Human FLSs were extracted from patients with RA undergoing total joint replacement. All patients with RA met the American College of Rheumatology 1987 revised criteria. FLSs were cultured in high-glucose Dulbecco's Modified Eagle Medium (DMEM) without pyruvate, supplemented with 10% fetal bovine serum (FBS), L-glutamine (8 mM; 4 mM for mice FLS), and penicillin/streptomycin (100 U/ml) (Gibco, ThermoFisher Scientific, Inc., Whaltham, MA).

Cell line Human embryonic kidney 293A (HEK293A) was purchased from ThermoFisher Scientific, Inc. (Whaltham, MA), and was cultured in high-glucose DMEM without pyruvate, supplemented with 10% FBS, L-glutamine (8 mM), and penicillin/streptomycin (100 U/ml) (Gibco, ThermoFisher Scientific, Inc., Whaltham, MA). Human peripheral blood mononuclear cells (PBMCs) were isolated from whole blood by Ficoll density gradient using Ficoll Plaque Premium (Sigma GE Healthcare, #17–544-02) or BD Vacutainer CPT Tubes (BD, #362753) as described somewhere else.

Platelet derived growth factor subunit BB (PDGF)-BB) was obtained from R&D Systems, Inc. (Minneapolis, MN). Methyl jasmonate (MJ), CoCl₂, methotrexate (MTX), and tofacitinib (Tofa) were purchased from Sigma-Aldrich (Sant Louis, MO).

Ad-mHK2 FL, Ad-mHK2ΔN (N-terminal deletion mutant), Ad-15G (association to mitochondria peptide), and Ad-GFP were developed as previously described (40). Supernatants were collected and freeze-thawed three times for cell disruption and 0.22-µm–filtered for cell debris elimination. Titration of adenoviral vectors was assessed by cell lysis assay with HEK293A cells to obtain the multiplicity of infection (MOI) quantification. Overexpression experiments were performed by infecting FLS with 500 MOI. Experiments were performed 48 h after infection except specified otherwise.

FLSs were plated 100,000 cells per well and were disrupted in lysis buffer after treatment and collection. Proteins were separated by Sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to a nitrocellulose membrane through the semi-dry system (Invitrogen). After blocking for 1 h with 5% milk, blots were incubated overnight at 4°C with the following antibodies: rabbit anti-human/mouse HK2 at 1:500 (#2867, C64G5, Cell Signaling Technology, Danvers, MA), mouse anti-human HK2 at 1:500 (sc-374091, B8, Santa Cruz Co, Santa Clara, CA), mouse anti-human tubulin at 1:1,000 (Santa Cruz Co., Santa Clara, CA), rabbit anti-TOM20 at 1:500 (Cell Signaling Technology, Danvers, MA), and Glyceraldehyde-3-Phosphate Dehydrogenase (GAPDH) and tubulin (Santa Cruz Co., Santa Clara, CA). Horseradish peroxidase–conjugated anti-mouse Immunoglobulin G (IgG) or anti-rabbit IgG (Cell Signaling Technology, Danvers, MA) was used as secondary antibody at 1:2,000 dilution. Membranes were developed using a chemiluminescence system (Clarity Western ECL, BIORAD, Hercules, CA). Images were captured by the ChemiDoc™ XRS+ System (BIORAD, Hercules, CA).

RA FLSs were plated to 70% confluence in two 143-cm² plates per condition and stimulated according to the figure legends. Mitochondrial protein was isolated using a mitochondria isolation kit (#89801 from ThermoFisher Scientific). Mitochondrial fraction was lysed in 2% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS) in Tris-Buffered Saline (TBS) buffer and quantified using Bradford. All subsequent steps are according to the Western blot protocol mentioned above.

RA FLSs were plated 100,000 per well in a six-well plate. After treatment with MJ or vehicle, cells were collected, and cell fractions were obtained as above but diluted according to a cytochrome c quantikine ELISA kit (R&D Biosystems, #DCTC0), and ELISA was performed according to the manufacturer’s directions.

RA FLS and PBMCs were collected and diluted to 1 × 10⁶ cells/ml, and HK activity assay was performed according to the manufacturer’s instructions (Sigma-Aldrich, #MAK091).

Cells (8 × 10³) per well of RA FLS were plated on Lab-Tek brand chamber slides (n = 3) and were either incubated with CoCl₂ (150 µM), PDGF (20 ng/ml), MTX (1 µM), and Tofa (1 µM) or infected with Ad-GFP, Ad-HK2, and Ad-HK2ΔN. Cells were fixed with 4% formaldehyde, permeabilized with 0.1% Triton X-100, and then blocked with 1% donkey serum in Phosphate buffered saline (PBS) for 1 h at Reverse transcriptase (RT). Then, cells were stained with anti-ATP5B (MAB3494, Millipore) to visualize the mitochondria and anti-HK2 (NBP2-16814, Novus) in blocking solution overnight at +4°C, followed by anti-mouse Alexa Fluor 594 and anti-rabbit Alexa Fluor 488 for 1 h at RT. Nuclei were counterstained with 4',6-diamidino-2-phenylindole (DAPI) and mounted with Fluosave. Cells were imaged at ×40 magnification using a DM6000 Leica microscope, images were analyzed, and HK2 colocation in the mitochondria was quantified using ImageJ Fiji software.

One million FLSs were trypsinized with 0.05% trypsin (Gibco, ThermoFisher Scientific, Inc., Whaltham, MA) and resuspended in 40 µl of 1% FBS DMEM and 40 µl of Matrigel (356231, BD Biosciences, San Diego, CA). Then, 4-µl drops of the mixture were plated in a six-well plate. After 5 min of incubation at 37°C, medium was added in the presence or absence of PDGF (10 ng/ml). The assay was stopped 24 h later, by adding 4% paraformaldehyde (PFA). Cells were stained with 0.05% crystal violet (Sigma-Aldrich, San Luis, MO), and images were captured with a 4× objective inverted microscope. For quantification, four images were taken per condition, and invasive area was measured by ImageJ software using arbitrary units.

FLSs were seeded onto six-well plates and allowed to come to confluence. A double–cross-scratch wound was done in each well with the end of a sterile pipette tip. Cells were subsequently exposed to 1% FBS DMEM in the presence or absence of PDGF (10 ng/ml) for 24 h. FLSs were fixed by adding 4% PFA. Crystal violet was used at a concentration of 0.05% in distilled water for 30 min. After staining, cells were washed three times with water, and FLS migration across the wound margins was assessed, photographed, and measured by ImageJ—the average of relative length between random spots in all eight arms of the two cross sections.

Cells were plated at 50,000 cells per well in a 12-well plate. Cells were starved in 0.1% medium and, the following day, given MJ (1 µM) for 1 h before adding H₂O₂ to final concentrations of 75 or 200 µM. After 4 h, cells were fixed by adding 4% PFA. Cells were stained with 0.05% crystal violet (Sigma-Aldrich, San Luis, MO). Images were captured with a 10× objective inverted microscope. For quantification, mean gray value was measured through ImageJ software as described (41).

FLSs were plated at 100,000 cells per well in a six-well plate using three-well replicates per condition. Cells were allowed to adhere overnight and then incubated 6 h with either ethanol (EtOH, vehicle) or MJ (1.5 mM). After the incubation time, cells were collected with TRIzol reagent (Invitrogen, 15596026). RNA was then extracted using chloroform, isopropyl alcohol, and 70% ethanol before adding 500 ng of each sample to do a reverse transcriptase reaction with the Bio-Rad iScript cDNA synthesis kit #1708891. The complementary DNA (cDNA) samples were then ran for real-time polymerase chain reaction (PCR) using primers from Integrated DNA technologies (IDT) and SsoAdvanced Universal SYBR green super mix from Bio-Rad Laboratories, Inc. Plates were analyzed with a Thermo Fisher QuantStudio 3 Real-Time PCR Systems machine. Relative mRNA was calculated in comparison with 18S gene as housekeeping. Primer sequences are available upon request.

FLSs were plated at 3,000 cells per well in a 96-well plate using three-well replicates per condition per cell line. Conditions were added comparing 1% and 10% FBS. After 24 h with treatment, 10% 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) was added for 4 h and dissolved in 200 µl of Dimethyl Sulfoxide (DMSO). Optical density was measured at 550 and 690 nm, and replicates were averaged per condition and cell line.

The C57BL/6 mice were obtained from Charles River. All mice utilized for experimental procedures were 8- to 10-week-old male mice. Single animals were regarded as experimental entities. Induction of serum transfer-induced arthritis (STIA) was done by intravenous injection consisting of 100 μl of serum derived from KRN mice (K/BxN). Bones with intact joints from hind limbs of day 0, peak days 7–9, resolving day 15, or resolved day 22 and STIA inflamed mouse joints (n = 3 biological replicate samples, each consisted of cells isolated from the joints of three animals) were dissected and transferred into RPMI-1640 (+2% Fetal calf serum (FCS)) containing collagenase D (0.1 g/ml; Roche) and DNase I (0.01 g/ml; Sigma-Aldrich). Samples were incubated at 37°C for 40 min, followed by incubation with medium containing collagenase dispase (0.1 g/ml; Roche) and Deoxyribonuclease I (DNase I) (0.01 g/ml) at 37°C for 20 min. Live CD45⁻ synovial cells were sort purified using the BD FACSaria and captured with the 10x Genomics Chromium system. Sequencing libraries were generated using the 10x Genomics Single Cell 3′ Solution (version 2) kit and subjected to Illumina sequencing (HiSeq 4000, read 2 sequenced to 75 bp). Alignment to mm10 was completed using CellRanger (10x Genomics, v2.1.1). Analysis was completed using and R (v4.1) and using Seurat (v4.0.3) (42). Data were then analyzed in Seurat, where the following quality control (QC) metrics were used for all samples: nFeature_RNA > 500, nFeature_RNA <5,000, nCount_RNA > 500, and nCount_RNA < 20,000 (except PEAK_A, PEAK_B, PEAK_C, RESING_A, RESING_B, and RESING_C, where the maximum nCount_RNA was 40,000, 40,000, 40,000, 25,000, 30,000, and 30,000, respectively). Differential expression between clusters was calculated using FindAllMArkers() and FindMarkers() (between time points) in Seurat. Gene ontology (GO) term analysis was completed using gsFisher (v0.2), and Volcano plots were plotted using EnahcedVolcano (v1.13.2) R packages. Murine data are available at GSE230145.

KRN T cell receptor–transgenic mice were a gift from Drs. D. Mathis (Harvard Medical School, Boston, MA) and C. Benoist (Institut de Génétique et de Biologie Moléculaire et Cellulaire, Strasbourg, France). All mice used in these experiments were bred on the C57BL/6 background and were 8–12 weeks old. All protocols involving animals received prior approval from the Institutional Review Boards and followed the Guide for the Care and Use of Laboratory Animals from the Institute for Laboratory Animal Research (National Research Council).

Sera from adult, arthritic K/BxN mice were pooled for use in serum transfer. Recipient mice were injected intraperitoneally with 150 µl of the K/BxN mouse serum. One hour before serum injection (d0) or 5 days after serum injection (d5), MJ or vehicle was injected into the mouse at either 25 or 100 mg/kg. Clinical arthritis scores were evaluated in the recipient mice after serum transfer, as described previously (43). On day 12 after serum injection, the mice were euthanized, and their paws were analyzed for histopathological changes.

A total of 1 × 10¹⁰ viral particles in 8 µl were injected in the knees of control C57BL/6 mice. Ad-HK2FL was injected in the right knee, whereas Ad-HK2ΔN was injected in the left knee. Fourteen days after the injection, mice were euthanized, and knees were collected for histological studies and quantified as below.

Experimental antigen-induced arthritis (AIA) was induced by a subcutaneous injection of 100 μg of methylated bovine serum albumin (mBSA) emulsified in 100 μl of complete Freund’s adjuvant (CFA) in the flank and, 1 week later, by an intradermal injection of 100 μg of mBSA/CFA in the tail base. Two weeks after these injections, arthritis was induced by intraarticular injection of 60 μg of mBSA in 10 μl of saline into the knee joint. After 3 days, a total of 1 × 10¹⁰ viral particles in 8 µl of Ad-GFP and Ad-15G were administered with intraarticular injection in the knees. Disease was assessed 14 days after intraarticular injection (11 days after adenovirus injection) by histological analysis as described below.

Joints were fixed in 10% formalin, decalcified in 10% EDTA for 2–3 weeks, and paraffin-embedded. Paraffin-embedded human synovium and mice joints sections were stained with hematoxylin and eosin and safranin O (43). A blinded semiquantitative scoring system was used to assess synovial inflammation, synovial hypertrophy, bone erosion, and cartilage damage (scale of 0–5) as previously described (43).

Statistical analysis was performed with Prism software (version 5; GraphPad). Results are expressed as the mean ± SEM. Normality of the variables was assessed using the Kolmogorov–Smirnov and D’Agostino–Pearson normality tests. For comparison between two groups, Student’s two-tailed unpaired t-tests or nonparametric Mann–Whitney tests were applied, depending on the normality of the distribution of the variables. We compared three or more groups with analysis of variance (ANOVA) using Kruskal–Wallis non-parametric test followed by Dunnett’s post-hoc test or with one-way ANOVA parametric test followed by Tukey’s post-hoc test, depending on the homogeneity of variances. Results were considered significant if the two-sided P-value was less than 0.05. In all the figures, * is p<0.05, ** is p <0.01, *** is p <0.005, **** is p<0.001.

We first determine whether HK2 binds to the mitochondria after stimulation with inflammatory mediators implicated in FLS responses. Stimulation of RA FLS with CoCl₂, which mimics hypoxia, and PDGF for 60 min triggered HK2 mitochondrial binding but not its HK activity ( Supplementary Figure 1A ), as detected by an overlap of HK2 in green and ATP5b (a marker of mitochondria) in red by immunofluorescent confocal microscopy ( Figures 1A, B ). Binding of HK2 to mitochondria is critical for RA FLS aggressive behavior

HK2 has motif at the N-terminal that allows binding to mitochondria (schematic representation in Supplementary Figure 1B ). We determined the role of the binding motif in FLS phenotype, by transducing the FLS with adenovirus carrying FL murine HK2, Ad-HK2 (FL), which has both kinase activity and mitochondrial binding motifs, and a truncated version of HK2, Ad-HK2ΔN, which has kinase activity but lacks of mitochondrial binding motif, which prevents HK2 from binding to the mitochondria. Adenovirus with GFP was used as control. Supplementary Figure 1C exhibits the upregulation of HK2, but not endogenous HK2, protein after infection with Ad-HK2FL and Ad-HK2ΔN. On the other hand, confocal immunofluorescent studies ( Supplementary Figure 1D ) indicate that there is a significant binding of HK2 to the mitochondria when incubated with Ad-HK2FL adenovirus rather than Ad-HK2ΔN adenovirus. Figures 2A-D demonstrate the potential role of HK2FL in aggressive behavior of RA FLS by looking at invasion ( Figures 2A, B ) and migration ( Figures 2C, D ). RA FLS invaded and migrated after infection with Ad-HK2FL and PDGF stimulation in comparison with GFP. This phenotype was not observed after the infection with Ad-HK2ΔN, suggesting that HK2 binding to the mitochondria plays a role in RA FLS aggressive phenotype.

Several drugs have shown to dissociate HK2 from mitochondria (44–46); among them, MJ is a plant stress hormone shown to dissociate HK2 from mitochondria in cancer cells as well as suppress proliferation and induce cell death (47, 48). Results from MTT viability assays revealed that MJ was nontoxic to RA FLS at a wide range of concentrations ( Supplementary Figure 2A ). In addition, MJ did not lead to a significant dissociation of cytochrome c from the mitochondria to the cytoplasm, which would be an indication of apoptosis ( Supplementary Figure 2B ). Mitochondrial fractionation protein analysis shows that HK2 expression in the mitochondrial fraction decreased after 60 min of incubation with MJ ( Supplementary Figure 2C ), which indicates MJ’s potential for dissociating HK2 from the mitochondria.

We next looked at the effect of MJ on invasion and migration, two important characteristics in RA FLS phenotype. In RA FLS, MJ significantly reduced both RA FLS invasion and migration after PDGF stimulation ( Figures 3A-D ). Because RA FLSs are also known to be resistant to apoptosis, we then conducted a death assay to determine the RA FLS resistance after addition of MJ ( Figures 3E, F ). MJ was found to potentiate H₂O₂-induced apoptotic cell death in comparison to EtOH control.

We also determined HK2 adherence to the mitochondria under the presence of approved drugs in the treatment of RA, which include MTX and Tofa. Using immunofluorescent confocal microscopy, the addition of Tofa but not MTX decreased the binding of HK2 to the mitochondria after CoCl₂ stimulation ( Figure 4 ). However, Tofa was not able to avoid the HK2 translocation after adenovirus overexpression of HK2 ( Supplementary Figure 3 ).

We next identified genes associated to HK2 synovial fibroblasts by examining scRNA-seq data from the STIA model. We first digested synovial tissue from hind limbs from mice at different time points of the disease stage (rest day 0, peak days 7–9, resolving day 15, and resolved day 22; Supplementary Figure 4A ). Live CD45⁻ synovial cells were sort purified. 3′ mRNA sequencing generated 20,216 cells that passed QC ( Figure 5B , Supplementary Figure 4B ). We then annotated the main cell types ( Figure 5A ) based on marker gene expression ( Supplementary Figures 4C–H ) and subset fibroblasts ( Figure 5B , Supplementary Figure 4I ), which included the lining layer ( Supplementary Figure 4K ). Of interest, fibroblasts were the CD45⁻ synovial cells with higher Hk2 expression ( Figure 5A ). We then selected Hk2 positive (cell with >0 Hk2 gene counts) and Hk2 negative (cells with 0 Hk2 gene counts) and examined marker gene expression. We found that Hk2-positive cells were enriched in pro-inflammatory molecules when compared with Hk2-negative cells ( Figure 5C , Supplementary Figure 4J ). We then examined differential gene expression across disease time points in Hk2-positive fibroblast and found a number of genes enriched in each time point when compared with rest ( Figure 5D ). Further examination of the differentially expressed genes revealed an upregulation of key inflammatory and matrix remodeling genes at peak (Ccl2, Timp1, Cxcl1, Postn, Col1a1, Col3a1, Acta2, Mmp3, and Cxcl2: Figure 5E ). Many of these genes retained their upregulation in resolving, and most were not significantly differentially expressed in the resolved time point compared with rest ( Figure 5E ). We then sought to validate the role of the mitochondrial binding in the upregulation of these genes in human RA FLS. MJ decreased the expression of several of the genes associated to HK2 expression ( Figure 5F ).

We finally determined the effect of HK2 with and without binding motif on arthritis in experiments in vivo. Adenovirus injection was given on each knee of a C57BL/6 mouse, with Ad-HK2FL on one knee and Ad-HK2ΔN into the other knee ( Figures 6A, B ). After injection with adenovirus particles, mice were euthanized after 14 days and analyzed with histological staining. Knees injected with Ad-FLHK2 had higher synovial hypertrophy and inflammation scores as compared with the Ad-ΔNHK2 ( Figures 6A, B ).

Selective dissociation of HK2 from mitochondria was achieved using Ad-15G, an adenovirus encoding the first 15 amino acids of the mitochondrial binding motif (schematic is displayed in Supplementary Figure 1A ). This competitively inhibits endogenous HK2 binding to mitochondria (40). In an AIA model of arthritis, the injection of Ad-15G reduced histological scores compared with Ad-GFP injection ( Figures 6C, D ).

To determine the effect in arthritis of dissociation of HK2 from mitochondria, MJ was injected either same day as K/BxN serum (d0) ( Figures 7A-C ) or 5 days ( Figures 7D, E ) in the K/BxN arthritis model, using doses of 25 and 100 mg/kg. MJ injection at day 0 served a preventative method of arthritis progression, whereas injection at day 5 demonstrated a treatment option at the peak of disease activity. Clinical score and joint diameter were significantly reduced in a dosage at 25 mg/kg over a dosage at 100mg/kg at both preventing arthritis (d0) and treating arthritis (d5). Interestingly, dosage at 100 mg/kg had increased arthritic clinical score compared with dosage at 25 mg/kg but still less than control, suggesting a reverse effect at higher doses (representative images are shown in Figure 7C ). Histological analysis revealed synovial hypertrophy, inflammation, cartilage damage, and bone damage scores were reduced in 25 mg/kg compared with control but not in 100 mg/kg in both d0 and d5 injections but was more prominent in d0.