ABSTRACT
Small noncoding RNA (sRNA) molecules are integral components of the regulatory machinery for many bacterial species and are known to posttranscriptionally regulate metabolic and stress-response pathways, quorum sensing, virulence factors, and more. The Yop-Ysc type III secretion system (T3SS) is a critical virulence component for the pathogenic Yersinia species, and the regulation of this system is tightly controlled at each step from transcription to translocation of effectors into host cells. The contribution of sRNAs to the regulation of the T3SS in Yersinia has been largely unstudied, however. Previously, our lab identified a role for the sRNA chaperone protein Hfq in the regulation of components of the T3SS in the gastrointestinal pathogen Yersinia pseudotuberculosis. Here we present data demonstrating a similar requirement for Hfq in the closely related species Yersinia pestis. Through deep sequencing analysis of the Y. pestis sRNA-ome, we found 63 previously unidentified putative sRNAs in this species. We identified a Yersinia-specific sRNA, Ysr141, carried by the T3SS plasmid pCD1 that is required for the production of multiple T3SS proteins. In addition, we show that Ysr141 targets an untranslated region upstream of yopJ to posttranscriptionally activate the synthesis of the YopJ protein. Furthermore, Ysr141 may be an unstable and/or processed sRNA, which could contribute to its function in the regulation of the T3SS. The discovery of an sRNA that influences the synthesis of the T3SS adds an additional layer of regulation to this tightly controlled virulence determinant of Y. pestis.
INTRODUCTION
Yersinia pestis is the causative agent of plague, a disease thought to be responsible for the mortality of over 200 million people throughout history and one that has continued to infect thousands of people worldwide throughout the 21st century (1–3). The plague bacillus lives a dual lifestyle, cycling between an arthropod vector and mammalian hosts, which provides a robust model for studying host-pathogen interactions and the regulation of virulence determinants (3–5). The coordinated production of virulence factors by Y. pestis as it transitions from flea to rat and back again must be stringently regulated in order for this pathogen to be successful in the host (6–9).
One critical virulence determinant under tight regulation in Y. pestis is the Yop-Ysc type III secretion system (T3SS), the genes for which are carried on the plasmid pCD1 (10). This T3SS is responsible for delivery of effector proteins, called Yops, to the cytosol of host macrophages, neutrophils, and dendritic cells, and Yops coordinately work to subvert the innate immune system (11). Regulation of the T3SS occurs at the transcriptional, posttranscriptional, and posttranslational levels (12). Activation of the T3SS is induced at mammalian host temperatures via conformational changes in the promoter regions of many yop-ysc genes (13). In addition, growth at 37°C permits the transcription and translation of LcrF, the master transcriptional activator of T3SS genes. Transcription of lcrF is prohibited at lower temperatures by DNA bends in its promoter and through association with the nucleoid-associated protein YmoA; upon temperature upshift, YmoA is degraded by Clp/Lon proteases to alleviate the transcriptional repression of the yscW-lcrF operon (13, 14). Furthermore, synthesis of the LcrF protein is repressed by a two-stem-loop structure in the transcript that sequesters the ribosome binding site (RBS) at 26°C, which then melts at 37°C as part of a thermosensing mechanism (15). In addition to the posttranscriptional regulation of LcrF, Yersinia modulates the synthesis of effector Yops in a posttranscriptional manner. In conjunction with the secretion chaperone LcrH, the pore-forming protein YopD has been shown to bind the yopK mRNA in the 5′ untranslated region (UTR) of the transcript (16). This binding may repress the translation of YopK, either by promoting the degradation of the yopK transcript or by competing with the ribosome for binding (17). Moreover, the half-lives of the yopH, yopE, and yscB transcripts are longer in a yopD deletion mutant of Y. pestis than the wild type, which implies that the posttranscriptional regulation of secreted effectors by YopD is not only limited to effects on yopK. Binding of the YopD-LcrH complex to target transcripts requires two specific AU-rich regions of mRNA common to many, but not all, yop transcripts, and the distance of the AU-rich regions from the RBS seems to affect the affinity of YopD for the transcript, suggesting a mechanism for a hierarchy of translation (17). Additionally, Y. pestis can degrade extracellular and/or mistargeted T3SS proteins via the activity of the Y. pestis-specific plasminogen activator protease Pla, thus providing a posttranslational layer of regulation to the T3SS in this species (18, 19).
Previous work from our laboratory has shown that the small noncoding RNA (sRNA) chaperone Hfq also contributes to the posttranscriptional regulation of the Y. pseudotuberculosis T3SS (20). sRNAs are important regulatory elements for many bacterial species, are generally 50 to 500 nucleotides (nt) in length, are encoded within intergenic regions, are transcribed from their own promoters, and contain Rho-independent terminators (21). These sRNAs frequently depend on Hfq in order to exert their regulatory effects, which generally occur through imperfect base pairing within the 5′ UTR of target mRNA sequences (22, 23). This can result in a variety of consequences, including translation initiation or repression, altered mRNA stability, or changes in protein activity. For example, in Escherichia coli the sRNA MicC binds within the untranslated leader sequence of the ompC mRNA adjacent to the RBS, which represses the synthesis of the OmpC protein (24), while the Vibrio cholerae Qrr sRNAs activate the translation of a diguanylate cyclase by relieving an inhibitory secondary structure in the 5′ UTR of this mRNA (25). Early work in this field established the significant impact of sRNAs on the regulation of outer membrane protein synthesis in E. coli, and more recently, the posttranscriptional regulation of virulence factors by Hfq-dependent sRNAs in bacterial pathogens has been described. For instance, Hfq and sRNAs participate in the regulation of biofilms produced by uropathogenic E. coli, V. cholerae, and Y. pestis, as well as in the regulation of alpha-toxin synthesis in Staphylococcus aureus (23, 25–29). Indeed, Hfq and sRNAs have been shown to contribute to the virulence of a number of pathogens, including both Y. pestis and Y. pseudotuberculosis (20, 30–32).
Recently, several studies have catalogued multiple cohorts of sRNAs expressed by Y. pestis and Y. pseudotuberculosis under various conditions (30, 31, 33–35), but a specific sRNA that contributes to the regulation of the T3SS in Y. pestis has not been identified. Here we present evidence for the posttranscriptional regulation of the synthesis of the T3SS effector protein YopJ by the Yersinia-specific sRNA Ysr141, as well as describe the identification of 63 additional putative sRNAs expressed by Y. pestis.
MATERIALS AND METHODS
Reagents, bacterial strains, and growth conditions.All reagents were purchased from Sigma-Aldrich (St. Louis, MO) unless stated otherwise. The bacterial strains used in this study are listed in Table S1 in the supplemental material. The oligonucleotide sequences used in this study are listed in Table S2 in the supplemental material. The plasmids used in this study are listed in Table S3 in the supplemental material. Y. pestis CO92 (36) and its derivatives were routinely grown at 26°C in liquid brain heart infusion (BHI) broth (Difco) or on BHI agar. The attenuated pgm-negative derivative of Y. pestis was used for this study, unless otherwise indicated. E. coli strains were grown at 37°C in Luria-Bertani (LB) broth or on LB agar. When necessary, these media were supplemented with kanamycin (50 μg/ml) or ampicillin (100 μg/ml).
Mutagenesis.To eliminate the effects of Pla on the T3SS, we generated a deletion of pla in the pgm-negative variant of Y. pestis using a modified form of lambda red recombination as described previously (37). A Y. pestis pgm-negative Δhfq mutant strain was created in the Δpla background as described previously (29). A Y. pestis pgm-negative Δpla Δysr141 double mutant was also generated by lambda red recombination using the primers ysr141 5′-509 and P1 ysr141 3′-1 to amplify a region approximately 500 bp upstream of ysr141 and primers ysr141 3′+535 and P4 ysr141 5′+1 to amplify a region approximately 500 bp downstream of ysr141.
Complementation of hfq and restoration of ysr141.The hfq deletion was complemented by amplifying the coding sequence (CDS) of hfq and a region including 492 bp upstream of the gene using primers hfq 5′-492 Xho and hfq 3′-305 Bam. This construct was cloned into pUC18-mini-Tn7-R6K-Km (38) and then integrated into the chromosome at the attTn7 site. The kanamycin resistance cassette, which was flanked by FLP recombination target sites and used for the selection of recombinants, was excised with the FLP recombinase as described previously (29). The ysr141 deletion was restored to the wild type through homologous recombination. A 1,171-bp region spanning approximately 500 bp on either side of the deleted portion of ysr141, including the deleted bases, was PCR amplified and cloned into the bacteriophage lambda pir-dependent vector pSR47S (39), the sequence of which was verified. The plasmid was subsequently electroporated into the Δysr141 strain. Transconjugants were selected for with kanamycin and resolved to the wild type by passage on sucrose. The restoration of ysr141 was confirmed by PCR.
Growth curves.Bacteria were subcultured from overnight cultures into 10 ml BHI (supplemented with 2.5 mM CaCl2, where indicated) in 125-ml Erlenmeyer flasks at an optical density at 620 nm (OD620) of 0.1. Cultures were grown at 26°C or 37°C for 12 h with shaking at 250 rpm. Aliquots were removed, and the OD620 was measured every 2 h.
Generation of the YopJ antibody.A 26-mer peptide (amino acid sequence CKNPLPHDKLDPYLPVTFYKHTQGKK) of YopJ from Y. pestis conjugated to the immunogenic carrier protein keyhole limpet hemocyanin was synthesized by Princeton BioMolecules. The peptide was administered to specific-pathogen-free rabbits, and polyclonal antibodies were raised by Covance.
Immunoblot analysis.Y. pestis strains were cultured in BHI with magnesium oxalate (MOX; 0.02 M Na2C2O4, 0.02 M MgCl2) to induce the T3SS in Erlenmeyer flasks at 26°C for 3 h and then transferred to 37°C for an additional 2 h (secretion-inducing conditions). Bacterial cells were collected by centrifugation, and culture supernatants were filtered through 0.2-μm-pore-size syringe filters. Volumes of filtered supernatants with equivalent ODs were precipitated by the addition of trichloroacetic acid to 10%. Precipitated proteins were resuspended in equal volumes of 1 M Tris (pH 9.0) and sample buffer, separated by SDS-PAGE, and transferred to nitrocellulose. Bacterial cell pellets were washed with phosphate-buffered saline (PBS), resuspended in PBS plus lysozyme (0.5 mg/ml), and incubated on ice for 30 min. Cells were lysed by sonication (three 30-s pulses), and cellular debris was removed by centrifugation. Whole-cell lysates mixed with sample buffer were separated by SDS-PAGE and transferred to nitrocellulose. Immunoblot analyses were performed using antibodies to YopE (J. Bliska), YopD (G. Plano), YopH (J. Bliska), YopJ, YopK (M. Marketon), YopM (S. Straley), YpkA (K. Schesser), YscF (M. Marketon), LcrF (G. Plano), and RpoA (as a loading control; M. Marketon).
RNA isolation for deep sequencing.For RNA isolation, overnight cultures of virulent Y. pestis CO92 were subcultured in BHI broth supplemented with 2.5 mM CaCl2 to an OD620 of 0.1 and were grown to early-log phase (OD620, 0.2; 2.5 h), mid-log phase (OD620, 0.8; 5.5 h), late-log phase (OD620, 1.8; 8.5 h), and stationary phase (OD620, 4.5; 15 h) at 26°C and at 37°C in a CDC-approved biosafety level 3 facility at the Ricketts Regional Biocontainment Laboratory, Argonne National Labs. At each time point, aliquots of bacteria were removed and immediately added to 2 volumes of RNAprotect Bacteria reagent (Qiagen). RNA was isolated and enriched for sRNAs using phenol-chloroform extraction, as previously described (31). Notably, the columns provided by the RiboPure Bacteria kit (Ambion) were omitted from the isolation procedure, as they are designed to eliminate small RNA species. The quality of the RNA was verified by use of an Experion automated electrophoresis system (Bio-Rad).
sRNA library preparation and deep sequencing.sRNA-enriched libraries were prepared according to a previously published protocol (31). Cluster generation was performed according to the manufacturer's instructions (Illumina), and 36-nt single-end reads were generated on a Solexa Genome Analyzer at the Institute for Genomics and Systems Biology (Argonne National Labs). The Solexa reads that passed the purity filtering and had a unique alignment were mapped to the Y. pestis CO92 reference chromosome (GenBank accession no. NC_003143) and plasmids (pCD1, GenBank accession no. NC_003131; pPCP1, GenBank accession no. NC_003132; pMT1, GenBank accession no. NC_003134).
Bioinformatics analysis of deep sequencing data.Bioinformatics was performed essentially as described previously (31). Illumina sequencing reads that overlap the annotated miscellaneous RNA, mRNA, rRNA, transfer-messenger RNA (tmRNA), and tRNA genes (based on the GenBank records of the Y. pestis CO92 chromosome and plasmids presented above) were extracted and counted. The reads not overlapping these annotations were considered to be intergenic and included 5′ and 3′ UTRs as well as potential sRNAs. Clusters of at least 50 bp in length that formed a continuous region of coverage were extracted from the intergenic category of reads on each strand of the chromosome or plasmids. Generated cluster sets were then analyzed using the Integrated Genome Browser (IGB; http://www.bioviz.org/igb/). Predicted sRNAs were examined for the presence of promoters and Rho-independent terminators using the BProm and TermFind/RNAFold programs (Softberry).
Northern blot analysis.sRNA-enriched total RNA isolated as described below (5 to 10 μg) was treated with DNase I (Ambion) and separated on 6% acrylamide gels (SequaGel reagents; National Diagnostics). RNA was transferred to Hybond N+ membranes; the hybridization conditions were those described previously (31), and washes and autoradiography were performed as described previously (31). The biotinylated oligonucleotide probes used to detect the Ysrs are listed in Table S2 in the supplemental material.
RNA isolation and qRT-PCR analysis of T3SS transcripts.Y. pestis strains were grown at 37°C under secretion-inducing conditions as described above for immunoblot analysis. Aliquots of the cultures were removed and immediately mixed with 2 volumes of RNAprotect Bacteria reagent. RNA was isolated using a RiboPure Bacteria kit and treated with DNase, and cDNA was synthesized using SuperScript II reverse transcriptase (Invitrogen) and random primers (Invitrogen). Quantitative reverse transcription-PCR (qRT-PCR) of the target genes was performed in triplicate with the SYBR green dye (Bio-Rad) using the primers listed in Table S2 in the supplemental material. The calculated threshold cycle (CT) was normalized to the CT of the gyrB gene (6) from the same cDNA sample before calculation of the fold changes using the ΔΔCT method (40).
Plasmid copy number determination.From bacteria cultured at 37°C under secretion-inducing conditions, as described above, genomic DNA (gDNA) was isolated using a Wizard gDNA isolation kit (Promega) according to the manufacturer's instructions. Quantitative PCR was performed on gDNA with primers specific for gyrB, pspA, and lcrV (see Table S2 in the supplemental material). Relative plasmid copy number was determined by the formula 2−ΔCT, where ΔCT is the threshold cycle difference between gyrB and the gene of interest. Data are the average and standard error of the mean (SEM) of 3 biological replicates.
5′/3′ RACE.Y. pestis was cultured under secretion-inducing conditions, and RNA was extracted as described above for deep sequencing. The quality of the RNA was verified by Experion. Simultaneous determination of the 5′ and 3′ ends of Ysr141 and RybB was performed using rapid amplification of cDNA ends (RACE) essentially as described previously (41). Clones were sequenced using M13 primers.
sRNA half-life determination.The decay rate and half-life of Ysr141 and RybB were determined as described previously (29). Briefly, Y. pestis was cultured in triplicate under secretion-inducing conditions as described above. After 2 h of growth at 37°C, rifampin (50 μg/ml) was added to the cultures to stop de novo transcription. Aliquots of bacteria were removed at the indicated times, RNA was extracted, cDNA was synthesized, and the relative levels of the transcripts at each time point were measured by qRT-PCR. The half-life was determined by plotting the percentage of the transcript remaining relative to the amount at time zero for each sRNA on a linear plot. The data points were then fit with a linear curve, and the equation t1/2 = 0.693/k was used, where t1/2 is the half-life and k is the slope of the line. Data are the average and SEM of 3 biological replicates. Statistical analysis was performed using the Student t test.
Ysr141 target identification.Reporter strains for target identification were constructed as described previously (29). Briefly, the PtetO promoter sequence lacking the 5′ UTR (42) was PCR amplified from pWL213, to which the 5′ UTR of sycO (43) or the 34 bases upstream of the start codon of yopJ were added via the 3′ primer used for amplification of PtetO. These products were subsequently joined to the CDS of the green fluorescent protein (GFP) gene by splicing by overlap extension (SOE)-PCR and cloned into the suicide vector pWL212 (37). All sequences were confirmed, and the reporter constructs were then introduced onto the chromosome of Y. pestis, Y. pestis Δhfq, Y. pestis Δysr141, or Y. pestis Δysr141 restored with ysr141 via Tn7-based integration, as described above. Overnight cultures of each strain were subcultured into 3 ml BHI in 15-ml culture tubes at an OD620 of 0.1. Bacteria were cultured for 2 h at 26°C; anhydrous tetracycline (ATc; 0.25 μg/ml) was then added, and the bacteria were subsequently cultured for an additional 4 h at 37°C. The OD620 and fluorescence of each culture were measured in a Tecan Safire2 microplate reader. The assay was performed in biological triplicate, and the results are representative of those from at least 3 unique assays. Statistical analysis was performed with the Student t test.
RESULTS
Hfq participates in the regulation of the Y. pestis T3SS.Previously, we demonstrated that Hfq is required for the production of multiple T3SS effector proteins by Y. pseudotuberculosis (20). As the Yop-Ysc T3SS is conserved between Y. pestis and Y. pseudotuberculosis, we hypothesized that Hfq may also contribute to the synthesis of Yops by Y. pestis. Multiple studies have shown, however, that the Pla protease of Y. pestis degrades extracellular Yops and, furthermore, that the levels of Pla are indirectly regulated via Hfq (1, 18, 19, 32). Therefore, to eliminate the effects of Pla on the T3SS, we deleted the CDS of pla from the pgm-negative strain of Y. pestis. In this context, we then generated an isogenic mutant lacking hfq; the mutation was complemented in trans by placing the native promoter and coding sequence of hfq onto the chromosome via the Tn7 transposase at the attTn7 site located downstream of glmS. To test the effects of Hfq on T3SS protein production and secretion, the parental strain, the Δhfq strain, and the Δhfq strain complemented with hfq were cultured under type III secretion-inducing conditions, cell lysates and culture supernatants were collected, and immunoblot analyses were performed using antibodies to YopD, YopE, YopH, YopJ, YopK, YopM, YpkA, YscF, and LcrF. The loss of Hfq resulted in reduced levels of all T3SS proteins, with the exceptions of YopM and LcrF, examined in both the cell lysate and the culture supernatant (>30% difference) (Fig. 1A). The levels of YopM appeared to be unchanged in the Δhfq mutant compared to those in the wild type in the culture supernatant, were elevated in the cell lysate, and were partially restored by the hfq complement. We observed that the abundance of LcrF in cell lysates was also slightly increased in the absence of Hfq.
Hfq is required for the synthesis of T3SS proteins. (A) Immunoblot analysis of cell lysates and culture supernatants from the Y. pestis wild type (WT), Y. pestis Δhfq, and Y. pestis Δhfq complemented with hfq cultured under secretion-inducing conditions. The antibody used to detect each protein is indicated to the left of the corresponding panel. The density of each band relative to that of the wild type is indicated beneath the corresponding panel. Data are representative of those from at least 3 biological replicates. (B) Relative fold change of T3SS transcripts from Y. pestis Δhfq and Y. pestis Δhfq complemented with hfq, measured by qRT-PCR, compared to the level for the wild type (set equal to 1). cDNA was generated from RNA isolated from strains cultured under secretion-inducing conditions. Error bars represent the SEMs of 3 biological replicates.
LcrF is the master transcriptional activator of the T3SS effector and structural genes (44). Given that LcrF levels are elevated in the Δhfq mutant, despite the decreased levels of other T3SS proteins, we hypothesized that the effects of Hfq are not due to decreased transcript levels. To verify this, we measured the relative steady-state levels of the yopD, yopE, yopH, yopJ, yopK, yopM, ypkA, yscF, and lcrF transcripts in the Δhfq strain compared to those in wild-type bacteria by qRT-PCR. In the absence of Hfq, we found that the transcript levels for most genes were increased (1.5- to 12-fold) compared to those in the wild type, suggesting that the decreased protein levels are due to a posttranscriptional, Hfq-dependent mechanism(s) (Fig. 1B). For instance, there were 4- and 6-fold increases in the levels of the yopJ and yscF transcripts, respectively (Fig. 1B). This highlights the posttranscriptional role of Hfq in the production of these proteins (Fig. 1A). In the absence of Hfq, transcript levels of yopM were increased over 6-fold compared to those in the wild type. Additionally, lcrF transcript levels were elevated in the Δhfq mutant, which is consistent with the increase in protein levels observed by immunoblotting. Complementation of hfq resulted in a reduction of transcript levels compared to those in the Δhfq mutant, and these levels were similar to those found in wild-type bacteria (Fig. 1B). Taken together, these data demonstrate that Hfq contributes to the regulation of the T3SS and suggest that an Hfq-dependent sRNA (or group of sRNAs) may be responsible for posttranscriptional regulation of the T3SS proteins tested in Y. pestis.
Identification of sRNAs expressed by Y. pestis.The observation that Hfq is required for the regulation of components of the T3SS in Y. pestis prompted us to search for sRNAs that could participate in this process. In order to identify potential candidate sRNAs (known as Ysrs, for Yersinia small RNAs), we performed a deep sequencing analysis of the sRNA-ome of Y. pestis cultured at both 26°C and 37°C in rich broth. In addition to finding regulators of the T3SS, we hoped to extend our previous study of global sRNA profiling in the related species Y. pseudotuberculosis (31). Size-selected cDNA libraries prepared from cultures grown to 4 distinct growth phases at both temperatures were analyzed by Illumina-based sequencing as described previously (31). Deep sequencing of these libraries resulted in 10 million to 16 million total reads per sample, of which 60 to 70% corresponded to mRNAs, and 25 to 43% of the reads were aligned to the intergenic regions (see Fig. S1 in the supplemental material). The size selection and library preparation were sufficient to keep the numbers of rRNA and tRNA reads low (0.05 to 0.7% and 0.5 to 1.2% of total reads, respectively). Prior to this study, 29 noncoding RNAs had been identified in the genome of Y. pestis. We detected 17 of these RNAs in our deep sequencing data set (see Data Set S1 in the supplemental material), and of the remaining 12, 3 were antisense RNAs, 1 was the tmRNA, and 1 was the 6S RNA, all of which were excluded by our filtering algorithm. Of these previously described sRNAs detected in our study, SsrS appeared to be the most abundant at all growth phases and temperatures (Fig. 2A).
Deep sequencing analysis of sRNA-ome of Y. pestis CO92. (A) Sequencing reads of previously annotated sRNAs found in Y. pestis. ncRNA, noncoding RNA. (B) Validation of Ysr expression by Northern blot analysis. The four times in panel A and for the four lanes in panel B correspond to early-log, mid-log-, late-log, and stationary-phase growth, respectively. The biotinylated probe used for each Ysr is listed next to each panel. (C) Ysrs identified in this study and a previous study (31). Green, 100% sequence identity between Y. pestis strain CO92, Y. pseudotuberculosis strain IP32953, Y. enterocolitica strain 8081, E. coli K-12 strain MG1655, and Salmonella enterica Typhimurium strain LT2; yellow, at least one nucleotide mismatch; red, absence of a homologous sequence. The gap corresponds to the Ysr numbers used by Beauregard et al. (33) that were not found in our analysis.
In order to identify additional putative sRNAs expressed by Y. pestis strain CO92, we used a modified version of the filtering algorithm described in our previous study (31), in that the 3-fold difference in expression of RNAs between the adjacent open reading frames (ORFs) and the intergenic clusters was omitted. By this method, we uncovered 63 new potential sRNAs and generated sequencing reads that corresponded to 144 sRNAs in Y. pestis that we previously identified in Y. pseudotuberculosis. We verified the expression of 10 of these newly identified sRNAs in Y. pestis by Northern blot analysis (Fig. 2B). Expression of most of the sRNAs varied over the course of the growth of the bacteria, while others, such as Ysr199, remained relatively constant (Fig. 2B). In addition, a significant number of the newly discovered sRNAs were absent from the genomes of or contained mismatches with the equivalent sequences in Y. enterocolitica, E. coli, and Salmonella enterica serovar Typhimurium (Fig. 2C).
Ysr141 contributes to the production of T3SS proteins in Y. pestis.One sRNA gene identified by our deep sequencing analysis, Ysr141, mapped to T3SS-carrying plasmid pCD1, encoded on the opposite strand within the intergenic region between yopH and the gene YPCD1.68c (Fig. 3A; see Data Set S1 in the supplemental material). On the basis of the genomic location of Ysr141, our ability to verify its expression by Northern blotting, and the knowledge that sRNA genes sometimes occur near their target genes (45), we hypothesized that Ysr141may contribute to the posttranscriptional regulation of the T3SS. The transcribed region of ysr141 predicted by our deep sequencing data suggests that this sRNA may overlap with the 5′ UTR, promoter, and LcrF binding site for yopH (46) (Fig. 3A). To avoid interfering with the transcription of yopH, then, we generated an isogenic mutant of Ysr141 that maintained the full yopH 5′ UTR and strong LcrF binding site, as mapped by Wattiau and Cornelis (46). This mutant contains a deletion of 182 bp (bases 48,628 to 48,810 of pCD1) of the region in which Ysr141 is encoded (Fig. 3A). We did not observe a significant difference in the growth between wild-type Y. pestis and the Δysr141 mutant when cultured in vitro at 26°C or 37°C with or without added CaCl2 (see Fig. S2 in the supplemental material). We subsequently attempted to trans-complement and/or overexpress Ysr141; however, several constructs failed to restore the wild-type phenotype to the Δysr141 mutant in our assays (see below). Therefore, to confirm that the strain containing the deletion of Ysr141 did not have a second-site mutation, the mutant was repaired to the parental genotype by restoring the deleted region back into its original locus on pCD1.
Ysr141 is a positive regulator of the T3SS. (A) Diagram illustrating the genomic context of predicted ysr141 coordinates and the ysr141 deletion. Numbers indicate coordinates on Y. pestis CO92 pCD1, as determined by Parkhill et al. (77). (B) Immunoblot analysis of cell lysates and culture supernatants from Y. pestis, Y. pestis Δysr141, Y. pestis Δysr141 restored with ysr141, and Y. pestis Δhfq cultured under secretion-inducing conditions. The antibody used to detect each protein is indicated to the left of the corresponding panel. The density of each band relative to that in the wild type is indicated beneath the corresponding panel. Data are representative of those from at least 3 biological replicates. (C) Relative fold change of T3SS transcripts from Y. pestis Δysr141 and Y. pestis Δysr141 restored with ysr141, measured by qRT-PCR, compared to that for the wild type (set equal to 1). cDNA was generated from RNA isolated from strains cultured under secretion-inducing conditions. Error bars represent the SEMs of 2 biological replicates. (D) Relative copy number of pCD1 (represented by lcrV) in Y. pestis, Y. pestis Δhfq, and Y. pestis Δysr141 compared to that in the wild type. pspA was used as a control for a chromosomal gene. Relative copy number was determined from gDNA isolated from cultures used in immunoblot/qRT-PCR analyses. Error bars represent the SEMs of 3 biological replicates.
Immunoblot analysis of cell lysates and culture supernatants collected from the parental strain, the Δhfq strain, the Δysr141 strain, and the Δysr141 strain restored with ysr141 grown under secretion-inducing conditions revealed Ysr141-dependent effects on the levels of various T3SS proteins (Fig. 3B). In the absence of Ysr141, the abundance of YopE, YopK, and LcrF was decreased ∼30 to 50% compared to that in the wild type, while YopJ, YpkA, and YscF levels were decreased ∼50 to 70%, as determined from a rough average of the densitometry of the cell lysate and culture supernatant. This suggests that Ysr141 stimulates production of these T3SS effector proteins. YopD and YopM levels appeared to be unchanged between the wild type and the Δysr141 mutant in the cell lysate, while the levels of YopH increased in the Δysr141 mutant in both the cell lysate and culture supernatant. The amount of Yop proteins detected in the culture supernatant was reduced in the absence of Ysr141 for most of the examined proteins, especially with YopJ, for example (Fig. 3B). Reduced production of YscF, the needle protein, in the Δysr141 mutant could be responsible for this phenotype. The level of production of all the Yops examined was returned to wild-type levels with the Ysr141-restored strain (Fig. 3B).
In order to determine the impact of Ysr141 on the steady-state abundance of these T3SS transcripts, we compared the relative mRNA levels between the wild-type and Δysr141 strains by qRT-PCR. We found that 7 of 9 transcripts were not decreased more than 2-fold (our threshold for biological significance) in the absence of Ysr141 compared to the level in the wild type, which is consistent with posttranscriptional effects (Fig. 3C). One exception was found with the transcriptional activator lcrF, the level of which was decreased 2.4-fold compared to the level in the wild type. In total, these data indicate that Ysr141 contributes to the production of multiple T3SS proteins and suggest that, with the exception of LcrF, these effects may occur at the posttranscriptional level. Both the immunoblot and transcript analyses were performed with a second independently derived mutant of Ysr141; this strain behaved identically to the original deletion strain in these experiments (data not shown). To confirm that the changes that we observed are not due to an effect of Ysr141 on the plasmid copy number of pCD1, we measured the abundance of the lcrV gene (carried by pCD1) in the Δysr141 mutant relative to that in the wild type and observed no significant differences (Fig. 3D).
Ysr141-dependent posttranscriptional regulation at the yopJ 5′ UTR.As sRNAs often exert their effects though interactions within the 5′ UTRs of regulated mRNA transcripts, we hypothesized that Ysr141 may target such sequences of the T3SS effector proteins that are decreased in the absence of Ysr141. To examine this, we chose to investigate the regulation of YopJ synthesis. The yopJ gene is the third gene in an operon with the T3S chaperone gene sycO and the gene encoding the secreted effector ypkA, and the transcriptional start site has been mapped to 28 bases upstream of the sycO translational start site (43). There is also a 396-bp transcribed, noncoding region between the stop codon of ypkA and the start codon of yopJ. Bioinformatic analysis revealed that within this region there is a putative promoter spanning 34 bp upstream of yopJ that also contains an alternative ribosome binding site. The translation of YopJ could be effected posttranscriptionally at either the sycO 5′ UTR or within the intergenic untranslated region. We therefore constructed a fluorescence-based reporter that contained the mapped sycO 5′ UTR cloned immediately downstream of a modified PtetO promoter that lacked the PtetO 5′ UTR, followed by the CDS for gfp (PtetO-sycO 5′ UTR-gfp), and a second, similar reporter which contained 34 bases of the ypkA-yopJ intergenic region immediately upstream of the yopJ start codon (PtetO-yopJ 5′ UTR34-gfp) (Fig. 4A). The transcription of these reporters is driven by the addition of ATc, and therefore, any differences in fluorescence levels between the strains can be attributed to posttranscriptional effects within the target region. These constructs were integrated onto the chromosomes of the parental strain, the Δhfq and Δysr141 strains, and the Δysr141 strain restored with ysr141 at the attTn7 site, the bacteria were cultured under secretion-inducing conditions in the presence of ATc, and the fluorescence of the strains was measured and normalized to the OD620 of the cultures. We observed a significant decrease in fluorescence in the Δhfq mutant, but not the Δysr141 strain, compared to that in the wild type with the PtetO-sycO 5′ UTR-gfp reporter, which suggests that while Hfq affects YopJ production via effects at this region, the sycO 5′ UTR is Ysr141 independent (Fig. 4B). In contrast, the PtetO-yopJ 5′ UTR34-gfp reporter showed significantly lower fluorescence for both the Δhfq and Δysr141 mutants than for the wild type, indicating that this 34-nt region is targeted for posttranscriptional activation in a Ysr141-dependent manner (Fig. 4C). This region upstream of yopJ is predicted to form a secondary stem-loop structure that could block access to an alternative RBS (AGGA) (Fig. 4D) (47). The IntaRNA alignment tool (48) showed a potential 7-nt-long base-pairing interaction between Ysr141 and the yopJ 5′ UTR34 region immediately upstream of the stem-loop (Fig. 4E).
Ysr141 posttranscriptionally targets a 34-bp region within the yopJ 5′ UTR. (A) Diagram of reporter constructs. (B) Relative fluorescence normalized to the OD620 of Y. pestis, Y. pestis Δhfq, Y. pestis Δysr141, and Y. pestis Δysr141 restored with ysr141 obtained with the sycO 5′ UTR reporter. (C) Relative fluorescence normalized to the OD620 of Y. pestis, Y. pestis Δhfq, Y. pestis Δysr141, and Y. pestis Δysr141 restored with ysr141 obtained with the yopJ 5′ UTR34 reporter. Error bars represent the SEMs of 3 biological replicates, and data are representative of those from at least 3 independent experiments. Statistical significance was calculated by the Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001. (D) Predicted mFOLD secondary structure of the yopJ 5′ UTR34 transcript. The most probable prediction with the lowest free energy state is shown. Circled nucleotides indicate a putative RBS. (E) Predicted IntaRNA interaction of the yopJ 5′ UTR34 transcript with the deep sequencing-predicted region of Ysr141.
Ysr141 is an unstable or processed sRNA.Our inability to trans-complement the ysr141 deletion suggested that this sRNA may be unstable or processed. To test this hypothesis, we employed simultaneous 5′/3′ rapid amplification of cDNA ends (RACE) to map the ends of this transcript. In order to ascertain the processing of Ysr141, we used a method of 5′/3′ RACE that does not distinguish between processed and unprocessed transcripts; therefore, these results represent the predominant ends in a pool of isolated RNA and not necessarily the start and end sites of the primary transcript. The ends were mapped from 72 sequences that were generated from 3 independently derived cultures of Y. pestis. We found 13 distinct 5′ ends (average at position 48,594 ± 30 on the CO92 pCD1 plasmid) and 12 different 3′ ends (average at position 48,778 ± 27), and the mean size was 189 nt (Fig. 5A). A 3′ end of Ysr141 predicted by the deep sequencing analysis (at position 48,786) was identified 10 times (or 13.5%) by our RACE method, and the nearby base at position 48,789 was identified 12 times; when combined, these account for 30% of the sequences. The predicted 5′ end was not found by RACE (at position 48,532). To verify this approach, we chose to examine as a control the well-studied, conserved sRNA RybB, for which we have previously mapped the 5′ and 3′ ends in Yersinia (31). In contrast to Ysr141, we identified only 5 different 5′ ends (average at position 1,491,837 ± 0.93 on the CO92 chromosome) and 6 different 3′ ends (at position 1,491,917 ± 1.89) of RybB when mapped from the same samples (Fig. 5B). Both the predicted 5′ and 3′ ends of RybB were found the majority of the time by RACE. Furthermore, an agarose gel of the products from 5′/3′ RACE of RybB showed a single band, whereas the same analysis of the Ysr141 5′/3′ RACE showed numerous products, consistent with the presence of a predominant RNA species versus multiple species, respectively (see Fig. S3 in the supplemental material).
Ysr141 is an unstable or processed RNA. (A) 5′/3′ RACE reads of Ysr141 present in pools of isolated RNAs. Black dots, biological replicate 1; dark gray dots, biological replicate 2; light gray dots, biological replicate 3. Numbers on the y axis indicate genomic coordinates. (B) 5′/3′ RACE reads of RybB present in the same pools of isolated RNAs. Numbers on the y axis indicate genomic coordinates. Symbols are as defined in the legend to panel A. (C) Half-lives of the RybB and Ysr141 transcripts. The percentage of mRNA remaining at each time point was measured by qRT-PCR and compared with that at time zero (t0) for each sRNA. The RybB half-life was 2.15 ± 0.24 min, and the Ysr141 half-life was 1.71 ± 0.10 min. P < 0.05. Data represent those from 3 biological replicates, with error bars representing SEMs.
These data suggest that the Ysr141 transcript may have an increased decay rate compared to that for RybB. Therefore, to measure the half-lives of these sRNAs, we cultured Y. pestis under secretion-inducing conditions and then treated the bacteria with rifampin to halt de novo RNA synthesis. Aliquots of the culture were taken at various time points, and we then isolated RNA from these aliquots to measure the relative transcript levels of RybB and Ysr141 over time. The transcript levels were normalized to the transcript level of stable gyrB mRNA (29), and the percentage of the transcript remaining was calculated for each sRNA compared to the level at time zero (Fig. 5C). The levels of both the RybB and Ysr141 transcripts decreased to between 2 and 6% of their original levels by 5 min, but they showed different decay rates. Ysr141 had a half-life of 1.71 ± 0.10 min, which was significantly shorter than the 2.15 ± 0.24-min half-life calculated for RybB.
DISCUSSION
The RNA chaperone protein Hfq is a necessary regulatory component for a wide variety of activities in many bacterial species; factors and processes affected by Hfq include outer membrane biogenesis, quorum sensing, virulence factor synthesis, and protein secretion (25, 49–51). Previously, we have shown that Hfq is involved in the positive posttranscriptional regulation of multiple components of the T3SS in Y. pseudotuberculosis, and in this report, we have demonstrated that this role for Hfq appears to be conserved in Y. pestis. In the absence of Hfq, Y. pestis was unable to synthesize wild-type levels of almost all T3SS proteins tested. Notably, the levels of YopM were unchanged in a Δhfq mutant, which suggests that the deficiency in the production of the other T3SS proteins may not be due to alterations in a pan-T3SS regulatory mechanism. In addition, the increase in transcript levels, particularly in the lcrF mRNA, in the absence of Hfq correlated with the microarray data presented in an earlier report (30). Geng et al. speculated that Hfq participates in the repression of this system under certain conditions (30); however, our study illustrates that the T3SS of Y. pestis is activated in an Hfq-dependent manner under the conditions examined here. Interestingly, the loss of Hfq does not influence the synthesis of pYV-encoded T3SS proteins in Y. enterocolitica, suggesting differences in the Hfq-dependent posttranscriptional control of virulence determinants among Yersinia species (52).
The transcription of T3SS genes is regulated through a negative-feedback loop, whereby low levels of T3SS proteins in the cell result in increased transcription, and vice versa (53, 54). We postulate that in the absence of Hfq, Y. pestis is unable to synthesize adequate levels of T3SS proteins, a condition which signals to the cell that more T3SS transcripts are needed. This leads to the elevated mRNA levels that both we and Geng et al. (30) observed by qRT-PCR and microarray analysis, respectively. The increased synthesis of the T3SS master transcriptional activator LcrF by the Δhfq mutant of Y. pestis may be an effort to enhance yop-ysc transcription; however, the end result is still a net decrease in protein levels, suggesting that Hfq and one or more sRNAs are acting downstream of LcrF to regulate the T3SS.
As Hfq typically acts in conjunction with sRNAs to exert its regulatory effects on the cell, we employed a genome-wide deep sequencing analysis to catalog the sRNA-ome of Y. pestis strain CO92 and, consequently, identified 63 additional putative sRNAs in this species. These data build upon our previous study as well as those from several other groups that have identified sRNAs in Yersinia (31, 33–35). While there exists some overlap in the sRNA sets discovered in each of these studies, there are a large number of potential sRNAs that are unique to each study. Recently, Beauregard et al. identified 15 previously unannotated sRNAs in the KIM6+ strain of Y. pestis (33). This group also identified several previously annotated sRNAs and multiple Ysrs described in our previous study; we subsequently detected in Y. pestis CO92 the expression of 2 Ysrs found in the KIM6+ strain (Ysr164 and Ysr172). A second report identified 104 sRNAs in Y. pestis strain 201 from both in vitro- and in vivo-derived samples, and 56 of these were not identified by our previous study or by the study of Beauregard et al. (33). In the current study, only one of the sRNAs identified (Ysr222 or sR049) was also detected by Yan et al. (35). The predicted region of a second sRNA (Ysr196) identified in the current study overlaps significantly with an sRNA identified by Yan et al. (35) (sR064); however, it is encoded on the opposite strand.
The variability in these data sets may stem from the different strains and methodologies used for each study. Indeed, all strains examined are associated with distinct biovars, each of which has unique genetic traits (55). The RNA used in our deep sequencing analysis was isolated from the fully virulent strain of Y. pestis CO92, which is a clinical isolate from a fatal case of pneumonic plague (56). Use of this strain allowed us to uncover the sRNAs that are expressed under the full genetic battery of a bacterium that is virulent to humans and allowed the identification of plasmid-encoded Ysrs, specifically, Ysr141, which could not have been identified using a strain that lacks the pCD1 plasmid (such as KIM6+).
This report brings the total number of potential sRNAs that we have identified between Y. pestis and Y. pseudotuberculosis to 216. As our current analysis and the study in which we identified sRNAs expressed by Y. pseudotuberculosis (31) were performed essentially identically, our data sets permit a direct comparison of sRNAs between these highly similar species and will be useful for understanding the evolution of Yersinia species through changes in sRNA content and expression. While the majority of the candidate sRNA genes that we identified are common to both Y. pestis and Y. pseudotuberculosis, the loci encoding 5 sRNAs expressed by Y. pestis CO92 are missing from the Y. pseudotuberculosis IP32953 genome. When one considers that Y. pseudotuberculosis also contains six sRNA genes that are absent from Y. pestis, one of which is required for virulence (31), these differences suggest the possibility that the disparate disease manifestations caused by each species may be due in part to their divergent sRNA-omes (57). It is also possible that the targets of conserved sRNAs are differentially controlled between the two species, as the expression patterns of the conserved sRNAs are in some cases distinct. For example, Ysr214 is encoded on the genomes of both Y. pestis and Y. pseudotuberculosis but is detectable both by deep sequencing and by Northern blotting only in Y. pestis (see Data Set S1 in the supplemental material). Additionally, thermal control of sRNA gene expression is likely to contribute to differences, as temperature plays an important role in Yersinia pathogenesis. Furthermore, many of the sRNAs that are encoded and expressed by both species contain single nucleotide variations. Regulation of common targets by sRNAs with differences in nucleotide sequence could result in distinct interactions with the mRNA between the species.
This is the first report of a specific sRNA linked to the regulation of the T3SS in Yersinia. Transcript levels of the T3SS genes generally did not change more than 2-fold, considered the threshold for biological significance, between the Δysr141 and wild-type strains. This indicates that the decreases in protein levels in the absence of Ysr141 may not be entirely attributable to decreased mRNA but, rather, may be attributable to a deficiency in protein production through transcript degradation, translation efficiency, or posttranslational effects. Whether Ysr141 targets transcripts directly or through an intermediate factor that then acts posttranscriptionally on the targets is not yet known. Furthermore, while we have demonstrated a link between Ysr141 and the 34 most proximal bases of the yopJ 5′ UTR, our assays also indicate that the 5′ UTR of the sycO-ypkA-yopJ operon is not activated in an Ysr141-dependent manner, yet YpkA levels are decreased in the absence of Ysr141. However, the actions of sRNAs are not always confined to the 5′ UTR; therefore, Ysr141 or another sRNA could affect ypkA translation by base pairing within the coding region; indeed, discoordinate regulation of operons by sRNAs has been demonstrated in other bacterial species (58–60). Alternatively, the translation of YopJ could be the only direct target of Ysr141, and the other observed changes could be due to the dysregulation of YopJ production, as the Yersinia T3SS is known to be sensitive to changes in effector protein levels (61).
In addition, the disparity between the Δysr141 and Δhfq mutant phenotypes suggests that there may be additional sRNAs that contribute to the regulation of the T3SS. A second pCD1-encoded putative sRNA gene that could be an additional regulator of the system was identified by Yan et al. (35). Alternatively, the Δhfq mutant has a general growth defect that is not seen with the Δysr141 strain, and the elaboration of the T3SS is known to be an energy-rich process, which could account for the significant decrease in T3SS protein production in this mutant. Interestingly, Ysr141 does not seem to affect YopM levels, which indicates that this effector is not part of the Ysr141-controlled regulon.
Several examples of Hfq- and sRNA-dependent regulation of type III secretion systems in other bacteria have recently been reported. For example, Hfq has been shown to be involved in the regulation of locus of enterocyte effacement-encoding T3SS genes in enterohemorrhagic E. coli (EHEC) (62, 63). In addition, in the plant pathogen Xanthomonas campestris pv. vesicatoria, the sRNA xS13, which is responsive to high salt concentrations and high temperatures, has been suggested to be an indirect positive regulator of secretion by affecting the levels of a key transcriptional regulator of the T3SS (64). In Shigella dysenteriae, the iron-responsive sRNA RyhB has been implicated in the repression of the transcriptional activator VirB, which leads to reduced transcription of T3SS genes (65). Interestingly, Y. pestis encodes 2 RyhB sRNAs (RyhB1 and RyhB2), which could also influence the regulation of the T3SS (31, 66). It is of great interest to determine how Ysr141 links the posttranscriptional activation of the T3SS to environmental signals in Yersinia and whether these mechanisms are similar to those found in other species that regulate their T3SSs via Hfq and sRNAs. In addition, as some sRNAs are known to have multiple targets, it is possible that Ysr141 may contribute to the regulation of factors outside the T3SS.
We have shown through RACE analysis that the processed 5′ and 3′ ends of Ysr141 are variable. Our data are consistent with findings from other groups that have mapped the ends of sRNAs in Yersinia and in other bacterial species, insofar as many individual sRNAs seem to have multiple start and end sites within the cell at any given time (33, 67). The prevailing theory is that, unlike their eukaryotic counterparts, microRNAs, bacterial sRNAs are not generally processed (21). However, the fact that sRNAs are transcribed from their own promoters and do not depend on a Dicer homolog to become functional does not preclude these sRNAs from being processed as an integral part of their regulatory mechanism. For example, several sRNAs (SraC, SraF, GlmZ) are transcribed as long primary transcripts that are then processed into shorter active forms (68–70), and multiple active forms of DsrA and GadY can be detected in a cell, despite being transcribed from a single promoter (71–73). We were unable to identify the deep sequencing-predicted 5′ end of Ysr141, which could be due to misidentification of the end by deep sequencing analysis, or it is possible that rapid processing of this transcript may prevent detection by our 5′/3′ RACE method. Processing or degradation could remove Ysr141 when it is no longer needed, allow it to fold differently upon cleavage of its 5′ or 3′ ends, or transform it into an active species. In addition, the 3 replicate bacterial cultures grown for 5′/3′ RACE each had distinct patterns of 5′ and 3′ ends for Ysr141, even though the ends of RybB were considerably more stable, suggesting that small changes in culture conditions, medium composition, or other variables can have a profound effect on the stability of Ysr141. In addition, our inability to trans-complement the deletion of ysr141, either by use of a plasmid or on the chromosome, may be due to its transient stability or complex processing strategy. If Ysr141 is as unstable as our data indicate, trans-complementation may produce the sRNA at intracellular locations too distal from the target transcripts, which would hinder the ability of Ysr141 to interact with them before it is degraded. However, we cannot exclude the possibility that the deletion of Ysr141 may also have polar effects, such as modification of the DNA coiling near this site.
The Yop-Ysc T3SS is a critical component of virulence for pathogenic Yersinia species (10). It is regulated in a specific, multilayered manner to ensure that effectors are synthesized and secreted only when their delivery to host cells is needed. In addition, Yersinia, through the action of YopK, ensures that only a certain amount of effectors is delivered to each cell (74). Within this framework, then, it is not surprising that sRNAs participate in the regulation of effector production. The regulation of Yop levels by Ysr141 could contribute to the virulence of Y. pestis, given that an excess or dearth of a given effector can disrupt the balance that Yersinia achieves during the host-pathogen interaction (75). Furthermore, the gene encoding Ysr141 is also present in Y. pseudotuberculosis and Y. enterocolitica, which raises the possibility that the function of this sRNA may be a mechanism to regulate the T3SS conserved among the pathogenic Yersinia species. Ysr141 may link environmental cues to the modulation of the T3SS in these species, and although the exact role of host cell contact in the induction of translocation is still not fully understood (76), the inclusion of sRNAs such as Ysr141 in the model may aid in the resolution of this question.
ACKNOWLEDGMENTS
We thank members of the W. W. Lathem laboratory for insightful discussions of this work and Kurt Schesser, Melanie Marketon, James Bliska, Susan Straley, and Greg Plano for the kind donation of antibodies. We thank Trevis Alleyne, Hannah Imlay, Lauren Bellows, and Jay Schroeder for their contributions to this work. Finally, we thank Laurianne Quenee for the isolation of Y. pestis for sRNA identification and Marc Domanus for Illumina-Solexa sequencing.
This work used resources of the Northwestern University Structural Biology Facility, which is generously supported by NCICCSG P30 CA060553 awarded to the Robert H. Lurie Comprehensive Cancer Center. This work was sponsored by the NIH/NIAID Regional Center of Excellence for Bio-Defense and Emerging Infectious Diseases Research (RCE) Program. We acknowledge membership within and support from the Region V Great Lakes RCE (NIH award U54 AI057153). This work was also supported by NIH/NIAID grants R21 AI103658 and R01 AI093727 to W.W.L.
FOOTNOTES
- Received 17 December 2013.
- Accepted 8 February 2014.
- Accepted manuscript posted online 14 February 2014.
- Address correspondence to Wyndham W. Lathem, lathem{at}northwestern.edu.
↵* Present address: Jovanka T. Koo, Wheaton College, Department of Biology, Wheaton, Illinois, USA.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/JB.01456-13.
REFERENCES
- Copyright © 2014, American Society for Microbiology. All Rights Reserved.
