Clerodendrum japonicum Extract Suppresses the Pathogenicity of Rheumatoid Arthritis Fibroblast‐Like Synoviocytes by Inhibiting METTL3‐Mediated ICAM2 mRNA Methylation Authors: Ying Huang, JiangCun Wei, Yan Ma, XiaoPing Mei, MingTao Huang, Liang Lu, JianLong Zhou, ZuJie Qin Published: 2026 ABSTRACT ABSTRACT Clerodendrum japonicum (Thunb.) Sweet (Clerodendrum japonicum) is a traditional medicinal plant with reported anti‐inflammatory use, but its anti‐rheumatoid arthritis (RA) activity and molecular basis remain insufficiently characterized. Activated fibroblast‐like synoviocytes (FLSs) contribute to persistent inflammation and joint destruction in rheumatoid arthritis (RA). This study investigated whether extracts of Clerodendrum japonicum attenuate pathogenic RA‐FLS phenotypes and examined the possible involvement of METTL3‐dependent N6‐methyladenosine (m6A) regulation of ICAM2 mRNA. Tumor necrosis factor‐α‐stimulated human RA‐FLSs (MH7A cell line) were used as an in vitro model to evaluate the anti‐RA effects of Clerodendrum japonicum extracts. Cell proliferation, apoptosis, migration, invasion, and inflammatory cytokine secretion were evaluated. The underlying mechanisms were further investigated using real‐time quantitative polymerase chain reaction, Western blot, and RNA immunoprecipitation. In addition, a collagen‐induced arthritis (CIA) mouse model was employed to validate the anti‐RA effects of the extracts in vivo. Drug‐containing serum derived from Clerodendrum japonicum extracts attenuated TNF‐α‐induced proliferation, migration, invasion, and pro‐inflammatory cytokine release in MH7A cells while increasing apoptosis. The ethyl acetate fraction (CJEA) produced the most pronounced effects among the tested fractions. CJEA‐containing serum reduced global m6A abundance and METTL3 expression. METTL3 overexpression partially reversed the cellular effects associated with CJEA‐containing serum. METTL3 knockdown reduced m6A enrichment and stability of ICAM2 mRNA, whereas METTL3 or ICAM2 overexpression partially restored ICAM2 expression and pathogenic RA‐FLS phenotypes. Oral CJEA also reduced clinical and histological arthritis severity in CIA mice and was associated with lower synovial METTL3 and ICAM2 expression. Our findings support the involvement of METTL3‐dependent m6A regulation of ICAM2 mRNA in the response of RA‐FLSs to CJEA‐containing serum. CJEA may attenuate pathogenic RA‐FLS behavior and experimental arthritis at least partly through modulation of this pathway. Because the active constituent(s), direct molecular target, pharmacokinetics, and comprehensive safety profile remain undefined, further chemical and pharmacological validation is required.Keywords: Clerodendrum japonicum, fibroblast‐like synoviocytes (FLSs), ICAM2, METTL3, rheumatoid arthritis (RA) Introduction Rheumatoid arthritis (RA) is a chronic autoimmune inflammatory disease characterized by synovitis, progressive cartilage damage, and bone erosion [1]. RA has a global prevalence of approximately 0.5%–1.0% [2]. Immunological abnormalities can precede clinically apparent joint swelling, and a substantial proportion of patients show inadequate or unsustained responses to conventional, biological, or targeted synthetic disease‐modifying antirheumatic drugs [3]. Identification of additional disease‐relevant pathways and pharmacologically tractable mechanisms, therefore, remains important. Identification of additional disease‐relevant pathways and pharmacologically tractable mechanisms, therefore, remains important. RA pathogenesis reflects interactions among genetic susceptibility, environmental exposure, immune dysregulation, and stromal‐cell reprogramming [4, 5]. Fibroblast‐like synoviocytes (FLSs) located in the synovial lining acquire persistent tumor‐like characteristics, including enhanced proliferation, migration, invasion, inflammatory mediator production, and resistance to apoptosis [6, 7]. Activated RA‐FLSs promote synovial hyperplasia, recruit and modulate immune cells, release matrix‐degrading enzymes, and directly contribute to cartilage and bone destruction [8, 9]. These features make RA‐FLSs a relevant cellular target for mechanistic and therapeutic studies. N6‐methyladenosine (m6A), one of the most prevalent RNA modifications in eukaryotic cells, regulates RNA processing, translation, and decay, thereby influencing cell differentiation and tissue development [10]. A growing body of research highlights a contributing role for m6A modification in the pathophysiology of RA [11, 12, 13]. m6A is deposited by writer complexes containing methyltransferase‐like 3 (METTL3), methyltransferase‐like 14 (METTL14), Wilms' tumor 1‐associated protein (WTAP), RNA binding motif protein 15 (RBM15/RBM15B), and VIRMA; removed by the erasers fat mass and obesity‐associated protein (FTO) and AlkB homolog 5 (ALKBH5); and interpreted by reader proteins, including YTH‐domain proteins and insulin‐like growth factor 2 (IGF2BP) family members. Increasing evidence implicates altered m6A regulation in RA and in the abnormal activation of RA‐FLSs [14, 15, 16, 17]. The functional link between m6A modification, RA‐FLSs, and RA pathogenesis not only provides new insights into the molecular mechanisms of the disease but also opens new avenues for developing novel therapeutic strategies against RA. Clerodendrum japonicum (Thunb.) Sweet (C. japonicum) in southern and eastern Asia and has long been used in regional traditional medicine for inflammatory and swelling‐related conditions [18]. Phytochemical studies have reported terpenoids, lignans, phenylethanoid glycosides, phenolic acids, flavonoids, and alkaloids in this species [19], providing a material basis for explaining its systemic pharmacological effects. In traditional medicine, C. japonicum is often used to treat inflammation‐related conditions such as upper respiratory tract infections, fever, cough, and conjunctivitis [20]. Its traditional functions of “expelling wind‐dampness and unblocking meridians” align closely with the treatment principles for RA in modern medicine. Natural products and plant‐derived small molecules can suppress pathogenic RA‐FLS phenotypes and ameliorate experimental arthritis through multiple metabolic and inflammatory pathways [21, 22, 23, 24, 25]. However, the composition, active constituents, and molecular actions of C. japonicum preparations in RA remain poorly defined. Accordingly, this study compared serum obtained from rats administered different polarity fractions of C. japonicum and selected the ethyl acetate fraction (CJEA) for further mechanistic analysis. Tumor necrosis factor‐alpha (TNF‐α)‐stimulated MH7A cells were used to assess RA‐FLS proliferation, apoptosis, migration, invasion, and cytokine production. The potential involvement of METTL3‐dependent m6A regulation of Intercellular Adhesion Molecule 2 (ICAM2) mRNA was examined using loss‐ and gain‐of‐function approaches, transcript‐specific methylation and binding assays, and mRNA stability analysis. The anti‐arthritic activity of CJEA was additionally evaluated in CIA mice. This study provides preliminary insights into the mechanism by which C. japonicum extract exerts anti‐RA effects and may offer a scientific basis for its further investigation as a potential anti‐RA agent. Methods Preparation of C. japonicum Extracts The medicinal material of C. japonicum was sourced from the Chinese Medicine Pharmacy of International Zhuang Medicine Hospital Affiliated to Guangxi University of Chinese Medicine and authenticated by a Chinese pharmacist in the Department of Pharmacy as the roots of C. japonicum (Thunb.) Sweet from the Verbenaceae family. A total of 8.5 kg of dried material was pulverized and macerated in 75% ethanol at a solvent‐to‐material ratio of 15:1 (mL/g) for 24 h, followed by reflux extraction for 1.5 h. The extraction was repeated three times. The combined filtrates were concentrated under reduced pressure and dried to a constant weight, yielding 1200.2 g of the 75% ethanol total extract (CJEE), corresponding to an extraction yield of 14.12%. A portion of CJEE was suspended in distilled water and successively partitioned with petroleum ether and ethyl acetate. The ethyl acetate layers were combined, concentrated under reduced pressure, and dried to a constant weight, yielding 195.5 g of the ethyl acetate fraction (CJEA), equivalent to 2.30% of the starting dry crude drug. The residual aqueous fraction was subsequently concentrated under reduced pressure and dried to a constant weight, yielding 647.7 g of the water‐soluble fraction (CJWE), corresponding to an extraction yield of 7.62% relative to the starting dry plant material. Extraction yield was calculated using the following equation: extraction yield (%) = dry extract mass/starting dry crude‐drug mass × 100%. All dried extracts were stored in airtight, light‐protected containers at 4°C until use. To provide preliminary phytochemical information, HPLC analysis was performed on the parent 75% ethanol extract of C. japonicum roots. Ten batches yielded 12 common peaks, of which peaks 2, 4, and 12 were identified as caffeic acid, acteoside, and kaempferol, respectively, by comparison with authentic standards (Figure S1). The similarity of each batch to the reference chromatogram ranged from 0.965 to 0.995 (Table S1). Precision, repeatability, and 24‐h stability tests showed relative standard deviation (RSD) values for peak areas not exceeding 2.90% (Table S2). These results support qualitative reproducibility of the parent root ethanol extract. Preparation of Blank and Drug‐Containing Serum Six‐week‐old male Wistar rats (weight 180 ± 200 g) were purchased from Hunan Slake Kingda Laboratory Animal Co. Ltd. (Changsha, China). The rats were housed under controlled conditions: temperature 22 ± 2°C, humidity 55 ± 2%, and a 12‐h light/dark cycle, with free access to water and standard feed. Twenty‐four rats were randomly divided into four groups using a random number generator to ensure unbiased distribution (n = 6 per group): blank control group, CJEE administration group, CJWE administration group, and CJEA administration group. Based on a clinical crude‐drug dose of 15–30 g/day for a 60‐kg adult and body‐surface‐area conversion, a rat‐equivalent dose of approximately 1.6–3.2 g/kg was estimated. Following preliminary dose‐ranging experiments, a final dose of 12 g/kg (approximately 4 times the equivalent dose) was selected for administration. Extract doses were converted to dried‐extract doses using the experimentally determined extraction yields and freshly suspended in 0.5% sodium carboxymethylcellulose (CMC‐Na). For drug‐containing serum preparation, rats received 12 g/kg/day crude equivalent by gavage (10 mL/kg), which corresponded to dried‐extract doses of 1694.4 mg/kg (CJEE), 914.4 mg/kg (CJWE), and 276.0 mg/kg (CJEA); accordingly, dosing suspensions were prepared at 169.44, 91.44, and 27.60 mg/mL. Rats were treated daily with the respective extract or vehicle (blank control) for 7 consecutive days. Two hours after the final administration, rats were anesthetized, and blood was collected from the abdominal aorta. Serum was separated after clotting and centrifugation, heat‐inactivated at 56°C for 30 min, filtered through a 0.22‐μm membrane, and stored at −80°C. Sera from the six rats within each treatment group were pooled to obtain a group‐specific serum preparation for the cell experiments. Cell Culture The human MH7A cell line (RRID: CVCL_0427; species: Homo sapiens; sex: female; tissue of origin: synovium) was obtained from BioVector NTCC Inc. (Beijing, China) in 2024. Cells were cultured in Dulbecco's Modified Eagle Medium (DMEM; Gibco, BRL, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (FBS; Gibco) and 1% penicillin‐streptomycin (Gibco), and maintained at 37°C in a humidified incubator with 5% CO2. Before use, the cells were authenticated by short tandem repeat profiling and tested negative for mycoplasma contamination using a Mycoplasma Detection Kit (Invitrogen, Carlsbad, CA, USA). The cell line is not listed as misidentified in the International Cell Line Authentication Committee database. Cell Treatment MH7A cells were assigned to an unstimulated control group, a TNF‐α plus blank‐serum group, extract‐containing‐serum groups, and a methotrexate (MTX) positive‐control group, as applicable to each experiment. Cells were pretreated for 24 h with DMEM containing 20% blank rat serum or 20% serum from rats administered CJWE, CJEE, or CJEA, and were then stimulated with TNF‐α (20 ng/mL; PeproTech, Rocky Hill, NJ, USA) for 24 h. The positive‐control group was pretreated with MTX (2 μM; Shanghai Yuanye Bio‐Technology Co. Ltd., Shanghai, China) for 24 h before TNF‐α stimulation [26, 27]. Blank serum was included to distinguish effects of serum exposure from extract‐associated effects. To complement the drug‐containing serum experiments and determine whether CJEA directly modulated RA‐FLS responses, MH7A cells were treated with CJEA at concentrations ranging from 25 to 100 μg/mL. CJEA was dissolved in dimethyl sulfoxide (DMSO) and subsequently diluted in culture medium. The final DMSO concentration was maintained at ≤ 0.1% in all corresponding groups. MH7A cells were pretreated with CJEA for 2 h and then stimulated with TNF‐α (20 ng/mL) for an additional 24 h. Cells treated with TNF‐α and an equivalent concentration of DMSO served as the vehicle control, whereas methotrexate (MTX, 2 μM) was used as the positive control. Untreated MH7A cells without TNF‐α stimulation served as the normal control. Cell Transfection METTL3 overexpression (OE‐METTL3), ICAM2 overexpression (OE‐ICAM2), METTL3 short hairpin RNA (sh‐METTL3), ICAM2 shRNA (sh‐ICAM2), and the corresponding empty‐vector or scrambled shRNA controls were obtained from GenePharma Co. Ltd. (Shanghai, China). The sequences of the shRNA targeting regions and the open reading frames (ORFs) of the overexpression constructs are listed in Tables S3 and S4, respectively. MH7A cells were seeded in 6‐well plates and transfected at approximately 60% confluence using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions. Cells receiving an overexpression construct were compared with the corresponding empty‐vector control, and cells receiving shRNA were compared with sh‐NC. Transfection efficiency was verified by RT‐qPCR and/or Western blotting before functional analysis. Cell Counting Kit‐8 (CCK‐8) The assay was performed according to the instructions of the CCK‐8 kit (Solarbio, Beijing, China). Cells were seeded overnight in 96‐well plates at a density of 2 × 104 cells per well, followed by the respective treatments. Then, 10 µL of CCK‐8 solution was added directly to 100 µL of DMEM in each well, and the plates were incubated at 37°C for 2 h. Absorbance was measured at 450 nm using a fully automated microplate reader (Bio‐Rad Laboratories Inc., Hercules, CA, USA). 5‐Ethyl‐2′‐Deoxyuridine (EdU) EdU incorporation was assessed using an EdU Cell Proliferation Detection Kit (RiboBio, Guangzhou, China). After labeling, fixation, and staining, images were acquired using identical microscope settings. EdU‐positive and total nuclei were counted in three randomly selected non‐overlapping fields per sample using ImageJ. The proliferation rate was expressed as the percentage of EdU‐positive nuclei. Flow Cytometry Cells were digested with 0.25% trypsin, collected in flow cytometry tubes, centrifuged, and the supernatant was discarded. After washing three times with phosphate‐buffered saline (PBS), the cells were resuspended. According to the instructions of the Annexin V‐FITC Apoptosis Detection Kit (Beyotime, Shanghai, China), 500 µL of loading buffer was added and mixed gently, followed by the addition of 5 µL Annexin V‐FITC and 10 µL propidium iodide solution. Samples were incubated for 15–20 min at room temperature in the dark and analyzed using a BD flow cytometer (Franklin Lakes, NJ, USA). At least 10,000 cellular events were acquired per sample after exclusion of debris and doublets. Total apoptosis was defined as the sum of early apoptotic cells (Annexin V‐positive/PI‐negative) and late apoptotic or secondary necrotic cells (Annexin V‐positive/PI‐positive). Wound Healing Assay After counting, cells were seeded into 6‐well plates (3 × 105 cells/well) and cultured for 24 h. A sterile 200 µL pipette tip was used to create a straight scratch perpendicular to the surface of the well, guided by a line on the back of the plate. After scratching, the cells were washed three times with PBS to remove dislodged cells, leaving a clear, visible gap. Fresh medium was then added, and the cells were returned to the incubator. The scratch width was observed and measured under a microscope at 0 and 48 h post‐scratch. Wound closure was calculated as [(initial wound width − wound width at 48 h)/initial wound width] × 100%. Transwell Cell invasion ability was detected using Matrigel‐coated Transwell chambers (Becton Dickinson, Franklin Lakes, NJ, USA). Cells were digested, washed twice with serum‐free medium, resuspended, and counted. Then, 200 µL of cell suspension containing 1 × 105 transfected cells was added to the upper chamber, while 500 µL of DMEM medium containing 10% FBS (acting as a chemoattractant) was added to the lower chamber. After 24 h, cells remaining on the upper membrane surface were removed. Invaded cells on the lower surface were fixed with 4% paraformaldehyde, stained with crystal violet, photographed, and counted in five randomly selected non‐overlapping fields using ImageJ. Three independent biological experiments were performed. Enzyme‐Linked Immunosorbent Assay (ELISA) The levels of interleukin‐1 beta (IL‐1β) and interleukin‐6 (IL‐6) concentrations in independent cell‐culture supernatants and individual mouse serum samples were measured using ELISA kits (R&D Systems, Minneapolis, MN, USA) according to the manufacturer's instructions. Each sample was assayed in technical duplicate, and cytokine concentrations were calculated from standard curves and reported as pg/mL. The mean of technical duplicates was used as one biological observation. Western Blot Total protein was extracted using a protein extraction kit (Beyotime), and protein concentration was determined using a bicinchoninic acid (BCA) protein assay kit (Beyotime). Approximately 40 µg of protein per sample was loaded for sodium dodecyl sulfate‐polyacrylamide gel electrophoresis (SDS‐PAGE) using 10% separating gels and 4% stacking gels. Electrophoresis was performed at 120 V in the separating gel and 80 V in the stacking gel. Proteins were then transferred onto polyvinylidene fluoride (PVDF) membranes using a semi‐dry transfer system at 4°C (0.8 mA/cm2). After transfer, the membranes were blocked with 5% skim milk at room temperature for 1 h. Subsequently, they were incubated overnight at 4°C with the following primary antibodies: METTL3 (1:2000, 67733‐1‐Ig; Proteintech, Wuhan, China), ICAM2 (1:1000, sc‐9987; Santa Cruz Biotechnology, Santa Cruz, CA, USA), METTL14 (1:500, A8530; Affinity, Jiangsu, China), WTAP (1:500, DF3282; Affinity), FTO (1:1000, DF8421; Affinity), ALKBH5 (1:500, DF2585; Affinity), and β‐actin (1:4000, ab115777; Abcam, Cambridge, UK). After washing three times for 10 min each with Tris‐Buffered Saline with Tween 20 (TBST), the membranes were incubated with horseradish peroxidase (HRP)‐conjugated anti‐rabbit or anti‐mouse IgG secondary antibody (Sangon Biotech, Shanghai, China) at room temperature for 1 h. Protein bands were visualized using an enhanced chemiluminescence (ECL) detection system. Densitometric analysis was performed using ImageJ, target proteins were normalized to β‐actin, and quantification was based on three independent protein preparations. Real‐Time Quantitative Polymerase Chain Reaction (RT‐qPCR) Total RNA was extracted using Trizol reagent (Invitrogen). RNA was reverse transcribed into complementary DNA (cDNA) using the PrimeScript™ RT reagent Kit with gDNA Eraser (TaKaRa, Tokyo, Japan) according to the manufacturer's instructions. Primer sequences were designed (Table 1) and synthesized by Shanghai GeneChem Co. Ltd. qPCR was performed using the 7500 Real‐Time PCR System (Applied Biosystems, Foster City, CA, USA) and SYBR Green Master Mix (Applied Biosystems). The messenger RNA (mRNA) levels of target genes were normalized to β‐actin and analyzed using the 2‐∆∆Ct method. Table 1Primer sequences.GenesPrimer Sequence (5′ → 3′)METTL3F: TTGTCTCCAACCTTCCGTAGTR: CCAGATCAGAGAGGTGGTGTAGICAM2F: CCACAGCCACATTCAACAR: AAGCAGAGCAGGACAGATMETTL14F: TGCCTGTGATGGGTCCTTAGR: ACAGGTGCCTATGCCATGTAWTAPF: AAGCAACAACAGCAGGAGTCR: TCGCTGGGTCTACCATTGTTFTOF: AGACACCTGGTTTGGCGATAR: GTTCCTGTTGAGCACTCTGCALKBH5F: TCTGCACTTGGTTGAGGTCTR: AGGGTGTTTGCATGAGCTTGβ‐actinF: CCCTGGAGAAGAGCTACGAGR: GGAAGGAAGGCTGGAAGAGT Primer sequences. Quantification of Total m6A Global m6A abundance was measured using the EpiQuik m6A RNA Methylation Quantification Kit (Colorimetric; EpigenTek, Farmingdale, NY, USA). Equal RNA input (200 ng/well) was used for all samples. Kit‐supplied negative and positive controls were included, and absorbance was measured at 450 nm. m6A content was calculated from the assay standard curve according to the manufacturer's formula: m6A (%) = (sample OD − negative‐control OD)/(slope × RNA input) × 100%. mRNA Stability Assay Transfected cells were cultured for 24 h, then treated with 4 μmol/L actinomycin D (Sigma‐Aldrich, St. Louis, MO, USA) to inhibit transcription. Cells were collected at 0, 3, 6, and 12 h post‐treatment. RNA was extracted, and ICAM2 mRNA levels were detected by RT‐qPCR. Methylated RNA Immunoprecipitation‐qPCR (MeRIP‐qPCR) The experiment was performed using the Magna MeRIP m6A Kit (Millipore, Bedford, MA, USA). Briefly, cells were lysed with complete RIP lysis buffer. The cell lysates were incubated overnight at 4°C with magnetic beads pre‐coated with anti‐m6A antibody (ab208577; Abcam) or anti‐IgG antibody (ab172730; Abcam) as a negative control. The beads were washed three times, and bound RNA was eluted using elution buffer for 2 h. The purified RNA, obtained using the RNeasy Mini Kit (Qiagen, Hilden, Germany), was subjected to RT‐qPCR analysis. ICAM2 enrichment was calculated relative to input and compared with the corresponding IgG control. RIP‐qPCR The RIP assay was conducted according to the instructions of the Magna RIP RNA‐Binding Protein Immunoprecipitation Kit (Millipore). Cells were lysed with complete RIP lysis buffer. The lysates were incubated overnight at 4°C with magnetic beads coated with anti‐METTL3 antibody (ab195352; Abcam) or control anti‐IgG antibody (ab172730; Abcam). RNA was extracted from the immunoprecipitates and analyzed by RT‐qPCR to determine the enrichment level of ICAM2 mRNA. Dual‐Luciferase Reporter Assay The potential m6A modification sites on ICAM2 mRNA were predicted using the online bioinformatics tool SRAMP. A wild‐type (WT) ICAM2 reporter plasmid (ICAM2‐WT) containing the predicted m6A site was synthesized. A mutant (MUT) ICAM2 reporter plasmid (ICAM2‐MUT) was constructed by site‐directed mutagenesis of the binding site. Plasmids were extracted using a commercial kit (Promega, Madison, WI, USA). Experimental cells were co‐transfected with either the ICAM2‐WT or ICAM2‐MUT plasmid, along with either sh‐METTL3 or its corresponding negative control (sh‐NC). 48 h post‐transfection, cells were collected and lysed. Luciferase activity was measured using a dual‐luciferase reporter assay kit (BioVision, San Francisco, CA, USA). Establishment of the CIA Model and Experimental Grouping Male DBA/1 mice (aged 6–8 weeks, weighing 18–20 g) were obtained from Hunan Slake Kingda Laboratory Animal Co. Ltd. All mice were housed under standard laboratory conditions with ad libitum access to food and water. Following a 1‐week acclimation period, the CIA model was induced. All animal experiments were reviewed and approved by the Animal Experimental Ethics Committee of Guangxi University of Chinese Medicine (Approval No. DW20250924‐01). All procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and reported in accordance with the ARRIVE guidelines. The CIA model was established as previously described [28, 29]. Briefly, bovine type II collagen (2 mg/mL, Chondrex, Redmond, WA, USA) was emulsified with an equal volume of Complete Freund's Adjuvant (Sigma‐Aldrich) for the primary immunization and with Incomplete Freund's Adjuvant for the booster, resulting in a final collagen concentration of 1 mg/mL. On Day 0, each mouse received a 100 µL subcutaneous injection of the emulsion at the tail base. A booster injection of 100 µL emulsion was administered at the same site on Day 21. Mice in the normal control (Normal) group received an equal volume of sterile PBS via the same procedure. Beginning on Day 22 post‐primary immunization, arthritis severity was assessed clinically by two independent, blinded observers using an established scoring system [30]: 0 = no signs; 1 = mild redness and swelling in one or more ankle/wrist joints or digits; 2 = moderate redness and swelling extending from the ankle/wrist; 3 = severe redness and swelling involving the entire paw; 4 = maximal inflammation and joint involvement. The scores from all four limbs were summed for each mouse, yielding a maximum possible score of 16. Mice with a total score ≥ 2 were deemed successfully modeled. These successfully modeled mice were then randomly allocated into three experimental groups (n = 6 per group, based on resource feasibility and preliminary studies): the model control group (CIA), the treatment group (CIA + CJEA), and the positive control group (CIA + MTX). An additional six mice served as the Normal group. Starting from the confirmation of successful modeling, mice in the CIA + MTX group were administered methotrexate (1 mg/kg) via intraperitoneal injection every 3 days. This dosage regimen was established with reference to existing protocols [31, 32] and was further confirmed and optimized through preliminary experiments. Mice in the CIA + CJEA group received CJEA daily via oral gavage at a dose of 20 g/kg, which was determined by converting the human equivalent dose based on body surface area and subsequently optimized through preliminary experiments. Treatments continued for 28 consecutive days. The Normal and CIA groups received an equal volume of vehicle daily via gavage. Oral CJEA was selected to reflect the intended administration route of the botanical preparation, whereas intraperitoneal MTX was used according to established CIA protocols to provide reliable systemic exposure. MTX, therefore, served as a positive reference rather than a route‐matched comparator. Body weight, survival, food intake, general appearance, activity, and overt adverse events were monitored throughout treatment. Serum alanine aminotransferase (ALT) and creatinine levels were measured using commercially available assay kits (Nanjing Jiancheng Biotechnology Research Institute, Jiangsu, China) according to the manufacturer's instructions as preliminary indicators of hepatic and renal function, respectively. Sample Collection Following the final treatment, mice were fasted for 12 h and subsequently anesthetized with isoflurane to minimize suffering prior to sacrifice. Blood was collected from the abdominal aorta, allowed to clot at room temperature for 30 min, and centrifuged at 3000 rpm for 10 min. The separated serum was stored at −80°C for subsequent analysis. Knee and ankle joints from each mouse were either fixed in 4% paraformaldehyde for histopathological examination or snap‐frozen and stored at −80°C for further molecular analyses. Hematoxylin and Eosin (H&E) and Safranin O/Fast Green Staining Following routine deparaffinization, rehydration through a graded alcohol series, and rinsing with deionized water, the paraffin‐embedded sections were subjected to staining. H&E staining was performed using a kit from Solarbio (Beijing, China) according to the manufacturer's instructions. Safranin O/Fast Green staining was carried out using reagents from Sangon Biotech Co. Ltd. (Shanghai, China). Subsequently, all stained sections were observed under an optical microscope (Olympus, Tokyo, Japan). Finally, synovial inflammation and cartilage damage were evaluated according to previously established criteria [32]. Two investigators independently scored coded sections while blinded to group allocation; the mean of the two scores was used for analysis. Statistical Analysis All data were analyzed using GraphPad Prism 10.0 software (GraphPad Software, San Diego, CA, USA). Normality was assessed by the Shapiro‐Wilk test and homogeneity of variances by the Brown‐Forsythe test. Parametric tests were used only when the relevant assumptions were satisfied. Otherwise, non‐parametric tests were applied. For comparisons among multiple groups, one‐way analysis of variance (ANOVA) followed by Tukey's post hoc test was used for normally distributed data with equal variance; Welch's ANOVA followed by Games‐Howell test was used when variances were unequal; and Kruskal–Walli's test followed by Dunn's post hoc test was used for non‐normally distributed data. CIA clinical scores were analyzed by two‐way repeated‐measures ANOVA, with treatment group as the between‐subject factor and time as the within‐subject factor. Animal identification was specified as the repeated‐measures subject. The Geisser‐Greenhouse correction was applied to adjust for potential violations of sphericity. When a significant treatment‐by‐time interaction was detected, Dunnett's multiple‐comparison test was used to compare the CIA + CJEA and CIA + MTX groups with the untreated CIA group at prespecified time points. A p < 0.05 was considered statistically significant. Independent cell experiments or individual animals served as the experimental units; technical replicate wells or fields were averaged before analysis and were not treated as independent observations. Unless otherwise stated, n = 3 for cell experiments denotes three independent biological experiments conducted on different days using independently prepared cell cultures and samples. Results C. japonicum extract‐containing serum suppresses the aggressive phenotype and inflammatory cytokine release in RA‐FLSs A TNF‐α‐induced human MH7A cell model was established to evaluate the therapeutic effects of drug‐containing serum derived from various C. japonicum extracts on RA‐FLSs. CCK‐8 and EdU assays revealed that the extract‐containing serum significantly suppressed TNF‐α‐stimulated proliferation of RA‐FLSs, with the ethyl acetate fraction (CJEA) exhibiting the most potent inhibition, comparable to the positive control (Figure 1A,B). Furthermore, the extract‐containing serum promoted apoptosis of TNF‐α‐induced MH7A cells (Figure 1C). Wound healing and Transwell assays confirmed that treatment with the serum also markedly impaired the migratory and invasive capacities of RA‐FLSs (Figure 1D,E). In addition, ELISA results demonstrated that the extracts significantly inhibited the TNF‐α‐induced upregulation of IL‐1β and IL‐6, indicating effective suppression of pro‐inflammatory cytokine release (Figure 1F). Consistently, CJEA again showed the strongest anti‐inflammatory efficacy among all extracts. Clerodendrum japonicum extract‐containing serum suppresses the aggressive phenotype and inflammatory cytokine release in RA‐FLS. MH7A cells were assigned to the following groups: unstimulated control (Ctrl), TNF‐α + blank serum (Blank), TNF‐α + the indicated extract‐containing serum (CJWE, CJEE, or CJEA), and TNF‐α + methotrexate (MTX). Blank serum served as the vehicle control for extract‐containing serum groups; MTX served as the positive control. (A) Cell viability (%) measured by CCK‐8 and expressed relative to the unstimulated control. (B) Representative EdU images and percentage of EdU‐positive nuclei. Scale bar: 100 µm. (C) Annexin V‐FITC/PI flow cytometry plots and quantification of total apoptosis. Total apoptosis rate was defined as the sum of early apoptotic cells (Annexin V‐positive/PI‐negative) percentage and late apoptotic or secondary necrotic cells (Annexin V‐positive/PI‐positive) percentage. At least 10,000 events were acquired per sample. (D) Representative wound‐healing images and wound closure percentage. Wound width was measured at 0 and 48 h after scratching; wound closure rate (%) = [(initial width − width at 48 h)/initial width] × 100. Three randomly selected fields per well were quantified. Scale bar: 100 µm. (E) Representative Matrigel‐Transwell images and number of invaded cells counted in five randomly selected non‐overlapping fields. Scale bar: 100 µm. (F) Concentrations of IL‐1β and IL‐6 in cell culture supernatants (pg/mL). Data are presented as mean ± SD from three independent biological experiments; technical replicates were averaged. Brackets indicate the specific group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant. To determine whether the effects observed with CJEA‐containing serum could also be reproduced by direct extract exposure, MH7A cells were treated directly with defined concentrations of CJEA. Direct CJEA treatment significantly attenuated the TNF‐α‐induced increase in MH7A cell proliferation, as demonstrated by the CCK‐8 and EdU assays (Figure S2A,B). CJEA also increased the proportion of apoptotic cells and reduced the migratory and invasive capacities of TNF‐α‐stimulated MH7A cells (Figure S2C‐E). In addition, CJEA decreased the secretion of IL‐1β and IL‐6 into the culture supernatant (Figure S2F). These findings provide complementary evidence that CJEA‐related constituents can directly modulate inflammatory RA‐FLS responses under the present in vitro conditions. These results collectively suggest that C. japonicum extracts, especially CJEA, attenuate the aggressive phenotype and inflammatory response of RA‐FLS. CJEA‐Containing Serum Is Associated With Altered m6A Regulation in RA‐FLS Given the emerging role of RNA methylation in immune dysregulation, we investigated m6A modification in RA‐FLSs. Global m6A levels were significantly elevated in MH7A cells upon TNF‐α induction (Figure 2A). Profiling of key regulatory enzymes revealed that TNF‐α upregulated several “writers,” with METTL3 and WTAP being the most responsive (Figure 2B,C). Notably, C. japonicum extract‐containing serum, especially CJEA, abolished this TNF‐α‐induced m6A hypermethylation and concurrently suppressed METTL3 expression, but not METTL14 or WTAP (Figure 2A–C). Consistent with the results obtained using CJEA‐containing serum, direct CJEA exposure reduced METTL3 protein expression (Figure S3A). Thus, METTL3‐dependent m6A regulation may contribute to the anti‐RA activity of these extracts, with CJEA showing the most pronounced METTL3‐suppressive effect. CJEA‐containing serum is associated with altered m6A regulation in RA‐FLS. (A) Global RNA m6A percentage determined using equal RNA input and kit‐supplied controls. (B) RT‐qPCR analysis of METTL3, METTL14, WTAP, FTO, and ALKBH5 expression normalized to β‐actin using the 2−ΔΔCt method. (C) Representative Western blots and densitometric quantification of METTL3, METTL14, WTAP, FTO, and ALKBH5 expression normalized to β‐actin. Data are mean ± SD from three independent biological experiments; technical replicates were averaged. Brackets indicate the specific group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant. METTL3 Contributes to the Cellular Response Associated with CJEA‐Containing Serum To determine if METTL3 mediates the modulation of RA‐FLS pathogenic activities by CJEA, we transfected TNF‐α‐induced MH7A cells with a METTL3 overexpression plasmid or negative control. Western blot analysis confirmed successful METTL3 overexpression (Figure 3A). Ectopic METTL3 expression partially rescued the CJEA‐mediated suppression of RA‐FLS proliferation (Figure 3B–D) and similarly abrogated the inhibitory effects of CJEA on cell migration and invasion (Figure 3E,F). Furthermore, METTL3 overexpression counteracted the suppression of inflammatory cytokine release induced by CJEA (Figure 3G). The partial rescue supports METTL3 as a functional mediator of the cellular response to CJEA‐containing serum, while leaving open the possibility that additional pathways contribute. METTL3 contributes to the cellular response associated with CJEA‐containing serum. MH7A cells were transfected with METTL3 overexpression vector (OE‐METTL3) or empty vector (OE‐NC), treated with CJEA‐containing serum, and stimulated with TNF‐α. (A) Representative Western blot confirming METTL3 overexpression. (B) Cell viability (%) measured by CCK‐8 and expressed relative to the unstimulated control. (C) Representative EdU images and percentage of EdU‐positive nuclei. Scale bar: 100 µm. (D) Annexin V‐FITC/PI flow cytometry plots and quantification of total apoptosis. Total apoptosis rate was defined as the sum of early apoptotic cells (Annexin V‐positive/PI‐negative) percentage and late apoptotic or secondary necrotic cells (Annexin V‐positive/PI‐positive) percentage. At least 10,000 events were acquired per sample. (E) Representative wound‐healing images and wound closure percentage. Wound width was measured at 0 and 48 h after scratching; wound closure rate (%) = [(initial width − width at 48 h)/initial width] × 100. Three randomly selected fields per well were quantified. Scale bar: 100 µm. (F) Representative Matrigel‐Transwell images and number of invaded cells counted in five randomly selected non‐overlapping fields. Scale bar: 100 µm. (G) Concentrations of IL‐1β and IL‐6 in cell culture supernatants (pg/mL). The key rescue comparison is CJEA + OE‐METTL3 versus CJEA + OE‐NC. Data are mean ± SD from three independent biological experiments; technical replicates were averaged. Brackets indicate the specific group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant. METTL3 Maintains ICAM2 mRNA Stability in an m6A‐Dependent Manner Further investigation into the downstream mechanism of METTL3 was conducted. Previous research [33] indicated that METTL3 showed the strongest positive correlation with ICAM2 in RA‐FLSs. The potential m6A modification sites on ICAM2 mRNA were predicted using the Sequence‐based RNA Adenosine Methylation site Predictor (SRAMP, https://www.cuilab.cn/sramp), which employs a whole‐transcriptome model and RNA secondary structure analysis (Figure 4A). MeRIP‐qPCR confirmed that the m6A abundance on ICAM2 mRNA was significantly increased in RA‐FLSs, and this increase was reduced by knocking down METTL3 (sh‐METTL3) (Figure 4B). RIP‐qPCR assays showed significant enrichment of ICAM2 mRNA in samples immunoprecipitated with an anti‐METTL3 antibody compared to control IgG (Figure 4C). A dual‐luciferase reporter assay demonstrated that knocking down METTL3 significantly reduced the luciferase activity of the WT‐ICAM2 reporter but had no significant effect on the MUT‐ICAM2 reporter (Figure 4D). mRNA stability analysis performed on transfected cells further revealed that METTL3 knockdown markedly promoted ICAM2 mRNA degradation, thereby reducing its stability (Figure 4E). Western blot analysis confirmed that ICAM2 protein expression was downregulated in METTL3‐deficient RA‐FLSs (Figure 4F). Together, these findings support METTL3‐dependent m6A regulation of ICAM2 mRNA stability. METTL3 maintains ICAM2 mRNA stability in an m6A‐dependent manner. (A) SRAMP‐predicted m6A‐containing region of the ICAM2 transcript. (B) MeRIP‐qPCR analysis of ICAM2 m6A enrichment. Enrichment was normalized to input and compared with the IgG control, then expressed as fold change relative to the Control group. (C) METTL3 RIP‐qPCR analysis of ICAM2 enrichment normalized to input and compared with IgG. MH7A cells were transfected with METTL3 short hairpin RNA (sh‐METTL3) or scrambled shRNA controls (sh‐NC) (D) Firefly luciferase activity normalized to Renilla luciferase activity after co‐transfection of WT or MUT reporters with sh‐METTL3 or sh‐NC. (E) ICAM2 mRNA remaining after actinomycin D treatment, expressed relative to the 0‐h time point. (F) Western blots and densitometric quantification of ICAM2 normalized to β‐actin. Data are mean ± SD from three independent biological experiments; technical replicates were averaged. Brackets indicate the specific group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant. CJEA‐Containing Serum Modulates RA‐FLS Phenotypes Partly through the METTL3‐m6A‐ICAM2 Axis Subsequent MeRIP‐qPCR analysis showed that CJEA‐containing serum significantly decreased ICAM2 mRNA m6A methylation (Figure 5A). RT‐qPCR and western blot analyses further demonstrated that CJEA‐containing serum markedly suppressed TNF‐α‐induced ICAM2 upregulation, an effect that was reversed by METTL3 overexpression (Figure 5B,C). Consistently, direct CJEA exposure also reduced ICAM2 protein levels (Figure S3B). To functionally validate ICAM2 involvement, TNF‐α‐induced MH7A cells were transfected with an ICAM2 overexpression plasmid or corresponding negative control, with transfection efficiency confirmed by western blot (Figure 5D). ICAM2 overexpression partially rescued the inhibitory effects of CJEA‐containing serum on RA‐FLS proliferation, migration, and invasion, and attenuated CJEA‐containing serum‐induced apoptosis (Figure 5E–I). A comparable reversal was observed for the suppression of inflammatory cytokine release (Figure 5J). Taken together, these data suggest that CJEA may attenuate the aggressive phenotype and inflammatory response in RA‐FLS, at least partly through METTL3‐mediated m6A methylation of ICAM2 mRNA. CJEA‐containing serum modulates RA‐FLS phenotypes partly through the METTL3‐m6A‐ICAM2 axis. (A) MeRIP‐qPCR analysis of ICAM2 m6A enrichment. Enrichment was normalized to input and compared with the IgG control, then expressed as fold change relative to the TNF‐α + blank serum (Blank) group. (B, C) Relative ICAM2 mRNA and protein expression after CJEA‐containing serum treatment with or without METTL3 overexpression. MH7A cells were transfected with ICAM2 overexpression vector (OE‐ICAM2) or empty vector (OE‐NC), treated with CJEA‐containing serum, and stimulated with TNF‐α. (D) Representative western blot confirming ICAM2 overexpression. (E) Cell viability (%) measured by CCK‐8 and expressed relative to the unstimulated control. (F) Representative EdU images and percentage of EdU‐positive nuclei. Scale bar: 100 µm. (G) Annexin V‐FITC/PI flow cytometry plots and quantification of total apoptosis. Total apoptosis rate was defined as the sum of early apoptotic cells (Annexin V‐positive/PI‐negative) percentage and late apoptotic or secondary necrotic cells (Annexin V‐positive/PI‐positive) percentage. At least 10,000 events were acquired per sample. (H) Representative wound‐healing images and wound closure percentage. Wound width was measured at 0 and 48 h after scratching; wound closure rate (%) = [(initial width − width at 48 h)/initial width] × 100. Three randomly selected fields per well were quantified. Scale bar: 100 µm. (I) Representative Matrigel‐Transwell images and number of invaded cells counted in five randomly selected non‐overlapping fields. Scale bar: 100 µm. (J) Concentrations of IL‐1β and IL‐6 in cell culture supernatants (pg/mL). The principal rescue comparison is CJEA + OE‐ICAM2 versus CJEA + OE‐NC. Data are mean ± SD from three independent biological experiments; technical replicates were averaged. Brackets indicate the specific group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant. CJEA Alleviates CIA in Mice We next evaluated therapeutic efficacy in a CIA mouse model. Both MTX and CJEA significantly reduced arthritis clinical scores (Figure 6A). Serum levels of IL‑1β and IL‑6 were markedly elevated in CIA mice and were suppressed by MTX or CJEA treatment (Figure 6B). Histological analysis (Figure 6C,D) revealed intact joint architecture in normal controls, whereas CIA mice showed pronounced synovial inflammation and cartilage damage, both of which were attenuated by MTX and CJEA. These results indicate that CJEA alleviates disease severity in CIA mice by reducing joint inflammation and structural damage. Regarding safety, no treatment‐related deaths or obvious behavioral abnormalities were observed, and all animals survived to the scheduled endpoint. Body weight remained stable, with no significant loss in the CIA + CJEA group compared with the CIA group (Figure 6E). Serum ALT and creatinine levels did not differ significantly between CJEA‐treated mice, untreated CIA mice, and normal controls (Figure 6F). CJEA alleviates CIA in mice. (A) Arthritis clinical scores assessed after the first immunization. (B) Serum levels of IL‐1β and IL‐6 measured by ELISA (pg/mL). (C) Representative H&E‐stained joint sections and quantitative analysis of synovial inflammation. Scale bar: 100 µm. (D) Representative Safranin O/Fast Green‐stained joint sections and quantitative analysis of cartilage damage. Scale bar: 100 µm. (E) Body weight recorded after the first immunization. (F) Serum ALT and creatinine levels. (G) Representative western blots and densitometric quantification of synovial METTL3 and ICAM2 normalized to β‐actin. Data are presented as mean ± SD. n = 6 mice per group; technical replicates were averaged. Brackets indicate the specific group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant. Notably, western blot analysis of synovial tissues from CIA mice further showed that CJEA downregulated the protein expression of both METTL3 and ICAM2 (Figure 6G), consistent with the trends observed in vitro, thereby reinforcing the role of the METTL3–ICAM2 axis in the anti‐RA mechanism of CJEA. Discussion Using TNF‐α‐stimulated MH7A cells, we found that serum obtained after administration of different C. japonicum fractions attenuated proliferation, migration, invasion, and inflammatory cytokine release while increasing apoptosis. CJEA‐containing serum produced the most pronounced effects among the tested fractions, and direct CJEA treatment produced effects consistent with those observed using CJEA‐containing serum. The differential bioactivity among extracts likely stems from their distinct chemical profiles governed by solvent polarity [34, 35]. The aqueous extract, enriched with hydrophilic constituents such as polysaccharides, proteins, and amino acids, primarily contributed to immunomodulation and baseline anti‐inflammatory effects. The ethanol extract, benefiting from the medium polarity of ethanol, captured a broader spectrum of compounds, including both water‐soluble components (e.g., flavonoid glycosides, phenylethanoid glycosides) and some medium polarity liposoluble components, resulting in more comprehensive anti‐inflammatory and antioxidant activities, albeit through relatively dispersed targets. In contrast, ethyl acetate, as a medium‐to‐low polarity solvent, further concentrated lipophilic constituents such as flavonoid aglycones, terpenoids, and aromatic acids. These components demonstrated superior membrane permeability, enabling efficient cellular uptake and interaction with critical intracellular targets, thereby conferring enhanced efficacy in suppressing inflammation and aberrant cell proliferation. Therefore, the anti‐RA bioactive principles in C. japonicum, particularly those capable of potently inhibiting RA‐FLS pathogenic activities, might be preferentially enriched in the ethyl acetate‐soluble fraction. Although the present study did not isolate individual active compounds, the HPLC fingerprint of the parent ethanol extract identified caffeic acid, acteoside, and kaempferol among the common peaks. Caffeic acid exhibits anti‐oxidative, anti‐inflammatory, anti‐tumor, and immunomodulatory properties and has shown anti‐inflammatory potential relevant to RA [36, 37]. Acteoside promotes B‐cell‐derived IL‐10 production and ameliorates autoimmunity [38]. Kaempferol, a natural flavonol, displays excellent anti‐inflammatory and immunomodulatory effects that are considered important for RA treatment [39, 40, 41]. Structurally, these or related phenylpropanoids and flavonoids are plausible contributors to the observed activity [42, 43]. This phytochemical information provides a foundation for the targeted isolation and characterization of specific anti‐RA compounds from this medicinal plant and prompted us to subsequently explore the molecular mechanisms underlying CJEA‐mediated anti‐RA effects. The m6A regulatory network has an increasingly recognized role in RA‐FLS activation [44, 45, 46]. METTL3 has been linked to NF‐κB signaling, AMIGO2 regulation, synovitis, bone erosion, and ferroptosis resistance [47, 48, 49], whereas METTL14 and ALKBH5 regulate additional pathogenic transcripts and phenotypes [50, 51, 52, 53]. Consistent with previous reports [48], we observed an increase in global m6A methylation in TNF‐α‐stimulated MH7A model, accompanied by elevated expression of the key m6A regulators METTL3 and WTAP [54, 55, 56]. Notably, CJEA treatment significantly suppressed METTL3 expression, while showing little effect on the other m6A enzymes examined. Functional rescue experiments further demonstrated that overexpression of METTL3 counteracted the inhibitory effects of CJEA on the tumor‐like characteristics and inflammatory responses of RA‐FLS, suggesting that METTL3 is a critical mediator of CJEA action. Although the initial screen focused on selected writers and erasers to identify changes in m6A deposition, reader proteins ultimately determine the fate of methylated transcripts. Published RA studies implicate readers such as YTHDF2 and IGF2BP2 in transcript stability and pathogenic FLS behavior [57, 58]. The present study did not identify the reader that recognizes methylated ICAM2 mRNA. Accordingly, reduced ICAM2 stability after METTL3 knockdown supports an m6A‐associated mechanism but does not define the complete writer–reader pathway. Future experiments should test candidate YTH‐domain and IGF2BP proteins using loss‐ and gain‐of‐function approaches and transcript‐specific RIP assays. ICAM2, also known as CD102, is a transmembrane glycoprotein belonging to the immunoglobulin superfamily. It facilitates leukocyte adhesion by binding to integrins LFA‐1 (CD11a/CD18) and MAC‐1 (CD11b/CD18) on immune cells, thereby participating in essential immunological processes including T‐cell activation, neutrophil migration, natural killer cell cytotoxicity, and cellular proliferation. Through these mechanisms, ICAM2 plays a pivotal role in regulating immune responses and inflammatory pathways [59]. Previous work implicated METTL3‐dependent m6A regulation of ICAM2 mRNA in the response of RA‐FLSs to artemisitene [33]. Consistent with that framework, our data support ICAM2 as a functional downstream component of METTL3 in the present model. However, partial rescue by METTL3 or ICAM2 overexpression indicates that other pathways may also contribute. From a translational perspective, oral CJEA significantly attenuated arthritis progression in CIA mice, as evidenced by reduced joint inflammation and improved synovial pathology, with therapeutic effects broadly similar in direction to those observed with MTX. Nevertheless, these findings should be interpreted in light of several limitations. First, CJEA remains a chemically incompletely defined fraction. The available HPLC fingerprint was generated from the parent ethanol extract rather than the final ethyl acetate fraction, and the specific constituent or combination of constituents responsible for the observed activity has not yet been identified. Second, although direct CJEA treatment supported an intrinsic effect on RA‐FLSs, the drug‐containing serum approach cannot distinguish among the contributions of prototype compounds, circulating metabolites, and serum‐mediated systemic changes. In addition, the use of pooled sera prevented assessment of inter‐animal variability and batch‐to‐batch differences. Third, the in vitro experiments were conducted primarily in the MH7A cell line and have not yet been validated in primary RA‐FLSs derived from patients, which limits the translational generalizability of the findings. Fourth, the mechanistic data support the involvement of the METTL3‐m6A‐ICAM2 axis but do not demonstrate direct target engagement by CJEA or by any defined constituent. Fifth, CJEA and MTX were administered via different routes, which limits direct pharmacological comparison between the two treatments. Moreover, the CJEA dose was expressed as a crude‐herb‐equivalent dose for preclinical efficacy evaluation and should not be interpreted as directly equivalent to a clinically applicable human dose. Finally, the current safety assessment was limited to short‐term tolerability, body‐weight monitoring, survival, general observations, and selected serum biochemical indicators. Comprehensive pharmacokinetic, dose‐response, repeated‐dose toxicity, and long‐term safety studies remain necessary. Therefore, the present study should be regarded as an initial step toward elucidating the anti‐rheumatic potential of C. japonicum. Further work is required to isolate and characterize the active constituents, validate the METTL3‐m6A‐ICAM2 pathway in primary patient‐derived RA‐FLSs, and establish the pharmacological, toxicological, and translational profiles of CJEA before clinical application can be considered. Conclusion In conclusion, CJEA attenuated the pathogenic phenotype of RA‐FLS in vitro and alleviated disease severity in CIA mice. These effects may be mediated, at least in part, through METTL3‐dependent m6A modification of ICAM2 mRNA. The study identifies a potentially relevant regulatory axis but does not define a direct molecular target or active compound. Further investigation is needed to isolate the active principles and validate these findings in more physiologically relevant models. Author Contributions Ying Huang, JiangCun Wei, Yan Ma, and XiaoPing Mei designed the research study. XiaoPing Mei, MingTao Huang, and Liang Lu performed the research. JianLong Zhou and ZuJie Qin provided help and advice on the experiments. MingTao Huang and Liang Lu analyzed the data. Ying Huang, JiangCun Wei, and Yan Ma wrote the manuscript. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. Ethics Statement All animal experiments were reviewed and approved by the Animal Experimental Ethics Committee of Guangxi University of Chinese Medicine (Approval No. DW20250924‐01). All procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and are reported in accordance with the ARRIVE guidelines. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Figure S1: HPLC fingerprint of the 75% ethanol extract of Clerodendrum japonicum roots. (A) Overlaid chromatograms of ten batches and the generated reference fingerprint. (B) Reference fingerprint showing 12 common peaks. (C) Chromatogram of the mixed reference standards. Peaks 2, 4, and 12 were assigned as caffeic acid, acteoside, and kaempferol, respectively; peak 4 (acteoside) was used as the reference peak. Detection wavelength: 320 nm. Figure S2: Direct CJEA treatment attenuates pathogenic phenotypes and inflammatory cytokine release in RA‐FLS. MH7A cells were assigned to the following groups: unstimulated control (Ctrl), TNF‐α + DMSO (Vehicle), TNF‐α + the indicated concentrations of CJEA (25, 50, or 100 μg/mL), and TNF‐α + methotrexate (MTX). DMSO served as the vehicle control for extract‐containing serum groups; MTX served as the positive control. (A) Cell viability (%) measured by CCK‐8 and expressed relative to the unstimulated control. (B) Representative EdU images and percentage of EdU‐positive nuclei. Scale bar: 100 µm. (C) Annexin V‐FITC/PI flow cytometry plots and quantification of total apoptosis. Total apoptosis rate was defined as the sum of early apoptotic cells (Annexin V‐positive/PI‐negative) percentage and late apoptotic or secondary necrotic cells (Annexin V‐positive/PI‐positive) percentage. At least 10,000 events were acquired per sample. (D) Representative wound‐healing images and wound closure percentage. Wound width was measured at 0 h and 48 h after scratching; wound closure rate (%) = [(initial width − width at 48 h)/initial width] × 100. Three randomly selected fields per well were quantified. Scale bar: 100 µm. (E) Representative Matrigel‐Transwell images and number of invaded cells counted in five randomly selected non‐overlapping fields. Scale bar: 100 µm. (F) Concentrations of IL‐1β and IL‐6 in cell culture supernatants (pg/mL). Data are mean ± SD from three independent biological experiments; technical replicates were averaged. Brackets indicate the specific group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant. Figure S3: Direct CJEA treatment reduces METTL3 and ICAM2 protein expression in RA‐FLS. MH7A cells were assigned to the following groups: unstimulated control (Ctrl), TNF‐α + DMSO (Vehicle), TNF‐α + the indicated concentrations of CJEA (25, 50, or 100 μg/mL), and TNF‐α + methotrexate (MTX). DMSO served as the vehicle control for extract‐containing serum groups; MTX served as the positive control. (A) Western blot analysis of METTL3. (B) Western blot analysis of ICAM2. Data are mean ± SD from three independent biological experiments; technical replicates were averaged. Brackets indicate the specific group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant. Table S1: Similarity of individual batches to the reference HPLC fingerprint. Table S2: Validation of the HPLC fingerprint method. Table S3: shRNA sequences used for gene knockdown. Table S4: Plasmid sequences used for gene overexpression. Acknowledgments This work was funded by the National Natural Science Foundation of China (No. 82460825). Youth Science Fund Project of Guangxi Natural Science Foundation (No. 2023GXNSFBA026072). High‐Level Key Disciplines in Traditional Chinese Medicine Construction Project by the National Administration of Traditional Chinese Medicine‐Ethnic Minority Pharmacy (Zhuang Medicine) (No. zyyzdxk‐2023165). National Project for the Construction of Senior Pharmacists' Inheritance Studios by the National Administration of Traditional Chinese Medicine [Zhong Wen] (Guo Zhong Yao Ren Jiao Han [2024] No. 255). Three‐year Action Plan Project for High‐level Talent Team Building of Guangxi International Zhuang Medicine Hospital in 2023 (No. GZCX20231203). The Third Batch of “Qihuang Project” High‐level Talent Team Cultivation Project of Guangxi University of Chinese Medicine (No. 202414). The discipline of “Pharmacology of Ethnic Medicines” at Guangxi International Zhuang Medicine Hospital. Huang Y., Wei J., Ma Y., et al., “ Clerodendrum japonicum Extract Suppresses the Pathogenicity of Rheumatoid Arthritis Fibroblast‐Like Synoviocytes by Inhibiting METTL3‐Mediated ICAM2 mRNA Methylation,” Cell Biochemistry and Function 44 (2026): e70303, 10.1002/cbf.70303. Ying Huang, JiangCun Wei, Yan Ma, and XiaoPing Mei contributed equally to this study. Contributor Information JianLong Zhou, Email: zzjianlong034@outlook.com. ZuJie Qin, Email: ZuJie63Qin@outlook.com. Data Availability Statement The datasets used and/or analyzed during the present study are available from the corresponding author on reasonable request. References Associated Data Supplementary Materials Figure S1: HPLC fingerprint of the 75% ethanol extract of Clerodendrum japonicum roots. (A) Overlaid chromatograms of ten batches and the generated reference fingerprint. (B) Reference fingerprint showing 12 common peaks. (C) Chromatogram of the mixed reference standards. Peaks 2, 4, and 12 were assigned as caffeic acid, acteoside, and kaempferol, respectively; peak 4 (acteoside) was used as the reference peak. Detection wavelength: 320 nm. Figure S2: Direct CJEA treatment attenuates pathogenic phenotypes and inflammatory cytokine release in RA‐FLS. MH7A cells were assigned to the following groups: unstimulated control (Ctrl), TNF‐α + DMSO (Vehicle), TNF‐α + the indicated concentrations of CJEA (25, 50, or 100 μg/mL), and TNF‐α + methotrexate (MTX). DMSO served as the vehicle control for extract‐containing serum groups; MTX served as the positive control. (A) Cell viability (%) measured by CCK‐8 and expressed relative to the unstimulated control. (B) Representative EdU images and percentage of EdU‐positive nuclei. Scale bar: 100 µm. (C) Annexin V‐FITC/PI flow cytometry plots and quantification of total apoptosis. Total apoptosis rate was defined as the sum of early apoptotic cells (Annexin V‐positive/PI‐negative) percentage and late apoptotic or secondary necrotic cells (Annexin V‐positive/PI‐positive) percentage. At least 10,000 events were acquired per sample. (D) Representative wound‐healing images and wound closure percentage. Wound width was measured at 0 h and 48 h after scratching; wound closure rate (%) = [(initial width − width at 48 h)/initial width] × 100. Three randomly selected fields per well were quantified. Scale bar: 100 µm. (E) Representative Matrigel‐Transwell images and number of invaded cells counted in five randomly selected non‐overlapping fields. Scale bar: 100 µm. (F) Concentrations of IL‐1β and IL‐6 in cell culture supernatants (pg/mL). Data are mean ± SD from three independent biological experiments; technical replicates were averaged. Brackets indicate the specific group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant. Figure S3: Direct CJEA treatment reduces METTL3 and ICAM2 protein expression in RA‐FLS. MH7A cells were assigned to the following groups: unstimulated control (Ctrl), TNF‐α + DMSO (Vehicle), TNF‐α + the indicated concentrations of CJEA (25, 50, or 100 μg/mL), and TNF‐α + methotrexate (MTX). DMSO served as the vehicle control for extract‐containing serum groups; MTX served as the positive control. (A) Western blot analysis of METTL3. (B) Western blot analysis of ICAM2. Data are mean ± SD from three independent biological experiments; technical replicates were averaged. Brackets indicate the specific group comparisons. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant. Table S1: Similarity of individual batches to the reference HPLC fingerprint. Table S2: Validation of the HPLC fingerprint method. Table S3: shRNA sequences used for gene knockdown. Table S4: Plasmid sequences used for gene overexpression. Data Availability Statement The datasets used and/or analyzed during the present study are available from the corresponding author on reasonable request.