Objective Despite the fact that rituximab depletes B cells in all treated patients with RA, not all patients show a favourable clinical response. The goal of this study was to provide insight into pharmacological changes in peripheral blood that are associated with clinical response to rituximab.
Methods Gene expression profiling was performed on peripheral blood RNA of 13 patients with RA (test group) using Illumina HumanHT beadchip microarrays. An independent group of nine patients was used for validation using TaqMan quantitative PCR. Clinical responder status was determined after 6 months using change in 28-joint Disease Activity Score (ΔDAS28) and European League Against Rheumatism (EULAR) response criteria. Significance analysis of microarrays and ontology analysis were used for data analysis and interpretation.
Results Pharmacogenomic analyses demonstrated marked interindividual differences in the pharmacological responses at 3 and 6 months after start of treatment with rituximab. Interestingly, only differences in the regulation of type I interferon (IFN)-response genes after 3 months correlated with the ΔDAS28 response. Good responders (∆DAS>1.2; n=7) exhibited a selective increase in the expression of type I IFN-response genes, whereas this activity was unchanged or hardly changed in non-responders (∆DAS<1.2; n=6) (p=0.0040 at a cut-off of 1.1-fold induction). Similar results were obtained using EULAR response criteria. These results were validated in an independent cohort of nine patients (five non-responders and four responders, p=0.0317).
Conclusions A good clinical response to rituximab in RA is associated with a selective drug-induced increase in type I IFN-response activity in patients with RA. This finding may provide insight in the biological mechanism underlying the therapeutic response to rituximab.
Statistics from Altmetric.com
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 systemic autoimmune disease characterised by chronic inflammation of the joints that may cause permanent cartilage and bone destruction. Although the aetiology is still unknown it is assumed that the interplay between genetic background, immunological and environmental factors is important in disease pathogenesis.1
No curative treatment is currently available, and patients are subjected to a prolonged course of treatment. Because of the increased insight in the role of proinflammatory mediators and cellular principles of immune tolerance a number of therapeutic options for RA have been developed and are currently available.2 3 However, a major challenge in the successful treatment of RA is disease heterogeneity, which is also presented in the diverse effects of treatment responses.4
One of the treatment strategies is targeting of B cells making use of rituximab, a chimaeric monoclonal antibody directed against the B cell marker CD20. This treatment was shown to be highly effective for suppression of disease activity in RA.5,–,7 Clinical studies have demonstrated that in analogy to anti-tumour necrosis factor (TNF) therapies not all patients show a favourable response to rituximab treatment.8 Despite the fact that rituximab directly depletes specific B cell populations in all patients treated, the existence of interindividual differences in clinical outcome of rituximab treatment has raised questions regarding the mechanism of action.9 10 In order to provide insight in the biological basis for the clinical response towards rituximab we evaluated the pharmacological effects of rituximab in relation to the clinical outcome using genome-wide gene expression technology in whole blood of patients with RA.
Materials and methods
Patients with established RA according to the revised American College of Rheumatology (ACR) criteria for the diagnosis of RA11 were consecutive recruited from two rheumatology outpatient clinics in The Netherlands (VU University Medical Center and Reade/Jan van Breemen Institute, Amsterdam). Inclusion criteria for this study are according to the guidelines of the Dutch Society for Rheumatology for treatment with rituximab, that is, active disease status despite previous treatment with methotrexate and one other disease-modifying antirheumatic drugs (DMARDs) and despite TNF-blocking agents, unless contraindicated in the opinion of the treating doctor. The study was approved by the local medical ethics committees and patients provided written informed consent. Patients had a delay of at least 4 weeks between last dose of TNF blocker and first injection of rituximab.
Treatment and clinical evaluation
Patients received rituximab 1000 mg intravenously on days 1 and 15, in combination with clemastine (2 mg intravenously), methylprednisolone (100 mg intravenously) and acetaminophen 1000 mg orally, as premedication. At 4 weeks after the first infusion and from 12 weeks on every 3 months the 28-joint Disease Activity Score (DAS28) was assessed for disease activity status and blood was drawn. The use of concomitant DMARDs, prednisolone or non-steroidal anti-inflammatory drugs during the study duration was permitted. Response to treatment was classified at 6 months following the start of treatment according to European League Against Rheumatism (EULAR) response and to change to DAS28 (ΔDAS28).12
Blood sampling and RNA isolation
For RNA isolation, 2.5 ml blood was drawn in PAXgene tubes (PreAnalytiX, Hilden, Germany) and stored at −20°C. After overnight thawing at room temperature total RNA was isolated using Bio robot MDX (Qiagen Benelux BV, Venlo, The Netherlands) according to the manufacturer's instructions (PAXgene Blood RNA Mdx kit). Samples were cleaned using Qiagen RNA MinElute (Qiagen, Venlo, The Netherlands). Total RNA concentration was measured using the NanoDrop spectrophotometer (NanoDrop Technologies, Wilmington, Delaware, USA) and RNA purity and integrity was verified using lab-on-chip technology (Agilent 2100 Bioanalyzer; Agilent, Palo Alto, California, USA).
The Illumina TotalPrep RNA amplification kit (Ambion, Austin, Texas, USA) was used to synthesize biotin-labelled cRNA from 500 ng total RNA. A total of 750 ng of biotinylated cRNA was hybridised onto the HumanHT-12 v3 Expression BeadChip (Illumina, San Diego, California, USA).
Amplification and hybridisation were performed at the outsourcing company ServiceXS (Leiden, The Netherlands). Bead summary intensities were log2 transformed and normalised using quantile normalisation.13 14
Calculation of interferon (IFN) type I response scores
Expression of six IFN-response genes (IFI44, IFI44L, HERC5, RSAD2, LY6E and Mx1) was quantitated using the log2 transformed and quantile normalised expression values. IFN type I response scores were calculated as the mean of the expression values of the six IFN response genes for every single patient and at every single timepoint.
cDNA synthesis and quantitative real-time PCR
RNA (0.5 μg) was reverse transcribed into cDNA using a Revertaid H-minus cDNA synthesis kit (MBI Fermentas, St Leon-Rot, Germany). Quantitative real-time PCR was performed using an ABI Prism 7900HT Sequence detection system (Applied Biosystems, Foster City, California, USA). Gene expression levels were determined using TaqMan Gene expression assays following manufacturer's guidelines. To calculate arbitrary values of mRNA levels and to correct for differences in primer efficiencies for each gene a standard curve was constructed. Expression levels of target genes were quantified relative to the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) housekeeping gene.
To determine the relative amount of peripheral T and B lymphocytes, whole blood was stained for 30 min with monoclonal antibodies conjugated to fluorescein isothiocyanate (FITC), phycoerythrin (PE), peridinin chlorophyll protein (PerCP) and allophycocyanin (APC), and directed against lymphocyte subset-associated surface molecules. Four colour antibody combinations used were (FITC/PE/PerCP/APC): CD3/CD8/CD45/CD4 and CD3/CD16+56/CD45/CD19 (all from BD Biosciences, San Jose, California, USA). Subsequently, red cells were lysed (Lysing Solution; BD Biosciences) and lymphocyte subsets were analysed by flow cytometry on a standard four-colour fluorescence-activated cell scanner (FACSCalibur; BD Biosciences). Data were analysed using CellQuest Pro software (BD Biosciences). Care was taken to analyse only viable cellular events based on light scatter properties. All analyses were performed on lymphocytes, based on bright CD45 staining and low sideward scatter.
Statistical analysis on microarray data was performed using Significance Analysis of Microarray (SAM), version 3.09.15 Two class paired analysis using SAM at a false discovery rate (FDR) of less than 5% between pretreatment and post-treatment data was applied to identify genes that were significantly changed in expression after rituximab treatment. Cluster analysis was used for the categorisation of coordinately differentially expressed genes.16 TreeView was used to visualise differentially expressed genes. Gene Set Enrichment Analysis (GSEA; http://www.broad.mit.edu/gsea) was used for pathway analysis.17 18 It uses gene set permutation to adjust for multiple testing, indicated by a FDR. A minimal gene set size of 15 genes per pathway was applied, and pathways with a p value of <0.05 and a FDR of <0.05 were considered significant. A total of 282 pathways from Biocarta (http://www.biocarta.com) and the Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.genome.jp/kegg) were applied in this analysis. In addition, we incorporated the IFN-response gene set from Baechler et al19 (genes with at least a fivefold upregulation in peripheral blood mononuclear cell after treatment with type I IFN).
For ontology analysis of gene sets identified by cluster analysis we used MetaCore Pathway analysis, using the MetaCore Ontology tools, developed by GeneGo (St Joseph, Michigan, USA; http://www.genego.com/). Data mining in MetaCore is based on a manually curated database of human protein–protein, protein–DNA interactions, transcription factors, signalling pathways and metabolic pathways. Calculation of statistical significance are based on p values which are defined as the probability of a given number of genes from the input list to match a certain number of genes in the functional GeneGo Ontology categories.
Differences in gene expression levels of IFN type I response score between patients with a ∆DAS28>1.2 versus ∆DAS28 <1.2 or between EULAR good versus moderate versus non-responders were analysed using Student unpaired t test or Mann–Whitney U test were appropriate. Correlations between ∆DAS28 and induction of IFN type I response score (ratio t3/t0) were tested using Pearson correlation test.
Pharmacological response to rituximab treatment in RA
To determine the pharmacological effects of rituximab we analysed the changes in peripheral blood gene expression profiles of 13 patients with RA at baseline, and 3 and 6 months following the start of treatment. Baseline characteristics are shown in table 1. To search for single genes that were significantly regulated in all patients after 3 months of treatment with rituximab we applied two-class paired analysis using SAM at a FDR of less than 5% between pretreatment and post-treatment (table 2). The analysis revealed 16 B cell specific genes that were significantly downregulated in all patients, confirming observations by others of an effective and selective B cell depletion after 3 months of treatment.9 20 21 All patients reached a comparable low level of expression of B cell related genes, indicative that the pharmacological depletion at 3 months was reached irrespective of the clinical response status. Accordingly, pathway level analysis using GSEA, identified ‘B cell-mediated immunity’ as the only significantly downregulated pathway (p=0.0020). At 6 months following treatment only six of the B cell related genes were still significantly downregulated. At both time points, no genes were significantly upregulated. These data are indicative for a gradual rise in B cell markers from 3 to 6 months following the start of treatment. These findings were confirmed by CD19 based fluorescence-activated cell sorting analysis (data not shown). Altogether, this indicates that under the influence of rituximab only changes in B cell related processes were consistently regulated in all patients.
Variation in the pharmacological response to rituximab between patients with RA
Given the heterogeneous nature of RA and the relative low number of differentially regulated genes in the group-based analysis we questioned how consistent the pharmacological response to rituximab was between patients with RA. Therefore, we analysed the pharmacological effects of each individual patient by comparison of the ratio of the post-treatment (3 months) versus pretreatment expression level for each gene (log2 ratios). To search for differences in the pharmacological response between patients, 154 genes were identified that revealed at least a twofold difference in the rituximab-induced response in at least 3 patients. Cluster analysis categorised these 154 genes in 6 clusters (figure 1A–F) that were differentially regulated by rituximab among patients with RA. MetaCore network analysis (supplementary table 1) revealed that cluster A is characterised by the expression of genes involved in IFN signalling (eg, MxA, IFIT3, IFI44, RSAD2) and relates to antiviral actions of IFN. Genes mainly involved in translation (eg, EEF1A1, RPL15, RPL39, RPS3a), macrophage migration inhibitory factor (MIF) and T cell receptor signalling (eg, CD8, ITK) and natural killer (NK) cell cytotoxicity (eg, ITPR3, CD2) were characteristic of cluster B. Cluster C represents genes involved in B cell immunology (eg, CD79a, CD79b, CD19, Igk chain V). Genes involved in extracellular matrix (ECM) remodelling (eg, CTSG) and iron transport (eg, LTF, LCN2) are characteristic for cluster D. Cluster E involves genes related to several processes including chemotaxis (eg, CXCR1, CXCR2, FPR1), cell adhesion and angiogenesis. Finally, cluster F consisted of genes involved in cytoskeleton and coagulation. Essentially, similar genes are regulated at the 3–6 month time period, albeit that the extent of the changes was not related to the 0–3 month time period.
Altogether, these analyses show that pharmacological responses in patients with RA under the influence of rituximab treatment are highly heterogeneous between patients.
Pharmacodynamics in relation to clinical response
Next we investigated the pharmacological differences between patients in relation to clinical response. Therefore, patients were stratified based on ∆DAS at 6 months after the start of treatment in good responders (∆DAS>1.2; n=7) and non-responders (∆DAS <1.2; n=6) (figure 1B,C). Subsequently, we performed a cluster analysis using the set of 154 genes to search for genes that were differentially regulated by rituximab between responders and non-responders. Remarkably, the analysis revealed a selective increase in the expression of only cluster A genes (type I IFN-response genes) at 3 months following the start of rituximab treatment in those patients who had a good clinical response, whereas the ones with a similar or decreased expression of these genes exhibited a poor response. This association was most prominent for genes that constitute a subcluster of six genes consisting of RSAD2, IFI44, IFI44L, HERC5, LY6E and Mx1, which were used for further analyses.
Subsequently, we investigated the relationship between the rituximab-induced changes in the expression of this gene set and the clinical response status. Therefore, the expression levels of IFI44, IFI44L, HERC5, RSAD2, LY6E and Mx1 were averaged for each individual patient at each time point to reach an IFN type I response score. Rituximab-induced changes in the IFN type I response score over the 3 month time period, expressed as the ratio at 3 months versus the baseline score (ratio t3/t0), were compared between the responders and non-responders and revealed a significant increase in the IFN type I response score in the responders compared to the non-responders (p=0.0492, figure 2A). Excluding patients who were seronegative from the analysis yielded essentially similar results (p=0.05, data not shown). Division of patients into two groups based on a cut-off of 1.1-fold induction (0.15 log2 based) of IFN type I response activity resulted in a clear separation of good responders (high ΔDAS) and non-responders (low ΔDAS) (p=0.0040, figure 2B). In addition, a trend was observed for the correlation between improvement in ΔDAS and the increase in IFN type I response score (ratio t3/t0) (p=0.09, data not shown). Accordingly, similar results were observed when response status was assessed by the EULAR response criteria in good responders (n=4), intermediate responders (n=4) and non-responders (n=5) (p=0.048, figure 2C). The change in IFN-response score during rituximab treatment negatively correlated with the corresponding baseline level, although no significance was reached (p=0.0576 and R=−0.53). The IFN-response score returned to baseline values at 6 months after the start of treatment (figure 3).
We also studied the relationship between the type I IFN signature and clinical and laboratory parameters that are listed in table 1. This analysis revealed no associations between the baseline type I IFN-response activity and laboratory parameters. This is in line with results from other studies that reported on the absence of a relationship between the type I IFN signature and clinical and laboratory parameters.20,–,22 Also no significant associations were observed when we studied the relationship between IFN type I response activity and erythrocyte sedimentation rate/C reactive protein as continuous variables at baseline and at 3 and 6 months. The IFN t3/t0 ratio did show a significant correlation with the decrease in CRP at 3 months (t3/t0 ratio) (Pearson r=-0.6095), p=0.0270) (data not shown). Moreover, no association of the time between the last anti-TNF injection and start of rituximab treatment (at least 4 weeks), and baseline IFN score or induction (ratio t3/t0) was observed, indicative that our results are not influenced by differences in the delay between treatments.
To provide further evidence that differential regulation of type I IFN-response activity during rituximab is associated with the clinical response status, we tested an independent cohort of nine patients (five non-responders and four responders based on ΔDAS). Therefore, we measured the expression of RSAD2, which is a representative IFN type I response gene that has an excellent correlation with the mean expression IFN type I score at baseline and 3 months after the start of treatment (Pearson r=0.97; p≤0.0001). Analysis of only RSAD2 expression data (ratio t3/t0) for the analysis of pharmacological differences between ∆DAS28-based responders and non-responders in the test cohort revealed essentially similar results compared to usage of the six genes set (p=0.0406). Analysing the t3/t0 ratio of RSAD2 in the validation cohort revealed a significant increase in the responders compared to the non-responders after 3 months of treatment (p=0.0317, figure 2D). Moreover, a correlation was observed between improvement of ΔDAS and increase in RSAD2 expression (ratio t3/t0) (p=0.045). These results confirm that differential regulation of type I IFN-response activity during rituximab is associated with the clinical response status.
Thus, whereas rituximab depletes B cells in all patients treated irrespective of their clinical response, our data show that a drug-induced increase of type I IFN-response activity is associated with clinical response (figure 3).
Pharmacogenomic analyses demonstrated that despite the overall decrease in the expression of B cell markers, patients with RA exhibited interindividual differences in their pharmacological responses upon rituximab treatment. Among these we observed a difference in the pharmacodynamics of only one gene signature during rituximab treatment between responders and non-responders. This signature represents type I IFN-response genes. Responders exhibited an increase in IFN type I-response activity after 3 months treatment with rituximab, whereas the IFN type I-response activity was unchanged or hardly changed during treatment in the non-responders. The differential response correlated with baseline levels of IFN type I-response genes, which were low in responders and high in non-responders. These findings lead us to conclude that an increase in IFN type I-response activity during rituximab treatment is associated with a favourable response and may provide insight in the biological mechanism underlying the therapeutic response.
RA is a heterogeneous condition that is reflected by a heterogeneous response to treatment. As an exponent of that we observed that periodic analyses of the patients' blood revealed rituximab-induced interindividual differences in the gene expression profiles. Our results revealed six gene signatures that are differentially regulated after rituximab treatment between the patients. These signatures contained gene sets that represent several distinct biological processes, including IFN-response gene activity, humoral immunity, cytotoxic T and NK cell immunity and chemotaxis. The rituximab induced expression of genes in cluster B (T and NK cell mediated immunity) was inversely correlated with the expression of genes involved in ECM remodelling and connective tissue degradation (cluster D) (r=−0.62, p<0.0233) and cluster E (chemotaxis, cell adhesion) (r=−0.96, p=<0.0001). For patients with systemic lupus erythematosus (SLE) treated with rituximab an increase in NK cell activity was found for patients who exhibited a sustained response.23 Such an association between an increased NK cell activity and clinical outcome could not be found for RA (data not shown). Differences in the reduction of B cell numbers have been reported by Dass and colleagues,24 who claimed a significant association between less reduction and a poor clinical outcome. However, our data, based on cytometric and gene expression analysis, did not reveal significant differences in the expression of B cell markers and genes between responders and non-responders at 3 and 6 months after the start of treatment.
Our data reveal that rituximab-induced differential regulation of only the IFN-response genes correlated with the ΔDAS28 and EULAR outcome. Regarding the pharmacodynamics of rituximab in relation to the type I IFN activity, two interesting observations were made in this study. First, non-responders displayed an activated type I IFN system prior to the start of treatment, which remains active during treatment. Second, good responders have low or absent IFN-response activity at baseline and develop IFN-response activity during 3 months of treatment that is comparable to that of non-responders. The correlation between baseline type I IFN levels and clinical response is in line with previous findings.20
Concerning the correlation between baseline level and clinical outcome, a simple explanation could be that the pathogenesis in IFNhigh patients is less dependent on B cells compared to IFNlow patients. Alternatively, one might speculate that a high IFN activity is associated with protection from depletion of pathogenic B cells, especially located in tissues, due to concomitant increased expression of B cell survival factors such as B cell activating factor (BAFF) and B lymphocyte stimulator (BLyS).25 Additional effects of IFNs on B cell differentiation, such as in situ differentiation in CD20 negative plasma blasts,26 could contribute to diminish the effects of anti-CD20 depletion, which processes remain unseen in the peripheral blood compartment.
With respect to the increase of IFN type I activity we know that regulation of type I IFN production and the consecutive IFN-response activity is initially mediated via the Toll-like receptors, which are triggered by exogenous (infectious) agents and endogenous agents, such as nucleic acids and apoptotic/necrotic material.27 Hence, subsequent release from apoptotic/necrotic material from depleted CD20+ B cells may promote IFN production and release, which might selectively take place in the IFNlow patients. Thus, interindividual differences in the regulatory processes involved in the IFN biology, such as mentioned above, may be held responsible for the rituximab induced in the divergent regulation of IFN type I activity. Such differences could have its origin in genetic variation in the type I IFN biology. Single nucleotide polymorphisms in several key components of the type I IFN system (such as interferon regulatory factor 5 (IRF5), tyrosine kinase 2 (Tyk2) and signal transducer and activator of transcription 4 (STAT4)) have recently been identified to play a role in the differential of the IFN activity in for example, SLE.28 29
The pharmacological induction of type I IFN activity could be an important factor in the ameliorative effect of B cell depletion treatment in RA. Such a role for type I IFN activity in RA is highlighted by Treschow et al,30 who showed that IFNβ deficiency prolonged experimental arthritis. Additional evidence for a beneficial effect of type I IFNs in RA has been provided by de Hooge et al31 who demonstrated that STAT1 deficiency resulted in exacerbation of experimental arthritis. Moreover, transfer of IFN-competent FLS was able to ameliorate arthritis in IFNβ-deficient recipients.30 However, although treatment with recombinant IFNβ revealed promising results in experimental arthritis, treatment of patients with RA with IFNβ has been unsuccessful so far, which may be due to issues with dosing and pharmacokinetics.32 Conversely, a pharmacological increase in the type I IFN activity by rituximab may lead to disease progression and/or an increase in disease activity in IFN type I-driven diseases such as SLE. Recent randomised, placebo-controlled trials of rituximab failed to meet their primary or secondary clinical end points for renal and non-renal SLE.33 Our data suggest that rituximab might be less effective in those patients with SLE who experience an increase in their type I IFN response activity levels during rituximab treatment.
A relation between the dynamics of the IFN system and clinical responsiveness has also been reported for infliximab, one of the TNF blockers.34 Data from an explorative study in RA revealed that an increase in IFN-response activity during treatment with infliximab is associated with a poor clinical outcome. The dynamics of this response do not seem to correlate with the baseline IFN-response activity as was seen for rituximab. Moreover, the response pattern in relation to clinical outcome that was observed for infliximab is not in line with that for rituximab. Recent observations indicate that TNF blockade involves complex crosstalk between TNF and IFN that may exert stimulating and inhibiting effects on IFN-response activity.35 36 Unlike rituximab, the differential effects on type I IFN-response activity associated with infliximab are believed to be mainly linked to intrinsic differences in this crosstalk. As a consequence, in the context of infliximab treatment the increased IFN system seems to accompany maintenance of inflammatory activity. This is in line with earlier observations that reveal an association between synovial inflammation and synovial IFN-response gene activity.37 Overall, these findings indicate that the dynamics of the type I IFN-response activity might have different clinical consequences in RA depending on the specific biological, thus consequently type of modulated immune pathway, used.
In conclusion, our data provide insight in the pharmacological features associated with the clinical outcome of rituximab. An increase in type I IFN during treatment correlated with a favourable clinical outcome in RA. Further studies are required to define the underlying mechanism for this response association.
SV and HGR contributed equally to this work
Funding This study was partly supported by grants from the European Community's FP6 funding (AUTOCURE), MS-Research (grant no. 04-549 MS) and the Centre for Medical Systems Biology (CMSB, a centre of excellence from The Netherlands Genomics Initiative).
Competing interests MTN has received research and speaking fees from Roche. WFL is member of advisory board van Roche, Abbott and Shering Plough (MSD). BACD received speaking fees from Pfizer and Abbott. CLV is an inventor on a patent application on the predictive value of IFN type I response activity for the clinical outcome of B cell depletion treatment. SV, HGR, TCTMvdPK, MWJS, BMEvB and AEV have no competing interests to declare.
Ethics approval This study was conducted with the approval of the local medical ethics committee.
Provenance and peer review Not commissioned; externally peer reviewed.