Increased nerve growth factor expression and osteoclast density are associated with subchondral bone marrow lesions in osteoarthritic knees

Objectives Subchondral bone marrow lesions (BMLs) detected on magnetic resonance imaging in knee osteoarthritis (OA) are associated with knee pain, though the mechanisms remain unknown. Increased nerve growth factor (NGF) expression and osteoclast density in subchondral bone appear to be the key features associated with bone pain in knee OA. Therefore, we aimed to identify associations among NGF, osteoclasts, and BMLs in knee OA. Methods Twenty tibial plateaus were obtained from patients undergoing total knee arthroplasty for medial knee OA with BMLs at the medial tibial plateau (MTP). Osteochondral tissue samples from the weight-bearing part of the MTP, with and without BML, and from the weight-bearing part of the lateral tibial plateau (LTP), without BML, were collected. NGF expression and density of osteoclasts were compared among the three osteochondral tissue types. Results MTP bone with BMLs exhibited significantly higher NGF expression in bone marrow space and osteochondral channel, and higher osteoclast density than MTP bone without BML and LTP bone. The mean differences in NGF-positive area in the bone marrow space and the percentage of NGF-positive channels between MTP bones with and without BML were 9.0% (95% confidence interval [CI]: 5.9–12.1%) and 23.1% (95% CI: 11.3–35.0%), respectively. The difference in osteoclast density between MTP bones with and without BML was 0.6 osteoclasts per mm (95% CI: 0.3–0.9 osteoclasts per mm). Conclusions Increased NGF expression and osteoclast density are associated with subchondral BMLs in knee OA, contribute to understanding the mechanisms underlying BML-related bone pain in knee OA.


Introduction
Knee pain is the most common symptom and reason for seeking medical care in patients with knee osteoarthritis (OA).Knee OA is characterized by aberrations in the structural integrity of the joint including cartilage degeneration, synovial inflammation, and subchondral sclerosis with osteophyte formation.Recent clinical evidence [1][2][3] indicates that the subchondral bone is involved in the development of joint pain in OA.On magnetic resonance imaging (MRI), patients with knee OA exhibited subchondral marrow lesions (BMLs) more frequently than control patients without arthritis [1].Subchondral BMLs are characterized by ill-defined hypointensity on T1-weighted, non-fat-suppressed images and hyperintensity on fluid-sensitive, T2-weighted, proton density-weighted, intermediate-weighted, fat-suppressed, and short tau inversion recovery images [4,5].Subchondral BMLs are strongly associated with knee pain.Specifically, larger baseline subchondral BMLs were associated with greater baseline knee pain, and higher total subchondral BML volume was associated with increased knee pain [3].The presence or size of BMLs can predict future BML incidence and progressive pain [6].For varus knee OA, we have previously demonstrated that subchondral BML size in the medial femorotibial joint compartment was associated specifically with weight-bearing pain (rather than non-weight-bearing pain), more so than any other OA-related MRI features, such as effusion-synovitis, Hoffa's synovitis, cartilage defects, osteophytes, meniscus extrusions, anterior cruciate ligament tears, or age, sex, or body mass index [7].
The exact pathophysiology of subchondral BMLs is still subject to debate, and the cellular and molecular factors that mediate the association between subchondral BMLs and knee OA pain remain unknown.Subchondral BMLs are histologically characterized by bone marrow necrosis, trabecular abnormalities, bone marrow fibrosis, edema, cellular infiltration, and vascular proliferation [8,9].Microarray analysis of subchondral BMLs in OA has revealed the upregulation of genes implicated in neurogenesis, osteochondral turnover, and inflammation [9].We recently demonstrated that subchondral pathology is associated with knee OA pain independent of chondropathy and synovitis and that increased nerve growth factor (NGF) expression at the osteochondral junction and increased osteoclast density are key features associated with bone pain in knee OA [10].Subchondral BMLs may contribute to OA pain via the generation of chemical factors that sensitize the nerves within the subchondral bone [11].Therefore, we hypothesized that elevated NGF expression and osteoclast density are associated with BMLs in knee OA.A recent histological study reported an increase in the tartrate-resistant acid phosphatase (TRAP)-positive cells and the expression of general neuronal marker protein-encoding gene product 9.5 (PGP9.5) in samples comprising subchondral BMLs combined with cysts [12]; however, the results of that study were influenced by systemic differences in bone metabolism, as the study compared individuals with and without BMLs.
Therefore, in this study, we aimed to identify the associations among NGF expression, osteoclast density, and subchondral BMLs in symptomatic knee OA, by comparing subchondral bone with and without BML in the same individual.

Patient samples
To identify associations among NGF, osteoclasts, and BML in knee OA, tibial plateaus were obtained from patients undergoing total knee arthroplasty (TKA) for medial knee OA with BMLs in the medial tibial plateau (MTP).Patients were recruited between August 2019 and September 2020.The exclusion criteria were as follows: insufficient amount of osteochondral tissue for evaluation; lateral knee OA; medial knee OA without BML; use of medication for osteoporosis; neuropathic arthropathy; psychological disorders; and inability to fill in the questionnaires.Twenty patients with medial knee OA with BMLs undergoing primary TKA were included in this study.We collected osteochondral tissue of approximately 1 Â 1.5 cm in size with and without BML from the weight-bearing part of the MTP and osteochondral tissue without BML from the weight-bearing part of the lateral tibial plateau (LTP) (Fig. 1).Written informed consent was obtained from all patients, and the protocols were approved by the Kochi University Research Ethics Committee (IRB:31-74).

Magnetic resonance imaging (MRI)
The knees were scanned using a 1.5T MRI machine with a knee coil (SIGNA Architect 3.0T; GE Healthcare, Chicago, IL, USA).The imaging sequences included sagittal, coronal, and axial T2-weighted fat-saturated fast spin echo (repetition time: 2500 ms, echo time: 90 ms, slice thickness: 3 mm, gap between slices: 1.0 mm).Bone marrow lesions (BML) were identified as areas of irregular hyperintense signal in the subchondral bone on the fat-saturated T2-weighted images (Fig. 1) [4].The MRI scan was performed within one month before total knee arthroplasty (TKA)

Sample processing
Osteochondral tissues were fixed in neutral-buffered formalin and then decalcified by immersing them in Osteosoft solution® (Sigma-Aldrich Japan G.K) for three months at 4 C before being embedded in paraffin wax.Formalin-fixed tibial plateau sections (5 μm) were stained with hematoxylin and eosin or Safranin O-fast green.

Immunohistochemistry for NGF
Sections were subjected to antigen retrieval (10 mM citrate buffer, 90 C, 20 min) and were blocked with 5% bovine serum albumin containing goat serum, followed by incubation with rabbit monoclonal antibody against NGF (EP1320Y, Abcam, Cambridge, UK) and biotinylated goat anti-rabbit IgG secondary antibody (BA1000, Vector, Peterborough, UK).Immunoreactivity to NGF was visualized using avidin-biotin-peroxidase complex (Vector, Peterborough, UK) with nickel-enhanced diaminobenzidine development [15].The sections were counterstained with hematoxylin so that different regions were more apparent.The areas positive for NGF immunoreactivity in the subchondral bone were quantified.Briefly, the regions of interest (ROIs) in the subchondral bone were manually outlined.Positive staining was differentiated from the background by thresholding the image using hues to create a mask.Areas of positive staining and ROIs were measured.The fractional area was determined as the percentage of the area positive for NGF immunoreactivity within the ROI (Supplementary Materials).The proportion of NGF-positive channels at osteochondral junctions was measured.

Tartrate-resistant acid phosphatase (TRAP) staining
Differentiated osteoclasts were identified via TRAP staining using a commercially available kit (Sigma-Aldrich 387A, St. Louis, MO) following the manufacturer's protocol.TRAP-positive osteoclasts were counted within 400 μm of the cement line in the osteochondral junction and divided by the length of the subchondral bone to give osteoclast density, expressed as TRAP-positive cells per millimeter.One dark-purplish or reddish cell with !3 nuclei was recorded as one osteoclast [10].

Image analysis
Histological scoring and quantification were performed using an allin-one fluorescence microscope BZ-X800 with analysis using BZ-X800 Analyzer (Keyence), by a single observer (KA) blinded to the details of the sections.

Statistical analysis
All statistical analyses were performed using JMP 10 (SAS Ins.Cary, NC, USA) and IBM SPSS 26.0 (IBM Corp. Armonk, NY, USA).Osteoarthritis bone scores, OARSI grade and Mankin scores were analyzed using Friedman tests with a post-hoc Bonferroni test.Comparison of NGF expression in the subchondral bone and osteochondral channel, channel density, and osteoclast density was done using one-way repeated measures ANOVA with a post-hoc Tukey test among the three osteochondral tissue types.Comparison of NGF expression and osteoclast density between MTP bone with and without cysts was done using unpaired t-test.Spearman rank correlation coefficient assessed associations between NGF expression and osteoclast density.Differences with p < 0.05 were considered statistically significant.In a previous study [10] comparing samples from symptomatic and asymptomatic knee OA, 0.23 osteoclasts mm À1 was considered a clinically significant difference.Therefore, a difference of 0.23 osteoclasts mm À1 was used as the threshold here.To demonstrate a difference of 0.23 osteoclasts mm À1 with 90% confidence and α ¼ 0.05, it was determined that 20 patients were needed.

Results
The demographic details of the study participants are presented in Table 1, and the histological characteristics in Fig. 2 and Table 2.The OA bone score, OARSI OA cartilage histopathology assessment system and Mankin score of BML þ MTP bone were significantly higher than those of BML À MTP and BML À LTP bones (p < 0.01 versus BMLÀ MTP and BMLÀ LTP bones); further, these scores were significantly higher in BML À MTP bone than in BML À LTP bone (p < 0.01 versus BMLÀ LTP bone) (Table 2).
In the MTP samples, NGF immunoreactivity was detected in multinucleate osteoclast-like cells adherent to the bone, in fibroblast-like cells in fibrotic connective tissue, and mononuclear cells in the bone marrow space in the subchondral bone (Fig. 2).The NGF-immunoreactive cells were also found in osteochondral channels (Fig. 2).The NGF-positive area in the subchondral bone marrow space and NGF expression within the osteochondral channel were significantly higher in BML þ MTP bone than in BML À MTP and BML À LTP bones (p < 0.01), and the NGF-positive area and channel were significantly higher in BML À MTP bone than in BML À LTP bone (NGF-positive area: p < 0.05, NGF-positive channel: p < 0.01).For subchondral MTP bone, the mean differences in NGFexpressing area and percentage of NGF-positive channels between bone with and without BMLs were 9.0% (95% confidence interval [CI]: 5.9-12.1%)and 23.1% (95% CI: 11.3-35.0%),respectively (Table 2, Fig. 3).
TRAP-positive multinucleated osteoclasts were observed on the surface of the subchondral trabecular bone (Fig. 2).Subchondral bone osteoclast density was significantly higher in BML þ MTP bone than in BML À MTP and in BML À LTP bones (p < 0.01).For subchondral MTP bone, the mean difference in osteoclast density between the groups with and without BML was 0.6 osteoclasts per mm (95% CI: 0.3-0.9osteoclasts per mm).No significant difference in osteoclast density was observed between BML À MTP and BML À LTP bones in the LTP (Fig. 3).
Based on the analyses of the relationship of NGF expression and osteoclast density in subchondral MTP bone, there were significant correlations between NGF-positive area and channel, and osteoclast density (area: Spearman's r ¼ 0.50, 95% CI: 0.24-0.71;channel: Spearman's r ¼ 0.52, 95% CI: 0.23-0.70).Cyst-like lesions are frequently observed in BML þ MTP bone, and TRAP-positive osteoclasts which increase mainly on the surface of the cysts (Fig. 2).Based on analyses of associations between NGF expression and osteoclast density with subchondral bone cysts, MTP bone with cysts had a significantly higher percentage of NGFpositive area and channels, as well as greater osteoclast density, compared to MTP bone without cysts (NGF-positive area and channel: p < 0.01, osteoclast density: p < 0.05).The differences in the percentage of NG-positive area and channel were 7.0% (95% CI: 3.2-10.7%)and 22.4% (95% CI: 11.3-33.4%),respectively, and the difference of osteoclast density was 0.4/mm (95% CI: 0.02-0.8/mm)(Fig. 4).

Discussion
In this study, we demonstrated increased NGF expression and osteoclast density in subchordal bone with BMLs compared with those in subchondral bone without BMLs in the same individual.NGF -positive cells and osteoclasts in subchondral bone with BMLs can induce sensitization of bone afferent neurons, and changes in subchondral bone quality may increase mechanosensitivity.BML-like histopathological changes are associated with subchondral innervation [13].For subchondral bone with BMLs, we observed elevated NGF expression in bone marrow space and osteochondral channel, and levels of osteoclasts, which can promote nerve growth; these findings advance the earlier findings on innervation in subchondral bone in knee OA.BMLs may contribute directly to OA pain, and our results contribute to the understanding of the cellular and molecular factors that mediate the association between subchondral BMLs and knee OA pain.Our findings showed elevated NGF expression in the subchondral bone marrow space and at the osteochondral junction in patients with BML.Furthermore, the elevated NGF expression was also associated with subchondral bone cysts.Our current results extend findings from a previous study [12], wherein subchondral BMLs with cysts were shown to exhibit increased osteoclastogenesis and nerve distribution.Increased NGF immunoreactivity was detected in multinucleate osteoclast-like cells adherent to the bone, fibroblast-like cells in the fibrotic connective tissue, mononuclear cells, and cartilage islands in the subchondral bone.NGF may directly activate sensory neurons that express TrkA and modulate the expression of TrkA or the p75 receptor [16].Anti-NGF antibodies can reduce OA pain, indicating the importance of NGF in pain generation [17,18].Increased numbers of NGF-immunoreactive cells in the osteochondral channels and bone marrow space could contribute to OA pain by increasing colocalized sensory nerve activity.Microarray studies in OA BML tissues have revealed the upregulation of genes implicated in neurogenesis [9].Anti-NGF monoclonal antibodies provide significant improvements in pain and function; however, the use of tanezumab leads to more frequent adverse events, namely abnormal peripheral sensation and rapidly progressive OA, than the use of nonsteroidal anti-inflammatory drugs [19,20].
Studies of the effects of osteoclast inhibitors, such as bisphosphonates, denosumab, and strontium ranelate [23], have revealed joint pain reduction in people with knee OA.Zoledronic acid, a bisphosphonate, reduced knee pain and BML size in patients with OA [24], although findings from a meta-analysis of randomized controlled trials did not support the analgesic effects of bisphosphonates in knee OA [25].Our data suggest that there are multiple mechanisms underlying OA knee pain with multiple sources and that targeting osteoclasts has clinically important benefits only in cases of subchondral BML, in which osteoclast activity is the predominant driver of pain.Further studies are needed to develop treatments targeting BML-related pain in patients with knee OA.Subchondral cyst-like lesions are sometimes observed in knee OA and commonly develop within regions of BML and adjacent to cartilage abnormalities [26].Previous histological studies showed subchondral cysts containing fibrous connective tissue, osteoblasts, adipocytes [27], and TRAP-positive cells [12], which is consistent with our results.Another study demonstrated that subchondral BMLs with cysts had increased osteoclastogenesis compared with subchondral BMLs without cysts.However, in this study, there were no significant differences in osteoclast density between subchondral BMLs with and without cysts, possibly due to the small sample size (data not shown).Therefore, we compared osteoclast density and NGF expression between MTP bone with and without cysts.We found that MTP bones with cysts had a significantly higher NGF expression (reported for the first time in this study) and osteoclast density, compared to MTP bones without cysts.In a large-scale observational study, MRI-assessed subchondral cysts were associated with the development of knee pain in knee OA [28].Our results may explain one mechanism by which subchondral cysts are related to knee pain in knee OA.
Recently, interventions directly targeting subchondral bone BMLs have been attempted.Subchondroplasty, which utilizes an injectable synthetic calcium phosphate bone void filler to treat BMLs, significantly improved pain relief and knee function [29].Extracorporeal shockwave therapy achieved good results in treating the treatment of BMLs [30,31].Injecting mesenchymal stem cells into the subchondral bone can potentially treat BMLs [32], while the evidence for these interventions remains limited.However, our histological results, which reveal increased levels of pain-related factors in subchondral BMLs, provide support for interventions and treatments directly targeting BMLs.
A previous study has suggested that subchondral BMLs might mediate mechanically induced pain, such as during weight-bearing activity [7].Biomechanical factors may contribute reciprocally to the pathogenesis of  subchondral BMLs.Increased mechanical load due to malalignment of the knee joint is a risk factor for the occurrence or enlargement of femorotibial joint subchondral BMLs [33,34].The meniscus and cartilage can act as shock absorbers that protect the subchondral bone from overload.The increase in subchondral BML size with worsening meniscal pathology [35,36] or cartilage loss [35,37] may be mediated by changes in biomechanical forces through the subchondral bone.Therefore, biomechanical unloading is an attractive treatment option for the treatment of subchondral BMLs.High tibial osteotomy, which alters the load distribution between the medial and lateral compartments of the knee, reduces BML size in the femorotibial compartments [38].In patients with painful patellofemoral OA, patellofemoral bracing reduces the number of BMLs in the patellofemoral joint compartments [39].
This study has several limitations.While tissue samples with BMLs were obtained from the weight-bearing part of the MTP, the findings might differ for other joint regions such as the femoral condyles.All subjects included in this study had BMLs in the medial tibial plateaus and subjects without BMLs in the medial tibial plateaus were not evaluated.However, patients with knee OA who exhibit BMLs are known to experience greater pain compared to those without BMLs, therefore we focused on pain-related histological changes of BMLs.Osteoclast activity itself was not examined in this study; however, cells with !3 nuclei were counted as a single osteoclast to estimate the number of active osteoclasts, as resorption activity has been shown under some circumstances to correlate with the number of nuclei present [40].Despite these limitations, these limitations, our study clarifies how subchondral BMLs may cause joint pain.

Conclusions
Increased NGF expression and osteoclast density appear to have important associations with bone pain in knee OA with BMLs.This study will contribute to understanding the pathophysiology of subchondral BMLs in OA and identifying new therapeutic targets for the management of BML-related OA pain.

Fig. 1 .
Fig. 1.Co-localization of MRI Images and joint tissue obtained at surgery.Representative images of axial views of target knee for co-localization of knee biopsy tissue by MRI identification before tissue harvest at joint replacement surgery.Yellow boxes represent BML þ MTP bone, BML À MTP bone and BML À LTP bone that were analyzed.MTP: medial tibia plateau, LTP: lateral tibial plateau.

Fig. 2 .
Fig. 2. Histopathologic features in BML þ MTP bone, BML ¡ MTP bone and BML ¡ LTP bone Adjacent sections were stained with Safranin O and fast green, H&E, TRAP, and were processed for NGF.Sections from BML þ MTP bone show severe damage on cartilage surface with loss of proteoglycans, thickened trabeculae, cartilage islands (appears pink on Safranin O stain (A; asterisk), cyst-like lesions (C, D; double asterisk), vascular proliferation (d) bone marrow fibroses (C and e), cellular infiltration (C and f) and increased presence of osteoclasts (F, a and b).NGF immunoreactivity was detected in multinucleate TRAP positive osteoclasts adherent to bone (a and b), fibroblast-like cell in fibrotic connective tissue (H), mononuclear cells (b) and cartilage islands in bone marrow space (G).BML þ MTP bone exhibits an NGF-positive osteochondral channel (c; sharp) while BML À MTP bones exhibits an NGF-negative osteochondral channel (N).Cyst-like lesions frequently observed in BML þ MTP bone (C, D; double asterisk), and contained fibrous connective tissue (e), cartilage islands (A; asterisk) and TRAP positive osteoclasts which increased mainly on the surface of cysts (a).SafO: Safranin O and fast green, H&E: hematoxylin and eosin, SafO: Safranin O and fast green, TRAP: tartrate-resistant acid phosphatase, NGF: nerve growth factor, MTP: medial tibia plateau, LTP: lateral tibial plateau.Bars ¼ 200 μm

Fig. 4 .
Fig. 4. NGF expression in the subchondral bone marrow space and the osteochondral channel, and TRAP-positive osteoclasts in MTP bone with and without cysts.Each symbol represents an individual sample; bars show the mean and 95% CI.TRAP: tartrate-resistant acid phosphatase, NGF: nerve growth factor, MTP: medial tibia plateau, LTP: lateral tibial plateau.*P < 0.05 versus MTP bone without cyst.**P < 0.01 versus MTP bone without cyst.

Table 1
Subject demographics.