Differential control of dNTP biosynthesis and genome integrity maintenance by the dUTPase superfamily enzymes

dUTPase superfamily enzymes generate dUMP, the obligate precursor for de novo dTTP biosynthesis, from either dUTP (monofunctional dUTPase, Dut) or dCTP (bifunctional dCTP deaminase/dUTPase, Dcd:dut). In addition, the elimination of dUTP by these enzymes prevents harmful uracil incorporation into DNA. These two beneficial outcomes have been thought to be related. Here we determined the relationship between dTTP biosynthesis (dTTP/dCTP balance) and the prevention of DNA uracilation in a mycobacterial model that encodes both the Dut and Dcd:dut enzymes, and has no other ways to produce dUMP. We show that, in dut mutant mycobacteria, the dTTP/dCTP balance remained unchanged, but the uracil content of DNA increased in parallel with the in vitro activity-loss of Dut accompanied with a considerable increase in the mutation rate. Conversely, dcd:dut inactivation resulted in perturbed dTTP/dCTP balance and two-fold increased mutation rate, but did not increase the uracil content of DNA. Thus, unexpectedly, the regulation of dNTP balance and the prevention of DNA uracilation are decoupled and separately brought about by the Dcd:dut and Dut enzymes, respectively. Available evidence suggests that the discovered functional separation is conserved in humans and other organisms.

Proper control of the intracellular concentration of deoxyribonucleoside-5-triphosphates (dNTPs), the building blocks of DNA, is critically important for efficient and high-fidelity DNA replication and genomic stability 1,2 . Three of the four canonical dNTPs are synthesized from their respective ribonucleoside diphosphate (NDP) counterparts 3 . The direct precursor for dTTP, however, is missing from the ribonucleoside pool and is synthesized via separate routes (Fig. 1).
The de novo synthesis of dTTP occurs through uracil base-containing precursors: dUMP is the direct input into the thymidylate synthase reaction (Fig. 1). In most organisms, the main dUMP supply is provided by the deamination of a cytosine deoxyribonucleotide (dCMP or dCTP) while other possible routes, e.g. the dephosphorylation of dUDP, are considered to be minor supplements [4][5][6] . When cytosine deamination occurs at the triphosphate level, the resulting dUTP is then converted into dUMP. The enzymes that catalyze these conversions belong to the dUTPase superfamily comprising dCTP deaminase (Dcd), dUTPase (Dut) and the bifunctional dCTP deaminase/dUTPase (Dcd:dut) (Fig. 1). These enzymes share the same quaternary structure as shown in Fig. 2A.
In addition to dUMP production, the dUTPase reaction also serves to eliminate excess dUTP to prevent uracil incorporation into DNA in place of thymine 7,8 . Although not mutagenic when replacing thymine, the uracil in DNA is considered to be an error and induces uracil-excision repair mechanisms 9 . In high dUTP/dTTP ratios, however, DNA polymerases keep re-incorporating dUTP and the repair process becomes overwhelmed. Dut is ubiquitous and essential in most investigated cases [10][11][12][13][14][15][16] . Recently, novel functions of Dut emerged in gene expression regulation as well 15,[17][18][19][20] . Our genetic experiments also suggested that the mycobacterial Dut has a yet unknown but essential moonlighting function 11 .
In summary, Dut catalyzes the break-down of dUTP to dUMP and with this action it potentially takes part i) in dTTP biosynthesis, ii) in the maintenance of low dUTP/dTTP ratio to prevent uracil incorporation into DNA and iii) in interactions with regulatory proteins. The various roles now attributed to Dut and the large amount of knock-out and knock-down data on the dUTPase superfamily enzymes in various genetic backgrounds create a confusing picture of the contribution of Dut to the physiological processes in which it may be involved. As dTTP biosynthesis is an essential process and a major target in several current drug therapies, it is important to pinpoint those pathways in which Dut is a key contributing enzyme.
We therefore set-out to dissect the contributions of dUTP-hydrolyzing enzymes, Dut and Dcd:dut, to dTTP biosynthesis and to the prevention of DNA uracilation. For this reason, we searched for a simple model in which the obligatory dTTP precursor, dUMP, is produced exclusively by Dut and Dcd:dut in lack of salvage pathways and dCMP deamination (Fig. 1). This favorable set of conditions naturally occurs in the genus Mycobacteria 11,21 . Due to the exclusive biosynthetic role of Duts in these organisms, they present potential targets for drug development, as well. Earlier mutagenesis studies found the presence of the bifunctional Dcd:dut to be dispensable for growth in M. tuberculosis 10,22 while the intact Dut protein is essential in Mycobacteria 10, 11, 22 . In the present study, we created M. tuberculosis Dut mutant proteins in which the enzyme activity is gradually tuned down. We then carried out genetic experiments in the fast-growing M. smegmatis in which we created the same Dut mutations and also included an inactive dcd:dut mutant strain in the experiments. We found that the dUTPase activity of either Dut or Dcd:dut can support cell growth. The double mutant M. smegmatis strain lacking the complete dUTPase activity, however, is inviable. We investigated the mutation-induced effects including  Both structures contain the non-hydrolysable substrate analog α-β-imido-dUTP (dUPNPP) in the active sites. (B) Enlarged view of the active site of M. tuberculosis Dut showing the C-terminal arm in green. The side chains of the amino acids in case of point mutations and C-terminal arm truncation are shown with atomic colored stick representation similarly to the dUPNPP molecule and with green cartoon representation, respectively. The catalytic water is shown as a red sphere while the yellow sphere denotes the Mg 2+ ion that coordinates the nucleotide. dNTP pool changes, the mutation rate and the uracil content of DNA in M. smegmatis strains conferring various Dut and Dcd:dut mutants. Unexpectedly, the lack of Dut activity did not influence the biosynthesis of dTTP. We arrived to the conclusion that dTTP biosynthesis and the maintenance of genomic integrity by dUTP elimination are under differential control.

Results
Tuning down the activity of M. tuberculosis Dut. On the basis of our previous investigations on the human and Escherichia coli (E. coli) Duts 23, 24 , we planned and created three mutants of the M. tuberculosis Dut enzyme by site-directed mutagenesis. These mutants were chosen to represent enzymatic activity loss from one order of magnitude to the practical inactivity.
The D83N substitution aims at compromising the coordination of the catalytic water by mutating the catalytic aspartate residue (Asp90 in E. coli Dut) (Fig. 2B). In effect, this mutant presents an extremely low catalytic activity (Fig. 3A, Table 1) similarly to what was observed in the E. coli enzyme 24 . The determination of the Michaelis constant (K M ), however, is uncertain due to the limitations of the activity measurements at such low activities. The K M could be better estimated in this case from the dissociation constant (K d ) of the protein complexed with the non-hydrolysable substrate analog α, β-imido-dUTP (dUPNPP). We measured the K d of the D83N.dUPNPP complex to be similar to that of the WT.dUPNPP complex (Table 1, Fig. S1B). Its catalytic efficiency being 0.0002 compared to the wt means that the D83N Dut is a practically inactive mutant.
The efficient dUTPase catalysis requires conserved sequence motifs I-V from all three monomers of the Dut homotrimer 7 . Motifs I-IV constitute the active sites at the intermonomer clefts while motif V, located at the C-terminal arm, leaves the globular core of its monomer to associate with the neighboring active site and shield it from the solvent (C-terminal arm shown in green in Fig. 2B). This P-loop-like motif V changes conformation upon substrate binding and positions the phosphate chain of the nucleotide for efficient hydrolysis. The lack of this motif results in a nearly inactive enzyme in all investigated species 23,[25][26][27][28][29][30][31][32][33] . We created a mutant lacking conserved motif V by the truncation of the 154 amino acid long protein at position 138 (T138Stop). The T138Stop mutant exhibits 3-fold higher enzymatic activity than that of the D83N mutant the catalytic efficiency still being  extremely low compared to the WT (Fig. 3A, Table 1). Both the K d of the enzyme.dUPNPP complex (Fig. S1A) and the K M are 4-fold higher than that of the WT complex (Table 1). We created the S148A mutation to distort one of the hydrogen bonding interactions between the P-loop-like motif and the γ-phosphate of the substrate dUTP (Fig. 2B). The S148A mutant showed one order of magnitude loss in the enzyme activity (Fig. 3A Fig. S1A).
The A115F mutation inactivates the Dcd:dut enzyme. We cloned and expressed the M. tuberculosis and M. smegmatis Dcd:dut enzymes and determined their steady-state kinetic parameters. In lack of significant difference between the behaviors of the M. tuberculosis and M. smegmatis Dcd:duts, we report the parameters for the M. tuberculosis enzyme for comparison with M. tuberculosis Dut constructs (Fig. 4A, Table 1). As expected 21 , the bifunctional enzyme is a relatively low-efficiency dUTPase compared to the monofunctional Dut (Table 1). In order to inactivate Dcd:dut without perturbing the overall structure of the enzyme, we introduced a bulky Phe into amino acid position 115 in place of an Ala (A115F) to prevent substrate binding to the active site. The Phe side chain in position 115 can only be accommodated within the active site cavity where it occupies the binding site of the uracil base (Fig. 4B). This mutant has the advantage that the cells keep synthetizing a structurally intact Dcd:dut protein. Fig. 4C shows that the Dcd:dut A115F does not exhibit enzyme activity. We have previously created this mutation in the structurally homologous human Dut and obtained an inactive but structurally intact mutant, as well 34 .
The hydrolysis activity of either Dut or Dcd:dut supports the growth of M. smegmatis. We aimed to investigate the effect of dUTPase activity loss in the living cell. We therefore created the above mutations (S148A, T138Stop and D83N) within the genome of M. smegmatis. The M. tuberculosis and M. smegmatis Duts share 85% amino acid sequence identity (100% within the conserved motifs) 11 and thus it is expected that the two enzymes behave similarly. In a previous paper, we established a method in which the disruption of the endogenous dut was rescued by a functional (complementing) copy of dut inserted into the genome on an integrating vector 11 . We used this scheme and introduced the mutations into the complementing copy of M. smegmatis dut. We obtained M. smegmatis strains that carried the mutant dut in a dut knock-out background (i.e. no wt copy present). Successful allele exchange was verified by Southern blot analysis (Fig. S2) and the mutations on the complementing dut copy were verified by sequencing of the appropriate genome region.
All three dut mutant strains were viable and unexpectedly, showed no growth defects when grown in liquid culture under stress-free conditions (Fig. 3B). This result suggests that the fully functional Dcd:dut in these dut mutant strains produces enough dUMP for the synthesis of more than limiting amounts of dTTP.
We also created a M. smegmatis strain carrying the non-functional A115F dcd:dut in a wt dut background. We constructed the A115F dcd:dut strain by allele exchange of the endogenous dcd:dut gene to a GFP-tagged copy carrying a point mutation. The A115F mutation was introduced either in wt or in inactive dut (D83N) background. Successful allele exchange was verified by Southern blot analysis (Fig. S2), and the mutation was verified by the sequencing of the appropriate genome region. The A115F dcd:dut strain encoding wt dut was viable and its growth rate was similar to that of the wt when grown in liquid culture (Fig. 4D). In contrast, we could not obtain any double mutant strains carrying both the dut D83N and the dcd:dut A115F mutations. In these cells, the dUTPase activity is completely abolished due to the fact that only inactive Duts are encoded. However, these inactive proteins are structurally intact and could still potentially mediate functions that are independent from their enzymatic activity or can operate in an inactive state (e.g. the essential surface loop in Dut 11 and examples from other species 15,[18][19][20]35 ). This implies that the dUTPase activity is essential for viability and reinforces that the dUTPase activity has exclusive role in dTTP biosynthesis in Mycobacteria (cf. Fig. 1).
Mutator phenotype of the dut and dcd:dut mutant strains. As we could not reveal any obvious defects in the enzymatically compromised dut or dcd:dut mutant M. smegmatis strains, we investigated possible long term effects of these mutations. We measured the mutation rates in each of the mutant strains 36,37 . We found that the mutation rates increased remarkably in the dut mutant strains (7-fold, 9-fold and 15-fold in the S148A, T138stop and D83N dut mutant strains, respectively). The mutation rate of the dcd:dut A115F strain appeared two-fold higher than the wt, however, the difference did not prove to be significant (Fig. 5A).
Dut mutant strains accumulate uracil in their genome. We also measured the genomic uracil content of the mutant strains by a q-PCR based method developed in our laboratory 38 . This assay is based on the fact that the Pfu polymerase does not amplify uracil-containing template DNA while other polymerases do (e.g. Taq). It is therefore possible to calculate the relative uracil content of the sample on the basis of PCR efficiencies driven parallel by Pfu and by Taq polymerases. We found that dut mutants have elevated genomic uracil content compared to the wt strain and that the increase in uracil content correlates with the in vitro measured activity loss (50,250 and 340 uracil/million base pairs in the S148A, T138stop and D83N dut mutant strains, respectively.) (Fig. 5B). For comparison, the 340 uracil/million base pairs genomic uracil content matches that measured in the ung-E. coli and the ung-MEF cells deficient in uracil misincorporation repair. The dut-/ung-E. coli strain contains even 20-times more uracils in its genome 38 . We could not detect any change in the genomic uracil content in the A115F dcd:dut mutant strain compared to the wt (Fig. 5B). To investigate if these marked effects are not simply due to an underlying difference in gene expression, we measured the mRNA levels of the various constructs. The expression of the T138stop and D83N dut constructs in the M. smegmatis cells seems to decrease compared to the wt, however, the difference proved not to be significant (Fig. 5C). On the other hand, the expression level of the A115F dcd:dut construct increased significantly compared to the wt (Fig. 5C). This finding emphasizes the differential effects in DNA uracilation and mutagenicity between the dut and dcd:dut mutations even more.
Pyrimidine nucleotide pool changes in the dut and dcd:dut mutant strains. To reveal the mechanism of uracil accumulation and mutation rate changes in our mutant strains, we measured the pyrimidine nucleotide pools in each strain. We used a DNA polymerase-based method to determine the concentration of dUTP, dTTP and dCTP in cell extracts 39,40 . The method is based on the incorporation of radiolabeled dATP into a nucleotide-specific template limited by the concentration of the quantifiable dNTP. The concentration of dUTP in the wt strain proved to be too low to be accurately quantified in our assay (<0.5 pmol/10 8 cells). We found that the concentration of dUTP became significantly elevated in the T138stop and D83N dut mutants (Fig. 6A). The dTTP:dUTP ratio changed from the >>10:1 ratio in the wt to 5:1 in the S148A dut mutant strain and to 4:1 in the T138stop and D83N dut mutant strains. The dTTP and dCTP concentrations remained quasi unchanged and consequently, the dTTP:dCTP ratio also remained in the wt range. However, the total pyrimidine content increased moderately but significantly in the D83N dut strain (Fig. 6, dCTP and dTTP concentrations increased 1.6-and 1.4-fold, respectively). Interestingly, however, the A115F dcd:dut strain had a normal low concentration of dUTP in its nucleotide pool while the dTTP:dCTP ratio became greatly imbalanced (Fig. 6 inset). The dCTP concentration of the A115F cell extract increased more than 6-fold while the dTTP concentration decreased to half of the wt (Fig. 6A).
These results clearly suggest two different roles for dcd:dut and dut in the regulation of pyrimidine nucleotide concentration and in the maintenance of genome integrity, respectively.

Discussion
Previous knock-out studies found dut essential and dcd:dut dispensable for Mycobacteria 10,11,22 . We reported that a complete knockout of dut is lethal 11 . We also reported that lethality is not due to the loss of the catalytic function but is mediated by a surface loop of unknown function independently of the dUTPase activity 11 . A well-folded and enzymatically active Dut enzyme lacking only this 5 amino acid long mycobacterium-specific surface loop does not support growth 11 . These previous studies did not provide information about the in vivo function of the enzymes bearing dUTPase activity, dut and dcd:dut. It was not clear, for example, how dUTPase activity affects the dNTP pool and downstream genomic processes. In the present study, we combined enzymology with genetics to address the role of dUTPase activity within the mycobacterial cell.
The mutant enzymes we created to tune down enzyme activity behaved in a well predictable manner in our in vitro experiments. The S148A and the T138stop dut mutations that compromised one or all interactions of the P-loop-like motif with the phosphate chain of the substrate resulted in small or severe activity loss, respectively, accompanied by weaker binding of the substrate (proportionally increased K d and K M , Table 1). The D83N mutation impaired the coordination of the catalytic water which left the binding of the substrate unaffected (Table 1). However, the subsequent catalytic reaction was compromised resulting in practical inactivity ( Table 1).
The activity-compromised dut point mutant M. smegmatis strains exhibit an increased mutation rate that accompanies the increase in the genomic uracil content (20-150 folds, Fig. 5). Mycobacteria encode three uracil DNA glycosylase enzymes (ung, udgB and udgX) 41-43 enabling effective uracil excision repair in these organisms 42 . This functional redundancy reinforces that the accumulation of uracil in DNA is an unwanted process. The potential reason why uracil could still accumulate in the DNA of our dut mutants is that DNA polymerases re-incorporate uracil despite the excision repair mechanism constantly excising it from the mycobacterial DNA. Polymerases usually do not distinguish between dTTP and dUTP and thus, they will likely incorporate dUTP from a nucleotide pool containing highly elevated dUTP concentrations (Fig. 6). Uracils (U·A pairs) are generally not considered as mutagenic compounds 44 while there is a controversy in the literature about the mutagenicity of abasic sites [44][45][46] . However, constantly excised and re-incorporated uracils probably cause stress and genome instability in the bacteria. In an ndk, dut double mutant E. coli strain, increased dUTP levels and replication intermediates from the uracil excision process caused several thousand-fold elevations in the mutation rate. The Ung mutation, which enables stable incorporation of uracil into DNA, could only partially alleviate the mutagenic effect 47 . This suggests that among other possible mechanisms, an increased frequency of uracil repair processed by either short patch or long patch base excision repair (BER) mechanism 48 may be responsible for the observed elevation in the mutation rates. In addition, the dut mutation had a modest effect on dCTP and dTTP levels 47 . In our experiments, the most severe D83N dut mutation also resulted in a modest increase of cellular pyrimidine levels. The overproduction of dNTPs in response to endogenous DNA damage is a general stress response 3, 49, 50 and most likely serves to promote tolerance against genotoxic stress. The elevated dNTP levels, in turn, increase the mutation rate 50-53 as the activity balance of DNA polymerases is transiently altered allowing the proofreading activity to decrease for the benefit of the nucleotide incorporating function (i.e. next nucleotide effect) 52 . These considerations are in agreement with our observations as well. The equally large increase in the mutation rates of the various dut mutants also suggests that this phenomenon is part of a general stress response and not specific to dUTPase activity loss. Interestingly, although the dTTP:dCTP ratio was highly imbalanced in the dcd:dut mutant strain, it displayed only two-fold increase in the mutation rate possibly indicating that dCTP is a relatively poor mutagenic precursor (Figs 5, 6). These results are in good accordance with previous literature data also reporting 2-fold increase in the mutation rate using the rifampicin resistance method in dcd mutant E. coli 46 . However, Schaaper and Matthews also found that the experimental system greatly affects the observation of mutagenicity. In their study, the 2-fold increase in the mutation rate observed using the rifampicin resistance assay appeared much higher (<42-fold) using a different assay (only available for E. coli for the moment) 46 . Moreover, Kumar and his colleagues found that even mild dNTP pool imbalances were mutagenic in Saccharomyces cerevisiae. However, the mutagenic potential of different imbalances did not directly correlate with their extent 53 . Nordman and Wright proposed that dNTP imbalances were not responsible for increased mutation rate in the ndk, dut mutant E. coli 47 . Our findings suggest that the dCTP:dTTP imbalance results in lower mutagenic potential than that of DNA uracilation.
Unexpectedly, the reduced dTTP concentration in the dcd:dut mutant strain did not limit the proliferation rate of the bacteria (Fig. 4D). This indicates that the activity of either of the two dUTPases is sufficient to support dTTP synthesis for the efficient growth of M. smegmatis. Our further results shown in Fig. 6 suggest, however, that dUMP production and dTTP biosynthesis are mainly under the control of the bifunctional Dcd:dut enzyme. The fact that Dcd:dut is a hundred-fold less efficient dUTPase than Dut (Table 1 and ref 21,54) suggests that the mechanistic differences between the two dUTPase enzymes are more important than simply is their catalytic efficiency. Dcd:dut is able to bind dTTP which inhibits its activity 55 . This negative feedback inhibition allows for the regulation of the cellular dCTP:dTTP concentration ratio. Dut, however, can only accommodate dUTP and does not show any allosteric features 56 . Based on our results shown in Figs 5, 6, we propose that Dut is responsible for the efficient elimination of dUTP, while the role of the bifunctional Dcd:dut is to maintain the proper dCTP:dTTP ratio. There are clear advantages to such functional diversion. dUTP is constantly generated from dUDP by the nucleoside diphosphate kinase (Ndk) and also by spontaneous dCTP deamination in mycobacteria (and by other enzymatic pathways in eukaryotes and in some prokaryotes ( Table 2). The accumulation of dUTP is efficiently prevented by the monofunctional Dut. The purpose here is to sanitize the dNTP pool from dUTP to prevent genome uracilation. The regulatory capabilities are built in the other, closely related, Dcd:dut enzyme that proved to be an inefficient dUTP sanitizer but keeps the pyrimidine DNA building blocks in a correct concentration ratio (Fig. 6). The preventive aspect of the bifunctional enzyme may be that dangerous dUTP is not released following the dCTP deamination reaction but is converted to dUMP on the same enzyme 21,57 .
We compiled the available literature data on the in vivo effects of dUTPase activity loss in various organisms in Table 2. This comprehensive data set supports a similar partition between the dUTP eliminating and the dNTP balancing functions in other organisms that bear additional dTTP producing pathways, as well. In Saccharomyces cerevisiae and Caenorhabditis elegans (bearing dCMP deaminase and a salvage pathway), the inactivation of dut is lethal. The inhibition of ung, however, can rescue the observed phenotype in these organisms 14,58 indicating the deleterious effect of genome uracilation in the dut mutant. The numerous reports on E. coli dut mutants also indicate that the major effect of Dut inactivation is genome instability and not a short supply of dTTP (Table 2). While the dut-1/ung-1 mutant E. coli strain can be maintained, the dut-1 phenotype is lethal 59 . However, in Trypanosoma brucei, in which there is no dCTP/dCMP deaminase and dUMP production strongly depends on dUDP/dUTP hydrolysis, the inactivation of ung could not rescue dut silencing but instead increased the cytotoxic effects conveyed by the low dUMP levels 13 . Efficient dut silencing in human cell lines resulted in a decline in clonogenic survival due to genome instability ( Table 2). All these literature data support the major role of dut in dNTP sanitizing. The regulation of dTTP concentration, however, seems to be exclusively committed to dCTP and/or dCMP transforming enzymes (Dcd, Dcd:dut, Dctd) which may be structurally unrelated from each other but are all allosterically regulated 55,57,[60][61][62] . A further in-depth investigation of the de-coupling of these functions may shine light on mechanisms supporting the appearance of T in and the exclusion of U from DNA. The recombinant Dut carrying an N-terminal hexa-His tag was cloned into pET19-b vector, and the recombinant M. smegmatis Dcd:dut amplified from p2NIL_dcdWT and p2NIL_dcdA115F (created in this study) was cloned into pET45-b vector with restriction sites BamHI and HindIII (Table S1). Both proteins were expressed in E. coli BL21(DE3)pLysS cells. For protein overexpression, the cells were grown to an OD 600 of 0.4, treated with 0.5 mM isopropyl-b-D-thiogalactopyranoside (IPTG) at 37 °C for 3 hours for Dut and at 30 °C for 6 hours for Dcd:dut expression.
Steady-state colorimetric dUTPase assay. Protons released in the dUTPase reaction were detected by phenol red pH indicator in 1 mM HEPES pH 7.5 buffer also containing 100 mM KCl, 40 μM phenol red (Merck) and 5 mM MgCl 2 . A Specord 200 (Analytic Jena, Germany) spectrophotometer and 10 mm path length thermostatted cuvettes were used at 20 o C for measuring dUTPase activity of the wt and mutant enzymes. The absorbance was recorded at 559 nm. V 0 was extracted from the raw absorbance vs. time curves followed by fitting the   (Table S1) and verified by Southern blot (Fig. S2) and by sequencing the appropriate genome region. Three parallel strains from each mutant were chosen and used for forward experiments. Primers used for cloning, mutagenesis and screening are compiled in Table S1.
Construction of the dcd:dut mutant strains. Dut KO strains carrying the wt 11 or the D83N mutant complementing dut copy were used to construct the A115F dcd:dut mutant and the D83N dut/A115F dcd:dut double mutant strains, respectively. A 3.5 kb fragment containing the dcd:dut gene and its flanking regions was cloned into p2NIL using HindIII restriction sites (Table S1) generating p2NIL_dcdWT. The A115F mutant dcd:dut containing vector was created by a modified QuikChange method 63 . A green fluorescent protein from pLL192 64 was C-terminally fused to the dcd:dut. The 6.1 kb PacI cassette carrying the lacZ and sacB selection markers from pGOAL17 was cloned into the sole PacI site of p2NIL to yield p2NIL_dcdA115F. p2NIL_dcdA115F was electroporated into electrocompetent cells. DCOs carrying the mutant dcd:dut were selected by colony PCR and verified by Southern blot (Fig. S2), then finally by sequencing the appropriate genome region. Three parallel strains were chosen and used for forward experiments. While the dcd:dut allele exchange worked in the wt dut background, it did not work in the D83N dut background even after several trials. Primers used for cloning, mutagenesis and screening are compiled in Table S1.
Growth assays. M. smegmatis mutant strains were grown in liquid media. OD 600 was measured every 3 hours. Three parallel strains were used from each investigated strain (9 parallel from each mutation) in these experiments. For quantitative comparison, growth curves were fitted with the y = a/(1 + exp(−k*(x − x c ))) equation that yielded the best fit keeping the function as simple as possible.
Determination of the spontaneous mutation rate. To determine the spontaneous mutation rates, three rifampicin sensitive independent colonies of each of the three strains/applied mutations were used to inoculate cultures that were grown at 37 °C, 150 rpm. Saturated cultures were serially diluted in sterile broth and plated onto agar plates to determine total CFUs or onto agar plates supplemented with 100 µg/ml rifampicin. The mutation rate for each of the 9 (3 × 3) independent cultures/applied mutation was determined as follows, where m 0 is the observed number of mutants at time point 0, m t is the observed number of mutants at the next time point, and N 0 and N t are the numbers of cells at time points 0 and t, respectively. The mean mutation rate was calculated for each mutant 36 .
Genomic DNA isolation. 10 ml liquid culture was grown until OD 600 = 0.5 and harvested. The cells were resuspended in 1 ml 10 mM Tris, pH 7.5 and 0.1 mm glass beads were added to 2 ml volume. The cells were disrupted by vortex and incubation on ice by turn. After centrifugation, the supernatant was manipulated routinely to purify DNA by phenol:chloroform:IAA (25:24:1) extraction followed by isopropanol precipitation.
Determination of the genomic uracil content. In order to quantify the uracil content of DNA, a real-time quantitative PCR-based assay was used 38 . Genomic DNA was isolated and digested with BamHI. DNA fragments of 5 kb were purified from gel. Real-time PCR was performed on a Mx3000P qPCR System (Agilent Technologies) using EvaGreen dye (Biotium) and PfuTurbo Hotstart DNA polymerase (Stratagene) and Mytaq Hotstart DNA polymerase (Bioline). A segment with 1017 base length defined by the primers (Table S41) was amplified during the PCR reaction. Two-fold dilution series were prepared from the DNA samples. Three parallel strains were used for each mutation in the experiments. dNTP extraction. Exponential phase cells were grown with appropriate antibiotics until OD 600 = 0.6. The total CFUs were determined for each culture, and cells were centrifuged for extraction. Washed pellets were extracted in 0.5 ml ice-cold 60% methanol overnight at −20 °C. Cells were removed by centrifugation (15-20 min, 13.000 rpm) the methanolic supernatant was boiled for 5 min and centrifuged. The supernatant containing the soluble dNTP fraction was vacuum-dried (Eppendorf) at 45 °C, 1h. Extracted dNTPs were dissolved in 50 μl dUTPase buffer (30 mM Tris-HCL, pH 7.5, 10 mM MgCl 2 , 50 mM NaCl, 1 mM EDTA) and stored at −80 °C.
The quantitation of dcd:dut and dut expression levels. Cells were grown in 50 ml liquid culture until saturation, washed in ice cold PBS and harvested by centrifugation (3100 g, 20 min). Bacterial pellets were resuspended in 1 ml Trizol (Life Technologies), and the cell wall was disrupted by repetitive vortexing with glass beads (6 × 1min). Nucleic acid recovered in the aqueous phase after addition of 0.2 ml chloroform was precipitated with the addition of 0.5 ml isopropanol. The RNA preparations were DNase-treated (10 min, 37 o C) and purified with the Nucleospin RNA Clean-up kit according to the instructions of the manufacturer. Mycobacterial RNA yield were assayed using the Nano-Drop ND-2000 Spectrophotometer (NanoDrop Technologies). RNA samples were amplified from 1 µg total RNA by random hexamer primers using the Transcriptor First Strand cDNA Synthesis Kit (Roche). The resulting cDNA was quantified by Quantitative PCR using EvaGreen (Bioline) and MyTaq PCR master mix (Bioline) in a Stratagene Mx3000P instrument. sigA (MSMEG_2758), an endogenous reference gene was used to normalize input cDNA concentration 65 . The relative expression ratios of the examined genes were calculated using the comparative Ct method (ΔΔCt). Primers used to measure cDNA of sigA, dcd and dut are compiled in Table S1.
Determination of the pyrimidine nucleotide pool size. The determination of the pyrimidine nucleotide pool size in each extract was based on DNA polymerase-catalyzed incorporation of radioactive dNTP into the synthetic oligonucleotide template method described in ref. 66. The reaction mixture (50 μl) contained Klenow buffer, 0.5 unit exonuclease negative Klenow-fragment (Fermentas), 0.25 μM dTTP/dCTP specific template, 0.25 μM primer (Table S1), 2.5 μM [3H] dATP (1,5 Ci/mmol) (American Radiolabeled Chemicals, Inc.) and 8 μl dNTP-extract or premixed dNTP for calibration. Calibration curve was prepared using 0, 0.1, 0.5, 1, 2, 4 and 8 pmol of each dNTPs/reaction mixture. Incubation was carried out for 60 min at 37 °C and the reaction mix was spotted onto DE81 paper. The papers were dried, washed (3 × 10 min) with 5% Na 2 HPO 4 , and rinsed once with distilled water and once with 95% ethanol. After drying, radioactivity on the papers was measured in a liquid scintillation counter (Beckman). In case of the dCTP measurement, we used Taq polymerase (RedTaq, Sigma), the incubation was carried out at 48 °C for 1 hour, as Klenow polymerase is capable of incorporating CTP and GTP from nucleotide extracts 39 . dUTP concentration was measured according to Koehler et al. 67 . Half of the samples for dTTP measurement were treated with 40 ng recombinant Dut at 37 °C, 45 min. The Dut enzyme was precipitated with 60% methanol. The dUTP in half of the extract was enzymatically hydrolyzed while in the parallel sample it was not. Then dUTP concentration was measured using the same protocol as for dTTP determination.
Two-fold dilution series were prepared from cell extracts in the polymerase reactions. Four samples for each of the three parallel strains were used for each mutation in the experiments (i.e. 12 data points).

Statistical analysis.
Statistical analysis was carried out using the STATISTICA.13 software. The non-parametric Kruskal-Wallis test or the one-way ANOVA test with Student-Newman-Keuls multiple comparison post-hoc test was used when samples passed the equal variance (Bartlett's) criterion.