Background

Axial spondyloarthritis (axSpA) is a chronic inflammatory disease that is characterized by inflammation in the sacroiliac joints and spine, and over time some patients progress from non-radiographic axSpA to radiographic axSpA (ankylosing spondylitis (AS)) [1, 2]. There is an unmet need to identify biomarkers that reflect disease activity in patients with axSpA [3]. To monitor disease activity in axSpA patients, the AS Disease Activity Score (ASDAS) was developed and is now widely used [4]. It combines a questionnaire with items regarding back pain, peripheral pain, and duration of morning stiffness with blood C-reactive protein (CRP) levels. However, there are some drawbacks of ASDAS, as it mainly reflects patients’ perspectives, and CRP is highly weighted in the score [5]. Magnetic resonance imaging (MRI) has emerged as a reliable imaging modality in axSpA to detect bone marrow edema (BME) and structural lesions. However, whether patients are scanned by MRI is highly dependent on the availability of MRI devices [6], and repeated MRIs in routine care are often not feasible. There is a lack of commonly available and objective measures to monitor changes in disease activity in patients diagnosed with axSpA. We hypothesized that biomarkers of extracellular matrix (ECM) remodeling might assist in the evaluation of disease activity and response to treatment in axSpA.

During the disease progression of axSpA, inflamed tissue is infiltrated by immune cells, which creates a disturbance in tissue homeostasis, where ECM proteins are formed but also degraded by the overexpression and activation of proteases, such as metalloproteinases (MMPs) and cathepsins [7]. Both the formation and degradation of the ECM are reflected in changes in circulating biochemical markers of ECM fragments, so-called neo-epitopes [8]. The 28 types of collagen are the major proteins in the body and are expressed in various tissues, including joint tissues [9]. In particular, type I, III, IV, VI, VII, and X collagen are expressed in joint tissues, including tendons, bone, and connective tissue. Type I collagen is the most abundant protein, as it is the major structural protein in bone. PRO-C1, measuring type I collagen formation (N-terminal pro-peptide of type I procollagen [PINP]), has been shown to be a biomarker of bone formation [10]. Type I and III collagen are the major ECM proteins in soft tissue, while type IV collagen is the major constituent of the basement membrane located below the epithelial/endothelial cells. The biomarker PRO-C3, which measures type III collagen formation and is a marker of fibrogenesis, has previously been shown to be associated with progressive systemic sclerosis [11]. Turnover of type IV collagen, quantified by PRO-C4, and the two MMP-mediated fragments of type III and IV collagen, C3M and C4M, respectively, are associated with response to TNF-a inhibitor therapy in inflammatory bowel disease [12]. Type VI collagen formation, quantified by PRO-C6, is elevated in patients with rheumatoid arthritis and predicts response to an anti-IL-6 receptor antibody in combination with methotrexate (MTX) after 16 weeks [13]. The MMP-mediated degradation fragment of type VI collagen, C6M, has previously been shown to be upregulated in AS [14]. Type VII collagen is a minor collagen anchoring the fibrils that connect the basement membrane to the underlying interstitial matrix. Degradation of type VII collagen can be quantified by the MMP-mediated fragment of type VII collagen, C7M, which has been found to be upregulated in systemic sclerosis [15]. Type X collagen is primarily expressed by hypertrophic chondrocytes [16], and the biomarker C10C, reflecting cathepsin-degraded type X collagen, is elevated in patients with osteoarthritis [17]. Neutrophil elastase-degraded elastin (EL-NE) is a potential biomarker for differentiating inflammatory bowel disease from irritable bowel syndrome [18]. It has been suggested that the MMP-mediated metabolite of CRP (CRPM) reflects local tissue inflammation [19, 20], and it has been associated with changes in disease activity in patients with AS [19]. Finally, vimentin (VICM), an MMP-derived metabolite of citrullinated vimentin, has been associated with disease activity and radiographic progression in AS [21].

The aim of this study was to investigate a panel of ECM biomarkers to link ECM turnover with disease activity measures to identify candidate serum markers of tissue homeostasis and disturbance associated with disease activity. Such markers could potentially enable better monitoring of disease development in patients with axSpA. Therefore, we investigated the association of the ECM metabolite biomarkers PRO-C1, PRO-C3, PRO-C4, PRO-C6, C3M, C4M, C6M, C7M, C10C, CRPM, EL-NE, and VICM with disease activity in patients with axSpA, as measured clinically and by MRI, and their relationship with the treatment response to TNF inhibitory therapy for 46 weeks.

Methods

Subjects

The BIOSPA study has been described previously [22, 23]. Briefly, 60 TNFα inhibitor-naïve patients with axSpA initiating TNF-α inhibitor therapy were recruited from nine Danish rheumatology departments. Clinical and biochemical assessments were performed at baseline, and patients were followed for 46 weeks. Seven patients discontinued before week 22, and another six patients discontinued before week 46 due to side effects or lack of efficacy. Clinical response was defined as a reduction in the Bath AS Disease Activity Score (BASDAI) of 50% or 20 mm (on a VAS scale of 0–100 mm) at week 22 [24]. Since the study was performed before ASDAS was introduced, this was calculated post hoc. ASDAS responses were defined as follows: no clinically important improvement (ΔASDAS<1.1), clinically important improvement (1.1≤ΔASDAS<2.0), and major improvement in ASDAS (ΔASDAS≥2.0) [25]. The Spondyloarthritis Research Consortium of Canada (SPARCC) MRI Spine and Sacroiliac Joint Inflammation score, which quantitatively assesses bone marrow edema of the spine and sacroiliac joint, was evaluated by an experienced reader (UW )[22]. Serum was taken according to standard operating procedures and stored at −80°C prior to biomarker measurement [26].

Biomarker analysis

Biomarkers were assessed in 55 patients with available blood samples. Twelve different ECM metabolite biomarkers were assessed by competitive ELISAs developed by Nordic Bioscience. The assays measure type I collagen (PRO-C1), type III collagen (PRO-C3), type IV collagen (PRO-C4), and type VI collagen (PRO-C6) formation; MMP-generated type III collagen (C3M), type IV collagen (C4M), type VI collagen (C6M), and type VII collagen (C7M) degradation; along with Cathepsin K-generated type X collagen, MMP-mediated metabolite of C-reactive protein (CRPM), neutrophil elastase-degraded elastin (EL-NE), and citrullinated vimentin (VICM). All nine biomarkers were measured in serum or EDTA plasma at baseline (n=55), week 2, week 22, and week 46 after treatment with TNF-inhibitory therapy (infliximab, etanercept, or adalimumab). The assays were previously developed and technically validated [15, 17, 27,28,29,30,31,32,33,34]. The inter- and intra-assay coefficients of variation were <15% and <10%, respectively. Samples below the lower limit of quantification (LLOQ) were assigned the value of LLOQ.

Statistical analysis

Baseline characteristics were presented as numbers (frequency) and percentages for categorical variables and as medians (interquartile range) for continuous variables. Differences between the biomarker levels at the four timepoints were calculated using a repeated measures analysis of variance (ANOVA). Comparison of baseline variables was performed with the Mann–Whitney test, while correlations were performed with Spearman’s rank correlation coefficient. Before the analyses, a total SPARCC MRI Spine and Sacroiliac Joint Inflammation score was calculated as the sum of the SPARCC MRI Spine Inflammation score and the SPARCC MRI Sacroiliac Joint Inflammation score. Analysis was performed in MedCalc (version 14.8.1) and GraphPad Prism (version 8) software. A p-value below 0.05 was considered significant.

Results

Baseline demographics

Clinical and demographic characteristics at baseline are shown in Table 1. The patients in the study were mostly (82%) male. Furthermore, 84% of the patients were HLA-B27 positive, and the median age was 40 years (inter quartile range (IQR): 32–48). The patients had a median disease duration of 5.5 years (IQR: 2–10) and a median baseline ASDAS of 3.9 (IQR: 3.0–4.5). At baseline, weak correlations were observed between PRO-C1, PRO-C4, VICM, and ASDAS (ρ=−0.286, p=0.038; ρ=0.315, p=0.022; and ρ=0.327, p=0.017, respectively), while moderate correlations were observed for C6M versus ASDAS (ρ=0.538, p<0.001) and CRP versus ASDAS (ρ=0.654, p<0.0001). At baseline, no biomarkers were correlated with BASDAI or swollen joint count.

Table 1 Demographics and baseline clinical characteristics
Full size table

After 22 weeks of treatment, 10 patients were classified as showing clinically important improvement in ASDAS, while 23 patients were classified as showing major improvement in ASDAS. After 46 weeks of treatment, 13 and 17 patients, respectively were classified as having achieved a clinically important and a major improvement in ASDAS, respectively.

Changes in PRO-C1, C6M, EL-NE, VICM, and CRP during TNF-α inhibitory therapy

Compared to baseline, circulating levels of PRO-C1 significantly increased after 2 and 22 weeks of treatment (p<0.0001 and p=0.037, respectively; Fig. 1A), whereas C6M levels significantly decreased after 2 and 22 weeks of treatment (p=0.001 and p=0.002, respectively; Fig. 1G). Compared to baseline, EL-NE levels were significantly decreased after 2 weeks of treatment (p=0.0008; Fig. 1K). VICM levels decreased after 2 and 22 weeks of treatment compared to baseline (p=0.0163 and p=0.0374, respectively; Fig. 1L). CRP levels decreased after 2 and 22 weeks of treatment compared to baseline (both p=0.0001; Fig. 1M). In the other biomarkers, no significant changes were observed from baseline to weeks 2, 22, and 46. In comparison, there was a significant decrease in the ASDAS and BASDAI scores from baseline to 2, 22, and 46 weeks (both p<0.0001) as well as a decrease in SPARCC scores from baseline to 22 weeks (p=0.005) and 46 weeks (p=0.0005).

Fig. 1
figure 1

Time course of biomarkers in response to TNF-α inhibitor therapy. A Type I collagen formation, PRO-C1; B type III collagen formation, PRO-C3; C type IV collagen turnover, PRO-C4; D type VI collagen formation, PRO-C6; E type III collagen degradation, C3M; F type IV collagen degradation, C4M; G type VI collagen degradation, C6M; H type VII collagen degradation, C7M; I type X collagen degradation, C10C; J C-reactive metabolite, CRPM; K elastin degradation, EL-NE; L citrullination and degradation of vimentin, VICM; and M C-reactive protein, CRP. Time course is shown as median ± 95% CI. *p<0.05 **p<0.01, ***p<0.001, ****p<0.0001

Full size image

Association of PRO-C1, PRO-C3, C6M, VICM, and CRP with improvement in ASDAS

After 22 weeks, 17 individuals did not achieve a clinically important or major improvement in ASDAS (ΔASDAS <1.1), while 10 patients achieved a clinical important improvement in ASDAS (ΔASDAS ≥1.1) and 23 individuals achieved a major improvement (ΔASDAS ≥2.0). After 46 weeks of treatment, 11 patients did not achieve a clinically important or major improvement in ASDAS, while 13 and 17 patients achieved clinically important and major improvement in ASDAS, respectively. Levels of PRO-C1, PRO-C3, C6M, VICM and CRP were all associated with clinically and major improvement after 22 weeks (p=0.006, p=0.008, p<0.001, <0.001, and p<0.001, respectively; Table 2), while C6M, VICM, and CRP were the only three biomarkers associated with a change in ASDAS score after 46 weeks (p=0.002, p=0.044, and p<0.001, respectively; Table 2). There was no association between the other biomarkers and improvement in ASDAS scores at weeks 22 and 46.

Table 2 Changes in collagen biomarker levels and ASDAS improvement categories from baseline to weeks 22 and 46, respectively
Full size table

Association of PRO-C1 and C6M with BASDAI response

Compared to non-responders, responders had significantly higher levels of PRO-C1 after 22 weeks, while C6M levels were significantly lower at baseline for patients who were responding to TNF-a inhibitory therapy compared to non-responders (p=0.020 and p=0.049, respectively; Table 3). No association was found for responders after 46 weeks. There was no association between the other biomarkers and BASDAI response at weeks 22 and 46.

Table 3 Response to TNF-α treatment. Percent changes of biomarkers from baseline to weeks 22 and 46 in patients with and without a BASDAI response, respectively
Full size table

Association of PRO-C1, PRO-C4, C6M, and C7M with SPARCC MRI sacroiliac joint and spine inflammation scores

Baseline and week 22 levels of PRO-C4 and C6M were mildly to moderately correlated with the total SPARCC MRI Spine and Sacroiliac Joint Inflammation score (PRO-C4: ρ=0.279, p=0.043 and ρ=0.317, p=0.021, respectively, and C6M: ρ =0.496, p=0.0002 and ρ =0.297, p=0.031, respectively; Table 4). Week 22 and week 46 levels of PRO-C1 and C7M were mildly correlated with the total SPARCC Spine and Sacroiliac Joint Inflammation score (PRO-C1: ρ=− 0.318, p=0.021 and p=-0.285, p=0.05, respectively; C7M: ρ=0.379, p=0.005 and ρ=0.319, p=0.029, respectively; Table 4). No correlations were found between CRP and SPARCC at baseline (ρ=− 0.205, p=0.153), week 22 (ρ=0.263, p=0.057), or week 46 (ρ=− 0.127, p=0.430) or the other ECM biomarkers investigated in the study.

Table 4 Correlation between the biomarkers, CRP, and total SPARCC score. Spearman’s correlation was performed between extracellular matrix metabolites and MRI scores. Spearman’s rho (ρ) and p-values are shown
Full size table

Discussion

We evaluated changes in ECM metabolites (PRO-C1, PRO-C3, PRO-C4, PRO-C6, C3M, C4M, C6M, C7M, C10C, CRPM, EL-NE, and VICM) as potential biomarkers of disease activity and treatment response in patients with axSpA initiating treatment with TNF-α inhibitor and followed for 46 weeks. The main findings were as follows: (i) PRO-C1 was increased at weeks 2 and 22 after the initiation of TNF-inhibitory therapy, while C6M, EL-NE, and VICM were decreased at week 2, and both C6M and VICM remained decreased after week 22; (ii) PRO-C1, PRO-C3, C6M, and VICM were associated with a clinical improvement in the ASDAS score; (iii) PRO-C1 and C6M were associated with a BASDAI response to TNF-inhibitory therapy; and (iv) PRO-C4 and C6M were associated with total SPARCC MRI Spine and Sacroiliac Joint scores at baseline.

In the past decade, progress has been made on the development of biologics for treating axSpA. However, reliable and commonly available tools to guide treatment decisions are still needed. Candidate biomarkers reflecting disease activity, response to treatment, and structural progression may help close this gap [3, 35,36,37]. An essential property for an ECM metabolite biomarker in axSpA is a relationship with MRI inflammation and treatment response scores, thus potentially reflecting the impact of disease activity on the tissue level. In this study, we demonstrated a correlation between ECM metabolites and the total SPARCC MRI Spine and Sacroiliac Joint Inflammation score at baseline as well as associations with measures of clinical improvement and response to treatment, indicating that these metabolites may represent valuable biomarkers of disease activity in axSpA. In this study, C6M was reduced by TNF-inhibitory therapy, associated with clinically important and major improvement in ASDAS disease activity at weeks 22 and 46, and it moderately correlated with the total SPARCC MRI Spine and Sacroiliac Joint Inflammation score, and differentiated responders from non-responders after 22 weeks. Type VI collagen (COL6) is an ECM protein located in the interface between the basement membrane and interstitial matrix, where it binds to other ECM proteins and supports cell-to-cell interactions [38,39,40,41]. C6M is a metabolite of COL6 that is released from the inflamed tissue and can be measured in serum as a biomarker of connective tissue remodeling [27]. Previously, C6M has been associated with AS in a cross-sectional setting, where it was used to distinguish between healthy controls and AS patients, and it has been suggested as a candidate marker for monitoring disease activity in interventional studies [14, 42]. C6M separated AS patients from healthy individuals with an area under curve (AUC) of 0.78, but it was not correlated with radiographic progression in the spine (mSASSS). In another radiographic axSpA (r-axSpA) clinical study, C6M was shown to be highly correlated with CRP but not mSASSS [43]. However, the associations between C6M and the composite measure of disease activity for patients with axSpA (ASDAS) have not been studied previously. While chronic inflammation is a hallmark of axSpA, the nature of inflammation is not fully elucidated. Inflammation typically accelerates the turnover of collagens [44], and thus it was expected that C6M would be associated with MRI inflammation and CRP at baseline, which was demonstrated in this study. In addition, C6M is a marker of connective tissue breakdown, whereas CRP is a non-specific acute phase reactant, which are increased in many conditions. The potential of C6M as a biomarker of treatment response has also recently been shown in patients with psoriatic arthritis (PsA) [45].

Another interesting biomarker in this study is PRO-C1, which increased 2 and 22 weeks following TNF-inhibitory therapy and showed associations with disease activity and response to treatment. Type I collagen is the major collagen in the human body and the major constituent of bone. PRO-C1, also known as PINP, can be used to measure bone formation. The bone phenotype in axSpA is a mix of processes involving bone loss and bone formation. Bone loss (osteoporosis and bone erosion) is caused by increased osteoclast activity, although the link between TNF-inhibitory therapy and osteoclast activity has been debated; however, most findings indicate that TNF-inhibitory therapy leads to the inhibition of osteoblast differentiation [46,47,48]. An observational study of AS patients treated with TNF inhibitor reported that structural progression continued after the initiation of therapy, and retardation of structural damage was reported only after 4 years of treatment [49]. This may support the increase of PRO-C1 after 2 and 22 weeks. A third biomarker, VICM, was decreased after 2 and 22 weeks of TNF-inhibitor therapy and was also associated with clinically important and major improvements after 22 and 46 weeks. Vimentin is a type III intermediate filament protein that is expressed by various cells as a constituent of the cytoskeleton. It is secreted by macrophages and known to be citrullinated intercellularly by PAD enzymes [50]. VICM measures a citrullinated and MMP-mediated fragment of vimentin, and it has been reported to be prognostic for radiographic progression in AS [21], whereas VICM has not been shown to be modulated by TNF-inhibitory therapy [21].

Conclusion

In conclusion, the study demonstrates that extracellular matrix metabolites are associated with clinically important and major improvement in ASDAS, MRI inflammation, and response to TNF-inhibitory treatment in patients with axSpA. Selected biomarkers reflecting bone formation (PRO-C1), soft tissue degradation (C6M), and macrophage activity (VICM) are candidate soluble tissue turnover biomarkers of disease activity in axSpA.