Temperature-responsive in vitro RNA structurome ofYersinia pseudotuberculosisFrancesco Righettia, Aaron M. Nussb, Christian Twittenhoffa, Sascha Beelea, Kristina Urbana, Sebastian Willc,d,Stephan H. Bernhartc,d, Peter F. Stadlerc,d,e,f,g, Petra Derschb, and Franz Narberhausa,1

aDepartment of Microbial Biology, Ruhr University Bochum, 44801 Bochum, Germany; bDepartment of Molecular Infection Biology, Helmholtz Centre forInfection Research, 38124 Braunschweig, Germany; cDepartment of Computer Science, University of Leipzig, 04107 Leipzig, Germany; dInterdisciplinaryCenter for Bioinformatics, University of Leipzig, 04107 Leipzig, Germany; eMax Planck Institute Mathematics in the Sciences, 04103 Leipzig, Germany; fSantaFe Institute, Santa Fe, NM 87501; and gFraunhofer Institute Cell Therapy and Immunology, 04103 Leipzig, Germany

Edited by Gisela Storz, National Institutes of Health, Bethesda, MD, and approved May 12, 2016 (received for review November 20, 2015)

RNA structures are fundamentally important for RNA function. Dy-namic, condition-dependent structural changes are able to modulategene expression as shown for riboswitches and RNA thermometers.By parallel analysis of RNA structures, we mapped the RNA structur-ome of Yersinia pseudotuberculosis at three different temperatures.This human pathogen is exquisitely responsive to host body tempera-ture (37 °C), which induces a major metabolic transition. Our analysisprofiles the structure of more than 1,750 RNAs at 25 °C, 37 °C, and42 °C. Average mRNAs tend to be unstructured around the ribosomebinding site. We searched for 5′-UTRs that are folded at low tempera-ture and identified novel thermoresponsive RNA structures from di-verse gene categories. The regulatory potential of 16 candidates wasvalidated. In summary, we present a dynamic bacterial RNA structur-ome and find that the expression of virulence-relevant functionsin Y. pseudotuberculosis and reprogramming of its metabolismin response to temperature is associated with a restructuring ofnumerous mRNAs.

RNA structure | RNA thermometer | temperature | virulence | translationalcontrol

RNA structures play a pivotal role in the function of non-coding RNAs (ncRNAs) comprised of rRNAs, tRNAs, and

small regulatory RNAs (sRNAs) and in the expression of protein-coding mRNAs. Structured segments affect the entire RNA lifecycle from transcription, maturation, and translation to degrada-tion and determine the specificity of interactions with other RNAs,proteins, or ligands. Although some RNAs, such as ribozymes andrRNAs, adopt rather stable secondary and tertiary structures, manyregulatory RNA elements are dynamic and undergo structural rear-rangements in the physiological temperature range. Well-known ex-amples are metabolite-sensing riboswitches (1) and temperature-sensing RNA thermometers (RNATs) (2). Bacterial RNATs areusually located in the 5′-UTR or the intercistronic region (ICR) of anmRNA and differentially control translation of the downstream ORFin response to temperature (3). Typically, an RNAT folds into astructure that occludes the ribosome binding site (RBS). A temper-ature upshift liberates the RBS and allows the ribosome to bind andinitiate translation. Structurally diverse RNATs are located upstreamof many bacterial heat shock and virulence genes. ThermosensitiveRNA structures playing a role in infection and host adaptation pro-cesses have been documented in Listeria monocytogenes (4), Yersiniapestis (5), Yersinia pseudotuberculosis (6), Leptospira interrogans (7),Shigella dysenteriae (8), Neisseria meningitidis (9), Pseudomonas aeru-ginosa (10), and Vibrio cholerae (11).In contrast to ligand-binding riboswitches, RNATs show poor, if

any, conservation in sequence and structure. This diversity has ham-pered the in silico identification of novel RNATs, but recently de-veloped genome-wide RNA structure-probing approaches offer newopportunities (12). Global structure probing maps the structures ofthe entire pool of expressed RNA molecules and provides a snapshotof the RNA structurome (13). Probing at different temperaturesprofiles dynamic RNA structures and revealed a large number ofRNAT candidates in yeast (14).

Global RNA structure-probing approaches have been usedwith yeast cells (13, 15), Drosophila melanogaster and Caenorhabditiselegans (16),Arabidopsis thaliana (17–19), mouse (20–22), human cells(23, 24), and, most recently, with Escherichia coli (25). We appliedthe parallel analysis of RNA structure (PARS) and experimen-tally determined single- and double-stranded regions in theY. pseudotuberculosis YPIII transcriptome at three differenttemperatures. Y. pseudotuberculosis is a foodborne pathogen closelyrelated to Y. pestis and causes enteritis, diarrhea, and mesentericlymphadenitis in animals and humans. Its metabolism and virulenceare strictly temperature controlled by various posttranscriptionalmechanisms (6, 26). An intercistronic RNAT is present upstream oflow calcium response gene F (lcrF) coding for the master regulator ofYersinia virulence (6). Temperature and nutrient changes experiencedduring the early stage of infection, when the bacterium enters a warm-blooded host, are the major cues of a complex gene expression cas-cade inducing not only virulence factors but also a metabolic switch toadapt to the host environment (27). A recent transcriptome studyrevealed 324 differentially regulated Y. pseudotuberculosis geneswhen cells were grown at 25 °C or 37 °C (28). Apart from theexpected arsenal of virulence-associated genes, almost half of thetemperature-regulated genes are linked to metabolic functions in-cluding enzymes of the energy metabolism, β-oxidation of fatty acids,and sugar and amino acid transporters. To address whether tem-perature-sensitive RNA structures are involved in this profoundmetabolic shift, we treated RNA with nuclease S1 and RNase V1 atthree different temperatures reflecting environmental (25 °C), viru-lence (37 °C), and heat shock (42 °C) conditions. We found that theY. pseudotuberculosis genome encodes a plethora of RNATs.

Significance

The RNA structure is critical for RNA function in all domains oflife. We determined the transcriptome-wide RNA structurome ofYersinia pseudotuberculosis, a food-borne pathogen, at threephysiologically relevant temperatures. Our analysis shows thatmRNAs tend to have a poorly structured ribosome binding site.Transcripts that deviate from this general principle are very goodcandidates as translational repressor elements, andwe identified16 RNA thermometers able to control gene expression in atemperature-dependent manner. Our analysis demonstrates thepower of high-throughput RNA structure probing approaches toidentify new sensory and regulatory RNA structures.

Author contributions: F.R., A.M.N., P.D., and F.N. designed research; F.R., A.M.N., C.T., S.B.,and K.U. performed research; F.R., A.M.N., C.T., S.W., S.H.B., P.F.S., P.D., and F.N. analyzeddata; and F.R., S.W., P.F.S., P.D., and F.N. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The sequences reported in this paper have been deposited at the home-page of the Bioinformatics Institute at the University of Leipzig, www.bioinf.uni-leipzig.de/publications/supplements/16-001.1To whom correspondence should be addressed. Email: franz.narberhaus@rub.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523004113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1523004113 PNAS | June 28, 2016 | vol. 113 | no. 26 | 7237–7242

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 3,

202

0

ResultsThe Y. pseudotuberculosis Transcriptome and RNA Structurome. Tocatalog RNA structures of the Y. pseudotuberculosis transcriptome,we performed PARS (13) as illustrated in Fig. 1. Total RNA wasisolated from a Y. pseudotuberculosis YPIII culture grown in LBmedium at 37 °C to an OD600nm of 0.5 (exponential phase). Ali-quots of rRNA-depleted total RNA were refolded in vitro at 25 °C,37 °C, or 42 °C. RNA samples were then partially digested eitherwith nuclease S1 or RNase V1, enzymes that preferentially cleavephosphodiester bonds 3′ of single-stranded (ss) and double-stranded(ds) nucleotides, respectively. Before this treatment, pilot experi-ments had been conducted to reach single-hit kinetics at the dif-ferent temperatures. To prevent capture of degradation products,our library synthesis protocol was optimized for fragments resultingfrom nucleases digestion (Materials and Methods). From each li-brary, between 7.6 and 12.7 Mio cDNA reads were generated andmapped to the YPIII chromosome (NC_010465) and the pYVvirulence plasmid (NC_006153). Mapping statistics are detailed inDataset S1. Each read is supposed to represent a S1 or V1 cleavageproduct. The PARS score, defined as the log ratio between the V1and the S1 events, was calculated for each nucleotide of the tran-scriptome of Y. pseudotuberculosis. A value above zero indicatesthat the nucleotide is base-paired, and a score below zero repre-sents an unpaired nucleotide.For our analysis, we used 5′ ends of 1,642 protein-coding RNAs,

1,614 mapped on the chromosome and 28 on the virulence plasmidpYV (Dataset S2). This number includes the 1151 previouslyreported transcriptional start sites (TSSs) (28) and 5′ ends of 498transcripts identified by manual revision of pilot RNAseq experi-ments (SI Appendix). By using the MicrobesOnline resource (29), weidentified 1,046 monocistronic RNAs and 568 polycistronic operonson the chromosome and 14 monocistronic and 14 polycistronic op-erons on plasmid pYV. 5′-UTRs have a median length of 73 nt andcomprise almost 168,000 nt on the chromosome and about 2,400 onthe pYV plasmid. Eight hundred ninety-three ICRs account forabout 58,000 nt of the chromosome, with a median length of 34 nt.To evaluate the accuracy of our approach, we first examined

the PARS profiles of well-studied RNA molecules. Our PARS

analysis recapitulates ds and ss nucleotides of the 5S rRNA andsuggests a thermostable structure (SI Appendix, Fig. S1). Also thePARS scores obtained for the 84 tRNAs at 25 °C match theexpected secondary structures. Almost all tRNAs have an averagePARS score above zero (from 1.75 to −0.35, with a mean of 0.76)indicating the prevalence of stable structures (SI Appendix, Fig. S1).

The sRNA Structurome. The overall architecture is critical for sRNAfunction and target recognition (30). However, RNA usually adoptsmany different structures at physiological temperatures. We ex-amined the structural ensembles of 19 (SI Appendix, Fig. S2) of 78previously described sRNAs in Y. pseudotuberculosis YPIII (28),which were sufficiently abundant for PARS profiling. They arecharacterized by stable secondary structures at 25 °C, with averagePARS profiles ranging from 1.56 to −0.26 and a mean of 0.28(Datasets S3 and S4). The CsrB and CsrC RNAs are functionallyrelevant for virulence (31). They are protein-binding sRNAsthat interact with the CsrA regulator via GGA motives (32). ThePARS-derived structures of CsrB and CsrC at virulence tem-perature suggest a complex architecture exposing multiple GGAsequences (Fig. 2).

The mRNA Structurome Reveals Numerous RNATs. Global RNAstructurome studies showed that the region upstream of the startcodon is characterized by low structural stability (13, 14, 25). Wealigned 1,642 Y. pseudotuberculosis mRNAs at the translationalstart site and calculated the average PARS score per nucleotide. Ingeneral, we observed little differences in the folding of 5′-UTRsand their downstream ORFs (PARS scores of −0.004 and 0.01,respectively; Fig. 3A) and the folding of ICRs and their downstreamORFs (−0.008 and 0.01, respectively; Fig. 3B). Strikingly, bothclasses of genes show a clear local minimum around the Shine–Dalgarno (SD) sequence (−10 ± 4 nt from the start codon). Theyexhibit an average PARS score of −0.035 (P < 0.01) in 5′-UTRsand −0.038 (P < 0.01) in ICRs, indicating that accessibility of theRBS is a typical feature of protein-coding transcripts.Translational control elements, like RNATs, are expected to

deviate from this general trend as they block the RBS by basepairing. To identify temperature-responsive RNA structures, wecompared the PARS profiles of all SD regions (−10 ± 4 from thestart codon) at 25 °C, 37 °C, and 42 °C and generated a hierarchicallist of UTRs and ICRs, in which the RBS is folded at 25 °C butdestabilized at 37 °C and 42 °C (ranked according to the PARSscore differences between 25 °C and 37 °C in Dataset S5). Seven-teen candidate UTRs and three ICRs located upstream of genescoding for a variety of different cellular functions were selected fromthat list for further analysis by reporter gene fusions (Table 1,ranked according to gene categories and PARS differences; sodAwas selected as negative control). A strikingly high number of5′-UTRs with temperature-sensitive structures are involved in pro-tection against oxidative stress (grxC, ahpC, trxA, katA, sodB, andsodC). Other candidates are positioned upstream of genes for sugar(manX) or oligopeptide (oppA) uptake, amino acid biosynthesis orutilization (cysK-2, putA, and pepN), other metabolic processes(fdoG-1 and aldB), or the heat shock response (yobF). Two candi-dates are associated with Yersinia virulence: ailA (attachment in-vasion locus protein) and cnfY (cytotoxic necrotizing factor). Three

Fig. 1. Overview of the PARS approach. Y. pseudotuberculosis was cultivatedat 37 °C, RNA was isolated, and rRNA was depleted and in vitro refolded at25 °C, 37 °C, and 42 °C. RNA samples were treated with S1 or V1 nucleases thatcleave single- or double-stranded RNA, respectively. After library preparationand Illumina sequencing, the count of the cleavage events generated S1 andV1 profiles. The PARS score per each nucleotide is calculated as the log ratio ofV1 cuts/S1 cuts.

Fig. 2. PARS-derived structures of CsrB and CsrC sRNAs at 37 °C. Red regionsdepict exposed GGA motifs.

7238 | www.pnas.org/cgi/doi/10.1073/pnas.1523004113 Righetti et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 3,

202

0

representative ICRs derive from upstream regions of iscS-1 (cysteinedesulfurase), cpoB/ybgF (tol/pal system protein), and dnaJ (chap-erone). The PARS profiles of these candidates measured at 25 °C,37 °C, and 42 °C are shown in Dataset S6, and the PARS-drivenstructures are shown in SI Appendix, Fig. S3.Each region was cloned as translational fusion to bgaB encoding a

heat-stable β-galactosidase. For two genes (trxA and manX), fusionsto a short and long UTR were constructed due to two alternativeTSSs (28). In the bgaB reporter gene system (33), transcription fromthe PBAD promoter was induced by arabinose, and β-galactosidaseactivity was measured at 25 °C, 37 °C, and 42 °C. The ICR betweenyscW and lcrF, which contains a well-characterized 4U thermometer(6), and sodA served as positive and negative control, respectively.Because gene expression generally increases with temperature, onlya more than threefold induction was considered a functional RNAT.Except for grxC, ahpC, and aldB and the long fusions to trxA andmanX, all selected 5′-UTRs and the ICRs were able to represstranslation at 25 °C compared with 37 °C and/or 42 °C (Fig. 4;corresponding Western blots showing the His-tagged BgaB proteinin SI Appendix, Fig. S4A).To validate the PARS-guided predictions of virulence-related

transcripts, point mutations designed to stabilize the putative RNATswere introduced into the cnfY and ailA 5′-UTRs (Fig. 5). Dependingon individual or combined mutations, gene expression was either lesstemperature responsive or fully repressed, respectively (Fig. 5; cor-responding Western blots showing the His-tagged BgaB protein in SIAppendix, Fig. S4B). In vitro-transcribed ailA RNA was subjected tostructure probing and toeprinting analysis. Consistent with the tem-perature regulation in vivo, the nucleotides in the SD sequence—inparticular G53 and G54, which are predicted to be paired with Cresidues—were not accessible to the S1 and T1 nucleases (the lattercuts at single-stranded Gs). Cleavage of these residues at 37 °C and42 °C (Fig. 6 A and B) suggests heat-induced destabilization of thestructure. Accordingly, ribosome binding to the RNA was inefficientat 25 °C but increased with increasing temperature (Fig. 6C). Overall,this study shows that PARS profiling is an excellent tool to identifytranslational repressor elements.

DiscussionWe report the genome-scale landscape of RNA structures of thehuman pathogen Y. pseudotuberculosis at three physiologically rel-evant temperatures reflecting environmental (25 °C), host body(37 °C), and heat shock (42 °C) conditions at single-nucleotideresolution. Comparison of the static snapshots of the RNA struc-turome at different temperatures disclosed the dynamic temperature-responsive reorganization of a bacterial transcriptome. The averageSD sequence in Y. pseudotuberculosis is unfolded regardless ofwhether it is located in the 5′-UTR or an ICR. This finding is fully

consistent with the report that translation efficiency is correlatedwith the accessibility of the ribosome binding site (34). The globalinformation on the 5′-UTR structures was exploited to search forRNA signatures not conforming to this general rule with the ex-pectation that such regions might act as translational repressorelements. We paid particular attention to the region overlappingthe SD sequence and identified numerous structures that aredestabilized at elevated temperature. More than 25,000 thermo-sensitive bases at 1,800 RNA sites were reported in yeast. In thiseukaryotic organism that is not known to contain RNATs astranslational control elements, RNA hairpin structures that melt

Fig. 3. Functional analysis of the average PARS profile across protein-coding transcripts. Average PARS score for each nucleotide across the 5′-UTR and first50 nt of the coding sequence (CDS) (A) and across the ICR and first 50 nt of the downstream CDS (B) of all protein-coding transcripts at 25 °C were aligned atthe start codon. The position of the SD region is indicated, and average (avg) values of PARS scores for 5′-UTR, ICR, CDSs and SD are shown.

Table 1. PARS analysis of the SD sequence of selected RNATcandidates

Genename

SD PARS* PARS difference†Thermalcontrol‡25 °C 37 °C 42 °C 37–25 °C 42–25 °C

cnfY 0.18 −0.37 −0.12 −0.55 −0.30 YailA 0.09 0.61 −0.25 0.52 −0.34 YgrxC 0.39 −0.56 −0.50 −0.95 −0.89 NtrxA 0.32 −0.53 −0.42 −0.85 −0.74 s:Y, l:NahpC −0.60 −1.23 0.44 −0.63 1.04 NkatA 0.08 −0.37 −0.02 −0.45 −0.10 YsodB 0.00 −0.22 −0.29 −0.22 −0.29 YsodC 0.14 −0.08 −0.47 −0.22 −0.61 YsodA 0.09 0.07 0.02 −0.02 −0.07 NoppA 0.64 −0.77 −0.13 −1.41 −0.77 YfdoG-1 0.55 −0.55 0.12 −1.10 −0.43 YpepN 0.36 −0.26 0.05 −0.62 −0.31 YputA 0.48 0.11 −0.05 −0.37 −0.53 YaldB 0.00 −0.27 0.22 −0.27 0.22 NyobF −0.08 0.01 0.02 0.09 0.10 YcysK-2 −0.04 0.07 −0.05 0.11 −0.01 YmanX −0.29 −0.11 −0.12 0.18 0.17 s:Y, l:NcpoB/ybgF 0.86 −0.33 −0.20 −1.19 −1.06 YiscS-1 0.43 0.16 −0.22 −0.27 −0.65 YdnaJ 0.18 −0.05 −0.10 −0.23 −0.28 Y

*Average PARS of the SD region (−14 to −6 from the start codon) measuredat 25 °C, 37 °C, and 42 °C.†Difference in the PARS score of the SD region from 37 °C to 25 °C and 42 °Cto 25 °C, respectively.‡Experimental confirmation of temperature-dependent translational regulationin reporter gene studies (Fig. 4 and SI Appendix, Fig. S4). N, no thermal control;s and l, short and long variants of the 5′-UTR, respectively; Y, thermal control.

Righetti et al. PNAS | June 28, 2016 | vol. 113 | no. 26 | 7239

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 3,

202

0

between 30 °C and 37 °C were shown to be selectively degraded bythe exosome, whereas relatively stable structures were not (14).Our follow-up experiments on the PARS-derived tempera-

ture-labile structures of Y. pseudotuberculosis demonstrated thatmost, but not all, of them are capable of controlling translation

in a temperature-dependent manner. Only the short forms of twoalternative transcripts of trxA andmanX were temperature controlled.This observation is reminiscent of the pqsA transcript of P. aeruginosawhere only one of two alternative mRNA isoforms folds into atranslation-inhibiting structure (35). Like lcrF (6), the ailA and cnfYgenes are virulence related. In addition to thermoregulated tran-scription, they are under RNAT control. The cytotoxic necrotizingfactor CNFγ acts as toxin and is an essential virulence determinant,as it modulates inflammatory responses and counteracts attack byinnate immune effectors (36, 37). Similar to YadA, the adhesin AilApromotes binding to and subsequent killing of neutrophils via type IIIsecretion-mediated translocation of effector proteins into the targethost cell and promotes resistance against complement-mediatedkilling (38–41).The response to oxidative stress seems to be another important

trait that is, at least in part, regulated by RNATs. Although theglutaredoxin gene grxC and the alkyl hydroperoxide reductase geneahpC were in the top list of the temperature-responsive RNAs, theydid not differentially control reporter gene expression pointing outthat structural rearrangements in the 5′-UTR are not necessarilyassociated with translational control. Alternatively, the original se-quence context including the ORF might be crucial, which is absentin the bgaB fusion. Consistent with their PARS profiles, thioredoxin,catalase, and superoxide dismutases were regulated in a tempera-ture-dependent fashion, and several nutrient uptake and utilizationsystems were controlled by temperature. Overall, the RNA struc-turome revealed that Y. pseudotuberculosis coordinates a wide va-riety of cellular processes by mRNA-intrinsic structural information.To rapidly induce metabolic pathways, the pathogen seems to fullyexploit the potential of RNATs that propagate a temperature signalimmediately into a translational response.We foresee multiple future applications for RNA structuromics.

PARS profiling or other genome-scale strategies such as FragSeq (20)or SHAPE-Seq (42) can easily be extended to other physiologicallyrelevant conditions and temperatures. RNA preparations fromcells grown in minimal media might reveal novel riboswitches,which were underrepresented in our dataset presumably becauseYersinia was grown in rich medium. The comparison of RNAstructuromes from different bacterial strains, for example clinicalisolates, might uncover functionally important differences in RNAstructures, so called riboSNitches (24). A recent report on aP. aeruginosa small colony variant showed that a point mutation in

Fig. 4. Temperature-responsive RNA structures confer temperature-dependent reporter gene expression. The RNA elements located within the 5′-UTRs or ICRsupstream the genes indicated on the x axis were translationally fused to the bgaB reporter gene, and temperature-dependent gene expression was measured asdescribed in Materials and Methods (mean ± standard deviation; n = 3). β-Galactosidase activity values were normalized to the value measured for lcrF at 25 °C.

Fig. 5. Point mutations within the 5′-UTRs of cnfY and ailA impair RNATfunctionality. WT and mutated 5′-UTRs (SI Appendix) were translationallyfused to the bgaB reporter gene, and gene expression was measured asdescribed in Fig. 4 (mean ± standard deviation; n = 3). Proposed schematicstructures and point mutations are displayed.

7240 | www.pnas.org/cgi/doi/10.1073/pnas.1523004113 Righetti et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 3,

202

0

the 5′-UTR of a fatty acid biosynthesis gene is sufficient to cause astable RNA structure with consequence on membrane composi-tion and multicellular behavior (43).Our ultimate goal will be to correlate the in vitro structurome

with the structures as they exist in the living cell. In the crowdedintracellular environment, where an RNAmolecule encounters otherRNAs, proteins, ions, and ligands, its structure might adopt a con-formation different from the one mapped in vitro. The membrane-permeable, adenine- and cytosine-modifying probe dimethyl sulfatewas used to read out in vivo RNA structures in yeast (23) andA. thaliana (19). A cell-permeable SHAPE reagent that modifiesaccessible 2′-hydroxy groups of all four nucleotides allowed invivo probing of the mouse transcriptome (22). In principal, suchapproaches are also applicable to bacteria. Apart from charting theintracellular RNA structurome at a given condition, in vivo ap-proaches bear the additional potential to address the contribution ofRNA-binding proteins, e.g., Hfq, or sRNAs on a global scale byprofiling RNA structures in the corresponding mutant strains. The in-depth exploration of the currently available and future RNA struc-turome datasets is going to provide unprecedented information on the

functional importance and dynamics of RNA structures in gene reg-ulation and RNA functionality.

Materials and MethodsDetails of methodologies are provided in SI Appendix.

Strains, Oligonucleotides, Plasmids, and Growth Conditions. Strains, oligonu-cleotides, and plasmids are listed in SI Appendix. Y. pseudotuberculosis andE. coli were grown aerobically in LB medium at the indicated temperatures.Media were supplemented with ampicillin (150 μg/mL) if required andL-arabinose [0.01% (wt/vol)] to induce the PBAD promoter. Recombinant DNAwork was performed according to standard protocols.

RNA Sample Preparation, Nuclease Treatment, and cDNA Library Synthesis.Total RNA was isolated from an exponential (OD600 = 0.5) culture ofY. pseudotuberculosis YPIII grown aerobically in LB at 37 °C. Two microgramsof rRNA-depleted RNA was refolded and treated with S1 nuclease (ThermoFisher Scientific) or V1 RNase (Thermo Fisher Scientific), at 25 °C, 37 °C, and42 °C under single-hit kinetics conditions according to the PARS protocol (44).Reactions were stopped by immediate phenol:chloroform:isoamyl alcoholaddition and ethanol precipitation. Strand-specific RNA-seq cDNA librarypreparation and barcode introduction was based on RNA adapter ligation as

Fig. 6. Temperature affects the secondary structureand ribosome binding efficiency of the ailA 5′-UTR.(A) 5′-end-labeled in vitro transcribed RNA wastreated with nucleases S1 and T1 at 25 °C, 37 °C, and42 °C. RNA fragments were separated on an 8%polyacrylamide gel. Lane L, alkaline ladder. CleavedGs, the SD region and the AUG start codon are in-dicated. (B) PARS-derived structure model of the ailA5′-UTR (+30 nt of the coding region). Gs accessible toRNase T1 cleavage already at 25 °C are indicated withwhite arrows, Gs that become accessible at 37 °C and42 °C are marked with black arrows. Nucleotides ofthe SD region and the AUG start codon are colored inblack and gray, respectively. (C ) Temperature-dependent binding of ribosomes to the ailA 5′-UTR.Toeprint analysis was performed with in vitro tran-scribed RNA, at 25 °C, 37 °C, and 42 °C, as described inMaterials and Methods. TGCA indicate the corre-sponding DNA sequencing reactions and the positionof the ATG start codon is indicated. Reverse tran-scription reactions were performed with (+) andwithout (−) 30S ribosomal subunit. The signals corre-sponding to the full-length transcript (FL) and thetoeprint (toe, corresponding to the position +14 to+16 from the AUG) are indicated.

Righetti et al. PNAS | June 28, 2016 | vol. 113 | no. 26 | 7241

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Dec

embe

r 3,

202

0

described previously (28), with some modifications to enrich RNA fragmentsderiving from nuclease cleavage events, as described in SI Appendix.

Sequencing and Data Analysis. Single-end sequencing on the HiSeq2500 andGenome Analyzer IIx followed a standard protocol. All sequenced librarieswere clipped and mapped to the YPIII genome (NC_010465) and the pYVplasmid (NC_006153) using Segemehl (45). In-house scripts were used tocount the cleavage sites per each base, normalize for coverage, and cal-culate the PARS score at each base as the log ratio of V1/S1 numbers ofcuts. PARS profiles of all RNA species were extracted for each of the threeexperiments at the temperatures 25 °C, 37 °C, and 42 °C. All PARS profilesare deposited at www.bioinf.uni-leipzig.de/publications/supplements/16-001. The final analysis of the PARS sums and computations of the averageprofiles were done with Microsoft Excel 2010. Student t tests were computedusing R software (version 3.2.2). Thermodynamically optimal RNA structureswith and without soft-constraints due to the extracted PARS scores were cal-culated as described in SI Appendix. The implementation of this method in theVienna RNA package (46) is available since version 2.2.0. Each PARS profile wastransformed to profiles of pairing probabilities applying the logistic function toeach score. Optimal perturbation vectors were derived from the probabilityvectors (RNApvmin with a τ/σ ratio of 10). The optimal structures under softconstraints were calculated using RNAfold (shape method W), which internallytransformed the perturbation vectors to pseudoenergies. For computed struc-tures of ncRNAs, 5′-UTRs (+30 nt of the CDS) and ICRs (+100 nt of the upstream

region and +30 nt of the CDS), at 25 °C, 37 °C, and 42 °C, see www.bioinf.uni-leipzig.de/publications/supplements/16-001. Secondary structures arevisualized using VARNA applet (version 3.93) (47) and the Vienna RNApackage (46).

β-Galactosidase Activity Assays. E. coli DH5α cells harboring the reporterplasmids were grown to an OD600 of 0.5, reporter gene transcription was in-duced by 0.01% (wt/vol) L-arabinose, and cultures were split and transferred to25 °C, 37 °C, and 42 °C. After 30 min, cell samples were taken and used forgalactosidase assays as described previously (48) and for Western blot analysis(SI Appendix). Standard deviations were calculated from three technical rep-licates, and each experiment was performed at least three times.

In Vitro RNA Methods. Enzymatic structure probing of the in vitro-transcribed and5′-end–radiolabeled YPK_1268 (ailA) 5′-UTR was performed at 25 °C, 37 °C, and42 °C, with S1 and T1 nucleases, according to ref. 44, with minor changes (SIAppendix). Primer extension inhibition (toeprinting) analysis of YPK_1268 (ailA) 5′-UTR was conducted according to ref. 49, with minor modifications (SI Appendix).

ACKNOWLEDGMENTS. We thank Ronny Lorenz for help with bioinformatics.This work was supported by a grant from the German Research Foundation(FN 240/10-1), a fellowship from the Cluster of Excellence RESOLV (EXC 1069)(to F.R.), and support within de.NBI by the German Federal Ministry ofEducation and Research (BMBF; support code 031A538B) (to P.F.S.).

1. Serganov A, Nudler E (2013) A decade of riboswitches. Cell 152(1-2):17–24.2. Kortmann J, Narberhaus F (2012) Bacterial RNA thermometers: Molecular zippers and

switches. Nat Rev Microbiol 10(4):255–265.3. Krajewski SS, Narberhaus F (2014) Temperature-driven differential gene expression by

RNA thermosensors. Biochim Biophys Acta 1839(10):978–988.4. Johansson J, et al. (2002) An RNA thermosensor controls expression of virulence genes

in Listeria monocytogenes. Cell 110(5):551–561.5. Hoe NP, Goguen JD (1993) Temperature sensing in Yersinia pestis: Translation of the

LcrF activator protein is thermally regulated. J Bacteriol 175(24):7901–7909.6. Böhme K, et al. (2012) Concerted actions of a thermo-labile regulator and a unique

intergenic RNA thermosensor control Yersinia virulence. PLoS Pathog 8(2):e1002518.7. Matsunaga J, Schlax PJ, Haake DA (2013) Role for cis-acting RNA sequences in the

temperature-dependent expression of the multiadhesive lig proteins in Leptospirainterrogans. J Bacteriol 195(22):5092–5101.

8. Kouse AB, Righetti F, Kortmann J, Narberhaus F, Murphy ER (2013) RNA-mediatedthermoregulation of iron-acquisition genes in Shigella dysenteriae and pathogenicEscherichia coli. PLoS One 8(5):e63781.

9. Loh E, et al. (2013) Temperature triggers immune evasion by Neisseria meningitidis.Nature 502(7470):237–240.

10. Grosso-Becerra MV, et al. (2014) Regulation of Pseudomonas aeruginosa virulence fac-tors by two novel RNA thermometers. Proc Natl Acad Sci USA 111(43):15562–15567.

11. Weber GG, Kortmann J, Narberhaus F, Klose KE (2014) RNA thermometer controlstemperature-dependent virulence factor expression in Vibrio cholerae. Proc Natl AcadSci USA 111(39):14241–14246.

12. Righetti F, Narberhaus F (2014) How to find RNA thermometers. Front Cell InfectMicrobiol 4:132.

13. Kertesz M, et al. (2010) Genome-wide measurement of RNA secondary structure inyeast. Nature 467(7311):103–107.

14. Wan Y, et al. (2012) Genome-wide measurement of RNA folding energies. Mol Cell48(2):169–181.

15. Talkish J, May G, Lin Y, Woolford JL, Jr, McManus CJ (2014) Mod-seq: High-throughput sequencing for chemical probing of RNA structure. RNA 20(5):713–720.

16. Li F, et al. (2012) Global analysis of RNA secondary structure in two metazoans. CellReports 1(1):69–82.

17. Zheng Q, et al. (2010) Genome-wide double-stranded RNA sequencing reveals thefunctional significance of base-paired RNAs in Arabidopsis. PLoS Genet 6(9):e1001141.

18. Li F, et al. (2012) Regulatory impact of RNA secondary structure across the Arabidopsistranscriptome. Plant Cell 24(11):4346–4359.

19. Ding Y, et al. (2014) In vivo genome-wide profiling of RNA secondary structure revealsnovel regulatory features. Nature 505(7485):696–700.

20. Underwood JG, et al. (2010) FragSeq: Transcriptome-wide RNA structure probingusing high-throughput sequencing. Nat Methods 7(12):995–1001.

21. Incarnato D, Neri F, Anselmi F, Oliviero S (2014) Genome-wide profiling of mouse RNAsecondary structures reveals key features of the mammalian transcriptome. GenomeBiol 15(10):491.

22. Spitale RC, et al. (2015) Structural imprints in vivo decode RNA regulatory mecha-nisms. Nature 519(7544):486–490.

23. Rouskin S, Zubradt M,Washietl S, Kellis M,Weissman JS (2014) Genome-wide probing ofRNA structure reveals active unfolding of mRNA structures in vivo. Nature 505(7485):701–705.

24. Wan Y, et al. (2014) Landscape and variation of RNA secondary structure across thehuman transcriptome. Nature 505(7485):706–709.

25. Del Campo C, Bartholomäus A, Fedyunin I, Ignatova Z (2015) Secondary structureacross the bacterial transcriptome reveals versatile roles in mRNA regulation andfunction. PLoS Genet 11(10):e1005613.

26. Herbst K, et al. (2009) Intrinsic thermal sensing controls proteolysis of Yersinia viru-lence regulator RovA. PLoS Pathog 5(5):e1000435.

27. Heroven AK, Dersch P (2014) Coregulation of host-adapted metabolism and virulenceby pathogenic yersiniae. Front Cell Infect Microbiol 4:146.

28. Nuss AM, et al. (2015) Transcriptomic profiling of Yersinia pseudotuberculosis revealsreprogramming of the Crp regulon by temperature and uncovers Crp as a masterregulator of small RNAs. PLoS Genet 11(3):e1005087.

29. Price MN, Huang KH, Alm EJ, Arkin AP (2005) A novel method for accurate operonpredictions in all sequenced prokaryotes. Nucleic Acids Res 33(3):880–892.

30. Storz G, Vogel J, Wassarman KM (2011) Regulation by small RNAs in bacteria: Ex-panding frontiers. Mol Cell 43(6):880–891.

31. Nuss AM, et al. (2014) A direct link between the global regulator PhoP and the Csr regulonin Y. pseudotuberculosis through the small regulatory RNA CsrC. RNA Biol 11(5):580–593.

32. Vakulskas CA, Potts AH, Babitzke P, Ahmer BM, Romeo T (2015) Regulation of bac-

terial virulence by Csr (Rsm) systems. Microbiol Mol Biol Rev 79(2):193–224.33. Klinkert B, et al. (2012) Thermogenetic tools to monitor temperature-dependent

gene expression in bacteria. J Biotechnol 160(1-2):55–63.34. de Smit MH, van Duin J (1990) Secondary structure of the ribosome binding site determines

translational efficiency: A quantitative analysis. Proc Natl Acad Sci USA 87(19):7668–7672.35. Brouwer S, et al. (2014) The PqsR and RhlR transcriptional regulators determine the level of

Pseudomonas quinolone signal synthesis in Pseudomonas aeruginosa by producing two

different pqsABCDE mRNA isoforms. J Bacteriol 196(23):4163–4171.36. Schweer J, et al. (2013) The cytotoxic necrotizing factor of Yersinia pseudotuberculosis

(CNFY) enhances inflammation and Yop delivery during infection by activation of Rho

GTPases. PLoS Pathog 9(11):e1003746.37. Wolters M, et al. (2013) Cytotoxic necrotizing factor-Y boosts Yersinia effector

translocation by activating Rac protein. J Biol Chem 288(32):23543–23553.38. Felek S, Krukonis ES (2009) The Yersinia pestis Ail protein mediates binding and Yop

delivery to host cells required for plague virulence. Infect Immun 77(2):825–836.39. Tsang TM, Felek S, Krukonis ES (2010) Ail binding to fibronectin facilitates Yersinia

pestis binding to host cells and Yop delivery. Infect Immun 78(8):3358–3368.40. Kolodziejek AM, Hovde CJ, Minnich SA (2012) Yersinia pestis Ail: Multiple roles of a

single protein. Front Cell Infect Microbiol 2:103.41. Paczosa MK, Fisher ML, Maldonado-Arocho FJ, Mecsas J (2014) Yersinia pseudotu-

berculosis uses Ail and YadA to circumvent neutrophils by directing Yop translocationduring lung infection. Cell Microbiol 16(2):247–268.

42. Lucks JB, et al. (2011) Multiplexed RNA structure characterization with selective2′-hydroxyl acylation analyzed by primer extension sequencing (SHAPE-Seq). ProcNatl Acad Sci USA 108(27):11063–11068.

43. Blanka A, et al. (2015) Constitutive production of c-di-GMP is associated with muta-tions in a variant of Pseudomonas aeruginosa with altered membrane composition.

Sci Signal 8(372):ra36.44. Wan Y, Qu K, Ouyang Z, Chang HY (2013) Genome-wide mapping of RNA structure

using nuclease digestion and high-throughput sequencing. Nat Protoc 8(5):849–869.45. Hoffmann S, et al. (2009) Fast mapping of short sequences with mismatches, inser-

tions and deletions using index structures. PLOS Comput Biol 5(9):e1000502.46. Lorenz R, et al. (2011) ViennaRNA Package 2.0. Algorithms Mol Biol 6:26.47. Darty K, Denise A, Ponty Y (2009) VARNA: Interactive drawing and editing of the RNA

secondary structure. Bioinformatics 25(15):1974–1975.48. Gaubig LC, Waldminghaus T, Narberhaus F (2011) Multiple layers of control govern ex-

pression of the Escherichia coli ibpAB heat-shock operon. Microbiology 157(Pt 1):66–76.49. Hartz D, McPheeters DS, Traut R, Gold L (1988) Extension inhibition analysis of

translation initiation complexes. Methods Enzymol 164:419–425.

7242 | www.pnas.org/cgi/doi/10.1073/pnas.1523004113 Righetti et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 3,

202

0