Genome-wide analysis reveals regulatory role of G4 DNA in gene transcription

Abstract

G-quadruplex or G4 DNA, a four-stranded DNA structure formed in G-rich sequences, has been hypothesized to be a structural motif involved in gene regulation. In this study, we examined the regulatory role of potential G4 DNA motifs (PG4Ms) located in the putative transcriptional regulatory region (TRR, -500 to +500) of genes across the human genome. We found that PG4Ms in the 500-bp region downstream of the annotated transcription start site (TSS; PG4M(D500)) are associated with gene expression. Generally, PG4M(D500)-positive genes are expressed at higher levels than PG4M(D500)-negative genes, and an increased number of PG4M(D500) provides a cumulative effect. This observation was validated by controlling for attributes, including gene family, function, and promoter similarity. We also observed an asymmetric pattern of PG4M(D500) distribution between strands, whereby the frequency of PG4M(D500) in the coding strand is generally higher than that in the template strand. Further analysis showed that the presence of PG4M(D500) and its strand asymmetry are associated with significant enrichment of RNAP II at the putative TRR. On the basis of these results, we propose a model of G4 DNA-mediated stimulation of transcription with the hypothesis that PG4M(D500) contributes to gene transcription by maintaining the DNA in an open conformation, while the asymmetric distribution of PG4M(D500) considerably reduces the probability of blocking the progression of the RNA polymerase complex on the template strand. Our findings provide a comprehensive view of the regulatory function of G4 DNA in gene transcription.

Figure 1.

Relationship between PG4Ms and gene expression level. Changes in the frequency of PG4M_U500 (A) and PG4M_D500 (B) with the increased gene expression level are plotted. Genes are ranked by expression level, with each point representing the mean expression level and the frequency of PG4Ms calculated for every 100 genes.

Figure 2.

Influence of PG4M_D500 on gene expression for each tissue/cell. Comparison of the gene expression levels between PG4M_D500-positive (black squares) and PG4M_D500-negative genes (open squares) in each human tissue/cell type. Error bars represent the 95% confidence interval of the mean expression level.

Figure 3.

Comparison of expression levels between PG4M_D500-positive and PG4M_D500-negative genes when gene family and function are controlled. Human genes are clustered according to gene family (A) and gene function (B), and mean expression levels are compared between PG4M_D500-positive (X-axis) and PG4M_D500-negative genes (Y-axis) in each cluster. The angle bisector represents equal expression levels. The squares below the angle bisector indicate that the expression level of PG4M_D500-positive genes is higher than that of PG4M_D500-negative genes, and vice versa.

Figure 4.

Strand asymmetry of PG4M_D500. (A) Comparison of the frequency between PG4M_cod (gray line) and PG4M_tem (black line) in the10-kb TSS-flanking region of human RefSeq genes (Ref) (n = 13,276) and pseudogenes (Pse) (n = 824). (B) Ratios of genes containing (1) PG4M_cod but not PG4M_tem (PG4M_cod+tem–); (2) PG4M_tem but not PG4M_cod (PG4M_cod–tem+); (3) more PG4M_cod than PG4M_tem (PG4M_cod>tem); and (4) fewer PG4M_cod than PG4M_tem (PG4M_cod<tem) in the D500 regions (black squares). The corresponding values for PG4M_U500 are plotted as a control (gray squares).

Figure 5.

A model of G4 DNA-mediated stimulation of transcription. Double-stranded DNA is denatured locally during transcription, forming a transcription bubble and exposing the template strand for nascent RNA synthesis (A, top). As RNAP II machinery (gray dashed box) moves along the template strand, the bubble moves with it, and the DNA rewinds to form the duplex behind the bubble (A, bottom). For the PG4M_D500-positive genes, the transient separation of the duplex DNA during transcription and transcription-derived negative supercoiling considerably increases the opportunities for the formation of G4 DNA. Such high-order structures are extremely stable, and thus, probably block rehybridization with the complementary strand, holding the structure open and thereby enabling a higher rate of transcription (B). For the same reason, the presence of high numbers of PG4M_D500 can help maintain the initial region of the transcript unpaired, which could facilitate the reinitiation of transcription (black dashed box) and also contribute to a higher level of transcription (B, bottom). The presence of PG4Ms in the template strand may hinder the progression of RNAP II. However, the intrinsic asymmetric distribution of PG4Ms between strands in the D500 region (PG4M_D500cod > PG4M_D500tem, represented in gray and black, respectively; see also Fig. 4) achieves a balance and minimizes the negative effects.