Termination Factor Rho and Its Cofactors NusA and NusG Silence Foreign DNA in E. coli

Abstract

Transcription of the bacterial genome by the RNA polymerase must terminate at specific points. Transcription can be terminated by Rho factor, an essential protein in enterobacteria. We utilized the antibiotic bicyclomycin, which inhibits Rho, to assess its role on a genome-wide scale. Rho is a global regulator of gene expression that matches E. coli transcription to translational needs. We found that genes in E. coli that are most repressed by Rho are prophages and other horizontally-acquired portions of the genome. Elimination of these foreign DNA elements increases resistance to bicyclomycin. Although rho remains essential, such reduced-genome bacteria no longer require Rho cofactors NusA and NusG. Deletion of the cryptic rac prophage in wild-type E. coli increases bicyclomycin resistance and permits deletion of nusG. Thus, Rho termination, supported by NusA and NusG, is required to suppress the toxic activity of foreign genes.

The complete genome sequence of E. coli revealed that 90% of its nucleotide sequence could encode protein (1). The remaining non-coding genome is densely packed with regulatory signals for transcription initiation and termination. This high information density requires that transcription terminate precisely at operon ends to avoid interference with neighboring transcription units. Based on sequence, approximately half of the transcription units, or operons, in E. coli are predicted to end with a specific structure, an intrinsic terminator, consisting of a hairpin followed by several U residues at the 3′ terminus of the RNA. This structure alone is sufficient to dissociate the polymerase elongation complex in vitro (2, 3). In contrast, transcription termination of the remaining half of operons could not be predicted from DNA sequence, and has been generally assumed to rely on an ATP-dependent RNA-DNA helicase known as Rho factor. In the decades since its discovery (4) Rho has been well-studied biochemically and structurally (2, 5–8), but its role as a biological regulator is still unclear. Rho factor recognizes no specific consensus but rather binds to naked, untranslated RNA, favoring C-rich sites that contain little secondary structure (9–11). Rho-dependent termination sites occur frequently in operons. For example, Rho can stop transcription when the end of the coding information is reached (12), attenuate transcription conditionally at the beginning of operons (13), and even within open reading frames when messenger RNA is uncovered by a nonsense mutation (14). In each case, the hypothesized roles of Rho are to prevent transcription from impinging on neighboring operons, to prevent the wasteful production of unusable transcripts, and to recycle polymerases promptly to locations where they are needed. However, since only a handful of Rho terminators (less than ten) have been actually located and characterized (15), there is still much to be learned about the role of Rho-dependent termination in vivo.

To investigate the biological role of Rho, we assayed gene expression using the Affymetrix E. coli Genome 2.0 array, an in-situ synthesized oligonucleotide array covering the entire genome of four evolutionarily divergent E. coli strains: the laboratory strain K-12 MG1655, the enterohemorrhagic strains O157:H7 (EDL933 and Sakai), and uropathogenic CFT073. Specific and potent inhibition of Rho can be achieved rapidly by treatment with the antibiotic bicyclomycin (BCM) (16). An advantage of chemical over genetic intervention is that the transcriptome content of control and experimental cultures remains identical until the moment the inhibitor is added. Indeed, total inhibition of Rho termination activity cannot be achieved by genetic manipulation since rho is an essential gene (17). BCM is highly specific to Rho; it rapidly permeates cells and has no other known in vivo targets (16, 18). Changes in gene expression in response to BCM reflect, therefore, a snapshot of Rho activity. Treatment of MG1655 with a series of concentrations of BCM for short time intervals revealed a pervasive change in gene expression (Fig. 1). One theme which emerges from the array data is a widespread increase in the expression of genes derived from recent horizontal transfer into the genome of K-12 from other species or from defective bacteriophage (Fig. 1B, 1D, and S2B). Based on whole-genome alignment, approximately 14–18% of the K-12 genome differs from other families of E. coli, tending to occur in contiguous blocks known as K-islands (19, 20). K-islands are characterized by an altered GC/AT content, distinct codon preference, and reduced evolutionary conservation. The genomic islands are enriched in defective prophages, transposons, and insertion sequences (21). Comparing MG1655 with the enterohemorrhagic strain O157:H7 (EDL933) shows that the two strains possess 3658 genes that are nearly identical in sequence, as well as 648 and 1769 unique genes, respectively (22). As shown in Fig. 1B and D, the genes unique to each strain and prophage genes tended to be upregulated, with half of these genes increasing expression more than threefold. By contrast, a quarter of orthologous genes, common between the two strains, were upregulated greater than threefold (compare orthologous and K-12-specific genes in Fig 1E).

Fig. 1. Genomic response of divergent E. coli strains to Rho inhibition.

(A) Hierarchical cluster analysis of a concentration gradient of BCM (doses of 10, 25, and 100 μg/ml) in E. coli K-12 strain MG1655 showing only genes orthologous between K-12 and enterohemorrhagic E. coli. Arrays (columns) are shown in biological duplicates, normalized so that average of each gene on the untreated control arrays is equal to 1 and expression in treated cultures is displayed as a ratio of treated to untreated. Yellow blocks represent upregulation by BCM and blue represents downregulation. (B) Response to BCM of genes present in K-12 E. coli but absent from enterohemorrhagic E. coli, displayed as in (A). (C) Hierarchical cluster analysis of the response of orthologous genes in enterohemorrhagic E. coli O157:H7 strain EDL933. (D) Response to BCM treatment of genes present in enterohemorrhagic E. coli but absent from K-12. (E) Expression of ORFs in response to BCM displayed as a scatterplot of probe intensity in the control array (X-axis) and BCM-treated array (Y-axis) from a representative pair of arrays. The diagonal line represents equal probe hybridization intensity between both arrays; points above the diagonal are genes upregulated by treatment with BCM and points below the diagonal are downregulated. The red lines located at 100 intensity units represent the threshold below which probe-level analysis is 90% likely to call the probe absent. Therefore, probes in the upper left quadrant are ORFs whose expression was induced de novo. Grey points are orthologous genes and violet points are K-12-specific genes. (F) Scatterplot of probe intensity for intergenic (IG) regions of MG1655 after treatment by BCM.

We find that expression of the non-coding intergenic (IG) regions is in general increased by Rho inhibition (Fig. 1F). Of the IG probes which were reproducibly measured, as selected by Significance Analysis of Microarrays at the 1% false discovery rate (23), 72% were increased at least three-fold and only 1% were decreased at least three-fold. The general upregulation of IG regions confirms that Rho has a global role in preventing synthesis of untranslated transcripts. Taken together, the array data from BCM treatment of E. coli indicate that Rho is intimately involved in operon regulation throughout the genome and is not only acting on a rare subset of genes or when translation terminates abnormally.

We next sought to determine if this extensive perturbation in the transcriptome was reflected in the proteome. We utilized the technique of difference gel electrophoresis (DIGE) to analyze the protein complement of MG1655 cells treated under the same conditions used in the microarray experiments (24). The workflow for this analysis is shown in Fig. S6A. Two-dimensional gels of fluorescently-labeled proteins show that of 3341 unique spots analyzed, 101 were increased greater than two-fold and 8 were decreased greater than two-fold by BCM treatment. Altered spots were robotically excised from gels and the proteins were identified by mass spectrometry (Fig. 2). As shown in Fig. 2 and Tables S1 and S2, among the most affected unique proteins is Rho itself and the RecE protein of the Rac prophage. For reasons not understood, most of the other proteins identified are involved in anaerobic metabolism and the response to acidic pH. Based on the microarray result, that many de novo transcripts of unique genes were being produced, we expected to see many new spots appearing on the gel from the BCM treated sample. However, this did not occur. The proteomic results corroborate the role of Rho as a general inhibitor of transcription under normal growth conditions. There is a profound excess of transcriptional output over translational needs when Rho activity is reduced. The lack of perturbation of the proteome also suggests that protein expression is frequently controlled post-transcriptionally.

Fig. 2. Proteomic response to Rho inhibition as detected by difference gel electrophoresis (DIGE).

Two-dimensional electrophoresis gel of protein extracted from BCM-treated and control cultures. Control protein is pseudocolored green and BCM-treated protein as red. Differentially-expressed proteins (indicated by name) were identified by mass spectrometry.

Since Rho strongly represses transcription of horizontally-transferred genes, we investigated a synthetic E. coli strain, MDS42, that lacks these genes. 14% of the MDS42 genome has been removed by targeted deletion of prophages, IS elements, and K-island clusters (25, 26). Fig. 3B shows that MDS42 was ~10⁴-fold more resistant to BCM (25 μg/ml) than the parent strain, MG1655. MG1655 contains the remnants of a lambdoid bacteriophage known as rac. This defective prophage carries a kil gene encoding an inhibitor of cell division. Deletion of rac alone produced levels of BCM resistance comparable to the MDS42 strain. The resistance was conferred by deletion of kil and the remaining downstream operon, but not DNA downstream of kil (Fig. 3B).

Fig. 3. Reduced-genome E. coli is resistant to Rho inhibition and deletion of elongation factors NusA and NusG.

(A) Hierarchical cluster analysis of ORF gene expression in strain MDS42, BCM-treated MDS42, MDS42 ΔnusA, and MDS42 ΔnusG. Probe intensity is normalized to the untreated MDS42 strain. (B) Efficiency of colony formation assay of the indicated strains. Cultures at dilutions of 10⁻², 10⁻⁴, and 10⁻⁶ were spotted onto a control plate or a plate containing BCM at 25 μg/ml. int-kilR and int-ydaE are fragments of the rac prophage that were deleted.

Studies of λ phage revealed that endogenous host proteins NusA, NusB, NusE (ribosomal protein S10) and NusG were required for the λ N protein to suppress transcription termination on the phage chromosome (27, 28). NusA and NusG have been implicated in both Rho-dependent and intrinsic termination and are essential for E. coli to survive (29, 30). It was possible to delete both nusA and nusG in strain MDS42, although stationary phase survival was poor, and the strains were highly sensitive to BCM (Fig. 3B). Deleting nusA or nusG also adversely affected growth rate, increasing the doubling time in rich media from 32 min to 57 min and 68 min, respectively. Unexpectedly, it was possible to transfer the nusG knockout allele to a wild-type MG1655 strain lacking the rac prophage alone, indicating that suppression of rac gene expression is the critical function of NusG.

Strains lacking NusA or NusG are highly similar in their overall pattern of gene expression as shown by the hierarchical cluster analysis in Fig. 3A and the scatterplot of intergenic region expression in supplemental Fig. S2C and D. We therefore conclude that these proteins normally act in concert, recognizing the same elongation and termination signals.

To understand the basis of how Rho inhibition could affect gene expression on such a pervasive scale we examined two specific operons, one in the rac prophage and one in λ. The maps in Fig. 4A and B show that the leftward operons of rac and λ are homologous, implying that there should be a Rho-dependent terminator (t_imm) in rac following the racR gene, as there is in λ. Addition of BCM enables the RNA polymerase to continue through this terminator and express downstream genes, including the toxic kil gene. RT-PCR analysis reveals the elongated transcript (Fig. 4C). Similarly, the leftward operon of λ (Fig. 4B) exhibits readthrough of the t_imm terminator into the downstream N::lacZ reporter fusion in the presence of BCM (Fig. 4D). The HK022 Nun termination protein, which blocks transcription elongation at the λ nutL site, prevents reporter gene expression, consistent with a transcript originating from the P_RM promoter (31).

Figure 4. Effect of BCM on the leftward operons of rac and λ phages.

(A) Map of the leftward operon of the rac prophage. Gray arrows: genes; open-headed arrows: PCR primers; brackets: deletions of intR-ydaE and intR-kil; bent arrow: the operon’s promoter, (P_RM). (B) Map of the homologous operon of λ phage. Dashed lines show proposed transcripts produced. (C) RT-PCR using primer pairs indicated on the map in panel A shows that BCM treatment yields an elongated transcript. RT: reverse transcriptase. (D) Average β-galactosidase activity (lacZ expression) from the phage strain shown in panel B. The standard error is less than 5%.

As shown by the maps of prophages in Fig S4, genes that are upregulated by treatment with BCM tend to occur in consecutive series in the same strand orientation, suggesting that preventing read-through into neighboring operons is an important function of Rho. Rho’s bias toward suppressing foreign DNA could be related to the lower density of the Rho-independent intrinsic terminators in the K-island regions. Utilizing the terminator-prediction model of Lesnik et al. (32), there is an intrinsic terminator on average every 4.0 kb in the conserved regions of the genome, but only every 8.5 kb in the K-islands (Table S7) (32). Moreover, the genes upregulated by treatment with BCM tend to be more AT rich than the genome as a whole (Fig. S1) and have a lower Codon Adaptation Index (Fig. S5B). The lower secondary structure of AT-rich RNA could make it a favored target of Rho-dependent termination in spite of Rho’s in vitro binding affinity for C-rich RNA, whereas the suboptimality of translation in genes with poor codon preference leaves them open to Rho.

In summary, our results reveal Rho factor as a global regulator of bacterial gene expression under normal growth conditions. Rho serves the crucial role of maintaining transcriptional boundaries throughout the genome. In particular, Rho is responsible for silencing horizontally transferred DNA elements, some of which are detrimental to the host. Recently, H-NS protein has been implicated in selective silencing of foreign DNA in Salmonella by acting at the level of promoter initiation (33). Rho-dependent termination may represent a separate “immunity” system that protects bacterial cells from the harmful activity of certain foreign genes. The existence of such different defensive tools against new acquisitions to the genome underscores the importance of this phenomenon for bacterial evolution.