Toward a Comprehensive Analysis of Posttranscriptional Regulatory Networks: a New Tool for the Identification of Small RNA Regulators of Specific mRNAs

ABSTRACT A number of computational or experimental tools have been developed to identify targets of small RNA (sRNA) regulation. Here, we modified one of these methods, based on in vivo proximity ligation of sRNAs bound to their targets, referred to as rGRIL-seq, that can be used to capture sRNA regulators of a gene of interest. Intracellular expression of bacteriophage T4 RNA ligase leads to a covalent linking of sRNAs base-paired with mRNAs, and the chimeras are captured using oligonucleotides complementary to the mRNA, followed by sequencing. This allows the identification of known as well as novel sRNAs. We applied rGRIL-seq toward finding sRNA regulators of expression of the stress response sigma factor RpoS in Escherichia coli, Pseudomonas aeruginosa, and Vibrio cholerae. In E. coli, we confirmed the regulatory role of known sRNAs and discovered a new negative regulator, asYbiE. When applied to P. aeruginosa and V. cholerae, we identified two novel sRNAs (s03661 and s0223) in P. aeruginosa and two known sRNAs (TfoR and Vcr043) in V. cholerae as direct regulators of rpoS. The use of rGRIL-seq for defining multiple posttranscriptional regulatory inputs into individual mRNAs represents a step toward a more comprehensive understanding of the workings of bacterial regulatory networks.

IMPORTANCE With the recognition of the importance of posttranscriptional regulation mediated by bacterial small RNAs (sRNAs), their contribution to global gene expression regulatory networks needs to be addressed in a truly comprehensive manner. While a single sRNA targets multiple RNAs, an mRNA can be regulated by multiple sRNAs that can be either transcribed individually or derived by processing of mRNAs. In this paper, we developed a tool (referred to as rGRIL-seq) to identify sRNAs that regulate mRNAs regardless of their origin. We demonstrated the utility of this approach by identifying positive and negative sRNA regulators of the rpoS mRNA in three bacterial species. We not only described known sRNAs of E. coli or P. aeruginosa that control rpoS but also identified several new rpoS regulators in V. cholerae. Therefore, rGRIL-seq can be used to identify species-specific sRNAs targeting a conserved mRNA, and they likely play an important role in bacterial adaptation to specific environmental niches.
KEYWORDS discovery of multiple small RNA regulators, rpoS, E. coli, P. aeruginosa, V. cholerae, rGRIL-seq, proximity ligation B acterial small regulatory RNAs (sRNAs) are typically ;40 to 400 nt in length and act as gene-expression regulators by various mechanisms (1,2). Extensive studies over the past 3 decades have shown that they modulate a wide range of physiological responses in bacteria. The discovery of the large number of bacterial sRNAs led to an appreciation of the role of posttranscriptional regulation in controlling prokaryotic gene expression, rivaling transcription factors in their abundance and in the number and functional diversity of their regulatory targets.

RESULTS
The rGRIL-seq method. In the previous study, we developed a method for the identification of direct targets of sRNAs in living cells (GRIL-seq), where an sRNA is ectopically expressed in bacteria coexpressing T4 RNA ligase, and the enrichment of the chimeric RNAs (sRNA-target RNA ligation products) was accomplished using a bead-immobilized single oligonucleotide complementary to the specific sRNA (13). In order to identify sRNA regulators targeting a specific mRNA, we altered the GRIL-seq protocol at the enrichment step by reversing the chimera capture strategy; we refer to this modified method as reverse GRIL-seq (rGRIL-seq) (Fig. 1). For the rGRIL-seq approach, we designed multiple oligonucleotides complementary to the different regions of an mRNA sequence and utilize these for bead capture and enrichment of mRNA fragments that are covalently linked to their sRNA regulators by the action of T4 RNA ligase (Fig. 1A). We reasoned that in the absence of any information on ribosome occupancy or the degradation patterns of a specific mRNA, spacing of the segments complementary to the capture oligonucleotides should be random, with a preference given to the 59 region of mRNAs including the 59 UTR (Fig. 1B).
To test whether rGRIL-seq is a reliable method for identifying sRNA regulators of specific mRNAs, we chose to study the regulation of the rpoS transcript in three bacterial species: E. coli, Vibrio cholerae, and Pseudomonas aeruginosa. The stress-responsive sigma factor RpoS has a similar regulatory function in all Proteobacteria. In E. coli, several sRNAs controlling the expression of rpoS translation and stability have been identified and extensively studied; this allowed for validation of the rGRIL-seq with known as well as previously uncharacterized sRNAs. The E. coli rpoS gene is found adjacent to the nlpD gene, encoding an outer membrane lipoprotein (Fig. 2). The coding sequence of Identification of sRNA Regulators of mRNAs ® NlpD contains the major rpoS promoter located 567 bp from the start codon for RpoS. This overall genetic organization is conserved among many Gram-negative microorganisms, including P. aeruginosa and V. cholerae, where the transcription start site and the unusually long 59 UTR are also located within the nlpD gene. At the amino acid level, RpoS is conserved (67 to 74% [see Fig. S1A and B in the supplemental material]) as expected from proteins which carry out the same function. However, the nucleotide sequences of the 59 UTR (55 to 62%), the open reading frame (ORF, 62 to 68%), and 39 UTR (60 to 63%) are less conserved (Fig. S1C to H). The low conservation of the regulatory sequences, particularly in the 59 UTR that serves as potential target sequence for base-paring regulatory sRNAs, suggests that they may be recognized by different sRNAs in each species. Not surprisingly, homologs of the known sRNA regulators of E. coli rpoS cannot be identified in Vibrio or Pseudomonas species.
We carried out the rGRIL-seq procedure as described in detail in Materials and Methods. Briefly, for each rpoS transcript, we designed 4 oligonucleotides complementary to sequences in the rpoS mRNA to be used for the enrichment of the rpoS-sRNA chimeras, with their approximate location indicated in Fig. 1B and Fig. 2. We isolated total RNAs from the cells ectopically expressing T4 RNA ligase in actively dividing cells (optical density at 600 nm [OD 600 ] of ;0.4), referred to as exponential-phase cultures (Ex.) and high-density cells (OD 600 of ;3.5), referred to as stationary-phase cultures (St.). The rpoS chimeric RNAs were enriched from the total RNAs using the four oligonucleotides and were captured on magnetic beads. The cDNAs of these chimeras were sequenced and analyzed by mapping them to the genomes of each bacterial species.
sRNAs regulating E. coli rpoS. We first carried out the rGRIL-seq in E. coli by extracting RNA from three independent cultures, each grown to exponential (Ex.) and early stationary (St.) phases. We obtained an average of 9,264 and 14,052 rpoS-chimeric reads for logarithmic and stationary phases, respectively (Table S1A).
Mapping the chimeric reads on the genome of E. coli MG1655 allowed us to identify several transcripts preferentially ligated to the rpoS mRNA (Fig. 3A). The enrichment for several RNAs was detected in both growth phases, while certain sRNA transcripts ligated to the rpoS mRNA were isolated in either the exponential or stationary phase. Table 1 shows the list of top 5 sRNAs ranked by number of reads from sequencing of the rpoS-containing chimeras. We identified a total of 6 different rpoS chimeric RNAs in the top 5 chimeric reads in each growth phase (Table 1A and Fig. 4A). These include the two known sRNA regulators of rpoS (DsrA and ArcZ [19,21]), while the other two well-characterized sRNAs, RprA and CyaR (18,19,23,24), were also identified but in lesser abundance and ranked 12th and 14th in the stationary phase, respectively (Table S1B). Previously, Moon and Gottesman identified additional regulators of rpoS following overexpression of a set of sRNAs and determined their effect on an rpoS Ec -lacZ translational fusion (24). Among these sRNAs, MgrR and GadY were also found in FIG 2 The rpoS locus in E. coli, P. aeruginosa, and V. cholerae. The transcription start site of the rpoS (red) gene in the nlpD gene with arrows was determined by EcoCyc (for E. coli MG1655) and our unpublished RNA-seq data (for P. aeruginosa PAO1 and V. cholerae C6706). The four lines below each rpoS gene show the approximate locations of the binding sites for the multiple complementary oligonucleotides used for the enrichment of chimeras.
the rpoS chimeras, demonstrating that their effect on the rpoS Ec -lacZ expression was the consequence of a direct base-pairing interaction between the sRNA and the rpoS 59 UTR. Interestingly, they also showed that the strongest repressive effect on the rpoS Ec -lacZ fusion was exerted by overexpression of OxyS; yet, the rpoS chimera with this sRNA was detected in only a very small number of reads in rGRIL-seq (7 and 3 chimeras of OxyS in Ex. and St., respectively), presumably supporting the indirect regulatory mechanism of this sRNA (25).
We also found two transcripts (sAspA and asYbiE) ligated to E. coli rpoS mRNA in the rGRIL-seq analysis ( Fig. 3A and Table 1A). Both rpoS chimeras were detected in exponential as well as stationary phases at similar levels (Fig. 3A). The sAspA corresponds to the 39 region of the aspA gene encoding aspartate ammonia-lyase (Fig. 4A), which Identification of sRNA Regulators of mRNAs ® carries out the reversible conversion of L-aspartate to fumarate and ammonia under catabolite repression. The aspA transcript (;1,600 nt) terminates at an inverted repeat immediately downstream of the aspA gene (26,27), as shown in Fig. S2A. We speculate that the sAspA is an sRNA derived from the 39 end of the aspA mRNA; such regulatory transcripts have been previously identified in several other bacteria (28)(29)(30)(31)(32), and they are orginated either by transcription from an internal promoter of the gene (Type I) or by processing of the parental mRNA (Type II) (33). Our analysis of this gene identified a putative promoter sequence upstream of the stop codon of the aspA gene (Fig. S2A); this would make the sAspA a Type I 39-derived sRNA. Interestingly, unlike other sRNA-rpoS chimeric transcripts, mapping of the rpoS sequences in the rpoS-sAspA ligation products on the rpoS gene showed an enrichment of the chimeras at the 39 end, near the terminator sequence of rpoS mRNA ( Fig. S2B and Fig. S4). This implies that the  Identification of sRNA Regulators of mRNAs ® sAspA may interact with the 39 region of rpoS mRNA, allowing us to predict the likely base-pairing between them using RNA-RNA interaction algorithm IntaRNA (34). The predicted base-pairing includes the sequence following the stop codon of rpoS and part of the 39 UTR (hybridization energy, 211.14 kcal/mol, Fig. S2C). Another rpoS ligation product was enriched at the adjacent region of the ybiE gene that has been recently shown to encode a small hydrophobic peptide (;2 kDa) in an intergenic region between ybiC and ybiJ (35) (Fig. 4A). However, the analysis of the sequences in these rpoS chimeric transcripts showed that majority of these are antisense transcripts of YbiE, created by ligation of the 39 OH of rpoS to 59 of antisense transcripts of YbiE; we refer to this sRNA as asYbiE.
We monitored the expression of sAspA and asYbiE in LB (Luria-Bertani) rich medium by Northern blotting of total RNAs isolated from E. coli cultures at different growth phases. The levels of sAspA gradually increase at the stationary phase as ;110 nt in size, with a concomitant increase of a smaller, ;80-nt species. The asYbiE transcript is ;350 nt in size in mid-log and stationary phase (Fig. 4B). The small size of asYbiE (;100 nt) is detected only at the stationary phase with a weak signal, though overall it is hardly detectable under the LB medium condition without using a multicopy plasmid harboring the intergenic region between ybiC and ybiJ gene ( Fig. 4B; see pIG_ybiCJ).
In order to examine the effect of the sRNAs on RpoS regulation in E. coli, we employed a plasmid which is able to express both an sRNA gene and the rpoS Ec ::lacZ translational reporter gene fusion with different inducible promoters (P BAD for sRNA and P tac for rpoS Ec ::lacZ gene, Fig. 4C). We first cloned the sequences encompassing the 59 UTR and the codons for the first 80 amino acids of the E. coli rpoS gene into the plasmid pKH24::lacZ to create plasmid pKH24-rpoS Ec ::lacZ. We then cloned each sRNA gene into this vector, to enable induction of its expression by the addition of arabinose, and introduced both of the constructs into E. coli DlacIYDaraCD. We monitored the activity of the fusion when each sRNA (asYbiE or sAspA) was overexpressed. The previously identified sRNAs, DsrA and MgrR (24), were used as the positive and negative regulators, respectively, and as expected, they showed the predicted increase and decrease of b-galactosidase activity, respectively. The b-galactosidase activity was significantly reduced in the strain overexpressing asYbiE RNA, while in a strain over expressing the sAspA RNA, no effect was observed compared to the negative control (Fig. 4C). We also monitored the level of endogenous RpoS, using E. coli RpoS antibody, when asYbiE or sAspA was overexpressed. Consistent with the previous results obtained from the b-galactosidase activity assay, the Western blot analysis showed that overexpression of asYbiE led to a reduction of the endogenous RpoS protein levels ( Fig. 4D). However, overexpression of sAspA had no effect on RpoS protein levels, though a possible RNA-RNA interaction between the 39 UTR region of rpoS mRNA and sAspA can be identified (Fig. S2C). We predicted the likely base-pairing interactions between the asYbiE and the rpoS 59 UTR sequence using IntaRNA or the RNA secondary structure of the chimeric RNA with the algorithm of the CLC Genomics Workbench package. The most likely interaction of asYbiE with its target is at the Shine-Dalgarno sequence of the rpoS mRNA (Fig. 4E), suggesting that the reduction of b-galactosidase activity of rpoS::lacZ and the RpoS protein is caused by interfering with initiation of translation by asYbiE, and therefore, it directly represses rpoS expression. In contrast, we cannot at this time assign a regulatory role for sAspA although potential base-pairing at the termination sequence of the coding region of rpoS could be computationally determined FIG 4 Legend (Continued) pKH24, which contains both the IPTG-inducible rpoS Ec ::lacZ reporter gene and the L-arabinose-inducible sRNA gene. The plasmid (Vec.) lacking sRNA was used as a negative control. Data are shown as mean 6 SD for three biological replicates. Statistical comparisons were performed using Student's t test. **, P # 0.01; ***, P # 0.001. (D) Western blot analysis of endogenous RpoS Ec levels detected by anti-RpoS Ec antibodies, using the total proteins isolated from the same cultures used for the b-galactosidase assay. RpoA Ec was used as a loading control. (E) Base-pairing between asYbiE and rpoS mRNA. The Shine-Dalgarno sequence of rpoS mRNA is boxed, and the numbers (224 and 27) over the sequence of rpoS mRNA are the nucleotide distances relative to the translational start site. (Fig. S2C). Therefore, sAspA could be regulating other mRNAs, and the interaction detected by rGRIL-seq could represent a regulatory mechanism, where the rpoS mRNA is sequestering ("sponging") the sAspA, reversing the effect of this sRNA on its other targets.
sRNAs regulating P. aeruginosa rpoS. When applying the rGRIL-seq method toward identifying sRNA regulators of P. aeruginosa rpoS, the number of chimeric reads from enriched samples was substantially lower than that with E. coli or V. cholerae. We recovered 296 sequencing reads in exponential phase and 3,467 reads in stationary phase, representing ;31-fold less in exponential phase and ;4-fold less in stationary phase compared to E. coli and ;28-fold less in exponential phase and ;3-fold less in stationary phase compared to V. cholerae (Table S1A). This poor recovery of rpoS-chimeric reads was accompanied by a substantial enrichment for chimeras containing rRNAs (Fig. 3B). They are likely created by a nonspecific ligation between random fragments of rRNAs and rpoS mRNA. The chimeras, generated by preferential base-pairing at regulatory sites, allow the sequencing reads to be visualized as peaks when mapped on the genome in the graphic display (Fig. 4A, Fig. 5A, and Fig. 6A and Fig. S3, S5, and S7). In contrast, nonspecific ligation products to rRNA are randomly distributed over the length of the gene rather than being enriched at a specific location (example shown in Fig. S4).
In our analysis of the P. aeruginosa chimeric reads, we identified a total of 7 different transcripts within the top 5 rpoS-chimeric RNAs (Table 1B and Fig. 5A). In spite of poor enrichment for the chimeras using the rpoS-complementary capture oligonucleotides, among the mapped sequences we found the ReaL sRNA (ranked at top no. 3 in Ex. and no. 2 in St., Table 1B), which has been previously reported as a negative regulator of rpoS expression in P. aeruginosa (36). This suggests that the our rGRIL-seq data may have other small RNA regulators for the rpoS mRNA but with lower numbers of chimeric reads. In our analysis of the 7 sRNA-rpoS chimeras, the most abundant rpoS-chimeric RNAs, in both exponential and stationary phase, contained ErsA, which has been previously shown to negatively regulate AlgC and OrpD and positively regulate AmrZ (37)(38)(39). Four sRNA-rpoS chimeric reads (s3661, s0223, sRmf, and sAdhC) were enriched near the stop codon or the 39 UTR of mRNAs (PA3661, PA0223, Rmf, and AdhC, respectively; Fig. 5A), suggesting that these sRNAs represent the class of sRNAs derived by processing of mRNAs or internal starts within the coding sequences analogous to the sAspA RNA in E. coli. However, unlike the sAspA-rpoS sequences, none of these sRNA-rpoS chimeras showed enrichment of the reads at the 39 region of the rpoS gene in P. aeruginosa (Fig. S5). In order to determine whether the candidate sRNAs can be identified in bacterial cells, we performed Northern blotting with total RNA isolated from P. aeruginosa PAO1 at different growth stages (Fig. 5B), and 6 of 7 total RNA transcripts were detected. With the exception of sAdhC, we were able to detect 3 transcripts (s3661, s0223, and sRmf) when they were probed with the radiolabeled oligonucleotides designed to bind to the 39 region of the mRNAs (Fig. 5B and Fig. S6). The observation of the s3661, s0223, and sRmf small-size transcripts suggests that they are all derived from the 39 UTR and all contain the rho-independent terminators.
We next constructed strains that allowed us to monitor the regulatory effect of overexpression of the putative sRNAs on rpoS. Based on the Northern blot analysis and the sequence of the rho-independent terminator, we predicted the 59 and 39 ends of each RNA transcript and cloned the putative sRNA genes into an arabinose-inducible expression plasmid (pKH6). We transformed each sRNA expression construct into P. aeruginosa carrying either the rpoS Pa ::lacZ reporter or the gene for rpoS Pa ::6XHis codons, inserting them into a chromosomal site (f CTX), and induced their expression with arabinose (Fig. 5C). In the b-galactosidase assay, we found that the ReaL sRNA shows a strong reduction in enzymatic activity of the reporter, in agreement with the previous report (36). Additionally, overexpression of four small RNAs (sr0161, s0223, s3661, and ErsA) also showed a moderate reduction of b-galactosidase activity compared with the control strain carrying the empty plasmid vector, suggesting that they  Identification of sRNA Regulators of mRNAs ® might also function as negative regulators of RpoS. We further tested the effects of overexpression of these selected sRNAs on expression of the RpoS protein, using the C-terminal 6ÂHis-tagged RpoS with a Western blot analysis probing with an antibody to the tag (Fig. 5D). In all cases, a similar reduction in RpoS Pa ::6ÂHis expression was detected as seen in the b-galactosidase assay, further indicating that the sRNAs identified by rGRIL-seq are negative regulators of rpoS mRNA, leading to a reduction in the levels of the RpoS protein. Focusing on the three sRNAs (ReaL, s3661, and s0223) showing the most pronounced reduction of RpoS following their overexpression, we predicted their base-pairing with rpoS mRNA using the IntaRNA algorithm (34) and analysis of the secondary RNA structure of the sRNA-rpoS chimeric RNA (Fig. 5E). The predictions show that ReaL base pairs with the Shine-Dalgarno sequence of rpoS mRNA as shown in the previous report (36). However, s3661 and s0223 were predicted to base-pair mRNA at regions 2325 to 2319 for s3661 and 1138 to 1150 for s0223 from the translational start codon of rpoS, suggesting the possibility of the presence of other regulatory sites and different mechanisms to repress the rpoS translation by these sRNAs.
sRNAs regulating V. cholerae rpoS. To identify sRNA regulators of rpoS in V. cholerae C6706 by rGRIL-seq, we mapped all rpoS chimeric reads on the two chromosomes (referred to as Ch1 and Ch2) of V. cholerae N16961, since the genes have been well annotated and are closely related to the C6706 strain. Applying the rGRIL-seq procedure to V. cholerae, we recovered a similar number of chimeric reads as in E. coli (Table S1A): on average, 8,331 reads in late exponential phase and 10,132 reads in early stationary phase. Mapping all of the rpoS chimeric reads onto the genome, within the top 5, we identified a total of 8 different transcripts under our two experimental growth conditions ( Fig. 3C and Table 1C). Among them, 5 rpoS chimeric transcripts were created in the intergenic regions (TfoR, Vcr090, Vcr043, sVca0838, and RyhB Vc ), while three were created with the transcripts from start codon regions of mRNAs (rplU, Vc0573, and obgE; Fig. 6A). The inspection of sequences at the ligation junctions revealed that 5 (TfoR, Vcr090, rplU, Vcr043, and RyhB Vc ) of the 8 transcripts formed chimeric products with fragments near the 59 UTR of rpoS (Fig. S7), suggesting that they are directly regulating either the expression of rpoS mRNA stability or its translation. The other three chimeras (sVc0573, sVca0838, and obgE) were created by ligation to rpoS sequences near the rpoS stop codon and the 39 UTR of the mRNA.
We assessed the expression of the 5 sRNAs (TfoR, Vcr090, Vcr043, sVca0838, and RyhB Vc ), transcribed from intergenic regions by Northern blot analysis in rich medium (LB) over several growth phases (Fig. 6B). All sRNAs were detected; they varied in size from ;90 nt (Vcr090) to ;350 nt (sVca0838) during the growth in LB; some showed accumulation in the stationary phase (sVca0838 and RyhB Vc ) while others (TfoR, Vcr090, and Vcr043) were found at similar levels through all growth phases. TfoR is an ;100-nt sRNA which has been shown previously to target and positively regulate tfoX mRNA, encoding the key regulator of natural competence and type VI secretion in V. cholerae (40)(41)(42). We observed a major ;100-nt TfoR RNA in the Northern blot experiment, while two more minor TfoR transcripts (;130 nt and ;250 nt) were also detected. Vcr090 and Vcr043 were previously reported as ;90and ;160-nt sRNAs, respectively, in wild-type V. cholerae C6706 (43), and they were detected as well at similar sizes in the Northern blot assay. Based on the size of sVca0838 in the Northern blot and the presence of the rho-independent terminator sequence, this new sVca0838 sRNA is transcribed as a 346-nt RNA. We also detected RyhB Vc at the corresponding size of 214 nt; its function ranges from control of iron homeostasis to regulation of biofilm formation (44). FIG 6 Legend (Continued) In order to assess the effect of each sRNA on regulation of RpoS expression in V. cholerae, we employed the same plasmid vector (pKH24::lacZ) previously used in E. coli for measuring b-galactosidase activity. We amplified the sequences encompassing the 59 UTR and codons for the first 80 amino acids of the V. cholerae rpoS gene to construct the plasmid pKH24-rpoS Vc ::lacZ. After inserting the 5 sRNA genes individually into the vector, we introduced each plasmid into the V. cholerae C6706 recA-lacZ* mutant cell. In the analysis of b-galactosidase activity in strains with the lacZ reporter, the most pronounced effect was observed following overexpression of TfoR, which significantly reduced the expression of the rpoS-lacZ fusion protein (;6-fold). In contrast, the overexpression of three of the sRNAs (Vcr043, sVca0838, and RyhB Vc ) resulted in a modest induction of the rpoS Vc ::lacZ translation (;3-fold for Vcr043, ;1.4-fold for sVca0838, and ;1.6-fold for RyhB Vc ). In order to monitor the level of the RpoS protein in V. cholerae, we constructed the plasmid pKH24-rpoS Vc ::6H expressing RpoS Vc fused to the His tag at its C terminus and then inserted each of the 5 sRNA genes into this vector. Following induction with arabinose, TfoR overexpression resulted in a significant reduction of RpoS::6ÂHis as measured by Western immunoblotting. In contrast, Vcr043 overexpression led to an increase in RpoS::6ÂHis levels (Fig. 6D). Based on the mapping of the chimeric RNAs with the TfoR and Vcr043 on the rpoS transcript (Fig. S7), combined with predictions of likely interactions using IntaRNA, base-pairing models were generated and are shown in Fig. 6E. The most likely base-pairing site of TfoR on the rpoS transcript is near the initiation codon (217 to 118). This predicted interaction suggests that TfoR sRNA negatively regulates RpoS expression by base-pairing with the Shine-Dalgarno sequences and/or the start codon. The predicted base paring model between TfoR and rpoS Vc shows that the region on the sRNA involved in basepairing is located between nucleotides 179 and 1113. This region differs from that predicted for the interaction of TfoR with tfoX where the base-pairing region is at the 59 end of TfoR, located between nucleotides 11 and 161 (45). In the predicted basepairing model for Vcr043 and the rpoS mRNA, two possible base-pairing regions can be identified (Fig. 6E) and both are found at two different sites at the 59 UTR (2414 to 2402 and 2218 to 2199) on the rpoS mRNA. In summary, our rGRIL-seq data along with these results show that in V. cholerae, TfoR functions as a negative regulator of rpoS in addition to a positive regulator of tfoX, while Vcr043 functions as a positive regulator of rpoS expression.

DISCUSSION
In the work presented here, we describe a modification of the previously developed proximity ligation method (GRIL-seq) for studying sRNA interactions with their targets. The main difference in the current rGRIL-seq approach and previous GRIL-seq is in the enrichment step, where in this case chimeras are captured using oligonucleotides complementary to the mRNA target sequences, thus allowing the identification of sRNA regulators of specific mRNAs. Combined with GRIL-seq, rGRIL-seq allows for a more comprehensive probing of sRNA regulons, potentially leading to the identification of new sRNAs that could not be predicted using computational tools or through the analysis of the transcriptomes due to their low abundance. Currently, a list of sRNAs transcribed in intergenic regions can be readily derived by combining transcriptome sequencing (RNA-seq) with computational analyses. However, recent findings that sRNA regulators can be also generated by fragmentation of mRNAs or transcription initiation within coding sequences (3,33) make it difficult to analyze their roles in posttranscriptional regulatory networks. Therefore, rGRIL-seq represents another method where the identity of an sRNA is revealed by becoming a substrate for in vivo ligation to an mRNA target and does not depend on any prior knowledge of its origin. Additionally, rGRIL-seq represents a useful approach for studying the regulatory activities of multiple sRNAs in controlling the expression from a single mRNA target in response to different environmental conditions. Finally, rGRIL-seq has an added advantage over other experimental methods developed for detecting sRNA-mRNA interactions such as RIL-seq and Hi-GRIL-seq (9,38) because the complementary oligonucleotide enrichment step allows the detection of interactions between sRNA and their target mRNAs in low abundance.
Here, we applied rGRIL-seq to identify sRNA regulators of expression of the stationary-phase sigma factor RpoS in E. coli, P. aeruginosa, and V. cholerae. This approach allowed us to validate rGRIL-seq in an organism (E. coli) where posttranscriptional regulation of rpoS expression has been extensively studied. Additionally, we probed sRNA regulation of rpoS in two organisms (P. aeruginosa and V. cholerae) that, because of their distinct natural environments, have likely evolved unique stress survival strategies. The summary of the distinct regulatory mechanisms for the control of rpoS expression in the three organisms studied is shown schematically in Fig. 7.
Positive regulators of RpoS expression. Using rGRIL-seq in E. coli, we confirmed that the three previously characterized sRNA activators (ArcZ, DsrA, and RprA) interact with rpoS mRNA and create RpoS chimeric RNAs in a reaction catalyzed by RNA ligase. We applied this method to P. aeruginosa and V. cholerae and identified Vcr043 sRNA as an sRNA activator of V. cholerae rpoS, while finding no activators of P. aeruginosa rpoS under our conditions. We noted that the formation of RNA chimeras seems to require the cleavage of the 59 UTR of rpoS mRNA or the sRNA, or both. When the activator sRNAs interact with the internal region of the long 59 UTR, the cleavage of the 59 UTR appears to be a prerequisite to the formation of the chimeric RNAs by the coexpressed T4 RNA ligase. For example, in our analysis of the DsrA-rpoS chimeric transcripts, the major ligation reaction in the DsrA-rpoS chimeric RNAs occurred in the internal region (2107 nt relative to translation initiation site) of the previously characterized DsrA base paring site (2120 to 290 nt) (19), suggesting that base-pairing, concomitant cleavage, and ligation are important factors to create chimeric RNAs under in vivo proximity ligation conditions. Indeed, it has been reported that base-pairing of DsrA with the 59 UTR of rpoS leads to a concomitant cleavage of both the 59 UTR of rpoS and DsrA by RNase III (46,47). We have previously also observed this form of chimeric RNAs in RyhB Ec -shiA chimeric RNAs in E. coli (13). In the rGRIL-seq analysis in P. aeruginosa, we were unable to find activators of rpoS under the experimental condition used to grow the bacteria. It is not clear at present whether these sRNA activators are not found in this organism or we are unable to detect them due to the absence of the RNase-mediated cleavage event in the 59 UTR as shown in E. coli DsrA, which is necessary to create the chimeric RNAs between the activator sRNA and rpoS in P. aeruginosa. Additionally, further studies are needed to determine whether rGRIL-seq is capable of identifying the activators such as RydC that base-pair in the middle of the 59 UTR of cfa transcript, stabilizing it by inhibiting the cleavage from RNase E (48).
In the rGRIL-seq analysis of sRNA-rpoS interactions in V. cholerae, we identified Vcr043 as a strong activating sRNA of rpoS expression ( Fig. 6C and D). Our prediction of a Vcr043-rpoS duplex based on the two forms of chimeric RNAs suggested two base-pairing sites for Vcr043 on the 59 UTR of rpoS. The Vcr043-rpoS chimeric RNAs identified in rGRIL-seq also appear to be created following the cleavage of the 59 UTR of rpoS and ligation to Vcr043 as previously shown in E. coli DsrA RNA. However, it remains to be determined whether base-pairings of the sRNA to both sites on the rpoS transcript are necessary for the activation of expression and what RNase is involved in the cleavage of the 59 UTR. The Vcr043 was previously identified in a transcriptomic analysis in V. cholerae (43). Interestingly, its gene is located upstream of the VC1045 gene encoding RNA polymerase sigma factor 70 and oppositely transcribed to the VC1045 gene. In our analysis of their transcription start sites (TSSs) based on transcriptomic data, the distance between their TSSs is 25 bp and their predicted promoters are partially overlapped. However, further studies are needed to establish whether the housekeeping sigma factor 70 (s 70 ) is also regulated with the sRNA activator to control sigma factor S (s S ) in V. cholerae.
Negative regulators of RpoS expression. Using rGRIL-seq in E. coli, we identified the sRNA asYbiE as a repressor of rpoS translation. Resembling the mode of action of many negative sRNA regulators studied previously, asYbiE appears to base-pair with the translation initiation site of rpoS and represses its translation. Though the rGRIL-seq shows a relatively high number of asYbiE-rpoS chimeric reads (ranked no. 5 in total), it seems to be expressed at a low level in rich media (Fig. 4B). The asYbiE is an antisense RNA of ybiE which encodes a small peptide (2 kDa), YbiE, composed of 19 hydrophobic amino acids (35). Since YbiE is known to be expressed in both exponential and stationary growth phases in rich media (35), asYbiE seems to be negatively regulated by the expression of YbiE. In P. aeruginosa, we confirmed the previously characterized ReaL sRNA as a repressor of rpoS. In addition, we also identified two 39-derived sRNAs (s3661 and s0223) as repressors of rpoS expression. As seen previously (36), overexpression of ReaL shows a significant repression of rpoS, while two 39-derived sRNAs demonstrated only moderate repression. The predicted location of the base-pairing regions of the 39derived repressor sRNAs, at the upstream region of the 59 UTR (for s3661) and in the coding region of rpoS for s0223, is uncommon for sRNAs interfering with the initiation of translation. This suggests that s3661 and s0223 may negatively regulate rpoS expression by mechanisms differing from the canonical action of sRNAs targeting the Shine-Dalgarno sequence and the start of translation.
Using rGRIL-seq in V. cholerae, we identified the new regulatory function of the TfoR sRNA, as a repressor of rpoS. TfoR sRNA has been known as a positive regulator of tfoX by base-pairing with the 59 UTR of tfoX mRNA and activating the translation (45). RpoS and TfoX are known to be required for natural transformation of V. cholerae (49). In the light of our finding that TfoR regulates the two mRNAs in an opposite manner, many questions remain about the physiological relevance of reciprocal expression of natural competence and stress response by V. cholerae.
In addition to finding positive and negative regulators of rpoS, we also observed a few featured interactions at the 39 region of the rpoS transcript whose regulatory consequence were not addressed in this study. In E. coli, sAspA RNA seems to interact with the 39 UTR of rpoS, while sVca0838 seems to interact with the 39 coding region of rpoS in V. cholerae (Fig. S3 and S7). However, when examining the effect of these sRNAs when overexpressed on RpoS protein levels, they appear to have no effect on rpoS translation. It is still an open question whether the transcripts originating from the 39 region of rpoS might function as new 39-derived sRNAs sequestering (sponging) sAspA and sVca0838.
Our work presented here also highlights two general concepts applicable to sRNA regulation. First, as illustrated with rpoS, sRNAs and their targets can follow an evolutionary pathway unique to each organism reflecting the adaption to their specific environmental niches. Second, the expression of sRNAs is likely regulated and under the control of various transcription factors, highlighting the need to consider the interplay of transcriptional and posttranscriptional inputs when elucidating the workings of regulatory networks. The analysis of sRNA-mRNA interactions should therefore take full advantage of complementary tools such as GRIL-seq and rGRIL-seq to provide a comprehensive picture of the consequences of regulatory events leading to effective response to changing environmental conditions.

MATERIALS AND METHODS
Oligonucleotides and plasmids. The oligonucleotides and plasmids used in this study are listed in Table S2A and B in the supplemental material, respectively. The construction of recombinant plasmids was carried out using an In-Fusion cloning kit (Clontech, catalog no. 639648), and T4 DNA ligase when required. To construct the dual expression plasmids (pKH24::Z and pKH24::6H), the linear vector pKH24XS (Amp r ) was amplified from pBAD24 by PCR with R_pKH24Xblvec and F_KH24SphIvec, adding new restriction enzyme sites, XbaI and SphI, at the site following the P BAD promoter site. The amplified linear vector (pKH24XS) was circularized by T4 DNA ligase and transformed into E. coli competent cells (Clontech, catalog no. 636763). The plasmid pKH24XS was purified and linearized by the restriction enzyme NsiI. The lacZ gene and 6ÂHis tag sequence with the lacI q gene were amplified by PCR from plasmids pmini-Ptac::lacZ and pmini-Ptac::6His (13), respectively, with a primer pair (F_DelNsil_lacI/ R_Nsil_lacZ_6H). These PCR products were ligated with the previous pKH24XS (NsiI digested) using the In-Fusion cloning kit to construct pKH24::Z and pKH24::6H. The ligated plasmids were transformed into competent E. coli, and the plasmids were purified to construct E. coli or V. cholerae rpoS translational fusions. To construct the rpoS::lacZ translational fusion vectors (pKH24-rpoS Ec/Vc ::lacZ), the sequences from the transcription start site (TSS) to the 80th amino acid codon of the E. coli and V. cholerae rpoS gene were amplified from E. coli MG1655 and V. cholerae C6706 genomes with primer pairs F_ERI_ecrpoS180aa/R_Hd3_ecrpoS180aa and F_ERI_vcrpoS180aa/R_Hd3_vcrpoS180aa, respectively. The plasmid pKH24::lacZ was linearized with restriction enzymes (EcoRI and HindIII), and the rpoS genes from each of the three bacterial species, previously obtained by PCR, were inserted by ligation. To construct the His-tagged fusion vectors (pKH24-rpoS Ec/Vc ::6H), the sequence from the TSS to the sequence corresponding to rpoS Ec/Vc up to the stop codon was amplified from E. coli and V. cholerae chromosomal DNA. The PCR products were cloned into the linearized pKH24::6H (EcoRI/HindIII). The ligated vectors were introduced into competent E. coli by electroporation, and the plasmids were purified and used to clone each sRNA gene into these vectors (pKH24-rpoS Ec/Vc ::lacZ and rpoS Vc ::6H). Each E. coli or V. cholerae sRNA gene was amplified with their corresponding primer pairs (Table S2A), and the vectors were digested with XbaI and SphI. The PCR products were ligated with the vectors and transformed into competent E. coli. All sRNA sequences used for expression are listed in Table S2C. All plasmids carrying sRNA and rpoS Ec/Vc ::lacZ (or rpoS Vc ::6H) genes were purified and electroporated into E. coli MG1655 DlacIY DaraCD and the V. cholerae C6706 recA-lacZ* mutant for measurements of b-galactosidase assays or Western blotting. To construct the vector for expressing P. aeruginosa sRNAs, their corresponding sRNA genes were amplified from the P. aeruginosa PAO1 genome and cloned into pKH6 (Gm r ) as described previously (13). For chromosomal expression of rpoS Pa ::lacZ fusion or the 6ÂHis-tagged proteins, the two different lengths of rpoS Pa genes were amplified as carried out in E. coli and cloned into the digested plasmids (EcoRI/HindIII; pPtac-miniCTX::lacZ or pPtac-miniCTX::6H) as performed previously (13).
Bacterial strains. The list of bacterial strains created for this study is provided in Table S2B. E. coli MG1655, P. aeruginosa PAO1, and V. cholerae C6706 are referred to as the wild-type strains. When required, they were used for a mutant construction. In order to prevent catabolism of arabinose and X-Gal (5-bromo-4-chloro-3-indolyl-b-D-galactopyranoside) in b-galactosidase assays, E. coli MG1655 DaraDC DlacAI was engineered from the MG1655 DlacAI strain lacking the genes lacAYZI using the lambda Red recombineering system as described previously (50). Briefly, the genes araDABC were disrupted by homologous recombination with a PCR product consisting of a kanamycin resistance gene (Kan r ) flanked by 40-bp regions homologous to araD and araC genes on the E. coli chromosome. For engineering a chromosomal deletion of a gene in P. aeruginosa, the plasmid pEXG2 was used as described previously (51). For the b-galactosidase assays and plasmid stability in V. cholerae, the spontaneous mutant of lacZ (referred to as lacZ*) in the C6706 recA 2 strain was used to prevent catabolism of X-Gal. For rGRIL-seq, E. coli MG1655, P. aeruginosa PAO1, and V. cholerae C6706 recA-lacZ* strains were used.
b-Galactosidase assays. Assays of b-galactosidase activity in E. coli and V. cholerae were carried out in the same manner. Each overnight culture of E. coli or V. cholerae containing the plasmid pKH24-rpoS Ec/Vc ::lacZ/sRNA was grown in LB with 50 mg/ml carbenicillin. The overnight culture was diluted to an optical density at 600 nm (OD 600 ) of ;0.02 in 1 ml LB containing the same concentration of antibiotics and then grown to an OD 600 of ;1.0. To express the sRNA and rpoS mRNA, L-arabinose and IPTG (isopropyl-b-D-thiogalactopyranoside) were added to the final concentrations of 0.02% and 50 mM, respectively. b-Galactosidase activity was measured using 50 ml of cells after 3.3 h of induction with L-arabinose and IPTG. Miller units were calculated as described previously (52). For b-galactosidase assays in P. aeruginosa, an overnight culture of P. aeruginosa (pPtac-miniCTX-rpoS::lacZ) carrying either the empty vector (pKH6) or pKH6-sRNA was grown in LB with 37.5 mg/ml tetracycline and 75 mg/ml gentamicin. The overnight culture was diluted as done for E. coli, and L-arabinose and IPTG were added to the final concentrations of 0.2% and 100 mM, respectively. b-Galactosidase activity was measured using 50 ml of cells after 3.5 h of induction with L-arabinose and IPTG.
Western blot analysis. To prepare the protein samples for Western blot analysis, E. coli strains containing the plasmids with the rpoS gene and expressing different sRNAs (pKH24-rpoS Ec ::lacZ/sRNA) were grown as described for the b-galactosidase assay. After 2.3 h of induction, cells from a 300-ml culture were collected by centrifugation (15,000 Â g; for 1 min), and the total protein samples were prepared as described previously (13). To monitor the levels of endogenous RpoS Ec , monoclonal antibody against E. coli RpoS (BioLegend, catalog no. 663104) was used according to the manufacturer's instructions. For the protein analysis in V. cholerae, the strains carrying pKH24-rpoS Vc ::6H were grown as described for the b-galactosidase assay. The preparation of protein samples from P. aeruginosa and Western blotting was described previously (13). To monitor the C-terminal His-tagged RpoS protein levels in V. cholerae and P. aeruginosa, the His tag antibody (GenScript, catalog no. A00186) was used according to the manufacturer's instruction.
rGRIL-seq. The rGRIL-seq was carried out as described for GRIL-seq (13) with the following modifications. Cultures of E. coli MG1655, P. aeruginosa PAO1, or V. cholerae C6706 strains carrying the plasmid pKH13-t4rnl1 were grown overnight aerobically with shaking at 37°C, in LB (Luria-Bertani) broth with carbenicillin (for E. coli and V. cholerae, 50mg/ml; for P. aeruginosa, 150 mg/ml). The overnight cultures were diluted into the same medium to an OD 600 of 0.01. For induction of the T4 RNA ligase, IPTG was added to 1 mM when the culture reached a low (OD 600 of ;0.4) or high (OD 600 of ;3.5) cell density. After induction for E. coli and V. cholerae (40 min) and for P. aeruginosa (60 min), cells from 0.6 ml of each culture were used for total RNA isolation by the hot-phenol-chloroform extraction method. The cultures were directly treated with a phenol mixture at 65°C, which contained 0.1 volume of RNA extraction buffer (20 mM sodium acetate, 0.5% SDS, 1 mM EDTA, pH 5.2) and the same volume (0.6 ml) of phenol saturated with the RNA extraction buffer. The phenol-cell mixture was vigorously mixed and incubated for 10 min with an intermittent vigorous mix. After centrifugation (21,000 Â g) for 10 min at 4°C, the aqueous phase was isolated and mixed with 0.55 ml of phenol-chloroform. After a vigorous mix, the two phases were separated by centrifugation for 5 min. The aqueous phase was mixed with 0.45 ml of chloroform to remove the residual phenol in the aqueous phase. After a vigorous mix, the two phases were separated by centrifugation for 5 min, and the aqueous phase containing purified total RNAs (final volume of 0.4 ml) was mixed with 2.5 volumes of ethanol (1 ml) and stored at 280°C. The total RNA was collected by centrifugation (10 min, 21,000 Â g, 4°C) and washed with 70% ethanol (0.7 ml). The dried RNA pellets were dissolved with nuclease-free water (40 ml), and 12 mg of total RNA was used for the enrichment of chimeric RNAs. For the enrichment of rpoS chimeric RNAs, 4 multiple DNA oligonucleotides were designed and 20 pmol (5 pmol for each) of oligonucleotides was mixed and used in the enrichment step of rGRIL-seq. All steps for the enrichment and the following precipitation of the rpoS chimeric RNAs were carried out as described for GRIL-seq (13). The enriched chimeric RNAs were precipitated at 280°C, collected by centrifugation (10 min, 21,000 Â g, 4°C), and washed with 70% ethanol (0.5 ml). The enriched RNAs were dissolved in 13 ml of nuclease-free water, and 100 ng of enriched RNAs was used for RNA library preparation using the NEBNext Ultra Directional RNA library prep kit (NEB, catalog no. E7420s). A total of 11 PCR cycles was performed for amplification of the library, and then the library was purified twice with AMPure XP (Beckman Coulter, catalog no. A63880) according to the NEBNext Ultra Directional RNA library prep kit. The library derived from this RNA was subjected to sequencing on a NextSeq500 instrument (Illumina) with 150-bp paired-end reading.
Data analysis. Data analysis for rGRIL-seq was performed using CLC Genomics Workbench 7.0 as used for GRIL-seq (13); the detailed protocol is described in Text S1 in the supplemental material.
RNA isolation and Northern blot analysis. Overnight cultures of the bacteria grown at 37°C were diluted in LB to an initial OD 600 of ;0.02. Cell growth was monitored at different time points, and total RNA was isolated by the hot-phenol method as described above for rGRIL-seq. For Northern blotting, total RNA (10 mg) was mixed with RNA loading buffer II (Ambion) and denatured at 95°C for 3 min. The RNA was fractionated on a 5% polyacrylamide-7 M urea gel in 1Â Tris-borate-EDTA (TBE). The fractionated RNAs were electrotransferred to a Hybond N1 membrane (GE Healthcare) and cross-linked on the membrane using the Stratalinker UV cross-linker on the autocrosslink setting (2 times, 120,000 mJ/cm 2 ). After prehybridization with Rapid-hyb buffer (GE Healthcare), a 32 P-labeled oligonucleotide probe was added, and after overnight hybridization at 43°C, the membrane was washed with 2Â SSC-0.1% SDS and 0.5Â SSC-0.1% SDS buffer (1Â SSC is 0.15 M NaCl plus 0.015 M sodium citrate). If required for deprobing, the blots were incubated in hot 0.1% (wt/vol) SDS solution twice for 15 min with gentle agitation and then the residual SDS was removed with distilled water. The signal was visualized on Typhoon FLA 7000 (GE Healthcare). Each oligonucleotide used to prepare a radioactive probe is listed in Table S2A.
In silico prediction of base-pairing between sRNA and target RNAs. IntaRNA software 2.0 (34) was used to predict base-pairing interactions between sRNA and target RNAs. In the seed parameters, the default for the minimum number of base pairs in the seed region was set as 7. When required, it was set as 6. For the prediction of secondary RNA structure of the chimeric RNA, generally a single chimeric RNA sequence was extracted for the chimeric DNA reads, and the secondary RNA structure algorithm of CLC Genomics Workbench 7.0 was used according to the manufacturer's instructions.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, PDF file, 0.05 MB.