Global discovery of small RNAs in Yersinia ...Global discovery of small RNAs in Yersinia pseudotuberculosis identifies Yersinia-specific small, noncoding RNAs required for virulence

Global discovery of small RNAs in Yersiniapseudotuberculosis identifies Yersinia-specificsmall, noncoding RNAs required for virulenceJovanka T. Kooa, Trevis M. Alleyneb,1, Chelsea A. Schianoa, Nadereh Jafarib, and Wyndham W. Lathema,2

aDepartment of Microbiology-Immunology and bCenter for Genetic Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL, 60611

Edited by Susan Gottesman, National Cancer Institute, Bethesda, MD, and approved June 7, 2011 (received for review January 28, 2011)

A major class of bacterial small, noncoding RNAs (sRNAs) acts bybase-pairing with mRNAs to alter the translation from and/orstability of the transcript. Our laboratory has shown that Hfq, thechaperone that mediates the interaction of many sRNAs with theirtargets, is required for the virulence of the enteropathogen Yersi-nia pseudotuberculosis. This finding suggests that sRNAs playa critical role in the regulation of virulence in this pathogen, butthese sRNAs are not known. Using a deep sequencing approach,we identified the global set of sRNAs expressed in vitro by Y.pseudotuberculosis. Sequencing of RNA libraries from bacteriagrown at 26 °C and 37 °C resulted in the identification of 150 un-annotated sRNAs. The majority of these sRNAs are Yersinia spe-cific, without orthologs in either Escherichia coli or Salmonellatyphimurium. Six sRNAs are Y. pseudotuberculosis specific andare absent from the genome of the closely related species Yersiniapestis. We found that the expression of many sRNAs conservedbetween Y. pseudotuberculosis and Y. pestis differs in both timingand dependence on Hfq, suggesting evolutionary changes in post-transcriptional regulation between these species. Deletion of mul-tiple sRNAs in Y. pseudotuberculosis leads to attenuation of thepathogen in a mouse model of yersiniosis, as does the inactivationin Y. pestis of a conserved, Yersinia-specific sRNA in a mousemodel of pneumonic plague. Finally, we determined the reguloncontrolled by one of these sRNAs, revealing potential virulencedeterminants in Y. pseudotuberculosis that are regulated in a post-transcriptional manner.

Ysr29 | RybB | stress | Illumina-Solexa | 2D-DIGE

Historically it was assumed that only proteins act as regulatorsof gene expression. There has been increasing recognition,

however, that small, noncoding RNAs (sRNAs) serve as majorcomponents of diverse regulatory circuits in bacteria. sRNAs areRNA molecules that typically are encoded in intergenic regions,are independently transcribed, contain their own promoters andρ-independent terminators, and range from 50–500 nt in length(1). Many were discovered initially by computational methodsinvolving homology searches of closely related bacterial speciesand more recently have been identified by a variety of experi-mental approaches (i.e., genetic screens, deep sequencing, tilingmicroarrays, coimmunoprecipitation with proteins) (2). MostsRNAs control gene expression at the posttranscriptional levelby base-pairing with target mRNAs, resulting in alterations inmRNA target translation or half-life (3, 4). The predominantoutcome of the sRNA–target mRNA interaction is the down-regulation of gene expression [e.g., repression of outer mem-brane protein synthesis by the sRNAs MicA and RybB (5)], butpositive regulation by sRNAs also has been described [e.g.,regulation of the stationary phase Sigma factor RpoS by thesRNAs DsrA and RprA (6)]. In most cases, the RNA chaperoneprotein Hfq is required, presumably to stabilize the sRNA–

mRNA interaction, because the sRNA contact on the targettypically is short and imperfect (7).Like protein regulators of gene expression, sRNAs act to in-

tegrate extracellular signals that aid bacteria in adjusting to their

environment. This regulation includes the adaptation of patho-genic bacteria to the host and the coordination of expression ofvirulence determinants. sRNAs can regulate the expression ofvirulence genes directly [e.g., RNAIII regulation of staphylo-coccal protein A and the α-toxin gene mRNAs in Staphylococcusaureus (8, 9)] or control global regulators [e.g., quorum regula-tory sRNAs regulate the hemagglutinin/protease regulator hapRmRNA in Vibrio cholerae (10)]. The end result is the fine-tuningof metabolic requirements of pathogenic bacteria to endure thestress imposed by the host and the synthesis of virulence factors.There are three pathogenic Yersinia species that cause disease

in humans: Y. pestis, Y. pseudotuberculosis and Y. enterocolitica.Y. pestis is thought to have evolved from Y. pseudotuberculosis∼1,500–20,000 y ago. Y. enterocolitica is more distantly related,having diverged from a common ancestral Yersinia species 41–186 million y ago (11, 12). Y. pestis is the causative agent of thedisease plague. The bubonic form of the disease occurs followingthe transmission of Y. pestis via fleabite and multiplication ofbacteria in the lymph nodes; pneumonic plague, the most severeform of disease, occurs if the bacteria colonize and multiply inthe lungs (13). Y. pestis can be spread from person to person byinfectious respiratory droplets, and without early treatment,pneumonic plague is 100% fatal.In contrast to Y. pestis, the closely related Y. pseudotuberculosis

is a soil- and water-borne enteropathogen that primarily infectswild animals and birds. In humans it causes a mild, self-limitinggastrointestinal disease called “yersiniosis” and is transmitted bythe fecal–oral route. Yersiniosis caused by Y. pseudotuberculosisis characterized by ileitis and mesenteric lymphadenitis, as wellas fever and diarrhea (14). Although Y. pseudotuberculosis and Y.pestis are highly genetically related (they share >97% identity in75% of their chromosomal genes), they differ radically in theirpathogenesis and disease etiologies (15).Despite these differences, recent evidence indicates that the

sRNA chaperone Hfq is required for the full virulence of bothY. pestis (16) and Y. pseudotuberculosis (17) in mouse models ofinfection (s.c. or i.v. and intragastric, respectively) and suggestsa role for sRNAs in the regulation of pathogenesis. The arsenalof sRNAs expressed by Y. pestis and Y. pseudotuberculosis thatcontrol gene expression has not been examined experimentallyon a global level, however. Thus, we sought to gain a greaterunderstanding of the roles that sRNAs play in manipulating the

Author contributions: J.T.K. and W.W.L. designed research; J.T.K. and W.W.L. performedresearch; J.T.K. and C.A.S. contributed new reagents/analytic tools; J.T.K., T.M.A., N.J., andW.W.L. analyzed data; and J.T.K. and W.W.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1Present address: École Polytechnique Fédérale de Lausanne, CH-1015 Lausanne, Switzer-land.

2To whom correspondence should be addressed. E-mail: [email protected].

See Author Summary on page 15029.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1101655108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1101655108 PNAS | September 13, 2011 | vol. 108 | no. 37 | E709–E717

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

9, 2

021

synthesis of virulence factors in these bacteria. To do so, we useda deep sequencing approach to perform a transcriptome-wideanalysis to identify sRNAs encoded by the chromosome andplasmids of Y. pseudotuberculosis strain IP32953. We identifieda total of 150 unannotated sRNAs, the majority of which areYersinia specific. The deletion of multiple sRNAs led to the at-tenuation of Y. pseudotuberculosis in a mouse model of yersi-niosis, suggesting that these sRNAs regulate factors involved inthe pathogenesis of this species. In addition, we found that thedeletion of one of these virulence-associated, Yersinia-specificsRNAs that is conserved between Y. pseudotuberculosis and Y.pestis resulted in the reduced virulence of the plague pathogen inan intranasal mouse model of infection. These results may haveimplications for the evolution of mechanisms that regulate vir-ulence in Yersinia.

ResultsGlobal Identification of sRNAs Expressed by Y. pseudotuberculosis.Our goal was to identify sRNAs expressed by Y. pseudotubercu-losis in an unbiased fashion. For this purpose we generatedsRNA libraries of Y. pseudotuberculosis grown to early-log, mid-log, late-log, and stationary phase in broth culture at 26 °C and37 °C and subjected these libraries to deep sequencing based onthe Illumina-Solexa platform (18). This analysis resulted in ∼2.4–17 million 36-nt-long reads that were aligned exactly to theY. pseudotuberculosis chromosome and plasmids. The mappedreads were categorized into mRNA, ribosomal RNA (rRNA),tRNA, tmRNA, and miscellaneous (misc.) RNAs, and the re-mainder of the reads was classified as intergenic RNA sequences.The distribution of mapped reads into categories at the twotemperatures is shown in SI Appendix, Fig. S1. The miscellaneousRNAs category contains all previously annotated sRNAs in Y.pseudotuberculosis (the annotation of which was based on se-quence homology with Escherichia coli and Salmonella typhimu-rium sRNAs), and our analysis was able simultaneously toconfirm and map the changing expression patterns of thesesRNAs over time (Fig. 1A). Spot 42 (Spf) is a previously anno-tated sRNA for which the numbers of deep sequencing readscorrelated well with its expression as determined by Northernblot (Fig. 1B). Additionally, the sensitivity of Illumina-baseddeep sequencing in identifying sRNAs was validated, becauselow-abundance sRNAs such as SraG, which was represented byas few as 191 reads and could not be detected by Northern blotusing as much as 10 μg of total RNA, are detected by thismethod. Notably, the major RNA species in the cell, rRNA, wasreduced to ∼0.01–0.7% of total reads by rRNA depletion withMicrobExpress (Ambion). The proportion of tRNA reads in thetotal also is small (0.3–19%) compared with previous globalsRNA identification efforts (19). The intergenic reads mapped tothe chromosome of Y. pseudotuberculosis (29.6–43.3% of totalreads) contain the 5′ and 3′ UTRs of transcribed mRNAs, cis-encoded antisense RNAs, and putative trans-encoded sRNAs.To identify these sRNAs specifically, the intergenic reads weregrouped further into clusters at least 50 nt in length (e.g., thisgrouping yielded ∼3,200 clusters for the 6-h time-point at 37 °C).A filtering algorithm was developed to eliminate the majority ofthe 5′ and 3′ UTRs in the clustered intergenic groups by ana-lyzing the expression level of clusters compared with the sur-rounding ORFs. The clusters whose expression levels differedfrom the expression levels of a neighboring ORF by less thanthreefold were removed from further analysis. The remainingintergenic clusters (e.g., 1,313 clusters for the 6-h time point at37 °C) then were assessed using the Integrated Genome Browser(20). Computational prediction of putative promoter and ρ-in-dependent terminator sequences, which are characteristic fea-tures of bacterial sRNAs (21), led to the identification of 150putative unannotated sRNAs in Y. pseudotuberculosis. Analysisof the intergenic clusters did not reveal any potential sRNAs on

the Y. pseudotuberculosis plasmid pYptb and identified only asingle sRNA on the plasmid pYV.

Yersinia-Specific sRNAs. The 150 clusters that satisfied the criteriafor presumed sRNAs are referred to hereafter as “Ysrs” (forYersinia small RNAs). The predicted genomic coordinates andsizes, orientation with respect to surrounding ORFs, and num-bers of deep sequencing reads generated at the different tem-peratures and time points for all the Ysrs are listed in DatasetS1. For 32 Ysrs, BlastN analysis revealed orthologous sequencesin the E. coli and S. typhimurium genomes; these sequences in-clude the previously characterized sRNAs MicA (Ysr7), FnrS/Stnc520 (Ysr11), RprA (Ysr40), GcvB (Ysr45), RybB (Ysr48),MicM (Ysr145), RyhB (Ysr146.1 and Ysr146.2), GlmY (Ysr147),GlmZ (Ysr148), and OmrA/B (Ysr149) (3). Seventy-nine percent(118/150) of the Ysrs are specific to Y. pseudotuberculosis andY. pestis in that they do not show sequence conservation withother bacterial species (Fig. 2, highlighted in red). Ninety-twoYsrs also are absent from other members of the Yersinia genussuch as Y. ruckeri, Y. frederiksenii, and Y. intermedia and alsofrom the closely related genus Photorhabdus (Dataset S1, sheet2). We identified six Ysrs that are Y. pseudotuberculosis specific(Ysr29, Ysr53, Ysr70, Ysr84, Ysr94, Ysr118), having no homol-ogous sequences in the genome of Y. pestis. In addition, 63/144Ysrs encoded by both Y. pseudotuberculosis and Y. pestis containsingle or multiple differences in sequence (i.e., mismatches,deletions, or insertions highlighted in yellow in Fig. 2).At 26 °C, the majority of identified sRNAs are not expressed

in earlier stages of growth (based on number of deep sequencingreads) and start to accumulate at later time points (SI Appendix,Fig. S2). The exceptions are Ysr8, Ysr32, and Ysr45/GcvB. Also,at this temperature, Ysr3 levels peak in late-log phase and thendecrease in stationary phase. Although there are more excep-tions to this trend at 37 °C than at 26 °C, a large proportion ofsRNAs accumulate over the growth curve (SI Appendix, Fig. S2).

Fig. 1. Experimental identification of annotated sRNAs in Y. pseudotu-berculosis. (A) All previously annotated Y. pseudotuberculosis sRNAs wereidentified by Illumina-Solexa sequencing. Percent of total reads for anno-tated RNAs identified at 37 °C shows a flux in sRNA levels over the course ofbacterial growth. (B) The expression levels of the sRNA spf (Spot 42) as de-termined by deep sequencing reads (red bars in A) correlate with the levelsof the sRNA as measured by Northern blot analysis.

E710 | www.pnas.org/cgi/doi/10.1073/pnas.1101655108 Koo et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

9, 2

021

The most notable exception at 37 °C is Ysr45/GcvB, the levels ofwhich are highest at the start of growth and decrease over time.The most abundant sRNA at 26 °C is Ysr45/GcvB, whereas at37 °C, the most sequencing reads were obtained for Ysr7/MicAand Ysr149/OmrA/B (SI Appendix, Fig. S2).

Verification of sRNA Expression. Northern blot analysis was used todetect newly identified sRNAs in Y. pseudotuberculosis. Weperformed the analysis for 49 Ysrs and detected 29 of theseRNAs using biotinylated oligonucleotide probes (Dataset S1).For most of the sRNAs detected, the band sizes correlated withthe lengths predicted by the deep sequencing (i.e., Ysr4, Ysr7,and Ysr11). Some Ysrs, however, yielded bands corresponding tolarger transcripts (i.e., Ysr1, Ysr2, and Ysr9). For nine of theseYsrs, we performed simultaneous 5′ and 3′ RACE to map theboundaries of primary transcripts and to distinguish them fromthe processing of longer transcripts. The genomic coordinatesand lengths of these sRNAs are listed in Dataset S1, and foreight of the nine Ysrs, the coordinates are identical to or fallwithin 20 nt of the borders predicted by deep sequencing.Because many genes in Yersiniae are under thermal regulation

and are expressed at either 26 °C or 37 °C (22), it was of interest toexamine the expression of Ysrs over the time course of growth atboth these temperatures. Additionally, because most sRNAsidentified in enteric bacteria are dependent on Hfq for theirfunctions, we examined whether the absence of hfq from the Y.pseudotuberculosis genome led to differential expression and/orstability of the Ysr in question. The expression levels of all Ysrstested increase over time and culminate in stationary phase at26 °C (Fig. 3A). At this temperature only Ysr4 shows strongdependence on Hfq for its expression or stability. All the Ysrstested at 37 °C are expressed similarly at 26 °C, with the exception

of Ysr11, which exhibits the highest expression level at late-logphase and then decreases in stationary phase. This sRNA dis-played a similar pattern of expression when examined in S.typhimurium (STnc520) (23). Unlike S. typhimurium, however, theabundance of Ysr11 is not dependent on Hfq at either tempera-ture. Ysr4 and Ysr48/RybB are the only sRNAs examined whoseexpression levels are Hfq dependent at 37 °C. For all the sRNAstested at both temperatures, only the expression pattern of Ysr45/GcvB at 37 °C by Northern blot appears to be different from thatpredicted by the numbers of deep sequencing reads.Because all but six sRNAs identified in Y. pseudotuberculosis

contain an at least partially homologous sequence in the Y. pestisgenome, Northern blot analysis was used to examine the expres-sion of the seven sRNAs tested above in Y. pestis grown under thesame conditions (26 °C and 37 °C, wild-type and hfq mutant) (Fig.3B). In Y. pestis, at 26 °C the sRNAs examined show mostly steady-state levels over time and little dependence on Hfq. At 37 °C,however, these Ysrs are expressed differently than in Y. pseudo-tuberculosis: Most show stable levels or accumulation over timewith a peak at late-log phase and are almost undetectable in sta-tionary phase. Additionally, unlike in Y. pseudotuberculosis, all thesRNAs tested require Hfq for their stability/expression.Ysr29 is one of six Y. pseudotuberculosis specific sRNAs. In-

terestingly, this sRNA is unique in that it is specific for theIP32953 strain of Y. pseudotuberculosis and is absent from othersequenced Y. pseudotuberculosis isolates. It is encoded in theintergenic region between the adenylate cyclase gene cyaA andantisense to a portion of the 3′ end of YPTB0186 (encodinga putative transposase for IS285). Even though there are multi-ple copies of this transposase in the genomes of Y. pseudotu-berculosis and Y. pestis, the Ysr29 locus is the only occurrence ofthis sRNA sequence. The Ysr29 transcript accumulates in sta-

Fig. 3. Verification of Ysr expression. (A) Northern blot analysis was performed to examine the expression of identified Ysrs. (B) Northern blot analysis ofY. pseudotuberculosis and Y. pestis wild-type and Δhfq strains. Representative blots of at least two replicate reproducible experiments are shown. Blacktriangles above the blots indicate different time points on the bacterial growth curve at which the cultures were collected for RNA isolation (i.e., early-log,mid-log, late-log, and stationary phase). 5S rRNA is shown as a loading control.

Fig. 2. Yersinia small RNAs (Ysrs). The sequences of sRNAs identified by deep sequencing in Y. pseudotuberculosis strain IP32953 were compared with thegenomes of Y. pestis strain CO92, Y. enterocolitica 8081, E. coli strain K12 (substrain MG1655), and Salmonella enterica serovar Typhimurium strain LT2 byBlastN analysis. Ysrs present in the genome and identical in sequence are shown in green, and Ysrs present in the genome with sequence differences areshown in yellow. For an sRNA to be considered present in a genome, at least 50 nt of its sequence had to contain homology in the genome of interest. Ysrsabsent from the genome are shown in red. Detailed information for each Ysr is given in Dataset S1.

Koo et al. PNAS | September 13, 2011 | vol. 108 | no. 37 | E711

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

9, 2

021

tionary phase of growth at both 26 °C and 37 °C, with slightlyelevated levels at the optimal growth temperature for Y. pseu-dotuberculosis (Fig. 3A). This sRNA may be classified as Hfqdependent, because its levels decrease in the absence of hfq,particularly at 26 °C.

Contribution of Newly Identified Yersinia sRNAs to Virulence. Pre-vious work from our laboratory has established that Hfq is crit-ical for the pathogenesis of Y. pseudotuberculosis, and thedeletion of hfq results in an ∼10,000-fold attenuation of thebacteria in a mouse model of yersiniosis (17). This result suggeststhat Hfq, together with sRNAs, regulates essential virulencedeterminants in Y. pseudotuberculosis. To test this hypothesis, theHfq-dependent sRNAs Ysr29 and Ysr48/RybB and the Hfq-in-dependent sRNAs Ysr23 and Ysr35 (SI Appendix, Fig. S3A) weredeleted from the Y. pseudotuberculosis chromosome by homol-ogous recombination. Deletions of these Ysrs were confirmed byPCR (SI Appendix, Fig. S3C). The strains carrying deletions ofYsr23, Ysr29, Ysr35, and Ysr48/RybB grew comparably to thewild-type strain at both 26 °C and 37 °C (SI Appendix, Fig. S3Eand F). In addition, the deletion of Ysr23, Ysr29, Ysr35, andYsr48/RybB did not alter significantly the expression of thesurrounding ORFs as measured by quantitative RT-PCR (qRT-PCR) (SI Appendix, Fig. S4). To determine whether thesesRNAs contribute to the virulence of Y. pseudotuberculosis, wild-type and Ysr-deleted strains were used to infect BALB/c mice viathe intragastric route. Although 90% of mice infected with 2.0 × 105

cfu of wild-type Y. pseudotuberculosis succumbed by day 8 post-infection, 10% of mice infected with ΔYsr48/RybB and 50% ofmice infected with ΔYsr29 or with ΔYsr35 survived for the du-ration of the experiment (Fig. 4A). Infection of mice with theΔYsr23 strain did not result in a significant difference in survivalcompared with the wild-type but showed a trend toward atten-uation. We also measured the weights of the animals as an in-

dicator of their overall health. Mice infected with wild-typebacteria displayed significant weight loss by day 6 before succumb-ing to infection, whereas the animals infected with Ysr mutants thatsurvived the infection were able to recover after an initial weightdecline and continued to maintain or gain weight (Fig. 4B).Ysr35 is 339 nt long, and the ExPasyTranslate Tool (24)

predicts that a small ORF may be encoded within the sRNA. Todetermine if a peptide is synthesized from the predicted ORF,we generated by allelic replacement a version of the ORF withthe influenza HA epitope tag on the predicted C-terminal end.Under the conditions in which Ysr35 is expressed, we were un-able to detect a HA-tagged peptide by immunoblot (SI Appendix,Fig. S3B).Because Ysr23, Ysr35, and Ysr48/RybB (but not Ysr29) also

are encoded in the genome of Y. pestis and are transcribed(Dataset S1 and Fig. 3), we deleted these sRNAs from Y. pestis(SI Appendix, Fig. S3D) and examined their contributions to thedevelopment of pneumonic plague. Similarly to Y. pseudotuber-culosis, the inactivation of both Ysr23 (although affecting theexpression of one of the neighboring ORFs, ybcI) (SI Appendix,Fig. S4) and Ysr48/RybB does not affect the ability of Y. pestis tocause disease and the subsequent death of mice when deliveredvia the intranasal route (Fig. 4C). However, much as in Y.pseudotuberculosis, the deletion of Ysr35 results in the attenua-tion of Y. pestis following intranasal infection, significantlydelaying the time to death of mice as compared with the fullyvirulent, wild-type strain (Fig. 4C). This observation suggests thatYsr35 may play a role in virulence that is conserved between thetwo Yersinia species.

Identification of sRNA-Regulated Proteins. Considering the unique-ness of Ysr29 to the IP32953 strain of Y. pseudotuberculosisand its contribution to virulence, we performed a proteomicanalysis using 2D differential gel electrophoresis (2D-DIGE) to

Fig. 4. Contribution of Ysrs to the virulence of Y. pseudotuberculosis and Y. pestis. (A) Groups of 10 mice were inoculated via oral gavage with Y. pseu-dotuberculosis wild-type, ΔYsr23, ΔYsr29, ΔYsr35, and ΔYsr48/RybB strains (∼2.0 × 105 cfu). Survival of infected mice was monitored over 21 d. P values weredetermined by Mantel–Cox survival analysis log-rank test. P = 0.2254 for ΔYsr23 compared with wild-type (not significant); *P = 0.0202 for ΔYsr29 comparedwith wild-type; ***P = 0.0002 for ΔYsr35 compared with wild-type; P = 0.9154 for ΔYsr48/RybB compared with wild-type (not significant). Data are repre-sentative of three independent experiments. (B) Body weight over 21 d of mice infected with Y. pseudotuberculosis. The plot shows median weight, indicatedby a solid line; boxes represent the 25th and 75th percentiles, and whiskers represent the range. Significance was calculated by student’s unpaired t-test. (C)Groups of 10 mice were inoculated intranasally with Y. pestis wild-type, ΔYsr23, ΔYsr35, and ΔYsr48/RybB strains (∼1.0 × 104 cfu). Survival of infected micewas monitored over 7 d. P values were determined by Mantel–Cox survival analysis log-rank test. P = 0.9066 for ΔYsr23 compared with wild-type and P =0.0946 for ΔYsr48/RybB compared with wild-type (both not significant); ***P < 0.0001 for ΔYsr35 compared with wild-type. Data are representative of twoindependent experiments.

E712 | www.pnas.org/cgi/doi/10.1073/pnas.1101655108 Koo et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

9, 2

021

determine the regulated targets of this sRNA. Protein profilesfrom whole-cell lysates of wild-type Y. pseudotuberculosis werecompared with those of the ΔYsr29 strain grown to stationaryphase at 26 °C, the time point at which this sRNA is mostabundant. The comparison of protein profiles between the wild-type and ΔYsr29 strains showed 16 spots with 1.5-fold or moredifference in fluorescence intensity (Fig. 5A). Thirteen spotswere analyzed successfully by MALDI-TOF mass spectroscopyand correspond to eight proteins: 30S ribosomal protein S1(RpsA), outer membrane protein A (OmpA), the chaperoninGroEL, glutathione-S-transferase (GST), the molecular chap-erone DnaK, peroxidase (AhpC), ribosome recycling factor(RRF), and urease (UreC) (SI Appendix, Table S1). Ysr29 doesnot appear to regulate the identified proteins at the levelof transcription, because qRT-PCR analysis shows that the

transcript levels for the genes in question are equivalent in thewild-type and ΔYsr29 strains (Fig. 5B).Therefore, we examined the effects of Ysr29 at the post-

transcriptional level by generating chromosomal in-frame fusionsof the GST, RpsA, OmpA, and GroEL coding regions with theHA-epitope tag in both wild-type and ΔYsr29 strains. Levels offusion proteins were measured by immunoblot analysis using ananti-HA antibody (Fig. 5C). In accordance with 2D-DIGE, wefound that GST is more abundant in the ΔYsr29 strain than inthe wild-type background, whereas RpsA, OmpA, and GroELare elevated in the wild-type compared with the ΔYsr29 strain,demonstrating posttranscriptional regulation by this sRNA. Be-cause the overexpression of sRNAs can result in more dramaticeffects on regulated target protein levels than sRNA deletions,we generated an isopropyl-β-D-thiogalactopyranoside (IPTG)-

Fig. 5. Posttranscriptional regulation of targets by Ysr29. (A) Proteomic comparison of Y. pseudotuberculosis wild-type and ΔYsr29 strains by 2D-DIGE. Wild-type proteins were labeled with Cy3, and ΔYsr29 mutant proteins were labeled with Cy5. The gel image shows 1.5× or greater differences in spot volume ratiofor 16 marked spots. Blue indicates spots with a Cy3/Cy5 ratio >0 (increased expression in ΔYsr29 compared with the wild-type); red indicates spots with a Cy3/Cy5 ratio <0 (increased expression in wild-type compared with ΔYsr29). Molecular masses in kilodaltons, indicated by the numbers to the right of the image,are approximate. (B) Expression of Ysr29 targets. Cultures of wild-type and ΔYsr29 strains were grown to stationary phase at 26 °C, and transcript levels oftargets identified by 2D-DIGE were examined by qRT-PCR. The fold change in RNA levels is relative to wild-type, which was set to 1. (C) Western blots ofwhole-cell lysates from wild-type and ΔYsr29 strains expressing chromosomal HA fusions of GroEL, OmpA, RpsA, and GST. (Upper) Anti-HA antibody. (Lower)Anti-RpoA antibody (loading control). (D) The effect of Ysr29 overexpression on GroEL and GST. ΔYsr29 strains with the GroEL-HA and GST-HA fusions wereelectroporated with either the pACY177 vector or a Ptac-Ysr29 overexpression construct. (Upper) Northern blots showing overexpression of Ysr29 uponaddition of 1 mM IPTG to the growth medium. Note: The band in the pACYC177 samples is nonspecific and migrates at a slightly higher molecular weightthan Ysr29. 5S rRNA is shown as a loading control. (Lower) Immunoblots showing the effect of Ysr29 overexpression on GST-HA, GroEL-HA, and RpoA (loadingcontrol). (E) IntaRNA predictions of base-pairing between Ysr29 and the target mRNAs. The software predicts at least hepta-nucleotide pairing of Ysr29 withthe 5′ UTRs of all four target mRNAs.

Koo et al. PNAS | September 13, 2011 | vol. 108 | no. 37 | E713

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

9, 2

021

inducible Ysr29 plasmid and introduced this construct into theΔYsr29 strains carrying the GST-HA and GroEL-HA fusions.As expected, upon induction of Ysr29 expression with IPTG(Fig. 5D, Upper), the abundance of GroEL-HA was elevated, andthe levels of GST-HA were reduced (Fig. 5D, Lower). BecauseYsr29 expression is not fully repressed in the GST-HA strain(Ptac-Ysr29, 0-min Northern blot in Fig. 5D), GST-HA protein isnot detected even at the 0-min time point, demonstrating thesubstantial effect of extended Ysr29 overexpression on GSTsynthesis. Finally, we used a computational approach to predict ifYsr29 could interact with the 5′ UTRs of these target mRNAs.The IntaRNA alignment tool (25) predicts at least 7-nt-longbase-pairing interactions of Ysr29 with the 5′ UTRs of all fourvalidated targets (Fig. 5E).

DiscussionAlthough small regulatory RNAs have become increasinglyrecognized as major modulators of gene expression in bacteria(26), only a limited number of studies have sought to identify theglobal set of expressed sRNAs, particularly in bacterial patho-gens (27–29). Additionally, the expression of only a small num-ber of sRNAs has been verified in Yersiniae relative to otherenteric bacteria (3). Here, we describe the deep sequencing-based identification of sRNAs in the human gastrointestinalpathogen Y. pseudotuberculosis. A similar approach has beenused to identify sRNAs and mRNA targets that interact with Hfqin S. typhimurium (23) and in an RNomic analysis of Bur-kholderia cenocepacia (30). Analysis of sRNA libraries generatedfrom Y. pseudotuberculosis grown at 26 °C and at 37 °C led to theidentification of 150 putative sRNAs as well as confirming theexpression of 15 previously annotated sRNAs. The ability toidentify all the “miscellaneous” RNAs in an impartial mannervalidated the deep sequencing approach for discovery of pre-viously unidentified sRNAs. The number of putative sRNAs ismuch smaller than the 1,478 sRNAs predicted for the closelyrelated Y. pestis by Livny et al. (31) who used a computer algo-rithm that relies on conservation of intergenic sequences andρ-independent terminators, but is comparable to the number ofsRNAs predicted and verified in other enteric bacteria (3). It ispossible, however, that additional sRNAs may be expressed un-der conditions not examined here, such as upon host-cell contactor during animal infection.Thirty-two identified sRNAs are orthologs of potential sRNAs

encoded in the genomes of E. coli and S. typhimurium. Thisfinding is not surprising, because sequence conservation inintergenic regions of closely related bacteria was one of theoriginal criteria used to identify sRNAs (32). The unexpectedobservation is that many of the well-characterized sRNAs fromE. coli and S. typhimurium are not encoded by the genome ofY. pseudotuberculosis as determined by BlastN alignment analy-sis. These sRNAs include MicC, OxyS, ArcZ, and DsrA, amongothers. Although informative, this type of analysis also was un-able to identify SgrS in the genome of Y. pseudotuberculosis, ansRNA that has been described previously in this species (33, 34).Interestingly, the filtering algorithm we used to identify sRNAsfrom our deep sequencing data also did not reveal SgrS, butwhen the parameters were relaxed (i.e., the position of thecluster with respect to any ORF was moved within 1 kb on thesame strand), the deep sequencing reads for SgrS were indeedpresent (Ysr150 in Dataset S1). This observation suggests thatusing multiple tools may enhance the comparison of sRNAsequences between species and that adjusting the parameters ofdeep sequencing data analysis could reveal additional sRNAsencoded in the genome of Y. pseudotuberculosis.Given the limitations discussed above, it is interesting to hy-

pothesize how Y. pseudotuberculosis may have evolved to regu-late the expression of genes that are conserved between theenteric bacteria. The fact that Y. pseudotuberculosis may not

express a DsrA ortholog, for example, raises the question of howY. pseudotuberculosis regulates the stationary phase Sigma factorσS synthesis. One possibility is that some of the Yersinia-specificsRNAs that we have identified functionally replace the above-mentioned enteric sRNAs. Alternatively, RpoS in Yersiniae maynot require regulation by sRNAs for its synthesis. Another dif-ference is that Y. pseudotuberculosis encodes two distinct copiesof the iron level-regulated sRNA RyhB (Ysr146.1 and Ysr146.2).Because the expression levels of these sRNAs in Y. pseudotu-berculosis are low in iron-replete conditions, it would be in-teresting to examine the expression of the two RyhB copies in Y.pseudotuberculosis in the iron-limiting conditions that are knownto induce RyhB in E. coli (26, 35).We also identified 118 Yersinia-specific sRNAs (78.7%) with-

out a sequence match in available databases. Similarly, ananalysis of S. typhimurium-specific genetic islands by Padalon-Brauch et al. (36) led to the identification of 28 candidatesRNAs, 19 of which were verified. These observations suggestthat related bacterial species have evolved distinct species-spe-cific regulatory networks that rely on sRNAs that are requiredfor the adaptation to a particular niche or, in the case ofpathogens or symbionts, for the interaction with the host. Ad-ditionally, we identified sRNAs that are conserved between Y.pseudotuberculosis and Y. pestis, two species of Yersiniae thatshare 97% identity in 75% of their protein-coding genes (15).Remarkably, a substantial number of these sRNAs (43.75%)contain nucleotide mismatches, insertions, or deletions. Al-though these hypotheses are not yet tested, it is interesting tospeculate that these differences in sRNA sequences might alterthe secondary structure of the molecules or that a single-nucle-otide mismatch between the sRNA and its target mRNA mightaffect the outcome of the interaction. To determine whether thenucleotide differences present in Ysrs in Y. pseudotuberculosisand Y. pestis are localized or are spread across the entire se-quence of the sRNA, we performed ClustalW2 alignments of allYsrs containing mismatches between the two species (SI Ap-pendix, Table S2). The differences do not localize to either the 5′or the 3′ end of Ysrs but rather are spread throughout the lengthof the sRNAs. Thus, although the mismatches in the conservedsRNAs could lead to differential regulation of targets by the twoclosely related species, it also is possible that these differencesmay have little or no effect on sRNA function.We also observed that sRNAs conserved between Y. pseudo-

tuberculosis and Y. pestis demonstrated divergent expressionpatterns. Most Ysrs accumulate over time in Y. pseudotubercu-losis, but in Y. pestis the levels of those same Ysrs peak in late-logphase and decrease in stationary phase. This difference suggeststhat the temporal regulation of Ysr gene expression or stabilitymay affect the outcomes of Ysrs in modulating target mRNAexpression. This hypothesis is supported by the work of Bai et al.(37), which showed that the deletion of hfq causes a severegrowth defect at 37 °C in Y. pestis but a less drastic effect in Y.pseudotuberculosis. Thus, our findings, together with those of Baiet al. (37), further support the hypothesis that these two genet-ically linked species could differentially regulate gene expressionvia sRNAs to result in divergent disease manifestations.Most trans-acting sRNAs in bacteria require the RNA chap-

erone Hfq for their functions. Although the exact role of Hfq isnot fully understood, it appears that the chaperone promotesbase-pairing interactions of the sRNA and the mRNA target byincreasing the rate of sRNA–mRNA association in addition toprotecting the sRNAs from degradation (38, 39). Here we noteanother difference between Y. pseudotuberculosis and E. coli andSalmonella. In E. coli and Salmonella, MicA and GcvB areknown to rely on Hfq for proper expression/stability and also forthe regulation of their mRNA targets (40, 41), but in Y. pseu-dotuberculosis, expression levels of Ysr7/MicA, Ysr20, Ysr23,and Ysr45/GcvB do not differ between the wild-type and Δhfq

E714 | www.pnas.org/cgi/doi/10.1073/pnas.1101655108 Koo et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

9, 2

021

strains. Our finding that the majority of sRNAs tested in Y.pseudotuberculosis do not require Hfq for their expression and/orstability suggests the possibility of an alternative sRNA chaper-one. Y. pseudotuberculosis may rely on SmpB, an Sm-like chap-erone of tmRNA (42), or an as yet unidentified protein tostabilize sRNAs and promote interaction with mRNAs.Hfq is known to play a role in the virulence of Y. pseudotu-

berculosis (17), and in other species a majority of sRNAs requireHfq for their function (reviewed in ref. 3). Therefore, we ex-amined the contribution of several sRNAs, whose stability/ex-pression either relies on or does not require Hfq, to the virulenceof Y. pseudotuberculosis. The deletion of both Ysr29 and Ysr35resulted in increased survival of mutant-infected mice comparedwith those infected with wild-type bacteria, whereas the deletionof Ysr23 delayed the overall time to death of mice but did notimpact overall survival. Interestingly, deletion of the Hfq-de-pendent sRNA Ysr48/RybB did not result in a significant de-crease in virulence of Y. pseudotuberculosis, suggesting that not allHfq-binding sRNAs play a direct role in virulence. The attenua-tion of individual sRNA-deleted strains is not as dramatic as thatof a Δhfq strain (17), suggesting that in the absence of Hfq thedysregulation of multiple sRNA-dependent targets may result inthe severe attenuation of Y. pseudotuberculosis. It is possible thata double mutant of Ysr35 and Ysr29 may have a phenotype thatmore closely resembles that of the Δhfq mutant.Because Ysr35 is a Yersinia-specific sRNA that is conserved

between Y. pseudotuberculosis and Y. pestis, we deleted thissRNA from the genome of the plague bacillus and tested its rolein virulence. The Ysr35 deletion is attenuated in a mouse modelof pneumonic plague (Fig. 4C), suggesting that the sRNA mayplay a similar role in virulence in Y. pseudotuberculosis and Y.pestis. It will be of interest to determine whether Ysr35 regulatesthe same or distinct targets in the two species. In addition, al-though we were unable to detect the production of a Ysr35-encoded peptide by immunoblot in vitro, our results do notdiscount the possibility that the ORF still may be translatedunder other conditions.Hfq-binding sRNAs in E. coli and S. typhimurium exert a variety

of physiological roles mediated by the base-pairing interactions ofthe sRNA to the mRNA target. We sought to determine howYsr29 may control virulence in Y. pseudotuberculosis by identifyingthe regulated targets of this unique sRNA. Proteomic analyses ofwild-type and ΔYsr29 strains resulted in identification of eightputative targets of Ysr29, although it is not yet known whetherYsr29 regulates these factors directly or acts indirectly to altertheir synthesis. Nevertheless, Ysr29 may act on these targetsposttranscriptionally, because the absence of the sRNA does notaffect their transcript levels but instead alters the levels of protein.This finding suggests that 2D-DIGE accompanied by target-fusion(or direct Western blot analysis, if antibodies are available) is anappropriate approach for validating target gene expression, be-cause transcriptional profiling by microarray probably would notreveal the targets identified by our analysis.All the Ysr29-regulated targets validated by immunoblot

analysis have the potential to be involved in the virulence of Y.pseudotuberculosis. For example, GST participates in protectingcells against the damage of oxidative stress (43, 44), and Ysr29repression of GST levels may prevent an aberrant response tothis stress. In addition, OmpA is an outer membrane proteinknown to play a role in the pathogenesis of E. coli K1 and actsas an adhesin, invasin, and immune evasin (45–47). Interestingly,in E. coli, the expression of OmpA is posttranscriptionally re-pressed by MicA (48). It is not known if MicA acts similarly inY. pseudotuberculosis, but we show here that Ysr29 acts to en-hance OmpA levels. There are other examples of targets beingdifferently regulated by multiple sRNAs [e.g., RpoS regulationby DsrA, RprA, ArcZ, and OxyS (49, 50)] and also of multipletargets under the control of one sRNA [e.g., RybB regulation of

outer membrane protein synthesis (51)]. In sum, the globalidentification of sRNAs in Y. pseudotuberculosis provides anopportunity to discover mechanisms of virulence gene regulationin this pathogenic bacterium, particularly in comparison withclosely related species such as Y. pestis and in contrast to otherbacteria such as E. coli and Salmonella, the prototype species forstudying sRNAs.

Materials and MethodsRNA Isolation. Overnight cultures of Y. pseudotuberculosis strain IP32953were subcultured in brain heart infusion (BHI) broth supplemented with 2.5mM CaCl2 to an OD620 of 0.1 and grown to early-log (OD620 0.3, 1.5 h), mid-log (OD620 0.65, 3.5 h), late-log (OD620 4.3, 6 h), and stationary phase (OD620

7.6, 11 h) at either 26 °C or 37 °C. Five OD equivalents were collected, andRNA was stabilized with the RNAprotect Bacteria reagent (Qiagen). RNAenriched for sRNAs was extracted using the RiboPure Bacteria kit (Ambion)with modifications. Specifically, the provided columns (which are designedto eliminate small RNAs) were omitted, and instead the RNA was pre-cipitated overnight with isopropanol. Isolated RNA was treated with DNase I(Ambion), and the quality was assessed using the Experion automatedelectrophoresis system (Bio-Rad). RNA also was isolated as above from Y.pestis strain CO92 Δpgm for experiments comparing the expression ofidentified sRNAs by Northern blot analysis. For Y. pestis RNA isolation,overnight cultures diluted to an OD620 of 0.1 were grown to early-log (OD620

0.2, 2.5 h), mid-log (OD620 0.8, 5.5 h), late-log (OD620 1.8, 8.5 h), and sta-tionary phase (OD620 4.5, 15 h) at 26 °C and at 37 °C.

sRNA Library Preparation and Illumina-Solexa Sequencing. Ribosomal RNA waseliminated by the MicrobExpress kit (Ambion), and the integrity of the RNAwas reanalyzed by Experion electrophoresis. sRNA libraries were generatedusing the Illumina small RNA v1.5 kit with the following modifications: RNAwas treated with 5′ RNA polyphosphatase (Epicentre) followed by the liga-tion of RNA adapter oligonucleotides, cDNA synthesis, and PCR amplificationfor 12 cycles. The library then was separated on 6% acrylamide gels, andfragments between 90–500 nt were excised. cDNA was eluted from the gelsand precipitated. Cluster generation was performed according to the man-ufacturer’s protocols (Illumina), and 36-nt single-end reads were generatedon a Solexa Genome Analyzer at the Institute for Genomics and SystemsBiology (Argonne National Laboratory, Argonne, IL). The Solexa reads thatpassed the purity filtering and had a unique alignment were mapped to theY. pseudotuberculosis IP32953 reference chromosome (NC_006155.1) andplasmids (pYV, NC_006154; pYtb, NC_006153).

Bioinformatics. Solexa sequencing reads that overlapped the annotatedmiscellaneous RNA, mRNA, rRNA, tmRNA, and tRNA genes (based on theabove GenBank records of Y. pseudotuberculosis IP32953 chromosome andplasmids) were extracted and counted. The reads not overlapping theseannotations were considered to be intergenic and include 5′ and 3′ UTRs, cis-encoded antisense RNAs, and potential trans-encoded sRNAs. Clusters of atleast 50 bp in length that form a continuous region of coverage wereextracted from the intergenic category of reads on each strand of thechromosome or plasmids. These clusters were analyzed further by calculat-ing the expression level in reads per kilobase for each intergenic cluster andall miscellaneous RNA, mRNA, rRNA, tmRNA, and tRNA genes. To aid ineliminating 5′ and 3′ UTRs from the analysis of intergenic clusters, intergenicclusters differentially expressed with respect to nearby ORFs were de-termined. The candidate clusters were required to have a difference in ex-pression level above or below a threshold of at least threefold with respectto any ORF within 1 kb on the same strand. Generated cluster sets then wereanalyzed using the Integrated Genome Browser (http://www.bioviz.org/igb/).Predicted sRNAs were inspected for the presence of promoters and ρ-independent terminators using the BProm and TermFind/RNAFold programs(Softberry).

Animal Experiments. All procedures involving animals were carried out incompliance with protocols approved by the Northwestern University in-stitutional animal care and use committee. For Y. pseudotuberculosisinfections, 8-wk-old female BALB/c mice were purchased from Harlan Lab-oratories and allowed to acclimate in the animal facility for 5–7 d beforeinfection. To prepare the inocula, Y. pseudotuberculosis strains were cul-tured overnight in BHI broth at 26 °C, diluted to an OD620 of 0.1 in BHI broth,and incubated at 26 °C with shaking (250 rpm) to an OD620 of 0.6. The cellswere harvested by centrifugation, washed once with sterile PBS, and dilutedto the appropriate OD620 in PBS. Groups of 10 mice were inoculated intra-

Koo et al. PNAS | September 13, 2011 | vol. 108 | no. 37 | E715

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

9, 2

021

gastrically using a 22-gauge feeding needle with ∼2 × 105 cfu of Y. pseu-dotuberculosis and monitored for 21 d. The weights of individual mice wererecorded every third day as an indicator of health status. Input colony-forming units were determined by serial plating onto BHI agar. Animalsinfected with wild-type Y. pseudotuberculosis that demonstrated weightloss at any point during the infection were included in the analysis. For Y.pestis infections, pathogen-free 6- to 8-wk-old female C57BL/6 mice wereobtained from the Jackson Laboratory and allowed to acclimate as above.Bacteria were grown in BHI broth as described in SI Appendix, SI Materialsand Methods, washed once with sterile PBS, and maintained at 37 °C. Groupsof 10 mice were anesthetized lightly with ketamine and xylazine and in-oculated by the intranasal route with wild-type, fully virulent Y. pestis or Ysrdeletion strains (1 × 104 cfu) diluted in PBS as described previously (52). Micewere monitored twice daily for 7 d. All animal infections were repeated atleast twice, unless stated otherwise. Statistical analysis was performed usingthe log-rank (Mantel–Cox) test. The weights of infected animals werecompared using the student’s unpaired t test. A P value of ≤0.05 wasconsidered significant.

sRNA Target Identification. Overnight cultures of wild-type and ΔYsr29 Y.pseudotuberculosis were diluted to an OD620 of 0.1 in BHI broth and grownto stationary phase (OD620 ∼7, 11 h) at 26 °C. Ten OD equivalents of culturewere collected, bacteria were centrifuged, washed once in ice-cold PBS, andincubated with lysozyme (10 mg/mL) for 30 min on ice. Cells were sonicated(three 20-s pulses) on ice, and whole-cell lysates were clarified by centrifu-gation. Protein content of the lysates was quantified by the Bradford assay(Bio-Rad). The wild-type sample was labeled with Cy3 dye, and the ΔYsr29sample was labeled with Cy5 dye using the CyDye DIGE Fluor Labeling Kit

(GE Healthcare) according to the manufacturer’s instructions. Both sampleswere run in a single polyacrylamide gel but were imaged separately ata wavelength of 570 nm for Cy3 and at a wavelength of 670 nm for Cy5.Protein spots were selected based on >1.5-fold differential expression be-tween the wild-type and ΔYsr29 mutant and were subjected to robotic spotexcision, trypsin digestion, and MALDI-TOF. Protein identification was basedon highly accurate masses and MS/MS sequence data from multiple trypticpeptides. The 2D-DIGE and subsequent analyses were performed at theW. M. Keck Foundation Biotechnology Resource Laboratory (Yale University,New Haven, CT).

Additional Methods. Bacterial strains and plasmids (SI Appendix, Table S3),oligonucleotides (SI Appendix, Table S4), and growth conditions andreagents, as well as methods for sRNA and hfq deletions, Northern blot, 5′/3′RACE, qRT-PCR, sRNA target verification, and Ysr29 overexpression are de-tailed in SI Appendix, SI Materials and Methods.

ACKNOWLEDGMENTS. We thank Brian Fritz for guidance and technicalsupport and Marc Domanus for Illumina-Solexa sequencing. We also thankTerence Wu for assistance with 2D-DIGE and mass spectroscopy analysis. Wethank Lauren Bellows for outstanding technical assistance and Hank Seifertfor helpful discussions. This work was sponsored by the Northwestern Uni-versity Feinberg School of Medicine and the National Institutes of Health/National Institute of Allergy and Infectious Diseases Regional Center of Ex-cellence (RCE) Research Program for Bio-defense and Emerging InfectiousDiseases. The authors acknowledge membership within and support fromthe Region V ‘Great Lakes’ RCE (National Institutes of Health Award U54-AI-057153).

1. Waters LS, Storz G (2009) Regulatory RNAs in bacteria. Cell 136:615–628.2. Sharma CM, Vogel J (2009) Experimental approaches for the discovery and

characterization of regulatory small RNA. Curr Opin Microbiol 12:536–546.3. Gottesman S, Storz G (2010) Bacterial small RNA regulators: Versatile roles and rapidly

evolving variations. Cold Spring Harb. Perspect Biol, 10.1101/cshperspect.a003798.4. Gottesman S (2004) The small RNA regulators of Escherichia coli: Roles and

mechanisms. Annu Rev Microbiol 58:303–328.5. Vogel J, Papenfort K (2006) Small non-coding RNAs and the bacterial outer

membrane. Curr Opin Microbiol 9:605–611.6. McCullen CA, Benhammou JN, Majdalani N, Gottesman S (2010) Mechanism of

positive regulation by DsrA and RprA small noncoding RNAs: Pairing increasestranslation and protects rpoS mRNA from degradation. J Bacteriol 192:5559–5571.

7. Jousselin A, Metzinger L, Felden B (2009) On the facultative requirement of thebacterial RNA chaperone, Hfq. Trends Microbiol 17:399–405.

8. Huntzinger E, et al. (2005) Staphylococcus aureus RNAIII and the endoribonuclease IIIcoordinately regulate spa gene expression. EMBO J 24:824–835.

9. Morfeldt E, Taylor D, von Gabain A, Arvidson S (1995) Activation of alpha-toxintranslation in Staphylococcus aureus by the trans-encoded antisense RNA, RNAIII.EMBO J 14:4569–4577.

10. Lenz DH, et al. (2004) The small RNA chaperone Hfq and multiple small RNAs controlquorum sensing in Vibrio harveyi and Vibrio cholerae. Cell 118:69–82.

11. Achtman M, et al. (1999) Yersinia pestis, the cause of plague, is a recently emergedclone of Yersinia pseudotuberculosis. Proc Natl Acad Sci USA 96:14043–14048.

12. Skurnik M, Peippo A, Ervelä E (2000) Characterization of the O-antigen gene clustersof Yersinia pseudotuberculosis and the cryptic O-antigen gene cluster of Yersiniapestis shows that the plague bacillus is most closely related to and has evolved fromY. pseudotuberculosis serotype O:1b. Mol Microbiol 37:316–330.

13. Perry RD, Fetherston JD (1997) Yersinia pestis—etiologic agent of plague. ClinMicrobiol Rev 10:35–66.

14. Mayer L, Greenstein AJ (1976) Acute yersinial ileitis: A distinct entity. Am JGastroenterol 65:548–551.

15. Chain PS, et al. (2004) Insights into the evolution of Yersinia pestis through whole-genome comparison with Yersinia pseudotuberculosis. Proc Natl Acad Sci USA 101:13826–13831.

16. Geng J, et al. (2009) Involvement of the post-transcriptional regulator Hfq in Yersiniapestis virulence. PLoS ONE 4:e6213.

17. Schiano CA, Bellows LE, Lathem WW (2010) The small RNA chaperone Hfq is requiredfor the virulence of Yersinia pseudotuberculosis. Infect Immun 78:2034–2044.

18. Srivatsan A, et al. (2008) High-precision, whole-genome sequencing of laboratorystrains facilitates genetic studies. PLoS Genet 4:e1000139.

19. Liu JM, et al. (2009) Experimental discovery of sRNAs in Vibrio cholerae by directcloning, 5S/tRNA depletion and parallel sequencing. Nucleic Acids Res 37:e46,10.1093/nar/gkp080.

20. Nicol JW, Helt GA, Blanchard SG, Jr., Raja A, Loraine AE (2009) The Integrated GenomeBrowser: Free software for distribution and exploration of genome-scale datasets.Bioinformatics 25:2730–2731.

21. Gottesman S (2005) Micros for microbes: Non-coding regulatory RNAs in bacteria.Trends Genet 21:399–404.

22. Straley SC, Perry RD (1995) Environmental modulation of gene expression andpathogenesis in Yersinia. Trends Microbiol 3:310–317.

23. Sittka A, et al. (2008) Deep sequencing analysis of small noncoding RNA and mRNAtargets of the global post-transcriptional regulator, Hfq. PLoS Genet 4:e1000163.

24. Gasteiger E, et al. (2003) ExPASy: The proteomics server for in-depth proteinknowledge and analysis. Nucleic Acids Res 31:3784–3788.

25. Busch A, Richter AS, Backofen R (2008) IntaRNA: Efficient prediction of bacterial sRNAtargets incorporating target site accessibility and seed regions. Bioinformatics 24:2849–2856.

26. Massé E, Gottesman S (2002) A small RNA regulates the expression of genes involvedin iron metabolism in Escherichia coli. Proc Natl Acad Sci USA 99:4620–4625.

27. Bohn C, et al. (2010) Experimental discovery of small RNAs in Staphylococcus aureusreveals a riboregulator of central metabolism. Nucleic Acids Res 38:6620–6636.

28. Kumar R, et al. (2010) Identification of novel non-coding small RNAs fromStreptococcus pneumoniae TIGR4 using high-resolution genome tiling arrays. BMCGenomics, 11:350, 10.1186/1471-2164-11-350.

29. Postic G, et al. (2010) Identification of small RNAs in Francisella tularensis. BMCGenomics, 11:625, 10.1186/1471-2164-11-625.

30. Yoder-Himes DR, et al. (2009) Mapping the Burkholderia cenocepacia niche responsevia high-throughput sequencing. Proc Natl Acad Sci USA 106:3976–3981.

31. Livny J, Brencic A, Lory S, Waldor MK (2006) Identification of 17 Pseudomonasaeruginosa sRNAs and prediction of sRNA-encoding genes in 10 diverse pathogensusing the bioinformatic tool sRNAPredict2. Nucleic Acids Res 34:3484–3493.

32. Wassarman KM, Repoila F, Rosenow C, Storz G, Gottesman S (2001) Identification ofnovel small RNAs using comparative genomics and microarrays. Genes Dev 15:1637–1651.

33. Wadler CS, Vanderpool CK (2009) Characterization of homologs of the small RNASgrS reveals diversity in function. Nucleic Acids Res 37:5477–5485.

34. Horler RS, Vanderpool CK (2009) Homologs of the small RNA SgrS are broadlydistributed in enteric bacteria but have diverged in size and sequence. Nucleic AcidsRes 37:5465–5476.

35. Massé E, Vanderpool CK, Gottesman S (2005) Effect of RyhB small RNA on global ironuse in Escherichia coli. J Bacteriol 187:6962–6971.

36. Padalon-Brauch G, et al. (2008) Small RNAs encoded within genetic islands ofSalmonella typhimurium show host-induced expression and role in virulence. NucleicAcids Res 36:1913–1927.

37. Bai G, Golubov A, Smith EA, McDonough KA (2010) The importance of the small RNAchaperone Hfq for growth of epidemic Yersinia pestis, but not Yersiniapseudotuberculosis, with implications for plague biology. J Bacteriol 192:4239–4245.

38. Mohanty BK, Maples VF, Kushner SR (2004) The Sm-like protein Hfq regulatespolyadenylation dependent mRNA decay in Escherichia coli. Mol Microbiol 54:905–920.

39. Brennan RG, Link TM (2007) Hfq structure, function and ligand binding. Curr OpinMicrobiol 10:125–133.

40. Figueroa-Bossi N, et al. (2006) Loss of Hfq activates the sigmaE-dependent envelopestress response in Salmonella enterica. Mol Microbiol 62:838–852.

41. Pulvermacher SC, Stauffer LT, Stauffer GV (2009) Role of the Escherichia coli Hfqprotein in GcvB regulation of oppA and dppA mRNAs. Microbiology 155:115–123.

42. Wower J, Zwieb CW, Hoffman DW, Wower IK (2002) SmpB: A protein that binds todouble-stranded segments in tmRNA and tRNA. Biochemistry 41:8826–8836.

43. Allocati N, Federici L, Masulli M, Di Ilio C (2009) Glutathione transferases in bacteria.FEBS J 276:58–75.

E716 | www.pnas.org/cgi/doi/10.1073/pnas.1101655108 Koo et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

9, 2

021

44. Hayes JD, Flanagan JU, Jowsey IR (2005) Glutathione transferases. Annu RevPharmacol Toxicol 45:51–88.

45. Shin S, Lu G, Cai M, Kim KS (2005) Escherichia coli outer membrane protein A adheresto human brain microvascular endothelial cells. Biochem Biophys Res Commun 330:1199–1204.

46. Prasadarao NV, et al. (1996) Outer membrane protein A of Escherichia coli contributesto invasion of brain microvascular endothelial cells. Infect Immun 64:146–153.

47. Selvaraj SK, Prasadarao NV (2005) Escherichia coli K1 inhibits proinflammatorycytokine induction in monocytes by preventing NF-kappaB activation. J Leukoc Biol78:544–554.

48. Rasmussen AA, et al. (2005) Regulation of ompA mRNA stability: The role of a smallregulatory RNA in growth phase-dependent control. Mol Microbiol 58:1421–1429.

49. Altuvia S, Weinstein-Fischer D, Zhang A, Postow L, Storz G (1997) A small, stable RNAinduced by oxidative stress: Role as a pleiotropic regulator and antimutator. Cell 90:43–53.

50. Soper T, Mandin P, Majdalani N, Gottesman S, Woodson SA (2010) Positiveregulation by small RNAs and the role of Hfq. Proc Natl Acad Sci USA 107:9602–9607.

51. Papenfort K, et al. (2006) SigmaE-dependent small RNAs of Salmonella respond tomembrane stress by accelerating global omp mRNA decay. Mol Microbiol 62:1674–1688.

52. Lathem WW, Crosby SD, Miller VL, Goldman WE (2005) Progression of primary pneu-monic plague: A mouse model of infection, pathology, and bacterial transcriptionalactivity. Proc Natl Acad Sci USA 102:17786–17791.

Koo et al. PNAS | September 13, 2011 | vol. 108 | no. 37 | E717

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

9, 2

021