A brain support cell protein involved in stress responses may contribute to depression-like behavior in mice

To investigate whether the m6A-modifying is aberrant in depression, we performed real-time quantitative PCR (qRT-PCR) analysis of the peripheral blood of 30 MDD patients and 36 healthy controls. The results showed that ALKBH5 was significantly upregulated and FTO was downregulated in the patient group among RNA methyltransferases (METTL3, METTL14, and WTAP), demethylases (FTO and ALKBH5) and m6A binding protein (YTHDF1 and YTHDF2) (Fig. 1a). Similarly, we reanalyzed published MDD datasets (GSE102556↗), and the results revealed that mRNA expression of ALKBH5 was significantly increased in the dlPFC of MDD patients compared with the healthy controls (Fig. 1b)²⁰. Next, we used chronic social defeat stress (CSDS) to mimic depressive-like behavior in preclinical studies of depression²¹. In the 10 days CSDS, 8-week-old male C57BL/6J male mice were exposed to an aggressive CD-1 mouse for social defeat and separated into susceptible (Sus) and resilient (Res) subpopulations as assessed with the social interaction (SI) test (Supplementary Fig. 1a, b). After CSDS, we collected the peripheral blood and the tissues of mPFC, a candidate site for impaired function in depression^13,19, and examined the mRNA expression of m6A-modifying enzymes. The mRNA expression of ALKBH5 in the peripheral blood of Sus mice was significantly upregulated, whereas the RNA expression levels of METTL3 and FTO were downregulated both in Sus and Res mice compared with control (Ctrl) mice (Fig. 1c). Besides, the ALKBH5 and FTO mRNA were also upregulated in the mPFC of Sus mice, whereas the METTL3 and METTL14 levels were changed in both of Sus and Res mice (Fig. 1d). Notably, ALKBH5 mRNA levels were increased in human and mouse peripheral blood as well as in brain tissues; and the ALKBH5 mRNA levels in the mPFC of Ctrl, Sus and Res mice correlated with SI Ratio in the SI test, respectively (Fig. 1e). Thus, we directly analyzed ALKBH5 protein levels in several mood-related brain regions, including the mPFC, nucleus accumbens (NAc), amygdala (Amy), striatum (Stri), hippocampus (Hip) and dorsal raphe nucleus (DRN) after CSDS. Western blotting analysis showed that ALKBH5 protein was specifically increased in the mPFC of Sus mice (Fig. 1f and Supplementary Fig. 1d), and the ALKBH5 protein level was correlated with SI ratio (Supplementary Fig. 1c). Furthermore, the total m6A level was significantly decreased in the mPFC of Sus mice rather than Ctrl mice (Fig. 1g). In addition, to determine whether stressed female mice also showed such increases in ALKBH5 expression, we used a protocol for CSDS in female mice. Following the CSDS paradigm, mPFC tissue was isolated from the mice and used for qPCR analysis. The result showed that ALKBH5 in the mPFC of female Sus mice was also increased, as in male mice (Supplementary Fig. 1e). In the following, we measured the expression of ALKBH5 in lipopolysaccharide (LPS)-induced mouse model of depression²². A total of 7 days LPS treatment (0.5 mg kg⁻¹ day⁻¹, intraperitoneal) induced a depressive-like phenotype in the adult male C57BL/6J mice in the forced swim test (FST) (Supplementary Fig. 1f). In the LPS-treated mice, ALKBH5 and FTO levels in the mPFC were also observed to have significantly increased (Fig. 1h and Supplementary Fig. 1g). These results suggest that ALKBH5 levels are consistently elevated in human and mouse peripheral blood and brain tissue.

Neurons, glial cells, and endothelial cells are representative cell types in the adult brain²³. To identify ALKBH5 in which subtype of cells responds primarily to stress, we detected ALKBH5 protein levels in vitro and in vivo. First, the GR agonist dexamethasone (DXMS) or LPS was applied in a dose- and time-dependent manner to cultured neurons, endothelial cell line (bEnd3), or astrocytes representing glial cells. Western blotting showed an increase in ALKBH5 protein levels in astrocytes treated with low concentrations of DXMS (1 μM) or LPS (0.5 μg mL⁻¹), and neurons showed an increase in ALKBH5 protein levels only when treated with high concentrations of DXMS (5 μM) or LPS (1.5 μg mL⁻¹), whereas endothelial cells showed no significant change in ALKBH5 protein in response to DXMS or LPS treatment (Fig. 1i, j). Meanwhile, the upregulation of ALKBH5 in astrocytes was faster than that in neurons and endothelial cells when treated with DXMS (1 μM) or LPS (1 μg mL⁻¹), suggesting that ALKBH5 in glial cells is more sensitive to stress than that in neurons and endothelial cells (Supplementary Fig. 2a, b). Furthermore, to confirm the change of glial ALKBH5 in vivo, astrocytes, microglia, or oligodendrocytes in the mPFC were collected by fluorescence-activated cell sorting (FACS) or Magnetic cell sorting (MACS) after CSDS (Supplementary Fig. 1h). qRT-PCR analysis showed that ALKBH5 was decreased in microglia in the mPFC of Sus mice and ALKBH5 was no significant change in oligodendrocytes (Supplementary Fig. 3a, b). Meanwhile, simple western blotting analysis showed that astrocytic ALKBH5 was significantly increased in the mPFC of Sus mice, consistent with the change in total ALKBH5 in Fig. 1 (Supplementary Fig. 1i). Together, these results suggest that the elevation of astrocytic ALKBH5 is a major contributor to the upregulation of total ALKBH5 in animal models.

To examine how ALKBH5 affects depressive-related phenotypes, we generated astrocyte-specific, neuron-specific, and endothelial cell-specific conditional ALKBH5 deletion mouse lines by crossing the ALKBH5 allele in which loxP sites flank exon 1 with Fgfr3-iCreER^T2, CaMKIIα-creER^T2 or Cdh5-iCreER^T2 transgenic lines (Fig. 2a and Supplementary Fig. 4a). Astrocyte cKO, Neuron cKO, EC cKO and littermate Ctrl mice exhibited no changes respectively in growth rate, body weight, brain size as well as in brain structure (Supplementary Figs. 4b–d, 6a–c, 7a). Western blotting analysis showed that ALKBH5 protein was significantly decreased in Astrocyte cKO mice (Supplementary Fig. 4f) and Neuron cKO (Fig. 2h) mice after treatment with tamoxifen. Immunostaining revealed that ALKBH5 were attenuated in S100β positive astrocytes in the mPFC of Astrocyte cKO mice (Fig. 2b, c and Supplementary Fig. 4e), NeuN positive neurons in the mPFC of Neuron cKO mice (Fig. 2i and Supplementary Fig. 6d) and Cdh5 positive endothelial cells in the mPFC of EC cKO mice (Fig. 2m, n) compared with Ctrl mice. We co-stained ALKBH5 and NeuN, Iba1, and Sox10 in the mPFC slices of Astrocytic ALKBH5 cKO mice and littermate controls. Confocal imaging and quantification analysis showed that the ALKBH5 positive neurons, oligodendrocytes, and microglias were not significantly changed in the astrocytic ALKBH5 cKO mice, suggesting that selective knockout of ALKBH5 in astrocytes (Supplementary Fig. 5a–c).

For behavioral studies, we first investigated the behavioral performance of Astrocyte cKO mice and littermate controls. In the force swimming test (FST), the duration of immobility was decreased in Astrocyte cKO mice compared with Ctrl mice (Fig. 2d). In the CSDS experiment, Astrocyte cKO and Ctrl mice showed no difference in social interaction in the baseline—pre-defeat conditions (Fig. 2e). However, after social defeat stress for 10 consecutive days, Ctrl mice spent less interaction time with aggressive CD1 target in SI test, whereas Astrocyte cKO mice prevented the development of the social avoidance (Fig. 2e and Supplementary Fig. 4i). In the sucrose preference test (SPT), Astrocyte cKO and Ctrl mice showed no difference in the sucrose preference test before CSDS. After 10 days CSDS paradigm, the sucrose preference was decreased in Ctrl mice, whereas Astrocyte cKO mice prevented the development of anhedonia (Fig. 2f). After mice were exposed to anxiety-related behavioral tests, Astrocyte cKO mice spent less time in the open arms of elevated plus maze (EPM) (Fig. 2g), but no behavioral differences were observed in the light-dark box (LD) (Supplementary Fig. 4k). Astrocyte cKO mice showed a slight decrease in locomotor performance in the open field test (OFT) (Supplementary Fig. 4g, j), but no differences were observed in motor coordination in the rotarod test (Supplementary Fig. 4h).

Then, we investigated the behavioral performance of Neuron cKO and EC cKO mice (Fig. 2a). No obvious behavioral changes were observed between the Neuron cKO and Ctrl mice in FST, SI test before and after CSDS and OFT (Fig. 2j, k and Supplementary Figs. 6e–h). Neuron cKO mice spent less time in the open arm in the EPM (Fig. 2l) and more time in the dark box in the LD (Supplementary Fig. 6i–k). For EC cKO mice and littermate controls, no obvious changes were observed in the FST, SI test before and after CSDS, OFT, LD, and EPM (Fig. 2o–q and Supplementary Fig. 7b–h). Together, these data indicate that selective knockout of ALKBH5 in adult astrocytes produces antidepressant-like behaviors, not in neurons or endothelial cells.

To determine the brain region-specific effect of ALKBH5 on behaviors, we specifically knocked out ALKBH5 in the mPFC of adult Alkbh5^loxP/loxP mice by using bilateral injection of AAV-gfaABC1D-GFP-iCre or AAV-gfaABC1D-eGFP (Fig. 3a). Confocal imaging and Western blotting analysis showed that viral transfection resulted in reduced ALKBH5 expression in the mPFC astrocytes (Fig. 3b, c). To address the specificity of AAV-gfaABC1D-eGFP-iCre expression, we immunostained S100β of the mPFC slice from Alkbh5^loxp/loxp mice injected with GFAP-icre virus. Confocal imaging and quantification analysis showed that mostly eGFP-positive cells co-stained with S100β in the slices (Fig. 3b and Supplementary Fig. 8a). We next examined behavioral performance and found that the mice infected with the AAV-gfaABC1D-GFP-iCre virus (GFAP^△Alkbh5) exhibited antidepressant-like behaviors, including reduced immobility time in the FST (Fig. 3d) and prevented the development of the social avoidance in SI after CSDS paradigm (Fig. 3e and Supplementary Fig. 8e). No differences were observed between GFAP^△Alkbh5 and Ctrl mice in EPM, OFT and LD (Fig. 3f–h and Supplementary Fig. 8b–d).

We then tested whether overexpression of ALKBH5 in the mPFC astrocytes (OE-Alkbh5) was sufficient to induce depressive-like behaviors by using a viral expression approach (Fig. 3i). Confocal imaging and Western blotting analysis confirmed the upregulation of ALKBH5 in the mPFC astrocytes (Fig. 3j, k). In behavioral tests, a significant increase in immobility time in FST was observed in OE-Alkbh5 mice (Fig. 3l). After that, we employed a 3-day subthreshold social defeat stress (SSDS) paradigm. Before SSDS, OE-Alkbh5, and Ctrl mice showed no difference in social interaction, but after SSDS, OE-Alkbh5 mice spent statistically less time in the interaction zone in the SI test compared to Ctrl mice (Fig. 3m and Supplementary Fig. 8j). No differences were observed between OE-Alkbh5 and Ctrl mice in EPM, OFT and LD (Fig. 3n–p and Supplementary Fig. 8f–i).

To determine whether restoration of astrocytic ALKBH5 is sufficient to reverse the antidepressant-like behaviors induced by astrocytic ALKBH5 knockout, we injected AAV-DIO-ALKBH5 or Ctrl virus into the mPFC of Astrocyte cKO mice (Fig. 3i). As expected, the Astrocyte cKO mice overexpressed ALKBH5 in mPFC showed the increase of immobility in FST (Fig. 3q) and spent less time in SI test compared with Astrocyte cKO mice infected with Ctrl virus (Fig. 3r and Supplementary Fig. 8n). No differences were observed in EPM, OFT and LD (Fig. 3s–u and Supplementary Fig. 8k–m). These findings suggest that the gain or loss of astrocytic ALKBH5 in the mPFC bidirectionally modulates depression-related behaviors.

Previous studies show that Fgfr3 is also expressed in the interneurons of the olfactory bulb²⁴. To rule out the effect of the Fgfr3-iCre mouse line in driving recombination in the olfactory bulb interneuron, we injected AAV2/9-Syn-eGFP-WPRE-pA virus into the olfactory bulb of Alkbh5^loxp/loxp mice (iCre) (Supplementary Fig. 9a). Western blotting analysis showed that viral transfection resulted in reduced ALKBH5 expression in olfactory bulb (Supplementary Fig. 9b). Behavioral tests were performed, and results showed that no significant difference was observed in FST (Supplementary Fig. 9c). The iCre mice showed a decrease in locomotor performance in OFT (Supplementary Fig. 9d–g). And the iCre mice spent less time in the open arms of EPM and less time in LD (Supplementary Fig. 9h–k). These results exclude the influence of ALKBH5 in the olfactory bulb interneuron in depression-related behaviors.

To investigate the mechanism by which astrocytic ALKBH5 regulates depression-related behavior, we collected mPFC tissues and performed m6A epitranscriptomic analysis by MeRIP-Seq (Supplementary Fig. 10a, b). Consistent with previous studies²⁵, m6A modification was increased mainly in the 3’ UTR, rather than in the 5’ UTR or coding region (Fig. 4a, b), and ‘GGAC’, the most frequent m6A consensus motifs, were contained in the enriched motifs in all samples (Fig. 4c). Next, we analyzed the differentially methylated peaks (DMPs) between Astrocyte cKO and littermate controls (Fig. 4d). Most DMPs were distributed in exons and the 3’ UTR, and the overall differentially expressed genes (DEGs) profiles of the Astrocyte cKO and Ctrl groups were obtained with a clear separation after hierarchical cluster analysis (Fig. 4e). Furthermore, by way of pathway analyses with Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) the Glutamatergic synapse pathway was significantly enriched in the dataset (Fig. 4f, g). The pathway includes the glutamate transporter SLC1A2 (also known as GLT-1 or EAAT2), a glutamate transporter primarily expressed on astrocytes that scavenges glutamate from excitatory synapses and prevents glutamate excitotoxicity^26,27. The m6A modification of GLT-1 was significantly increased in the mPFC of Astrocyte cKO mice compared to the Ctrl group, according to quantitative analysis of consensus peaks (Fig. 4h). These findings suggest that astrocytic ALKBH5 may be involved in glutamatergic synapses via the m6A modification of GLT−1.

To investigate whether the GLT-1 m6A sites regulate the expression of GLT−1, we cloned two mutants at the m6A modification site of the 3’UTR of GLT−1 (Fig. 5a) into the psiCHECK-2 double luciferase expression vector. The luciferase activity was reduced in cells transfected with the SLC1A2 mutant vector compared with the empty vector control, suggesting that GLT-1 3’ UTR m6A site regulate GLT−1 expression (Fig. 5b). Then, we used a single-base editing to construct dCas9-mCherry and GLT-1-targeting-sgRNA-puro-eGFP and then transfected them into mixed primary cultured astrocytes to alter the 3’ UTR region m6A modification site of GLT-1 (Fig. 5c). Western blotting analysis showed that GLT-1 protein levels were significantly decreased in the astrocytes with GLT-1-targeting-sgRNA-puro-eGFP compared with the Ctrl plasmid (Fig. 5d). Then we determined the glutamate uptake capacity of astrocytes by measuring glutamate levels in the medium after 4 h of l-glutamate (200 M) treatment and found that the concentration of glutamate in the sgRNA group was almost 40% higher than that in the Ctrl group, indicating 201 mutant GLT-1 disrupted the glutamate uptake ability of astrocytes (Fig. 5e).

To examine the contribution of astrocytic ALKBH5 in the m6A modification of GLT-1, we then examined the expression and uptake ability of GLT-1 in ALKBH5 cKO astrocytes in vitro. Western blotting analysis showed a decrease in ALKBH5 and an increase in GLT-1 in the primary cultured astrocytes from Alkbh5^loxP/loxP mouse infected with pLenti-GFAP-Cre virus (Fig. 5f). Meanwhile, the level of uptaken glutamate was increased in ALKBH5 cKO astrocytes compared with the Ctrl (Fig. 5g). Furthermore, we examined GLT-1 expression and extracellular glutamate dynamics in the mPFC of Astrocyte cKO mice. Western blotting analysis revealed increased GLT-1 in the mPFC of Astrocyte cKO mice compared with the littermate control (Fig. 5h).

To detect extracellular glutamate dynamics in vivo, we employed a glutamate sensor (iGluSnFR, A184S) containing a high-affinity SF-iGluSnFR variant that enhances the detection of stimulus-evoked glutamate release²⁸. After the Astrocyte cKO and Ctrl mice were further treated through injections of AAV-Syn-iGluSnFR(A184S) in the mPFC, fiberoptic implants were placed above the infected cells (Fig. 5i). Two weeks after the viral expression, photometry recordings were performed in chronically stressed mice with a forced interaction test (FIT) assay (Fig. 5j). Before CSDS paradigm, the heatmaps and maximum fluorescence change (Z-score) showed that similar amounts of glutamate was released in astrocytic cKO and Ctrl mice when interacting with CD1 in FIT. After 10 days of CSDS, the amount of glutamate was increased in the Ctrl mice to a large extent as well during CD1 contact, but the extent of glutamate increase in the astrocytic cKO mice was notably reduced (Fig. 5k–m).

Furthermore, we specifically knocked down astrocytic GLT-1 by using AAV- gfaABC1D-GLT-1-shRNA. GLT-1-shRNA or Ctrl virus was injected into the mPFC of GFAP^△Alkbh5 mice (Supplementary Fig. 11a). Western blotting analysis showed that GLT-1-shRNA reversed the elevation of GLT-1 level in the mPFC of GFAP^△Alkbh5 mice (Supplementary Fig. 11b). The GFAP^△Alkbh5 mice infected with GLT-1-shRNA showed an increase in immobility in the FST (Fig. 5m) and spent less time in the SI test after CSDS (Fig. 5n and Supplementary Fig. 11g), suggesting that GLT-1-shRNA reversed the antidepressant-like effects of GFAP^△Alkbh5 mice. No differences were observed in EPM, OFT, and LD (Supplementary Fig. 11c–f). These results suggest that astrocytic ALKBH5 mediates depression-related behaviors via GLT-1 m6A methylation.

To investigate whether ALKBH5 in astrocytes resists stress by affecting synaptic transmission, we performed the whole-cell patch clamp to record the action potentials from layer IV-V pyramidal neurons in the astrocytic ALKBH5 cKO and Ctrl mice. We found no difference in the amplitude and number of action potentials in Ctrl and astrocytic ALKBH5 cKO mice with the presence or absence of stress (Fig. 6a–c). Next, we recorded spontaneous excitatory postsynaptic currents (sEPSCs) in the mPFC. After 10 days CSDS, the frequency of sEPSCs in the Ctrl mice was significantly decreased, but remained unchanged in the astrocytic ALKBH5 cKO mice in response to stress (Fig. 6d, e and Supplementary Fig. 12a, b), and meanwhile, the amplitude of sEPSCs was observed with no difference in the mice of two groups (Fig. 6f).

Then, miniature excitatory postsynaptic currents (mEPSCs) in the pyramidal neurons were determined to investigate the effects of basal glutamatergic synaptic transmission in the mPFC of Astrocyte ALKBH5 cKO and Ctrl mice. The results showed that the frequency and amplitude of mEPSC were slightly decreased after conditional knockdown of ALKBH5 in the astrocytes under basal conditions (Fig. 6g, i). After CSDS, the frequency of mEPSCs was significantly decreased in the Ctrl mice, but slightly increased in the astrocytic ALKBH5 cKO even under stress (Fig. 6g, h and Supplementary Fig. 12c, d), and meanwhile, the amplitude of mEPSCs was not affected by stress (Fig. 6g, i and Supplementary Fig. 12c, d). Furthermore, we analyzed the decay times of individual mEPSCs. Under basal conditions, we observed that the decay time was decreased after the conditional knockdown of ALKBH5 in the astrocytes (Fig. 6j, k). After CSDS, the decay time of mEPSCs was elevated in the Ctrl mice, indicating longer glutamate clearance from synapses (Fig. 6j, k), and it was slightly increased in the astrocytic ALKBH5 cKO even under stress (Fig. 6j, k). Next, mEPSCs in the pyramidal neurons of GFAP^△Alkbh5 mice infected with GLT-1-shRNA were also recorded, and the results showed that both the frequency and amplitude of mEPSCs were slightly decreased after conditional knockdown of ALKBH5 in astrocytes, which was reversed by GLT-1-shRNA (Supplementary Fig. 12e, f). These findings implied that the knockout of astrocytic ALKBH5 prevented attenuation of glutamatergic synaptic transmission during chronic stress.

As excitotoxicity induced by chronic stress or abnormally high levels of extracellular glutamate is reported to cause dendritic atrophy and spinal loss, leading to neuronal dysfunction^29–31, we used sparingly labeled layer II/III pyramidal neurons in the mPFC of Astrocyte cKO and Ctrl mice with a viral cocktail (1:1) of AAV-CaMKII-FLP and AAV-nEf1-FDIO-EYFP to evaluate neuronal morphological alterations in response to chronic stress. Morphological analysis showed that both Astrocyte cKO and Ctrl mice exhibited no significant differences in total dendritic length, dendritic complexity, and spine density of pyramidal neurons before the CSDS experiment, but all of them were significantly decreased in Ctrl mice after the CSDS experiment (Fig. 7a–e). Notably, the total dendritic length, dendritic complexity, and spine density in the Astrocyte cKO mice were relatively intact after CSDS (Fig. 7d, e). These findings indicated that loss of astrocyte ALKBH5 function reduced dendritic and spinal loss of pyramidal neurons in response to social stress.

Next, to determine the neuronal activity under stress, we employed in vivo fiber photometry and the Ca²⁺ indicator, GCaMP6s. In order to do this, we injected the AAV-Syn-GCaMP6s virus into the mPFC, and implanted optical fibers above the infected cells (Fig. 7f, g). After two weeks, photometry recordings were performed in chronically stressed mice with the FIT (Fig. 7f). Before the CSDS experiment, heatmaps and maximum fluorescence change (Z-score) showed that both Astrocyte cKO and Ctrl mice demonstrated similar stimulus-evoked intracellular Ca²⁺ elevations while interacting with a CD1 aggressor in FIT (Fig. 7h, i). After 10 days of CSDS, GCaMP6s fluorescence was substantially reduced in the Ctrl mice during the contact period, and the Ca²⁺ activity in the Astrocyte cKO mice notably showed no significant change during CD1 exposure (Fig. 7i–k), indicating that astrocyte ALKBH5 deficiency maintained Ca²⁺ activity in the mPFC pyramidal neurons under social stress.

To explore astrocytic ALKBH5 modulate depression-related behaviors via glutamatergic neuronal activity, we bidirectionally injected the AAV-CamkIIα-hM4D(Gi)-mCherry virus into the mPFC of Astrocyte cKO mice to specifically manipulate the activity of glutamatergic neurons. Confocal images showed that hM4Di was expressed in glutamatergic neurons (Supplementary Fig.). Electrophysiologically, the CNO (1 μm) significantly inhibited the firing of hM4Di-expressing glutamatergic neurons (Supplementary Fig.). Behaviorally, intraperitoneal injection of CNO dramatically in Astrocyte cKO mice reversed the decrease of immobility of Astrocyte cKO mice in FST (Supplementary Fig.). No differences were observed in LD and EPM (Supplementary Fig.). These results suggest that the m6A modification of GLT-1 plays a critical role in regulating depression-related behaviors. 13a 13b 13c 13d, e

S-adenosylmethionine (SAMe), a universal methyl donor for most of the methylation reactions, is used to sustain RNA methylation (including m6A modification) via RNA methyltransferases and regulate RNA splicing, processing, translation, and decay (Fig. 8a)^4,32. To investigate the role of the SAMe in m6A modification dysfunction in depression, we first examined plasma SAMe levels in MDD patients. The results showed that the plasma SAMe levels were lower in the MDD patients compared with healthy controls (Fig. 8b). The plasma SAMe levels were also decreased in the LPS-induced depressed mice compared to the Ctrl mice (Fig. 8c). In addition, the SAMe concentrations in the mPFC were significantly decreased in both the LPS-induced depression model and the CSDS paradigm (Fig. 8d, e). Furthermore, we found that total m6A level and GLT-1 m6A level in the adult C57BL/6 J mice was increased after administration of SAMe (300 μg mL⁻¹, 7 d, i.p.) (Fig. 8f, g).

To test whether SAMe could induce an antidepressant-like effect, we employed the FST. It was found that the total duration of immobility in the adult C57BL/6J mice was decreased after administration of SAMe after 1 week or 3 weeks (300 μg mL⁻¹, 7 d, i.p.) (Fig. 8h, i). For CSDS experiments, the interaction time was increased in the SI test (Fig. 8j and Supplementary Fig. 14d); and less time was spent in the open arm in the EPM after SAMe treatment compared with the vehicle group (Fig. 8k). No difference in OFT and LD was observed between the SAMe and Vehicle groups (Supplementary Fig. 14a–c, e–g). In addition, we tested whether SAMe could reverse the depressive-like behaviors caused by the overexpressed ALKBH5 in the mPFC astrocytes (Fig. 8l). In FST, the immobility time caused by overexpressed ALKBH5 was reduced after SAMe treatment (Fig. 8m). After SSDS, SI test showed the OE-Alkbh5 mice spent less time on interaction zone with target while the OE-Alkbh5 mice administered with SAMe spent more time on the interaction zone with target (Fig. 8n). Taken together, these findings demonstrated that SAMe produces antidepressant-like effects through m6A methylation.

Thirty patients with MDD (19 women and 11 men) aged 24.4 ± 6.61 (SD) years and healthy controls (20 women and 18 men) aged 25.08 ± 2.53 (SD) years were recruited from the Department of Psychology and Behavior, Guangdong 999 Brain Hospital, Guangzhou, China. None of these patients with MDD or healthy controls had taken any psychotropic medications within 4 weeks. All of the patients were clinically diagnosed by at least two psychiatrists in accordance with the Diagnostic and Statistical Manual of Mental Disorders Fourth Edition (DSM-IV) criteria and the 17-item Hamilton Rating Scale for Depression (HAMD-17). Their clinical information is detailed in Supplementary Data. The study with the use of peripheral blood from MDD patients was approved by the Ethics Committee of Guangdong 999 Brain Hospital, Guangzhou (NO. 2021-01-087). The participants have signed the informed consent form. There were no sex- or gender-based analysis have been performed in our study. The research was performed in accordance with the Declaration of Helsinki (ethical principles for medical research involving human subjects). 1

Four to five C57BL/6J mice or CD1 mice were housed in an EVC cage (300 × 170 × 120 mm) at 23 ± 1 °C, humidity 40% under standard laboratory conditions with a 12 h light/dark cycle (lights on from 8:00 a.m. to 8:00 p.m.) and with free access to food and water. C57BL/6J mice (aged 8–12 weeks) were obtained from the Southern Medical University Animal Center (Guangzhou, China). The adult male CD1 mice (CD-1®(ICR)Mice, NO. VM0011) (older than 5 months of age) were purchased from Charles River Laboratories (Beijing). Before the behavioral tests, the C57BL/6 J mice were handled every day for 3 days, and double-blind behavioral experiments were performed between 1:00 p.m. and 5:00 p.m. All of the experiments were conducted in accordance with the Regulations for the Administration of Affairs Concerning Experimental Animals (China) and were approved by the Southern Medical University Animal Ethics Committee (NO. 2016104).

Alkbh5^loxP/loxP mice were purchased from the Shanghai Model Organisms Center (No. NM-CKO-190004). CaMKIIα-creER^T2 mice were purchased from Jackson Laboratories (Stock No. 012362). Fgfr3-iCreER^T2 mice (C57BL/6 J background) were generously provided by William D Richardson (University College London). Cdh5-CreER^T2 mice were purchased from Biocytegen (Stock No. 110140). Fgfr3-iCreER^T2; Alkbh5^loxP/loxP mice were generated by crossing the Alkbh5^loxP/loxP mice with the Fgfr3-iCreER^T2 mice and so it was with the CaMKIIα-creER^T2; Alkbh5^loxP/loxP mice by crossing the Alkbh5^loxP/loxP mice with the CaMKIIα-creER^T2 mice. Cdh5-CreER^T2; Alkbh5^loxP/loxP mice by crossing the Alkbh5^loxP/loxP mice with the Cdh5-CreER^T2 mice. To excise the loxP sites by Cre recombination, 2-month-old male mice were i.p. injected with TAM (Sigma-Aldrich: T5648) with 100 mg kg⁻¹ of body weight once a day for 5 consecutive days. TAM was dissolved in corn oil (Sigma-Aldrich: C8267) at a final concentration of 10 mg mL⁻¹. The littermate Alkbh5^loxP/loxP mice injected with TAM were used as the controls. Only male mice (8–13 weeks old) with normal appearance and weight were used for all tests. Behavioral tests were conducted 28 days after the first TAM injection by experimenters who were blinded to the experimental group.

Mice were anesthetized with pentobarbital sodium and perfused with ice-cold sterile PBS, followed by dissecting and coarsely chopping their brains. Then, the cells from the chopped brains were dissociated for 30 min at 37 °C in 5 mL EBSS solution containing 2 mg mL⁻¹ Papain (Sigma), 1 mg mL⁻¹l-cysteine (Sigma), 0.5 mg mL⁻¹ EDTA (Sigma), and 100 μg mL⁻¹ DNase I (Sigma) for incubation. In the middle of the incubation, the brain tissues were dissociated with an 800 μm pipette, with the cell pellet resuspended in 5 mL of a 26% Percoll solution and centrifuged at 900×g (4 °C) for 15 min to enrich the astrocytes. After the cell pellet was passed through a 40-μm mesh, the cells were incubated in FACS buffer with antibody for 30 min on the ice at 4 °C in the dark and FACS analysis. The cell surface markers were stained for 30 min at 4 °C with FcR blocking reagent (Miltenyi) and treated with ACSA-2-PE of the FACS antibodies (Miltenyi Biotec, 130-102-365). Data collection was performed on a BD LSRFortessa X-20, and subsequent analysis was completed using FlowJo software.

Mice were euthanized after CSDS, and the astrocytes were isolated. The astrocytes were lysed in RIPA with 1% protease inhibitor (PMSF). The samples were then mixed with Simple Western sample buffer and standards to a final concentration of 1 μg μL⁻¹. For protein detection, we used the Simple Western system (Protein Simple, Bio-Techne). Target proteins were identified with primary antibodies against ALKBH5 (Sigma: HPA007196, 1:200) and GAPDH (Proteintech: 60004, 1:500) and subsequent immunodetection using horseradish peroxidase (HRP)-conjugated secondary antibody and chemiluminescent substrate.

We isolated microglial and neuronal cellular populations utilizing MACS column-based protocols according to the manufacturer’s instructions (Miltenyi Biotec: 130-107-677). The cortical tissue was enzymatically and mechanically lysed at 37 °C for 35 min to obtain a cell suspension. Then, the suspension was passed through a 40-µm cell strainer to remove debris and ensure a single-cell suspension. After a 10 min centrifugation at 300×g, the pellet was gently resuspended with ice-cold PBS with 0.5% BSA buffer and incubated for 15 min at 4 °C with Myelin Removal Beads II (Miltenyi Biotec). After washing with PBS + 0.5% BSA and 10 min centrifugation at 300 × g, the pellet was resuspended with PBS + 0.5% BSA. The single-cell suspension was applied to a column placed on a magnetic stand (LS column and magnets from Miltenyi Biotec) to deplete the myelin fragments by magnetic separation and washed with PBS + 0.5% BSA. Finally, the cells were centrifuged for 5 min at 300 × g and the pellet was resuspended with 1 mL of PBS with 0.5% BSA. After obtaining a myelin-free single-cell suspension, the microglia were labeled with anti-CD11b immunomagnetic beads (Miltenyi Biotec: 130-093-634) and isolated using magnetic columns. To isolate the oligodendrocytes, the cell suspension depleted of microglia cells was incubated with anti-oligodendrocytes immunomagnetic beads (Miltenyi Biotec: 130-094-543).

The 2-month-old mice were anesthetized with sodium pentobarbital (50 mg kg⁻¹, i.p. injection) for unilateral and fixed in a stereotactic frame (RWD) before surgery. Then they were treated with a bilateral stereotaxic injection of virus into the mPFC (AP: +1.75; ML: ±0.3; DV: −2.7 mm relative to bregma; AP, ML, and DV denote anteroposterior, mediolateral and dorsoventral distances from bregma, respectively). The coordinates were measured from bregma according to the mouse atlas. The virus (300 nL) was injected into each location at a rate of 100 nL min⁻¹. After each injection, the needle was left in place for 6 min and then slowly withdrawn.

The virus purchased from Taitool Bioscience containing iCre under a gfaABC1D promoter (AAV2/9-gfaABC1D-eGFP-iCre-SV40pA) or control virus (AAV2/9-gfaABC1D-eGFP-bGHpA) was bilaterally injected into the mPFC of Alkbh5^loxP/loxP mice to specially knockdown ALKBH5 in the astrocytes.

Then, to overexpress ALKBH5 in the astrocytes, the virus (AAV2/8-nEf1α -DIO-ALKBH5-EGFP-WPRE-PA) or control virus (AAV2/8-nEf1α-DIO-EGFP- WPRE-PA) was bilaterally injected into the mPFC of Fgfr3-iCreER^T2 mice, followed by administering TAM at 5 days after virus injection.

The viruses AAV2/9-CaMKIIα-hM4D(Gi)-mCherry-WPRE-pA, AAV5-gfaABC1D-mCherry-5’miR-30a-shRNA(GLT-1)-3’miR-30a-WPREs, and AAV2/9-hSyn-eGFP-iCre-WPRE-pA were purchased from BrainVTA (Wuhan, China). The shRNA sequences targeting GLT-1 were GCTCTCACTGACTGTGTTT.

For sparse labeling of pyramidal neurons in the mPFC (AP: −1.75; ML: ±0.3; DV: −2.5 mm) of Fgfr3-iCreER^T2; Alkbh5^loxP/loxP cKO and control mice, the mice were bilaterally injected with a total 400 nL viral cocktail (1:1) of AAV2/9-CaMKIIα-FLP-WPRE-PA and AAV2/9-nEf1α-FDIO-EYFP-WPRE-PA. Two weeks after the injection, both Astrocyte cKO and Ctrl mice were subjected to 10 days of CSDS or non-CSDS. The brain sections were collected for confocal imaging. The basal dendritic complexity from the cell body to the surrounding 120 μm and the spine density of pyramidal neurons were analyzed with Imaris 8.0 by double-blind investigators.

Fgfr3-iCreER^T2; Alkbh5^loxP/loxP mice were bilaterally injected with hM4Di into the mPFC following TAM injection one week later. After 4 weeks of virus expression, mice were injected with clozapine-N-oxide (CNO; 2 mg kg⁻¹, i.p.; Sigma-Aldrich) and placed into the test room to habituate for 60 min. Stimulation-induced behaviors were scored manually by a human observer inspecting the videos post hoc (EthoVision by Noldus) while blinded to the underlying conditions.

Following injection of the virus (300 nL) (AAV2/9-iGluSnFR(A184S)-WPRE-hGHpA) or (AAV2/9-hSyn-GCaMp6s-WPRE-hGHpA) into the mPFC, a fiberoptic implant was advanced and secured at the same location of Fgfr3-iCreER^T2; Alkbh5^loxP/loxP mice.

To record fluorescence signals, a laser beam from a laser tube (488 nm) was reflected by a dichroic mirror focused by a 10× (NA of 0.3) lens and coupled to an optical commutator. Continuous video and fiber photometry acquisition were recorded during the FIT with an aggressive CD1 mouse. The fluorescence was bandpass filtered (MF525-39, Thorlabs) and collected by a photomultiplier tube (R3896, Hamamatsu). An amplifier (C7319, Hamamatsu) was used to convert the photomultiplier tube current output to voltage signals, which were further filtered through a low-pass filter (40 Hz cutoff; Brownlee 440). To minimize photobleaching, the laser power at the fiber tip was adjusted to 30 μW.

Bulk fluorescence signals were acquired and analyzed with MATLAB software. The z-score for a population of astrocytes was calculated using the following formula: z-score = (F_Signal − F_Basal)/STD (F_Basal).

Tissues were isolated from the cerebral cortex of postnatal day 1 mouse washed in ice-cold PBS and transferred to a 50 mL Falcon tube containing 0.5 mL of PBS for dissociation using a pair of sterile operating scissors. Then, they were incubated with 0.25% trypsin (Gibco) in 0.5 mM EDTA at 37 °C for 10 min. In the following step, culture medium was added to inhibit the reaction, and the cell suspension was transferred into 15 mL tubes and centrifuged at 900 × g for 6 min. The pellet was resuspended in 10 mL of culture medium (for neurons: Neurobasal Medium-A + 1% B27 + 1% Glutamax; for astrocytes, DMEM F12 + 10% FBS). Finally, the cells were placed in a culture flask at a density of 5 × 10⁶ cells per 5 mL and incubated in a humidified incubator containing 5% CO₂ air at 37 °C.

Cultures of mPFC glia and neurons were prepared from E18 mice following previously described methods. Briefly, cells were plated at a density of 60 × 10³ cells per mL on poly-l-lysine precoated coverslips. Cultures were kept in a neurobasal medium (Invitrogen) with 3% horse serum (Invitrogen) for several days before changing to serum-free neurobasal medium (Invitrogen). Cultures were maintained at 37 °C in 5% CO₂ for 20 d in vitro at maximum.

Cultured cells and mouse brain tissues were homogenized in lysis buffer (RIPA) in the presence of protease inhibitor (PMSF) on ice for 30 min and centrifuged at 10,000 × g for 20 min at 4 °C. Protein samples (30–60 μg) were separated using 10% SDS-PAGE gels and subsequently immunoblotted onto polyvinylidene difluoride (PVDF) membranes (Millipore) in ice-cold buffer (25 mM Tris HCl, 192 mM glycine, and 20% methanol) by electrotransfer for 2 h. The membranes were blocked with 5% nonfat milk powder dissolved in TBST and then incubated with the indicated antibodies at 4 °C overnight (ALKBH5, Sigma: HPA007196, 1:1000; GLT-1 Abcam: Ab205247↗, 1:1000; GAPDH, Proteintech: 60004, 1:5000). Horseradish peroxidase (HRP)-conjugated goat-anti-rabbit (Abclonal: AS014, 1:5000) and goat-anti-mouse Abclonal: AS003, 1:5000) secondary antibodies were incubated at room temperature for 1 h. Blots were detected using enhanced chemiluminescence (Pierce). Protein abundance was quantified by analyzing the Western blot bands using Image Lab (Bio-Rad) software. Quantified band intensities were normalized to GAPDH levels and averaged from at least three independent experiments.

The mice were anesthetized with sodium pentobarbital. The transverse mPFC was cut from each mouse brain. The tissues were washed in ice-cold PBS and then homogenized in fresh lysis buffer. The homogenates were centrifuged for 5 min at 10,000 × g. The supernates were immediately collected and assayed according to the manufacturer’s protocols. The SAMe ELISA kits were purchased from Dogesce (Human: DG94114Q; Mouse: DG91579Q).

For the m6A RNA methylation analysis, the SAMe ELISA kits were purchased from Epigentek (P-9005-96).

Animals were anesthetized and perfused with saline followed by 4% PFA in 0.1 M PBS, pH 7.4. The brains of the animals were removed, post-fixed overnight in 4% PFA at 4 °C and transferred to 30% sucrose in 0.1 M PBS, pH 7.4, followed by cutting the coronal sections (40 μm) on a freezing microtome (Leica CM3050 S). After being seeded on coverslips to the appropriate density, the cultured cells were fixed in 4% paraformaldehyde for 15 min and permeabilized in 0.5% Triton X-100 solution for 5 min at room temperature. Then, the cells and sections were washed with PBS three times and incubated first in a blocking buffer containing 3% bovine serum albumin in 0.2% Triton X-100/PBS for 2 h at room temperature and then with primary antibodies against ALKBH5 (1:300; Sigma: HPA007196), NeuN (1:300; Millipore: MAB377), Iba1 (1:500; Abcam: ab178846), Sox10 (1:300; Abcam: ab155279), and S100β (1:300; Abcam: ab52642) in blocking buffer overnight at 4 °C. After washing the treated cells and sections with PBS again for three times, they were further incubated with Alexa Fluor 488- or Alexa Fluor 568-conjugated secondary antibodies at room temperature for 2 h. Alexa fluor 488-anti-Mouse secondary antibody (Invitrogen, A32723, 1:500), Alexa fluor 568-anti-Mouse secondary antibody (Invitrogen, A11004, 1:500), Alexa fluor 488-anti-Rabbit secondary antibody (Invitrogen, A32731, 1:500), Alexa fluor 568-anti-Rabbit secondary antibody(Invitrogen, A11036, 1:500). The nuclei were counterstained with DAPI and the coverslips were mounted onto glass slides with anti-fade solution and visualized using a Nikon fluorescence microscope (Nikon Instruments Inc.).

The total RNA from the whole blood and brain tissues was extracted using TRIzol (Thermo Fisher Scientific) in line with the manufacturer’s instructions, and the RNA samples were stored at −80 °C. Quantitative PCR was performed by using the Transcriptor First Strand cDNA Synthesis kit (TAKARA) and TB Green® Premix Ex Taq™ II (TAKARA) on a LightCycler 96 Real-Time System (Roche). Differences in gene expression were calculated by the 2^−ΔΔCT method and were presented as the fold change. The relevant primers was provided in the Supplementary Data 5.

dCas9-m Cherry and GLT-1-targeting-sgRNA-puro-eGFP plasmids (BrainVTA) were constructed and transfected into primary cultured astrocytes to alter the 3’ UTR region 201 m6A modification site in GLT-1. Then, the astrocytes were transfected using TransIT-X2 (Mirus Bio) for 24 h at 37 °C according to the manufacturer’s protocol.

The SLC1A2-3’ UTR WT was PCR-amplified directly from mouse cDNA and then cloned into the 3’ end of the Renilla luciferase gene of the psiCHECK-2 dual-luciferase vector containing firefly luciferase driven by a different promoter as an internal control (Promega, Madison, WI, USA) via the Xho I and Not I sites. Mutant reporter constructs (SLC1A2-3’ UTR-A, SLC1A2-3’ UTR-C) were obtained from the SLC1A2-3’ UTR WT reporter using a QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). Primary astrocytes were cultured, and transfection using Lipofectamine 3000 reagents (Life Technologies), and dual-luciferase activity (Promega) was measured with the Wallac Victor V 1420 Multilabel Counter (PerkinElmer, San Jose, CA, USA) to yield the ratio of Renilla luciferase activity to firefly luciferase activity. All luciferase readings were taken from more than three individual biological repeats.

l-Glutamate (200 μM) (Sigma) was added, and glutamate uptake ability was determined by measuring glutamate levels in the medium after 4 h. The astrocytes treated with rLV-GFAP or Ctrl virus in each group were treated with 200 μM l-glutamate and incubated for 4 h at 37 °C. At each time point, the concentration of glutamate in the medium was measured with a glutamate assay kit (Abcam: ab83389) according to the manufacturer’s protocol, and the glutamate uptake ability was calculated by (Concentration _{4 h} - Concentration _{0 h})/ Concentration _{0 h} × 100%.

The mice were anesthetized with pentobarbital sodium. Their brains were quickly removed and cut into 300-μm horizontal slices containing the mPFC using a Leica LS1200s vibrating microtome. which were prepared in ice-cold oxygenated cutting artificial cerebrospinal fluid (aCSF) containing (in mM) 220 Sucrose, 2.5 KCl, 1.3 CaCl₂, 2.5 MgSO₄, 1 NaH₂PO₄, 26 NaHCO₃, and 10 Glucose. After that, the slices were transferred to a chamber with oxygenated recording aCSF containing (in mM) 126 NaCl, 26 NaHCO₃, 3.0 KCl, 1.2 NaH₂PO₄, 2.0 CaCl₂, 1.0 MgSO₄, and 10 Glucose and held at 32 °C for 30 min and then at room temperature for 1 h. The mPFC pyramidal neurons were viewed with a 40× water immersion objective (Zeiss). Recording electrodes with a resistance of 3–5 MΩ were pulled from borosilicate glass capillaries.

To record spontaneous excitatory postsynaptic current (sEPSC), the internal pipette solution for the recording electrodes contained (in mM) 125 cesium methanesulfonate, 5 CsCl, 10 HEPES, 0.2 EGTA, 1 MgCl₂, 4 Mg-ATP, 0.3 Na-GTP, 10 phosphocreatine and 5 QX314 (pH 7.40, 290 mOsm). Recordings were made with a HEKA EPC10 amplifier with the signals filtered at 5 kHz and digitized at 10 kHz. Finally, the sEPSCs with 20 μM bicuculine (Sigma) added to the cerebrospinal fluid in the voltage clamp (Vclamp = −70 mV) and the mEPSCs in the presence of 1 μM tetrodotoxin (Aladdin) and 20 μM bicuculine (Sigma) were recorded.

The action potentials were assessed under the current-clamp mode, with the pipette solution including (in mM): 125 K-gluconate, 5 KCl, 10 HEPES, 0.2 EGTA, 1 MgCl₂, 4 Mg-ATP, 0.3 Na-GTP, and 10 phosphocreatine (pH 7.40, 285 mOsm). A series of depolarizing currents (from 0 to 200 pA, at a step of 10 pA) were injected to induce action potentials.

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

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

The MeRIP-seq data generated in this study have been deposited in the Genome Sequence Archive under accession code CRA012006 (https://ngdc.cncb.ac.cn/gsa↗)^63,64. Previously described data was downloaded from the GEO database (GSE102556↗). Source data are provided with this paper.