Blood levels of matrix metalloproteinase generated C-reactive protein metabolite (CRPM) are a sensitive measure of chronic inammation in OA

The heterogeneous nature of osteoarthritis (OA) and the need to subtype patients is widely accepted. The biomarker CRPM, a metabolite of C-reactive protein (CRP), is released to the circulation during inammation. Blood CRPM levels have shown to be associated with disease activity and response to treatment in rheumatoid arthritis (RA). We hypothesized that circulating levels of CRPMs could be used to identify OA patients with an inammatory phenotype. We investigated the level and the prognostic effect of CRPM using combined data from two phase III OA and two RA studies (N = 1591). The association between serum CRPM levels and radiographic progression was investigated in knees of OA patients without radiographic OA in contra-lateral knee at baseline by dividing the patients into cases (knees with two-year radiographic progression) and controls (knees without radiographic progression).


Background
Osteoarthritis (OA) is a heterogeneous, painful and serious disease. The heterogeneity may be attributed to the existence of several phenotypes and underlining molecular subtypes also called endotypes. These phenotypes are not fully characterized, however it has been suggested that presence of systemic and/or local in ammation may be descriptive of one phenotypes [1,2]. The lack of understanding of OA phenotypes may be a major limitation for interventional trials, as patients with different phenotypes are expected to respond differently to a given drug. This limitation may be one of the main reasons for the lack of approved disease modifying drugs in OA (DMOADs). Anti-in ammatory treatment options (anti-TNF-alpha and anti-IL-1R1) [3][4][5] have been tested interventional trial including patient populations with de ned OA, however with no or limited description to their phenotype that may indicate differential response across different sub-populations. It has been suggested that a drug like lutikizumab (monoclonal antibody against IL1α/β) would work in patients with an in ammatory phenotype, while sprifermin (recombinant and truncated form of broblast growth factor 18) would work in patients with cartilage repair endotype [6][7][8]. There is a clear medical need to identify phenotypes in OA to enable development of more e cacious drugs.
Biochemical, blood-based and objective markers could be attractive tools for identifying sub-populations in OA; however, to date there are not such marker available for diagnostic use. Acute in ammation can be quanti ed in blood by the acute phase reactant C-reactive protein (CRP). CRP is highly elevated in rheumatic disorders such as rheumatoid arthritis (RA) [9], where is being used as part of the ACR-EULAR diagnosis criteria [10]. Blood CRP level have been tested in several studies of OA [11][12][13], however with differential results and with no clear conclusion to its applicability as a diagnostic marker in OA. One explanation may be the large biological variation associated with the role of CRP as an acute reactant.
CRP is mainly expressed and released from the liver in response to injury and infection, where from it is released in its pentameric form and binds to one of the Fc receptors on cells at site of injury or infection [14]. After exerting its effect, CRP is metabolized resulting in the release of the CRP metabolites. One such metabolite is the matrix metalloproteinase (MMP) degradation product CRPM [15]. Serum CRPM levels have been shown to be higher in subjects with radiographic knee OA [16], and to be dose-and timedependently inhibited in response to anti-TNF and anti-IL-6 treatment in RA [17,18]. In addition, in RA CRPM have been shown to be signi cantly correlated with disease activity measures such as DAS28, CDAI, SDAI, HAQ, ESR and CRP [18,19].
The primary hypothesis of the study, was that a notable proportion of OA patients have in ammation levels comparable to that of RA patients, as quanti ed by CRP or CRPM. The second hypothesis is that one of such in ammatory reactants would be associated with progression, and thus possibly re ect and in ammatory endotype with a distinctive clinical prognosis.

Results
The rst objective was to investigate and compare the level of CRPM and CRP in early RA (E-RA), moderate to severe RA (MS-RA) and osteoarthritis (OA). There were more female subjects in the moderate to severe (MS) RA cohort (p = 0.031) and signi cantly more Caucasian subjects in the OA cohort (p = 0.023) as compared to E-RA cohort (Table 1). Mean age was signi cantly higher in the OA cohort compared to E-RA (p < 0.0001) ( Table 1). DAS28 was higher in the MS-RA cohort compared to the E-RA cohort (p < 0.0001). Both the MS-RA and the OA cohorts had symptomatic and radiographic disease as observed by DAS28, WOMAC total and radiographic severity scores. There was no difference between the OA-ALL cohort and the OA-CC cohort (Table 1).  1C). The level of both CRP and CRPM were marginally higher in the OA patients as compared to the central lab established reference levels (Fig. 1B, B).
The second objective was to investigate the level of CRPM in knee OA patients and whether a level was associated with radiographic progression. Different cut-offs have been de ned for CRP, which are either used formally or informally for diagnoses and monitoring of RA. These cut-offs or references levels are listed in Table 2. In addition, we used the interquartile ranges of the E-RA cohort, the reference cohort, to provide exploratory cut-offs discriminating between low, moderate, high and very high CRP and CRPM levels. More than ninety percent of the MS-RA population had moderate to very high CRP and CRPM levels, con rming the expected that both CRP and CRPM is high in RA patients moderate to severe disease ( Table 2). Twenty-nine percent of the OA subjects had CRP levels higher lower than the reference level 3 mg/L and 15% had levels higher than the reference level 15% (Table 2). Using the IQR cut-offs then 51% of the OA patients had moderate to very high levels of CRP, whereas 31% of the OA patients had moderate to very high CRPM levels ( Table 2).
Independent of which cut-off was tested for CRP, it was not associated with two-year incidence of OA in the investigated knee (Table 3). Patients with moderate to very high CRPM (≥ 9 ng/mL) were 2.2-times more likely to develop OA in the knee with no OA at baseline (Table 3).

Discussion
We hypothesised that a sizeable number of OA patients in fact could have been diagnosed as RA patients if CRP was the primary diagnostic tool. We found that more than a third of the OA patients had CRPM and half had CRP levels corresponding to the levels observed in RA. Secondly, we hypothesized an in ammatory biomarker could be associated with an in ammatory endotype, associated with progression. Interesting only CRPM but not CRP was prognostic for progression of OA.
CRPM is a metabolite of C-reactive protein, which is released from the in amed tissue after CRP has exerted its effect on its target cells [20,21]. Both CRP and CRPM are reduced in response to antiin ammatory treatments such as anti-IL1 or anti-IL-6 receptor antibodies [22,23]. Presently, no subset analyses for identi cation pro les associated with progression in OA have been presented. In addition, the literature on CRP as both a biomarker for progression and diagnosis of OA is very contradicting with both positive and negative observations (ref ref). This is of interest in respect disorders like OA, as these disorders are not classically thought of as acute or systemic in ammatory diseases, however a subpart of patients may have an in ammatory endotype that could be associated with progression.
A decade ago, Kerkhof et al. [24]. and Attur et al. [25], showed that speci c genotypes linked with the IL-1β and IL-1 receptor antagonist were associated with radiographic severity of OA. They did not nd an association with radiographic progression. These data was further substantiated a year later by Attur et al. [26], that demonstrated that two distinct subgroups existed amongst OA patients where IL-1β expression was increased as compared to controls. This IL-1β endotype was associated with worse pain and function scores as well as radiographic progression. In the OAI-FNIH biomarker consortium, Kraus et al. found that blood markers of cartilage and bone remodelling as well as the in ammation-associated marker hyaluronic acid (HA) was associated with radiographic and pain progression [27,28]. Of note, as in our investigation, the OAI study was designed to include both progressors and non-progressors, where all subjects had OA at baseline. These data clearly point in the direction of the existence of an in ammatory endotype in OA, and such endotype could be de ned by speci c risk associated polymorphism and the level of one or more biochemical or transcriptomics markers that assess in ammatory activity.
Progression in OA in clinical trial settings is commonly de ned as changes to joint space width or increase in KLG over a time frame of one to three years. However, drivers of progression most likely have different origins amongst patients [1,29,30]. Thus, we know how the patients feel and function, but we don't know why and what drive a worsening of disease. This may be key in understanding why certain drugs fail in development: if only 30% of the trial population will bene t from targeting as speci c pathway, and with a plausible 50% response rates, the hurdle for documenting a clinical bene t is high, or impossible. Thus, there is a medical need for better subgrouping of patients by developing tools for endotyping. We propose that emphasis should be put on identifying sub-groups or subtypes of OA and to target drug development to each group followed by investigation of the e cacy [31]. We have recently published that CRPM together with other markers of tissue turnover may further subgroup patients into different endotypes, and that these endotypes display different radiographic progression types in RA [32]. Interestingly, we also included OA patients in the analysis that approximately 25% of the OA patients clustered with a subset of RA patients which were primarily in bone and cartilage markers, whereas about 10% of the patients clustered with RA patients that were high in bone, cartilage, macrophage and interstitial matrix turnover markers including CRPM. CRPM is most likely just one signal of many that could be investigated and combined to enable endotyping of patients and thereby better characterization of the patient pro les and planning of appropriate treatment regimen. CRPM have also recently been investigated in idiopathic pulmonary brosis [33] and spondyloarthropathy [34] and have been associated with progression. This may suggest that there is a general in ammatory endotype associated with progression of chronic tissue diseases, which may worthwhile investigating for the bene t of patients responding.
A limitation of the OA studies is that all patients have OA at baseline, in contrast to general demographic studies, thus the results cannot be extended to classi cation of CRPM as diagnostic biomarker (i.e. nding knee OA patients in a risk population). Instead, the analysis is limited to the identi cation of OA patients likely to develop bilateral and radiographic knee OA. Another limitation is the selection bias which is introduced in a well-de ned trial protocol. From a clinical perspective the population was homogeneous, where most subject had KLG 2 and moderate to severe pain; subjects with mild or very severe pain were excluded, and thus the variation in the dataset is narrow and not representative of all types of patients.

Conclusion
This study provides two major ndings: i) a subset of OA patients appears to have tissue in ammation comparable to that of RA; and ii) high CRPM levels are prognostic of incident knee OA. These data suggest that CRPM is a candidate biomarker of disease activity and for patient pro ling. The perspective is to target current bottlenecks in OA drug development and contribute to a strati ed medicine approach for more e cacious treatment of OA. There is an urgent medical need for patient strati cation in OA; rstly to facilitate better patient recruitment in clinical drug trials and secondly to facilitate better clinical decision-making concerning treatment of OA patients [31,35].

Patients
The early rheumatoid arthritis cohort (E-RA) Ninety-two early arthritis patients were enrolled in the prospective early arthritis 'Synoviomics' cohort at the Academic Medical Center (AMC) in Amsterdam between April 2004 and January 2013 in this study [36]. Of the 92 patients, 60 patients ful lled the ACR/EULAR 2010 criteria for RA classi cation [10] and were selected for current investigations. All patients enrolled in the study had less than one-year disease duration, as measured from the rst clinical evidence of joint swelling. Patients had active arthritis in at least one joint and were disease-modifying anti-rheumatic drug (DMARD) naïve. All patients provided written informed consent. The study was performed according to the Declaration of Helsinki and approved by the Medical Ethics Committee of the AMC. Only baseline measures were used in the current study, and demographic data were collected, and the following clinical and laboratory parameters were obtained: serum levels of C-reactive protein (CRP); erythrocyte sedimentation rate (ESR); 68 tender and 66 swollen joint counts (TJC68 and SJC66). Other parameters were recorded; however, this data was not used in current study. The prognostic data was published in 2016 by Maijer KI et al. [19].
The moderate to severe RA cohort (MS-RA) This is a post-hoc analysis of the available data from the LITHE sub-study, which have previously been thoroughly described by Blair et al. and Bay-Jensen et al. [23,32]. In brief, LITHE is a 2-year phase III, multicentre, randomized, three-arm, placebo-controlled, parallel group trial in patients with moderate to severely active RA who had an inadequate response to MTX [37]. Current analysis only apply baseline data from the biomarker sub-study of LITHE consisting of serum samples from a one year, doubleblinded treatment study where 704 patients were randomized 1:1:1 to one of 3 treatment groups: 4 mg/kg or 8 mg/kg TCZ, or placebo (PBO) in combination with a stable dosage of MTX (10-25 mg/week). The clinical study and the biomarker measurements were approved by the ethics committee at each participating institution and was conducted according to the Principles of Good Clinical Practice and according to the Declaration of Helsinki. Fasted serum for biomarker research was scheduled to be collected from patients who provided additional informed written consent. Baseline clinical and demographic measures were used in the current study including age, gender, race, DAS28 and erosion score.

The osteoarthritis cohort (OA-ALL)
This post-hoc study also include pooled data from two, double-blinded, randomised, placebo-controlled and multi-center phase III clinical trials assessing the e cacy and safety of an oral formulation of 0.8 mg salmon calcitonin in patients with painful and radiographic knee OA (NCT00486434 (trial 1) and NCT00704847 (trial 2)) [38]. The trials were conducted in accordance with the Helsinki Declaration and ICH GCP, and were approved by all applicable Independent Review Boards, Ethics Committees, and regulatory bodies. Each independent trial recruited patients aged 51-80 years with painful OA in the target knee, de ned as a Visual Analogue Score of ≥ 150 mm on the Western Ontario and McMasters Osteoarthritis Index (WOMAC) pain subscale (500 mm being the maximum score). In study 2, patients scoring ≤ 150 mm on the pain sub-score could participate if they also scored ≥ 510 mm on the WOMAC function sub-scale (1700 mm being the maximum score). The radiographic inclusion criteria for target knees included Kellgren-Lawrence grades (KLG) 2 or 3, and a Joint Space Width (JSW) of ≥ 2.0 mm. A total of 2,206 patients, corresponding to 4430 knees, were recruited at 19 sites in 11 countries. Patients were followed for two years with regular clinic visits. Details regarding trial design and results are published elsewhere [38]. Fasted serum samples were collected and assessed for biomarker at baseline, 3, 6, 12 and 24 months, however in current study only baseline biomarker data were included.
De nitions of OA radiographic progression case-controls (OA-CC) The case-control (CC) was designed with inspiration from the biomarker analyses conducted by Kraus et al. on data from the OAI-FNIH biomarker consortium [28]. The CC study only included knees with no marked (KLG < 2) radiographic knee OA at baseline. After excluding the sCT treatment arm and patients with no baseline serum samples, the biomarker sub-population consisted of 785 subjects (Fig. 1). Five subjects at baseline and 14 at year two had no radiographic scores (X-ray) recorded. 614 subjects had KLG equal to or greater than two at baseline and therefore had radiographic OA in the non-target knee. A total of 152 subjects were included in the case-control study. Cases were de ned as those subjects who progressed to KLG ≥ 2 in the non-target knee at year 2, while controls were those that did not progress in the non-target knee (KLG < 2) (Fig. 3). The case-control study included knees with not radiographic OA at baseline.

Biomarker measurements
Serum CRPM were measured by a competitive, solid-phase ELISA [39] according to the manufacture protocol. Assessments were performed by Nordic Bioscience Laboratory, a College of American Pathologists (CAP) certi ed division of Nordic Bioscience. Assays were run blinded and quality controlled according to internal standard operating procedure. CRP was measured on by ADVIA® chemistry CardioPhase High sensitivity C-reactive protein assay (Siemens Healthineers, NY, US) at the Nordic Bioscience Laboratory. Brie y, CRPM samples were run in duplicates and samples with CVs above 15% were rerun. Failure of internal plate controls lead to rerun of whole ELISA plate or run. Samples outside the quanti cation range were either diluted and rerun or given the value of the lower quanti cation limit. Batch control was completed using 10 serum samples and master calibrators covering the linear range of the calibration curve.

Statistics
Summary statistics of general demographics, baseline subject characteristics and biomarker levels was depicted by median and interquartile range (IQR). Ethics approval and consent to participate Current investigation applies pseudo-anonymized and retrospective data, collected as part of the original study designs. No new raw data was generated as course of described work. All analyses conducted was post-hoc analysis of existing raw data from the individual clinical studies. Experimental design was approved by the IRB at Nordic Bioscience. The original clinical study designs, where from the raw data for the post hoc analysis was acquired, were approved by local authorities and regulated, and all participant consented (written) for their biomarker data and de ned clinical to be used for research purposes.

Consent for publication
Not applicable.
Availability of data and materials The datasets generated and/or analyzed during the current study are not publicly available due to GDPR policies at the respective institutions but are available from the corresponding author on reasonable request. That is; anonymized biomarker data of this study may be available to access. Proposals should be directed to the corresponding author. Access will be provided to researches after the proposal has been reviewed by the IRB and after ethical approvals have been acquired. Samples from the patients are no longer available.  IQR and differences are tested by ANCOVA using log transformed biomarker data adjusting for covariates (age, gender, race, VAS pain). Bonferroni corrected signi cance levels: ** p < 0.01 and **** p < 0.0001.

Figure 2
Correlation between serum CRP and CRPM levels in the RA cohorts (A) and the OA cohort (B). Correlation were tested by simple linear regression. Data is shown as individual patients (grey dots and regression line (black line) with 95% con dence interval (black dotted line). Regression coe cient R2 and slope (b) are given.