Autoantibodies to Posttranslationally Modified Type II Collagen as Potential Biomarkers for Rheumatoid Arthritis

Authors

  • Rocky Strollo,

    1. Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, UK
    2. University Campus Bio-Medico of Rome, Rome, Italy
    Search for more papers by this author
    • Drs. Strollo and Ponchel contributed equally to this work.

  • Frederique Ponchel,

    1. Leeds Institute of Rheumatic and Musculoskeletal Medicine, University of Leeds, Chapel Allerton Hospital, and NIHR Leeds Musculoskeletal Biomedical Research Unit, Leeds Teaching Hospitals NHS Trust, Leeds, UK
    Search for more papers by this author
    • Drs. Strollo and Ponchel contributed equally to this work.

  • Vivianne Malmström,

    1. Karolinska University Hospital and Karolinska Institute, Solna, Stockholm, Sweden
    Search for more papers by this author
  • Paola Rizzo,

    1. Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, UK
    2. Catholic University, Rome, Italy
    Search for more papers by this author
  • Michele Bombardieri,

    1. Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, UK
    Search for more papers by this author
  • Claire Y. Wenham,

    1. Leeds Institute of Rheumatic and Musculoskeletal Medicine, University of Leeds, Chapel Allerton Hospital, and NIHR Leeds Musculoskeletal Biomedical Research Unit, Leeds Teaching Hospitals NHS Trust, Leeds, UK
    Search for more papers by this author
  • Rebecca Landy,

    1. Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, UK
    Search for more papers by this author
  • David Perret,

    1. Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, UK
    Search for more papers by this author
  • Fiona Watt,

    1. Kennedy Institute of Rheumatology, London, UK, and University of Oxford, Oxford, UK
    Search for more papers by this author
  • Valerie M. Corrigall,

    1. King's College London School of Medicine, London, UK
    Search for more papers by this author
  • Paul G. Winyard,

    1. University of Exeter Medical School, Exeter, UK
    Search for more papers by this author
  • Paolo Pozzilli,

    1. Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, UK
    2. University Campus Bio-Medico of Rome, Rome, Italy
    Search for more papers by this author
  • Philip G. Conaghan,

    1. Leeds Institute of Rheumatic and Musculoskeletal Medicine, University of Leeds, Chapel Allerton Hospital, and NIHR Leeds Musculoskeletal Biomedical Research Unit, Leeds Teaching Hospitals NHS Trust, Leeds, UK
    Search for more papers by this author
  • Gabriel S. Panayi,

    1. King's College London School of Medicine, London, UK
    Search for more papers by this author
  • Lars Klareskog,

    1. Karolinska University Hospital and Karolinska Institute, Solna, Stockholm, Sweden
    Search for more papers by this author
  • Paul Emery,

    1. Leeds Institute of Rheumatic and Musculoskeletal Medicine, University of Leeds, Chapel Allerton Hospital, and NIHR Leeds Musculoskeletal Biomedical Research Unit, Leeds Teaching Hospitals NHS Trust, Leeds, UK
    Search for more papers by this author
    • Dr. Emery has received consulting fees, speaking fees, and/or honoraria from Novartis, Pfizer, UCB, MSD, Abbott, Bristol-Myers Squibb, and Roche (less than $10,000 each).

  • Ahuva Nissim

    Corresponding author
    1. Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, UK
    • Barts and The London School of Medicine and Dentistry, Queen Mary University of London, London, UK
    Search for more papers by this author

Bone and Joint Research Unit, William Harvey Research Institute, Barts and The London School of Medicine and Dentistry, Queen Mary University of London, Charterhouse Square, London EC1M 6BQ, UK. E-mail: a.nissim@qmul.ac.uk

Abstract

Objective

Type II collagen (CII) posttranslationally modified by reactive oxygen species (ROS-CII) that are present in the inflamed joint is an autoantigen in rheumatoid arthritis (RA). The aim of this study was to investigate the potential use of anti–ROS-CII autoantibodies as a biomarker of RA.

Methods

CII was exposed to oxidants that are present in the rheumatoid joint. Autoreactivity to ROS-CII was assessed by enzyme-linked immunosorbent assays in synovial fluid (SF) and serum samples obtained from patients during various phases of RA. This group included disease-modifying antirheumatic drug (DMARD)–naive patients with early RA (n = 85 serum samples) and patients with established RA (n = 80 serum and 50 SF samples), who were categorized as either DMARD responders or DMARD nonresponders. Control subjects included anti–citrullinated protein antibody (ACPA)–positive patients with arthralgia (n = 58 serum samples), patients with osteoarthritis (OA; n = 49 serum and 52 SF samples), and healthy individuals (n = 51 serum samples).

Results

Reactivity to ROS-CII among DMARD-naive patients with early RA was significantly higher than that among patients with ACPA-positive arthralgia, patients with OA, and healthy control subjects (P < 0.0001), with 92.9% of serum samples from the patients with early RA binding to anti–ROS-II. There was no significant difference in anti–ROS-CII reactivity between ACPA-positive and ACPA-negative patients with RA, with 93.8% and 91.6% of serum samples, respectively, binding to ROS-CII. The sensitivity and specificity of binding to ROS-CII in patients with early RA were 92% and 98%, respectively. Among patients with established RA, serum reactivity in DMARD nonresponders was significantly higher than that in DMARD responders (P < 0.01); 58.3% of serum samples from nonresponders and 7.6% of serum samples from responders bound to HOCl-ROS, while the respective values for SF were 70% and 60%. In patients with longstanding RA, autoreactivity to ROS-CII changed longitudinally.

Conclusion

Autoantibodies to ROS-CII have the potential to become diagnostic biomarkers of RA.

Rheumatoid arthritis (RA) is the most common chronic autoimmune arthritis and affects 0.5–1% of the population. RA is characterized by inflammation of the joints and is associated with synovitis and erosion of cartilage and bone. The damage involves the action of proinflammatory cytokines ([1]), free radicals ([2]), and matrix metalloproteinases ([3]). The high influx of metabolically active immune cells into inflamed joints consumes increased amounts of oxygen, in association with respiratory burst and the generation of reactive oxidants. The key reactive oxygen species (ROS) present in inflamed joints are O2·−, H2O2, ·OH, HOCl, NO·, and ONOO, all of which are involved in acute and chronic inflammation ([4, 5]). In addition, cartilage damage as a result of collagen oxidation by glycation and formation of advanced glycation end products is evident despite the absence of hyperglycemia ([6]).

Type II collagen (CII) is the principal component of human articular cartilage and thus is a prominent target for posttranslational chemical modification by ROS in inflamed joints. Native CII is a well-studied autoantigen in RA ([7, 8]). Nevertheless, we previously reported autoimmune reactivity against ROS-modified CII (ROS-CII) ([9]). Distinct from posttranslational chemical modifications, the relevance of posttranslational enzymatic modifications in modulating the immune response in RA has been demonstrated. Anti–citrullinated protein antibodies (ACPAs) have become important diagnostic and prognostic tools in RA ([10]). Notably, citrullinated CII is also a component of ACPA reactivity in many patients with RA ([11-13]). The origin of citrullinated protein and its contribution to disease pathogenesis are, however, still incompletely understood. In addition, the diagnostic sensitivity of ACPAs is ∼60%, with some centers reporting ACPA positivity as low as 40% at the time of the diagnosis of RA ([14-16]). In any case, a significant percentage of patients with RA are ACPA negative. Therefore, the discovery of additional tissue-specific biomarkers that will improve diagnosis and prognosis or monitor response to therapy remains an important unmet clinical need ([17]).

In the present study, we explored the diagnostic potential of anti–ROS-CII in patients with RA by testing samples obtained at various stages of the disease continuum, from DMARD-naive early RA to established RA. We show that anti–ROS-CII autoreactivity is high in DMARD-naive patients with early RA regardless of ACPA status, and that it is significantly higher than that in ACPA-positive patients with arthralgia, patients with osteoarthritis (OA; either inflammatory or noninflammatory), and healthy individuals. We further demonstrate a correlation between disease severity and ROS-CII autoreactivity in patients with established RA.

PATIENTS AND METHODS

Patients and clinical samples

Serum samples were obtained from the Karolinska Institute, Leeds Division of Rheumatology and Musculoskeletal Diseases, Barts Hospital, Kennedy Institute of Rheumatology, and University of Pavia School of Medicine. RA was defined according to the 1987 American College of Rheumatology (ACR) criteria for the classification of RA ([18]), and the diagnosis was made by a specialist in rheumatology. Ethics approval was obtained from all of the involved clinical centers, and informed consent was obtained from all subjects prior to the collection of blood or synovial fluid (SF) samples. SF samples were obtained during knee arthroscopy or directly by knee joint aspiration.

The patient groups were categorized as follows:

  1. DMARD-naive patients with early RA (symptom duration <12 months). Blood samples (n = 85) were obtained at the first visit to a specialist clinic and before treatment with any DMARD. Forty-nine patients were ACPA positive, and 36 were ACPA negative.
  2. Patients with established RA (disease duration >1 year), categorized as DMARD responders or DMARD nonresponders. DMARD responders, defined as achieving low disease activity (Disease Activity Score in 28 joints [DAS28] [19] of <3.2) after DMARD monotherapy (methotrexate or sulfasalazine) or combined therapy with methotrexate plus sulfasalazine or prednisolone (n = 26 serum and 10 SF samples). DMARD nonresponders were defined by a DAS28 that remained ≥3.2 despite treatment (n = 24 serum and 10 SF samples). At the time of sampling, nonresponders had already failed escalation therapy with DMARDs (n = 16) or had already received treatment with anti–tumor necrosis factor (n = 7) or rituximab (n = 1).
  3. Patients with established, longstanding RA (n = 30) who were followed up longitudinally for up to 43 years. Fifteen patients were ACPA positive, and 15 were ACPA negative. Patients were categorized as having high disease activity (DAS28 >5.1), medium disease activity (DAS28 >3.2 and ≤5.1), or low disease activity (DAS28 <3.2). All except 2 patients had medium/high disease activity. Most patients had already been treated with biologic agents at the time of sampling. Matched SF and serum samples were obtained (2–11 samples per patient).

In the groups of patients with early and established RA, the DAS28 was determined in the clinic at the time of blood sampling.

The control groups included the following:

  1. ACPA-positive patients presenting with arthralgia but with no clinical evidence of synovitis. Patients were recruited from among those with new musculoskeletal pain symptoms (usually involving 1 or 2 joints) or arthralgia. When patients were classified as being ACPA positive according to an anti–cyclic citrullinated peptide 2 (anti–CCP-2) antibody test, they were seen by an experienced rheumatologist who established the absence of clinical evidence of synovitis. Normal C-reactive protein (CRP) levels (<10 mg/liter) were also confirmed. Eleven of 58 individuals had signs of OA but during 36 months of followup showed no sign of synovitis. This group provided 58 serum samples.
  2. Patients with knee OA as defined by the ACR criteria ([20]) (n = 49 serum and 52 SF samples). When detailed clinical information was available, patients with OA were classified by the presence or absence of inflammatory disease according to the symptoms and the presence or absence of clinical synovitis and/or joint effusion. Patients who had severe knee pain that occurred during physical activity and that disturbed sleep on a daily basis or who had persistent joint effusion despite treatment with intraarticular steroid injection and oral antiinflammatory agents were classified as having inflammatory disease. Patients with only intermittent mild knee pain and/or swelling that responded to quadriceps-strengthening exercises and/or treatment with acetaminophen were classified as having noninflammatory OA.
  3. Healthy control subjects (n = 51 serum samples). Samples were collected from a range of volunteers with no reported inflammatory joint disease; these subjects were matched for age and sex with the other participants. A few of the older individuals may have had OA, but none required medication to alleviate their symptoms. Results of anti–CCP-2 antibody testing were negative in all control subjects. Data for rheumatoid factor (RF) status were not available.

Enzyme-linked immunosorbent assay (ELISA).

CII was chemically modified as previously described to generate CII posttranslationally modified by HOCl and ribose ([9]) (Table 1). Bovine serum albumin (BSA; Sigma) and human serum albumin (HSA; Sigma) were similarly modified and used as control antigens. Data for CII modified by glycation/glycoxidation (GLY-II) and CII modified by HOCl (HOCl-CII) as examples of ROS-CII compared with native CII are shown in the Results section. ROS-CII encompasses GLY-CII and HOCl-CII.

Table 1. Chemical posttranslational modifications of type II collagen (CII)*
OxidantChemical reactionKey amino acids modifiedResulting ROS-CII modification
  1. The main chemical reactions resulting from the exposure of proteins to the strong oxidant HOCl are oxidation of methionine, cysteine, and lysine residues ([44]). A minor reaction is chlorination of aromatic amino acids, in particular tyrosine residues, generating modified forms of this amino acid, including 3-chlorotyrosine, within the polypeptide backbone ([44]). Exposure of CII to ribose results in the formation of advanced glycation end products and glycoxidation products within the target protein ([45]). These products, which include pentosidine, form through Maillard reactions between reactive carbonyl groups and basic amino acid side chains such as lysine and arginine ([4]). ROS-CII = reactive oxygen species–modified CII.
HOClOxidationLysine Methionine CysteineModified amino acids within the polypeptide backbone of CII, including lysine chloramines, methionine sulfoxide, cysteine sulfinic acid, and disulfides
   Fragmentation
   Crosslinking
   Aggregation
RiboseGlycation/glycoxidationLysine ArginineModified amino acids within the polypeptide backbone of CII, in the form of advanced glycation end products and glycoxidation products, e.g., N-carboxymethyllysine and pentosidine
   Fragmentation
   Crosslinking
   Aggregation

An ELISA was performed using ROS-CII and native CII as targets, as described previously ([9]). Briefly, ELISA plates were coated with 10 μg/ml of ROS-CII or native CII as bait to bind autoantibodies from SF or serum samples. The ELISA optical density (OD) values obtained for BSA, ROS-modified BSA, HSA, and ROS-modified HSA were used as background controls to normalize the respective OD values for native and ROS-CII. In addition, to control assay fluctuation, we performed an ELISA using the same batch of modified CII for any groups that were going to be compared. Each assay included a known reference control sample (positive or negative). Longitudinal samples obtained from the same individuals (serum and SF) were tested on the same day using the same batch of ROS-CII.

In the absence of absolute standards (as for anti–CCP-2 kits), titers could not be measured by an ELISA; therefore, arbitrary OD values were used. Patients positive for anti–ROS-CII autoantibodies (“binders”) were therefore defined using the mean plus 3SD of the ELISA OD values of healthy control subjects as a cutoff, set to 0.24 units. When paired serum and SF samples were tested, arbitrary ELISA OD units were normalized to their respective IgG levels. IgG levels were measured using a Human IgG ELISA Quantitation Set (Cambridge Bioscience) according to the manufacturer's instructions.

An ACPA ELISA was performed with an anti–CCP-2 test kit according to the manufacturer's instructions (Axis-Shield for the UK samples and Euro-Diagnostica for the Swedish samples). The cutoff was set according to each manufacturer's instructions (5 IU/ml and 25 IU/ml for Axis-Shield and Euro-Diagnostica, respectively).

Statistical analysis

Variables were not normally distributed; therefore, nonparametric tests were used. Wilcoxon's signed rank sum test was used to compare the reactivity between native CII and ROS-CII, while Mann-Whitney tests were used for comparisons between the various groups. Correlation was measured by Spearman's test, which is a nonparametric correlation test. To determine diagnostic discrimination between patients with early RA, patients with arthralgia, patients with OA, and healthy control subjects, we used the cutoff point of 0.24 OD units to construct a contingency table of positive autoantibodies against clinical diagnosis (early RA versus healthy controls; early RA versus arthralgia; early RA versus OA) and tested it using Fisher's exact test. A nonparametric Wilcoxon-type test was used to assess the longitudinal followup of anti–ROS-CII reactivity across time for each individual ([21]). A mixed model was also fitted to the longitudinal data to investigate whether time since diagnosis was a significant predictor of anti–ROS-CII reactivity. Data were analyzed using GraphPad Prism and Stata version 12 software.

RESULTS

Binding to ROS-CII in samples from patients with early RA versus healthy control subjects, patients with arthralgia, and patients with OA

In DMARD-naive patients with early RA, regardless of ACPA status, reactivity to ROS-CII was significantly higher than that in ACPA-positive patients with arthralgia, patients with OA, and healthy control subjects (P < 0.0001) (Figure 1), regardless of ACPA status. Among patients with early RA, 92.9% of serum samples bound to HOCl-CII, and 64.7% bound to GLY-CII (Table 2). In contrast, only 18.8% of sera from patients with early RA bound to native CII; this percentage was significantly lower than the percentage of sera that bound to ROS-CII (P < 0.0001). There was no significant difference (P > 0.05) in binding to ROS-CII between sera from ACPA-positive and ACPA-negative patients with early RA; 93.8% and 91.6%, respectively, bound to HOCl-CII (Figure 1A and Table 2). Reactivity to GLY-CII was, however, slightly higher in ACPA-positive sera compared with ACPA-negative sera, with 71.4% and 55.5%, respectively, being bound (P = 0.024).

Figure 1.

Binding of serum samples to reactive oxygen species–modified type II collagen (ROS-CII), and correlation with the Disease Activity Score in 28 joints (DAS28), as assessed by enzyme-linked immunosorbent assay (ELISA). A, Reactivity in serum samples obtained from patients with early rheumatoid arthritis (RA), patients with arthralgia, patients with osteoarthritis (OA), and healthy controls (HC). Early RA samples were obtained <12 months after diagnosis and before treatment with any disease-modifying antirheumatic drugs and were categorized as either anti–citrullinated protein antibody (ACPA) positive (n = 49) or ACPA negative (n = 36). Reactivity in early RA samples was significantly higher than that in arthralgia, OA, and HC samples (P < 0.0001). Regardless of ACPA status, reactivity to ROS-CII was significantly higher than that to native CII (NT-CII) (P < 0.001). B, Correlation between levels of anti–ROS-CII reactivity and the DAS28 in ACPA-positive patients (triangles) and ACPA-negative patients (squares) (ρ = 0.03 and ρ = −0.10, respectively; P > 0.58). GLY = glycation/glycoxidation-modified CII; HOCl = HOCl-modified CII.

Table 2. Distribution of binders to ROC-CII*
 No. of samplesAge, mean (range) yearsFemale sex, %Binders, %
NT-CIIGLY-CIIHOCl-CII
  1. Type II collagen (CII) posttranslationally modified by reactive oxygen species (ROS-CII) encompasses CII modified by glycation/glycoxidation (GLY-II) and CII modified by HOCl (HOCl-CII). RA = rheumatoid arthritis; NT-CII = native CII; ACPA = anti–citrullinated protein antibody; DMARD-R = disease-modifying antirheumatic drug responder; DMARD-NR = DMARD nonresponder; SF = synovial fluid; OA = osteoarthritis.
RA serum      
Early RA8551 (22−70)7218.864.792.9
ACPA−3653 (29−70)6619.455.591.6
ACPA+4948 (22−69)7818.371.493.8
DMARD-R2655 (23−84)767.67.67.6
DMARD-NR2456 (31−81)7537.558.358.3
RA SF      
DMARD-R1046 (21−75)8030.050.060.0
DMARD-NR1048 (48−71)10060.070.070.0
Disease control serum      
Arthralgia, ACPA+5853 (27−76)831.73.46.8
OA4961.5 (27−81)728.134.620.4
No synovitis3561 (27−81)670.030.013.8
Synovitis1465 (29−75)7821.450.028.6
Disease control SF      
OA      
No synovitis3660 (39−89)500.00.00.0
Synovitis1658 (43−78)504.125.016.6
Healthy control serum5148 (26−70)660.00.00.0

In ACPA-positive patients with arthralgia, binding to HOCl-CII was significantly higher than that in healthy control subjects (P < 0.001); 6.8% of serum samples from these patients bound to HOCl-CII, while only 3.4% bound to GLY-CII and 1.7% bound to native CII (Figure 1A and Table 2). Although reactivity in OA sera was significantly lower than that in early RA sera (P < 0.001), 34.6% of samples bound to GLY-CII, and only 20.4% bound to HOCl-CII (Table 2). There were no significant sex or age differences between healthy control subjects and patients with OA, patients with arthralgia, and DMARD-naive patients with early RA (Table 2).

The sensitivity and specificity of autoantibody binding to ROS-CII (both GLY-CII and HOCl-CII) in patients with early RA were 92% and 98%, respectively, compared with healthy controls (odds ratio [OR] 671, 95% confidence interval [95% CI 78.4–5,745], P < 0.0001) and 92% and 93%, respectively, compared with patients with arthralgia (OR 177.8, 95% CI 47.8–660.1, P < 0.0001). Compared with OA, the specificity and sensitivity for anti–CII-HOCl reactivity in patients with early RA were 75% and 92%, respectively (OR 40.6, 95% CI 14.1–116.6, P < 0.0001). However, reactivity to GLY-CII was less specific in patients with early RA compared with patients with OA, with sensitivity and specificity of 64% and 65%, respectively (OR 3.45, 95% CI 1.65–7.2, P = 0.0011).

In order to examine whether the presence of autoreactivity to modified collagen was a novel diagnostic biomarker, we investigated whether autoantibodies were associated with disease activity or levels of inflammation as reflected by the DAS28 (Figure 1B). There was no relationship between ELISA-derived OD values for anti–ROS-CII and the DAS28, regardless of ACPA status (ρ = 0.03 for ACPA positive and ρ = −0.10 for ACPA negative [P > 0.58]).

Anti–ROS-CII reactivity in patients with established RA

Despite the similarity in age and sex between DMARD responders and nonresponders (Table 2), there was a striking difference in the observed autoimmune reactivity toward ROS-CII in the tested serum samples (Figure 2). The strongest reactivity was seen in DMARD nonresponders, with 58.3% of samples binding to HOCl-CII; in comparison, only 7.6% of sera from DMARD responders bound to HOCl-CII (P < 0.01). Similarly, although SF samples from DMARD responders and nonresponders displayed the same patterns of reactivity, the levels of anti–ROS-CII reactivity in SF from DMARD nonresponders and responders were significantly higher than the levels in serum (P < 0.05 and P < 0.007, respectively) (Figure 2A), with 70% of samples from nonresponders and 60% of samples from responders binding to HOCl-CII. To confirm that the increased binding in SF was not an artifact related to the difference in immunoglobulin levels in SF versus serum, we tested a set of paired SF and serum samples in which binding to HOCl-CII was normalized according to the corresponding levels of IgG. Increased anti–ROS-CII reactivity in SF compared with serum was further confirmed in the paired SF and serum samples (P = 0.001) (Figure 2B).

Figure 2.

Anti–ROS-CII reactivity in patients with established RA, as assessed by ELISA. A, Anti–ROS-CII reactivity in serum and synovial fluid (SF) samples from patients with established RA who were categorized as either disease-modifying antirheumatic drug (DMARD) responders (R) or nonresponders (NR). Reactivity was significantly higher in both serum and SF samples from DMARD nonresponders (P < 0.01 versus responders). In both groups, anti–ROS-CII reactivity in SF was higher than that in serum (P < 0.05). B, Anti–ROS-CII reactivity in matched SF and serum samples. Levels of anti–ROS-CII reactivity were normalized to total levels of IgG. Reactivity in SF was significantly higher than that in serum (P = 0.001). C, Association between anti–ROS-CII reactivity and ACPA levels (ρ = 0.32 and ρ = −0.22 for DMARD nonresponders [open triangles] and responders [solid triangles], respectively; P > 0.134), C-reactive protein (CRP) levels (ρ = 0.08 and ρ = 0.12 for DMARD nonresponders and responders, respectively; P > 0.74), and rheumatoid factor (RF) status. Reactivity in RF-positive samples was similar to that in RF-negative samples (P > 0.05). See Figure 1 for other definitions.

No association between anti–ROS-CII reactivity and ACPA status was observed in either DMARD nonresponders or responders (ρ = 0.32 for nonresponders and ρ = −0.22 for responders [P > 0.134]). Similarly, no association between anti–ROS-CII reactivity and the CRP level was observed in either DMARD nonresponders or responders (ρ = 0.08 and ρ = 0.12, respectively) (Figure 2C). In addition, anti–ROS-CII reactivity in RF-positive patients was similar to that in RF-negative patients (Figure 2C).

Anti–ROS-CII binding in inflammatory versus noninflammatory OA

The presence of anti–ROS-CII antibodies was examined in more detail in OA with respect to the presence of clinical evidence of synovitis, severe pain, or persistent effusion (Table 2 and Figure 3A). Reactivity to HOCl-CII in serum samples from patients with severe inflammatory OA and patients with mild noninflammatory OA was low, with 28.6% and 13.8% of samples, respectively, binding to HOCl-CII. Reactivity to GLY-CII was higher, with 50% of samples from OA patients with synovitis and 30% of samples from OA patients without synovitis binding to GLY-CII (Figure 3A and Table 2). In addition, the pattern of binding to ROS-CII in OA was different from that in RA, with a tendency toward higher reactivity in serum than in SF, which is the opposite of the pattern observed in RA. In contrast to RA, matched OA SF and serum samples showed no tendency toward higher reactivity in SF, with both higher and lower reactivity observed in SF versus serum samples (P = 0.272) (Figure 3B).

Figure 3.

Anti–ROS-CII reactivity in patients with OA, as assessed by ELISA. A, Anti–ROS-CII reactivity in serum (S) and synovial fluid (SF) samples from patients with OA categorized as either inflammatory (INF) or noninflammatory (NON). Reactivity was higher in serum compared with SF and was significantly higher in inflammatory OA compared with noninflammatory OA, in both SF (P < 0.005) and serum (P = 0.05). Each data point represents a single sample. B, Anti–ROS-CII reactivity in matched SF and serum samples from patients with OA. Levels of anti–ROS-CII reactivity in matched serum and SF were normalized to the respective total levels of IgG. Reactivity in SF was not significantly different from that in serum (P = 0.272). See Figure 1 for other definitions.

Longitudinal study of ROS reactivity in patients with chronic disease

To further study the correlation between anti–ROS-CII reactivity and disease evolution, we analyzed matched SF and serum samples from 30 patients with chronic RA that were collected longitudinally. As shown in Figures 4A and B, anti–HOCl-CII reactivity varied considerably; similar variability was also observed for reactivity with GLY-CII (data not shown). A trend toward higher reactivity was observed in SF from ACPA-positive patients compared with SF from ACPA-negative patients. Anti–ROS-CII reactivity in sera from both groups was, however, similar (Figure 4B). Nevertheless, no correlation between the DAS28 and ROS-CII autoreactivity at the time of sampling was observed in either ACPA-negative patients (ρ = 0.3006, P = 0.2173 and ρ = 0.31718, P = 0.1736 for serum and SF, respectively) or ACPA-positive patients (ρ = −0.04581, P = 0.8134 and ρ = 0.0696, P = 0.5346 for serum and SF, respectively) (Figure 4C). Corresponding DAS28 values were not available for all of the tested samples.

Figure 4.

Longitudinal followup of anti–ROS-CII autoreactivity in patients with longstanding RA, as assessed by ELISA. A, Longitudinal changes in anti–ROS-CII reactivity in synovial fluid (SF) (squares) and serum (circles) from 4 different patients with chronic RA who were categorized as having high or medium/low disease activity. Reactivity in SF was higher than that in serum in the ACPA-positive samples but was similar to or lower than that in serum in the ACPA-negative samples. B, ELISA OD values for matched SF (F) and serum (S) samples from each patient, shown as pairs. There was a trend toward higher OD values in SF from ACPA-positive patients compared with ACPA-negative patients. C, Association between the DAS28 and anti–ROS-CII reactivity in serum and SF from both ACPA-negative patients (ρ = 0.31718, P = 0.1736 and ρ = 0.3006, P = 0.2173 for serum and SF, respectively) and ACPA-positive patients (ρ = −0.04581, P = 0.8134 and ρ = 0.0696, P = 0.5346 for serum and SF, respectively). See Figure 1 for other definitions.

A nonparametric test for trend showed no evidence of a trend in anti–ROS-CII reactivity over time (P = 0.634). Similarly, when a mixed model was fitted to the data, time since diagnosis was not a significant predictor of anti–ROS-CII reactivity in either the unadjusted model (P = 0.834) or the model adjusted for ACPA status and DAS28 category (P = 0.631).

DISCUSSION

A common biomarker for the diagnosis of RA is the presence of ACPAs ([10]) and, to a lesser extent, RF. Elevated cytokine/chemokine and CRP levels are also predictive of the development of RA ([22]). Only ∼60–70% of patients with RA are ACPA positive ([15, 23]), with some centers reporting that the sensitivity of an ACPA test in individuals with the recent onset of symptoms of RA is as low as ∼40% ([24, 25]). Furthermore, ACPA levels do not change significantly during disease progression, even after treatment, although a limited reduction (∼20%) in ACPA levels has been observed after rituximab therapy ([26-28]). Therefore, the use of additional tissue- and disease-specific biomarkers of RA may facilitate improved diagnosis and more accurate monitoring of treatment efficacy.

CII is the predominant cartilage collagen and is a known autoantigen ([7, 8]). Thus, testing for antibodies to CII could potentially be the most relevant approach to the diagnosis of RA. Unfortunately, anti-native CII antibodies are present in only 3–27% of patients with RA ([29-32]). Recent studies demonstrated that antibodies to citrullinated CII are common in RA ([11-13]). Our group previously developed an approach that tests autoimmune reactivity toward CII neoepitopes that result from some of the pathogenic processes in the inflamed joint, namely, CII posttranslationally modified by reactive oxidants (ROS-CII). RA serum samples showed increased binding to ROS-CII, but no anti–ROS-CII reactivity was detected in sera from patients with other inflammatory arthritis conditions such as psoriatic arthritis, systemic lupus erythematosus, ankylosing spondylitis, palindromic arthritis, scleroderma, Behçet's disease, primary Sjögren's syndrome, fibromyalgia, inflammatory arthritis, tendinitis, and reactive arthritis ([9]).

In the current study, we analyzed anti–ROS-CII autoreactivity in patients with RA at various stages of the disease and studied the potential of anti–ROS-CII as a diagnostic/prognostic biomarker of RA. Only 18% of sera from patients with early RA (before DMARD treatment) bound to native CII. In contrast, a high proportion of DMARD-naive patients with early RA showed autoreactivity to ROS-CII (92.9% were binders), suggesting that a routine test would have high detection power. Importantly, in the ACPA-negative group, the proportion of binders to HOCl-CII was similar to that in the ACPA-positive group (91.6% and 93.8%, respectively), suggesting the potential use of anti–ROS-CII as an ACPA-independent serum biomarker.

To assess the specificity of anti–ROS-CII reactivity in RA, we evaluated not only healthy control subjects but also 2 disease control groups: ACPA-positive patients with arthralgia who had no clinical evidence of synovitis and patients with OA with or without knee inflammation. Reactivity to ROS-CII in DMARD-naive patients with early RA was significantly higher than that in ACPA-positive patients with arthralgia, patients with OA, and healthy control subjects. Therefore, anti–HOCl-CII reactivity is highly specific and sensitive, with specificity and sensitivity of >75%. For GLY-CII, sensitivity and specificity were reduced to 64% and 65%, respectively, due to background anti–GLY-CII reactivity in OA, thus suggesting the potential of developing an anti–HOCl-CII test for further validation as a diagnostic marker of RA.

Only low levels of anti–ROS-CII reactivity were detected in 6.8% of ACPA-positive patients with arthralgia, suggesting that the appearance of anti–ROS-CII may occur closer to the time of onset of clinical synovitis, possibly with different dynamics compared with ACPAs, which in many cases appear years before the onset of clinical RA ([33, 34]). In the absence of full information regarding whether and when the onset of clinical symptoms may occur in all of these subjects (this cohort is still being followed up), it is difficult to estimate precisely the predictive value of the low ROS-CII autoreactivity observed in ACPA-positive patients with arthralgia. So far, 2 of the 4 anti–ROS-CII–positive patients progressed to RA within 1 month of followup, and RA developed in a third ACPA-positive patient 3 months later. In contrast, 18 anti–ROS-CII–negative individuals experienced progression to RA over longer periods of time (median 7 months). Thus, our data showed that anti–ROS-CII reactivity is independent of ACPA status and suggest the need to perform a more comprehensive study to establish when anti–ROS-CII reactivity occurs and whether it provides additional insight into the development of clinical RA.

Although OA is not a systemic inflammatory disease, synovial inflammation is highly prevalent, although it is not as severe as that in RA ([35]). Chondrocytes, synoviocytes, and infiltrating immune cells produce inflammatory mediators in OA joints similar to those present in inflamed RA joints. OA chondrocytes are metabolically active and produce high levels of ROS ([35, 36]). In fact, in both OA and RA, cartilage damage as a result of collagen oxidation by glycation and formation of advanced glycation end products is evident ([37]). Therefore, although OA is not considered an autoimmune inflammatory disease, it was interesting to study whether posttranslational modification of CII and thus formation of neoepitopes could stimulate an immune response in OA. Although the percentage of OA samples binding to ROS-CII was significantly lower than the percentage of RA samples binding to ROS-CII, reactivity to ROS-CII in serum samples from patients with inflammatory OA was higher compared with that of samples from patients with noninflammatory OA. The elevated reactivity to GLY-CII is interesting and might represent a screening test for the presence of arthropathy as opposed to the presence of chronic pain/arthralgia; this will be addressed in a future study. ACPA positivity was also observed in some of the patients with OA, more frequently in SF (21%) than in serum (6%). In fact, the presence of ACPAs in OA and other joint diseases ([38]) is slowly being acknowledged.

Based on findings in our cohort of patients with established RA who were receiving DMARDs (despite the small numbers), it appeared that most of those without reactivity to ROS-CII had a better outcome to DMARD treatment. The strongest reactivity toward ROS-CII was in patients who were DMARD nonresponders, independently of ACPA, CRP, and RF status. In addition, 34% of patients who were both ACPA negative and RF negative were anti–ROS-CII positive.

Interestingly, anti–ROS-CII reactivity in SF was higher than that in serum, although the patterns were similar. The fact that higher levels of anti–ROS-II were present in SF may be explained by the local presence of the target CII neoepitopes in the inflamed joint and thus local production and/or local retention of these antibodies ([39]). Due to high infiltration of immune cells in the inflamed RA joint (including B cells), as well as the formation of ectopic lymphoid structures in RA ([40]), local production of anti–ROS-CII in RA joints is possible. In addition, the presence of cartilage debris, possibly modified by the high local levels of ROS ([35, 41]), may trigger localization of anti–ROS-CII antibodies in the inflamed joint. Interestingly, reactivity in OA SF was similar/lower than that in OA serum. The lower reactivity in OA SF may reflect the lack of local ectopic lymphoid formation and local anti–ROS-CII formation. It is probable that systemic autoimmunity to ROS-CII is stimulated by a breach in tolerance as a result of posttranslational modification and formation of neoantigens. This process spreads to the inflamed OA joint but is intensified only in RA joints due to the formation of ectopic germinal centers.

To investigate the changes in anti–ROS-CII autoreactivity, samples collected longitudinally over 1–43 years after the onset of RA were analyzed, showing that anti–ROS-CII reactivity changed over time with at least 1 peak. This is in contrast to anti–native CII antibodies that occur in some patients with RA at the time of diagnosis, after which the levels tend to decrease during the first few years of disease ([42]). In our study, only 2 patients displayed low anti–native CII reactivity at the late stage of disease. In this group, we again observed anti–ROS-CII reactivity in both ACPA-positive and ACPA-negative patients. There was no correlation between serum levels of ROS-CII autoreactivity and the DAS28. Nevertheless, only 2 patients had low disease activity, and only 5 patients with medium/high disease activity had a DAS28 of <3.2 at some time point. Therefore, we were unable to carry out a statistical analysis comparing patients with low disease activity (DAS28 <3.2) with those with high/medium disease activity (DAS28 ≥3.2) in this cohort. Furthermore, higher levels of anti–ROS-CII in SF were observed in ACPA-positive patients with high disease activity, suggesting that ROS-CII autoantibodies in SF may more closely reflect tissue/cartilage damage.

Our data suggest a lack of direct correlation between anti–ROS-CII reactivity and the DAS28 in patients with active disease. A potential interpretation of this observation is that once disease was active (DAS28 ≥3.2), the redox state was unbalanced (i.e., oxidative stress), such that a burst in oxidative reactions resulted in the formation of ROS-CII neoepitopes. In turn, this perhaps led to an autoimmune response regardless of further fluctuations in the DAS28. In patients who experienced a response to DMARDs, i.e., achieving a DAS28 of <3.2, the redox state was rebalanced, with a consequent reduction in the production of ROS-CII neoepitopes and thus a decrease in anti–ROS-CII autoimmunity. Therefore, anti–ROS-CII reactivity does not simply reflect the degree of inflammation, as suggested by the fact that anti–ROS-CII reactivity was not directly correlated with either the DAS28 or the CRP level at any stage of disease (early, established, or longstanding RA).

The lack of association between anti–ROS-CII and inflammation is supported by our previous study ([9]), in which no anti–ROS-CII reactivity was detected in other inflammatory conditions. In addition, in a recent parallel study ([43]), we detected anti–ROS-CII reactivity in autoimmune patients with type 1 diabetes mellitus who were HLA–DRB1*4 positive and carried the shared epitope alleles associated with susceptibility to both type 1 diabetes mellitus and RA. Interestingly, anti–ROS-CII reactivity in HLA–DRB1*4–negative patients with type 1 diabetes mellitus and patients with type 2 diabetes mellitus was low/absent. Our findings will need to be addressed in greater depth in future studies using much larger cross-sectional and longitudinal cohorts versus large numbers of disease control patients with inflammatory arthritis distinct from RA, to exclude the possibility that the current study lacked the power to detect a meaningful biologic association between anti–ROS-CII and inflammatory markers.

In conclusion, our results suggest that anti–ROS-CII autoantibodies may provide a novel serologic biomarker that can facilitate the diagnosis of RA, particularly in ACPA-negative patients; lead to better subgrouping of patients with RA; and provide an additional criterion with which to define remission. Identification of the exact ROS-CII neoepitope(s), the mechanism that results in anti–ROS-CII reactivity, and validation of its potential value as a novel RA-specific biomarker will require further studies. ROS-CII appears to represent a promising new member of the family of posttranslationally modified proteins that are able to elicit autoimmune responses.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Nissim had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Ponchel, Bombardieri, Perret, Corrigall, Winyard, Conaghan, Panayi, Nissim.

Acquisition of data. Strollo, Ponchel, Rizzo, Wenham, Watt, Conaghan, Klareskog, Emery, Nissim.

Analysis and interpretation of data. Ponchel, Malmström, Rizzo, Landy, Perret, Watt, Winyard, Pozzilli, Conaghan, Panayi, Klareskog, Emery, Nissim.

Acknowledgments

We would like to thank Arthritis Research UK for supporting 2 intercalated BSc studentships for Harsha Mistry and Mariam Adeola Balogun, both of whom contributed to this study. We thank members of Professor Pitzalis' group, including Drs. Frances Humby, Francesco Carubbi, Stephen Kelly, and Becki Hands (William Harvey Research Institute), Dr. Antonio Manzo (Pavia School of Medicine), Sarah R. Kingsbury (Leeds Institute of Molecular Medicine), Lena Israelsson (Karolinska Institute), and Professor Alison McGregor (Imperial College London) for providing study patients. We extend special thanks to Dr. Amar Ahmad and Professor Peter Sasieni for their help with the statistical analysis of our data.

Ancillary