WO2023089339A2

WO2023089339A2 - Method for treating rheumatoid arthritis

Info

Publication number: WO2023089339A2
Application number: PCT/GB2022/052948
Authority: WO
Inventors: Costantino Pitzalis; Myles J LEWIS
Original assignee: Queen Mary University Of London
Priority date: 2021-11-19
Filing date: 2022-11-21
Publication date: 2023-05-25
Also published as: WO2023089339A3; GB202116748D0

Abstract

A method for determining whether a Rheumatoid Arthritis (RA) patient is susceptible or refractory to treatment with a B cell targeted therapy and/or an agent that downregulates IL-6 mediated signalling.

Description

METHOD FOR TREATING RHEUMATOID ARTHRITIS

FIELD OF THE INVENTION

The invention relates to a method for determining whether a Rheumatoid Arthritis (RA) patient is susceptible or refractory to treatment with a B cell targeted therapy, such as rituximab, and/or an agent that downregulates IL-6 mediated signalling, such as tocilizumab. The invention also relates to methods for treating RA patients that are determined to be susceptible or refractory to B cell targeted therapy and/or an agent that downregulates IL-6 mediated signalling.

BACKGROUND TO THE INVENTION

Inflammatory arthritis is a prominent clinical manifestation in diverse autoimmune disorders including rheumatoid arthritis (RA), psoriatic arthritis (PsA), systemic lupus erythematosus (SLE), Sjogren's syndrome and polymyositis.

RA is a chronic inflammatory disease that affects approximately 0.5 to 1% of the adult population in northern Europe and North America. It is a systemic inflammatory disease characterized by chronic inflammation in the synovial membrane of affected joints, which ultimately leads to loss of daily function due to chronic pain and fatigue. The majority of patients also experience progressive deterioration of cartilage and bone in the affected joints, which may eventually lead to permanent disability. The long-term prognosis of RA is poor, with approximately 50% of patients experiencing significant functional disability within 10 years from the time of diagnosis. Life expectancy is reduced by an average of 3-10 years.

Inflammatory bone diseases, such as RA, are accompanied by bone loss around affected joints due to increased osteoclastic resorption. This process is mediated largely by increased local production of pro-inflammatory cytokines, of which tumor necrosis factor-a (TNF-a) is a major effector.

In RA specifically, an immune response is thought to be initiated/perpetuated by one or several antigens presenting in the synovial compartment, producing an influx of acute inflammatory cells and lymphocytes into the joint. Successive waves of inflammation lead to the formation of an invasive and erosive tissue called pannus. This contains proliferating fibroblast-like synoviocytes and macrophages that produce proinflammatory cytokines such as TNF-a and interleukin-1 (IL-I). Local release of proteolytic enzymes, various inflammatory mediators, and osteoclast activation contributes to much of the tissue damage. There is loss of articular cartilage and the formation of bony erosions. Surrounding tendons and bursa may become affected by the inflammatory process. Ultimately, the integrity of the joint structure is compromised, producing disability.

B cells are thought to contribute to the immunopathogenesis of RA, predominantly by serving as the precursors of autoantibody-producing cells but also as antigen presenting cells (APC) and pro-inflammatory cytokine producing cells. A number of autoantibody specificities have been identified including antibodies to Type II collagen and proteoglycans, as well as rheumatoid factors and most importantly anti citrullinated protein antibodies (ACPA). The generation of large quantities of antibody leads to immune complex formation and the activation of the complement cascade. This in turn amplifies the immune response and may culminate in local cell lysis.

Current standard therapies for RA which are used to modify the disease process and to delay joint destruction are known as disease modifying anti-rheumatic drugs (DMARDs). Methotrexate, leflunomide and sulfasalazine are traditional DMARDs and are often effective as first-line treatment.

Biologic agents designed to target specific components of the immune system that play a role in RA are also used as therapeutics. There are various groups of biologic treatments for RA including; TNF-a inhibitors (etanercept, infliximab and adalimumab), human IL-1 receptor antagonist (anakinra) and selective co-stimulation modulators (abatacept).

RA patients receive highly-targeted biologic therapies without prior knowledge of target expression levels in the diseased tissue. Approximately 40% of patients do not respond to individual biologic therapies and 5-20% are refractory to all. The mechanisms of non-response are largely unknown and, unlike other medical fields such as cancer where molecular pathology guides the use of targeted therapies, RA targeted therapeutics are prescribed “blindly” and irrespectively of the target expression levels in the diseased tissue.

Biologic therapies for RA may be associated with various safety issues, especially infusion- related adverse events and are also very expensive, for example rituximab costs approximately USD 10000 per treatment course.

Accordingly, there is a need for methods of predicting whether a given RA patient is likely to respond to biologic therapy, such as rituximab or tocilizumab treatment. There is also a need for methods of treating RA patients who are non-responsive to DMARD and/or anti-TNF therapy.

SUMMARY OF THE INVENTION The present inventors carried out in-depth analyses of synovial-biopsies from the first biopsybased precision-medicine randomised-clinical-trial in RA (R4RA) and have identified signatures associated with response to rituximab and tocilizumab, and also a signature in patients refractory to all medications.

In particular, the inventors investigated the mechanisms of response and non-response against the primary end-point (CDAI>50%) for rituximab and tocilizumab through deep histopathological and molecular (RNA-Seq) characterisation of synovial tissue at baseline and longitudinally in post-treatment biopsies at 16 weeks. The inventors identified signatures associated with therapeutic response and developed machine learning classifier modules to predict treatment response. Additionally, by analysing patients classified as non-responders at 16 weeks, who were switched to the alternative biologic therapy, the inventors developed insights into the cellular and molecular pathways underpinning multi-drug resistance that define a refractory phenotype, characterised by a stromal/fibroblast signature.

The inventors have developed predictive molecular pathology signatures that may be integrated into clinical algorithms in order to optimise the use of existing medications.

In one aspect, the invention provides a method of detecting a biomarker in a rheumatoid arthritis patient, the method comprising detecting the expression level of a biomarker in a biological sample obtained from the patient, wherein the biomarker is in Table 1 , Table 2, or Table 3. In some embodiments, the biomarker is in Table 1. In some embodiments, the biomarker is in Table 2. In some embodiments, the biomarker is in Table 3. In some embodiments, the biomarker is in Table 1 and the patient is susceptible to treatment with a B cell targeted therapy. In some embodiments, the biomarker is in Table 1 and the patient is susceptible to treatment with a B cell targeted therapy, and the method further comprises administering to the patient an effective amount of a B cell targeted therapy. In some embodiments, the biomarker is in Table 2 and the patient is susceptible to treatment with an agent that downregulates IL-6 mediated signalling. In some embodiments, the biomarker is in Table 2 and the patient is susceptible to treatment with an agent that downregulates IL-6 mediated signalling, and the method further comprises administering to the patient an effective amount of an agent that downregulates IL-6 mediated signalling. In some embodiments, the biomarker is in Table 3, and the patient is refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling.

In one aspect, the invention provides a method of detecting a biomarker in a rheumatoid arthritis patient susceptible to treatment with a B cell targeted therapy comprising detecting: (a) a level of gene expression greater than a corresponding reference value of a first biomarker in a biological sample obtained from the patient, wherein the first biomarker is selected from the group consisting of XCR1 , TCN1 , PLEKHG6, TBC1 D3, MYH6, MESP1 , RARRES2, HIST2H2AA3, CXCL2, SAA2, AC068547.1 , PI16, C6orf58, KCNIP2, EIF3CL, MAL2, MDFI, and STAC3; and/or (b) a level of gene expression less than a corresponding reference value of a first biomarker in a biological sample obtained from the patient, wherein the first biomarker is selected from the group consisting of SHC3, DLX4, COLGALT2, ERICH3, MLXIPL, MLIC6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, NKX3.2, BAIAP3, FNDC1 , WIF1 , DEFA1 B, CDON, IGHV7.4.1 , DKK3, NOG, ITGA10, and FIBIN. In some embodiments, the method comprises detecting (a). In some embodiments, the method comprises detecting (b). In some embodiments, the method comprises detecting (a) and (b). In some embodiments, the corresponding reference value is the level of gene expression in a population of patients having rheumatoid arthritis that are not responsive to treatment with a B cell targeted therapy. In some embodiments, the method further comprises administering to the patient an effective amount of a B cell targeted therapy. In some embodiments, the B cell targeted therapy is selected from the group consisting of rituximab, ocrelizumab, veltuzumab, ofatumumab, epratuzumab, obinutuzumab, ibritumomab and tiuxetan. In some embodiments, the B cell targeted therapy is rituximab.

In one aspect, the invention provides a method of treating rheumatoid arthritis in a patient in need thereof, the method comprising: (i) detecting: (a) a level of gene expression greater than a corresponding reference value of a first biomarker in a biological sample obtained from the patient, wherein the first biomarker is selected from the group consisting of XCR1 , TCN1 , PLEKHG6, TBC1 D3, MYH6, MESP1 , RARRES2, HIST2H2AA3, CXCL2, SAA2, AC068547.1 , PI16, C6orf58, KCNIP2, EIF3CL, MAL2, MDFI, and STAC3; and/or (b) a level of gene expression less than a corresponding reference value of a first biomarker in a biological sample obtained from the patient, wherein the first biomarker is selected from the group consisting of SHC3, DLX4, COLGALT2, ERICH3, MLXIPL, MUC6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, NKX3.2, BAIAP3, FNDC1 , WIF1 , DEFA1 B, CDON, IGHV7.4.1 , DKK3, NOG, ITGA10, and FIBIN; and (ii) administering to the patient an effective amount of a B cell targeted therapy. In some embodiments, the B cell targeted therapy is selected from the group consisting of rituximab, ocrelizumab, veltuzumab, ofatumumab, epratuzumab, obinutuzumab, ibritumomab and tiuxetan. In some embodiments, the B cell targeted therapy is rituximab. In some embodiments, the method comprises detecting (a). In some embodiments, the method comprises detecting (b). In some embodiments, the method comprises detecting (a) and (b). In some embodiments, the corresponding reference value is the level of gene expression in a population of patients having rheumatoid arthritis that are not responsive to treatment with a B cell targeted therapy. In one aspect, the invention provides a method of treating rheumatoid arthritis in a patient in need thereof comprising administering to the patient an effective amount of a B cell targeted therapy; wherein a biological sample obtained from the patient has: (a) a level of gene expression greater than a corresponding reference value of a first biomarker in a biological sample obtained from the patient, wherein the first biomarker is selected from the group consisting of XCRI, TCN1 , PLEKHG6, TBC1 D3, MYH6, MESP1 , RARRES2, HIST2H2AA3, CXCL2, SAA2, AC068547.1 , PI16, C6orf58, KCNIP2, EIF3CL, MAL2, MDFI, and STAC3; and/or (b) a level of gene expression less than a corresponding reference value of a first biomarker in a biological sample obtained from the patient, wherein the first biomarker is selected from the group consisting of SHC3, DLX4, COLGALT2, ERICH3, MLXIPL, MLIC6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, NKX3.2, BAIAP3, FNDC1 , WIF1 , DEFA1 B, CDON, IGHV7.4.1 , DKK3, NOG, ITGA10, and FIBIN. In some embodiments, the B cell targeted therapy is selected from the group consisting of rituximab, ocrelizumab, veltuzumab, ofatumumab, epratuzumab, obinutuzumab, ibritumomab and tiuxetan. In some embodiments, the B cell targeted therapy is rituximab. In some embodiments, the biological sample has (a). In some embodiments, the biological sample has (b). In some embodiments, the biological sample has (a) and (b). In some embodiments, the corresponding reference value is the level of gene expression in a population of patients having rheumatoid arthritis that are not responsive to treatment with a B cell targeted therapy.

In one aspect, the invention provides a method of detecting a biomarker in a rheumatoid arthritis patient susceptible to treatment with an agent that downregulates IL-6 mediated signalling comprising detecting: (a) a level of gene expression greater than a corresponding reference value of a second biomarker in a biological sample obtained from the patient, wherein the second biomarker is selected from the group consisting of XCR1 , MYH6, PLEKHG6, CD36, CXCL14, SSC5D, AC093525.2, AC005943.1 , SAA2, PTGER3, AC068547.1 , MESP1 , HIST2H2AA3, VMO1 , CDC20, AC135068.9, KCNIP2, and MDFI; and/or (b) a level of gene expression less than a corresponding reference value of a second biomarker in a biological sample obtained from the patient, wherein the second biomarker is selected from the group consisting of SHC3, DLX4, TCN1 , AP001781.2, MUC6, AC009336.2, NPIPA3, SLC18A2, COLGALT2, GALNT15, HOXD11 , DEFA1 B, FAM180A, IGHV7.4.1 , TUBB1 , SCARA3, MUC7, COL5A1 , PTPRZ1 , NKX3.2, and FNDC1. In some embodiments, the method comprises detecting (a). In some embodiments, the method comprises detecting (b). In some embodiments, the method comprises detecting (a) and (b). In some embodiments, the corresponding reference value is the level of gene expression in a population of patients having rheumatoid arthritis that are not responsive to treatment with an agent that downregulates IL-6 mediated signalling. In some embodiments, the method further comprises administering to the patient an effective amount of an agent that downregulates IL- 6 mediated signalling. In some embodiments, the agent that downregulates IL-6 mediated signalling is selected from the group consisting of tocilizumab, sarilumab, satralizumab and siltuximab. In some embodiments, the agent that downregulates IL-6 mediated signalling is tocilizumab.

In one aspect, the invention provides a method of treating rheumatoid arthritis in a patient in need thereof, the method comprising: (i) detecting: (a) a level of gene expression greater than a corresponding reference value of a second biomarker in a biological sample obtained from the patient, wherein the second biomarker is selected from the group consisting of XCR1 , MYH6, PLEKHG6, CD36, CXCL14, SSC5D, AC093525.2, AC005943.1 , SAA2, PTGER3, AC068547.1 , MESP1 , HIST2H2AA3, VMO1 , CDC20, AC135068.9, KCNIP2, and MDFI; and/or (b) a level of gene expression less than a corresponding reference value of a second biomarker in a biological sample obtained from the patient, wherein the second biomarker is selected from the group consisting of SHC3, DLX4, TCN1 , AP001781.2, MUC6, AC009336.2, NPIPA3, SLC18A2, COLGALT2, GALNT15, HOXD11 , DEFA1 B, FAM180A, IGHV7.4.1 , TUBB1 , SCARA3, MUC7, COL5A1 , PTPRZ1 , NKX3.2, and FNDC1 ; and (ii) administering to the patient an effective amount of an agent that downregulates IL-6 mediated signalling. In some embodiments, the agent that downregulates IL-6 mediated signalling is selected from the group consisting of tocilizumab, sarilumab, satralizumab and siltuximab. In some embodiments, the agent that downregulates IL-6 mediated signalling is tocilizumab. In some embodiments, the method comprises detecting (a). In some embodiments, the method comprises detecting (b). In some embodiments, the method comprises detecting (a) and (b). In some embodiments, the corresponding reference value is the level of gene expression in a population of patients having rheumatoid arthritis that are not responsive to treatment with an agent that downregulates IL-6 mediated signalling.

In one aspect, the invention provides a method of treating rheumatoid arthritis in a patient in need thereof comprising administering to the patient an effective amount of an agent that downregulates IL-6 mediated signalling; wherein a biological sample obtained from the patient has: (a) a level of gene expression greater than a corresponding reference value of a second biomarker in a biological sample obtained from the patient, wherein the second biomarker is selected from the group consisting of XCR1 , MYH6, PLEKHG6, CD36, CXCL14, SSC5D, AC093525.2, AC005943.1 , SAA2, PTGER3, AC068547.1 , MESP1 , HIST2H2AA3, VMO1 , CDC20, AC135068.9, KCNIP2, and MDFI; and/or (b) a level of gene expression less than a corresponding reference value of a second biomarker in a biological sample obtained from the patient, wherein the second biomarker is selected from the group consisting of SHC3, DLX4, TCN1 , AP001781.2, MUC6, AC009336.2, NPIPA3, SLC18A2, C0LGALT2, GALNT15, H0XD11 , DEFA1 B, FAM180A, IGHV7.4.1 , TUBB1 , SCARA3, MUC7, C0L5A1 , PTPRZ1 , NKX3.2, and FNDC1. In some embodiments, the agent that downregulates IL-6 mediated signalling is selected from the group consisting of tocilizumab, sarilumab, satralizumab and siltuximab. In some embodiments, the agent that downregulates IL-6 mediated signalling is tocilizumab. In some embodiments, the biological sample has (a). In some embodiments, the biological sample has (b). In some embodiments, the biological sample has (a) and (b). In some embodiments, the corresponding reference value is the level of gene expression in a population of patients having rheumatoid arthritis that are not responsive to treatment with an agent that downregulates IL-6 mediated signalling.

In one aspect, the invention provides a method of detecting a biomarker in a rheumatoid arthritis patient refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling comprising detecting: (a) a level of gene expression greater than a corresponding reference value of a third biomarker in a biological sample obtained from the patient, wherein the third biomarker is selected from the group consisting of PLEKHG6, IGHV7.4.1 , DLX4, NTN1 , TCN1 , TPSD1 , CHAD, WIF1 , BIVM.ERCC5, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5.GP1 BB, FAM69C, CXCL14, CD36, SCD, SAA2, EDIL3, FNDC1 , PRRG3, AP001781.2, PTPRZ1 , MUM1 L1 , MYH6, PTGER3, TUBB1 , LEFTY2, SHC3, ITGA10, PADI4, SLC18A2, TNFRSF11 B, COL11A2, COLGALT2, and PDE4C; and/or (b) a level of gene expression less than a corresponding reference value of a third biomarker in a biological sample obtained from the patient, wherein the third biomarker is selected from the group consisting of AC012184.2, CDC20, AC093525.2, HIST2H2AA3, CCL4L2, AC005943.1 , XCR1 , LGALS2, G0S2, RASD1 , AL139300.1 , AC068547.1 , S100B, PPP1 R1A, AL121900.2, and DUSP2. In some embodiments, the method comprises detecting (a). In some embodiments, the method comprises detecting (b). In some embodiments, the method comprises detecting (a) and (b). In some embodiments, the corresponding reference value is the level of gene expression in a population of patients having rheumatoid arthritis that is refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling.

In one aspect, the invention provides a method for determining whether a Rheumatoid Arthritis (RA) patient is susceptible or refractory to treatment with a B cell targeted therapy and/or an agent that downregulates IL-6 mediated signalling, the method comprising the steps:

(a) (i) determining the level of one or more first biomarker in one or more sample obtained from the patient, wherein the one or more first biomarker is selected from T able 1 ; and (ii) comparing the level of the one or more first biomarker to one or more corresponding reference value; wherein the level of the one or more first biomarker compared to the corresponding reference value is indicative of the susceptibility to treatment with a B cell targeted therapy;

(b) (i) determining the level of one or more second biomarker in one or more sample obtained from the patient, wherein the one or more second biomarker is selected from Table 2; and (ii) comparing the level of the one or more second biomarker to one or more corresponding reference value; wherein the level of the one or more second biomarker compared to the corresponding reference value is indicative of the susceptibility to treatment with an agent that downregulates IL-6 mediated signalling; and/or

(c) (i) determining the level of one or more third biomarker in one or more sample obtained from the patient, wherein the one or more third biomarker is selected from Table 3; and (ii) comparing the level of the one or more third biomarker to one or more corresponding reference value; wherein the level of the one or more third biomarker compared to the corresponding reference value is indicative of the refractoriness to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling.

In another aspect, the invention provides a method for determining whether a Rheumatoid Arthritis (RA) patient is susceptible to treatment with a B cell targeted therapy and/or an agent that downregulates IL-6 mediated signalling, the method comprising the steps:

(a) (i) determining the level of one or more first biomarker in one or more sample obtained from the patient, wherein the one or more first biomarker is selected from T able 1 ; and (ii) comparing the level of the one or more first biomarker to one or more corresponding reference value; wherein the level of the one or more first biomarker compared to the corresponding reference value is indicative of the susceptibility to treatment with a B cell targeted therapy; and/or

(b) (i) determining the level of one or more second biomarker in one or more sample obtained from the patient, wherein the one or more second biomarker is selected from Table 2; and (ii) comparing the level of the one or more second biomarker to one or more corresponding reference value; wherein the level of the one or more second biomarker compared to the corresponding reference value is indicative of the susceptibility to treatment with an agent that downregulates IL-6 mediated signalling.

In another aspect, the invention provides a method for determining whether a Rheumatoid Arthritis (RA) patient is refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling, the method comprising the steps: (i) determining the level of one or more third biomarker in one or more sample obtained from the patient, wherein the one or more third biomarker is selected from Table 3; and (ii) comparing the level of the one or more third biomarker to one or more corresponding reference value; wherein the level of the one or more third biomarker compared to the corresponding reference value is indicative of the refractoriness to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling.

In another aspect, the invention provides a method for determining whether a Rheumatoid Arthritis (RA) patient is susceptible to treatment with a B cell targeted therapy, the method comprising the steps: (i) determining the level of one or more first biomarker in one or more sample obtained from the patient, wherein the one or more first biomarker is selected from Table 1 ; and (ii) comparing the level of the one or more first biomarker to one or more corresponding reference value; wherein the level of the one or more first biomarker compared to the corresponding reference value is indicative of the susceptibility to treatment with a B cell targeted therapy.

In some embodiments, the one or more first biomarker comprises 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 , 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 , 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 , 32, 33, 34, 35, 36, 37, 38, 39, or all 40 biomarkers from Table 1.

In some embodiments, the one or more first biomarker comprises at least 20 (for example at least 21 , at least 22, at least 23, at least 24, or preferably at least 25) biomarkers from Table 1.

In some embodiments, the one or more first biomarker comprises all 40 biomarkers from Table 1.

In some embodiments, the one or more first biomarker consists of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 , 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 , 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 , 32, 33, 34, 35, 36, 37, 38, 39, or all 40 biomarkers from Table 1. In some embodiments, the one or more first biomarker consists of at least 20 (for example at least 21 , at least 22, at least 23, at least 24, or preferably at least 25) biomarkers from Table 1.

In some embodiments, the one or more first biomarker consists of all 40 biomarkers from Table 1.

In some embodiments, the one or more first biomarker comprises or consists of a biomarker selected from the group consisting of: SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1 , RARRES2, NKX3-2, BAIAP3, FNDC1 , WIF1 , DEFA1 B, HIST2H2AA3, CXCL2, SAA2, AC068547.1 , CDON, IGHV7-4-1 , DKK3, NOG, P116, C6orf58, KCNIP2, EIF3CL, ITGA10, MAL2, MDFI, STAC3, and FIBIN.

In some embodiments, the one or more first biomarker comprises or consists of SHC3. In some embodiments, the one or more first biomarker comprises or consists of XCR1. In some embodiments, the one or more first biomarker comprises or consists of TCN1. In some embodiments, the one or more first biomarker comprises or consists of DLX4. In some embodiments, the one or more first biomarker comprises or consists of PLEKHG6. In some embodiments, the one or more first biomarker comprises or consists of COLGALT2. In some embodiments, the one or more first biomarker comprises or consists of ERICH3. In some embodiments, the one or more first biomarker comprises or consists of MLXIPL. In some embodiments, the one or more first biomarker comprises or consists of MLIC6. In some embodiments, the one or more first biomarker comprises or consists of TBC1 D3. In some embodiments, the one or more first biomarker comprises or consists of MYH6. In some embodiments, the one or more first biomarker comprises or consists of CXCL14. In some embodiments, the one or more first biomarker comprises or consists of AC009336.2. In some embodiments, the one or more first biomarker comprises or consists of RELN. In some embodiments, the one or more first biomarker comprises or consists of NPIPA3. In some embodiments, the one or more first biomarker comprises or consists of AC093525.2. In some embodiments, the one or more first biomarker comprises or consists of MESP1. In some embodiments, the one or more first biomarker comprises or consists of RARRES2. In some embodiments, the one or more first biomarker comprises or consists of NKX3-2. In some embodiments, the one or more first biomarker comprises or consists of BAIAP3. In some embodiments, the one or more first biomarker comprises or consists of FNDC1. In some embodiments, the one or more first biomarker comprises or consists of WIF1. In some embodiments, the one or more first biomarker comprises or consists of DEFA1 B. In some embodiments, the one or more first biomarker comprises or consists of HIST2H2AA3. In some embodiments, the one or more first biomarker comprises or consists of CXCL2. In some embodiments, the one or more first biomarker comprises or consists of SAA2. In some embodiments, the one or more first biomarker comprises or consists of AC068547.1 . In some embodiments, the one or more first biomarker comprises or consists of CDON. In some embodiments, the one or more first biomarker comprises or consists of IGHV7-4-1. In some embodiments, the one or more first biomarker comprises or consists of DKK3. In some embodiments, the one or more first biomarker comprises or consists of NOG. In some embodiments, the one or more first biomarker comprises or consists of PI16. In some embodiments, the one or more first biomarker comprises or consists of C6orf58. In some embodiments, the one or more first biomarker comprises or consists of KCNIP2. In some embodiments, the one or more first biomarker comprises or consists of EIF3CL. In some embodiments, the one or more first biomarker comprises or consists of ITGA10. In some embodiments, the one or more first biomarker comprises or consists of MAL2. In some embodiments, the one or more first biomarker comprises or consists of MDFI. In some embodiments, the one or more first biomarker comprises or consists of STAC3. In some embodiments, the one or more first biomarker comprises or consists of FIBIN.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, and XCRI .

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , and TCN1.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , and DLX4.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, and PLEKHG6.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, and COLGALT2.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, and ERICH3.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, and MLXIPL.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, and MUC6. In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, and TBC1 D3.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, and MYH6.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, and CXCL14.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, and AC009336.2.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, and RELN.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, and NPIPA3.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, and AC093525.2.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, and MESP1.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1 , and RARRES2.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1 , RARRES2, and NKX3- 2. In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1 , RARRES2, NKX3-2, and BAIAP3.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1 , RARRES2, NKX3-2, BAIAP3, and FNDC1.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1 , RARRES2, NKX3-2, BAIAP3, FNDC1 , and WIF1.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1 , RARRES2, NKX3-2, BAIAP3, FNDC1 , WIF1 , and DEFA1 B.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1 , RARRES2, NKX3-2, BAIAP3, FNDC1 , WIF1 , DEFA1 B, and HIST2H2AA3.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1 , RARRES2, NKX3-2, BAIAP3, FNDC1 , WIF1 , DEFA1 B, HIST2H2AA3, and CXCL2.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1 , RARRES2, NKX3-2, BAIAP3, FNDC1 , WIF1 , DEFA1 B, HIST2H2AA3, CXCL2, and SAA2.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1 D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1 , RARRES2, NKX3-2, BAIAP3, FNDC1 , WIF1 , DEFA1 B, HIST2H2AA3, CXCL2, SAA2, and AC068547.1. In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, and CDON.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, CDON, and IGHV7-4-1.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, CDON, IGHV7-4-1, and DKK3.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, CDON, IGHV7-4-1, DKK3, and NOG.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, CDON, IGHV7-4-1, DKK3, NOG, and PI16.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, CDON, IGHV7-4-1, DKK3, NOG, P116, and C6orf58.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, CDON, IGHV7-4-1, DKK3, NOG, PI16, C6orf58, and KCNIP2.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, CDON, IGHV7-4-1, DKK3, NOG, PI16, C6orf58, KCNIP2, and EIF3CL.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, CDON, IGHV7-4-1, DKK3, NOG, PI16, C6orf58, KCNIP2, EIF3CL, and ITGA10.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, CDON, IGHV7-4-1, DKK3, NOG, PI16, C6orf58, KCNIP2, EIF3CL, ITGA10, and MAL2.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, CDON, IGHV7-4-1, DKK3, NOG, PI16, C6orf58, KCNIP2, EIF3CL, ITGA10, MAL2, and MDFI.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, CDON, IGHV7-4-1, DKK3, NOG, PI16, C6orf58, KCNIP2, EIF3CL, ITGA10, MAL2, MDFI, and STAC3.

In some embodiments, the one or more first biomarker comprises or consists of the biomarkers SHC3, XCR1, TCN1, DLX4, PLEKHG6, COLGALT2, ERICH3, MLXIPL, MUC6, TBC1D3, MYH6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, MESP1, RARRES2, NKX3-2, BAIAP3, FNDC1, WIF1, DEFA1B, HIST2H2AA3, CXCL2, SAA2, AC068547.1, CDON, IGHV7-4-1 , DKK3, NOG, PI16, C6orf58, KCNIP2, EIF3CL, ITGA10, MAL2, MDFI, STAC3 and FIBIN.

In another aspect, the invention provides a method for determining whether a Rheumatoid Arthritis (RA) patient is susceptible to treatment with an agent that downregulates IL-6 mediated signalling, the method comprising the steps: (i) determining the level of one or more second biomarker in one or more sample obtained from the patient, wherein the one or more second biomarker is selected from Table 2; and (ii) comparing the level of the one or more second biomarker to one or more corresponding reference value; wherein the level of the one or more second biomarker compared to the corresponding reference value is indicative of the susceptibility to treatment with an agent that downregulates IL-6 mediated signalling.

In some embodiments, the one or more second biomarker comprises 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 , 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 , 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 , 32, 33, 34, 35, 36, 37, 38, or all 39 biomarkers from Table 2.

In some embodiments, the one or more second biomarker comprises at least 20 (for example at least 21 , at least 22, at least 23, at least 24, or preferably at least 25) biomarkers from Table 2.

In some embodiments, the one or more second biomarker comprises all 39 biomarkers from Table 2.

In some embodiments, the one or more second biomarker consists of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 , 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 , 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 , 32, 33, 34, 35, 36, 37, 38, or all 39 biomarkers from Table 2.

In some embodiments, the one or more second biomarker consists of at least 20 (for example at least 21 , at least 22, at least 23, at least 24, or preferably at least 25) biomarkers from Table 2.

In some embodiments, the one or more second biomarker consists of all 39 biomarkers from Table 2.

In some embodiments, the one or more second biomarker comprises or consists of a biomarker selected from the group consisting of: SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1 , HOXD11 , SAA2, PTGER3, DEFA1 B, AC068547.1 , MESP1 , FAM180A, IGHV7-4-1 , TUBB1 , SCARA3, HIST2H2AA3, MUC7, COL5A1 , VMO1 , PTPRZ1 , CDC20, NKX3-2, AC135068.9, KCNIP2, MDFI, and FNDC1. In some embodiments, the one or more second biomarker comprises or consists of SHC3. In some embodiments, the one or more second biomarker comprises or consists of XCR1. In some embodiments, the one or more second biomarker comprises or consists of DLX4. In some embodiments, the one or more second biomarker comprises or consists of MYH6. In some embodiments, the one or more second biomarker comprises or consists of TCN1. In some embodiments, the one or more second biomarker comprises or consists of In some embodiments, the one or more second biomarker comprises or consists of PLEKHG6. In some embodiments, the one or more second biomarker comprises or consists of AP001781.2. In some embodiments, the one or more second biomarker comprises or consists of MLIC6. In some embodiments, the one or more second biomarker comprises or consists of AC009336.2. In some embodiments, the one or more second biomarker comprises or consists of CD36. In some embodiments, the one or more second biomarker comprises or consists of NPIPA3. In some embodiments, the one or more second biomarker comprises or consists of CXCL14. In some embodiments, the one or more second biomarker comprises or consists of SSC5D. In some embodiments, the one or more second biomarker comprises or consists of SLC18A2. In some embodiments, the one or more second biomarker comprises or consists of COLGALT2. In some embodiments, the one or more second biomarker comprises or consists of AC093525.2. In some embodiments, the one or more second biomarker comprises or consists of GALNT15. In some embodiments, the one or more second biomarker comprises or consists of AC005943.1. In some embodiments, the one or more second biomarker comprises or consists of HOXD11. In some embodiments, the one or more second biomarker comprises or consists of SAA2. In some embodiments, the one or more second biomarker comprises or consists of PTGER3. In some embodiments, the one or more second biomarker comprises or consists of DEFA1 B. In some embodiments, the one or more second biomarker comprises or consists of AC068547.1. In some embodiments, the one or more second biomarker comprises or consists of MESP1. In some embodiments, the one or more second biomarker comprises or consists of FAM180A. In some embodiments, the one or more second biomarker comprises or consists of IGHV7-4-1. In some embodiments, the one or more second biomarker comprises or consists of TLIBB1. In some embodiments, the one or more second biomarker comprises or consists of SCARA3. In some embodiments, the one or more second biomarker comprises or consists of HIST2H2AA3. In some embodiments, the one or more second biomarker comprises or consists of MLIC7. In some embodiments, the one or more second biomarker comprises or consists of COL5A1. In some embodiments, the one or more second biomarker comprises or consists of VMO1. In some embodiments, the one or more second biomarker comprises or consists of PTPRZ1. In some embodiments, the one or more second biomarker comprises or consists of CDC20. In some embodiments, the one or more second biomarker comprises or consists of NKX3-2. In some embodiments, the one or more second biomarker comprises or consists of AC135068.9. In some embodiments, the one or more second biomarker comprises or consists of KCNIP2. In some embodiments, the one or more second biomarker comprises or consists of MDFI. In some embodiments, the one or more second biomarker comprises or consists of FNDC1 .

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, and XCR1.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , and DLX4.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, and MYH6.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, and TCN1.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , and PLEKHG6.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, and AP001781.2.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, and MUC6.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, and AC009336.2.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, and CD36.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, and NPIPA3.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, and CXCL14. In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, and SSC5D.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, and SLC18A2.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, and COLGALT2.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, and AC093525.2.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, and GALNT15.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, and AC005943.1.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1 , and HOXD11.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1 , HOXD11 , and SAA2.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1 , HOXD11 , SAA2, and PTGER3. In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1 , HOXD11 , SAA2, PTGER3, and DEFA1B.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1 , HOXD11 , SAA2, PTGER3, DEFA1B, and AC068547.1.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1 , HOXD11 , SAA2, PTGER3, DEFA1B, AC068547.1 , and MESP1.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11 , SAA2, PTGER3, DEFA1B, AC068547.1 , MESP1 , and FAM180A.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11 , SAA2, PTGER3, DEFA1B, AC068547.1 , MESP1 , FAM180A, and IGHV7-4-1.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11 , SAA2, PTGER3, DEFA1B, AC068547.1 , MESP1 , FAM180A, IGHV7-4-1 , and TUBB1.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11 , SAA2, PTGER3, DEFA1B, AC068547.1 , MESP1 , FAM180A, IGHV7-4-1 , TUBB1 , and SCARA3. In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11 , SAA2, PTGER3, DEFA1B, AC068547.1 , MESP1 , FAM180A, IGHV7-4-1 , TUBB1 , SCARA3, and HIST2H2AA3.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11 , SAA2, PTGER3, DEFA1B, AC068547.1 , MESP1 , FAM180A, IGHV7-4-1 , TUBB1 , SCARA3, HIST2H2AA3, and MUC7.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11 , SAA2, PTGER3, DEFA1B, AC068547.1 , MESP1 , FAM180A, IGHV7-4-1 , TUBB1 , SCARA3, HIST2H2AA3, MUC7, and COL5A1.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11 , SAA2, PTGER3, DEFA1B, AC068547.1 , MESP1 , FAM180A, IGHV7-4-1 , TUBB1 , SCARA3, HIST2H2AA3, MUC7, COL5A1, and VMO1.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11 , SAA2, PTGER3, DEFA1B, AC068547.1 , MESP1 , FAM180A, IGHV7-4-1 , TUBB1 , SCARA3, HIST2H2AA3, MUC7, COL5A1 , VMO1 , and PTPRZ1.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1 , DLX4, MYH6, TCN1 , PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11 , SAA2, PTGER3, DEFA1B, AC068547.1 , MESP1 , FAM180A, IGHV7-4-1 , TUBB1 , SCARA3, HIST2H2AA3, MUC7, COL5A1, VMO1 , PTPRZ1 , and CDC20. In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1, DLX4, MYH6, TCN1, PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11, SAA2, PTGER3, DEFA1B, AC068547.1, MESP1, FAM180A, IGHV7-4-1, TUBB1, SCARA3, HIST2H2AA3, MUC7, COL5A1, VMO1, PTPRZ1, CDC20, and NKX3-2.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1, DLX4, MYH6, TCN1, PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11, SAA2, PTGER3, DEFA1B, AC068547.1, MESP1, FAM180A, IGHV7-4-1, TUBB1, SCARA3, HIST2H2AA3, MUC7, COL5A1, VMO1, PTPRZ1, CDC20, NKX3-2, and AC135068.9.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1, DLX4, MYH6, TCN1, PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11, SAA2, PTGER3, DEFA1B, AC068547.1, MESP1, FAM180A, IGHV7-4-1, TUBB1, SCARA3, HIST2H2AA3, MUC7, COL5A1, VMO1, PTPRZ1, CDC20, NKX3-2, AC135068.9, and KCNIP2.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1, DLX4, MYH6, TCN1, PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11, SAA2, PTGER3, DEFA1B, AC068547.1, MESP1, FAM180A, IGHV7-4-1, TUBB1, SCARA3, HIST2H2AA3, MUC7, COL5A1, VMO1, PTPRZ1, CDC20, NKX3-2, AC135068.9, KCNIP2, and MDFI.

In some embodiments, the one or more second biomarker comprises or consists of the biomarkers SHC3, XCR1, DLX4, MYH6, TCN1, PLEKHG6, AP001781.2, MUC6, AC009336.2, CD36, NPIPA3, CXCL14, SSC5D, SLC18A2, COLGALT2, AC093525.2, GALNT15, AC005943.1, HOXD11, SAA2, PTGER3, DEFA1B, AC068547.1, MESP1, FAM180A, IGHV7-4-1, TUBB1, SCARA3, HIST2H2AA3, MUC7, COL5A1, VMO1, PTPRZ1, CDC20, NKX3-2, AC135068.9, KCNIP2, MDFI, and FNDC1.

In some embodiments, the one or more third biomarker comprises 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 , 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 , 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 , 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 , 42, 43, 44, 45, 46, 47, 48, 49, 50, 51 , 52, or all 53 biomarkers from Table 3.

In some embodiments, the one or more third biomarker comprises at least 20 (for example at least 21 , at least 22, at least 23, at least 24, or preferably at least 25) biomarkers from Table 3.

In some embodiments, the one or more third biomarker comprises all 53 biomarkers from Table 3.

In some embodiments, the one or more third biomarker consists of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 , 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 , 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 , 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 , 42, 43, 44, 45, 46, 47, 48, 49, 50, 51 , 52, or all 53 biomarkers from Table 3.

In some embodiments, the one or more third biomarker consists of at least 20 (for example at least 21 , at least 22, at least 23, at least 24, or preferably at least 25) biomarkers from Table 3.

In some embodiments, the one or more third biomarker consists of all 53 biomarkers from Table 3.

In some embodiments, the one or more third biomarker comprises or consists of a biomarker selected from the group consisting of: AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7- 4-1 , DLX4, NTN1 , HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, AC005943.1 , WIF1 , BIVM- ERCC5, XCR1 , LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1 BB, FAM69C, G0S2, RASD1 , CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1 , FNDC1 , PRRG3, AC068547.1 , S100B, AP001781.2, PTPRZ1 , MUM1 L1 , MYH6, PTGER3, TUBB1 , LEFTY2, SHC3, ITGA10, PPP1 R1A, PADI4, AL121900.2, SLC18A2, DUSP2, TNFRSF11 B, COL11A2, COLGALT2, and PDE4C. In some embodiments, the one or more third biomarker comprises or consists of AC012184.2. In some embodiments, the one or more third biomarker comprises or consists of CDC20. In some embodiments, the one or more third biomarker comprises or consists of AC093525.2. In some embodiments, the one or more third biomarker comprises or consists of PLEKHG6. In some embodiments, the one or more third biomarker comprises or consists of IGHV7-4-1. In some embodiments, the one or more third biomarker comprises or consists of DLX4. In some embodiments, the one or more third biomarker comprises or consists of NTN1 . In some embodiments, the one or more third biomarker comprises or consists of HIST2H2AA3. In some embodiments, the one or more third biomarker comprises or consists of TCN1 . In some embodiments, the one or more third biomarker comprises or consists of TPSD1. In some embodiments, the one or more third biomarker comprises or consists of CHAD. In some embodiments, the one or more third biomarker comprises or consists of CCL4L2. In some embodiments, the one or more third biomarker comprises or consists of AC005943.1. In some embodiments, the one or more third biomarker comprises or consists of WIF1. In some embodiments, the one or more third biomarker comprises or consists of BIVM-ERCC5. In some embodiments, the one or more third biomarker comprises or consists of XCR1. In some embodiments, the one or more third biomarker comprises or consists of LGALS2. In some embodiments, the one or more third biomarker comprises or consists of ITGA2B. In some embodiments, the one or more third biomarker comprises or consists of EMILIN3. In some embodiments, the one or more third biomarker comprises or consists of RSPO2. In some embodiments, the one or more third biomarker comprises or consists of MLIC6. In some embodiments, the one or more third biomarker comprises or consists of SEPT5-GP1 BB. In some embodiments, the one or more third biomarker comprises or consists of FAM69C. In some embodiments, the one or more third biomarker comprises or consists of G0S2. In some embodiments, the one or more third biomarker comprises or consists of RASD1. In some embodiments, the one or more third biomarker comprises or consists of CXCL14. In some embodiments, the one or more third biomarker comprises or consists of CD36. In some embodiments, the one or more third biomarker comprises or consists of SCD. In some embodiments, the one or more third biomarker comprises or consists of SAA2. In some embodiments, the one or more third biomarker comprises or consists of EDIL3. In some embodiments, the one or more third biomarker comprises or consists of AL139300.1. In some embodiments, the one or more third biomarker comprises or consists of FNDC1. In some embodiments, the one or more third biomarker comprises or consists of PRRG3. In some embodiments, the one or more third biomarker comprises or consists of AC068547.1. In some embodiments, the one or more third biomarker comprises or consists of S100B. In some embodiments, the one or more third biomarker comprises or consists of AP001781.2. In some embodiments, the one or more third biomarker comprises or consists of PTPRZ1. In some embodiments, the one or more third biomarker comprises or consists of MLIM1 L1. In some embodiments, the one or more third biomarker comprises or consists of MYH6. In some embodiments, the one or more third biomarker comprises or consists of PTGER3. In some embodiments, the one or more third biomarker comprises or consists of TLIBB1. In some embodiments, the one or more third biomarker comprises or consists of LEFTY2. In some embodiments, the one or more third biomarker comprises or consists of SHC3. In some embodiments, the one or more third biomarker comprises or consists of ITGA10. In some embodiments, the one or more third biomarker comprises or consists of PPP1 R1A. In some embodiments, the one or more third biomarker comprises or consists of PADI4. In some embodiments, the one or more third biomarker comprises or consists of AL121900.2. In some embodiments, the one or more third biomarker comprises or consists of SLC18A2. In some embodiments, the one or more third biomarker comprises or consists of DLISP2. In some embodiments, the one or more third biomarker comprises or consists of TNFRSF11 B. In some embodiments, the one or more third biomarker comprises or consists of COL11A2. In some embodiments, the one or more third biomarker comprises or consists of COLGALT2. In some embodiments, the one or more third biomarker comprises or consists of PDE4C.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, and CDC20.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, and AC093525.2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, and PLEKHG6.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, and IGHV7-4-1.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , and DLX4.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, and NTN1.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1 , and HIST2H2AA3. In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, and TCN1.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , and TPSD1.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , and CHAD.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, and CCL4L2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, and AC005943.1.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, AC005943.1 , and WIF1.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, AC005943.1 , WIF1 , and BIVM-ERCC5.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, and XCR1.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, AC005943.1 , WIF1 , BIVM-ERCC5, XCR1 , and LGALS2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, AC005943.1 , WIF1 , BIVM-ERCC5, XCR1, LGALS2, and ITGA2B.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, AC005943.1 , WIF1 , BIVM-ERCC5, XCR1 , LGALS2, ITGA2B, and EMILIN3.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, AC005943.1 , WIF1 , BIVM-ERCC5, XCR1 , LGALS2, ITGA2B, EMILIN3, and RSPO2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, AC005943.1 , WIF1 , BIVM-ERCC5, XCR1 , LGALS2, ITGA2B, EMILIN3, RSPO2, and MUC6.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, AC005943.1 , WIF1 , BIVM-ERCC5, XCR1 , LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, and SEPT5-GP1 BB.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, AC005943.1 , WIF1 , BIVM-ERCC5, XCR1 , LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, and FAM69C.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, AC005943.1 , WIF1 , BIVM-ERCC5, XCR1 , LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, and G0S2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1 , TPSD1 , CHAD, CCL4L2, AC005943.1 , WIF1 , BIVM-ERCC5, XCR1 , LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, and RASD1.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSP02, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, and CXCL14.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, and CD36.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, and SCD.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, and SAA2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, and EDIL3.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, and AL139300.1.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, and FNDC1. In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, and PRRG3.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, and AC068547.1.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, and S100B.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, and AP001781.2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, and PTPRZ1.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, and MLIM1L1. In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, and MYH6.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, and PTGER3.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, and TUBB1.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, TUBB1, and LEFTY2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, TUBB1, LEFTY2, and SHC3.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B,

AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, TUBB1, LEFTY2, SHC3, and ITGA10.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, TUBB1, LEFTY2, SHC3, ITGA10, and PPP1R1A.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, TUBB1, LEFTY2, SHC3, ITGA10, PPP1R1A, and PADI4.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, TUBB1, LEFTY2, SHC3, ITGA10, PPP1R1A, PADI4, and AL121900.2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, TUBB1, LEFTY2, SHC3, ITGA10, PPP1R1A, PADI4, AL121900.2, and SLC18A2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, TUBB1, LEFTY2, SHC3, ITGA10, PPP1R1A, PADI4, AL121900.2, SLC18A2, and DUSP2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, TUBB1, LEFTY2, SHC3, ITGA10, PPP1R1A, PADI4, AL121900.2, SLC18A2, DUSP2, and TNFRSF11B.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, TUBB1, LEFTY2, SHC3, ITGA10, PPP1R1A, PADI4, AL121900.2, SLC18A2, DUSP2, TNFRSF11B, and COL11A2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, TUBB1, LEFTY2, SHC3, ITGA10, PPP1R1A, PADI4, AL121900.2, SLC18A2, DUSP2, TNFRSF11B, COL11A2, and COLGALT2.

In some embodiments, the one or more third biomarker comprises or consists of the biomarkers AC012184.2, CDC20, AC093525.2, PLEKHG6, IGHV7-4-1, DLX4, NTN1, HIST2H2AA3, TCN1, TPSD1, CHAD, CCL4L2, AC005943.1, WIF1, BIVM-ERCC5, XCR1, LGALS2, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5-GP1BB, FAM69C, G0S2, RASD1, CXCL14, CD36, SCD, SAA2, EDIL3, AL139300.1, FNDC1, PRRG3, AC068547.1, S100B, AP001781.2, PTPRZ1, MUM1L1, MYH6, PTGER3, TUBB1, LEFTY2, SHC3, ITGA10, PPP1R1A, PADI4, AL121900.2, SLC18A2, DUSP2, TNFRSF11B, COL11A2, COLGALT2, and PDE4C. In some embodiments, the one or more biomarker (e.g. first and/or second biomarker) comprises or consists of a biomarker selected from the group consisting of: SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, MUC6, MYH6, CXCL14, AC009336.2, NPIPA3, AC093525.2, MESP1 , NKX3-2, FNDC1 , DEFA1 B, HIST2H2AA3, SAA2, AC068547.1 , IGHV7- 4-1 , KCNIP2, and MDFI.

In some embodiments, the one or more biomarker (e.g. first and/or second biomarker) comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, MUC6, MYH6, CXCL14, AC009336.2, NPIPA3, AC093525.2, MESP1 , NKX3-2, FNDC1 , DEFA1 B, HIST2H2AA3, SAA2, AC068547.1 , IGHV7-4-1 , KCNIP2, and MDFI.

In some embodiments, the one or more biomarker (e.g. first and/or third biomarker) comprises or consists of a biomarker selected from the group consisting of: AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, HIST2H2AA3, TCN1 , WIF1 , XCR1 , MUC6, CXCL14, SAA2, FNDC1 , AC068547.1 , MYH6, SHC3, ITGA10, and COLGALT2.

In some embodiments, the one or more biomarker (e.g. first and/or third biomarker) comprises or consists of the biomarkers AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, HIST2H2AA3, TCN1 , WIF1 , XCR1 , MUC6, CXCL14, SAA2, FNDC1 , AC068547.1 , MYH6, SHC3, ITGA10, and COLGALT2.

In some embodiments, the one or more biomarker (e.g. second and/or third biomarker) comprises or consists of a biomarker selected from the group consisting of: CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, HIST2H2AA3, TCN1 , AC005943.1 , XCR1 , MUC6, CXCL14, CD36, SAA2, FNDC1 , AC068547.1 , AP001781.2, PTPRZ1 , MYH6, PTGER3, TUBB1 , SHC3, SLC18A2, and COLGALT2.

In some embodiments, the one or more biomarker (e.g. second and/or third biomarker) comprises or consists of the biomarkers CDC20, AC093525.2, PLEKHG6, IGHV7-4-1 , DLX4, HIST2H2AA3, TCN1 , AC005943.1 , XCR1 , MUC6, CXCL14, CD36, SAA2, FNDC1 , AC068547.1 , AP001781.2, PTPRZ1 , MYH6, PTGER3, TUBB1 , SHC3, SLC18A2, and COLGALT2.

In some embodiments, the one or more biomarker (e.g. first, second and/or third biomarker) comprises or consists of a biomarker selected from the group consisting of: SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, MUC6, MYH6, CXCL14, AC093525.2, FNDC1 , HIST2H2AA3, SAA2, AC068547.1 , and IGHV7-4-1. In some embodiments, the one or more biomarker (e.g. first, second and/or third biomarker) comprises or consists of the biomarkers SHC3, XCR1 , TCN1 , DLX4, PLEKHG6, COLGALT2, MUC6, MYH6, CXCL14, AC093525.2, FNDC1 , HIST2H2AA3, SAA2, AC068547.1 , and IGHV7-4-1.

In some embodiments, the level of the one or more first, second and/or third biomarker is a nucleic acid level. In some embodiments, the nucleic acid level is an mRNA level.

In some embodiments, the step of determining the level of one or more first, second and/or third biomarker is performed by direct digital counting of nucleic acids, RNA-seq, RT-qPCR, qPCR, multiplex qPCR or RT-qPCR, microarray analysis, or a combination thereof.

In some embodiments, the step of determining the level of one or more first, second and/or third biomarker is performed by RNA sequencing.

In some embodiments, the step of determining the level of the one or more first, second and/or third biomarker comprises determining the level of gene expression of the one or more first, second and/or third biomarker.

In some embodiments, the one or more sample is a synovial sample. In some embodiments, the sample is a synovial tissue sample or a synovial fluid sample.

In some embodiments, the sample is obtained by synovial biopsy, preferably ultrasound- guided synovial biopsy. In some embodiments, the synovial biopsy is obtained by an arthroscopic procedure.

In some embodiments, the level of one or more biomarker is determined in 2, 3, 4, 5, 6, 7, 8 or more samples obtained from the patient. In some embodiments, the level of one or more biomarker is determined in 6, 7, 8 or more samples obtained from the patient. In some embodiments, the level of one or more biomarker is determined in 6-8 samples obtained from the patient. In some embodiments, the level of one or more biomarker is determined in 6 samples obtained from the patient. In some embodiments, the samples obtained from the patient are pooled before determination of the level of one or more biomarker.

In some embodiments, (i) when the level of the one or more first biomarker is greater than the corresponding reference value the patient is determined to be susceptible to treatment with the B cell targeted therapy; and/or (ii) when the level of the one or more first biomarker is less than the corresponding reference value the patient is determined to be resistant to treatment with a B cell targeted therapy. In some embodiments, when the level of the one or more first biomarker selected from the group consisting of XCR1 , TCN1 , PLEKHG6, TBC1 D3, MYH6, MESP1 , RARRES2, HIST2H2AA3, CXCL2, SAA2, AC068547.1 , PI16, C6orf58, KCNIP2, EIF3CL, MAL2, MDFI and STAC3 is greater than the corresponding reference value the patient is determined to be susceptible to treatment with the B cell targeted therapy.

In some embodiments, when the level of the one or more first biomarker selected from the group consisting of SHC3, DLX4, COLGALT2, ERICH3, MLXIPL, MUC6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, NKX3.2, BAIAP3, FNDC1 , WIF1 , DEFA1 B, CDON, IGHV7.4.1 , DKK3, NOG, ITGA10 and FIBIN is less than the corresponding reference value the patient is determined to be susceptible to treatment with the B cell targeted therapy.

In some embodiments, the patient is determined to be susceptible to treatment with the B cell targeted therapy when: (a) the level of the one or more first biomarker selected from the group consisting of XCR1 , TCN1 , PLEKHG6, TBC1 D3, MYH6, MESP1 , RARRES2, HIST2H2AA3, CXCL2, SAA2, AC068547.1 , PI16, C6orf58, KCNIP2, EIF3CL, MAL2, MDFI and STAC3 is greater than the corresponding reference value, and (b) the level of the one or more first biomarker selected from the group consisting of SHC3, DLX4, COLGALT2, ERICH3, MLXIPL, MUC6, CXCL14, AC009336.2, RELN, NPIPA3, AC093525.2, NKX3.2, BAIAP3, FNDC1 , WIF1 , DEFA1 B, CDON, IGHV7.4.1 , DKK3, NOG, ITGA10 and FIBIN is less than the corresponding reference value.

In some embodiments, (i) when the level of the one or more second biomarker is greater than the corresponding reference value the patient is determined to be susceptible to treatment with the agent that downregulates IL-6 mediated signalling; and/or (ii) when the level of the one or more second biomarker is less than the corresponding reference value the patient is determined to be resistant to treatment with an agent that downregulates IL-6 mediated signalling.

In some embodiments, when the level of the one or more second biomarker selected from the group consisting of XCR1 , MYH6, PLEKHG6, CD36, CXCL14, SSC5D, AC093525.2, AC005943.1 , SAA2, PTGER3, AC068547.1 , MESP1 , HIST2H2AA3, VMO1 , CDC20, AC135068.9, KCNIP2 and MDFI is greater than the corresponding reference value the patient is determined to be susceptible to treatment with the agent that downregulates IL-6 mediated signalling.

In some embodiments, when the level of the one or more second biomarker selected from the group consisting of SHC3, DLX4, TCN1 , AP001781.2, MUC6, AC009336.2, NPIPA3, SLC18A2, COLGALT2, GALNT15, HOXD11 , DEFA1 B, FAM180A, IGHV7.4.1 , TUBB1 , SCARA3, MUC7, C0L5A1 , PTPRZ1, NKX3.2 and FNDC1 is less than the corresponding reference value the patient is determined to be susceptible to treatment with the agent that downregulates IL-6 mediated signalling.

In some embodiments, the patient is determined to be susceptible to treatment with the agent that downregulates IL-6 mediated signalling when (a) the level of the one or more second biomarker selected from the group consisting of XCR1 , MYH6, PLEKHG6, CD36, CXCL14, SSC5D, AC093525.2, AC005943.1 , SAA2, PTGER3, AC068547.1 , MESP1 , HIST2H2AA3, VMO1 , CDC20, AC135068.9, KCNIP2 and MDFI is greater than the corresponding reference value, and (b) the level of the one or more second biomarker selected from the group consisting of SHC3, DLX4, TCN1 , AP001781.2, MUC6, AC009336.2, NPIPA3, SLC18A2, COLGALT2, GALNT15, HOXD11 , DEFA1 B, FAM180A, IGHV7.4.1 , TUBB1 , SCARA3, MUC7, COL5A1 , PTPRZ1 , NKX3.2 and FNDC1 is less than the corresponding reference value.

In some embodiments, when the level of the one or more third biomarker is greater than the corresponding reference value the patient is determined to be refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling.

In some embodiments, when the level of the one or more third biomarker selected from the group consisting of PLEKHG6, IGHV7.4.1 , DLX4, NTN1 , TCN1 , TPSD1 , CHAD, WIF1 , BIVM.ERCC5, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5.GP1 BB, FAM69C, CXCL14, CD36, SCD, SAA2, EDIL3, FNDC1 , PRRG3, AP001781.2, PTPRZ1 , MUM1 L1 , MYH6, PTGER3, TUBB1 , LEFTY2, SHC3, ITGA10, PADI4, SLC18A2, TNFRSF11 B, COL11A2, COLGALT2 and PDE4C is greater than the corresponding reference value the patient is determined to be refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling.

In some embodiments, when the level of the one or more third biomarker selected from the group consisting of AC012184.2, CDC20, AC093525.2, HIST2H2AA3, CCL4L2, AC005943.1 , XCR1 , LGALS2, G0S2, RASD1 , AL139300.1 , AC068547.1 , S100B, PPP1 R1A, AL121900.2 and DUSP2 is less than the corresponding reference value the patient is determined to be refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling.

In some embodiments, the patient is determined to be refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling when (a) the level of the one or more third biomarker selected from the group consisting of PLEKHG6, IGHV7.4.1 , DLX4, NTN1 , TCN1 , TPSD1 , CHAD, WIF1 , BIVM.ERCC5, ITGA2B, EMILIN3, RSPO2, MUC6, SEPT5.GP1 BB, FAM69C, CXCL14, CD36, SCD, SAA2, EDIL3, FNDC1 , PRRG3, AP001781.2, PTPRZ1 , MUM1 L1 , MYH6, PTGER3, TUBB1 , LEFTY2, SHC3, ITGA10, PADI4, SLC18A2, TNFRSF11 B, C0L11A2, C0LGALT2 and PDE4C is greater than the corresponding reference value, and (b) the level of the one or more third biomarker selected from the group consisting of AC012184.2, CDC20, AC093525.2, HIST2H2AA3, CCL4L2, AC005943.1 , XCR1 , LGALS2, G0S2, RASD1 , AL139300.1 , AC068547.1 , S100B, PPP1 R1A, AL121900.2 and DLISP2 is less than the corresponding reference value.

In some embodiments, the level of the one or more first biomarker compared to the corresponding reference value classifies the sample as B cell rich or B cell poor.

In some embodiments, (a) when the sample is B cell rich the patient is determined to be susceptible to treatment with a B cell targeted therapy; and/or (b) when the sample is B cell poor the patient is determined to be resistant to treatment with a B cell targeted therapy.

In some embodiments, the B cell targeted therapy is B cell depletion therapy.

In some embodiments, the B cell targeted therapy is selected from the group consisting of: rituximab, ocrelizumab, veltuzumab, ofatumumab, epratuzumab, obinutuzumab, ibritumomab and tiuxetan.

In some embodiments, the B cell targeted therapy is selected from the group consisting of: rituximab, ocrelizumab, veltuzumab, ofatumumab and epratuzumab.

In preferred embodiments, the B cell targeted therapy is rituximab.

In some embodiments, a patient determined to be resistant to treatment with the B cell targeted therapy is determined to be suitable for treatment with an agent that downregulates IL-6 mediated signalling.

In some embodiments, the agent that downregulates IL-6 mediated signalling is an IL-6 receptor antagonist.

In some embodiments, the agent that downregulates IL-6 mediated signalling is selected from the group consisting of tocilizumab, sarilumab, satralizumab and siltuximab.

In preferred embodiments, the agent that downregulates IL-6 mediated signalling is tocilizumab.

In some embodiments, the patient is refractory to DMARD and/or anti-TNF therapy. In some embodiments, the method further comprises administering to the patient a B cell targeted therapy when the patient is determined to be susceptible to treatment with a B cell targeted therapy.

In some embodiments, the method further comprises administering to the patient an agent that downregulates IL-6 mediated signalling when the patient is determined to be susceptible to treatment with an agent that downregulates IL-6 mediated signalling.

In some embodiments, the method further comprises administering to the patient an alternative therapeutic when the patient is determined to be refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling. The alternative therapeutic may be a therapeutic that is not a B cell targeted therapy or an agent that downregulates IL-6 mediated signalling

In some embodiments, the method (or the step of comparing the level of the one or more biomarker and/or the determining susceptibility or refractoriness to treatment) is carried out by a computer.

In another aspect, the invention provides a kit for use in the method of the invention.

In some embodiments, the kit comprises one or more reagent suitable for detecting the one or more first, second and/or third biomarker.

In some embodiments, the kit comprises reagents for RNA sequencing.

In some embodiments, the kit comprises one or more probe or antibody for detecting the one or more first, second and/or third biomarker.

In some embodiments, the kit is in the form of a microchip or microarray.

In another aspect, the invention provides a method for treating Rheumatoid Arthritis (RA), the method comprising:

(a) administering to a patient an effective amount of a B cell targeted therapy, wherein the patient is determined to be susceptible to treatment with a B cell targeted therapy by the method of the invention; or

(b) administering to a patient an effective amount of an agent that downregulates IL-6 mediated signalling, wherein the patient is determined to be susceptible to treatment with an agent that downregulates IL-6 mediated signalling by the method of the invention. In another aspect, the invention provides a method for treating Rheumatoid Arthritis (RA), the method comprising administering to a patient an effective amount of a B cell targeted therapy, wherein the patient is determined to be susceptible to treatment with a B cell targeted therapy by the method of the invention.

In another aspect, the invention provides a method for treating Rheumatoid Arthritis (RA), the method comprising administering to a patient an effective amount of an agent that downregulates IL-6 mediated signalling, wherein the patient is determined to be susceptible to treatment with an agent that downregulates IL-6 mediated signalling by the method of the invention.

In another aspect, the invention provides a method of identifying one or more biomarker for determining whether a Rheumatoid Arthritis (RA) patient is susceptible to treatment with a B cell targeted therapy and/or an agent that downregulates IL-6 mediated signalling. In another aspect, the invention provides a method of identifying one or more biomarker for determining whether a Rheumatoid Arthritis (RA) patient is refractory to treatment with a B cell targeted therapy and/or an agent that downregulates IL-6 mediated signalling. In another aspect, the invention provides a method of generating or optimising a model for determining whether a Rheumatoid Arthritis (RA) patient is susceptible to treatment with a B cell targeted therapy and/or an agent that downregulates IL-6 mediated signalling. In another aspect, the invention provides a method of generating or optimising a model for determining whether a Rheumatoid Arthritis (RA) patient is refractory to treatment with a B cell targeted therapy and/or an agent that downregulates IL-6 mediated signalling. The method of identifying one or more biomarker, or of generating or optimising a model may be part of any other method of the disclosure.

In some embodiments, the method is carried out by a computer.

Suitably, the method may be a machine learning method. Suitably the method may apply or develop a machine learning model. Suitably, the model may be a machine learning model.

In another aspect, the invention provides a data processing device comprising means for carrying out the method of the invention (or the step of comparing the level of the one or more biomarker and/or the determining susceptibility or refractoriness to treatment).

In another aspect, the invention provides a computer program product in which a computer program is stored in a non-transient fashion, which when executed on a processing device causes the processing device to carry out the method of the invention (or the step of comparing the level of the one or more biomarker and/or the determining susceptibility or refractoriness to treatment). In another aspect, the invention provides a computer-readable medium comprising instructions which, when executed by a computer, cause the computer to carry out the method of the invention (or the step of comparing the level of the one or more biomarker and/or the determining susceptibility or refractoriness to treatment).

DESCRIPTION OF THE DRAWINGS

FIGURE 1. Synovial histological patters (pathotypes) associate with response to Rituximab and Tocilizumab a, Classification into synovial pathotypes according to semi-quantitative scores for CD3+ T- cells, CD20+ B-cells, CD68+ macrophages and CD138+ plasma cells, with representative examples from patients classified as fibroid, diffuse-myeloid and lympho-myeloid and (right) 16 weeks CDAI50% response (primary trial endpoint) in patients stratified according to synovial pathotypes. Two-sided Chi-squared test comparing proportion of responders to rituximab and tocilizumab within each pathotype, actual patient numbers shown in addition to proportions. Exact p values shown when <0.05 b, Approach to in silico deconvolution of synovial tissue, using MCP-counter. c, MCP-counter scores for each cell type compared in CDAI50% responders and nonresponders, -Log 10 p values (positive) are shown for tocilizumab and Log 10 p values (negative) for rituximab, with dashed line corresponding at p=0.05 (two-sided Mann-Whitney test). On the right, individual graphs of MCPcounter scores for CD8 T cells and Macrophages, in patients classified as responders (R) and non-responders (NR) to rituximab and tocilizumab, as indicated. Exact nominal p values, two-sided Mann-Whitney test. d-e, 16weeks CDAI50% response in patients stratified into B and T-cell poor/rich (d) and Macrophage and mDC poor/rich (e) according to the median MCP-counter scores for the individual cells (rich if >median, poor if <median). Exact p values shown when <0.05, Chi- squared test comparing the proportion of responders to rituximab and tocilizumab. f, 16weeks CDAI50% response in patients stratified combining B-cell and macrophages/mDC scores as in e. Fisher test comparing the proportion of responders to rituximab and tocilizumab. g-i Longitudinal disease activity scores (CDAI) recorded monthly from baseline to 16 weeks in patients classified as B cell and T cell poor/rich(g), macrophage and mDC poor/rich (h) and combined B cells and macrophages poor/rich (i). Exact p values shown when <0.05 for the comparison of CDAI between the two medications at individual timepoints by two-sided Mann Whitney test (p values adjusted for multiple comparison by false discovery rate) and treatmenttime p = exact p values for the interaction between treatment and time (two-way ANOVA for repeated measures).

FIGURE 2. Molecular signature of response and non-response to rituximab and tocilizumab a, Monte Carlo reference-based consensus clustering of 22,256 most variable genes identified a high inflammatory consensus cluster 1 (blue) and low inflammatory cluster 2 (yellow). Heatmaps were produced for rituximab (n=68, left panel) and tocilizumab (n=65, right panel) treated patients using Pearson’s distance metric and complete linkage method using the ComplexHeatmap package in R. Upper tracks show consensus cluster, Cell type (B-cell rich/B-cell poor), the overall pathotype, CDAI 50% response, and EULAR response and the histological scores for CD20, CD138, CD68L, CD68SL and CD3. b-e, Volcano plots of DEGs using DESeq2 comparing CDAI50% responders versus non responders to rituximab (left a, c) and tocilizumab (right, b, d). Volcano plots c (RTX) and d (TOC) show DEGs after adjustment for principal component 1. Comparison between groups using Wald test and correcting for multiple testing Storey’s q value (q <0.05 = significant, shown in blue). Positive values represent upregulation in responders and negative values downregulation compared to non responders. f, g, Modular analysis applying Qusage for responders versus non-responders to rituximab (e) and tocilizumab (f). Log2-fold change of responders (positive values) and non responders (negative values) are plotted for blood microarray-based modules (Li et al. (2014) Nat. Immunol. 15: 195-204) and WGCNA modules summarized in one plot and dots are color coded for their q-value.

FIGURE 3. Identification of non-response (refractory) signature a, Classification of patients considering treatment switch (complete trial scheme in Figure 9): patients who responded to rituximab after failing tocilizumab (pro-RTX, blue), patients who responded to tocilizumab after failing rituximab (pro-TOC, yellow) and patients who failed both drugs sequentially (refractory, red). Numbers indicate all patients, numbers in brackets are patients with available RNA-Seq. b, Venn diagram showing the overlap of differentially expressed genes between patients classified as in (a). c 3-way differential gene expression analysis on baseline synovial biopsies of patients classified as in (a). Genes in blue show a significant difference in patients who failed tocilizumab and responded to rituximab (pro-RTX). Genes in yellow indicate a significant difference in patients who failed rituximab but responded to tocilizumab (pro-TOC). Genes in green show significant genes overlapping in pro-RTX and pro-TOC patients. Finally, genes in red are significantly upregulated in refractory patients, who failed both drugs. Genes in grey are not significantly changed, or not specific for any group of patients. Significance was internally estimated by the volcano3D package combining significance (q < 0.05) from both likelihood ratio test (LRT) and pairwise Wald test estimated via DESeq2. d, Three-way Quantitative gene enrichment set analysis for gene expression (QuSage) radial plot showing differential WGCNA module expression in patients classified as above). e, Histological semi-quantitative scores for immune cells in patients classified as refractory (non-responders to both medications, n=40) or responders to one of any two medications (n=24). Boxplots showing median and first and third quartiles. Two way Mann-Whitney test, exact p values false-discovery rate adjusted for multiple comparisons. f, Deconvolution of immune cells using MCP counter in patients classified as refractory or responders as in (a). Boxplots showing median and first and third quartiles, dotplots showing individual patients. Two way Mann-Whitney test, exact p values false-discovery rate adjusted for multiple comparisons. g, Fibroblast and macrophages single cell subsets enrichment scores in patients classified as refractory (non-responders to both medications, n=40) or responders to either rituximab or tocilizumab (n=24), as per scheme in Figure 2g. Boxplots showing median and first and third quartiles, whiskers extending to the highest and lowest values. Exact p values are shown, two sided Mann-Whitney test. h, Representative images of DKK3+ fibroblasts (upper) and SPP1+ macrophages (bottom) in refractory and responder patients. Immunofluorescence with DNA in blue, CD45 in red, CD90 in green and DKK3 in yellow (upper panel) and DNA in blue, CD68 in green and SPP1 in red (bottom panel).

FIGURE 4. Digital spatial profiling of refractory RA a, Scheme showing the approach to digital spatial profiling, including selection of ROIs: CD68+ lining and superficial sublining, CD20-CD3- deep sublining and CD3+CD20+ lymphoid aggregates. b, MA plots showing mean expression (Iog2) on the x axis and fold change on the y axis comparing responders and refractory patients across all ROIs. Genes that are significantly upregulated in responders are shown in blue (upper half) and upregulated in refractory in red (lower half). In Grey genes with a q value >0.05; P-values were false discovery rate (FDR) adjusted using Storey’s q-value with a cut-off of q < 0.05. n= 12 patients, 6 ROIs per patient. c, Example of individual genes that are differentially expressed in refractory (red) or responders (green). Scatterplots showing individual ROIs, boxplots showing median and first and third quartiles. Exact p values are shown, differential expression analysis using DeSeq2, p values were false discovery rate (FDR) adjusted using Storey’s q-value. d, MA plots showing mean expression (Iog2) on the x axis and fold change on the y axis comparing responders and refractory patients in the different ROIs, as indicated. Genes that are significantly upregulated in responders are shown in blue (upper half) and upregulated in refractory in red (lower half). In Grey genes with a q value >0.05; P-values were false discovery rate (FDR) adjusted using Storey’s q-value with a cut-off of q < 0.05. n= 12 patients, 6 ROIs per patient. e, Venn diagram showing the number of differentially expressed genes that are specific for each ROIs and overlaps f, Examples of individual genes that are differentially expressed in refractory (red) or responders (green) in the different ROIs. Scatterplots showing individual ROIs, boxplots showing median and first and third quartiles. Exact p values are shown, differential expression analysis using DeSeq2, p values were false discovery rate (FDR)-adjusted using Storey’s q- value.

FIGURE 5. Longitudinal differential gene expression analysis a, Scatter plot showing change in gene expression over 16 weeks comparing differences in the change in gene expression in 88 paired biopsies at baseline and 16 weeks in 44 patients following rituximab (n=29) or tocilizumab (n=15) treatment. Log2 fold change in expression following rituximab is represented on the x axis and Iog2 fold change in expression following tocilizumab is represented on the y axis. Fold change and statistical analysis of longitudinal differential gene expression were calculated by negative binomial general linear mixed effects model. Genes in green show significant (FDR < 0.05) overall change in expression over time. Genes in blue/yellow show significantly differential change in expression over time between the two drugs based on significant (FDR < 0.05) interaction term time:medication (see Example 1 Methods). Genes with greater absolute fold change following rituximab are shown in blue, while genes with greater absolute fold change following tocilizumab are shown in yellow. b, Scatter plots for selected genes showing fitted model with 95% confidence intervals (fixed effects). c-f, Longitudinal mixed-effects model on (c) rituximab (58 samples, 29 individuals) and (e) tocilizumab treated patients (30 samples, 15 individuals) showing differential gene expression between responders and non-responders categorised by CDAI50% response. Genes showing greater absolute gene expression change in non-responders are in red, while genes showing greater absolute gene expression change in responders are in blue for rituximab responders (d) and yellow for tocilizumab responders (f). Scatter plots of representative genes. g-i, Pathway analysis at baseline and 16 week in patients treated with rituximab (g), responders and non-responders to rituximab (h) and responders to tocilizumab (i). The dashed line indicates p=0.05 adjusted p-value (Bonferroni adjustment).

FIGURE 6. Predictive models using nested 10x10-fold cross-validation for response to Rituximab and Tocilizumab. a, Machine learning pipeline to predict CDAI 50% response to RTX and/or TOC using gene expression, clinical data and histological data as features (n=133). Data processing (box i) involved selecting protein-coding genes with the highest variance and removing highly correlated genes. Data was split into 10 inner and 10 outer folds for building machine learning models (box ii). In models built using gene expression, recursive feature elimination (RFE) or univariate filtering was used to select the most important/predictive features for each model. Each model was evaluated on both the test set and the set left-out during cross-validation (box iii). The average tuned parameters from the outer folds were used to fit to the whole data set to determine the importance of features selected for each model. b, Grid of plots showing the optimal predictive models for different treatments (glmnet RTX response prediction left; glmnet TOC response prediction middle; and gradient boosting machine Refractory response prediction right) using gene expression and baseline clinical parameters as features. From top to bottom plots show: ROC curves for the best model on the test data (from outer-fold) set; ROC curves on the left-out (from inner-fold) set; and the variable importance when fit to the whole data set.

FIGURE 7. Histological analyses a, Atlas of semi-quantitative synovial IHC scores for immune cells. b, Distribution of semiquantitative scores at baseline in all patients, individually shown in the y axis. The total on the x axis represents the sum of the individual scores (Immune score) c, Baseline semi-quantitative IHC scores, Krenn synovitis score (Synovitis score) and total Immune score in patients stratified according to 16 weeks CDAI50% response to rituximab (top) and tocilizumab (bottom). Two-sided Mann Whitney test. ns= p value >0.05

FIGURE 8. Unsupervised Principal Component Analysis shows association primarily with cell types present and consequently also pathotype. a, Clinical features and their degree of association with Principal Components (PC) 1-10 with coloring indicating the -log(p) (left) and FDR corrected -log(q) value (right). RF, Rheumatoid Factor; CCP, anti-Cyclic Citrullinated Protein; CRP, C-Reactive Protein; ESR, Erythrocyte Sedimentation Rate; SJC, Swollen Joint Counts; TJC, Tender Joint Count b, PC 1 and 3 gene expression variance with coloring by (b) pathotypes showing fibroid (blue), lymphoid (red), myeloid (pink) and ungraded (grey) patients. Ellipses indicate 80% confidence interval. c and d, PC1 and 2 colored by response to treatment. Patients allocated to treatment group rituximab are displayed in c and to tocilizumab in d, with non response colored in red, response to RTX in blue and response to TOC in gold. Ellipses shown for all PCs represent the 80% confidence interval. e, differential expression of genes important for B-cells (MS4A1 , CD79A, CD79B, PIK3CA, BTK and SYK) and Weighted Gene Correlation Network Analysis (WGCNA) cell modules (B- cells, M1 macrophage cytokine signalling, Fibroblast 2a THY1+) in Rituximab treated patients, according to the consensus clusters shown in (Fig.2a, left panel).

Boxplots show median with upper and lower hinges and whiskers extending to highest and lowest point, but at most 1.5x the interquartile range, p-values stated for Kruskal-Wallis test. f, II-6 related genes (IL6R, IL6, IL6ST, JAK1 JAK2 and STAT3) and WGCNA cell modules expression in tocilizumab (Fig.2a, right panel) treated patients based on consensus clusters. Boxplots as above. g, Boxplots showing median with upper and lower hinges for semiquantitative histological scores of CD3, CD20, CD68L, CD68SL, CD138 and CD79a for all patients split into consensuscluster 1 and consensuscluster 2. Kruskal-Wallis test p-values are shown.

FIGURE 9. Influence of immune cells on consensusclusters a-d, Volcano plots showing differential gene expression analysis using DESeq2 comparing consensuscluster 1 and 2 of patients treated with rituximab (left) or tocilizumab (right). While a and b were analyzed without covariates, c and d were adjusted for principal component (PC1). Comparison between groups were tested for significance using Wald test and multiple testing was corrected for with Storey’s q value (q <0.05 = significant, shown in blue). Positive Iog2fold changes represent upregulation in consensuscluster 2, negative Iog2fold changes represents upregulation in consensuscluster 1. e, Correlation plot highlighting relation between PC1 , histology markers and genes involved in the mode of action of RTX and TOC. Positive correlation is shown in blue while red would indicate negative correlation. For all correlations without significance, the p-value is shown.

FIGURE 10. Trial scheme

Tocilizumab (TOC), rituximab (RTX), responder (R), non-responder (NR), n = samples available for histology, (n) = samples available for RNAseq

FIGURE 11. Longitudinal histological and in silico analysis of paired pre- and posttreatment synovial biopsies a, Schema showing an overview of longitudinal analysis of matched pre and post-treatment synovial biopsies, with number of samples for each medication (in brackets samples with available RNA-Seq). b, Semi-quantitative scores of synovial immune cells at baseline and 16 weeks in patients treated with rituximab (b) and tocilizumab (c), in all patients (ALL) and stratified by CDAI50% response (NR= non responders, R= responders). Boxplots showing median and first and third quartiles. Wilcoxon signed-rank test, p values shown when <0.05, two sided Wilcoxon signed- rank test (paired) comparing baseline and 16 weeks, adjusted for multiple testing by false discovery rate. c, Semi-quantitative scores at baseline and 16 weeks in patients stratified according to treatment with rituximab or tocilizumab. Mean ± SEM. Exact p values from analysis of covariance testing the difference in the changes from baseline between treatments, with treatment as factor and baseline score as covariate. d,e, MCP-counter scores in baseline and 16 weeks samples. In e, overview of comparison between rituximab (- Iog10 p values) and tocilizumab (log 10 p values) treated patients, with dashed lines at p=0.05, and in e, detailed graphs for each cell, with scatterplots showing individual samples and boxplots showing median and first and third quartiles, whiskers extending to the highest and lowest values no further than 1.5 * interquartile range. Two sided Wilcoxon signed-rank test (paired), comparing baseline and 16 weeks, adjusted for multiple testing by false discovery rate Sample sizes as shown in (a).

FIGURE 12. Immunofluorescence of DKK3+ fibroblasts and SPP1+ macrophages

DKK3+ fibroblasts (upper) and SPP1+ macrophages (bottom) in refractory and responder patients (representative image our of 3 refractory and 3 responders for each marker). Immunofluorescence with DNA in blue, CD45 in red, CD90 in green and DKK3 in yellow (upper panel) and DNA in blue, CD68 in green and SPP1 in red (bottom panel). In the upper panels (DKK3) asterisks correspond to CD45+ lymphocytes (red) positive for DKK3 (yellow), while arrowheads to CD90+ fibroblasts (green) expressing DKK3 (yellow). In the bottom panel (SPP1), arrowheads correspond to CD68+ (green) macrophages expressing SPP1 (red). Lines at 250pm in larger panels and 50pm in higher magnification.

FIGURE 13. Comparison of negative binomial mixed-effects model and gaussian mixed- effects model applied to log count data a, Correlation plots of -Iog10 P values for analysis comparing effect of each drug over time using model gene ~ drug * time + (1 | patient). Gaussian p values for model parameters time, drug and interaction term time:drug are plotted on the x axis, while negative binomial mixed- effects model p values are plotted on the y axis, with p values -Iog10 transformed. Each point represents a single gene. b, Similar analysis comparing Gaussian mixed model and negative binomial mixed model for the comparison of responders vs non-responders over time for the rituximab treated cohort. c, Similar analysis comparing Gaussian mixed model and negative binomial mixed model for the comparison of responders vs non-responders over time for the tocilizumab treated cohort. d-f, QQ plots showing observed vs expected -log 10 p values for each of the 3 analyses in (a- c) demonstrating substantially increased power of negative binomial mixed models compared to Gaussian mixed models for detecting significant genes.

FIGURE 14. Plots showing the evaluation models using gene expression and clinical variables as features. a, Venn diagrams showing gene names and the number of genes overlapping between the three main predictive models in Fig. 6b. b, From top to bottom: ROC curves on the test data set (outer CV folds) for the three best models using clinical and histological data only; ROC curves for clinic-histological models on the left-out inner CV folds; variable importance for the best clinical models.

DETAILED DESCRIPTION OF THE INVENTION

The terms “comprising”, “comprises” and “comprised of” as used herein are synonymous with “including” or “includes”; or “containing” or “contains”, and are inclusive or open-ended and do not exclude additional, non-recited members, elements or steps. The terms “comprising”, “comprises” and “comprised of” also include the term “consisting of”.

Rheumatoid arthritis (RA)

Rheumatoid arthritis (RA) is a chronic, systemic inflammatory disorder that may affect many tissues and organs, but principally attacks synovial joints. It is a disabling and painful condition, which can lead to substantial loss of functioning and mobility if not adequately treated.

The disease process involves an inflammatory response of the synovium, secondary to massive immune cell infiltration and proliferation of synovial cells, excess synovial fluid, and the development of fibrous tissue (pannus) in the synovium that attacks the cartilage and subchondral bone. This often leads to the destruction of articular cartilage and the formation of bone erosions with secondary ankylosis (fusion) of the joints. RA can also produce diffuse inflammation in the lungs, the pericardium, the pleura, the sclera, and also nodular lesions, most commonly in subcutaneous tissue. RA is considered a systemic autoimmune disease as autoimmunity plays a pivotal role in its chronicity and progression.

A number of cell types are involved in the aetiology of RA, including T cells, B cells, monocytes, macrophages, dendritic cells and synovial fibroblasts. Autoantibodies known to be associated with RA include those targeting Rheumatoid factor (RF) and anti-citrullinated protein antibodies (ACPA).

RA therapy

A typical patient with newly diagnosed RA is often treated initially with nonsteroidal antiinflammatory drugs and disease-modifying anti-rheumatic drugs (DMARDs), such as hydroychloroquine, sulfasalazine, leflunomide or methotrexate (MTX), alone or in combinations. Patients who do not respond to general DMARDs may be termed DMARD- refractory. DMARD-refractory patients are traditionally often progressed to biological therapeutic agents, for example TNF-a antagonists such as Adalimumab, Etanercept, Golimumab and Infliximab. Patients who do not respond to TNF-a antagonist therapy may be termed TNF-a antagonistrefractory or inadequate responders (ir).

The method of the invention may be performed on a sample from a RA patient who has previously been determined to be refractory to DMARD-therapy and/or TNF-a antagonist therapy. The method may also be performed on a sample from a RA patient unable to tolerate TNF-a antagonist therapy.

B cell targeted therapy

The method of the invention may determine an RA patient as being susceptible to treatment with a B cell targeted therapy.

The term “B cell targeted therapy”, as used herein, may refer to the administration of an agent that interferes with or inhibits the development and/or function of B cells. The B cell targeted therapy may cause B cell depletion or the inhibition of B cell development and maturation. Advantageously, the B cell targeted therapy is directed against B cells in all stages of development other than undifferentiated stem cells and terminally differentiated antibodyproducing plasma cells.

The agent may be a small molecule drug, such as a Bruton's tyrosine kinase (BTK) inhibitor or other agent which targets B cell signalling pathways.

Direct depletion of B cells may be performed through the use of antibodies, such as monoclonal antibodies (mAbs), directed against cell surface markers (e.g. CD20 and CD22). Such antibodies bind to the target antigen and kill the cell by initiating a mixture of apoptosis, complement dependent cytotoxicity (CDC), and antibody-dependent cell-mediated cellular cytotoxicity (ADCC).

The B cell targeted therapy used in the invention may be an agent directed against CD20, for example Rituximab, Ocrelizumab, Veltuzumab or Ofatumumab, or an agent directed against CD22 such as Epratuzumab.

Rituximab is a chimeric mouse/human immunoglobulin G1 (lgG1) monoclonal antibody to CD20 that stimulates B cell destruction upon binding to CD20. Rituximab depletes CD20 surface-positive naive and memory B cells from the blood, bone marrow and lymph nodes via mechanisms which include antibody-dependent cellular cytotoxicity (ADCC), complement dependent cytotoxicity (CDC). It does not affect CD20-negative early B cell lineage precursor cells and late B lineage plasma cells in the bone marrow.

Ocrelizumab is a humanized anti-CD20 monoclonal antibody that causes CD20+ B cell depletion following binding to CD20 via mechanisms including ADCC and CDC.

Veltuzumab is a humanized, second-generation anti-CD20 monoclonal antibody that causes CD20+ B cell depletion following binding to CD20 via mechanisms including ADCC and CDC.

Ofatumumab is a human monoclonal lgG1 antibody to CD20 and may inhibit early-stage B lymphocyte activation. Ofatumumab targets a different epitope located closer to the N- terminus of CD20 compared to the epitope targeted by rituximab and includes an extracellular loop, as it binds to both the small and large loops of the CD20 molecule. Ofatumumab stimulates B cell destruction through ADCC and CDC pathways.

Epratuzumab is a humanized monoclonal lgG1 antibody to CD22. It contains a murine sequence comprising 5-10% of the molecule, the remainder being human framework sequences. Epratuzumab binds to the CD22 third extracellular domain (epitope B), without blocking the ligand binding site, with measured affinity of Kd = 0.7 nm. In vitro studies showed epratuzumab induces CD22 phosphorylation by binding to its surface. It results in modulation, mostly negative, of BCR activation.

IL-6 mediated signalling

The method of the invention may determine an RA patient as being susceptible to treatment with an agent which downregulates interleukin-6 (IL-6) signalling.

IL-6 is a cytokine that provokes a broad range of cellular and physiological responses, including inflammation, hematopoiesis and oncogenesis by regulating cell growth, gene activation, proliferation, survival, and differentiation. It is able to directly influence B cell activation state and late stage differentiation towards plasma cells.

IL-6 signals through a receptor composed of two different subunits, an alpha subunit that produces ligand specificity and GP (Glycoprotein) 130, a receptor subunit shared in common with other cytokines in the IL-6 family. Binding of IL-6 to its receptor initiates cellular events including activation of JAK (Janus Kinase) kinases and activation of Ras-mediated signalling. Activated JAK kinases phosphorylate and activate STAT transcription factors, particularly STAT3 and SHP2. Phosphorylated STAT3 then forms a dimer and translocates into the nucleus to activate transcription of genes containing STAT3 response elements. STAT3 is essential for GP130-mediated cell survival and G1 to S cell-cycle-transition signals. Both c- Myc and Pirn have been identified as target genes of STAT3 and together can compensate for STAT3 in cell survival and cell-cycle transition. SHP2 links cytokine receptor to the Ras/MAP (Mitogen-Activated Protein) kinase pathway and is essential for mitogenic activity.

The Ras-mediated pathway, acting through SHC, GRB2 (Growth Factor Receptor Bound protein-2) and SOS1 (Son of Sevenless-1) upstream and activating MAP kinases downstream, activates transcription factors such as Elk1 and NF-IL-6 (C/EBP-P) that can act through their own cognate response elements in the genome.

In addition to JAK/STAT and Ras/MAP kinase pathways, IL-6 also activates PI3K (Phosphoinositide-3 Kinase). The PI3K/Akt/NF-KappaB cascade activated by IL-6, functions cooperatively to achieve the maximal anti-apoptotic effect of IL-6 against TGF-p. The anti- apoptotic mechanism of PI3K/Akt is attributed to phosphorylation of the BCL2 family member BAD (BCL2 Associated Death Promoter) by Akt. The phosphorylated BAD is then associated with 14-3-3, which sequesters BAD from BCLXL, thereby promoting cell survival. Regulating the BCL2 family member is also considered as one of the anti-apoptotic mechanisms of STAT3, which may be capable of inducing BCL2 in pro-B cells. The termination of the I L-6- type cytokine signalling is through the action of tyrosine phosphatases, proteasome, and JAK kinase inhibitors SOCS (Suppressor of Cytokine Signaling), PIAS (Protein Inhibitors of Activated STATs), and internalization of the cytokine receptors via GP130.

An agent which downregulates IL-6 signalling may interfere with or inhibit any of the above stages involved in IL-6 mediated signalling such that IL-6 signalling and responses are diminished. For example, the agent may be an IL-6 receptor antagonist such as Tocilizumab, which is a humanized monoclonal antibody against the IL-6 receptor. An IL-6 receptor antagonist refers to an agent that reduces the level of IL-6 that is able to bind to the IL-6 receptor.

Tocilizumab is a humanized monoclonal lgG1 antibody against the IL-6 receptor that binds to soluble and membrane-bound IL-6 receptor. Tocilizumab inhibits the induction of biological activity due to IL-6 in cells that have expressed both membrane-bound IL-6 receptor and gp130 molecules, and also inhibits the induction of biological activity due to IL-6/IL-6 receptor complex formation in cells that express gp130 alone. Furthermore, since it has the capacity to dissociate IL-6/IL-6 receptor complexes that have already formed, it is able to block IL-6 signal transduction.

B cells

B cells play a central role in the pathogenesis of RA. Immature B cells are produced in the bone marrow. After reaching the lgM⁺ immature stage in the bone marrow, these immature B cells migrate to secondary lymphoid tissues (such as the spleen, lymph nodes) where they are called transitional B cells, and some of these cells differentiate into mature B lymphocytes and possibly plasma cells.

B cells may be defined by a range of cell surface markers which are expressed at different stages of B cell development and maturation (see table below). These B cell markers may include CD19, CD20, CD22, CD23, CD24, CD27, CD38, CD40, CD72, CD79a and CD79b, CD138 and immunoglobulin (Ig).

Immunoglobulins (Ig) are glycoproteins belonging to the immunoglobulin superfamily which recognise foreign antigens and facilitate the humoral response of the immune system. Ig may occur in two physical forms, a soluble form that is secreted from the cell, and a membranebound form that is attached to the surface of a B cell and is referred to as the B cell receptor (BCR). Mammalian Ig may be grouped into five classes (isotypes) based on which heavy chain they possess. Immature B cells, which have never been exposed to an antigen, are known as naive B cells and express only the IgM isotype in a cell surface bound form. B cells begin to express both IgM and IgD when they reach maturity - the co-expression of both these immunoglobulin isotypes renders the B cell “mature” and ready to respond to antigen. B cell activation follows engagement of the cell bound antibody molecule with an antigen, causing the cell to divide and differentiate into an antibody producing plasma cell. In this activated form, the B cell starts to produce antibody in a secreted form rather than a membrane-bound form. Some daughter cells of the activated B cells undergo isotype switching to change from IgM or IgD to the other antibody isotypes, IgE, IgA or IgG, that have defined roles in the immune system.

CD19 is expressed by essentially all B-lineage cells and regulates intracellular signal transduction by amplifying Src-family kinase activity.

CD20 is a mature B cell-specific molecule that functions as a membrane embedded Ca²⁺ channel. Expression of CD20 is restricted to the B cell lineage from the pre-B-cell stage until terminal differentiation into plasma cells.

CD22 functions as a mammalian lectin for a2,6-linked sialic acid that regulates follicular B-cell survival and negatively regulates signalling.

CD23 is a low-affinity receptor for IgE expressed on activated B cells that influences IgE production.

CD24 is a GPI-anchored glycoprotein which was among the first pan-B-cell molecules to be identified.

CD27 is a member of the TNF-receptor superfamily. It binds to its ligand CD70, and plays a key role in regulating B-cell activation and immunoglobulin synthesis. This receptor transduces signals that lead to the activation of NF-KB and MAPK8/JNK.

CD38 is also known as cyclic ADP ribose hydrolase. It is a glycoprotein that also functions in cell adhesion, signal transduction and calcium signalling and is generally a marker of cell activation.

CD40 serves as a critical survival factor for germinal centre (GC) B cells and is the ligand for CD154 expressed by T cells.

CD72 functions as a negative regulator of signal transduction and as the B-cell ligand for Semaphorin 4D (CD100).

CD79a/CD79b dimer is closely associated with the B-cell antigen receptor, and enables the cell to respond to the presence of antigens on its surface. The CD79a/CD79b dimer is present on the surface of B-cells throughout their life cycle, and is absent on all other healthy cells.

CD138 is also known as Syndecan 1. Syndecans mediate cell binding, cell signalling and cytoskeletal organisation. CD138 may be useful as a cell surface marker for plasma cells.

In some embodiments, the method further comprises a step of analysing the presence of B cells in one or more sample (preferably a synovial sample) from the RA patient and determining if the RA patient is B cell rich or B cell poor by histological analysis. This analysis may involve determining the presence of cells expressing one or more of the markers detailed in the table above.

The presence of B cells may be determined by analysing the level and pattern of B cells.

The histological identification of RA patients who are B cell rich or B cell poor may be performed by using a system for grading lymphocytic aggregates known to those skilled in the art, for example as disclosed in the Examples herein.

For example, sections may undergo semi-quantitative scoring (0-4) to determine expression of CD20+ B-cells, CD3+ T cells, CD138+ plasma cells and CD68+ lining (I) and sub lining (si) macrophages as previously described and validated (Rivellese F et al. Arthritis Rheumatol 2020; 72: 714-25; Kraan MC et al. Rheumatology 2000; 39: 43-9; Krenn V et al. Histopathology 2006; 49: 358-64). Synovial tissue with a CD20 score <2 may be classified histologically as B-cell-poor, while tissues with CD20 score >2 and with CD20+ B-cell aggregates may be classified histologically as B-cell-rich.

Response to therapies in RA patients

Methods of assessing a subject’s response to a therapy for rheumatoid arthritis are known in the art and would be familiar to a skilled person.

Byway of example, well known measures of disease activity in RA include the Disease Activity Score (DAS), a modified version DAS28, and the DAS-based ELILAR response criteria.

The assessment of response to a therapy for rheumatoid arthritis may use the Clinical Disease Activity Index (CDAI), for example as disclosed in the Examples herein.

Susceptibility or refractoriness to treatment of rheumatoid arthritis may, for example, be achievement or not of a CDAI > 50%.

Other measures of assessment of response to a therapy for rheumatoid arthritis include CDAI- remission, DAS28(ESR)/(CRP) moderate/good EULAR-response, DAS28(ESR)/(CRP) low- disease-activity, DAS28(ESR)/(CRP) remission and patient reported outcomes, such as fatigue.

Biomarkers

Susceptibility to treatment with a B cell targeted therapy In one aspect, the invention provides a method for determining whether a Rheumatoid Arthritis (RA) patient is susceptible to treatment with a B cell targeted therapy, the method comprising the steps: (i) determining the level of one or more first biomarker in one or more sample obtained from the patient, wherein the one or more first biomarker is selected from Table 1 ; and (ii) comparing the level of the one or more first biomarker to one or more corresponding reference value; wherein the level of the one or more first biomarker compared to the corresponding reference value is indicative of the susceptibility to treatment with a B cell targeted therapy.

In some embodiments, the one or more first biomarker comprises 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 , 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 , 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 , 32, 33, 34, 35, 36,

37, 38, 39, or all 40 biomarkers from Table 1.

In some embodiments, the one or more first biomarker consists of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 , 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 , 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 , 32, 33, 34, 35, 36,

37, 38, 39, or all 40 biomarkers from Table 1.

In some embodiments, the one or more first biomarker consists of all 40 biomarkers from

Table 1.

Table 1. Rituximab model genes.

Susceptibility to treatment with an agent that downregulates IL-6 mediated signalling

In one aspect, the invention provides a method for determining whether a Rheumatoid Arthritis (RA) patient is susceptible to treatment with an agent that downregulates IL-6 mediated signalling, the method comprising the steps: (i) determining the level of one or more second biomarker in one or more sample obtained from the patient, wherein the one or more second biomarker is selected from Table 2; and (ii) comparing the level of the one or more second biomarker to one or more corresponding reference value; wherein the level of the one or more second biomarker compared to the corresponding reference value is indicative of the susceptibility to treatment with an agent that downregulates IL-6 mediated signalling. In some embodiments, the one or more second biomarker comprises 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 , 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 , 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 , 32, 33, 34, 35, 36, 37, 38, or all 39 biomarkers from Table 2.

Table 2. Tocilizumab model genes.

Refractoriness to treatment with a B cell targeted therapy and an agent that downregulates IL- 6 mediated signalling

In one aspect, the invention provides a method for determining whether a Rheumatoid Arthritis (RA) patient is refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling, the method comprising the steps: (i) determining the level of one or more third biomarker in one or more sample obtained from the patient, wherein the one or more third biomarker is selected from Table 3; and (ii) comparing the level of the one or more third biomarker to one or more corresponding reference value; wherein the level of the one or more third biomarker compared to the corresponding reference value is indicative of the refractoriness to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling.

53

In some embodiments, the one or more third biomarker comprises 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 , 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 , 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 , 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 , 42, 43, 44, 45, 46, 47, 48, 49, 50, 51 , 52, or all 53 biomarkers from Table

3. In some embodiments, the one or more third biomarker comprises all 53 biomarkers from Table 3.

In some embodiments, the one or more third biomarker consists of 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 , 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 , 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 , 32, 33, 34, 35, 36, 37, 38, 39, 40, 41 , 42, 43, 44, 45, 46, 47, 48, 49, 50, 51 , 52, or all 53 biomarkers from

Table 3.

In some embodiments, the one or more third biomarker consists of all 53 biomarkers from

Table 3.

Table 3. Refractory model genes.

The methods of the invention may apply statistical methods as would be understood by the skilled person. For example, the methods of the invention may apply a model such as an elastic net regression model or a gradient boosting machine model, such as a model disclosed herein in the Examples (e.g. using one or more coefficient or variable importance disclosed therein). In some embodiments, the method for determining whether a Rheumatoid Arthritis

(RA) patient is susceptible to treatment with a B cell targeted therapy applies an elastic net regression model (e.g. as disclosed herein in the Examples, for example using one or more coefficient as disclosed therein). In some embodiments, the method for determining whether a Rheumatoid Arthritis (RA) patient is susceptible to treatment with an agent that downregulates IL-6 mediated signalling applies an elastic net regression model (e.g. as disclosed herein in the Examples, for example using one or more coefficient as disclosed therein). In some embodiments, the method for determining whether a Rheumatoid Arthritis (RA) patient is refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling applies a gradient boosting machine model (e.g. as disclosed herein in the Examples, for example using one or more variable importance as disclosed therein).

In some embodiments, the method further comprises combining the level of the one or more biomarkers with one or more anthropometric measures of the patient. Preferably the anthropometric measure is selected from the group consisting of gender, weight, height, age and body mass index, more preferably the anthropometric measure is age (particular preferably when the method is for determining whether a Rheumatoid Arthritis (RA) patient is susceptible to treatment with an agent that downregulates IL-6 mediated signalling, or for determining whether a Rheumatoid Arthritis (RA) patient is refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling).

Determining the level of one or more biomarkers

Methods for determining biomarker levels are well known in the art and would be familiar to the skilled person.

For example, the level of a biomarker may be determined by measuring gene expression for the biomarker gene (for example, using RTPCR) or by detecting the protein product of the biomarker gene (for example, using an immunoassay).

In some embodiments, the step of determining the level of the one or more biomarkers comprises determining the level of gene expression of the one or more biomarkers.

In some embodiments, the level is a nucleic acid level. In some embodiments, the nucleic acid level is an mRNA level.

In some embodiments, the level of the one or more biomarkers is determined by direct digital counting of nucleic acids (e.g. by Nanostring), RNA-seq, RT-qPCR, qPCR, multiplex qPCR or RT-qPCR, microarray analysis, or a combination thereof.

In preferred embodiments, the level of one or more biomarker is determined by RNA sequencing

In some embodiments, the level is a protein level. In some embodiments, the level of the one or more biomarkers is determined by an immunoassay, liquid chromatography-mass spectrometry (LC-MS), nephelometry, aptamer technology, or a combination thereof.

In some embodiments, the level of the one or more biomarkers is an average of the level of the one or more biomarkers. In some embodiments, the average of the level of the one or more biomarkers is an average of a normalised level of the one or more biomarkers.

In some embodiments, the level of the one or more biomarkers is a median of the level of the one or more biomarkers. In some embodiments, the median of the level of the one or more biomarkers is a median of a normalised level of the one or more biomarkers.

In some embodiments, the level of the one or more biomarkers is the level of the one or more biomarkers normalised to a reference gene.

Reference values

The method of the invention may comprise the step of comparing the level of one or more biomarker to one or more corresponding reference value.

As used herein, the term “reference value” may refer to a level against which another level (e.g. the level of one or more biomarker disclosed herein) is compared (e.g. to make a diagnostic (e.g. predictive and/or prognostic) and/or therapeutic determination).

For example, the reference value may be derived from level(s) in a reference population (preferably the median level in a reference population), for example the population of patients disclosed in the Examples herein; a reference sample; and/or a pre-assigned value (e.g. a cut-off value which was previously determined to significantly separate a first subset of individuals who are susceptible to treatment with a B cell targeted therapy and/or an agent that downregulates IL-6 mediated signalling and a second subset of individuals who are resistant to the treatment; or a cut-off value which was previously determined to significantly separate a first subset of individuals who are refractory to treatment with a B cell targeted therapy and/or an agent that downregulates IL-6 mediated signalling and a second subset of individuals who are susceptible to the treatment).

In some embodiments, the cut-off value may be the median or mean (preferably median) level in the reference population. In some embodiments, the reference level may be the top 40%, the top 30%, the top 20%, the top 10%, the top 5% or the top 1 % of the expression level in the reference population. The reference value may, for example, be based on a mean or median level of the biomarker in a control population of subjects, e.g. 5, 10, 100, 1000 or more subjects (who may be age- and/or gender-matched, or unmatched to the test subject).

In certain embodiments the reference value may have been previously determined, or may be calculated or extrapolated without having to perform a corresponding determination on a control sample with respect to each test sample obtained.

The increase in the level of the one or more biomarker compared to the corresponding reference value (when the level of the one or more biomarker is greater than the corresponding reference value) may, for example, be an increase in the level of at least about 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, 95%, 96%, 97%, 98% or 99% or greater relative to the reference value. The increase in the level of the one or more biomarker compared to the corresponding reference value (when the level of the one or more biomarker is greater than the corresponding reference value) may, for example, be an increase in the level of at least about 1.1x, 1.2x, 1.3x, 1.4x, 1.5x, 1.6x, 1.7x, 1.8x, 1.9x, 2x, 2.1x, 2.2x, 2.3x, 2.4x, 2.5x, 2.6x, 2.7x, 2.8x, 2.9x, 3x, 3.5x, 4x, 4.5x, 5x, 6x, 7x, 8x, 9x, 10x, 15x, 20x, 30x, 40x, 50x, 100x, 500x or 1000x relative to the reference value.

The decrease in the level of the one or more biomarker compared to the corresponding reference values (when the level of the one or more biomarker is less than the corresponding reference value) may, for example, be a decrease in the level of at least about 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, 95%, 96%, 97%, 98% or 99% or greater relative to the reference value.

Sample

The method of the invention may be carried out on one or more sample obtained from a subject, for example a patient with RA.

Samples may be obtained from a joint of a subject, for example from a biopsy. Samples may be obtained from a synovial tissue sample from a subject.

In preferred embodiments, the one or more sample is a synovial sample. In some embodiments, the synovial sample is a synovial tissue sample or a synovial fluid sample.

As used herein, the term “synovial sample” refers to a sample derived from a synovial joint. Typically, the synovial sample will be derived from a synovial joint of a RA patient. A synovial sample may be a synovial tissue biopsy and the synovial joint may display active inflammation at the time the sample is taken. Samples may be biological samples taken from a patient. In some embodiments, a sample is blood. In some embodiments, a sample is serum (e.g. the fluid and solute component of blood without the clotting factors). In some embodiments, a sample is plasma (e.g. the liquid portion of blood).

Methods for obtaining samples, such as synovial tissue samples are well known in the art and are familiar to the skilled person. For example, techniques such as ultrasound (US)-guided biopsies may be used to obtain tissue samples.

In some embodiments, the sample is obtained by synovial biopsy, preferably ultrasound- guided synovial biopsy.

Patient

In preferred embodiments, the patient is a human.

In preferred embodiments the patient is an adult human. In some embodiments, the patient may be a child or an infant.

In preferred embodiments, the RA patient is refractory to DMARD and/or anti-TNF therapy

Antibodies

The term “antibody” is used herein to relate to an antibody or a functional fragment thereof. By functional fragment, it is meant any portion of an antibody which retains the ability to bind to the same antigen target as the parental antibody.

As used herein, “antibody” means a polypeptide having an antigen binding site which comprises at least one complementarity determining region (CDR). The antibody may comprise 3 CDRs and have an antigen binding site which is equivalent to that of a domain antibody (dAb). The antibody may comprise 6 CDRs and have an antigen binding site which is equivalent to that of a classical antibody molecule. The remainder of the polypeptide may be any sequence which provides a suitable scaffold for the antigen binding site and displays it in an appropriate manner for it to bind the antigen. The antibody may be a whole immunoglobulin molecule or a part thereof such as a Fab, F(ab)’2, Fv, single chain Fv (ScFv) fragment or Nanobody. The antibody may be a conjugate of the antibody and another agent or antibody, for example the antibody may be conjugated to a polymer (e.g. PEG), toxin or label. The antibody may be a bifunctional antibody. The antibody may be non-human, chimeric, humanised or fully human.

Methods of treatment The invention also provides a method for treating Rheumatoid Arthritis (RA), the method comprising administering to a patient an effective amount of a B cell targeted therapy, wherein the patient is determined to be susceptible to treatment with a B cell targeted therapy by the method of the invention.

In some embodiments, the B cell targeted therapy is B cell depletion therapy.

In preferred embodiments, the B cell targeted therapy is rituximab.

The invention also provides a method for treating Rheumatoid Arthritis (RA), the method comprising administering to a patient an effective amount of an agent that downregulates IL- 6 mediated signalling, wherein the patient is determined to be susceptible to treatment with an agent that downregulates IL-6 mediated signalling by the method of the invention.

Kits

The present invention also provides a kit suitable for performing the method as disclosed herein. In particular, the kit may comprise reagents suitable for detecting the biomarkers disclosed herein, or a biomarker combination as disclosed herein.

The kit may also comprise instructions for use.

The kit may also comprise a B cell targeted therapy or an agent that downregulates IL-6 mediated signalling. The skilled person will understand that they can combine all features of the invention disclosed herein without departing from the scope of the invention as disclosed.

Preferred features and embodiments of the invention will now be described by way of nonlimiting examples.

The practice of the present invention will employ, unless otherwise indicated, conventional techniques of chemistry, biochemistry, molecular biology, microbiology and immunology, which are within the capabilities of a person of ordinary skill in the art. Such techniques are explained in the literature. See, for example, Sambrook, J., Fritsch, E.F. and Maniatis, T. (1989) Molecular Cloning: A Laboratory Manual, 2nd Edition, Cold Spring Harbor Laboratory Press; Ausubel, F.M. et al. (1995 and periodic supplements) Current Protocols in Molecular Biology, Ch. 9, 13 and 16, John Wiley & Sons; Roe, B., Crabtree, J. and Kahn, A. (1996) DNA Isolation and Sequencing: Essential Techniques, John Wiley & Sons; Polak, J.M. and McGee, J.O’D. (1990) In Situ Hybridization: Principles and Practice, Oxford University Press; Gait, M.J. (1984) Oligonucleotide Synthesis: A Practical Approach, IRL Press; and Lilley, D.M. and Dahlberg, J.E. (1992) Methods in Enzymology: DNA Structures Part A: Synthesis and Physical Analysis of DNA, Academic Press. Each of these general texts is herein incorporated by reference.

EXAMPLES

EXAMPLE 1

RESULTS

Histopathology and in silico deconvolution identify cell lineages associated with response/non- response to rituximab and tocilizumab

To identify potential associations with response/non-response to rituximab and tocilizumab against the primary end-point (CDAI>50%), we initially assessed the infiltration of main immune cell lineages in synovial biopsies at baseline by immunohistochemistry (IHC), using IHC semi-quantitative scores (Fig. 7a-b). It can be seen that immune cell types did not differ between responders and non-responders to the individual medications (Fig. 7c). However, when we compared (Fig.1a) the two medications in patients stratified according to histological patterns (pathotypes) that we previously described the diffuse-myeloid (DM) pathotype, which is B-cell poor but rich in macrophages, was associated with a significantly higher response to tocilizumab: 13/16 (81 %) tocilizumab vs 7/20 (35%) rituximab p=0.008, OR 7.53, 95%CI 1.4- 55.7. In contrast, patients with the lympho-myeloid (LM) pathotype and the fibroid/pauci- immune (FPI) pathotype showed a similar response to the two medications. To complement the IHC analysis and more precisely dissect the specific role of synovial immune cells, we used an in silico approach based on lineage-specific transcript analysis (MCP-counter) to deconvolute immune cells from the bulk synovial RNA-Seq (Fig.1b). When comparing responders and non-responders to the individual medications, we found significantly higher levels of CD8+T-cells in responders to rituximab and, in line with the enhanced response to tocilizumab in the myeloid pathotype determined by IHC, higher macrophage-monocytes and myeloid dendritic cells (mDC) transcripts in responders to tocilizumab (Fig.1c).

Moreover, when we stratified patients in B-cell poor/rich by MCP-counter scores and we compared the response to the two medications, consistently with the primary results of the trial, B-cell poor patients showed significantly higher response rates to tocilizumab (Fig.1 d), while no difference was found in the B-cell rich patients. Similar results were observed for T- cells (Fig.ld). In contrast, macrophage and myeloid dendritic cell (mDC) rich individuals showed higher response to tocilizumab, with a significant difference observed in the latter (Fig.le). Strikingly, combined scores for lymphoid and myeloid cells (Fig.1 f) demonstrated that patients poor in B-cells but rich in macrophages/mDC had a significantly higher response to tocilizumab (77% responders to tocilizumab vs 14% responders to rituximab, p=0.017, OR 16.48, 95%CI 1.29-1000.5), indicating that complex multicellular networks, rather than individual cell lineages, are relevant for specific therapeutic responses. Furthermore, by analysing disease activity over time using two-way ANOVA for repeated measures, we found a statistically significant interaction effect between treatments and time assessed longitudinally (0-16 weeks) in B cell poor patients (p=0.003), T cell poor (p=0.022) (Fig.1g), mDC rich (p=0.029) (Fig. 1h) and B cell poor/Macrophages-mDC rich patients (p=0.006) (Fig. 1 i). Comparing disease activity at individual timepoints showed significantly lower CDAI at weeks 6, 12 and 16, was found in tocilizumab treated patients who were also B cell poor and macrophage/mDC rich (Fig.1 i). Overall, these results point to myeloid cell infiltration in synovia as one of the key factors explaining the enhanced response to tocilizumab in B-cell poor patients.

Unsupervised clustering defines treatment response differences linked to drug-target genes and immune cell infiltration

Next, we used RNA-Seq analyses to explore the relationship of multiple genes/pathway interactions with response to treatment. To look for evidence of underlying subgroup structure within the data, we applied principal component analysis (PCA) to the synovial RNA-Seq whole dataset. We observed that PC1 , 3 and 4 correlated with the degree of inflammatory cell infiltration in synovial biopsies, and PC1 and 3 showed association with histological pathotypes primarily separating the lymphoid and fibroid pathotype (Fig. 8a and b). All the first six PCs did not associate with demographics, treatment and its associated response or clinical disease features such as disease activity or anti-CCP antibody status (Fig. 8a, c and d).

Applying unsupervised clustering on the whole tissue baseline synovial RNA-Seq data (Fig.2a) using a Monte Carlo reference-based consensus clustering algorithm (M3C), 71 % of rituximab responders (n=24) were found in cluster 1 compared to only 29% of responders (n=10) in cluster 2 (p=0.0004, Fig.2a, left). Genes relevant for B-cell biology (MS4A1 , CD79A, CD79B, PIK3CA, BTK and SYK) were significantly higher in cluster 1 in rituximab patients (Fig. 8e, Fig. 9a). This was also reflected in a significant upregulation of the B-cell gene module S136 derived from Weighted Gene Correlation Network Analysis (WGCNA), together with the upregulation of the pro-inflammatory M1 macrophage module-(S39) and down-regulation of the fibroblast module-(S115) (Fig. 8e).

On the other hand, clustering of tocilizumab-treated patients was less distinctive, with 46% responders (n=21) in cluster 1 and 54% in cluster 2 (n=25) (Fig.2a, right). However, cluster 1 was significantly associated with IL-6 pathway genes (IL6R, IL6, IL6ST, JAK1 , JAK2, STAT3) (Fig. 8f, Fig. 9b), together with a similar upregulation of B-cell and M1 macrophages modules and downregulation of fibroblast modules. In keeping with the increase of immune cell-related modules in cluster 1 for both treatments, when looking at the semi-quantitative IHC scores for all patients, synovial immune cells were also significantly higher in cluster 1 (Fig. 8g). This indicates that the level and type of immune cell infiltration is causative of the related gene expression in cluster 1 , as documented also by the loss of significance when adjusting DEG analysis between consensus cluster 1 and 2 for immune cell content using principal component 1 as a covariate (Fig. 9c, d) due to the high correlation of PC1 , histology markers and immune cell-related genes (Fig. 9e).

Differential gene and module expression analysis defines treatment response signatures linked to inflammatory/immune cell genes

Next, we performed differentially expressed gene (DEG) analysis from synovial RNA-Seq as a complementary approach to identify genes associated with response to each medication. To this aim, we analysed all patients treated with either rituximab or tocilizumab throughout the trial, including non-responders who, as per trial protocol, switched to the alternative medication at week 16, as shown in the trial overview in Fig. 10. A total of 6625 genes were significantly different (FDR < 0.05) in rituximab responders compared to non-responders (Fig.2b), and 85 in tocilizumab (Fig.2c). Genes upregulated in the synovial tissue of rituximab responders included members of the immunoglobulin (Ig) superfamily (e.g. CD86, IGHV3- 64D), and a wealth of additional leukocyte-related genes (e.g. CXCL2, CCL2, MS4A7, IL6). Non-response to rituximab, on the other hand, was associated with complement genes (e.g. CR2), bone morphogenic proteins (e.g. BMP2), fibroblast related genes (e.g. FGFR3) and several Hox genes (e.g. HOXB1). Interestingly, lymphocyte and Ig genes were also upregulated in the synovial tissue of tocilizumab responders (e.g. LY6D and IGKV1 D-43). Finally, both non-responder groups showed an upregulation of extracellular matrix (ECM) associated genes, including integrin-binding sialoprotein (IBSP, a major bone matrix protein), aggrecan (ACAN), and collagen gene COL2A1 , genes linked to tissue remodelling, cell infiltration and cell-cell interaction. In contrast to the above unsupervised consensus clustering, following adjustment for immune cell infiltration by PC1 , DEG analysis largely retains its differential gene expression and in the case of tocilizumab even increases the number of identified DEGs (Fig.2d for RTX and Fig.2e for TOC). This highlights that DEG analysis provides an additional dimension to the inflammatory cell infiltrate alone that differentiates responders from non-responders. Of note, covariates such as age, gender and ethnicity did not cause major changes in DEGs.

To investigate the functional role of the above genes, we next used Quantitative Set Analysis for Gene Expression (QuSAGE) gene modules and synovium-specific WGCNA-based gene modules annotated against single-cell RNA-Seq of RA synovium cells (Fig.2f&g). Rituximab responders had several antigen presentation and lymphocyte modules significantly increased (FDR < 0.05), including T-cell signalling, surface activation and B-cells, togetherwith interferon signalling genes. On the other hand, Hox gene, fibroblast and ECM modules were associated with non-response to rituximab (Fig.2f).

In tocilizumab-treated patients, myeloid cell cytokine module, PPAR signalling and metabolic pathways were upregulated in responders (Fig.2g). Notably, although none of the modules was significantly modulated in non-responders to tocilizumab, we observed an overlap in non- responder signatures between the two medications: namely, fibroblast modules were detected in non-responders to both rituximab and tocilizumab, suggesting a common non-response signature, consistent with the possibility of a treatment-resistant disease signature linked to refractory RA.

Refractory disease is linked to a stromal/fibroblast signature

To further explore the hypothesis of a refractory signature, following the switch at 16 weeks as per trial protocol (Fig. 10), we identified 3 groups: patients who failed both medications (multi-drug resistant/refractory), patients who responded to rituximab after failing tocilizumab (pro-RTX), and patients who responded to tocilizumab after failing rituximab (pro-TOC) (Fig.3a). By comparing these three populations, we identified 1980 genes upregulated both in pro-RTX and pro-TOC patients, while 175 genes were exclusive to the pro-RTX group and 306 were exclusive to pro-TOC (Fig.3b). Among the genes upregulated in responders to both medications (Fig.3c side view and 3d top view), we found a mixture of lymphoid genes (e.g. CD3D, CD8A, CD8B, CD52, CD69, CD72), myeloid genes (e.g. CD33, CD86) and a multitude of cytokine genes (IL7, IL10, IL10RB, IL18, IL21 R).

Patients who responded exclusively to rituximab (pro-RTX) had upregulation of chemokines (CCL8, CCL2), Fc receptor (FCGR3A) and lymphocyte-related genes (CD226, CD58). On the other hand, patients who exclusively responded to tocilizumab (pro-TOC), showed an upregulation of both lymphocyte and myeloid lineage genes (e.g. CD1 E, CD79B, TNFSF12 also known as TWEAK, PDK1). Modules associated with response to either drug encompassed antigen presentation, dendritic cell, macrophage and plasma cell infiltration (Fig.3e). In proximity to the pro-RTX axis, we found the CD8 and Tph T-cell module, while tolllike receptor signalling and Chemokine and cytokine signalling of macrophages modules were specifically associated with rituximab responders. By comparison, the pro-TOC group had upregulation of T-cells, plasma cells and TNF receptor superfamily gene modules.

Notably, 1277 significant genes were identified as unique to the multidrug-resistant refractory patients (Fig.3c), who did not respond to both rituximab and tocilizumab within the study and, as per trial inclusion criteria, had already failed conventional synthetic disease-modifying antirheumatic drugs (csDMARDs) and at least one anti-TNF. Thus, by study-end, this group of patients display resistance to three biologic therapies, notably, targeting three distinct immunological pathways: TNF, B-cells and IL-6R. The genes upregulated in refractory patients included the fibroblast growth factor (FGF) family (e.g. FGF2, FGFR1 and FGFRL1), homeobox (HOX) family (e.g. HOXA13), NOTCH family (NOTCH 1-3) and cell adhesion molecules including 26 cadherin genes and multiple collagen genes (e.g, COL11A2). A large proportion of the genes identified in refractory patients can be linked to fibroblasts and cell interaction with the ECM. Accordingly, modular analysis (Fig.3e) highlighted Hox genes, chondrocyte differentiation and fibroblast modules.

In line with the above molecular signatures, synovial immune cells assessed by histology, in particular CD3 T-cells, CD79a and CD138 plasma cells and the total ImmuneScore, were significantly lower at baseline in refractory patients compared to responders to any of the two medications (Fig.3f). Accordingly, in silico deconvolution showed significantly lower levels of immune cells (CD8+T-cells, monocytes and myeloid dendritic cells) and an increase in endothelial cells and neutrophils in refractory patients compared to responders (Fig.3g). Next, we aimed to better characterize the association of fibroblast and macrophage genes in synovia with treatment response and multi-drug resistance. To this aim, we complemented the MCP-counter deconvolution analysis by examining the transcriptome for enrichment in synovium-specific fibroblast and macrophage subtypes using gene modules derived from the recently described single-cell (sc)RNA-Seq from RA synovial tissue (Zhang et al. (2019) Nat. Immunol. 20: 928-942). Strikingly, (Fig.3h) the signature for HLA-DRAhi sublining fibroblasts (SC-F2), considered a pro-inflammatory subset and associated with leukocyte-rich synovial infiltration in RA, was significantly more expressed at baseline in responders (p=0.027), as opposed to CD34+sublining fibroblasts (SC-F1) and, in particular, the newly described DKK3+sublining fibroblasts (SC-F3) that were increased in refractory patients (p=0.036 and 0.00055, respectively). On the other hand, signatures of C1QA+ (SC-M3) and IFN-activated (SC-M4) macrophages were significantly increased in responders (Fig ,3h , p=0.034 and 0.01 respectively).

To orthogonally confirm these findings at protein level, we assessed the presence of DKK3+ fibroblasts and SPP1+ macrophages by immunofluorescence. In line with the transcriptomic signature, DKK3 was expressed in the synovial lining and sublining of refractory patients, including in areas rich in CD90+ fibroblasts (Figure 3i and Fig. 11). Interestingly, DKK3 was also expressed by CD45+ lymphocytes, which is in line with its expression in CD8+ T cells. On the other hand, SPP1+ macrophages were expressed in the synovial lining layer of responders (Figure 3i and Fig. 11).

Together, these results show that baseline histological and molecular signatures identify response to biologic therapies, including specific response signatures linked to the individual drugs. Importantly, while response was associated with myeloid cells, refractory patients resistant to multiple biologies displayed a specific baseline signature associated with fibroblasts and ECM.

Digital Spatial Profiling of refractory RA reveals differential expression of genes and cell types in synovial regions in association to treatment response

Since both fibroblasts and macrophages are known to exhibit positional identity relevant to their contribution to the pathogenesis of RA, next, we used digital spatial profiling (DSP) to better characterise the spatial positioning of immune and stromal cell signatures and determine their association with treatment response/resistance. To this aim, we applied GeoMx DSP (NanoString) that enables the combination of protein lineage markers to define Region of Interests (ROIs) with whole-transcriptomic spatial RNA expression (Fig.4a). First, we compared gene expression in responders and refractory patients across all Region of interest (ROIs): lining/superficial sublining, deep sublining and lymphoid aggregates (Fig.4b). Consistently with the above bulk RNA-Seq modules and protein expression, a number of fibroblast genes, including the ones related to the DKK3+ subset (DKK3, PRELP, OGN, CAM1 KD), were significantly higher in refractory patients, while macrophage genes such SPP1 and ATF3 were significantly higher in responders (Fig.4c).

When looking at individual ROIs, we found several differentially expressed genes in each synovial region (Fig.4d). More specifically, 41 genes were exclusively modulated in lining/superficial sublining, 146 in the sublining and 371 in lymphoid aggregates (Fig.4e). For example, FAP (Fibroblast Activation Protein) was significantly increased in the deep sublining of refractory patients. SPP1 , on the other hand, was significantly increased in the lining of responders, while CD24 (a B-cell related gene) was significantly higher in the lymphoid aggregates of responders (Fig .4f) . Multiple genes were modulated in unison across all regions: for example, the chemokine CCL13 was significantly higher in refractory patients, while the metalloprotease ADAM 15 was significantly higher in responders (Fig.4f).

Longitudinal pre- and post-treatment histopathological and molecular analyses identify specific immunological signatures and pathways differentiating drug-treatment and drugresponse over time

To explore the differential effects of each drug on synovial immune cell infiltration and gene expression overtime, we compared paired samples at baseline and 16 weeks (Fig. 12a). First, we analysed immune cells by histology in matched pre- and post-treatment biopsies, showing a significant reduction of synovial CD20+ total B-cells [-1.53 (-81%), p<0.001], CD79a+ B-cells [-0.87 (-49%) p<0.001] and CD138+ plasma-cells [-0.76 £ (-45%) p<0.05] in rituximab-treated patients, in line with the mechanism of action of rituximab in targeting CD20+B-cells (Table 4 and Fig. 12b). Conversely, tocilizumab treated patients had a significant reduction of CD68+sub-lining macrophages [-1.04 (-54%) p<0.05] but not B-lineage cells (Table 4 and Fig. 12b). Analysis of covariance (ANCOVA) showed a significant treatment effect for CD20, CD79a and CD68SL, with a significantly higher reduction of B-cells and CD79a in rituximab- treated patients and a significantly higher reduction of macrophages in tocilizumab-treated patients (Fig. 12c).

Notably, when patients were stratified according to 16 weeks response, a significant reduction of CD138 and CD79a was only observed in rituximab responders, while a significant reduction of CD68SL macrophages was only observed in responders to tocilizumab (Fig. 12b). Altogether, this indicates that the reduction of B and synovial plasma cells and macrophages are associated with response to rituximab and tocilizumab, respectively. Also notably, similar results were obtained when comparing immune cell signatures deconvoluted by MCP counter in matched samples at baseline and 16 weeks. Namely, rituximab-treated patients showed a significant reduction of B-cells, T-cells and monocyte/macrophages, while tocilizumab-treated patients showed a significant reduction of monocyte/macrophages, T-cells, but also neutrophils, myeloid dendritic cells and, interestingly, an increase in fibroblast signature (Fig. 12d and e). This suggests that both medications have an effect on immune cells, but tocilizumab can potentially also affect stromal cells.

T o further dissect the molecular signatures longitudinally, as the most widely used mainstream differential gene expression analysis tools edgeR, DESeq2 and limma voom are all unable to fit mixed-effects linear models, we developed a custom-made analytical method to fit negative binomial mixed-effects models at individual gene level. Using mixed-effect models we compared gene expression differences over time in paired biopsies at baseline and 16 weeks in 44 patients following rituximab or tocilizumab treatment (Fig. 12a). Fig.5a compares the fold change in gene expression over time across the transcriptome between rituximab and tocilizumab treated individuals on the x and y axes respectively. A substantial number of genes (7316 genes shown in green) were significantly up or downregulated by both drugs, with genes equally affected by each drug lying along the line of identity in this plot. Also, we observed that 345 genes (Fig.5a shown in blue/yellow) were found to be differentially affected by either drug based on significance (FDR < 0.05) of the interaction term time:medication. Genes that were more strongly influenced by rituximab therapy are shown in blue, while genes for which tocilizumab had the greater effect are shown in yellow. Of note, we observed that MS4A1 (which encodes CD20), PAX5, BLK and immunoglobulin chain genes such as IGLV4-3 were significantly downregulated in response to rituximab, consistent with its B-cell depletion mechanism and the results by histology (Fig. 12b), while tocilizumab had significantly less effect on these genes. Thus, consistently with their mechanism of action, rituximab is more potent at downregulating B-cell specific genes in synovial tissue than tocilizumab, while tocilizumab is most effective on IL6 transcripts. Moreover, differences were observed on metalloproteinases expression, which were more strongly reduced following tocilizumab therapy. In addition, differential metalloproteinase expression was observed between the two drugs, with tocilizumab therapy strongly reducing MMP10, expressed in chondrocytes and synovial fibroblasts, but increasing MMP7, whereas rituximab caused a moderate reduction in both. Comparison of the generalised mixed model using negative binomial distribution against Gaussian linear mixed-effects model on log count data showed very similar results with strong correlation between P values generated using either distribution (Fig. 13a-c). QQ plots suggested that the negative binomial mixed model showed greater power for identifying significant effects (Fig. 13d-f).

The mixed-effects model allowed us to examine differences in change in gene expression after therapy between responders and non-responders for each drug (Fig.5c). Rituximab had a general effect on 1796 genes (Fig.5c, shown in green), with 349 genes (Fig.5c, shown in blue/red) showing significant (FDR < 0.05) differential expression change over time between responders (blue) and non-responders (red). Rituximab responders showed a greater decrease in SAA1 and SAA2 (serum amyloid proteins 1 and 2), as well as greater drops in immunoglobulin chain genes IGHV3-64D and IGKV1-13 suggesting that a drop in antibodysecreting B-cells is associated with response to rituximab (Fig.5d). The chemokine CXCL11 , the citrullination enzyme PADI2 (peptidyl arginine deiminase 2), HP, which encodes haptoglobin, and the key Th17 and mucosal-associated invariant T (MAIT) cell transcriptional regulator RORgamma (RORC) were also more strongly modulated in rituximab responders. Both S100A1 , which is involved in altered chondrocyte differentiation, FOXD3, a transcription factor involved in mesenchymal stem cell differentiation, decreased in responders over time but increased in non-responders, suggesting that interplay between the B-cell milieu and chondrocyte differentiation may be important in determining response to rituximab and the effect on structural damage.

Tocilizumab treatment resulted in up or downregulation of 1609 genes (Fig.5e in green) with an additional 136 genes (Fig.5e in yellow/red) showing differential change in gene expression between responders (yellow) and non-responders (red). Importantly reduction in pro-lymphoid follicle development genes lymphotoxin A (LTA), complement receptor 2 (CR2), the chemokine receptor and lymphoid-tissue resident dendritic cell marker XCR124 and prolactin (CLEC17A), expressed on specialized, proliferating germinal centre B-cells, augured response to tocilizumab.

To further investigate the pathway modulation induced by treatments with rituximab and tocilizumab, the genes identified in the longitudinal mixed-effects model analysis were used to perform Gene Ontology/pathway enrichment analysis using ClueGO. Treatment with rituximab, as expected, induced a significant downregulation of B-cell receptor signalling and associated pathways, including Igs and humoral immune response (adjusted p < 0.05, Fig.5g). Interestingly, bone resorption and remodelling pathways were significantly downregulated in rituximab treated patients, providing an indication of its mechanism in the reduction of bone erosions. When stratifying patients according to response (Fig.5h), the significant reduction in B-cell signalling and related Ig pathways observed in all rituximab treated patients was also observed in non-responders due to B cell depletion. Responders to rituximab also showed significant downregulation of T-cell receptor complex, lymphocyte chemotaxis and migration, chemokines and IL-1 related pathways (Fig.5h), suggesting that response to rituximab is linked to additional immunomodulation in addition to direct B-cell depletion. As for tocilizumab, there were no significant pathways in all tocilizumab-treated patients, but responders had a significant modulation of humoral immune response, Ig related pathways and B-cell-mediated immunity and complement activation in line also with the known effect of IL-6 with B-cell growth (Fig.5i). Altogether, longitudinal analyses of matched pre and post-treatment biopsies indicate that specific biological changes predict response to the individual treatments.

Machine learning classifier models predict treatment response to rituximab and tocilizumab

To establish in an unbiased fashion the ability of synovial RNA-Seq to predict treatment response to rituximab and tocilizumab or refractory state, we developed machine learning predictive models with the dataset partitioned for training and testing using 10x10-fold nested cross-validation (CV). Model hyperparameters were tuned by inner 10-fold cross-validation, with model accuracy determined in separate outer cross-validation folds to give an unbiased estimate of model accuracy. For comparison, model accuracy in the left-out inner CV folds was also compared, although these are known to give a biased estimate of model prediction ability. As shown schematically in Fig.6a, multiple predictive models were tested including elastic net penalized regression (glmnet), support vector machine (SVM) with radial or polynomial kernel, flexible discriminant analysis including penalized discriminant analysis (PDA) and mixture discriminant analysis (MDA), random forest (RF) and gradient boosted machine (GBM). In addition, fully nested within the inner nested-CV, three types of parameter filters were tested: none, univariate filter and recursive feature elimination (RFE). Of these, RFE frequently failed to reduce variable selection and was thus abandoned when fitting final models. Table 8 shows the performance of models used to predict i) rituximab response, ii) tocilizumab response and iii) refractory state (no response to both drugs) ordered by area under receiver operating characteristic (ROC) curve (AUC). The tuning parameters for the final model were the mean over all 10 outer folds (Table 9). The final model was trained on the entire data set to extract the variable importance (Fig.6a iv, Table 10)

The optimal rituximab response predictive model was a 40-gene elastic net regression model which produced an AUC of 0.744 (Fig.6b). Of note, no clinical or histological features were selected by the final model (Table 10j). In comparison, predictive models built using clinical and histology parameters alone performed poorly (Fig. 14b). The optimal model for predicting response to tocilizumab, which also used elastic net regression in a 39-gene model, achieved an AUC of 0.681 (Fig.6b), which was higher than the AUC of 0.582 achieved by clinical and histology data alone (Fig. 14b). The optimal model for predicting the refractory state was a 53- gene GBM model which reached an AUC of 0.686. AUC values in the left-out inner CV folds were consistent with the AUC results in the true test folds. The number of genes required for all three models could have been a limitation of these models. However, multiple genes were shared across models, so that only 85 genes are required to build all three prediction models and 32 genes were shared between at least one model (Fig. 14a, Table 10). It was notable that 15 genes were shared across all three prediction models including the top six genes in the rituximab prediction model which was the most accurate model, suggesting that a universal subset of genes may be linked to future response outcome, while additional genes are required to hone prediction for individual drugs and the refractory state. Key shared prediction genes included: dendritic cell marker XCR1 ;24 chemokine CXCL14; SAA2 (serum amyloid A2); immunoglobulin heavy chain gene IGHV7-4-1 reflecting the presence of plasma cells; and collagen glycosylation enzyme COLGALT2. The rituximab and tocilizumab predictive models shared a further seven genes, while the refractory predictive model was the most exclusive with 28 unique genes. The machine learning models spontaneously selected key genes involved in RA pathogenesis and cartilage biology including chondrocyte gene NKX3- 2, sublining fibroblast marker DKK3, collagen-binding scavenger receptor CD36, CD8 T cell gene PI16, chemokines CXCL2 and CCL4L2, TNFRSF11 B (osteoprotegerin), CHAD (chondroadherin), WIF1 (WNT inhibitory factor 1) and the citrullination enzyme PADI4 (PAD4). Although the clinical parameter and histology based predictive models (Fig. 14) were much weaker in prediction ability compared to gene expression models, we observed that CD68 sublining histology score had the highest variable importance as a predictor of tocilizumab response in keeping with our earlier results regarding sublining macrophages, while age, which has been identified as a predictor of tocilizumab response, was a feature of both tocilizumab response and refractory clinical predictive models. These results show that predictive models can harness molecular information from synovial biopsies at baseline prior to treatment and thus are of potential clinical utility for predicting response to therapy.

DISCUSSION

Our study provides an in-depth molecular and histological profiling of the joint tissue from the first biopsy-driven randomised-clinical-trial in RA (R4RA). This has afforded major novel insights in the cellular and molecular pathways underpinning the diverse treatment response to two commonly used targeted biologic therapies. It has also enhanced our understanding of mechanisms of multi-drug resistance and led to the identification of specific histopathological/molecular signatures and the development of machine learning classifier modules predictive of treatment response. Building on previous studies by our group in treatment-naive early RA patients: the Pathobiology of Early Arthritis Cohort (PEAC), which showed that infiltrating immune cells were the strongest drivers of heterogeneity of synovial gene expression and strongly correlated with clinical and radiographic measures of concurrent disease activity, we analysed and compared whole tissue RNA-Seq and histology data in the R4RA biopsies both at baseline and longitudinally at 16 weeks primary end-point CDAI>50%. Using both conventional histology and in silico deconvolution, we observed that in addition to CD20+B-cells, CD3+ and CD8+T-cells were also associated with pro-rituximab response, while myeloid cells were consistently associated with response to tocilizumab. The importance of synovial macrophages as a predictor of response to anti-TNF is well established, however, to our knowledge, this is the first report that within the synovial B-cell poor group, only patients with high levels of infiltrating myeloid cells (macrophages or mDCs) responded well to tocilizumab. Strikingly, the combined score: B-cell-poor/macrophages-mDC-rich defined a subset of patients displaying major response rates differences, namely only 14% response to rituximab versus 77% to tocilizumab.

Through advanced molecular analyses, we next elucidated the mechanisms of treatment response heterogeneity to the individual drugs, focusing both on response and non- response/resistance mechanisms. We applied unsupervised clustering using M3C and WGCNA, followed by DEG analysis. Interestingly, as predicted, genes associated with response were linked to the cognate drug targets, namely, genes associated with enhanced response with rituximab included genes relevant for B-cell biology such as PIK3CA, BTK and SYK, together with members of the immunoglobulin (Ig) superfamily and a number of chemokines and leukocyte related genes, but also IL6, a known B-cell growth factor. Correspondingly, tocilizumab response was significantly associated with IL6 pathway genes, but also lymphocyte and Ig genes.

The functional role of these associations was explored by QuSAGE gene modules and synovium specific WGCNA-based gene modules that demonstrated in rituximab responders significantly increased (FDR < 0.05) antigen presentation and expression of lymphocyte- related modules, together with interferon signalling genes, in keeping with previous reports linking increased type I interferon response gene expression in blood with enhanced response to rituximab. In tocilizumab responders, in line with the prominent role of myeloid cells identified by histopathology, myeloid cell cytokine module was strongly upregulated together with PPAR signalling and metabolic pathways. In contrast, non-response appeared to be associated with an increased number of shared pathways to both drugs, though more genes and related pathways were found to be linked with rituximab compared to tocilizumab. For example, both non-responder groups showed an upregulation of ECM associated genes, including bone matrix protein I BSP and aggrecan (ACAN), and notable complement genes such as CR2. By QuSAGE and WGCNA analysis, Hox gene and ECM modules were more associated with non-response to rituximab than nonresponse to tocilizumab. The common non-response signature corresponds to the fibroid pauci-immune pathotype, also demonstrated to be associated with poor response to both synthetic-DMARDs and TNF-inhibitors, supporting the concept that the fibroid pauci-immune phenotype is linked with treatment-resistant disease.

This notion was greatly strengthened by analysing a group of patients who failed to respond to both drugs capitalising on the cross-over R4RA study design that permitted the switch to the alternative biologic for patients deemed to be non-responders at 16 weeks. Since these patients had also not adequately responded to csDMARDs and TNF-inhibitors, as per trial entry criteria, they meet the definition of multi-drug resistant/refractory RA. This analysis identified for the first time a true refractory signature including over 1000 genes associated with multi-drug resistance with strong representation from fibroblast and collagen genes consistent with a pro-fibroblastic phenotype. As recent studies, using single-cell RNA-Seq on RA synovium identified specific fibroblast subsets with critical roles in the pathogenesis of RA, it was important to determine which fibroblast subtypes were involved in such multidrug resistance. Using modular analysis based on genes specific for single-cell RNA-Seq of fibroblast subsets and orthogonal validation by IHC, we identified the DKK3+ fibroblasts as the synovial fibroblast subtype most strongly associated with the refractory phenotype, while IFN-activated (SC-M4) macrophages were associated with treatment response. Furthermore, Digital Spatial Profiling of refractory RA revealed differentially expressed genes in synovial regions in association to treatment response, suggesting that the spatial organization of immune infiltrates is relevant for determining treatment response/resistance.

Another significant contribution of this study is the longitudinal analysis of the synovial tissue in matched pre-/post-treatment biopsies available at 16 weeks in 40% of the patients, which enabled to investigate the drug effect directly on synovial pathology and gene expression. Histological analysis showed that rituximab reduced synovial CD20+ B-cells in both responders and non-responders but, notably, non-responders failed to show a significant reduction in CD79a+B-cells and plasma cells in comparison to responders. This suggests that treatment with rituximab reduces synovial B-cells in all patients, but this is not sufficient to translate into a definite improvement of disease activity. To achieve a clinically relevant response, a broader/deeper impact on differentiated CD79a+ plasma-blasts and CD138+ plasma-cells over and above CD20+ B-cell depletion appears to be required. These observations are in agreement with some of the findings of previous observational studies assessing longitudinal synovial samples following treatment with rituximab. For example, in line with our findings on CD79a cells and plasma cells, it has previously been found that the changes in plasma cells at week 16 differ significantly between responders and nonresponders. However, other studies showed a variable depletion of B cells and plasma cells in the synovia and an unclear association with treatment response, nonetheless, the small sample size and the analysis of the biopsy at different time points (4, 8, 12 or 16 weeks) makes it difficult to make comparisons and draw conclusions from those studies. In our study, the repeated synovial biopsy was performed at 16 weeks following one cycle of rituximab (2x 1g infusion), thus, it is plausible that a repeated biopsy at the time of the second infusion at 6 months, as per standard protocol, could have detected a wider effect on B cell lineages linked to a clinical response at a later timepoint. However, insufficient biopsies were performed at that later time-point to be informative.

Following tocilizumab therapy, in contrast to responders, non-responders were characterised by a failure to reduce sublining macrophage levels, which is in line with previous literature, suggesting that the reduction of sublining macrophages is associated with response to treatment.

To assess the relationship of gene expression changes and treatment response following biologic treatment, we developed a novel pipeline for mixed-model analysis of RNA-Seq data. This has the major advantage that, by taking into account random effects between individuals, it revealed patterns of change in gene expression over time that were not detectable by previous standard analytical pipelines, while interaction analysis allowed us to identify genes that were significantly differentially affected by each drug specifically. For example, the biological differences in synovial gene expression following treatment with each drug are consistent with the specific pathways targeted by each individual drug, B-cell depletion and IL-6 receptor blockade, but also revealed unexpected differences such as differential changes in metalloproteinase gene expression in response to each drug. Other genes more strongly modulated by tocilizumab than rituximab included, as expected, IL-6 and IL-8 production, but also IRGM, which regulates inflammasome activation and P116, which is expressed in connective tissue homing CD8+T cells. Interaction analysis was particularly informative when comparing responders and non-responders in each drug cohort. This showed that rituximab responders demonstrated a greater decrease in serum amyloid proteins, immunoglobulin chain genes, the chemokine CXCL11 , the citrullination enzyme PADI2 and transcriptional regulator RORgamma, whereas chondrocyte differentiation genes S100A1 and FOXD3 increased in non-responders over time. In tocilizumab responders, a greater reduction in prolymphoid follicle development genes was observed in keeping with the important role of IL-6 as a B-cell growth factor driving in situ ectopic lymphoid structure development within inflamed tissues. Pathway analysis was largely consistent with these results.

With regard to the interpretation of the results related to the second biopsy, it is important to point out that it was an optional procedure and, thus, at risk of potential selection bias. Although there were no major differences between treatments (Table 5) and no significant differences in baseline demographics between patients who did or did not have a second biopsy, there were lower response rates in patients who underwent a repeated biopsy (Table 6). This is to be expected, as responder patients, being clinically well, would have been less likely to consent to a second biopsy. Additionally, repeating a biopsy in responders with little residual synovitis would have been challenging and thus less likely to have been performed.

In translational terms, the importance of molecular studies is measured on their ability to enhance disease understanding but also on their clinical impact. Thus, in order to determine the predictive value of deep molecular characterization in foretelling treatment response, we applied a number of successful machine learning methodologies to the R4RA RNA-Seq data resulting in the selection of models with high AUC for predicting CDAI50% response as tested by nested cross-validation. To reduce the parameter space for predictive models, genes were restricted to protein-coding genes and limited based on variance and collinearity. The purpose of testing multiple models was to determine whether non-linear decision boundaries as used by SVM, MDA, PDA and tree-based prediction algorithms such as GBM could outperform linear regression based prediction systems such as glmnet. In practice, for prediction of response to each drug individually, elastic net regression performed best. However, a GBM model was the superior refractory model, consistent with the notion that a model with greater complexity of its decision boundary is required to optimally predict the refractory state. Of relevance to future clinical practice, gene expression models were significantly superior to models built using clinical and histological data alone. For rituximab response, a 40-gene elastic net predictive model demonstrated an AUC of 0.744, for tocilizumab response a 39- gene elastic net model had AUC 0.681 and for refractory patients the GBM predictive model had AUC 0.686. Furthermore, each of these models selected multiple genes of biological relevance to synovial tissue inflammatory and repair responses as well as bone and cartilage biology. Key prediction genes shared between all three models included: XCR1 which is a marker of DC1 migratory dendritic cell subset; chemokine CXCL14; acute phase reactant SAA2 (serum amyloid A2); and IGHV7-4-1 whose presence likely reflects specific tissue- resident plasma cell populations. Other noteworthy predictor genes included the sublining fibroblast marker DKK3, scavenger receptor CD36 involved in collagen binding and macrophage phagocytosis, P116 which is expressed in tissue-resident CD8+T cells. The refractory state model, which contained the largest number of unique genes, included several genes linked to the fibroid pathotype such as TNFRSF11 B which encodes the osteoclast negative regulator osteoprotegerin, the chondrocyte adhesion mediator CHAD (chondroadherin), WIF1 (WNT inhibitory factor 1) but also the citrullination enzyme PADI4 (PAD4) consistent with a role of persistent tissue destruction and remodelling in the refractory RA state.

In summary, these data strongly support the existence of specific biological states in the disease tissue of individual patients that influence the ability to respond to specific targeted therapies. We describe specific inflammatory pathway associated genes linked with response to two important commonly used biologies, rituximab and tocilizumab, while highlighting that some pathways are upregulated in responders to both drugs. This suggests there is likely a group of patients with multiple, overlapping upregulated immune pathways with combined IB- cell and macrophage infiltration who are prone to respond to several biologies as suggested by the similar response rates obtained in most clinical trials regardless of the mechanism of action of the drug used. Due to the nature of the R4RA study design in which patients who responded to their initial drug stayed on that drug and did not switch, it was not possible within the current study to identify this group of ‘double-responder’ patients however, the cross-over design did make it possible to identify the ‘double non-responders’. The invaluable data generated in these patients highlight the existence of a refractory state to multiple biologic therapies largely characterized by the fibroblast pathotype and their related gene expression, which mediate resistance to multiple biologic therapies. Altogether, these results represent a paradigm shift that supports the concept that disease endotypes, driven by a diverse molecular pathology in the diseased tissue, determine diverse clinical and treatment-response phenotypes. It also supports the need for the development of new drugs targeting the refractory endo-phenotype driven by the fibroid/pauci-immune pathotype, as current medications principally target immune-related pathways and, therefore, predictively they are not as effective in these patients.

METHODS

Patients and intervention

A total of 164 patients aged 18 years or over, fulfilling 2010 ACR/EULAR classification criteria for RA who were eligible for treatment with rituximab therapy according to UK NICE guidelines, i.e. failing or intolerant to csDMARD therapy and at least one biologic therapy (excluding trial IMPs) were recruited when fulfilling the trial inclusion/exclusion criteria (for the full study protocol and baseline patient characteristics see Humby et al. (2021) Lancet 397: 305-317). Briefly, patients underwent a synovial biopsy of a clinically active joint at entry to the trial performed according to the expertise of local centre as either ultrasound-guided or arthroscopic procedure, as previously described (Kelly et al. (2015) Ann. Rheum. Dis. 74: 611- 617). Following synovial biopsy, patients were randomised to receive rituximab as two 1000mg intravenous infusions 2 weeks apart or intravenous tocilizumab at a dose of 8mg/kg administered at 4 weekly intervals. Patients were followed up every 4 weeks throughout the 48-week trial treatment period where RA disease activity measurements and safety data were collected. An optional repeated synovial biopsy of the same joint sampled at baseline was performed at 16 weeks (Tables 5 and 6). The study protocol has been published online (http://www.r4ra-nihr.whri.qmul.ac.uk/docs/r4ra_protocol_version_9_30.10.2017_clean.pdf) and was registered on the ISRCTN database, ISRCTN97443826, and EudraCT, 2012- 002535-28. All patients provided written informed consent. The study was done in compliance with the Declaration of Helsinki, International Conference on Harmonisation Guidelines for Good Clinical Practice, and local country regulations. The protocol was approved by the institutional review board of each study centre or relevant independent ethics committees (UK Medical Research and Ethics Committee (MREC) reference: 12/WA/0307).

Response criteria and treatment switch

The primary end-point was defined as Clinical Disease Activity Index (CDAI) >50% improvement from baseline at 16 weeks. CDAI is calculated by adding up the number of tender joints (0-28), the number of swollen joints (0-28), the patient global health assessment on a 0- 10 visuoanalogic scale and the care provider global health assessment on 0-10 Visual Analogue Scale (VAS).

As shown in Fig. 9, CDAI50% non-responders at 16 weeks were switched to the alternative biologic agent, and their response was assessed at 16 weeks following the switch as determined by CDAI50% improvement. Including cross-over patients, a total of 108 patients were treated with rituximab and 117 with tocilizumab. Of those treated with rituximab, 43 were defined responders (40%) while 46 responded to tocilizumab (45%). Among all responders, 9 responded to rituximab after failing tocilizumab and were classified as exclusive responders to rituximab (pro-RTX), while 12 patients responded to tocilizumab after failing rituximab, thus classified as pro-TOC. Patients who failed both drugs throughout the study were classified as multi-drug resistant/refractory (n=40). Histological analysis

A minimum of 6 synovial biopsies were processed in an Excelsior tissue processor before being paraffin-embedded en masse at Queen Mary University of London Core Pathology department. 3-5pm thick tissue sections stained for Hematoxylin and Eosin (H&E), and immunohistochemical markers CD20 (B-cells), CD138 (plasma cells), CD21 (follicular dendritic cells), and CD68 (macrophages) in an automated Ventana Autostainer machine. CD79A (B-cells) and CD3 (T-cell) staining was performed in-house on deparaffinised tissue after antigen retrieval (30 mins at 95°C) followed by peroxidase and protein blocking steps. Primary antibodies [CD79A (clone JCB117, Dako), CD3 (clone F7.238, Dako), CD20 (clone L26, Dako), CD68 (clone KP1 , Dako) and CD138 (clone Ml 15, Dako)] were used for 60 mins at room temperature. Visualisation of antibody binding was achieved by a 30 min incubation with Dako EnVisionTM+ before completion by addition of 3,3’-diaminobenzidine (DAB)+ substrate chromogen for 10 secs, followed by counterstaining with haematoxylin. Following immunohistochemical staining, sections underwent semi-quantitative scoring (0-4), by a minimum of two assessors, to determine levels of CD20+ and CD79a+ B-cells, CD3+ T-cells, CD138+ plasma cells and CD68+ lining (I) and sublining (si) macrophages adapted from a previously described score (Bugatti et al. (2014) Rheumatology (Oxford) 53: 1886-1895) and recently validated for CD20 (Rivellese et al. (2020) Arthritis Rheumatol. 72: 714-725). Haematoxylin and eosin-stained slides also underwent evaluation to determine the level of synovitis, according to the Krenn Synovitis Score (0-9) (Krenn et al. (2006) Histopathology 49: 358-364). The sum of semi-quantitative scores for Krenn Synovitis Score (0-9), CD20 (0-4), CD3 (0-4), CD138 (0-4), and CD68 (0-4) is reported as immune score (0-25).

RNA-Sequencing and Molecular Classification /Analysis

A minimum of 6 synovial samples per patient were immediately immersed in RNA-Later and RNA was extracted from synovial tissue using one of two protocols: either using Phenol/Chloroform isolation or via a Zymo Direct-zol RNA MicroPrep - Total RNA/miRNA Extraction kit. In both methods, tissue was lysed in Trizol solution using a LabGen125 homogeniser. Briefly, for phenol/chloroform extraction method, 1-1 Omg of tissue was lysed and then sheared using a 21G needle. The tissue lysate was then mixed vigorously with chloroform before centrifugation. The aqueous phase was removed and mixed with ice-cold isopropanol for 30 minutes. After further centrifugation, the RNA pellet was washed in 70% ethanol before air-drying and re-suspension in RNAse free water. Samples extracted using Zymo Direct-zol Miniprep kits were done so as per the manufacturer’s instructions. Briefly, 1- 10mg of tissue lysate was run through the Zymo-Spin IC Column. Columns were then washed using the appropriate kit wash buffers before RNA was eluted and re-suspended in RNAse free water. Quality control was carried out by quantifying samples by spectrophotometer readings on a Nanodrop ND2000C. RNA integrity was measured using Pico-chip technology on an Agilent 2100 Bioanalyzer to determine a RIN (RNA integrity number). 214 synovial tissue samples were available for RNA extraction and were subsequently sent for RNA- Sequencing to Genewiz (South Plainfield, NJ, USA). RNA sequencing libraries were prepared using NEBNext Ultra RNA Library Prep kit for Illumina following the manufacturer’s instruction (NEB, Ipswich, MA, USA). Briefly, mRNAs were initially enriched with Oligo d(T) beads followed by limited PCR cycles. The sequencing library was validated on the Agilent TapeStation (Agilent Technologies, Palo Alto, CA, USA), and quantified by using Qubit 2.0 Fluorometer (Invitrogen, Carlsbad, CA) as well as by quantitative PCR (KAPA Biosystems, Wilmington, MA, USA). The sequencing libraries were clustered on Illumina flowcells. Sequencing was performed on an Illumina HiSeq instrument according to the manufacturer’s instruction. The samples were sequenced using a 2x150bp Paired End configuration.

RNA-Seq data processing

214 paired-end RNA-Seq samples of 50 million reads of 150 base pair length were trimmed to remove the Illumina adaptors using bbduk from the BBMap package version 37.93 using the default parameters. Transcripts were then quantified using Salmon40 version 0.13.1 and an index generated from the Gencode release 29 transcriptome following the standard operating procedure. Tximport version 1.13.10 was used to aggregate the transcript level expression data to genes, counts were then subject to variance stabilizing transformation (VST) using the DESeq2 version 1.25.9 package (Love et al. (2014) Genome biology 15: 550). Following RNA-Seq quality control 36 samples were excluded due to poor mapping or RNA quality. Using unsupervised principal component analysis (PCA) and plotting the first 5 eigenvectors in pairs one outlier was identified and removed from further analysis. Thus 133 patients had RNA-Seq data available for subsequent analysis at baseline and 44 patients for the follow-up time point. Baseline characteristics of patients with available RNAseq are shown in Table 7. Starting with length scaled transcripts per million (TPM) counts derived using R package tximport, limma-voom was used for normalisation of data and calculation of weights for linear modelling (Law et al. (2014) Genome biology 15: R29).

Cluster analysis

For cluster analysis, after removing low expressed genes the VST data was filtered using a coefficient of variation cut-off of > 0.075 to select the most variable 22,256 (of 56,809) genes. These genes were used for cluster analysis of all baseline patients (n=133) using the M3C algorithm (John et al. (2020) Scientific reports 10: 1816) with Partitioning Around Medoids (PAM) clustering and 1000 iterations. The lowest Penalised Cluster Stability Index was used to select the number of clusters. After cluster assignment, the patients were split into treatment groups using Pearson’s distance metric and complete linkage method and plotted using the ComplexHeatmap package (version 2.2.0) in R. Chi-squared test was applied to test significance between clusters and response to treatment based on the trial primary outcome measure CDAI50% and additionally ELILAR CRP response (ELILAR response) as another commonly used criterion.

Differential expression and modular analysis of RNA-Seq data

Low expressed genes were excluded from analysis and the remaining 30,841 genes were used for Differential Expression Gene (DEG) analysis. This was based on negative binomial distribution using regression models of normalized count data using DESeq2 and a Wald test to compare variation between treatment response groups in synovium RNA-Seq samples. P- values were false discovery rate (FDR) adjusted using Storey’s q-value with a cut-off of q < 0.05 used to define significantly DEGs. Distributions of DEGs were illustrated in volcano plots and DESeq2 outputs were used for further modular analysis using Bioconductor package Quantitative Set Analysis for Gene Expression (QuSAGE, version 2.10.0). Gene modules from Li et al. (Li et al. (2014) Nat. Immunol. 15: 195-204) and weighted gene correlation network analysis (WGCNA) modules were selected for gene set enrichment.

Deconvolution

MCP counter (Becht et al. (2016) Genome biology 17: 218) was used to deconvolute synovial RNA-Seq, using the package Immunedeconv. Following deconvolution, patients were classified into rich/poor according to the median value of the individual cell type (e.g. B cellrich if >median value of MCP B cells, poor if < median value). For the enrichment of four fibroblast subtypes (SC-F1 : CD34+ sublining, SC-F2: HLA+ sublining, SC-F3: DKK3+ sublining and SC-F4: CD55+ lining), we used average expression of gene signatures obtained from differential gene expression analysis and known markers that are previously described by scRNA-Seq study (Zhang et al. (2019) Nat. Immunol. 20: 928-942). Module scores for each subtype were calculated using the AddModuleScore function from R package Seurat. The top 5 differentially expressed genes were considered subtype-specific gene sets. These gene sets did not have genes in common. Wilcoxon test was used for the statistical assessment of the module scores when comparing responders and non-responders.

Crossover analysis on patients who underwent treatment switch The drug-crossover analysis was performed on baseline RNA-Seq samples of patients who underwent treatment switch (Fig.2g). RNA-Seq counts of protein-coding genes (n=19,508) were used to perform a likelihood ratio test (LRT) that was calculated in comparison to a reduced model with DESeq2 R package (v1.24.0). 3D volcano plot and radial plot were generated using the volcano3D (v1.0.3) R package (Fig.2i). QuSAGE was applied using WGCNA derived gene modules and radial plots were created using the volcano3D package with a p-value significance threshold of p < 0.05 (Fig.2j).

Multiplex Immunofluorescence

Immunofluorescence staining was performed on 3pm formalin-fixed paraffin-embedded (FFPE) human sections obtained from synovial tissues of RA patients. Tissue sections were deparaffinised in sequential changes of xylene and ethanol chambers before washing and placing into preheated pH 6 target retrieval solution (Dako, S1699) in a pressure cooker for 15 minutes. Tissue sections were allowed to cool down at room temperature (RT) before being washed in Tris-buffered saline (TBS). Endogenous peroxidase and biotin activity were blocked with peroxidase block (Dako, S2023) for 10min at RT.

For the CD90/CD45/DKK3 staining, protein block (Dako, X0909) was applied for 1 h, slides were then stained with the first primary antibody (CD45, Dako M0701 , mouse lgG1), washed 3x in TBS and then incubated with Anti-Mouse Envision system HRP (Dako, K4001) for 30min at RT. After 3x washes in TBS, the Cy5/Alx647-conjugated Tyramide reagent (1 :100 dilution, Thermofisher, CatNumb B40958) was applied for 3 minutes. After 3x washes in TBS, antibody stripping was performed by placing the slides into preheated pH 6 target retrieval solution (Dako, S1699) in a pressure cooker for 15 minutes. This process was repeated for two additional primary antibodies: CD90 (1 :240 dilution, Abeam 133350, rabbit) or DKK3 (1 :150 dilution, Sigma-Aldrich, HPA011868, rabbit), followed by Anti-Rabbit Envision system HRP (Dako K4003), followed by Alx488-conjugated Tyramide reagent for CD90 (1 :100 dilution, Thermofisher B40953) or Alx555-conjugated Tyramide reagent for DKK3 (1 :100 dilution, Thermofisher B40955), with antibody stripping in between as described above.

For the CD68/SPP1 staining, following antigen retrieval and peroxidase block as above, protein block (Dako, X0909) was applied overnight at 4°C, slides were then stained with the first primary antibody, SPP1 (Abeam, ab8448, rabbit polyclonal, 1 :300), washed 3x in TBS and then incubated with anti-Rabbit Envision system HRP (Dako, K4003) for 30min at RT, followed by Cy5/Alx647-conjugated Tyramide reagent (1 :100 dilution, Thermofisher B40958). After 3x washes in TBS, slides were incubated with CD68 (CD45, M0814, mouse lgG1 , 1 :50) for 30min at RT, then washed 3x in TBS and incubated with secondary antibody AF488 Rabbit anti-mouse IgG (H+L) (Life Technologies, A11059, 1 :500 dilution) for 30min at RT.

For both stainings, 4',6-diamidino-2-phenylindole (DAPI, Thermofisher) nuclear counterstaining was applied for 10min at RT and slides were then mounted with Prolong Gold Antifade reagent (Thermofisher).

Images were captured using a NanoZoomer S60 Digital slide scanner (Hamamatsu, C13210- 01) at 20x magnification at a resolution of 440 nm/pixel (57727 DPI), with the following exposure times: CD45 alx647 Cy5 16ms; CD90 alx488 FITC 32ms; DKK3 alx555 TRITC 24ms; DAPI 224ms and CD68 alx488 FITC 224ms, SPP1 alx647 Cy 5 24ms, DAPI 96ms. Image analysis was performed using NDP.view 2 Software (Hamamatsu Photonics, U12388- 01).

GeoMx Digital Spatial Profiling

Formalin-fixed paraffin-embedded (FFPE) synovial tissue from 12 patients with rheumatoid arthritis prior to treatment with rituximab or tocilizumab were profiled using GeoMx digital spatial profiling (DSP) platform as previously described (Smolen et al. (2018) Nat Rev Dis Primers 4: 18001). Tissue morphology was visualized using fluorescent CD68-AF532 (clone KP-1 , Novus), CD20-DL594 (clone IGEL/773, Novus) and CD3-AF647 (clone UMAB54, Origene) antibodies and Syto13 (ThermoFisher).

For the NanoString GeoMx DSP WTA assay, slides were prepared following the automated Leica Bond RNA Slide Preparation Protocol (NanoString, MAN-10131-03). In situ hybridizations with the GeoMx Whole Transcriptome Atlas Panel (WTA, 18,677 genes) at a 4 nM final concentration were done in Buffer R (NanoString). Morphology markers were prepared for 4 slides at a time using Syto13 (DNA), CD20, CD3, CD68 in Buffer W for a total volume of 125 pL/slide. Slides incubated with 125 pL morphology marker solution at room temperature for 1 hour. Slides were then washed in SSC and loaded onto the NanoString DSP instrument.

On the DSP instrument each slide was scanned with a 20x objective with the scan parameters: 60 ms FITC/525 nm, 200 ms Cy3/568 nm, 250 ms Texas Red/615 nm, and 300 ms Cy5/666 nm.

Resulting immunofluorescent images were used to select six freeform polygon-shaped regions of interest (ROI) containing approximately 200 nuclei in the CD68+ve synovial tissue lining and superficial sublining, CD20-ve CD3-ve sublining and in CD20+veCD3+ lymphocyte aggregates.

After approval of the ROIs, the GeoMx DSP photocleaved the UV cleavable barcoded linker of the bound RNA probes and collected the individual segmented areas into separate wells in a 96-well collection plate.

The dataset had 72 ROIs from 12 patients (4 Refractory vs. 8 Responder) across the three ROI types. An NTC water well was used for quality control checks.

DSP Analysis

GeoMx WTA sequencing reads from NovaSeq6000 was compiled into FASTQ files corresponding to each ROI. FASTQ files were converted to Digital Count Conversion (DCC) files using the NanoString GeoMx NGS DnD Pipeline. Out of 18,677 genes, 17,065 exceeded the LOQ in more than 10% of the ROIs. Genes that did not exceed the LOQ were excluded from the analysis. For the normalisation, the counts were divided by sample-specific size factors determined by the median ratio of gene counts relative to geometric mean per gen. DESeq2 R package was used for this preprocessing step.

Differential expression analysis

We conducted differential expression analysis to compare responders and refractory patients using DESeq2 (Bhattacharya et al. (2021) Brief Bioinform 22). This analysis was done for all the patients together (responders n=48, refractory n=24) and separately for each location in synovial layer; CD68+ lining/superficial sublining [lining] (responders n=17, refractory n=8), CD20-CD3- deep sublining (responders n=21 , refractory n=12), CD3+CD20+ lymphoid aggregates (responders n=10, refractory n=4). Since the samples were collected from different locations, in the analysis of all samples, we included location as a covariate (~ Location + Response) to eliminate its influence on gene expression. The numbers of patients used in these DEG analyses are given in parenthesis, responders and refractory patients, respectively, qvalue R package implementing Storey’s qvalue method was used to correct for multiple testing effects and a cut-off of q < 0.05 was used to define significantly DEGs.

Longitudinal mixed effects model analysis

Longitudinal analysis of RNA-Seq on paired synovial biopsies was performed by fitting a negative binomial distribution general linear mixed effects model (GLMM) for each gene via the glmer function from R package Ime4 (version 1.1-25), with negative binomial family function from the MASS package (version 7.3-53). Models were fit by maximum likelihood estimation by Laplace approximation and using bound optimization by quadratic approximation (BOBYQA). To analyse the differential effects of the two trial medications over time, the following model was fitted for each gene individually:

Ty_g ~ NB ( iijg, cig) log( tijg) = Oij + Pgo + Pgitimeij + p_g2medicatiorii + Pgstimeijmedicationi + b_gi b_gi ~ 7V(0, o² _gb) where Yijg is the longitudinal raw count of gene g in individual I at timepoint j, ag is the dispersion parameter for each gene, oij is an offset term scaled to the logarithm of the total library size for each sample, and bgi are random effects between individual patients. TPM counts were used as input and only individuals with paired samples were included (88 samples, 44 individuals). The dispersion parameter for the negative binomial distribution for each gene was calculated using DESeq2 (version 1.28.1) estimateDispersions function. To reduce the problem of inflated model coefficients relating to zero counts, genes of low expression were removed using the Limma (version 3.44.3) function filterByExpr, and zero counts were adjusted to a pseudo-count of 0.125, equivalent to the “prior count” approach of edgeR and Voom (Law et al. (2014) Genome biology 15: R29) whose internal defaults are 0.125 and 0.5 respectively. Statistical testing of the fitted model coefficients was performed using Wald type 2 Chi-square test from the car package (version 3.0-10). P values were FDR adjusted using Storey’s q value and a cut-off of FDR < 0.05 was considered significant for each term in the model. Predictions were calculated for each fitted gene model based on the fitted linear model coefficients. 95% confidence intervals for the fixed effects of the fitted model were calculated from the standard deviations of the predictions, by extracting the prediction variances as the diagonal from the variance-covariance matrix of the predictions XVX’, where X represents the model matrix corresponding to the new data and V is the variance-covariance matrix of the model parameters. Similarly, to analyse the difference between CDAI50% responders and non-responders following drug exposure for each medication, the following model was fitted for each drug cohort (58 samples, 29 individuals for rituximab; 30 samples, 15 individuals for tocilizumab): log( tijg) = Oij + Pgo + Pgitimeij + p_g2responsei + p_g3timeijresponsei + b_gi

Longitudinal pathway analysis Genes that showed a significant change in the analysis described in the previous section were used for Gene Ontology/pathway enrichment analysis by means of clueGO (v2.5.5) Cytoscape plug-in. To allow an automated enrichment process, clueGO REST-enabled features were used in R using the following GO/pathway repositories: BiologicalProcess-EBI- UniProt-GOA (11.02.2020), CellularComponent-EBI-UniProt-GOA (11.02.2020), ImmuneSystemProcess-EBI-UniProt-GOA (11.02.2020), MolecularFunction-EBI-UniProt- GOA (11.02.2020), KEGG (27.02.2019), REACTOME (27.02.2019).

Building classifier models for prediction of rituximab and tocilizumab response

Baseline gene expression, clinical, and histological data were used as features for machine learning models build to predict CDAI50% response to either rituximab or tocilizumab treatment at the primary endpoint (16 weeks) or the refractory response, defined as response to either drug at the secondary endpoint (post treatment cross-over, 24 weeks). An overview of the pipeline is shown in Fig. 6a.

The model feature space was created using either clinical and histological parameters, or clinical data with gene expression. The gene expression data underwent a variance stabilisation transform (VST) and was subset to protein-coding genes (using gencode gene annotation v29) with the highest expression variance (top 10%). Highly correlated genes (r > 0.9) were removed using the findCorrelation function from the R package caret (version 6.0- 86) leaving 1438 genes remaining. Clinical features included: baseline tender joint count (TJC), swollen joint count (SJC), age, gender, clinical disease activity index (CDAI), erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), and disease activity score based on ESR and CRP (DAS28ESR and DAS28CRP respectively). Histology features included: CD3, CD68L, CD68SL, CD20, and CD138.

Following processing, data was split into a 10x10 nested folds (Fig.6aii). For models using gene expression features, filtering was performed using either recursive feature elimination (RFE) or univariate filtering from the caret package version 6.0. The number of features selected was chosen to maximise accuracy from 25, 30, 50, or 100.

Seven machine learning methodologies from the caret package were used to create the classifier models: elastic net (glmnet); random forest (rf); least squares support vector machine with radial basis function kernel (svmRadial); least squares support vector machine with polynomial kernel (svmPoly), gradient boosting machine (GBM), mixture discriminant analysis (MDA) and penalised discriminant analysis (PDA). Models which failed to converge during training were excluded from evaluation. In order to evaluate the model performance, receiver operating characteristic (ROC) curves were built using the plotROC R package version 2.2.1 to determine prediction accuracy in the outer fold test data and samples left-out for the inner-fold (LO). The area under the curve (AUC) was calculated to determine the prediction performance. The tuning parameters for the final model were finalised as the mean over all 10 outer folds. The final best model for each classification was fit to the entire data set, exported and feature importance ranked.

Table 4. Synovial histological analysis stratified according to treatment at baseline and 16 weeks

Unpaired analysis (all patients) Paired analysis

Baseline biopsyf

TOC Treatment

N=41 N=24 effect

RTX

Baseline Week 16 Absolute Baseline Week 16 Absolute Least Squares

N=82

Change Change mean difference (%) (%) (95% Cl)

CD20 1 .62 (1 .3) 1 .5 (1 .4) 1 .88 (1 .4) 0.35 (0.8) -1.53 ft 1 .67 (1 .3) 1 .33 (1 .3) -0.34 1 .02 (-81 %) (-20%) (0.52 to 1 .52) §

CD79a 1.54 (1.3) 1 .6 (1 .4) 1 .77 (1 .4) 0.9 (1.1) -0.87 ft 1 .54 (1 .3) 1 .47 (1 .2) -0.07 0.55 (-49%) (-5%) (0.04 to 1 .06) §

CD138 1.43 (1.3) 1 .42 (1 .4) 1 .68 (1 .3) 0.92 (1.1) -0.76 f 1 .58 (1 .4) 1.25 (1.1) -0.33 0.36 (-45%) (-21 %) (-0.16 to 0.88)

CD3 1.43 (1.1) 1 .47 (1 .2) 1.63 (1.1) 1 .52 (1 .2) -0.11 1 .58 (1.1) 1 .42 (1 .2) -0.16 -0.08 (-7%) (-10%) (-0.64 to 0.49)

CD68L 1 .11 (1) 1.2 (1.1) 1 .2 (1) 1 .07 (0.9) -0.13 1 .46 (1.1) 1.38 (1.1) -0.08 0.2 (-11 %) (-5%) (-0.27 to 0.66)

CD68SL 1 .67 (1) 1.75 (1.1) 1 .88 (0.8) 1 .3 (0.6) -0.58 $ 1.92 (1) 0.88 (0.7) -1.04 $ -0.43 (-31 %) (-54%) (-0.78 to -0.08) §

KRENN 3.99 (2.6) 3.88 (2.9) 4.63 (2.5) 3.23 (2) -1 .4 ft 4.38 (2.8) 3.46 (2.4) -0.92 0.32 (-30%) (-21 %) (-0.69 to 1.32)

Data shown as mean (SD) t No significant difference between treatments was observed for the presented values (tested through Mann-Whitney U test)

$ P<0.05 and $$ P<0.001 for the within group change from baseline (paired Wilcoxon test comparing baseline values with values at 16 weeks within the same patients).

§ P<0.05 for the comparison with Non-Responders of the change from baseline (analysis of covariance testing the difference in the changes from baseline between treatments, with treatment as factor and baseline score as covariate)

Table 5: Baseline characteristics of all of patients which consented to a second biopsy at 16 weeks h consented to a second biopsy at 16 weeks Rituximab (n=37) Tocilizumab (n=22) ^H^a^ue^

**6 patients in total used non-TNFi biologies (5 Abatacept and 1 “vaccine RA TNF-K-006” for a clinical study%).

Table 6: Demographics and disease activity of patients undergoing paired week 16 biopsy

Consent to a second biopsy

Baseline characteristics Total cohort No Yes p-value (N=161) (N=96) (N=65)

Gender = M 33 (20.5) 23 (24.0) 10 (15.4) 0.261

Age (years) 55.7 (12.9) 56.4 (12.4) 54.6 (13.7) 0.381

Disease Duration (years) 9.0 [4.0, 19.0] 9.0 [5.0, 21.0] 10.0 [4.0, 16.0] 0.357

CDAI 29.8 [21.7, 40.6] 29.4 [21.2, 41.8] 30.4 [24.5, 39.7] 0.707

ESR (mm/h) 31.5 [17.0, 48.0] 28.0 [17.0, 48.0] 35.0 [19.0, 50.0] 0.202

CRP (mg/L) 11.0 [5.0, 27.5] 14.4 [5.0, 32.2] 10.0 [5.0, 25.0] 0.462

RF positive 105 (67.3) 60 (65.2) 45 (70.3) 0.621

ACPA positive 119 (76.8) 70 (76.9) 49 (76.6) 1.000

Tender Joint Count (28) 11.0 [6.0, 18.0] 10.5 [5.8, 17.0] 11.0 [7.0, 18.0] 0.248

Swollen Joint Count (28) 6.0 [3.0, 10.0] 6.0 [3.0, 11.2] 6.0 [4.0, 9.0] 0.919

DAS28 (ESR) 5.8 (1.2) 5.7 (1 .4) 5.9 (1.0) 0.311

DAS28 (CRP) 5.3 (1.2) 5.3 (1.3) 5.3 (1.0) 0.829

Week 16 Treatment response, N (%)

CDAI >50% Response 81 (50.3) 56 (58.3) 25 (38.5) 0.016*

CDAI-MTR 52 (32.3) 40 (41.7) 12 (18.5) 0.002*

Baseline histological score, mean (SD)

CD20+ B cells 1.56 (1.3) 1 .38 (1 .3) 1 .8 (1 .3) 0.052

CD138+ plasma cells 1.43 (1.4) 1 .26 (1 .4) 1 .65 (1 .4) 0.077

CD68L+ macrophages 1.15 (1) 1.05 (1) 1 .29 (1.1) 0.141

CD68SL+ macrophages 1.71 (1) 1 .57 (1.1) 1 .89 (0.9) 0.04*

CD3+ T cells 1.45 (1.2) 1 .32 (1 .2) 1 .62 (1.1) 0.081

Synovitis score 3.93 (2.8) 3.48 (2.8) 4.54 (2.6) 0.019*

Table 7: Baseline characteristics of patients with available RNAseq

Gender (Male) 24(18%) 16(24%) 8(12%) 014

Age, years 55-1 (13-3) 54-7(13-7) 55-6(13-0) 069

Disease Duration, years 90 [50, 190] 80 [40, 210] 100 [50, 180] 085

Clinical disease activity index (CDAI) 309 [21 -7, 40-7] 310 [223, 41 1] 304 [21 7, 406] 083

Erythrocyte sedimentation rate (ESR), mm/h 320 [180, 460] 340 [178, 462] 280 [190, 450] 062

C-reactive protein (CRP), mg/L 110 [50, 230] 100 [50, 208] 150 [60, 290] 019

Rheumatoid factor (RF) OR Anti— citrullinated protein antibody .. ,„__n/x

(ACPA) positive 115(86/o) 59(87/o) 56 (86%) 1

Rheumatoid factor (RF) positive 96 (72%) 52 (76%) 44 (68%) 035

Anti— citrullinated protein antibody (ACPA) positive 104 (78%) 54 (79%) 50 (77%) 089

Creatinine (pmol/L) 610 [540, 710] 630 [540, 725] 587 [540, 672] 036

Alanine aminotransferase (ALT), U/L 160 [11 0, 220] 160 [110, 200] 160 [120, 230] 041

Aspartate aminotransferase (AST), U/L 190 [150, 222] 190 [150, 220] 190 [160, 225] 073

,, , .. 121 10 3 [1100, 1200 [1085, 1215[1110,

Haemoglobin, g/L J._{5] 131} ^l._{5] n R1}

1300] ⁰⁸¹

White Blood Cell count, 10⁹/L 80 [67, 102] 7-9 [62, 97] 8-5 [70, 103] 017

Platelets 10⁹/L 3040 [2545, 3030 [2530, 3040 [2558, _n __

Platelets, 10 /L 3940] 3645] 413-8] ⁰³³

Neutrophils, 10⁹/L 5-5 [44, 72] 57 [42, 71] 5-4 [4-6, 7-2] 028

Lymphocytes, 10⁹/L 1-7 [1-3, 23] 16 [12, 22] 1-7 [1-4, 24] 0-15

Synovial semi-quantitative scores

CD20 10 [00, 30] 1-5 [00, 30] 10 [00, 30] 0'85

CD138 10 [00, 30] 10 [00, 30] 10 [00, 30] 084

CD68 lining 10 [00, 20] 10 [08, 20] 10 [00, 20] 052

CD68 sub-lining 20 [10, 20] 20 [10, 20] 20 [10, 30] 038

CD3 10 [00, 20] 10 [10, 20] 20 [00, 30] 0-61

CD21 (Positive) 9 (8%) 5 (9%) 4 (7%) 1

Synovial Pathotype 022

Fibroid 24 (18%) 11 (16%) 13 (20%)

Lymphoid 74 (56%) 36 (53%) 38 (58%)

Myeloid 31 (23%) 17 (25%) 14 (22%)

Ungraded 4 (3%) 4 (6%) 0 (0%)

B cell status 0 25

B cell poor 65 (49%) 32 (47%) 33 (51 %)

B cell rich 55 (41 %) 27 (40%) 28 (43%)

Germinal Centre 9 (7%) 5 (7%) 4 (6%)

Unknown 4 (3%) 4 (6%) 0 (0%)

Number of tender joints, 0-28 12 0 [6 0, 18 0] 11 0 [7 0, 18 2] 12 0 [6 0, 16 0] 0 67

Number of swollen joints, 0-28 6 0 [4 0, 10 0] 6 0 [4 0, 9 0] 6 0 [3 0, 11 0] 0 73

28 joint count Disease Activity Score (DAS-28), ESR 5-8 (1 -2) 5-9 (1 -2) 5-8 (1 -3) 0 95

28 joint count Disease Activity Score (DAS-28), CRP 5-4 (1 -2) 5 3 (1 1) 5-4 (1 -2) 0 62

Patient’s global assessment, 0-100 VAS 72 0 [51 0, 85 0] 69 5 [48 5, 80 5] 74 0 [53 0, 86 0] 0 18

Physician’s global assessment, 0-100 VAS 60 3 (21 7) 60 4 (21 3) 60 2 (22 2) 0 96

Patient’s assessment of early morning stiffness, 0-100 VAS 32 5 [20 0, 100 0] 30 0 [17 5, 80 0] 45 0 [20 0, 100 0] 0 26

Patient’s assessment of tiredness, 0-100 VAS 68 0 [50 0, 83 0] 67 0 [45 5, 78 2] 70 0 [50 0, 91 0] 0 17

Patient’s assessment of pain, 0-100 VAS 71 0 [50 0, 87 0] 67 5 [49 8, 82 5] 73 0 [50 0, 89 0] 0 34

HAQ total score 1 -7 (0 6) 1 -7 (0 7) 1 -7 (0 6) 0 72

Functional Assessment of Chronic Illness Therapy (FACIT) score 22 0 [15 0, 33 2] 23 0 [15 0, 32 0] 21 -5 [13 0, 34 0] 0 76

Previous Methotrexate use 133 (100%) 68 (100%) 65 (100%)

Previous Prednisolone use 74 (56%) 35 (51 %) 39 (60%) 0 41

Number of previous biologies used, [anti-TNF/Other**] 0 19

1 93 (70%) [93/0] 51 (75%) [51/0] 42 (65%) [42/0]

2 32 (24%) [29/3] 12 (18%) [9/3] 20 (31 %) [20/0]

3+ 8 (6%) [5/3] 5 (7%) [3/2] 3 (5%) [2/1]

Data are n (%), median [IQR], mean (SD%). ECOG=Eastern Cooperative Oncology Group. BMI=body-mass index CDAI=Clinical disease activity index. DAS28 =28 joint count disease activity score. CRP=C-reactive protein. ESR=erythrocyte sedimentation rate. "6 patients in total used non-TNFi biologies (5 Abatacept and 1 “vaccine RA TNF-K-006” for a clinical study%).

Table 8: Machine learning model evaluation using nested cross-validation

glmnet = lasso and elastic-net generalized linear model; rf= random forest; gbm=gradient boosting machine; svmRadial=radial support vector machine; svmPoly = polynomial support vector machine.

* Final fit failed.

Table 9: Best tuned parameters for final models i) Rituximab glmnet model ii) Tocilizumab glmnet model

ill) Refractory gbm model

Table 10: Final fitted model coefficients and variable importance i) Rituximab glmnet model (n=40) ii) Tocilizumab glmnet model (n=39) iii) Refractory gbm model (n=53)

Table 11 : Antibodies used for immunofluorescence staining

All publications mentioned in the above specification are herein incorporated by reference. Various modifications and variations of the disclosed methods and kits of the invention will be apparent to the skilled person without departing from the scope and spirit of the invention. Although the invention has been disclosed in connection with specific preferred embodiments, it should be understood that the invention as claimed should not be unduly limited to such specific embodiments. Indeed, various modifications of the disclosed modes for carrying out the invention, which are obvious to the skilled person are intended to be within the scope of the following claims.

Claims

1 . A method for determining whether a Rheumatoid Arthritis (RA) patient is susceptible or refractory to treatment with a B cell targeted therapy and/or an agent that downregulates IL-6 mediated signalling, the method comprising the steps:

(b) (i) determining the level of one or more second biomarker in one or more sample obtained from the patient, wherein the one or more second biomarker is selected from Table 2; and (ii) comparing the level of the one or more second biomarker to one or more corresponding reference value; wherein the level of the one or more second biomarker compared to the corresponding reference value is indicative of the susceptibility to treatment with an agent that downregulates IL-6 mediated signalling; or

2. A method for determining whether a Rheumatoid Arthritis (RA) patient is susceptible to treatment with a B cell targeted therapy and/or an agent that downregulates IL-6 mediated signalling, the method comprising the steps:

(b) (i) determining the level of one or more second biomarker in one or more sample obtained from the patient, wherein the one or more second biomarker is selected from Table 2; and (ii) comparing the level of the one or more second biomarker to one or more corresponding reference value; wherein the level of the one or more second biomarker compared to the corresponding reference value is indicative of the susceptibility to treatment with an agent that downregulates IL-6 mediated signalling. A method for determining whether a Rheumatoid Arthritis (RA) patient is refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling, the method comprising the steps: (i) determining the level of one or more third biomarker in one or more sample obtained from the patient, wherein the one or more third biomarker is selected from Table 3; and (ii) comparing the level of the one or more third biomarker to one or more corresponding reference value; wherein the level of the one or more third biomarker compared to the corresponding reference value is indicative of the refractoriness to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling. The method of any preceding claim, wherein the level of the one or more first, second and/or third biomarker is a nucleic acid level, optionally wherein the nucleic acid level is an mRNA level. The method of any preceding claim, wherein the step of determining the level of one or more first, second and/or third biomarker is performed by direct digital counting of nucleic acids, RNA-seq, RT-qPCR, qPCR, multiplex qPCR or RT-qPCR, microarray analysis, or a combination thereof. The method of any preceding claim, wherein the step of determining the level of one or more first, second and/or third biomarker is performed by RNA sequencing. The method of any preceding claim, wherein the step of determining the level of the one or more first, second and/or third biomarker comprises determining the level of gene expression of the one or more first, second and/or third biomarker. The method of any preceding claim, wherein the one or more sample is a synovial sample. The method of any preceding claim, wherein:

(a) (i) when the level of the one or more first biomarker is greater than the corresponding reference value the patient is determined to be susceptible to treatment with the B cell targeted therapy; and/or (ii) when the level of the one or more first biomarker is less than the corresponding reference value the patient is determined to be resistant to treatment with a B cell targeted therapy;

(b) (i) when the level of the one or more second biomarker is greater than the corresponding reference value the patient is determined to be susceptible to treatment with the agent that downregulates IL-6 mediated signalling; and/or (ii) when the level of the one or more second biomarker is less than the corresponding reference value the patient is determined to be resistant to treatment with an agent that downregulates IL-6 mediated signalling; and/or

(c) when the level of the one or more third biomarker is greater than the corresponding reference value the patient is determined to be refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling. The method of any preceding claim, wherein the level of the one or more first biomarker compared to the corresponding reference value classifies the sample as B cell rich or B cell poor. The method of any preceding claim, wherein: (a) when the sample is B cell rich the patient is determined to be susceptible to treatment with a B cell targeted therapy; and/or (b) when the sample is B cell poor the patient is determined to be resistant to treatment with a B cell targeted therapy. The method of any preceding claim, wherein the B cell targeted therapy is B cell depletion therapy. The method of any preceding claim, wherein the B cell targeted therapy is selected from the group consisting of: rituximab, ocrelizumab, veltuzumab, ofatumumab, epratuzumab, obinutuzumab, ibritumomab and tiuxetan. The method of any preceding claim, wherein the B cell targeted therapy is rituximab.

106

15. The method of any preceding claim, wherein a patient determined to be resistant to treatment with the B cell targeted therapy is determined to be suitable for treatment with an agent that downregulates IL-6 mediated signalling.

16. The method of any preceding claim, wherein the agent that downregulates IL-6 mediated signalling is an IL-6 receptor antagonist.

17. The method of any preceding claim, wherein the agent that downregulates IL-6 mediated signalling is selected from the group consisting of tocilizumab, sarilumab, satralizumab and siltuximab.

18. The method of any preceding claim, wherein the patient is refractory to DMARD and/or anti-TNF therapy.

19. The method of any preceding claim, wherein the method further comprises:

(a) administering to the patient a B cell targeted therapy when the patient is determined to be susceptible to treatment with a B cell targeted therapy;

(b) administering to the patient an agent that downregulates IL-6 mediated signalling when the patient is determined to be susceptible to treatment with an agent that downregulates IL-6 mediated signalling; or

(c) administering to the patient an alternative therapeutic when the patient is determined to be refractory to treatment with a B cell targeted therapy and an agent that downregulates IL-6 mediated signalling.

20. A kit for use in the method of any preceding claim.

21. The kit of claim 20, wherein the kit comprises one or more reagent suitable for detecting the one or more first, second and/or third biomarkers.

22. The kit of claim 20 or 21 , wherein the kit comprises reagents for RNA sequencing.

23. The kit of any one of claims 20-22, which comprises one or more probe or antibody for detecting the one or more first, second and/or third biomarkers.

24. The kit of any one of claims 20-23 which is in the form of a microchip or microarray.

25. A method for treating Rheumatoid Arthritis (RA), the method comprising:

107 (a) administering to a patient an effective amount of a B cell targeted therapy, wherein the patient is determined to be susceptible to treatment with a B cell targeted therapy by the method of any one of claims 1-18; or

(b) administering to a patient an effective amount of an agent that downregulates IL-6 mediated signalling, wherein the patient is determined to be susceptible to treatment with an agent that downregulates IL-6 mediated signalling by the method of any one of claims 1-18.

108