Rho and NusG suppress pervasive antisense transcription in Escherichia coli

Abstract

Despite the prevalence of antisense transcripts in bacterial transcriptomes, little is known about how their synthesis is controlled. We report that a major function of the Escherichia coli termination factor Rho and its cofactor, NusG, is suppression of ubiquitous antisense transcription genome-wide. Rho binds C-rich unstructured nascent RNA (high C/G ratio) prior to its ATP-dependent dissociation of transcription complexes. NusG is required for efficient termination at minority subsets (~20%) of both antisense and sense Rho-dependent terminators with lower C/G ratio sequences. In contrast, a widely studied nusA deletion proposed to compromise Rho-dependent termination had no effect on antisense or sense Rho-dependent terminators in vivo. Global colocalization of the histone-like nucleoid-structuring protein (H-NS) with Rho-dependent terminators and genetic interactions between hns and rho suggest that H-NS aids Rho in suppression of antisense transcription. The combined actions of Rho, NusG, and H-NS appear to be analogous to the Sen1-Nrd1-Nab3 and nucleosome systems that suppress antisense transcription in eukaryotes.

Figure 1.

Regulators of transcript elongation in bacteria. The EC is comprised of RNAP (β′βα₂ω subunits), DNA template, and RNA transcript. The Rho hexamer binds nascent RNA in primary sites on each subunit and secondary site in the central pore. NusG contacts RNAP via its N-terminal domain (NTD) and Rho via a C-terminal domain (CTD) connected by a flexible linker. NusA binds RNAP via an NTD near the RNA exit channel and contains additional domains (KH1, KH2, S1, and a CTD present in some bacteria).

Figure 2.

Genome-wide analysis of Rho-dependent transcription termination. (A) Distribution of Rho-dependent terminators in the E. coli genome as detected by BSTs. Features on the + strand are shown above the solid black lines, and features on the − strand are shown below the lines. Genes are depicted as black boxes, and transcript abundance detected by tiling arrays is shown as blue bar graphs. Rho-dependent termination sites are shown as colored boxes; the colors correspond to BST annotations from B. (B) Rho-dependent terminator annotations. Terminators are divided into eight classes based on their locations relative to annotated genes. The diagram immediately to the right of the class number illustrates the orientation of transcripts and termination sites relative to annotated genes. Light-blue arrows represent genes, and the arrows point in the direction of translation of the gene. The blue line indicates the length of the transcript in untreated cells, and the violet dashed line shows extension of the transcript in BCM-treated cells. The bar graph to the right of the gene diagram shows the number and percentage (rounded) of terminators in each class. (C) Effects of Rho inhibition on sense and antisense transcription. The log₂ ratios of normalized read counts in BCM-treated versus untreated conditions are shown for the coding strand (sense) or the strand opposite the coding strand (antisense) for all E. coli K-12 genes in biological duplicate. Each row represents one gene. Fold increases in transcript abundance due to BCM treatment are shown in yellow, and decreases are shown in blue. The genes are clustered based on similar patterns of effects on antisense and sense transcription into arbitrarily ordered clusters by centroid linkage clustering using a Euclidean distance metric (see the Supplemental Material; Eisen et al. 1998).

Figure 3.

Effects of Rho inhibition at class I and class II Rho-dependent terminators. (A) BCM effects on the class I Rho-dependent terminator at the grxD locus. A statistically significant increase in transcript levels (magenta dashed brackets) between untreated (blue bars) and BCM-treated (violet bars) cells occurs at the 3′ end of the grxD gene, indicating that the grxD transcript is terminated by Rho. (B) BCM effects on two class II Rho-dependent terminators at the bgl locus. Rho terminates antisense transcripts that arise from within the bglF and blgH genes. Note that the bglH antisense transcript is essentially undetectable in untreated conditions.

Figure 4.

Spatial and functional associations between H-NS and Rho-dependent termination. (A) H-NS binding near Rho-dependent terminators. The median H-NS ChIP signal (see the Supplemental Material) is shown at specified distances from the 5′ end of all BSTs (violet), class II BSTs (black), or random chromosome positions (gray). (B) H-NS binding at the class I Rho-dependent terminator at the yhhJ locus. Colors are as in Figure 3, except that H-NS ChIP–chip data are shown in orange. Readthrough of the Rho-dependent terminator at the end of the yhhJ gene (indicated by the appearance of a BST shown as magenta dashed brackets) corresponds with a peak in H-NS ChIP–chip occupancy (orange). (C) H-NS binding at the class II Rho-dependent terminator within the xylG gene. Readthrough of the Rho-dependent terminator on the antisense strand of xylG corresponds with a peak in H-NS occupancy. (D) Genetic interactions between hns and rho. Fitness is expressed as the ratio of the doubling times (D) for wild-type (D_WT = 24.75 ± 0.33 min) versus mutant cells in liquid LB medium. Wild-type fitness is set at one. Predicted fitness of double mutants is based on the multiplicative model (fitness of mutant #1 × fitness of mutant #2 = predicted fitness of double mutant) (St Onge et al. 2007).

Figure 5.

Effects of ΔnusG and ΔnusA* on Rho-dependent termination. (A) Effects of ΔnusG and ΔnusA* on expression within MDS42 BSTs. The log₂ ratio of median intensity within MDS42 BSTs in BCM-treated or mutant cells versus untreated cells is shown in biological duplicate. Each row represents one MDS42 BST. Fold increases in transcript abundance due to either BCM treatment—ΔnusG or ΔnusA*—are shown in yellow, and decreases are shown in blue. The genes are clustered based on similar patterns of effects on antisense and sense transcription into arbitrarily ordered clusters by centroid linkage clustering using a Euclidean distance metric (see the Supplemental Material; Eisen et al. 1998). (B) Effects of ΔnusG or ΔnusA* on the class I Rho-dependent terminator at the grxD locus. Neither ΔnusG nor ΔnusA* has a significant effect on transcript abundance at the 3′ end of the grxD gene. Colors are as in Figure 3, except that transcripts from ΔnusG cells are shown in green, and transcripts from ΔnusA* cells are shown in red. (C) Effects of ΔnusG or ΔnusA* on the class II Rho-dependent terminator within the bglF gene. ΔnusG had a significant effect on transcript abundance of the bglF antisense transcript, but ΔnusA* had no significant effect. Note that the ΔnusG effect on termination of the bglF antisense transcript was not as potent as direct inhibition of Rho with BCM.

Figure 6.

Basis for NusG effects on Rho-dependent termination. (A) NusG enhancement of termination is not associated with terminator class. All BSTs and BSTs with an overlapping significant ΔnusG effect were not statistically distinguishable by terminator class (Fisher's exact test; P = 0.21). Colors are as in Figure 2B. (B) NusG enhancement of termination is associated with sequences present at the termination site. The median C/G ratio (see the Supplemental Material) is shown at specified distances from the 5′ end of all BSTs (violet), NusG-independent BSTs (black), NusG-dependent BSTs (green), or random chromosome positions (gray). The orange “pacman” represents putative processing of transcripts by 3′ → 5′ exonucleases.

Figure 7.

Models of antisense transcription termination by Rho. Rho terminates transcription at the end of genes, preventing antisense transcription into downstream genes (readthrough antisense, or class I terminators). Rho also terminates antisense transcription arising from within genes (internal antisense, or class II terminators). Termination sites are more C-rich and G-poor than random genomic DNA and thus facilitate Rho loading onto the nascent RNA (violet box labeled “C>G”). H-NS (orange ovals) is typically bound adjacent to sites of Rho termination (H-NS-binding sites are shown as orange boxes labeled “H-NS”) and is functionally synergistic with Rho. NusG enhances Rho termination at termination sites with reduced C and increased G content, which account for less than a quarter of all Rho-dependent terminators (dashed black arrow). The RNA-binding domains of NusA do not affect Rho termination of antisense transcription (black crossout). Class I terminators can also be associated with REP elements, which may stabilize the terminated mRNA (light-blue box labeled “REP”).