Identification of LSC-associated genes by RNA-Seq and in vivo data. (A) Experimental design (see our “Materials and methods” section for details). i.f., intrafemoral. (B) High LSC frequency is associated with poor survival (median overall survival 143 vs 369 days, P = .02 (log-rank), HR 3.2 (95% CI, 1.2-8.7), n = 56. (C, left) Average gene expression in LSC^high vs LSC^low samples (log(RPKM+0.001) transformed). GPR56 is highlighted in red. (Right) Volcano plot displaying the average fold difference in gene expression in LSC^high (n = 10) vs LSC^low samples (n = 26) for all genes with an average log10(RPKM+0.001) ≥1 in LSC^high samples in relation to the P value for each gene. P values were transformed as –log10 (P value). Candidate LSC genes are shown in red (criteria: average (log10(RPKM) in LSC^high samples ≥1, average ratio LSC^high/LSC^low ≥2, P < .01 [Mann-Whitney U test]). Font and dot size indicate the average expression level of the genes in LSC^high samples. (D) GPR56 gene expression in hematopoietic cell populations. (E) Waterfall plot displaying correlation coefficients for all genes correlated to GPR56 based on RNA-Seq data from 437 AML samples. Highlighted in red are selected genes positively correlating with GPR56. Blue-annotated genes inversely correlate with GPR56. (F) Correlation between GPR56 gene expression by RPKM and percentage of positive cells measured by flow cytometry in 45 AML samples (log-scale). (G) Fractions of GPR56 and CD34⁺ subpopulations determined by flow cytometry in samples with high, medium, and low LSC frequency. Box-and-whisker plots displaying median, lower (Q1) and upper (Q3) quartile, minimum and maximum without outliers, and outliers identified as below Q1 − 1.5 (IQR) or above Q3 + 1.5 (IQR) (Tukey method). IQR, interquartile range = Q3-Q1.