Introduction
In a biopsy-based, precision-medicine, randomized clinical trial in RA (R4RA; n = 164), patients with low/absent synovial B cell molecular signature had a lower response to rituximab (anti-CD20 monoclonal antibody) compared with that to tocilizumab (anti-IL6R monoclonal antibody) although the exact mechanisms of response/nonresponse remain to be established. Here, in-depth histological/molecular analyses of R4RA synovial biopsies identify humoral immune response gene signatures associated with response to rituximab and tocilizumab, and a stromal/fibroblast signature in patients refractory to all medications. Using ten-by-tenfold nested cross-validation, we developed machine learning algorithms predictive of response to rituximab, tocilizumab and, notably, multidrug resistance. This study supports the notion that disease endotypes, driven by diverse molecular pathology pathways in the diseased tissue, determine diverse clinical and treatment– response phenotypes. It also highlights the importance of integration of molecular pathology signatures into clinical algorithms to optimize the future use of existing medications and inform the development of new drugs for refractory patients.
Study
To address this hypothesis we carried out a biopsy-driven, randomized clinical trial in RA (R4RA)9 in which TNF-inhibitor-inadequate responders were randomized to either rituximab (anti-CD20 monoclonal antibody) or tocilizumab (anti-IL6R monoclonal antibody) after stratification according to synovial B cell signatures. The trial results demonstrated that only 12% of patients with a low synovial B cell molecular signature responded to rituximab while 50% responded to tocilizumab. In contrast, in patients with high synovial B cell lineage signature, the two drugs appeared comparably effective. Here, we investigated the mechanisms of response and nonresponse to these two targeted biologics through deep histopathological and molecular (RNA-sequencing (RNA-Seq)) characterization of synovial tissue at baseline, and longitudinally in post-treatment biopsies at 16 weeks. We identified specific signatures associated with therapeutic response and developed machine learning classifiers to predict treatment response. Additionally, we provide insights into the cellular and molecular pathways underpinning multidrug resistance defining a refractory phenotype, characterized by a stromal/fibroblast signature. Finally, digital spatial profiling of synovial biopsies highlighted differences in gene expression in specific synovial regions with relevance to disease pathogenesis and treatment response.
Results
Histological and in silico cell lineages
correlate with drug response
To assess the association of synovial immune cells with treatment response, we compared semiquantitative immunohistochemistry (IHC) scores (Extended Data Fig. 1a,b) in pretreatment synovial biopsies of responders (n = 28 for rituximab, n = 37 for tocilizumab) and nonresponders (n = 54 and n = 42, respectively), showing no differences (Extended Data Fig. 1c). However, when patients were stratified according to previously described, synovial histological patterns, also known as pathotypes (Fig. 1a), patients with a diffuse-myeloid pathotype, i.e. with myeloid lineage predominance but low in B/plasma cells, displayed a significantly higher response to tocilizumab (13/16, 81%) versus rituximab (7/20, 35%) (P = 0.008, odds ratio (OR) = 7.53, 95% confidence interval (CI) 1.4–55.7).
Next, we used unsupervised analyses to explore the relationship of multiple genes/pathways with response to treatment. First, we applied principal component analysis (PCA) to identify underlying subgroup structures. PC1 and PC3 correlated with inflammatory cell infiltration in synovial biopsies, while they also associated with histological pathotypes primarily separating the lympho-myeloid and fibroid pathotypes.
Molecular signatures of treatment
response
Next, we performed DEG analysis to identify genes associated with treatment response on all patients who at any point in the trial had received rituximab or tocilizumab. A total of 6,625 genes were significantly different (false discovery rate (FDR) < 0.05) in rituximab responders compared with nonresponders. Both nonresponder groups showed upregulation of extracellular matrix genes, including integrin-binding sialoprotein, aggrecan and collagen, and genes linked to tissue remodeling, cell infiltration and cell–cell interaction. Following adjustment for immune cell infiltration by PC1, DEGs for rituximab remained significant and, in the case of tocilizumab, the number of identified DEGs increased suggesting that DEG analysis provides an additional dimension to the inflammatory cell infiltrate alone that differentiates responders from nonresponders.
Refractory disease is linked to a
stromal/fibroblast signature
To further explore the hypothesis of a common refractory signature following treatment switch at 16 weeks
1) We compared patients in whom both rituximab and tocilizumab failed to induce response (multidrug resistant/refractory, n = 40 for histology, n = 32 for RNA-seq) with
(1) patients who responded exclusively to rituximab after tocilizumab failure (pro-rituximab, n = 11 for histology and n = 9 for RNA-seq) and
(2) patients who responded exclusively to tocilizumab after rituximab failure (pro-tocilizumab, n = 13 for histology and n = 12 for RNA-seq).
To further characterize the association of synovial fibroblast genes with multidrug resistance, we complemented MCP-counter deconvolution by examining enrichment in synovium-specific fibroblast gene modules derived from RA synovial single-cell RNA-seq14. The signature for HLA-DRAhigh sublining fibroblasts (SC-F2), a proinflammatory subset associated with leukocyte- rich synovial infiltration in RA. It was significantly higher in responders (P = 0.027) as opposed to CD34+ sublining fibroblasts (SC-F1) and, in particular, to the newly described DKK3+ sublining fibroblasts (SC-F3), both increased in refractory patients (P = 0.036 and 0.00055, respectively). For orthogonal confirmation of these findings at the protein level, we used multiplex immunofluorescence to detect DKK3+ fibroblasts in the synovial lining and sublining of refractory patients. Together, these results show that baseline histological and molecular signatures are associated with response to individual drugs, while nonresponse to multiple biologics is linked to a specific pretreatment signature associated with fibroblasts.
Digital spatial profiling of refractory RA
Because immune and stromal cells are known to exhibit positional identity relevant to the pathogenesis of RA15, we used digital spatial profiling (DSP) to characterize the spatial positioning of cell signatures in association with treatment response/resistance. We employed GeoMx DSP (NanoString), which uses a set of protein lineage markers to define regions of interest (ROIs) that undergo whole-transcriptomic spatial RNA expression (Fig. 2a). First, we compared gene expression in responders and refractory patients across all ROIs: lining/superficial sublining, deep sublining and lymphoid aggregates (Fig. 2b). Consistent with the above bulk RNA-seq modules and protein expression, multiple genes related to the DKK3+ fibroblast subset (PRELP, OGN, CAM1KD) were significantly higher in refractory patients (Fig. 2c).
Pre and post treatment histopathological
and molecular analyses
To explore the longitudinal effects of each drug on synovial immune cell infiltration and gene expression, we compared paired synovial samples at baseline and 16 weeks (rituximab, n = 41 for histology and n = 29 for RNA-seq; tocilizumab, n = 24 and n = 15, respectively).
Machine learning models predict drug
response and multidrug resistance
To establish the ability of synovial tissue gene expression in prediction of treatment response/resistance, we developed machine learning predictive models with the dataset partitioned for training and testing using ten-by-tenfold nested cross-validation, as detailed in Methods and schematically in Fig. 6.
Discussion
An enhanced response to tocilizumab is associated with the presence of myeloid cells. Response to rituximab was associated with IL-6 pathway genes, but also lymphocyte and immunoglobulin genes. Nonresponse to both drugs was defined by >1,000 genes and several shared pathways. Stromal cells may be a new drug target that helps overcome the complex problem of refractoriness in RA. Genes encoding for CD24, a B cell marker associated with response to biologic treatment, were found to be upregulated in refractory patients. Rituximab reduced synovial CD20+ B cells in both responders and nonresponders. A clinically relevant response required broader and deeper impacts on CD79a+ plasmablasts and CD138+ plasma cells. Nonresponders were characterized by a failure to reduce sublining macrophages. Biological differences in synovial gene expression following treatment with rituximab or tocilizumab were consistent with the cognate treatment targets: B cell depletion and IL-6 receptor blockade. However, they also revealed unexpected differences such as differential changes in metalloproteinases.
References:
1. Smolen, J. S. et al. Rheumatoid arthritis. Nat. Rev. Dis. Prim. 4, 18001 (2018).
2. Buch, M. H. Defining refractory rheumatoid arthritis. Ann. Rheum. Dis. 77, 966–969 (2018).
3. Borghaei, H. et al. Nivolumab versus docetaxel in advanced nonsquamous non-small-cell lung cancer. N. Engl. J. Med. 373, 1627–1639 (2015).
4. Kawalec, P., Łopuch, S. & Mikrut, A. Effectiveness of targeted therapy in patients with previously untreated metastatic breast cancer: a systematic review and meta-analysis. Clin. Breast Cancer 15, 90–100 (2015).
5. Pitzalis, C., Choy, E. H. S. & Buch, M. H. Transforming clinical trials in rheumatology: towards patient-centric precision medicine. Nat. Rev. Rheumatol. 16, 590–599.
6. Lewis, M. J. et al. Molecular portraits of early rheumatoid arthritis identify clinical and treatment response phenotypes. Cell Rep. 28, 2455–2470 (2019).
7. Humby, F. et al. Synovial cellular and molecular signatures stratify clinical response to csDMARD therapy and predict radiographic progression in early rheumatoid arthritis patients. Ann. Rheum. Dis. 78, 761–772 (2019).
8. Giacomelli, R. et al. The growing role of precision medicine for the treatment of autoimmune diseases; results of a systematic review of literature and experts’ consensus. Autoimmun. Rev. 20, 102738 (2021).
9. Humby, F. et al. Rituximab versus tocilizumab in anti-TNF inadequate responder patients with rheumatoid arthritis (R4RA): 16-week outcomes of a stratified, biopsy-driven, multicentre, open-label, phase 4 randomised controlled trial. Lancet 397, 305–317 (2021).
10. Becht, E. et al. Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression. Genome Biol. 17, 218 (2016).
11. John, C. R. et al. M3C: Monte Carlo reference-based consensus clustering. Sci. Rep. 10, 1816 (2020).
12. Li, S. et al. Molecular signatures of antibody responses derived from a systems biology study of five human vaccines. Nat. Immunol. 15, 195–204 (2014).
13. Stephenson, W. et al. Single-cell RNA-seq of rheumatoid arthritis synovial tissue using low-cost microfluidic instrumentation. Nat. Commun. 9, 791 (2018).
14. Zhang, F. et al. Defining inflammatory cell states in rheumatoid arthritis joint synovial tissues by integrating single-cell transcriptomics and mass cytometry. Nat. Immunol. 20, 928–942 (2019).
15. Wei, K. et al. Notch signalling drives synovial fibroblast identity and arthritis pathology. Nature 582, 259–264 (2020).
16. Croft, A. P. et al. Distinct fibroblast subsets drive inflammation and damage in arthritis. Nature 570, 246–251 (2019).
17. Crozat, K. et al. Cutting edge: expression of XCR1 defines mouse lymphoid-tissue resident and migratory dendritic cells of the CD8α+ type. J. Immunol. 187, 4411–4415 (2011).
18. Graham, S. A. et al. Prolectin, a glycan-binding receptor on dividing B cells in germinal centers. J. Biol. Chem. 284, 18537–18544 (2009).
19. Curran, A. M., Naik, P., Giles, J. T. & Darrah, E. PAD enzymes in rheumatoid arthritis: pathogenic effectors and autoimmune targets. Nat. Rev. Rheumatol. 16, 301–315 (2020).
20. Yanni, G., Whelan, A., Feighery, C. & Bresnihan, B. Synovial tissue macrophages and joint erosion in rheumatoid arthritis. Ann. Rheum. Dis. 53, 39–44 (1994).
21. Bresnihan, B. et al. Synovial tissue sublining CD68 expression is a biomarker of therapeutic response in rheumatoid arthritis clinical trials: consistency across centers. J. Rheumatol. 36, 1800–1802 (2009).
22. Teng, Y. K. et al. Immunohistochemical analysis as a means to predict responsiveness to rituximab treatment. Arthritis Rheum. 56, 3909–3918 (2007).
