Editorial: Plasma and B Cell Gene Signatures: Quantitative Targeting and Monitoring of B Cell–Depleting Therapies in Autoimmune Diseases in the Genomic Era
The role of B cells in autoimmunity is becoming increasingly apparent and is supported by many lines of evidence ([1-3]). Autoantibodies are a hallmark of autoimmune disease and are produced by autoreactive plasma cells with likely contributions to disease pathogenesis. B cells may also play a more integral role in the cascade of events that lead to autoimmune disease by acting in both effector and immunoregulatory functions (). Due to mounting evidence implicating B cells in the pathogenesis of autoimmune disease, there has been increased interest in B cell–depleting therapies.
The group of cells we term B cells includes those at many different stages of maturation, from the pro–B cell to the terminally differentiated plasma cell (for review, see refs. and). A number of different molecules are expressed on the surface of B cells and are used to determine the differentiation state of the cell. CD20 (encoded by the MS4A1 gene) is expressed at all stages of B cell development, from the pro–B cell to the mature B cell, but is not expressed on plasmablasts (precursor cells of short- and long-lived plasma cells) or plasma cells. The CD19 molecule is also expressed through all stages of B cell development except on terminally differentiated plasma cells. High levels of CD19 expression are associated with an activated pre–plasma cell phenotype. Increased numbers of this cell type are correlated with adverse outcomes in systemic lupus erythematosus (SLE), and these cells may then differentiate into plasma cells at sites of inflammation (). Plasma cells are generally defined in humans by expression of the surface markers CD27 and CD138.
B cell–depleting therapies such as anti-CD20 (rituximab) have shown benefits in rheumatoid arthritis (RA) ([7, 8]) and in studies of their off-label use in SLE ([9-11]). Unfortunately, the 2 large multicenter placebo-controlled trials of rituximab in SLE, one in moderately to severely active SLE and the other in active proliferative lupus nephritis, failed to meet their primary clinical end points (refs. and; for review, see ref.). Anti-CD20 trials in systemic sclerosis (SSc) have had mixed results. A trial by Lafyatis and coworkers found that rituximab depleted circulating and dermal CD20+ B cells but had no effect on circulating autoantibodies, and patients showed no significant improvement in skin disease (). In contrast, other trials have shown improvement in skin disease ([16, 17]) and lung function ([18, 19]). Successes with anti-CD20, coupled with data showing that CD19 is critical for fibrosis in mouse models of SSc ([20, 21]), suggest that B cell–depleting therapy for autoimmune diseases such as SSc is still an important therapeutic option. Although there are many possible reasons for the variable results in these trials, a simple explanation may be that one needs to target patients who are most likely to benefit (i.e., those with significant numbers of CD20+ B cells). The same could be argued for anti-CD19 therapy. In this issue of Arthritis & Rheumatology, Streicher et al () present a possible gene expression biomarker that could assist in identifying patients who have high levels of plasma cells (and consequently elevated numbers of CD19+ plasma cell precursors) and who may thus benefit from anti-CD19 therapy.
Streicher and colleagues report a sensitive gene expression–based plasma cell signature that does not require counting cells by flow cytometry. Flow cytometry has limited feasibility as a diagnostic tool for measuring plasma cells in the clinical setting (for review, see ref.), as the cells are present in very small numbers and are difficult to isolate from tissue. Moreover, significant loss of plasma cells generally occurs during processing for flow cytometry, and it is difficult to sample all surviving cells in the analysis. In contrast, samples for gene expression analysis usually undergo an immediate homogenization after sample collection that isolates and captures the state of the tissue by virtue of the genes being expressed, generally before molecular or cellular changes can occur. These analyses can be performed from both tissue and peripheral blood mononuclear cells (PBMCs), and if the measures are genome-wide, the data can be queried for gene expression signatures representing specific infiltrating cells or even deregulated molecular signaling pathways.
Streicher et al define a small set of genes that are primarily expressed in plasma cells (but not in B cells, granulocytes, monocytes, or T cells), and this plasma cell signature is sensitive enough to detect fewer than 400 plasma cells spiked into PBMCs from a healthy donor. Importantly, this plasma cell signature was sensitive to the number of plasma cells, whereas simply examining the transcript levels of either of the general B cell markers CD20 or CD19 showed no change. The authors additionally define a separate B cell signature that can distinguish terminally differentiated plasma cells from B cells. The significance of this finding is that it not only allows one to predict which patients may have higher numbers of plasma cells, but it also provides a quantitative measure of whether or not plasma cells or B cells are depleted by targeted therapies.
The authors use the plasma cell signature to query gene expression data from PBMCs, skin, and synovium of patients with different autoimmune diseases. Their findings indicate that 30–35% of patients with SLE and RA have increased plasma cell levels and therefore could be candidates for anti-CD19–directed therapies, which should target the pre–plasma cells that may ultimately result in self-recognition (). Much of the analysis of the plasma cell signature in different autoimmune diseases was performed using publicly available data from the public data repository, NCBI GEO, as well as data generated by the authors, highlighting the importance of the public release of published data for use by the scientific community. Although patients with several types of autoimmune disease may benefit from this therapy, data are reported for a phase I dose-escalation trial of MEDI-551, a humanized monoclonal antibody directed against CD19, in SSc (clinical trial no. NCT00946699; www.clinicaltrials.gov).
Targeting the correct patients and quantitative monitoring of effect
In general, successful therapeutic intervention in SSc has proven to be a significant challenge. The phenotypic manifestations of the disease are significantly heterogeneous, which has confounded not only clinical trials, but also basic science findings. To date, SSc has no widely accepted biomarkers apart from autoantibodies. There are no curative treatments, and 1 in 3 patients dies within 10 years of diagnosis (). Interest in SSc clinical trials is increasing dramatically, but heterogeneity in the patient population and lack of biomarkers make outcomes difficult to assess. One way to increase the likelihood of success is to target drug treatments to the subset of patients who will benefit most directly from them, while providing a quantitative measure of whether or not the treatment modulates its intended targets. The field of SSc may now be nearing a tipping point spurred by basic science advances, genome-wide profiling of gene expression, discovery of genetic variation, drug repositioning efforts, and the pharmaceutical industry making SSc a priority. Genomic approaches that have become commonplace in cancer are now being used in trials in autoimmune disease.
The major manifestations of SSc are dysfunction of the vascular and immune systems, skin fibrosis, and extracellular matrix (ECM) deposition (). Which of these processes is the initial event has not been clear, but genetic data implicate initial predisposing genetic variants that occur in components of the immune system ([26, 27]), including possible modulation of B cell function. Fibrosis that occurs in the skin and in affected organs is primarily the result of collagen and ECM deposition by fibroblasts.
Streicher and colleagues address a slightly different question from the global distribution of gene expression that may be found by genome-wide expression profiling ([28-30]). Instead, they focus on the plasma cell, which is expected to be depleted by anti-CD19 therapy. Using whole-genome analysis of gene expression in sorted cell fractions, they selected a panel of 5 genes that were primarily expressed in plasma cells. They make use of skin and PBMC samples obtained at various time points after a single dose of MEDI-551 in the phase I dose-escalation trial of this agent. The authors demonstrate a 98% reduction in the plasma cell signature and a 90% reduction in the B cell signature posttreatment, with decreases observed both in PBMCs and in skin. This shows a specific decrease in the genes associated with cells targeted by the therapy and provides a quantitative measure of the biologic effect of the therapy that can be applied both to target tissue as well as to peripheral blood samples in the absence of cell sorting. The decrease in the plasma cell signature strongly correlates with a decrease in the expression of COL1A1, COL3A1, and COL5A1, which suggests a decrease in ECM deposition.
Most importantly, this quantitative signature may provide a means to identify the patients most likely to benefit from anti-CD19 therapy by using these genes to select patients with high levels of expression of these markers. This would dramatically increase the likelihood of targeting the correct drug to the patient who is most likely to respond to the therapy. Streicher et al mention that the plasma cell signature is enriched in the immunoglobulin cluster discovered by Milano et al (), which tends to be heterogeneous across gene expression–based patient subsets. Previously, dense clusters of CD20+ B cells were thought to give rise to this signature (), but the results presented by Streicher et al suggest that this signature more likely arises from infiltrating plasma cells.
Cellular subset prediction by gene expression
Since the early days of genome-wide analysis of gene expression by DNA microarray, it was consistently observed that molecular signatures found by gene expression reflect the tissue of origin from which the sample was derived. This result is found in the gene expression profiling of liquid and solid tumors ([32, 33]), of cells in culture (), and, of course, in peripheral blood cells (). One determinant of breast cancer subtypes based on examination of whole tumor tissue is identifying which tumors have infiltrating immune cells that are readily discernible by gene expression and usually absent in normal tissue ([33, 36]). In peripheral blood cells, the rapidly changing immune cell subsets can be a major confounder of the overall gene expression patterns. Similar results have been found in autoimmune diseases, such as SSc ([28, 29, 31]) and SLE ([37, 38]).
Just as in the selection of cell surface markers, assigning the expression of a set of genes to a particular pathway, cell type, or process can be difficult as most genes are expressed in more than one cell type or can be induced by more than one pathway. Indeed, the results reported by Streicher et al show that although the transcripts selected are primarily expressed in plasma cells and B cells (see Table 1 in their article), some genes, such as IGJ, are predicted to have 15% of their expression contributed by B cells and 85% by plasma cells. Few genes are completely restricted to a single cell type (although the T cell receptor may be the exception). Nevertheless, there are subsets of genes that are enriched and can be used to identify cellular populations; lists of such cellular subsets are plentiful in the literature and can be found in public databases. For example, Palmer et al () isolated B cells, T cells (CD4+ and CD8+), granulocytes, and total lymphocytes by negative selection followed by gene expression profiling and selection of genes specific to each of the different cell subsets. A more recent study specifically examined cells of the myeloid lineage () and included gene sets for monocytes, neutrophils, eosinophils, myeloblasts, monoblasts, and megakaryocytes. Therefore, the work of Streicher et al extends and adds to this valuable literature of cell type–specific gene expression. Whole-tissue gene expression cannot distinguish between activation of a gene and increased gene expression due to infiltrating cells. Despite these caveats, using this approach to estimate changes in cellular composition in a complex tissue can be extraordinarily valuable, particularly in a rare disease, as the authors demonstrate.
Clinical and basic science implications
One of the most important opportunities emerging from the work of Streicher et al may be the ability to identify those patients with plasma cells and/or B cells in the skin and peripheral blood, differentiating the cellular subsets, and monitoring the depletion of these cells by targeted therapies in posttreatment samples in clinical trials. Additional work will be required to determine if monitoring a gene expression signature is sufficiently robust to overcome and replace the logistical considerations of counting the cells by flow cytometry or detecting them by immunohistochemical staining.
Implementing such a test can be a challenge, but one that we as a community should consider carefully and incorporate into clinical trials in the design stage rather than after the trial has been funded. Presumably, those running trials of B cell–depleting therapeutics will increase the likelihood of success by using these signatures to identify patients most likely to benefit from anti-CD19 or anti-CD20 therapy. A limitation of the current work is that it doesn't address whether or not the decrease in the plasma cell signature correlated with clinical improvement in the patients. It will be intriguing to see if the clinical data from the MEDI-551 trial parallel the decrease in the plasma cell signature and the concomitant decrease in collagen production, resulting in clinical improvement in the patients. If gene expression signatures prove to be successful in identifying patients who benefit from therapy, then this will provide one more piece of evidence that assessment of molecular markers should be standard in clinical trials of complex autoimmune disease, even if initially only to aid interpretation of the final results. Ultimately, after sufficient data have been generated and the results have been carefully validated, such measures may be useful for determining which patients may benefit from B cell–depleting therapies.
Dr. Whitfield drafted the article, revised it critically for important intellectual content, and approved the final version to be published.