IgMhi β7 integrinhi T2 cells are reduced in frequency in LN. (A) CITRUS trees generated from CD19+ cells from HCDs (n = 8) and LN patients (n = 8) and clustered according to the expression of CD5, CD9, CD10, CD24, CD27, CD38, CD45RB, IgD, IgM, and IgA (see also Fig. S1, A–C; and Fig. S4 A). CITRUS trees represent nodes redundantly so that the most peripheral nodes contain cell populations that are progressively shared by more central nodes with the central node containing all events. Red nodes indicate significantly different population abundances between HCDs and LN patients. A gray background is automatically assigned to aggregates of significant nodes. Arrows and roman numerals indicate nodes further analyzed in C. (B) CITRUS trees demonstrating the median expression of the clustering panel markers in the nodes. (C) Histograms demonstrating the abundance and expression of panel markers in selected nodes. The identity of these nodes can be inferred as (i) MZP, (ii) MZB, (iii) TS B cells, and (iv) class-switched IgA memory. (D) Minimal spanning tree generated by FlowSOM from events exported from CITRUS node (iii) representing TS B cells identified in A and C. This generated five metaclusters; the identification of each could be inferred by median expression of panel markers in each metacluster. (E) Minimal spanning trees generated by FlowSOM plots demonstrating CD38 expression in a representative HCD and LN patient. Prominent skewing of TS B cell subpopulations is evident in LN. (F) Relative abundances of metaclusters as a percentage of TS B cells indicate reduced frequency of events within metacluster 4 corresponding to IgMhi β7hi TS B cells in LN and increased frequency of events in metaclusters 1, 2, and 3 corresponding to T1 cells (medians, Mann–Whitney test). **, P < 0.01; ***, P < 0.001.