kg glucose solution intraperitoneally.

Immunofluorescence
Mice were sacrificed and VAT was dissected. VAT was then fixed in zinc formalin overnight and embedded in paraffin. Paraffin sections were deparaffinized, microwaved, and washed in PBS with 0.3% Triton (PBST). Subsequently, unspecific binding sites were blocked using PBST and 1% BSA for 1 h at room temperature. AT was stained with primary antibodies at 4°C overnight. For AT analysis, AT was stained using the macrophage marker Mac-2 (CL8942AP; 1:1000, Cedarlane; Burlington, Canada) and the fat cell marker Perilipin A (ab3526; 1:200; Abcam, Cambridge, UK). For proliferation studies, macrophages were stained against Mac-2 and the proliferation marker PCNA (ab15497; 1:200; Abcam). DAPI (1:10,000; Thermo Fisher Scientific, Schwerte, Germany) was used for nuclear staining. Appropriate secondary antibodies were selected and incubated for 1 h at room temperature. For AT analysis, images were taken using the Olympus BX40 epifluorescence microscope. For proliferation studies, AT was studied using an inverted confocal microscope (FV1000 Olympus, Hamburg, Germany).

Analysis of AT sections
For AT analysis, the tissue was stained for Mac-2 and Perilipin A as described above. AT integrity was evaluated by measuring fat cell diameter and formation of CLS. CLS were defined as adipocytes entirely surrounded by leucocytes as described by others [60,61]. CLS per field were counted in 10 randomly chosen fields on one section per animal. For the study of proliferating ATMs, staining for PCNA was performed as described above. ATMs were visualized by staining for Mac-2. Since it has been shown that up to 90% of ATMs in the inflamed AT reside in CLS [60] and proliferation was shown to occur predominantly in CLS [17], 20-30 photographs of CLS were taken per section and analyzed for content of proliferating cells.

Analysis of pancreatic tissue sections
For analysis of islet morphology, pancreatic tissue was stained with H&E. Subsequently, the area of pancreatic tissue was determined in three H&E-stained sections per animal (minimal distance 300 μm). Islets were counted in these sections (mean 76.8 ± 7.8 islets per mouse) and displayed as number per mm 2 of whole pancreatic tissue. Similarly, the diameter of islets was measured and averaged over all islets present in all three sections. Images were taken using a Keyence microscope (Neu-Isenburg; Germany).

Whole-mount staining
Whole mounts of AT from tdTomato (TDTO) reporter mice were used to visualize presence of CD3-and CD4-positive cells in AT. For that purpose, epididymal AT was immediately fixed after sacrifice for 20 min in zinc formalin (Polysciences, Hirschberg, Germany), washed in PBS, and cut into small pieces (<1 mm 3 ). These AT pieces were then washed in PBS, blocked with staining buffer (3% BSA in PBS) for 1 h at room temperature, and stained with pre-labeled antibodies in staining buffer (anti-CD3-PE-Cy7; 1:100 145-2C11; BioLegend or anti-CD4-AF647; 1:100; RM4-5; BioLegend) overnight. We used Hoechst (1:10,000 in PBS; Life Technologies) to stain the nuclei. AT pieces were then washed three times in PBS and subsequently transferred into cavities of microscope slides and mounted using Fluorescence Mounting Medium (Dako; Hamburg; Germany). AT was studied using the Olympus FV1000 confocal microscope.

Flow cytometry
For flow cytometry analyses, VAT and spleen were dissected. VAT was digested with collagenase type II (Worthington Biochemical, Lakewood, NJ) centrifuged and the lipid phase discarded in order to gain a cell suspension of stromal vascular cells (SVCs). This suspension was then filtered through a 75 μm mesh. Splenic tissue was minced using a razor blade and pressed through a 70 μm mesh using a syringe plunger. Fc receptors of splenic cells and SVCs were blocked using anti-CD16/32 (1:100; eBioscience, Frankfurt, Germany) for 10 min on ice. Surface staining was performed using anti-CD45-FITC ( μl BrdU 3 h prior to the experiment. After Fc-blocking, staining of surface markers, and fixation as described above, cells were further treated with DNase IV (Sigma-Aldrich, Deisenhofen, Germany). For IFN-γ detection, SVCs were cultured in RPMI 1640 medium (Sigma-Aldrich) supplemented with 10% fetal calf serum and 0.6% PSA (penicillin, streptomycin, und amphotericin B) at 5% CO 2 /21% O 2 and 37˚C. Later, leucocyte activation cocktail with BD GolgiPlug TM (BD Pharmingen) was applied according to the manufacturer's protocol and the cells were incubated overnight for 12 h. The following day, Fc-blocking, staining, and fixation were performed as described above. 7-Aminoactinomycin D (7-AAD; BD Pharmingen) was used for DNA staining. Appropriate isotype controls were duly performed for all experiments. Analysis was performed using an LSR II (BD Pharmingen) equipped with FACSDiva software 8.0. Quantification of flow cytometry data was implemented using FlowJo software 10.0.5 (Tree Star, Ashland, OR). Gating strategies are described in detail in the supporting information (Fig. S3). Guideline for the use of flow cytometry studies was duly followed [62].

Gene expression analysis
Analysis of relative gene expression was performed using quantitative real-time PCR (Maxima SYBR Green quantitative PCR master mix; Thermo Fisher Scientific) on a Bio-Rad CFX96 Manager system (Bio-Rad, Munich, Germany). RNA was extracted (TRI Reagent solution; Thermo Fisher Scientific), followed by synthesis of cDNA using oligo (dT) primers and a ProtoScript first-strand cDNA synthesis kit (New England Biolabs, Frankfurt am Main, Germany). Gene-specific primers (presented in Supporting Information Table S1) were designed with the Primer 3 software. Gene of interest mRNA levels was measured in duplicate and normalized to TATA-binding protein (TBP). The acquired data were analyzed with the Ct method by Pfaffl [63].

Enzyme-linked immunosorbent assay
Plasma concentrations of insulin were measured using ELISA according to the manufacturer's guideline (Mouse Insulin ELISA, ALPCO, Salem, USA).

Western blot analysis
Western blot was performed as described by others [64]. Total protein from pancreas samples were extracted using RIPA lysis and extraction buffer (Thermo Fisher Scientific) supplemented with protease and phosphatase inhibitor cocktail (Roche, Basel, Switzerland

Statistical analysis
Data are presented as means ± SEM and as column bar graphs or as pie charts of at least three animals evaluated by the Student's ttest, or the Mann-Whitney U test as calculated by GraphPad Prism (GraphPad Software, La Jolla, CA). A p-value < 0.05 was considered statistically significant.