Buffering of deoxyribonucleotide pool homeostasis by threonine metabolism

Abstract

Synergistically interacting gene mutations reveal buffering relationships that provide growth homeostasis through their compensation of one another. This analysis in Saccharomyces cerevisiae revealed genetic modules involved in tricarboxylic acid cycle regulation (RTG1, RTG2, RTG3), threonine biosynthesis (HOM3, HOM2, HOM6, THR1, THR4), amino acid permease trafficking (LST4, LST7), and threonine catabolism (GLY1). These modules contribute to a molecular circuit that regulates threonine metabolism and buffers deficiency in deoxyribonucleotide biosynthesis. Phenotypic, genetic, and biochemical evidence for this buffering circuit was obtained through analysis of deletion mutants, titratable alleles of ribonucleotide reductase genes, and measurements of intracellular deoxyribonucleotide pool concentrations. This circuit provides experimental evidence, in eukaryotes, for the presence of a high-flux backbone of metabolism, which was previously predicted from in silico modeling of global metabolism in bacteria. This part of the high-flux backbone appears to buffer deficiency in ribonucleotide reductase by enabling a compensatory increase in de novo purine biosynthesis that provides additional rate-limiting substrates for dNTP production and DNA synthesis. Hypotheses regarding unexpected connections between these metabolic pathways were facilitated by genome-wide but also highly quantitative phenotypic assessment of interactions. Validation of these hypotheses substantiates the added benefit of quantitative phenotyping for identifying subtleties in gene interaction networks that modulate cellular phenotypes.

Cells are complex genetic systems, having evolved compensatory molecular networks that provide growth homeostasis (robustness). Conceptually, gene interactions underlie robustness by buffering environmental or genetic perturbations (1–3). Synergistic effects on the phenotype resulting from two genetic deficiencies or chemical inhibition in combination with a genetic deficiency reveal buffering relationships when the double limitation is more severe than either single limitation. Genome-wide phenotypic analysis, as possible with RNAi or use of the complete set of yeast gene deletion mutants, has enabled new approaches to investigate buffering relationships systematically (2, 4, 5). It has been shown that quantitative (strength) and qualitative (pattern) aspects of gene interaction profiles reveal how genes organize in a pathway or cellular process (4, 6). Conceptually, such sets of genes represent genetic modules that contribute buffering capacity to the cell, providing insight into how molecular circuitry is arranged to achieve robustness (7, 8). Comprehensive and quantitative methods for genotype–phenotype analysis are becoming available for gaining a more global and precise understanding of buffering networks (4, 6, 9). These methods permit unbiased experimental investigation of growth homeostasis, systematically revealing how combinations of genetic and environmental variables result in phenotypic complexity. High-throughput genotype–phenotype data offer an opportunity to use the extensive and growing genome annotations to discover new connections between previously annotated genes and pathways, with respect to physiological homeostasis. Systematic, experimentally derived understanding of genetic interaction networks will advance efforts to map natural phenotypic variation, thereby aiding the dissection of genetic disease complexity (10).

This work tests a model constructed after unexpectedly finding threonine biosynthesis to play a role in buffering growth inhibition with the deoxyribonucleotide (dNTP) biosynthesis inhibitor, hydroxyurea (HU) (4). HU is a chemotherapy agent that limits cell proliferation by inhibition of ribonucleotide reductase (RNR) leading to dNTP pool deficiency and slow DNA synthesis (11). The results provide genetic, biochemical, and phenotypic evidence that growth homeostasis is maintained by a molecular circuit that regulates threonine metabolism to buffer depletion of dNTP pools. These findings shed light on systems-level observations about cellular metabolism, including function of a high-flux backbone of metabolism (12) and gating of DNA synthesis by oscillation of global transcription and redox metabolism (13, 14).

Functional Interactions Between dNTP and Threonine Metabolism.

This work focused on understanding genetic modules found to buffer RNR deficiency (4). Synergistic interactions between HU and threonine biosynthesis genes, but not genes that function in the synthesis of other amino acids, were uncovered (Fig. 1a). Deletion of AAT2 (aspartate aminotransferase) was also synergistic, suggesting buffering by tricarboxylic acid (TCA) cycle flux because AAT2 converts the TCA cycle intermediate, oxaloacetate, to aspartate, which is the substrate for synthesis of homoserine and ultimately threonine (www.yeastgenome.org). RTG1, RTG2, and RTG3, transcription factors regulating transcription of TCA cycle genes in response to mitochondrial stress (15, 16), were also synergistic with hydroxyurea growth limitation, further implicating TCA cycle involvement (Fig. 1b). The RTG and threonine biosynthesis modules were independently confirmed to buffer HU-induced stress by Pan et al. (17). Synergistic interaction between HU and deletion alleles of LST4 and LST7 implicated extracellular uptake of threonine as an alternative mechanism to augment threonine flux (Figs. 1c and 2a) because LST4 and LST7 regulate delivery of amino acid permeases between the vacuole and plasma membrane compartments of the cell (18).

Fig. 1.

HU chemical–genetic interactions. Interactions for three genetic modules are depicted. The WT control is compared with deletion strains representative of each module, with respect to their area under the growth curve (AUGC) vs. perturbing drug concentration (AUGC = 0 means undetectable growth after 4 days). (a) Deletion of threonine biosynthesis genes involved in the conversion of oxaloacetate to aspartate (aat2) and subsequent conversions of aspartate to homoserine (hom3, hom2, and hom6) and homoserine to threonine (thr1 and thr4). Interactions caused by deletion of hom3, hom6, and thr4 are shown in d. (b) rtg1, rtg2, and rtg3 deletion strains are deficient in regulating TCA cycle gene transcription. (c) lst4 and lst7 deletions deregulate permease trafficking and extracellular uptake of threonine. (d) RNR1 transcription was dialed down after introduction of tetracycline repressor control elements into the indicated deletion strains (see Methods). Representative curves are shown.

Fig. 2.

Extracellular threonine suppresses interactions between HU and threonine biosynthesis. (a) A model explaining interactions between HU and genes involved in threonine metabolism. In the context of dNTP pool deficiency, threonine metabolism is up-regulated by intracellular production and extracellular uptake, ultimately for catabolism and production of glycine, which is primarily used for de novo purine biosynthesis. (b–d) Threonine biosynthesis mutants and WT controls were grown with normal (black) and three times the normal amount of methionine (white), threonine (charcoal), or lysine (light gray) (3 aa synthesized from aspartate) in the absence of HU (b), and in the presence of 50 mM HU (c) or 150 mM HU (d). AUGC is plotted on the y axis for each culture/condition. Error bars represent standard deviation.

To confirm that chemical–genetic interactions with HU were caused by its known inhibitory effect on dNTP biosynthesis, a more specific method was used. Integrating plasmids were introduced into a variety of mutant strains to place RNR1 or RNR2 under transcriptional control by doxycycline (19). Deletion of homoserine or threonine biosynthesis genes was found to be synergistic with repression of RNR activity by using doxycycline in these mutants (Fig. 1d), confirming that interactions with HU were caused by its inhibitory effect on RNR.

Media supplementation with amino acids was used to test whether uptake of extracellular threonine was able to suppress the growth limitation of mutations in threonine biosynthesis in the presence of HU. Threonine was found to selectively suppress interaction between HU and disruption of threonine biosynthesis, in a concentration-dependent manner (Fig. 2 b–d). This finding led to the prediction that disabling both threonine biosynthesis and threonine uptake would be synergistic in the presence of HU growth limitation. To test this hypothesis, double deletion mutants of the four possible combinations of thr1 or thr4 and lst4 or lst7 were created. All combinations were synthetic lethal even in the absence of HU, consistent with the hypothesis that lst4 and lst7 compensate threonine biosynthetic deficiency through regulation of extracellular uptake (Fig. 3). A slow-growth phenotype observed for the hom6 deletion mutant was exacerbated by extracellular threonine, even in the absence of HU (Fig. 2b). The hom6 deletion mutant is unique among threonine biosynthesis mutants in that the resulting intermediate metabolite (aspartate β-semialdehyde) is toxic (20), although whether this toxicity could be related to its different phenotype in the context of HU perturbation is unexplained.

Fig. 3.

Deficiency in amino acid permease trafficking (LST4 or LST7) is synthetic lethal with deletion of threonine biosynthesis (THR1 or THR4). Interactions between permease trafficking (LST4 and LST7) and threonine biosynthesis (THR1 and THR4) were assessed by tetrad dissection of the four double heterozygous mutants, constructed by mating the respective haploid mutants. (a) A representative series of dissected tetrads (on YPD medium without drug selection) is shown. Examples of the nonparental ditype (NPD), tetratype (T), and parental ditype (PD) tetrad phenotypes are indicated. For each segregant, “−” indicates the WT allele and “+” the deletion allele at the THR1_thr1/LST7_lst7 loci, respectively. (b) All dissection results are summarized. There was evidence of gene conversion in 14 of 53 tetrads among the four loci analyzed (including G418^R, ClonNat^R, and the heterozygous MET17/met17 and LYS2/lys2 loci), indicating greater than normal genome instability. Construction of strains for tetrad analysis is described in Methods.

Biosynthesis and Extracellular Uptake of Threonine Contribute to dNTP Pool Homeostasis.

To test the effect of threonine metabolism on dNTP pools, pools were measured in threonine metabolism deletion mutants perturbed by doxycycline-conditional repression of RNR2 transcription (Fig. 4). Although rnr2 deletion is lethal in a haploid, repression of RNR2 transcription only reduced the growth rate when an otherwise non-growth-inhibitory concentration (5 mM) of HU was present (data not shown). In contrast, RNR1 repression was growth-limiting without HU (see Fig. 1d) and did not sensitize growth to 5 mM HU (data not shown). The specificity of low-dose HU for growth inhibition in combination with repression of RNR2 is explained by the mechanism of action of HU. HU scavenges a tyrosyl radical that is present on the Rnr2p and that is required as a cofactor for ribonucleotide reduction (21).

Fig. 4.

Effect of threonine metabolism on dNTP pool homeostasis. (a) Intracellular dNTP pool concentrations are depicted 90 (black) and 360 (gray) min after exposure to the perturbations indicated by each block. Block 1 is the unperturbed WT (BY4741) strain in YPD medium. Block 2 is BY4741 perturbed with 5 mM HU. Block 3 is strain 15004 (Tet-regulatable RNR2 without doxycycline repression). Blocks 4–6 show the effect of doxycycline repression (1, 3, and 10 μg/ml, respectively) on RNR2. Block 7 shows the combined effect of 10 μg/ml doxycycline and 5 mM HU. (b) Intracellular dNTP pool concentrations are plotted for the WT control and indicated deletions strains (each transformed with Tet-regulatable RNR2), before (black), and 120 (gray) or 240 (white) min after perturbation with 10 μg/ml doxycycline and 5 mM HU. (c) Relative increases in cell number (CNI), median cell size (CSI), and total cell volume (TCVI), for cultures shown in b.

Non-growth-inhibitory concentrations of HU paradoxically induced increased steady-state dNTP pool concentrations. The increase in pools was sustained over time and additive with the effect of modulating RNR transcriptional levels (Fig. 4a). As a result, growth inhibition, caused by RNR2 repression combined with low-dose HU treatment, occurred with dNTP levels slightly higher than those in the untreated wild-type (WT) control strain (endogenous RNR2 promoter). A possible explanation is that increases in dNTP pools are required for growth fitness in the setting of DNA damage, which is known to involve RNR regulation (22). However, production of DNA damage (requiring increased dNTP pools for DNA repair) would have been expected only at HU concentrations high enough to arrest DNA synthesis in the first place (23). The observation that low concentrations of HU led to increased pools could be explained if DNA damage occurs by a mechanism independent of the effect of HU on cytoplasmic pools. A possible mechanism could involve dNTP pool concentrations at replication forks being affected differentially from cytoplasmic pools; however, this difference is not thought to occur in eukaryotic cells (24). Thus, the paradoxical effect of low HU concentrations on increasing dNTP pools remains unexplained.

dNTP pools were increased by expression of RNR2 from the Tet promoter (in the absence of repression with doxycycline), presumably because of overexpression relative to the endogenous RNR2 promoter. Dox-conditional repression of RNR2 reduced dNTP pools in a concentration-dependent fashion (Fig. 4a) so that synergism from deletion of threonine metabolism genes could be tested. The rtg2, hom2, thr1, and lst4 deletion mutants all exacerbated the reduction in dNTP pools after RNR2 repression (Fig. 4b). The contribution of RTG2 for dNTP pool maintenance was quantitatively less than that of HOM2, THR1, or LST4, a result consistent with their effects on growth (Fig. 1). Consistent with low dNTP pools causing cell cycle arrest in each of the mutants, median cell size increased whereas the relative number of cells and total cell volume (median cell size × median cell volume) decreased as pools became depleted over time (Fig. 4c). The scs7 (functions in sphingolipid metabolism) deletion strain maintained dNTP pools comparable with WT, despite a greater fitness defect (Fig. 4 b and c), indicating a specific role of threonine metabolic genes in homeostatic regulation of dNTP pools.

Threonine Aldolase Is Rate-Limiting for dNTP Metabolism in Saccharomyces cerevisiae.

The genetic, phenotypic, and biochemical results presented above are consistent with a model whereby TCA cycle regulation (RTG genes), threonine biosynthesis (HOM and THR genes), and permease trafficking (LST genes) pathways coordinately buffer dNTP pool depletion by up-regulating threonine metabolism. The model postulates that threonine catabolism contributes glycine to augment de novo purine synthesis (Fig. 2a). HU has been shown to preferentially deplete dATP pools in mammalian cells (25, 26), and there was a tendency for purine (particularly dATP) pools to fluctuate acutely whenever threonine metabolism and RNR activity were perturbed in combination (Fig. 4 a and b). However, allosteric regulation of RNR would tend to distribute this effect across all pools (21). Threonine aldolase, encoded by GLY1 (EC 4.1.2.5), cleaves threonine into glycine and acetaldehyde (27). Notably, the gly1 deletion mutant exhibited slow growth (data not shown) even with glycine supplementation. This phenotype was found to be the result of limitation of dNTP metabolism. Basal dNTP pools were reduced in the gly1 deletion mutant, pools fell dramatically after treatment with 10 mM HU, and normal homeostatic increases in dNTP concentrations after treatment with 50 mM HU were delayed, particularly dATP pools (Fig. 5). CHA1 (EC 4.2.1.13) and ILV1 (EC 4.3.1.19) are deaminases that convert threonine to 2-oxybutanoate or other metabolic intermediates such as homoserine, cystathionine, or propionyl-CoA. However, deletion of neither CHA1 nor ILV1 modified the growth response to HU (4).

Fig. 5.

Threonine aldolase contributes to normal dNTP metabolism. Intracellular dNTP pools are shown for the WT control strain (BY4741) and gly1 (threonine aldolase) deletion mutant before (Left) and 120 or 360 min after perturbation with 10 mM (Center) or 50 mM HU (Right). Homeostatic rebound in pools exhibited by the WT strain after HU treatment is blunted in the gly1 deletion mutant.

Discussion

Computational analysis of global metabolism in Escherichia coli has suggested that threonine flux is of particular importance for global metabolism. These studies propose that threonine synthesis and its degradation to glycine for purine biosynthesis are part of a high-flux backbone (HFB) of metabolism (12). The HFB was defined by a subset of all metabolic reactions found to have sufficient flux for providing growth homeostasis in response to growth-limiting perturbations (such shifting to a poor carbon source). Utilization of threonine for buffering dNTP metabolism and growth homeostasis provides experimental evidence for the presence of the HFB in eukaryotes.

Glycine can be synthesized from sources besides threonine, such as alanine or serine; however, deletion of the appropriate biosynthetic genes, AGX1 (EC 2.6.1.44) or SHM1/SHM2 (EC 2.1.2.1), was not synergistic with HU for growth limitation (4). Glycine is a substrate for generation of 1-carbon equivalents in the form of tetrahydrofolate derivatives, which are also needed for de novo nucleotide biosynthesis (28); however, deletion of genes required for glycine cleavage (GCV1, GCV2, and GCV3) did not exhibit slow growth or synergistic growth inhibition with HU (4).

In addition to glycine, acetaldehyde is a product of GLY1-mediated threonine catabolism. Murray et al. (29) have described rhythmic oscillation of acetaldehyde levels in synchronized yeast cultures and demonstrated that acetaldehyde functions as an attractor (synchronization agent) for these rhythms. Preferential use of threonine aldolase for threonine catabolism, with its associated production of acetaldehyde during cycles of ribonucleotide reduction and DNA synthesis, could partially explain why acetaldehyde concentrations oscillate along with global transcription during oxidative-reductive cycles that gate DNA replication (13, 14).

Threonine is an essential dietary amino acid for humans. However, extracellular uptake and/or catabolism of threonine could be used for dNTP metabolism in human cells because the molecular machinery for regulating extracellular uptake of amino acids through trafficking of permeases is conserved in LST8. LST8 is part of the TOR (target of rapamycin) pathway, involved in cancer and other human diseases (30–32) and acts with LST4, LST7, and SEC13 in regulating amino acid fluxes in yeast (18, 33). Although relatively little is known about threonine aldolase in mammals, it appears to function in mice but not humans (34). Additional studies will be needed to examine the potential importance of these pathway interactions in human disease.

Discovery of new connections between dNTP and threonine metabolism demonstrates the value of quantitative high-throughput cellular phenotyping for identifying functional redundancies in gene networks by measuring interactions between genetic module metabolism. The ability to detect relatively small effects of individual modules and to order their relative quantitative impact aided hypotheses about how these modules might relate to one another (4). By this approach, genes involved in TCA cycle regulation, threonine biosynthesis, amino acid permease trafficking, threonine catabolism, and ribonucleotide reduction were found to function as a modular circuit to maintain robust dNTP pools for DNA synthesis even though these modules appear to function independently in other contexts (1, 8, 35).

In natural (outbred) populations, compensatory networks also buffer genetic and chemical growth perturbations; however, the amount of genotypic and phenotypic variation renders dissection of interactions relatively intractable. By contrast, systematic analysis of yeast deletion mutants exposes interactions on a fixed genetic background but does not survey natural variation. Recently, segregants from a cross of S288C (the background used for systematic gene deletion) and a natural isolate have been genotyped at high resolution (36, 37). Quantitative high-throughput cellular phenotyping, applied in parallel to these strains and the comprehensive collection of yeast gene deletion mutants, would provide a dual strategy to deconstruct gene networks that buffer growth perturbations, by systematic analysis of all deletion mutants in parallel with surveying for natural occurrence. Quantitative genetic dissection of buffering networks in yeast thus provides a way to model genotype–phenotype variation on a genomic scale, providing insight into functional interactions between conserved pathways that potentially modulate human disease.

Methods

Strains.

Deletion mutants were from the MATa collection, created by the Saccharomyces Genome Deletion Project (http://yeastdeletion.stanford.edu) and purchased from Research Genetics (Invitrogen, Carlsbad, CA). The background (BY4741) genotype was MATa/his3/leu2/met17/ura3. Tetrad analysis was performed after first switching drug resistance cassettes from KAN^R to ClonNAT^R in the MATa/lst4, lst7, and thr1 deletion mutants by using plasmid 4339 (38). Double heterozygous mutants were obtained by mating these strains to the deletion strains (marked with G418 resistance) of the BY4742 background (MATb/his3/leu2/lys2/ura3) followed by selection with G418 and ClonNAT. These mutants were sporulated in 1% potassium acetate and dissected, and segregants were scored for G418 and ClonNat resistance.

Doxycycline-Regulated Gene Transcription.

A plasmid (pJH023) was constructed with the Tet-repressor sequence and Tet-repressible transactivator-coding sequence in a tail-to-tail orientation within a large multiple-cloning site so that these control elements can be targeted to desired loci by PCR amplification and directional subcloning 500 bp of the 3′ end of the respective promoter (adjacent to the transactivator) and 500 bp of the 5′ end of the ORF 3′ (adjacent to the TetO7 sequence). Recombinant plasmids were constructed in this way, linearized, and integrated by lithium acetate transformation at the RNR1 and RNR2 loci. pJH023 was constructed from pNEB193 (New England Biolabs, Ipswich, MA). The multiple-cloning site was expanded by annealing complementary synthetic oligonucleotides encoding the following restriction sites: AscI, XhoI, NotI, AvrII, FseI, NheI, NgoMIV, BamHI, PacI, KasI, and SbfI, digesting, and ligating directionally into pNEB193 between AscI and SbfI (pJH002). The oligonucleotide sequences used to extend the multiple-cloning site were 5′-gcatggcgcgccctcgaggcggccgccctaggggccggccgctagcggatccttaattaaggcgcccctgcaggatgc-3′ and 5′-gcatcctgcaggggcgccttaattaaggatccgctagcggccggcccctagggcggccgcctcgagggcgcgccatgc-3′.

The original KasI restriction site of pNEB193 was disrupted by KasI digestion, treatment with T4 polymerase, and religation. The Tet-conditional transactivator was amplified from pCM188 (sequence at http://web.uni-frankfurt.de/fb15/mikro/euroscarf/data/pCM188.txt) and cloned directionally from the AscI to EcoRI of pJH002, using primers 5′-ttggcgcgccATGTCTAGATTAGATAAAAGTAAAGTGATTAACAG-3′ and 5′-ttgaattcTTATTACGATCCTCGCGCC-3′, to create plasmid pJH020 (capital letters indicate annealing sequences for PCR). The nourseothricin resistance cassette was amplified from pAG25 (sequence at http://web.uni-frankfurt.de/fb15/mikro/euroscarf/data/pAG25.html) using primers 5′-gatcgacgtcgggcccCGACATGGAGGCCCAGAAT-3′ and 5′-gatcgacgtcgggcccACACTGGATGGCGGCGTTA-3′ and cloned into the AatII site of pJH020 to create plasmid PJH021. The TetO7 element of pCM159 (sequence at http://web.uni-frankfurt.de/fb15/mikro/euroscarf/data/pCM159.txt) was PCR-amplified using primers 5′-gcgatcaagcttCACTTCTAAATAAGCGAATTTCTTATG-3′ and 5′-gcgatcttaattaaTTTAGTGTGTGTATTTGTGTTTGTGTGTC-3′, digested, and directionally cloned between the HindIII and PacI sites of pJH021 to create pJH023. For targeting doxycycline control elements to the RNR1 locus, the 3′ promoter was PCR-amplified from genomic DNA using primers 5′-gcatcctaggGCTTGTTTACGCGTTTTATCC-3′ and 5′-gcatggcgcgccGATGTTAATATATCAACAAATAAAGTGTTG-3′, digested, and ligated directionally between AvrII and AscI restriction sites of pJ023 to create plasmid pJH025. The 5′ ORF of RNR1 was ligated between PacI and NheI after using primers 5′-catgcacgttaattaaATGTACGTTTATAAAAGAGACGGTCG-3′ and 5′-gcatgctagcGACGTTCGGCCACTTGAC-3′ for amplification, to create plasmid pJH031. Similarly, genomic sequences were PCR-amplified and subcloned for targeting to the RNR2 locus: the 3′ promoter or RNR2 was PCR-amplified using primers 5′-gcatcctaggACTATGCGAAATCCGGAGC-3′ and 5′-gcatggcgcgccGGTAATTGGACAAATAAATACGTGTA-3′, digested, and ligated directionally between AvrII and AscI of pJ023 to create plasmid pJH027. The 5′ ORF of RNR2 was ligated between PacI and NheI after using primers 5′-catgcacgttaattaaATGTACGTTTATAAAAGAGACGGTCG-3′ and 5′-gcatgctagcGACGTTCGGCCACTTGAC-3′ for amplification, to create plasmid pJH033. Correct targeting was confirmed by PCR of genomic DNA from newly created strains. RNR1 was confirmed with primers 5′-CGTTACCAAGTCAATGCTGAAC-3′ and 5′-ATCTTATCGAATTGGACAGGTTCT-3′, and RNR2 was confirmed with 5′-CTTGACATCGCGCGATCTT-3′ and 5′-AAGTCGGACAATGCATCGG-3′. Correct targeting was also confirmed by growth-inhibitory effects of doxycycline treatment (Figs. 1d and 4c).

Cell Proliferation Measurements.

Experiments represented in Figs. 1 and 2 were performed in Hartwell complete agar medium. High-throughput kinetic phenotyping (by imaging and image analysis) and area under the growth curve (AUGC) calculations were performed as described previously (4). AUGC encapsulates the overall growth phenotype of a strain with respect to time under a particular condition. AUGC is affected by initial population size (no. of cells transferred in a spot culture), lag time (delay before log-linear growth), maximum specific rate (actual log-linear rate), total efficiency (saturation density), and duration of the assay. For assessing the strength of a genetic interaction, the change in the AUGC conferred by a particular deletion allele relative to its WT control allele is considered with respect to perturbation intensity, e.g., concentration of HU, as depicted in Fig. 1. AUGC values for all mutants perturbed with 0, 50, and 150 mM HU are available at (http://genomebiology.com/2004/5/7/R49/additional).

dNTP Pool Sample Collections.

Strains were grown overnight in liquid medium at 30°C to a concentration of ≈3 × 10⁶ cells/ml and diluted to prewarmed medium with HU or doxycycline to achieve the desired cell and drug concentrations in a final volume of 30 ml. Each time point was grown separately and harvested when the cell concentration was ≈3 × 10⁶ cells per ml. Twenty milliliters of culture was collected by vacuum filtration and immediately washed with ice-cold medium, and filters were transferred to 2 ml of ice-cold medium (dNTP concentrations remain stable in iced medium for several hours). Cells were removed from the filter by vortexing, the sample was divided in half for duplicate readings, cells were pelleted, medium removed by aspiration, and cells were lysed with 40 μl of 0.1 M perchloric acid and then snap-frozen.

Cell Volume Measurements.

Cell volumes were measured by size analysis with a Coulter Counter (Beckman–Coulter, Fullerton, CA). The total cell volume of each culture (median cell size × total cell number) was used for calculating intracellular dNTP pool concentrations. Before vacuum filtration and lysis of each culture for mass spectrometry analysis, 200 μl was collected into 10 ml of ice-cold isoton (Beckman–Coulter). Samples were sonicated at low power to separate nonspecifically adherent cells. To calculate relative changes in total cell volume (Fig. 4c), values for each strain were first normalized against self at time zero and then divided by the corresponding normalized WT (BY4741) values.

HPLC.

Samples were thawed by microcentrifugation (18,000 × g) for 15 min at 4°C. Sixteen microliters of lysate was added to 8 μl of 3× mobile-phase buffer [60 mM acetic acid/0.075% dimethylhydroxylamine (Sigma, St. Louis, MO)/pH adjusted to 7 with ammonium hydroxide], and 10 μl was injected onto an Agilent C-8 Zorbax column (part 883700-906) with a linear 5–30% methanol gradient from 2 to 11 min, 30–50% from 11 to 12 min, with final reequilibration for 5 min in 5% methanol (flow rate of 0.3 ml/min). Retention times of 4.5 (dTTP), 7.5 (dGTP and dTTP), and 9.5 min (dATP) were observed. dNTP-depleted lysate was obtained by lysis of saturation-density cultures after 30-min incubation in room temperature water. Dilution of standards in this lysate improved dCTP chromatography. Trace amounts of dNTPs remaining in the diluent were subtracted for standard curve calculations.

Mass Spectrometry.

Mass spectrometry was performed with electron spray ionization in negative ion mode. Two instruments were used: (i) an Agilent 1100 MSD [dNTPs were monitored as single ions at m/z 466 (dCTP)], 481 (dTTP), 490 (dTTP), and 506 (dGTP). The drying gas was N₂ at 340°C at 10 liters/min, and nebulizing pressure 25 psi (1 psi = 6.89 kPa). The fragmentor was set at 90 eV and capillary voltage 3500. (ii) An ABI API-4000 Q-trap triple quadrupole instrument was used [mass transition to a 189 fragment was monitored for each of the dNTP species, as described previously (39); N₂ gas was used for nebulization, drying, and collision and the ionization chamber temperature was 250°C]. New standard curves were created for every assay.

Calculation of Intracellular dNTP Concentrations.

Sample concentrations were determined from standard curves and adjusted to account for dilution by lysis and total cell volume [volume added for lysis + 2(tcv)] μL /tcv (μL). Standard curves showed high linear correlation (R² > 0.998), and variation from duplicate mass spec measurements was generally <10%.

Abbreviations

AUGC

area under the growth curve

dNTP

deoxyribonucleotide

HU

hydroxyurea

RNR

ribonucleotide reductase

TCA

tricarboxylic acid.

Note Added in Proof.

The requirement reported here of mitochondrial-to-nucleus retrograde signaling for dNTP pool homoeostasis in yeast may be of importance to a recent report that mutations in p53R2 cause human mitochondrial depletion syndromes (MDS) (40, 41). If compensatory/buffering relationships between RNR and retrograde signaling in yeast are evolutionarily conserved, then genetic variation in retrograde signaling may modulate MDS disease phenotypes resulting from deficiency in p53R2 activity.