A Screen for rfaH Suppressors Reveals a Key Role for a Connector Region of Termination Factor Rho

ABSTRACT RfaH activates horizontally acquired operons that encode lipopolysaccharide core components, pili, toxins, and capsules. Unlike its paralog NusG, which potentiates Rho-mediated silencing, RfaH strongly inhibits Rho. RfaH is recruited to its target operons via a network of contacts with an elongating RNA polymerase (RNAP) and a specific DNA element called ops to modify RNAP into a pause- and NusG-resistant state. rfaH null mutations confer hypersensitivity to antibiotics and detergents, altered susceptibility to bacteriophages, and defects in virulence. Here, we carried out a selection for suppressors that restore the ability of a ΔrfaH mutant Escherichia coli strain to grow in the presence of sodium dodecyl sulfate. We isolated rho, rpoC, and hns suppressor mutants with changes in regions previously shown to be important for their function. In addition, we identified mutants with changes in an unstructured region that connects the primary RNA-binding and helicase domains of Rho. The connector mutants display strong defects in vivo, consistent with their ability to compensate for the loss of RfaH, and act synergistically with bicyclomycin (BCM), which has been recently shown to inhibit Rho transformation into a translocation-competent state. We hypothesize that the flexible connector permits the reorientation of Rho domains and serves as a target for factors that control the motor function of Rho allosterically. Our results, together with the existing data, support a model in which the connector segment plays a hitherto overlooked role in the regulation of Rho-dependent termination.

a similar I382N substitution dramatically increased readthrough in vivo but not in vitro (11). In addition to these "expected" mutations, we isolated four mutations (two of them more than once) in a nonconserved region of Rho that connects the N-terminal and C-terminal domains (NTD and CTD, respectively) of Rho. These changes are immediately upstream and within the ␣ 6 helix in the CTD (residues 153 to 166) that is adjacent to the connector. Mutations in this region have not been reported previously; we hypothesize that these changes alter domain rearrangements required for Rho function (see Discussion). All rho alleles conferred increased sensitivity to novobiocin (see Fig. S2), consistent with a previous report (14).
Most SDS-resistant mutations are in the hns gene. To map the remaining mutants, we used random Tn5::Tet libraries (21). Using this approach, we mapped 16 mutations to hns (7 unique) and 1 to yciC (see Table S1). The remaining suppressor could not be mapped with different Tn5 libraries, suggesting that it contained more than one mutation required to confer the SDS resistance phenotype. The yciC gene encodes a predicted inner membrane protein. It is unclear how a missense mutation (A83T) in yciC restores SDS resistance in the rfaH background, as the function of YciC is unknown.
Mutations in hns could affect the expression of RfaH-dependent genes in several ways. First, H-NS silences the transcription of the waa operon (22) and loss-of-function hns mutants could compensate for the lack of RfaH by increasing the amount of waa RNA. Second, H-NS may cooperate with Rho to inhibit the expression of some genes (23). Changes at the dimer interface, which encompasses residues 2 to 47 and 58 to 84, will be expected to alter the structure of H-NS filaments, in turn affecting Rho function. Surprisingly, although L26P and E74K H-NS variants have been shown to restore viability to strains carrying defective rho and nusG alleles (24) and are thus expected to increase Rho-dependent termination, we isolated substitutions at identical (L26P) and  (15) indicated. The NTD, CTD, and connector are red, gray, and black, respectively. The key Rho elements are shown as colored boxes. The suppressor mutants isolated in this work are shown below the sequence. BS, binding site; ER, a motif of two residues, Glu and Arg. (B) Structure of the Rho monomer (PDB code 3ICE). The NTD, CTD, and key CTD regions are colored as in panel A; the bound ATP (orange) and RNA in the central pore (blue) are also shown. The C␣ atoms of the mutated residues are shown as red spheres. (C) The apo Rho hexamer exists in an open state in solution and undergoes a rigid-body rearrangement into the closed state in the presence of RNA and ATP (27), trapping the RNA in the central channel. The CC dinucleotide contacts with the NTD favor the transition, whereas BCM inhibits it (27). Other Rho ligands could affect the conformational change (as shown by a question mark). adjacent (L75Q) positions that likely reduce Rho-dependent polarity in the waa operon. Furthermore, the hns suppressor alleles were different in the extent of SDS resistance (see Table S1) and sensitivity to novobiocin (see Fig. S2), implying that their underlying mechanisms are distinct. Elucidation of the mechanism of suppression will necessitate an in-depth analysis of the effects of hns and other nucleoid-associated proteins on the regulation of RfaH-controlled genes, which we intend to undertake in the future. In this work, we focused on novel Rho variants identified by our suppression screen.
Substitutions in Rho reduce transcription termination and confer growth defects. It is logical to assume that the rho suppressors isolated compensate for the lack of RfaH by reducing Rho-dependent termination. To test this assumption, we compared the effects of rho alleles on reporter expression in vivo. Defects in Rho function lead to the overreplication of most ColE1 plasmids, which replicate via an R-loop intermediate, whereas pSC101 plasmids are stably maintained (24). We constructed a pSC101-based vector in which a lux operon from Photorhabdus luminescens was cloned downstream from the arabinose-inducible P BAD promoter and a synthetic (TC) 15 terminator (Fig. 3A) that functions efficiently in vivo and in vitro (25). Following transformation into selected suppressor strains, we measured the luciferase activity in exponentially growing live cells. We observed that all mutations in rho led to greater lux expression than that in the rho ϩ ΔrfaH strain (Fig. 3B). Since Rho is essential in E. coli, we expected that the termination-deficient rho mutants would exhibit growth defects. Consistently, all rho mutations reduced growth (see Fig. S3). The I382S and L285F alleles were the least and most defective, respectively, an observation consistent with their relative effects on termination; however, there was no correlation between the growth and termination defects of all mutants (Fig. 3B).
Suppressors do not dramatically decrease Rho levels. Substitutions in Rho could affect its function or reduce its levels/stability. An IS2 insertion in the rho leader would be expected to reduce rho expression, whereas substitutions in the conserved regions could interfere with Rho binding to RNA or inhibit ATP hydrolysis/translocation. To determine whether the defects observed are due to reduced Rho levels, we performed Western blotting with anti-Rho polyclonal antibodies. We found that none of the alleles changed Rho levels more than 2.5-fold relative to those of the WT; these small effects are likely explained by autoregulation of Rho expression (26). IS2 insertion and L157 opal substitutions decreased Rho levels to 40%, whereas Δ156-158 and S363F had smaller effects (see Fig. S4). In contrast, other substitutions increased Rho levels. Interestingly, we observed very efficient ribosome readthrough of the stop codon in the L157 opal mutant (see Fig. S4). Our initial attempts to identify the substituted residue were not successful, and since the reduced Rho level would be sufficient to explain the termination-altering phenotype of this mutant, we did not pursue this line of analysis. We conclude that the termination defects of most Rho mutants are conferred by the substituted residues.
Substitutions in the connector lead to defects at a natural terminator. The (TC) 15 element is a near-perfect rut (Rho utilization) element, whereas Rho is able to act on very diverse sequences in vivo, and strong Rho-dependent sites have not been identified in RfaH-controlled operons. To confirm that the isolated suppressors are defective in termination at a natural site, we used a chromosomal ␤-galactosidase reporter previously used to identify defective rho alleles (11). In this strain, a RS45 lysogen carries a T R1 terminator from lambdoid phage H19B between the P lac promoter and the lacZYA operon (Fig. 4A). We moved rho mutations linked to the ilvC::Kan marker into the test strain (rfaH ϩ ). The suppressor mutations in rho formed red colonies on MacConkey lactose agar (Fig. 4B), similarly to nusG G146D, which reduces the efficiency of Rho-dependent termination (24). ␤-Galactosidase activities measured in exponentially growing cells confirmed these observations (Fig. 4C).
These results show that the ␣ 6 and adjacent residues play an important role in the cellular function of Rho, and their substitution confers defects at both synthetic and natural terminators. These defects are comparable to that caused by the substitution of Ile382, a residue that was shown to be important for Rho-dependent termination (11). However, we were unable to transduce L285F, S325P, and S363F alleles into the test strain, suggesting that the original suppressor strains carry second-site mutations that allow the growth of these presumably strongly deleterious rho variants that contain substitutions in the invariant (15) residues of the Q loop, R loop, and Arg finger. Consistent with this interpretation, we could not transduce L285F, S325P, and S363F alleles (linked to ilvC or wzzE) into the original ΔrfaH mutant strain (IA228), whereas other alleles were moved easily, conferring resistance to SDS and exhibiting termination defects (see Fig. S5).
The nusG G146D allele suppresses rfaH. RfaH excludes NusG from RNAP transcribing ops-containing operons (5). In the absence of RfaH, NusG would be expected to take the place of RfaH and increase Rho-dependent termination (8). In this scenario, decreasing Rho-NusG interactions should compensate for the loss of RfaH. To test this prediction, we used a nusG G146D allele that has been shown to reduce Rhodependent termination (24). When transduced into a ΔrfaH mutant strain, the nusG G146D allele restored growth on 0.5% SDS (see Fig. S6). This result confirms that RfaH acts in part by directly competing with NusG. The loss of RfaH would unmask numerous potential Rho loading sites in the waa operon (and others) whose utilization may be dependent on NusG (2).

DISCUSSION
In this work, we screened for suppressors of the E. coli rfaH deletion phenotypes. On the basis of previous results with Salmonella (10) and the primary mode of RfaH action as an antagonist of Rho (9), we expected to identify suppressors in rho, rpoBC, and perhaps nusG that alleviate Rho-mediated polarity in waa, in which a polar mini-Tn5 insertion phenocopies the effects of rfaH (13). Loss-of-function mutations in hns could also compensate for the lack of RfaH because H-NS represses the transcription initiation of (22), and is enriched in (2), the waa operon, where it could act synergistically with Rho. Consistently, our analysis identified substitutions in key Rho and H-NS regions that could lead to a partial or total loss of function. However, we also identified changes in rho, rpo, and hns regions not expected to lead to defects in Rho-dependent termination (see above). In this study, we focused on novel mutations in the connector region of Rho, a nonconserved unstructured linker between the NTD and CTD of Rho. We hypothesize that these substitutions interfere with communication between the Rho domains, underscoring the importance of the allosteric control of Rho function illustrated by several recent reports (27)(28)(29).
Structural rearrangements of the connector. Rho is a hexameric RecA family translocase composed of two domains connected by a flexible~30-residue linker (Fig. 2B). The NTD harbors a primary RNA-binding site, and the CTD contains the secondary RNA-binding site and ATPase and helicase determinants. Rho-dependent termination is a highly tunable event that can be separated into four sequential steps (reviewed in reference 8). First, residues in the primary RNA-binding site interact with pyrimidine dinucleotides in a rut element; the rut composition is a key determinant of Rho termination, with Rho affinity ranging widely, depending on the rut sequence (8). Second, the growing RNA downstream from the bound rut site is threaded into the central channel, where the secondary RNA-binding site residues from the Q and R loops make direct contacts with RNA. Binding of RNA to the central channel triggers, in the presence of ATP, a conformational change from an open to a closed, translocationcompetent state in which the RNA is trapped inside the hexamer (Fig. 2C). Third, the closed hexamer engages in stepwise ATP-powered translocation 5= to 3= along the RNA toward the RNAP while maintaining the contacts with the rut element in a tetheredtracking mode. Finally, upon reaching the paused RNAP, Rho extracts the nascent transcript to trigger the TEC dissociation. Following the RNA release from the RNAP, the Rho ring must open to reset the cycle.
The transition from the open to the closed state is thought to represent the rate-limiting step and is thus expected to be a key target for regulation. Recent studies showed that Rho exists in the open state in solution even in the presence of physiological levels of ATP and that the ring closes only upon the binding of RNA to the secondary sites in the CTD (27). While U 12 RNA bound in the central pore was sufficient to induce the transition, occupancy of the primary sites in the NTD promoted ring closure around a suboptimal A 12 RNA, implying allosteric communication between the two domains. Accordingly, comparison of the open and closed Rho structures reveals that the two domains and the connector may undergo significant rearrangements during ring closure. The N and C termini become more flexible, while the two "handles" of the connector, residues 126 to 129 and 147 to 152, become ordered in the closed state. The ␣ 6 helix (residues 153 to 166) located at the end of the connector is rotated 40°in a closed Rho structure obtained with a long 30-mer RNA (30) but not in those with shorter RNAs (27,31), suggesting that different ligands could promote different structural transitions of the ring. The observed changes in the connector are likely essential for the attainment of the translocation-competent Rho conformation. Substitutions or ligands that restrict these movements would be expected to inhibit Rhodependent termination.
We identified four defective Rho variants with changes in and immediately upstream of ␣ 6 , which is sandwiched between ␣ 7 and ␣ 16 . This region (residues 150 to 158) is poorly conserved among Rho homologs (20) but is enriched (with a loose consensus, GNGSTEDLT) in residues that are favored in natural and engineered protein linkers (32). Glycine is strongly preferred in flexible regions, but charged residues are also tolerated. Gly150 and Gly152 are located in the region disordered in the open hexamer structure (33), and their substitutions for Asp would be expected to rigidify the ␣ 6 junction. Changes in ␣ 6 , such as a Leu157X substitution and the deletion of three residues (156 to 158), could lead to repositioning of the P loop located at the end of ␣ 7 . Mori et al. isolated a defective D156N substitution that did not compromise Rho binding to RNA, suggesting a postbinding defect (40). Substitution of the first ␣ 6 residue, S153Y, which was isolated in combination with a P103L substitution, could not be engineered alone (41). We hypothesize that the S153Y substitution restricts the mobility of the connector-␣ 6 junction, acting similarly to G150D and G152D.
Tuning Rho-dependent termination. In contrast to intrinsic termination, which depends primarily on a signal in the nascent RNA, Rho-dependent termination is only loosely dependent on the RNA sequence. Rho preferentially interacts with short, C-rich sequences in both the primary and secondary sites (28), but Rho affinities for C-rich sequences vary greatly and Rho can terminate transcription at sites with lower C content (2). Broad sequence specificity is likely a prerequisite for many diverse roles that Rho has been shown to play. In addition to terminating the transcription of some structural genes, Rho terminates the transcription of poorly translated RNAs, such as horizontally transferred foreign genes, antisense RNAs, or mRNAs bearing early stop codons, resolves R loops, and ensures genome stability by reducing replicationtranscription collisions (see reference 8 and references therein). Thus, Rho has to act at emerging problem sites in addition to genetically programed sites, and its recruitment to the nascent RNA is strongly context dependent.
Productive recruitment of Rho requires an extended Ͼ70-nucleotide-long segment of RNA that is devoid of strong secondary structures and RNA-bound proteins, followed by ring closure. Termination by Rho can be tuned by nascent RNA sequences and trans-acting RNA and protein cofactors that can act directly or indirectly (8,34,35). E. coli NusG is the best-characterized activator of Rho. Among more than 1,000 Rho-dependent sites in MG1655, those with poor rut elements (that possess low CϾG ratios) require NusG for efficient termination (2), implying that NusG binding to the Rho CTD may mimic an effect of an optimal C-rich signal or ligand binding to the primary sites (27). H-NS, which colocalizes to many Rho release sites, also potentiates Rhomediated RNA release (2). Diverse regulators that inhibit Rho have been characterized (8,34,35). Some antiterminators are recruited only to their target operons to modify the RNAP into a termination-resistant state (36). Others, such as Psu and Hfq (37,38), appear to directly inhibit Rho via protein-protein interactions. While their detailed mechanisms remain to be determined, some of these regulators may utilize several independent modes of inhibition to ensure potent activation of their target genes.
Rho modulators may act allosterically. Recent structural and biochemical studies suggest that many, if not most, Rho ligands act allosterically. BCM, which has been argued to act as a noncompetitive ATP-binding inhibitor of Rho, has recently been shown to block ring closure (27). Conversely, NusG (29) and RNA bound to the primary sites (27) appear to promote ring closure. Whereas the sensitivity to substitutions signifies the mechanistic role of the connector, mounting evidence also points to the key role of this region in Rho regulation. Psu, a phage-encoded Rho antagonist, binds to the region encompassing residues 139 to 153 (38). The Psu dimer has been proposed to bridge two Rho protomers to sterically block the central RNA-binding channel and thereby translocation (38) but may also restrict connector movement and thus inhibit the closed-to-open transition. Interestingly, a P167L substitution at the C terminus of ␣ 6 restored the function of defective Psu (38). NusG potentiates Rho-mediated termination at suboptimal rut sites (2). Several underlying mechanisms have been proposed, but the existence of substitutions at the opposite ends of the connector that mimic (29,39) or abolish (33) the effect of NusG suggests that NusG may stimulate Rho activity by realigning the domains to facilitate ring closure. We speculate that a recently discovered RARE element that likely binds to the NTD (34) and Hfq (37) may antagonize Rho activity by inhibiting ring closure.
These data support a model in which the connector segment plays a hitherto unknown regulatory role. Structural evidence shows that the connector is a mobile tether, and our present findings suggest that it may be what enables facile transitions between the open and closed states. The connector is present in diverse Rho homologs, across which the majority of its sequence is not conserved (20). If our conjecture regarding the functional importance of the connector is correct, the variability of its sequence would further suggest that it may be a target for species-specific regulators that modulate Rho activity. Such modulation could be used to respond to unique cellular contexts that may differ among different bacteria, particularly those in which Rho is dispensable. Our isolation of mutations in the variable region of the connector might justify extending this study to other bacterial species to explore the possibility of such specificity.
Changes in the connector act synergistically with BCM. We propose that connector flexibility is essential for Rho ring closure; we hypothesize that Gly-to-Asp substitutions at positions 150 and 152 reduce this flexibility but still permit the domain rearrangements, whereas more "drastic" changes will likely be lethal, e.g., S153Y in the absence of compensatory changes in the primary RNA-binding region (41).
If this were true, the G150D and G152D substitutions would be expected to act similarly to BCM, inhibiting the transformation into the active, translocation-competent state and thereby conferring increased sensitivity to BCM in vivo. Indeed, we found that the G150D and G152D alleles and, to a lesser extent, the Δ156-158 rho allele caused severely impaired growth at low (10 g/ml) BCM concentrations that fully supported the growth of the WT strain (Fig. 5). In contrast, reducing the level of WT Rho (⍀IS2), inhibiting Rho interactions with NusG (G146D substitution), or inhibiting Rho function by substitution of Ile382 did not lead to BCM hypersensitivity (Fig. 5). These observations are consistent with our hypothesis that BCM and the connector affect the same step in the Rho mechanism.
Why did we isolate connector mutants? Most screens for defective rho mutants have been carried out with strong terminators placed in front of a reporter gene. These screens identified mutations in several key regions of Rho ( Fig. 2A). In our selection, mutations in these regions were also recovered, along with a new class of mutations in the connector (4 out of 10). We note that a screen for suppressors of polarity in Salmonella also identified a connector mutation (39). Although our selection was by no means saturating, the repeated recovery of two of these mutants argues against a serendipitous explanation and implies that a different basis for our selection could explain this bias. We assume that compromised waa operon expression is the underlying reason for the SDS sensitivity of the rfaH mutant strain (see above). In WT MG1655, waa is a poor target for Rho; it has a low frequency of C residues (17.4%) and is devoid of NusG that is excluded by RfaH. The loss of RfaH would be expected to permit NusG binding to RNAP and abolish ribosome recruitment, potentiating Rho-depending termination. Because of the absence of demonstrably strong Rho termination sites (2), we propose that, rather than using its most potent mode of gene silencing (at a single, dominant early site), Rho silences waa expression by inducing termination at many weak sites along the operon. The absence of a strong RNA-binding site should increase the dependence of regulation on the allosteric communication between the two domains and therefore the functionality of the linker.
Mapping of mutant alleles. To map the rfaH suppressors, linkage to selected markers (linked to the Kan resistance gene in the Keio collection [18]) was tested by P1vir transduction. Suppressors in rho (linked to ilvC::Kan and wzzE::Kan) and rpoC (thiH::Kan) were identified by this approach. To map the remaining suppressors, we used random Tn5::Tet libraries (a gift from Natacha Ruiz). A P1 lysate from a pool of mutants carrying randomly inserted mini-Tntet cassettes in the chromosome of WT strain MC4100 was used as a donor of WT alleles in P1 transductions where the recipients were ΔrfaH mutant strains carrying the SDS-resistant suppressors. We then screened for transductants that lost resistance to SDS. PCR of the chromosomal DNA was performed with an arbitrary primer (5=-GGCCACGCGTCGACTAGTAC NNNNNNNNNNACGGC) and a transposon-specific primer (5=-CCTTCATGTTAACCCCTCAAGCTCAGGGG), and the resulting PCR products were sequenced to identify the mutated region. P1vir transduction was then used to finely map the suppressors. To identify the mutations, PCR amplification of the affected region with locus-specific primers and sequencing were used. To confirm that these rfaH sup mutations conferred the suppressor phenotype, cotransduction was used to move the nearby marker and the mutant alleles into IA228.
Disc diffusion assays. Disc diffusion assays were performed by pouring a mixture of 100 l of an overnight culture and 4 ml of LB top agar over LB agar plates. After the top agar solidified, 6.5-mm antibiotic disks (BD BBL Sensi-Disk Susceptibility Test Disks) containing novobiocin (30 g) were placed on top. After overnight incubation at 37°C, zones of clearance around the disks were measured. The data shown are representative of at least three independent experiments.
Luciferase reporter assays. Strains defective in Rho-dependent termination are susceptible to killing because of the overreplication of many common plasmids; in contrast, pSC101-based vectors can be stably maintained in these strains (24). We therefore constructed reporter plasmid pHK2, in which the pSC101 origin of replication and spectinomycin resistance gene were combined with the araC-P BAD TC 15 luxCDABE region of pIA1250, a derivative of pIA955 with a synthetic TC 15 cassette cloned into the leader region (9). pHK2 was transformed into selected strains with TSS (Epicentre) and plated on 20 g/ml spectinomycin. The single colonies were inoculated into 3 ml of LB supplemented with spectinomycin and incubated at 37°C with aeration. After 8 h of growth, cultures were diluted 1:50 into EZRDM-G supplemented with 20 g/ml spectinomycin and 0.1% arabinose and allowed to grow for 2 h. Luminescence was measured in 200-l aliquots in triplicate on a FLUOstar Optima plate reader (BMG Labtech GmbH) and normalized by cell density. Results were analyzed with Microsoft Excel.
␤-Galactosidase assays. Single colonies were inoculated into 3 ml of LB and incubated at 37°C with aeration overnight. Cultures were diluted 1:50 in 2 ml of EZRDM supplemented with 0.2% galactose and 1 mM isopropyl-␤-D-thiogalactopyranoside (IPTG) and grown at 37°C to the mid-log phase (optical density at 600 nm [OD 600 ] of~0.5 to 0.6). The cells were pelleted and resuspended in 1 ml of Z-Buffer, 2 drops of chloroform and 1 drop of 0.1% SDS were added, and the mixture was vortexed for 10 s. Equal amounts of permeabilized cells (adjusted on the basis of cell culture density) were mixed with Z-Buffer in a 150-l final volume in 96-well microplates (Costar). One hundred microliters of 5 mg/ml o-nitrophenyl-␤-D-galactopyranoside was added, and ␤-galactosidase activity was determined from the rate of increase of the o-nitrophenol concentration measured every 10 s for 10 min at 420 nm in an xMark spectrophotometer (Bio-Rad). Individual cultures were assayed in triplicate, and average values of three independent cultures determined with the Microplate Manager software (Bio-Rad) are reported.