Proteolytic processing of PRG4 by tryptase β
We hypothesized that PRG4 is susceptible to proteolytic processing and we explored if tryptase β was able to cleave PRG4 as increased tryptase β activity and mast cells are abundant in OA-affected joints29,32. Enzyme kinetic analyses of recombinant human proteins using an SDS-PAGE gel followed by silver staining (Fig. 1a) revealed that human PRG4 was efficiently cleaved by tryptase β within 5 min of incubation at 37 °C generating fragments of ~ 50 kDa (Fig. 1b). The cleaved PRG4 form was stable for at least 240 min. (Fig. 1b). PRG4 was also cleaved by tryptase in a dose-dependent manner when incubated for 1 hour at 1:10, 1:100, and 1:1000 tryptase β:PRG4 molar ratios (Fig. 1c). PRG4 cleavage by tryptase β was prevented by the serine protease inhibitor, AEBSF, at a dose of 1, 10, and 100 µM (Fig. 1d). As the mucin domain of PRG4 is extensively glycosylated and is known to provide its boundary lubrication property, we wanted to explore if a change in glycosylation would impact tryptase β processing of PRG4. PRG4 and tryptase β were incubated for 1 h or 18 h after being subjected to two deglycosylation enzymes: PNGase F, to remove N-linked glycans, and an Enteroccocus faecalis O-glycosidase, to remove O-linked glycans. Deglycosylation of the processed PRG4 fragments resulted in complete degradation of the ~ 50 kDa PRG4 fragments and generated additional lower molecular weight fragments (Supplementary Fig. 1a). Therefore, glycosylation of PRG4 decreases its susceptibility to be processed by tryptase β.
Next, to determine the N-terminal sequences of the potential multiple protein fragments that we detected, we used a mass spectrometry approach called Amino-Terminal Oriented Mass Spectrometry of Substrates (ATOMS)33,34 (Fig. 1e). A mixture of PRG4 was dimethylated with light formaldehyde (CH2O, + 28 Da), while a mixture of PRG4 and tryptase β was dimethylated with heavy formaldehyde (CD2O, + 34 Da). Both mixtures were combined before being subjected to LC-MS/MS followed by MaxQuant35,36 analysis (Fig. 1e). After a 5 min incubation, eight cleavage sites were identified: 146K↓K147, 147K↓V148, 251K↓I252, 303K↓T304, 1224R↓D1225, 1284R↓R1285, 1330K↓G1331, and 1345K↓G1346 (Fig. 1f). Using ATOMS to identify cleavage sites of PRG4 by tryptase β at 1h and 24 h incubation, we found 28 and 31 cut sites, respectively (Supplementary Fig. 1b, c), suggesting that tryptase β can degrade PRG4. Thus, we identified PRG4 as a novel substrate of tryptase β.
Reduced PRG4 lubricating ability after tryptase β cleavage
PRG4 is a glycoprotein with important lubricating functions5,18,37. We next examined if processed PRG4 would result in a loss of lubrication. Using a tribology test with a glass-polydimethylsiloxane (PDMS) polymer interface38, we performed a velocity sweep analysis at different lubricating regimes (Fig. 2a). Our tribological results demonstrated that the addition of tryptase β to PRG4 causes loss of boundary lubrication (low velocity) and mixed lubrication (medium velocity) but not hydrodynamic lubrication (high velocity) (Fig. 2b, Supplementary Fig. 2). This loss could be partially rescued when AEBSF was added to the mixture to inhibit tryptase activity. The loss of PRG4 lubrication was observed after 1 h, 6 h, and 24 h addition of tryptase (Fig. 2c). Thus, processing of PRG4 by tryptase β resulted in a loss of lubricating ability.
Tryptase β Cleavage Of PRG4 Associates With Articular Cartilage Degeneration
To study the impact of PRG4 processing by tryptase β, we used an OA-induced destabilization of the medial meniscus (DMM) rat model39 that resembles clinical meniscal injury in humans (Fig. 3a). Utilizing three different antibodies (tryptase β (red), antibody to PRG4 mucin domain (blue), and antibody to PRG4 C-terminal (green)) and DAPI (white) (Fig. 3b), we stained rat knee joints over a period of 1 thru 4 weeks post-DMM. Interestingly, no signal for the PRG4 C-terminal antibody was detected in the uninjured cartilage, but the PRG4 mucin domain antibody showed intense staining of the superficial layer of the cartilage as expected40 (Fig. 3c). One-week post- DMM, loss of mucin domain PRG4 with concurrent appearance of C-terminal PRG4 and Tryptase was observed on the surface of articular cartilage. We also detected a disrupted/abnormal expression of mucin and C-terminal PRG4 in the cartilage along with colocalization of tryptase β in the areas where PRG4 was detected (Fig. 3d, Supplementary Fig. 3a-b). Two weeks post- DMM, diminished PRG4 and tryptase β staining was detected (Fig. 3e, Supplementary Fig. 3c). The mucin domain antibody to PRG4 (blue) was undetectable at three weeks post-DMM (Fig. 3f, Supplementary Fig. 3d) but by four weeks post-DMM, reactivity to the mucin domain antibody was re-observed at the cartilage surface (Fig. 3g, Supplementary Fig. 3e).
Next, we wanted to see if we could prevent this loss of PRG4 in the DMM model by injecting recombinant PRG4 at the time of DMM surgery (Fig. 3h, Supplementary Fig. 4a). Four weeks post DMM, increased PRG4 staining was detected on the cartilage surface of rats treated with recombinant PRG4 as compared to control saline injections (Fig. 3i-j and Supplementary Figs. 4a-c). Therefore, injections of PRG4 could partially prevent this loss of PRG4 levels. Also, the localization of tryptase β and PRG4 corresponded to regions of cartilage injury and areas where PRG4 was lost.
Tryptase β Cleavage Of PRG4 Enhances NF-κb Activity Via TLR Receptors
We previously demonstrated that PRG4 binds directly to TLR2,-4, and − 5 and activates NF-κB8. Next, we investigated if tryptase β processing of PRG4 changes its ability to activate NF-κB. We used a HEK blue media reporter cells system that either lack TLRs (TLR-null) or have overexpression of TLR2, -4, or -5 (Fig. 4a). TNFα was used as a positive control. In the TLR-null cells, no activation of NF-κB was detected, except for the positive control TNFα (p < 0.001), suggesting that PRG4 requires a TLR for NF-κB activation in this cell type (Fig. 4b). In TLR2+ cells, addition of PRG4 + tryptase β resulted in a significant increase of NF-κB when compared to PRG4 (p < 0.001) or TNFα (p = 0.03) (Fig. 4c). Addition of PRG4 + tryptase β in the presence of AEBSF significantly decreased NF-κB activation (p < 0.001) (Fig. 4c). In TLR4+cells, no significant difference was detected between PRG4 + tryptase β and TNFα, but PRG4 + tryptase β significantly activated NF-κB as compared to PRG4 (p < 0.001) and PRG4 + tryptase β + AEBSF (p < 0.001) (Fig. 4d). In TLR5+cells, PRG4 + tryptase β was able to activate NF-κB above the levels observed in the TNFα (p = 0.03), PRG4 (p < 0.001) or PRG4 + tryptase β + AEBSF (p < 0.001) groups (Fig. 4e). Therefore, tryptase β processing of PRG4 resulted in a significant increase of NF-κB activation as compared to full length PRG4 and this activation was achieved either via TLR2, -4, or -5.
PRG4 Is Expressed In Synovial Lining Fibroblasts
OA presents a dynamic cellular landscape throughout the disease course; therefore, we wanted to know which specific cell types within the joints are selectively expressing PRG4, TLR2, -4, and − 5. We mined the literature for rodent arthritis models and used a dataset with the associated number of GSE184609 for single-cell RNA sequencing (scRNA-seq) analysis41. Population distribution of scRNA-seq data of hind paw joint cells isolated at the indicated time points of glucose-6-phosphate isomerase (GPI)-induced arthritis were determined. After excluding low-quality cells, our new analysis included 24,496 cells from 5 mice. Uniform manifold approximation and projection (UMAP) based clustering separated cells into 10 individual clusters (Fig. 5a and Supplementary Fig. 5a-c). These clusters predominantly belonged to three classes, namely fibroblasts, monocytes/neutrophils, and macrophages (Fig. 5a). We next analyzed the percentages of cell types in each group. We found that the abundance of various cell types differs widely between naive, Day 6, 14, and 25. We detected a reduced number of cells between fibroblasts and synovial lining fibroblasts between naïve and the GPI-arthritis-induced groups (Day 14 vs. naive) (Fig. 5b-d). Interestingly, Prg4 was explicitly expressed specifically in synovial lining fibroblasts as seen previously42 (Fig. 5c). Subsequently, we looked at the expression pattern of Prg4, Thy1(CD90), Tlr2, Tlr4, and Tlr5, and identified Tlr4 as a predominant gene in the synovial lining fibroblasts, whereas Tlr2 and Tlr5 appear to be expressed by other cell types (Fig. 5c). Additionally, Pgr4’s expression is downregulated over time from the naive group to Day 6, 14, and 25 (Fig. 5e). As we have shown previously7, PRG4 is known to interact with other cell surface receptors such as CD44 (Fig. 5f); therefore, we next investigated the crosstalk between synovial lining fibroblasts and other cell types in the hind paw joint cells using Squidpy. The level of interaction between lining fibroblasts and monocytes/neutrophils was the most elevated in the naive group (Fig. 5g). On Day 14, the most significant communications between synovial lining fibroblasts and macrophages and monocytes/neutrophils were mediated via the CD44/HBEGF pairs (Fig. 5h). Overall, in a mouse model of inflammatory arthritis, our data analysis suggests that PRG4 could signal via TLR4 and/or CD44 to activate NF-κB.
Comparison Of The Proteomes Of Non-OA vs OA Primary Human Synovial Cells
To characterize the biological differences occurring in OA, we isolated primary human synovial CD90+/THY1+ cells using cell sorting (Fig. 6a). As these fibroblasts were demonstrated in mice to have minimal to no PRG4 expression41 (Fig. 5c), we subjected them to the addition of recombinant PRG4, PRG4 + tryptase β, tryptase β, or buffer alone. In parallel, we used a similar CD90+ cell sorting approach in OA patients (Fig. 6a). All five different conditions (N = 3 biological replicates, n = 2 technical replicates) were subjected to a quantitative shotgun proteomics workflow using tandem mass tags (TMT) 6-plex labeling (Fig. 6a). After LC-MS/MS analysis, database search for the identification of peptides and proteins was performed with MaxQuant35 at 1% false discovery rate (FDR). Statistical analysis was performed with MSstatsTMT43. We identified 2,206 proteins and quantified 2,166 proteins (Supplementary Table 1). We found a loss of PRG4 in OA patients’ cells as compared to buffer treated non-OA cells. As expected, we found elevated levels of PRG4 in the PRG4 treated cells (Fig. 6b), but these levels were reduced when PRG4 was added in combination with tryptase β which is consistent with our data that tryptase β degrades PRG4 (Fig. 1f, Supplementary Fig. 1c). In a similar fashion to OA patients’ cells, multiple proteins levels were reduced when PRG4 was added with tryptase β, including phosphatidylinositol-binding clathrin assembly protein (PICALM), inositol 1,4,5-trisphosphate receptor type 3 (ITPR3), hepatoma-derived growth factor (HDGF), alpha-L-iduronidase (IDUA), and CD44 (Fig. 6b). Interestingly, HDGF was previously reported to exert mitogenic activity on fibroblasts and PRG4 to regulate proliferation of fibroblasts44,45, therefore, suggesting a loss of this activity in OA or when tryptase β was added with PRG446,47. Interestingly, there was also a loss of ossification components (SLC39A14, CCDC154, THBS3, IGF2, MMP14, RBP4) in OA fibroblasts when compared to buffer-treated non-OA fibroblasts, as analyzed by Metascape48 (Fig. 6c, Supplementary Fig. 7a-b). Using Metascape, the OA fibroblasts had an enrichment of signaling by interleukins and expressed a significant increase in the interleukin-13 receptor subunit alpha-2 (IL13RA2) and both cell groups shared an enrichment for cell-cell adhesion (Fig. 6c, Supplementary Fig. 7a-b). When non-OA fibroblasts were treated with PRG4 + tryptase, an enrichment of retrograde endocannabinoid signaling (GNB3, ANXA6, PDHA1, NDUFB3, FGF2, GNB1) and mesenchyme development (DPPA4, PEF1, ANXA6) was identified over those treated with PRG4, using Metascape (Fig. 6d-f). Genome-wide pathway-based association analysis identified retrograde endocannabinoid signaling pathway associated with inflammatory arthritis49. Also, enrichment of genes associated with mesenchyme development including DPPA4 was associated with changes in phenotypes of cellular niche during early OA50. Overall, tryptase processing of PRG4 impact the proteomes of primary healthy human fibroblast promoting a phenotype that resemble OA phenotypes.
Tryptase β Cleaves PRG4 In Human Synovial Fluid
The molecular mechanisms implicated in OA initiation and progression remain elusive51. Importantly, PRG4 is found in high concentrations in the synovial fluid of joints (~ 400 µg/ml)52, which often changes in OA5,6. As proteases play key biological roles in synovial fluid and in OA53 − 55, we investigate if tryptase β can cleave PRG4 in complex ex vivo synovial fluid. As the collection of synovium fluid is challenging and limited for healthy patients, collected synovial fluid from cadaveric normal joints (n = 7) within 4 h of death, with no evidence of cartilage pathology on dissection, and incubated them with tryptase or vehicle (Fig. 7a). Proteins from healthy synovial fluid treated with vehicle were labeled with light formaldehyde (+ 28 Da dimethylation), while synovial fluid incubated with tryptase β for 1 h were labeled with heavy formaldehyde (+ 34 Da dimethylation) (Fig. 7a). To identify the cleavage sites (neo-N-termini), we subjected the treated synovial fluid to an N-terminomics/TAILS protocol, where the N-termini are enriched using the dendritic polyglycerol aldehyde TAILS polymer56 (Fig. 7a). The global proteomes were compared by a shotgun (pre-enrichment TAILS) proteomics analysis57 (Fig. 7a). After sample acquisition LC-MS/MS analysis, data were analyzed using MaxQuant35 at 1% FDR. Shotgun/preTAILS analysis of seven healthy synovial fluid samples incubated with vehicle or tryptase yielded 2,524 unique peptides corresponding to 1,498 unique proteins, and the TAILS analysis yielded 310 unique peptides, where 80 were identical to the pre-enrichment TAILS analysis (Fig. 7b, Supplementary Tables 2–6). In the preTAILS data, we identified a significant change of 4.8% of peptides in the tryptase β-treated synovial fluid and 5.8% in the buffer control (Fig. 7c, Supplementary Tables 2–3). We next analyzed the N-terminal processing in the tryptase-treated samples and identified predominantly internal N-termini (87.3%), in addition to other proteoforms, including signal peptide removal (7.7%) and alternative start sites (5.5%) (Fig. 7d, Supplementary Tables 4–6). Next, we generated IceLogos to determine cleavage site preferences between tryptase β-treated and buffer-treated synovial fluids58. As expected, we identified a preference for P1 lysine and arginine residues, and a preference for glycine in P1’ and P2’ position in the tryptase β-treated group (Fig. 7e). In the buffer-treated groups, we found a preference for P1 tryptophan and arginine residues (Fig. 7f). In the TAILS data of tryptase β-treated synovial fluid, we identified a cleavage site of PRG4 at position 1330K↓G1331 (Fig. 7g, Supplementary table 6), which was similarly identified by our ATOMS experiment (Fig. 1f). We additionally identified 14 new potential tryptase substrates including CXCL7/PPBP, ORM1, vimentin, fibrinogen alpha/beta/gamma chains, SAA1, SAA2, APOC3, APOE, fibronectin, gelsolin, sulfhydryl oxidase 1, and procollagen C-endopeptidase enhancer 1 (Fig. 7h, Supplementary table 6). Using Metascape, we generated a protein-protein interaction networks by merging all significantly changed proteins (Supplementary Fig. 8b) and found minimal overlap between the tryptase β and buffer-treated cells (Supplementary Fig. 8a,c). Using STRING-db59, we identified an enrichment of the activation of C3 and C5, ECM proteoglycans, and toll-like receptor cascade in the buffer-treated synovial fluid (Supplementary Fig. 8d). When analyzing the pre-enrichment TAILS data, we identified an enrichment of cytokine signaling in immune system and adaptive immune response in buffer that is lost in tryptase-treated synovial fluids, which suggests an effect on the N-termini processing as identified the TAILS data (Fig. 7k). Therefore, we identified new potential human tryptase β substrates and validated that PRG4 is an endogenous tryptase β substrate.