Deep immune profiling by mass cytometry links human T and NK cell differentiation and cytotoxic molecule expression patterns

Abstract

The elimination of infected or tumor cells by direct lysis is a key T and NK cell effector function. T and NK cells can kill target cells by coordinated secretion of cytotoxic granules containing one or both pore-forming proteins, perforin and granulysin and combinations of granzyme (Gzm) family effector proteases (in humans: Gzm A, B, K, M and H). Understanding the pattern of expression of cytotoxic molecules and the relationship to different states of T and NK cells may have direct relevance for immune responses in autoimmunity, infectious disease and cancer. Approaches capable of simultaneously evaluating expression of multiple cytotoxic molecules with detailed information on T and NK differentiation state, however, remain limited. Here, we established a high dimensional mass cytometry approach to comprehensively interrogate single cell proteomic expression of cytotoxic programs and lymphocyte differentiation. This assay identified a coordinated expression pattern of cytotoxic molecules linked to CD8 T cell differentiation stages. Coordinated high expression of perforin, granulysin, Gzm A, Gzm B and Gzm M was associated with markers of late effector memory differentiation and expression of chemokine receptor CX3CR1. However, classical gating and dimensionality reduction approaches also identified other discordant patterns of cytotoxic molecule expression in CD8 T cells, including reduced perforin, but high Gzm A, Gzm K and Gzm M expression. When applied to non-CD8 T cells, this assay identified different patterns of cytotoxic molecule co-expression by CD56^hi versus CD56^dim defined NK cell developmental stages; in CD4 T cells, low expression of cytotoxic molecules was found mainly in TH1 phenotype cells, but not in Tregs or T follicular helper cells (TFH). Thus, this comprehensive, single cell, proteomic assessment of cytotoxic protein co-expression patterns demonstrates specialized cytotoxic programs in T cells and NK cells linked to their differentiation stages. Such comprehensive cytotoxic profiling may identify distinct patterns of cytotoxic potential relevant for specific infections, autoimmunity or tumor settings.

Keywords: Cytotoxic; Granzymes and perforin; Mass cytometry; NK cell; T cell; T cell differentiation.

Fig. 1. Lineage gating

PBMC from healthy donors were stained with the panel outlined in Table 1 and analyzed by mass cytometry. (A) Representative example of the gating strategy used to identify lymphocyte lineages and subpopulations. (B) (upper) Representative examples of the gates used to identify naïve T cells (positive for CD27, CD45RA and CCR7). (lower) overlay plot of CD27+CD45RA+CCR7+ naïve phenotype cells (orange) over total CD8 cells, indicating little expression of markers associated with non-naïve T cells (CD95, CD57, T-bet, Eomes, PD-1 and TIGIT). (C) Healthy donors were analyzed for T cell and NK cell populations. Frequencies are displayed by box and whiskers plots with 5–95% interval.

Fig. 2. Cytotoxic molecule expression on T and NK cells

(A) Representative example of the gating strategy used to identify expression of cytotoxic molecules. Example is gated on CD8 T cells. (B) Scattered dot plots indicating frequency of cytotoxic molecule expression by T cell and NK cell subpopulations in the peripheral blood (population colors: CD4-blue, CD8-black, NK-red).

Fig. 3. Cytotoxic molecule expression patterns are linked to CD8 T cell differentiation

(A) CD8 T cell differentiation subsets were distinguished using a classification based on CD27, CD45RA and CCR7. (B) Scattered dot plots indicating frequency of cytotoxic molecule expression by CD8 T cell differentiation populations in the peripheral blood reveals distinct patterns of cytotoxic molecule expression along developmental trajectories.

Fig. 4. Distinct coexpression patterns of cytotoxic molecules partly overlap across lineages

(A) Representative bivariate coexpression plots of cytotoxic molecules analyzed. CD27+CD45RA+CCR7+ naïve phenotype cells (blue) were overlaid on total CD8 T cells (red). (B) tSNE-based dimension reduction with viSNE was performed on CD8 T cells using the high-dimensional information of cytotoxic molecule channels. Expression of cytotoxic molecules is presented as heatmap overlay over cells arranged by their tSNE coordinates to illustrate high-dimensional population features in two dimensions (upper two rows). This approach reveals populations of CD8 T cells with distinct coexpression patterns that are linked to the expression of CX3CR1, CD57 and CD28 (not used for the viSNE analysis). An extended analysis in (C) was performed on CD8 and NK cells to facilitate cross-comparison. (D) Populations of CD56^hi and CD56^dim CD16+ cells were analyzed by the tSNE dimensions, revealing distinct, shared, but less diverse cytotoxic phenotypes compared to CD8 T cells. (E) Overlay heatmaps of Perforin, Gzm B and Gzm K reveal key differences between CD56^hi and CD56^dim CD16+ NK cells.