Galectin-9 inhibition of the MIF-CD74/CD44 pathway suppresses chronic arthritis

Abstract

The destructive potential of rheumatoid arthritis (RA) lies in the aggressive behavior of fibroblast-like synoviocytes (FLSs), which actively contribute to the erosion of cartilage and bone and may persist even in the face of apparent clinical remission. Therapeutic approaches targeting RA-FLSs have been developed to treat RA; however, there are no clinically approved drugs available at present. Here, single-cell RNA sequencing of RA-FLSs identified a distinct macrophage migration inhibitory factor (MIF)^high subset with mitochondrial and endoplasmic reticulum dysfunction. MIF^high conditions led to increased survival, proliferation, and migration of FLSs, along with the upregulation of CD44 and the CD44v6 isoform expression. We next explored whether a stable, recombinant form of galectin-9 (sGal-9), which acts as a CD44 blockade, regulates the MIF-induced aggressive phenotype of RA-FLSs. We found that sGal-9 remarkably reduced the increased proliferation, migration, and invasion of RA-FLSs by inhibiting the MIF-CD44 pathway. Moreover, both local and systemic administration of sGal-9 substantially inhibited excessive cartilage and bone destruction by RA-FLSs in a xenotransplantation arthritis model and alleviated the severity of collagen-induced arthritis in mice, comparable to Enbrel and tofacitinib. Conclusively, these data suggest that sGal-9 is effective at repressing destructive phenotypes of RA-FLSs as a novel anti-MIF agent.

Graphical abstract

Fibroblast-like synoviocytes (FLSs) are key drivers of rheumatoid arthritis pathology, yet no targeted therapy is available. Kim and colleagues identify a distinct MIF^high subset of FLSs and show that galectin-9, an MIF-pathway inhibitor, suppresses their aggressive behavior and alleviates chronic arthritis in mice, supporting its potential as a novel therapy.

Introduction

Rheumatoid arthritis (RA) is a chronic autoimmune disease characterized by inflammation of the synovial joints. A hallmark feature of RA pathology is the formation of a synovial pannus, an invasive and proliferative tissue consisting of fibroblast-like synoviocytes (FLSs), immune cells, and newly formed vessels.^1,2 The destructive potential of RA lies in the aggressive behavior of FLSs,^2,3 which actively contribute to the erosion of cartilage and bone within affected joints and may persist even in the face of apparent clinical remission.^2,3,4 The FLSs of RA patients (RA-FLSs) show resistance to apoptotic death, enhanced adhesive properties, and invasive capabilities reminiscent of tumor cells.^2,5 Moreover, these features persist ex vivo and likely reflect the influence of long-term epigenetic modification.³ RA-FLSs secrete an array of mediators, including cytokines, prostaglandins, and matrix metalloproteinases, leading directly to joint destruction.^2,6 Therefore, understanding the intricate molecular and cellular mechanisms driven by RA-FLSs is crucial for the development of targeted therapies to halt disease progression and improve patient outcomes.

High expression alleles of the innate cytokine macrophage migration inhibitory factor (MIF) have been linked to more severe RA pathology.⁷ MIF is expressed by various cell types, including FLSs, promoting activation and survival of monocytes/macrophages.^8,9 Neutralization of MIF with an anti-MIF antibody or genetic deletion can effectively impede progression of inflammatory arthritis across multiple animal models of RA.^10,11 In patients with RA, MIF levels are elevated in the sera and synovial fluids.^12,13 Moreover, MIF perpetuates inflammation cascades in RA joints by sustaining mitogen-activated protein kinase (MAPK) activation, hampering the pro-apoptotic function of p53, and promoting the production of arachidonic acid.⁷

The initiation of MIF-mediated signal transduction occurs through its binding to the cell surface receptor CD74, leading to intracytoplasmic phosphorylation of its co-receptor CD44 and subsequent activation of Src-family kinases.^14,15 CD44, a polymorphic glycoprotein, mediates cell-cell adhesion and interactions with the extracellular matrix, and has been implicated in cellular homing, tumor invasiveness, metastasis, and angiogenesis.^16,17,18 The expression of CD44 is elevated in rheumatoid synovium and is upregulated in RA-FLSs with high-genotypic MIF expression.¹⁹ The MIF-CD74/CD44 axis thus facilitates the adhesion and invasion of RA-FLSs.¹⁹

Galectin-9 (Gal-9) is a 36-kDa β-galactoside containing lectin with metabolic roles in diabetes and obesity, cancer, and the immune response.^20,21,22 In the immune system, Gal-9 regulates the activity and function of natural killer (NK) cells, neutrophils, and T cells via interaction with CD44.^23,24,25 Here, we postulate that Gal-9 may be utilized to repress MIF activity by blocking CD44 function. We describe that a stable, recombinant form of Gal-9, referred to as sGal-9 (Gal-9 mC10-HPPY),²⁶ suppresses the pathologic phenotypes of RA-FLSs, including increased survival, migration, and invasion—possibly by the inhibition of the MIF-CD74/CD44 axis—and ameliorates the progression of chronic arthritis in mice. sGal-9 may be a promising treatment for RA, particularly in high-genotypic MIF-expressing patients.

Results

Identification of MIF^high subset in RA-FLSs

Recent studies have demonstrated FLS heterogeneity in the synovia of RA patients, which appears to be associated with the functional diversity of FLS subsets.^27,28 To compare MIF expression levels in the diverse FLS subsets, we performed single-cell RNA sequencing (scRNA-seq) on six samples of RA-FLSs from three donors. Our analysis of 17,220 cells found 9 different clusters depending on molecular markers, including LIM domain only (LMO), high-temperature requirement A serine peptidase 1 (HTRA1), CD81 molecule (CD81), clusterin (CLU), MIF, carboxypeptidase B1 (CPB1), dystonin (DST), PCNA clamp-associated factor (PCLAF), and guanylate binding protein 1 (GBP1) (Figure 1A). MIF expression levels were similar overall among the clusters, but one cluster exhibited an increased MIF expression (Figure 1B). In particular, the MIF^high cluster displayed the upregulation of genes related to mitochondrial function (NDUFB9, UQCRQ, NDUFB11, NDUFA3, NDUFB2, UQCR11, COX6A1, ATP5MF, ATP5ME, and COX6B1), endoplasmic reticulum (ER) function (WDR83OS, GAPDH, FKBP8, PPIA, and TIMP1), negative regulation of apoptosis (FKBP8, PPIA, TIMP1, HSPB1, MIF, and GAS6), and actin filament organization (TPM2, RHOC, and TMSB10) (Figure 1C). Gene set enrichment analysis (GSEA) revealed that the MIF^high cluster represented cellular processes related to the respiratory electron transport chain, mitochondrial ATP synthesis, oxidative phosphorylation, protein targeting to ER, and establishment of protein localization to ER, indicating abnormal activation of mitochondria and ER function in this subset (Figure S1).

Figure 1.

Expression and functional characterization of the MIF^high subset in RA-FLSs

(A) t-SNE plot showing clusters analyzed by single-cell RNA seq using 10x Genomics in cultured RA-FLSs isolated from three patients. (B) Violin plot representing the expression levels of MIF across the nine clusters. (C) Heatmap depicting differentially upregulated genes in MIF^high FLS associated with mitochondria function, ER function, apoptosis, and actin filament organization across the nine clusters of RA-FLSs. Rows represent genes, and columns represent clusters. The color of each cell of the heatmap indicates the expression levels, where red signifies higher expression of genes associated with the gene set in the respective cluster compared to other clusters, while blue indicates lower expression. (D) Changes in ER stress and mitochondria-related proteins in RA-FLSs treated with recombinant MIF (100 ng/mL). The cells were stimulated with recombinant MIF for the indicated times (hours), and the expression of GRP78, CHOP, BAX, and BCL2 was assessed by western blot analysis. (E) Expression levels of BCL2 and GRP78 in RA-FLSs with lower (<25^th percentile) and higher (>75^th percentile) intracellular MIF expression, as determined by flow cytometry analysis. The data were quantified as the mean fluorescence intensity. Histograms are representative of experiments conducted at least three times. The bar graphs show the mean ± SD. Statistical differences were assessed using an unpaired two-tailed t test (E). ∗p < 0.05; ∗∗∗∗p < 0.0001.

To further characterize the MIF^high cluster, we conducted an additional set of experiments to analyze ER- and mitochondria-related protein expression under MIF^high conditions. The results showed that the addition of recombinant MIF to cultured RA-FLSs resulted in changes in ER and mitochondrial protein expression, including GRP78, CHOP, and Bcl-2 expression, as determined by western blot analysis (Figure 1D). Moreover, flow cytometry analysis revealed that the expression levels of BCL2 and GRP78, which are essential for inhibiting the apoptotic death of RA-FLSs,^5,29 were higher in RA-FLSs with greater intracellular MIF expression (>75^th percentile) than in those with lower MIF expression (<25^th percentile) (Figure 1E). This indicates that the MIF^high subset may have anti-apoptotic properties that result in synovial hyperplasia.

We next investigated how this MIF^high cluster is linked to the FLS subsets previously identified in RA synovia. To achieve this, we imported AMP/RA phase 1 and 2 datasets, which profile diverse synovial tissue cells.^27,28 From the phase 1 data,²⁷ we obtained 5,265 cells, among which 1,844 fibroblasts were grouped into 4 subsets (Figure S2A). We analyzed them in association with the “MIF signature,” defined as the top 10 upregulated differentially expressed genes (DEGs), including transgelin (TAGLN), cyclin D1 (CCND1), cathepsin S (CTSS), retinoic acid receptor responder 2 (RARRES2), myosin light chain 9 (MYL9), MIF, cathepsin Z (CTSZ), profilin-1 (PFN1), ribosomal protein S26 (RPS26), and heat shock protein beta-1 (HSPB1). The results showed that the MIF signature was stronger in HLA-DRA^high sublining and CD34⁺ sublining FLSs (Figures S2B and S2C). Notably, HLA-DRA^high sublining FLSs, the primary source of interleukin-6 (IL-6), expanded 18-fold in leukocyte-rich RA synovium compared to leukocyte-poor RA and osteoarthritis.²⁷ The MIF signature was enriched in sublining FLSs, whose proliferation is a pathological hallmark of RA synovium (Figure S2D).³⁰ From the phase 2 data,²⁸ we also obtained 77,819 FLSs. The MIF signature was enriched in CD74^high HLA^high sublining FLSs, NOTCH3⁺ sublining FLSs, and POSTN⁺ sublining FLSs, which are all key pathogenic subsets (Figure S2E)^28,31,32; CD74^high HLA^high sublining FLSs in a phase 1 study correspond to HLA-DRA^high sublining FLSs, which represent the inflammatory FLSs.²⁸ In addition, studies have shown that NOTCH3 drives the pathological differentiation of FLSs,³¹ while POSTN promotes their migration and invasion.³² Consistent with the analysis of phase 1 data, sublining FLSs exhibited a higher MIF signature score than intermediate and lining FLSs (Figure S2F). However, such associations of the MIF^high subset with pro-inflammatory phenotypes (AMP/RA phases 1 and 2) and pro-migratory/pro-invasive phenotypes (AMP/RA phase 2) were not observed in the other seven subsets, except for the HTRA1^high subset, thereby indicating a specific role for the MIF^high subset in RA pathology (data not shown). Overall, the MIF^high FLSs appear to be associated with the pathogenic FLS subtypes in the AMP/RA phases 1 and 2 study.

These findings suggest the presence of a unique MIF^high subset in RA-FLSs, which is related to abnormally increased ER and mitochondria function of FLSs. Given that ER and mitochondria dysfunction is required for the enhanced survival and proliferation of RA-FLSs,^29,33 leading to the “pannus formation” (a pathologic hallmark of RA), the MIF^high subset could be a major FLS endotype mediating pannus formation in RA joints.

Inhibition of the interaction between MIF and CD44 by sGal-9

MIF binds to its cognate receptor CD74, leading to the recruitment of its co-receptor CD44 and signal transmission for proliferative, pro-migratory, and invasive activity.^15,19 Intervention in this pathway could offer a means to selectively target the aggressive features of RA-FLSs. To develop new drugs to intervene in the interaction between MIF and CD74/44, we generated a stable form of Gal-9 (sGal-9), which has been shown to influence different immune cells by direct interaction with CD44.^20,21,22 We confirmed the specific binding of sGal-9 to CD44 by conducting a plate-based binding assay with purified Fc-fusion proteins of CD44 (CD44-Fc). This binding was dose dependently increased with increasing concentrations of sGal-9 (Figures 2A and 2B). Moreover, the binding of sGal-9 to CD44-Fc was almost completely inhibited by the addition of an anti-CD44 antibody (Figures 2A and 2C). These results, together with the earlier reports,^{15,19,23,24,29} demonstrate that sGal-9 specifically binds to CD44.

Figure 2.

Solid-phase ELISA for the interaction between CD44 and sGal-9

(A) Illustration of the plate-based binding assay. The plate was coated with 20 μM (100 μL) sGal-9, followed by incubation with CD44 Fc chimera protein (CD44-Fc) or human IgG1 Fc protein (IgG1-Fc) at different concentrations in the absence or presence of the anti-CD44 antibody (α-CD44 Ab; 3.5 μg/mL). Fc-Bio, biotinylated anti-human IgG Fc antibody; HRP, streptavidin-conjugated horseradish peroxidase. (B) Dose-dependent increase in sGal-9 binding to CD44-Fc. (C) Inhibition of the interaction between sGal-9 and CD44 by addition of (α-CD44 Ab). Data are presented as mean ± SD. Statistical differences were assessed using one-way ANOVA with Tukey’s multiple comparisons test (C). ∗∗∗∗p < 0.0001.

sGal-9 suppression of survival, migration, and invasion of MIF-stimulated RA-FLSs

Based on the binding assay data, we next investigated whether sGal-9, as a CD44 inhibitor, suppresses the MIF-driven survival benefit in RA-FLSs that appears to be associated with ER and mitochondria dysfunction (Figures 1C and S1). As shown in Figure 3A, the MIF-induced increase in the expression of GRP78, a master regulator of the unfolded protein response that protects against apoptosis,²⁹ was partially blocked by the addition of sGal-9. Moreover, MIF downregulation of caspase-3 expression in RA-FLSs was restored by sGal-9 co-treatment. In parallel, MIF protected the starvation-induced apoptosis of RA-FLSs, as determined by the APOPercentage apoptosis assay, which was almost completely abrogated by sGal-9 addition (Figure 3B). Together, these data support the notion that sGal-9 renders RA-FLSs more susceptible to apoptosis by depriving them of the anti-apoptotic effect of MIF.

Figure 3.

Suppression of MIF-stimulated survival, migration, and invasion of RA-FLSs by sGal-9

(A) Changes in MIF-induced expression of GRP78 and caspase-3 by sGal-9. RA-FLSs were treated with recombinant MIF (100 ng/mL) in the absence or presence of sGal-9 (1 nM) for 24 h. Levels of GRP78 and caspase-3 were determined by western blot analysis. (B) sGal-9-induced apoptosis in MIF-stimulated RA-FLSs. Apoptosis levels were assessed using the APOPercentage apoptosis assay, which utilizes an intense, pink-colored dye reagent that is taken up by apoptosis-committed cells. Apoptotic RA-FLSs, therefore, appeared pink. Scale bar, 400 μm. (C and D) Effect of sGal-9 on wound migration and invasion of RA-FLSs induced by MIF. Cells were incubated with recombinant MIF (100 ng/mL) in the absence or presence of sGal-9 (1 nM) for 16 h. The number of migrated cells in wound area (C) and invaded cells in Matrigel chamber (D) were manually counted. Scale bars, 1,000 μm for (C) and 100 μm for (D). (E and F) Reduction of MIF-induced CD44 and CD44v6 expression by sGal-9 as determined by real-time PCR (E) and western blot analysis (F). Data are representative of experiments conducted at least three times and presented as mean ± SD. Statistical differences were assessed using one-way ANOVA followed by Tukey’s multiple comparisons test (A–F) and Kruskal-Wallis test with Dunn’s multiple comparisons test (E). ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

We next sought to determine whether sGal-9 also controls other pathologic responses induced by MIF in RA joints, such as the increased migratory and invasive capacity of RA-FLSs.³⁴ sGal-9 treatment led to a substantial decrease in wound migration of RA-FLSs stimulated with 100 ng/mL MIF (Figure 3C). Additionally, utilizing Matrigel invasion assay chambers, we observed a marked decrease in MIF-induced FLS invasion following treatment with sGal-9 (Figure 3D). However, sGal-9 failed to reduce the production of pro-migratory cytokine/chemokines, including IL-6, IL-8, and CCL2, indicating that sGal-9 suppression of FLS migration is not dependent on these cytokines (Figure S3). Our group previously demonstrated that MIF-induced FLS invasion is mediated by the increase in CD44 expression by MIF, particularly of the alternatively spliced CD44v3 and CD44v6 isoform.¹⁹ Accordingly, we found that sGal-9 almost completely repressed the MIF-induced upregulation of CD44 and CD44v6 expression at both mRNA and protein levels (Figures 3E and 3F). These observations, together with our previous study,¹⁹ suggest that sGal-9 regulates MIF-induced pro-invasive capacity of RA-FLSs by reducing CD44 and CD44v6 overexpression as well as by blocking MIF interaction with CD44 (see Figure 2).

RA-FLSs are exposed to a variety of pro-inflammatory cytokines and growth factors in inflamed joints, including tumor necrosis factor α (TNF-α), IL-1β, and transforming growth factor β (TGF-β).^2,5 We therefore investigated whether sGal-9 is also effective at repressing FLS activity under other stimulatory conditions in addition to MIF. To address this question, we first investigated whether sGal-9 shows a direct anti-FLS activity in the absence of exogenously added MIF. The results showed that sGal-9 substantially inhibited FLS viability comparable to the parental form of sGal-9 (G9Null), and this effect was dose dependent (Figure 4A). Moreover, TNF-α- and TGF-β-induced proliferation of RA-FLSs was almost completely repressed by sGal-9 in a dose-dependent manner, as determined by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) and bromodeoxyuridine assays (Figures S4A and S4B). In parallel, sGal-9 remarkably reduced lamellipodia formation, migration, and invasion of RA-FLSs in the absence of exogenous MIF, potentially through inhibiting endogenous MIF produced by RA-FLSs (Figures 4B–4D). The migration and invasion of RA-FLSs stimulated with IL-1β or TGF-β were also significantly diminished by sGal-9 co-treatment (Figures 4E and 4F). The observed reduction in cell migration and invasion occurred independently of the anti-proliferative effect of sGal-9 (Figure S4C). Anti-FLS activity of sGal-9 is not limited to MIF-stimulated conditions and remains effective under IL-1β-, TNF-α-, and TGF-β-elevated conditions.

Figure 4.

Anti-FLS effects of sGal-9 under non-MIF stimulatory conditions in vitro

(A–D) sGal-9 suppresses the survival, migration, and invasion of RA-FLSs stimulated with media alone (without MIF). (A) Dose-dependent inhibition of FLS viability by sGal-9 in vitro. Cell viability was assessed using the MTT assay. (B) Dose-dependent reduction of lamellipodium formation by sGal-9. RA-FLSs were stained with Alexa Fluor 488-conjugated phalloidin (green) for visualization of F-actin, and nuclei were stained with DAPI (blue). Arrowheads indicate filopodia and lamellipodia. (C) Inhibition of wound migration of RA-FLSs by sGal-9. Cells were incubated in DMEM containing sGal-9 for 16 h. The migrated cells beyond the reference line were then quantified. Scale bar, 1,000 μm. (D) Reduction of RA-FLS invasion by sGal-9. RA-FLSs were incubated with sGal-9 for 16 h, and invaded cells were stained using violet solution. Scale bar, 100 μm. (E and F) sGal-9 suppression of IL-1β or TGF-β-induced migration and invasion of RA-FLSs. Wound migration (E) and invasion (F) of RA-FLSs treated with recombinant IL-1β (1 ng/mL) or TGF-β (10 ng/mL) in the presence or absence of sGal-9 (1 nM) for 16 h. Scale bar, 1,000 μm for (E) and 100 μm for (F). “None” images in Figures 3D and 4F are derived from the same experiment. Data are representative of more than three independent experiments and presented as mean ± SD. Statistical differences were assessed using two-way ANOVA with Tukey’s multiple comparisons test (A, E, and F) and Friedman test with Dunn’s multiple comparisons test (B–D). ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

Anti-FLS effects of sGal-9 in vivo

sGal-9 (Gal-9 mC10-HPPY) possesses a 4-fold increase in solubility after modifying the remaining linker region of the parental form of sGal-9 (G9Null) with a truncated linker peptide, which facilitates scalable recombinant production.²⁶ Our formulation study demonstrated that sGal-9 exists in a monomeric form with a purity exceeding 96%, as confirmed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and size-exclusion high-performance liquid chromatography (HPLC) analyses (Figures S5A and S5B). Moreover, anti-sGal-9 neutralizing antibodies (kindly provided by GBIOLOGICS, Seongnam, Korea) restored sGal-9-mediated inhibition of FLS migration (Figure S5C). Given the high purity of sGal-9 as demonstrated by HPLC (Figure S5B), these results may alleviate concerns regarding the potential off-target effects caused by contaminants in the recombinant sGal-9 preparation.

Based on the aforementioned in vitro findings demonstrating the direct anti-FLS activity of sGal-9, we sought to examine the specific impact of sGal-9 on cartilage destruction and invasion mediated solely by RA-FLSs in vivo. To this end, we utilized the severe combined immunodeficiency (SCID) mouse xenograft model of arthritis, which is a humanized model of arthritis that induces cartilage destruction but without inducing inflammation or pathogenic action of lymphocytes and macrophages.³⁵ In this model, human cartilage along with RA-FLSs are implanted into left flank of SCID mice, while the right flank contains only cartilage. As illustrated in Figure 5A, both the primary site (left lateral) and contralateral (right lateral) flanks exhibited cartilage destruction, with increased degradation observed at the RA-FLS-loaded site. In contrast, cartilage destruction was notably reduced in both left and right implants treated locally with sGal-9. Furthermore, intraperitoneal administration of sGal-9 (2 mg/kg twice per week) significantly mitigated RA-FLS-induced cartilage degradation on both ipsilateral and contralateral sides (Figure 5B), which is concurrent with the in vitro findings in Figure 4. The number of migrated cells from the original implantation site was markedly reduced by both local and systemic administration of sGal-9 (Figures 5A and 5B). These findings suggest that sGal-9 suppresses the invasiveness of RA-FLSs in vivo and hinders their migration from the affected joint to unaffected cartilage.

Figure 5.

sGal-9 inhibition of RA-FLS migration and invasion in a humanized synovitis model

(A) Human cartilage implants in left flanks (primary) were co-implanted with sGal-9-treated RA-FLSs for 60 days in SCID mice. Cartilage implants of the same size were placed in the right flank (contralateral) without RA-FLSs (n = 2). Scale bars, 200 μm (overview) and 50 μm (magnified image). Bar graphs on the right depict the severity of cartilage invasion and degradation. (B) Human cartilage was co-transplanted into left flanks of SCID mice along with RA-FLSs, while equivalent-sized human cartilage was transplanted into the right flank. sGal-9 (2 mg/kg) was subsequently administered intraperitoneally twice per week for 60 days (n = 4). Afterward, the transplanted tissues were retrieved and analyzed. The implants underwent H&E staining. The invaded regions with perichondrocytic degradation are marked by black arrowheads. Scale bars, 200 μm (overview) and 50 μm (magnified image). Bar graphs on the right depict the severity of cartilage invasion and degradation. Data are presented as mean ± SD. Statistical differences were assessed using two-way ANOVA with Tukey’s multiple comparisons test (A and B). ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

Anti-arthritic effects of sGal-9 in mice with collagen-induced arthritis

sGal-9 has pleiotropic activities, showing diverse modes of action in various types of immune cells, including T cells, B cells, neutrophils, and NK cells.^{21,22,23,24,25} RA is an autoimmune disease involving diverse immune cells in its pathogenesis.¹ To determine the therapeutic relevance of sGal-9 to RA, we examined whether sGal-9 ameliorates the severity of collagen-induced arthritis (CIA), a model in which multiple types of immune cells, particularly autoreactive T and B lymphocytes, are implicated.³⁶ We first found that human sGal-9 exhibited cross-reactivity with rodent sGal-9 in binding to CD44, while maintaining considerable binding affinity for mouse CD44 (Figure S6). We next sought to determine the optimal dose of sGal-9 for suppressing CIA. As shown in Figure S7, sGal-9 inhibited the progression of CIA in a dose-dependent manner, with the maximum suppressive effect observed starting at 2 mg/kg administered once per week. There was no significant difference in body weight between the sGal-9- and vehicle-treated mice. Based on these findings, a dose of 2 mg/kg weekly was used for the remainder of the study.

As depicted in Figures 6A and S8A, subcutaneous injection of sGal-9 (2 mg/kg) once per week for 3 weeks substantially inhibited arthritis progression in mice with CIA, as assessed by gross inspection. This demonstrates the excellent clinical efficacy of sGal-9 in alleviating the progression of chronic autoimmune arthritis, similar to Enbrel and tofacitinib. Histopathologic examination of the affected joints also revealed lower degrees of synovial hyperplasia, cartilage destruction, pannus formation, and bone erosion in mice treated with sGal-9, also comparable to Enbrel and tofacitinib (Figures 6B, 6C, and S8B). Moreover, immunohistochemical staining showed that sGal-9 treatment potently reduced the infiltration of CD3⁺ (for T cells) and F4/80⁺ cells (for macrophages) in the affected joints (Figures 6D and 6E) and markedly decreased the proliferation of mouse FLSs in the inflamed synovia, as determined by Ki-67 immunostaining (Figure 6F). Additionally, consistent with the in vitro data, the number of MIF-expressing synovial fibroblasts was remarkably lower in mice treated with sGal-9 (Figure 6G). No apparent adverse effects were observed (data not shown). Collectively, concurrent with the in vitro and SCID mouse data, sGal-9 injection effectively suppresses the clinical and pathologic severity of autoimmune arthritis in mice.

Figure 6.

Amelioration of collagen-induced arthritis by sGal-9

(A) From 3 weeks after the primary immunization, mice with collagen-induced arthritis (CIA) were subcutaneously administered with sGal-9 (2 mg/kg, n = 6) once per week for 3 weeks. Arthritis severity was assessed at the indicated time point. Enbrel (10 mg/kg, 3 times per week, n = 6), serving as a positive control, was injected subcutaneously, while tofacitinib (6.2 mg/kg, once daily, n = 6) was fed orally. (B) Histology score for the synovial hyperplasia, cartilage destruction, pannus formation, and bone erosion in mice treated with vehicle alone versus with sGal-9, Enbrel, or tofacitinib. (C) Toluidine blue staining for assessment of cartilage damage. Scale bar, 100 μm. (D and E) Infiltration of CD3⁺ and F4/80⁺ cells in the affected joints of CIA mice treated with vehicle alone, sGal-9, Enbrel, and tofacitinib, as determined by immunostaining. Scale bar, 50 μm. (F) Cell proliferation determined by immunostaining for Ki-67⁺ cells. Scale bar, 50 μm. (G) Immunostaining for MIF expression in the joint tissues of CIA mice treated with either vehicle alone or sGal-9. Tissue sections were stained with an anti-MIF antibody, and the number of MIF-expressing FLSs was manually counted. Data are representative of more than three independent experiments and presented as mean ± SD. Statistical differences were analyzed using two-way ANOVA with Tukey’s multiple comparisons test (A), Kruskal-Wallis test with Dunn’s multiple comparisons test (B–D and F), one-way ANOVA with Tukey’s multiple comparisons test (E), and Mann-Whitney U test (G). ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001.

Gal-9 reduces the activation and expansion of pathogenic T and B lymphocytes in mice, mitigating generation of T helper 17 (Th17) cells and autoantibody production,^21,22,25 both of which are required for the initiation and perpetuation of CIA.³⁶ Finally, we aimed to verify whether sGal-9 modulates lymphocyte functions in our experimental setting. A substantial decrease in arthritis severity was reproduced by sGal-9-treated mice with CIA in a separate experiment (Figure S8C). In these mice, serum levels of anti-type II collagen antibody (anti-CII immunoglobulin G [IgG]), as well as proinflammatory cytokines, such as IL-6 and TNF-α, were significantly reduced by sGal-9 (Figure S8D). The production of IL-6 by splenocytes stimulated with anti-CD3/CD28 antibodies for 3 days was also decreased in sGal-9-treated mice (Figure S8E). Moreover, flow cytometry analysis of the stimulated splenocytes demonstrated that the frequency of CD4⁺IL-17⁺ Th17 cells was significantly lower in sGal-9-treated mice than in vehicle-treated mice; meanwhile, CD4⁺FOXP3⁺ T regulatory cells tended to be more abundant in the splenocytes of sGal-9-treated mice (Figures S8F and S8G). In summary, these results suggest that the regulation of CIA by sGal-9 results from its multiple effects on autoantibody, pro-inflammatory cytokines, and activated Th cells, in addition to those on FLSs.

Discussion

In this study, scRNA-seq analysis was employed to identify a unique MIF^high subset in RA-FLSs, which was associated with ER and mitochondrial dysfunction. Focusing on this synovial fibroblast subpopulation, we next investigated whether sGal-9, a CD44 blockade, regulates the MIF-CD74/CD44 axis in RA-FLSs. We found that sGal-9 was effective at reducing the increased migration, invasion, and proliferation of RA-FLSs induced by exogenous (recombinant) and presumably endogenous MIF in vitro. Moreover, both local and systemic administration of sGal-9 substantially inhibited excessive bone destruction by RA-FLSs in a xenotransplantation arthritis model and alleviated the severity of CIA in mice, comparable to Enbrel and tofacitinib. Collectively, these results suggest that sGal-9 controls the aggressive phenotype of RA-FLSs at least by inhibiting the MIF-CD74/44 pathway and prompts consideration for application to RA patients, particularly in those with high expression MIF alleles.

Recent studies have established that FLSs exhibit considerable morphological and functional diversity.^29,33 Given their pivotal role in RA pathogenesis,^1,2,3,5 we conducted unsupervised clustering analysis of RA-FLSs, uncovering significant heterogeneity with nine subclusters, consistent with earlier studies on synovial tissues.^37,38 Notably, our investigation with early passage (e.g., 3–6) FLSs identified a distinct subset of MIF-high expressing cells with the pathologic characteristics of ER and mitochondrial dysfunction. Given the previous studies of the role of ER and mitochondrial dysfunction in RA-FLSs,^29,33 this subset may play an important role in the increased survival, migration, and invasion of RA-FLSs that lead to bone and cartilage destruction in disease-affected joints.³⁹

In our analysis of the AMP/RA I and II datasets, the “MIF signature” of the MIF^high FLSs was associated with highly pathogenic FLS subtypes, such as CD74^high HLA^high FLSs, NOTCH3⁺ FLSs, and POSTN⁺ FLSs, underscoring its relevance to RA. Given the roles of these three subtypes in RA pathology,^27,28,31,32 the MIF signature may underlie their pro-inflammatory, pro-migratory, and pro-invasive properties, supporting the notion that it could serve as a therapeutic target for RA. In this regard, it would be interesting to see whether the MIF signature, consisting of the top 10 upregulated DEGs, or its pathway contributes to poor responsiveness to anti-rheumatic drugs and whether it can be modified in RA-FLSs following sGal-9 treatment. Further investigation is warranted.

Multiple pre-clinical studies have advanced the development of MIF inhibitors for autoimmune inflammatory diseases. For example, anti-MIF monoclonal antibody treatment profoundly and dose-dependently inhibits adjuvant arthritis in rats.⁴⁰ Several small molecules targeting MIF, including ISO-1, 4-1PP, and ibudilast, also have been developed.⁴¹ The small-molecule MIF antagonist ISO-1, for instance, protects against glomerulonephritis in two distinct experimental models of lupus,⁴² and ibudilast has shown efficacy at retarding disease progression in a phase 2 study of progressive multiple sclerosis.⁴³ The present study using sGal-9 provides a novel approach to disrupt the functional consequences of MIF binding to CD74, which leads to the recruitment and signaling activation of CD44. Interestingly, MIF-independent pathways can also be involved in sGal-9 regulation of RA-FLSs of sGal-9 under certain conditions, as evidenced by sGal-9 suppression of survival, migration, and invasion of RA-FLSs stimulated with IL-1β and TGF-β, which further highlights its therapeutic potential.

Our previous work has shown that MIF induces CD44 intracellular phosphorylation, leading to the activation of downstream MAPK and RhoA signaling, followed by initiation of CD44 alternative exon splicing, and an increase in the CD44v3 and -v6 isoforms that promote tumor cell invasiveness and oncogenic progression.^19,44 In this study, we demonstrate that sGal-9 remarkably reduces survival, migration, and invasion of RA-FLSs, which was accompanied by the downregulation of CD44 and CD44v6 expression. These findings indicate that targeting the MIF-CD74/CD44 axis with sGal-9 may offer a specific means to retard local invasion and migration of RA-FLSs by repressing a MIF^high FLS subset. Such a therapeutic approach could be envisioned as important adjunct to conventional immunosuppressive therapy in RA patients given the observations that synovial joint erosion may continue despite evident clinical remission.⁴

Currently used disease-modifying anti-rheumatic drugs (DMARDs) have revolutionized RA treatment. Many RA patients, however, discontinue conventional DMARDs within the first year due to drug toxicity or therapy-independent relapse.⁴⁵ Moreover, given that current therapeutic agents targeting T cells, B cells, and cytokines (e.g., TNF-α)—so-called biologic DMARDs—show limited success in achieving complete remission, RA-FLSs represent an attractive therapeutic target. Unfortunately, no drug currently targets the FLS population in RA, which might contribute to the non-responsiveness or inadequate responses observed in the clinic.^46,47 sGal-9 is a β-galactoside-binding lectin involved in immune regulation, inflammation, and cell signaling.^{20,21,22,23,24,25} Here, we identified sGal-9 as a novel drug candidate that directly kills or incapacitates RA-FLSs. The sGal-9 is more resistant to proteolysis when compared to native Gal-9.²⁶ In our formulation study, sGal-9 also exhibits increased solubility and stability, preventing aggregation or polymerization, and it is amenable to large-scale production. No significant toxicity or adverse effects were noted in rodents treated with sGal-9. We plan to conduct a clinical trial (phase 1/2) to evaluate the efficacy of sGal-9 in patients with autoimmune diseases, including RA.

A striking finding of this study is that the anti-arthritic effect of sGal-9 is comparable to two representative biologic DMARDs (Enbrel and tofacitinib), suggesting that sGal-9, while targeting FLS-driven pathologies, can provide an alternative strategy for RA treatment. Since CIA is a model of T cell-dependent autoimmune arthritis, sGal-9 appears to suppress a dominant feature of pathologic T cell activation via its effect on the MIF-CD74/44 pathway in FLSs. Nevertheless, it should be noted that sGal-9 exerts additional actions in CIA, including reducing the generation of Th17 and increasing regulatory T cell numbers. Moreover, the administration of Gal-9 reduces the activation and expansion of T and B cells and autoantibody generation in lupus-prone mice.⁴⁸ It is also possible that sGal-9 inhibition may block the binding of other CD44 ligands, such as hyaluronan, to control the aggressive behavior of RA-FLSs.⁴⁹ Thus, the therapeutic efficacy of sGal-9 in mice with CIA may result from additional mechanisms, including inhibitory action on autoimmune lymphocytes, as evidenced by our data regarding sGal-9 suppression of anti-CII IgG, proinflammatory cytokines (e.g., IL-6, TNF-α), and Th17 lymphocytes.

In conclusion, our study provides a comprehensive resource of single-cell transcriptomics in cultured RA-FLSs, which can be applied in future studies to aid in the discovery of novel regulators of pathological processes mediated by RA-FLSs. Moreover, our work highlights the importance of an MIF^high FLS subset and the MIF-CD74/44 pathway as a therapeutic target for reducing the invasive and destructive character of rheumatoid synovitis. Finally, we propose sGal-9 as a promising therapeutic agent for RA directly targeting RA-FLSs with aggressive phenotypes. We anticipate that our data will significantly enrich the understanding of the MIF-driven pathogenesis of RA and provide a novel approach for clinical intervention.

Materials and methods

Patients

Synovial tissues or fluids were collected during synovectomy or total joint arthroplasty from 12 patients, all of whom met the 2010 American College of Rheumatology (ACR)/European League Against Rheumatism classification criteria for RA.⁵⁰ The mean age (±standard error of the mean) was 57.8 ± 3.5 years, and 75.0% (9/12) were female. The mean duration of disease was 15.4 ± 3.5 years. A total of 75.0% (9/12) were treated with anti-rheumatic drugs. Further details regarding patient demographics, clinical characteristics, and medications are provided in Table S1.

This study was approved by the institutional review board of the Catholic Medical Center, Catholic University of Korea (approval number KC20TIS0643). All study participants gave written informed consent.

Sample collection and preparation

FLSs were isolated from synovial tissues or synovial fluids as previously described.¹⁹ All experiments were carried out with FLSs from passages 3 through 6.

scRNA-seq analysis in RA-FLSs

The 10x Genomics Chromium platform was employed for single-cell capture and library preparation, following the manufacturer’s instructions. Sequencing libraries were generated using the 10x Genomics Single Cell 3′ Library Kit and sequenced on an Illumina platform. All procedures were conducted at Macrogen (Seoul, Korea). Raw sequencing data were processed using the Cell Ranger pipeline to generate gene expression matrices for individual cells. Quality control, filtering, and normalization of scRNA-seq data were then performed using the Seurat version 3.0 package in R. Cells with low gene counts, high mitochondrial gene expression, or ambiguous cell-type identification were filtered out. Data were normalized, scaled, and subjected to principal-component analysis and dimensionality reduction using t-distributed stochastic neighbor embedding (t-SNE) to visualize cellular clusters. FindMarkers was used to identify DEGs for each cluster, with significant DEGs selected based on an absolute log2(fold change) > 0.3 and an adjusted p < 0.05. GSEA was conducted using the clusterProfiler package in R.⁵¹

To investigate the relationship between MIF^high FLS and previously reported FLS subtypes, we defined MIF^high signature genes as the top 10 upregulated DEGs ranked by log₂ fold change. These included TAGLN, CCND1, CTSS, RARRES2, MYL9, MIF, CTSZ, PFN1, RPS26, and HSPB1. Subsequently, we evaluated the expression of this signature across known synovial FLS subsets from the AMP/RA phase 1 and 2 datasets, which profile heterogeneous synovial tissue cells using scRNA-seq and single-cell Cellular Indexing of Transcriptomes and Epitopes by Sequencing, respectively.^27,28 The phase 1 dataset is available on ImmPort (SDY998),²⁷ and the phase 2 dataset can be accessed via Synapse (https://doi.org/10.7303/syn52297840).²⁸ We calculated the score for each FLS subset using Seurat’s AddModuleScore function. Signature scores were also calculated in broader FLS subsets generated by merging detailed subsets based on their sublining, intermediate, or lining identity. To assess differences in scores, we used the Wilcoxon rank-sum test for pairwise comparisons and the Kruskal-Wallis test followed by Dunn’s post hoc test for comparisons involving three or more subsets.

Western blot analysis

RA-FLSs were lysed in radioimmunoprecipitation assay buffer (50 mM Tris-HCl pH 7.6, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, and 0.1% SDS) supplemented with complete protease inhibitor. Supernatants were then obtained after centrifugation at 20,000 × g for 20 min at 4°C. Protein concentrations were determined using a BCA assay kit (Bio-Rad, Hercules, CA). Proteins were resolved via SDS-PAGE on approximately 4%–10% gradient acrylamide gels and transferred to polyvinylidene fluoride membranes (0.45 μm pore size; GE Healthcare, Milwaukee, WI). Membranes were incubated with anti-GRP78 (1:1,000; Invitrogen, Carlsbad, CA), anti-CHOP (1:1,000; Abcam, Cambridge, MA), anti-XBP-1 (1:1,000; Abcam), anti-BAX (1:1,000; Santa Cruz Biotechnology, Santa Cruz, CA), anti-BCL2 (1:1,000; Cell Signaling Technology, Danvers, MA), anti-caspase-3 (1:1,000; Cell Signaling Technology), anti β-tubulin (1:1,000; Abcam), anti-CD44 (1:1,000; Cell Signaling Technology), and anti-CD44v6 (1:1,000; Bio-Rad). After washing three times in Tris-buffered saline with 0.1% Tween 20 detergent (TBS-T), membranes were incubated with a peroxidase-conjugated goat anti-rabbit or anti-mouse antibody (1:1,000; Santa Cruz Biotechnology) in TBS-T containing 5% skim milk for 2 h at room temperature. They were subsequently washed three times in TBS-T and exposed to enhanced chemiluminescence substrate (Santa Cruz Biotechnology). The chemiluminescence signal was detected using a Luminescent Image Analyzer LAS-4000 (Fuji Film, Osaka, Japan).

Flow cytometry

For the intracellular staining of RA-FLSs, the cells were collected, washed, fixed, and permeabilized, followed by staining with anti-MIF (R&D Systems, Minneapolis, MN), anti-BCL2 (BD Pharmingen, San Jose, CA), or anti-GRP78 (Invitrogen) antibodies for 1 h at room temperature. For the staining of T cells, the cells were stained with anti-CD3 (BD Pharmingen), anti-CD4 (BD Pharmingen), and anti-CD25 (BD Pharmingen) antibodies for 30 min at room temperature and then subjected to intracellular staining with anti-IL-17 (BD Pharmingen) or anti-FOXP3 antibodies (BD Pharmingen) for 1 h at room temperature. The stained cells were analyzed using an FACS Canto II system (BD Biosciences, San Jose, CA).

Binding assay of sGal-9 with CD44

The interaction between sGal-9 and CD44 was determined by enzyme-linked immunosorbent assay (ELISA). Briefly, the plates were coated with sGal-9 (20 μM) in PBS at 4°C overnight and blocked with 1% BSA in PBS 2 h. After washing with PBS containing 0.01% Tween 20, they were incubated with human IgG1 Fc protein (R&D Systems) or recombinant human CD44 Fc chimera protein (R&D Systems) in incubation buffer for 2 h at room temperature. After washing with PBS, the wells were incubated with biotinylated anti-human IgG Fc antibody (mouse IgG1; 1:5,000, R&D Systems) for 2 h. Subsequently, streptavidin-conjugated horseradish peroxidase (HRP) was added to each well and incubated again for 20 min. TMB substrate solution was added, and the reaction was allowed to proceed for 20 min before being terminated by the addition of H₂SO₄. Absorbance at 450 nm with a reference wavelength of 570 nm was then determined using a microplate reader (Molecular Devices, San Jose, CA).

Generation of sGal-9

Recombinant human sGal-9 (Gal-9 mC10-HPPY), supplied from GBIOLOGICS, was generated as previously described.²⁶ sGal-9 has increased solubility and biologic activity due to modifications in the amino acid in the linker region as compared to the parental form of sGal-9 (G9Null), which has a truncated linker peptide.²⁶

Real-time qPCR

Total RNA was extracted using an RNeasy Mini Kit (Qiagen, Valencia, CA), followed by cDNA synthesis from the isolated RNA using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, Waltham, MA) according to the manufacturer’s instructions. Real-time qPCRs were performed with the cDNA, specific primer sets, and SYBR Green Master Mix (Bio-Rad) using a MyCycler (Bio-Rad). The real-time qPCR primers (Bioneer, Daejeon, Korea) were as follows: human CD44 (5′-CAGACCTGCCCAATGCCTTTGATGGACC-3′; 5′-CAAAGCCAAGGCCAAGAGGGATGCC-3′), human CD44v6 (5′-CAGGCAACTCCTAGTAGTAC-3′; 5′-CCAAGATGATCAGCCATTCTGG-3′), human glyceraldehyde 3-phosphate dehydrogenase (GAPDH) (5′-AAGGTGAAGGTCGGAGTCAA-3′; 5′-AATGAAGGGGTCATTGATGG-3′). The qPCR conditions for mRNA quantification were as follows: 95°C for 3 min, followed by 40 cycles of denaturation at 95°C for 10 s and annealing/extension at 60°C for 30 s. The cycle threshold values of each sample were normalized by the value of GAPDH.

Cell viability

RA-FLS viability was determined by MTT assays. Briefly, RA-FLSs were seeded into 48-well plates at a density of 1 × 10⁴ cells per well in DMEM supplemented with 10% fetal bovine serum (FBS). The cells were treated with sGal-9 for 24 or 48 h. MTT solution (Sigma-Aldrich, St. Louis, MO) was added and then incubated for 3 h at 37°C. Optical density was measured at a wavelength of 540 nm using a microplate reader (Molecular Devices, Sunnyvale, CA).

APOPercentage assay

To quantify the extent of apoptosis in RA-FLSs, we utilized the APOPercentage apoptosis assay kit (Biocolor Ltd., Belfast, UK) in accordance with established protocols. After inducing apoptosis with sGal-9, the cells were labeled with an intense, pink-colored APOPercentage dye and examined under a light microscope. Photographic images of the stained cells in pink were captured using an EVOS Cell Imaging System (Thermo Fisher Scientific), and the degree of apoptosis was assessed by manually counting in three randomly selected microscopic fields.

Wound migration assay

RA-FLSs (1 × 10⁵ cells/well) were seeded onto 6-well plates. Then, upon reaching approximately 90% confluence, the cells were wounded by scratching cells with sterile pipette tips and subsequently incubated with DMEM supplemented with 1% FBS containing sGal-9 in the absence or presence of MIF (100 ng/mL) for 16 h. The cells were fixed with 4% paraformaldehyde (Wako Pure Chemicals, Osaka, Japan) and stained with crystal violet solution. Images were captured using an EVOS Cell Imaging System (Thermo Fisher Scientific). The cells that migrated over the reference lines were manually counted in three random microscopic fields.

Matrigel invasion assay

To assess the invasion of RA-FLSs, the BD BioCoat Matrigel invasion chamber assay systems (Becton Dickinson, Bedford, MA) were used according to the manufacturer’s protocol. Briefly, RA-FLSs were seeded in the upper chamber of the Matrigel invasion chamber and allowed to migrate for 16 h in DMEM containing sGal-9. Non-invading cells then were removed by gentle scrubbing with a cotton-tipped swab, and the cells on the lower surface of the membrane were stained with crystal violet solution. The invaded cells were manually counted in eight randomly selected fields for quantification.

Immunofluorescence staining of F-actin

Immunofluorescence staining of F-actin was performed on chamber slides to visualize the lamellipodium-containing RA-FLSs. In brief, RA-FLSs (5 × 10³ cells per well) were seeded onto 8-well chamber slides (Nalgene Nunc International, Rochester, NY) and incubated at 37°C for 24 h in DMEM supplemented with 1% FBS containing sGal-9. For all experiments, cells were fixed with 4% paraformaldehyde in PBS (Wako Pure Chemicals) for 20 min and permeabilized with 0.1% Triton X-100 in PBS for 5 min at room temperature. After washing twice with PBS, the cells were blocked with 1% BSA in PBS at room temperature for 60 min and then incubated with Alexa Fluor 488-conjugated phalloidin (1:400; Invitrogen) diluted in PBS for 1 h at room temperature. In some experiments, mitochondria also were stained with Mitotracker Red (Invitrogen). After washing with PBS, nuclei were stained with DAPI (2 μg/mL; Roche, Mannheim, Germany). Coverslips were mounted on glass slides with Antifade Mounting Medium (Vector Laboratories, Burlington, CA). Cells were examined using a confocal microscope LSM 800 (Carl Zeiss, Oberkochen, Germany). Lamellipodium-containing cells were manually counted in three random fields.

Cytokine ELISA

The levels of IL-6, IL-8, and CCL2 were measured by ELISA kits (R&D Systems) according to the manufacturer’s instructions.

Mice

SCID and DBA1 mice were purchased from The Jackson Laboratory (Bar Harbor, ME). They were held under specific pathogen-free (SPF) and germ-free conditions, respectively, in the animal facilities at the Catholic University of Korea. Mice were randomly allocated into experimental groups. All surgical interventions and presurgical and postsurgical animal care was conducted in accordance with the Animal Protection Act, the Laboratory Animal Act, and the Guide for the Care and Use of Laboratory Animals. Additionally, the Guidelines and Policies for Rodent Survival Surgery provided by the Institutional Animal Care and Use Committee at the School of Medicine, Catholic University of Korea, were followed. The animal study was approved and received approval number CUMS-2022-0348.

Humanized synovitis in vivo model using SCID mice

Male SCID mice, aged 8 weeks, were sourced from The Jackson Laboratory and housed in an SPF facility. RA-FLSs were implanted with normal human cartilage to induce synovitis, as previously described.⁵² Briefly, RA-FLSs (2 × 10⁶ cells/100 μL) were mixed with sGal-9 (50 nM) and co-implanted with human cartilage into the left flank of the SCID mice. The right flanks received only cartilage implants. The mice then were administered sGal-9 intraperitoneally at a dosage of 2 mg/kg twice weekly for 60 days post-implantation. The implants were harvested after the designated period, and the extent of cartilage destruction was assessed through hematoxylin and eosin (H&E) staining, employing established methodologies.⁵²

Cartilage degradation and invasion were assessed by the scoring system described previously.⁵² Briefly, the cartilage degradation score was evaluated as follows: visible degradation was scored as 1 point, moderate degradation as 2 points, and intensive degradation as 3 points. The invasion score was assessed based on the number and extent of invasion cell regions; a single invasive cell region with 3–5 layers was scored as 1 point, 2 invasion cell regions with 6–10 layers as 2 points, a single invasion cell region with 11 layers as 3 points, and 2 or more invasion cell regions with 11 or more layers as 4 points.

Induction and evaluation of CIA in mice

Male DBA/1 mice were immunized with bovine type II collagen (Chondrex, Redmond, WA) according to the established protocols.^36,53 Booster immunization was administered if necessary. From 3 weeks after primary immunization, the mice received subcutaneous injections of sGal-9 at a dosage of 2 mg/kg once per week for 3 weeks. For the positive control, anti-TNF-α etanercept (Enbrel; 10 mg/kg, 3 times per week) was administered via subcutaneous injection, while JAK inhibitor tofacitinib (6.2 mg/kg, once daily) was orally administered. The incidence and severity of arthritis were determined by visual inspection, as described previously.³⁶ Briefly, mice were observed 2–3 times per week for the onset, duration, and severity of the joint inflammation over an 8-week period following primary immunization. Each limb was assessed on a 0- to 4-point scale, for a maximum possible arthritis score of 16. If necessary, booster immunization was carried out, and the hind foot that received the booster was excluded from the evaluation.

We also assessed the pathological severity of arthritis in the joint tissues stained with H&E and toluidine blue. As described earlier,⁵⁴ four distinct areas of each tissue sample were captured and evaluated independently by two or more researchers. Synovial hyperplasia, pannus formation, cartilage destruction, and bone erosion were assessed based on the following criteria: synovial hyperplasia was scored as 0 for no hyperplasia, 1 for mild hyperplasia with very few inflammatory cells, 2 for well-marked inflammatory cellular proliferation with thickening of the synovium, and 3 for extensive inflammatory cell proliferation with severe synovial thickening. Pannus formation was scored as 0 for no pannus formation, 1 for indistinct pannus formation, 2 for weak invasion of pannus into the bone, and 3 for strong invasion of pannus into the bone. Cartilage destruction was scored as 0 for no destruction, 1 for an irregular cartilage surface with focal erosions, 2 for a corrugated cartilage surface, and 3 when more than 50% of the cartilage was destroyed. Bone erosion was scored as 0 for no erosion, 1 for a roughened bony surface, 2 for cell invasion into the bone with marked loss of bone surface integrity, and 3 for nearly complete disruption of bone structure. For the toluidine blue-stained tissues, surface regularity and cartilage thickness were assessed. Surface regularity was scored as 0 for a smooth surface (75%–100% intact), 1 for moderate irregularity (50%–75% intact), 2 for an irregular surface (<50% intact), and 3 for a severely irregular surface. Cartilage thickness was scored as 0 when thickness was more than two-thirds of the average depth relative to the ground control, 1 for thickness between half and two-thirds of the ground control, and 2 for thickness less than half of the ground control.

Immunohistochemical analysis

Paraffin-embedded block of the ankle joints from mice with CIA were used for immunohistochemical analysis. The sections underwent deparaffinization in xylene, followed by rehydration through a series of ethanol solutions. Antigen retrieval was performed by microwave heating in citrate buffer. The sections were treated with a 3% H₂O₂ solution for 30 min at room temperature to block endogenous peroxidases and subsequently incubated with 10% normal donkey serum (Jackson Immunoresearch Laboratories, West Grove, PA) for 1 h at room temperature to block nonspecific binding. Primary antibodies against CD3 (Santa Cruz Biotechnology), F4/80 (Abcam), Ki-67 (Abcam), or MIF antibodies (kindly provided by Dr. Richard Bucala) were applied to the sections and incubated overnight at 4°C. After washing with PBS, the sections were exposed to biotinylated secondary antibodies. Visualization was achieved using 3,3′-diaminobenzidine tetrahydrochloride (Vector Laboratories) as the chromogenic substrate. Nuclei were counterstained with Mayer’s hematoxylin. Following dehydration, clearing, and mounting, images were captured using a Panoramic MIDI slide scanner (3DHISTECH, Budapest, Hungary).

Statistical analyses

Statistical analyses were conducted using GraphPad Prism version 10.0.1. The data are expressed as mean ± standard deviation (SD). Missing values were handled using pairwise deletion. Data normality was assessed using the Shapiro-Wilk test. Statistical differences between groups were evaluated using one-way ANOVA with Tukey’s multiple comparisons test or the Kruskal-Wallis test with Dunn’s multiple comparisons test, depending on the normality of the data. For analyses involving multiple groups with significantly different variances, two-way ANOVA with Tukey’s multiple comparisons test was used. p < 0.05 was considered significant.

Data availability

The datasets used and/or analyzed during the present study are available from the corresponding author upon reasonable request.

Acknowledgments

This work was supported by grants from the National Research Foundation of Korea (NRF) funded by the Korean government (MSIT) (RS-2024-00442793 to W.-U.K.); the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (RS-2024-00512879 to W.-U.K.); the NIH (AR078334); and the ACR Rheumatology Research Fund. The authors also acknowledge the financial support of the Catholic Medical Center Research Foundation (program year 2020 to S.-A.Y.) and a grant from the Jiangsu Health Commission, China (MQ2024017 to M.L.).

Author contributions

W.-U.K., J.J., and S.-A.Y. designed the experiments. W.-U.K., D.S., and S.-A.Y. drafted the manuscript. M.L., M.-K.N., J.K., S.-H.P., S.-H.L., C.K., and J.G.K. performed the experiments and data analyses. W.-U.K., J.J., R.B., and S.-A.Y. interpreted the results from data analyses and experiments. R.B. provided the research reagents. W.-U.K, J.J., S.-A.Y., and R.B. edited the manuscript and supervised the research. All authors commented on the manuscript.

Declaration of interests

This study was supported by a grant from GBIOLOGICS, which holds a US Patent for sGal-9 (no. US20150307574A1). R.B. serves on the advisory boards of Apaxen SA and OnCooNe.