Lubricin binds cartilage proteins, cartilage oligomeric matrix protein, fibronectin and collagen II at the cartilage surface

Lubricin, a heavily O-glycosylated protein, is essential for boundary lubrication of articular cartilage. Strong surface adherence of lubricin is required given the extreme force it must withstand. Disulfide bound complexes of lubricin and cartilage oligomeric matrix protein (COMP) have recently been identified in arthritic synovial fluid suggesting they may be lost from the cartilage surface in osteoarthritis and inflammatory arthritis. This investigation was undertaken to localise COMP-lubricin complexes within cartilage and investigate if other cartilage proteins are involved in anchoring lubricin to the joint. Immunohistochemical analysis of human cartilage biopsies showed lubricin and COMP co-localise to the cartilage surface. COMP knockout mice, however, presented with a lubricin layer on the articular cartilage leading to the further investigation of additional lubricin binding mechanisms. Proximity ligation assays (PLA) on human cartilage biopsies was used to localise additional lubricin binding partners and demonstrated that lubricin bound COMP, but also fibronectin and collagen II on the cartilage surface. Fibronectin and collagen II binding to lubricin was confirmed and characterised by solid phase binding assays with recombinant lubricin fragments. Overall, COMP, fibronectin and collagen II bind lubricin, exposed on the articular cartilage surface suggesting they may be involved in maintaining essential boundary lubrication.

in the synovial fluid (SF), synovial membrane and superficial zone of articular cartilage [6][7][8][9] . Lubricin holds predominantly core 1 O-glycan structures with sialic acid giving the protein a negatively charged central domain capped by positively charged termini 10,11 . This glycan component likely contributes to the protein's low friction and non-adherent lubricating properties 12 . The extreme load and shear forces that the cartilage surface withstands means that the surface lubricin layer must adhere strongly to the cartilage surface 13,14 .
Articular cartilage, found in synovial joints such as the knee, hip and shoulder, functions foremost for load-supporting and load transferring between bones 15 . It is an avascular tissue with minimal cellularity resulting in a low proliferative environment making restoring arthritic damage difficult 15 . Articular cartilage is composed predominately of water, up to 70-80%, with the matrix forming component of the extracellular matrix (ECM) including a range of often highly post translationally modified proteins including collagens, proteoglycans such as aggrecan, laminins and fibronectin, reviewed in detail elsewhere 15,16 .
Collagen II is the most abundant collagen in articular cartilage and forms extended fibrils creating a network with other proteins including other collagens and proteoglycans 17,18 . Another fibril forming protein important for matrix formation is the glycoprotein fibronectin, which consists of two disulfide bound subunits of approximately 250 kDa 19 . Fibronectin aids ECM formation by interacting with a range of ECM components including collagen, where it may organise the ECM network 20 . COMP, a homopentameric glycoprotein of the thrombospondin family consisting of 100-110 kDa subunits synthesised by chondrocytes 21 , is of lower abundance in cartilage, however, its complex forming capabilities remain pivotal [22][23][24] . The globular C-terminal region of COMP is important for binding to aggrecan 22 , fibronectin 23 and collagen I, II and IX [24][25][26][27] , as well as forming non-covalent bonds with lubricin. COMP also forms covalent, disulfide bonds with lubricin, however, it is the N-terminal region (aa 102-191 and 287-504) involved in forming these bonds 28 . Lubricin has been shown to bind to other cartilage components including fibronectin in surface force apparatus experiments 29 , and to galectin-3 at the cartilage surface 30 . Lubricin, fibronectin and collagen II have been identified co-localised at the surface of engineered meniscal tissue 31 .
Our previous analysis has shown specific disulfide bonds and non-covalent interactions that complex lubricin, a protein of the cartilage and SF, with the cartilage protein COMP 28 . These complexes were identified in the SF of arthritic patients, suggesting that the complexes could have been lost from the articular cartilage surface as part of the disease process. Here, the nature and location of COMP-lubricin complex in the cartilage tissue was investigated to determine if this bond adheres lubricin to the cartilage surface. Analysis was also performed to identify other cartilage binding partners that may also be involved in the adherence of lubricin to the cartilage surface.

Results
Immunohistochemical co-localisation of lubricin and COMP on the cartilage surface. The co-localisation of COMP and lubricin was investigated by immunohistochemistry ( Fig. 1). Human cartilage biopsies including the subchondral bone were obtained from non-weight bearing ventral areas of the femoral trochlea peripheral to the femoropatellar joint during medial unicompartmental knee arthroplasty surgery. Both patients, (n = 2, both male aged 67 and 79 years of age at surgery) had primary, anterior idiopathic OA and intact cruciate ligaments. All biopsies were acquired from the border between macroscopically normal appearing cartilage and areas with obvious cartilage degradation. Matched isotype and normal serum negative controls confirmed staining specificity (Fig. 1f-h).
The representative tissue section in Fig. 1 contains an area of smooth, relatively normal cartilage surface (left side) and an adjoining distorted, highly OA affected area including disrupted cartilage with tear and cartilage flap (right side). There is a reduction in cells in the superficial zone and oedema, particularly under the disrupted cartilage surface. COMP ( Fig. 1c and e) was distributed throughout the interstitial matrix with stronger staining on the superficial layer; approximately 10 µm deep in the healthy tissue, up to 30 µm in the OA affected area. Lubricin ( Fig. 1d and e) staining in the less severely OA affected tissue (left hand side of slide) was strong on the very edge of the cartilage with lesser staining up to 25 µm in depth. In the OA affected area, strong lubricin staining was present up to 150 µm into the tissue. This shows that in more normal appearing cartilage surface tissue, lubricin and COMP are co-localised only on the superficial surface of the cartilage. In OA affected tissue, lubricin production appeared to be upregulated, possibly to re-establish a functional lubricative layer.
The distribution of Lubricin on the cartilage surface of COMP Knockout (KO) mice. Given that COMP KO mice do not show massive deformity compared to wildtype (WT) animals 32 , and it is known that lubricin is essential for boundary lubrication, immunohistochemistry (IHC) was used to evaluate the presence of lubricin on the cartilage of COMP KO mice. WT mice showed lubricin staining with the most intense staining in the superficial layer chondrocytes with little staining of the cartilage surface ( Fig. 2a and c). COMP KO tissue also stained with lubricin with clear staining in chondrocytes deeper into the tissue than observed in WT tissue as well as more lubricin apparent on the cartilage surface illustrated by the sharper staining at the surface edge ( Fig. 2b and d). Negative control is shown in Fig. 2e.
Although lubricin appeared more intense in COMP KO mice, it should be noted that the cartilage tissue in the COMP KO mouse seemed generally less dense as observed by the dense ECM obscuring the DAPI-stained nuclei in the WT tissue compared to the clear nuclei in the KO tissue ( Fig. 2c and d). This reduced density in KO mice could have allowed more efficient lubricin staining. It is clear, however, that lubricin was present on the cartilage surface in both WT and COMP KO mice, suggesting interactions other than COMP may be important in the adherence of lubricin to the cartilage surface in mice.

Proximity ligation assay (PLA) of lubricin and possible binding partners. Given this indication
that other binding partners may be involved in lubricin adherence to the cartilage surface, two proteins important in cartilage ECM formation, fibronectin and collagen II, were investigated in human tissue sections. PLA was used to directly identify protein complexes formed between lubricin and other likely cartilage binding proteins fibronectin and collagen II, as well as COMP, in situ 33 . Negative controls are shown in Fig. 3d-h.
As expected the PLA of lubricin and COMP (Fig. 3a) confirmed an interaction between the two proteins at the cartilage surface. Interestingly, the PLA signal from the COMP-lubricin complex was strong up to 40 µm into the cartilage, particularly intense where the surface layer was diminished. PLA signals were also detected for both the lubricin and fibronectin (Fig. 3b) and lubricin and collagen II (Fig. 3c) pairs. The latter two pairs were only identified at the cartilage surface with no protein interactions apparent deeper in the tissue. This suggests that both fibronectin and collagen II, along with COMP may be involved in the adherence of lubricin to the cartilage surface.
Solid phase binding assays. To verify that collagen II and fibronectin bind lubricin and to determine the lubricin binding region, solid phase binding assays were performed using a range of lubricin fragments that have been previously described 28 . Statistical analyses compared the binding of each recombinant lubricin fragment to bovine serum albumin (BSA) binding. Collagen II isolated from bovine cartilage bound the full length lubricin (lub) with a truncated mucin domain, the L871-1078 mucin domain construct and the N-terminal fragment L25-221 (Fig. 4b) to a similar extent. Four smaller N-terminal lubricin fragments were then tested and collagen II was found to bind only the L105-160 fragment (Fig. 4c) when compared to BSA. Human blood fibronectin bound with similar intensity to the full length-truncated mucin domain lubricin and L871-1078 mucin domain fragment; however, fibronectin binding was more than twice as intense to the C-terminal fragment L1078-1404 (Fig. 4d). Together, these results verify that lubricin binds collagen II and fibronectin and that the N-terminal of lubricin is important for binding with collagen II and the C-terminal for fibronectin.

Identification of lubricin from digested cartilage biopsy tissue.
To confirm the observed superficial lubricin was directly at the cartilage surface, pieces of biopsy tissue were digested with matrix metalloproteinase-9 (MMP-9). Tissue was digested with activated MMP-9 as well as incubated with the pro-enzymes. MMPs effectively digest proteins of the cartilage ECM. MMP-9 (gelatinase B) cleaves elastin, aggrecan, laminin, COMP and collagens IV, V, XI, XIV 34,35 . MMP-9 released lubricin from the pieces of decalcified OA patient articular cartilage biopsies into the buffer (Fig. 5a, complete Western blot shown in supplementary Fig. S1). Purified lubricin was also incubated with activated MMP9, which demonstrated that lubricin itself is not a substrate for the enzyme (Fig. 5b, complete Western blot shown in supplementary Fig. S1), eliminating the possibility that lubricin being released from the cartilage occurred through its direct digestion.

Discussion
Co-localisation of COMP and lubricin on the cartilage surface suggested the two are bound on the surface, an observation confirmed by in situ PLA. The presence of lubricin on the outer surface was confirmed by human biopsy digestion with the ECM destructive enzyme MMP-9 which released lubricin from the cartilage surface as identified by Western blot. Further PLA analyses identified that lubricin also binds fibronectin and collagen II at the cartilage surface. These data suggest a robust redundant mechanism where COMP, fibronectin and collagen II anchor lubricin to the cartilage surface to aid in the creation of the essential boundary lubrication of the joint.
The disulfide-bound COMP-lubricin complex was initially identified in arthritic SF by mass spectrometry 28 , however, the concentration of COMP is far greater in cartilage compared to SF 36 , suggesting the complex originates in the cartilage. The superficial layer is lost in OA through physical destruction or perhaps by enzymatic means such as increased MMP-3 degradation 37 . Lubricin and COMP co-localised at the superficial zone of cartilage, with lubricin expression deeper in areas of damage, indicating damaged cartilage tissue may attempt to reform the surface structure. Lubricin is expressed in the superficial zone of the cartilage in healthy tissue 6 , however, lubricin protein expression has been shown to be upregulated after mechanical injury 38 . Bovine explants showed lubricin staining deeper into the cartilage tissue in areas of injury 38 , as observed in this study in the damaged tissue. Altered COMP expression patterns in patients with arthritis have been identified 39 , and stronger expression observed in the superficial fibrillated cartilage was shown here. In situ PLA corroborated this result and confirmed that COMP and lubricin are bound at the cartilage surface. A large accumulation of COMP-lubricin complexes identified deeper into the cartilage tissue suggests that the complex may be produced in the cartilage as a precursor reservoir, and migrate through the tissue to be held on the cartilage surface providing a renewable source of the lubricin layer.
Given the necessity of the lubricin layer in maintaining joint articulation, it is possible that a redundant mechanism is involved. This is evidenced by the finding that COMP KO mice have normal development and histology of the skeleton 32 . Investigation of the superficial lubricin layer in COMP KO mice in our study demonstrated that the mice retain lubricin on articular cartilage. The lubricin layer even appeared more intense than in wildtype mice, although this is likely due to the loss of cartilage tissue density in the KO mice given COMP does not appear to be compensated for by other members of the thrombospondin family 32 . It is important to consider that the significance of COMP may be diminished in smaller animals, for instance, mice do not have COMP in their tendons 32 unlike larger animals including humans 40 and bovine 41 . Nevertheless, the identification of a lubricin layer on COMP KO mice articular cartilage suggests that other mechanisms may also be involved in lubricin retention in humans. Disruption to COMP folding by mutation causes devastating human diseases including multiple epiphyseal dysplasia (MED) 42 , and pseudoachondroplasia (PSACH) 43 , which can be replicated in mice 44 . PSACH causes disproportionate short stature, early onset OA and debilitating joint pain from childhood 45 , a result of the intracellular accumulation of misfolded COMP in the rough endoplasmic reticulum of chondrocytes 46,47 . This leads to the premature death of growth plate chondrocytes 48 and it is hypothesised that this, rather than the loss of COMP itself, results in PSACH 32,43 . Overall, although it is clear that COMP and lubricin complex at the cartilage surface, the loss of COMP does not lead to complete loss of boundary lubrication, hence additional mechanisms appear to be involved in the adherence of lubricin to the cartilage surface.
The two proteins chosen for investigation as possible lubricin binding partners, fibronectin and collagen II, are both important in ECM formation, and fibronectin has been reported to co-purify during enrichment of synovial lubricin 10 . Using in situ PLA, we demonstrated that both proteins are capable of forming proximity complexes in situ with lubricin. Unlike COMP, the fibronectin-lubricin and collagen II-lubricin complexes were only identified on the outer surface of articular cartilage from OA patients. The role of fibronectin in the cartilage superficial zone is not well characterised, however, it is found throughout the cartilage tissue, increased at the articular surface 49 and upregulated in OA 49,50 . Lubricin has been shown to tether to a fibronectin layer and this dual-protein layer Figure 3. In situ PLA of lubricin with possible cartilage binding partners. In situ PLA was performed on OA cartilage biopsy cryosections with the following pairs. In situ PLA uses two primary antibodies targeting proteins of interest and secondary antibodies conjugated to DNA oligonucleotides (PLA probes). If the two proteins of interest are in a complex the two PLA probes will be in close enough proximity to bind and facilitate formation of a circular DNA molecule using two additional DNA oligonucleotides which then undergo rolling circle amplification to enhance the signal. Fluorescently (red) labelled complementary DNA is then added in order to visualise the DNA. (a) Lubricin (PA3-118) and COMP (16F12). Bound lubricin and COMP (red) were identified at the cartilage surface. A second layer was also apparent below the surface. (b) Lubricin (5C11) and fibronectin (ab32419). Bound lubricin and fibronectin (red) were predominately identified at the cartilage surface. Only when the PLA signal was thinner at the surface was signal observed just below the surface. (c) Lubricin (5C11) and collagen II (sc-7763). Bound lubricin and collagen II (red) were identified only at the surface with no PLA signal further into the tissue. PLA signal are red, DAPI staining in blue. For the 3 antibody pairs, 5 separate antibodies were used as anti-lubricin 5CII was used for 2 pairs. Negative controls were performed by omitting each primary antibody for all of the five antibodies used. (d) lubricin negative control (PA3-118), (e) COMP negative control (16F12), (f) lubricin negative control (5CII), (g) fibronectin negative control (ab32419) and (h) collagen II negative control (sc-7763). Interaction between collagen II isolated from bovine cartilage and RC lubricin fragments, full length lubricin and BSA by solid phase binding assay. (c) Interaction between collagen II isolated from bovine cartilage and RC N-terminal lubricin fragments and BSA by solid phase binding assay. (d) Interaction between fibronectin isolated from human blood and RC lubricin fragments, full length lubricin and BSA by solid phase binding assay. For all assays n = 3 and error bars are standard deviation. N-designates the amino terminus and -C designates the carboxy terminus. The signal peptide (1-24) is shown in grey. All statistical analyses compare the binding of lubricin fragments to the binding of the BSA standard. * is defined as p ≤ 0.05, and *** is defined as p ≤ 0.001. provided coefficients of friction similar to that observed for lubricin alone suggesting strong binding between the two proteins resulting in effective wear protection 29 .
Collagen II is abundant throughout the articular cartilage 17 . Collagen has also been shown to interact with lubricin by quartz crystal microbalance with dissipation experiments. The lubricin bound spontaneously and could not be removed by interference with serum albumin 51 , again suggesting a strong bond resulting in a surface with a low coefficient of friction. Collagen II, along with fibronectin, was identified to co-localise with lubricin on engineered meniscal tissue, collagen gel seeded with fibrochondrocytes 31 . Here we have confirmed that both fibronectin and collagen II form complexes with lubricin in situ and these complexes are present at the superficial zone of articular cartilage. It has also been shown that Galectin-3, a carbohydrate-recognition-binding-domain containing protein, binds lubricin at the cartilage surface, enhancing boundary lubrication 30 . Overall, this redundant binding mechanism shows that the retention of lubricin on the cartilage surface is essential for cartilage health.
The binding of collagen II and fibronectin to lubricin was confirmed by solid phase binding assay using collagen II from bovine cartilage and fibronectin from human blood binding to recombinant lubricin fragments. Both proteins were shown to bind to some extent to the mucin domain, however, given the possibility of the involvement of glycosylation in this interaction, further analyses are needed to better understand the recombinant glycosylation compared to normal lubricin to determine if the binding observed is specific. Lubricin bound to fibronectin strongly via the C-terminal fragment (L1078-1404). Lubricin binding to collagen II, on the other hand, is focused to the N-terminal of lubricin and when smaller N-terminal fragments were used could be narrowed to the L105-160 fragment. The unglycosylated N-terminal of lubricin 11 has been shown to be important for noncovalent and covalent bonding with COMP 28 . In fact, solid phase binding assays between recombinant COMP and the same lubricin fragments used here showed that the lubricin amino acid range of 105 to 202 was responsible for non-covalent binding with COMP, overlapping with the binding regions (L105-160) important for binding to collagen II. The adherence at the termini of lubricin to other cartilage proteins including COMP, collagen II and fibronectin would allow the heavily O-glycosylated negatively charged region 11 , suggested to provide the low friction properties of lubricin 12 , to move freely, extending from the articular cartilage surface into the SF (as illustrated in Fig. 6). This extended nature would create a flexible surface that, even with multidirectional movement, would flex to always cover the surface.
In summary, lubricin complexes with COMP, collagen II and fibronectin at the articular cartilage surface, as shown by in situ PLA, leaving the lubricin mucin domain exposed at the surface. The COMP-lubricin complexes were also identified deeper into the tissue in this report, suggesting they are produced in the cartilage and migrate to the surface. Although we do not know the binding mechanism involved in these complexes, or how synovial lubricin may also play into this complex formation, it is clear that lubricin is able to bind collagen II as well as fibronectin in situ on the cartilage surface. COMP also binds fibronectin 23 and collagen II 24 , suggesting that a large multi-protein complex may be created at the surface. Understanding the mechanism of the attachment of lubricin to the cartilage surface becomes even more significant as the use of lubricin as a biologic to treat osteoarthritic disease becomes more likely 52 . Overall, it is clear that a multifaceted network is involved to ensure the adherence of lubricin, and as we continue to elucidate the mechanisms involved in retaining boundary lubrication, we become closer to the ability to restore it for the millions of people who need it.

Methods
Human SF and cartilage and mouse samples and preparation. Human cartilage biopsies (n = 2), both male aged 67 and 79 years of age, including the subchondral bone were obtained from non-weight bearing ventral areas of the femoral trochlea peripheral to the femoropatellar joint during medial unicompartmental Figure 6. Illustration of the cartilage surface showing protein interactions. Lubricin is blue with glycosylation shown in light blue, COMP is red shown as the pentameric structure, fibronectin is green shown as the dimeric structure and collagen II is purple shown as collagen II fibrils. The SF is shown in light straw, the cartilage in yellow and the cartilage surface in pink. COMP, collagen II and fibronectin adhere lubricin, primarily by the termini of lubricin, to the cartilage surface creating an exposed lubricin layer. The ECM proteins also bind to each other and are shown in the cartilage. Lubricin is also found in the SF. knee arthroplasty surgery with primary, anterior idiopathic OA and intact cruciate ligaments. Biopsies were collected at the Orthopaedic Clinic at Danderyds Hospital (Stockholm, Sweden) and were acquired from the border between macroscopically normal appearing cartilage and areas with obvious cartilage degradation. Biopsies were demineralised until soft enough to section in 175 mM EDTA, 10 mM Tris (pH 6.95), and 3.75% polyvinylpyrrolidone-40 at 4 °C. Enriched lubricin from SF samples was used as control for the MMP assay. SF sample from arthritic patients (n = 1 was collected during aspiration of knee joints at the Rheumatology Clinic, Sahlgrenska University Hospital (Gothenburg, Sweden). All OA and RA patients gave informed consent and all the procedures were approved by the regional ethical review board in Gothenburg (172-15,13/5-2015). All methods were performed in accordance with the relevant guidelines and regulations.
Paw tissue samples from 7 month old COMP deficient 129/sv mice described here as COMP KO mice 32 and equivalent wild type (WT) mice (n = 2) were used for immunohistochemistry analyses. This work, undertaken at Lund University, Lund, Sweden, was approved by the regional Lund-Malmö laboratory animal ethics committee (M9-08). All methods were performed in accordance with the relevant guidelines and regulations.
COMP KO mouse samples. Mouse paw tissue paraffin embedded sections from 7 month old COMP deficient 129/Sv mice 32 and age matched WT mice were analysed. Paraffin removal and then antigen retrieval were obtained by incubation in an alkaline solution (K8004 EnVision FLEX Target Retrieval Solution, High pH 9.0, DAKO, Denmark A/S, Glostrup, Denmark) at 85 °C for 40 min. Sections were cooled in washing buffer (K8007 EnVision FLEX Wash Buffer, DAKO) and blocked for 1 h at room temperature (RT) with 2% normal donkey serum in PBS (Jackson ImmunoResearch Laboratories, INC., West Grove, PA), also used for antibody dilution. Primary rabbit anti-lubricin antibody (P3-118, Thermo Scientific) at 2 µg/ml was incubated overnight at 4 °C followed by incubation with the secondary antibody for 1 h at RT with 3 × 5 min intermediary washes with washing buffer. The secondary antibody was either Rhodamine Red-X-conjugated, donkey anti-rabbit (711-296-162, Jackson ImmunoResearch, diluted 1 in 1000), or biotin-conjugated, donkey anti-rabbit (Jackson ImmunoResearch, diluted 1 in 1000) visualised after incubation with streptavidin-FITC for 30 min at RT. Negative control was performed with primary antibody omitted. Finally, sections were mounted as above and images were captured in a Zeiss Axioscope 2 Plus fluorescence microscope and images were cropped and contrast adjusted for the complete image in Adobe Photoshop CC.
Proximity ligation assay (PLA). Human cartilage biopsy tissue sections were permeabilised with 0.05% Triton in Tris-buffered saline (TBS) (50 mM Tris pH 7.6, 150 mM NaCl) for 15 min and rinsed twice in TBS. Tissue sections for collagen II antibody were also treated with hyaluronidase from bovine testes (H3506, Sigma-Aldrich) at 8000 U/ml in PBS, pH 5.5, for 1 h at 37 °C. Slides were blocked with Duolink Blocking solution (Olink Bioscience) for 45 min at 37 °C incubated in a humidity chamber. Primary antibodies were diluted in Duolink Antibody Diluent (Olink Bioscience) as follows: mouse anti-COMP (clone 16F12, Biovendor Research and Diagnostic Products) at 1 µg/ml, rabbit anti-lubricin (PA3-118, Thermo Scientific) at 1 µg/ml, mouse anti-lubricin (clone 5C11, Millipore MABT400) at 1 µg/ml, goat anti-collagen II α1 (COL2A1 (C-19), sc-7763, Santa Cruz Biotechnology) at 1 µg/ml, rabbit anti-fibronectin (ab32419, abcam) at 0.06 µg/ml, and incubated overnight at 4 °C. The slides were washed in TBS with 0.05% Tween-20 (TBS-T) for 3 × 5 min with gentle agitation as for all following washes. Secondary probes (Olink Bioscience) were diluted 1:5 in Duoink Antibody Diluent (Olink Bioscience) and incubated in a humidity chamber for 90 min at 37 °C. The slides were washed and ligation solution (Olink Bioscience) was added for 30 min at 37 °C, followed by washing and incubation with amplification solution for 90 min at 37 °C. Finally, the slides were washed with Buffer-B (Olink Bioscience) and counterstained with Hoechst (Life technologies) and FITC-conjugated phalloidin (Sigma) at RT for 10 min. Appropriate technical controls with primary antibody omission were included to test background levels. Images were captured using a Zeiss Imager Z2 microscope with AxioCam MRm Rev.3 camera and Zen pro 2011 software.
Solid phase binding assays. Recombinant lubricin fragments were produced and characterised as described previously 28 . FLAG-tagged recombinant human lubricin with a truncated mucin-like domain (without AA 403-870) was produced in 293 F cells using p3xFLAG-CMV-8 vector (Sigma-Aldrich) and purified on anti-FLAG beads. The same method was used to produce four lubricin fragments named L25-221 (all molecular weights are calculated, MW 21.9 kDa), L220-402 (MW 19. , were produced using pGEX-5X-3 vector (GE Healthcare) and Rosetta 2 E. coli (Novagen), purified in native conditions using glutathione beads (Pierce). Lubricin fragments have been recently characterised 28 . Fibronectin was purified from human blood 53 and collagen II was extracted from bovine cartilage 54 as previously described.
Fibronectin or collagen II (2 µg/ml in PBS) were coated overnight on a 96-well plate, rinsed twice with TBS and blocked for 1 h with 5% BSA in TBS. After a TBS rinse, the plates were incubated with lubricin fragments at 2 µg/ml in wash buffer (TBS + 0.1% BSA + 0.1% Tween-20) or BSA as a negative control, for 2 h. Plates were then washed three times with wash buffer and incubated with mouse anti-FLAG (Genscript) or rabbit anti-GST (Abcam) antibodies at 1 µg/ml in wash buffer for 1 h. The plates were then incubated with anti-mouse or anti-rabbit HRP-conjugated antibodies (DAKO) at 0.2 µg/ml in wash buffer for 1 h. After washing, binding was detected using TMB substrate (ThermoFisher), signal stopped with sulphuric acid and absorbance read at 450 nm. Assays were performed in triplicate and repeated twice, error bars are standard deviation. Data was analyses by one-way ANOVA followed by Bonferroni's Multiple Comparison Test to compare each of the fragments with the BSA standard using GraphPad Prism version 5.
MMP assay and Western blot analysis. MMP assay. Human MMP-9 protease was obtained as proenzymes (R&D systems, Inc.) and activated with 1 mM ρ-aminophenylmercuric acetate at 37 °C in reaction buffer (50 mM Tris, 10 mM CaCl 2 , 150 mM NaCl, 0.05% Brij 35, pH 7.5) according to the manufacturer's instruction. Four pieces of cartilage (20-25 mg) were removed from intact biopsy specimens after softening for sectioning. Cartilage pieces were rinsed with reaction buffer before the addition of 300 µl of reaction buffer and 1 µg of activated or pro-enzyme. Aliquots of each supernatant were removed at 1, 2 and 4 h. Enriched synovial lubricin was tested under the same conditions to evaluate its susceptibility to MMP-9.
Western blotting analysis. MMP assay samples were separated by 3-8% Tris/acetate gels. Gels were transferred to PVDF membrane, blocked (3% bovine serum albumin in PBS), and probed with mouse anti-human lubricin (mAb 13, Pfizer Research) followed by HRP conjugated rabbit anti-mouse immunoglobulins (DakoCytomation). Images are shown without alteration and complete Western blots shown in supplementary information Fig. S1. Data availability. All data generated or analysed during this study are included in this published article and its supplementary information.