Macrophage Proliferation Sustains Adipose Tissue Inflammation in Formerly Obese Mice

Obesity induces a state of chronic, low-grade inflammation characterized by qualitative and quantitative changes in the leukocytes of metabolic tissues including the hypothalamus, liver, and adipose tissue (1–3). In particular, inflammation within visceral adipose tissue (AT) of obese mice and humans has been shown to be associated with diabetes and to contribute to the development of insulin resistance; multiple cellular sources contribute to the inflammatory environment, including adipocytes, stromal cells, and leukocytes (4,5). In mice, a proinflammatory CD11c⁺ macrophage population accumulates in visceral AT during obesity and assumes a metabolically activated phenotype that is associated with the development of systemic insulin resistance (6–8). Recruitment of circulating bone marrow–derived monocytes as well as proliferation contribute to CD11c⁺ AT macrophage (ATM) accumulation and maintenance (9–11). AT T-cell–dependent signals have also been shown to promote CD11c⁺ ATM accumulation and inflammation with obesity (12,13).

While the understanding of how AT inflammation is generated with obesity is fairly detailed, relatively little is known regarding how AT leukocytes respond to weight loss (WL) following obesity. Previous studies have shown varying degrees of change in AT inflammation after WL through different regimens. Interventions such as caloric restriction or bariatric surgery can improve metabolic dysfunction and reduce markers of inflammation within AT (14–16). However, several studies suggest that WL may not fully resolve AT inflammation and insulin sensitivity. Inflammation-related gene expression in subcutaneous AT of formerly obese human subjects remains high compared with expression in lean individuals (17,18). Similar results have been observed in mice: WL has been shown to be associated with the persistent expression of inflammatory cytokines such as interleukin (IL)-6, IL-1β, and tumor necrosis factor (TNF)-α in AT (19–22). Schmitz et al. (21) recently demonstrated that WL improved glucose tolerance but incompletely resolved AT inflammation and insulin sensitivity in mice and humans. Few studies have specifically examined how WL influences the quantity and quality of AT leukocytes. Weight cycling in mice suggests that interferon (IFN)-γ–expressing T cells accrue during successive exposures to a high-fat diet (HFD), indicating an incomplete resolution of lymphocyte activation with WL (23). WL has also been shown to induce short-term ATM recruitment in response to lipolytic signals (24).

Understanding the cellular and molecular events that reshape AT leukocytes during the resolution of obesity may identify pathways that are important in promoting WL and improving insulin resistance. Therefore, our objective was to perform a detailed investigation of the effects of WL on inflammatory leukocyte activation and composition within AT to understand the mechanisms by which metainflammation may resolve. We found that WL improved glucose tolerance; however, abnormal insulin resistance persisted systemically and in visceral AT. This was associated with long-term alterations in the composition of AT leukocytes, including retention of proliferating CD11c⁺ ATMs, expansion of AT lymphocytes, and persistent activation of ATMs despite WL. Understanding which inflammatory characteristics of obesity remain despite WL helps inform our understanding of the relationship between AT inflammation and metabolic function.

Male Rag1^−/− (B6.129S7-Rag1tm1Mom/J) and C57BL/6J mice were purchased from The Jackson Laboratory. Mice, at 6 weeks of age, were started on a normal (control) diet (ND; 4.09 kcal/g; 29.8% protein, 13.4% fat, 56.7% carbohydrate; PicoLab 5L0D; LabDiet) or an HFD (5.24 kcal/g; 20% protein, 60% fat, 20% carbohydrate; D12492; Research Diets Inc.).

Glucose tolerance tests (GTTs) and insulin tolerance tests (ITTs) were performed after 6 h of fasting. Mice were injected i.p. with d-glucose (0.7 g/kg) for GTTs and human recombinant insulin (1 U/kg) for ITTs. Insulin and leptin were measured using ELISA (Crystal Chem Inc.). Energy metabolism was measured using Comprehensive Lab Animal Monitoring System analysis (Columbus Instruments International) and body composition using nuclear magnetic resonance (Minispec LF90II; Bruker Optics). All mice procedures were approved by the University of Michigan Committee on Use and Care of Animals and were conducted in compliance with the Institute of Laboratory Animal Research’s Guide for the Care and Use of Laboratory Animals.

Excised AT was digested in RPMI medium with 0.5% BSA and 1 mg/mL type II collagenase for 25min at 37°C, and the stromal vascular fraction (SVF) was separated from adipocytes by centrifugation. The following antibodies were used for flow cytometry: anti-CD45 (30-F11), anti-CD3e (145–2C11), anti-CD4 (GK1.5), anti-CD8a (53–6.7), anti-Foxp3 (FJK-16s), anti-IFN-γ (XMG1.2), anti-Ki67 (SolA15), anti-TNF-α (MP6-XT22), anti-CD40 (1C10), anti-CD80 (16–10A1), anti-CD86 (GL1), and anti-CD11c (N418) (eBioscience), and anti-IL-6 (MP5–20F3) and CD64 (X54–5/7.1) (BD Pharmingen; BD Biosciences). Analysis was performed on a BD FACSCanto II system and sorting was performed on a FACSAria III (BD Biosciences).

RNA was extracted from adipose using Trizol LS (Life Technologies) and cDNA was generated using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). SYBR Green PCR Master Mix (Applied Biosystems) and the StepOnePlus System (Applied Biosystems) were used for real-time quantitative PCR. Gapdh expression was used as an internal control for data normalization. Samples were assayed in duplicate, and relative expression was determined using the 2^−ΔΔ CT method. For microarray experiments, analysis was done using a Mouse Gene 2.1 ST plate and a WT Pico kit (Affymetrix, Santa Clara, CA). Expression values were calculated using a robust multi-array average. Data were filtered to remove probe sets with a variance less than 0.1 over all samples and then fit to a linear model (25,26). The false discovery rate was set at an adjusted P value of 0.05. Oligo_1.24.2 and limma_3.16.7 packages from Bioconductor were used for data analysis in the R statistical environment (R version 3.0.0).

AT was homogenized in radioimmunoprecipitation assay lysis buffer with phosphatase inhibitors (Roche). Protein concentration was determined using the Bio-Rad Protein Assay Dye Reagent. Proteins were labeled and visualized using an Odyssey infrared imaging system (LI-COR Biosciences). Antibodies used for immunoblotting included anti–peroxisome proliferator–activated receptor-γ (81B8), anti–insulin receptor substrate-1, anti–phospho-Akt (Ser473), anti-AKT, anti-adiponectin (C45B10), anti-caveolin (D46G3) (Cell Signaling Technology), and anti–β-actin (AC-40) (Sigma-Aldrich).

Release of glycerol was stimulated in minced 100-mg pieces of AT cultured for 5 h with or without isoproterenol (1 μmol/L), according to the manufacturer’s instructions for the Glycerol Detection Kit for Explants (ZenBio). Explants were cultured in serum-free AIM V media with AlbuMax and BSA (Life Technologies). Additional samples were hydrolyzed in 6 mol/L HCl, according to manufacturer’s instructions for the Total Collagen Assay Kit (QuickZyme BioSciences). Culture supernatant glycerol and adipose hydroxyproline contents were measured using colorimetry. The Mouse Cytokine Antibody Array (R&D Systems) was used to evaluate cytokine output from AT explants after 48 h and pooled from 3 samples. Quantitation was performed using ImageJ software after subtracting the background and normalizing to reference controls.

Ki67 immunohistochemistry and hematoxylin-eosin (H-E) staining were performed by the University of Michigan’s Comprehensive Cancer Center Histology Core. A Picrosirius Red Stain Kit (Polysciences, Inc.) was used following the manufacturer’s instructions. Antibodies used for immunofluorescence included polyclonal anticaveolin (BD Pharmingen; BD Biosciences) and anti-Mac2 (Galectin-3; eBioM3/38; eBioscience).

A PKH26 Cell Linker Kit for phagocytic cell labeling (Sigma-Aldrich) was used for in vivo macrophage labeling experiments, per the manufacturer’s instructions. Briefly, 500 μL of a 1 μmol/L solution of PKH26 dye mixed with a diluent was injected i.p. into two sets of mice (those fed an ND and those fed an HFD) before WL. The first set was euthanized 1 day later to check labeling efficiency, and the second set was euthanized after WL (8 weeks); PKH26 uptake was evaluated using flow cytometry.

All values are reported as the mean ± SEM unless otherwise stated. Statistical significance of differences between ND controls and other diet groups were determined using the unpaired two-tailed Student t test or one-way ANOVA for multiple groups, with the Fisher least significant difference test for planned statistical comparisons to the ND group (unless otherwise noted) using GraphPad Prism V6.05.

We established a model of WL based on the withdrawal of the HFD. Male C57BL/6J mice were fed an HFD (60% kcal from fat) or an ND (13.5% kcal from fat) for 12 weeks. Obese mice were then either maintained on the HFD or switched to the ND to induce WL over a period of 2–24 weeks (Fig. 1A). After 8 weeks of the switched diet, body weight of mice that showed WL decreased ∼28%, from 44.8 g (SD ±2.8 g) to 31.8 g (SD ±2.3 g), and was similar to WL in age-matched mice fed the ND (29.6 g [SD ±1.8 g]) (Fig. 1B). Body composition analysis showed no significant differences in the percentages of fat and lean mass between ND-fed mice and mice with WL (data not shown). Epididymal white AT (eWAT) and inguinal white AT (iWAT) masses and adipocyte size in eWAT normalized by 4 weeks off the HFD (Fig. 1C and D).

GTTs and ITTs were performed to assess metabolism after WL. After 8 weeks of WL, fasting glucose and glucose tolerance decreased to levels similar to those in ND-fed mice (Fig. 1E). Fasting serum insulin concentrations decreased but remained elevated compared with ND-fed controls (Fig. 1F). ITTs revealed improvement in insulin sensitivity in mice with WL compared with those fed the HFD, but insulin tolerance remained abnormal compared with ND-fed mice (Fig. 1F). Leptin levels were lower in mice with WL compared with those fed the HFD, but they remained elevated compared with the ND group (Fig. 1G). Food intake and energy expenditure increased in mice with WL compared with HFD-fed mice but remained lower than ND-fed controls (Fig. 1H and I). The respiratory exchange ratios for ND-fed mice and those with WL were not significantly different, and were higher in both groups than in HFD-fed mice. No differences in physical activity were noted between groups. Overall, these studies demonstrated that formerly obese mice retained persistent abnormalities in insulin sensitivity despite normalization of body weight, fat mass, and glucose tolerance.

We next evaluated whether AT structure normalized with WL concomitant with the normalization of depot mass. Crown-like structures (CLSs), a characteristic histologic feature of AT in obese mice, were induced with 12 weeks of an HFD. Despite WL, CLSs remained a prominent feature of eWAT and were similar in quantity to those in mice maintained on the HFD for 20 weeks (Fig. 2A). Picrosirius red staining demonstrated an association between the persistence of CLSs and AT fibrosis in mice with WL (Fig. 2B). Biochemical quantification of hydroxyproline content in AT supported this finding of increased fibrosis (Fig. 2C).

We next examined several measures of adipocyte function in mice with WL. To assess lipolysis, glycerol release from explants at baseline was examined and found to be similar between ND, 20-week HFD, and 8-week WL explants (Fig. 2D). Isoproterenol-stimulated glycerol release was decreased in HFD-fed mice compared with ND-fed mice and was restored after WL. Cytokine release was assessed in eWAT explants using cytokine arrays (Fig. 2E). AT secretion of CCL3, CCL4, CCL5, and CXCL9 was increased in eWAT of mice with WL compared with those fed the HFD. IL-1RA and CCL12 were increased in both HFD-fed mice and mice with WL compared with ND-fed mice.

Immunoblots demonstrated decreases in peroxisome proliferator–activated receptor-γ, insulin receptor substract-1, and total Akt protein expression in eWAT from HFD-fed mice and mice with WL compared with the ND group (Fig. 2F and G). Caveolin and 27-kDa adiponectin were similar between groups. Adipose AKT serine 473 phosphorylation (pAKT) was decreased in HFD-fed mice and mice with WL at baseline (Fig. 2G and H), but pAKT in mice with WL was not significantly different from ND-fed mice after insulin injection. Overall, these data demonstrate that formerly obese mice experience persistent derangements in AT architecture, fibrosis, adipogenic protein expression, and cytokine production.

Flow cytometry was used to profile AT leukocyte changes with WL. The percentage of CD45⁺ cells in the SVF of eWAT was increased with the HFD. Mice with WL had fewer CD45⁺ leukocytes in the SVF than HFD-fed mice but remained increased compared with ND-fed mice, despite similar AT weights. ATMs (CD45⁺, CD64⁺ [27]) were reduced with WL but remained significantly elevated compared with ND-fed mice when expressed as either total ATMs per eWAT pad, ATMs per gram eWAT, or as a frequency of CD45⁺ leukocytes (Fig. 3B–D). While total ATM content was reduced with WL, the frequency of CD11c⁺ ATMs remained consistently elevated (Fig. 2E), indicating that formerly obese mice have long-term perturbations in ATM composition.

During obesity, CD11c⁺ ATMs show a proinflammatory gene expression profile (7,28). To determine whether CD11c⁺ ATMs from mice with WL retained proinflammatory characteristics, FACS-sorted ATMs were evaluated using gene expression microarrays. Microarrays revealed that all ATMs during obesity and after WL had similar gene expression profiles, regardless of CD11c (M1-like marker) or CD301 (M2-like marker) expression; they were thus combined for the subsequent analyses. Pathway analysis identified enrichment of genes involved in cytokine–cytokine receptor interaction (mmu04060) and the chemokine signaling pathway (mmu04062) in ATMs from mice with WL compared with ND-fed mice (29). ATMs in mice with WL maintained a proinflammatory expression profile compared with those in ND-fed mice, including increased Il-1β, Il-6, Tnfα, Cxcl1, Ccl4, Ccl5, Ccl11, and Ccl12 (Fig. 3F). Intracellular cytokine labeling of nonstimulated ATMs verified increased IL-6 (Fig. 3G) and TNF-α (Fig. 3H) protein expression despite WL. Immunofluorescence localization of active IL-1β revealed enrichment surrounding CLSs in HFD-fed mice that was sustained after WL (Fig. 3I). Overall this demonstrates a persistence of proinflammatory ATMs in mice with WL and suggests that WL is insufficient to deactivate ATMs.

To establish how long the effects of obesity might persist in AT after WL, we extended our WL model to 24 weeks off the HFD. Body weights in mice with 24 weeks of WL were 35.8 g (SD ±1.3 g) compared with HFD-fed mice at 60.8 g (SD ±4.3 g) and ND-fed mice at 32.8 g (SD ±1.2 g) (Fig. 4A), whereas eWAT weight completely normalized (Fig. 4B). Total ATMs and CD11c⁺ ATM content remained significantly elevated compared with ND-fed mice and was comparable to that in mice with 8 weeks of WL (Fig. 4C and D). Overall the data indicate prolonged maintenance of leukocyte population changes induced by obesity despite WL. Immunofluorescence revealed continued maintenance of CLSs (Fig. 4F). However, H-E staining revealed that these CLSs were surrounded by less dense collagen deposition than CLSs found in mice with 8 weeks of WL. Mice with 24 weeks of WL also no longer retained the abnormal systemic insulin responsiveness (Fig. 4E) observed in those with 8-week WL.

We next evaluated whether the degree of ATM persistence is dependent on the duration of HFD exposure before switching the diet. Mice were placed on the HFD for 6, 9, or 12 weeks before switching to the ND for 8 weeks. All HFD groups gained weight (6 weeks: 37.45 g [SD ±3.5 g]; 9 weeks: 38.6 g [SD ±3.6 g]; and ND: 23.67 g [SD ±0.9 g]), but groups fed the HFD for 6 and 9 weeks gained less than the mice on the 12-week HFD (46.6 ± 2.0 g) (Fig. 4G). After WL, eWAT weights among all HFD durations were similar to that in ND-fed controls (Fig. 4H). Total ATM content after WL was higher than that in age-matched ND-fed mice and increased with longer time spent on the HFD (Fig. 4I). CD11c⁺ ATMs after WL in mice fed the HFD for 6 and 9 weeks remained slightly increased compared with ND-fed controls (Fig. 4J). However, mice fed the HFD for 12 weeks before WL had significantly more CD11c⁺ ATMs after WL than the groups exposed to the HFD for shorter times. Immunofluorescence after WL revealed fewer CLSs in the mice fed the HFD for 6 and 9 weeks (Fig. 4K) compared with those on the 12-week HFD (Fig. 2A). These data indicate that the degree to which ATMs and CD11c⁺ ATMs persist in AT after WL is dependent on the duration of the HFD or the degree of adiposity before the WL intervention.

Subcutaneous iWAT was also evaluated to determine how this AT depot responds to WL. iWAT weights remained significantly elevated after WL compared with ND-fed mice (Fig. 5A). Mice with WL had fewer CD45⁺ leukocytes in iWAT than HFD-fed mice, and the values were similar to those in the ND group (Fig. 5B). Total ATMs trended toward being increased in HFD-fed mice and mice with WL but were not significantly different from those in the ND group (Fig. 5C). The frequency of all ATMs and CD11c⁺ ATMs were not increased in the HFD-fed or WL mice compared with the ND group (Fig. 5D and E). Histology revealed few or no CLSs and little or no fibrosis in iWAT of HFD-fed or WL mice (Fig. 5F). Overall the results indicate that immune infiltration to murine iWAT during obesity is blunted compared with eWAT and remains low with WL. Liver histology revealed that steatosis after 12 weeks of the HFD resolved in mice with WL (Fig. 5G).

Cross-talk between macrophages and T cells is critical for the activation of both cell types in obesity (30,31). We evaluated costimulatory marker expression in eWAT ATMs from mice with WL. Gene expression from sorted ATMs revealed reduced Cd40 expression with the HFD compared with ND-fed controls, which increased with WL (Fig. 6A). WL induced Cd80 expression in ATMs (Fig. 6B), whereas Cd86 was reduced in ATMs of both HFD-fed mice and mice with WL compared with ND-fed controls (Fig. 6C). Surface expression of these markers was assessed using flow cytometry. CD40⁺ and CD80⁺ ATMs were significantly higher in mice with WL compared with ND-fed mice, whereas CD86⁺ ATMs were elevated in both ND-fed and WL mice (Fig. 6D).

We next investigated how WL influenced AT T cells. Similar to ATMs, conventional CD4⁺ T-cell and CD8⁺ T-cell numbers increased with obesity (Fig. 6E). CD4⁺ AT T cells decreased after 2 weeks and transiently increased 4 weeks after the diet was switched. CD8⁺ AT T cells were decreased in mice with WL but remained significantly elevated compared with the ND-fed group up to 8 weeks after the diet was switched (Fig. 6F). Total Foxp3⁺ regulatory AT T-cell numbers were also increased with obesity and returned to levels similar to those in ND-fed mice by 8 weeks of WL (Fig. 6G).

Type-1 polarization of T cells and IFN-γ has been implicated in obesity and insulin resistance (32–35). Evaluation of IFN-γ expression by flow cytometry demonstrated a reduction in the percentage of IFN-γ⁺CD4⁺ and IFN-γ⁺CD8⁺ AT T cells in HFD-fed compared with ND-fed mice (Fig. 6H and I). Decreased per-cell IFN-γ expression was also observed in obese mice. The frequency of IFN-γ⁺ CD4⁺ AT T cells returned to the levels of those in the ND group after 2 weeks of WL. The frequency of IFN-γ⁺ CD8⁺ AT T cells returned to the levels of those in the ND group after 4 weeks of WL and was increased after 8 weeks. These data suggest that continued activation of CD4⁺ and CD8⁺ AT T cells occurs in AT in the setting of WL.

Given the increase in AT T cells with WL, we evaluated whether T cells were required for the persistence of ATMs in formerly obese mice. T- and B-cell–deficient Rag1^−/− mice and wild-type (WT) controls were fed the HFD for 12 weeks and then switched to the ND (Fig. 7A). Rag1^−/− mice weighed less than WT mice before the diet was switched. Surprisingly, Rag1^−/− mice with WL had significantly worse glucose tolerance compared with Rag1^−/− ND-fed controls, despite similar body weights (Fig. 7B), which contrasts with the normalized GTTs of WT mice with WL (Fig. 1E). Similar to WT mice, Rag1^−/− mice had significantly elevated total and CD11c⁺ ATMs despite 8 weeks of WL (Fig. 7C and D). Histology demonstrated persistence of CLSs in WT and Rag1^−/− mice with WL (Fig. 7H and I).

To evaluate the contribution of T and B cells toward ATM proinflammatory activity during WL, we assessed gene expression from flow-sorted ATMs in WT mice and Rag1^−/− mice with WL. Il6, Tnfa, and Il1b were all significantly reduced in Rag1^−/− mice with WL compared with WT mice with WL (Fig. 7E–G). Rag1^−/− adipose insulin signaling was evaluated. Baseline phospho-AKT was decreased in Rag1^−/− mice with WL but, similar to WT mice, was not significantly different from ND-fed mice after insulin injection (Fig. 7J). Taken together, the data show that T cells are not essential for the maintenance of CD11c⁺ ATMs or the continual development of adipose fibrosis with WL in WT mice. However, signals from lymphocytes seem to contribute to the ongoing activation of ATMs in formerly obese mice.

We next examined the mechanisms responsible for ATM persistence during WL. To determine whether recruitment was a significant contributor to ATM maintenance during WL, we used PKH26 pulse-labeling in ND- and HFD-fed mice before switching the diet. This method labels tissue macrophages but not blood monocytes, allowing identification of recruited ATMs as PKH26⁻ cells (28). Mice fed the HFD for 12 weeks and the ND-fed controls were injected i.p. with PKH26. Labeling efficiency of ATMs assessed 1 day after PKH26 injection was similar in lean and obese mice (Fig. 8A and B). The frequency of PKH26⁺ ATMs was examined after 8 weeks and was lower in ND-fed mice compared with the HFD or WL groups, demonstrating a higher rate of ATM retention in HFD-fed mice and mice with WL (Fig. 8C). The data indicate that new monocyte recruitment may not be the primary mechanism for the maintenance of CD11c⁺ ATMs during WL.

Reduced apoptosis was also considered a potential mechanism of ATM maintenance. Annexin V labeling of ATMs was significantly reduced in HFD-fed mice and after 1 week of the switched diet compared with ND-fed mice, indicating these ATMs had a relatively low level of apoptosis, which may contribute to ATM maintenance in HFD-fed mice and mice with WL (Fig. 8D). This is surprising given the reduction in total ATM content observed during the early stages of WL (Fig. 3D) and suggests that ATM exfiltration may be responsible for the decrease.

Another mechanism for how ATMs may be retained without continual recruitment is proliferation. We evaluated ATM proliferation by Ki67 expression. Immunohistochemistry demonstrated an increase in Ki67⁺ nuclei surrounding CLSs in mice with 8 weeks of WL compared with mice fed the HFD for 12 weeks and was comparable to that in mice fed the HFD for 20 weeks (Fig. 8E and F), suggesting a continual increase in proliferating cells despite WL. Flow cytometry demonstrated that >90% of the Ki67⁺ cells in the SVF were CD45⁺ leukocytes (Fig. 8G). ATMs made up >50% of these Ki67⁺ SVF cells in both HFD-fed and WL mice, and both were significantly elevated compared with ND-fed controls (Fig. 8H). The overall frequency of Ki67⁺ ATMs in HFD-fed and WL mice was also significantly higher than in the ND group (Fig. 8I). In addition, the frequency of Ki67⁺CD11c⁺ ATMs in HFD-fed and WL mice was significantly elevated compared with Ki67⁺CD11c^– ATMs (Fig. 8J). Taken together, the data indicate that CD11c⁺ ATM maintenance during WL may primarily be the result of a combination of increased proliferation and reduced apoptosis.

In this study we used a mouse model of WL to understand leukocyte dynamics in AT of formerly obese mice. Our major finding is that chronic obesity leads to long-term alterations in AT leukocyte populations associated with persistent abnormalities in insulin tolerance that persist despite WL. Maintenance of CD11c⁺ ATMs was observed as long as 6 months after the HFD was stopped. ATMs in mice with WL retained a proinflammatory profile, with elevated cytokine (IL-6, IL-1β, and TNF-α) and costimulatory marker expression. AT T cells were required for ATM activation, but were not required for sustaining ATMs in abdominal depots.

This persistent pattern of myeloid activation in AT with WL contrasts with tissue such as the hypothalamus, which shows a reversal of activation with the withdrawal of an HFD (36). ATMs were sustained in formerly obese mice primarily through local proliferation and decreased apoptosis, and not ongoing recruitment of blood monocytes. The mechanism for the induction and maintenance of ATM proliferation remains unclear at this time, but potential mechanisms include MCP-1 and IL-4 (10,37). The concentration of proliferating ATMs in CLSs suggests that lipid species may also regulate proliferation. Our findings are in agreement with those of other studies suggesting that WL only partially improves AT inflammation after it is established (19–22). The persistence and transient increase in CLSs is more robust than reported in other studies. One potential difference is our use of a 60% HFD for 12 weeks, a widely used model of chronic overnutrition (6,13,28,38,39). The leukocyte changes observed after removal of the HFD were dependent on the duration of HFD feeding. We have focused on the inflammatory activation of individual leukocyte subsets since few studies have compared the activation profiles of sorted ATMs after WL. We observed reduced Tnfα and Il1β in WL ATMs compared with maintained HFD feeding, in agreement with other observations (6). However, Il6 expression was substantially increased, and we found that the overall expression of inflammatory cytokines and chemokines remained significantly elevated in mice with WL compared with ND-fed mice.

Clinical studies have found persistent subcutaneous adipose inflammation after WL despite improvements in insulin sensitivity (17,19,21,40). Our observations are consistent with a growing literature that suggests a broader role for AT leukocyte activation. Thermoneutrality, for example, potentiates AT inflammation and ATM accumulation without abnormalities in glucose regulation (41). Low-level adipose inflammation is important for maintaining metabolic homeostasis (39). Overall, the literature indicates that the role played by AT inflammation is complex, with both detrimental and potentially beneficial effects on AT function.

We observed a persistent decrease in systemic and AT insulin sensitivity in WL mice; this was associated with architectural changes, more ATM content, and fibrosis, in agreement with other recent studies (6,17,21). Several proteins involved in adipocyte differentiation and insulin signaling were substantially reduced in formerly obese mice, despite a return to normal weight. Interestingly, the 24-week WL experiment shows that adipose fibrosis can be cleared with time. It is worth speculating that the retained ATMs may have beneficial functions related to resolving the excess extracellular matrix deposition, consistent with the role macrophages play in wound healing and matrix remodeling (42,43).

Given the connection between antigen-presenting cells and T-cell activation in AT, we evaluated T-cell changes with WL (31,34,35). Weight cycling was shown to alter T-cell composition in AT (23), and we observed dynamic changes in AT T cells in our model as well. IFN-γ⁺ T cells, particularly CD8⁺ T cells, were retained during WL. However, experiments in Rag1^−/− mice revealed that T cells were not required for CD11c⁺ ATM recruitment and maintenance. After WL, Rag1^−/− mice remained glucose intolerant, which may relate either to ongoing adipose tissue inflammation caused by a lack of regulatory T cells (13), or to dysfunction in other metabolic tissues such as islets or the liver. Our findings instead indicate that local lymphocyte signals are necessary for the inflammatory activation of ATMs. This suggests lymphocyte stimulatory signals (i.e., CD40L or IFN-γ) could be targeted to attenuate inflammation while still preserving beneficial ATM functions.

Clinical studies have shown a persistent risk for cardiometabolic disease in formerly obese adults (17,21,45). Our study suggests that obesity elicits damage responses in AT that persist despite WL and that may contribute to this risk. We posit that maintenance of these features within AT, particularly inflammation, could contribute to the persistence of metabolic disease risk observed in formerly obese patients. It will be critical to compare our results using dietary manipulation with the induction of WL by other mechanisms such as bariatric surgery, which may impart different outputs in the inflammatory state of AT (45). A potential weaknesses of this study is that our WL protocol did not use isocaloric diets and did not control for food intake by pair feeding. It is possible that caloric restriction while maintaining an HFD to induce WL may provide different results. Our paradigm was designed to mimic current WL strategies that use reduced-calorie meal replacement to achieve WL. Future work will continue to investigate the physiologic consequences of these maintained features and whether the CD11c⁺ ATM pool has potentially beneficial functions that could be separated from their deleterious effects.

Acknowledgments. The authors thank Drs. Cheong-Hee Chang, Philip King, John Osterholzer, Darleen Sandoval (University of Michigan), and Alan Saltiel (University of California, San Diego) for critical evaluation of data.

This work used Core Services from the University of Michigan’s Flow Cytometry Core, University of Michigan’s DNA Sequencing Core, University of Michigan’s Animal Phenotyping Core, and University of Michigan’s Comprehensive Cancer Center Histology Core.

Funding. This work was carried out with support from the National Institutes of Health (NIH)/National Institute of Diabetes and Digestive and Kidney Diseases (2R01DK090262 to C.N.L.) and the American Diabetes Association (07-12-CD-08). Trainees were supported by NIH/National Institute of Allergy and Infectious Diseases (Experimental Training in Immunology grant 5T32AI007413), the National Institute of Diabetes and Digestive and Kidney Diseases (5F31DK103524 [to B.F.Z.] and T32DK101357 and 1F32DK105676 [to L.A.M.]), an American Heart Association Predoctoral Fellowship (to B.F.Z.), and an NIH Minority Training Supplement (2R01DK090262 to G.M.-S.).

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

Author Contributions. B.F.Z. researched data and wrote the manuscript. T.A.M. and D.L. researched data. K.W.C., G.M.-S., J.L.D., and L.M.G. researched data and reviewed and edited the manuscript. K.S. and L.A.M. contributed to the discussion and reviewed and edited the manuscript. C.N.L. wrote, reviewed, and edited the manuscript. C.N.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.