Cohort-aware single-cell analysis of PBMCs identifies an interferon program specific to primary Sjögren’s syndrome and limited immune convergence with periodontitis

Periodontitis (PD) and primary Sjögren’s syndrome (pSS) are epidemiologically linked, and shared peripheral immune dysregulation has been proposed. Published comparisons rest on bulk transcriptomes or on tests that treat cells as replicates, and in public data disease is confounded with study. We reanalyzed four public single-cell RNA-seq cohorts of peripheral blood mononuclear cells (PBMCs; two PD, two pSS) with the donor as the unit of inference and cases compared with same-cohort controls. After removal of eight duplicated samples, 48 donors (24 controls, 10 PD, 14 pSS) and 317,205 cells in 28 subsets were analyzed for composition, pseudobulk expression and consensus regulon activity. A finding was considered replicated when significant in the pooled model (false discovery rate <0.05) and in both cohorts of a disease; differences between the PD and pSS effects were tested directly. PD showed no replicated change in any layer, and its transcriptome-wide perturbation did not exceed the permutation null. pSS showed a type I interferon program across lineages: STAT1(+), STAT2(+), ETV7(+) and IRF7(+) accounted for 25 of 44 replicated regulon–subset pairs, their targets were the interferon-stimulated genes that dominated the replicated differentially expressed genes, and an interferon score separated pSS from PD (area under the ROC curve 0.90–0.95). In the larger pSS cohort, 93% of the PBMC interferon signal arose from higher expression within cells, not from composition. Naive/central memory CD4 + T cells were reduced in both pSS cohorts. Cross-disease concordance was weak but exceeded permutation nulls (regulons, rho = 0.38, P = 0.031; genes, rho = 0.13, P = 0.012) and had two components: fewer CD56dim NK cells and more cells of a mixed T-like/adaptive-like NK subset in all four cohorts, and lower expression of glucocorticoid- and stress-responsive genes in cases across lineages. With pooled controls, the same data yielded “shared” regulons that did not survive cohort-aware testing, and 39%–68% of regulon–subset pairs differed between the controls of the PD and pSS studies. The replicated peripheral immune change is specific to pSS; what the two diseases share is limited to a compositional shift in the NK lineage and a stress-response signature of uncertain origin.