Post-transcriptional cross- and auto-regulation buffer expression of the human RNA helicases DDX3X and DDX3Y

Abstract

The Y-linked gene DDX3Y and its X-linked homolog DDX3X survived the evolution of the human sex chromosomes from ordinary autosomes. DDX3X encodes a multifunctional RNA helicase, with mutations causing developmental disorders and cancers. We find that, among X-linked genes with surviving Y homologs, DDX3X is extraordinarily dosage sensitive. Studying cells of individuals with sex chromosome aneuploidy, we observe that when the number of Y Chromosomes increases, DDX3X transcript levels fall; conversely, when the number of X Chromosomes increases, DDX3Y transcript levels fall. In 46,XY cells, CRISPRi knockdown of either DDX3X or DDX3Y causes transcript levels of the homologous gene to rise. In 46,XX cells, chemical inhibition of DDX3X protein activity elicits an increase in DDX3X transcript levels. Thus, perturbation of either DDX3X or DDX3Y expression is buffered: by negative cross-regulation of DDX3X and DDX3Y in 46,XY cells and by negative auto-regulation of DDX3X in 46,XX cells. DDX3X-DDX3Y cross-regulation is mediated through mRNA destabilization-as shown by metabolic labeling of newly transcribed RNA-and buffers total levels of DDX3X and DDX3Y protein in human cells. We infer that post-transcriptional auto-regulation of the ancestral (autosomal) DDX3X gene transmuted into auto- and cross-regulation of DDX3X and DDX3Y as these sex-linked genes evolved from ordinary alleles of their autosomal precursor.

Figure 1.

DDX3X is highly dosage-sensitive and expressed broadly among human tissues. (A) Among human X–Y pair genes, DDX3X ranks highest in combined sensitivity to overexpression (as judged by PCT percentile among all Chr X genes) and diminished function (as judged by LOEUF percentile among all Chr X genes). (B,C) DDX3X (B) and DDX3Y (C) and their chicken ortholog display the highest expression breadth among, respectively, the X and Y members of human X–Y pairs. Note that expression breadth data were not available for the chicken ortholog of KDM5C/D.

Figure 2.

DDX3X and DDX3Y transcript levels are negatively responsive to Chr Y and Chr X copy numbers, respectively. Scatterplots show DDX3X and DDX3Y transcript levels in cultured fibroblasts with the indicated sex chromosome constitutions. Each point represents a primary fibroblast culture from one individual. (A,B) DDX3Y transcript levels are significantly elevated and DDX3X transcript levels significantly reduced in fibroblasts with multiple Chr Y. (C,D) DDX3X transcript levels are significantly elevated and DDX3Y transcript levels significantly reduced in fibroblasts with multiple Chr X. R-values and statistical significance were calculated using Pearson's correlation.

Figure 3.

DDX3X and DDX3Y each respond to perturbations in the other's expression. (A) Schematic diagram of naturally occurring Chr Y (AZFa) microdeletion of DDX3Y and USP9Y. (B) DDX3X transcript levels are significantly higher in AZFa-deleted 46,XY LCLs compared with 46,XY LCLs with intact Chr Y. Each point represents a sample from one individual. Statistical significance determined by Mann–Whitney U-test: (***) P < 0.0001, (*) P < 0.05 (C) CRISPRi-mediated knockdown of DDX3Y using two independent gRNAs in three unrelated 46,XY fibroblast cultures results in significantly elevated DDX3X transcript levels. Conversely, DDX3X knockdown results in significantly elevated DDX3Y transcript levels. (D) Reanalysis of CRISPRi knockdown of ZFX or ZFY (San Roman et al. 2024) demonstrates that knockdown of either gene does not result in significant elevation of the homolog's transcripts. Statistical significance determined by ANOVA: (**) P < 0.001.

Figure 4.

Increased expression of DDX3X fully compensates, at transcript and protein levels, for CRISPRi knockdown of DDX3Y, but the inverse is not true. (A) Stacked bar graph showing summed TPM of DDX3X and DDX3Y transcripts in knockdowns using two independent gRNAs in three independent 46,XY fibroblast cultures. Statistical significance calculated by ANOVA: (**) P < 0.001. (B) Bar graph showing abundance of shared DDX3X and DDX3Y peptides in CRISPRi knockdowns with three technical replicates in two independent 46,XY fibroblast cultures. (C) Differential gene expression analysis of control versus DDX3X knockdown reveals significant expression changes in 397 target genes across the genome, including DDX3X. Genes with P < 0.05 (after multiple hypothesis correction) are indicated in blue, with exception of DDX3X (orange) and DDX3Y (purple). (D) Differential gene expression analysis of control versus DDX3Y knockdown reveals only six genes, including DDX3Y, that change significantly.

Figure 5.

DDX3X is negatively auto-regulated in 46,XX cells. (A) DDX3X’s allelic ratio (AR) is significantly higher than its ΔE_X value in LCLs, setting it apart from all other Xi/Xa/Y-expressed X–Y pair genes, whose AR values approximate their ΔE_X values. Statistical significance determined via one sample t-test: P = 0.02. (B) DDX3X transcript levels (by qPCR) in 46,XX fibroblasts are significantly elevated in a dose-responsive manner upon treatment with DDX3 helicase inhibitor RK-33. Statistical significance determined by one-sided t-test on delta Ct values. Error bars indicate the standard deviation of three technical replicates. (*) P < 0.05, (**) P < 0.01, (***) P < 0.001.

Figure 6.

DDX3X mRNA stability is regulated. (A) Schematic of experiment to determine half-lives of mRNAs. 46,XY and 49,XYYYY LCLs were incubated with 5-ethyl uridine (5-EU) to obtain nascent mRNAs. (B) DDX3X has an mRNA half-life of 0.5 h in 49,XYYYY versus 1.3 h in 46,XY LCLs.

Figure 7.

Not only protein-coding sequences but also gene regulatory mechanisms were preserved during the evolution of sex chromosomes from ordinary autosomes. The auto- and cross-regulation of DDX3X and DDX3Y reported here likely originated from the auto-regulation of ancestral (autosomal) DDX3X. Together with published studies of two other X–Y gene pairs—EIF1AX-EIF1AY and ZFX-ZFY (Godfrey et al. 2020; San Roman et al. 2024)—our findings suggest that an array of gene-specific regulatory schemes operative on the ancestral autosomes persist today on human Chr X and Chr Y.