Uncovering the ligandome of low-density lipoprotein receptor-related protein 1 in cartilage: a top-down approach to identify therapeutic targets

The low-density lipoprotein receptor-related protein 1 (LRP1) is a cell-surface receptor ubiquitously expressed in adult tissues. It plays tissue-specific physiological roles by mediating endocytosis of a diverse range of extracellular molecules. Dysregulation of LRP1 is involved in multiple conditions including Alzheimer’s disease, atherosclerosis and osteoarthritis (OA). However, little information is available about the specific ligand profile (ligandome) for each tissue, which would lead to better understanding of its role in disease states. Here, we investigated adult articular cartilage where impaired LRP1-mediated endocytosis leads to tissue destruction. We used a top-down approach involving analysis of human chondrocyte secretome, direct binding assays and validation in LRP1-deficient fibroblasts, as well as a novel Lrp1 conditional knockout (KO) mouse model. We found that inhibition of LRP1-mediated endocytosis results in cell death, alteration of the entire secretome and transcriptional modulations in human chondrocytes. We have identified more than 50 novel ligand candidates and confirmed direct LRP1 binding of HGFAC, HMGB1, HMGB2, CEMIP, SLIT2, ADAMTS1, IGFBP7, SPARC and LIF. Our in vitro endocytosis assay revealed the correlation of their affinity for LRP1 and the rate of endocytosis. Moreover, a conditional LRP1 KO mouse model demonstrated a critical role of LRP1 in regulating the high-affinity ligands in cartilage in vivo. This systematic approach revealed the extent of the chondrocyte LRP1 ligandome and identified potential novel therapeutic targets for OA.


Introduction
Endocytosis is the process by which cells control the composition of their cell surface and extracellular environment and is thus essential for cellular signaling and metabolism. The low-density lipoprotein receptor-related protein 1 (LRP1, CD91) is a type 1 transmembrane protein consisting of a 515 kDa heavy-chain containing the extracellular ligand-binding domains, and a non-covalently associated 85 kDa light-chain containing a transmembrane and cytoplasmic domain. LRP1 is ubiquitously expressed in different tissues and cell types [1,2], and mediates clathrin-dependent endocytosis of structurally and functionally diverse array of molecules including lipoproteins, extracellular matrix (ECM) proteins, growth factors, cell surface receptors, proteinases, proteinase inhibitors and secreted intracellular proteins [3,4]. LRP1 is not only involved in ligand uptake but also in modulation of cellular signalling pathways by binding extracellular ligands including growth factors and subsequent intracellular interaction with scaffolding and adaptor proteins. The LRP1 intracellular domain contains two tyrosine phosphorylation sites, which can directly transmit the signals [5].
OA is a major cause of pain and disability and is the most prevalent age-related degenerative joint disease. With an expanding elderly population, OA imposes a major socio-economic burden on society [20]. Despite its prevalence, there is currently no diseasemodifying OA treatment available, except joint replacement surgery at the late stage of the disease. The hallmark of OA is loss of articular cartilage, which results in impairment of joint function. Chondrocytes, as the only cell type present in articular cartilage, are essential for balancing in anabolic and catabolic processes. Therefore, chondrocytes are key regulators of cartilage homeostasis and joint health. Our previous study has shown that LRP1 plays a key major role in maintaining the ECM turnover of healthy adult cartilage by controlling extracellular levels of matrix-degrading metalloproteinases such as aggrecanases [21,22], collagenases [23,24], and their endogenous inhibitor tissue inhibitor of metalloproteinases (TIMP)3 [25,26]. However, this endocytic process is impaired in cartilage with OA, resulting in prolonged presence of LRP1 ligands at extracellular milieu shifting homeostatic cartilage matrix turnover to a catabolic state [19]. The impairment of LRP1 is due to an increase in the proteolytic shedding of the ectodomain of the receptor from the cell surface and we have previously identified matrix metalloproteinase 14 (MMP14) and a disintegrin and metalloproteinase 17 (ADAM17) as the major LRP1 sheddases in human OA cartilage.
Importantly, inhibition of LRP1 shedding recovered the endocytic capacity of the cells and reduced ECM degradation in OA cartilage [19]. Dysregulated LRP1 shedding is also observed in various pathological conditions such as rheumatoid arthritis, systemic lupus erythematosus [27], and in cancer [28,29].
To date, more than 80 LRP1 ligands have been reported in the literature [4,30] and the list is still growing. However, very limited information is available on how LRP1 ligand profiles, i.e., its ligandome, differ in specific tissues including cartilage. This is partly due to that tightly regulated LRP1 ligands are rarely detectable in the tissue unless LRP1-mediated endocytosis is blocked [22,23,26,31]. Uncovering the depth of this "hidden" ligandome is essential to identify novel molecular interactions, and possibly therapeutic opportunities. In this study, we took a systematic top-down approach to uncover the LRP1 ligandome in healthy cartilage. We have identified 38 novel LRP1 interaction partners whose excessive activities exerts detrimental consequences for chondrocytes and cartilage tissue integrity.

Inhibition of LRP1-mediated endocytosis induces chondrocyte cell death
Inhibition of LRP1-mediated endocytosis is essential to enable the identification of tightly regulated LRP1 ligands. We thus decided to block LRP1-mediated endocytosis in human chondrocytes using either 10 nM soluble form of full-length LRP1 (sLRP1) or 500 nM receptor-associated protein (RAP). sLRP1 binds to LRP1 ligands and acts as a soluble decoy receptor competing with the endogenous LRP1, whereas RAP antagonises ligand binding by competitively occupying the ligand binding region of LRP1 (Fig 1A). Initially, we evaluated the effect of sLRP1 and RAP on chondrocytes morphology and viability. We found that 72-h incubation with either the inhibitor or soluble receptor resulted in altered chondrocyte morphology and a significant number of the cells detaching from the plate surface (Fig 1B). The MTS cell metabolic assay showed that compared to 0-h, approx. 53.7% and 57.7% of the cells were dead after 72-h incubation with sLRP1 or RAP, respectively (Fig 1C). The time-dependent effect of sLRP1 and RAP on the cell death ( Fig   1D) indicates that the accumulation of LRP1 ligands gradually leads to chondrocyte cell death.
The ligand binding regions in LRP1 lie in four clusters (clusters I-IV), containing 2 to 11 individual ligand binding cysteine-rich complement-like repeats. Clusters II and IV are considered to be responsible for most ligand binding [32][33][34]  and TIMP3 [35]. Thus, we next tested the effect of the soluble form of LRP1 ligand-binding cluster II (sLRP1-II) on chondrocyte cell death. After 24-h incubation with 40 nM sLRP1-II, approx. 60% of the cells were dead compared to 0-h (Fig 1E). We further examined the dose-dependent effect of sLRP1-II and found that 24-h incubation with 20 nM and 10 nM sLRP1-II resulted in approx. 25% and 9% of cell death, respectively, whereas 5 nM and 2.5 nM sLRP1-II showed negligible effect (Fig 1E). The highest concentration of sLRP1-II that does not induce the cell death up to 24-h incubation (5 nM) was thus employed for LRP1 ligand identification.

Alteration of the entire chondrocyte secretome by inhibition of LRP1-mediated endocytosis
We performed a complete chondrocyte secretome analysis to determine the effect of inhibition of LRP1-mediated endocytosis and identify LRP1 ligands whose excess activity exerts detrimental consequences for chondrocytes. This analysis identified 635 proteins including 197 secreted, 415 intracellular and 23 transmembrane molecules according to the Uniprot annotation. Twenty-four hour-treatment of 5 nM sLRP1-II resulted in significant alteration (p-value <0.05) of 417 molecules with an increase in 34 secreted, 253 intracellular, and 10 transmembrane proteins and a decrease in 104 secreted, 9 intracellular and 7 transmembrane proteins (Fig 2A-C and Suppl Table I -III). The secreted proteins significantly (>2.5-fold difference and p-value <0.05) increased or decreased in the presence of sLRP1-II were listed on Table I or Table II, respectively. Among the 34 secreted proteins increased by the sLRP1-II treatment, only 10 molecules have previously been reported as LRP1 ligands (Fig 2C). Unexpectedly, a total of 21 previously reported LRP1 ligands were decreased (12) or unaltered (9) by the sLRP1-II treatment (Fig 2D). These results suggest that the identified chondrocyte LRP1 ligandome consists of mostly novel ligands and that the 21 previously reported LRP1 ligands are not regulated by LRP1-mediated endocytosis at least in human chondrocytes.
To validate the chondrocyte secretome analysis, hyaluronan-binding protein (CEMIP, also known as KIAA1199 and HYBID), slit homolog 2 protein (SLIT2), high-mobility group protein (HMG)B2 and tumour necrosis factor (TNF)-inducible gene 6 protein (TSG6, TNFAIP6 in the plot) were further investigated by Western blot analysis. Endogenously produced CEMIP and HMGB2 were rarely detectable in the conditioned medium of human chondrocytes after 24-h incubation without sLRP1-II but markedly increased in the presence of 5 nM sLRP1-II (Suppl Fig 1A and B). The sLRP1-II treatment increased the levels of SLIT2 and TSG6 in the medium of chondrocytes approx. 2.5-and 2.8-fold, respectively, compared to the control (Suppl Fig 1C and D).

Inhibition of LRP1-mediated endocytosis induces transcriptional modulations.
It has been reported that RAP increases mRNA levels of pro-inflammatory mediators such as TNF, interleukin-6 and C-C motif chemokine ligand 2 in macrophages [36]. We next investigated effect of inhibition of LRP1-mediated endocytosis on transcriptional modulation of several LRP1 ligands and a non-ligand. Our recent [24] study identified MMP1, a collagenase highly elevated in knee joint and synovial fluid from patients with arthritis [37,38], as an LRP1 ligand. We found that endogenously produced MMP1 was not detectable in the medium of chondrocytes in the absence of RAP, whereas it was increased after 24 hincubation with 200 nM RAP (Fig 3A). Similar results were obtained for MMP3 (Fig 3B), which is not a LRP1 ligand but relevant to arthritis pathogenesis [37,38]. Notably, the quantitative PCR analysis revealed that mRNA levels of MMP1 and MMP3 in human chondrocytes are increased time-dependently in the presence of 200 nM RAP or 10 nM sLRP1 (Fig 3C and D). In contrast, denatured sLRP1 did not affect their mRNA expression, suggesting that functional sLRP1 is required for upregulation of mRNA levels of these proteinases. Similar results were obtained for other LRP1 ligands, MMP13 and ADAMTS4 (Fig 3E and F), whereas mRNA levels of ADAMTS5 or TIMP3 were not affected by inhibition of LRP1-mediated endocytosis (Fig 3G and H). These studies suggest that excessive activity of LRP1 ligands alters transcription of several molecules. In addition, a fraction of the molecules increased upon sLRP1-II treatment in our secretome analysis (Fig   2) might not be LRP1 ligands but due to transcriptional upregulation.

The chondrocyte LRP1 ligandome comprises a variety of novel ligands.
To enrich and identify molecules that directly interact with LRP1, we have employed coimmunoprecipitation assay for sLRP1-II containing a C-terminal 6xHis/c-Myc tag and c-Myc antibody. The isolation of LRP1 ligands using this method was first validated using ADAMTS5 (Suppl Fig 2). Purified ADAMTS5 (5 nM) was mixed with 5 nM sLRP1-II and anti-c-Myc antibody-coupled paramagnetic beads, recognising sLRP1-II. Complexes of sLRP1-II, ADAMTS5 and the magnetic beads were then isolated using a magnetic separator. Most of ADAMTS5 and sLRP1-II were detected in the bound fraction (Suppl Fig  2A). As expected, RAP competitively inhibited ADAMTS5 binding to sLRP1-II. Since LRP1 ligands dissociate from LRP1 in early endosome due to acidic conditions, we have tested whether acidic conditions (pH 3.0) can dissociate molecules bound to sLRP1-II/anti-c-Myc antibody complexes. As expected, ADAMTS5 were eluted under acidic conditions as effectively as 1% sodium dodecyl sulphate (SDS) or 100 nM RAP (Suppl Fig 2B). To identify LRP1 ligands produced by human chondrocytes, cells were incubated without (control) or with 5 nM sLRP1-II for 24 h and the molecules bound to sLRP1-II were isolated as described above (Fig 4A). The molecules isolated by the same procedure without sLRP1-II were considered as a negative control and used to estimate a fold-difference for the ligand identification. Western blot analysis of the isolated proteins using anti-His tag antibodies has confirmed the isolation of sLRP1-II added to the culture (Suppl Fig 3).
Mass spectrometry analysis identified 276 molecules co-immunoprecipitated with sLRP1-II from the medium of human chondrocytes. These molecules include 51 secreted proteins, whereas the remainder of the molecules are either intracellular or transmembrane proteins according to the Uniprot annotation (Suppl Table IV), supporting a notion that LRP1 plays a role in the clearance of cellular debris [30]. Among the 51 secreted proteins identified, 13 molecules were previously reported LRP1 ligands, while 38 molecules were novel LRP1 ligand candidates (Fig 4B and Table III). These 38 ligand candidates included 23 proteins that were significantly increased (19) or decreased (4) (>2.5-fold difference and p-value <0.05) by sLRP1-II treatment in the secretome analysis, respectively. The Protein ANalysis THrough Evolutionary Relationships (PANTHER) [39] gene ontology analysis on biological processes of these molecules that interact with LRP1 showed that 27, 20, 16 and 14 of them are involved in "Cellular processes", "Biological regulation", "Response to stimulus" and "Metabolic processes", respectively (Fig 4C). The Search Tool for Retrieval of Interacting Genes/Proteins (STRING) analysis [40] of protein-protein interactions further revealed functional and physical interactions of these molecules illustrated in (Fig 4D).

Newly identified LRP1 ligands exhibit a wide range of affinities
Based on the mass-spectrometry scores, reported pathophysiological functions and a feasibility to obtain purified proteins, we selected 11 novel ligand candidates for further characterization: hepatocyte growth factor activator (HGFA, gene name: HGFAC), inflammatory mediators high-mobility group protein (HMG)B1 and HMGB2, CEMIP, slit homolog 2 protein (SLIT2), secreted growth factor progranulin (gene name: GRN), proteoglycanase ADAMTS1, which was increased in medium of HEK293 cells treated with RAP in our previous study [41], tumour necrosis factor (TNF)-inducible gene 6 protein (TSG6, gene name: TNFAIP6), ECM glycoprotein fibulin1C (gene name: FBLN1), insulin-like growth factor-binding protein (IGFBP)7, secreted protein acidic and rich in cysteine (SPARC, also known as osteonectin), which is previously identified as one of the proteins co-isolated with sLRP1 binding clusters II and IV [30] but no evidence of direct LRP1 binding provided, and leukemia inhibitory factor (LIF). To confirm their direct binding and estimate their affinity for LRP1, solid-phase binding assay were performed using purified ligand candidates and full-length sLRP1. We found that HMGB1 (6.4 nM), HMGB2 (3.9 nM), CEMIP (2.6 nM), SLIT2 (2.0 nM), ADAMTS1 (20.2 nM), TSG6 (11.0 nM), IGFBP7 (11.0 nM) and SPARC (41.0 nM) directly bind to immobilised LRP1 with high affinity (apparent binding constant (KD,app) indicated in parentheses) (Fig 5B-E, GH and JK). HGFA (>500 nM) and LIF (>200 nM) also directly bind to immobilised LRP1 but their affinities for LRP1 were much weaker than the high affinity ligands mentioned above (Fig 5A and L). In contrast, LRP1 binding to progranulin or fibulin-1C was not observed at all (Fig 5F and I). These results indicate that the molecules co-isolated with sLRP1-II include not only direct LRP1 binders but also non LRP1 binders that may interact with LRP1 indirectly, possibly via true ligands.

LRP1-mediated endocytic clearance of novel LRP1 ligands
Newly identified proteins that directly bind to LRP1 were further tested for their endocytic clearance by cells via LRP1. Recombinant HGFA, HMGB2, CEMIP, SLIT2, ADAMTS1, TSG6, IGFBP7 and SPARC (each 10 nM) were incubated with wild-type (WT) mouse embryonic fibroblasts (MEFs) or LRP1-deficient MEFs [42] for 8 h and these molecules in conditioned medium and cell lysate were detected by Western blot analysis using anti-FLAG or His tag antibody. The amounts of exogenously added HMGB2, CEMIP, SLIT2, ADAMTS1, TSG6 and IGFBP7 in the conditioned medium were reduced in WT MEFs by 59%, 92%, 65%, 62%, 80% and 45%, respectively, compared to LRP1-deficient MEFs, whereas no changes were observed for HGFA (Fig 6A-G). SPARC was also reduced in WT MEFs by 25% compared to LRP1-deficient cells but the differences appeared to be not statistically significant (Fig 6H). Exogenously added SLIT2, ADAMTS1, TSG6 and IGFBP7 were also detected in the cell lysate and the slightly lower amounts of SLIT2, TSG6 and IGFBP7 were detected in WT compared to LRP1-deficient MEFs (Fig 6D-G). These results indicate that if the same concentration of LRP1 ligands is present in the extracellular milieu, the ligands with higher affinity are taken up via LRP1-mediated endocytosis more rapidly than low affinity binders (Fig 6I and Table IV).
LRP1-mediated endocytic regulation of endogenously produced ADAMTS1 and IGFBP7 was further investigated by Western blot analysis of conditioned medium of human chondrocytes treated with 5 nM sLRP1-II for 24 h. The sLRP1-II treatment increased the levels of ADAMTS1 in the medium of chondrocytes approx. 10-fold compared to the control (Suppl Fig 4A), whereas no significant changes were observed for IGFBP7 (Suppl Fig 4B).

High-affinity LRP1 ligands are tightly regulated by LRP1 in knee articular cartilage tissue in vivo
Candidate ligands were further validated in vivo. Given the complexity of extracellular environment and a possible competition between LRP1 and sulphated glycosaminoglycans (GAGs) in articular cartilage [43,44], we investigated newly identified high-affinity LRP1 qPCR analysis of total RNA extracted from the mouse skin tissue showed >79% reduction of Lrp1 mRNA in the KO compared to wild-type tissue (Fig 7B). Histological investigation of LRP1 in knee joint tissue revealed that LRP1 is abundantly expressed in the cell membrane of chondrocytes at the superficial area of cartilage and of the cells in the medial meniscus but these immunosignals were markedly diminished in the KO mice (Fig 7C). SLIT2 and CEMIP were detected in some chondrocytes in the control whereas stronger immunosignal of these ligands was observed in LRP1 KO tissue (Fig 7D). Like SLIT2 and CEMIP, connective tissue growth factor (CTGF), an important LRP1 ligand [45,46], showed a negative correlation between absence of LRP1 and its level in cartilage in vivo.

Discussion
This study found that inhibition of LRP1-mediated endocytosis in human chondrocytes results in cell death, alteration of the entire secretome and transcriptional modulation. Our systematic approach revealed the extent of the chondrocyte LRP1 ligandome that illustrates novel biological roles of LRP1 in cartilage and potentially other tissues, and the pathological conditions when LRP1 ligands are present in the tissue for extended periods.
We showed that both sLRP1, which acts as a soluble decoy receptor and competes with endogenously expressed LRP1, and RAP, which binds to LRP1 and competes with endogenous ligand binding, induce chondrocyte cell death (Fig 1). Previous studies have indicated a role of LRP1 in cell survival and death [47][48][49]. In agreement with these studies, we found pro-cell survival role of LRP1 in human articular chondrocytes. Furthermore, several studies demonstrated cell survival and death functions for both previously reported and newly identified LRP1 ligands such as TIMP3 [50][51][52][53][54][55], ADAMTS5 [56], CTGF [57,58], CEMIP [59], SLIT2 [60,61], IGFBP7 [62][63][64] and HMGB2 [65]. The mechanisms by which LRP1 ligands regulate cell survival in chondrocytes remain unclear but the prolonged presence of one or several LRP1 ligands in the extracellular milieu may gradually alter cellular environment and metabolism leading to cell death.
The correlation between the affinity of LRP1 and the rate of endocytosis was previously demonstrated for ADAMTS4 and 5 isoforms [21]. The data presented here also supported this notion for various LRP1 ligands. Importantly, several of LRP1 ligands bind to both LRP1 and sulphated GAGs, with their extracellular availability determined by their relative affinity for each [43,44,66]. The extracellular environment in articular cartilage is a complex matrix where chondrocytes are surrounded by a thin layer of pericellular ECM followed by an abundant ECM mainly consisting of aggrecan proteoglycan and type II collagen fibrils. Nonetheless, markedly increased SLIT2, CEMIP and CTGF in knee articular cartilage from conditional LRP1 knockout mice (Fig 7), indicate that LRP1 is a master regulator for extracellular availability of high-affinity ligands in the tissue in vivo.
The chondrocyte secretome analysis revealed a huge impact of LRP1 blockade with significant changes in 368 molecules. LRP1 ligands can affect each other, as well as molecules that do not interact with LRP1, in several ways including transcriptional modulation, complex formation and proteolytic degradation. It has been reported that RAP increases mRNA levels of pro-inflammatory mediators such as TNF, interleukin-6 and C-C motif chemokine ligand 2 in macrophages [36]. Furthermore, considering functional diversity of LRP1 ligands, transcriptional modulation of several molecules shown in this study could be the tip of the iceberg. Negligible binding of progranulin and fibulin-1C to LRP1 suggests that these molecules indirectly interact with LRP1 via LRP1 ligands (Fig 5). This notion is supported by the interaction networks formed by co-isolated proteins (Fig 4D). To date, several extracellular proteolytic enzymes, which cleave a broad range of substrates, have been identified as LRP1 ligands [4,5,44]. Proteolytic degradation by LRP1 ligands is thus likely to contribute to reduction of the number of secreted proteins in particular ECM molecules. Indeed, levels of versican, nidogen-2 and biglycan, all previously reported substrates for ADAMTS1 [67][68][69], were reduced in the medium of chondrocytes treated with sLRP1-II (Table II).
This study generated new information about the pathobiological functions of the key molecules that are regulated by LRP1-mediated endocytosis and their connection to the development of OA. Extracellular HMGB molecules function as alarmins, which are endogenous molecules released upon tissue damage to activate the immune system. HMGB2 KO mice exhibit earlier onset of OA and a more severe progression, which is associated with a profound reduction in cartilage cellularity [65]. By contrast, HMGB1 acts as a late mediator of inflammation and contributes to prolonged and sustained systemic inflammation in subjects with rheumatoid arthritis (see [70] for review). Importantly, hyaluronan degradation is highly associated with the risk of OA progression [71,72]. CEMIP functions as a hyaluronan depolymerase [73] and is highly relevant to OA pathogenesis [74][75][76]. Expression levels of CEMIP and CTGF are correlated with disease severity in OA patients [77][78][79], and we found that their levels are increased in cartilage from LRP1 KO mice in vivo (Fig 7). Notably, hyaluronan is endocytosed via clathrin-coated pits and degraded in vesicles in HEK293 cells overexpressing CEMIP [80]. LRP1 not only regulates extracellular availability of CEMIP but may also play a role in transporting both CEMIP and hyaluronan to the coated pits where hyaluronan is degraded. TSG6 also binds to hyaluronan [81] and its protein and mRNA levels are increased in OA cartilage [82,83]. In the presence of inter-α-inhibitor (IαI), which was identified as a LRP1 ligand candidate (Table III), TSG6 catalyzes formation of anti-inflammatory stabilised hyaluronan-aggrecan assembly [84][85][86].
However, it blocks hyaluronan-aggrecan assembly in the absence of IαI [87]. Furthermore, our previous study demonstrated that IαI is proteolytically cleaved by ADAMTS5, MMP3, MMP7 and MMP13 [88]. Considering that CEMIP, TSG6, ADAMTS5 and MMP13 are all regulated by LRP1, increased LRP1 shedding in OA effectively leads to cartilage destruction by both increasing extracellular activity of cartilage-degrading proteinases [19] and destabilizing cartilage ECM assembly.
Accumulating evidence suggests a role of LRP1 in the regulation of Wnt signalling, which governs a myriad of biological processes underlying the development and maintenance of adult tissue homeostasis. Macrophage LRP1 increases Wnt/-catenin pathway by directly binding to and effectively removing secreted frizzled-related protein 5, which prevents Wnt binding to its receptor [86]. LRP1 also stimulates Wnt/-catenin pathway and prevents intracellular cholesterol accumulation in fibroblasts [87]. A recent study also demonstrated that LRP1 regulates expression of Wnt family member 4 in chondrocytes through transforming growth factor-β1 signalling in chondrocytes [88]. On the other hand, it has been reported that LRP1 interacts with Frizzled1 and downregulates Wnt/-catenin pathway in HEK293T cells [89]. Newly identified LRP1 ligands in this study are also involved in the Wnt pathway. For example, HMGB2 enhances the binding of lymphoid enhancerbinding factor 1 (Lef-1), a crucial transcription factor for Wnt/-catenin pathway, to its target sequence and potentiates transcriptional activation of the Lef-1--catenin complex [89]. The HMG domain within HMGB2 is crucial for interaction with Lef-1, suggesting that both HMGB2 and HMGB1 may be involved in this function. CEMIP also enhances Wnt/-catenin signalling and bone formation of osteoblastic stem cells [90]. Moreover, we have identified WNT5a and WNT11 as LRP1 ligand candidates (Table III). These studies emphasise a role of LRP1 in the Wnt pathway and we are currently investigating how the interaction between LRP1 and these WNTs affects Wnt signalling pathways.
This study provided important information about possible roles of LRP1 in other tissue contexts and diseases. SLIT2 binds to Roundabout 1 and 2 receptors [91] and plays a crucial role in neuronal outgrowth in embryogenesis [92,93]. HGFA, which directly binds to LRP1 and was increased in the medium of chondrocytes by sLRP1-II treatment, is a serine protease responsible for converting HGF into the active two-chain form [94][95][96]. HGF is considered to play a major role in the repair and regeneration of various tissues, including the liver, kidney, lung, and stomach [97]. Notably, both HGF activation and tissue regeneration were markedly impaired in injured intestinal tissue of HGFA-deficient mice [95,96], indicating the important function of HGFA during tissue repair process. Although HGFA is the major HGF activator, previously reported LRP1 ligands including urokinase-type plasminogen activator, tissue plasminogen activator and coagulation factor XI were also able to activate HGF in vitro [98,99]. Another newly identified LRP1 ligand, ADAMTS1, is broadly expressed and essential for normal growth, structure, and function of the kidney, adrenal gland, female reproductive-organ, and myocardial trabeculation during heart development (see [100] for review), and limb joint development [101]. Interestingly, recent studies revealed emerging roles of CEMIP in bone development [90,102], inflammation and antimicrobial activity [103]. A study using CEMIP-deficient mice revealed that hyaluronan metabolism by CEMIP is involved in endochondral ossification during postnatal development by modulation of angiogenesis and osteoclast recruitment at the chondro-osseous junction [102]. In another study, CEMIP-deficient mice showed no hyaluronan digestion and significantly less evidence of skin infection after an intradermal bacterial challenge by Staphylococcus aureus [103]. This information about the physiological functions of newly identified ligands implies LRP1 in organ morphogenesis, tissue regeneration and infection.
The methodology established in this study could be useful for other tissue contexts and diseases. One of the limitations of such an approach is that several LRP1 ligands were discovered with conflicting functions. For example, both proteases and their specific inhibitors were identified as LRP1 ligands. It is thus difficult to predict any net outcome of LRP1-mediated endocytosis and its dysregulation from the results of this study. However, our elucidation of a cartilage-specific ligandome opens potential new avenues of therapeutic intervention in diseases characterized by dysregulated LRP1-mediated endocytosis. By designing agents able to modulate the endocytosis of cartilage specific LRP1 ligands, it will be possible to counteract the negative effects of LRP1 dysregulation. For example, variants of TIMP3 resistant to LRP1-mediated endocytosis [104] and soluble LRP1 subcluster that selectively block TIMP3 endocytosis [35] have been already engineered to combat excessive ADAMTS5 activity. Combination of functional studies using cells or animal models and more complete tissue specific LRP1 ligandome identification is required for better understanding of the physiological functions of LRP1, and the detrimental effects of its dysregulation.

Reagents and antibodies
The sources of materials used were as follows: Pierce™ Anti-c-Myc magnetic beads, RNase (E200A) [105], ProMMP3 (E202A) [105] were prepared as previously reported. All other reagents used were of the highest analytical grade available.
Purification of full-length sLRP1 according to the manufacturer's protocol. LRP1 was extracted from fresh human placenta as previously described [106], and the placental mixture was applied to the RAP affinity column

Isolation and culture of human articular chondrocytes
Healthy (normal) articular cartilage was obtained from the Stanmore BioBank, Institute of Orthopaedics, Royal National Orthopaedic Hospital, Stanmore from patients following informed consent and approval by the Royal Veterinary College Ethics and Welfare Committee (Institutional approval URN 2010 0004H). Articular cartilage was obtained from the femoral condyles of the knee following amputation due to soft tissue sarcoma and osteosarcoma with no involvement of the cartilage. Tissues were obtained from 5 patients (3 males aged 18, 23 and 57 years; 2 females aged 19 and 68 years) and chondrocytes were isolated as described previously [26]. The cells were grown and maintained in DMEM/F12 media supplemented with 10% FBS at 37 °C with 5% CO2. Both primary and passaged (< 3 times) human cells were used in the study.

Cell-survival assay
Human chondrocytes were grown on 24 well plate with DMEM/F12 containing 10% FBS.
Once the cells reached 90% confluency, the medium was removed and the cells were rested Statistical analysis using the search results and intensities from Proteome discover was made with Perseus (v1.6.5.0) [107]. Initially the results were filtered based on the Protein FDR to remove low confidence IDs. The raw intensities were log2 transformed and missing values were replaced based on the normal distribution of the entire dataset. A two-sample ttest was performed to identify significant changes and calculate the fold change. Volcano plot analysis was performed using GraphPad Prism 9. Several proteins were identified in the sLRP1-II preparation thereby all intensities were normalized to the intensity of sLRP-II preparation in the sample. Exogenously added sLRP1-II and associated RAP were excluded from the data set of secretome and co-immunoprecipitation analysis.

Western blot analysis of secreted proteins produced by human chondrocytes
Human chondrocytes were incubated for 24 h with SF DMEM/F12 containing CT1746 (100 µM) and the protease inhibitor cocktail (1/500) in the presence and the absence of 5 nM sLRP1-II. After incubation, media were collected, and the protein was precipitated with trichloroacetic acid and dissolved in 20 μl of 1x SDS-sample buffer (50 mM Tris-HCl pH 6.8, 5% 2-mercaptoethanol (2ME), 2% SDS and 10% glycerol). All samples were analyzed by SDS-PAGE and Western blotting using a specific antibody against each protein.
Immunoreactive bands were quantified using NIH ImageJ, and the relative amount of each ligand in the medium in the presence and absence of sLRP1-II was calculated.

Quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR)
Human chondrocytes were incubated for 24 h with SF DMEM/F12 containing CT1746 (100 µM) and the protease inhibitor cocktail (1/500) in the presence and the absence of 200 nM RAP, 10 nM full-length sLRP1 or 10 nM denatured full-length sLRP1 (preincubated at 98 °C for 10 min) for 0-24 h. cDNA was generated from these cells using a reverse-transcription kit (Applied Biosystems, CA, USA) and random primers from RNA extracted and prepared using the RNeasymini kit (Qiagen, CA, USA) following the manufacturer's guidelines. cDNA was then used for real time PCR assays using TaqMan technology. The  threshold cycle (Ct) method of relative quantitation was used to calculate relative mRNA levels for each transcript examined. The 60S acidic ribosomal protein P0 (RPLP0) gene was used to normalize the data. Pre-developed primer/probe sets for MMP1, MMP3, MMP13, ADAMTS4, ADAMTS5, TIMP3 and RPLP0 were purchased from Applied Biosystems.

Functional analysis of proteomics data
The PANTHER Classification System [39] was used to determine gene ontology and biological processes of the identified proteins. The functional and physical networks analysis for the identified proteins were performed using the STRING version 11.0 [40]. Extrapolated KD,app values were estimated based on one-phase decay nonlinear fit analysis using GraphPad Prism 9.

Endocytosis assay using WT and LRP1 deficient MEFs
WT and LRP1 KO MEFs were generated as described previously [42] and kindly provided containing 10% 2ME. The medium was then completely removed and the cells were lysed with 50 μl of 2x SDS-sampling buffer containing 5% 2ME. All samples were analyzed by SDS-PAGE under reducing conditions and immunoblotting using anti-His tag or anti-FLAG M2 antibody. Immune signals for exogenously added proteins in the medium were quantified using ImageJ and the amount of the protein remaining in the medium after 8 h incubation was calculated as a percentage of the amount before incubation.

Tissue processing for histology
Animals were killed by CO2 and knee joints were removed by using sanitised scissors. The tissues were fixed in 4% paraformaldehyde for 24 h, stored in 70% ethanol followed by decalcification with 10% EDTA (pH 8.0) for 2 weeks. The decalcified tissues were then processed through graded ethanol and xylene before being embedded in paraffin wax. 4.5 μm sections were cut using a rotary microtome RM2235 (Leica), adhered to microscope slides, then dried overnight at 37°C. Sections were dewaxed and rehydrated with xylene followed by a series of decreasing ethanol concentrations prior to the histological analysis.

Immunohistochemical staining of LRP1 and LRP1 ligands
Slides were incubated with the basic-based antigen unmasking solution by placing a polypropylene Coplin staining jar filled with the respective retrieval solution into a water bath at 92-95 °C for 3-8 minutes. Slides were washed with distilled water followed by PBS and blocked with 0.3% hydrogen peroxide for 15 min at 37 °C. Slides were then incubated with Avidin solution for 15 minutes followed by the Biotin solution for 15 minutes at room temperature. After washing with PBS, slides were incubated with PBS containing 10% goat serum and 0.1% BSA for 3 hours at room temperature. Slides were then incubated with anti-LRP1 β-chain rabbit monoclonal antibody (ab92544), anti-SLIT2 rabbit monoclonal antibody (ab134166), anti-CEMIP rabbit polyclonal antibody (21129-1-AP) or anti-CCN2 rabbit polyclonal antibody (ab6992) for overnight at 4 °C. Rabbit IgG was used as an isotype control. Slides were washed with PBS containing 0.1% Tween 20 (PBST) three times for 10 minutes each. Slides were then incubated with secondary antibody ImmPRESS peroxidasemicropolymer conjugated horse anti-rabbit IgG for 30 minutes at room temperature followed by incubation with Vectastain solution (Vector Labs, PK-6100) for 30 minutes at room temperature. After wash with PBST three times for 10 minutes each slide was stained with 3,3'-diaminobenzidine and counterstained with Methyl green solution for 5 minutes followed by dehydration and mounted with DPX mounting media. Images were acquired using a Nikon Eclipse Ti-E microscope (Nikon, Tokyo, Japan).      A, Workflow of isolation and identification of LRP1 ligands in human chondrocytes from four different donors. B, Venn diagrams showing secreted proteins that were co-isolated with sLRP1-II (red), secreted proteins increased (orange) or decreased (blue) in medium of chondrocytes treated with sLRP1-II (Fig 2). C, The PANTHER gene ontology analysis for biological processes of 51 secreted proteins co-isolated with sLRP1-II from the medium of human chondrocytes. D, The STRING functional and physical interaction map of 51 secreted proteins co-isolated with sLRP1-II from the medium of human chondrocytes.     TGFB2  THBS2  C1R  TIMP1  SERPINEA3  TIMP2   C3  MDK  CALR  THBS1  MMP1  SERPINC1  PZP  IGFBP3  A2M   HPX  EMILIN3  SEPP1  SLIT2  TIMP3  CEMIP  GARS  LIF  PLAT  HGFAC  GRN  CTGF  HMGB2   CILP2  SMOC2  WNT11  HSP90AB1  HSP90AA1  HMGB1  ANXA1  PPIA  GDF15  TXN  LGALS1  CCDC80  TNFAIP6   ANXA2  C4A  TNC  PSAP  CTSD  ITIH2  F2  SERPINF2   HPX  TIMP3  PLAT  CTGF  HSP90AB1  HSP90AA1  PSAP  CTSD  F2