Article Text

Extended report
MRP8/14 serum levels as a strong predictor of response to biological treatments in patients with rheumatoid arthritis
  1. Ivy Y Choi1,
  2. Danielle M Gerlag1,
  3. Marieke J Herenius1,
  4. Rogier M Thurlings1,
  5. Carla A Wijbrandts1,
  6. Dirk Foell2,
  7. Thomas Vogl3,
  8. Johannes Roth3,
  9. Paul P Tak1,4,5,
  10. Dirk Holzinger2,3
  1. 1Division of Clinical Immunology and Rheumatology, Academic Medical Center/University of Amsterdam, Amsterdam, The Netherlands
  2. 2Department of Paediatric Rheumatology and Immunology, University Children's Hospital Muenster, Muenster, Germany
  3. 3Institute of Immunology, University Hospital Muenster, Muenster, Germany
  4. 4University of Cambridge, Cambridge, UK
  5. 5GlaxoSmithKline, Stevenage, UK
  1. Correspondence to Professor Paul P Tak, Department of Clinical Immunology and Rheumatology, Room F4-105, Academic Medical Center, University of Amsterdam, P.O. box 22700, Amsterdam 1100 DE, The Netherlands; p.p.tak{at}


Background One-third of rheumatoid arthritis (RA) patients treated with biological therapy show lack of response. The use of predictive biomarkers to identify responders to treatment may provide guidance in optimising treatment strategies and reduce unnecessary side effects and costs.

Objective To test the ability of myeloid-related proteins (MRP)8/14 protein complexes, an endogenous TLR-4 receptor agonist, to predict and monitor response to biologics in RA patients.

Methods 170 RA patients treated with adalimumab (n=86), infliximab (n=60) or rituximab (n=24) were categorised into clinical responders (n=123) and non-responders (n=47). MRP8/14 serum complexes were measured at baseline, and 4 and 16 weeks after initiation of treatment and related to response outcome.

Results Before initiation of treatment, responders showed significantly higher MRP8/14 protein complex levels compared with non-responders in each prospective cohort (p=0.010, p=0.001 and p<0.001, respectively). Logistic regression analysis showed that having high MRP8/14 baseline levels increased the odds of being a responder by 3.3 up to 55. In responders to adalimumab or infliximab treatment, MRP8/14 levels decreased after 4 weeks of treatment by 46% and 60% and after 16 weeks by 61% and 68%, respectively. In contrast, MRP8/14 levels were stable in non-responders. In patients treated with rituximab, MRP8/14 levels decreased by 59% after 16 weeks in responders and increased by 89% after 16 weeks in non-responders.

Conclusion Serum concentrations of MRP8/14 protein complex are a promising biomarker to predict response to biological therapy in active RA patients at baseline and could be used to monitor response to treatment across different mechanisms of action.

  • Anti-TNF
  • Rheumatoid Arthritis
  • Disease Activity

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Rheumatoid arthritis (RA) is a chronic inflammatory disease affecting synovial tissue in multiple joints leading to joint destruction, deformity and disability.1 Biological therapies have dramatically improved the treatment and outcome of RA patients who experience persistent disease activity despite conventional disease-modifying antirheumatic drugs (DMARDs).2–5 However, a significant subset of patients does not respond to treatment with biological agents such as tumour necrosis factor (TNF) inhibitors, tocilizumab, abatacept or rituximab.6–10 The use of biomarkers predictive of response to biological treatments may provide guidance in determining optimal treatment in the context of stratified or precision medicine and may help to reduce unnecessary side effects and costs.11

We have previously shown that clinical improvement is consistently correlated with the change in synovial macrophages after initiation of antirheumatic treatment, independent of the specific mechanism of action.12–14 From a discovery perspective, these studies highlight the potential use of biomarkers related to monocyte/macrophage infiltration into the synovial compartment to predict and monitor the response to targeted treatment, but this approach may not be practical in clinical practice because it requires synovial biopsy. Therefore, we investigated the use of a serum biomarker reflecting the role of monocytes in RA.

Myeloid-related proteins (MRP) 8 (S100A8) and 14 (S100A9) are endogenous TLR-4 ligands that are expressed in granulocytes and monocytes.15 ,16 The extracellular complex of MRP8 and MRP14 (MRP8/14) is also known as calprotectin. They cause strong proinflammatory effects on phagocytes and endothelial cells in vitro and promote inflammatory processes in vivo.17–23 An important contribution to joint inflammation was shown in experimental antigen-induced arthritis, where MRP8 and MRP14 significantly contribute to joint inflammation and leucocyte infiltration.21 Moreover, MRP8/14 are released at the local site of inflammation and serum levels have been proposed as biomarkers to monitor disease activity in many inflammatory diseases and are able to predict the risk of a relapse in juvenile idiopathic arthritis by detecting subclinical inflammation.24–26

Here we studied the relationship between serum levels of MRP8/14 at baseline and the clinical response to treatment with anti-TNF antibodies or rituximab in a prospective, open design. In addition, we examined whether measuring serum levels of MRP8/14 could be used to monitor the response to targeted treatment.


Study population

Serum samples from RA patients participating in three previously described prospective cohort studies were used.27–29 Patients fulfilled the 1987 American College of Rheumatology (ACR) classification criteria for RA and had active disease despite the use of a stable dose of methotrexate (5–30 mg/week).30 Active disease was defined by having a disease activity score (DAS28) ≥3.2 in patients who started treatment with adalimumab (ADA group) or infliximab (IFX group). In patients who started rituximab treatment (RTX group), active disease was defined by swollen and tender joints out of 28 joints ≥4 and at least one of the following erythrocyte sedimentation rate (ESR) >28 mm/h, serum C-reactive protein level (CRP) >15 mg/L or morning stiffness >45 min. Stable prednisone (≤10 mg/day) and use of non-steroidal anti-inflammatory drugs were allowed in all groups. The studies were performed in the Academic Medical Centre/University of Amsterdam and approved by the local medical ethics committee. All patients gave written informed consent.

The ADA group (included between 2004 and 2010) was treated with adalimumab 40 mg subcutaneously every other week.29 The IFX group (included between 2002 and 2004) was treated with infliximab 3 mg/kg intravenously at baseline, week 2, week 6 and subsequently every 8 weeks.28 At week 16, response was determined using the European League Against Rheumatism (EULAR) response criteria in the ADA and IFX groups.31 The RTX group (included between 2005 and 2006) was treated with two intravenous infusions of 1000 mg rituximab at day 1 and day 15 after premedication with 2 mg clemastine fumarate and 1000 mg acetaminophen.27 Of note, these patients did not receive pretreatment with methylprednisolone as this would have influenced the mechanistic studies performed in this clinical trial.27 ,32 Response to RTX treatment was determined at week 24 according to the EULAR response criteria.

Determination of MRP8/14 serum complex levels

Serum levels of MRP8/14 complexes were determined by double-sandwich ELISA system, performed in the laboratory of the University of Münster, as described previously.25 ,33 For comparison with earlier studies, internal control sera were used as a reference in all ELISA studies. The readers of the laboratory assay were blinded for the diagnosis and the inflammatory activity of the patients. Data are expressed in ng/mL.

Statistical analysis

Continuous data were described as mean and SD if normally distributed, and as median and IQR if not normally distributed. The χ2 test was used to compare categorical characteristics between responders and non-responders and between three treatment groups. Mann–Whitney U test or Kruskal–Wallis test was used to compare continuous variables. Correlations were assessed using the Spearman's rank correlation coefficient. The Wilcoxon signed ranked test was used for analysing the differences in MRP8/14 levels at two paired time points. To examine the relationship between MRP8/14 baseline levels and clinical response to biological treatment, we performed binary logistic regression. Stepwise backward multivariable logistic regression analyses were performed to obtain estimates of the ORs. A receiver operating characteristic (ROC) curve was generated by plotting the sensitivity against one-specificity, and the area under the curve (AUC) with 95% CI was calculated. All statistical analyses were performed with SPSS V.19.0 software (IBM Corp., Armonk, New York, USA). A p value <0.05 was considered statistically significant.


Baseline characteristics

In total, 170 patients were included in the three different treatment groups. The baseline characteristics are summarised in table 1. Age and gender were representative of typical RA cohorts. IgM rheumatoid factor in kUA/L (IgM-RF) positivity was lower in the ADA treatment group compared with the IFX and RTX treatment groups. Baseline DAS28, ESR levels and percentage of glucocorticoid use were higher in patients treated with RTX than those treated with ADA or IFX. In the RTX group, baseline CRP level was significantly higher compared with the ADA group.

Table 1

Baseline characteristics of patients enrolled in the ADA, IFX and RTX groups

MRP8/14 serum levels are correlated with clinical parameters at baseline

There was a statistically significant correlation between MRP8/14 serum levels and clinical parameters at baseline. Baseline MRP8/14 serum levels were statistically significantly correlated with the swollen joint count of 68 joints (SJC68 at baseline (Spearman's rho 0.227, p=0.003), baseline DAS28 (Spearman's rho 0.195, p=0.012), baseline ESR levels (Spearman's rho 0.336, p<0.001) and baseline CRP levels (Spearman's rho 0.509, p<0.001).

Clinical response to biological treatment

In the ADA group (n=86), 36 patients were good, 29 moderate and 21 non-responders. In the IFX group (n=60), there were 25 good responders, 20 moderate responders and 15 non-responders. In the RTX group (n=24), 4 patients achieved good response, 9 patients moderate response and 11 patients non-response.

When comparing baseline patient characteristics in responders and non-responders, the mean DAS28 in the ADA group was significantly higher in responders (5.49±0.97) compared with non-responders (4.97±0.92; p=0.035; see also online supplementary table S1). Baseline characteristics were not different between responders and non-responders to IFX and RTX treatment (see online supplementary tables S2 and S3).

Baseline MRP8/14 serum levels are higher in responders to biological treatment

Baseline MRP8/14 serum levels were higher in responders (good and moderate responders together) compared with non-responders to ADA (figure 1, 1100 ng/mL (IQR 711.5–1615) vs 730 ng/mL (IQR 575–1065); p=0.010). In the IFX group, the median baseline level was 2650 ng/mL (IQR 1483–4120) in responders and 1220 ng/mL (IQR 1053–1533) in non-responders (p=0.001). Also in the RTX group, baseline MRP8/14 levels were higher in responders compared with non-responders (2811 ng/mL (IQR 1945–4525) versus 1050 ng/mL (IQR 780–1290); p<0.001). No significant differences in levels were seen between the (small) subgroups of good versus moderate responders (3390 ng/mL (IQR 1978–4608) versus 2811 ng/mL (IQR 1783–4825); p=0.940). Thus, MRP8/14 serum levels were consistently higher in responders to targeted treatment, independent of the specific mechanism of action.

Figure 1

Baseline levels of myeloid-related proteins (MRP)8/14 (ng/mL) in European League Against Rheumatism (EULAR) good and moderate responders are significantly higher compared with baseline levels of MRP8/14 in EULAR non-responders to (A) adalimumab treatment, (B) infliximab treatment and (C) rituximab treatment. Each dot represents a patient, and data are presented as the median with IQR. Significance of the comparison is determined by the Mann–Whitney test.

Baseline MRP8/14 serum levels predict clinical response to treatment

When dividing the baseline MRP8/14 serum complex levels in low and high in patients treated with ADA (cut-off at median level of 995 ng/mL; median chosen because of the relatively small sample size and the non-normal distribution of the data), having a high level of MRP8/14 at baseline increased the odds of being a responder by a factor of 3.30 (95% CI 1.14 to 9.60, p=0.028, R2=0.088). Positive predictive value (PPV) of a high MRP8/14 serum level at baseline for response was 86%, and negative predictive value (NPV) of a low MRP8/14 serum level at baseline for no response was 35%. Sensitivity was 57% and specificity was 71% (table 2). High level of MRP8/14 serum level in the IFX group (cut-off at median level of 2027 ng/mL) was significantly associated with response (OR 9.75, 95% CI 1.93 to 49.33, p=0.006, R2=0.249) with PPV of 93%, NPV of 43%, sensitivity of 62% and specificity of 86% (table 2). In the RTX group, logistic regression revealed that high levels (cut-off at median level of 1665 ng/mL) were significantly associated with being a responder (OR 55, 95% CI 4.30 to 703.43, p=0.002, R2=0.633). The PPV of a high MRP8/14 level for predicting response was very high (92%) with a NPV of 83%, a sensitivity of 85% and a specificity of 91% (table 2).

Table 2

Value of high baseline level of MRP8/14 serum complex in predicting good or moderate EULAR response to therapy with ADA, IFX and RTX

We performed the ROC curve AUC analysis to determine the accuracy of baseline TJC68, baseline SJC66, baseline levels of CRP and ESR, baseline DAS28 and baseline MRP8/14 in separating good/moderate responders from non-responders. As shown in figure 2, baseline MRP8/14 was the best predictor for response. The AUC for baseline MRP8/14 levels in the ADA, IFX and RTX group was 0.688 (95% CI 0.571 to 0.804), 0.791 (95% CI 0.575 to 0.907) and 0.984 (95% CI 0.945 to 1.000), respectively.

Figure 2

Receiver operating characteristic (ROC) curve analyses for achievement of European League Against Rheumatism (EULAR) good or moderate response after treatment with (A) adalimumab, (B) infliximab or (C) rituximab. The increasing area under the ROC curve (area under the curve ) corresponds to a higher diagnostic test yield. The diagnostic accuracy of baseline tender joint count of 68 joints (TOTTJC68 baseline—thick dotted black line), baseline swollen joint count of 68 joints (TOTSJC68 baseline—thin dotted black line), baseline erythrocyte sedimentation rate level (thin black line), baseline C-reactive protein level (dashed black line), baseline disease activity score 28 (dashed grey line) and baseline myeloid-related proteins 8/14 levels (thick black line) are presented.

An univariate logistic analysis was also used to assess the relationship between IgM-RF positivity, anti-citrullinated peptide antibody (ACPA) positivity, baseline TJC68, baseline SJC66, baseline levels of CRP and ESR, and baseline DAS28, on the one hand, and clinical response to treatment, on the other hand. Baseline DAS28 was related to response to ADA treatment (OR 1.840, 95% CI 1.031 to 3.285, R2=0.08, p=0.039), as was the TJC68 at baseline to response to IFX treatment (OR 0.916, 95% CI 0.841 to 0.998, R2=0.99, p=0.046).

Next, we performed stepwise backward multivariate logistic regression analysis with DAS28 and TJC68 in combination with baseline MRP8/14 serum levels to assess the additional value of MRP8/14 in predicting clinical response. In this model, high baseline MRP8/14 serum level was the only statistically significant independent determinant of therapy response in patients treated with ADA (OR 3.14, 95% CI 1.06 to 9.32, p=0.040), with IFX (OR 7.82, 95% CI 1.49 to 40.95, p=0.006) and RTX (OR 210.21, 95% CI 3.48 to 12716.88, p=0.002).

Decrease of MRP8/14 serum complex levels in responders to treatment

Treatment with ADA, IFX or RTX lowered the levels of MRP8/14 significantly in the combined group of good and moderate responders, but not in non-responders (figure 3). This change already reached statistical significance after 4 weeks for the ADA and IFX groups, suggesting that MRP8/14 could be used to monitor the early response to anti-TNF treatment and to predict subsequent clinical response.

Figure 3

Changes of myeloid-related proteins (MRP)8/14 levels after 4 and 16 weeks of treatment with adalimumab, infliximab and rituximab. Levels of MRP8/14 (in ng/mL) in European League Against Rheumatism (EULAR) good and moderate responders to (A) adalimumab and (B) infliximab treatment decreased after 4 and 16 weeks of therapy, while levels in EULAR non-responders to (D) adalimumab and (E) infliximab treatment were not significantly different after 4 or 16 weeks of treatment. Levels of MRP8/14 (in ng/mL) decreased significantly after 16 weeks of treatment in (C) EULAR good and moderate responders to rituximab treatment and (F) increased significantly after 16 weeks in EULAR non-responders. Each dot represents a patient, and data are presented as the median with IQR. Significance of the comparison is determined by the Wilcoxon ranked t test.

For the analysis of MRP8/14 serum complex levels after ADA treatment, paired samples of all 65 responders and 20 non-responders were available for the week 4 analysis and paired samples of 62 responders and 21 non-responders were available for the week 16 analysis. For the ADA responders, the median baseline MRP8/14 level decreased by 46% at week 4 (figure 3A, p<0.0001) and by 61% at week 16 (figure 3A, p<0.0001), with a significantly higher median delta change (Δ) in responders than non-responders (p=0.0002 and p<0.0001, respectively). In the good responders subgroup, MRP8/14 serum levels were significantly decreased after 4 and 16 weeks by 37% and 57%, respectively (see online supplementary table S4). Also in the moderate responders subgroup, there was significant decrease of MRP8/14 serum levels after 4 week (50%) and 16 weeks (68%). In patients who did not respond, the MRP8/14 serum levels did not decrease (figure 3D). The median Δ after 4 and 16 weeks was higher in responders than non-responders (p=0.0002 and p<0.0001, respectively).

For analysis at 4 weeks, paired samples of 31 responders and only 3 non-responders to IFX were available, whereas for analysis at 16 weeks, paired samples of 35 responders and 6 non-responders to IFX were available. We observed similar changes as described above in the ADA cohort. In the IFX group, a decrease of MRP8/14 serum level was seen in the responding patients after 4 weeks (60%, median Δ 840 ng/mL; IQR 170–1170, p<0.0001) and after 16 weeks of treatment (68%, median Δ970 ng/mL; IQR 530–1830, p<0.0001). When separating the good and moderate responders, the decrease in MRP8/14 serum levels after treatment was still statistically significant. There was a decrease of 56% after 4 weeks and 58% after 16 weeks in good responders and in moderate responders the decrease after 4 and 16 weeks was 63% and 68%, respectively. In non-responders, MRP8/14 levels did not change after 4 and 16 weeks. The median Δ after 16 weeks was also significantly higher in responders compared with non-responders (p=0.0056).

For the MRP8/14 analysis at 4 weeks, paired samples of 11 responders and 9 non-responders to RTX were available. Paired samples after 16 weeks of treatment were available in all responders to RTX and in 9 non-responders to RTX. In the RTX group, a decrease in MRP8/14 serum levels was found after 4 weeks in responders (60%, median Δ 1230 ng/mL; IQR 286–1931, p=0.123), but this decrease did not reach statistical significance, possibly due to the small number of patients. However, comparing the median Δ after 4 weeks between responders and non-responders, there was a statistically significant difference between responders (decrease of 1300 ng/mL (IQR 333–4030)) and non-responders (increase of 80 ng/mL (IQR −210 to 290)) (p=0.0150). Sixteen weeks after treatment, a significant decrease of MRP8/14 serum levels was observed in the responders (59%, p=0.0005), but not in non-responders (increase of 89%, p=0.0039).


The results from this study show that (1) serum levels of MRP8/14 are related to clinical signs and symptoms; (2) MRP8/14 at baseline might be used to predict future response to targeted treatment independent of the specific mechanism of action; and (3) serial measurement of MRP8/14 levels in serum could be used to monitor the early response to treatment, predictive of the clinical response over time.

The major source of MRP8/14 is believed to be the synovial compartment of arthritis patients,34 and indeed MRP8/14 proteins are upregulated in the inflamed synovial tissue and synovial fluid.20 ,28 ,34 ,35 In an earlier study, we demonstrated the expression of MRP8 and MRP14 in inflamed synovial tissue of RA patients who started infliximab treatment. Increased expression of MRP8 and MRP14 was observed on the group level in patients who subsequently responded well to anti-TNF treatment.28 Although miniarthroscopy is a very valuable research tool that is safe and well tolerated,36 it is not feasible in every outpatient clinic. Therefore, measurement in serum of the MRP8/14 complex as a predictive biomarker is a far more attractive strategy.

The correlation between serum levels of MRP8/14 and measures of disease activity described here is consistent with previous reports in RA and juvenile idiopathic arthritis diseases.25 ,37 ,38 In RA patients, the concentrations of MRP8/14 serum complexes were also shown to be associated with laboratory and clinical markers of disease activity, like CRP, SJC and DAS28.39

To reduce unnecessary side effects and costs in patients treated with biologicals but are non-responders, it would be helpful to have a test with high specificity, meaning it can predict non-response to a biological with high degree of certainty. As shown by multivariate logistic regression, MRP8/14 at baseline seems to have the ability to predict response to treatment independently of other disease activity parameters with a moderate to high specificity ranging from 71% to 90%.

In patients with recent onset RA, normalisation of MRP8/14 serum levels has previously been described in those who achieved remission after conventional antirheumatic treatment.39 In line with previous findings in adalimumab treatment,40 we have shown here that concentrations of MRP8/14 serum complexes decrease during effective treatment with biologic agents with different mechanisms of action. A significant decrease of MRP8/14 levels was observed in the responders to biologicals after 4 weeks of treatment while this effect could not be observed in the patients who did not respond to treatment. Reports about early changes in biomarkers other than MRP8/14 after initiation of biological treatment have revealed variable results.41 ,42 One study in 207 RA patients suggested that failure to suppress CRP levels after the first infusion identified the majority of patients who were non-responders at week 12.41 Another small study with 18 RA patients showed that the reduction in serum CRP levels after the first infusion of infliximab was similar between subsequent responders and non-responders.42 The results from our study in three independent cohorts show that measurement of MRP8/14 levels may be used to monitor the early changes after biological treatment. Conceivably, this approach can be developed to facilitate discontinuation of treatment early on in patients who are unlikely to exhibit clinical improvement over time. Since MRP8/14 is a relatively stable protein and can be measured in serum without need for cold storage, unlike cytokines, it is a feasible candidate for multicentre studies to validate its potential.

There are several limitations to our study. First, our findings are limited by a relatively small sample size, especially regarding the paired comparisons. Of importance however, the results are consistent across the three cohorts tested. Second, the cut-off points used to predict clinical response are at this point arbitrary. It remains to be clarified which cut-off value is most clinically relevant. This study provides the rationale for future studies in independent cohorts across a variety of mechanisms of action, investigating the value and added value of measurement of MRP8/14 levels relative to different variables of clinical improvement.

In conclusion, our study shows that measurement of serum levels of MRP8/14 complexes at baseline might be used to predict response to anti-TNF antibody and rituximab treatment in RA patients as well as in monitoring response early after initiation of treatment. Next steps should include studies in patients treated with other medicines, including abatacept, tocilizumab and tofacitinib, and further standardisation of the laboratory test.


We thank our study subjects for participating in the study.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Handling editor Tore K Kvien

  • Contributors IYC was involved in study design, data collection, data analysis, interpretation of data and preparing the manuscript. DMG was involved in data analysis, interpretation of data and preparing the manuscript. MJH was involved in study design and data collection. RMT was involved in data collection and interpretation of data. CAW was involved in data collection. DF and TV were involved in revising the manuscript critically. PPT and DH were involved in study design, data collection, interpretation of data and preparing the manuscript.

  • Funding This research was supported by the Center for Translational Molecular Medicine (project TRACER, grant 04I-202), the Innovative Medicines Initiative project BTCure (grant agreement number 115142-1), EU FP7 grants EUTRAIN and MIAMI and the Interdisciplinary Centre of Clinical Research at the University of Muenster (IZKF CRA04).

  • Competing interests PPT is currently also an employee at GlaxoSmithKline, Stevenage, U.K. GSK was not involved in this study.

  • Ethics approval AMC Medical Ethical Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.