The relationship between dNTP pool levels and mutagenesis in an Escherichia coli NDP kinase mutant

Abstract

Loss of nucleoside diphosphate kinase (Ndk) function in Escherichia coli results in an increased frequency of spontaneous mutation and an imbalance in dNTP pool levels. It is presumed that the imbalance in dNTP pool levels is responsible for the mutator phenotype of an E. coli ndk mutant. A human homologue of Ndk and potential suppressor of tumor metastasis, nm23-H2, can complement the mutagenic phenotype of an E. coli ndk mutant. Here, we show that the antimutagenic property of nm23-H2 in E. coli is independent of dNTP pool levels, indicating that dNTP pool imbalance is not responsible for the mutator phenotype associated with the loss of ndk function. We have identified multiple genetic interactions between ndk and genes involved in the metabolism of dUTP, a potentially mutagenic precursor of thymidine biosynthesis. We show that loss of ndk function is synergistic with a dut-1 mutation and synthetically lethal with the loss of thymidine kinase function. Our results suggest that Ndk prevents the accumulation of dUTP in vivo. Based on these results and biochemical studies of Ndk, we propose that the mutagenic phenotype of an ndk mutant is caused by excess misincorporation of uracil in place of thymidine combined with a defect in the uracil base excision pathway.

In Escherichia coli the DnaA protein initiates bidirectional DNA replication at a unique origin of replication, oriC, in a cell-cycle dependent manner (1). In addition to its role as the initiator of DNA replication, DnaA is a transcriptional regulator, coordinately regulating the expression of numerous genes with respect to the cell cycle (2, 3). One set of genes directly regulated by the DnaA protein encodes the enzyme ribonucleotide reductase, known as NrdAB or RNR (4). Because reduction of nucleoside diphosphates (NDPs) is the rate-limiting step in the production of dNTPs, regulated expression of RNR with respect to initiation of DNA replication ensures sufficient dNTPs will be produced during the replication period (5).

dNTP pools are limiting for DNA replication, their levels in the cell being only ≈1% of the amount necessary for a single round of chromosomal replication (1, 6). Furthermore, reducing the concentration of a single dNTP can cause a decrease in the rate of replication fork progression in vivo (7). Improperly regulated synthesis of dNTPs by altered expression and/or activity of RNR in Saccharomyces cerevisiae can be mutagenic and affect cell-cycle progression (8, 9).

We have previously shown that mutations in ndk, which encodes nucleoside diphosphate kinase (Ndk), decrease the relative rate of replication fork progression in vivo, suggesting an important link between proper regulation of dNTP synthesis and replication fork elongation (10). NDP kinases are a broad and well conserved family of enzymes present in all domains of life. In vitro, NDP kinases are capable of generating all (d)NTPs from their corresponding (d)NDPs through a well characterized phosphotransferase reaction (11). Based on the ability of NDP kinase to generate all (d)NTPs in vitro it is assumed to be the major enzyme involved in dNTP synthesis in vivo. But NDP kinases from many organisms are dispensable; their loss has little or no effect on cell growth (12–16).

The human genome contains eight Ndk paralogues, Nm23 H1–H8, which have been implicated in multiple cellular processes. For example, human NDP kinase (nm23-H1) was initially identified as a suppressor of tumor metastasis by using metastatic melanoma cell lines (17). Nm23-H2 has been shown to cleave DNA in a mechanism similar to that of the AP lyase family of DNA repair enzymes, and directly regulate expression of the c-MYC oncogene (18, 19). Mutational loss of NDP kinase function in Drosophila leads to developmental defects (20). Interestingly, restoration of NDP kinase enzymatic activity is necessary but not sufficient to complement this mutant phenotype (21).

E. coli ndk mutants have an alteration in dNTP pool levels, which is thought to give rise to a weak mutator phenotype associated with the loss of ndk function (13). Expression of human nm23-H2 complements the mutagenic phenotype of an E. coli ndk mutant (22). Ndk has also been shown to cleave DNA and functionally interact with the uracil glycosylase, Ung (23, 24). The interaction between Ndk and Ung is reported to lower the K_m of Ung for its substrate, uracil containing DNA (23). Ung is responsible for the excision of uracil, which can be misincorporated in place of thymidine (25). Although dUTP is an essential precursor in the de novo synthesis of thymidine, it is necessary that its concentration be kept low to prevent such misincorporation (26). Indeed, because of the action of the essential enzyme dUTPase, which converts dUTP to dUMP, the dUTP concentration is estimated to be <1% of the total dTTP concentration (27, 28).

In an effort to better understand the role of Ndk in DNA replication and mutagenesis, we used a human nm23-H2 construct to determine whether its antimutagenic effect was caused solely by NDP kinase phosphotransferase activity. Here, we show that human nm23-H2 is able to complement the mutagenic phenotype of an E. coli ndk mutant without restoring dNTP pool levels. This finding provides definitive evidence that the dNTP pool imbalance associated with loss of ndk function is not responsible for its mutator phenotype. Furthermore, we show that ndk genetically interacts with dut and tdk, two genes specifically involved in dUTP/dUMP metabolism, and provide evidence that increased mutagenesis in an ndk mutant is likely occurring from misincorporaton of uracil in place of thymidine. Based on these results, we propose that Ndk functions to prevent excess dUTP accumulation in vivo and that E. coli Ndk and human Nm23-H2 possess DNA repair activities specific to uracil base excision repair.

Results

dNTP Pool Imbalances Are Not Responsible for Increased Mutation Frequency in an ndk Mutant.

To gain further insight into the role of Ndk in DNA replication and mutagenesis, we fused the human ndk homologue, nm23-H2, to the E. coli ndk promoter and cloned this construct in a low-copy plasmid (≈10 copies per cell). Nm23-H2 has been shown to complement the mutator phenotype of an E. coli ndk mutant (22). To test the functionality of our nm23-H2 construct, we introduced the pnm23-H2 plasmid into an E. coli ndk mutant and measured the frequency of spontaneous mutagenesis (see Materials and Methods). The frequency of spontaneous rifampicin (Rif) resistance in the ndk mutant was ≈6- to 15-fold increased relative to that in WT cells (Fig. 1A and Table 1). Expression of E. coli ndk and human nm23-H2 in trans reduced the frequency of spontaneous mutagenesis to levels comparable with that of WT cells (0.55- and 0.84-fold increase, respectively; Fig. 1A and Table 1). These results confirm that nm23-H2 complements the mutator phenotype of an ndk mutant, as previously reported, and indicate that our nm23-H2 construct is functional (Fig. 1A).

Fig. 1.

Human nm23-H2 does not fully complement an E. coli ndk mutant. (A) Frequency of spontaneous Rif resistance in four to seven independent cultures for each strain tested. Median and interquartile ranges are presented. The median values are 2.5, 38.46, 1.38, and 2.10 for WT, ndk, ndk/pndk, and ndk/pnm23-H2, respectively. (B) Cells of a dnaA(cos) or dnaA(cos) ndk double mutant harboring the indicated plasmid were grown to log phase at 42°C, 10-fold serially diluted, and incubated at 42°C or 30°C for 18 or 24 h, respectively.

Table 1.

dNTP pool imbalances are not responsible for the increased mutagenic frequency in an ndk mutant

Assay	WT	ndk	ndk/pndk	ndk/pnm23-H2
dCTP*	1	2.38 (±0.58)	1.02 (±0.26)	2.71 (±0.45)
dGTP*	1	2.19 (±0.32)	1.49 (±0.17)	3.06 (±0.49)
dATP*	1	0.27 (±0.02)	1.08 (±0.19)	0.76 (±0.16)
dTTP*	1	2.76 (±0.90)	1.47 (±0.20)	2.69 (±0.26)
RifR /108 CFU†	1	15.38	0.55	0.84
NDK activity‡	1	0.091 (±0.02)	0.797 (±0.13)	0.167 (±0.03)

*For relative dNTP pool measurements, values represent the mean of three independent cellular extracts. The values in parentheses indicate the SEM.

^†For Rif^R resistance frequencies, values represent the fold differences relative to WT of four to seven independent cultures. The actual median values are represented in Fig. 1.

^‡For NDP kinase activity, values represent the mean of three independent extract preparations. Values in parentheses indicate the SEM.

We previously identified loss of Ndk function as a suppressor of dnaA(cos)-mediated overinitiation that acts independently of initiation of DNA replication. Furthermore, we showed that loss of Ndk function results in a decreased rate of replication fork progression, which is likely responsible for the suppression of overinitiation (10). To test whether mutagenesis has a role in this novel phenotype associated with loss of Ndk function in E. coli, we introduced the pnm23-H2 plasmid into a dnaA(cos) ndk double mutant to ask whether nm23-H2 could complement the double mutant and restore cold sensitivity. As shown in Fig. 1B, the pnm23-H2 construct did not restore cold sensitivity to the dnaA(cos) ndk double mutant. This finding suggests that Nm23-H2 lacks an activity associated with E. coli Ndk function.

Loss of Ndk in E. coli results in an imbalance in dNTP pool levels, which is thought to give rise to the mutator phenotype. Specifically, the level of dCTP was reported to increase ≈20-fold and the level of dGTP ≈7-fold with only modest changes in dATP and dTTP when compared with WT cells (13). However, the fluctuation in dNTP pool levels and their role in mutagenesis became unclear when it was reported that strain background differences affected dNTP pool levels, but not the frequency of spontaneous mutagenesis in an ndk mutant (29). Furthermore, mutational bias in an ndk mutant is inconsistent with the notion that changes in dNTP pool levels, specifically an increase in dCTP and dGTP, are responsible for mutagenesis (22).

To address these issues, dNTP pool levels were assayed in WT and ndk mutant cells complemented with either E. coli ndk or nm23-H2. As shown in Table 1, loss of ndk function gives rise to a modest change in dNTP pool levels, with ≈2- to 3-fold increases in dCTP, dTTP, and dGTP and a ≈4-fold decrease in dATP. These values are consistent with results reported by Shen et al. (29) and support the notion that dNTP pool measurements are affected by strain background differences. Imbalances in dNTP pool levels could be corrected by providing E. coli ndk in trans, but not nm23-H2. These results demonstrate that an imbalance in dNTP pool levels is not responsible for the increased frequency of spontaneous mutation in an ndk mutant. dNTPs are limiting for DNA replication, the dNTP pool being only ≈1% of what is needed for one round of chromosomal replication (1, 6). Lack of complementation of the dnaA(cos) ndk double mutant by nm23-H2 (Fig. 1B) suggests that the reduction in the dATP pool level is responsible for the decreased rate of replication fork progression associated with the loss of ndk function. Limitation of the intracellular thymidine pool is known to have a similar effect on the rate of replication fork progression (7).

Because an ndk mutant strain expressing nm23-H2 behaves like an ndk null mutant with respect to dNTP pool levels, but no longer has the mutator phenotype of an ndk mutant, we asked whether we could detect NDP kinase activity in cellular extracts derived from the nm23-H2-expressing strain (see Materials and Methods). Nm23-H2-expressing cells had only ≈2-fold increased NDP kinase activity compared with the ndk mutant, whereas the ndk strain complemented with E. coli ndk showed NDP kinase activity similar to that of WT cells (79% relative to WT; Table 1). Taken together, these results indicate that the antimutagenic activity of nm23-H2 is independent of dNTP pool levels and NDP kinase activity. Thus, low-level expression of nm23-H2 does not provide all of the functions missing in an ndk mutant.

A dut-1 Mutation Synergistically Increases the Spontaneous Mutation Frequency of an ndk Mutant.

Biochemical studies of E. coli Ndk have identified a functional interaction between Ndk and uracil glycosylase, Ung (23). This interaction has been shown to lower the K_m of Ung for its substrate, uracil-containing DNA, which led to the hypothesis that Ndk could have a role in repair of uracil-containing DNA. Human Nm23-H2 has been shown to cleave DNA by a mechanism similar to that of AP lyases, and E. coli Ndk has also been shown to cleave DNA, but the specific amino acid residue necessary for this activity has not yet been defined (18, 24). However, the residue necessary for Nm23-H2 DNA cleavage activity (K11) is conserved in E. coli Ndk (K12), suggesting both enzymes cleave DNA in a similar manner. Taken together, these results suggested to us that Ndk could have a role in uracil metabolism. To test this hypothesis, we combined an ndk mutation with known mutations that affect uracil metabolism and assayed the level of spontaneous mutagenesis in the resulting strains.

For the de novo synthesis of thymidine, dCTP is deaminated to form dUTP, which is rapidly hydrolyzed to dUMP by the enzyme dUTPase (Dut). dUMP is then methylated to form dTMP (26). Dut is an essential protein (30, 31), but a dut-1 mutant strain, which retains only 2–3% the activity of WT Dut activity is capable of supporting growth (30, 32). A dut-1 mutant has a modest effect on spontaneous mutation frequency and loss of Ndk function results in ≈6-fold increase in mutation frequency compared with WT cells (Table 2). An ndk dut-1 double mutant, in contrast, was found to have a ≈3,000-fold increase in mutagenesis (Table 2). This synergistic genetic interaction suggests that loss of Ndk function results in increased levels of dUTP and that normal Dut activity is necessary to maintain the dUTP pool at a lower level. Expression of Ndk in trans completely reversed this increase in mutation frequency, whereas reversal by Nm23-H2, expressed in trans, was incomplete (≈8-fold increase in mutagenesis relative to WT; Table 2). These results are consistent with the idea that Nm23-H2 functions independently of dNTP pool regulation.

Table 2.

Frequency of spontaneous Rif resistance

Strain	RifR CFU/108 CFU*	Fold increase†
WT	1.67 (0.48–3)	1
ndk	10.63 (7.6–29.3)	6.37
dut-1	1.28 (0.39–3.2)	0.77
dut-1 ndk	5,005 (3,682−15,623)	2,997‡
dut-1 ung	3.39 (0.8–10.2)	2.03
dut-1 ndk ung	524 (428–3,348)	314‡
ung	2.48 (1.99–12.8)	1.49
ung ndk	8.90 (6.8–11.3)	5.33
dut-1 ndk/pndk	1.04 (0.66–1.52)	1
dut-1 ndk/pnm23-H2	7.69 (5.96–9.56)	7.69
dut-1 ndk/pBAD33	3,703 (3,272–12,151)	3,578

*Values represent the median from six to 12 independent cultures for each strain tested and parentheses represent the 95% CI of the median or the 85% CI of the median for the complemented strains.

^†Fold increase relative to WT (ED8566).

^‡A t test indicates P = 0.0065 when comparing the dut-1 ndk double mutant and the dut-1 ndk ung triple mutant.

To determine whether the synergistic increase in mutation frequency in the dut-1 ndk double mutant was caused by excessive uracil incorporation, we introduced an ndk mutation into a dut-1 ung double mutant. The resulting triple mutant had a substantial increase in mutation frequency (≈314-fold relative to WT; Table 2) but had a ≈10-fold reduced frequency compared with the dut-1 ndk double mutant (P = 0.0065). Taken together, these results suggest that the synergistic effect on mutagenesis observed in the dut-1 ndk double mutant is likely caused by replication intermediates generated from the uracil excision process, because stable incorporation of uracil into DNA can partially alleviate this mutagenic effect.

To further address the relationship between dNTP pool imbalance and mutagenesis, we measured the dNTP pool levels in ndk, dut-1, dut-1 ndk, and dut-1 ung ndk mutant cells and compared them with those of WT cells. As previously shown, ndk cells had only a slight alteration in dNTP pool levels relative to WT cells (Tables 1 and 3). Also, a dut-1 mutation had only a modest effect on dNTP pool levels, resulting in a slight increase in dCTP, dATP and dTTP (1.27-, 1.64-, and 1.78-fold, respectively) and a slight decrease in dGTP (31% reduction). The dut-1 ndk double mutant and the dut-1 ndk ung triple mutant had very similar dNTP pool levels, a ≈2-fold increase in dCTP being the only significant difference when compared with ndk mutant cells (Table 3). Although there was no difference in the dNTP pool levels between the double and triple mutant strains, there was a ≈10-fold difference in their mutation frequency.

Table 3.

Relative dNTP pool measurements

dNTP	WT	ndk	dut-1	dut-1 ndk	dut-1 ndk ung
dCTP	1	4.51 (±0.46)	1.27 (±0.06)	7.05 (±0.80)	8.32 (±1.04)
dGTP	1	1.86 (±0.13)	0.69 (±0.10)	1.03 (±0.06)	1.36 (±0.17)
dATP	1	0.79 (±0.05)	1.64 (±0.11)	0.54 (±0.08)	0.79 (±0.14)
dTTP	1	3.68 (±0.16)	1.78 (±0.30)	3.87 (±0.05)	4.34 (±0.83)

Values represent the average of three independent cellular extracts. The values in parentheses indicate the SEM.

To further analyze the severe mutagenic effect in the dut-1 ndk double mutant, we asked whether there was a mutational bias associated with that strain. Mutational bias was previously determined for an ndk mutant by using mutation to Rif resistance (22), but the limited range of mutations known to give rise to Rif resistance led us to screen instead for lac constitutive mutants that result from loss of lacI function (33). If the mutational event that gives rise to the lac^const phenotype is caused by excessive dUTP incorporation and uracil excision, mutations affecting thymidine residues (AT base pairs) should be overrepresented. Indeed, of 47 independent lacI^const mutants isolated, 92% (43/47) resulted from mutations originating at an AT base pair with only 2% (1/47) originating at a GC base pair (Table 4).

Table 4.

Bias of mutations in the lacI gene leading to a lac^const phenotype in a dut-1 ndk⁻ double mutant

Base change	Occurrence, %	AT or GC bias, %
AT→GC	79 (37/47)
AT→CG	4 (2/47)	92
AT→TA	9 (4/47)
GC→AT	2 (1/47)
GC→TA	0 (0/47)	2
GC→CG	0 (0/47)
Other*	6 (3/47)	6

The lacI gene = 43.7% AT, 56.3% GC. n = 47.

*Other indicates (−1) frameshifts or a nucleotide insertion.

Loss of Thymidine Kinase Function Is Synthetically Lethal with the Loss of Ndk Function.

Results presented in this study, as well as biochemical studies by Goswami et al. (23), indicate that Ndk has a direct role in uracil metabolism. Based on this hypothesis we performed a genetic screen to identify genes that are synthetically lethal with loss of ndk function to better understand the physiological role of Ndk in vivo. For this screen we used a miniF plasmid that is unstable in the absence of antibiotic selection; plasmid pJN65 contains the lacZ gene and the ndk gene under control of its own promoter. JN319/pJN65 cells were subjected to EZ::TN <KAN-2> insertional mutagenesis (Epicentre) and then plated on LB medium supplemented with kanamycin and X-Gal. A total of ≈60,000 colonies from two independent experiments were screened for nonsectored solid blue colonies. Four mutations were identified in this primary screen, and two were verified by their inability to backcross the EZ-Tn5 insertion into the parental ndk strain (JN319). Sequence analysis showed that the mutations were caused by independent insertion events within the tdk gene, which encodes thymidine kinase.

To validate the synthetic lethal interaction between ndk and tdk, we constructed a strain containing an unmarked deletion of the tdk ORF (see Materials and Methods) and transformed this strain with plasmids expressing either ndk or nm23-H2 or with the plasmid vector alone. We then attempted to introduce an ndk null allele, ndk::EZ-Tn5, into these strains by P1 transduction. A xerD::EZ-Tn5 allele was used as a control for P1 transduction efficiency. Consistent with a synthetic lethal effect, efficient transduction yielding the tdk ndk double null mutant occurred only in the case where Ndk was expressed in trans from the plasmid (Fig. 2). In the case of cells expressing Nm23-H2, a reduced number of transductants, forming smaller than normal colonies was obtained, suggesting that Nm23-H2 only partially complemented the ndk/tdk synthetic lethal phenotype (Fig. 2). Furthermore, the Nm23-H2-complemented tdk ndk strain displayed a temperature-sensitive growth defect when plated at 42°C [supporting information (SI) Fig. S1].

Fig. 2.

Loss of tdk function is synthetically lethal with loss of ndk function, which is only partially complemented by human nm23-H2. A single representative P1 transduction quantifying the efficiency that an ndk::EZ-Tn5 allele was transduced into a Δtdk mutant harboring the indicated plasmids is shown. A xerD::EZ-Tn5 allele was used as a control for P1 transduction frequencies. * indicates no colonies were detected.

Thymidine kinase is able to use dU or dT interchangeably. However, the tdk/ndk synthetic lethality likely reflects a dUMP-specific interaction, because loss of thymidylate synthase (thyA) or thymidine phosphorylase (deoA) function had no effect when combined with an ndk mutation (data not shown).

In an attempt to determine whether suppression of dUTP formation affected any of the genetic interactions described above, we introduced a dcd (dCTP deaminase) null mutation into an ndk mutant. Approximately 75% of the dUTP in the cell is derived from deamination of dCTP to form dUTP, a process catalyzed by dCTP deaminase (26). To our surprise, we observed a genetic interaction between dcd and ndk. Cells of a dcd ndk double mutant gave rise to small heterogeneous colonies, whereas neither single mutant did so (Fig. S2). Furthermore, propagation of a dcd ndk double mutant rapidly gave rise to nonmutagenic, fast-growing derivatives, likely because of extragenic suppression (unpublished data).

Discussion

Our motivation for studying the role of NDP kinase in DNA replication and mutagenesis in E. coli has been 2-fold. First, it provides a genetically tractable system to study how dNTP synthesis and dNTP pool imbalances affect DNA replication and mutagenesis. Second, it serves as a model system to study the physiological role of Ndk in vivo, which is apparently more complex than predicted, based solely on the known biochemical properties of Ndk.

Studies of E. coli ndk mutants led to the paradigm that dNTP pool imbalances are responsible for the increased frequency of spontaneous mutagenesis (13). Our results indicate that expression of the human Ndk homologue, Nm23-H2, can complement the mutagenic phenotype of an E. coli ndk mutant, but is unable to correct the imbalance in dNTP pool levels. Therefore, dNTP pool imbalance is not responsible for the increased mutagenesis seen in an ndk mutant.

In attempts to address the role of Ndk in mutagenesis, we have uncovered multiple genetic interactions between ndk and genes involved in the metabolism of dUTP that suggest that Ndk prevents the formation of excess dUTP in vivo. We propose that an intermediate generated by the excision of uracil is responsible for mutagenesis in an ndk mutant, which is exacerbated when the dUTP pool is elevated by a dut-1 mutation. AP (abasic) sites are known to be highly mutagenic intermediates of the uracil excision pathway and could provide a mechanistic explanation for the increase in mutagenesis. In support of this model, stable uracil incorporation in a dut-1 ndk ung triple mutant displays a reduction in mutagenesis when compared with the dut-1 ndk double mutant. The fact that the dut-1 ndk ung triple mutant still has an elevated mutation frequency likely reflects an increase in dUTP pool levels caused by the combination of a dut-1 and ndk mutation. These observations, in combination with the known biochemical interaction of Ndk with Ung (23), and the fact that Ndk is able to cleave DNA (24), suggest that Ndk is at the interface of dUTP formation and uracil base excision repair (Fig. 3). Attempts to measure the frequency of uracil incorporation into DNA in the dut-1 ndk double mutant strain by using a sensitive plasmid based assay developed by Kouzminova and Kuzminov (34) were unsuccessful because of the high level of plasmid instability in this strain background (unpublished data).

Fig. 3.

E. coli NDP kinase genetic and biochemical interactions suggest that ndk could function at the interface between dNTP synthesis and base excision repair. The uracil glycosylase base excision repair pathway (Left) and the pathway of de novo thymidylate biosynthesis pathway (Right) are shown. Arrows with asterisk indicate genetic interactions identified in this study. The Ndk/Ung functional biochemical interaction was identified by Goswami et al. (23). Additionally, Ndk and human Nm23-H2 have been shown to cleave DNA and suggested to do so in a mechanism similar to AP lyases (18). The schematic representation of the uracil BER pathway was adapted from ref. 25.

If mutagenesis were occurring through uracil misincorporation in place of thymidine and/or defects in excision of uracil, we would expect an overrepresentation of mutations originating at AT base pairs, because U:G mismatches and abasic sites are both mutagenic. Indeed, 92% of all of the mutations we identified were caused by mutational events at AT base pairs. These data agree with an analysis of mutational bias in an ndk mutant by Miller et al. (22), who also found that ndk mutants are hypermutable upon loss of mismatch repair (MMR) activity. Excess uracil misincorporation could give rise to U:G mispairs, which would subsequently be recognized by the MMR system. ndk mutants and ndk mutS double mutants have nearly identical mutational biases, which supports this hypothesis. But, we cannot rule out the possibility that the dNTP pool imbalance in an ndk mutant gives rise to a low level of mispairs that are efficiently detected by the MMR system. It is important to note that this low level of mispairing is distinct from the phenotypic mutagenesis seen in a MMR-proficient ndk mutant. We have found that Nm23-H2 only modestly alleviates mutagenesis in an ndk mutS double mutant (Table S1). Regardless, because all methods used to detect mutational bias show an increase in mutations originating at AT base pairs it is likely that misincorporation of uracil is the underlying cause.

Our finding that a mutation in the thymidine kinase gene, tdk, is synthetically lethal with an ndk mutation, supports our hypothesis that Ndk is involved in uracil metabolism. Under conditions where dUTP metabolism is compromised, Tdk is thought to degrade dUMP to dU to help decrease the dUTP pool (26). In support of this hypothesis, we find that tdk mutants are sensitive to sublethal doses of fluorodeoxyuridine (Fig. S3). Also, loss of tdk function is synthetically lethal with loss of recA function, and although the mechanism is not entirely clear, it is likely caused by alterations in dUTP/dUMP metabolism because amplification of tdk can suppress dut-1 recA synthetic lethality (35).

Why do ndk mutants have an imbalance in dNTP pool levels but no effect on rNTP pools? To date we are unaware of a proposed mechanism to explain the dNTP pool imbalance. Based on the observation that Ndk likely acts to prevent dUTP formation in vivo, either directly or indirectly, we can propose a mechanism leading to dNTP pool imbalances in an ndk mutant. Because the enzyme dCTP deaminase, which converts dCTP to dUTP, is feedback-inhibited by dUTP, an increase in the dUTP concentration would inhibit dCTP deaminase activity, leading to an increase in dCTP pool levels. Alternatively, if an ndk mutant accumulated excess dCTP, it would be deaminated, resulting in an increase in both dCTP and dUTP pool levels. Both mechanisms would predict a concomitant increase in dTTP formation, which is observed in an ndk mutant. dTTP is an allosteric regulator of RNR activity and would, in turn, alter the specificity of RNR to reduce GDP to dGDP. Because dTTP is continually being synthesized because of increased dUTP formation, insufficient dGTP would be available to specify the reduction of ADP, leading to a decrease in dATP pool levels. This model is consistent with the dNTP pool measurements associated with loss of Ndk function presented in this study and those reported by Mathews and coworkers (29).

NDP kinases are able to synthesize all (d)NTPs in vitro, but loss of Ndk function in organisms as diverse as E. coli, Bacillus anthracis, Pseudomonas aeruginosa, S. cerevisiae, and Schizosaccharomyces pombe has little or no effect on cell growth (12–16). Given this fact, combined with the observation that loss of Ndk function has no effect on ribonucleotide pools, which are in vast excess over dNTP pools (13), we propose that NDP kinase is not the primary determinant of (d)NTP synthesis in vivo. Adenylate kinase of E. coli is known to be capable of synthesizing all (d)NTPs in vitro (36) and is thus a candidate for performing this function in vivo.

Materials and Methods

Bacterial Strains, Media.

E. coli K-12 strains used in this study are listed in Table S2. All strains were grown in L broth (LB) or M9 minimal medium supplemented with tryptophan and glucose. P1 transductions were carried out as described (37). Kanamycin (Kan; 30 μg/ml), 10 μg/ml chloramphenicol, 10 μg/ml tetracycline, 10 mM uracil, 40 μg/ml X-Gal, 25–50 μg/ml ampicillin, 250 μM isopropyl-β-d-thiogalactoside, and 100 μg/ml Rif were added when necessary.

Plasmid Constructions.

All restriction enzymes and T4 DNA Ligase were from New England Biolabs. All primers for PCR were synthesized by Integrated DNA Technologies (IDT), and DNA polymerase was Triple Master (Eppendorf). All DNA constructs were verified by DNA sequencing at the Tufts University Core Facility. The nm23-H2 ORF was amplified from plasmid pQE60nm23-H2 (22) with the primers 1331 (5′-TTTACAGAGGTAAAAATGGCCAACCTGGAGCGCACC-3′) and 1333 (5′-TTATTCATAGACCCAGTCATGAGCAC-3′). The E. coli ndk promoter region was amplified by using the primers 1299 (5′-CTGGATCCCGCGACAGTGAAATTTGTCATGC-3′) and 1332 (5′-GGTTGGCCATTTTTACCTCTGTAAATTGTTCTGTTGTTG-3′). These two PCR products were used together in an overlapping primerless PCR and then supplemented with primers 1299 and 1333. The resulting PCR product containing the E. coli ndk promoter region fused to the nm23-H2 ORF was ligated into pCR2.1 TOPO TA vector (Invitrogen), resulting in plasmid pJN15. The ndk^Pr-nm23-H2 containing HindIII and XbaI fragment from pJN15 was ligated into HindIII/XbaI-digested pBAD33 (38), resulting in pJN22. To create an inducible tdk allele, tdk was cloned into plasmid pDSW204 (39). The tdk ORF was PCR-amplified by using primers 1520 (5′- CCTCTAGACATTGAGGGCCTGTGGCTGATG-3′) and 1521 (5′- GGAAGCTTTTAATCGTGGCGATGCCTTTCC-3′), digested with XbaI/HindIII, and ligated into XbaI/HindIII-digested pDSW204.

The unstable lac⁺/ndk⁺ miniF plasmid (pJN65) was created by PCR amplifying the ndk gene and upstream promoter region with the oligo 1299 (5′- CTGGATCCCGCGACAGTGAAATTTGTCATGC-3′) and oligo 1515. The resulting PCR product was digested with BamHI and HindIII and ligated into BamHI/HindIII-digested pFZY1 (40).

Spontaneous Mutation Frequency Determination.

To determine the spontaneous mutation frequency, independent colonies were used to inoculate cultures that were grown overnight at 37°C with aeration. Cultures were serially diluted in sterile saline and plated onto LB plates for total CFUs or LB plates supplemented with 100 μg/ml Rif. The total number of Rif^R CFUs was divided by the total CFUs to determine mutation frequency. The mutation frequencies for the dut-1 ndk double mutant (JN264) and the dut-1 ung ndk triple mutant (JN266) was determined by using independent purified transductants from a cross of JN57 × JN263 or JN57 × JN265 and selecting for Kan^R (ndk::EZ-Tn5) transductants. Because of rapid suppression of the mutator phenotype it was necessary to limit manipulation of these strains. Complementation analysis of the dut-1 ndk double mutant was carried out by crossing JN57 with JN263/pBAD33, JN263/pJN15, or JN263/pJN22, selecting for Kan^R, and assaying independent transductants.

dNTP Pool Measurements.

Overnight cultures were diluted 1:200 in LB and grown to an OD₆₀₀ ≈0.2. To ensure balanced growth, these cultures were then diluted 1:10 in 10 ml of LB and grown at 37°C to an OD₆₀₀ ≈0.1. Cultures were then either immediately filtered through a 0.45-μm nitrocellulose filter (Millipore) to collect the cells or chilled rapidly in an ice-water bath before filtration. The total CFUs were determined for each culture, and Rif^R CFUs were determined for highly mutagenic cultures. Three independent cultures were used for each strain.

Immediately after filtration, filters were vortexed in 5 ml of ice-cold 60% methanol and placed at −20°C overnight. The methanolic suspension of cells was boiled for 5 min and cells were removed by centrifugation. Because of volume differences in the final evaporation step, extracts were assayed for their absorbance at OD₂₆₀ to quantitatively compare the nucleotide concentrations between each extract.

dNTP pool measurements were conducted as described (41). For each measurement, three extract concentrations were used to ensure that the extract did not inhibit the enzymatic reaction. The relative amounts of dNTPs were calculated by: average cpm /μl extract × OD₂₆₀⁻¹ for WT and divided by the value obtained of the relative extracts.

NDP Kinase Assay.

Overnight cultures were diluted 1:200 in 100 ml of LB and grown to OD₆₀₀ ≈0.5. Cultures were rapidly chilled on an ice water bath and centrifuged. Pellets were washed in10 mM Tris (pH 7.4), 10 mM MgCl₂,10 mM NaCl, 10 mM EDTA,and 1 mM DTT and stored at −80°C overnight. Pellets were thawed on ice and resuspended in 2 ml of the above buffer, lysed by sonication, and centrifuged (≈16,000 × g) to remove cell debris. Lysates were dialyzed overnight at 4°C in 10 mM Tris (pH 7.4), 10 mM NaCl, 10 mM MgCl₂, 10 mM EDTA, and1 mM DTT by using 3,500 MW dialysis tubing (Spectra/Por). Extract protein concentration was determined by a Bradford assay. A pyruvate kinase/lactate dehydrogenase-coupled assay was used to determine NDP kinase activity as described (42) with CDP as the acceptor nucleoside diphosphate.

Synthetic Lethal Screen.

pJN65 (lac⁺/ndk⁺) was introduced by transformation into a lac ndk strain of E. coli (JN319), rendering it phenotypically Lac⁺/Ndk⁺ in the presence of antibiotic selection. Once antibiotic selection is withdrawn, pJN65 is readily lost because ndk is a nonessential gene, giving rise to blue/white sectored colonies on LB plates supplemented with X-Gal. Synthetic lethal interactions can be identified by looking for mutant cells unable to lose the plasmid and that form solid blue colonies on medium supplemented with X-Gal. Strain JN319/pJN65 was subjected to EZ::TN <KAN-2> transposon mutagenesis according to the manufacturer's recommendation (Epicentre) and plated on LB containing Kan and X-Gal. Candidate EZ-Tn5 insertions were sequenced by random PCR (43).