The RihA, RihB, and RihC Ribonucleoside Hydrolases ofEscherichia coli

Pyrimidine-requiring cdd mutants ofEscherichia coli deficient in cytidine deaminase utilize cytidine as a pyrimidine source by an alternative pathway. This has been presumed to involve phosphorylation of cytidine to CMP by cytidine/uridine kinase and subsequent hydrolysis of CMP to cytosine and ribose 5-phosphate by a putative CMP hydrolase. Here we show that cytidine, in cdd strains, is converted directly to cytosine and ribose by a ribonucleoside hydrolase encoded by the previously uncharacterized gene ybeK, which we have renamedrihA. The RihA enzyme is homologous to the products of two unlinked genes, yeiK and yaaF, which have been renamed rihB and rihC, respectively. The RihB enzyme was shown to be a pyrimidine-specific ribonucleoside hydrolase like RihA, whereas RihC hydrolyzed both pyrimidine and purine ribonucleosides. The physiological function of the ribonucleoside hydrolases in wild-type E. coli strains is enigmatic, as their activities are paralleled by the phosphorolytic activities of the nucleoside phosphorylases, and a triple mutant lacking all three hydrolytic activities grew normally. Furthermore, enzyme assays andlacZ gene fusion analysis indicated that rihBwas essentially silent unless activated by mutation, whereasrihA and rihC were poorly expressed in glucose medium due to catabolite repression.

In Escherichia coli and Salmonella enterica serovar typhimurium, exogenous ribonucleosides are predominantly metabolized by nucleoside phosphorylases, which phosphorolytically cleave the N-glycosidic bond, yielding ribose 1-phosphate and the corresponding nucleobase ( Fig. 1) (for reviews, see Refs. 1 and 2). Uridine phosphorylase encoded by the udp gene is specific for uridine, whereas purine-nucleoside phosphorylase encoded by the deoD genes is capable of cleaving all the purine ribonucleosides, except xanthosine, which is metabolized by xanthosine phosphorylase, the xapA product. Nucleoside phosphorylases capable of cleaving cytidine are not known; however, cytidine is efficiently converted to uridine by cytidine deaminase, the cdd gene product.
Some ribonucleosides may also be phosphorylated directly to the corresponding nucleoside 5Ј-monophosphates by nucleoside kinases, encoded by the udk and gsk genes (Fig. 1). However, these enzymes generally constitute a minor pathway of nucleoside salvage because they are tightly feedback-inhibited by nucleotides and compete poorly for their nucleoside substrates against the nucleoside phosphorylases and cytidine deaminase (3)(4)(5). A kinase specific for adenosine has not been found in E. coli, but adenosine is predominantly converted to inosine by adenosine deaminase (6).
Although the general picture of nucleoside metabolism as outlined here has not changed considerably over the last 20 years, there has been some indications that additional unknown enzymes might be involved in the assimilation of nucleosides in E. coli. Thus, it is well known that pyrimidinerequiring mutants that are deficient in both the udk and udp genes still have the ability to utilize uridine as a pyrimidine source by some unknown pathway (6). It has also been known for many years that pyrimidine-requiring cdd mutants of E. coli deficient in cytidine deaminase can grow slowly in glucose minimal medium with cytidine as a pyrimidine source (7). In this case, an alternative pathway for conversion of cytidine to uracil nucleotides has been inferred to involve phosphorylation of cytidine to CMP by cytidine/uridine kinase, followed by hydrolysis of CMP to cytosine and ribose 5-phosphate by a putative CMP hydrolase. Cytosine could then be further metabolized to uracil nucleotides by cytosine deaminase and uracil phosphoribosyltransferase encoded by the codA and upp genes, respectively. The evidence for the operation of this alternative pathway of cytidine salvage rested on the finding that growth of pyr cdd mutants on cytidine was eliminated by additional mutational blocks in the udk, codA, or upp genes (7).
In this work, we attempted to identify the gene for the putative CMP hydrolase by subjecting a pyrF cdd strain to transposon mutagenesis and screening for mutants that had lost the ability to utilize cytidine as a pyrimidine source. Contrary to our expectations, all the obtained mutants contained normal levels of CMP hydrolase activity, but they were deficient in a cytidine hydrolase activity encoded by the previously uncharacterized ybeK gene, which we have renamed rihA. Thus, the major pathway for conversion of cytidine to uracil nucleotides in a cdd mutant appears to be initiated by direct hydrolysis of cytidine to cytosine and ribose catalyzed by the RihA hydrolase. Interestingly, this enzyme was also found to be an efficient uridine hydrolase and a major contributor to the unknown pathway of uridine salvage operating in the absence of uridine kinase and uridine phosphorylase.
Further analyses revealed that the E. coli genome contains two additional ribonucleoside hydrolase genes, yeiK and yaaF, which we have renamed rihB and rihC, respectively. To investigate the possible physiological function of the three nucleoside hydrolases, we have determined their substrate specificity and have characterized mutant strains in which each of the structural genes has been disrupted singly or in combination, as well as strains that overproduce the hydrolytic activities from multicopy plasmids. Finally, we have mapped the corresponding promoters of these genes and present initial studies of their regulation using lacZ gene fusions.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Media-The bacterial strains used in this study are all derivatives of E. coli K12 and are listed in Table I. Generalized transductions with lysates of bacteriophage P1vir were performed as described (8). Minimal medium plates contained AB minimal medium (15) solidified with 2% of Difco Bacto-agar and supplemented with 0.2% glucose or glycerol as a carbon source, 1 g/ml thiamin, and 15 g/ml nucleobases or 30 g/ml nucleosides when required. Rich medium was Luria broth (LB medium).
Isolation of a rihA::cam Mutant, CN2258 -The pyrF cdd strain CN1930 (Table I) was subjected to transposon mutagenesis with mini-Tn10 cam from NK1324 as described (16). Approximately 20,000 chloramphenicol-resistant colonies selected on LB medium ϩ 20 g/ml chloramphenicol were pooled and subjected to penicillin treatment (resulting in ϳ99% killing) during exponential growth in glucose minimal medium with cytidine as a pyrimidine source. After washing and appropriate dilution in AB minimal medium, survivors were plated on glucose minimal medium supplemented with cytosine (and 10 g/ml chloramphenicol). 13 clones that had lost the ability to utilize cytidine as a pyrimidine source were subsequently identified by replica plating onto glucose minimal medium supplemented with cytidine. The cam insert from one of these mutants (CN2258) was cloned into pBR322 by selection for chloramphenicol resistance, and DNA sequencing revealed that the chloramphenicol resistance gene was inserted at positions 8643-8651 in the rihA DNA sequence (GenBank TM /EBI accession number AE000169) in the opposite direction of rihA.
Isolation of a rihB::cam Mutant, CN2324 -The genome of CN2264 (cdd::Tn10 rihB, IS50-activated) ( Table I) was mutagenized with mini-Tn10 cam as described above. A phage P1 lysate prepared on a pool of ϳ20,000 chloramphenicol resistant transposants was subsequently used to transduce CN2300 (cdd ϩ rihB, IS50-activated) ( Table I) to tetracycline resistance. By replica plating onto LB medium ϩ tetracycline ϩ chloramphenicol, we isolated a total of 327 transductants, which had received a cam insertion by co-transduction with cdd::Tn10. Among these transductants, we found three clones that had lost the ability to grow rapidly on glucose ϩ cytidine while retaining the ability to use cytosine as a pyrimidine source. The cam insert in one of these strains (CN2324) was cloned from genomic DNA by selection for chloramphenicol resistance and found to be located at positions 8224 -8232 in the rihB gene (GenBank TM /EBI accession number AE000305) in the opposite orientation of rihB. One of the two other clones contained an identical insertion, whereas the last clone had the cam gene inserted at positions 8093-8101 in the same orientation as rihB.
Isolation of a rihC::cam Mutant, CN2403-The genome of CN1930 (Table I) was subjected to transposon mutagenesis with mini-Tn10 cam as described above. A phage P1 lysate prepared on a pool of ϳ20,000 chloramphenicol resistant transposants was subsequently used to transduce CN2316 (car-403::Tn10) (Table I) to pyrimidine prototrophy. By replica plating onto glucose minimal plates with chloramphenicol, we isolated a total of 220 clones presumed to have the chloramphenicol resistance gene inserted in the vicinity of the car ϩ allele. By PCR 1 screening of these clones using the primer pairs cam-down ϩ rihC-BamHI or cam-down ϩ rihC-EcoRI (Table II), we identified one clone (CN2403) that gave a PCR product of ϳ400 base pairs with cam-down ϩ rihC-BamHI. Sequencing of this product revealed that the cam gene was inserted in the rihC reading frame in the same orientation as rihC, at nucleotides 7271-7279 (GenBank TM /EBI accession number AE000113).
Isolation of a deoD gsk::kan Mutant, CN1980 -The gsk-3 mutant CN1932 (Table I) was mutagenized with mini-Tn10 kan from NK1316 as described by Kleckner et al. (16). A phage P1 lysate prepared on a pool of ϳ20,000 kanamycin resistant transposants was subsequently used to transduce CN1879 (adk-2(Ts) gsk ϩ zbb-2419::Tn10) to kanamycin resistance at 42°C, resulting in the selection of clones with kan insertions in the vicinity of the adk ϩ allele. The gsk gene is located between the adk gene and zbb-2419::Tn10, so most tetracycline-sensitive transductants had received the gsk-3 allele, which enabled them to utilize guanosine as the sole source of purines (5,9). However, rare tetracycline-sensitive transductants, such as CN1980, which had received a disrupted gsk::kan allele, could be identified by their inability to use guanosine as a purine source. Transductants that had retained the wild-type gsk allele also failed to use guanosine as a purine source, but they could be distinguished from the gsk knockouts by their ability to grow well with adenine ϩ guanosine as a purine source.
Verification of Genomic Mutations by Colony PCR Amplification-An aliquot of a bacterial colony was suspended in 25 l of a standard PCR mixture containing 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl 2 , 0.01% gelatin, 0.1 mM deoxynucleotides, 0.2 M oligonucleotide primers, and 0.25 units of Taq polymerase (PerkinElmer Life Sciences). PCR amplifications were performed for 40 cycles (94°C for 30 s, 55°C for 1 min, and 72°C for 1.5 min), followed by 72°C for 7 min. The amplified products were separated on agarose gels and eluted overnight in 50 l of water at room temperature. Direct sequencing of the eluted fragment was performed using ThermoSequenase (U. S. Biochemical Corp.) and a 32 P-labeled primer. Plasmid Constructions-DNA manipulations, transformations, and restriction analyses were performed according to standard procedures (17). PCR amplifications were performed on 1 g of genomic DNA using Pfu polymerase (Stratagene) according to the manufacturer's recommendations. The DNA oligonucleotides used as primers in PCRs are listed in Table II. Plasmids containing cloned PCR fragments were verified by DNA sequencing.
BamHI primers and inserted between the EcoRI and BamHI sites of the medium-copy vector pBR322 (19).
For construction of low-copy rih-lacZ gene fusions, the promoters and N-terminal coding regions of rihA, rihB, and rihC (nucleotides 9481-9933, 8254 -8485, and 6444 -6677, respectively) were PCR-amplified from genomic DNA with the rihA-EcoRI ϩ rihA-HindIII, rihB-EcoRI ϩ rihB-HindIII, and rihC-EcoRI ϩ rihC-HindIII primer pairs, respectively. These PCR products were cloned into the low-copy lacZ vector pCN2423 between a unique EcoRI site upstream of the lac promoter and a unique HindIII site at codon 5 of the lacZ gene. Plasmid pCN2423 has the same overall structure as the previously described mediumcopy vector pCNP5 (20), except that it contains the pSC101 replicon derived from pLG339 (21). High-copy gene fusions of the rih genes to the ␣-lacZ gene were constructed by subcloning of the promoter containing fragments from the low-copy plasmids between the unique EcoRI and HindIII sites of the high-copy vector pUC8 (22).
Assays of Nucleoside Hydrolase Activities-Unless otherwise noted, cells were grown in glycerol minimal medium, with uracil as a pyrimidine source when required. 30 ml of bacterial culture was harvested on ice at A 436 ϭ 0.5. Cells were collected by centrifugation, washed, and resuspended in 50 mM Tris-HCl (pH 7.5) to A 436 ϭ 20. Cells were disrupted by sonic treatment, and the extract was cleared by centrifugation at 20,000 ϫ g for 3 min in a refrigerated microcentrifuge. Assays were performed at 37°C by mixing appropriately diluted extract with 14 C-labeled nucleoside substrate at a final concentration of 1 mM in a total volume of 75 l of 40 mM Tris-HCl (pH 7.5). At time intervals, 15-l samples were taken out, boiled for 2 min, and cooled on ice. After a 3-min centrifugation at 20,000 ϫ g, 5 l of supernatant was applied to a thin-layer chromatography plate for separation of substrate and reaction products, which were subsequently quantitated by counting in an Instant Imager (Packard Instrument Co.). The enzymatic activities were calculated from the initial slope of a plot of the amount of radioactive substrate remaining as a function of time. The reported activities are averages of two independent determinations, which generally deviated Ͻ10% from the average.
Assays of cytidine and uridine hydrolase activities were performed with uniformly 14 C-labeled nucleosides, and thin-layer chromatography was performed with cellulose-coated aluminum sheets (Merck), which were developed in mixtures of pyridine, ethyl acetate, and water. For cytidine hydrolase assays, we used 1:2:2 (v/v) pyridine/ethyl acetate/ water, and for uridine hydrolase, 25:30:50 (v/v) pyridine/ethyl acetate/ water. Assays of purine-nucleoside hydrolase activities were performed with [8-14 C]adenosine, [8-14 C]inosine, [2-14 C]xanthosine, and [U- 14 C] guanosine. Substrates and products were separated by thin-layer chromatography on polyethyleneimine plates developed in water. The 14 Clabeled nucleosides were from several different suppliers; similar products can be obtained from Moravek Biochemicals Inc.
Assays of CMP Hydrolase Activity-Assays of CMP hydrolase activity were performed as described for the nucleoside hydrolase assays, except that [U-14 C]CMP was used as substrate, and reaction products were separated from the substrate by chromatography on polyethyleneiminecellulose plates developed in 1 M acetic acid to 2 cm above the origin and then in 1 M acetic acid and 3 M LiCl (9:1) to 15 cm above the origin.
d The presence of the mutant rih alleles was verified by PCR analysis. e The add mutation in this strain arose spontaneously and was not recognized in the original publication.
side was not required because the plasmid host strain CN2349 is deleted for the lacI gene. RNA Isolation and Primer Extension Analysis-Bacterial RNA was prepared from strains grown in glycerol minimal medium by hot phenol extraction, and primer extension analysis was performed on 5 g of total RNA as described (23). We used a 32 P-labeled DNA primer complementary to nucleotides 41-64 of the lacZ coding sequence (primer 1224) (Table II). Primer extension products were separated on 60-cmlong 8% standard sequencing gels and visualized by autoradiography.

RESULTS
Isolation of rihA::cam Mutants Deficient in Cytidine Salvage-With the aim of identifying the putative CMP hydrolase gene of E. coli, we subjected a pyrF cdd mutant strain, CN1930 (Table I), to transposon mutagenesis with mini-Tn10 cam (16). Following penicillin enrichment, we obtained, by replica plat-ing, 13 mutants that had lost the ability to utilize cytidine as a pyrimidine source while retaining the ability to utilize cytosine. For each mutant, we verified, by back-transduction into CN1930, that chloramphenicol resistance and the cytidinenegative phenotype co-transduced at a frequency of 100%, showing that the transposons were indeed responsible for the inability to utilize cytidine.
Cloning of the cam insert from one of these mutants (CN2258) revealed that the transposon had disrupted a previously uncharacterized open reading frame (ybeK) located at 15 min on the genetic map. Subsequent PCR analysis revealed that all of the original 13 cytidine-negative mutants had a mini-Tn10 insertion in this gene, representing at least six different insertion points (data not shown). A BLAST homology search (24) revealed that ybeK encodes a protein with strong homology to ribonucleoside hydrolases from protozoan parasites. Thus, we have renamed the ybeK gene rihA.
Cytidine Hydrolase Activity of RihA-We originally anticipated that the cytidine-negative mutants would be deficient in CMP hydrolase. However, as shown in Fig. 3a, CN1930 and the rihA derivative CN2258 contained equal levels of CMP hydrolase activity, and this activity was not significantly increased by introduction of the multicopy plasmid pRihA, containing the intact rihA gene. In contrast, the rihA disruption almost eliminated the cellular cytidine hydrolase activity, and introduction of the pRihA plasmid increased it ϳ40-fold relatively to the wild-type level (Fig. 3b). Thus, we concluded that rihA encodes a cytidine hydrolase with no activity for CMP.
As shown in Fig. 4a, the predominant products of the reaction in the extract of CN1930 were ribose and uracil, the latter being formed from the primary reaction product cytosine by cytosine deaminase. When cytidine hydrolase activity was assayed in extracts of ⌬codA strains deficient in cytosine deaminase, the cytidine substrate was converted quantitatively to the primary products cytosine and ribose, with no sign of uracil production (Fig. 4, b and c). Interestingly, cytosine also accu-   mulated as the major reaction product in extracts of pRihA transformants, even if they were CodA ϩ (data not shown), presumably because cytosine deaminase could not keep pace with the increased cytidine hydrolase activity. Unlike cytidine deaminase, the RihA cytidine hydrolase appeared to be strictly specific for ribonucleosides. Even the pRihA transformant with 40-fold increased cytidine hydrolase activity showed no enzymatic activity with deoxycytidine as substrate, and it was completely unable to grow with deoxycytidine as a pyrimidine source.
Role of Cytidine/Uridine Kinase in the Salvage of Cytidine by a cdd Mutant-The genetic evidence for the existence of a CMP hydrolase rested on the finding that growth of pyr cdd strains on cytidine was prevented by mutational inactivation of the udk gene (7). This was very puzzling since our results indicated that cytidine was converted directly to cytosine by the RihA hydrolase. To reinvestigate this problem, we introduced a udk mutation into the pyrF cdd strain CN1930 and compared its effect on the efficiency of cytidine salvage with that of the rihA disruption. The efficiency of cytidine salvage was estimated from the bacterial growth rate with cytidine as a pyrimidine source relative to the growth rate with cytosine.
In glucose minimal medium, the udk mutation did impair cytidine salvage, almost to the same extent as the rihA disruption in CN2258 (Fig. 5a), in agreement with the previous study (7). However, the negative effect of both the udk and rihA mutations was more than eliminated by introduction of pRihA, to the extent that cytidine salvage was no longer growth-limiting. Thus, cytidine/uridine kinase per se was not required for salvage of cytidine in a cdd mutant background, provided the RihA activity was sufficiently high. It should also be noted that even the parental strain (CN1930) was substantially growthlimited by the rate of cytidine salvage, unless transformed with pRihA.
The negative effect of the udk mutation on cytidine salvage in the pyrF cdd background may be rationalized by considering that there are two different pathways for salvage of cytidine in a pyr cdd mutant. The cytidine hydrolase pathway only needs to satisfy the requirement for uracil nucleotides, whereas cytosine nucleotides can be synthesized directly from cytidine via the udk reaction (Fig. 1). In the udk mutant, however, the entire cellular pyrimidine requirement would have to be channeled via the cytidine hydrolase reaction, which was already limiting for the supply of uracil nucleotides in the parental strain (CN1930 ).
With glycerol as a carbon source, on the other hand, the endogenous RihA activity was not limiting for growth on cytidine either for CN1930 or for the udk derivative CN2269 (Fig.  5b). The rihA mutant, however, was still severely compromised for growth on cytidine. These results emphasized that cytidine/ uridine kinase was not necessary for salvage of cytidine in a pyr cdd mutant and further suggested that rihA might be derepressed during growth on glycerol. Accordingly, CN1930 and CN2269 contained 3-fold higher cytidine hydrolase activity when cells were grown on glycerol rather than on glucose (Fig.   FIG. 4. Reaction products of the RihA-mediated turnover of cytidine. a, cytidine hydrolase assays for the CodA ϩ strains CN1930 and CN2258. The position of the cytidine substrate is marked with an arrowhead; the three bands in the marker lane (M) are, from top to bottom, uracil, ribose, and cytosine, respectively. Numbers above each lane indicate the assay time in minutes. b, similar assays for the ⌬codA strains CN1928 and CN2428. c, plot of the radioactivity in the bands of b given as a fraction of the total radioactivity in each lane. f, cytidine; E, cytosine; ‚, ribose; q, cytidine ϩ cytosine ϩ ribose.
FIG. 5. Effect of the rihA and udk mutations on cytidine salvage and cellular cytidine hydrolase activity in glucose and glycerol media. The efficiency of cytidine salvage was estimated from the growth rate with cytidine (CR) as a pyrimidine source relative to that obtained with cytosine (C). All strains had a growth rate with cytosine of ϳ1.1 and 0.7 generations/h in glucose and glycerol media, respectively. a, efficiency of cytidine salvage in glucose medium; b, efficiency of cytidine salvage in glycerol medium; c, cellular cytidine hydrolase activity of cells grown in glucose or glycerol medium (gray and black bars, respectively). 5c). The udk mutation by itself, however, had no significant effect on the cellular cytidine hydrolase activity either in glycerol or glucose medium. Thus, the poor growth of the udk mutant on glucose ϩ cytidine (Fig. 5a) was not caused by a lower activity of cytidine hydrolase compared with the parental strain, but rather by a greater demand for the reaction.
Uridine Hydrolase Activity of RihA-We suspected that the RihA protein, if endowed with uridine hydrolase activity, might constitute the unknown pathway of uridine salvage in udp udk double mutants. To test this possibility, we introduced the rihA::cam disruption into a pyrimidine-requiring udp udk mutant strain (CN2389) to create the udp udk rihA triple mutant (CN2390) ( Table I). As shown in Fig. 6a, the cellular uridine hydrolase activity was more than halved by the rihA disruption in CN2390, whereas introduction of the pRihA plasmid caused a 40-fold increase in activity compared with the RihA ϩ strain CN2389. These results showed that RihA is a uridine hydrolase, and comparison with the data in Fig. 3b indicated that the enzyme was approximately equally efficient with uridine and cytidine as substrates.
Growth experiments confirmed that the uridine hydrolase activity of RihA is a major contributor to the pathway of uridine salvage operating in udp udk mutants (Fig. 6b). In glucose medium, uridine salvage by CN2389 was clearly limited by the endogenous uridine hydrolase activity, which, in this case, had to satisfy the entire pyrimidine requirement. This limitation was further aggravated by the rihA disruption, but was completely eliminated by introduction of pRihA. In glycerol medium, CN2389 was not limited by the rate of uridine hydrolysis, presumably because of the de-repression of the rihA gene (Fig.  5c). In contrast, the rihA derivative CN2390 remained pyrimidine-restricted, although the residual uridine hydrolase activity in this strain (Fig. 6a) allowed for a fairly high growth rate compared with glucose ϩ uridine medium (Fig. 6b). These results suggested glucose-mediated repression of the residual activity, which can probably be ascribed to the RihC nucleoside hydrolase described below.
Selection of Mutants with an Activated rihB Gene and Isolation of rihB::cam Disruption Mutants-Mutants of CN1930 with increased cytidine hydrolase activity were readily obtained by selection for rapid growth in glucose minimal medium with cytidine as a pyrimidine source. One group of mutants were found to contain an amplification of the rihA region and were not characterized further. Another group of three independent mutants, however, seemed to contain a mutation that activated another cytidine hydrolase gene since growth of these mutants on cytidine was not eliminated by introduction of the rihA::cam disruption. The total cytidine hydrolase activity in one of these mutants (CN2264) was increased 3-fold, which allowed for nearly unrestricted growth on glucose ϩ cytidine (at 80% of the growth rate with cytosine as a pyrimidine source).
The mutation responsible for the increased cytidine hydrolase activity in CN2264 was coarsely mapped by conventional methods to be near the cdd gene at 48 min of the genetic map (data not shown). To identify the particular gene affected by the mutation, we subjected the 48 min region of the genome in CN2264 to random insertion mutagenesis with the mini-Tn10 cam transposon and isolated three mutants that had lost the ability to grow rapidly on cytidine. All three strains were found to have a cam insertion at either of two different positions within the yeiK gene, which we have renamed rihB (see "Experimental Procedures"). Since rihB was one of two genes that showed strong homology to rihA in a BLAST search of the E. coli genome, we inferred that the increased cytidine hydrolase activity of the original up-mutant (CN2264) was caused by activation of the rihB locus.
The mutation responsible for activation of the rihB gene in CN2264 was discovered fortuitously during cloning of the cam insert from one of the rihB::cam disruption mutants (CN2324). The restriction pattern of the resulting plasmid deviated from what we expected from the genomic DNA sequence, and DNA sequencing revealed that it was because an IS50 insertion element had integrated 23 base pairs upstream of the rihB gene with the transposase gene in the opposite orientation of rihB. Subsequent PCR analyses revealed that this IS50 insertion was present in the original up-mutant (CN2264) as well as in the two other independent mutants selected for rapid growth on cytidine. In contrast, no insertion was present in the parental strain (CN1930). These findings strongly indicated that the IS50 element activated the rihB gene in the fast-growing mutants, presumably because IS50 contains an outwardly directed weak constitutive promoter (25).
Isolation of a rihC Mutant and a rihA rihB rihC Triple Mutant-Homology searches revealed that rihA and rihB are highly homologous to a third gene, rihC (previously called yaaF), located at 0.6 min. To study the function of this putative nucleoside hydrolase gene, we isolated a rihC::cam mutant, CN2403 (Table I), by PCR screening of a collection of mutants with cam insertions in the vicinity of the carAB operon at 0.6 min. As the rihC::cam mutation was isolated in a prototrophic background, it had no detectable phenotypic consequences. To facilitate studies of the individual nucleoside hydrolases, the disrupted rihA, rihB, and rihC alleles were combined in a prototrophic strain, CN2573 (Table I), which carried additional mutations in the deoD, add, udp, cdd, and udk genes (see Fig.  1). Thus, crude extracts of CN2573 were essentially devoid of nucleoside-metabolizing enzymatic activities, which might interfere with assays of the individual nucleoside hydrolases produced from recombinant plasmids. Despite the nearly complete elimination of nucleoside catabolism in CN2573, it was fully viable and grew with a normal generation of 70 min in glycerol minimal medium. Thus, none of the nucleoside hydrolase genes were essential, either alone or in combination.
Determination of the Substrate Specificity of the Three Nucleoside Hydrolases-To determine the substrate specificity of the three nucleoside hydrolases, we transformed the multiple mutant CN2573 (Table I) with recombinant plasmids pRihA, pRihB, and pRihC, containing the corresponding structural genes (see "Experimental Procedures"). Measurements of hydrolytic activities with different nucleoside substrates in extracts of the transformed strains revealed that both the RihA and RihB enzymes were essentially pyrimidine-specific (Fig. 7). The RihB enzyme, however, did have a clear preference for cytidine over uridine, whereas RihA was equally efficient with either substrate, in agreement with the previous results (Figs.  3b and 6a). The RihC enzyme, on the other hand, was characterized by a remarkably broad substrate specificity for both purine and pyrimidine ribonucleosides, with decreasing activity in the order uridine Ͼ xanthosine Ͼ inosine Ͼ adenosine Ͼ cytidine Ͼ guanosine (Fig. 7). Like RihA, the RihB and RihC enzymes were specific for ribonucleosides, as the transformed strains showed no detectable activity with deoxycytidine or deoxyadenosine (Fig. 7).
Expression and Physiological Capacity of the RihB and RihC Activities-Plasmids pRihA and pRihC are medium-copy plasmids based on the replicon from pBR322, whereas pRihB is a high-copy number plasmid derived from the pGEM3 vector. When cloned into a medium-copy vector, the rihB gene gave rise to a much lower cytidine hydrolase activity (2.6 mol/min/g (dry weight) in glycerol medium), approximating the activity expressed from the chromosomal rihA gene in CN1930 (Fig.  3b). This result suggested that the chromosomal RihB activity was 40-fold lower than the endogenous RihA activity based on the approximate magnitude of the gene dosage effect obtained by cloning in a medium-copy vector (Figs. 3 and 6a). Taking the gene dosage effect into account, the endogenous RihC activity in glycerol medium could be estimated from the data in Fig. 7 to be on the order of 1.3 and 0.3 mol/min/g (dry weight) with uridine and cytidine, respectively, i.e. at least 5-fold higher than the corresponding RihB activities. Thus, the low residual hydrolytic activities for cytidine and uridine in the rihA mutants CN2258 and CN2390 (Figs. 3 and 6a) could probably be accounted for by the endogenous RihC activity.
The physiological capacity of the RihC enzyme was investigated specifically by measuring the efficiency of inosine salvage in a purine-requiring strain, CN1980 (Table I), in which the two other pathways of inosine metabolism had been blocked by mutations in the deoD and gsk genes (see Fig. 1). In glycerol medium, growth of CN1980 with inosine as a purine source was severely limited by the endogenous RihC activity (Fig. 8). This growth limitation was further aggravated by disruption of the rihC gene in CN2545, but was completely eliminated by introduction of pRihC. In glucose medium, on the other hand, both CN1980 and its rihC derivative were essentially unable to grow with inosine as a purine source, and even the pRihC transformant was markedly growth-limited by the rate of inosine hydrolysis. These results indicated that rihC was repressed in glucose medium. Moreover, they demonstrated that the endogenous RihC activity was far from sufficient to satisfy the cellular purine requirement by hydrolysis of inosine, even in glycerol medium.
Regulation of rihA and rihC by Catabolite Repression-The repression of rihA and rihC expression in glucose medium suggested that they might be subject to catabolite repression (reviewed in Ref. 26). This possibility was investigated by the use of gene fusions of the rihA, rihB, and rihC genes to lacZ. As shown in Fig. 9, expression of both the rihA-lacZ and rihC-lacZ fusions was stimulated 3-fold by cAMP in glucose medium, albeit not quite to the level of expression seen in glycerol minimal medium. In contrast, the rihB-lacZ fusion was stimulated Ͻ20% by cAMP, indicating that the low expression of this construct in glucose medium was not caused by catabolite repression. Furthermore, the poor expression in general of the rihB-lacZ fusion supported previous indications that the low endogenous RihB activity was caused by poor expression of the rihB gene, rather than a low intrinsic activity of the enzyme.
Many genes involved in nucleoside catabolism in E. coli, such as cdd and udp and the deo operon, form a regulon that is induced by cytidine via the CytR repressor and the cAMP FIG. 7. Substrate specificity of the RihA, RihB, and RihC enzymes. The hydrolytic activities with different substrates were determined in crude extracts of the multiple mutant CN2573 (Table I) transformed with the indicated plasmids. CR, cytidine; UR, uridine; AR, adenosine; HxR, inosine; XR, xanthosine; GR, guanosine; CdR, deoxycytidine; AdR, deoxyadenosine.
receptor protein (reviewed in Ref. 27). The rih genes, however, did not appear to be part of this regulon. None of the rih-lacZ fusions were significantly induced by cytidine in a wild-type background strain, nor were they induced by any of their other nucleoside substrates added to a concentration of 30 g/ml in glycerol minimal medium (data not shown).
Mapping of the rihA, rihB, and rihC Promoters-The promoters of the rihA, rihB, and rihC genes were tentatively identified by primer extension analysis of mRNAs produced from high-copy gene fusions of the nucleoside hydrolase genes to the N-terminal portion of the lacZ gene (␣-lacZ) in the vector pUC8. As shown in Fig. 10, we obtained one major primer extension signal for each gene fusion, although the signal for the rihB-␣-lacZ construct was very weak. The corresponding mRNA 5Ј-ends were appropriately positioned relative to likely Ϫ10 promoter signals and probably corresponded to the primary transcripts initiated at the rih promoters. None of the promoters contained Ϫ35 signals with strong homology to the consensus sequence 5Ј-TTGACA-3Ј, a feature typical of promoters that depend on specific transcriptional activator proteins. In the rihA promoter, a putative cAMP/cAMP receptor proteinbinding site centered 60 nucleotides upstream of the transcription start site (Fig. 10) did show similarity to the cAMP/cAMP receptor protein consensus sequence (28,29). However, the rihC promoter contained no obvious cAMP/cAMP receptor protein-binding sites, even though this promoter responded as strongly as the rihA promoter to the addition of cAMP (Fig. 9).

DISCUSSION
Enzymology of Nucleoside Hydrolases-Nucleoside hydrolases catalyzing the irreversible hydrolysis of nucleosides to ribose and the free nucleobase are widespread in nature and have been identified in bacteria, fungi, protozoan parasites, fish, and plants, but so far not in mammals (30,31). In most organisms, however, the physiological functions of the nucleoside hydrolases are unknown. A notable exception is the protozoan parasites, which rely on nucleoside hydrolases for salvage of purine nucleosides from the host organism, as they are unable to synthesize purines de novo (32,33).
Based on their substrate specificity, the nucleoside hydro-lases characterized so far may be divided into four different classes (30). One major class consists of relatively nonspecific hydrolases acting on both purine and pyrimidine ribonucleosides. A prominent member of this class is the Crithidia fasciculata inosine/uridine-preferring nucleoside hydrolase (Swiss-Prot accession number O27546), which has been extensively characterized structurally and mechanistically (34 -36).
The second class consists of enzymes that are strictly specific for purine nucleosides, as exemplified by the inosine/adenosine/ guanosine-preferring nucleoside hydrolase from Trypanosoma brucei brucei (GenBank TM /EBI accession number AF017231) (37). The third class consists of the uridine hydrolase encoded by the URH1 gene of Saccharomyces cerevisiae (Swiss-Prot accession number Q04179), an enzyme that is specific for uridine and ribothymidine, but shows no activity for cytidine and purine ribonucleosides (38). Finally, the fourth class is defined by a purine deoxyribonucleoside-specific hydrolase from Leishmania donovani (39). No structural or genetic information is available for this enzyme. The RihC enzyme clearly belongs to the class of nonspecific hydrolases, whereas the RihA and RihB enzymes, being active with both cytidine and uridine, represent a new type of substrate specificity. Functionally, these enzymes are most naturally grouped in the third class with the pyrimidine-specific uridine hydrolase of S. cerevisiae, although this enzyme is inactive with cytidine. Structurally, however, the pyrimidinespecific RihA and RihB enzymes are much more homologous to the nonspecific enzymes RihC and inosine/uridine-preferring nucleoside hydrolase of C. fasciculata (Fig. 11). The greatest homology was found between RihA and RihC (homology score of 44), but all other pairwise homology scores within the group of the inosine/uridine-preferring nucleoside hydrolase and the three E. coli enzymes fell in the range of 35-38, whereas alignments of each of these proteins with Urh1 from yeast gave rise to homology scores only in the range of 25-28. On the other hand, the three E. coli enzymes and the inosine/uridine-preferring nucleoside hydrolase are more homologous to the Urh1 enzyme than they are to the strictly purine-specific inosine/ adenosine/guanosine-preferring nucleoside hydrolase (Fig. 11), which obtained homology scores of only 12-22 in pairwise alignments with each of the five other enzymes.
Thus, from a structural point of view, the pyrimidine-specific hydrolases appear to group with the nonspecific hydrolases; their inability to hydrolyze purine nucleosides may simply derive from a small number of amino acid residues that sterically restrict access of purine nucleosides to the active site. In line with this view, we found that the RihB enzyme did have a low but significant activity for the purine nucleoside xanthosine (Fig. 7). Furthermore, all three E. coli hydrolases were inactive with deoxyribonucleosides as substrates, in analogy with the inosine/uridine-preferring nucleoside hydrolase, which forms specific hydrogen bonds with the 2Ј-hydroxyl of the ribose ring from the conserved residues Asp-14, Asn-39, and Asp-242 in the active site (36) (see Fig. 11). The protozoan hydrolases are not active with nucleoside 5Ј-phosphates as substrate (35); and given the similarity to the inosine/uridine-preferring nucleoside hydrolase, it is hardly surprising that the three E. coli enzymes had no detectable CMP hydrolase activity.
Role of the Putative CMP Hydrolase Pathway in Cytidine Salvage-The present results raise some serious doubts about the involvement of the CMP hydrolase activity in conversion of cytidine to uracil nucleotides in cdd mutants. In glycerol medium, the putative CMP hydrolase pathway was fully dispensable for growth on cytidine, which was efficiently metabolized via the RihA reaction in the cdd udk mutant CN2269 (Fig. 5b). The partial requirement for cytidine/uridine kinase in glucose FIG. 8. Efficiency of inosine salvage via the rihC reaction in a deoD gsk double mutant. The bacterial growth rate with inosine (HxR) as a purine source relative to the growth rate with hypoxanthine (Hx) is taken as an indicator of the efficiency of inosine hydrolysis to hypoxanthine. Gray bars, glucose medium; black bars, glycerol medium. medium (Fig. 5a), on the other hand, might simply be explained by its ability to provide an independent pathway for synthesis of cytosine nucleotides when flux through the RihA reaction was limited by catabolite repression. There is no genetic evidence that any of the cytidine channeled via the udk reaction in glucose medium was converted to uracil nucleotides via the CMP hydrolase reaction. In glycerol medium, very little cytidine appeared to be converted to cytosine via this reaction, as shown by the poor growth of the rihA mutant on glycerol ϩ cytidine (Fig. 5b).

Physiological Function and Regulation of the Ribonucleoside
Hydrolases in E. coli-The physiological role of the three ribonucleoside hydrolases in wild-type E. coli strains is enigmatic, as their activities (except for the hydrolysis of cytidine) are paralleled by the reversible nucleoside phosphorylases (Fig. 1). A similar redundancy exists for the 3Ј-exonucleolytic RNases, where the hydrolytic activities of RNases II and D are paralleled by the phosphorolytic enzymes polynucleotide phosphorylase and RNase PH, respectively (reviewed in Ref. 41). Unfortunately, the physiological rationale for the redundancy of the RNases is no less elusive than it is for the nucleosidecatabolizing enzymes.
The finding that both rihA and rihC were subject to catabolite repression might suggest a role for these genes in the provision of ribose for utilization as a carbon source. However, their low levels of expression were far from sufficient to allow utilization of nucleosides as carbon sources in strains lacking the nucleoside phosphorylase activities (data not shown). The low capacity of the nucleoside hydrolase reactions might hint that the true natural substrates of these enzymes are not the common ribonucleosides, but rather some low-level nucleoside analogs, such as modified nucleosides derived from turnover of tRNA and rRNA. Alternatively, the nucleoside hydrolases might be sufficiently induced to function in bulk metabolism of the common nucleosides under certain physiological conditions. However, we found no evidence for induction of these genes by their ordinary nucleoside substrates, so the hypothetical inducing conditions are presently unknown.
It is noteworthy that the silent rihB gene is located immediately upstream of an open reading frame, yeiJ (Fig. 2b), which encodes a protein with strong homology to the nucleoside transporter NupC. Preliminary results indicate that YeiJ is indeed a nucleoside transporter and that the yeiJ gene is activated coordinately with rihB by the IS50 insertion upstream of the rihB gene. 2 Thus, it is likely that these genes constitute an inducible operon devoted to the transport and catabolism of some nucleoside that is only occasionally present in the environment.
Apart from this, the genetic organization of the rih genes gives no clues to their physiological function. The rihA gene is flanked by the gltJKL operon and an open reading frame (ybeW) that encode a transport system for aspartate and glutamate and a homolog of the DnaK chaperone, respectively (Fig. 2a). The identification of the rihA promoter suggested that expression of this gene is independent of the upstream gltJKL cluster, but it is presently unclear if ybeW might be co-transcribed with rihA. The rihC gene apparently constitutes a monocistronic operon (Fig. 2c). The downstream dapB gene contains a promoter of its own (42), and our identification of the rihC promoter corroborated previous studies that indicated that rihC is not co-transcribed with the upstream genes of the ileS gene cluster (43).
The preferential isolation of mutants with amplifications of the rihA gene or IS50 insertions upstream of rihB in selections for increased cytidine hydrolase activity suggested that these genes are not controlled by simple repressor proteins. So far, attempts to identify gene-specific regulatory loci for the rih genes by genetic selection have been unsuccessful, but we expect that further studies on the regulation of these genes may eventually reveal their true physiological function.