3T3-L1 preadipocytes were cultured in the Dulbecco's Modified Eagle Medium (DMEM) (Gibco 11995073, Carlsbad, CA) with 10% newborn calf serum (Gibco 26010074, Carlsbad, CA) and 1% penicillin-streptomycin (PS) (Gibco 15140163, Carlsbad, CA). About 2 days after confluency, cells were fed with MDI induction media containing DMEM, 10% fetal bovine serum (FBS) (Gibco 10099141C, Carlsbad, CA), 1% PS, 0.5 mM IBMX (Sigma I7018, St. Louis, Missouri), 1 μM dexamethasone (Sigma D4902, St. Louis, Missouri), and 10 μg/ml insulin (Sigma I0516, St. Louis, Missouri) for 3 days. Then, cells were replenished with maintenance media supplied with 10% FBS, 1% PS, and 10 μg/ml insulin every 2–3 days until the fully differentiated adipocyte-like cells were obtained. 20 μM rhein treatment was also carried out in this study.

Mature adipocytes were rinsed with phosphate-buffered saline (PBS) and fixed with 10% formalin at room temperature for 1 h. The fixed cells were then rinsed with PBS and equilibrated with 60% isopropanol diluted in double-distilled water (ddH₂O). After staining with the ORO working solution at room temperature for 10 min, cells were washed with ddH₂O until the supernatant was clear. The stained cells were photographed by using a microscope (Nikon, Tokyo). ImageJ software was applied to analyze the relative lipid accumulation.

Lentiviral shRNA templates were cloned into shRNA expression vector termed as the DECIPHER pRSI9-U6-(sh)-UbiC-TagRFP-2A-Puro (Cellecta, California, USA). shRNA lentivirus was produced by using Lenti-X 293T cells and harvested 48 h after transfection. 3T3-L1 cell lines were infected by the filtered viral media for 48 h with 10 μg/ml polybrene (Sigma, St. Louis, Missouri). The complete growth media containing 2 μg/ml puromycin were used to select cells. Lentiviral shRNA targeting sequences are listed below:

Total RNA was extracted by using the TRIzol reagent (Invitrogen 15596018, Waltham, Massachusetts) and the first strand complementary DNA (cDNA) was synthesized via the HiScript III 1st Strand cDNA Synthesis Kit (Vazyme R312-02, Nanjing, Jiangsu). qPCR analysis was carried out by using the ChamQ SYBR Green qPCR Master Mix (Vazyme A331-02, Nanjing, Jiangsu) by using the QuantStudio^TM 3 Real-Time PCR Instrument (Applied Biosystems, Waltham, Massachusetts). Primers used for amplifying target genes were presented in Supplementary Table 1.

Cells were rinsed with PBS and lysed in sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 25% glycerol, 5% β-mercaptoethanol, and 0.01% bromophenol blue) followed by heating at 100°C for 8 min. Then, samples were loaded on SDS-PAGE gels and separated by electrophoresis before being transferred to polyvinylidene difluoride (PVDF) membranes (Millipore IPVH00010, Burlington, Massachusetts). The membranes were first blocked by 5% non-fat milk diluted in 0.1% Tris buffered saline, 0.1% Tween (TBST) (1 × TBS, 0.1% Tween-20) for 1 h at room temperature followed by incubating with desired primary antibodies diluting in 5% non-fat milk at 4°C overnight. After that, membranes were rinsed three times with 0.1% TBST and incubated with corresponding horseradish peroxidase (HRP)-conjugated secondary antibodies for 1 h at room temperature. The primary antibody signals were visualized after reaction with enhanced chemiluminescence (Vazyme E412-02, Nanjing, Jiangsu). The primary antibodies used in western blot analysis were presented below: anti-β-actin (Proteintech 60008-1-Ig, Rosemont, Illinois), antiperoxisome proliferator-activated receptor γ (PPARγ) (Proteintech 16643-1-AP, Rosemont, Illinois), antifatty acid-binding protein 4 (FABP4) (Santa Cruz sc-271529, Santa Cruz, CA), anti-FTO (Abcam ab92821, Cambridge, England), anti-ALKBH5 (Proteintech 16837-1-AP, Rosemont, Illinois), anticyclin A (Proteintech 18202-1-AP, Rosemont, Illinois), and anticyclin B1 (Proteintech 55004-1-AP, Rosemont, Illinois).

Total RNA was extracted by using the TRIzol Reagent first and mRNA was isolated by using the Oligo d(T)25 Magnetic Beads (NEB S1419S, Ipswich, Massachusetts). A 2-fold serial dilution starting from 500 ng of mRNA was spotted on a nylon membrane and crosslinked under 254 nm UV light (0.12 J/cm²). Subsequently, the membrane was blocked by 5% non-fat milk diluted in 0.1% PBST (1 × PBS, 0.1% Tween-20) for 1 h at room temperature followed by incubating with anti-m⁶A antibody (Abcam ab151230, Cambridge, England) 1:1000 dilution at 4°C overnight. After that, the membrane was rinsed three times with 0.1% PBST and incubated with HRP-conjugated antirabbit immunoglobulin G (IgG) for 1 h at room temperature. The m⁶A blotting signal was visualized after reaction with enhanced chemiluminescence.

For RNA-seq and m⁶A RNA immunoprecipitation followed by high-throughput sequencing (MeRIP-seq), cells treated with rhein were collected at day 0 (preadipocytes), day 3 (MCE), and day 10 (differentiation), respectively. Total RNA was extracted by using the TRIzol Reagent followed by chemical fragmentation by using the freshly prepared fragmentation buffer as previously described approaches (32–34). A portion of 5 μg untreated fragmented RNA was served as input control for RNA-seq and 600 μg fragmented RNA was subjected to MeRIP-seq. For MeRIP-seq, the fragmented RNA was incubated with 12 μg anti-m⁶A antibody (Abcam ab151230) in 1 × immunoprecipitation (IP) buffer [10 mM Tris-HCl, pH 7.4, 150 mM sodium chloride (NaCl), 0.1% NP-40, and 40 U/μl RNase inhibitor] with continuous rotation for 2 h at 4°C. In the meanwhile, 15 μl protein A and 15 μl protein G mixed beads were blocked in 1 × IP buffer supplemented with bovine serum albumin (BSA) (0.5 mg/ml). After that, the mixtures were incubated with the prior blocked beads on a rotating wheel for another 2 h at 4°C. Following that, the reaction mixtures were spinned down and sediments were eluted twice in 100 μl of elution buffer with 6.7 mM m⁶A 5′-monophosphate sodium salt (Santa Cruz sc-215524, Santa Cruz, CA) in 1 × IP buffer followed by ethanol precipitation. The input fragmented RNA and precipitated m⁶A elute were used for cDNA library construction and deep sequencing described below.

The desired m⁶A-enriched RNA elute and input control were used to construct cDNA library in parallel as described in our previous studies (35, 36). Briefly, samples were first dephosphorylated and 40–60 nt region was selected according to 15% polyacrylamide Tris-borate-EDTA (TBE)-urea gel separation. Desired RNA species were excised and soaked overnight in gel elution buffer [300 mM sodium acetate (NaOAc), pH 5.5, 1 mM ethylenediaminetetraacetic acid (EDTA), and 0.1 U/ml SUPERase_In) followed by ethanol precipitation. Then, the purified RNA was processed for poly-(A) tailing reaction and reverse transcription afterward. Eventually, PCR amplification was proceeded by using the Phusion High-Fidelity DNA Polymerase (NEB M0530L, Ipswich, Massachusetts). The quantification of PCR products was evaluated on the Agilent 2100 Bioanalyzer System (Santa Clara, CA) before being subjected to cluster generation followed by deep sequencing on an Illumina Novaseq platform (San Diego, CA).

The Trim Galore (v0.6.5) was initially used to trim the low-quality sequences and retain paired reads under—length 100—paired parameters. The paired-end reads were aligned to the mouse genome (GRCm38) with the HISAT2 aligner (v2.1.0) with default parameters and Ensembl genes (GRCm38.102) transcriptome annotation (37). Then, gene expression was quantified by using the StringTie (v2.1.4) (38) and differentially expressed genes were identified by using the DESeq2 (v1.30.1) (39). The fold change > 2 or <0.5 and p < 0.05 genes were considered as differentially expressed genes. The transcripts per kilobase of exon model per million mapped reads (TPM) was used to produce heatmap.

Quality control of MeRIP-seq data was performed by using the FastQC (v.0.11.9). Low-quality bases and adaptor contamination were removed by the Trim Galore. Remaining reads were aligned to mouse genome version mm10 with the HISAT2 aligner with default parameters. The mapped reads were filtered with the Picard (v2.23.5) to remove PCR duplicates and the SAMtools (v1.9) to remove the reads with mapping quality below 30. m⁶A peaks were identified by using the exomePeak2R package with default parameters. Peaks were annotated by the R package ChIPseeker (v.1.26) (40). The distribution of m⁶A peaks was plotted by the R package Guitar (v.2.6.0). Motif finding was performed by using the HOMER (v.4.11) findMotifs.pl with parameter (-rna -p 10 -len 5, 6, 7, 8 -S 10). The final bam files were transformed to bigwig format by using the deepTools (v3.5.0) with parameter [–normalizeUsing Bins Per Million (BPM)]. The Integrative Genomics Viewer (IGV) (v2.94) was used to visualize the aligned data.

The Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were done on the DAVID (https://david.ncifcrf.gov) with default parameters. The gene set enrichment analysis (GSEA) software (v4.1.0) was used for the analysis of gene expression data with (c5.go.bp.v7.3.symbols.gmt) gene set with default settings (41).

Statistical analysis was carried out by using the GraphPad Prism. Unless otherwise specified, all the values were conducted as the mean ± SEM. Differential analysis was calculated by the Student's t-test. The data were considered as statistically significant when ^*p < 0.05, ^**p < 0.01, or ^***p < 0.001.

The raw and processed high-throughput sequencing data in this study are publicly available in the Gene Expression Omnibus (GEO) (reference numbers GSE171785 and GSE173702).

Rhein has been suggested to regulate adipogenesis and mice body weight in the previous studies. To systematically evaluate the effect of rhein on adipocyte differentiation, we induced adipogenesis of growth-arrested 3T3-L1 preadipocytes with the MDI medium in the presence or absence of rhein. Rhein treatment was carried out at the different stages including the MCE stage (rhein T1), the differentiation stage (rhein T2), and continuous treatment at both the stages (rhein T3) (Figure 1A). Remarkably, as shown in ORO staining results, rhein treatment at the early stage effectively inhibited the mature adipocytes formation, as lipid droplets accumulation in the mature adipocytes reduced 80% in the rhein T1 group compared to the dimethyl sulfoxide (DMSO) control, whereas supplement of the phytochemical at the later stage (rhein T2 group) could not attenuate the subsequent adipocytes formation (Figures 1B,C). Consistent with this, the expressions of the master regulator PPARγ in adipogenesis were significantly decreased in the rhein T1 group compared to the rhein T2 group. Spontaneously, three major downstream target genes of Pparγ, Fabp4, and Adiponectin (Adipoq) were also responsively downregulated (Figures 1D–F). To the best of our knowledge, the prolonged treatment of rhein at both the MCE stage and the differentiation stage had negligible influence on the adipogenic progression (Supplementary Figures 1A–C), indicating that the MCE stage is the crucial period for rhein to exert its detrimental effects on adipocytes development. Therefore, we conducted the rhein treatment at the MCE stage in the subsequent experiments.

Next, we investigated whether the different doses of rhein could similarly influence the progression of preadipocyte differentiation. As shown in Figures 1G–I, the newly formed adipocytes coupled with the mRNA and protein expressions of three marker genes (Pparγ, Fabp, and Adipoq) were gradually decreased with increasing concentration of rhein. All the results above suggest that rhein modulates the adipogenesis process in stage-specific and dose-dependent manners.

Rhein has been identified to inhibit FTO demethylase activity. To investigate whether rhein suppresses the adipogenic process through targeting FTO, we initially verified the functional role of FTO in adipogenesis. As examined in ORO staining, knockdown of Fto resulted in a substantial decrease in the accumulation of the lipid droplets of the mature adipocytes (Supplementary Figure 2A). Meanwhile, the RNA levels of adipogenesis marker genes Fabp4 and Adipoq were significantly downregulated upon FTO deficiency (Supplementary Figures 2B,C), confirming the positive functional role of FTO in adipocyte formation.

Next, we sought to compare the global gene expression profiles in rhein-treated cells with those in Fto-knockdown cells during the process of adipocyte differentiation. At the MCE and the differentiation stages, knockdown of Fto led to over 4,000 differentially expressed genes (DEGs) compared to the control cells (shScram) (Supplementary Figures 3A,B). However, only hundreds of genes changed expression in response to rhein treatment (Supplementary Figures 3C,D). There were 53 and 102 genes commonly downregulated in both the groups at the MCE stage and differentiation stage, respectively (Figures 2A,B). As expected, the GO analysis revealed that these overlapped genes were involved in the fat cell differentiation process, fatty acid metabolic process, and PPARγ signaling pathway (Figures 2C,D and Supplementary Figures 4A,B). Consistent with our previous results, the primary adipogenic marker genes and their regulatory genes, Fabp4, Adipoq, C/ebpβ, and Pparγ, were included in the overlap group (Figure 2E). A close inspection of the MCE stage revealed that only 16.5% (53 out of 320) of genes downregulated upon rhein also showed reduced expression in response to FTO deficiency (Figure 2A). The similar pattern was also observed at the differentiation stage. Only 21.5% of genes in the rhein-downregulated group bore decreased expression upon Fto knockdown (Figure 2B), suggesting that rhein might primarily target FTO-independent regulatory pathway to interfere adipocyte differentiation.

We, therefore, focused on the genes that are specifically downregulated in the rhein-treated group. As overlapped genes, rhein-specific downregulated genes were associated with positive regulation of fat cell differentiation by using the GSEA (Supplementary Figures 4C,D). In addition, the GSEA analysis also revealed that these genes participated in the cell cycle and cell division pathway (Figure 2F), suggesting that rhein treatment might disturb mitotic progression. We further took a close look at the mitotic-related genes. In contrast to the adipogenic genes, they were obviously downregulated upon rhein treatment rather than FTO deficiency, especially at differentiation stage (Figure 2G). These observations collectively indicate that besides minimal FTO-dependent regulatory genes, rhein could limit the expressions of FTO-independent mitotic genes as well to attenuate the adipocyte formation.

We have shown that rhein treatment results in remarkable reduced expressions of genes involved in the mitotic pathway. To confirm whether rhein plays a role in the regulation of MCE progression, we detected two typical mitotic markers, cyclin A and cyclin B1, during MDI-induced adipogenic formation. As shown in Figure 3A, both cyclin A and cyclin B1 proteins accumulated at highest levels at 24 h after the addition of MDI medium (Figure 3B). The presence of rhein delayed the maximal induction of cyclin proteins with their expression peaks at 48 h, suggesting that rhein treatment retards the progression of the cell cycle.

To further validate the functional role of rhein in the regulation of mitotic progression, we initially synchronized 3T3-L1 cells at G1/S boundary by using a double thymidine block. After release by addition of fresh medium, cells went through the cell cycle synchronously from the G1/S boundary to mitosis until the subsequent G1 phase. In the absence of rhein, levels of both cyclin expression fluctuated during the cell cycle. Specifically, the cyclin A protein was present at low levels at 6 h, while cyclin B1 abundance increased gradually (Figure 3C). Intriguingly, rhein treatment at the MCE stage remarkably prolonged cyclin A degradation and retarded cyclin B1 expression, highlighting the negative function of rhein in the regulation of cell cycle progression. Further supporting this notion, rhein-treated 3T3-L1 cells proliferated more slowly compared to the control cells (Figure 3D). Collectively, these results provide strong evidence indicating that rhein retards the MCE progression to suppress adipogenic differentiation.

We next characterized the m⁶A dynamics during the adipogenesis process. As shown in dot blot assay (Supplementary Figure 5A), m⁶A levels underwent dynamic increase during adipogenic differentiation, which is consistent with the previous study (26). Global m⁶A methylome analysis revealed typical enrichment of m⁶A in 3′ untranslated region (UTR) and near stop codon along the transcripts at the three different stages during adipogenesis (Figure 4A). Besides, de-novo motif analysis by using program Multiple Em for Motif Elicitation (MEME) identified the consensus sequence “GGAC” (Figure 4B). Surprisingly, a close inspection of methylomes showed that m⁶A read density increased dramatically at 5′ UTR, but not 3′ UTR or coding regions (CDs) of mRNAs during adipogenic differentiation (Figure 4A). Consistent with this, the percentage of m⁶A peaks within 5′ UTR was relatively high at the later two stages compared to the growth arrest stage (Supplementary Figure 5B). Altogether, these results suggest that m⁶A plays crucial roles in the regulation of adipogenic formation.

We tend to investigate how rhein regulates the global m⁶A methylome. As a potential FTO inhibitor, rhein treatment triggered an increased abundance of m⁶A, but not dysregulated expressions of m⁶A writer and eraser proteins (Supplementary Figures 6A,B), suggesting that rhein mainly interferes with the catalytic activity rather than protein abundance of FTO. We focused on the comparative m⁶A methylome in the presence and absence of rhein at the MCE stage, as rhein treatment substantially suppresses the mitotic progression. Unexpectedly, global m⁶A peak analysis revealed a specific increase of m⁶A read density at 3′ UTR after rhein treatment (Figure 4C), suggesting that rhein and differentiation signal possibly perturb the global m⁶A methylome in a distinct manner.

Next, we sought to assess whether rhein modulates gene expression in an m⁶A-dependent manner. As rhein may also induce the upregulation of m⁶A levels at the other regions of individual transcripts through acting on FTO, we analyzed all the mRNAs with significantly increased m⁶A at any regions upon rhein. These transcripts showed similar abundance before and after rhein treatment (Figure 4D). Individually, the majority of m⁶A-containing transcripts were insensitive to be dysregulated upon rhein such as H1f0, Zfp50, and Chpf (Figures 4E,F). On the other hand, we deeply analyzed the 320 MCE stage-specific rhein-downregulated transcripts, which are involved in the mitotic pathway. Only 10.94% of these mRNAs contained m⁶A peaks (Figure 4G), while the majority of transcripts did not bear m⁶A peaks as exemplified by Uba52, Rpl36, and receptor expressing enhancing protein 3 (Reep3) (Figure 4H). Furthermore, m⁶A-bearing transcripts were much less sensitive to rhein compared to non-m⁶A transcripts (Figure 4G). Collectively, the available evidence suggests that rhein modulates global RNA methylation landscape and adipogenic differentiation in a separate manner.

We focused on one of m⁶A-indepdendent targets of rhein, REEP3. We initially examined Reep3 expression levels in cells cultured in the presence of rhein. At the MCE stage, the expression of Reep3 increased gradually in the control cells (Figure 5A). Upon rhein treatment, its expression levels substantially reduced in a dose-dependent manner (Figures 5A,B).

To address the role of Reep3 in regulating adipogenesis, we generated 3T3-L1 cells depleting Reep3. As determined by the cell proliferation assay, knockdown of Reep3 obviously slowed down the cell proliferation rate (Supplementary Figure 7 and Figure 5C). As shown in adipogenic formation assay, ~60–70% control cells were able to differentiate into adipocytes, as determined by ORO staining (Figure 5D). Remarkably, depleting of Reep3 resulted in severe adipogenesis defects with dramatic reduction of lipid droplets accumulation (Figure 5D). Consistent with the phenotype, knockdown of Reep3 significantly downregulated the adipogenesis marker genes Pparγ, Fabp4, and Adipoq (Figures 5E,F). Altogether, these data indicate that REEP3 plays an essential role in driving adipogenic differentiation.