The association of DNA damage response and nucleotide level modulation with the antibacterial mechanism of the anti-folate drug Trimethoprim
© Sangurdekar et al; licensee BioMed Central Ltd. 2011
Received: 12 August 2011
Accepted: 28 November 2011
Published: 28 November 2011
Trimethoprim is a widely prescribed antibiotic for a variety of bacterial infections. It belongs to a class of anti-metabolites - antifolates - which includes drugs used against malarial parasites and in cancer therapy. However, spread of bacterial resistance to the drug has severely hampered its clinical use and has necessitated further investigations into its mechanism of action and treatment regimen. Trimethoprim selectively starves bacterial cells for tetrahydrofolate, a vital cofactor necessary for the synthesis of several metabolites. The outcome (bacteriostatic or bactericidal) of such starvation, however, depends on the availability of folate-dependent metabolites in the growth medium. To characterize this dependency, we investigated in detail the regulatory and structural components of Escherichia coli cellular response to trimethoprim in controlled growth and supplementation conditions.
We surveyed transcriptional responses to trimethoprim treatment during bacteriostatic and bactericidal conditions and analyzed associated gene sets/pathways. Concurrent starvation of all folate dependent metabolites caused growth arrest, and this was accompanied by induction of general stress and stringent responses. Three gene sets were significantly associated with the bactericidal effect of TMP in different media including LB: genes of the SOS regulon, genes of the pyrimidine nucleotide biosynthetic pathway and members of the multiple antibiotic resistance (mar) regulon controlled by the MarR repressor. However, the SOS response was identified as the only universal transcriptional signature associated with the loss of viability by direct thymine starvation or by folate stress. We also used genome-wide gene knock-out screen to uncover means of sensitization of bacteria to the drug. We observed that among a number of candidate genes and pathways, the effect of knock-outs in the deoxyribose nucleotide salvage pathway, encoded by the deoCABD operon and under the control of the DeoR repressor, was most informative.
Transcriptional induction of DNA damage response is an essential feature of the bactericidal effect of trimethoprim. Either the observation of the transcriptional response or DNA damage itself, or both, is made possible by thymine starvation when other folate-dependent metabolites are not limited. The effect of DNA damage by the drug takes place prior to its bactericidal effect, at the beginning of the lag stage of the treatment. Mutations in the deoxyribose nucleotide salvage pathway can affect duration of the lag as well as the rate of killing. This information can be used to postulate certain mechanistic differences between direct thymine starvation in thymidylate synthase deficient mutants and thymine starvation by anti-folate inhibitors.
2,4,-diamino-5-(3', 4', 5'-trimethoxybenzyl)-pyrimidine (Trimethoprim, TMP) is a folate analog that inhibits the reduction of dihydrofolate to tetrahydrofolate (THF) by competitively binding the active site of the enzyme dihydrofolate reductase (DHFR) [1, 2]. The drug is highly specific against bacterial and malarial enzymes , has a wide antibacterial spectrum and, despite emerging resistance [4–7], has been extensively used against urinary and respiratory tracts infections and against enteric pathogens and methicillin resistant S. aureus.
To answer these and related questions, we treated Escherichia coli with TMP in a variety of media and supplementation conditions and tracked global transcriptional responses using microarrays. We also applied a genome-wide genetic screen to identify gene knock-out mutants with decreased or increased susceptibility to TMP in rich media. We observed that the drug treatment, either in minimal medium supplemented with requisite amino acids and adenine or in complex medium (LB), resulted in the loss of viability and this loss was correlated with the induction of DNA damage repair genes. Under limited supplementation conditions (only amino acids or purines) in minimal media, the primary signature was that of stress response, but when both were added in absence of thymine, the stress response signature was not observed. Addition of individual supplements (amino acids or purine) also led to the repression of the respective biosynthetic genes, even in the presence of TMP. We also identified a signature of gene set activity, including induction of DNA damage response, pyrimidine nucleotide biosynthesis and antibiotic stress responsive MarR targets to be most associated with cell lethality in minimal media. However, only the SOS response was associated with the bactericidal outcome across all tested nutrient conditions, including thymine starvation in a thymine auxotroph.
Using a genome-wide collection of knock-out mutants to screen for genetic, rather than environmental, factors that can affect TMP susceptibility, we found that the deoxyribose nucleotide salvage pathway has the capacity to modulate and mediate the effect of the drug. The deoA and deoC gene knock-out mutants were found to be resistant to TMP in complex medium, whereas the loss of their transcriptional repressor deoR rendered bacteria more sensitive to the drug. However, this relationship was reversed in minimal medium with supplements. This reversal was also observed at the transcriptional level wherein the deoCA genes were down-regulated during TMP treatment in LB medium and up-regulated in M9 minimal medium. Using this data and other available information we propose that the nucleotide salvage pathway encoded by deoCABD genes is a critical modulator of physiological outcomes of the TMP treatment in E. coli.
Escherichia coli MG1655 F - lambda - ilvG - rfb-50 rph - was used as a wild-type strain in all experiments. Trimethoprim (Fluka (Sigma-Aldrich), St. Louis, MO) stock solution was prepared by dissolving the drug in distilled water with 1% v/v glacial acetic acid (for Keio screen) or by dissolving in 50% v/v ethanol-chloroform mixture (for transcriptional profiling) and stored in small aliquotes at -20°C. Complex medium for cell growth was LB (Luria Bertani broth) and minimal medium was M9 minimal medium with 0.4% glucose . In LBTMP experiments, cells were treated with 50 μg/ml of TMP, while for M9 experiments the concentration used was 25 μg/ml, corresponding to the MIC50 determined in the respective media after 18 hrs of treatment. Amino acids for supplementation (methionine and glycine) were purchased from Fluka (Sigma-Aldrich, St. Louis, MO) and stock solutions were prepared in water. Standard cultures were grown in flasks at 37°C under aeration in a rotary shaker either in M9 minimal medium containing 0.4% glucose or in LB liquid medium. For M9 with folate dependent supplementation experiments, the following chemicals were added with a final concentration of 50 μg/ml: adenine, amino acid (methionine and glycine) mixture, and thymine.
Overnight cultures were resuspended in fresh medium and TMP was added when the optical density (O.D. 600 nm) of the culture reached 0.3-0.4. Time-point samples were taken every 15-30 minutes from 15 minutes to 2 hours post treatment. The samples were spun down and snap-frozen in liquid nitrogen before storing at -80°C until required. Total RNA samples were purified using the Qiagen RNeasy kit (Chatsworth, CA) according to the manufacturer's protocol. To identify genome-wide gene expression profiles, relative mRNA levels were determined by parallel two color hybridization at single-gene resolution to whole-genome E. coli K-12 MG1655 spotted DNA microarrays, designed, printed and probed as described , and containing discrete sequence elements corresponding to 98 and 3/4% of all annotated open reading frames (ORFs). Fluorescence DNA probes were synthesized from 10~15 μg of total RNA with random hexamers and Cy-5 dUTP or Cy-3 dUTP dyes (Amersham) and hybridized to microarrays for 6 hours at 65°C followed by a washing step as described previously . All microarray experiments were designed as time courses, using the 0 min time point sample as a common reference. The hybridized arrays were then scanned and the raw fluorescence intensity data were normalized in R (http://cran.r-project.org/) using the Limma (Bioconductor) package  to remove spatial and array based biases. The data was converted to log-ratios of normalized intensity at each time point to time point 0. The microarray data was uploaded to NCBI GEO database and is available for download through accession number GSE32562.
Gene Set Analysis
To analyze gene expression data in the context of known gene sets and pathways, we obtained gene functional classification information from databases (GenProtEC , RegulonDB , EcoCyc ) to assemble a total of 480 gene sets (Additional File 1). These comprise of sets of genes whose products are involved in the same pathway and sets of genes that share common regulators or other properties. Normalized gene expression data was analyzed for gene set activity using a method modified from Tian et al . We then modeled the normalized set scores across the entire time series to analyze the effect of individual supplements on gene set activity and to infer set activity associated with the lethal phenotype. The raw gene set score for k th gene set was calculated as:
where ESi is the vector of set scores of set i; AAi, Adi and Thyi are factors with level 1 (if the respective supplements are added) or 0. M9TMP condition has all three factors at level 0 since none of the supplements were added. A positive coefficient for any factor indicates that the addition of the supplement is associated with increase in set score relative to its baseline level.
where ESi is the vector of set scores for set i, SV i are surrogate variables and Phe i represents a vector of phenotypes associated with the set (factor with two levels: "cidal" or "static"). Parametric p-values associated with the estimated coefficients were obtained by ANOVA, and False Discovery Rates (q-value ) was calculated to infer significant gene sets. A q-value of 5% indicates all sets with the same or lower q-value are called significant at 5% False Discovery Rate. All calculations were done in R. Non-parametric classification based on Receiver-Operator Characteristics (ROC) curves were done by using the R package "caTools". AUC values of 0.85 and higher were considered significant, and it was used as the second threshold to filter significant gene sets. The entire data and the R code to reproduce the analysis can be found in Additional File 2.
Genome-wide genetic screen
The Keio collection was a kind gift of Prof. Hirotada Mori and consists of knock-out mutants of 3985 genes of E. coli strain BW25113 . The knock-out mutants contain KanR cassettes inserted in genes using the method of Datsenko and Wanner . For mutant screening, a copy of the mutant plates was made by replica inoculating the frozen stocks into 96-well plates containing 150 μ l of LB medium and incubating them at 30°C for 12-18 hours with intermittent shaking. For growth and inhibition experiments, the optical density in the wells of the 96-well plate was measured using Victor3 plate reader (Perkin-Elmer). The actual O.D. reading in the plate reader depends on the medium and the volume of liquid in the wells and is correlated with actual culture density measured by conventional light scattering techniques (r2 > 0.99) for O.D. 0.1-1.0. Since each plate was inoculated as a set, there was plate-to-plate variation in O.D. that was accounted for by normalizing across plates. Resistant and sensitive mutants were identified by normalized O.D. readings. Mutants in Keio collection are in BW25113 background. For individual mutant analysis, the Kan R insertion alleles were transferred into wild-type MG1655 background using P1-transduction . An E. coli thyA715 mutant MG1655 was obtained from CGSC. Different deo mutations were similarly introduced into the thyA- parental strain. thyA-deo- double mutants had different thymine/thymidine growth requirements; thyA- 20 μ g/ml thymine, thyA-deoA- 20 μ g/ml thymidine, thyA-deoC- 2 μ g/ml thymine, thyA-deoR- 50 μ g/ml thymine [35, 36] For viability experiments and RNA sampling, the cultures were handled as follows. Cells were grown to stationary phase in LB or M9 minimal media (with glucose). They were then inoculated in fresh medium and grown till O.D. 0.4-0.6 before treatment. For TMP treatment in LB, TMP was added to appropriate final concentrations. For TMP treatment in minimal medium, growing cultures were diluted in pre-warmed fresh minimal medium containing TMP and methionine, glycine, adenine supplements. For viability counts, samples were diluted in 0.9% NaCl, spread or spotted on LB plates and incubated 12-16 hours before colony counts were taken.
Results and Discussion
Folate supplementation regimes determine bacterial viability phenotypes
Treatment regimens used for gene expression analysis
+ (90% loss in 3 hours)
Met, Gly, Ad
+ (>70% loss in 6 hours)
Met, Gly, Ad, Thy
Effects of individual supplementation regimes
Given the diverse nature of phenotypes generated by the combination of TMP and media conditions, and the complex nature of folate metabolism being perturbed, it is essential to identify key molecular events during each treatment that are associated with the observed phenotype. We first followed transcriptional responses to TMP treatment in these conditions as a time series using cDNA microarrays and obtained gene expression profiles as ratios of transcript abundances after the treatment (10 - 120 minutes) normalized to the before-treatment levels (0 minutes). To develop a biological interpretation of these responses, we examined transcriptional activity of sets of genes . Briefly, for each time point in each condition, the activity score (S k ) for each gene set was calculated as the weighted average of all the genes within the gene set (Methods). The score was then normalized using background null distributions  to get the enrichment scores (ES) and the p-values for these scores were then converted to a posterior probability (PP, ) to adjust for multiple testing. The PP value (1 for strong enrichment, 0 for no enrichment), along with the enrichment score ES, allowed for quantitative comparison of different gene sets in different conditions. PP values of greater than 0.90 were considered to be highly significant for gene set activity, corresponding to false discovery rates of less than 5% for the conditions.
The effect of TMP in different supplementation conditions can be viewed and analyzed as an interaction between the effects of the drug effect and the condition. Traditionally, the effect of a drug would be ascertained in cultures adapted over multiple generations to their growth environments. While such approach may seem physiologically most sound, it does not allow determining the media specific effects in any straightforward way, in part because of a prolonged adaptation and resulting steady state are accompanied by macroscopic growth and metabolic differences which can mask nutrient-specific regulation. This has not been seen as a serious problem since for most actively studied antibacterials, the media composition either does not have an effect on the outcome of the treatment or its effects are not known. This however would be a serious problem for TMP treatments where media composition effects have been well documented. Therefore we designed our experiment in a way that yielded physiologically meaningful, and consistent with historical data, media-dependent outcomes of the treatments and also allowed detecting direct regulatory effects of the supplementations. According to such design, the drug was added to the cell cultures along with the supplements and transcriptional responses were recorded immediately afterwards. Thus the ensuing regulatory and physiological changes in the cells will be, by definition, a result of the combinations, or interactions, of the effects of the drug and supplements. Modeling these transcriptional responses as estimated additive effects of each supplement could help us explain these interactions as simple cause and effect representation for each individual factor-gene set pair.
Identification of a transcriptional signature associated with lethality
To further compare transcriptional sensitivities of these gene sets, we scored their responses in another thymine starvation condition, triggered by thymine withdrawal from a thyA strain ("Thymine starvation" in Figure 3). We observed that the SOS response followed the up-regulation trend following the treatment, while pyrimidine biosynthesis was down-regulated and MarR targets were not affected. Thus, we propose that the induction of the DNA damage response is the only transcriptional effect associated with thymineless death phenomenon observed across a variety of experimental conditions including thymine starvation of auxotrophic mutants, starvation of dTMP using TMP in the presence of supplements or using TMP in complex growth media; whereas up-regulation of pyrimidine synthetic genes and MarR targets is associated with lethality only in TMP treatments. Of note, when we examined a compendium of expression profiles for activity of the MarR regulon, we observed that other top conditions in which MarR targets were up-regulated were treatments with antibacterials and antibiotics including Norfloxacin, Streptomycin, Kanamycin, Indole-acrylate and Chloramphenicol (data not shown). Since all of these treatments were carried out in LB, and the current identification of the MarR signature was driven by the LBTMP condition, it is possible that de-repression of MarR targets is a result of interaction(s) between yet unidentified factors associated with growth in a rich medium and the generic antibiotic induced stress.
We previously reported that the DNA damage response was the primary transcriptional response that separated thymine starvation leading to thymineless death (TLD) in thyA auxotrophs from the thymine limitation (TLM) wherein cells were supplied with suboptimal amounts of thymine . Here, we showed that the SOS response was also significantly associated with cell death by TMP in minimal medium supplemented with folate metabolites or in complex medium. The genes involved in SOS response were observed to be significantly induced in bactericidal conditions as compared to static conditions after 30 minutes of treatment (One-sided t-test p-value = 0.004) and at later time points. No loss of viability was observed in cidal conditions in either media at this time point, indicating that DNA damage precedes viability loss and thus may be causative to cell lethality.
Genome-wide genetic screen for TMP susceptibility
List of mutants yielding high biomass in presence of TMP at sub-lethal concentrations in LB medium
Resistant knock-out candidates
Molybdopterin (MPT) synthase, large subunit; chlorate resistance; dimer of dimers with MoaD
Phosphocarrier protein for glucose of the PTS; Enzyme IIA(Glc); formerly EIII(glc)
Single-stranded DNA-specific exonuclease, 5'-3'
Function unknown; induced by alkali; putative membrane transport or efflux protein
Na+/serine (threonine) symporter
N-Acetylneuraminate lyase (aldolase)
Fosfomycin sensitivity; sugar P transport system; transport protein for hexose P's
Excision nuclease subunit A; repair of UV damage to DNA; LexA regulon; binds Zn(II)
Site-specific recombinase for fimA promoter segment inversion; bias for ON to OFF phase switching
Deoxyribose-phosphate aldolase; deoxyriboaldolase; binds selenium
List of mutants yielding low biomass in presence of TMP at sub-lethal concentrations in LB medium
Sensitive knock-out candidates
Repressor for deo operon, nupG and tsx; binds deoxyribose-5-phosphate inducer
Flagellar synthesis, basal body L-ring protein
Hook-associated protein 2, axial family; flagellar regulon
Cytoplasmic membrane ATPase involved in flagellar assembly; involved in export of flagellar axial protein subunits
Flagellar switch protein
Required for regulatory inactivation of DnaA; multicopy suppressor of dnaN(ts)
The gene tdk encodes thymidine kinase, which converts thymidine to thymidine monophosphate dTMP (Figure 1). In the absence of tdk, cells exclusively rely on tetrahydrofolate dependent conversion of dUMP to dTMP, which is inhibited during TMP treatment. Cells without tdk should be therefore unable to grow, and indeed a tdk knock-out was identified as a sensitive mutant in the assay. This was verified by the observation of rapid loss of viability of tdk mutant after 50 μg/ml TMP addition in liquid culture (data not shown). Other mutants imparting sensitivity to TMP included several knock-outs of the genes of chemotaxis pathway (fliD, fliL, fliN, flgH) and of hda, a regulatory inhibitor of the replication initiator protein DnaA. Inactivation of Hda is a major cause of DnaA-dependent over-initiation of the chromosomal replication , and recent studies have proposed a link between aberrant DNA replication initiation and TLD [37, 39, 40]. High-yielding mutants could in principle be genuine TMP resistant mutants or strains filamenting in the presence of the drug. Among these mutants were knock-outs of uvrA, a nuclease component of the nucleotide excision repair, and of recJ, a single strand DNA exonuclease. A loss of recJ function was reported to alleviate the lethal effect of the thymine starvation [37, 39, 41], which is believed to be the primary bactericidal mechanism of TMP. Knock-outs of two of the genes involved in deoxypyrimidine nucleotide salvage pathway, deoC and deoA, were found to be among resistant candidates, whereas a knock-out of the repressor of the deoCABD operon, deoR, was found to be more susceptible to TMP. deoC, deoA and deoR mutants grew in LB without the drug as well as their isogenic parental strain. Of note, the deoCABD operon was down-regulated during LBTMP treatment and was among the top down-regulated gene sets/pathways in the condition. This enrichment of a single metabolic pathway and its regulator in the TMP phenotypic screen along with its transcriptional signature, and given the central role of the pathway in the conversion of thymidine and uridine nucleosides into nucleotides, led us to investigate these genes in more detail.
In minimal medium supplemented with amino acids and purine, cells cannot utilize exogenous thymine or thymidine like in LB. dTTP depletion during thymine starvation induces the ribonucleoside-diphosphate reductases, which increases flux through the dTTP biosynthetic pathway and causes an accumulation of dCTP and dUMP [45, 46]. In M9TMPAdAA condition, we see an up-regulation of the nrdAB, genes (log-ratio > 0.5) responsible for the reduction of pyrimidine ribonucleotides to dUTP, indicating that similar conditions may exist during this treatment. The excess accumulated dUTP can then get incorporated into DNA or can be degraded to dUMP and further to deoxyuridine and dR1P via the dUMP phosphatase YjjG [42, 43, 47]. During thymine starvation, excess ribose sugar is secreted into the medium, particularly in deoC- mutants . In a deoA- mutant, the cells lack the ability to degrade deoxyuridine and this may lead to the increased accumulation of dUMP and dUTP in the cells, possibly due to the conversion of the accumulated deoxyuridine to dUMP via thymidine kinase .
Similarly, in deoC- mutant, dUMP cannot be catabolized leading to dUTP accumulation. The slight difference in sensitivity in deoA and deoC to TMP (Figure 4B) could be attributed to the activity of uridine phosphorylase (Udp) which also degrades deoxyuridine at much lower rate than DeoA . Finally, the deoR- mutant should be similar to the wild-type since in the wild-type, the deo operon is derepressed due to accumulation of dR5P, and the loss of viability in deoR is similar to that in the wild-type strain.
In the current study, we observed that the kinetics of killing by TMP was similar to that by thymine starvation: 1-2 hour lag is followed by a relatively rapid loss of viability. Moreover, induction of the SOS response in both conditions occurred in the first half of the lag stage, indicating that DNA becomes damaged long before the cells loose their ability to divide. The duration of the lag stage can in principle be determined by two factors: 1- the time it takes for dTTP to drop below a certain level to cause replication fork arrest; 2 - the time it takes for DNA damage to become irreversible. While the first factor can be postulated a priori; the second is derived from the observation that the loss of RecA or RecBC function shortens or eliminates the lag , suggesting that double-strand break repair pathway has a capacity to reverse some effects of the starvation. It is likely that the shortening of the lag in a deoR- mutant in the rich medium was due to faster depletion of the dTTP pool, and the phenotype of deoC- and deoA- mutants in the rich medium can be viewed as an extension of the lag stage due to the pool maintenance in the mutant cells effectively scavenging thymi(di)ne. When both dTTP and dUTP pools become fully exhausted, DNA damage becomes persistent and its processing by some repair enzymes makes the damage irreversible. Cells from this stage in the thymineless death pathway cannot be recovered and that accounts for a rapid decline in viability. In support of this view, mutants that can't process initial DNA damage, e.g., recF - , loose viability at a much-reduced rate following the lag [39, 41].
It appears that the primary factor in establishing the rate of starvation-induced killing is a number of chromosomal replication events in a bacterial population. The higher the rate of initiation and the greater the number of growing replication forks per chromosome, the higher the probability of the irreversible chromosomal collapse . In principle, if the dTTP pool is efficiently replenished with dUTP, that would prevent replication fork arrest (at least in the current replication cycle) and thus would reduce the probability of the subsequent chromosomal collapse. Such replenishment does not happen under conventional thymine starvation in thyA auxotrophs because of the activity of dUTPase, which efficiently hydrolyzes dUTP to dUMP and PPi  and because of the yjjG-deo pathway, which catabolizes excess dUMP and dU [42, 43, 47]. That is probably why the thymineless death triggered by thymine starvation in thymidylate auxotrophs is not affected by the activity of uracil-N-glycosyase, the first enzyme in the uracil base excision pathway [41, 52]: although uracil may be occasionally found in DNA from some mutant TLD cells, its post-replicative excision would not matter much since the chromosome would be already on its way toward collapse from the fork arrest, i.e., for uracil to be incorporated in a part of DNA, the replication fork must traverse that part of DNA in the template, incorporation of U signifies exhaustion of dTTP and since dUTP are readily hydrolyzed, the fork that started incorporating U's is subject to an eminent arrest. As a result, the kinetics of the viability loss would be determined by the frequency of fork arrests and not by the frequency of other DNA lesions.
However, in the conditions that maybe prone to a slower exhaustion of the dTTP pool, due to indirect or limited inhibitory effect on the thymidylate synthase and/or significant increase of dUTP over dTTP, incorporation of uracil may become a factor in determining the outcome of dTTP starvation. For example, chromosomes that are fully replicated in the presence of sufficient dUTP may eventually collapse as a result of uracil excision repair, which could be initiated but not completed in the absence of dTTP or dUTP. That appears to be the case in yeast treated with the antifolate inhibitor aminopterin  and in animal cells treated with the antifolate drug methotrexate . We argue that at least in deoC- and deoA- cultures treated with TMP, a fraction of cells can complete the ongoing rounds of DNA replication, using dUTP in combination with slowly decaying dTTP supply, and these chromosomes would collapse primarily because of uracil processing and not because of replication fork arrest. Consistent with that is the observation that the lag stage was not affected in the deoC- and deoA- mutants; instead, the loss of viability was relatively accelerated in these mutants - pointing at an additional source(s) of irreversible DNA damage. It is possible that TMP treatment of E. coli, independent of the growth medium and similar to anti-folate treatments in higher organisms, results in dUTP levels sufficient to support DNA replication. Processing uracil out of such DNA with UNG would be detrimental to the cells. This assertion is supported in part by our observation that an ung- mutant yielded in our screen 3 times as much biomass as an average strain (data not shown) and by observations from a chemical genomics screen where dut- and ung- mutants showed fitness defect, while dcd- showed fitness advantage in response to the TMP treatment . Thus, it is likely that depending on the mechanism of thymine starvation (TMP versus direct thymidylate deficiency) and despite apparent similarities between them, different components of DNA metabolism may contribute to the starvation outcomes.
Treatment of E. coli bacteria with the antibiotic Trimethoprim results in regulatory responses that can be associated with the drug mechanism as well as treatment conditions. Depending on the treatment condition, the drug can have bacteriostatic or bactericidal effect. The DNA damage response can be used to discriminate between the bacterial killing, which is preceded by the damage response, and the bacteriostasis, during which the damage response does not develop. The drug induces DNA damage at the beginning of the characteristic lag stage of the treatment. Mutations in the deoxyribose nucleotide salvage pathway can affect duration of the lag as well as the rate of killing. The duration of the lag is likely determined by the capacity of the cells to buffer the effect of thymine starvation caused by the drug, e.g., through dTTP pools. The rate of killing, however, is determined by the distribution of drug sensitivities across a bacterial population. The sensitivity of individual cell in turn should be determined by its stage in the replication and division cycles. We speculate that Trimethoprim, by affecting replication and possibly division cycles, makes uracil incorporation into DNA an important source of the offending lesion in the drug-induced thymine starvation.
We thank Drs. Hirotada Mori for the gift of a knock-out library, Friedrich Srienc for financial support, Patrick Higgins for the gift of P1 phage. This work was supported in part by grant GM066098.
- Bushby S, Hitchings G: Trimethoprim, a sulphonamide potentiator. British journal of pharmacology and chemotherapy. 1968, 33 (1): 72-PubMed CentralPubMedView ArticleGoogle Scholar
- Burchall JJ, Hitchings GH: Inhibitor binding analysis of dihydrofolate reductases from various species. Molecular pharmacology. 1965, 1 (2): 126-PubMedGoogle Scholar
- Burchall JJ: Mechanism of action of trimethoprim-sulfamethoxazole. II. J Infect Dis. 1973, 128 (Suppl): 437-441. 10.1093/infdis/128.Supplement_3.S437.PubMedView ArticleGoogle Scholar
- Amyes S, Towner K: Trimethoprim resistance; epidemiology and molecular aspects. Journal of medical microbiology. 1990, 31 (1): 1-19. 10.1099/00222615-31-1-1.View ArticleGoogle Scholar
- Huovinen P: Trimethoprim resistance. Antimicrob Agents Chemother. 1987, 31 (10): 1451-1456.PubMed CentralPubMedView ArticleGoogle Scholar
- Huovinen P: Increases in rates of resistance to trimethoprim. Clin Infect Dis. 1997, 24 (Suppl 1): S63-66.PubMedView ArticleGoogle Scholar
- Murray BE, Alvarado T, Kim KH, Vorachit M, Jayanetra P, Levine MM, Prenzel I, Fling M, Elwell L, McCracken GH: Increasing resistance to trimethoprim-sulfamethoxazole among isolates of Escherichia coli in developing countries. The Journal of Infectious Diseases. 1985, 152 (6): 1107-1113. 10.1093/infdis/152.6.1107.PubMedView ArticleGoogle Scholar
- Huovinen P, Sundström L, Swedberg G, Sköld O: Trimethoprim and sulfonamide resistance. Antimicrobial Agents and Chemotherapy. 1995, 39 (2): 279-PubMed CentralPubMedView ArticleGoogle Scholar
- Blakley RL: The biochemistry of folic acid and related pteridines. 1969, North-Holland Publishing Co.Google Scholar
- Harvey R, Dev I: Regulation in the folate pathway of Escherichia coli. Advances in enzyme regulation. 1975, 13: 97-124.View ArticleGoogle Scholar
- Smith RJ, Midgley JE: The effect of trimethoprim on macromolecular synthesis in Escherichia coli. Biochem J. 1973, 136 (2): 225-234.PubMed CentralPubMedView ArticleGoogle Scholar
- Kwon YK, Higgins MB, Rabinowitz JD: Antifolate-induced depletion of intracellular glycine and purines inhibits thymineless death in E. coli. ACS Chem Biol. 2010, 5 (8): 787-795. 10.1021/cb100096f.PubMed CentralPubMedView ArticleGoogle Scholar
- Dale BA, Greenberg GR: Effect of the folic acid analogue, trimethoprim, on growth, macromolecular synthesis, and incorporation of exogenous thymine in Escherichia coli. J Bacteriol. 1972, 110 (3): 905-916.PubMed CentralPubMedGoogle Scholar
- Miovic M, Pizer LI: Effect of trimethoprim on macromolecular synthesis in Escherichia coli. Journal of Bacteriology. 1971, 106 (3): 856-PubMed CentralPubMedGoogle Scholar
- Amyes SG, Smith JT: Trimethoprim action and its analogy with thymine starvation. Antimicrobial Agents and Chemotherapy. 1974, 5 (2): 169-178.PubMed CentralPubMedView ArticleGoogle Scholar
- Then R, Angehrn P: Nature of the bactericidal action of sulfonamides and trimethoprim, alone and in combination. The Journal of Infectious Diseases. 1973, 128: 498-501. 10.1093/infdis/128.Supplement_3.S498.PubMedView ArticleGoogle Scholar
- Then R, Angehrn P: Sulphonamide-induced 'thymineless death' in Escherichia coli. Microbiology. 1973, 76 (2): 255-10.1099/00221287-76-2-255.Google Scholar
- Sambrook J, Russell DW: Molecular cloning: a laboratory manual. 2001, Cold Spring Harbor, N.Y.: Cold Spring Harbor Laboratory Press, 3Google Scholar
- Khodursky AB, Bernstein JA, Peter BJ, Rhodius V, Wendisch VF, Zimmer DP: Escherichia coli spotted double-strand DNA microarrays. Functional genomics Humana, Totowa. 2003, 61-78.View ArticleGoogle Scholar
- Smyth G: Limma: linear models for microarray data. Bioinformatics and computational biology solutions using R and Bioconductor. 2005, 397-420.View ArticleGoogle Scholar
- Serres MH, Riley M: MultiFun, a multifunctional classification scheme for Escherichia coli K-12 gene products. Microbial & comparative genomics. 2000, 5 (4): 205-222.View ArticleGoogle Scholar
- Gama-Castro S, Salgado H, Peralta-Gil M, Santos-Zavaleta A, MuÒiz-Rascado L, Solano-Lira H, Jimenez-Jacinto V, Weiss V, GarcÌa-Sotelo JS, LÛpez-Fuentes A: RegulonDB version 7.0: transcriptional regulation of Escherichia coli K-12 integrated within genetic sensory response units (Gensor Units). Nucleic Acids Research. 2011, 39 (suppl 1): D98-PubMed CentralPubMedView ArticleGoogle Scholar
- Keseler IM, Collado-Vides J, Santos-Zavaleta A, Peralta-Gil M, Gama-Castro S, MuÒiz-Rascado L, Bonavides-Martinez C, Paley S, Krummenacker M, Altman T: EcoCyc: a comprehensive database of Escherichia coli biology. Nucleic Acids Research. 2011, 39 (suppl 1): D583-PubMed CentralPubMedView ArticleGoogle Scholar
- Tian L, Greenberg SA, Kong SW, Altschuler J, Kohane IS, Park PJ: Discovering statistically significant pathways in expression profiling studies. Proceedings of the National Academy of Sciences of the United States of America. 2005, 102 (38): 13544-10.1073/pnas.0506577102.PubMed CentralPubMedView ArticleGoogle Scholar
- Zare H, Sangurdekar D, Srivastava P, Kaveh M, Khodursky A: Reconstruction of Escherichia coli transcriptional regulatory networks via regulon-based associations. BMC Syst Biol. 2009, 3: 39-10.1186/1752-0509-3-39.PubMed CentralPubMedView ArticleGoogle Scholar
- Adler P, Peterson H, Agius P, Reimand J, Vilo J: Ranking Genes by Their Co expression to Subsets of Pathway Members. Annals of the New York Academy of Sciences. 2009, 1158 (1): 1-13. 10.1111/j.1749-6632.2008.03747.x.PubMedView ArticleGoogle Scholar
- Sangurdekar DP, Srienc F, Khodursky AB: A classification based framework for quantitative description of large-scale microarray data. Genome Biology. 2006, 7 (4): R32-10.1186/gb-2006-7-4-r32.PubMed CentralPubMedView ArticleGoogle Scholar
- Storey JD, Akey JM, Kruglyak L: Multiple locus linkage analysis of genomewide expression in yeast. PLoS Biology. 2005, 3 (8): 1380-View ArticleGoogle Scholar
- Leek JT, Storey JD: Capturing heterogeneity in gene expression studies by surrogate variable analysis. PLoS Genet. 2007, 3 (9): 1724-1735.PubMedView ArticleGoogle Scholar
- Storey JD: The positive false discovery rate: A Bayesian interpretation and the q-value. Annals of Statistics. 2003, 2013-2035.Google Scholar
- Tuszynski J: caTools: Tools: moving window statistics, GIF, Base64, ROC AUC, etc. R package version. 2008, 1:Google Scholar
- Baba T, Ara T, Hasegawa M, Takai Y, Okumura Y, Baba M, Datsenko KA, Tomita M, Wanner BL, Mori H: Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Molecular systems biology. 2006, 2 (1):
- Datsenko KA, Wanner BL: One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. Proceedings of the National Academy of Sciences of the United States of America. 2000, 97 (12): 6640-10.1073/pnas.120163297.PubMed CentralPubMedView ArticleGoogle Scholar
- Lennox E: Transduction of linked genetic characters of the host by bacteriophage P1. Virology. 1955, 1 (2): 190-206. 10.1016/0042-6822(55)90016-7.PubMedView ArticleGoogle Scholar
- Alikhanian S, Iljina TS, Kaliaeva ES, Kameneva SV, Sukhodolec V: A genetical study of thymineless mutants of E. coli K12. Genet Res. 1966, 8 (1): 83-100. 10.1017/S0016672300009939.PubMedView ArticleGoogle Scholar
- O'Donovan GA: Thymidine metabolism in bacteria. 1977, DNA synthesis, present and future Plenum Publishing Corp, New York, 219-253.Google Scholar
- Sangurdekar DP, Hamann BL, Smirnov D, Srienc F, Hanawalt PC, Khodursky AB: Thymineless death is associated with loss of essential genetic information from the replication origin. Mol Microbiol. 2010, 75 (6): 1455-1467. 10.1111/j.1365-2958.2010.07072.x.PubMedView ArticleGoogle Scholar
- Camara JE, Breier AM, Brendler T, Austin S, Cozzarelli NR, Crooke E: Hda inactivation of DnaA is the predominant mechanism preventing hyperinitiation of Escherichia coli DNA replication. EMBO Rep. 2005, 6 (8): 736-741. 10.1038/sj.embor.7400467.PubMed CentralPubMedView ArticleGoogle Scholar
- Fonville NC, Bates D, Hastings PJ, Hanawalt PC, Rosenberg SM: Role of RecA and the SOS response in thymineless death in Escherichia coli. PLoS Genet. 2010, 6 (3): e1000865-10.1371/journal.pgen.1000865.PubMed CentralPubMedView ArticleGoogle Scholar
- Martin CM, Guzman EC: DNA replication initiation as a key element in thymineless death. DNA repair. 2011, 10 (1): 94-101. 10.1016/j.dnarep.2010.10.005.PubMedView ArticleGoogle Scholar
- Kuong KJ, Kuzminov A: Stalled replication fork repair and misrepair during thymineless death in Escherichia coli. Genes Cells. 2010, 15 (6): 619-634. 10.1111/j.1365-2443.2010.01405.x.PubMed CentralPubMedView ArticleGoogle Scholar
- Ahmad S, Pritchard R: A regulatory mutant affecting the synthesis of enzymes involved in the catabolism of nucleosides in Escherichia coli. Molecular and General Genetics MGG. 1971, 111 (1): 77-83. 10.1007/BF00286556.PubMedView ArticleGoogle Scholar
- Hammer Jespersen K, Munch Petersen A, Nygaard P, Schwartz M: Induction of enzymes involved in the catabolism of deoxyribonucleosides and ribonucleosides in Escherichia coli K 12. European Journal of Biochemistry. 1971, 19 (4): 533-538. 10.1111/j.1432-1033.1971.tb01345.x.PubMedView ArticleGoogle Scholar
- Koch AE, Burchall JJ: Reversal of the antimicrobial activity of trimethoprim by thymidine in commercially prepared media. Applied and Environmental Microbiology. 1971, 22 (5): 812-Google Scholar
- Neuhard J, Thomassen E: Turnover of the deoxyribonucleoside triphosphates in Escherichia coli 15 T during thymine starvation. European Journal of Biochemistry. 1971, 20 (1): 36-43. 10.1111/j.1432-1033.1971.tb01359.x.PubMedView ArticleGoogle Scholar
- Mollgaard H, Neuhard J: Biosynthesis of deoxythymidine triphosphate. Metabolism of nucleotides, nucleosides, and nucleobases in microorganisms. 1983, Academic Press, London, United Kingdom, 149-198.Google Scholar
- Weiss B: YjjG, a dUMP phosphatase, is critical for thymine utilization by Escherichia coli K-12. Journal of Bacteriology. 2006, JB. 01645-01606v01641Google Scholar
- Breitman T, Bradford R: The absence of deoxyriboaldolase activity in a thymineless mutant of Escherichia coli strain 15: a possible explanation for the low thymine requirement of some thymineless strains. Biochimica et biophysica acta. 1967, 138 (1): 217-PubMedView ArticleGoogle Scholar
- Ives DH, Morse PA, Potter VR: Feedback inhibition of thymidine kinase by thymidine triphosphate. J Biol Chem. 1963, 238: 1467-1474.PubMedGoogle Scholar
- Leer JC, Hammer Jespersen K, Schwartz M: Uridine phosphorylase from Escherichia coli. European Journal of Biochemistry. 1977, 75 (1): 217-224. 10.1111/j.1432-1033.1977.tb11520.x.PubMedView ArticleGoogle Scholar
- Kornberg A, Baker TA: DNA replication. 1992, WH Freeman New York, 5:Google Scholar
- Kunz BA, Glickman BW: Mechanism of mutation by thymine starvation in Escherichia coli: clues from mutagenic specificity. J Bacteriol. 1985, 162 (3): 859-864.PubMed CentralPubMedGoogle Scholar
- Tinkelenberg BA, Hansbury MJ, Ladner RD: dUTPase and uracil-DNA glycosylase are central modulators of antifolate toxicity in Saccharomyces cerevisiae. Cancer Res. 2002, 62 (17): 4909-4915.PubMedGoogle Scholar
- Goulian M, Bleile B, Tseng BY: The effect of methotrexate on levels of dUTP in animal cells. J Biol Chem. 1980, 255 (22): 10630-10637.PubMedGoogle Scholar
- Nichols RJ, Sen S, Choo YJ, Beltrao P, Zietek M, Chaba R, Lee S, Kazmierczak KM, Lee KJ, Wong A, et al: Phenotypic landscape of a bacterial cell. Cell. 2011, 144 (1): 143-156. 10.1016/j.cell.2010.11.052.PubMed CentralPubMedView ArticleGoogle Scholar