The vacB Gene Required for Virulence inShigella flexneri and Escherichia coli Encodes the Exoribonuclease RNase R*

vacB, a gene previously shown to be required for expression of virulence in Shigella and enteroinvasive Escherichia coli, has been found to encode the 3′–5′ exoribonuclease, RNase R. Thus, cloning of E. coli vacB led to overexpression of RNase R activity, and partial deletion or interruption of the cloned gene abolished this overexpression. Interruption of the chromosomal copy ofvacB eliminated endogenous RNase R activity; however, the absence of RNase R by itself had no effect on cell growth. In contrast, cells lacking both RNase R and polynucleotide phosphorylase were found to be inviable. These data indicate that RNase R participates in an essential cell function in addition to its role in virulence. The identification of the vacB gene product as RNase R should aid in understanding how the virulence phenotype in enterobacteria is expressed and regulated. On the basis of this information we propose that vacB be renamed rnr.

Exoribonucleases play an important role in RNA maturation, turnover, and degradation (for reviews, see Refs. 1 and 2). In Escherichia coli eight distinct exoribonucleases have been characterized. Most of them display a degree of overlap in their function. For example, six of the eight, including RNases II, D, BN, T, PH, and polynucleotide phosphorylase (PNPase), 1 participate in the 3Ј-maturation of tRNA precursors (3). Recently, the maturation of the small stable RNAs, M1 RNA, 10Sa RNA/tmRNA, 6S RNA and 4.5S RNA, was examined and found to involve many of the same exoribonucleases (4). It is also known that strains lacking RNases II, D, BN, T, and PH in combination are inviable, but the presence of any one of the five enzymes is sufficient to confer viability, although with varying degrees of effectiveness (5).
RNase R is one of the eight exoribonucleases. It acts nonspecifically on poly(A), poly(U), and ribosomal RNAs (rRNA) in vitro (1, 6 -8). The enzyme was initially identified 20 years ago in an E. coli strain deficient in RNase II (6). Whereas RNase II accounts for more than 95% of the activity against poly(A) and poly(U) in crude cell extracts, the residual activity against these substrates and rRNA is due primarily to RNase R (1,5,7). Based on its gel filtration properties, RNase R is apparently a protein of ϳ85 kDa (8). However, despite all of this biochemical information, essentially nothing was known about the gene encoding RNase R other than that it mapped to the last quarter of the E. coli chromosome. 2 In this paper we report the identification and characterization of the gene that encodes RNase R and show that it is the E. coli vacB gene. vacB was originally described in Shigella flexneri as a chromosomal gene required for expression of the virulence genes carried on the large plasmid of this organism (9). We were led to consider vacB as a candidate for the gene encoding RNase R because (a) sequence analysis revealed that it is homologous to the rnb gene that encodes another exoribonuclease with similar properties, RNase II (10); (b) the deduced size of the VacB protein (ϳ92 kDa) agreed closely with that known for RNase R; and (c) vacB is located at 95 min on the E. coli chromosome (9), a position consistent with the earlier mapping studies of the gene encoding RNase R. Based on these considerations, we cloned and characterized the E. coli vacB gene. The data obtained from these studies demonstrate that the vacB gene does indeed encode RNase R, and we propose that it be renamed rnr. The identification of VacB as an exoribonuclease has important implications for the understanding of virulence associated with enterobacteria.
The E. coli genomic clone, DD947, which contains a 20-kb fragment carrying vacB, was a gift from Dr. Frederick R. Blattner (University of Wisconsin-Madison) and was used to prepare DNA for subcloning. Plasmid pBS(ϩ) (Stratagene) was used as the cloning vector. Plasmids pUC4K (Pharmacia) and pBR325 provided the Kan r and Cam r cassettes used to interrupt vacB.
Plasmid pBSV was constructed by subcloning a 4.58-kb EcoRI-NheI fragment from DD947 into the EcoRI-XbaI sites of pBS(ϩ). To interrupt the vacB gene on pBSV, a 1.23-kb PstI fragment was deleted from vacB, and the remaining fragment was self-ligated to generate plasmid pBSVD. The Kan r cassette, a 1.25-kb PstI fragment from pUC4K, was inserted into the PstI site of pBSVD to create plasmid pBSVK. Likewise, plasmid pBSVC was constructed by inserting the 1.23-kb Cam r cassette from pBR325 into the PstI site of pBSVD (see Fig. 1). Plasmid pBSR was constructed by transferring the 2.72-kb XmnI fragment of pBSV into the HincII site of pBS(ϩ) such that vacB was placed under the control of the lac promoter. Because of a stop codon upstream of vacB, its gene product is not translationally fused to lacZ. In this plasmid the upstream gene, yjeB, and most of the downstream gene, yjfH, are deleted (see Fig. 1).
* This work was supported by Grant GM16317 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom reprint requests should be addressed. Interruption of vacB on the Chromosome-Plasmids pBSVK and pBSVC, which contain deletion-insertion mutations in vacB, were cleaved with ScaI and FspI, respectively, and introduced into strain CF881 by linear transformation. Kan r Amp s or Cam r Amp s transformants were selected, and they served as a source for preparation of bacteriophage P1 lysates. The interrupted vacB genes were transferred to wild type, CA265, and the RNase II Ϫ strain, CAN20 -12E, by P1mediated transduction. Extracts of the resulting Kan r or Cam r strains were prepared and assayed against [ 3 H]poly(A) to verify the decrease in RNase R activity. To ensure that the wild type vacB gene was replaced by one containing the Kan r or Cam r cassettes, chromosomal DNA of these strains was prepared and analyzed by Southern hybridization.
Preparation of Cell Extracts-Cells were grown to the indicated A 550 and collected by centrifugation. Extracts were prepared by sonication of cells suspended in 0.1 volume of a buffer containing 20 mM Tris-chloride, pH 7.5, 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride, 300 mM KC1. The protein concentration of the extract was determined by the Bradford method.
RNase R Assay-RNase R activity was measured under optimal conditions by determination of acid-soluble radioactivity released from the substrates, [ 3 H]rRNA or [ 3 H]poly(A), as described (8). Assays were carried out in 50-l reaction mixtures containing: 20 mM Tris-chloride, pH 8.0, 0.25 mM MgCl 2 , 300 mM KC1, 5 g of [ 3 H]rRNA, or 40 g of [ 3 H]poly(A). Twenty microliters of extract were added to each reaction, and incubations were carried out at 37°C for 5-30 min.

RESULTS AND DISCUSSION
Cloning of the vacB Gene-To ascertain whether the vacB gene encodes RNase R, it was subcloned from the E. coli genomic clone, DD947, and placed into pBS(ϩ) (Fig. 1). The resulting plasmid, pBSV, harbors a 4.58-kb insert containing three genes, yjeB (426 bp), vacB (2,442 bp), and yjfH (732 bp). pBSV was transformed into an RNase II Ϫ strain, CA265II Ϫ , and assays were carried out to determine whether RNase R activity was elevated. Transformation with the pBS(ϩ) vector served as a control. Inasmuch as RNase II accounts for Ͼ95% of the nonspecific exoribonucleolytic activity in an E. coli extract (1,5,7), it was necessary to use the RNase II Ϫ background to accurately detect changes in RNase R activity. Transformed cells were grown to an A 550 Ϸ 1, and sonicated extracts were prepared and assayed using [ 3 H]poly(A) and [ 3 H]rRNA, substrates of RNase R (7,8). Under these conditions, activity against poly(A) was elevated 2-3-fold and against rRNA, 5-7fold when extracts from pBSV-transformed cells were assayed. No elevation of activity was observed for extracts from pBS(ϩ)transformed cells (data not shown). These initial findings supported the conclusion that one of the three genes present in the insert in pBSV encodes RNase R.
Because the elevation of RNase R activity was lower than expected for expression from a multicopy plasmid, we examined whether the plasmid may have been lost during growth. In fact, based on plating on YT/ampicillin and on analysis of DNA minipreps, only 10 -15% of the cells retained plasmid pBSV; in contrast, all the cells retained pBS(ϩ). These data suggested that overexpression of at least one of the genes on the pBSV insert is deleterious to cells. Accordingly, we repeated the RNase R assays in cells grown only to an A 550 Ϸ 0.3 in an attempt to minimize plasmid loss (Table I). In several experiments, one of which is presented, activity using poly(A) as substrate was elevated 3-5-fold, and with rRNA as substrate, 7-12-fold.
To show that vacB was the gene responsible for the elevated RNase R activity, two additional plasmids, pBSVD, containing a deletion in the vacB gene, and pBSVK, containing both a deletion and a Kan r insertion in vacB (Fig. 1), were also examined for their ability to elevate RNase R activity (Table I). No increase in activity against poly(A) was observed in the presence of these plasmids; moreover, these plasmids were stably maintained in cells. In another experiment to demonstrate that vacB was responsible for the elevation of RNase R activity, an additional plasmid, pBSR, in which the two adjacent genes had been removed (Fig. 1), was also examined. In this plasmid vacB is under the control of the lac promoter. In the presence of 0.2% lactose and 1 mM isopropylthio-␤-D-galactoside, up to a 100-fold increase in activity against poly(A) was observed in the CAN20 -12E background (data not shown), indicating that the adjacent genes are not needed for elevation of RNase R activity.
To confirm that the elevated RNase activity actually corresponds to RNase R, samples were analyzed by gel filtration on Ultrogel AcA44. Extracts prepared from CA265II Ϫ cells transformed with either pBSV or pBS(ϩ) were fractionated, and RNase activity was determined. Assays using [ 3 H]poly(A) as substrate revealed a peak of activity eluting with a molecular weight of ϳ86,000 from the pBSV extract; however, this peak of activity was non-detectable from extracts of cells transformed with pBS(ϩ) (data not shown). Taken together, these data strongly support the conclusion that vacB encodes RNase R. Inasmuch as the predicted size of the VacB protein is ϳ92,000, these data also indicate that RNase R is a monomer. Based  upon the information presented here, we propose that vacB be renamed rnr.
Interruption of Chromosomal rnr-The deletion-interruption mutations of rnr present in plasmids pBSVK and pBSVC were introduced into the chromosome of strain CF881 by linear transformation; Kan r Amp s and Cam r Amp s transformants, respectively, were selected for further study. The mutated rnr genes were then transferred to strains CA265 and CAN20 -12E by P1-mediated transduction. RNase activity assays (Table II) revealed that the residual activity in strain CAN20 -12E, amounting to Ϸ2% of wild type, was decreased even further and rendered undetectable upon introduction of the interrupted rnr genes.
To confirm that the chromosomal rnr gene had, in fact, been substituted with the deletion-interruption mutation, chromosomal DNA from strain CAN20 -12Ernr::kan was subjected to Southern analysis using probes specific for the rnr gene and the Kan r cassette. DNA from the parental strain, CAN20 -12E, and from CA265II Ϫ were used as controls (Fig.  2). When hybridized with the probe specific for rnr, EcoRIdigested chromosomal DNA from all of the strains, including strains CA265II Ϫ and CAN20 -12E that carry wild type rnr and strain CAN20 -12Ernr::kan that contains the Kan r interruption, gave rise to a 9.6-kb band ( Fig. 2A, lanes 2-5). This is as expected because the interrupted rnr gene is a deletioninsertion mutant and the fragment deleted is almost identical in size to the Kan r fragment inserted (see "Experimental Procedures"). When hybridized with the probe specific for the Kan r cassette, no band was detected with strains CA265II Ϫ and CAN20 -12E (Fig. 2B, lanes 2 and 3), whereas the same 9.6-kb band was detected with strain CAN20 -12Ernr::kan (Fig. 2B,  lanes 4 and 5). These data demonstrate that the rnr gene in the latter strain has been interrupted. Identical results were obtained with the strain containing the Cam r interruption (data not shown).
The isolation of an E. coli strain with a null mutation in rnr indicates that RNase R is not an essential enzyme for cells cultured in the laboratory even when several other exoribonucleases are also absent. To determine whether the absence of RNase R has any effect on cell growth, strain CAN20 -12Ernr::kan was grown on rich medium (YT) and on minimal medium (M9/0.2% glucose) plates at 31, 37, and 42°C. No growth defect was detected compared with the parental strain, CAN20 -12E, under any of these conditions. Moreover, the doubling time of the mutant strain at 37°C in YT, 0.2% glucose was 30 min, the same as that of the parent. Strain CAN20 -12Ernr::kan also recovered from a 24-h carbon source starvation in M9 salts with the same kinetics as that of the parent, indicating no defect in recovery from starvation. These data suggest that whatever function is served by RNase R, it can be rescued completely by the exoribonucleases that are still present in strain CAN20 -12Ernr::kan. Although this strain lacks RNases II, D, BN, and R, RNases T and PH, PNPase, and oligoribonuclease still remain.
To test whether any of the remaining, known exoribonucle-ases can compensate for the absence of RNase R, the interrupted rnr gene was introduced into mutant strains already lacking either RNase T, RNase PH, or PNPase using phage P1-mediated transduction. As shown in Table III, viable transductants can be isolated when the rnr mutation is combined with mutations leading to the absence of either RNase T or RNase PH. However, double mutants lacking RNase R and PNPase do not grow. Inviability of the rnr,pnp double mutant was also seen when the pnp mutation was introduced into a RNase R Ϫ strain (Table III). Thus, these data indicate that of the known exoribonucleases (except oligoribonuclease), only PNPase overlaps with RNase R to the extent that at least one of them needs to be retained to maintain cell viability. Sequence Analysis of rnr and RNase R-Based on sequence analysis of the E. coli rnr (vacB) gene, its coding region encompasses 2,442 nt starting with an AUG initiator codon and ending at a UGA termination codon. This open reading frame would encode a protein of 813 amino acids with a calculated molecular mass of 92,109 Da, in close agreement with the size of RNase R determined by gel filtration. The translation start site proposed here is that predicted in the SWISS-PROT data base (accession number P21499) and differs from that indicated in the GenBank data base (accession number G1790622), which begins at a GUG initiation codon 42 nt upstream. We favor the former assignment because a good Shine-Dalgarno sequence, GGAGG, is located 7 nt upstream of the proposed AUG codon, whereas no Shine-Dalgarno sequence is evident upstream of the GUG codon. Also, plasmid pBSR, which leads to elevated RNase R activity (see above), lacks the GUG and other upstream codons.   2. Southern blot analysis of chromosomal DNA. Samples were digested with EcoRI and resolved on a 0.7% agarose gel. DNA digested with HindIII and labeled with 32 P at the 5Ј ends was used as the size marker (lane 1). The sizes of the marker bands are indicated (in kilobases). Lanes 2 and 3 were loaded with chromosomal DNA from strains CA265II Ϫ and CAN20 -12E, respectively, both of which carry wild type rnr. Lanes 4 and 5 were loaded with DNA from CAN20 -12Ernr::kan. A, probed with the 1.01-kb StuI-PstI fragment of rnr; B, probed with the 1.25-kb Kan r cassette, both prepared by random primer extension using the Prime-a-Gene Labeling System. It is likely that rnr is cotranscribed with the adjacent genes, yjeB and yjfH. First, no promoter sequence is evident in the short intergenic sequence between yjeB and rnr or between rnr and yjfH, and second, no transcription terminator is seen downstream of rnr. On the other hand, a possible 70 promoter, TAGCGA (18 nt) TATCAT, is present upstream of yjeB, and a likely rho-independent terminator, a 7-bp stem followed by 9 U residues, is located 20 nt downstream of the termination codon of yjfH. If these predictions are confirmed, it would indicate that rnr is part of an operon together with the two adjacent genes. Although the identity and functions of these two genes have not yet been established, we have found that yjeB is distantly related to a number of transcriptional repressors and contains a helix-turn-helix motif. yjfH is homologous to a family of RNA methyltransferase genes, including the E. coli spoU (trmH) gene encoding a tRNA 2Ј-O-methyltransferase (13). Computer analysis also revealed the presence of a REP sequence (14) in the intergenic region between rnr and yjfH.
Based upon its deduced amino acid sequence, RNase R is a basic protein with a pI ϭ 8.78. Whereas basic amino acid residues are distributed throughout the protein, there is a particularly high positive charge density in the C-terminal region. In fact, 40% of the 73 C-terminal residues of RNase R are basic (5 Arg, 24 Lys). Also identified in the C-terminal region is one copy of the S1 RNA binding domain (10). Interestingly, this domain is also present in the C-terminal region of two other E. coli exoribonucleases, RNase II and PNPase (10), both of which have substrate specificities similar to RNase R (1). In addition, as noted earlier, we have now found that RNase R Ϫ , PNPase Ϫ double mutant strains are inviable, and earlier work had shown that RNase II Ϫ , PNPase Ϫ strains also do not survive (15). Moreover, E. coli RNase II and RNase R display a high degree of sequence similarity, approaching 60% if conservative amino acid replacements are considered. These observations strongly suggest that RNase II, RNase R, and PNPase may constitute a subfamily within the group of eight E. coli exoribonucleases.
Homologues of E. coli rnr are found in essentially all the sequenced genomes, extending from Mycoplasma to humans (Ref. 16 and this work). Sequence similarities extending over wide regions of these derived proteins range upward from 30% identity and 40% when conservative amino acid replacements are included. These data suggest that the function carried out by RNase R may have been maintained over a wide range of organisms. On the other hand, we have not found homologues of rnr in the sequenced Archaeal genomes.
E. coli rnr is clearly orthologous to the vacB gene of S. flexneri (9). However, upon comparing the VacB and RNase R protein sequences, we were surprised to find two interruptions in the near perfect alignment. The first is a 52-amino acid stretch between residues 177 and 228. We found that by inverting a 150-bp EcoRV fragment (bases 1199 to 1349 of Gen-Bank D11024) and introducing two single frameshift corrections near the EcoRV sites, amino acid sequence identity could be restored. Although the inverted vacB segment could be a natural variant because it causes a major disruption within a region conserved among multiple species, we suspect that vacB sequencing errors are the cause of this difference. The second discrepancy, which we also attribute to a likely Shigella vacB sequencing error, is caused by a C-terminal frameshift (a missing G after position 2858 of GenBank D11024). We resequenced this region in E. coli and found perfect agreement with the published E. coli sequence. The reconstructed Shigella VacB and the E. coli RNase R sequences are now 99% identical with only 7 amino acid differences and 29 nucleotide differences. After the C-terminal frameshift reversal, the Shigella vacB sequence extends to the end of GenBank D11024. The last 43 amino acids of the reconstructed Shigella VacB remain unsequenced. The reconstructed partial Shigella VacB protein sequence will appear in the SWISS-PROT data base as a revised entry P30851.
Earlier work showed that the vacB gene product is required for the expression of the virulence phenotype of Shigella and enteroinvasive E. coli (9). A mutation in vacB was found to reduce the expression of several plasmid-encoded virulence antigens, and it was suggested that this deficiency was because of an effect at the posttranscriptional level (9). However, the molecular processes affected by the vacB product were not understood. The new information presented here that vacB encodes the 3Ј-5Ј exoribonuclease, RNase R, narrows the possibilities for VacB action in the expression of virulence and should aid in clarifying its role in this process. However, it appears that the function of RNase R extends beyond just affecting virulence. The fact that mutant E. coli K-12 strains lacking PNPase and RNase R are inviable suggests that these enzymes carry out an essential function in RNA metabolism that cannot be taken over by any of the other cellular exoribonucleases, even the closely related RNase II. It will be of considerable interest to identify this role as well.