Regulation of Dinucleoside Polyphosphate Pools by the YgdP and ApaH Hydrolases Is Essential for the Ability of Salmonella enterica serovar Typhimurium to Invade Cultured Mammalian Cells*

The ygdP and apaH genes of Salmonella enterica serovar Typhimurium (S. Typhimurium) encode two unrelated dinucleoside polyphosphate (NpnN) hydrolases. For example, YgdP cleaves diadenosine tetraphosphate (Ap4A) producing AMP and ATP, while ApaH cleaves Ap4A producing 2ADP. Disruption of ygdP, apaH individually, and disruption of both genes together reduced intracellular invasion of human HEp-2 epithelial cells by S. Typhimurium by 9-, 250-, and 3000-fold, respectively. Adhesion of the mutants was also greatly reduced compared with the wild type. Invasive capacity of both single mutants was restored by transcomplementation with the ygdP gene, suggesting that loss of invasion was due to increased intracellular NpnN. The normal level of 3 μm adenylated NpnN (ApnN) was increased 1.5-, 3.5-, and 10-fold in the ygdP, apaH and double mutants, respectively. Expression of the putative ptsP virulence gene downstream of ygdP was not affected in the ygdP mutant. Analysis of 19 metabolic enzyme activities and the ability to use a range of carbohydrate carbon sources revealed a number of differences between the mutants and wild type. The increase in intracellular NpnN in the mutants appears to cause changes in gene expression that limit the ability of S. Typhimurium to adhere to and invade mammalian cells.

The dinucleoside polyphosphates (Np n N) 1 are a ubiquitous family of nucleotides found at micromolar to submicromolar concentrations in which two nucleoside moieties are linked 5Ј-5Ј through a polyphosphate chain containing from two to seven phosphoryl groups. The most widely studied are diadenosine 5Ј,5ٞ-P 1 ,P 3 -triphosphate (Ap 3 A) and diadenosine 5Ј,5ٞ-P 1 ,P 4 -tetraphosphate (Ap 4 A), for which several functions have been suggested, although none yet conclusively proved (1)(2)(3). In Escherichia coli, Ap 4 A has been proposed to couple DNA replication to cell division (4,5) and to participate in stress responses by modulating protein refolding by chaperones (3,6,7). Ap 4 A and related adenylated dinucleotides (e.g. Ap 3 N and Ap 4 N, where N ϭ any nucleoside) are synthesized mainly by aminoacyl-tRNA synthetases, although other ligases have been shown to synthesize them in vitro (8,9). In Gram-negative bacteria, the predominant enzyme believed to be responsible for Ap 4 N hydrolysis is the symmetrically cleaving diadenosine tetraphosphatase, ApaH. This enzyme, which is active toward many Np n N nucleotides (n Ն 3), degrades Ap 4 A to two moles of ADP and Ap 5 A to ADP and ATP (10 -12) and is structurally related to serine/threonine protein phosphatases (13,14). Deletion of the E. coli apaH gene leads to a 10 to 100-fold increase in intracellular Ap 4 N (15,16).
Recently, a second prokaryotic dinucleoside polyphosphate hydrolase was discovered. The IalA protein from the invasive pathogen Bartonella bacilliformis is a member of the Nudix (nucleoside diphosphate linked to X) hydrolase family and hydrolyzes Ap 4 A asymmetrically to AMP and ATP and Ap 5 A to ADP and ATP (17,18). It is closely related to the eukaryotic Ap 4 A hydrolases, particularly those from plants, which also hydrolyze many Np n N species, where n Ն 4 (12,18). The ialA gene has been implicated indirectly in the process of cellular invasion by this bacterium; expression of B. bacilliformis ialA in non-invasive E. coli renders it invasive (19). Furthermore, the orthologous ygdP gene from E. coli K1 may be required for the invasion of human brain microvascular endothelial cells as its expression is up-regulated by invasion-enhancing growth conditions and down-regulated by invasion-repressing conditions (20). YgdP and the related InvA protein from Rickettsia prowazekii preferentially hydrolyze Ap 5 A (21,22). Since the intracellular levels of several Ap n N species are known to increase substantially under conditions of oxidative stress (23,24), we previously suggested that the ability to metabolize Np n N may be necessary for invasion in the face of an oxidative attack by the invaded cell (18). If that were so, then the apaH gene would be expected to be essential for invasion as well. Indeed, a DNA fragment from the oral pathogen Actinobacillus actinomycetemcomitans that confers an invasive ability on E. coli contains the apaH gene (25).
We have, therefore, examined the effects of deleting the ygdP and apaH genes both singly and doubly on the invasive ability of Salmonella enterica serovar Typhimurium (S. Typhimurium), a facultative intracellular parasite that can invade and multiply within various cell types, including phagocytes and epithelial cells, and so establish a chronic infection. Our results provide the first direct evidence of the involvement of these bacterial genes in intracellular invasion.

EXPERIMENTAL PROCEDURES
Reagents, Bacterial Strains, and Plasmids-Recombinant human Ap 4 A hydrolase was prepared as described for the Caenorhabditis elegans enzyme (26). Strains and plasmids used and produced in this study are listed in Table I. Plasmids were introduced into E. coli by transformation and into S. Typhimurium strain LT2 by electroporation using a Bio-Rad Gene Pulser II.
DNA Manipulation and Analysis-DNA ligation, restriction analysis, and gel electrophoresis were carried out as described by Sambrook et al. (28).
Cloning, Expression, and Purification of ApaH and YgdP-The coding regions of the apaH and ygdP genes were PCR-amplified from S. Typhimurium genomic DNA. ApaH was amplified using the 5Ј primer (5Ј-ATTATACATATGGCAACTTATCTCATC-3Ј) containing an NdeI site and the 3Ј primer (5Ј-TTTCGGGATCCTGGAGCGTC-3Ј) containing a BamHI site and inserted after restriction digestion between the NdeI and BamHI sites of pET15b (Novagen) to give pET-ApaH. YgdP was amplified using the 5Ј primer (5Ј-AAGGTTTATCCATGGG-TAGTCCGGTG-3Ј) containing an NcoI site and the 3Ј primer (5Ј-AATATTCAGCGCCTCGAGCAGAC-3Ј) containing a XhoI site and inserted after restriction digestion between the NcoI and XhoI sites of pET32b (Novagen) to give pET-YgdP. For expression, plasmids were transformed into E. coli BL21(DE3). The His-tagged recombinant proteins were expressed and purified on NiCAM TM -HC resin (Sigma) as described previously (29).
Construction of apaH and ygdP Disruption Cassettes-Gene disruption cassettes were generated by PCR according to Wach et al. (30). An apaH::kan cassette was constructed as follows. First, a 300-bp 5Ј segment of the apaH gene was amplified from S. Typhimurium genomic DNA using the 5Ј primer HP1 (5Ј-TCCCCCGGGATGAACGTTACGTT-TTTGC-3Ј) and the 3Ј primer HP2K (5Ј-TGCAGGTCGACGGATCCGG-CGATCAGTTCGTCGTAG-3Ј), where HP1 corresponds to part of the 5Ј non-coding region of apaH and HP2K contains the 5Ј end of the coding region of apaH and the 5Ј end of the non-coding region of the kanamycin resistance gene from pUC4K. A 285-bp 3Ј segment of the apaH gene was also amplified using the 5Ј primer HP3K (5Ј-GCTCGATGAGTTT-TTCTAATGCGCTGGGAAGATAAACAG-3Ј) and the 3Ј primer HP4 (5Ј-CATGTCTAGACGCAGAATTCTCACAGCTATTG-3Ј), where HP3K contains the 3Ј end of the coding region of apaH and the 3Ј end of the non-coding region of the kanamycin-resistance gene and HP4 corresponds to part of the 3Ј non-coding region of apaH. In the second step, the complete kanamycin resistance gene was amplified from pUC4K using HP1, HP4, and the two PCR products from the first step as primers to give the apaH::kan cassette in which the kanamycin resistance gene is flanked by Ͼ250 bp segments of the apaH gene. A ygdP::kan disruption cassette was produced in a similar manner using primers YP1 (5Ј-CATGTCTAGACTTAGATGTGATGCTGGTCA-3Ј), YP2K (5Ј-TGCAGGTCGACGGATC ATCATCAATCACCGGAC-3Ј), YP3K (5Ј-GCTCGATGAGTTTTTCTAAGCTCAG GATAATCC-3Ј), and YP4 (5Ј-CATGTCTAGAATGATCGGCACACCGAG-3Ј). Finally, an apaH::cat cassette containing the chloramphenicol resistance gene from pLysS was produced using the primers HP1, HP2C (5Ј-GGGACACCA-GGATTTATGGCGATCAGTTCGTCGTAG-3Ј), HP3C (5Ј-GTGGCAGG-GCGGGGCGTAATGCGCTGGGAAGATAAACAG-3Ј), and HP4.
Construction of Single and Double apaH and ygdP Deletion Mutants of S. Typhimurium-ApaH::kan and ygdP::kan disruption cassettes were first cloned into the pGEM®-T Easy vector (Promega) to give plasmids pGEM-AK and pGEM-YK respectively and the plasmids transformed into TOP10 E. coli for propagation. Recovered plasmids were digested with ApaI and SpeI and the cassettes purified after gel electrophoresis using a Qiagen purification kit. Cassettes were then ligated into the cut suicide vector pMRS101 (31). The ligation mixture was transformed into E. coli K12 CC118 (pir) and the resulting plasmids, pTI201 and pTI202, electroporated into S. Typhimurium LT2. The transformants were isolated on LB medium supplemented with 50 g/ml kanamycin and 50 g/ml streptomycin and subcultured again on LB medium containing 10% sucrose to select deletants resulting from a double crossover event. Surviving colonies were then subcultured again on kanamycin-containing medium to give strains STYA201 (⌬apaH::kan) and STYY202 (⌬ygdP::kan). The apaH::cat chloramphenicol disruption cassette was cloned into pGEM®-T Easy in a similar manner to give pGEM-AC then transferred to pMRS101 to give pTI203, which was introduced into STYY202. Colonies of the ⌬apaH::cat ⌬ygdP::kan double mutant STYAY203 were selected as above on sucrose and on medium containing 25 g/ml chloramphenicol and 50 g/ml kanamycin. All deletions were confirmed by PCR analysis of three independent isolates.
Invasion and Adhesion Assays-HEp-2 epithelial cells and U937 macrophage-like cells were maintained in Eagle's minimal essential medium (Invitrogen) supplemented with 5% (v/v) fetal calf serum and 0.15% Na 2 HCO 3 . HEp-2 cells were seeded on coverslips in vials at 1 ϫ 10 5 cells/coverslip and grown overnight at 37°C in 1 ml of medium. U937 cells were split into 1-ml cultures at a density of 2 ϫ 10 5 cells/ml on the day of use. After inoculation with 50 l of overnight bacterial cultures (4 ϫ 10 9 /ml), cells were incubated at 37°C in a 5% CO 2 for 3 h. For invasion assays, cells were incubated for 1 h with 25 g/ml gentamicin, washed five times with 1 ml of phosphate-buffered saline, then lysed in 0.5 ml 0.5% sodium deoxycholate. Lysates were diluted in phosphate-buffered saline and the viable count determined on LB agar plates (containing 50 g/ml kanamycin for mutants). For adhesion assays, incubation with gentamicin was omitted. In addition, HEp-2 monolayers were washed with phosphate-buffered saline after adhesion, fixed with methanol, and stained with a 10% solution of Giemsa prior to examination by light microscopy.
Metabolic Phenotype-Nineteen different enzyme activities of the wild type (WT) S. Typhimurium and its apaH, ygdP, and apaH ygdP double null mutants were examined using the API ZYM kit (bio-Mérieux, Basingstoke, UK). The ability of the WT and mutant strains to metabolize a variety of different carbohydrates was determined by using the API 50 CH system (bioMérieux). This system tests assimilation, oxidation, and fermentation of the carbohydrate sources.
Extraction and Assay of Ap n N-Cells from 50-ml cultures of S. Typhimurium (WT and mutants) in mid log phase (OD 0.6 -0.7) were collected by rapid centrifugation (5000 ϫ g for 5 min). Pellets were resuspended in 5 ml of ice-cold 0.4 M trichloroacetic acid and shaken for 15 min. Neutralization, alkaline phosphatase digestion, and purification of the dinucleotide-containing fraction were as described previously (32). The freeze-dried extract was dissolved in 0.1 ml 30 mM Hepes-NaOH, pH 7.7, 5 mM magnesium acetate and triplicate 25-l samples each mixed with 25 l of luciferin/luciferase ATP-monitoring reagent (Bio-Orbit). After measuring the background luminescence, 1 ng of recombinant human Ap 4 A hydrolase was added to generate ATP and the increase in luminescence determined. This generalized assay measures all ATP-generating nucleotides of the form Ap n N, where n Ն 4. Samples (10 l) of the neutralized acid extract were also retained for luminometric ATP determination before alkaline phosphatase digestion. These were mixed with 90 l of 30 mM Hepes-NaOH, pH 7.7, 5 mM magnesium acetate, and triplicate 25-l samples then each added to 25 l of luciferin/luciferase ATP-monitoring reagent and the luminescence determined. Previous ATP determinations had revealed no significant differences in intracellular ATP between any of the strains (values Ϯ S.E. were 10.1 Ϯ 1.5, 9.5 Ϯ 0.6, 10.9 Ϯ 1.6, and 9.7 Ϯ 1.3 nmol/mg of protein (n ϭ 3) in wild type LT, STYA201, STYY202, and STYAY203, respectively). Therefore, the ATP content of the extracts acts as an internal standard for the extraction process. The intracellular concentration of Ap n N (n Ն 4) was then estimated from the Ap n N/ATP ratio making the following assumptions: (i) that, for all strains, the intracellular ATP concentration is 3 mM (33); (ii) that Ap n A is 50% of the Ap n N pool (34); (iii) that the human Ap 4 A hydrolase used in the assay cleaves equally efficiently at either end of each Ap n N and so 1 mol of Ap n A yields 1 mol of ATP, while 1 mol of Ap n N yields 0.5 mol of ATP. The latter two assumptions require the figure for Ap n N concentration to be multiplied by 1.33 to compensate for the 50% efficiency in ATP production from Ap n N (N A).
Assay of ApaH and YgdP Activity-Activity of YgdP with Ap 4 A, Ap 5 A, and Ap 6 A was measured luminometrically by direct continuous assay of the ATP product and kinetic parameters determined by nonlinear regression analysis (26). Activity of ApaH with Ap 5 A was measured in the same way, while activity with Ap 4 A required the inclusion of 2 mM phosphoenolpyruvate and 5 g of pyruvate kinase in the assay to convert the ADP product to ATP (35). In addition, all ApaH assays contained 100 M CoCl 2 (10,11). Product identification was by high performance liquid chromatography as described previously (26). Reverse

Properties of the ApaH and YgdP
Proteins-To confirm that the S. Typhimurium ApaH and YgdP proteins had the enzymic activities predicted from their sequences, they were cloned and expressed in E. coli BL21(DE3) cells: ApaH in pET15b as a His-tagged 33.6-kDa protein and YgdP in pET32b as a Histagged thioredoxin fusion protein of total mass 39.2 kDa. When purified to homogeneity (Fig. 1A), the enzymes had the expected activities. ApaH efficiently hydrolyzed Ap 4 A, Ap 5 A, and Ap 6 A, always producing ADP as one product, while YgdP hydrolyzed the same nucleotides, with a preference for Ap 5 A, like the E. coli and R. prowazekii enzymes (21,22), and always producing ATP as one product. Both enzymes followed Michaelis-Menten kinetics with all substrates tested; representative plots for the hydrolysis of Ap 4 A by both enzymes are shown in ApaH, YgdP, and Ap n N in WT and Mutant Cells-⌬apaH (STYA201) and ⌬ygdP (STYY202) null mutants and a ⌬apaH ⌬ygdP double null mutant (STYAY203) were generated by replacement of the genes with antibiotic resistance cassettes (Table I). Gene deletion was confirmed by PCR and by measurement of Ap 4 A hydrolytic activities in cell extracts. ApaH and YgdP have predicted pI values of 4.8 and 10.0, respectively. Thus, they can be measured independently after separation by batch anion-exchange chromatography at pH 7.5. Table II confirms the absence of the enzymes in the appropriate mutants. The total concentration of nucleotides of general structure Ap n N (n Ն 4) was also measured using a luciferase-based assay in which human Ap 4 A hydrolase is used to generate ATP from Ap n N (n Ն 4) compounds. WT cells had 3.6 M Ap n N, which compares favorably with previous figures of 3 M in both S. Typhimurium (23) and E. coli (15). Deletion of ygdP led to a slight, 1.5-fold increase in Ap n N and deletion of apaH to a 3.5-fold increase, while deletion of both genes led to a 10-fold increase. These results show that both YgdP and ApaH contribute to control of the Ap n N pool in S. Typhimurium. As non-adenylated Np n N, which are not detected by the assay, are also substrates for these two hydrolases, it is likely that their levels also increase. Hence the Ap n N pool measurements provide a rough indication of the effects of deleting the hydrolase genes but do not yet convey the detailed picture.
Invasion-The ability of the mutants to invade HEp-2 epithelial cells was determined using a gentamicin protection assay (36). Deletion of ygdP (STYY202) reduced invasion by 9-fold compared with the WT, while deletion of apaH (STYA201) reduced invasion by 250-fold. Deletion of both genes (STYAY203) produced a dramatic 3000-fold reduction (Fig. 2). Importantly, transformation of both STYY202 and STYA201 with the YgdP expression plasmid pTrc-YgdP restored full invasive capacity. This strongly suggests that loss of invasion is primarily related to a common property of YgdP and ApaH, i.e. the hydrolysis of Np n N, rather than to some protein-specific function. Transcomplementation of apaH in STYA201 partially restored invasion (18-fold) but appeared to have no effect on STYY202 (Fig. 2). This incomplete restoration may indicate poor expression of the cloned apaH gene or that the specific Ap n N or Np n N responsible for creating the non-invasive phenotype are better substrates for YgdP. It also suggests that the full phenotypes of the single mutants have elements that are distinct, in addition to the common feature and consequences of increased Ap n N and Np n N. Similar data were obtained with U-937 macrophage-like cells.
Adhesion of the single mutants to both HEp-2 and U-937 cells was measured by recovery of colony-forming units from washed cell cultures. Both STYA201 and STYY202 showed a dramatically reduced ability to adhere to either cell type (Fig.  3). This was confirmed by microscopic examination after Giemsa staining (36).
Cell Growth, Morphology, and Motility-Apparent doubling times in LB determined from optical density measurements were similar for WT, STYA201, and STYY202 (20 -22 min). However, STYA201 showed a much longer lag-phase after inoculation compared with the others (4.5 versus 1.5 h) and did not attain as high a final cell density. Microscopic examination showed STYA201 formed long filaments, 20 -30-times normal length. Filamentous growth has previously been observed as a consequence of apaH deletion in E. coli (16).
Metabolic Activities-When tested for expression of a range of metabolic enzyme activities using the API ZYM and API 50 CH systems, WT cells were found reproducibly (n ϭ 3) not to express cystine arylamidase, ␣-glucosidase, or ␤-glucuronidase activity, whereas such activity was detectable in each of the three mutants. The ␣-glucosidase activity was relatively lower in STYY202 compared with STYA201 and STYAY203. No differences were observed in the expression of the remaining enzymes (alkaline phosphatase, C4 esterase, C8 esterase/ lipase, C14 lipase, leucine arylamidase, valine arylamidase, trypsin, ␣-chymotrypsin, acid phosphatase, naphthol phosphohydrolase, ␣-galactosidase, ␤-galactosidase, ␤-glucosidase, N-acetyl-␤-glucosaminidase, ␣-mannosidase, and ␣-fucosidase). With regard to carbohydrate utilization the only differences were between the WT and STYA201 and STYAY203. The WT and STYY202 strains were able to utilize myo-inositol and D-tagatose, whereas STYA201 and STYAY203 could not (n ϭ 3). These results indicate that apaH and ygdP deletion has substantial but specific effects on the cells.
PtsP Expression-In S. Typhimurium and other enterobacteria, ygdP is in an operon with the downstream ptsP gene. E. coli ptsP encodes Enzyme I Ntr , a component of a P-enolpyru-vate-dependent phosphotransferase system believed to be involved in the regulation of RpoN-dependent operons (37). PtsP has been shown to be a virulence factor in Pseudomonas aeruginosa (38) and Legionella pneumophila (39), hence, it was important to show whether insertional deletion of ygdP had also disrupted ptsP expression. Reverse transcription-PCR analysis showed that ptsP expression was unaffected. An identical PCR  product was amplified from WT and STYY202 total RNA using ptsP gene-specific primers (Fig. 4). Thus, the phenotype of STYY202 is not simply due to loss of PtsP expression. DISCUSSION Many genes and gene products are required for the successful invasion of mammalian cells by S. enterica and similar bacteria (40,41). To these we can now add ygdP and apaH. Previously, we suggested that the B. bacilliformis YgdP homologue, IalA, was required for intracellular invasion to enhance intracellular survival by restricting potentially deleterious increases in Np n N levels caused by oxidative attack by the invaded cell (18). This hypothesis was based on the well established increases in the levels of Np n N in cells exposed to oxidising agents (23,42). However, the data here show that, at least in S. Typhimurium, an increase in Np n N has profound effects on the bacteria in the absence of mammalian cells which subsequently reduce their ability to adhere to and then invade the cells. This does not support a role for these enzymes in combating host cell-mediated oxidative stress. The most convincing argument for the involvement of Np n N in the invasive phenotype is the transcomplementation of the ⌬apaH mutant by ygdP.
The increases in total Ap n N measured in the mutants were less than initially expected, given the 10 -100-fold increases previously reported upon deletion of the E. coli apaH gene alone (15,16). However, it is not known if ygdP is expressed in the E. coli strains used in these studies; its lack of discovery until recently (and then only as a recombinant product) suggests that it may not be. In contrast, it is clearly expressed in S. Typhimurium LT2 (Table II). The modest 3.5-fold increase in Ap n N upon apaH deletion appears to be sufficient to cause the filamentous, non-invasive phenotype. However, it is possible that this overall figure for Ap n N (comprising predominantly Ap 4 N) hides a more dramatic rise in one or more specific minor species that is/are primarily responsible for the phenotype or that a non-adenylated, and therefore undetected, Np n N is the critical species involved. A detailed analysis of the specific Np n N content of the mutants is clearly required.
Of the two Np n N hydrolases, loss of ApaH has much the greater effect. In E. coli, deletion of apaH has previously been shown to cause filamentation and to decrease transcription of motility and chemotaxis genes by inhibiting expression of the RpoF alternative sigma factor (16). RpoF is itself regulated by the cAMP/cAMP-binding protein complex via the flhDC master operon (43), and since the transcription of other cataboliterepressible, cAMP/CAP-controlled genes such as lacZ and galK is also substantially reduced, increased Np n N may directly interfere with the production or function of the cAMP⅐CAP complex (16). RpoF also regulates flagellar operons in S. Typhimurium (44), so similar changes may occur in S. Typhimurium when Np n N is increased. Indeed we have observed a complex pattern of changes in flagellar and fimbrial expression in the mutants. 2 This could be a major factor in the loss of adhesion and invasion and is currently under investigation.
With regard to a possible functional connection between ygdP and ptsP, both of which are virulence genes and which together comprise an operon, it is possible that increased Np n N may affect PtsP (Enzyme I Ntr ) directly by binding to its regulatory domain (37). Expression of ygdP along with ptsP may be necessary to ensure the proper control of Np n N concentration under conditions where the activity of Enzyme I Ntr is required. Thus, increased Np n N may affect the transcription of RpoN-, as well as RpoF-, regulated genes. Such genes are significant in number (43,45). What is clear is that extensive but specific metabolic changes have occurred in the null mutants as a consequence of an increase in some Np n N species. This is shown by the changes in metabolic enzymes and carbohydrate utilization that would be expected consequences of changes in transcription factor activity. In R. prowazekii, the YgdP homologue InvA is co-expressed with a putative two-component response regulator protein Rrp (46). A clear role for InvA in the entry of rickettsiae into animal cells has not yet been shown, so it is too early to draw a comparison between these two rather different systems.
It will now be of interest to determine whether the above effects of ygdP and apaH deletion on the invasion of cultured cells extend to a reduction in virulence in an animal model system. If so, then YgdP and particularly ApaH, which has no known mammalian orthologue, may represent useful targets for new antibacterial agents. Substrate analogue inhibitors of E. coli ApaH have already been synthesized that could serve as starting points for the design of inhibitors of intracellular invasion (47,48).