Hfq binds directly to the ribosome‐binding site of IS 10 transposase mRNA to inhibit translation

Summary Hfq is a critical component of post‐transcriptional regulatory networks in most bacteria. It usually functions as a chaperone for base‐pairing small RNAs, although non‐canonical regulatory roles are continually emerging. We have previously shown that Hfq represses IS 10/Tn10 transposase expression through both antisense RNA‐dependent and independent mechanisms. In the current work, we set out to define the regulatory role of Hfq in the absence of the IS 10 antisense RNA. We show here that an interaction between the distal surface of Hfq and the ribosome‐binding site of transposase mRNA (RNA‐IN) is required for repressing translation initiation. Additionally, this interaction was critical for the in vivo association of Hfq and RNA‐IN. Finally, we present evidence that the small RNA ChiX activates transposase expression by titrating Hfq away from RNA‐IN. The current results are considered in the broader context of Hfq biology and implications for Hfq titration by ChiX are discussed.


Introduction
Hfq is an abundant RNA-binding protein that acts at the core of complex post-transcriptional regulatory networks in many bacteria and is critical for stress and virulence responses (Storz et al., 2011;Vogel and Luisi, 2011;Sobrero and Valverde, 2012). It is found in at least 50% of sequenced bacteria (Sun et al., 2002) and has been predicted to be involved in the regulation of 269 mRNAs in Escherichia coli and at least 20% of all genes in Salmonella Typhimurium (Guisbert et al., 2007;Sittka et al., 2008;Ansong et al., 2009). Hfq is important for the function of trans-encoded small regulatory RNAs (sRNAs) that basepair with partially complementary mRNAs. Hfq binds sRNAs and their partner mRNAs and facilitates intermolecular base-pairing. This typically affects translation and/or transcript stability. Hfq contains three RNA-binding surfaces, all of which play a role in promoting base-pairing between RNAs. The 'top' and 'bottom' of the toroidalshaped Hfq homohexamer are termed the proximal and distal RNA-binding surfaces respectively. The proximal surface binds short U-rich sequences typically found in sRNAs while the distal surface binds longer ARN repeats (where A is an adenine, R is a purine and N can be any nucleotide) typically found in mRNAs (Mikulecky et al., 2004;Link et al., 2009). The proximal surface is proposed to be critical for sRNA stability through interactions with the 3′poly(U) tract following a Rho-independent terminator (Sauer and Weichenrieder, 2011;Ishikawa et al., 2012). The third, less defined surface consists of the outer rim or lateral RNA-binding surface. This surface connects the proximal and distal RNA-binding sites. The lateral surface is extremely basic in E. coli and may be important for binding internal U-rich sequences of sRNAs (Sauer et al., 2012). One model for Hfq-catalyzed pairing predicts simultaneous binding of a cognate sRNA and mRNA pair via the proximal and distal surfaces respectively. Hfq-binding sites are often just outside of RNA pairing sequences so simultaneous binding would tether the RNAs to Hfq while keeping seed regions available for pairing (Panja and Woodson, 2012). The RNAs can then initiate pairing by interacting in either the lateral or proximal surfaces and the RNAs are released as pairing proceeds (Hopkins et al., 2011;Hwang et al., 2011;Panja et al., 2013).
In addition to a role in sRNA-based regulation, Hfq has been shown to directly affect translation. In the case of sdhC mRNA, the sRNA Spot42 recruits Hfq to an AU-rich region in the translation initiation region (TIR) to inhibit translation. As the Spot42 pairing region in sdhC is too far upstream of the TIR to influence translation, it was inferred that stable association of Hfq with sdhC was sufficient to compete with 30S ribosomal subunit binding (Desnoyers and Massé, 2012). In another example, Hfq was shown to bind to a translational enhancer in cirA mRNA and block translation. Interestingly, in this case, translation repression was relieved by the upstream binding of an sRNA (RyhB) that caused restructuring of the mRNA within the 5′ untranslated region (5′UTR), which ultimately prevented Hfq binding (Salvail et al., 2013). Finally, evidence has been presented in two different organisms that Hfq autoregulates expression by binding its own TIR (Vecerek et al., 2005;Sobrero and Valverde, 2011). No sRNAs have been implicated in this autoregulatory loop, supporting the contention that Hfq can act directly to inhibit translation. In the above examples, Hfq binding to the TIR of an mRNA is the effector of translational control, in contrast to sRNAdependent regulation where the stable sRNA-mRNA duplex is responsible for blocking ribosome binding. Unlike sRNA-dependent regulation, the role of each RNA-binding surface of Hfq in direct translational repression is largely unknown. However, Hfq binding to the TIR of target mRNAs would presumably require the distal surface, which preferentially binds purine-rich sequences such as the Shine-Dalgarno sequence.
In addition to its role as an important regulator of endogenous gene expression, Hfq was recently found to suppress Tn10/IS10 transposition in E. coli (Ross et al., 2010). Tn10 is a composite transposon containing genes encoding for tetracycline resistance (Fig. 1A). Its component insertion sequence IS10-Right encodes a functional transposase that catalyzes the chemical steps in Tn10/IS10 transposition (Foster et al., 1981;Halling et al., 1982;Chalmers et al., 2000). Expression of IS10 transposase is regulated by Dam methylation as well as a 69-nt antisense RNA (asRNA) that is transcribed from the opposite strand of DNA relative to the transposase (Simons and Kleckner, 1983;Roberts et al., 1985). The first 35-nt of this asRNA (RNA-OUT) is perfectly complementary to the TIR of the transposase mRNA (RNA-IN), and pairing of these two RNAs inhibits translation by preventing ribosome binding ( Fig. 1A) (Ma and Simons, 1990). Antisense control of transposase expression increases with IS10 copy number, a phenomenon termed 'multi-copy inhibition' (MCI). MCI can be explained by the fact that transposase is a cisacting protein whereas the asRNA is trans acting (Jain and Kleckner, 1993). Accordingly, increasing transposon copy number essentially serves to increase the amount of transacting inhibitor while the effective concentration of transposase per element remains constant. Importantly, a single-copy IS10 element is not subject to antisense control of transposase expression (Kleckner, 1990).
Hfq was initially linked to Tn10/IS10 transposition when it was found that IS10 transposition increased in the order of 80-fold in an hfq − strain of E. coli harboring IS10 on a multi-copy plasmid. In contrast, the impact of Hfq deficiency on transposition was greatly reduced (sevenfold increase), but not completely abrogated, when transposition was measured for IS10 in single copy. These obser- vations were consistent with Hfq contributing to MCI, but also playing a role in down-regulating IS10 transposition independent of the MCI pathway (Ross et al., 2010). Subsequent work demonstrated that Hfq bound both RNA-IN and RNA-OUT in vitro, and accelerated the rate of IN-OUT pairing almost 20-fold, observations that are consistent with Hfq working through its prototypical RNA pairing pathway to promote the MCI response (Ross et al., 2013). It remains to be determined how Hfq regulates Tn10/IS10 transposition when MCI is not in play. However, it has been shown that: (i) in the absence of RNA-OUT transposase, expression increased sixfold in hfq − ; (ii) there is an Hfq-binding site in RNA-IN that overlaps the TIR; and (iii) Hfq status has only a subtle effect on steady-state RNA-IN transcript level (Ross et al., 2010;2013). Taken together, it seems likely that Hfq functions in the MCI-independent pathway by inhibiting translation of RNA-IN. The goal of the current work was to test this hypothesis and to further characterize how Hfq interacts with RNA-IN.

Antisense-independent regulation of transposase expression requires the distal surface of Hfq
To gain further insight into how Hfq regulates IS10 transposase expression (independent of its cis-encoded sRNA), we asked if regulation was maintained when RNAbinding surfaces of Hfq were mutated. Our expectation was that if a trans-encoded sRNA is involved in this pathway, all three surfaces would be critical for regulation. In particular, we expected the proximal surface to be important for stabilizing any involved sRNAs and the lateral surface for catalyzing pairing with RNA-IN.
To assess the function of each Hfq-binding surface in antisense-independent regulation of transposase expression, we assayed the ability of wild-type (WT) and mutant forms of Hfq to complement an hfq − phenotype. We constructed a chromosomal IS101-339-lacZ translational fusion with a single bp change (HH104) in the promoter for RNA-IN, which increases transcription ∼ 100-fold ( Fig. 1B) (Case et al., 1988). The HH104 mutation also destabilizes RNA-OUT; however, a singlecopy transposase-lacZ translational fusion would not normally be regulated by the cis-encoded RNA-OUT so the net effect of this mutation is to simply increase RNA-IN expression to detectable levels (Case et al., 1989). Expression of transposase-lacZ was measured in an hfq − strain of E. coli harboring plasmids expressing WT Hfq or Hfq deficient in RNA binding at the distal (Y25A), proximal (K56A) or lateral (R17A) surface (Mikulecky et al., 2004;Panja et al., 2013).
We show in Fig. 2A  A. Transposase expression was measured in the context of a chromosomal transposase-lacZ translational fusion (parent strain DBH298) with the indicated forms of Hfq expressed or in the absence of Hfq expression. The bars show β-galactosidase activity (Miller units) with standard error of the mean, measured in mid-exponential phase in LB (n = 8). Where indicated, the hfq − strain (DBH299) was transformed with a low-copy plasmid encoding Hfq from its native promoter (P3). The mean relative expression observed for each strain is indicated at the top of the graph, where transposase-lacZ expression in hfq + was set at 1. B. Transposition of chromosomal IS10 HH104 -kan was measured by the conjugal mating out assay (see Experimental procedures) in an hfq − strain (DBH337) transformed with one of the indicated Hfq-encoding plasmids. The mean relative transposition frequency for each strain is indicated at the top of the graph, where transposition in the presence of Hfq WT was set at 1. Error bars indicate the standard error of the mean for two independent experiments (n = 8).
regulation was fully restored when Hfq WT was expressed from a plasmid. In contrast, none of the Hfq variants were able to fully complement hfq − , with Hfq Y25A being the most impaired and Hfq R17A functioning the most like WT. Importantly, the reduced function of the mutant proteins cannot be attributed to protein expression as the Hfq levels were not significantly different for WT versus the mutant forms of Hfq (Supporting Information Fig. S1). As the results showed that the integrity of the lateral site is not important for regulation and the integrity of the proximal site is less important than that of the distal site, our data suggested that Hfq repression was sRNA independent. We also performed the Hfq complementation experiment in the context of an IS10 transposition assay. As IS10 transposition frequency is directly proportional to transposase expression (Morisato et al., 1983), this experiment allowed us to indirectly measure the effect of Hfq mutations on native transposase expression. Transposition of a single-copy IS10 HH104 -kan element ( Fig. 1B) in an hfq − strain of E. coli was repressed 13-fold in the presence of Hfq WT and close to full repression was achieved in the presence of Hfq K56A and Hfq R17A (Fig. 2B). However, in accordance with the expression data ( Fig. 2A), Hfq Y25A was the least effective of the mutant Hfq forms in repressing transposition.
It is not clear why the Hfq K56A functioned essentially as WT in the transposition assay ( Fig. 2B) while exhibiting a moderate defect in repressing transposase-lacZ expression ( Fig. 2A). However, the concordance of the other Hfq variants between the two experiments leads us to conclude that the distal surface is critical for repressing transposase expression in the absence of RNA-OUT, while the proximal and lateral surfaces are dispensable for regulation.

The distal surface of Hfq binds the RBS of RNA-IN in vitro
We have previously shown that Hfq binds a 14-nt A-rich region of RNA-IN, which overlaps with the ribosomebinding site (RBS), as well as an 8-nt U-rich region overlapping with codons 5-8 of the transposase coding region (Ross et al., 2013). Based on the results from the previous section, we anticipated that the A-rich-binding site within the TIR would be the critical site for Hfq regulation as it has the signature of a distal binding site. To further characterize the two Hfq-binding sites in the 5′ segment of RNA-IN, we carried out chemical and enzymatic RNA footprinting. Purified WT, Y25A and K56A Hfq were incubated with an RNA corresponding to the first 160-nt of  followed by partial digest with lead acetate (Pb 2+ ), RNase T1 or RNase V1.
Within the first 70-nt of RNA-IN, three regions of cleavage reagent protection were detected in the presence of Hfq WT (Fig. 3). The most upstream region (relative to the start codon) spans the RBS, extending roughly from nt −20 to −4; hereafter, this site will be referred to as site 1. Additional regions of protection downstream of this include residues 13-20 and 32-34 (hereafter referred to as sites 2 and 3 respectively). Another region spanning nt 3-7 became hypersensitive to RNase T1 and showed reduced RNase V1 cleavage, consistent with Hfq binding inducing a structural transition in this region from dsRNA to ssRNA. In contrast, in reactions with Hfq Y25A , site 1 showed greatly reduced protection from both V1 and lead cleavage, consistent with the distal surface of Hfq making contacts with this site. The V1 and lead cleavage pattern in site 1 for Hfq K56A was very similar to that of Hfq WT . As Hfq K56A retains a fully functional distal binding surface, this result further supports our conclusion that the distal surface of Hfq contacts site 1.
At sites 2 and 3, the K56A mutation greatly reduced protection from both V1 and lead cleavage whereas the Y25A mutation did not. This is consistent with the proximal surface of Hfq binding both sites 2 and 3.
Finally, the T1 hypersensitivity at positions 3-7 was lost only in the reaction with Hfq Y25A . Accordingly, we infer that the structural transition in this region is dependent on the distal face of Hfq binding to site 1.

Hfq binding to site 1 is critical for regulation of transposase expression in vivo
As the distal surface of Hfq is most critical for repression of transposase expression and transposition, our footprinting data suggest that Hfq primarily exerts its regulatory role by binding site 1, which overlaps the RBS of RNA-IN. We set out to further test this hypothesis by analyzing the impact of nucleotide changes in site 1 and site 2 on Hfq interactions and ultimately transposase expression.
Our first objective was to define the relative binding affinity of sites 1 and 2 for Hfq. We show by electrophoretic mobility shift assay (EMSA) in Fig Based on these results, as well as footprinting experiments (Supporting Information Fig. S2 K56A Hfq (750 nM Hfq hexamer) before treatment with Pb 2+ , RNase T1 or RNase V1. Untreated RNA was also incubated with or without WT Hfq (lanes 1-2). Nucleotide numbering is relative to the translational start codon (AUG), where the A is position 1. Positions that were protected from cleavage by WT and K56A but not Y25A are indicated with red arrows, while positions protected by WT and Y25A but not K56A are indicated with blue arrows. RNase T1 hypersensitivities are highlighted with black arrows. The nucleotide sequence of the first 160-nt of RNA-IN is shown below the gel image with distal and proximal specific-binding sites indicated in red and blue respectively. The RBS and start codon are underlined. This figure is available in colour online at wileyonlinelibrary.com.
with a much higher affinity than it binds site 2, and Hfq binding to site 2 appears to be dependent on Hfq first binding site 1. As formation of IN:Hfq-2 increased in a concentration-dependent manner, it is likely that occupancy of site 2 depends on recruitment of a second Hfq hexamer to IN:Hfq-1, as opposed to the unoccupied proximal surface of an Hfq hexamer bound at site 1 and engaging site 2. We next looked in vivo at the impact of disrupting Hfq interactions with sites 1 and 2. To do this, we introduced the M5 and M2 mutations into an IS10 1-242-lacZ translational fusion on a multi-copy plasmid. A multi-copy transposase-lacZ translational fusion would normally be highly repressed by the cis-encoded RNA-OUT. To separate the role of Hfq in direct repression of transposase expression from its role in facilitating antisense pairing, we introduced a single nucleotide mutation (R5) into the promoter region of RNA-IN, which destabilizes RNA-OUT while having only a subtle effect on RNA-IN transcription ( Fig. 1B) (Case et al., 1988). Based on our earlier results ( Fig. 2A and B), our expectation was that the M5 mutation would make IS10-lacZ expression insensitive to Hfq status. As the two mutants have multiple nucleotide substitutions, we were concerned that these changes could have indirect effects on transposase expression. Accord-ingly, we also isolated RNA from cells used in the reporter assays and performed primer extension analysis to monitor steady-state transcript levels (Fig. 5).
We show in Fig. 5 that both M5 and M2 mutants exhibited reduced Hfq regulation, with the degree of dysregulation being stronger for M5 versus M2. In this reporter setup, there was a 3.2-fold decrease in transposase expression in the presence versus the absence of Hfq, consistent with Hfq having a negative regulatory role. In contrast, for the M5 reporter, expression levels were essentially the same in hfq + and hfq − and for the M2 reporter, expression decreased about 1.5-fold in the presence of Hfq.
In the above experiment, the ratio of transposase-lacZ expression (β-galactosidase assay) to the steady-state level of fusion transcript provides a measure of the translation efficiency. For example, if expression was low and transcript levels were high, this would be indicative of low translation efficiency. The presence of Hfq in cells expressing the WT reporter decreased translation efficiency approximately 2.5-fold, consistent with Hfq interfering with transposase translation. For the M5 reporter, translation efficiency was greatly reduced compared with the WT reporter, which is not unexpected given that 5 of the 8 nt changes in this construct are in the RBS. Importantly, the translation efficiency did not further decrease in  pauses at nts +17/+18 as has previously been reported (Ma and Simons, 1990). These observations are consistent with a stable translation initiation complex forming on RNA-IN. When Hfq was added prior to addition of the 30S ribosome and initiator tRNA (lanes 7-12), there was a decrease in the toeprint signal, the magnitude of which was dependent on the Hfq concentration. For example, at an Hfq concentration of 200 nM, where Hfq and RNA-IN are present at a 1:1 molar ratio, the toeprint signal decreased greater than 90% relative to the signal observed in the absence of Hfq ( Fig. 6A and B). In contrast, when the same experiment was performed with lpp or usg mRNA, inhibition of the toeprint signal was significantly weaker (Supporting Information Fig. S3A). For example, at the same ratio of Hfq:mRNA that gave greater than 90% inhibition for IS10, only 50% inhibition was observed for lpp and usg (Fig. 6B). Although Hfq plays a role in repressing lpp expression by stabilizing the sRNA micL (Guo et al., 2014), we did not detect lpp as an Hfq-binding mRNA ( Fig. 9) and there is conflicting evidence in the literature regarding an interaction between Hfq and lpp in vivo (Chao et al., 2012;Bilusic et al., 2014;Tree et al., 2014). However, Hfq does not interact with usg mRNA in vivo (Beisel et al., 2012). Accordingly, the toeprinting results in Fig. 6A are consistent with Hfq acting specifically to block translation initiation in the IS10 system. We presume that the relatively low level of toeprint inhibition observed in the lpp and usg experiments is the result of non-specific interactions between Hfq and components of the 30S ribosome and thus represents 'background noise' in the assay. We reasoned that because the distal surface of Hfq interacts with site 1 of RNA-IN, then the distal surface of Hfq would be critical for blocking 30S ribosome binding to RNA-IN. Accordingly, a competitor RNA that is specific for the distal RNA-binding face of Hfq should relieve the 'toeprint repression' afforded by Hfq. We show in Fig. 6C and D that when Hfq was pre-incubated with A18, a distal site-binding RNA, prior to its addition to RNA-IN, the toeprint signal increased approximately 2.5-fold relative to a control reaction where no competitor was added (compare lanes 6 and 7). Also, if a proximal face-binding RNA (U7) was used instead of A18, there was no increase in the toeprint signal (compare lanes 6 and 8). We also show in Fig. 6C and D that Hfq Y25A failed to reduce the toeprint signal (compare lanes 9 and 10 with lanes 11 and 12). Taken together, the results in this section show that Hfq can inhibit 30S ribosome binding to RNA-IN in vitro and that this inhibition requires an available distal surface on Hfq. Furthermore, as we have shown that there is a high-affinity Hfq-binding site (site 1) that spans the RBS of RNA-IN and engages the distal binding surface of Hfq, we conclude that Hfq binding to this site is responsible for blocking translation initiation.

Hfq binds native RNA-IN in vivo and this binding is inhibited by site 1 mutations
As Hfq is a pleiotropic regulator of gene expression in E. coli, there is a concern that any phenotype observed in an hfq − strain might be the result of dysregulation of a factor under Hfq control. As we are proposing that Hfq binds directly to the 5′UTR of RNA-IN to block translation, we thought it important to look for Hfq-RNA-IN binding in vivo where RNA-IN would have to compete with other cellular RNAs for Hfq binding. We performed an Hfq-RNA immu-noprecipitation experiment (RIP) with hfq − cells containing a chromosomal copy of IS10 HH104 -kan and a plasmid encoding FLAG-tagged Hfq (Fig. 7A). Importantly, this experiment used the same full-length RNA-IN as that used in our transposition experiments. RNA recovered from an Hfq IP was subject to reverse transcription polymerase chain reaction (RT-PCR) where RNA-IN was amplified from total cDNA with gene-specific primers (see Experimental procedures and Supporting Information Fig. S4). PCR reactions were then analyzed on an agarose gel. We show that RNA-IN was strongly A. 30S ribosome binding to RNA-IN +/− Hfq is shown in a 'toeprint' assay (for details, see Experimental procedures). Addition of 30S ribosomal subunits and initiator tRNA is indicated by + . The toeprint signal is indicated (+16/+17/+18) with numbering relative to the translational start codon. CUAG refers to sequencing reactions generated from the same RNA used for toeprinting. B. Quantitation of RNA-IN (A), lpp and usg (Supporting Information Fig. S3) toeprints. Toeprint signal was normalized to the combined band intensity (+16/+17/+18 for RNA-IN; +15/+16 for lpp and usg) in the absence of Hfq, which was set at 100. The dashed line highlights 50% inhibition of the toeprint signal. C. Toeprint analysis of RNA-IN with distal (A18) and proximal (U7) site-specific competitor RNAs. In addition, RNA-IN toeprint signal was also analyzed for reactions containing Y25A versus WT Hfq. Note that for this latter comparison, Hfq contained a C terminal 6x-His epitope tag. The competitor RNAs alone had no effect on ribosome binding (not shown). D. Band intensities in C were quantified and normalized to the toeprint signal in lane 5 where no competitor or Hfq was added.
enriched in the Hfq IP compared with control reactions (Fig. 7A). Specifically, when cells expressing an untagged Hfq (Hfq WT ) were subject to RIP, no Hfq was detected in the IP fraction (Supporting Information Fig. S4A) and accordingly no RNA-IN was detected by RT-PCR (Fig. 7A). We therefore conclude that RNA-IN is a bona fide Hfq-binding partner in vivo.
We also performed an Hfq IP with cells containing either the WT or M5 IS10-lacZ translational fusion on a multi-copy plasmid. RT-PCR analysis of these samples revealed a 2.3-fold reduction in the recovery of IN M5 -lacZ compared with IN WT -lacZ in the IP. This result is consistent with the Hfq-binding site within the 5′UTR of RNA-IN (site 1), providing important determinants for Hfq binding in vivo (Fig. 7B).

ChiX overexpression titrates Hfq away from RNA-IN
Overexpression of Hfq-binding RNAs can impinge on other regulatory networks by titrating available Hfq away from other mRNAs and sRNAs (Papenfort et al., 2009;Hussein and Lim, 2011;Moon and Gottesman, 2011). Given our finding that RNA-IN can compete with other cellular RNAs for Hfq binding (Fig. 7), we wondered if induction of an sRNA might increase transposase expression by sequestering Hfq away from RNA-IN. Although most sRNAs interact with the proximal surface of Hfq, an overexpressed sRNA that can bind the distal surface of Hfq might block Hfq's association with the RBS of RNA-IN and therefore increase transposase translation by allowing the ribosome to bind. To test this prediction, we screened a library of Hfq-binding sRNAs to see if any increased transposase-lacZ expression.
We next quantified the level of IS10 transposase up-regulation by measuring β-galactosidase activity with overexpression of ChiX. As a control, we included SgrS, as this sRNA did not give a blue colony color in our screen. As shown in Fig. 8A, overexpression of ChiX increased transposase-lacZ expression almost 12-fold compared with a vector control, while SgrS overexpression had no effect on transposase expression. We also analyzed RNA extracted immediately before the Miller assay by Northern blot and primer extension to measure sRNA induction and transposase-lacZ transcript levels respectively ( Fig. 8B and C). Consistent with previous results, a low amount of endogenous ChiX was detected in all samples (Vogel et al., 2003;Figueroa-Bossi et al., 2009), but there was a large increase in the presence of the ChiX plasmid (52-fold). This induction is comparable with that seen as cells transition to stationary phase (Vogel et al., 2003). Importantly, ChiX induction resulted in only a twofold increase in transposase-lacZ transcript levels relative to the vector control (Fig. 8C). As ChiX overexpression increased transposase expression 12fold while having only a subtle effect on steady-state L -+ -+ -+ -+ RT  A. hfq + cells containing a chromosomal IS101-339-lacZ translational fusion were transformed with a plasmid expressing ChiX (pDH765), SgrS (pDH764) or a vector control (pDH763). Transformants were grown to mid-exponential phase in LB media and β-galactosidase activity was measured. Error bars show the standard error of the mean for two independent experiments (n = 7) and the relative expression is shown above the graph, where transposase-lacZ expression in the presence of vector was set to 1. RNA was extracted immediately before the Miller assay. B. A 3.5 μg of total RNA (three biological isolates) was used for a Northern blot using a 5′ 32 P-labeled oligonucleotide (SgrS) or internally labeled antisense RNA probe (ChiX, 5S rRNA). C. Primer extension analysis of 10 μg of total RNA (four biological isolates) was used to detect RNA-IN-lacZ transcript and lpp was analyzed as an internal control. The ratio of IN-lacZ:lpp was normalized to the vector control and is shown with standard error of the mean as a graph above the gel images. transcript levels, we conclude that ChiX increases transposase translation and not transcription or mRNA stability. ChiX may increase transposase translation by one of two mechanisms: (i) ChiX may base-pair with RNA-IN to increase ribosome accessibility, as seen in the rpoS system (Brown and Elliott, 1997;Soper et al., 2010), or (ii) ChiX may bind Hfq with high affinity and block Hfqbinding to RNA-IN. We used in vitro lead footprinting with 5′labeled ChiX and purified Hfq to define Hfq-binding sites on ChiX. We also included RNA-IN in the footprinting reactions to determine if ChiX base-pairs with RNA-IN.
In the absence of Hfq, ChiX exhibited high reactivity to lead with the exception of nucleotides 17-22, 56-65 and 70-78, which is consistent with a mostly unstructured RNA containing a 5′stem-loop and a Rhoindependent terminator (Fig. 8D, lane 5 and right panel).
In the presence of Hfq, two regions of reduced lead cleavage consisting of nucleotides 24-33 and 48-52 were observed (Hfq-I and Hfq-II, compare lanes 5 and 6). Binding reactions were also analyzed by EMSA (shown beneath footprinting gel image) and showed that Hfq forms a single complex with ChiX sRNA (ChiX:Hfq, lanes 2 and 3).
We also performed footprinting experiments with Hfqbinding face mutants and ChiX RNA (Supporting Information Fig. S5). We show that the distal surface of Hfq binds Hfq-I, while binding of Hfq-II requires an intact proximal surface. In addition, a third Hfq binding site (Hfq-III) was identified that includes the poly(U) tract following the Rhoindependent terminator, which accordingly interacts with the proximal surface of Hfq.
In Based on our in vitro data that RNA-IN can compete with ChiX for Hfq binding, we performed an RIP in hfq − cells containing the IS101-339-lacZ translational fusion, FLAGtagged Hfq and the ChiX overexpression plasmid or a vector control. RNA recovered from the Hfq IP was analyzed directly by Northern blot or primer extension to detect ChiX and IN-lacZ respectively; the 5S rRNA and lpp were also analyzed as negative controls (Fig. 9). Overexpression of ChiX resulted in a 4.5-fold reduction in the amount of RNA-IN associated with Hfq. This experiment also allowed us to compare the relative binding affinities of ChiX and RNA-IN for Hfq. In the presence of the vector control, ChiX binds Hfq about 30-fold better than RNA-IN in vivo. Based on the differences in the amount of RNA analyzed (10 μg input RNA, 0.3 μg IP RNA), we calculated that ChiX was enriched 365-fold in the Hfq IP, while RNA-IN was enriched 12-fold. We presume that the relatively low amount of 5S rRNA and lpp mRNA that were detected in the IP represents non-specific interactions with Hfq in vivo or during the IP procedure.
As ChiX was the only sRNA in our screen that titrated Hfq away from RNA-IN and ChiX is unique among sRNAs in that it contains a distal Hfq-binding site, we wondered if other RNAs that interact with the distal surface of Hfq would increase transposase expression through an Hfqtitration mechanism. Most Hfq-binding mRNAs (including sodB, ptsG and maeA) interact solely with the distal surface of Hfq (Zhang et al., 2013). We overexpressed the first 300-nt of sodB, ptsG or maeA mRNA and measured the impact on transposase expression. Note that the mRNAs were expressed from the same plasmid background as the sRNA overexpression library. Unexpectedly, transposase expression was mostly unaffected by mRNA overexpression (2.4-fold increase for sodB; 1.3fold increase for ptsG and maeA) (Supporting Information Fig. S6) cells containing the chromosomal IS101-339-lacZ translational fusion (DBH299) were co-transformed with a plasmid encoding Hfq WT -3xFLAG (pDH909) and a plasmid expressing ChiX (pDH765) or vector control (pDH763). Hfq was immunoprecipitated from cell lysates with ANTI-FLAG ® M2 magnetic beads. Total input RNA (10 μg) or RNA recovered from the IP (0.3 μg) was analyzed by Northern blot for ChiX and the 5S rRNA, or primer extension of RNA-IN and lpp. Band intensities were quantified using ImageQuant. and in vivo. Because overexpression of mRNAs containing a distal-binding site did not affect transposase expression, we think it likely that the ability of ChiX to up-regulate transposase expression is due to the fact that this sRNA possesses both distal and proximal Hfqbinding sites.

Direct repression of IS10 translation by Hfq
Hfq typically regulates translation by catalyzing pairing of an sRNA to the TIR of an mRNA. In the simplest model, Hfq simultaneously binds an sRNA and cognate mRNA near or overlapping the pairing sequences (also known as seed regions), and as pairing proceeds, Hfq is released from the sRNA-mRNA duplex, whereupon it can catalyze additional pairing reactions (Fender et al., 2010;Hwang et al., 2011;Panja and Woodson, 2012;Tree et al., 2014). The role of each Hfq-binding surface in sRNA-dependent regulation has been studied extensively. A study of seven different sRNA-mRNA pairs found that the proximal surface (in particular lysine 56) is critical for Hfq chaperone activity and sRNA stability (Zhang et al., 2013). The lateral surface of Hfq also interacts with sRNAs and is important for sRNA stability as well as providing a favorable surface for nucleating pairing (Sauer and Weichenrieder, 2011;Sauer et al., 2012;Panja et al., 2013); however, this surface of Hfq is dispensable for some systems. Unlike the proximal surface, the distal RNA-binding site is not an absolute requirement for sRNA-dependent regulation and in some cases may simply serve as a way to tether Hfq to target mRNAs (Sauer, 2013;Zhang et al., 2013). The role of each RNA-binding surface of Hfq in sRNA-independent regulation has not been studied.
We first measured the impact of Hfq-binding face mutations on IS10 transposase expression and transposition. The finding that the proximal surface is only partially required for regulation suggested that Hfq is functioning independent of a trans-encoded sRNA. The moderate effect of the K56A mutation on Hfq is likely a result of reduced binding to site 2 in RNA-IN, which is supported by decreased regulation in the presence of the M2 mutations. Additionally, the lateral surface was not required for repressing transposase expression or transposition. Alone, the R17A phenotype does not exclude sRNAdependent regulation. In the case of rpoS, multiple mutations to the lateral surface resulted in the strongest decrease in sRNA-mediated activation of rpoS expression (Panja et al., 2013). Additionally, the lateral surface was only important for about half of the sRNA/mRNA pairs tested previously (Zhang et al., 2013). However, a dispensable lateral surface is consistent with sRNA-independent regulation in the IS10 system. Unlike most sRNA-dependent regulation, the distal surface is critical for repressing transposase translation and mutations that block the interaction between the TIR of RNA-IN and the distal site on Hfq strongly de-repressed transposase expression in vivo. Additionally, an available distal surface on Hfq was required for blocking 30S ribosome binding to RNA-IN in vitro. We therefore suggest that sRNAindependent regulation by Hfq requires mRNA binding through the distal surface and not the proximal or lateral surfaces.
For Hfq to be an effective direct inhibitor of translation, we think some very specific requirements must be met. Firstly, there needs to be an Hfq-binding site in the TIR of an mRNA. Interestingly, a recent survey of RNA sequences bound by Hfq in vivo included repeated trinucleotide motifs (ARN) that were frequently associated with the Shine-Dalgarno sequence. Notably, 18% of all Hfqassociated mRNAs contained Hfq-binding sites located within the TIR (Tree et al., 2014). Secondly, the Hfq-binding site within the TIR should be a high affinity site to ensure in vivo binding. Thirdly, there must be sufficient available Hfq to act stoichiometrically on TIRs. Given that Hfq is a highly expressed protein, this third factor might not appear to be a limitation. However, there is growing evidence that despite being highly abundant (5000-10 000 hexamers per cell) (Kajitani et al., 1994;Ali Azam et al., 1999;Argaman et al., 2012), the amount of unbound Hfq at any given time might in fact be limiting for RNA binding (Hussein and Lim, 2011;Moon and Gottesman, 2011). Hfq has a large number of specific mRNA and sRNA targets and may also be sequestered through mostly non-specific DNA interactions Updegrove et al., 2010). Accordingly, for translational repression where a sustained interaction with an mRNA is required to block 30S ribosomal subunit binding, it is likely critical that Hfq binds with extremely high affinity to the TIR, which might compensate for limited availability of Hfq. We found this to be the case in the IS10 system as Hfq bound the TIR with an affinity of approximately 0.2 nM. This represents one of the highest affinity interactions between Hfq and an mRNA [cf. ompA 1 nM, ompC 0.9 nM, ompF 4 nM (Fender et al., 2010), sodB 0.3 nM (Geissmann and Touati, 2004) and rpoS ∼50 nM (Soper et al., 2011;Peng et al., 2014)]. Moreover, we found that a moderate increase in expression of ChiX sRNA was sufficient to de-repress transposase expression. Given our evidence that (i) ChiX possesses a distal Hfq-binding site; (ii) its overexpression did not substantially influence RNA-IN steady-state levels; (iii) ChiX does not base-pair with RNA-IN; and (iv) ChiX overexpression reduces the amount of RNA-IN bound by Hfq in vivo, we think the results of the ChiX overexpression experiment are most easily explained by a ChiX-Hfq titrating mechanism. That is, despite the high affinity of site 1 in RNA-IN for Hfq, moderate overexpression of ChiX was sufficient to deplete the pool of available Hfq such that there were insufficient amounts to repress RNA-IN translation. The implications of Hfq titration by ChiX are discussed below.

Hfq regulation of IS10 in single and multi-copy
All of the in vivo experiments reported in this work were performed under conditions where the naturally occurring antisense RNA (RNA-OUT) was not produced. This was meant to mimic a situation where IS10 is in single copy and the antisense RNA has little to no effect on transposase expression (Kleckner, 1990). The HH104 mutation was introduced to the single-copy IS10 elements to increase transposase transcription to detectable levels, and although this mutation also eliminates the small amount of cis-encoded RNA-OUT, there would be little antisense control to begin with. We propose that under these conditions, Hfq acts in a stoichiometric manner to limit RNA-IN translation (Fig. 10). However, when IS10 is present on a multi-copy plasmid, stoichiometric action of Hfq on RNA-IN might not be sufficient to limit translation because of the increase in the number of RNA-IN transcripts. In support of this, when we compared Hfq regulation of a chromosomal IS10-lacZ fusion with a high-copy translational fusion, the extent of Hfq-mediated repression was attenuated from 13-to 3-fold (Figs 2A and 5). With a high-copy IS10 element, the MCI pathway comes into play and RNA-OUT becomes an important negative regulator of RNA-IN translation and stability (Case et al., 1990). We have previously shown that Hfq binds RNA-OUT and promotes its restructuring to expose sequences important for RNA-IN pairing (Ross et al., 2013). As Hfq facilitates RNA-IN:OUT pairing in a catalytic cycle, we think that when IS10 is in multi-copy, this mechanism would predominate over the 'stoichiometric' inhibition pathway. We eliminated RNA-OUT from the multi-copy IS10 1-242-lacZ translational fusion so we could study antisense-independent regulation by Hfq, and our results show that stoichiometric repression can still occur albeit regulation is weaker for multi-copy IS10. We therefore think that Hfq is a negative regulator of IS10 regardless of copy number. This model may also explain previous results where it was found that Hfq is a stronger negative regulator of transposition when IS10 is present in multicopy compared with single copy (Ross et al., 2010). Our model would predict that in this scenario, Hfq is repressing predominantly by a catalytic mechanism involving Hfqmediated RNA-IN:OUT pairing, which presumably is more efficient than the stoichiometric repression that is confined to the single-copy situation.
This dual model for Hfq repression may be applicable to other members of the Hfq regulon. In the absence of a cognate sRNA, we suggest that Hfq would repress translation of mRNAs containing an Hfq-binding site in the TIR by a stoichiometric mechanism. The strength of this regulation would be governed primarily by the affinity of Hfq for that site relative to other cellular mRNAs. This model might also explain the conflicting data concerning regulation of ompA expression. Work in the Bläsi lab suggested that Hfq directly represses ompA translation by binding the TIR in a manner analogous to that described here for  IS10 (Vytvytska et al., 2000). This is supported by several surveys of Hfq-binding mRNAs that have identified ompA as an Hfq-binding mRNA (Zhang et al., 2003;Sittka et al., 2008;Tree et al., 2014). The characterization of the stationary-phase sRNA MicA by the Wagner lab suggested that the primary role of Hfq in regulating ompA expression was to promote sRNA-mRNA pairing. Our model would combine both mechanisms and suggest that Hfq exerts some basal repression of ompA translation that is strengthened in stationary phase by promoting MicA pairing with ompA. Additionally, our model is applicable to the sdhC system where Hfq had an sRNA-independent effect on expression that is presumably a result of direct repression of translation (Desnoyers and Massé, 2012).

Hfq titration by the sRNA ChiX
Induction of Hfq-binding sRNAs can impinge on other Hfq-dependent post-transcriptional networks. In the simplest model, induction of an sRNA would provide enough sRNA molecules to bind all available Hfq and even compete with other Hfq-binding RNAs. Hfq titration was first proposed as a mechanism for OxyS repression of rpoS expression, which was later verified (Zhang et al., 1998;Moon and Gottesman, 2011). ChiX overexpression also resulted in increased rpoS-lacZ expression, and this effect was proposed to be a result of Hfq titration . Additionally, overexpression of ArcZ in Salmonella was shown to have a pleiotropic effect on gene expression (altering expression of 757 genes) in part by decreasing the number of mRNAs bound to Hfq as well as specific competition with the sRNAs CyaR and InvR (Papenfort et al., 2009). Our studies with ChiX overexpression have provided another example of an sRNA sequestering sufficient amounts of Hfq to produce a biological effect: de-repression of RNA-IN translation. ChiX (previously named SroB, RybC and MicM) is a negative regulator of genes involved in chitobiose utilization. ChiX is constitutively expressed but its levels increase substantially in stationary phase (Vogel et al., 2003;Figueroa-Bossi et al., 2009;Papenfort and Vogel, 2014). The fact that ChiX was the only sRNA of 14 screened to have an impact on expression of RNA-IN under antisense-independent conditions fits fully with our model of stoichiometric inhibition resulting from Hfq binding the TIR of RNA-IN (site 1) through its distal binding site. ChiX is somewhat unique among E. coli sRNAs, as (according to our footprinting data) it possesses both distal and proximal Hfq-binding sites. Moreover, in a recent study that compared the ability of a set of sRNAs to compete for Hfq binding, it was established that ChiX was at the top of the hierarchy, while SgrS was at the bottom. This same work also showed that the A-rich region of ChiX that we designated Hfq-I interacts with the distal surface of Hfq and provides important determinants for Hfq competition (Małecka et al., 2015). This fits fully with our data showing that ChiX overexpression titrated Hfq away from RNA-IN while SgrS did not.
ChiX is unique among sRNAs as it acts catalytically to repress its target mRNA (chiP, previously known as ybfM), and ChiX levels are regulated by an 'anti-sRNA', chbBC (Overgaard et al., 2009). It is therefore unexpected that ChiX levels would increase so dramatically during stationary phase. Given the current work, it is tempting to speculate that ChiX has a yet unidentified role during the transition to stationary phase; notably, ChiX constitutes 24-26% of Hfq-bound sRNAs during early stationary phase (Chao et al., 2012) . Based on ChiX's distinct interaction properties with Hfq, it is worth considering the possibility that ChiX expression could influence the entire Hfq regulon during the transition to stationary phase.

Bacterial strains, phage, plasmids and oligonucleotides
All bacterial strains, phage and plasmids used in this study are listed in Supporting Information Table S1 and oligonucleotides are listed in Supporting Information Table S2.
The Hfq expression plasmids used for complementation experiments were made by amplifying the hfq gene (including the P3 promoter) from pDH700, pDH701 and pDH713 (Ross et al., 2013) with primers oDH518 and oDH519. The PCR product was digested with HaeIII and cloned into the XmnI/ScaI sites of pACYC184. The R17A mutation was first introduced into pDH700 by overlap PCR using primers oDH518, oDH519, oDH520 and oDH521. The PCR product was cloned into the XbaI/HindIII sites of pDH700 to make pDH874, and Hfq R17A was then subcloned into pACYC184 as above. Hfq was amplified from pDH904 with oDH184 and oDH479 to add a C-terminal 3xFLAG tag and this amplicon was cloned directly into XmnI/ScaI-digested pACYC184.
The multi-copy IS10-lacZ translational fusion is a derivative of pNK2974 (Jain, 1995). First, IS10 was amplified with primers oDH502 and oDH503 and this amplicon was cloned into the EcoRI/HindIII sites of pNK2974 to produce pDH858, which contains the first 242-nt of IS10R fused in frame to codon 10 of lacZ. All subsequent mutations were introduced into pDH858 using overlap PCR with primers oDH505 and oDH13 and the relevant mutagenic primers: R5 (oDH506, oDH507), M2 (oDH498, oDH499) and M5 (oDH508, oDH509). PCR products were digested with EcoRI and HindIII and cloned into the same sites in pDH858. M5 and M2 were originally identified as increased expression mutants in a transposase expression screen. The transposase gene used was derived from a library of sequences generated by mutagenic PCR.
Plasmids overexpressing sodB, ptsG and maeA were constructed as previously described (Zhang et al., 2013). MC4100 genomic DNA served as a template for PCR with the following primers: oDH558 and oDH559, sodB; oDH560 and oDH561, ptsG; and oDH562 and oDH563, maeA. PCR products were digested with AatII and EcoRI and cloned into the same sites of pDH765.

RNA extraction and primer extension analysis
Cells were grown in LB supplemented with ampicillin (100 μg ml −1 ) to OD600 = 0.6 at which time 600 μl of cells was added to 300 μl of RNA lysis buffer [1.5% (w/v) sodium dodecyl sulfate (SDS), 300 mM sodium acetate, 30 mM EDTA] and boiled for 1 min. Samples were chilled on ice for 30 s and then sequentially extracted twice with acid phenol (pH 4.3), once with phenol:chloroform:isoamyl alcohol (25:24:1) and once with 2-butanol followed by ethanol precipitation. Residual genomic DNA was removed with TURBO DNase (Ambion, Life Technologies) prior to primer extension analysis. Five or 10 μg of total RNA (Figs 5B, 8C and 9 respectively) was subject to primer extension analysis with 5′ 32 P-labeled oDH511 (RNA-IN) and oDH482 (lpp) and SuperScript III (Invitrogen) according to the manufacturer's instructions.

Toeprinting
The 30S ribosomal subunit was prepared as previously described (Fechter et al., 2009). In vitro transcribed RNA (2 pmol) was annealed to 5′ 32 P-labeled oDH511 (RNA-IN), oDH482 (lpp) and oDH555 (usg) in toeprint buffer (10 mM Tris-acetate, pH 7.6, 1 mM DTT, 100 mM potassium acetate) by heating to 95 o C for 1 min followed by snap-cooling on ice for 2 min. While on ice, magnesium acetate was added to a final concentration of 10 mM and dNTPs to 0.5 mM. Reactions were then incubated at 37 o C for 5 min. Hfq (0.5-4 pmol of hexamers) or buffer was added to reactions, which were incubated for another 15 min at 37 o C, followed by addition of 30S ribosome (3.6 pmol for RNA-IN and usg; 2.7 pmol for lpp) and incubation at 37 o C for 5 min. Initiator fMet-tRNA (10 pmol; Sigma-Aldrich) was added and reactions were incubated for a further 15 min before addition of 200 U of