ALKBH5-mediated microglia-neuron crosstalk inhibits sympathetic innervation and adipocyte browning in male mice

Abstract

Interactions between nervous system and adipose tissue are involved in the regulation of adipocyte browning. However, whether microglial activation participates in the regulation of cold exposure-induced sympathetic innervation and adipocyte browning is unclear. Here, we demonstrate that RNA m6A demethylase ALKBH5 in the microglia nucleus of hypothalamus (PVH) region inhibits cold exposure-induced adipocyte browning in the subcutaneous white adipose tissue (sWAT) of male mice. The single nuclear-RNA sequencing, brain slice electrophysiological assay and pathological analyses show that microglia-specific ALKBH5 knockout (MCKO) enhances sympathetic activation and adipocyte browning in the sWAT during cold exposure. Mechanistically, MCKO enhances m6A methylation of Ntf5 mRNA and upregulates Ntf5 expression in the microglia. Ntf5 then activates PVH-TrkB neurons and downstream sympathetic innervation in the sWAT. During aging, MCKO prevents brown adipocyte decline in the sWAT, and promotes metabolic activity. These findings fill a gap in the research of microglia in neuron-mediated adipocyte browning, and suggest that microglia may be a potential therapeutic target for the treatment of individuals with metabolic disorders.

Interactions between nervous system and adipose tissue are involved in the regulation of adipocyte browning, but the role of microglia is incompletely understood. In this study, the authors demonstrate that RNA m6A demethylase ALKBH5 in the microglia of the nucleus of hypothalamus (PVH) region inhibit cold exposure-induced adipocyte browning in the subcutaneous white adipose tissue of male mice.

Introduction

Adipose tissues are important organs for maintaining energy homeostasis. White adipose tissue (WAT) stores energy with triglycerides and mobilizes to release fatty acids¹. Brown adipose tissue (iBAT) consumes energy by non-shivering thermogenesis, dependent on mitochondrion-associated uncoupling protein 1 (UCP1), which uncouples oxidative phosphorylation to thermogenesis². Beige adipocytes are inducible brown adipocytes residing in WAT and are functionally very similar to classical brown adipocytes, which are induced by responding to stimuli, such as cold exposure³. Therefore, enhancing adipocyte browning in WAT may be a promising strategy for preventing metabolic diseases⁴.

Interactions between the nervous system and adipose tissue are involved in the regulation of energy balance⁵. On cold exposure, sympathetic nervous system (SNS) activity in subcutaneous WAT (sWAT) is effectively enhanced, and norepinephrine (NE) is released from nerve terminals³. NE binds to the β3 adrenergic receptor (β3 AR) and activates downstream signaling to induce adipocyte browning in sWAT and adaptive thermogenesis^6–8. The central nervous system (CNS) communicates with peripheral SN activity in the sWAT to control adaptive thermogenesis⁹. The hypothalamus in the CNS is the core of energy homeostasis, including the neurons in the paraventricular nucleus of the hypothalamus (PVH), dorsomedial hypothalamus (DMH), ventromedial hypothalamus (VMH), lateral hypothalamus (LH), and arcuate nucleus of the hypothalamus (ARH) region¹⁰. Activation of BRS3-expressing neurons or TrkB neurons in the DMH region promotes thermogenesis¹¹. PDYN-expressing neurons in the VMH region are involved in cold exposure-induced adaptive thermogenesis¹². POMC neurons in the ARH region promote adipocyte browning, sympathetic innervation, and thermogenesis¹³.

It is worth noting that many immune cells in the CNS are involved in the regulation of neuronal function¹⁴. Among these, alterations in the activity of microglia, the main immunocompetent cells of the CNS, are linked with major human diseases¹⁵. Microglia are capable of monitoring and modulating neuronal activity through direct cell membrane contact or secreting neurotrophic factors to regulate neuronal activity^16,17. Recently, Microglial inflammatory signaling has been well-documented to affect CNS parasympathetic tone and plays a critical role in metabolic pathologies such as blood pressure elevation, insulin release dysfunction, and deterioration of glucose tolerance^18–20. However, whether microglial activation is involved in regulating cold exposure-induced sympathetic innervation, adipocyte browning, and adaptive thermogenesis remains unclear. Here, using cell-specific models combined with approaches to delete microglia in PVH or remove the sympathetic nerve of sWAT, we reveal an unexpected benefit of microglia-neuron crosstalk-mediated adipocyte browning via activating sympathetic nerve activity in the sWAT. Mechanically, RNA m6A demethylase ALKBH5 downregulation in the microglia mediates the expression and release of neurotrophin-4 (NT-4, or Ntf5), which binds to TrkB receptors on PVH neurons, and then activates the sympathetic nerve of sWAT after cold exposure.

Results

Microglial activation is required for cold exposure-induced adipocyte browning

The Cx3cr1^CreERT mouse has been reported in multiple studies to effectively knock out target genes in CNS microglia^{15,16,21–24}. In order to explore the potential roles of microglia in cold-induced thermogenesis, we housed Cx3cr1-GFP reporting mice, which express GFP under the control of the endogenous Cx3cr1 promoter to visualize microglia in the brain, at 6 °C for 5 days, and adipocyte browning in different adipose depots, sympathetic nervous system (SNS), and changes in microglia and neuron in central nervous system (CNS) were examined. Consistent with previous studies^25,26, cold exposure resulted in a significant increase of adipocyte browning in interscapular BAT (iBAT), sWAT and epididymal white adipose tissue (eWAT) (Supplementary Fig. 1a–d). Correspondingly, sympathetic nerve activity (TH immunostaining) had a significant increase in both iBAT and sWAT, but no increase in eWAT (Supplementary Fig. 1e, f). We detected GFP expression and found it could specifically label microglia (Iba1 immunostaining) (Supplementary Fig. 2a). More importantly, we showed that the number of CX3CR1⁺ microglia in the PVH region of the hypothalamus was significantly increased, while the other regions, including DMH, VMH, LHA, and ARC, had no significant changes after cold exposure (Fig. 1a, b and Supplementary Fig. 2b–e). We also observed a reduction of microglia endpoints (Fig. 1c), as well as shortened branch length of microglia (Fig. 1d), which were typical morphological characteristics of reactive microglia²⁷. Besides, we detected an increased expression of both pro-inflammatory and anti-inflammatory factors in the GFP⁺ microglia after cold exposure (Fig. 1e).

Fig. 1. Cold exposure induced-activation of microglia mainly occurs in the PVH region.

a Representative images of GFP expression in different hypothalamic regions of 12-week-old Cx3cr1-GFP male mice under room temperature (RT) and cold exposure for 5 days. n = 6 per group. Scale bars, 50 μm. b Quantification of (a). c, d Quantification of microglia morphology. n = 6 PVH, DMH or VMH regions per group. e Heatmap of mRNA levels of inflammatory cytokines in isolated microglia of mice under RT and cold exposure. f Schematics of neural projection circuit of PRV injection sWAT. Created in BioRender. Hua, Y. (2026) https://BioRender.com/r27d50c. g Representative images and quantification of PRV+ neurons in the different hypothalamic regions after PRV injection into the sWAT of Cx3cr1-GFP male mice under RT and cold exposure. n = 5 RT group, n = 4 cold group. Scale bars, 50 μm. h Representative images and quantification of HSV+ neurons of PVH and sWAT in 12-week-old WT (C57BL/6 J) male mice under RT and cold exposure. n = 5 RT group, n = 6 cold group. Scale bars, 50 μm. i 3D imaging and statistical maps of microglia and neurons in PVH region of WT mice under RT and cold exposure. n = 20 cells. Scale bars, 5 μm. j Representative images of patch-clamp neuron of PVH in WT mice under RT and cold exposure. Scale bars, 20 μm. k Representative traces and quantification of half-widths (ms), AP post-hyperpolarization and threshold of action potentials (AP) of PVH neurons in WT mice under RT and cold exposure. n = 8 cells. PVH: paraventricular hypothalamic nucleus, DMH: dorsomedial hypothalamic nucleus, VMH: ventromedial hypothalamic nucleus. Values represent mean ± SEM. Source data are provided as a Source Data file. Two-tailed Unpaired ttest with Welch’s correction (b–d, g, h(PVH), i, k), Two-tailed Mann Whitney test (h(sWAT)).

To further determine whether the PVH region is involved in regulating adipocyte browning, we injected PRV-CAG-RFP (Pseudorabies virus, a neurotropic retrograde tracer expressing red fluorescent protein (RFP)) into sWAT, followed by detection of RFP+ cells in the spinal cord, brainstem, mesencephalon, and hypothalamus. We observed RFP positive staining in the T5-6, RPa, Gi, MdV, NTS, LPAG, ARC, LHA, PVH, DMH, and VMH regions at room temperature (RT) (Fig. 1f, g and Supplementary Fig. 3a, b). Inconsistent with increased microglia number and reactivity in PVH (Fig. 1a–d), we detected a significant increase of PRV+ cells in PVH after cold exposure (Fig. 1g), we also found an increase of PRV+ cells in T5-6, but not in the other regions (Supplementary Fig. 3a and b). We speculated that the PVH might regulate cold-induced sWAT browning through enhanced neural output by T5–6. We also used the anterograde projective virus (HSV) and injected it into the PVH region (Fig. 1h). We found that RFP signals were detected in sWAT, and a significant increase of RFP signals in the sWAT after cold exposure (Fig. 1h). While the number of HSV+ cells in other brain regions, including the DMH and VMH showed no significant changes (Supplementary Fig. 3c). By using 3D analysis, we further demonstrated microglia and neurons in the PVH region, but not DMH and VMH, had an increased direct contact after cold exposure (Fig. 1i). More importantly, brain slice electrophysiological assay on neurons from different hypothalamic regions showed that cold exposure enhanced the activity of neurons in the PVH region (Fig. 1j, k), while having no significant effect on neurons in the DMH and VMH regions (Supplementary Fig. 3d, e). These results indicate that microglia-neuron crosstalk in the PVH region of the hypothalamus may contribute to the regulation of adipocyte browning via promoting sympathetic nerve activity in the sWAT.

In order to further verify the regulatory role of microglia under cold exposure, we crossed R26^iDTR mice with Cx3cr1^CreERT/+ mice, that has been well demonstrated to specifically knock out the target gene in microglia²⁸. To selectively ablate microglia, we first injected these mice with tamoxifen and then waited for 4 weeks to allow the myeloid cell populations originating from the bone marrow to fully turn over and revert to the wild-type genotype^16,29. Then we injected diphtheria toxin (DT) to PVH region via brain stereotaxic injection to deplete microglia (Supplementary Fig. 4a), while peripheral macrophages had no significant changes (Supplementary Fig. 4b). We found that microglia depletion significantly increased weight of sWAT compared with control mice after cold exposure (Supplementary Fig. 4c). Whereas the weight of iBAT and eWAT, as well as the weight of heart and liver, total body weight, had no significant changes after microglia depletion in the PVH region (Supplementary Fig. 4d–h). More importantly, microglia depletion inhibited cold exposure-induced sWAT browning, as determined by morphological changes of adipocytes (Supplementary Fig. 4i), and decreased sympathetic nerve activity (TH staining and expression) (Supplementary Fig. 4j, k). Besides, we also found that microglia depletion inhibited neuronal activity (Supplementary Fig. 4l).

We recognized the technical limitations of Cx3Cr1^CreERT mouse line, CX3CR1 is not only expressed in microglia but also in brain macrophages. Based on previous studies indicating that brain macrophages^30,31, but not microglia, specifically express Lyve1, we generated Lyve1^creERT2/+; Csf1r ^flox/flox mice to selectively ablate Lyve1+ macrophages. Our results demonstrated that ablation of Lyve1⁺ brain macrophages had no significant effect on sympathetic nerve activity and adipocyte browning (Supplementary Fig. 5a–c), providing strong evidence that the observed phenotype is primarily mediated by microglia.

Collectively, these suggest that microglia in the PVH region regulate sympathetic nerve activity and adipocyte browning in the sWAT.

Microglial deletion of ALKBH5 promotes sympathetic activity and subcutaneous adipocyte browning

N6-methyladenosine (m6A) is a common RNA modification that is implicated in many biological processes. The brain has been shown to have higher m6A abundance than the other organs³². We found that cold exposure increased the methylation level of microglia (Fig. 2a), but not neurons (Supplementary Fig. 6a). We then detected the genes involved in the regulation of m6A modification in the microglia after cold exposure, and found a significant downregulation of Alkbh5 mRNA compared with the other genes (Fig. 2b), consistent with decreased ALKBH5 expression (Supplementary Fig. 6c, d). Besides, we did not find a significant change of ALKBH5 expression in adipose tissues, including iBAT and sWAT (Supplementary Fig. 6b).

Fig. 2. Specific depletion of Alkbh5 from microglias (MCKO) promotes thermogenesis, sympathetic innervation, browning of sWAT under cold exposure.

a m6A dot blot assessed m6A mRNA methylation in the microglia of 12-week-old WT (C57BL/6 J) male mice under RT and cold exposure for 5 days. n = 6 per group. b Heatmap of transcription levels of m6a modification-related genes in microglia of WT mice under RT and cold exposure. Representative infrared thermal images (c) and iBAT temperature (d) of Cx3cr1^CreERT/+ (Ctrl) and Cx3cr1^CreERT/+; Alkbh5 ^flox/flox (MCKO) male mice under cold exposure. e Quantification of the weight of sWAT in Ctrl and MCKO mice under cold exposure. f H&E staining and quantification in sWAT of the indicated mice. Scale bars, 50 μm. g Immunostaining images and quantification with anti-UCP1 and anti-PLIN1 antibody in sWAT of the indicated mice. Scale bars, 50 μm. h TH, UCP1 protein expression and quantification in sWAT of the indicated mice. i sWAT-related thermoregulation gene expression levels in the indicated mice. n = 6 per group (a, d, e–i). Scale bars, 50 μm. PVH: paraventricular hypothalamic nucleus. Values represent mean ± SEM. Source data are provided as a Source Data file. Two-tailed Unpaired ttest with Welch’s correction (a, e–i), Two-way ANOVA with Šídák’s multiple comparisons test (d).

To explore the role of microglial ALKBH5 in adipocyte browning, we crossed Alkbh5 ^flox/+ mice with Cx3cr1^CreERT/+ (Ctrl) mice, to obtain Cx3cr1^CreERT/+; Alkbh5 ^flox/flox mouse model (MCKO). The 8-week-old male Ctrl and MCKO mice were injected intraperitoneally with Tamoxifen for 5 consecutive days. Following a four-week period to allow for the turnover of the circulating monocyte/macrophage population, the mice underwent the subsequent experiments. The knockout efficiency and specificity were verified by detecting the protein and mRNA level of Alkbh5 in MCKO compared with Cx3cr1^CreERT/+ (Ctrl) mice (Supplementary Fig. 6e–h). Interestingly, under cold exposure, the body temperature of MCKO mice was higher than Ctrl mice (Fig. 2c, d), and decreased the weight of sWAT (Fig. 2e). Whereas, it had no significant effect on the weight of body, heart, liver, brain, iBAT, eWAT (Supplementary Fig. 6i–n). At room temperature, MCKO had no significant effect on iBAT and sWAT browning and sympathetic nerve activity (Supplementary Fig. 7a–f). Whereas, MCKO promoted cold exposure-induced sWAT browning, as determined by morphological changes of adipocytes (Fig. 2f), increased UCP1 staining and expression and thermogenesis-related genes mRNA levels (Fig. 2g–i). In contrast, MCKO had little effect on iBAT adipocytes size and thermogenesis- related genes after cold exposure (Supplementary Fig. 7g–j). These suggest that microglial ALKBH5 specifically regulates adipocyte browning in the sWAT.

Besides, to further explore the role of PVH neurons on sWAT browning, we utilized an inhibitory chemogenetic strategy. Specifically, we injected a retrograde Cre-expressing virus (Retro-ScAAY-hSyn-CRE-WPREs) into the sWAT. And we performed targeted microinjection of a Cre-dependent inhibitory DREADD virus (rAAV-hSyn-DIO-hM4D(Gi)-EGFP-WPRE-hGH) into the PVH region. Then we subjected Ctrl and MCKO mice (Vehicle or DCZ drinking, final concentration 0.005 mg/mL) to cold exposure. We detected EGFP⁺ signals in the PVH, indicating that PVH neurons are involved in regulating sWAT browning (Supplementary Fig. 8a). Furthermore, chemogenetic inhibition of these EGFP⁺ neurons in the PVH significantly attenuated enhanced sWAT browning and sympathetic innervation in MCKO mice (Supplementary Fig. 8b, c). Collectively, these results confirm that the PVH neuron-adipose SNS neural circuit is involved in the regulation of adipocyte browning in sWAT.

To determine the microglia-neuron crosstalk during the sWAT browning process, we performed single-nucleus RNA-sequencing (snRNA-seq) in the PVH region of Ctrl and MCKO mice. Unsupervised clustering analysis of the snRNA-seq data identified multiple cell clusters, including neurons, microglia, astrocytes, endothelial cells and so on (Fig. 3a). Gene Ontology (GO) and KEGG pathway analyses indicated that these genes were predominantly involved in the regulation of neuron activity, such as synaptic membrane and synapse organization (Supplementary Fig. 9a, b). Analysis of cellular interaction revealed the stronger interactions between microglia and neurons (Fig. 3b–d). Next, we subclustered the neuron group into 15 clusters (Fig. 3e). Notably, we found marker genes of neuronal activity and neurogenesis were highly expressed, among them, Egr1, Nr4a1, Junb are related to the sympathetic nervous system (Fig. 3f). These genes are highly associated with neuron axons/synapses/survival. More importantly, we showed that MCKO significantly enhanced the direct contact between microglia and neurons by 3D analysis (Fig. 3g), as well as a substantial increase in neuronal excitability by performing brain slice electrophysiological assay in the PVH region after cold exposure (Fig. 3h–j). Consistently, we detected an enhanced sympathetic nerve activity (TH staining and expression) (Fig. 3k, l), and increased NE levels in sWAT without affecting plasma NE level in MCKO mice after cold exposure (Fig. 3m, n).

Fig. 3. Specific depletion of Alkbh5 from microglias (MCKO) promotes neuronal activity under cold exposure.

a UMAP plot displaying cells clusters from brain of Ctrl and MCKO mice. b Plot of cellular signaling interactions from brain of Ctrl and MCKO mice. c Chord diagrams displaying cells clusters interactions from brain of Ctrl and MCKO mice. d Chord diagrams displaying cell clusters and interactions between microglia and other subsets. e UMAP plot displaying cell clusters from neurons of Ctrl and MCKO mice. f UMAP plot of genes related to different neuronal functions. g 3D imaging and statistical maps of microglia and neurons in the PVH region under cold exposure. n = 20 cells. Scale bars, 5 μm. h–j Representative traces and quantification of half-widths (ms), AHP post-hyperpolarization and threshold of action potentials (AP) of neurons with Ctrl and MCKO mice. n = 8 cells. k Representative immunostaining images and quantification of TH expression in Ctrl and MCKO mice under cold exposure. n = 6 per group. Scale bars, 50 μm. l TH gene mRNA expression levels of sWAT in Ctrl and MCKO mice. n = 6 per group. Quantification of NE levels in blood (m) and sWAT (n) of Ctrl and MCKO mice under cold exposure. n = 6 per group. AST Astrocytes, RGC RadialGlialCells, OD Oligodendrocytes, ODPC OligodendrocytePrecursorCells, NR Neurons, MC MicroglialCells, FB Fibroblasts, EDC EpendymalCells, ETC EndothelialCells, CPEC ChoroidPlexusEpithelialCells. Values represent mean ± SEM. Source data are provided as a Source Data file. Two-tailed Unpaired ttest with Welch’s correction (g–n).

To determine the role of ALKBH5-mediated microglia-sympathetic nerve axis in sWAT browning, a denervation model was used to remove the sympathetic nerve of sWAT in Ctrl and MCKO mice (Fig. 4a–d). Removal of the sympathetic nerve downregulated sympathetic nerve activity and inhibited sWAT browning in Ctrl mice after cold exposure. More importantly, the increased sWAT browning of MCKO mice was not observed after removal of the sympathetic nerve compared with Ctrl mice (Fig. 4e–h). To further confirm these findings, we injected 6-OHDA into the sWAT to specifically ablate the sympathetic nerves in sWAT. The results showed that 6-OHDA treatment reduced adipocyte browning and sympathetic innervation in sWAT in Ctrl mice. Importantly, the MCKO-mediated enhancement of sWAT browning was also abolished following 6-OHDA injection (Fig. 4i–l). These suggest that microglial MCKO drives sWAT browning via directly regulating sympathetic nerve activity.

Fig. 4. Denervation of sWAT inhibits adipose tissue browning with MCKO under cold exposure.

a Scheme of operation in the experimental mouse. Created in BioRender. Hua, Y. (2026) https://BioRender.com/154xuj5. b Representative images of sWAT denervation with sham and operation group of Ctrl and MCKO mice. Scale bars, 100 μm. c Representative images and quantification (d) of sWAT with sham and operation group of Ctrl and MCKO mice under cold exposure. Scale bars, 500 μm. e H&E staining and quantification of sWAT with sham and operation group of the indicated mice. Scale bars, 50 μm. Immunostaining images and quantification with anti-TH antibody (f), anti-UCP1 and PLIN1 antibody (g) in sWAT with sham and operation group of the indicated mice. Scale bars, 50 μm. h TH and UCP1 protein expression and quantification in sWAT with sham and operation group of the indicated mice. n = 4 Ctrl group with sham, n = 5 MCKO group with sham, n = 6 Ctrl group with operation, n = 5 MCKO group with operation (d–h). i H&E staining and quantification of sWAT with VC and 6-OHDA group of Ctrl and MCKO mice under cold exposure. Scale bars, 50 μm. j Immunostaining images and quantification with anti-TH antibody (j), anti-UCP1 and PLIN1 antibody (k) in sWAT with VC and 6-OHDA group of the indicated mice. Scale bars, 50 μm. l TH and UCP1 protein expression and quantification in sWAT with VC and 6-OHDA group of the indicated mice. n = 6 per group (i–l). Values represent mean ± SEM. Source data are provided as a Source Data file. One-way ANOVA with Tukey’s multiple comparisons test (d–f, i, j), One-way ANOVA with Dunnett’s T3 multiple comparisons test (g, h, k, l).

Microglial ALKBH5 KO, but not tissue resident or circulating macrophage ALKBH5 KO, promotes adipocyte browning

CX3CR1 is indeed expressed not only in microglia but also in macrophages in both the brain and periphery, and these CX3CR1⁺ macrophages have also been reported to be involved in the regulation of adipocyte metabolism^33,34. Therefore, we designed a series of experiments to clarify whether ALKBH5 in other CX3CR1⁺ macrophages affects adipocyte browning.

To determine the role of ALKBH5 in adipose tissue resident CX3CR1⁺ macrophages, we utilized an adipose transplantation experiment combined with cold exposure. We transplanted sWAT from Cx3cr1^CreERT/+; Alkbh5^flox/flox (MCKO) mice or control Cx3cr1^CreERT/+ (Ctrl) mice into recipient WT mice. Cold exposure was then conducted on the recipient mice. In these mice, only the Alkbh5 gene in resident CX3CR1⁺ macrophages within the transplanted adipose tissue was knocked out, while ALKBH5 expression in microglia remained normal. We found that sWAT browning in recipient (WT) mice that received MCKO sWAT showed no significant difference compared to those that received control (Ctrl) sWAT under cold exposure (Supplementary Fig. 10a–e). These results indicate that knockout of ALKBH5 in CX3CR1⁺ resident macrophages within sWAT does not affect adipocyte browning.

To investigate the potential role of ALKBH5 in circulating CX3CR1⁺ macrophages, we utilized a parabiosis model combined with cold exposure. We conducted parabiosis between Cx3cr1^CreERT/+; Alkbh5^flox/flox (MCKO); Rosa26^Td mice and WT (MCKO) mice, Cx3cr1^CreERT/+ (Ctrl); Rosa26^Td mice and WT (Ctrl) mice. Adipocyte browning was assessed in each mouse under cold exposure. Since the circulatory systems were shared, circulating CX3CR1⁺ macrophages derived from the MCKO mice were recruited into the WT mice (Supplementary Fig. 11a, b). If the ALKBH5 in these circulating cells were responsible for regulating adipose browning, then there should be no significant difference in sWAT browning between the Ctrl and WT (Ctrl) mice, MCKO and WT (MCKO) mice in the parabiosis pairs. However, we observed that adipose browning in MCKO mice was significantly more pronounced than that in Ctrl mice. Additionally, CX3CR1⁺ macrophages from MCKO mice could enter the WT mice through the shared circulation. If the ALKBH5 in these cells played a role, then the WT mice paired with MCKO mice should exhibit more browning compared to the WT mice paired with Ctrl mice. However, we found no significant difference in adipocyte browning between 2 groups of WT mice (Supplementary Fig. 11c, d). These results demonstrate that ALKBH5 in circulating CX3CR1⁺ macrophages do not participate in the regulation of adipose browning or sympathetic innervation.

In addition, Tmem119 has been reported to be specifically expressed in microglia, and previous studies have utilized Tmem119^CreERT/+ mice to specifically delete genes in the microglia^15,18,35. Therefore, we employed Tmem119^CreERT/+ mice and crossed them with Alkbh5 ^flox/flox mice to generate microglia-specific ALKBH5 knockout mice Tmem119^CreERT/+; Alkbh5^flox/flox (MCKO2). We first confirmed that MCKO2 did not affect ALKBH5 expression in adipose tissue but significantly reduced ALKBH5 expression in the microglia (Supplementary Fig. 12a, b). Subsequently, we demonstrated that, similar to the results observed in Cx3cr1^CreERT/+; Alkbh5 ^flox/flox (MCKO) mice, MCKO2 mice significantly promoted sWAT browning after cold exposure (Supplementary Fig. 12c, d). These results indicated that ALKBH5 in microglia, rather than other CX3CR1⁺ macrophages in adipose tissue resident or circulating CX3CR1⁺ macrophages, is involved in the regulation of adipose browning and sympathetic innervation.

ALKBH5-deleted microglia activate TrkB neurons by promoting Ntf5 expression

To determine the molecular mechanism of microglial ALKBH5-mediated sympathetic nerve activity, we first detected the expression levels of pro- and anti-inflammatory factors in the CX3CR1⁺ microglia, and found no significant difference between Ctrl and MCKO mice after cold exposure (Fig. 5a). This suggests the possibility of a microglia-neuron cross-talk independent of the inflammatory response. Therefore, we performed methylated (m6A) RNA immunoprecipitation sequencing (m6a-RIPseq) and RNA sequencing (RNA-seq) assay in Ctrl and MCKO microglia, as well as analyzed total brain RNA-seq data after RT and cold exposure (GSE172021). The intersection analysis of the 3 assays suggests that Ntf5 is potential target of ALKBH5(Fig. 5b, c). Bioinformatic analysis showed that the GGAC motif was highly enriched within m6A locus of Ntf5 mRNA in ALKBH5 knockout microglia (Fig. 5d). ALKBH5 knockout upregulated Ntf5 mRNA (Fig. 5e) and protein level (Fig. 5f) in microglia. We also detected increased protein level of Ntf5 in microglia of MCKO mice after cold exposure (Fig. 5g). While removal of sympathetic nerve in sWAT had no significant effect on Ntf5 expression in microglia of PVH region (Supplementary Fig. 13a).

Fig. 5. Alkbh5-depletion in microglia activates TrkB neurons by promoting Ntf5 expression.

a Heatmap of transcription levels of inflammatory cytokines in microglia of Ctrl and MCKO mice under cold exposure. b Scheme of intersection analysis of the m6a-RIPseq and RNA-seq, brain RNA-seq assays of differential gene. c The heatmap of trophic factor mRNA levels in WT mice brain at RT and cold exposure (GSE172021). d The enrichment of GGAC motif within m6A locus of Ntf5 mRNA of Ctrl and MCKO mice. e Ntf5 transcription levels in microglia of Ctrl and MCKO mice. n = 6 per group. f Ntf5 protein expression and quantification in microglia of Ctrl and MCKO mice. n = 6 per group. The samples derive from the same experiment and the blots were processed in parallel. g Immunostaining images and quantification with anti-Iba1 and anti- Ntf5 antibody in PVH of Ctrl and MCKO mice under cold exposure. n = 10 cells of ctrl group, n = 9 cells of MCKO group. Scale bars, 50 μm. Immunostaining images and quantification with anti-NeuN and anti-p-TrkB antibody (h), anti-NeuN and anti-SYN1 antibody (i) in PVH of Ctrl and MCKO mice under cold exposure. n = 6 per group. Scale bars, 20 μm. j p-TrkB protein expression and quantification in primary neurons with Ctrl and MCKO microglia cultured medium treatment. n = 6 independent experiment per group. The samples derive from the same experiment and the blots were processed in parallel. k Representative traces and quantification of numbers in 300 ms and amplitude (mV), half-widths (ms) of action potentials of primary neurons with Ctrl and MCKO microglia cultured medium treatment. n = 12 cells of per group. l Immunostaining images and quantification with anti-NeuN and anti-p-TrkB antibody of primary neurons with Ctrl and MCKO microglia cultured medium treatment. n = 10 cells of ctrl group, n = 11 cells of MCKO group. Scale bars, 50 μm. m Immunostaining images and quantification with anti-PSD95 and anti-SYN1 antibody of primary neurons with Ctrl and MCKO microglia cultured medium treatment. n = 10 cells of per group. Scale bars, 50 μm. Values represent mean ± SEM. Source data are provided as a Source Data file. Two-tailed Unpaired ttest with Welch’s correction (e, g–m). Two-tailed Mann Whitney test (f).

Ntf5 mainly binds to its receptor TrkB on neurons, promotes TrkB phosphorylation, and then regulates downstream neuronal function and sympathetic nerve activity³⁶. We detected p-TrkB and SYN1 level significantly increased in NeuN⁺ neurons of PVH region in MCKO mice compared with Ctrl mice after cold exposure (Fig. 5h, i). While removal of sympathetic nerve in sWAT had no significant effect on TrkB phosphorylation in microglia of PVH region (Supplementary Fig. 13b), suggesting that cold-induced sympathetic nerve activation in the sWAT is dependent on TrkB phosphorylation in the neurons of the PVH region. Next, we detected TrkB activity in cultured neurons, and found that MCKO microglia-conditioned medium significantly promoted TrkB phosphorylation in primary neurons (Fig. 5j). We then detected the direct effects of microglia on neuronal electrophysiological activity, and found that MCKO microglia-conditioned medium stimulated promoted action potential firing of neurons compared with Ctrl microglia (Fig. 5k). Then, further cell experiment showed MCKO microglia had higher Ntf5 expression level (Supplementary Fig. 14a). And MCKO microglia-conditioned medium significantly promoted MAP2 expression (Supplementary Fig. 14b), TrkB phosphorylation (Fig. 5l), SYN1 and PSD95 expression (Fig. 5m) of primary neuron. These results demonstrate that Alkbh5-depletion in microglia activates TrkB neurons and synaptic plasticity by promoting Ntf5 expression.

We then determined whether Ntf5 is required for MCKO-mediated sympathetic nerve activation and adipocyte browning after cold exposure. We crossed Ntf5 ^flox mice with Cx3cr1^CreERT/+; Alkbh5 ^flox/flox (MCKO) mice, ultimately obtaining Cx3cr1^CreERT/+; Alkbh5 ^flox/flox; Ntf5 ^flox/flox (MDKO) mice, Cx3cr1^CreERT/+; Ntf5 ^flox/flox (MNKO) mice and Cx3cr1^CreERT/+ mice (Ctrl) allowing for the specific knockout of Ntf5 in microglia or the double knockout of both ALKBH5 and Ntf5 in microglia (Fig. 6a). We found the enhanced cold-induced SNS activity and adipocyte browning in sWAT observed in MCKO mice was not detected in MDKO mice (Fig. 6b-e). Also, we found p-TrkB levels in PVH neurons of MNKO mice were lower than in Ctrl mice under cold exposure, while MCKO mice exhibited increased p-TrkB levels. However, in MDKO mice, we did not detect any increase in p-TrkB levels (Fig. 6f, h). Also, the increased neuron synaptic plasticity in MCKO mice was also not observed in MDKO mice under cold exposure (Fig. 6g). We specifically knocked out ALKBH5 in microglia, employed RNA methylation sequencing and transcriptome profiling to identify Ntf5 as a downstream target of ALKBH5. Subsequently, we confirmed TrkB as the specific receptor for Ntf5. Through a combination of in vivo and in vitro experiments, we demonstrated that microglial ALKBH5 depletion promotes the expression of Ntf5, enhances the phosphorylation and activity of TrkB neuron, and ultimately augments sympathetic nerve activity and browning of sWAT.

Fig. 6. Ntf5-depletion in microglia attenuates browning of adipose tissue with MCKO.

a Scheme of experimental mouse strains. The image of mouse is created in BioRender. Hua, Y. (2026) https://BioRender.com/ggd2y0o. b H&E staining and quantification in sWAT of Cx3cr1^CreERT/+ (Ctrl) mice, Cx3cr1^CreERT/+; Ntf5 ^flox/flox (MNKO) mice, Cx3cr1^CreERT/+; Alkbh5 ^flox/flox (MCKO) mice and Cx3cr1^CreERT/+; Alkbh5 ^flox/flox; Ntf5 ^flox/flox (MDKO) mice under cold exposure. Scale bars, 50 μm. c Immunostaining images and quantification with anti-UCP1 and anti-PLIN1 antibody in sWAT of the indicated mice. Scale bars, 50 μm. d Immunostaining images and quantification with anti-TH antibody in sWAT of the indicated mice. Scale bars, 50 μm. e TH and UCP1 protein expression and quantification in sWAT of the indicated mice. Immunostaining images and quantification with anti-NeuN and anti-p-TrkB antibody (f), anti-NeuN and anti-SYN1 antibody (g) in PVH of the indicated mice. n = 6 Ctrl and MCKO group, n = 7 MNKO and MDKO group. h p-TrkB protein expression and quantification in primary neurons following treatment with conditioned medium from Ctrl, MNKO, MCKO and MDKO microglia. n = 6 independent experiment per group. The samples derive from the same experiment and the blots were processed in parallel. Scale bars, 50 μm. Values represent mean ± SEM. Source data are provided as a Source Data file. One-way ANOVA with Tukey’s multiple comparisons test (b–f, h), One-way ANOVA with Dunnett’s T3 multiple comparisons test (g).

ALKBH5 upregulation contributes to the aging-induced decline of brown adipocytes

Aging induces a steep decline of brown-like adipocytes in the sWAT, resulting in decreased energy expenditure and heat generation, and contributes to age-related metabolic disorders³⁷. Although MCKO did not affect adipocyte function in the sWAT in young mice at normal room temperature conditions, we further investigated the effects of MCKO on sWAT during aging. We detected a smaller adipocyte size (Fig. 7a), increased sympathetic innervation (Fig. 7b), and higher expression of TH and UCP1 of the sWAT in MCKO mice compared with age-matched Ctrl mice (Fig. 7c, d). While these effects of MCKO were not observed in iBAT and eWAT in aged mice (Supplementary Fig. 15a, b). In addition, we also detected increased expression of Ntf5 in microglia (Fig. 7e) and p-TrkB in neuron (Fig. 7f) of aged MCKO mice. Then, we measured the metabolic parameters and energy expenditure using indirect calorimetry in Ctrl and MCKO aged mice at room temperature. MCKO tended to have an increased systemic oxygen consumption, carbon dioxide release and energy expenditure, heat production (Fig. 7g-j), more water intake (Supplementary Fig. 15c) and locomotor activity (Supplementary Fig. 15f), but no significant increase of food intake (Supplementary Fig. 15d) or RER (Supplementary Fig. 15e) (at both day and night) compared with Ctrl mice. These results indicate microglial ALKBH5 is involved in regulating brown adipocyte decline in the sWAT during aging.

Fig. 7. Alkbh5-depletion in microglia delayes aging-induced decline of brown adipocyte.

a Immunostaining images and quantification with anti-PLIN1 antibody in sWAT of aging Ctrl and MCKO mice. Scale bars, 50 μm. b Immunostaining images and quantification with anti-TH antibody in sWAT of aging Ctrl and MCKO mice. Scale bars, 50 μm. c TH, UCP1 protein expression and quantification in sWAT of aging Ctrl and MCKO mice. d Th, Ucp1 gene mRNA expression levels in sWAT of aging Ctrl and MCKO mice. e Immunostaining images and quantification with anti-Iba1 and anti-Ntf5 antibody in PVH of aging Ctrl and MCKO mice. n = 10 cells of per group. Scale bars, 20 μm. f Immunostaining images and quantification with anti-NeuN and anti-p-TrkB antibody in PVH of aging Ctrl and MCKO mice. n = 4 per group. Scale bars, 20 μm. Representative images and quantification of VCO2 (g), energy expenditure (h), VO2 (i), heat production (j) with metabolic cage experiment of Ctrl and MCKO mice. n = 4 per group (g–j). Values represent mean ± SEM. Source data are provided as a Source Data file. Two-tailed Mann Whitney test (a, c(TH)), Two-tailed Unpaired ttest with Welch’s correction (b, c(UCP1), d–f), Two-way ANOVA with Šídák’s multiple comparisons test (g, i, j), ANCOVA analysis with body weight included as a covariate (h).

Discussion

Here we show that activation and proliferation of microglia in hypothalamic PVH are involved in cold exposure-induced sympathetic activation and adipocyte browning in the sWAT. Microglial-specific deletion of RNA m6A demethylase ALKBH5 enhances sensitivity to cold exposure-induced adaptive thermogenesis. ALKBH5 deletion increases the Ntf5 mRNA stability, protein expression and secretion, which phosphorylates the neuronal TrkB receptor in the PVH and activates downstream SNS in the sWAT. These findings fill in a missing but important gap in the research of microglia in neuron-mediated adipocyte metabolism, and refine the microglia-neuron-sWAT multi-organ regulatory axis.

N6-methyladenosine (m6A) is the most prevalent, abundant and conserved internal transcriptional RNA modification, which is modified by m6A methyltransferases, demethylases and methyl-recognition enzymes. The m6A modification is observed highly enriched within the brain and markedly increased throughout brain development, and implicated in multiple physiological and pathological processes of the brain^38,39. ALKBH5 and FTO are the major demethylases that regulate eukaryotic RNA m6A modification. Hypothalamic neuronal FTO promotes high-fat diet-induced leptin resistance via directly increasing CX3CL1 expression⁴⁰. Enhanced FTO expression promotes the differentiation of appetite-controlling neurons, leading to increased calorie intake and a greater threat of obesity⁴¹. ALKBH5 is ubiquitously expressed across the brain during the embryonic period and dramatically decreases in the late development. Hepatic ALKBH5 ablation improves glycolipid metabolism and attenuates MASLD progression⁴². However, little is known about the function of ALKBH5 in the CNS to control metabolic health and disorders. We herein showed that microglial ALKBH5 is involved in cold exposure-induced adipocyte browning and adaptive thermogenesis. Cold exposure results in decreased ALKBH5 expression in the microglia, but has no effect on the neurons of the PVH region. Microglial ALKBH5 knockout directly promotes the expression and secretion of Ntf5, which activates TrkB neurons, leading to increased SNS activity and adipocyte browning in the sWAT.

Adipocyte browning is triggered by stimuli, such as cold exposure and peroxisomal biogenesis. On cold exposure, SNS activity in adipose tissue at subcutaneous depots is effectively enhanced and norepinephrine is released from nerve terminals. Norepinephrine binds to the β3 adrenergic receptor on adipocytes and activates downstream signaling to stimulate the expression of UCP1 and fuel thermogenesis⁴³. Therefore, sympathetic activation is critical for cold exposure-induced adipocyte browning. Recently, several studies have shown that the adipose tissue local microenvironment, including adipocytes and immune cells, could directly regulate sympathetic activation via releasing cytokines, trace elements and neurotrophic factors. It is worth noting that sWAT is mainly innervated by postganglionic sympathetic neurons. By using a viral transneuronal tract tracer, innervation of sWAT has been demonstrated to originate from the central nervous system. Our findings further confirmed that the PVH region of the hypothalamus is a key brain region governing cold exposure-induced sympathetic innervation by detecting RFP signaling after PRV injection into sWAT. We showed that the activation and proliferation of microglia in the PVH region were stronger than those in other regions through the brain sections of Cx3cr1-GFP mice under cold exposure, suggesting that microglia are also involved in the CNS-SNS axis-mediated adipocyte browning and adaptive thermogenesis. Microglia clearance in the hypothalamus attenuates cold exposure-induced adipocyte browning. Selective peripheral denervation blocks the effects of ALKBH5 knockout-mediated microglia activation on adipocyte browning. Collectively, these results highlight a critical role of microglia in promoting sWAT sympathetic innervation.

ALKBH5/Ntf5-mediated microglia-neuron crosstalk mainly modulates cold exposure-induced SNS activity in the sWAT, but little in the iBAT and eWAT. We speculate this might be attributed to the density of sympathetic nerves in different adipose tissue depots. It is well known that iBAT is densely innervated by sympathetic nerves to maintain basal body thermogenesis, and UCP1 expression could be detected evenly under normal room temperature^44,45. While eWAT is an inert adipose tissue, it exhibits a weak response to adipocyte browning during cold exposure⁴⁶. Brown/Beige adipocytes, as well as UCP1 expression in sWAT, are barely detectable under room temperature, while SNS activity and UCP1 expression are dramatically increased under cold exposure⁴⁷. These suggest sympathetic plasticity in the sWAT and its critical role in maintaining body temperature homeostasis when cold is encountered. Since long-term systemic sympathetic activation also drives vasoconstriction and high blood pressure, microglial ALKBH5/ Ntf5-mediated sympathetic activity being selectively restricted to the sWAT is safe and efficient.

We have demonstrated that ALKBH5 deletion affects microglia-neuron proximity in the PVH, neuronal firing dynamics, and synaptic density and connectivity between the PVH and sWAT. Through literature review, we found that microglia-neuron interactions include both membrane contact and the secretion of soluble factors by microglia that act on neurons^17,48. In our experiments, MCKO was observed to enhance microglia-neuron membrane contacts in the PVH, as well as promote the secretion of Ntf5 by microglia. These findings suggest that microglia exert a stronger regulatory influence on neurons in MCKO mice. Given that Ntf5 primarily acts on TrkB-expressing neurons, we further investigated its underlying mechanism and found that Ntf5 promotes the phosphorylation of TrkB neurons. The primary phosphorylation sites of TrkB are tyrosine 515 and tyrosine 816, where phosphorylation at tyrosine 515 activates the MAPK and PI3K pathways, while phosphorylation at tyrosine 816 activates the PLCγ pathway⁴⁹. Western blot analysis revealed that MCKO specifically enhances phosphorylation at tyrosine 816. Since the PLCγ pathway is critical for sustained neuronal firing, we performed electrophysiological recordings and confirmed that microglia in MCKO mice increase neuronal firing activity. Sustained neuronal firing is essential for functions, such as working memory and synaptic plasticity^50,51. Consistent with this, synaptic staining for SYN1 and PSD95 showed that MCKO increases synaptic density, a key hallmark of synaptic plasticity. While this study focuses on the role of Ntf5 signaling, we do not exclude the potential contribution of enhanced microglia-neuron membrane contacts in the PVH of MCKO mice, which may function in parallel or interact with the Ntf5-mediated mechanism. We acknowledge that these phenotypes may not be mediated exclusively by the ALKBH5–Ntf5 signaling pathway. Nevertheless, our core finding, which is supported by Ntf5 knockout mouse models, confirms that the ALKBH5–Ntf5 axis in the PVH promotes sympathetic activation in adipose tissue and adipose browning via TrkB neuronal activation. We plan to further investigate in future studies whether the effects of microglial ALKBH5 on other neural activities also depend on Ntf5 release.

Microglia is critical for regulating neuroinflammation by directly releasing pro-inflammatory substances, such as cytokines, prostaglandins, and reactive oxygen and nitrogen species. In the process of metabolic diseases, it has been reported that microglia, located in the ARH region of the hypothalamus, can be activated to a pro-inflammatory phenotype in response to nutrient excess, leading to hypothalamic inflammation and neuronal dysfunction. For instance, microglial activation increases TNFα secretion, leading to dysfunction of adjacent POMC neurons and thus contributing to the development of obesity. On the contrary, increased physical activity reduces hypothalamic microglial activation and improvement of glucose tolerance⁵². Microglial activation modulates hypothalamic glucose sensing via a TNF-α-, melanocortin-, and parasympathetic-dependent pathway, enhancing insulin secretion and mitigating glucose dysregulation in diet-induced obesity¹⁸. Besides, microglia exhibit lower m6A levels than other brain cell types, which may reflect their rapid immune responsiveness⁵³. We here provide evidence for a paracrine pattern of microglia on hypothalamic neurons through ALKBH5-mediated release of Ntf5, a neurotrophic factor. Neurotrophins could directly bind to tropomyosin-related kinase receptors and then regulate the fate of neurons, synaptic plasticity and other neuronal functions. Our findings demonstrated that ALKBH5 deletion increases the production of microglia-derived Ntf5, which binds to neuronal TrkB and promotes its phosphorylation. TrkB activation in the hypothalamic neurons have been identified critical roles in energy homeostasis, including energy expenditure, adaptive thermogenesis, adipose tissue browning and sympathetic innervation^54,55. Collectively, these suggest that microglia may be a potential therapeutic target in the future to for the treatment of individuals with metabolic disorders.

Methods

Animals

The animal experiments were approved by the Animal Care and Use Committee of the School of Basic Medical Sciences, Fudan University, and the study adhered to the Guide for the Care and Use of Laboratory Animals by the U.S. National Institutes of Health. Animals were maintained on the 12 h light/dark cycle, with access to chow diet and water ad libitum. The Alkbh5 ^flox/+ mice (CKOCMP-268420-Alkbh5-B6N-VA), R26^iDTR mice (C001477) and Ntf5 ^flox/+ mice (S-CKO-16808) were purchased from Cyagen company. Cx3cr1^CreERT/+ mice (JAX 021160), Cx3cr1-GFP mice (JAX 005582) and Rosa26^Td mice (JAX 007914) were obtained from the Jackson Laboratory. C57BL/6 J (WT) mice (N000013), Tmem119^CreERT/+ mice (T050776) and Csf1r ^flox/+ mice (T009476) were obtained from GemPharmatech company. Lyve1^CreERT2 mice were obtained from Prof Qingbo Xu in Zhejiang University⁵⁶. The 8-week-old Cx3cr1^CreERT/+ (Ctrl) and Cx3cr1^CreERT/+; Alkbh5 ^flox/flox (MCKO) male mice were intraperitoneally injected with Tamoxifen (Sigma, T5648) dissolved in corn oil to a final concentration of 10 mg/mL, at a dose of 75 mg/kg body weight, for 5 consecutive days. Following a four-week period to allow for the turnover of the circulating monocyte/macrophage population, the mice underwent the subsequent experiments. The same tamoxifen induction protocol was applied to 8-week-old male mice of the following genotypes: Tmem119^CreERT/+ (Ctrl), Tmem119^CreERT/+; Alkbh5 ^flox/flox (MCKO2), Cx3cr1^CreERT/+; Ntf5 ^flox/flox (MNKO), and Cx3cr1^CreERT/+; Alkbh5 ^flox/flox; Ntf5 ^flox/flox (MDKO). After a four-week waiting period, all mice underwent subsequent experiments. For the aging-related experiments, 14–16 months old Cx3cr1^CreERT/+ (Ctrl) and Cx3cr1^CreERT/+; Alkbh5 ^flox/flox (MCKO) male mice were employed. All mice were maintained on C57BL/6 J background and bred and housed in a specific pathogen-free environment at the University Laboratory Animal Services Center of Fudan University.

Cold exposure

To induce adipose tissue browning and sympathetic innervation, the above male mice of the indicated genotypes were subjected to cold exposure at 6 °C for 5 days, with mice maintained at room temperature (RT) serving as controls. Following the exposure period, mice were euthanized and transcardially perfused with PBS. Adipose tissues and brains were collected concurrently. The tissues were then post-fixed in 4% paraformaldehyde (PFA) at 6 °C overnight and processed for frozen sectioning, staining, and subsequent imaging analysis.

Microglia morphology analysis

Microglial morphology in brain sections from Cx3cr1-GFP mice was assessed based on endogenous GFP signal. Using ImageJ software with the “Analyze Skeleton 2D/3D” plugin, we quantified the GFP+ cells as a measure of microglial coverage, as well as morphological parameters, such as the number of endpoints and branch length for cells.

Body temperature measurements

To assess iBAT thermogenic activity, the dorsal skin hair covering the iBAT depot was carefully removed from both Ctrl and MCKO mice. For measurements at room temperature, mice were individually housed in a cage and infrared images of the iBAT region were captured using a thermal camera. For the cold challenge, mice were transferred to individual cages within a 4 °C climate-controlled chamber. Infrared imaging was performed hourly over a 5-hour period from 10:00 to 15:00. The temperature over the iBAT region was subsequently quantified using FLIR analysis software.

PRV and HSV trace

To retrograde label adipose-projecting neurons in the CNS, 12-week-old Cx3cr1-GFP male mice were anesthetized, and an incision was made in the groin to expose sWAT. A total of 2 μl PRV-CAG-RFP (Pseudorabies virus, a neurotropic retrograde tracer expressing RFP, BrainVTA, P03002) was injected into different sites of sWAT with a stereotaxic apparatus using Hamilton syringes. Following the injection, the needle was kept in place for an additional 10 min to minimize the backflow of the viral solution. The skin incision was subsequently closed with surgical sutures. Post-surgery, the mice were monitored daily, animals underwent cold exposure (or RT for control), and were then euthanized on day 7 post-infection.

For anterograde labeling of neuron-projecting adipose tissues, 12-week-old WT male mice were anesthetized and placed in a brain stereotaxic device. An HSV (100 nL, Anterograde projective virus, BrainVTA, H01002) was injected into the PVH location (AP: −0.80, DV: −4.90, ML: ±0.25). The mice were subjected to RT or cold exposure and were then euthanized.

Chemogenetic strategy

To specifically target and inhibit the neural circuit that regulates sWAT browning, a dual-virus chemogenetic strategy was used in male Ctrl and MCKO mice. Gene deletion was induced by 5 consecutive days of tamoxifen injections. Four weeks later, a retrograde Cre-expressing virus (5 μl, ScAAV-hSyn-CRE-WPREs; BrainVTA, PT-2334) was administered into different sites of the sWAT. After 2-weeks, mice were stereotaxically injected with a Cre-dependent inhibitory DREADD virus (100 nL, rAAV-hSyn-DIO-hM4D(Gi)-EGFP; BrainVTA, PT-7157) into the PVH region. Following another four-week interval, the mice were given water with or without DCZ (final concentration 0.005 mg/mL, MCE, HY-42110) under cold exposure for 5 days.

Microglia and brain macrophage depletion

For microglia depletion, 8-week-old Cx3cr1^CreERT/+ and Cx3cr1^CreERT/+; R26^iDTR male mice were intraperitoneally injected with Tamoxifen consecutively for 5 days. After 4 weeks^57,58, the mice were completely anesthetized and placed into the brain stereotaxic device, injection DT (Aladdin, D381867, 10 ng/μL) or vehicle into brain location of PVH (AP: −0.80, DV: −4.90, ML: ±0.25) of the Cx3cr1^T/+ and Cx3cr1^CreERT/+; R26^iDTR mice. Then these mice underwent cold exposure for 5 days.

For brain resident macrophage depletion, 8-week-old Lyve1^creERT2/+ and Lyve1^creERT2/+; Csf1r^flox/flox male mice were intraperitoneally injected with Tamoxifen consecutively for 5 days. After 4 weeks, the mice underwent cold exposure for 5 days.

Single nucleus RNA-sequencing

Nuclei isolation sorting

The PVH brain tissues were harvested from Ctrl and MCKO mice and were washed in pre-cooled PBS (PBS buffer containing 2 mM EGTA). Nuclei isolation was carried out using GEXSCOPE® Nucleus Separation Solution (Singleron Biotechnologies, Nanjing, China), as referred to in the manufacturer’s product manual. Isolated nuclei were resuspended in PBSE to 10⁶ nuclei per 400 μl, filtered through a 40 μm cell strainer, and counted with Trypan blue. Nuclei enriched in PBSE were stained with DAPI (1:1000) (ThermoFisher Scientific, D1306). Nuclei were defined as DAPI-positive singlets.

Single-nucleus RNA-sequencing library preparation

The concentration of singlenucleus suspension was adjusted to 3-4 × 105 nuclei/mL in PBS. Single-nucleus suspension was then loaded onto a microfluidic chip (GEXSCOPE® Single NucleusRNA-seq Kit, Singleron Biotechnologies) and snRNA-seq libraries were constructed according to the manufacturer’s instructions (Singleron Biotechnologies). The resulting snRNA-seq libraries were sequenced on an Illumina Novaseq 6000 with 150 bp paired-end reads.

Primary analysis of raw read data

Raw reads from scRNA-seq were processed to generate gene expression matrixes using CeleScope (https://github.com/singleron-RD/CeleScope) v1.9.0 pipeline. Briefly, raw reads were first processed with CeleScope to remove low-quality reads with Cutadapt v1.17⁵⁹ to trim poly-A tail and adapter sequences. Cell barcode and UMI were extracted. After that, we used STAR v2.6.1a⁶⁰ to map reads to the reference GRCm38 (ensembl version 92 annotation). UMI counts and gene counts of each cell were acquired with featureCounts v2.0.1⁶¹ software, and used to generate expression matrix files for subsequent analysis.

Pathway enrichment analysis

To investigate the potential functions of DEGs, the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were used with the “clusterProfiler” R package 3.16.1^62. Pathways with p_adj value less than 0.05 were considered as significantly enriched. Gene Ontology gene sets, including molecular function (MF), biological process (BP), and cellular component (CC) categories, were used as reference.

m6A dot plot

Neurons of the PVH region of WT male mice under RT and cold exposure were isolated with the Adult Neuron Isolation Kit (Miltenyi, 130-126-602). For microglia isolation, brain tissue cells of PVH region of the Cx3cr1^CreERT/+; Rosa26^Td mice under RT and cold exposure were dissociated with Adult Brain Dissociation Kit (Miltenyi, 130-107-677), and RFP+ microglia was isolated by flow cytometry sorting. RNA from the isolated neuron and microglia were extracted with RNA Purification Kit (EZBioscience, B0004D, China). To assess the global m6A modification level, an RNA dot-blot assay was performed. Briefly, total RNA was purified from neurons and microglia, denatured, and spotted onto a Hybond-N+ membrane. After UV crosslinking, the membrane was probed with an anti-m6A antibody (abcam, ab151230) followed by an HRP-conjugated secondary antibody for chemiluminescent detection. Methylene blue staining of the membrane served as the loading control.

Primary microglia isolation

Newborn mice (aged 0–3 days) were euthanized and soaked in 75% alcohol. The mouse head was then excised and placed in pre-cooled D-Hanks’ solution. The skin and skull were carefully removed, and the brain was extracted and transferred to a sterile petri dish. The olfactory bulb, posterior regions and meninges were carefully dissected away, and the remaining brain tissue was minced using ophthalmic scissors. The tissue fragments were suspended with Dulbecco’s Modified Eagle Medium (DMEM), filtered through a 100 µm filter, and subjected to a 1000 g centrifugation at 5 min. The resulting pellet was resuspended in DMEM supplemented with 10% fetal bovine serum (FBS) and 1× penicillin-streptomycin (PS) and deposited into a culture bottle, and the medium was replaced every two days. After 2 weeks, the culture bottle was placed on a shaker set 200 rpm/min, 2 hs, the supernatant was collected and precipitated into microglia followed by centrifugation at 1000 g for 5 min. The resulting cell suspension was plated onto poly-D-lysine (PDL)–coated culture plates (0.1 mg/mL) and maintained in complete DMEM medium. The microglias were subsequently harvested for other experiments.

Primary neuron isolation

Pregnant WT mice at E18.5-21 day were sacrificed, and the abdominal skin was disinfected by immersion in 75% ethanol. Fetal mice were carefully retrieved and the cerebral cortex tissues were dissected into 0.5 cm³ sections using ophthalmic scissors and incubated in 0.125% trypsin solution. After digestion at 37 °C for 15 min, the enzymatic reaction was terminated by the addition of complete DMEM culture medium. The cell suspension was aspirated using a sterile pipette, and the tissue fragments were triturated 10 times gently. The resulting suspension was evenly triturated using a 1 mL pipette tip, and the cell suspension was collected and counted. 1 × 10⁶ cells were seeded with DMEM solution in 3.5 cm dish pre-treated with PDL (0.1 mg/mL) and cultured at 37 °C for 1 h. The medium was then replaced with Neurobasal medium supplemented with 1×B-27. The supernatants from microglia with Ctrl and MCKO mice were applied to neuronal cultures, and subsequent electrophysiological measurements were performed.

Sympathetic denervation

The Ctrl and MCKO male mice were completely anesthetized, and a 2 cm incision was made on one side at the junction between the dorsal lumbar spine and the tail vertebrae of the mice. Blunt dissection with forceps exposed the major dorsal blood vessels that support adipose tissue, and the nerve that binds the arteriovenous was peeled off and excised. A 2 cm incision was made in the groin of the mice, the blood vessel connecting to sWAT was found, and the nerve connecting to sWAT was also excised. Then the skin was sutured, and the other side was sutured as well. The sham group also carried out the same experimental procedures, except that the dissected nerves were not removed. After 1 week of postoperative recovery, cold exposure was performed.

The Ctrl and MCKO male mice were completely anesthetized, small incisions were made on the skin of the sWAT, the adipose tissue was exposed, VC or 6-OHDA (8 mg/mL) was injected into each side by the microinjector, the injection time was 5 min, the needle was stopped for 10 min, the needle was removed, and the skin was sutured with a medical suture. Until the wound healed, the mice were subjected to cold exposure.

NE measurement

Adipose tissue samples of the Ctrl and MCKO male mice were lysed for 20 min at 4°C in RAPI lysis buffer (Beyotime, Cat#: P0013B; China) to isolate total proteins. NE levels of adipose tissue and plasma were measured by NE Elisa kit (mlbio, ml063805).

Parabiosis

Parabiosis assay was conducted as described earlier⁶³, in briefly, the mice of indicated genotypes were completely anesthetized, each mouse was shaved along a contiguous line from the elbow, the flank, to the knee on the side to be joined, and the skin incisions were performed along the flank, ranging from the knee to the elbow without damaging the muscle tissue underneath, the triceps and quadriceps of the parabionts were joined with two interrupted sutures, then the ventral and dorsal skin of the parabionts was closed with interrupted sutures, after the parabiosis surgery, each pair was housed individually. We conducted parabiosis between Cx3cr1^CreERT/+; Alkbh5 ^flox/flox(MCKO); Rosa26^Td mice and WT (MCKO) mice, Cx3cr1^CreERT/+(Ctrl); Rosa26^Td mice and WT (Ctrl) mice, after two weeks, and the heterochronic parabiosis group intraperitoneally injected Tamoxifen, then cold exposure was performed.

Adipose tissue transplantation

The mice of indicated genotypes were completely anesthetized, skin incision of ventral sWAT of WT mice were made, the same location and volume (Length: 1 cm) prepared sWAT of Cx3cr1^CreERT/+; Alkbh5 ^flox/flox (MCKO) mice or Cx3cr1^CreERT/+ (Ctrl) mice were transplanted into ventral sWAT of WT mice by human cellulose proteinogen, after 5-10 min, the skin was sutured with a medical suture, cold exposure was performed.

MeRIP-seq

Shake and mix Protein A/G beads (Repligen), absorb 100 μL Protein A/G beads into 1.5 mL centrifuge tube, and use 0.5 mL IP buffer (150 mM NaCl and 0.1% (vol/vol) NP-40, 10 mM Tris-HCl, pH 7.4) (stored at 4 °C) were washed twice, and Protein A/G beads were suspended with IP buffer. The total RNA of microglia of Ctrl and MCKO mice from cells was extracted with RNA Purification Kit (EZBioscience, B0004DP) as above described. The RNA was chemically fragmented into 100-nucleotide-long fragments at 94 °C for 5 min in fragmentation buffer (10 mM ZnCl2, 10 mM Tris-HCl, pH 7), remove the tubes and immediately add one-tenth volumes of 0.5 M EDTA, add one-tenth volumes of 3 M sodium acetate (pH 5.2), glycogen (100 μg/ml final) and 2.5 volumes of 100% ethanol for precipitation. Centrifuge the tubes at 15,000 g for 25 min at 4 °C, discard the supernatant, wash the pellet with 1 ml of 75% (vol/vol) ethanol and centrifuge again at 15,000 g for 15 min at 4 °C. The fragmented RNA (200 μg) was incubated with 30 μg anti-m6a antibody (Abcam, ab151230) in IP buffer at 4°C for 2-3 h supplemented with RNase inhibitors. Then the mixture was incubated with Protein A/G beads at 4°C for 2 h. After extensive washing, bound RNA was eluted from the beads with 0.5 mg/ml N6-methyladenosine (Sigma-Aldrich, 1867-73-8) in IP buffer, followed by ethanol precipitated. The RNA was resuspended in RNase-free water and used for sequencing.

Electrophysiological experiment

The neurons were placed in ACSF solution on a 6 cm dish table and oxygenated, and the glass microelectrode (4–8 MΩ) was immersed in artificial cerebrospinal fluid (ACSF: 130 mM NaCl, 3 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, 10 mM Glucose, pH7.4). ACSF is continuously oxygenated with 95% O2/5% CO2 and maintained at 37 °C. The patch microelectrode was filled with intracellular solution (150 mM K gluconate, 2 mM MgSO₄ 7H₂O, 0.1 mM CaCl₂, 1 mM EGTA, 2 mM K₂ATP, 0.1 mM Na₃GTP, 10 mM HEPES, pH7.4). In seal formation (>1 GΩ) and resting membrane potential and membrane capacitance were measured immediately after membrane rupture. The action potential (AP) was recorded in a current clamp configuration from −400 to +150 pA with 50 pA steps and a 400 ms duration per pulse. The spontaneous postsynaptic current was recorded in a voltage clamp with a holding potential of −65 mV. Signal recorded at room temperature, 2 kHz digital low-pass filtering, 10 kHz sampling, using Axopatch 700B patch-clamp amplifier (Axon Instruments, Foster City, CA, USA), Digidata 1322 A (Molecular Devices) and stored on a PC using pCLAMP software (Version 6.0.4, Axon Instruments). Data was analyzed offline using Clampfit 10.0 (Axon Instruments) software.

To detect brain electrophysiological activity in vivo, mice were deeply anesthetized before sacrifice by decapitation, the brain was quickly removed and immersed in ice-cold sucrose-based ACSF (10 mM glucose, 213 mM sucrose, 26 mM NaHCO₃, 1.25 mM NaH₂PO₄, 2.5 mM KCl, 2 mM MgSO₄, and 2 mM CaCl₂). The brain slices with a thickness of 200 μm were cut in sucrose-based ACSF by a vibratome (Leica VT 1200S). Afterward, brain slices were immediately transferred to an incubation chamber filled with 95% O₂ and 5% CO₂ equilibrated normal ACSF (25 mM glucose, 126 mM NaCl, 26 mM NaHCO₃, 1.25 mM NaH₂PO₄, 2.5 mM KCl, 2 mM MgSO₄, and 2 mM CaCl₂) at 34 °C for 45 min. Slices were then transferred to 95% O₂ and 5% CO₂ equilibrated normal ACSF at RT before recording. Slices were then transferred to a recording chamber continuously perfused in 95% O₂ and 5% CO₂ equilibrated normal ACSF, with the temperature maintained at 34 °C ± 1 °C. An infrared-differential interference contrast (IR-DIC) microscope (Olympus BX-51WI) was used for visualization of individual neurons. The glass microelectrode (4–8 MΩ) was used in normal ACSF; the patch microelectrode was filled with intracellular solution. In seal formation ( > 1 GΩ) and resting membrane potential were measured and membrane rupture, when the membrane potential was held at –70 mV, the AP is recorded in a current clamp configuration from −400 to +50 pA with 50 pa per pulse step and 400 ms duration. Membrane voltage and current were sampled at 10–25 kHz and low-pass filtered at 2–10 kHz using the patch-clamp amplifier MultiClamp 700B (Molecular Devices, LLC), digitized and sampled by Micro 1401 with Spike2 software (Cambridge Electronic Design). The data were analyzed using MATLAB R2023a.

Metabolic cage

Body weight of the Ctrl and MCKO mice was recorded prior to the experiment using TSE PhenoMaster software (V7.5.7, 2020-6039). Mice were then housed in individual metabolic cages with ad libitum access to food and water. The system was programmed to initiate monitoring, and parameters including oxygen consumption (VO₂), carbon dioxide production (VCO₂), food intake, and water intake were automatically recorded over a predetermined period. The collected data were analyzed and presented as line graphs.

HE staining

Following fixation in 4% PFA, mouse adipose tissue was dehydrated through a graded alcohol series (30%, 50%, and 75%). The tissues were then embedded in paraffin and sectioned at a thickness of 5 µm. The resulting sections were baked overnight at 60 °C and subsequently stained with hematoxylin and eosin (H&E) for the assessment of adipocyte browning.

Immunofluorescence staining

Brain tissues were fixed in 4% PFA and sectioned at 30 µm thickness. Adipose tissues were fixed in 4% PFA and frozen sectioned at a thickness of 30 or 60 μm for Tyrosine hydroxylase (TH) antibody staining, and paraffin sectioned at a thickness of 5 µm for other antibodies following paraffin embedding. The sections were conducted antigen retrieval in sodium citrate at 95 °C for 15 min. The sections were then permeabilized and blocked with 0.3% Triton X-100 and 5% bovine serum albumin (BSA) at room temperature for 1 h, and incubated with the following primary antibodies: mouse anti-Iba1 (Abcam, ab283319, 1:400), rabbit anti-Iba1 (Abcam, ab178846, 1:500), rabbit anti-Ntf5 (Millipore, AB1781SP, 1:100), rabbit anti-TH (Millipore, AB152, 1:500), rabbit anti-UCP1 (Abcam, ab234430, 1:200), goat anti-PLIN1 (Abcam, ab61682, 1:300), rabbit anti-ALKBH5 (Proteintech, 16837-1-AP, 1:200), rat anti-F4/80 (Abcam, ab6640, 1:200), rabbit anti-CX3CR1 (Abcam, ab308614, 1:200), rabbit anti-LYVE1 (CST, 67538S, 1:200), rabbit anti-NeuN (Abcam, ab177487, 1:300), mouse anti-NeuN (Abcam, ab104224, 1:300), and rabbit anti-p-TrkB (Tyr816) (Millipore, ABN1381, 1:200) overnight at 4 °C. After washing with PBST, the sections were incubated with the following secondary antibody: Alexa Fluor^TM488 donkey anti-rabbit IgG(H + L) (Invitrogen, A21206, 1:1000), Alexa Fluor^TM594 donkey anti-rabbit IgG(H + L) (Invitrogen, A21207, 1:1000), Alexa Fluor^TM488 donkey anti-mouse IgG(H + L) (Invitrogen, A21202, 1:1000), Alexa Fluor^TM488 donkey anti-goat IgG(H + L) (Invitrogen, A11055, 1:1000) for 1 h in dark. The sections were then stained with DAPI for 10 min at room temperature. Fluorescent images were captured using a Leica SP8 confocal microscope, and the fluorescent area was quantified using ImageJ software.

Microglia or neurons were fixed in 4% PFA for 15 min, washed with PSBT, and then permeated and blocked with 0.1% Triton X-100 and 5% BSA at room temperature for 30 min, the sections were then incubated with primary antibodies overnight at 4 °C, washed with PBST, and incubated with secondary antibody for 1 h in dark, followed by DAPI staining for 10 min at room temperature. The following primary antibodies were used: rabbit anti-MAP2 (Millipore, AB5622, 1:400), rabbit anti-Synapsin1 (SYSY, 106103, 1:500), mouse anti-PSD95 (Abcam, ab2723, 1:500). Fluorescent images were captured using a Leica SP8 confocal microscope, and the fluorescent area was quantified using ImageJ software.

RNA extraction and quantitative real-time PCR analysis

Total RNA was extracted from cells using the RNA Purification Kit (EZBioscience, Cat#: B0004DP). For tissue samples, RNA was extracted using the Tissue RNA Purification Kit (EZBioscience, Cat#: RN001-Plus). Reverse transcription of the isolated RNA was carried out using the Reverse Transcription Kit (EZBioscience, Cat#: A0010CGQ), following the manufacturer’s instructions. Quantitative real-time PCR analysis was performed using SYBR Green qPCR master Mix (EZBioscience, Cat#: A0012-R2).

Western blot

Cells and tissue samples were lysed for 20 min at 4 °C in RAPI lysis buffer (Beyotime, Cat#: P0013B; China) to isolate total proteins. Protein concentration was determined using bicinchoninic acid reagent (Beyotime, Cat#: P0010; China). The proteins were then separated by SDS-PAGE and transferred to PVDF membranes. The membranes were blocked with 5% BSA and incubated with primary antibody: rabbit anti-ALKBH5 (Proteintech, 16837-1-AP, 1:1000), rabbit anti-Ntf5 (Millipore, AB1781SP, 1:1000), rabbit anti-TH (Millipore, AB152, 1:1000), rabbit anti-UCP1 (abcam, ab234430, 1:2000), rabbit anti-p-TrkB (Millipore, ABN1381, 1:1000), goat anti-TrkB (R&D systems, AF1494, 1:1000), rabbit anti-GAPDH (Servicebio, GB15004, 1:2000), rabbit anti-β ACTIN(Servicebio, GB11001, 1:2000) overnight at 4 °C. Following incubation, the membranes were washed with TBST and then incubated with second antibody: HRP-conjugated Affinipure Goat Anti-Mouse IgG (H + L) and Anti-Rabbit IgG secondary antibody (Proteintech, SA00001-1 and SA00001-2; 1:4000) for 1 h. The membranes were then washed with TBST, and ECL substrate was applied for detection.

Statistical analysis

Statistical analyses were performed using GraphPad prism 9.0 software (GraphPad Software, San Diego, CA). The normality of continuous variables was assessed with the Shapiro-Wilk test and all data passed normality and equal variance tests. Two-tailed Unpaired t test, Two-tailed Unpaired ttest with Welch’s correction, Two-tailed Mann-Whitney test for 2-group comparisons were used to determine statistical significance. One-way ANOVA followed by Tukey’s multiple comparison analysis or Dunnett’s T3 multiple comparisons analysis was applied for three or more group comparisons. Two-way ANOVA, followed by Šídák’s multiple comparisons test, was used to compare multiple groups when 2 experimental factors were involved. All data are presented as means ± SEM (standard error of the mean), and values of P < 0.05 was considered statistically significant. All the statistical methods used for each analysis were indicated in the corresponding figure legends. No data were excluded from the analyses. The sample ID was blinded to the investigator for image quantification.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Y.-F.Y. designed and performed the experiments, analyzed the data, and wrote the manuscript. Y.L. designed and performed the experiments, analyzed the data. M.-H.C., T.M., X.-D.H., Y.-J. H., X.-D.S., D.-H.H. and C.-Y.S. performed the experiments. T.Z. and X.-R.Z. provided suggestions on the project design and revised the manuscript. Y.L. designed and performed the experiments, analyzed the data, validated results and revised the manuscript. R.-G.L. and C.-C.R. conceived the project, designed and supervised the experiments, and wrote and revised the manuscript.

Peer review

Peer review information

Nature Communications thanks Chan Hee Lee and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. [A peer review file is available].

Funding

This work was supported by the Noncommunicable Chronie Diseases-National Science and Technology Major Project (2023ZD0503304), National Key R&D Program of China (2022YFA1104200), the Medical-Engineering Cross-disciplinary Research Fund of the “Star of Jiao Tong” Program, Shanghai Jiao Tong University (Grant No. YG2026ZD26), National Natural Science Foundation of China (92368105, 32271233, 82300513, 82570329, 82270298 and 82500512).

Data availability

The Single nucleus RNA-sequencing data in Fig. 2, MeRIP-sequencing data and mRNA-sequencing data in Fig. 5 are deposited in Genome Sequence Archive database (CRA039090, CRA039091, CRA039092). All data supporting the findings of the paper are available within the paper and the Supplementary Information. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77050-w.