Erasing m⁶A-dependent transcription signature of stress-sensitive genes triggers antidepressant actions

Abstract

Emerging evidence has shown that stress responsivity and psychiatric diseases are associated with alterations in N⁶-methyladenosine (m⁶A) mRNA epigenetic modifications. Fat mass and obesity-associated protein (FTO) is an m⁶A demethylase that has been linked to increased body mass and obesity. Here, we show that tricyclic antidepressants (TCAs) with weight-gain side effects, such as imipramine and amitriptyline, directly increased FTO expression and activated its epigenetic function in the ventral tegmental area (VTA). VTA-specific genetic disruption of FTO increased stress vulnerability and abolished the antidepressant activity of TCAs, whereas erasing m⁶A modification in the VTA by FTO overexpression or cycloleucine led to significant antidepressant activity. Mechanistically, both transcriptome sequencing and quantitative PCR revealed that overexpression of FTO in the VTA decreased the transcription of stress-related neuropeptides, such as cocaine- and amphetamine-regulated transcript peptide and urocortin, in the social defeat model, which was mimicked by imipramine, suggesting an m⁶A-dependent transcription mechanism of stress-related neuropeptides may underlie the responses to antidepressant. Collectively, our results demonstrate that inhibiting m⁶A-dependent transcription of stress-related genes may work as a novel antidepressant strategy and highlight a previously unrecognized activator of FTO-dependent epigenetic function that may be used for the treatment of other neurological diseases.

Highlights

• •TCAs erase m⁶A epigenetic modification by activating FTO.
• •FTO mediates the antidepressant activity of TCAs.
• •FTO in the VTA confers stress resistance.
• •FTO in the VTA limits m⁶A-dependent transcription of stress-sensitive genes.

1. Introduction

N⁶-methyladenosine (m⁶A), the most abundant and conserved internal transcript modification on eukaryotic messenger RNA (mRNA) molecules, plays important roles in various neurobiological processes (Chen et al., 2019; Edens et al., 2019; Merkurjev et al., 2018; Widagdo et al., 2016; Wu et al., 2019), such as learning and memory, synaptic transmission, neurogenesis and neural differentiation, by regulating mRNA transcript processing and translation. The formation of m⁶A is catalyzed by a methyltransferase complex (Methyltransferase-like 3, METTL3; methyltransferase-like 14, METTL14 and Wilms tumor 1-associated protein, WTAP) and erased by demethylases, including fat mass- and obesity-associated (FTO) protein and the α-ketoglutarate-dependent dioxygenase alkB homolog 5 (ALKBH5) protein (Jia et al., 2011; Liu et al., 2014; Ping et al., 2014; Wang et al., 2016; Zheng et al., 2013). Emerging evidence indicates that aberrant m⁶A may be considered a new layer of epigenetic mechanisms underlying the pathogenesis of psychiatric diseases, such as cocaine addiction (Hess et al., 2013), alcohol use disorder (Bohnsack et al., 2019) and Alzheimer's disease (Shafik et al., 2021). For major depressive disorder (MDD), a recent study demonstrated that the location and time of stress exposure alter m⁶A modifications (Engel et al., 2018). Another recent report revealed that m⁶A methylation of fatty acid amide hydrolase messenger RNA in the hippocampus (Hip) may be involved in the regulation of MDD (Huang et al., 2020). However, precisely how m⁶A methylation in different brain regions is involved in MDD is far from clear.

The discovery of FTO as the first m⁶A eraser establishes the concept of reversible RNA modification. FTO knockout mice exhibit neuropsychiatric phenotypes that include anxiety-like behaviors (Spychala and Ruther, 2019), memory deficits (Li et al., 2017a; Spychala and Ruther, 2019), abnormal fear coping (Engel et al., 2018), decreased brain size (Li et al., 2017a), reduced body weight (Li et al., 2017a), and impaired neurogenesis (Gao et al., 2020; Li et al., 2017a) and neurotrophic signaling (Spychala and Ruther, 2019). These lines of evidence point to a possible antidepressant role of FTO. Although some pharmacological inhibitors of FTO have been identified, including rhein (Chen et al., 2012), R-2-hydroxyglutarate (Su et al., 2018) and entacapone (Peng et al., 2019), very little is known about pharmacological activators of FTO. Before its role as a demethylase was confirmed, FTO originally attracted attention as an obesity- and diabetes-related protein due to the link between its genetic polymorphism and the body mass index of humans (Frayling et al., 2007). Interestingly, FTO overexpression leads to increased food intake (Church et al., 2010), which are often observed as side effects in patients taking antipsychotic medicines, such as tricyclic antidepressants (TCAs) (Berken et al., 1984; Salvi et al., 2018). Given the predicted association of FTO with antipsychotic-induced weight gain (Shing et al., 2014) and MDD-related brain functions, exploring its role in TCAs pharmacology is a promising research pursuit.

As the first class of antidepressants has been used clinically for over half a century, the real pharmacological mechanisms of TCAs are not fully understood. The previously held hypothesis that TCAs act by inhibiting monoamine uptake has been questioned because their rapid effect on monoamines contrasts with delayed antidepressant effects. The rise of several other theories in TCA pharmacology, including the neurogenesis hypothesis, the acid sphingomyelinase-ceramide hypothesis and the autophagy hypothesis, profoundly inspired research on the pathophysiological mechanism of MDD (Gulbins et al., 2013, 2018; Kornhuber et al., 2014; Santarelli et al., 2003). Here, we identified TCAs as previously unrecognized activators of FTO-dependent epigenetic function and highlighted FTO as a primary antidepressant target that reverses stress-induced transcriptome changes.

2. Material and methods

2.1. Animals and behavioral experiments

Adult male C57BL/6J mice (6–8 weeks old) were purchased from Hunan SJA Laboratory Animal (Changsha, Hunan, China). Male CD-1 retired breeder mice ranging from 8 to 10 months old were obtained from Vital River Laboratory Animal Technology Co. Ltd (Beijing, China) and individually housed. Animals were maintained under standard laboratory conditions with a 12-h light/dark cycle at a constant temperature (22 ± 2 °C) and humidity (40%–60%), and provided ad libitum access to food and water. All procedures were conducted following the Declaration of Helsinki and the Guide for Care and Use of Laboratory Animals as adopted and promulgated by the National Institutes of Health and approved by the Review Committee for the Use of Human or Animal Subjects of the Huazhong University of Science and Technology. The mice from all experimental groups were imaged in a random order. The required sample size was estimated based on prior experiments to satisfy the 3Rs (replacement, reduction, and refinement) principle of ethical use of experimental animals.

The procedure of chronic social defeat stress (CSDS) was performed as we previously reported (He et al., 2021; Li et al., 2018). Briefly, aggressive CD-1 mice were used as residents to defeat C57BL/6J mice (intruders) for 5–10 min per day over 10 consecutive days. C57BL/6J mouse were then housed with CD-1 mouse in a cage divided by a transparent clipboard. The shared home cage allowed sensory (visual, olfactory, auditory) contact between the C57BL/6J mouse and CD-1 mouse to enable continuous sensory defeat. After the 10 day, C57BL/6J mouse was returned to a single cage overnight and carry out social interaction testing (SIT) and a sucrose preference test (SPT) on the following day. For subthreshold social defeat stress (SSDS) (Li et al., 2017b), mice were exposed to three social defeat sessions: 5 min defeat bouts followed by a 15 min inter trial interval between the defeats.

SIT and SPT were performed to evaluate the depressive-like behavior after CSDS according to previous studies from our group and other labs (Friedman et al., 2016; He et al., 2021; Li et al., 2018; Yohn et al., 2019). After CSDS treatment, animals were subjected to SIT and SPT to identify the susceptible mice. Only the mice with both interaction ratio <1.0 in SIT and sucrose preference <75% in SPT were considered susceptible, and 14 days of vehicle or drug injection was indistinctly administered to susceptible mice. Immobility time in tail suspension test (TST) and forced swim test (FST) of normal animals are widely used models in antidepressants screening, and a previous study has indicated that social defeat stress exerted little effects on the TST and FST (Klinsey et al., 2007). Thus, TST, FST, open field test (OFT) and elevated plus maze (EPM) were performed in normal mice as we previously reported (He et al., 2021; Li et al., 2018). Detailed methods for behavioral experiments were available in Supplemental Information.

2.2. Cell culture and drug treatments

Mouse neuroblastoma 2A (N2a) cell line was obtained from the China Biological Part Stock Center (Beijing, China) and cryopreserved after subculturing different passages. N2a cells were routinely propagated using Dulbecco's Modified Eagle Medium (DMSO, Gibco Laboratories, Grand Island, NY, USA), supplemented with 10% fetal bovine serum (Gibco Laboratories, Grand Island, NY, USA) and 100 U/mL penicillin and 100 μg/mL streptomycin (Gibco Laboratories, Grand Island, NY, USA) in a humidified incubator of 5% CO₂ at 37 °C. Cultured cells were incubated for 48 h with pharmacological agents at the following concentrations: 0.1% DMSO (vehicle), 10 μM imipramine (IMI; Sigma-Aldrich, St. Louis, MO, USA), 10 μM amitriptyline (AMT; Aladdin, Shanghai, China), 10 μM mirtazapine (TCI, Tokyo, Japan), 10 μM fluoxetine (Sigma-Aldrich, St. Louis, MO, USA). Then, RNA was extracted from the N2a cells with Trizol (Takara Bio, Tokyo, Japan).

2.3. Tissue collection

Coronal brain slices (300 μm) containing the ventral tegmental area (VTA), medial prefrontal cortex (mPFC), Hip and nucleus accumbens (NAc) were cut by a vibratome (Leica, Wetzlar, Germany) in ice-cold artificial cerebrospinal fluid (ACSF) containing 119 mM NaCl, 1.3 mM MgSO₄, 3.5 mM KCl, 11.0 mM glucose, 26.2 mM NaHCO₃, 1.0 mM NaH₂PO₄ and 2.5 mM CaCl₂, pH 7.4, which was oxygenated with 95% O₂ + 5% CO₂. The certain brain subregions, including the mPFC, Hip and VTA, were dissected thoroughly under the dissecting microscope using a special blade in accordance with the histology atlas of Paxinos and Franklin (2001, The Mouse Brain in Stereotaxic Coordinates 2nd edn, San Diego, CA: Academic) to perform molecular experiments.

2.4. Stereotaxic surgery

After CSDS exposure, mice were subjected to SIT and SPT before the cannulation to identify the susceptible group, and only the mice with both interaction ratio <1.0 in SIT and sucrose preference <75% in SPT were considered susceptible and performed cannulation. Susceptible or control mice were anesthetized with an intraperitoneal injection (i.p.) of pentobarbital sodium (45 mg/kg; Sinopharm, Shanghai, China). Subsequently, mice were placed in a small-animal stereotaxic device, which was washed with swabs containing 75% ethanol. Surgical tools were sterilized with 75% ethanol. The skull was exposed and gently thinned using a dental drill around the VTA region. Twenty-two-gauge stainless steel guide cannulas (RWD, Shenzhen, China) were bilaterally implanted in the VTA region (AP = −3.3 mm, ML = 0.5 mm, DV = − 4.4 mm). The mice were allowed to recover for 1 week after surgery. Cycloleucine (40 μg/μL/side, Sinopharm, Shanghai, China), betaine (20 μg/μL/side, Sinopharm, Shanghai, China), rhein (1 μg/μL/side, Aladdin, Shanghai, China), recombinant urocortin (1 μg/μL/side, Cloud-Clone Crop, Wuhan, China) and recombinant cocaine and amphetamine regulated transcript (CART, 0.5 μg/μL/side, Immuno Clone Biosciences Co., Ltd, Huston, USA) were microinjected into the VTA through the cannulas with the microsyringe pump. Saline was used as the vehicle for IMI, AMT, mirtazapine and fluoxetine. The ACSF was used as the vehicle for cycloleucine, betaine, recombinant urocortin and recombinant CART. The ACSF containing 1‰ DMSO was used as the vehicle for rhein.

2.5. Genetic approaches

The third generation of self-inactivating lentivirus vector, which contains a CMV-driven enhanced green fluorescent protein (GFP) reporter and a U6 promoter upstream of cloning restriction sites (HpaI and XhoI) to allow the introduction of oligonucleotides encoding short hairpin RNAs (shRNAs), was purchased from Shanghai Genechem Co., Ltd. The targeted sequence of FTO (referred as “LV-shFTO”) was 5′-ATACAAACTTTGCACCGAT-3′, and for control scrambled shRNA (referred as “LV-shCON”) was 5′-TTCTCCGAACGTGTCACGT-3’. For in vivo lentiviral microinjections, lentiviral vector (2 μL) was bilaterally targeted to the VTA region at a rate of 0.1 μL/min followed by a 10 min rest period to prevent backflow. Mice were allowed to recover for 2–3 weeks before experiments.

For FTO overexpression, AAV-CMV-βGlobin-FTO-3Flag-SV40Poly A (referred as “AAV-FTO”) or AAV-CMV-βGlobin-MCS-EGFP-MCS- 3Flag-Poly A (referred as “AAV-GFP”) was purchased from Shanghai Genechem Co., Ltd. AAV-FTO (0.3 μL) was infused into the VTA at a rate of 0.05 μL/min followed by a 10 min rest period to prevent backflow. After 4 weeks, mice were sacrificed and GFP fluorescence in target brain regions were estimated to confirm the viral infection.

2.6. Immunofluorescence staining

To measure m⁶A level by immunofluorescent staining, N2a cells were seeded onto coverslips in a 24-well plate and administrated with IMI or AMT for 48 h. Then cells were washed in phosphate buffer saline, fixed using 4% paraformaldehyde for 15 min at room temperature and treated with 3% bovine serum albumin (BSA) and 0.1% Triton X-100 in phosphate buffer saline for 30 min. The coverslips were incubated with mouse anti-m⁶A antibody (202111, Synaptic Systems, Gottingen, Germany) at 1:250 dilution in 1% BSA overnight at 4 °C. Alexa-Fluor 594 (red, 1:1000; A21203, Invitrogen, California, USA) tagged anti-mouse monoclonal secondary antibody at 1:1000 dilution in 1% BSA was added for all fluorescent labeling. Images were acquired with a laser scanning microscope (FV1000, Olympus, Tokyo, Japan). A semi-quantitative analysis was conducted based on different staining intensities using Image J software. Images from confocal microscopes were acquired in a 16-bit setting. The level of m⁶A was calculated as staining intensity/staining area. Results were expressed as a relative intensity ratio and normalized to the control group.

2.7. m⁶A dot blot

The m⁶A dot blot assay was performed as previously described with some modifications (Li et al., 2017b). Briefly, the indicated amount of total RNA (400 ng, 200 ng, 100 ng and 50 ng) was denatured in 1-fold volume of RNA incubation buffer (22.5% formaldehyde solution in 10 × Saline Sodium Citrate buffer) 50 °C for 20 min. And then RNA samples were applied to nitrocellulose membrane (BS-NC-22, Biosharp, Hefei, China) with a Bio-Dot Apparatus (BioRad, California, USA). After UV crosslinking, membrane was stained with 0.02% methylene blue in 0.3 M sodium acetate. After washing with 1 × TBST buffer for three times (10 min every time), membranes were blocked with 5% BSA in TBST for 1 h at room temperature. After that, blots were incubated with anti-m⁶A antibody overnight at 4 °C. After rinsing in TBST, the blots were incubated for 1 h at room temperature with horseradish peroxidase-conjugated secondary antibodies (1:1000; Thermo Fisher Scientific, Massachusetts, USA). Finally, the immunoblot was completely immersed in enhanced chemiluminescence reagents (SuperSignal ECL kit; Thermo Fisher Scientific, Massachusetts, USA), and the blot reaction was revealed by MicroChemi (DNR Bio-Imaging Systems, Jerusalem, Israel).

2.8. Methylated RNA immunoprecipitation-quantitative PCR (meRIP-qPCR)

MeRIP assay was adjusted from reported protocol (Dominissini et al., 2013; Yue et al., 2015). Briefly, around 4–5 μg intact poly-A-purified RNA was fragmented into ∼100-nt using RNA Fragmentation Reagents (AM8740, Invitrogen, California, USA) under 70 °C for 10 min followed by ethanol precipitation and collection. The m⁶A-specific antibody was incubated with 60 μL dynabeads anti-mouse IgG (11033, Invitrogen, California, USA) in 0.5 mL immunoprecipitation buffer (10 mM sodium phosphate, 1M NaCl and 0.05% Triton-X) for overnight at 4 °C with rotation. Then the fragment mRNAs were incubated with antibody-beads mixture in 0.5 mL immunoprecipitation buffer for 4 h at 4 °C. Next, the mRNA-antibody-beads complex was eluted with elution buffer (5 mM Tris-HCL, pH 7.5, 1 mM EDTA, PH 8.0, 0.05% SDS, 20 mg/mL Proteinase K) at 50 °C for 1.5 h with vibrating. Then, supernatant was collected and precipitated with 1 μg glycogen, one-tenth volumes of 3 M sodium acetate in 2.5 vol of 100% ethanol at −80 °C overnight. Final mRNA sample was brought up in water and m⁶A enrichment was detected by qPCR analysis. RNA from MeRIP was quantified by 2^−ΔΔct against non-immunoprecipitated input RNA. Primers used for MeRIP-qPCR are as follows: Ucn: 5′-CAAGGCGTCTTCAGCCCGT-3′, 5′-CATGGTGCCGCAGGTTATCG-3’; Cartpt: 5′-GCGCTATGTTGCAGATCGAA-3′, 5′-CAGTCACACAGCTTCC CGAT-3’.

2.9. Statistical analysis

All data are expressed as means ± SEM. Statistical analysis was performed using GraphPad Prism 7.0. Sample sizes were according to those used in previous publications from our group and others reporting CSDS. All values use biological replicates and are indicated by group size (n) in figure legends or within bar graphs. Student's t-test was used between two-group analysis, and one-way or two-way ANOVA test followed by the Bonferroni's test was used for multi-group analysis. Assessments were considered significant with P < 0.05 and non-significant with P > 0.05. The detailed statistical information were provided in the Supplemental Information.

3. Results

3.1. TCAs erase m⁶A epigenetic modification via activating FTO function

To address the effect of antidepressants on m⁶A modification, in the first section, we incubated N2a neuronal cell line with IMI and AMT for 48 h. By immunofluorescence, we found that both IMI and AMT strongly reduced m⁶A level in N2a cells (Fig. 1A). By dot blotting, we found that at a therapeutic dosage, intraperitoneal injection (i.p.) of IMI (20 mg/kg, dissolved in saline) and AMT (10 mg/kg, dissolved in saline) for 14 days significantly decreased the global m⁶A levels in the VTA (Fig. 1B), but did not reduce m⁶A levels in other brain regions, such as Hip and mPFC (Figs. S1C and D). To elucidate whether IMI also demethylate m⁶A RNA in the VTA of stressed subjects, CSDS, a social defeat stress-induced depressive model, was used. In the CSDS model, 14 days of IMI was administered indistinctly to susceptible mice that subjected to CSDS, and the global m⁶A levels in the VTA were also decreased compared with that in the vehicle group (Fig. 1C). We tested the possible mechanisms underlay the effect of TCAs. We incubated N2a neuronal cell line with various antidepressants, including IMI, AMT, mirtazapine and fluoxetine, and qPCR analysis indicated that only IMI and AMT increased the mRNA level of FTO (Fig. S1A). Western blotting analysis revealed that IMI and AMT (Fig. 1D) increased, but fluoxetine reduced the FTO expression (Fig. S1B) in the N2a cells. Both IMI and AMT did not affect the expression of Mettl3, Mettl14 and ALKBH5 (Fig. 1E, F, G), highlighting the role of FTO in the epigenetic function of TCAs. Both IMI and AMT are TCAs, and the common tricyclic structure appears to give similar mechanism properties resembling IMI and AMT.

Fig. 1.

TCAs erase m⁶A epigenetic modification via activating FTO function. (A) Immunofluorescence imaging and quantitative analysis of m⁶A abundance in N2a cells treated with IMI (10 μM), AMT (10 μM) or vehicle for 48 h. Scale bar: 50 μm (n = 10–13 from 4 independent biological replicates, one-way ANOVA, Bonferroni's test). (B) Dot blot shows significant decrease in m⁶A levels in VTA of mice following 14 days treatment with IMI (20 mg/kg, i.p.) or AMT (10 mg/kg, i.p.). (C) 14 days of IMI injection indistinctly to susceptible mice subjected to CSDS and m⁶A levels in VTA was detected by dot blot. (D–G) IMI and AMT incubation increased FTO expression in the N2a cells (n = 10–12, one-way ANOVA, Bonferroni's test), but not affected the expression of Mettl3, Mettl14 and ALKBH5 (n = 4–8, one-way ANOVA, Bonferroni's test). (H) IMI or AMT significantly increase FTO protein in the VTA (n = 5–6 mice/group, one-way ANOVA, Bonferroni's test), but not in Hip (n = 4–6 mice/group, one-way ANOVA, Bonferroni's test) or mPFC (n = 5–6 mice/group, one-way ANOVA, Bonferroni's test). (I) IMI was administered indistinctly for 14 days to susceptible mice that subjected to CSDS, and FTO levels in the VTA were analyzed by western blots (n = 6–10 mice/group, one-way ANOVA, Bonferroni's test). (J) Genetically knockdown of FTO in the VTA abolished the effect of IMI on m⁶A RNA methylation. Data are presented as mean ± SEM. *P < 0.05, **P < 0.01. ANOVA, analysis of variance; N2a, neuroblastoma 2A; IMI, imipramine; AMT, amitriptyline; m⁶A, N6-methyladenosine; FTO, fat mass and obesity-associated protein; VTA, ventral tegmental area; Hip, hippocampus; mPFC, medial prefrontal cortex; LV, lentivirus; MB, methylene blue; CON, control; CSDS, chronic social defeat stress.

Then, we detected FTO levels in response to antidepressants in different tissues, including the VTA, Hip, mPFC, NAc and liver of mice. Both IMI and AMT increased FTO protein levels in the VTA but not in the Hip, mPFC, NAc, and liver (Fig. 1H, Figs. S2A and B). In the susceptible mice, after IMI treatment for 14 days, FTO also increased significantly in the VTA (Fig. 1I). Genetic disruption of FTO in the VTA by LV-shFTO (Figs. S3A and B) reduced the global m⁶A levels (Fig. S3C) and largely abolished the reducing effect of IMI on m⁶A epigenetic modification (Fig. 1J), suggesting a critical role of FTO in the epigenetic function of TCAs.

3.2. TCAs act by erasing m⁶A modification via an FTO-dependent manner

We proceeded to investigate whether m⁶A erasing effects and FTO mediates the antidepressant effects of TCAs. We pharmacologically mimicked the effect of m⁶A RNA demethylation using cycloleucine, an inhibitor of RNA methylation. Cycloleucine erased m⁶A (Fig. S1E) by reducing the cellular levels of S-adenosyl-l-methionine (SAM), the major methyl group donator for DNA and RNA methylation, via competitively inhibiting methionine adenosyltransferase (Jani et al., 2009; Lombardini et al., 1970) that catalyzes the production of SAM from methionine and ATP.

Immobility time in FST and TST of control animals are widely used in antidepressants screening. Microinjection of cycloleucine in the VTA decreased the immobility time of TST and FST (Fig. 2A), indicating that erasing m⁶A in the VTA produces antidepressant-like behaviors. Sweet-tasting solutions are a well characterized natural reward and most classic antidepressants did not affect natural reward in control animals, but only reversed stress-induced reduction in sucrose preference, which was considered to be a core anhedonia symptom of depression. After 10 consecutive days of social defeat, the susceptible mice were microinjected with cycloleucine or vehicle into the VTA, and it was shown that cycloleucine reversed CSDS-induced deficits in sucrose preference and social interactions (Fig. 2B).

Fig. 2.

TCAs act by erasing m⁶A modification via an FTO-dependent manner. (A, top) Schematic timeline of cycloleucine injection and behavioral tests. Bottom, reducing m⁶A RNA methylation in the VTA by microinjection of cycloleucine (40 μg/μL/side) decreased the immobility time of TST and FST in mice (n = 8–10 mice/group, Student's t-test). (B, top) Schematic timeline of CSDS, cycloleucine treatment and behavior tests. Animals were subjected to SIT and SPT before the cannulation to identify the susceptible group, and only the mice with both interaction ratio <1.0 in SIT and sucrose preference <75% in SPT were considered to be susceptible and given vehicle or cycloleucine. Bottom, cycloleucine increases social interaction ratio and sucrose preference in susceptible mice (n = 7–9 mice/group, two-way ANOVA, Bonferroni's test). (C) Intraperitoneal delivery of rhein (120 mg/kg) blocks the effect of IMI and AMT on immobility time of FST and TST (n = 11–12 mice/group, one-way ANOVA, Bonferroni's test). (D) Local bilateral infusion with rhein (1 μg/μL, 1μL/side) eliminated the antidepressant effect of IMI and AMT (n = 5–10 mice/group, two-way ANOVA, Bonferroni's test). (E) Genetically knockdown of FTO in the VTA abolished the effect of IMI on immobility time of TST and FST in mice (n = 5–7 mice/group, two-way ANOVA, Bonferroni's test). (F) IMI reverses the decrease in social interaction and sucrose preference of stressed mice, while FTO knockdown blocked the therapeutic effect of IMI (n = 8 mice/group, one-way ANOVA, Bonferroni's test). Data are presented as mean ± SEM. *P < 0.05, **P < 0.01. FST, forced swim test; TST, tail suspension test; SPT, sucrose preference test; SIT, social interaction test.

Next, we explored the functions of FTO in the effect of antidepressants. Intraperitoneal infusion of rhein (120 mg/kg, dissolved in 0.1% DMSO), an FTO inhibitor, 30 min before administration of IMI and AMT dissipated the antidepressant action of both IMI and AMT in mice (Fig. 2C). Intra-VTA infusion with rhein abolished the effect of IMI on immobility time of TST and FST in mice (Fig. 2D). Furthermore, genetically knockdown of FTO in the VTA abolished the effect of IMI on immobility time of TST and FST (Fig. 2E). After viral microinjection of GFP-tagging LV-shFTO or GFP-tagging LV-shCON into the VTA of mice, the mice were subjected to 10 consecutive days of social defeat stress, and the depressed subjects were treated with IMI or a vehicle for two additional weeks. We observed that IMI only reversed the decrease in social interaction and sucrose preference in the mice injected with LV-shCON in the VTA. In contrast, FTO genetic knockdown blocked the therapeutic effect of IMI (Fig. 2F). Thus, FTO in the VTA may play a crucial role in the pharmacology of TCAs.

3.3. FTO in the VTA exerts antidepressant-like action and mediates stress resilience

We sought to determine whether FTO emerges as an endogenous mechanism to mediate stress resilience. First, the effects of AAV-mediated FTO overexpression within the VTA on behavioral despair were measured by the immobility time in two behavioral tests, the TST and FST. Mice were microinjected AAV-FTO or AAV-GFP into the VTA (Fig. 3A). Four weeks later, a marked elevation in the FTO protein level was observed by Western blot (Fig. 3A), and the global m⁶A levels in the VTA significantly decreased in the FTO-overexpressing mice (Fig. S4A). FTO overexpression in the VTA significantly reduced the immobility time in the TST and FST models (Fig. 3B) and alleviated the anxious behaviors in mice, evaluated by the OFT (Fig. S4B) and EPM (Fig. 3B). Anorexia is a common symptom in depressed individuals, and we found that FTO overexpression in the VTA promoted 2 h and 24 h food intake in control mice with no change in body weight (Figs. S4C and D).

Fig. 3.

FTO in the VTA confers antidepressant activity and stress resilience. (A, top) Schematic timeline of adeno-associated virus injection and behavioral tests. Bottom, representative photomicrograph of injection sites (left) in the VTA. Scale bars, 100 μm. Western blot analysis of FTO protein level in VTA (right) four weeks after injection with AAV-GFP/AAV-FTO (n = 4–5 mice/group, Student's t-test). (B) AAV-FTO viral expression in the VTA decreased the immobility time of TST and FST, increased open arm time and distance of EPM (n = 10–13 mice/group, Student's t-test). (C) FTO mRNA expression in the VTA of control, susceptible and resilient mice (n = 7–8 mice/group, one-way ANOVA, Bonferroni's test). (D–F) Western blots analysis of FTO protein levels in the VTA, Hip and mPFC of control, susceptible and resilient mice (n = 4–12 mice/group, one-way ANOVA, Bonferroni's test). (G) LV-shFTO viral expression in the VTA decreased social interaction time with target and sucrose preference following SSDS (n = 11–12 mice/group, two-way ANOVA, Bonferroni's test). (H) AAV-FTO viral expression in the VTA increased sucrose preference and social interaction time with target in the stressed mice (n = 9–11 mice/group, two-way ANOVA, Bonferroni's test). Data are presented as mean ± SEM. *P < 0.05, **P < 0.01. AAV, adeno-associated virus; EPM, elevated plus maze; GFP, green fluorescent protein; SSDS, subthreshold social defeat stress.

To gain further insight into the role of FTO in the stress resilience, we measured FTO expression in the VTA, Hip and mPFC of mice after social defeat stress. In consistent with the results from susceptible animals in Fig. 1I, CSDS did not alter FTO expression in susceptible group. Interestingly, after CSDS, resilient mice expressed an even higher level of FTO in the VTA than susceptible mice and control mice (Fig. 3C and D), but not in the Hip and mPFC (Fig. 3E and F, Figs. S5A and B). These findings suggest that inducible expression of FTO in the VTA may contribute to stress resilience. After viral microinjection of LV-shFTO or LV-shCON into the VTA, the mice were subjected to a SSDS (three social defeat sessions: 5 min of physical defeat followed by 15 min with no physical defeat). We found that mice injected with LV-shFTO spent less time in the interaction zone and displayed a decreased sucrose preference over that of the LV-shCON mice (Fig. 3G), indicating that FTO may represent an endogenous resilience mechanism.

Then, we evaluated the effect of FTO on the CSDS mice. After 10 consecutive days of social defeat, the susceptible mice were microinjected with AAV-FTO or AAV-GFP into the VTA. FTO overexpression in the VTA significantly increased the time spent with the social target in the interaction zone, ameliorating the social avoidant behavior in stressed individuals (Fig. 3H). Similarly, FTO overexpression also reversed the CSDS-induced decrease in sucrose preference, a behavioral symptom that reflects anhedonia (Fig. 3H). Thus, FTO in the VTA may serve as a target that mediates antidepressant activity.

3.4. Overexpression of FTO oppositely regulates the stress-coupled modifications of transcripts in the VTA

A recent report indicated that stress induced the accumulation of m⁶A RNA in various brain areas, including the Hip, mPFC and NAc (Engel et al., 2018), but little is known about the VTA. We performed a dot blot analysis and revealed a significant increase in the m⁶A levels in the VTA of CSDS mice, but not in the Hip or mPFC (Fig. 4A). We hypothesized that the role of VTA m⁶A may underlie stress susceptibility. To test this idea, we utilized betaine, an activator of m⁶A RNA methylation, and sought to determine its effect on mice subjected to SSDS. Betaine facilitates m⁶A formation (Fig. S1F) by increasing the cellular levels of SAM, the major methyl group donator for DNA and RNA methylation, via serving as a substrate for the transmethylation of homocysteine (Lever and Slow, 2010; Narayan and Rottman, 1992). Analogously, we found that mice microinjected with betaine displayed a decreased sucrose preference (Fig. 4B) after SSDS and spent less time in the interaction zone (Fig. 4B).

Fig. 4.

FTO overexpression in the VTA reverses the altered transcriptional signature under social defeat stress. (A) Dot blot detects m⁶A levels in the VTA, Hip and mPFC of susceptible mice. (B) Locally bilateral infusion of betaine (20 μg/μL, 1 μL per side), an activator of m⁶A RNA methylation, decreases sucrose preference (n = 12–14 mice/group, Student's t-test) and social interaction time (n = 12–14 mice/group, one-way ANOVA, Bonferroni's test) of mice underwent subthreshold social defeat stress. (C, top) Heatmap of CSDS-induced expression changes in the VTA as compared to control mice. Bottom, heatmap of FTO overexpression-regulated expression changes in the VTA compared to mice injected with AAV-GFP. (D, top) Venn diagram depicts the overlap of different expression genes (DEGs) between downregulated in depressed mice and upregulated by FTO overexpression. Bottom, bubble chart displaying the top ten gene ontology categories with downregulated in depressed mice and upregulated by FTO overexpression (P < 0.05 or fold-change ≥ 2.0). (E, top) The Venn diagram showing common and specific DEGs upregulated in the susceptible mice and downregulated by FTO overexpression. Bottom, the top ten significantly enriched categories associated with upregulated in depressed mice and downregulated by FTO overexpression. (F) qPCR analysis of mRNAs in the VTA of stressed mice. (n = 7–8 mice/group, Student's t-test). (G) qPCR analysis of mRNAs in the VTA of susceptible mice injected with AAV-FTO. (n = 7–12 mice/group, Student's t-test). Data are presented as mean ± SEM. *P < 0.05, **P < 0.01. Ucn, urocotin; Cartpt cocaine- and amphetamine-regulated transcript prepropeptide; Cdh1, cadherin 1; Rgs14, regulator of G protein signaling 14; Htr1d, 5-hydroxytryptamine receptor 1D; Cox6a2, cytochrome c oxidase subunit 6A2; Marco, macrophage receptor with collagenous structure.

The m⁶A modification can alter mRNA expression via multiple mechanisms (Bartosovic et al., 2017; Berulava et al., 2013; Yue et al., 2015). We performed RNA-seq using RNA derived from the VTA of control and susceptible mice to test transcriptional signatures under social stress (Fig. 4C). Using fold-change as a measurement, 235 putative mRNAs were remarkably changed at least 2.0-fold in the mouse transcriptome (Table S1). We hypothesized that the stress-altered transcriptional signature can be oppositely regulated by FTO via its epigenetic role. We observed the transcriptional signature using RNAs derived from the VTA of mice overexpressing AAV-FTO or AAV-GFP, and RNA-seq results revealed that 2032 putative mRNAs were remarkably changed at least 2.0-fold in the mouse transcriptome (Fig. 4C, Table S2). Combining these two RNA-seq results, we found 163 genes in the VTA that were regulated by both FTO and social stress (Fig. S6A, Table S3). We found 23 genes (e.g., Phf19, Pyroxd2, etc. Fig. 4D, Fig. S6B), which were downregulated by social defeat stress but upregulated by FTO overexpression, and 19 genes (e.g., Cldn2, Ndnf, etc. Fig. 4E, Fig. S6C), which were up-regulated in susceptible mice but down-regulated by FTO overexpression. Gene Ontology (GO) analysis (Fig. 4D and E) showed that these different expression genes were highly enriched in pathways related to biological processes, and cellular components. Taken together, these results indicate that FTO may reverse the stress-coupled gene expression network, which may be central to its antidepressant activity.

We selected some candidate overlapping genes (e.g., Cdh1, Rgs14, Cox6a2, Ucn, Cartpt, Htr1d and Marco) that most significantly changed in the RNA-seq results of depressed mice as stress-sensitive genes (Fig. S6D) for further verification. Consistent with the RNA-seq data, qPCR analysis (Fig. 4F) found that Cartpt (associated with addiction, depression, obesity and insulin resistance), Htr1d (associated with type 2 diabetes, depression, schizophrenia, and anorexia) and Ucn (associated with anorexia, anxiety, depression, and alcohol drinking) mRNA levels were significantly increased in the VTA of stressed mice, and that Cdh1 (associated with nerve injury and memory), Cox6a2 (associated with obesity and insulin resistance), and Rgs14 (associated with fear memory) mRNA levels were markedly decreased. No significant changes in Marco mRNAs were observed (Fig. 4F). Overexpression of FTO oppositely regulates the stress-coupled transcripts changes in the VTA of stressed mice (Fig. 4G).

3.5. Erasing m⁶A-dependent methylation of mRNAs encoding stress-sensitive peptides may underlie the antidepressant activity of FTO and TCAs

Since FTO expression was promoted by TCAs, we next explored whether IMI could alter the expression of overlapping candidate genes. After CSDS, susceptible mice were treated by IMI or vehicle for 14 days, and the overlapping candidate genes was tested. After 14 days of IMI treatment, decreased levels of Cartpt and Ucn and increased levels of Cad1, Cox6a2 and Rgs14 in the VTA were observed (Fig. 5A), consistent with the results of FTO overexpression (Fig. 4G), suggesting that both FTO and antidepressants inversely regulated the social stress-related transcriptional profile in the VTA. Interestingly, after 14 days, only two candidate genes encoding stress-sensitive neuropeptides, Cartpt and Ucn, remained significantly altered in the vehicle group, which was reversed to normal only by IMI. Genetic inhibition of FTO in the VTA via LV-shFTO eliminated the effect of IMI on Ucn and Cartpt in the stressed mice (Fig. 5B). Then, we explored whether Ucn and Cartpt were associated with stress resilience. Local bilateral infusion of recombinant UCN (1 μg/μL, 1 μL per side) and CART (0.5 μg/μL, 1 μL per side) protein into the VTA reduced the time spent in the interaction zone and resulted in decreased sucrose preference (Fig. 5C) after SSDS. These lines highlight a critical role of stress-sensitive neuropeptides such as CART and UCN in the antidepressant activity of TCAs. Considering the important role of FTO in RNA demethylation, the m⁶A status was detected via MeRIP-qPCR (Fig. 5D). SRAMP, a sequence-based m⁶A modification site predictor (Zhou et al., 2016), revealed five m⁶A modification sites in the mRNA sequence of Ucn and three m⁶A modification sites in the mRNA sequence of Cartpt (Fig. 5E). Specific primers against m⁶A sites located in the CDS region were designed to amplify products. Increased m⁶A levels were found in fragmented RNA from the VTA of stressed mice (Fig. 5F). Overexpression of FTO alleviated m⁶A methylation of Ucn and Cartpt in the VTA of stressed mice (Fig. 5G). In summary, IMI may act as an antidepressant by activating FTO-dependent m⁶A demethylation to regulate the transcription of stress-sensitive genes such as Cartpt and Ucn (Fig. 5H).

Fig. 5.

Erasing m⁶A-dependent methylation of mRNAs that encoding stress-sensitive peptides underlies the antidepressant activity of FTO and TCAs. (A) qPCR analysis of overlapping candidate genes in the VTA of control and CSDS mice after chronic IMI treatment (n = 7–19 mice/group, one-way ANOVA, Bonferroni's test). (B) Genetically down-regulation of FTO in the VTA of stressed mice eliminate the changes in Ucn and Cartpt mRNA levels induce by IMI (n = 5–9 mice/group, two-way ANOVA, Bonferroni's test). (C) Schematic timeline (left) and behavioral consequences of UCN and CART recombinant protein in the SIT (middle), SPT (right) after subthreshold social defeat stress (n = 6–8 mice/group, one-way ANOVA, Bonferroni's test). (D) A Flowchart for MeRIP-qPCR. (E) Predictive m⁶A sites of Ucn and Cartpt mRNA in SRAMP. (F) The m⁶A status of Ucn and Cartpt increased in the stressed mice (n = 5–6 mice/group, Student's t-test). (G) MeRIP-qPCR showed a decrease in m⁶A modification of Ucn and Cartpt in the stressed mice with FTO overexpression in the VTA (n = 4–7 mice/group, Student's t-test). (H) Schematic diagrams of tricyclics produce an antidepressant activity via initiating an FTO-dependent regulation of stress-related m⁶A modification. Data are presented as mean ± SEM. *P < 0.05, **P < 0.01. MeRIP, methylated RNA immunoprecipitation.

4. Discussion

FTO is traditionally thought to work in epigenetic regulation as a vital nucleic acid demethylase. Here, we found a novel and unexpectedly beneficial function of FTO under social stress. Another surprising finding is that a previously unrecognized link between m⁶A modification and antidepressant pharmacology has been revealed. Tricyclics may produce antidepressant activity via initiating an FTO-dependent regulation of stress-related m⁶A modification, followed by a reversal of a stress-coupled transcriptional signature in the VTA (Fig. 5H). Thus, our findings shed new light on conceptually novel antidepressants that function through the potentiation of FTO function.

The evidence that fostered the general TCA pharmacology mainly originated from the monoamine hypothesis. However, the monoamine hypothesis has been questioned because the antidepressant effect of TCAs is not clearly associated with the monoaminergic effect. The investigation of TCA pharmacology profoundly inspired and supported the findings of MDD mechanisms, such as the dysfunction in neurogenesis and the activation of acid sphingomyelinase (Gulbins et al., 2013; Kornhuber et al., 2014; Santarelli et al., 2003; Warner-Schmidt and Duman, 2006). Here, we report that the FTO-mediated demethylation of m⁶A emerged as a novel target for TCAs. TCAs increased the expression of FTO both in vitro and in vivo, and a disruption of FTO by pharmacological approaches or VTA-specific genetic approaches led to a significant abolishment of the antidepressant effect of TCAs. The increase of FTO expression occurred after TCA treatment for two weeks, which was in consistent with the antidepressant effect delay. The promoting effect on FTO expression seems to be closely related to the structure and general pharmacological properties of TCAs. Interestingly, weight gain is a common adverse effect of TCAs (Berken et al., 1984; Salvi et al., 2018; Serretti and Porcelli, 2018). FTO is an obesity- and diabetes-related protein owing to the linkage between its genetic polymorphism and body mass index (BMI). Previous reports and our results have demonstrated that FTO overexpression leads to increased food intake (Church et al., 2010), which may result from the alterations of dopamine D2-like receptor function to influence reward-related food intake (Hess et al., 2013) and FTO-associated increase in body mass may result primarily from increased food intake. More recent studies have indicated that FTO may facilitate weight gain by shifting the endocrine balance from the satiety hormone leptin toward the hunger-promoting hormone ghrelin (Benedict et al., 2014; Karra et al., 2013). Our study has revealed that TCAs increased FTO expression in the VTA, a central brain region for the control of food-related processes, and raised a previously unrecognized link between the weight effect and antidepressant pharmacology. Thus, an activation of FTO function may represent a previously unrecognized link between the weight effect and antidepressant pharmacology. Given the predicted association of FTO with antipsychotic-induced weight gain (Shing et al., 2014) and our findings about FTO-mediated antidepressant effect, FTO may be critical to determine whether antidepressant exerts therapeutic effect or adverse effect. Blockade of FTO downstream genes that participate in weight gain but not in stress resistance possibly alleviate the drug-induced weight gain, which may confer a new strategy for much safer treatment for depression in the future. Our findings raise novel insights into the TCAs pharmacology.

The present study highlights a novel neurobiological function of FTO, the RNA demethylase that has attracted much attention in recent years. An increasing number of studies have demonstrated that FTO plays vital roles in many biological processes, ranging from brain development (Ma et al., 2018) and cell metabolism (Merkestein et al., 2015; Speakman, 2015) to tumor procession (Li et al., 2017b). We reported that FTO-dependent m⁶A demethylation determines the vulnerability to social stress. These results, along with recent observations that FTO controls neurogenesis (Li et al., 2017a) and regulates memory formation (Spychala and Ruther, 2019), raised the possibility that FTO may serve as a target for developing novel rapid antidepressants. In our study, virus-mediated overexpression was used to investigate the action of FTO, and we found that overexpression of FTO in the VTA significantly reduced the immobility time in the TST and FST, alleviated the anxious behaviors in mice (Fig. 3B) and social avoidant behavior in stressed individuals (Fig. 3H), and promoted food intake (Figs. S4C and D). Virus-mediated overexpression of FTO increased FTO level to 6.83 ± 0.25 fold in the VTA, and FTO expression was increased to 1.33 ± 0.06 fold in the VTA in resilient mice. Meanwhile, IMI treatment increased FTO expression to 1.44 ± 0.05 fold in the VTA. Although stress resilience and TCAs could not completely mimic the effect of virus-mediated overexpression, they both produced some similar behavioral effects, such as improved social interactions or reduced behavioral despair.

It should be noted that FTO may affect various brain functions, thus, identification of FTO downstream genes to mediate antidepressant effects may bring novel targets for development of antidepressants. We observed that an overexpression of FTO altered the expression of more than 2195 genes. GO analysis implied that numerous genes regulate cell proliferation, differentiation, or metabolism, among which the expression of 398 RNAs were under tight regulation by social stress. Thus, FTO may work via oppositely regulation of the stress-coupled transcriptional signature in the VTA. In Fig. 4F, qPCR analysis found that Cartpt, Htr1d, Ucn, Cdh1, Cox6a2, and Rgs14 mRNA levels were markedly altered by CSDS. However, in Fig. 5A, only Cartpt and Ucn displayed significant alterations in the vehicle group. The discrepancies between Fig. 4, Fig. 5G may be resulted from that, animals in Fig. 4F were sacrificed after CSDS, but animals in Fig. 5A were sacrificed after a two-week treatment of vehicle or IMI following CSDS, which may buffer some CSDS-induced alterations of gene expression. Additionally, injections of vehicle or IMI for two weeks induced some levels of other stress. Thus, only genes that were more strongly dysregulated by social defeat stress, such as Cartpt and Ucn, were detected in the experimental conditions of Fig. 5G. The peptide encoded by Cartpt has been revealed as a candidate biomarker for MDD because of its effects on emotion and distribution covering brain areas involved in the mood regulation, via interactions with neurotransmitter receptors, such as dopamine receptors, glutamate receptors and γ-aminobutyric acid receptors (Ahmadian-Moghadam et al., 2018; Upadhya et al., 2012; Wiehager et al., 2009), which may bridge the interaction between the stress-related events and the neurobehavioral response (Chaki et al., 2003; Walker et al., 2021). Urocortin peptides, including Ucn1, Ucn 2 and Ucn3, acts on the corticotropin-releasing factor receptors to elicit stress responses (Ma et al., 2020), modulate social behavior (Deussing and Chen, 2018) and controll stress-related disorders (Romero-Leguizamon and Kohlmeier, 2020; Wang et al., 2007), including anxiety and depression. As a member of the corticotropin-releasing factor (CRF) family, Ucn1 acts as a CRF1 receptor agonist to elicit anxiety- and depression-like behaviors via activating HPA axis (Kormos and Gaszner, 2013; Tsatsanis et al., 2007). However, until now, the precise mechanisms underlying stress responses regulation by Ucn and Cartpt remain unclear. FTO overexpression inhibited CSDS-triggered transcription of Cartpt and Ucn, which may underlie the pathophysiology of depression. TCAs reduced CSDS-triggered transcription of Cartpt and Ucn, which may be involved in their pharmacology.

Epigenetic changes that cause modifications of a gene's functional state without changing its coding sequence have emerged as possible pivotal regulatory mechanisms in the pathogenesis of psychiatric diseases (Hess et al., 2013; Zheng et al., 2013), and psychoactive drugs may act by impacting epigenetic machinery, such as modulating stress-induced DNA methylation and histone deacetylation (Bohnsack et al., 2019; Shafik et al., 2021). Many studies have revealed that changes in DNA methylation, histone modifications and noncoding RNAs contribute to stress-induced depression-like behavior (Lin and Gregory, 2014; Yue et al., 2015; Zhao et al., 2014). As a novel epigenetic machinery, m⁶A has been shown to regulate gene expression by modulating pre-mRNA splicing, mRNA stability, RNA structure, and pre-miRNA processing (Lopez et al., 2014; Sales et al., 2011; Tadic et al., 2014). Here, we found that stress-coupled m⁶A modifications of transcripts in the VTA altered the expression of Cartpt and Ucn. Reversal of stress-induced m⁶A modification by FTO replenished the alterations in the transcript network. Our study successfully showed in vivo evidence of m⁶A writing/erasing in CNS diseases.

It should be noted that several problems needed to be resolved in the next study. In the present study, the mechanisms underlying stress resilience-coupled FTO expression in the VTA remains unknown. Considering that dopamine neurons in the VTA are sensitive to stress, which mediates stress susceptibility and resilience, and a recent study (Hu et al., 2020) has reported that glucocorticoid receptor-dependent FTO transactivation induces lipid accumulation in hepatocytes, we hypothesized that a glucocorticoid-regulated FTO expression in the VTA may contribute to stress resilience. Further investigation is required. Although FTO may reverse the social stress-coupled modifications of candidate gene transcripts in the VTA, much less is known about the precise mechanism. For instance, FTO down-regulated the m⁶A methylation of Carpt and Ucn, but decreased their expression. m⁶A modifications altered gene expression via multiple mechanisms. A m⁶A-IP-seq should be performed to confirm these alterations and clarify the mechanism in the future. We speculated that this was probably due to the multiple functions of m⁶A modification in RNA splicing, nuclear export, and protein translation. Although stress-related peptides including Ucn and Cartpt were important downstream genes of FTO-mediated antidepressant effects, RNA-seq data and qPCR analysis revealed other stress-related genes that regulated by FTO, such as Htr1d and Rgs14. Htr1d control serotonin release, and its alterations may contribute to the increased synaptic serotonin levels underlying depression and antidepressant response (Honda et al., 2004; Whale et al., 2001). Rgs14 is a multifunctional signaling protein that regulates postsynaptic plasticity in neurons to affect memory, emotion, and stimulus-induced behaviors (Harbin et al., 2021; Lee et al., 2010). These altered genes may also be involved in the pathogenesis of depression.

In conclusion, we discovered that targeting stress-coupled m⁶A mRNA epigenetic modifications to reverse transcriptional signature may work as a novel antidepressant strategy. Furthermore, our results suggest that FTO and its downstream genes are molecular targets for the development of novel antidepressants.

Declaration of competing interest

The authors declare no competing financial interests.

CRediT authorship contribution statement

Peng-Fei Wu: Conceptualization, Methodology, Data curation, Formal analysis, Writing – original draft, Funding acquisition. Qian-Qian Han: Methodology, Investigation, Writing – original draft. Fu-Feng Chen: Investigation, Methodology, Validation. Tian-Tian Shen: Investigation, Methodology, Validation. Yi-Heng Li: Investigation, Methodology, Validation. Yu Cao: Investigation, Methodology, Validation. Jian-Guo Chen: Writing – review & editing, Project administration, Supervision, Funding acquisition. Fang Wang: Writing – review & editing, Project administration, Supervision, Funding acquisition.

Appendix A. Supplementary data

The following are the Supplementary data to this article: