Role of the terminator hairpin in the biogenesis of functional Hfq-binding sRNAs

Rho-independent transcription terminators of the genes encoding bacterial Hfq-binding sRNAs possess a set of seven or more T residues at the 3′ end, as noted in previous studies. Here, we have studied the role of the terminator hairpin in the biogenesis of sRNAs focusing on SgrS and RyhB in Escherichia coli. We constructed variant sRNA genes in which the GC-rich inverted repeat sequences are extended to stabilize the terminator hairpins. We demonstrate that the extension of the hairpin stem leads to generation of heterogeneous transcripts in which the poly(U) tail is shortened. The transcripts with shortened poly(U) tails no longer bind to Hfq and lose the ability to repress the target mRNAs. The shortened transcripts are generated in an in vitro transcription system with purified RNA polymerase, indicating that the generation of shortened transcripts is caused by premature transcription termination. We conclude that the terminator structure of sRNA genes is optimized to generate functional sRNAs. Thus, the Rho-independent terminators of sRNA genes possess two common features: a long T residue stretch that is a prerequisite for generation of functional sRNAs and a moderate strength of hairpin structure that ensures the termination at the seventh or longer position within the consecutive T stretch. The modulation of the termination position at the Rho-independent terminators is critical for biosynthesis of functional sRNAs.


INTRODUCTION
Hfq-binding small RNAs (sRNAs), major regulatory RNAs in bacteria, are induced under specific physiological and/or stress conditions and regulate, along with an RNA chaperone Hfq, the expression of target genes at the post-transcriptional level (Waters and Storz 2009;Gottesman and Storz 2010;Vogel and Luisi 2011;Wagner and Romby 2015). The primary role of Hfq is to accelerate base-pairing between sRNAs and target mRNAs to regulate their translation. The second role of Hfq is to recruit RNase E near target mRNAs leading to rapid degradation of sRNA-mRNA hybrids (Massé et al. 2003;Morita et al. 2005). Hfq also plays a role in stabilization of sRNAs by protecting them from the attack of ribonucleases (Massé et al. 2003).
The biosynthesis of sRNAs in cells is regulated primarily at the transcription initiation step. The promoter of the individual sRNA gene is under the control of at least one transcription factor that is modulated by the cognate stress. SgrS and RyhB of Escherichia coli are among well-characterized sRNAs. In response to the glucose-phosphate stress, such as accumulation of glucose-6-phosphate, a transcription factor SgrR is activated to stimulate the transcription of sgrS encoding SgrS (Vanderpool and Gottesman 2007). The transcribed SgrS pairs with target mRNAs, such as the ptsG mRNA encoding the major glucose transporter, to either down-or up-regulate their expression to attenuate the glucose-phosphate stress (Vanderpool and Gottesman 2004;Papenfort et al. 2013). The promoter of ryhB encoding RyhB is under the control of the Fur repressor. Depletion of Fe 2+ inactivates Fur, resulting in induction of RyhB, which in turn regulates the translation of several mRNAs encoding Fe-binding proteins through base-pairing (Massé and Gottesman 2002;Massé et al. 2003).
The production of sRNAs is regulated not only at the step of transcription initiation but also at the step of transcription termination (Morita et al. 2015). The sRNA genes possess a typical Rho-independent or factor-independent or intrinsic transcription terminator encoding a GC-rich RNA hairpin followed by a run of T residues (d'Aubenton Carafa et al. 1990;Ray-Soni et al. 2016). We demonstrated previously that transcription termination at the sRNA genes is enhanced under stress conditions and thereby contributes to an efficient production of active sRNAs (Morita et al. 2015). A striking feature of the Rho-independent terminators of sRNA genes is that the length of the T residue stretch is longer than seven (Otaka et al. 2011;Ishikawa et al. 2012). This feature is a prerequisite for generation of functional sRNAs because a poly(U) tail longer than seven is essential for sRNAs to bind efficiently to Hfq (Otaka et al. 2011). To generate surely the functional sRNAs, however, transcription termination must occur at the seventh or longer position within the T residue stretch by preventing "premature" termination. This implies that the modulation of the termination position at the Rho-independent terminators of sRNA genes is quite important for biosynthesis of functional sRNAs. The stability of the terminator RNA hairpin and the length of the T residue stretch are known to be major determinants for the efficiency of transcription termination, although DNA sequences around terminators also affect the termination efficiency (Lynn et al. 1988;Cheng et al. 1991;Reynolds et al. 1992;Ray-Soni et al. 2016). When the terminator hairpin is too weak, elongating RNA polymerase mostly reads through the terminator, resulting in extended forms of sRNAs, which are nonfunctional (Morita et al. 2015). On the other hand, the termination efficiency is expected to increase when the terminator hairpin is stabilized.
It is an interesting possibility that the stability of the terminator RNA hairpin affects not only the termination efficiency but also the termination position within the polythymidine stretch. The aim of the present work is to examine this possibility by using SgrS and RyhB as model sRNAs. We constructed variant sgrS and ryhB genes in which the GC-rich inverted repeat sequences of terminators are extended to stabilize the terminator hairpins. We showed by Northern blotting and 3 ′ RACE experiments that the extension of the stem leads to production of heterogeneous shorter transcripts possessing shortened poly(U) tails. The shortened transcripts can be generated in an in vitro system with purified RNA polymerase, confirming our working hypothesis that the stabilization of the terminator hairpin causes premature termination. We conclude that the control of the termination position within the T residue stretch is critical for the biogenesis of functional sRNAs, and that the terminator hairpin structure of sRNA genes is optimized to produce efficiently functional sRNAs under stress conditions.

Stabilization of the terminator hairpin enhances termination efficiency
We demonstrated previously by using a "double terminator system" that RNA polymerase frequently reads through the Rho-independent terminator of sgrS under normal growth conditions (Morita et al. 2015). The moderate termination efficiency of the sgrS terminator could be primarily due to a moderate thermodynamic stability (ΔG) of the terminator hairpin. If so, it is expected that stabilization of the hairpin structure of the sgrS terminator enhances the termination efficiency. To test this possibility, we constructed plasmid pSgrS-S-LS1-rplLT ("LS" stands for "long stem") carrying the sgrS-S-LS1-rplLT by inserting four GC base pairs into the sgrS hairpin stem on plasmid pSgrS-S-rplLT. The full DNA sequence of sgrS-S-rplLT and the sequence around the terminator region of sgrS-S-LS1-rplLT are shown in Figure 1A. The predicted terminator RNA hairpin structures, along with their ΔG (kcal/mol), are shown in Figure 1B. The sgrS-S-rplLT is a hybrid gene in which the second terminator derived from rplL is placed just downstream from the sgrS-S (Morita et al. 2015). SgrS-S, the 3 ′ portion of SgrS, represents a minimal functional region of SgrS consisting of the basepairing region and the Hfq-binding module including the Rho-independent terminator sequence (Otaka et al. 2011;Ishikawa et al. 2012). The insertion of GC pairs increases ΔG of the terminator RNA hairpin of SgrS-S from −10.7 to −24.0. Each plasmid was introduced into TM772 (ΔsgrS Δhfq) cells (Table 1). The sgrS-S gene is under the control of an arabinose-inducible P BAD promoter in these plasmids. Cells were grown in LB medium and the expression of sRNAs was induced by arabinose. Total RNAs were prepared and analyzed by Northern blotting. We first confirmed the previous observation (Morita et al. 2015) that the FIGURE 1. Effect of extension of the terminator stem of sgrS on transcription termination. (A) DNA sequences of sgrS-S-rplLT and sgrS-S-LS1-rplLT. The sequence corresponding to the sgrS-S is shown as regular letters, whereas the terminator sequence derived from the rplL is shown as italic letters. The inverted repeat sequences are indicated by horizontal arrows. Nucleotides are numbered from the site corresponding to the 5 ′ end of sgrS-S. The inserted sequences to stabilize the terminator stem are shown as bold letters. (B) The predicted secondary structures and the thermodynamic stabilities (ΔG, kcal/mol) of terminator RNA hairpins without poly(U) sequence were determined according to the Mfold program (Zuker 2003). (C) Analysis of transcription termination. TM772 (ΔsgrS Δhfq) cells harboring indicated plasmids were grown in LB medium. At A 600 = 0.6, 0.02% arabinose was added and incubation was continued for 5 min. Total RNAs were prepared, and 10 or 0.25 µg of RNA samples was subjected to Northern blotting using the SgrS-S probe and tmRNA probe, respectively. readthrough occurs frequently resulting in a significant amount of the readthrough product, SgrS-S-rplLT, along with the terminated product, SgrS-S (Fig. 1C, lane 1). When the RNAs generated from the sgrS-S-LS1-rplLT were analyzed, the readthrough product markedly decreased, whereas the relative abundance of terminated products increased (Fig. 1C,lane 2). This implies that stabilization of the terminator hairpin enhances the termination efficiency, as expected. Interestingly, the sgrS-S-LS1-rplLT generates heterogeneous shorter transcripts (Fig. 1C, lane 2), suggesting that the extension of the terminator stem causes premature termination within or before the polythymidine stretch, resulting in transcripts in which the poly(U) tail is shortened.
Generation of shorter transcripts is caused by stabilization of the terminator hairpin and enhanced by glucose-phosphate stress To focus more on the effect of the strength of the terminator hairpin on the generation of the heterogeneous shorter transcripts, we constructed plasmid pSgrS-S-LS1 carrying the sgrS-S-LS1 by inserting four GC base pairs into the hairpin stem of the sgrS-S on plasmid pSgrS-S ( Fig. 2A). We also constructed plasmid pSgrS-S-LS2 carrying the sgrS-S-LS2 by inserting seven GC base pairs ( Fig. 2A). The ΔG of the RNA hairpin of SgrS-S-LS1 is -24.0, whereas that of SgrS-S-LS2 is -32.2 (Fig. 2B). Each plasmid was introduced into TM772 (ΔsgrS Δhfq) cells. Cells were grown in LB medium and expression of sRNAs was induced by arabinose. Total RNAs were prepared and analyzed by Northern blotting. The sgrS-S-LS1 generated heterogeneous shorter transcripts along with the full-length SgrS-S-LS1 as expected (Fig. 2C, lane 2). The generation of heterogeneous shorter transcripts from the sgrS-S-LS2 (lane 3) was more significant compared with the sgrS-S-LS1, indicating that the extent of the production of the heterogeneous shorter transcripts correlates with the strength of the terminator hairpin.
Because transcription termination is enhanced under stress conditions (Morita et al. 2015), we assume that the glucosephosphate stress may also enhance the production of shorter transcripts. Then, we tested the effect of glucose-phosphate stress on the generation of shorter transcripts at the sgrS terminator by using ΔsgrS Δhfq cells harboring either pSgrS-S, pSgrS-S-LS1, or pSgrS-S-LS2. Cells were grown in LB medium to exponential phase and exposed to nonmetabolizable glucose analog α-methylglucoside (αMG), and then to arabinose. Total RNAs were prepared and subjected to Northern blotting using the SgrS-S probe (Fig. 2D). Both the sgrS-S-LS1 and sgrS-S-LS2 generated the heterogeneous shorter transcripts along with the full-length transcripts (Fig. 2D, lanes 2,3). Again, the generation of shorter transcripts from the sgrS-S-LS2 was more significant compared with the sgrS-S-LS1. In addition, the production of shorter transcripts from the sgrS-S-LS1 and sgrS-S-LS2 was clearly enhanced under the glucose-phosphate stress. The generation of shorter transcripts was not observed from the wild-type sgrS-S even under the stress condition (Fig. 2D, lane 1). We also tested the effect of replacing the terminator hairpin of sgrS-S with the corresponding portion derived from strong Rho-independent terminators. For this, we constructed plasmid pSgrS-S-LS3 and pSgrS-S-LS4 in which the terminator hairpin was replaced with those derived from rrnB and rplL, respectively ( Fig. 2A). The thermodynamic stabilities of these hairpins are −27.4 and −20.4, respectively (Fig. 2B). Cells harboring each plasmid were grown in LB medium to exponential phase and exposed to αMG, and then to arabinose. Total RNAs were prepared and subjected to Northern blotting using the SgrS-S probe (Fig. 2E). The data indicate that heterogeneous shorter transcripts are generated efficiently from pSgrS-S-LS3 and pSgrS-S-LS4 (Fig. 2E, lanes 2,3). This suggests that the stabilization of the terminator hairpin itself rather than specific sequences within the hairpin is responsible for the generation of heterogeneous shorter transcripts.

Properties of the heterogeneous shorter transcripts
We expect that the poly(U) tail is shortened in the heterogeneous shorter transcripts. If so, the levels of shorter transcripts would not be affected by the hfq backgrounds because SgrS-S possessing a shorter poly(U) tail no longer binds to Hfq (Otaka et al. 2011). Then, we investigated the expression of SgrS-S in TM542 (hfq + ) and TM772 (Δhfq) cells (Table 1) under the glucose-phosphate stress condition in which the generation of shorter transcripts is enhanced. The abundance of SgrS-S increased in hfq + cells compared with in Δhfq cells (Fig. 3A, lanes 1,2). On the other hand, the levels of shorter transcripts were not affected by the hfq backgrounds, whereas the abundance of full-length SgrS-S-LS1 and SgrS-S-LS2 was significantly elevated in hfq + cells ( Fig. 3A, lanes 3-6). This strongly suggests that the shorter transcripts lose the ability to bind to Hfq.
To examine directly the Hfq-binding ability of shorter transcripts, we carried out a pull-down assay using cell extracts of hfq-Flag cells harboring pSgrS-S-LS2. Cell extracts were incubated with anti-Flag M2-agarose beads. Proteins bound to the agarose beads were analyzed by Western blotting using anti-Flag antibodies. The affinity-purified Hfq-Flag was treated with phenol and subjected to Northern blotting. As shown in Figure 3B, the full-length SgrS-S-LS2 but not shorter transcripts copurified with Hfq-Flag. This indicates that the shorter transcripts generated from the sgrS-S-LS2 are not able to bind to Hfq. Thus, it is highly likely that the poly(U) tail is shortened in the heterogeneous shorter transcripts.
The abundance of full-length active SgrS-S variants is markedly decreased with accumulation of inactive shorter transcripts in cells harboring pSgrS-S-LS1 or pSgrS-S-LS2, The predicted secondary structures and the thermodynamic stabilities (ΔG, kcal/mol) of terminator RNA hairpins without a poly(U) sequence were determined according to the Mfold program (Zuker 2003). (C ) The effect of extension of the terminator stem of sgrS-S on the generation of shorter transcripts. TM772 (ΔsgrS Δhfq) cells harboring indicated plasmids were grown in LB medium. At A 600 = 0.6, 0.2% arabinose was added and incubation was continued for 10 min. Total RNAs were prepared, and 20 or 0.25 µg of RNA samples was subjected to Northern blotting using the SgrS-S probe and tmRNA probe, respectively. Arrowheads represent the full-length transcripts. (D) The effect of extension of the terminator stem of sgrS-S on the generation of shorter transcripts under glucose-phosphate stress. TM772 (ΔsgrS Δhfq) cells harboring indicated plasmids were grown in LB medium. At A 600 = 0.6, 0.01% αMG was added and incubation was continued for 10 min. Then, 0.2% arabinose was added and incubation was continued for 10 min. Total RNAs were prepared, and 20 or 0.25 µg of RNA samples was subjected to Northern blotting using the SgrS-S probe and tmRNA probe, respectively. Arrowheads represent the full-length transcripts. (E) The effect of replacement of the terminator hairpin of sgrS-S with those derived from rrnBT and rplLT on the generation of shorter transcripts. TM772 (ΔsgrS Δhfq) cells harboring indicated plasmids were grown in LB medium and treated as described in D. Total RNAs were prepared, and 20 or 0.25 µg of RNA samples was subjected to Northern blotting using the SgrS-S probe and tmRNA probe, respectively. Arrowheads represent the full-length transcripts.
in particular under the glucose-phosphate stress. Thus, it is expected that the down-regulation of ptsG mRNA, the target of SgrS, is impaired in these cells. To examine this, we performed Northern blot analysis by using the ptsG probe. First, we analyzed the expression of ptsG mRNA in cells harboring the vector plasmid pAraX, pSgrS-S, pSgrS-S-LS1, or pSgrS-S-LS2 under normal growth conditions (Fig. 3C). The ptsG mRNA was highly expressed in cells harboring pAraX (Fig. 3C, lane 1), whereas the abundance of the fulllength ptsG mRNA was dramatically reduced when SgrS-S was expressed (Fig. 3C, lane 2) because SgrS-S accelerates the RNase E-dependent degradation of ptsG mRNA . The down-regulation of ptsG mRNA was also observed when SgrS-S-LS1 or SgrS-S-LS2 was expressed ( Fig. 3C, lanes 3,4). These results suggest that the abundance of the full-length SgrS-S-LS1 and SgrS-S-LS2 is still sufficient to down-regulate the ptsG mRNA under nor-mal growth conditions. Then, we carried out Northern analysis by using RNA samples prepared from cells exposed to the glucose-phosphate stress (Fig. 3D). The expression of ptsG mRNA was not affected by the stress in cells harboring pAraX (Fig. 3D, lane 1). The dramatic reduction of the ptsG mRNA was observed in cells harboring pSgrS-S as expected (Fig. 3D, lane 2). On the other hand, the abundance of ptsG mRNA clearly increased in cells harboring pSgrS-S-LS1 or pSgrS-S-LS2 under glucose-phosphate stress (Fig.  3D, lanes 3,4), indicating that the down-regulation of ptsG mRNA was partially suppressed in these cells. It is apparent that the marked reduction of the active full-length SgrS-S variants is responsible for this suppression of the down-regulation of ptsG mRNA.

Analysis of 3 ′ ′ ′ ′ ′ ends of transcripts
To verify that the poly(U) tail is indeed shortened in the heterogeneous shorter transcripts, we determined the 3 ′ ends of RNAs generated from the sgrS-S-LS2 in hfq + cells under the glucose-phosphate stress by the 3 ′ -RACE (rapid amplification of cDNA ends) experiment (Argaman et al. 2001;Kawano et al. 2005). In this experiment, total cellular RNAs and a linker DNA oligo are ligated in vitro and the RNA-DNA junction region is amplified by RT-PCR. The amplified cDNA fragments were cloned into a plasmid and subjected to DNA sequence analysis. The 3 ′ ends of transcripts can be determined by finding the junction between RNAs and the 5 ′ end of ligated linker oligo DNA. We first analyzed the 3 ′ ends of transcripts generated from the sgrS-S. The total RNAs were separated by electrophoresis on a polyacrylamide gel containing 8 M urea. The gel region corresponding to SgrS-S was cut out and RNAs were purified from the gel piece (Fig. 4A). The purified RNAs were subjected to the 3 ′ -RACE experiment. The sequences around the junction region of randomly isolated 15 cDNA clones indicate that 11, three, and one clones correspond to SgrS-S possessing 8U, 7U, and 6U tail, respectively (Fig. 4B). Thus, the sgrS-S is able to produce quite efficiently active SgrS-S possessing a long poly(U) tail. Then, we analyzed the 3 ′ ends of transcripts generated from the sgrS-S-LS2. The gel region corresponding to heterogeneous RNA bands was divided into three pieces as shown in Figure 4C. RNAs prepared from each piece were subjected to the 3 ′ -RACE experiment (Fig. 4D). The sequence analysis of randomly isolated cDNA clones indicates that RNAs derived from the upper region exclusively correspond to SgrS-S-LS2 possessing a long poly(U) tail consisting of eight or seven uridine residues. The length of the poly(U) tail of RNAs derived from the middle region varies from four to eight. The length of the poly(U) tail of RNAs derived from the lower region was <3. The 3 ′ ends of several RNAs obtained from the lower region were mapped in the sgrS terminator region preceding the poly (U) tail. Thus, we conclude that the poly(U) tail of the heterogeneous shorter transcripts is indeed shortened. TM542 (ΔsgrS) and TM772 (ΔsgrS Δhfq) cells harboring indicated plasmids were grown in LB medium. At A 600 = 0.6, 0.1% αMG was added and incubation was continued for 10 min, and then 0.02% arabinose was added and incubation was continued for 5 min. Total RNAs were prepared, and 10 or 0.25 µg of RNA samples was subjected to Northern blotting using the SgrS-S probe and tmRNA probe, respectively. Arrowheads represent the full-length transcripts. (B) An examination of Hfq binding of heterogeneous transcripts. TM803 (ΔsgrS hfq-Flag) cells harboring pSgrS-S-LS2 were grown in LB medium. At A 600 = 0.6, 0.1% αMG was added and incubation was continued for 10 min, and then 0.02% arabinose was added and incubation was continued for 5 min. Crude extract was prepared and subjected to the pull-down assay using anti-Flag agarose as described in Materials and Methods. (C,D) The effect of extension of the terminator stem of sgrS-S on SgrS function in the presence (D) and absence (C ) of glucose-phosphate stress. TM542 (ΔsgrS) cells harboring indicated plasmids were grown in LB medium. To prepare total RNAs from cells without the stress, 0.02% arabinose was added at A 600 = 0.6 and incubation was continued for 5 min (C ). To prepare total RNAs from cells with the stress, the culture of A 600 = 0.6 was exposed to 0.1% αMG for 10 min, and then 0.02% arabinose for 5 min (D). Ten micrograms of RNA samples was subjected to Northern blotting using the ptsG probe.

Effect of 3 ′ ′ ′ ′ ′ exoribonuclease on the generation of shorter transcripts and the stability of transcripts
Our hypothesis is that the shorter transcripts are generated by premature transcription termination within or before the terminator polythymidine stretch. However, it is also possible that the shorter products are generated through trimming of the primary transcripts by 3 ′ exoribonucleases. To test this possibility, we investigated the expression of sgrS-S-LS2 in cells lacking either PNPase (pnp), RNase II (rnb), or RNase R (rnr), respectively. The production of shorter transcripts was not significantly affected by the individual mutation (Fig. 5A). Although this result is consistent with our hypothesis, it does not prove that the shorter transcripts are generated primarily by premature termination because we do not rule out the involvement of 3 ′ exonucleases based on the results from pnp, rnb, and rnr single mutants.
We also performed a chase experiment to examine the fate of the heterogeneous transcripts in ΔsgrS Δhfq cells harboring pSgrS-S-LS2. Cells were grown to exponential phase, and expression of sRNAs was induced by arabinose. Rifampicin was added to prevent further initiation of transcription. RNAs were isolated at various times after the addition of rifampicin, and subjected to Northern blotting (Fig. 5B). The stability of shorter transcripts was essentially the same as that of the full-length transcripts. As expected, the full-length transcript but not the heterogeneous transcripts were markedly stabilized in the hfq + background (data not shown). We also examined the fate of the heterogeneous transcripts under the glucose-phosphate stress. In this case, cells were exposed to αMG prior to the addition of arabinose. Again, the stability of shorter transcripts was the same as that of the full-length transcripts, although both transcripts were significantly stabilized by the stress (Fig. 5C). These results suggest that the shorter transcripts are generated by transcription termination rather than by degradation of primary transcripts.
The shortened transcripts are generated in vitro with purified RNA polymerase To verify directly that the shortened transcripts are generated by premature termination, we tried to perform an in vitro transcription assay. For this, we constructed plasmids, pSgrS-S(tac) and pSgrS-S-LS2(tac), in which the sgrS-S and the sgrS-S-LS2 are placed under the constitutive tac promoter (P tac ), respectively (Fig. 6A). Each plasmid was introduced into ΔsgrS Δhfq cells. Cells were grown in LB medium, and total RNAs were prepared and analyzed by Northern blotting. As expected, the full-length SgrS-S was expressed in cells carrying pSgrS-S (tac), whereas heterogeneous shorter transcripts along with the full-length SgrS-S-LS2 were generated in cells carrying pSgrS-S-LS2(tac) (Fig. 6B). We purified Flag-tagged RNA polymerase (RNAP-Flag) from the strain carrying the rpoC-Flag allele (Fig. 6C). Then, we carried out an in vitro transcription experiment using RNAP-Flag, pSgrS-S(tac), or pSgrS-S-LS2(tac), and four nucleoside triphosphates (NTPs). If our hypothesis is correct, pSgrS-S-LS2(tac) should produce prematurely terminated transcripts. After the transcription reaction, transcripts were subjected to Northern blotting using the SgrS-S probe (Fig. 6D). When pSgrS-S (tac) was used as a DNA template, a transcript corresponding to the full-length SgrS-S generated in vivo was detected as a FIGURE 4. Analysis of the 3 ′ ends of transcripts derived from sgrS-S and sgrS-S-LS2. TM542 (ΔsgrS) cells harboring pSgrS-S (A) or pSgrS-S-LS2 (C) were grown in LB medium. At A 600 = 0.6, 0.1% αMG was added and incubation was continued for 10 min, and then 0.02% arabinose was added and incubation was continued for 5 min. Total RNAs were prepared and duplicate RNA samples (10 µg) were resolved side-by-side on a 12% polyacrylamide gel electrophoresis in the presence of 8 M urea. One side of the gel was subjected to Northern blotting (A,C). The region corresponding to transcripts was cut out from the other side of the gel (A,C). A single gel piece containing the SgrS-S band was used for transcripts from sgrS-S (A). The gel was divided into three gel pieces (upper, middle, and lower) for transcripts from sgrS-S-LS2 (C). The upper gel piece corresponds to the full-length transcript, whereas the middle and lower gel pieces correspond to heterogeneous shorter transcripts. RNAs were purified from the gel pieces and subjected to 3 ′ -RACE analysis. DNA sequences corresponding to the 3 ′ region of transcripts were determined by using randomly picked-up plasmid clones containing amplified cDNAs. The sequences and number of clones analyzed are shown (B,D). Others represent the clones in which the 3 ′ ends of RNAs were mapped in the region prior to the terminator T residue stretch (D). major RNA band (Fig. 6D, lane 1). Importantly, the heterogeneous shorter transcripts that are similar to those generated in vivo were produced when pSgrS-S-LS2(tac) was used as a template (Fig. 6D, lane 2). Thus, generation of the heterogeneous transcripts with shortened poly(U) tails is ascribed at least partly to termination events occurring earlier in the run of T residues.

Effect of stabilization of the ryhB terminator hairpin on expression of RyhB and sodB mRNA
We also examined the effect of stabilization of the terminator hairpin of the ryhB on expression of RyhB. To do this, we constructed plasmid pRyhB-LS1 carrying the ryhB-LS1 by inserting four GC base pairs into the hairpin stem of the ryhB on plasmid pRyhB (Fig. 7A). We also constructed plasmid pRyhB-LS2 carrying the ryhB-LS2 by inserting seven GC base pairs into the hairpin stem of the ryhB (Fig. 7A). The predicted RNA hairpin structures and the ΔG values are shown in Figure 7B. Each plasmid was introduced into TM635 (ΔryhB hfq + ) and TM820 (ΔryhBΔhfq) cells. Cells were grown in LB medium to exponential phase and exposed to 2,2 ′ -dipyridyl to deplete Fe 2+ , and then arabinose was added to induce the transcription of ryhB. Expression of RyhB and its variants was analyzed by Northern blotting. As shown in Figure 7C, shorter transcripts, along with a decreasing amount of the full-length transcripts, were generated from the ryhB-LS1 and the ryhB-LS2 but not from the wild-type ryhB genes. The generation of shorter transcripts was more significant in the case of ryhB-LS2 compared to ryhB-LS1. Consistent with the increase in shorter transcripts, the amount of the full-length transcripts decreases more significantly in the ryhB-LS2. The abundance of the full-length RyhB, RyhB-LS1, and RyhB-LS2 in the hfq + cells was significantly higher than that in the Δhfq cells, reflecting Hfq binding of the full-length RyhB variants. On the other hand, the hfq backgrounds did not affect the level of the shorter transcripts, suggesting that the poly(U) tails of these RyhB variants are shortened. Then, we examined the expression of sodB mRNA in the hfq + cells harboring the vector plasmid pAraX, RyhB, RyhB-LS1, or RyhB-LS2 under the Fe 2+ depletion. The sodB mRNA is well expressed in cells harboring pAraX (Fig. 7D, lane 1). A marked reduction of the sodB mRNA was observed in cells harboring pRyhB (Fig. 7D, lane 2). The reduction of sodB mRNA expression was moderate in cells harboring pRyhB-LS1 or pRyhB-LS2 (Fig. 7D,  lanes 3,4), indicating that the down-regulation of sodB mRNA was partially suppressed in these cells. The suppression of the sodB mRNA down-regulation was stronger in cells harboring pRyhB-LS2 in which the reduction of levels of the full-length active RyhB variant was more significant. Thus, we conclude that the stabilization of the terminator hairpin of the ryhB also leads to generation of the shortened transcript with concomitant reduction of the active fulllength transcript.

DISCUSSION
The poly(U) sequence at the 3 ′ end of sRNAs is a central element of the Hfq-binding module of sRNAs (Otaka et al. 2011;Ishikawa et al. 2012). Hfq hexamer binds to the poly(U) tail through its proximal face (Sauer and Weichenrieder 2011;Schu et al. 2015). We have previously found that the poly(U) tail must be seven residues or longer to achieve a stable binding to Hfq that is required for the regulatory function of sRNAs (Otaka et al. 2011). Therefore, terminators of sRNA genes are expected to fulfill the structural features that ensure the generation of functional sRNAs possessing a long poly(U) tail. In fact, the polythymidine stretch of Rho-independent terminators of sRNA genes are longer than seven (Otaka et al. 2011;Ishikawa et al. 2012), whereas those of many other genes vary from four to eight or more (d'Aubenton Carafa et al. 1990). A recent global study on sRNA-target interactions has also revealed that the poly(U) tails of Rho-independent terminators are longer than those , and TM896 (ΔsgrS Δrnr) cells harboring pSgrS-S-LS2 were grown in LB medium containing 0.2% arabinose. At A 600 = 0.6, total RNAs were prepared and 10 µg of RNA samples was subjected to Northern blotting using the SgrS-S probe. (B,C) Stability of heterogeneous transcripts in the presence (C) and absence (B) of stress. TM772 (ΔsgrS Δhfq) cells harboring pSgrS-S-LS2 were grown in LB medium. For the preparation of samples in the absence of stress, 0.2% arabinose was added at A 600 = 0.6 and incubation was continued for 10 min, and then rifampicin (250 µg/mL) was added. Total RNAs were prepared at the indicated time after the addition of rifampicin. For the preparation of samples in the presence of stress, cells were exposed to 0.01% αMG for 10 min prior to the addition of arabinose. The RNA samples (20 µg) were subjected to Northern blotting using the SgrS-S. Arrowheads represent the full-length transcripts. of mRNAs (Melamed et al. 2016). The presence of a long polythymidine stretch in the terminator is essential but not sufficient for generation of the functional sRNAs. To generate the functional sRNAs, the consecutive terminator A residues on the template strand must be transcribed up to the seventh position before termination occurs within the T residue stretch. Thus, the modulation of the termination position at the Rho-independent terminators of sRNA genes is critical for the biogenesis of sRNAs. We assumed that thermodynamic stability of the terminator RNA hairpin affects the termination position within the polythymidine stretch and therefore acts as a key element for generation of functional sRNAs.
In the present work, we studied the role of the terminator hairpin in the biogenesis of sRNAs, focusing on how the hairpin stability affects the termination position at Rho-independent terminators. We constructed variant sgrS and ryhB genes in which the inverted repeat sequences of terminators are extended to stabilize the terminator hairpin. The stabilization of a hairpin at a given terminator is expected to enhance the termination efficiency because the strength of the terminator hairpin, along with the length of the T residue stretch, is an important element to determine the efficiency of termination at Rho-independent terminators, although the DNA sequences within and near the hairpin also affect the termination efficiency (Lynn et al. 1988;Cheng et al. 1991;Reynolds et al. 1992;Ray-Soni et al. 2016). Indeed, we showed that the extension of the terminator stem markedly enhances termination at the sgrS terminator (Fig. 1). The central finding in the present study is that the extension of the stem leads to generation of significant amounts of heterogeneous transcripts in which the poly(U) tail is shortened (Figs. 1-4, 6, 7). The transcripts with shortened poly(U) tails lose the ability to bind to Hfq and therefore are no longer able to repress target mRNAs (Figs. 3, 7). Importantly, we found that heterogeneous transcripts can be generated in an in vitro transcription system with purified RNA polymerase (Fig. 6), confirming that premature transcription termination within and/or before the terminator polythymidine stretch is at least partly responsible for the generation of shortened transcripts. Thus, the stability of the terminator RNA hairpins of sgrS and ryhB genes is optimized (not too strong and not too weak) to ensure an effective termination at the seventh or longer position of the consecutive T stretch. The moderate thermodynamic stability of the terminator hairpin is associated with most well-characterized sRNAs (Table 2), whereas the hairpin stability fluctuates more widely in many other terminators (d'Aubenton Carafa et al. 1990). The present study, along with previous studies (Otaka et al. 2011;Ishikawa et al. 2012;Morita et al. 2015), led us to conclude that the Rho-independent terminators of sRNA genes possess two common features to generate functional sRNAs: a long T residue stretch (longer than seven) and a moderate strength of hairpin structure that allows the termination at the seventh or longer position of the consecutive T stretch.
Concerning the roles of the terminator hairpin and the T residue stretch in transcription termination at Rho-independent terminators, it was originally proposed that the formation of the RNA hairpin induces pausing of the elongation FIGURE 6. Analysis of transcription of sgrS-S and sgrS-S-LS2 directed by the tac promoter. (A) DNA sequences of the chimeric P tac -sgrS-S and P tac -sgrS-S-LS2 genes. The sequence corresponding to sgrS-S is shown as regular letters, whereas the tac promoter sequence is shown as italic letters. The sequences of the −10 and −35 regions of the tac promoter are boxed. The inverted repeat sequences of the sgrS-S terminator are indicated by horizontal arrows. The inserted sequences to stabilize the terminator stem are shown as bold letters. (B) In vivo expression of the P tac -sgrS-S and P tac -sgrS-S-LS2. TM772 (ΔsgrS Δhfq) cells harboring indicated plasmids were grown in LB medium. Total RNAs were prepared at A 600 = 0.4, and 20 µg of RNA samples was subjected to Northern blotting using the SgrS-S probe. Arrowheads represent the full-length transcripts. (C) SDS-PAGE analysis of the purified RNAP-Flag. The affinity-purified PNAP-Flag sample (5 µL) was analyzed by SDS-PAGE and Coomassie staining as described in Materials and Methods. The bands corresponding to RNAP subunits are indicated on the left. Protein size markers are shown on the right. (D) In vitro transcription of the P tac -sgrS-S and P tac -sgrS-S-LS2. The transcription reaction was performed as described in Materials and Methods. The samples were subjected to Northern blotting using the SgrS-S probe. Arrowheads represent the full-length transcripts.
complex, and the low stability of rU-dA pairs allows the release of the nascent transcript and RNA polymerase from the template DNA (Rosenberg and Court 1979;Platt 1986;Yager and von Hippel 1991). The later study suggests that the Utract is principally responsible for the pause prior to hairpin formation (Gusarov and Nudler 1999). Several models have been proposed about the detailed mechanism by which the terminator RNA hairpin, along with the T residue stretch, leads to transcription termination at Rho-independent terminators (Santangelo and Roberts 2004;Larson et al. 2008;Peters et al. 2011). These in vitro studies, which were performed using various terminators including those possessing a discontinuous T stretch disrupted with other nucleotides, suggest that different mechanisms operate depending on terminators. The sRNA genes and their variants, possessing a long continuous T residue stretch with different stability/ structure of the hairpin, would be useful to understand more deeply the roles of the terminator hairpin and the T residues stretch in transcription termination.
Although a number of analyses have been performed regarding the mechanism of transcription termination, little has been known about how the hairpin stability affects the termination position at Rho-independent terminators. We have addressed this question experimentally in this study and demonstrated clearly that the strength of the terminator hairpin controls the termination position within the polythymidine stretch. In other words, the terminator hairpin determines the length of the poly(U) tail of the transcripts depending on its thermodynamic stability. Thus, the control of the termination position by the terminator RNA hairpin is a critical event for the biogenesis of functional sRNAs. It remains to be studied whether the function of many other transcripts including mRNAs is affected by the length of the poly(U) tail. The stronger RNA hairpin would be more beneficial for efficient transcription termination and, therefore, for efficient generation of termination products such as mRNAs in many genes. It is apparent that this is not the case for transcription termination at the Rho-independent terminators of sRNA genes. The thermodynamic stability of the terminator hairpin of sRNA genes is not too strong to ensure the generation of transcripts possessing a long poly(U) tail by preventing premature termination and also not too weak to prevent too much readthrough. Thus, the terminator of sRNA genes is optimized to generate efficiently functional sRNAs. How the stability of the terminator hairpin controls the termination position is certainly an important question to be addressed for understanding the molecular mechanism of transcription termination at Rhoindependent terminators. In this connection, it is interesting to note that shortened transcripts were generated at a Rho-independent terminator in an in vitro transcription system with a low concentration of NTPs (McDowell et al. 1994).
Another interesting question regarding the mechanism of transcription termination is how the stress enhances the termination. We demonstrated previously that stress conditions lead to the enhancement of termination at not only sRNA genes but also a gene encoding an mRNA (Morita et al. 2015). It is interesting that low NTP concentrations are reported to enhance transcription termination at Rho-independent terminators in vitro (Reynolds et al. 1992). In addition, it is reported that shortened transcripts were generated at a Rho-independent terminator in vitro when a low concentration of NTPs was used (McDowell et al. 1994). Therefore, one plausible effect of stress is to reduce the level of NTPs in cells.
The simplest type of Hfq binding module consists of a long poly(U) tail, terminator hairpin, and an internal U-rich sequence just before the hairpin (Ishikawa et al. 2012). The poly(U) tail binds to the proximal face of the Hfq hexamer,  (Zuker 2003). (C) Generation of heterogeneous transcripts and effect of hfq backgrounds. TM635 (ΔryhB) and TM820 (ΔryhB Δhfq) cells harboring the indicated plasmid were grown in LB medium. At A 600 = 0.6, 250 µM of 2, 2 ′ -dipyridyl was added and incubation was continued for 10 min. Then, 0.02% arabinose was added and incubation was continued for 5 min. Total RNAs were prepared, and 1.6 or 0.25 µg of RNA samples was subjected to Northern blotting using the RyhB probe and tmRNA probe, respectively. Arrowheads represent the full-length transcripts. (D) Effect of extension of the terminator stem on RyhB function. Two micrograms of RNA samples described in Figure 7C was subjected to Northern blotting using the sodB probe.
whereas the internal U-rich sequence preferentially binds to the rim of the Hfq hexamer (Sauer and Weichenrieder 2011;Sauer et al. 2012;Schu et al. 2015). Thus, the roles of the poly(U) tail and internal U-rich sequence in Hfq binding are clear. On the other hand, it remains to be studied whether and how the terminator RNA hairpin is involved in Hfq binding. We showed previously that SgrS variants retain the ability to bind to Hfq and to repress the target ptsG mRNA even when the sequence corresponding to the stem-loop structure was dramatically changed as far as the hairpin structure was maintained (Otaka et al. 2011). It is interesting to note that a hairpin-less RNA possessing a base-pairing region and a poly(U) tail is able to base pair with a beacon probe RNA in the presence of Hfq in vitro (Panja et al. 2013). In addition, a structural analysis demonstrated that the hairpin structure seems to be not involved in interactions with Hfq (Dimastrogiovanni et al. 2014). These results suggest that the terminator RNA hairpin plays no role in Hfq binding itself. Further studies are needed to know whether the hairpin structure of the terminator hairpin plays any roles in Hfq binding.

Bacterial strains and plasmids
The E. coli K12 strains and plasmids used in this study are listed in Table 1. IT1568 (W3110mlc − ) was used as a parent strain. To con-struct TM894, TM895, and TM896, the alleles Δpnp::cat, Δrnb::cat and Δrnr::cat of ECK3152, ECK1281 and ECK4175, respectively, were moved to TM542 (Kawamoto et al. 2005) by P1 transduction. ECK3152, ECK1281, and ECK4175 were donated from H Mori, NAIST (pers. comm.). The rpoC-Flag-cat allele was constructed according to the modified Datsenko-Wanner protocol using pSU313 harboring the Flag-cat sequence (Uzzau et al. 2001). To construct TM905 in which the chromosomal rpoC gene was replaced with the rpoC-Flag encoding carboxy terminally Flag-tagged β ′ , the rpoC-Flag-cat allele was moved to IT1568 by P1 transduction. TM908 was constructed by removing the cat gene flanked by two FRT sequences from TM905.
The DNA primers used are listed in Table 3. Plasmid pSgrS-S-LS1-rplLT was constructed as follows: pSgrS-S-rplLT was used to amplify the DNA fragment containing the sgrS-S-LS1 sequence and rplLT sequence with primers 1585 and 1544. The amplified DNA fragment was digested with XbaI and HindIII, and cloned into pAraX. Plasmids pSgrS-S, pSgrS-S-LS1, pSgrS-S-LS2, pSgrS-S-LS3, and pSgrS-LS4 were constructed as follows: pSgrS (Otaka et al. 2011) was used to amplify the DNA fragment containing the sgrS-S, sgrS-S-LS1, sgrS-S-LS2, sgrS-S-LS3, or sgrS-S-LS4 sequence with primers 1127 and 1839, 1840, 1923 or 1924, respectively. The amplified DNA fragments were digested with XbaI and HindIII, and cloned into pAraX. Plasmids pSgrS-S (tac) and pSgrS-S-LS2 (tac) carrying the P tac -sgrS-S and P tac -sgrS-S-LS2, respectively, were constructed as follows: pSgrS-S or pSgrS-S-LS2 was used to amplify the DNA fragment containing the P tac -sgrS-S or P tac -sgrS-S-LS2 sequence with primers 1919 and 1839 or 1841, respectively. The amplified DNA fragments were digested with XbaI The DNA sequences of Rho-independent terminators of representative sRNA genes were taken from http://www.genome.jp/kegg/. The GC-rich inverted repeat sequences are underlined. The sequences corresponding to the polythymidine stretch are italicized. The thermodynamic stabilities (ΔG, kcal/mol) of the terminator RNA hairpins without the poly(U) sequence were determined according to the Mfold program (Zuker 2003).