This study revealed different molecular and cellular mechanisms behind non-response to MMF and AZA by analyzing retrospectively a longitudinal cohort of responder and non-responder SLE patients to both drugs.
The course of the disease is complex and unpredictable, alternating periods of inactivity, disease flares and progression to organ damage, with different underlying molecular mechanisms which may potentially differ between patients. This heterogeneity particularly hinders the effective discovery of robust biomarkers for both disease progression as for treatment responses17. Cross-sectional studies of patients with active disease limit the different scenarios to analyze, reducing reproducibility in other cohorts and/or disease conditions. Therefore, a longitudinal cohort was selected, with samples representing different disease states, with different clinical manifestations and treated with different routine treatments and doses. Robust non-response gene signatures to MMF and AZA were obtained across all the clinical and molecular heterogeneity of the disease. Maintenance drugs including HC and HC plus GC were analyzed demonstrating that MMF and AZA non-response patterns were drug-specific, not influenced by secondary SOC therapies. In addition, drug signatures were used to build ML-based models to predict drug responses obtaining high performance results (balanced accuracies higher than 0.75 in all cases).
One main limitation of our study is the small number of patients treated using for some specific drugs (mainly for AZA), making more difficult the interpretation of the AZA-associated data. A larger interventional clinical trial would be required in order to validate responsiveness and non-responsiveness mechanisms to the drugs alone and to test the predictive capacity of the non-response signatures defined. In lupus, it is particularly difficult to obtain public longitudinal transcriptome data and more so if a single drug is to be studied. SLE patients take, in most instances, combinations of multiple drugs, and response outcomes are often not shared. Validation could bring us closer to more personalized medicine, supporting more effective first-line therapy choice for LN patients.
Despite this, we obtained revealing and encouraging results. Analyzing cell profiles, we observed a depletion of T cells in non-responder patients and a worse response ratio was consistently observed for patients poor in various T cell subpopulations. In a previous study, T lymphocyte exhaustion was associated with LN25, but differences comparing response and non-response to drugs have never been reported before. Perhaps insufficient or abnormal T cell function could be influencing the lack of response26. For MMF, the non-response was mainly mediated by PCs, pDCs and ABCs, in line with the fact that the worst response ratios were obtained for patients showing rich memory B cell profiles. ABCs are a class-switched, antigen-specific memory-like B cell population expanded in SLE that contributes to autoimmunity through the production of autoantibodies and cytokines and regulating inflammatory T cells acting as APCs27. Their differentiation is driven by the toll-like receptor (TLR) 7 in an interleukin-21-mediated mechanism28. Recently, expansion of ABCs has been observed in the kidneys of LN patients29 and in SLE mouse models30, underscoring the importance of these cells. The question remains as to why are ABCs remaining high and if this might be due to resistance of these cells to MMF, mechanisms that would need to be experimentally tested.
The MMF non-response signature was also expressed in NKT cells, which regulate Th1/Th2 balance31. In fact, cross-regulation between Tregs and NKT cells was previously reported. Activated NKT cells modulate Treg function through IL-2-dependent mechanisms, whereas Treg can suppress proliferation, cytokine release and cytotoxic activity of NKT cells by cell-contact-dependent mechanisms32.
CD1C + cDC2 and non-classical monocytes also over-expressed the non-response signature to MMF. cDC2 influence aberrant T cell functions secreting interleukin-8 and other proinflammatory cytokines33. HLA class II genes, expressed by APCs and importantly expressed by the relevant non-response-related cell subtypes, modulate the interaction of T and B cells in the production of autoantibodies. The genetic association of the HLA class II genes with autoantibody production in SLE is well established, and our results suggest that CD1C + cDC2 may be importantly involved in this context34. These clusters seem to be playing an important role in renal damage control, showing functions related to complement-mediated phagocytosis22. Complement cascade proteins bind immune-complex deposits in the kidney glomerulus driving immunopathology leading to long-time scars35.
For AZA, the most notable finding is the exacerbated expression of a non-response signature in CD16+ and CD14+ monocytes with genes involved in migration related functions. The accumulation of CD16+ monocytes in the blood could reflect either an increase in their differentiation, which would lead to greater amounts of them migrating to the target tissue, or just the opposite, a deficit in the correct migration processes to the tissue36. Deconvolution of cell types from bulk transcriptome did not allow identification of CD16+ monocytes in blood, so future analyses would be necessary to validate the increase or lack of migration of these monocytes to the tissue in the AZA therapy context.
Therefore, we revealed different molecular signatures and different cellular subtypes associated with them for non-response to MMF and AZA. In fact, in silico inhibition of targets from regulatory networks regulating clusters associated to MMF or AZA non-response identified different response ratios for refractory patients for each drug. CCL2 inhibition has been previously proposed to reduce tissue infiltration of monocytes, minimizing the inflammatory phenotypes37, while belimumab, an anti-BAFF drug, is currently approved for SLE and LN. BAFF inhibition leads to a reduction in autoantibody production, depleting the differentiation of PCs from B cells38. In fact, growing studies show the effectivity of combining belimumab with other immunosuppressant drugs3. Here, we presented potential evidence that anti-BAFF could be more beneficial for non-responders to MMF by in silico analysis. Detailed analysis is required to test the efficacy of belimumab as an add-on therapy to MMF in real world terms.
Finally, there is extensive evidence showing the importance of IFN-I in SLE and other autoimmune diseases39,40. We herein report the co-expression of IFN-related genes and non-response signatures to LN drugs in the same cell subsets. Specifically, at least a handful of genes from the ISG and IFITM families of genes showed high expression scores in subsets expressing AZA and MMF non-response signatures, both of them for AZA, and IFITM genes particularly for MMF.
The IFITM-family of genes codify 3 anti-viral subfamilies of proteins, one of which is immune-related, including, in turn, 3 main proteins, IFITM1, IFITM2 and IFITM341, all of which evolved evolutionarily through their expansion and interaction with viral infections. Despite their protein sequence similarity IFITM1, 2 and 3 have different cellular localization and function, and different anti-viral specificity through mechanisms still poorly understood. While IFITM1 is exposed on the cell surface (former Leu-13 antigen-expressing cells, now CD225), IFITM2 and 3 are localized in endosomes and lysosomes. Interestingly, IFITM1 and IFITM3 have been found as part of the B cell signaling complex in the plasma membrane together with CD19 and CD21, as well as CD81. Upon B cell activation, IFITM3 protein is increased moving from the endosomes to the lipid rafts containing the B cell signaling complex. Most interestingly, several studies have addressed the role of IFITM3 in B cell activation with expansion and affinity maturation in germinal center B cells through amplification of the PI3K signaling pathway41. In B cell malignancies, expression of IFITM3 is associated with poor outcomes42. In addition, IFITMs expression is induced by IFN-I primarily in monocyte-derived macrophages. Transcription is induced by various pro-inflammatory cytokines and Toll-like receptors agonists. The IFITM1-3 genes have an IFN response element that confers responsiveness to type I and II IFNs. So, IFITM and IFN-I regulate each other. What the function of these genes and others identified in non-responders is in the context of SLE, requires further investigation.
This new knowledge shed light on the molecular and cellular patterns associated to the non-response to LN therapies, opening a new scenario for further investigation of the regulatory mechanisms between implicated cell subsets, the genes and cells involved, and the development of new therapeutic strategies for LN and drug response prediction.