Defective chemotaxis in GPI^negNK cells. (A) Cell surface expression of CD48, CXCR4, and CD15s on NK92 cells, with and without enzyme treatment. PI-PLC hydrolyses GPI anchors (removing CD48 but preserving CXCR4 and CD15s), whereas sialidase treatment leaves CD48 and CXCR4 expression intact but removes sialic acid and destroys the CD15s antigen. (B) Migration of enzyme-treated NK92 cells (left) or primary NK cells (right) to SDF-1. Cells were treated with either sialidase (S) or PI-PLC (P) or were left untreated (U). The data were analyzed using the Mann-Whitney nonparametric test. SDF-1 induced migration by a factor of 4 over media lacking the chemokine (data not shown). (C) Migration of NK cells from 6 PNH patients in response to SDF-1. Pie charts show the percentage of GPI+ (black slice) and GPI^neg (white slice) NK cells applied to the assay (top; input population) and those that migrated in response to the chemokine (bottom; migrated population). Patient p1.38 was an addition to the 47 patients in the cohort shown. (D) Analysis of migration of NK cells from the 6 PNH patients according to the size of the GPI+ population in the input and migrated fractions. The P value was calculated using the Wilcoxon matched-pairs signed rank test. (E) CXCR4, CCR7, and CXCR3 expression on the GPI+ and GPI^neg NK cells of 7 PNH patients. The lines link the NK cells from a particular patient. The data show the percentage of NK cells expressing these chemokine receptors (determined by flow cytometry) as a fraction of the total GPI+ or GPI^neg NK cell population. Analysis of the expression of these molecules on the CD56^dim and CD56^bright NK cell subsets is shown in supplemental Figure 3. P values were calculated using Wilcoxon matched-pairs signed rank test.