Slit3 secreted from M2-like macrophages increases sympathetic activity and thermogenesis in adipose tissue

Beiging of white adipose tissue (WAT) is associated with an increase of anti-inflammatory M2-like macrophages in WAT. However, mechanisms through which M2-like macrophages affect beiging are incompletely understood. Here, we show that the macrophage cytokine Slit3 is secreted by adipose tissue macrophages and promotes cold adaptation by stimulating sympathetic innervation and thermogenesis in mice. Analysing the transcriptome of M2-like macrophages in murine inguinal WAT (iWAT) after cold exposure, we identify Slit3 as a secreted cytokine. Slit3 binds to the ROBO1 receptor on sympathetic neurons to stimulate Ca2+/calmodulin-dependent protein kinase II signalling and norepinephrine release, which enhances adipocyte thermogenesis. Adoptive transfer of Slit3-overexpressing M2 macrophages to iWAT promotes beiging and thermogenesis, whereas mice that lack Slit3 in myeloid cells are cold-intolerant and gain more weight. Our findings shed new light on the integral role of M2-like macrophages for adipose tissue homeostasis and uncover the macrophage–Slit3﻿–sympathetic neuron﻿–adipocyte signalling axis as a regulator of long-term cold adaptation. Slit3 is shown to be secreted from M2-like macrophages resident in adipose tissue, where it enables cold adaptation by stimulating norepinephrine release from sympathetic neurons.

A dipose tissues are important organs for maintaining energy homeostasis. For energy-dealing purposes, adipose tissues have developed into different types. WAT stores energy in the form of triglycerides after nutrient load and readily mobilizes to release fatty acids on energy demands 1 . In contrast, brown adipose tissue (BAT) dissipates energy by non-shivering thermogenesis 2 . The thermogenic function is fulfilled by mitochondrion-associated uncoupling protein 1 (UCP1), which uncouples oxidative phosphorylation to generate heat instead of adenosine triphosphate (ATP) 3 . Classical BAT in humans is concentrated in interscapular areas in infants and disappears in adulthood 4 . For the past two decades, studies have revealed a third type of adipose tissue that is aroused in white fat depots 5,6 . Such adipose tissue is termed beige adipose tissue due to its brown-like activity and phenotype 7,8 . Because beige adipocytes contain abundant mitochondria and are active in dissipating energy by thermogenesis, they have drawn much attention as a therapeutic target for individuals with obesity.
Beige adipose tissue is induced by responding to stimuli such as cold, exercise-induced hormones [9][10][11] . On cold exposure, sympathetic nervous system (SNS) activity in adipose tissue at subcutaneous (s.c.) depots is effectively enhanced and norepinephrine (NE) is released from nerve terminals 12,13 . NE binds to the β3 adrenergic receptor (β3 AR) on adipocytes and activates downstream cyclic adenosine monophosphate/protein kinase A (cAMP/PKA) signalling. PKA phosphorylates hormone-sensitive lipase (HSL), which releases glycerol and fatty acids from lipids to fuel thermogenesis 14 . Moreover, PKA activates UCP1 transcription by phosphorylating and activating peroxisome proliferative activated receptor, gamma, coactivator 1 alpha (PGC1α), the key transcription coactivator for mitochondrial biogenesis 15 . As a result, sympathetic activation acquires adipocytes with a beiging appearance.
Immune cells in WAT play an important role in the process of beiging and adipocyte thermogenesis. There are two distinct phenotypes of macrophage subset: M1-like (classically activated) and M2-like (alternatively activated). Type 2 cytokines (for example, interleukin-4 (IL-4) and interleukin-13 (IL-13)) trigger M2 polarization of macrophages, which are related to tissue homeostasis 16 . Conversely, at the onset of obesity, proinflammatory mediators (for example, interferon-γ (INF-γ)) increase, which biases WAT resident macrophages to a more M1-like phenotype 17 . Proinflammatory M1 macrophage-derived cytokines (mainly tumour necrosis factor-α (TNF-α) and interleukin-1β (IL-1β)) suppress the induction of UCP1 gene expression in BAT and WAT and impair the ability to appropriately respond to thermogenic stimuli [18][19][20] . However, some recent studies have further shown that type 2 innate lymphoid cells (ILC2s) play a key role in the crosstalk between immune cells and the thermogenic activation of beige adipocytes 21,22 . ILC2s were also found to activate and recruit eosinophils by secreting interleukin-5 (IL-5) and IL-13, which led to the alternative activation of M2 macrophages 21,22 . In adipose tissue, sympathetic nerves are thought to be the main source of NE. One study reported that adipose tissue macrophages also released NE and promoted beiging 23 . However, the conclusion was then questioned by another study that demonstrated that deletion of tyrosine hydroxylase (TH) expression and inhibition of NE synthesis in macrophages failed to impair thermogenesis of adipose tissue 24 . Although whether macrophages release NE is controversial, the association between the expansion of anti-inflammatory macrophages (M2-like) and beiging has been shown in several studies 22,25,26 . Some chemokines and bioactive peptides can directly or indirectly promote polarization and activation of M2 macrophage and subsequent beiging 27,28 . So far, most studies have focused on revealing the way M2 macrophages are activated; however, the mechanism by which M2 macrophages promote beiging has been less well studied.
Adipose tissue is composed of heterogeneous cellular populations 29,30 , so the macrophage regulation of beiging may involve communication between macrophages and adipocytes, macrophages and sympathetic nerves, or even other cell types. Here, we found that M2 macrophages in the s.c. adipose tissue of mice were activated by cold exposure and were able to synthesize and release Slit3, a secretive protein. Slit3 promoted nerve growth and activated TH activity in sympathetic nerves and stimulated NE synthesis and release, which promoted thermogenesis in adipocytes and helped to maintain body temperature. Considering that other thermokeeping responses (for example, shivering and vasoconstriction) are transient, the expansion of resident M2 macrophages and Slit3 may represent a long-term prothermogenic mechanism that can help the body with cold acclimation. Therefore, our study reveals that macrophages regulate beiging and thermogenesis, and provides insight into the role of adipose tissue in cold adaptation.

results
Slit3 is cold sensitive and secreted by M2 macrophages. To identify the change in the proportion of M2 and M1 macrophages in adipose tissue after cold exposure, we isolated the stromal vascular fraction (SVF) of iWAT from wild-type mice housed at 22 °C or exposed to 4 °C for 3 d, followed by flow cytometry analyses. The percentage of M2 macrophages increased significantly after cold exposure (Fig. 1a) while the percentage of M1 macrophages was significantly reduced ( Supplementary Fig. 1a). Then, we sorted M2 macrophages of the two groups by fluorescence-activated cell sorting (FACS) and the RNA of each group was extracted for transcriptome sequencing. A total of 999 genes were identified to change by more than >1. 3 fold, of which 527 were elevated in the 4 °C cold exposure group M2-CM 37   is cold sensitive and secreted by M2 macrophages. a, Flow cytometry analysis for M2 macrophages (CD45 + F4/80 + CD206 + ) in iWAT from mice housed at 22 °C or exposed to 4 °C for 3 d (n = 6 for the 22 °C group and n = 5 for the 4 °C group). b, Gene ontology analysis of rNA-sequence data for differentially regulated genes in the KEGG pathway. One sample in each group was collected by FACSing M2 macrophages from pooled SVF in the inguinal adipose tissues of 30 mice housed at 22 °C or exposed to 4 °C for 3 d. c, The heat map was created on the basis of normalized read count of genes in the axon guidance pathway in b. d, Gene expression of Slit3 in M2 macrophages in iWAT from mice housed at 22 °C or exposed to 4 °C for 3 d (n = 3 per group). M2 macrophages were collected by FACS on total SVF pooled from 30 mice in each group; the experiment was repeated three times. e, Gene expression of Slit3, TNFα, iNOS, Arg1, Mrc1 and Chi3l3 in M0, M1 and M2 macrophages (n = 6 for Slit3 in M0 group, n = 5 for TNFα, iNOS, Arg1, Mrc1 and Chi3l3 in the M0 group; n = 5 for the M1 group and n = 6 for the M2 group). f, Slit3 levels in CM from M0, M1 and M2 macrophages was determined by enzyme-linked immunosorbent assay (n = 5 for the M0 group, n = 7 for the M1 and M2 groups). g, representative western blotting against Slit3, aP2 in M1 and M2 macrophages. h, Fold change of Slit3 and aP2 protein expression level in g (n = 3 per group). i, representative western blotting against Slit3 from concentrated M2 macrophage CM. j, Fold change of Slit3 protein expression level in i (n = 3 per group). Data are presented as mean ± s.e.m. P values were determined by a two-tailed Student's t-test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; exact P values are provided in the source data.
versus the 22 °C room temperature group (Supplementary Data 1). We performed enrichment analysis of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways to prioritize these candidates, which yielded the most significantly changed pathway of axon guidance including 19 genes (Fig. 1b). Three of the genes were the secretory factors, among which the expression abundance of Slit3 war the highest (Fig. 1c). Thus, we proposed that Slit3 was a macrophage cytokine.
To verify the derivation of the Slit members in adipose tissue, we first analysed their expression in adipocytes and SVF from three Ad-Slit3  . c, Slit3 levels in CM from M2 macrophages (n = 4 per group). d, Slit3 levels in CM from iWAT injected with M2 macrophages (n = 6 per group). e, Slit3 levels in the serum of mice after M2 macrophages were injected (n = 6-8 per group). f, Food intake of mice after M2 macrophages were injected (n = 6 for the Ad-GFP group and n = 8 for the Ad-Slit3 group). g, Body weight of mice before and after M2 macrophages were injected (n = 6 per group). h, Locomotor activity of mice after M2 macrophages were injected (n = 6 per group). i, OCr of iWAT isolated from mice after M2 macrophages were injected (n = 6 per group). j, H&E and IHC staining with anti-UCP1 antibody of iWAT collected from mice after M2 macrophages were injected. Scale bar, 20 μm. k, Western blot analysis of the iWAT of mice after M2 macrophages were injected. BAT was the positive control of UCP1. l, Fold change of protein expression level in k (n = 3 per group). m, qPCr analysis of thermogenesis, lipolysis and β-oxidation genes in the iWAT of mice after M2 macrophages were injected (n = 5 for MCAD in the M2 Ad-GFP group and CPT1β, MCAD in the M2 Ad-Slit3 group; n = 6 for UCP1 and Dio2 in the M2 Ad-GFP group; n = 7 for the rest of the group). n-q, Indirect calorimetry performed in a CLAMS system after bilateral s.c. M2 macrophages were injected for 7 d (n = 6 per group). n, O 2 consumption profile of mice during a 12-h light-dark cycle. o, Histogram representative of the results shown in n. p, Heat generation profile of mice during a 12-h light-dark cycle. q, Histogram representative of the results shown in p. Data are presented as mean ± s.e.m. P values were determined by a two-tailed Student's t-test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; exact P values are provided in the source data. adipose depots. Slit3 is highly expressed in the SVF of adipose tissues, especially in iWAT ( Supplementary Fig. 1b). Moreover, the Slit3 messenger RNA levels were inducible in the SVF of iWAT after cold exposure ( Supplementary Fig. 1c). As some literature suggests, cells cultured at 31 °C in vitro can mimic the conditions of cold exposure of animals 31,32 . Therefore, we cut the iWAT into small pieces that were placed in an incubator set at 31 °C. The Slit3 levels in the conditioned medium (CM) of iWAT cultured at 31 °C were observably elevated compared with those at 37 °C ( Supplementary Fig. 1d).
To determine the expression of Slit3 in macrophages, M2 and M1 groups of macrophages in the iWAT of mice housed at 22 °C or exposed to 4 °C for 3 d were sorted by FACS. The mRNA expression of Slit3 was inducible in M2 macrophages ( Fig. 1d) but not in M1 macrophages ( Supplementary Fig. 1e). Then we isolated bone marrow-derived macrophages (BMDMs) and induced their polarization towards M1 and M2 macrophages with lipopolysaccharide or interleukin-4 (IL-4), respectively. M1 and M2 macrophages were successfully differentiated, which was verified by detecting the expression of marker genes (Fig. 1e). The Slit3 mRNA levels ( Fig. 1e) and protein levels in CM (Fig. 1f) were both elevated in M2 macrophages compared with M0 or M1 macrophages, suggesting a distinct expression pattern of Slit3 in M2 macrophages. In addition, Slit3 protein levels in cells (Fig. 1g,h) and CM (Fig. 1i,j) secreted by M2 macrophages were both induced by cold stimulation treatment at 31 °C in vitro. These data point to a physiological function of Slit3 in adipose tissue and are suggestive of a link between Slit3 and thermogenesis.
Slit3 increases the thermogenic capacity of iWAT. To assess whether Slit3 secreted by M2 macrophages enhances thermogenesis in vivo, we first constructed a Slit3-overexpressing adenovirus, which significantly increased the expression of Slit3 in both M2 cells (Fig. 2a,b) and CM (Fig. 2c). The iWAT injected with Slit3-overexpressing M2 macrophages secreted more Slit3 protein into the CM when cultured in vitro (Fig. 2d). Moreover, local M2 macrophage injection into the iWAT did not affect the levels of Slit3 in the serum (Fig. 2e). M2 macrophages overexpressing green fluorescent protein (GFP) or Slit3 were then labelled with the bioactive fluorescent dye PKH26 ( Supplementary Fig. 2a) and then injected into the iWAT site of wild-type mice by contralateral s.c. injection. (Supplementary Fig. 2b). PKH26-labelled cells were analysed by flow cytometry and accounted for roughly 30% of the total M2 macrophages in iWAT after 1 week ( Supplementary Fig. 2c), suggesting that exogenous M2 macrophages can survive in iWAT.
Food intake (Fig. 2f), body weight ( Fig. 2g) and locomotor activity ( Fig. 2h) of mice were not affected after local injection of M2 macrophages into the iWAT; nevertheless, the oxygen consumption rate (OCR) was significantly increased (Fig. 2i). The volume of adipocytes was smaller on the side injected with Slit3-overexpressed M2 macrophages, as shown in the histological sections (Fig. 2j, top). At the same time, UCP1 staining suggested distinct beiging features in iWAT injected with Slit3-overexpressing M2 macrophages (Fig.  2j, bottom). As anticipated, the protein levels of UCP1, adipose triglyceride lipase (ATGL) and phosphorylation of HSL were induced by Slit3 with changes in total HSL (Fig. 2k,l). Under the same conditions, Slit3 also increased the mRNA levels of the thermogenic genes UCP1 and PGC1α, PR domain containing 16 (PRDM16), peroxisome proliferator activated receptor gamma (PPARγ) and cytochrome c, somatic (Cycs); the lipolysis genes ATGL and HSL and the β-oxidation genes carnitine palmitoyltransferase 1b Ad-Slit3  Medium culturing M2 macrophages were collected as CM to treat adipocytes. b, representative western blot analysis for UCP1, ATGL and phospho-and total HSL in C3H10T1/2 adipocytes on 24 h of treatment with CM from M2 macrophages. c, Fold change of UCP1, ATGL and phospho-and total HSL protein expression level in b (n = 3 per group). d, Normalized gene expression in C3H10T1/2 adipocytes on 24 h of treatment with CM from M2 macrophages (n = 6 per group). e, qPCr analysis of TH and Gap43 expression in the iWAT of mice after M2 macrophages were injected (n = 7 per group). f, representative western blot analysis of Gap43, phospho-and total TH in the iWAT of mice after M2 macrophages were injected. g, Fold change of Gap43, phospho-and total TH protein expression level in f (n = 4 per group). h, IHC staining with anti-TH antibody in the iWAT of mice after M2 macrophages were injected. Scale bar, 20 μm. i, NE levels in the iWAT of mice after M2 macrophages were injected. results were normalized to total protein levels (n = 7 per group). j, NE levels in the serum of mice after M2 macrophages were injected (n = 5 per group). Data are presented as mean ± s.e.m. Student's t-test was used for comparisons. P values were determined by a two-tailed Student's t-test. *P < 0.05, **P< 0.01, ***P < 0.001, ****P < 0.0001; exact P values are provided in the source data.
(CPT1b), acyl-coenzyme A dehydrogenase, long-chain (LCAD), acyl-coenzyme A dehydrogenase, very long-chain (VLCAD) and acyl-coenzyme A dehydrogenase, medium chain (MCAD) (Fig.  2m). Meanwhile, we noted that Slit3 increased iWAT lipolysis ( Fig. 2k-m). To rule out the effect of lipolysis, the ATGL inhibitor atglistatin was locally injected into iWAT before M2 macrophages were transferred. Food intake ( Supplementary Fig. 3a), body weight ( Supplementary Fig. 3b) and rectal temperature on 4 °C cold exposure ( Supplementary Fig. 3c) of mice were not affected by atglistatin injection. As expected, inhibition of lipopolysis significantly increased adipocyte volume ( Supplementary Fig. 3d, top) in both the GFP and Slit3 macrophage-transferred groups. Inhibition of lipolysis decreased the UCP1 level in the control group, but had no effect on the beiging induced by Slit3 as indicated by UCP1 staining (Supplementary Fig. 3d, bottom) and western blot analysis ( Supplementary Fig. 3e,f). At the same time, UCP1 levels and lipolysis in BAT were not altered by local injection of Slit3-overexpressing M2 macrophages or inhibitors into iWAT ( Supplementary Fig. 3g,h).
To study the metabolic effects of increased Slit3, M2 macrophages overexpressing Slit3 or GFP were injected into the iWAT of wild-type mice bilaterally, and whole-body energy expenditure was analysed over the following 60 h using a comprehensive laboratory animal monitoring system (CLAMS). Slit3 induced whole-body oxygen consumption (Fig. 2n,o) and elevated heat generation ( Fig.  2p,q). Taken together, these results demonstrate that Slit3 secreted by M2 macrophages can promote thermogenesis in vivo.
Slit3 activates sympathetic activity in iWAT. We next asked whether Slit3 can directly promote the thermogenesis of fat cells. The CM of Slit3-overexpressing M2 macrophages was collected and added to the culture medium of adipocytes differentiated from C3H10T1/2 cells (Fig. 3a). However, UCP1 protein levels in C3H10T1/2 adipocytes remained unchanged after supplementing Slit3 in the culture medium, and so did the important lipases ATGL and HSL (Fig. 3b,c). Similarly, the mRNA levels of the thermogenic genes UCP1, PGC1α, PRDM16 and PPARγ were not altered (Fig. 3d). To reinforce the results of this experiment, C3H10T1/2 adipocytes and primary adipocytes were treated with recombinant Slit3 protein. The UCP1, ATGL and HSL protein levels in both C3H10T1/2 adipocytes ( Supplementary Fig. 4a,b) and primary adipocytes ( Supplementary Fig. 4c,d) were not changed. Meanwhile, the mRNA levels of the thermogenic genes UCP1, PGC1α, PRDM16 and PPARγ were not altered in either C3H10T1/2 adipocytes (Supplementary Fig. 4e) or primary adipocytes ( Supplementary Fig. 4f). At the same time, we used bone morphogenetic protein 4 (BMP4), which has been shown in our previous work 33 , as a positive control for the beiging process. Here, we confirmed that C3H10T1/2 adipocytes have the same ability to express UCP1 as primary adipocytes ( Supplementary   Fig. 4a-d). Collectively, these results indicated that Slit3 may indirectly promote adipocyte thermogenesis through certain other cells.
Since sympathetic activation plays a critical role in thermogenesis and lipolysis in iWAT 9,34-36 , we sought to determine whether Slit3-overexpressing M2 macrophages stimulate local sympathetic activation and growth in iWAT. To this end, we first examined the expression levels of TH, a marker of sympathetic neurons, as well as growth-associated protein 43 (Gap43), a major marker of growing axons and developing nerve terminals that can be rapidly transported from the cell body to growth cones and nerve terminals in neurons 37,38 . We found a dramatic increase in both the mRNA levels ( Fig. 3e) and protein levels (Fig. 3f,g) of Gap43. Phospho-TH, which represents the activity of TH in the iWAT of mice, was also induced after Slit3-overexpressing M2 macrophage injection. We next examined nerve density by whole-mount immunohistochemistry (IHC) with an anti-TH antibody, and the results showed a significant increase in sympathetic nerve density on Slit3 overexpression in M2 macrophages (Fig. 3h). As anticipated, we found that the levels of NE, which is synthesized and released by sympathetic neurons, were significantly higher in the iWAT (Fig. 3i), but not in the serum (Fig. 3j) of mice after Slit3-overexpressing M2 macrophage injection, indicating local activation of the sympathetic nerve by Slit3 from M2 macrophages.

Thermogenesis induced by Slit3 depends on SNS innervation.
To further confirm that Slit3 can activate SNS activity locally, sympathetic denervation was performed, which was achieved by microinjection of 6-hydroxydopamine (6-OHDA), a selective neurotoxin to sympathetic nerves, into iWAT 39,40 . The food intake (Fig. 4a), body weight (Fig. 4b) and locomotor activity (Fig. 4c) of mice remained unchanged, while the NE level in iWAT (Fig. 4d) was dramatically decreased after 6-OHDA treatment, indicating that sympathetic denervation was successful. The rectal temperature on 4 °C cold exposure (Fig. 4e) of mice was significantly reduced on 6-OHDA treatment, although Slit3 was overexpressed. Local denervation significantly blocked the increase in OCR (Fig. 4f) and glycerol levels ( Fig. 4g) induced by Slit3 in iWAT. Concurrently, adipocyte volume (Fig. 4h, top) was increased and the beiging process induced by Slit3 was damaged as indicated by UCP1 staining (Fig. 4h, middle) and western blot analysis ( Supplementary Fig. 4i,j). The results of IHC with anti-TH antibody also indicated that sympathetic nerves were largely removed (Fig. 4h, bottom). Meanwhile, the protein expression levels of Gap43, phospho-and total TH, p-PKA substrate, ATGL, phospho-and total HSL (Supplementary Fig. 4i,j) were also inhibited by 6-OHDA treatment. The results of metabolic cages in mice revealed that whole-body oxygen consumption (Fig. 4k,l) and heat generation (Fig. 4m,n) induced by Slit3 were significantly reduced after 6-OHDA treatment. In brief, these results indicated Body weight of mice before and after M2 macrophages were injected with or without 6-OHDA treatment (n = 6 per group). c, Locomotor activity of mice with or without 6-OHDA treatment (n = 6 per group). d, NE levels in the iWAT of mice with or without 6-OHDA treatment (n = 6 per group). e, rectal temperature measurements of mice with or without 6-OHDA treatment, which were bred at 22 °C and subjected to 4 °C cold challenge for 200 min (n = 6 per group). f, OCr of iWAT isolated from mice with or without 6-OHDA treatment (n = 6 per group). g, Glycerol levels in the iWAT of mice with or without 6-OHDA treatment (n = 6 per group). h, H&E and IHC staining with anti-UCP1 and anti-TH antibody for the iWAT collected from mice with or without 6-OHDA treatment. Scale bar, 20 μm. i, representative western blot analysis for the iWAT from mice with or without 6-OHDA treatment. j, Fold change of protein expression level in i (n = 3 per group). k-n, Indirect calorimetry performed in a CLAMS system after bilateral s.c. M2 macrophages were injected with or without 6-OHDA treatment (n = 6 per group). k, O 2 consumption profile of mice during a 12-h light-dark cycle. l, Histogram representative of the results shown in k. m, Heat generation profile of mice during a 12-h light-dark cycle. n, Histogram representative of the results shown in m. results in d and g were normalized to total protein levels. Data are presented as mean ± s.e.m. Data in b and e were analysed using a two-way ANOVA with time and treatment type as covariates, using a multiple comparison to test for differences in individual time points. The asterisks represent M2 Ad-GFP + vehicle versus M2 Ad-GFP + 6-OHDA, the hashes represent M2 Ad-GFP + vehicle versus M2 Ad-Slit3 + vehicle and the dagger represents M2 Ad-Slit3 + vehicle versus M2 Ad-Slit3 + 6-OHDA. Data in a,c,d and f-n were analysed using a two-tailed Student's t-test for comparisons. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; exact P values are provided in the source data. that thermogenesis induced by Slit3 secreted from M2 macrophages depended on sympathetic activity.
Myeloid deletion of Slit3 impairs thermogenesis ability. We used floxed Slit3 mice (Slit3 f/f ) crossed with the lysozyme-Cre strain (Lyz2 Cre ), which has been shown to target cells of myeloid lineage monocytes and primarily macrophages to generate conditional myeloid deletion of Slit3 mice referred to herein as Slit3 f/f /Lyz2 Cre mice. The percentage of M2 macrophages remained unchanged in the iWAT of Slit3 f/f /Lyz2 Cre mice compared to Slit3 f/f control mice at both 22 and 4 °C ( Supplementary Fig. 5a).
The mRNA levels (Fig. 5a) and protein levels ( Fig. 5b) of Slit3 in the iWAT of Slit3 f/f /Lyz2 Cre mice were both significantly reduced, although there was no significant change in the serum (Fig. 5c). The food intake (Fig. 5d) and locomotor activity (Fig. 5e) of Slit3 f/f /Lyz2 Cre mice were not altered compared to those of Slit3 f/f control mice. At   NE level (serum) Slit3 level (serum) We found that Slit3 f/f /Lyz2 Cre mice gained weight more easily than control mice (Fig. 5f), suggesting a lower metabolic rate in Slit3 knockout mice. When exposed to the 4 °C cold challenge, the ability to sustain core temperature in Slit3 f/f /Lyz2 Cre mice was significantly impaired (Fig. 5g), indicating significant impairment of thermogenesis in mice after myeloid deletion of Slit3.
The levels of NE were dramatically reduced in the iWAT of Slit3 f/f /Lyz2 Cre mice at both 22 and 4 °C (Fig. 5h). NE levels were also reduced in the serum of Slit3 f/f /Lyz2 Cre mice at 22 °C, although they remained unchanged at 4 °C (Fig. 5i), indicating compensation of NE from other sources in Slit3 f/f /Lyz2 Cre mice. Moreover, histological examination revealed larger lipid droplets in the iWAT (Fig. 5j,  top), gWAT ( Supplementary Fig. 5c) and BAT ( Supplementary Fig.  5g, top) of Slit3 f/f /Lyz2 Cre mice at both 22 and 4 °C. The beiging process in the iWAT was inhibited by Slit3 deletion, which was revealed by UCP1 staining (Fig. 5j, middle) and western blot analysis (Fig.  5k,l). We next tested sympathetic innervation in adipose tissue; IHC results further showed reduced TH levels in both the iWAT (Fig. 5j, bottom) and BAT ( Supplementary Fig. 5g, bottom) of Slit3 f/f /Lyz2 Cre mice. Consistently, both phospho-and total TH were dramatically reduced in the iWAT (Fig. 5k,l) and BAT ( Supplementary Fig.  5h,i). Consequently, Gap43 expression was significantly decreased in the iWAT of Slit3 f/f /Lyz2 Cre mice, suggesting impaired nerve growth after Slit3 deletion (Fig. 5k,l). As anticipated, phosphorylated PKA substrates were dramatically reduced in the iWAT (Fig.  5k,l), gWAT ( Supplementary Fig. 5d,e) and BAT ( Supplementary  Fig. 5h,i) of Slit3 f/f /Lyz2 Cre mice at 22 °C and became more evident when exposed to 4 °C. In addition, the levels of the critical lipases ATGL and phospho-HSL were also reduced in the iWAT (Fig.  5k,l), gWAT ( Supplementary Fig. 5d,e) and BAT ( Supplementary  Fig. 5h,i) of Slit3 f/f /Lyz2 Cre mice at both 22 and 4 °C. Meanwhile, the iWAT OCR was significantly reduced in Slit3 f/f /Lyz2 Cre mice at both 22 and 4 °C (Fig. 5m). As anticipated, glycerol production was dramatically reduced in the iWAT of Slit3 f/f /Lyz2 Cre mice when exposed to 4 °C, although there was no significant difference at 22 °C (Fig.  5n). Furthermore, whole-body oxygen consumption (Fig. 5o,p) and heat generation (Fig. 5q,r) were both impaired in Slit3 f/f /Lyz2 Cre mice. In aggregate, these results indicate that Slit3 maintains the thermogenic capacity of adipose tissue by intensifying sympathetic nerve function.
Slit3 stimulates the phosphorylation of TH by PKA/CaMKII. Next, we sought to investigate the mechanism governing the sympathetic activation induced by Slit3. The CM of Slit3-treated PC12 sympathetic nerve cells was collected and added to the culture medium of C3H10T1/2 adipocytes (Fig. 6a). The expression of Gap43, a nerve growth marker gene, was increased after Slit3 treatment in PC12 cells; however, the mRNA levels of Ki67, a marker gene of cell proliferation, were not altered (Fig. 6b). TH phosphorylation at Ser40 was significantly induced by Slit3 in PC12 cells, although total TH remained unchanged (Fig. 6c,d). Given that PKA phosphorylates Ca 2+ /calmodulin-dependent protein kinase II (CaMKII) at Thr286, which can then phosphorylate TH at Ser40 both in vivo and in vitro 41,42 , we investigated the regulation of the PKA/CaMKII signalling pathway. The cAMP level was significantly increased in PC12 cells after Slit3 treatment (Fig. 6e). Both the phosphorylated PKA substrates and phosphorylated CaMKII were dramatically induced by Slit3 with no change in the level of extracellular signal-regulated kinase (ERK) (Fig. 6f,g), another major protein kinase phosphorylating TH at Ser31 (ref. 26 ). The phosphorylated PKA substrates, UCP1 and lipases in C3H10T1/2 adipocytes were both significantly induced by Slit3-treated PC12 CM (Fig. 6h,i).
To gain insight into whether PKA/CaMKII signalling is critical in Slit3-induced TH activation and NE production in sympathetic neurons, PC12 cells were treated with Slit3 followed by treatment with H-89, a PKA-specific inhibitor, or (and) KN-93, a CaMKII-specific inhibitor. We found that inhibiting PKA and/or CaMKII significantly suppressed Slit3-stimulated Gap43 expression, TH phosphorylation (Fig. 6j,k) and NE production (Fig. 6l), suggesting that PKA/CaMKII are the downstream kinases of Slit3 and play an important role in Slit3-induced nerve growth and NE production in sympathetic neurons. Consequently, phosphorylated PKA substrates, UCP1 and lipases in C3H10T1/2 adipocytes induced by Slit3-treated PC12 CM were significantly suppressed by PKA/ CaMKII pathway inhibition in PC12 cells (Fig. 6m,n). Furthermore, to study the effect of Slit3 directly secreted by M2 macrophages on  7 per genotype). b,c, Slit3 levels in the iWAT (b) and serum (c) of mice housed at 22 °C or exposed to 4 °C for 3 d (n = 4 for 22 °C group, n = 6 for Slit3 f/f in the 4 °C group and n = 5 for the Slit3 f/f Lyz2 Cre in the 4 °C group). d-f, Food intake (d), locomotor activity (e) and body weight (f) of Slit3 f/f and Slit3 f/f Lyz2 Cre mice (n = 6 per group in d and e, n = 8 per group in f). g, rectal temperature of mice when subjected to 4 °C (n = 10 per genotype). h-n, NE levels in iWAT (h) and serum (i), H&E and IHC staining of iWAT (j), representative western blot analysis for iWAT (k), fold change of protein expression level in k (l), OCr (m) and glycerol levels (n) of mice housed at 22 °C or exposed to 4 °C for 3 d (n = 5 for the 22  . results in b, h and n were normalized to total protein levels. Data are presented as mean ± s.e.m. Data in f and g were analysed using a two-way ANOVA using age and body weight (f) or time and temperature as covariates (g) and multiple comparison to test for differences in individual time points. Data in a-d and h-o were analysed using a two-tailed Student's t-test for comparisons. NS, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; exact P values are provided in the source data. PC12 sympathetic nerve cells, the CM of Slit3-overexpressing M2 macrophages was collected and added to the culture medium of PC12 cells, of which the CM was then collected and added to the culture medium of C3H10T1/2 adipocytes (Supplementary Fig. 6a). We found that the expression of Gap43, phospho-TH in PC12 cells was significantly induced by Slit3-overexpressing M2 macrophages  Fig. 6b,c). The phosphorylated PKA substrates, UCP1 and lipases in C3H10T1/2 adipocytes were both induced by CM from PC12 cells treated with Slit3-overexpressing M2 macrophages CM ( Supplementary Fig.  6d,e). Taken together, these results indicate that Slit3 stimulates nerve growth, phosphorylation of TH and NE production by the PKA/CaMKII signalling pathway in sympathetic neurons.

ROBO1 is the receptor for Slit3 both in vitro and in vivo.
Given that roundabouts (ROBOs) are canonical receptors for Slit3 (ref. 43 ), we determined whether ROBOs are receptors in PC12 sympathetic nerve cells for Slit3. The quantitative PCR (qPCR) data indicated that there was only one Robo receptor (Robo1) expressed in PC12 cells (Fig. 7a). Next, we knocked down ROBO1 in PC12 cells by small-interfering RNA interference; phosphorylation of TH (Fig.  7b,c) and NE production (Fig. 7d) induced by Slit3 were both dramatically suppressed by siROBO1 treatment, as well as phosphorylated PKA substrates and phosphorylated CaMKII (Fig. 7b,c). These results indicated that ROBO1 is the receptor for Slit3 on sympathetic neurons in vitro. Then, we assessed the cellular localization of ROBO1 in vivo by fluorescence imaging of anti-ROBO1 and anti-TH in iWAT. Significant colocalization of ROBO1 and TH was observed in iWAT, especially when mice were exposed to 4 °C (Fig. 7e).
To assess whether ROBO1 is the crucial receptor for Slit3 on sympathetic neurons in vivo, we constructed and purified a ROBO1 knockdown adenovirus, which was then locally injected into the s.c. fat pad. The ROBO1 knockdown adenovirus significantly decreased the expression of ROBO1 at both the mRNA (Fig. 8a) and protein levels, as indicated by IHC staining with an anti-ROBO1 antibody (Fig. 8i, second line). Food intake (Fig. 8b), body weight (Fig. 8c) and locomotor activity (Fig. 8d) of mice remained unchanged after ROBO1 was knocked down. When exposed to a 4 °C cold challenge, the ability to sustain core temperature in mice injected with short hairpin Robo1 (shRobo1) was significantly impaired (Fig. 8e), indicating significant impairment of thermogenesis in mice after knockdown of ROBO1. The NE level (Fig. 8f) and OCR (Fig. 8g) of iWAT were both significantly reduced in ROBO1 knockdown Robo receptor **** **** **** is the receptor for Slit3 to stimulate Ne production. a, Gene expression of Robo1, Robo2, Robo3 and Robo4 in PC12 sympathetic nerve cells (n = 11 for the Robo1 and Robo2 group, n = 5 for the Robo3 group and n = 10 for the Robo4 group). b-d, representative western blot analysis for PC12 cells (b), fold change protein expression level in b (c) and NE levels in PC12 CM (d) on treatment of small interfering negative control (siNC) or sirOBO1, with or without Slit3 for 24 h (n = 6 per group for c; n = 4 per group for d). e, Double immunostaining with TH and rOBO1 showed that TH-positive neurons were colabelled with rOBO1 in iWAT. Scale bar, 50 μm. Data are presented as mean ± s.e.m. P values were determined by two-tailed Student's t-test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; exact P values are provided in the source data. mice at both 22 and 4 °C. Glycerol production was also reduced in the iWAT of shRobo1-injected mice when exposed to 4 °C (Fig. 8h). Meanwhile, histological examination revealed a larger adipocyte volume (Fig. 8i, top) and reduced UCP1 staining (Fig. 8i, third line) and TH expression levels (Fig. 8i, bottom) in the iWAT of ROBO1 knockdown mice. Consistently, the protein levels of phospho-and total TH, p-PKA substrate, UCP1, ATGL and phospho-and total HSL were dramatically reduced in the iWAT of ROBO1 knockdown mice, especially under 4 °C cold exposure (Fig. 8j,k). At the same time, whole-body oxygen consumption (Fig. 8l,m) and heat generation (Fig. 8n,o) were both impaired in mice after ROBO1 was knocked down in iWAT. Collectively, our results indicate that the Slit3-ROBO1 signalling pathway is necessary for NE production in sympathetic nerve cells and thermogenesis both in vitro and in vivo.
Since it is controversial whether macrophages express TH 16,17 , we examined TH expression in macrophages. The results showed that both phospho-and total TH were not expressed in either M1 or M2 macrophages (Supplementary Fig. 7a,b), indicating that sympathetic neurons are a vital source of NE in adipose tissue.

Discussion
Here we report that when cold is encountered, the percentage of M2 macrophages in iWAT increased and more Slit3 was secreted. Slit3 activated the PKA signalling pathway in sympathetic neurons through the ROBO1 receptor, and promoted sympathetic nerve growth and the synthesis and release of NE through the phosphorylation of TH by the PKA/CaMKII pathway. NE subsequently activated the PKA signalling pathway in adipocytes and promoted thermogenesis. In adipocytes, activated PKA enhanced lipolysis and glycerol release through phosphorylation of HSL and promoted the expression of UCP1 to uncouple respiration to maintain adaptive thermogenesis in a cold environment ( Supplementary Fig. 8).
Adipose tissues are highly dynamic and readily change to adapt to the environmental alterations such as nutrient loads and cold exposure [44][45][46] . Since adipose tissues are heterogeneous and composed of different types of cells 29,30 , their function as metabolic organs should be fulfilled by orchestrating different cell populations, including adipocytes, immune cells, endothelial cells, neurons, fibroblasts and other undefined cells. On cold challenge, WAT is remoulded to beiging and adipocytes gain brown-like activity with high thermogenic capacity 7,8 , which is associated with changes represented by other cells, such as a denser vasculature, higher sympathetic nerve function and increased type 2 immune cells including anti-inflammatory M2 macrophages 9,22,[47][48][49] . These cells communicate with each other and construct a complicated network. Studies have been carried out and clarified some parts of the network. For example, sympathetic neurons release NE to function on adipocytes, and beige adipocytes secrete cytokines to stimulate angiogenesis or regulate the immune response 12,13,26,50,51 . In the present study, we reported that M2 macrophages were able to intensify sympathetic nerve function in a paracrine manner by relying on the secretive protein Slit3. Therefore, Slit3 served as the mediator of communication between macrophages and sympathetic nerves.
In the KEGG analysis of the differentially expressed genes, Slit3 was included in the axon guidance pathway. It was reported that Slit3 secreted from macrophages interacts with ROBO1 on Schwann cells and fibroblasts in the nerve bridge to control axon guidance 52 . Its identity and action as a macrophage-derived cytokine were also indicated by another report that Slit3 from osteoclast stimulated osteoblast migration and proliferation 53 . The Slit family in mammals consists of three members, Slit1, Slit2 and Slit3, all of which are secreted proteins with a molecular weight of approximately 150-200 kDa and a 33-amino acid signal peptide for classical secretion. Slits are ligands for ROBOs, with Slit receptors containing a single transmembrane domain. Slit-ROBO signalling has been studied mainly in the developing nervous system 43,54 . However, in recent years, Slit functions in other tissues have gradually been discovered. For example, Slit2, secreted from beige fat cells, regulates adipose tissue thermogenesis 55 . These results indicate that Slit family members might have distinct functions from each other.
The signalling activated by Slit3 in sympathetic neurons is PKA/ CaMKII. Considering that PKA activation is also involved in adipocytes by cold-induced beiging, signalling appears to be conserved in both sympathetic neurons and adipocytes, which might be an easier way to coordinate the different cell types. Although Slit3 activates PKA in sympathetic neurons, Slit3 has no effect on adipocytes and cannot induce UCP1 or lipases (HSL and ATGL). The absence of ROBO1 on mature adipocytes might be a cause, since the mRNA of ROBO1 cannot be detected in C3H10T1/2-derived adipocytes by qPCR.
WAT is innervated by postganglionic sympathetic neurons 56,57 . By using a viral transneuronal tract tracer, innervation of WAT has been demonstrated to originate from the central nervous system 51 . After sensing cold, the sympathetic system is activated. In addition to promoting thermogenesis of adipose tissue, sympathetic activation also drives muscle shivering for heat generation and causes vasoconstriction to restrict heat transfer from the core to the environment 58 . These responses may represent acute adaptation to cold. However, shivering and vasoconstriction should not last for too long considering that they are harmful by creating metabolic burden or high blood pressure, respectively. Adipose tissue macrophages through expansion may provide a sustained way to keep body temperature, thus adapting the body for cold acclimation. In our experiment, adoptive transfer of macrophages overexpressing Slit3 increased the local NE concentration, but did not obviously increase the levels in blood (Fig. 3i,j), suggesting that sympathetic activity is restricted at the local site level. Therefore, increasing thermogenesis activation of adipose tissue through macrophages and Slit3 can be safe and efficient.
The ability to maintain stable body temperature is essential for mammals when cold is encountered, which requires the coordination of multiple cell types in adipose tissue. Here, we propose that  ). b, Food intake of mice with or without 6-OHDA treatment (n = 6 per group). c, Body weight of mice before and after shLacZ or shrobo1 injected (n = 6 per group). d, Locomotor activity of mice with shLacZ or shrobo1 injected (n = 6 per group). e, rectal temperature of mice with shLacZ or shrobo1 injected that were subjected to 4 °C challenge for 3 h (n = 6 per group). f-j, NE levels in iWAT (f), OCr of iWAT (g), glycerol levels in iWAT (h), H&E and IHC staining of iWAT (i) and representative western blot analysis of iWAT (j) from mice with shLacZ or shrobo1 injected, which were housed at 22 °C or exposed to 4 °C for 3 d (n = 6 per group for g and h). Scale bar in i, 20 μm. k, Fold change of protein expression level in j (n = 3 per group). l-o, Indirect calorimetry performed in a CLAMS system after shLacZ or shrobo1 injected (n = 6 per group). l, O 2 consumption profile of mice during a 12-h light-dark cycle. m, Histogram representative of the results shown in l. n, Heat generation profile of mice during a 12-h light-dark cycle. o, Histogram representative of the results shown in n. results in f and h were normalized to total protein levels. Data are presented as mean ± s.e.m. Data in e were analysed using a two-way ANOVA with time and temperature as covariates and using multiple comparison to test for differences in individual time points. Data in a-d and f-o were analysed using a two-tailed Student's t-test for comparisons. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; exact P values are provided in the source data.   be manipulated in the future for the treatment of individuals with obesity and related metabolic disorders.

Animals.
All animal experiments were approved by the Fudan University Shanghai Medical College (no. 20180302-010). All mice were congenic to the C57BL6/J background and purchased from Nanjing University Model Animal Research Center. Mice were housed under a 12 h light-dark cycle in plastic cages and fed an irradiated chow diet. Housing temperature and humidity were kept within arange of 21.7-22.8 °C and 40-60%, respectively. Water and cages were autoclaved. Cages were changed once weekly and the health status of the mice was monitored using a dirty bedding sentinel programme. For the experiments at 22 °C (room temperature), mice were housed in cages exposed to room air. For the experiments at 4 °C, mice were housed in cages inside temperature-controlled chambers set to the indicated temperature. Mice had free access to food and water in the mouse vivarium. Slit3 f/f and Lyz2 Cre mice were purchased from Nanjing University Model Animal Research Center and bred in our vivarium to generate Slit3 f/f Lyz2 Cre mice. Unless otherwise indicated, male mice at 7-8 weeks of age were used in these experiments. For all in vivo studies, cohorts of greater than or equal to three mice per treatment or genotype were assembled and experiments were repeated two to three independent times.
Metabolic cage study. For the indirect calorimetry study, we housed and monitored mice for 60 h in metabolic cages (Comprehensive Lab Animal Monitoring System: OXYMAX-CLAMS, Columbus Instruments) at 22 °C. Mice were housed individually and maintained on a 12 h light-dark cycle with lights on from 7:00 to 19:00. The first 36 h were used for mice to acclimatize to the system; then we analysed oxygen consumption (VO 2 ) and heat generation during the next 24 h. The oxygen consumption (ml kg −1 h −1 ) and heat generation (kcal kg −1 h −1 ) of each mouse were calculated according to its body weight.
Cold tolerance tests. Mice were housed within a temperature-controlled chamber (MMM Friocell) set to 4 °C, three mice per cage, with free access to food and water. Mouse rectal temperature was measured every 40 min using a BAT-12 microprobe thermometer with RET-3 thermocouple (Physitemp).
Cell culture. C3H10T1/2 mesenchymal stem cells were donated by M.D. Lane from Johns Hopkins University and we tested for Mycoplasma before the experiments. Cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) (Gibco) supplemented with 10% (v/v) calf serum (Sigma-Aldrich) at a low density. Then, 2 d after cells reached confluence, the day was marked as day 0. Preadipocytes were differentiated into adipocytes with DMEM, supplemented with 10% foetal bovine serum (FBS) (Gibco), 0.5 mM 3-isobutyl-1-methylxanthine, 1 μM dexamethasone, 1 μg ml −1 insulin and 1 μM rosiglitazone for 2 d. The cells were then cultured in fresh medium supplemented with 1 μg ml −1 insulin and 1 μM rosiglitazone for another 2 d. From day 4, adipocytes were cultured with DMEM containing 10% FBS, and the medium was replaced every other day until adipocytes were used for the experiments on day 6.
The Rattus norvegicus pheochromocytoma PC12 cell line was purchased from the American Type Culture Collection (CRL-1721.1) and was cultured in RPMI-1640 medium (Sigma-Aldrich) supplemented with 10% heat-inactivated horse serum (Hyclone) and 5% FBS. PC12 cells were plated on collagen IV (Sigma-Aldrich) to differentiate into sympathetic nerve cells induced by nerve growth factor (Biosensis) for 5 d; during this period, the medium containing new nerve growth factor was replaced every other day.
All cells were cultured and maintained at 37 °C in a 5% CO 2 incubator, and the medium was supplemented with 1% penicillin and 1% streptomycin (Invitrogen).
Isolation and culture of primary adipocytes. Inguinal WAT was obtained from 6-week-old male C57Bl/6J mice. iWAT was minced and digested for 40 min at 37 °C with 0.075% collagenase VIII (Sigma-Aldrich) in PBS. The cell suspension was filtered, centrifuged and resuspended in growth medium (DMEM/F12 1:1 plus Glutamax (Thermo Fisher Scientific) containing 1% penicillin/streptomycin and 10% FBS). Primary adipocytes were grown to confluence (37 °C, 10% CO 2 ) followed by induction of differentiation using 0.5 mM 3-isobutyl-1-methylxanthine, 1 μM dexamethasone, 1 μg ml −1 insulin and 1 μM rosiglitazone in growth medium. After 2 d of induction, cells were maintained in culture medium containing 1 μg ml −1 insulin and 1 μM rosiglitazone. On day 4 of differentiation, cells were cultured in growth medium containing 10% FBS and the medium was replaced every other day until adipocytes were used for the experiments on day 6.

Culture of BMDMs.
Bone marrow cells were isolated from the femur and tibia of 6-7-week-old male C57BL/6J wild-type mice and differentiated to mature macrophages for 7 d as described 59 . Briefly, cells were maintained in DMEM with 10% FBS, containing 10 ng ml −1 M-CSF (PeproTech). On day 7, 100 ng ml −1 lipopolysaccharide (PeproTech) or 10 ng ml −1 IL-4 (PeproTech) were added for M1 or M2 polarization, respectively. The polarization was considered complete after 24 h of maintenance.
Cold exposure assay in vitro. To mimic the conditions of cold exposure of animals, the cell or small pieces of iWAT were cultured at 31 °C in an incubator in vitro. The M1 and M2 macrophages that had been cultured and differentiated in a 37 °C incubator were transferred into the 31 °C incubator and cultured for 12 h, then the cells and CM were collected and analysed.
M2 macrophage transfer assay. BMDMs were isolated and cultured according to a standard procedure and polarized into M2 macrophages by IL-4 (10 ng ml −1 ) for 24 h followed with GFP-or Slit3-adenovirus treatment for another 24 h, and then enzymatically dissociated and collected. The same number (3 × 10 6 ) of GFP-adenovirus or Slit3-overexpressed adenovirus-treated M2 macrophages were injected into the s.c. site adjacent to the inguinal fat every 2 d for a total of three times. The mice were then moved on to ametabolic cage study at day 7 after the third M2 macrophage transfer. At day 9, the mice were fasted for 5 h and killed, then tissues and sera were collected for further analyses.
CM transfer assay. In the CM transfer assay, the original medium for culturing different cells was discarded and the cells were gently washed with fresh blank medium. Then fresh medium without any additional agents was added and cultured for another 12 h, then the medium was collected as CM.
Lipolysis inhibition with atglistatin treatment. The specific ATGL inhibitor atglistatin was dissolved in dimethylsulfoxide and then diluted with PBS, which was injected bilaterally into s.c. fat pads locally at a dose of 300 µg per iWAT lobe. Mice were allowed to recover for 24 h and were then injected with M2 macrophages.
Chemical denervation of iWAT sympathetic nerve with 6-OHDA. iWAT sympathetic denervation was achieved by microinjection of 6-OHDA (Sigma-Aldrich), a selective neurotoxin to sympathetic nerves, into s.c. fat pads. Then 6-OHDA (10 mg ml −1 ) was dissolved in vehicle (0.15 mol l −1 NaCl and 1% ascorbic acid solution) and was injected bilaterally into iWAT depots at a dose of 350 µg per iWAT lobe. The syringe was held in place for 1 min to prevent efflux. Mice were allowed to recover for 7 d and were then injected with M2 macrophages.

Mature adipocytes and SVF isolation.
Freshly isolated adipose tissue was finely minced with scissors and incubated in digest buffer containing 0.075% collagenase VIII (Sigma-Aldrich) for 30-40 min at 37 °C with shaking and occasional vertexing. The digested samples were filtered through a 100-μm strainer, then centrifuged at 500g for 5 min. The adipocytes in the upper layer were removed to a new tube and washed with PBS, then centrifuged at 500g for 5 min for collection. The SVF in the bottom layer was resuspended with ammonium chloride lysis buffer (1.5 M NH 4 Cl, 100 nM KHCO 3 , 10 nM Na 2 EDTA), then centrifuged at 500g for 5 min for collection.
RNA extraction and real-time qPCR. RNA was extracted from cultured cells or frozen tissue samples using TRIzol (Invitrogen). Normalized RNA was reversed-transcribed using the RevertAid First Strand Complementary DNA Synthesis Kit (Thermo Fisher Scientific) and complementary DNA was analysed by real-time qPCR through the 7500 Fast Real-Time PCR System (Applied Biosystems). Relative mRNA levels were calculated using the comparative cycle threshold method and normalized to 18S ribosomal RNA mRNA. The average of the control group was set at 1 and all the results were represented as relative mRNA expression were. All primers used are listed with their sequences in Supplementary  Table 1.
Western blot. For western blotting, homogenized tissues, whole-cell lysates or concentrated serum-free CM were lysed in 2% SDS buffer containing protease inhibitor cocktail (Roche) and phosphatase inhibitor cocktail (Roche). Next, 20-50 μg of protein were separated by SDS-PAGE and then transferred onto a polyvinylidene fluoride (0.22 µm) transfer membrane using the wet transfer method. The membranes were blocked with 5% fat-free milk in Tris-buffered saline plus 0.1% Tween 20 (TBST) for 1 h at room temperature. The membranes were incubated in primary antibodies overnight at 4 °C. The next day, the membranes were washed in TBST (3 × 10 min) and then incubated with horseradish peroxidase-conjugated secondary antibody for 1 h at room temperature. After the