Figure 4.
Mass cytometry analysis of NK-cell maturation after haplo-HCT. Mass cytometry analysis was performed using cryopreserved PBMCs obtained 1 and 3 months after transplant (haplo-HCT n = 10, MD-HCT n = 6, MDs-HCT n = 4) and from HDs (n = 7). (A) FlowSOM analysis of all 1-month samples overlaid on viSNE map identified 20 distinct NK-cell metaclusters. (B) Heat map summarizes the expression of different markers in each NK-cell metacluster 1 month after HCT. (C) Relative representation of each NK-cell metacluster in haplo-HCT, MDs-HCT, and HD samples 1 month after HCT. Indicated columns compare representation of each meta-cluster in haplo-HCT vs HD and haplo-HCT vs MD-HCT samples. (D) One-month density viSNE map representation of all NK-cell metaclusters depicting data for all samples (n = 12), haplo-HCT (n = 6), HD (n = 3), and MD-HCT (n = 3). The location of each NK-cell metacluster is the same as in panel A. (E) FlowSOM analysis of all 3-month samples overlaid on viSNE map identified 16 distinct NK-cell metaclusters. (F) Heat map summarizes the expression of different markers in each NK metacluster 3 months after HCT. (G) Relative representation of each NK metacluster in haplo-HCT, MDs-HCT, MD-HCT, and HD samples. Indicated columns compare representation of each metacluster in haplo-HCT vs HD, haplo-HCT vs MDs-HCT, and haplo-HCT vs MD-HCT samples. (H) Three-month density viSNE map representation of all NK-cell metaclusters depicting data for all samples (n = 21), haplo-HCT (n = 8), HD (n = 4), MDs-HCT (n = 3), and MD-HCT (n = 6). For the heat maps median intensity was normalized for the highest value for each marker; red indicates the highest value, and blue represents the lowest value. For the tables shown in panels C and G, blue squares indicate a relative decrease, red squares represent a relative increase, and white squares indicate no change. Only statistically significant changes (P < .05) are reported. For comparison, the Wilcoxon rank-sum test for unpaired group was used. viSNE allows visualization of high-dimensional single-cell data and is based on the t-Distributed Stochastic Neighbor Embedding (t-SNE) algorithm.