De novo enrichment in H3K27ac in CLL. (A) Heat map depicting the H3K27ac signal intensities of genomic regions de novo enriched in H3K27ac in CLL. Signal intensities are presented as z scores (calculated per row) of the variance stabilizing transformation (VST) signal for CLL and B-cell samples. The mean log2 signal and the proportion of peaks detected in each sample are indicated below. The right bar shows the 2 signatures derived from hierarchical clustering. (B-C) Heat maps illustrating the H3K27ac signal intensities in signature S1 (B) and signature S2 (C). Signal intensities are represented as row z scores of VST signal. A box plot (right) compares the mean z scores between U-CLL and M-CLL samples in each cluster. Statistical significance was assessed using 2-tailed Student t tests. (D) Volcano plot showing hazard ratios (HRs) and P values for associations between H3K27ac enrichment in peaks and TTFT. Peaks from signatures S1 and S2 are differentiated by color. (E-F) The number of H3K27ac peaks (left) whose enrichment is associated with shorter or longer TTFT using univariate Cox scores in signature S1 (E) and signature S2 (F). Peaks’ association with shorter or longer TTFT was defined by the estimated HR >1 or <1, respectively. Kaplan-Meier curves (right) show the groups based on the H3K27ac enrichment levels of representative regions from each cluster; chr13:62.209.280-62.210.122 (E) and chr1:203.468.258-203-470.929 (F). Higher and lower H3K27ac groups were defined using the maxstat rank statistic–based cutoff. P values were obtained with log-rank tests. Note, HR and P values in panel D were computed with the coxph function from the survival31 R package, whereas the analyses in panels E-F used the samr30 R package for greater statistical power, potentially causing minor discrepancies between these analyses. ∗∗∗∗P < .0001. GCBC, germinal center B cell; MBC, memory B cell; ns, not significant; NBC-PB, naïve B cell from PB; NBC-T, naïve B cell from tonsil; PCT, plasma cell from tonsil.
