Altered Protein Levels in Bone Marrow Lesions of Hip Osteoarthritis: Analysis by Proteomics and Multiplex Immunoassays

Abstract

Aim:

To assess tissue level changes of proteome and cytokine profiles of subchondral bone in hip osteoarthritis (OA) affected by bone marrow lesions (BMLs). We compared significant protein level differences in osteoarthritic bone with BMLs to control bone without bone marrow lesions.

Methods:

Subchondral bone biopsies were taken from femoral heads of end-stage osteoarthritis patients with (BML, n = 21) and without (CON, n = 9) BMLs. Proteins were extracted through a standardized Trizol protocol and used in the subsequent analyses. Angiogenesis and bone markers were assessed using multiplex immunoassays (Luminex). Liquid chromatography tandem mass spectrometry (LC-MS/MS) was performed to detect significant differences in proteome and peptide profiles between BML and CON.

Results:

Multiplex immunoassays revealed increased tissue contents of vascular endothelial growth factors (A/C/D), endothelin-1, angiopoietin-2 and interleukin-6 in bone with BMLs compared to control bone, whereas osteoprotegerin levels were reduced. Mass spectrometry demonstrated pronounced increase in the levels of hemoglobin (73-fold), serum albumin (30-fold), alpha-1-antitrypsin (9-fold), apolipoprotein A1 (4.7-fold), pre-laminin-A/C (3.7-fold) and collagen-alpha1-XII (3-fold) in BMLs, while aggrecan core protein (ACAN) and hyaluronan and proteoglycan link protein 1 (HAPL1) decreased 37- and 29-fold respectively.

Conclusion:

Reduced osteoprotegerin, ACAN and HAPL1 are consistent with osteoclastic activation and high remodeling activity in BMLs. The pronounced increase in angiogenesis markers, hemoglobin and serum albumin support the presence of increased vascularity in subchondral bone affected by BMLs in OA. VEGFs and IL-6 are known nociceptive modulators, and increased levels are in keeping with pain being a clinical feature frequently associated with BMLs.

Introduction

Substantial challenges in early diagnosis and treatment of osteoarthritis (OA) still prevail. This is partly due to continued dispute regarding the relative roles of bone and cartilage in OA pathogenesis and limitations of diagnostic tools, including relevant imaging modalities (radiography and MRI)^{1, 2}. These limitations have led to the search for new biomarkers, including markers of bone, cartilage and synovial metabolism³. Despite evidence of differential expression of a number of genes between early and late OA, there is considerable overlap in the activated biological pathways⁴. Therefore, combined use of several markers seems to be necessary to facilitate early detection and improve the prediction of disease progression¹.

Significant research activity has centered around identifying different phenotypes of OA and several clinical phenotypes have been described. Clinical phenotypes of OA with chronic pain, inflammatory mechanisms, metabolic mechanisms of bone and cartilage local to the joint, metabolic syndrome, mechanical overload, and other phenotypes have been suggested^{5, 6}. OA affects the change of every articular tissue over time, leading to different clinical phenotypes depending on the most damaged tissue at any given time. When subchondral bone injury is the main event, bone pain due to BMLs is the prominent manifestation and is expressed as the typical water signal on MRI⁷. Classification based on the underlying disease process in bone, might therefore be a valuable to further elucidate OA pathophysiology and to optimize patient selection for treatment.

Using MRI imaging, BMLs have been observed in a wide range of non-inflammatory and degenerative pathologies, prompting the notion that BMLs may represent a universal response to injury⁸. BMLs are seen in up to 80% of symptomatic hip and knee OA patients^{9, 10} and their association with OA pain and progression is well established making BMLs an imaging biomarker for OA¹¹. This prompted us to look for other, molecular biomarkers of OA progression in BMLs. The purpose of this study was therefore to compare osteoarthritic samples with and without bone marrow lesions with respect to protein levels using shotgun proteomics and the bone and angiogenesis multiplex panels in the Luminex system.

Methods

Study population

The study population consisted of cases (BML) and controls (CON), and the characteristics of the subjects in each group is presented in table 1. In brief, thirty patients with end-stage primary hip OA were recruited. This study population had a mean age of 64.2 (9.9) years and met American College of Rheumatology criteria for OA¹³.

Table 1.

Characteristics of the study population

Patients with hip OA	Without BMLs (n = 9)	With BMLs (n = 21)
Age, yrs	62.1 (12.1)	65.3 (9.1)
Female, %	67	52
Body mass index, kg/m2	27.8 (5.0)	25.9 (4.3)

Among the thirty patients recruited, nine femoral heads were found to be without BMLs. These participants had a mean age of 62.2 (12) years and were considered as controls. Comparison of mean age and body mass index between BML and CON revealed no significant differences. The study was approved by the Regional Committees for Medical and Health Research Ethics in South-East Norway (2011/1089/REK). All patients were given oral and written information, and written informed consent was obtained from each of the participant patients, in accordance with the Declaration of Helsinki.

Sample preparation

Protein fractions were extracted from pulverized and homogenized subchondral bone biopsies taken out of the femoral heads. The extraction site for the biopsies was determined by visual evaluation of MRI (Fig. 1) in coronal and axial planes as described elsewhere¹². In brief, regions of subchondral bone affected by BMLs were demarcated in pairs of images from each plane prior to excision of the core biopsies. Articular cartilage was removed, and the bone tissue samples were snap-frozen with liquid nitrogen and ground manually using mortar and pestle. For each patient sample, proteins from two samples of 100 mg of homogenized bone were extracted using Trizol (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer instructions. Total protein content in each sample was measured using the BCA-assay (BCA Protein Assay kit, Thermo Fisher Scientific).

Fig. 1:

Biopsy site. The excision site of the biopsies (red circles) was determined by visual evaluation of MRI in coronal and axial planes. Regions of subchondral bone affected by BMLs were demarcated in three sequential images from each plane prior to excision of the core biopsies

Quantification of cytokines and angiogenesis markers

The levels of dickkopf-related protein 1 (DKK-1), fibroblast growth factor 23 (FGF-23), interleukin-1β (IL-1β), interleukin-6 (IL-6), insulin, leptin, osteocalcin (OC), osteopontin, (OPN), osteoprotegerin (OPG), sclerostin (SOST), parathyroid hormone (PTH), angiopoietin-2, endoglin, endothelin-1, vascular endothelial growth factor (VEGF)-A, VEGF-C and VEGF-D were quantified using a human bone (HBNMAG-51K) and angiogenesis panel (HAGP1MAG-12K) in the Luminex-200 system (Luminex, TX, USA). Assays were performed according to the manufacturers’ instructions.

Proteomics by liquid chromatography-tandem mass spectrometry (LC-MS/MS)

Precipitation

Protein extracts from individual samples in each group, CON and BML, respectively, were pooled prior to precipitation with four volumes of cold (−20 °C) acetone. Next, the samples were centrifuged at 13,000 RCF for 25 min, and the protein pellets were dissolved in 250 μL of denaturing buffer (8 M Urea in 50 mM TEAB, pH 8.5). Total protein concentrations were measured by a colorimetric protein assay using a microplate absorbance reader (Tecan Austria GmbH, Grödig, Austria). Dilutions of γ-microglobulin (Bio-Rad, Hercules, CA, USA) were used as standards.

Digestion

The protein extracts were reduced by adding 25 μL 100 mM dithiothreitol (DTT) solution to a final concentration of 10 mM, and incubated for 1 h at 37 °C. Alkylations of free sulfhydryl groups were done by adding 22 μL 250 mM iodoacetamide (IAA) solution to a final concentration of 20 mM, and incubated at 25 °C for 45 min in the dark. Urea concentration was reduced in the samples by adding 750 μL of 50 mM TEAB (pH 8.5) buffer, prior to digestion. Enzymatic digestion was performed with Lys-C/trypsin for 6 h at 37 °C. The digested samples were dried under nitrogen stream.

Dimethyl labeling

Protein extracts were dimethyl labeled using isotopomers of formaldehyde (CH₂O) and sodium cyanoborohydride (NaBH₃CN)¹⁴. Briefly, digested samples from each group were dissolved in 250 μL of 100 mM TEAB buffer (pH 8.5). 140 μL of 4% (v/v) CH₂O (light label) was added to the control group and 140 μL of CD₂O (intermediate label) was added to the stimulated group. Then, 140 μL of 0.6 M NaBH₃CN was added and the samples were incubated under rotation for 1 h at 22 °C. The labeling reaction was quenched by adding 560 μL of 1% (v/v) ammonia solution, mixed followed by brief centrifugation. Further quenching and acidification were performed by adding 280 μL of 5% formic acid. The labeled samples were then mixed in 1:1 ratio, evaporated and dissolved in 0.1% formic acid before nano LC-MS/MS analysis.

Nano LC-MS/MS

Tryptic digest separation of protein extracts was performed on a PepMap RSLC Easy-spray C18 column (2 μm, 100 Å, 75 μm x 150 mm) using the EASY-neck 1000 nano UHPLC system (ThermoFisher Scientific) connected to an LTQ-Orbitrap XL hybrid mass spectrometer (Thermo Fisher Scientific) equipped with a nano EASY-Spray source (Thermo Fisher Scientific). The analytical separation was run for 180 min using a multi-step gradient of 0.1% formic acid in water as solvent (A), and 0.1% formic acid in ACN as solvent (B). 0–25% eluent B was used in 150 min and 25–60% B in 20 min followed by 60% B in 10 min at a flow rate of 300 nL/min and a column temperature at 45 °C. The mass spectrometer was operated in positive mode with a spray voltage set at 2.0 kV and the heated capillary temperature was kept at 200 °C. The LTQ-Orbitrap XL was operated in data-dependent mode in which one cycle of experiments consisted of one full-MS survey scan using the Orbitrap mass analyzer and subsequently five sequential MS/MS events of the most intense peaks using collision-induced dissociation (CID) in the LTQ. The MS survey scans were performed on the high resolution Orbitrap (R = 30,000) with an m/z range of 350–2000. Precursor ions with charge 1 or unassigned charge were rejected, and the isolation width was set to 3 m/z.

Data analysis

Raw files were evaluated against a UniProt human database (June 2015) using Sequest HT in Proteome Discoverer 1.4 (Thermo Scientific). Precursor and fragment mass tolerances were set to 10 ppm and 0.6 Da, respectively. Only the peptides resulting from the tryptic cleavages were used, and two missed cleavage sites were allowed. Carbamidomethylation of cysteine (+57.021 Da) was selected as a fixed modification. The variable modifications were as follows; +15.995 Da for methionine oxidation, + 28.031 for dimethyl (K and N-term) light label and +32.056 for dimethyl (K and N-term) intermediate label. Peptide and protein false discovery rates (FDR) were set to 1% using default filters. Dimethyl datasets were quantified using peak area with the precursor ions quantifier node integrated in Proteome Discoverer with RT tolerance of isotope pattern multiplets set to 1 min. To correct for possible experimental bias, protein ratio distribution was normalized on protein median. The cut-off ratio for up- and down-regulated proteins was set at ≥ 2.0 and ≤ 0.5, respectively. Abundance ratios for proteins reported as differentially expressed in this study were confirmed by manual inspection of the MS spectra intensities of the labeled peptide pairs. Briefly, the criteria for passing the manual inspection were as follows: i) signal-to-noise ratios of both light and intermediate labeled peptide pairs ≥ 20, ii) light and intermediate labeled peptide ion spectra must show similar isotope patterns and expected mass shift between doublet clusters. Only unique peptides were considered for protein quantification.

Statistics

Patient age, body mass index and Luminex results were compared between BML and CON using one-way ANOVA followed by Šídák’s test for multiple comparisons. Normality was checked using D’Agostino-Pearson omnibus and Shapiro-Wilk tests. Values were presented as mean (SD) unless stated otherwise. In all instances, significance was assigned to p < 0.05.

Bioinformatics

In order to find known and predicted functional associations between differentially expressed proteins in the dataset, we used STRING database (Search Tool for the Retrieval of Interacting Genes/Proteins)^{15, 16}. Each set of differentially upregulated/down regulated proteins where uploaded individually using the UniProt accession number, the number of K-clusters was set to 6 and stringency was set to high to correlate detailed biological information from the genes encoding the differentially expressed proteins. Gene-GO term enrichment analysis with modified Fisher Exact P-Value (p = 0 (perfect enrichment) and p < 0.05 (strongly enriched) was performed.

Results

Cytokines and angiogenesis marker levels

Luminex multiplex immunoassays demonstrated reduced tissue levels of OPG (Fig. 2A), and increased levels of IL-6 and the angiogenesis markers VEGF-A, VEGF-C, VEGF-D, endothelin-1 and angiopoietin-2 (Fig. 2B), in hips with BMLs compared to those without. No statistically significant differences were found for endoglin, leptin, DKK-1, OPN, SOST, PTH, FGF-23 or insulin between the two groups (Fig. 2A). For many of the samples within both groups, the values for OC and IL-1β were found to be above and below the upper and lower values of the standard curve ranges of the assay, respectively. Therefore, no meaningful interpretation could be made for these factors.

Fig. 2:

Cytokine and angiogenesis marker levels. Biomarker levels are assessed by Luminex in hips with (red) and without (black) BMLs. Total protein is given by logarithmic transformation of μg/mL protein. Cytokine and angiogenesis marker levels are given by logarithmic transformation of pg/mg bone tissue. DKK-1, dickkopf-related protein 1; FGF‐23, fibroblast growth factor 23; IL-6, interleukin-6; OPG, osteoprotegerin; OPN, osteopontin; PTH, parathyroid hormone; SOST, sclerostin; VEGF, vascular endothelial growth factor. Multiplicity-adjusted P -values: P 1 = <.001, P 2 = .006, P 3 = .007, P 4 = .002

Significant differential changes in BMLs proteomic profiles

In total, 106 proteins were differentially expressed in BMLs; 23 proteins were significantly (p < 0.05) upregulated while 20 were down regulated in BML compared to control (Fig. 3). Several proteins were markedly upregulated in the BML group, with hemoglobin subunit beta, serum albumin and IgG1 chain C region exhibiting the highest upregulation, with 73- and 30- and 12-fold increases, respectively. The remaining upregulated proteins in BML showed fold-changes between 12 and 2. On the other hand, fibromodulin and the cartilage derived proteins, hyaluronan and proteoglycan link protein 1 (HAPL1) and aggrecan core protein, were downregulated 27-, 29- and 37-fold in BML compared to CON (Fig. 3).

Fig. 3:

Differentially regulated proteins. Protein levels in BML are illustrated as fold-changes compared to CON. Upregulated proteins are represented by the rows in red and the downregulated ones in blue

Bioinformatic analysis of differentially expressed proteins in BMLs indicated that most upregulated proteins were derived from red blood cells, serum, and plasma (Fig. 4 D–G). Moreover, upregulated peptides in BMLs had several biological functions. This included immunoglobulin kappa variable 3–20 (IGKV3–20), immunoglobulin kappa constant (IGKC), serine (or cysteine) peptidase inhibitor, clade F, member 1 (SERPINF1), apolipoprotein A-I (APOA1), apolipoprotein A-II (APOA2), transthyretin (TTR) and collagen type VI, alpha 3 (COL6A3) as signal peptides (Fig. 4 D). Some of the upregulated proteins were membrane components, including glyceraldehyde-3-phosphate dehydrogenase (GAPDH), profilin 1 (PFN1), ribosomal protein SA (RPSA), hemoglobin alpha 1 (HBA1) (Fig. 4 F). Others were associated with extracellular region, i.e. immunoglobulin lambda constant 2 (IGLC2), Immunoglobulin heavy chain (gamma polypeptide 2) (IGHG2) and albumin (Fig. 4 D). Few of the upregulated proteins were part of the extra cellular matrix (ECM), i.e. collagen type VI alpha 3 (COL6A3), collagen type XII alpha 1 (COL12A1) and SERPINA1 (Fig. 4 G). Some peptides were significantly (p< 0.01) associated with phagocytosis, proteolysis, receptor mediated endocytosis (Fig. 3 E). Downregulated proteins were mainly components of the ECM, including versican, collagen type I alpha 2 chain (COL1A2), clusterin (CLU), azurocidin 1 (AZU1), cathepsin G (CTSG), elastase, neutrophil expressed (ELANE), heat shock factor binding protein 1 (HSBP1) (Fig. 4 A), signal peptides such as heat shock factor binding protein 1 (HSBP1), HtrA serine peptidase 1 (HTRA1), calgranulin A (S100A8)) (Fig. 4 B), regulators of apoptosis, including S100 calcium binding protein A8 (calgranulin A) (S100A8), CLU and HSBP1 (Fig. 4 B) and participants in epigenetic regulation of transcription; histone cluster 1, H1c (HIST1HC), histone cluster 1, H1b (HIST1HB), H3 histone family member 3A (H3F3A) and heat shock protein 1 (HSPB1)) (Fig. 4 C). An overview over networks of protein-protein interactions and the biological or cellular processes associated with upregulated and downregulated proteins in this study are listed in figure 5A and 5B, respectively.

Fig. 4:

Functional annotation clustering of differentially regulated proteins in BML compared to CON. The heat map shows the protein-coding genes and their significant associated annotation term for the upregulated (red) or downregulated (green) proteins. Red/green squares: corresponding gene term association positively reported in the literature. White squares: corresponding gene-term association not reported yet

Fig. 5:

Protein-protein interaction networks. The network nodes represent the downregulated (A) and upregulated (B) proteins. The edges show protein-protein interactions with predicted functional partners. Blue lines represent known protein-protein interactions from curated databases. Purple lines represent experimentally determined interactions. Green lines represent gene neighborhood, while fading, grey lines represent protein homology. Black lines represent co-expression

Discussion

Bone plays an essential role in OA pathophysiology and the presence of BMLs in the subchondral bone is associated with pain, disease progression and important disease outcomes such as cartilage loss and joint replacement surgery^{7, 17}. These outcomes may be a consequence of dysregulated repair mechanisms associated with increased remodeling due to microdamage of subchondral bone tissue with BMLs. Our proteomics and multiplex data demonstrate significant differences in protein levels in hip bone affected by OA and BMLs. These differences indicate a central role for BMLs in the overarching context of OA pathogenesis. Biological pathways for angiogenesis, nociception and cartilage degeneration implicated by our findings are discussed in the two sections below.

Markers of angiogenesis and pain

Biomarkers in subchondral bone with osteoarthritic BMLs have not been measured previously. Despite the absence of directly comparable studies, our finding of increased expression of IL-6 in bone tissue with BMLs is in line with previous clinical research indicating its involvement in OA pain¹⁸ and several key elements of OA pathogenesis. IL-6 has been suggested to be as potent as VEGF in inducing vessel-sprouting¹⁹. The proinflammatory role of IL-6 in OA has been demonstrated in animal models, and increased levels have been found in non-calcified joint tissue^20–22. Serum levels of IL-6 are associated with increased knee BMLs in both females and males with OA²³. Our results are in line with other studies by demonstrating that levels of IL-6 are also increased in subchondral bone of the affected joints contributing to pathological angiogenesis and pain (Fig. 6). IL-6 induces angiogenesis and activation of the JAK-STAT3 signaling pathway, which is also activated by VEGF^{19, 24, 25}. IL-6 also induces matrix metalloproteinases, which play a role in ECM degradation^{26, 27}. In arthritis, IL-6 signaling in sensory neurons plays a role in nociception by increasing inflammatory swelling²⁸. Similarly, increased levels of IL-6 are related to pain caused by bone metastasis²⁹. Molecular mechanisms related to angiogenesis and pain, therefore, share common pathways, which may exert synergistic effects in arthritis. The role of IL-6 in serum as a biomarker for pain and progression of OA and cartilage loss has been demonstrated^{30, 31}. Longitudinal studies are still needed, however, to determine possible correlations between cytokine levels and measures of disease progression or severity. Interactions between the structural changes seen in the advanced states of the disease and the inflammatory response in the subchondral bone also need further investigation.

Fig. 6:

The biological role of differentially regulated cytokines and proteins in BMLs. The figure represents an overview of the differentially regulated cytokines and proteins in the biological context of angiogenesis, pain and bone remodeling

In the VEGF family, VEGF-A is the most widely studied and targeted isoform in the context of OA pathogenesis. Increased VEGF levels have been observed in osteoblasts from patients undergoing total hip replacement³², cartilage, synovial fluid, serum and meniscus of OA patients³³. Our observation of increased levels of all major isoforms of the VEGF family is therefore in agreement with previous studies, and supports our earlier findings of increased angiogenesis in BMLs. The proteins and growth factors involved in blood vessel growth, e.g. angiopoietins and the VEGF family, may contribute to inflammation^{34, 35}. Delivery of progenitor cells via blood vessels play a central role in endochondral ossification and perturbation of blood vessel growth, potentially through inflammatory pathways may lead to structural disease progression. Neovascularization may also be linked to pain, because it is accompanied by the growth of sensory nerves that penetrate non-calcified articular cartilage, osteophytes and the inner regions of menisci³⁶. The influence of VEGF on pain may occur indirectly from VEGF-mediated stimulation of angiogenesis and sensory neurogenesis or inflammation. Additionally, VEGF may be acting directly on sensory neurons to produce nociceptive sensitization. VEGF, VEGFR1, and VEGFR2 signaling have been directly associated with hyperexcitability of sensory neurons, and inhibition of VEGF signaling led to reduction of pain sensitivity³³. Therefore, angiogenesis is associated with OA pain and represents a possible therapeutic target³⁷. Mice with increased expression of VEGF-C in bone exhibit increased osteoclast number, bone loss, lymphatic vessels in bone and a similar phenotype to Gorham-Stout disease (GSD). Individuals with this disease demonstrate massive bone loss associated with a profound angiomatosis of blood/lymphatic vessels in their bones³⁸. VEGF-D is closely related to VEGF-C, and both have N- and C-terminal extensions that are not found in other VEGF family members. VEGF-D is a ligand for the tyrosine kinases VEGFR-2 (Flk1) and VEGFR-3 (Ftl4). However, specific biological effects of increased levels of VEGF-D in bone are unknown, and the molecular effects of anti-VEGF antibody targeting nociceptive signaling sensitized by VEGF in OA pain remain to be established.

Markers of extracellular matrix turnover

Previous attempts to identify changes in protein levels associated with OA and BMLs have relied on standard biochemical assays measuring markers of bone, cartilage, and synovium turnover in serum and/or urine^{39, 40}, without direct measurements of protein levels in bone tissue. These studies have shown increases in serum hyaluronan (HA) (+233%)⁴⁰, serum cartilage oligomeric matrix protein (COMP; +16%)⁴⁰ and increases in urinary N-terminal telopeptide of collagen I (NTx)³⁹. However, the observed increase in NTx contradicted the observed decline in other bone turnover markers, including C-terminal collagen crosslinks in both serum and urine and decreased serum osteocalcin, in patients with knee OA⁴⁰. Therefore, direct measurement of the bone protein content as performed in the current study may more accurately reflect tissue level changes of the proteome and underlying molecular mechanisms in bone affected by BMLs.

Detection of some upregulated proteins usually present in blood in high concentrations (hemoglobin, serum albumin, IgG) is consistent with previous findings using immunohistochemical methods where a 4-fold increase of vascularity within BML compared to CON bone¹². IL-6 may also be implicated in the crosstalk between bone and cartilage, as it has been shown that it induces a phenotype in normal osteoblasts similar to what is observed in osteoblasts from sclerotic subchondral bone of OA patients. These osteoblasts downregulated aggrecan but upregulate metalloproteinase expression by chondrocytes in vitro⁴¹. IL-6 also alters expression of different chemokines, such as regulated and normal T cell expressed and secreted, and monocyte chemoattractant protein-1⁴². Our observations of reduced aggrecan core protein and increased IL-6 levels in BMLs would be in keeping with BMLs being involved in cartilage degradation.

Our bioinformatics results suggest that many downregulated proteins (collagen I alpha 1 and alpha 2, biglycan, aggrecan core protein, COMP, CHAD) probably reflect the loss of bone extracellular matrix. We observed for instance a 10-fold decrease of cartilage oligomeric matrix protein (COMP) in the BML compared to CON bone samples. COMP plays a role in the structural integrity of cartilage via its interaction with other extracellular matrix proteins such as collagens and fibronectin. This interaction of chondrocytes with the extracellular matrix of cartilage is mediated through interaction with cell surface integrin receptors⁴³. Downregulation of COMP in BMLs may play a role in the pathogenesis of OA as COMP is a potent suppressor of apoptosis in primary chondrocytes by blocking the activation of caspase-3 and by inducing the Inhibitor of Apoptosis family of proteins. Previous studies show increased serum COMP in OA and indicate increased cartilage turnover, potentially leading to the elimination of cartilage matrix proteins through the increased vascularity. Chondroadherin (CHAD) promotes attachment of chondrocytes, fibroblasts, and osteoblasts. CHAD-deficient mice show altered cartilage biomechanical properties⁴⁴ and bone turnover⁴⁵, indicating its potential role in OA.

Collagen I was also decreased in bone with BMLs, while collagen type XII was increased. Type XII interacts with type I collagen-containing fibrils via the COL1 domain and may be associated with the surface of the fibrils, and the COL2 and NC3 domains may be localized in the perifibrillar matrix. These finding are in agreement with increased bone resorption as reflected in elevated urinary NTx levels³⁹. The decrease in cartilage-associated proteins⁴⁶ is consistent with the previous reports⁴⁰, where loss of proteins controlling HA levels in cartilage and bone is a possible mechanism underlying the increase in serum HA.

Endochondral bone formation as a response to microdamage in subchondral bone has been suggested to be associated with OA initiation and progression⁴⁷. Proteins primarily associated with cartilage may have shared functions in bone. The cartilage associated protein HAPL1, for instance, stabilizes aggregates of aggrecan and HA, giving cartilage its tensile strength and elasticity. Mutations in HAPL1 of developing mice resulted in defects in cartilage development and delayed bone formation⁴⁸. The observed downregulation of HAPL1 and other proteins shared by cartilage and bone matrix in the current investigation may therefore reflect altered cartilage properties and pathological endochondral bone formation. Enhanced angiogenesis, as observed histologically in advanced hip OA¹², may contribute to the observed reduction in ECM derived proteins and promote alterations of the mechanical properties of the newly formed bone.

Small leucine-rich proteoglycans (SLRP) like biglycan and fibromodulin are essential in ECM turnover. They interact with collagen fibrils and limit the access of the collagenases to their cleavage sites. SLRP have been shown to be involved in OA pathogenesis, with the evidence mostly coming from knockout mouse models⁴⁹. Biglycan deficiency increases osteoclast differentiation and activity due to defective osteoblasts⁵⁰ and fibromodulin affects the rate of fibril formation in collagen fibrillogenesis⁵¹. Our observation of a pronounced downregulation of fibromodulin and biglycan supports increased collagenase activity in BMLs, and is also in agreement with the histological observation of increased bone turnover¹² suggesting altered ECM composition. Peptides or proteins that are associated with programmed cell death, e.g. CLU (clusterin), that interact with APOA1 and PON1 were also downregulated^{52, 53}.

Monitoring or targeting SLRPs may offer new prognostic or therapeutic modalities for OA⁴⁹. Downregulation of histones (HIST1HC, HIST1HB, H3F3A and HSPB1) indicate a pertubation in epigenetic regulation of transcription. Nucleosomes are the basic units of chromatin and are connected to one another by linker DNA which are bound by H1. These histone variants play significant roles in modulating the chromatin architecture, thereby influencing important biological processes. Serine Peptidase 1 (HTRA1) is a highly conserved family of serine proteases found downregulated in many pathologies by epigenetic mechanisms⁵⁴. It is also known that the loss of HTRA1 function in cells causes increased rates of proliferation, delayed onset of senescence, centrosome amplification, and polyploidy that suggest, HTRA1 implication in regulation of the cell cycle⁵⁵.

It is a strength of this study that bone samples from the same hips were previously characterized using immunohistochemistry and bone histomorphometry and compared to results from the present study. The correlation between results of the different analysis methods may add to the validity and reliability of the findings from the current study. Another strength of this study is the tissue level analysis of the subchondral bone compared to previous studies which were based on serum, urine or soft tissue samples. This study also has several shortcomings, however. It would have been advantageous to include samples from less severe stages of OA. Another limitation is the inability to compare the levels of proteins detected using Luminex to the proteins detected by the proteomic analysis.

Conclusion

Our results support the notion that BMLs in advanced OA cause pronounced alternations in the proteome and cytokine profile of the subchondral bone. The alterations in protein levels in BMLs demonstrated in the present study may partly explain the association between the repair response to biomechanical injury in subchondral bone, progression of cartilage loss and the development of OA pain. Further research is required to verify these findings and to explore the temporal and spatial relation between the reported changes and the progression of osteoarthritis in the affected joint.