Identification of differential co-expressed gene networks in early rheumatoid arthritis achieving sustained drug-free remission after treatment with a tocilizumab-based or methotrexate-based strategy

Rheumatoid arthritis (RA) is an autoimmune disease characterized by synovial tissue inflammation of the peripheral joints, often followed by destruction of those affected joints [1, 2]. Although the cause of RA is not fully understood, it is recognized to be a multifactorial disease involving both environmental and genetic factors that influence the susceptibility and severity of the disease [3‐5]. Approximately 50% of the risk for developing RA is attributable to genetic predisposing factors. Previous studies have shown that, for example, the protein tyrosine phosphatase, non-receptor type 22 (PTPN22) and peptidyl arginine deiminase, type 14 (PADI4) genes and various human leucocyte antigen (HLA) class II alleles are associated with an increased risk [4, 6‐8]. Furthermore, the heterogeneity of RA is demonstrated by the variability in the presence of autoantibodies, the diverse clinical disease manifestations and different grades of responses to disease-modifying anti-rheumatic drugs (DMARDs). Adequately targeting the disease from the start is essential to preserve physical function and ensure long-term beneficial outcome, especially in early RA [9]. A better understanding of gene expression profiles could help to distinguish between patients with different pathogenic mechanisms and could lead to improved clinical outcomes when more individualised therapy is initiated from the start. An anchor drug in the treatment of RA is methotrexate, one effect of which is inhibition of dihydrofolate reductase that is not only important for cell proliferation and cell growth but also for inducing protein synthesis [10]. The various mechanisms of methotrexate involved in the treatment of RA still have not yet been fully elucidated, but it is believed that an anti-inflammatory effect also plays a large role as in vitro and in vivo studies have demonstrated previously [11]. Although methotrexate is recommended as an initial treatment strategy [12], a large number of patients needs additional treatment with a biological DMARD or withdraw from this therapy because of inefficacy or adverse effects [13‐15]. Tocilizumab is a humanized monoclonal antibody inhibiting interleukin (IL)-6 signalling by blocking the binding of IL-6 to its receptor, and has been proven in randomized controlled trials to be effective across different RA populations in reducing disease activity and inhibiting the progression of joint damage [16‐28]. IL-6 is a pro-inflammatory cytokine that is secreted by multiple cell types (e.g. T cells, B cells, monocytes, and osteoblasts) and stimulates the immune response and osteoclast formation [29, 30]. In the present study, we aimed to identify biological networks and signature protein coding genes that are associated with achieving sustained drug-free remission (sDFR) after initiating treatment with tocilizumab, methotrexate, or the combination of both, by performing whole transcriptome ribonucleic acid sequencing (RNA-seq) of positive cluster of differentiation 4 (CD4⁺) and CD14⁺ cells obtained from DMARD-naive early RA patients. Additionally, we were interested whether pathways related to achieving sDFR were treatment dependent by investigating gene expression profiles between the tocilizumab and methotrexate arms, or whether sDFR was dependent on general predisposing factors present within all strategy arms. As genotypes differ between RA patients, identification of biological pathways associated with achieving sDFR could enable application of more personalized treatment strategies which, in those who are newly diagnosed, would result in better disease activity control and prevention of disease progression.

In total, 60 individual patient samples (tocilizumab plus methotrexate, 19 (n = 14 achieved sDFR, n = 5 controls); tocilizumab, 24 (n = 13 achieved sDFR, n = 11 controls); methotrexate, 17 (n = 10 achieved sDFR, n = 7 controls)) were included in the present analysis. The mean (SD) age of all patients was 53 (14) years with a median (interquartile range (IQR)) symptom duration of 23 (18–40) days; 60% were rheumatoid factor positive and 60% anti-cyclic citrullinated peptide positive. The majority were female (68%) and patients had at baseline a mean (SD) DAS28 of 4.9 (1.1); median (IQR) CRP level was 9 (3–19) mg/L and erythrocyte sedimentation rate 20 (12–32) mm/h. No significant differences in clinical characteristics were found between those achieving sDFR and the controls within the strategy arms (p ≥ 0.07, Additional file 1: Table S1). For each sample, the expression of approximately 20,000 protein coding genes was measured in both CD4⁺ and CD14⁺ cells and normalized read counts were determined for each gene. Hierarchical cluster analyses of the relevant DEGs are shown in Fig. 2. In the sequenced CD4⁺ cell population, weighted network analyses identified two, four, and four modules after merging that were significantly correlated with achieving sDFR in the tocilizumab plus methotrexate, tocilizumab, and methotrexate strategy arms, respectively (Table 1). In the tocilizumab plus methotrexate arm, the salmon module was found to be most significantly related to sDFR (PCC 0.58, p = 0.009); in the tocilizumab arm it was the purple module (PCC 0.52, p = 0.009), and in the methotrexate arm the black module (PCC 0.60, p = 0.010). The number of genes included in the modules is 40, 26, and 49, respectively (Additional file 2: Table S2). When performing network analyses within the CD14⁺ cells, no modules were found that were significantly correlated with achieving sDFR in the tocilizumab plus methotrexate and methotrexate arms (Additional file 3: Table S3). In the tocilizumab arm, only one module (pink) was significantly correlated (PCC 0.41, p = 0.049). However, within this module, symptom duration was an important contributing trait as it was highly correlated (PCC 0.93, p < 0.001) with the average expression profile (E). Because the correlation between the pink module and achieving sDFR was low (<0.50) and symptom duration was a more contributing factor to the gene expression in this module, it was disregarded for further analyses. Pathway analyses were therefore only performed for the modules found relevant within the CD4⁺ cells of the three treatment arms.

In the present study, using analysis of networks of differentially co-expressed genes, several biological pathways were identified that were associated with achieving sDFR in DMARD-naive patients with early RA after initiating tocilizumab plus methotrexate-, tocilizumab-, or methotrexate-based treatment strategies. Across the three treatment arms, we found different networks, implying that there are several predisposing genes for achieving sDFR and, depending on the treatment strategy selected, patients have a higher chance of achieving sDFR if these are up- or downregulated. Some genes in particular that are included in the networks may play an important role as they are characterized by a high connectivity with other (signature) genes and therefore have the strongest influence on the pathways found relevant.

In the tocilizumab plus methotrexate strategy arm, the most important GO term was “nuclear-transcribed mRNA catabolic process, nonsense-mediated decay”, which is a pathway important in altering gene expression by degrading mRNA when converting amino acid-specific codons into premature stop codons (i.e. early termination of translation into proteins) [37]. After the genetic information from DNA is copied into RNA molecules during transcription, mRNA is decoded by ribosomes in the cytoplasm and is translated, using transfer RNA, into a specific sequence of amino acids (i.e. proteins). Translation starts with a start codon, which is the first codon of a mRNA transcript translated by a ribosome, and ends with a stop codon. Mutations, also known as single-nucleotide polymorphisms (SNPs), can occur in the sequence of nucleotide bases in DNA resulting in a different protein product. SNPs may not only increase the susceptibility to certain diseases but are also responsible for the differences in treatment response because of genetic variations. Other significant GO terms within this treatment arm are also related to processes involved in translation (GO:0006412) of mRNA by ribosomes (GO:0005840, GO:0003735) and is in accordance with the only identified significant KEGG pathway (“Ribosome”). Thus, it seems that pathways important for achieving sDFR in patients treated with tocilizumab plus methotrexate are in general related to processes involved with protein synthesis, especially during translation from mRNA into a sequence of amino acids. In the tocilizumab strategy arm, the most important GO terms in the purple module were related with migration of white blood cells (granulocytes and myeloid leukocytes) as a response to inflammation or to an external stimulus (i.e. leukocyte chemotaxis). Furthermore, G-protein-coupled receptor (GPCR) activity and signalling pathway (GO:0008227, GO:0007187) were significant GO terms within this module. GPCRs constitute a superfamily of receptor proteins that are involved in a many (patho)physiological processes and an important role of these receptors is to regulate the immune system and inflammation by activating internal signal transduction across the cell membrane, which eventually leads to cellular responses [38]. It is estimated that approximately two-third of all available drugs targets the GPCRs; the most commonly used in RA is methotrexate that works via a specific subtype of adenosine receptors, which is part of the large GPCR family [38]. The exact mechanism and the downstream effect ultimately leading to cellular responses of many GPCRs are still not fully elucidated, but they have been shown to be effective in several auto-immune diseases in reducing disease symptoms when specific receptors are targeted. Although it is known that tocilizumab inhibits the binding of IL-6 to its receptor, which belongs to the family of the type I cytokine receptors and thus does not seem to directly affect GPCRs, it might have an indirect effect on certain GPCRs. Further research is, however, required to determine and understand the potential effect of tocilizumab on these receptors. In the methotrexate arm, all selected GO terms were related to a response to some kind of bacterium or biotic (i.e. biological material) stimulus resulting in a change of state or activity of a cell. When analysing the GO ancestor chart, all terms were direct subtypes of “response to stimulus” (GO:0050896) indicating that the genes included in the black module may play a direct role in the response to therapy. However, the exact mechanisms of methotrexate in RA remain partially unclear, but multiple processes seem to play a role in suppressing disease activity. Although it is known, for example, that methotrexate inhibits cell proliferation and apoptosis, it was also found to inhibit cytokine production by inducing T-cell and B-cell activation [39]. Although many ex vivo and in vitro studies have been performed, the exact mechanism of action of cytokine level reduction by methotrexate remains largely not understood [39]. An important function of methotrexate may be the inhibiting effect on “Janus Kinase Signal Transducer and Activator of Transcription (JAK-STAT)” signalling, which was also a significantly expressed KEGG pathway in the black module (methotrexate arm). JAKs are enzymes transducing signals of cytokine receptors and, by inhibiting one or more enzymes, they prevent activation of cytokines important for the immune response [40]. A previous study has shown that methotrexate suppresses the JAK-STAT pathway; this could be an important mechanism for reducing RA disease activity [41]. Another significant KEGG pathway in the methotrexate arm was associated with signalling of the tumour protein p53 (“p53 signalling pathway”), which functions as a tumour suppressor and is thus not directly related to the pathogenesis of RA. This protein was, however, previously found to be involved with methotrexate-resistance in osteosarcoma and, therefore, it seems that genes important in the p53 pathway may be related to the response to methotrexate therapy in general [42].

In the present study, we analysed two different clusters of cells, CD4⁺ T-helper cells and CD14⁺ monocytes, for transcriptional profiling as both are important in modulating the immune response but have different functions and could thus have different pathways involved. T-helper cells are essential in the adaptive immunity by producing cytokines when pathogens are presented by monocytes and are therefore regulators of the immune response. Besides recruiting lymphocytes (antigen presentation), monocytes also perform phagocytosis after becoming macrophages, and are thus also important for the removal of pathogens. Although both clusters of cells are important within RA, when performing gene co-expression network analyses, no relevant modules were found in the CD14⁺ cells and only transcripts sequenced from CD4⁺ cells seems to play a role in achieving sDFR in DMARD-naive patients after initiating a tocilizumab- or methotrexate-based strategy. Other cells, such as B lymphocytes, also play an important role in auto-immune diseases and thus may contribute to the understanding of genetic variations within RA patients. However, these cells were not analysed in this study as they were not isolated because of the limited quantity of blood that had been collected at baseline and cost-related considerations.

In conclusion, by performing network analyses of co-expressed genes we have identified several pathways related to achieving sDFR within DMARD-naive early RA patients after initiation of tocilizumab plus methotrexate-, tocilizumab-, or methotrexate-based treatment strategies. Within each of these three strategy arms, we found different biological pathways implicating that are different networks of predisposing genes related to achieving sDFR which are dependent on the type of treatment strategy selected. In addition, we identified several important genes within the networks that could potentially act as prognostic biomarkers for RA.