Reverse Association Of Single Cells To Rheumatoid Arthritis Accounting For Mixed Effects Identifies An Expanded CD27- HLA-DR+ Effector Memory CD4+ T Cell Population

Defining disease associated cellular subsets from single-cell assays can be confounded by technical variation and inter-individual differences in subset frequencies. We present Mixed Effects Modeling of Associations of Single Cell Populations (MASC), a novel reverse single cell association strategy that tests whether case-control status influences the membership of single cells in any of multiple cellular subsets. MASC is able to account for random technical artifacts that confound clustering and person-to-person biological variation. After demonstrating appropriate type I error, we applied this method to mass cytometry data collected from 52 cases and controls to identify CD4+ memory T cells altered in the circulation of rheumatoid arthritis (RA) patients. MASC identified a population of effector memory T cells expanded in RA patients compared to non-inflammatory controls (OR = 1.7; p = 1.1 x 10^-3). This population was characterized by absence of CD27 and expression of HLA-DR and constituted 1.7% of the total resting CD4+ memory T cells in controls and 3.1% in cases. We replicated the expansion of CD27- HLA-DR+ T cells in an independent cohort of 39 RA patients (p = 0.011), and find that effective RA treatment reduces the frequency of these cells in circulation. Further, we find that CD27- HLA-DR+ cells infiltrate the joints of RA patients and produce abundant effector cytokines upon stimulation. We propose that MASC is a broadly applicable method to identify disease-associated cell populations in high-dimensional single cell data.