Giant yeast cells with nonrecyclable ribonucleotide reductase
- 132 Downloads
Ribonucleotide reductase (RNR) catalyzes the reduction of ribonucleotides to deoxyribonucleotides and thereby provides the precursors required for DNA synthesis and repair. In an attempt to test cell resistance to a permanent replicational stress, we constructed a mutant Saccharomyces cerevisiae strain containing exclusively nonrecyclable catalytic subunits of RNR that become inactivated following the reduction of one ribonucleoside diphosphate. In this rnr1C883A rnr3Δ mutant, the synthesis of each deoxyribonucleotide thus requires the production of one Rnr1C883A protein, which means that 26 million Rnr1C883A proteins (half the protein complement of a wild-type cell) have to be produced during each cell cycle. rnr1C883Arnr3Δ cells grow under constant replicational stress, as evidenced by the constitutive activation of the checkpoint effector Rad53, and their S phase is considerably extended compared to the wild type. rnr1C883Arnr3Δ mutants also display additional abnormalities such as a median cell volume increased by a factor of 8, and the presence of massive inclusion bodies. However, they exhibit a good plating efficiency and can be propagated indefinitely. rnr1C883Arnr3Δ cells, which can be used as a protein overexpression system, thus illustrate the robustness of S. cerevisiae to multiple physiological parameters.
KeywordsRibonucleotide reductase Saccharomyces cerevisiae DNA replication DNA checkpoints Protein overexpression
S phase is a period of great vulnerability for the genome of eukaryotic cells, due to several factors. First, chromosomes have to be completely unwound and duplicated. This process implies the formation of single-stranded DNA, which is chemically less stable than double-stranded DNA (Lindahl 1993). Second, replication forks can be blocked because of a depletion of deoxyribonucleoside triphosphates (dNTPs) or because of various impediments including DNA lesions, transcription complexes and replication slow zones in Saccharomyces cerevisiae (Mirkin and Mirkin 2007). Stalled forks have to be stabilized and restarted to prevent their collapse and the accumulation of DNA structures potentially leading to illegitimate recombination events and chromosome rearrangements (Branzei and Foiani 2007; Tourriere and Pasero 2007). Third, the logic of replication control makes eukaryotic cells especially sensitive to stresses able to erase the memory of cell cycle position during S phase (Nasmyth 1999). Indeed, eukaryotic genomes are replicated from large numbers of replication origins, the activity of which must be tightly coordinated. A two-step mechanism ensures that no origin can fire more than once in a cell cycle: (1) the assembly of prereplicative complexes (pre-RCs) at origins only occurs when cyclin-dependent kinase (Cdk) activity is low, and (2) origin firing, in contrast, can only occur when Cdks become active (Diffley 2004). Any temporary inactivation of the Cdks in S phase could lead to the assembly of pre-RCs and thereby to a partial, most probably lethal, rereplication of the genome. This could explain the observation that quiescent cells never arrest in the S phase and that the passage through the START or restriction points irreversibly commits the cells to the completion of the cycle up to exit, and entry into a new G1 phase.
Genomic stresses affecting specifically the S phase can be generated by inhibiting the DNA polymerases with drugs such as aphidicolin or by inducing a dNTP depletion, which causes fork stalling. Cell responses to acute replicational stresses inducing both an increase in fork stalling and an extension of S-phase length are usually observed during a brief period, shorter than a generation time. Here, we have sought to evaluate cell resistance over generations to a constitutive replication stress by mutating two genes encoding the catalytic subunits of ribonucleotide reductase (RNR) in S. cerevisiae. Ribonucleotide reductase [for reviews, see (Eklund et al. 2001; Kolberg et al. 2004)] reduces ribonucleoside diphosphates (NDPs) into deoxyribonucleoside diphosphates, thus catalyzing an essential step in the production of dNTPs required for DNA synthesis. In S. cerevisiae, RNR is a tetrameric enzyme composed of two dimeric subunits, R1 and R2. The catalytic R1 subunits, which hold the substrate-binding active site, are encoded by two homologous genes, RNR1 and RNR3, but the large majority of R1 subunits are Rnr1/Rnr1 homodimers (Elledge and Davis 1990; Domkin et al. 2002). The R2 subunit is an Rnr2/Rnr4 heterodimer. The reduction of one NDP by Rnr1 or Rnr3 is balanced by the formation of a disulfide bond between two cysteines of the active site (C218 and C443 in Rnr1), which inactivates the enzyme. The catalytic capacity of Rnr1 or Rnr3 is regenerated by the reduction of this disulfide bond by two C-terminal cysteines (C883 and C886 in Rnr1), which are ultimately reduced by thioredoxins or glutaredoxins (Mao et al. 1989, 1992; Camier et al. 2007).
We managed to construct an rnr1C883A rnr3Δ mutant strain containing nonrecyclable Rnr1C883A proteins as the only RNR catalytic subunits. rnr1C883Arnr3Δ cells grow under constant replicational stress and their S phase is largely extended compared to the wild type. However, they show a good plating efficiency and can be propagated indefinitely, thus demonstrating the tolerance of yeast cells to constitutive replication stress. rnr1C883Arnr3Δ mutants also exhibit other abnormalities such as a largely increased cell volume and the production of massive amounts of Rnr1C883A proteins, which accumulate in inclusion bodies. The viability of rnr1C883Arnr3Δ cells illustrates the robustness of yeast cells to multiple physiological parameters.
Materials and methods
Plasmids, strains and media
The pBAD54 (2μ, TRP1), pRS316/RNR1 (CEN, URA3, pRNR1-HA-RNR1) and pSC2/RNR1 (2μ, TRP1, pRNR1-HA-RNR1) plasmids have been described (Camier et al. 2007). The Quick-Change Site-Directed Mutagenesis Kit (Stratagene) was used to give the pSC2/rnr1C883A vector. The pRNR1-HA-rnr1C883A construct, harbored by the 4.7 kb BamHI-XmaI fragment of the pSC2/rnr1C883A plasmid, was inserted into the pRS424 GAL1 (2μ, TRP1, pGAL1) vector (Mumberg et al. 1994) digested with BamHI and XmaI to give the pRS424/rnr1C883A plasmid (we used the pRS424 GAL1 plasmid with the idea that the GAL1 promoter could further increase the transcription of the rnr1C883A allele if cells were grown on galactose, but this possibility has remained unexplored so far). The pSC2/rnr1C883A and pRS424/rnr1C883A plasmids gave similar results in all experiments.
To get the pRS424/rnr1C883A-green fluorescent protein (GFP) construct, the sequence encoding the Aequorea victoria GFP was amplified by PCR using the pFA6a-GFP(S65T)-kanMX6 vector (Wach et al. 1997) as a template and primers that were designed to allow the subsequent recombination of this sequence with the rnr1C883A gene, while destroying rnr1C883A stop codon. The PCR product was introduced into YPH499 cells along with the NotI-digested pRS424/rnr1C883A vector (a unique NotI site is located just after the stop codon of the rnr1C883A gene). pRS424/rnr1C883A-GFP constructs resulting from in vivo recombination between the pRS424/rnr1C883A plasmid and the GFP sequence were recovered from [Trp+] transformants. Additional details and primer sequences are available on request.
Yeast strains used in this study
MATa ura3-52 lys2-801amberade2-101ochretrp1-Δ63 his3-Δ200 leu2-Δ1
(Sikorski and Hieter 1989)
as YPH499 bar1Δ::LEU2
(Leroy et al. 2001)
as MCM185 rnr1Δ::kanMX rnr3Δ::his5++ pRS316/RNR1
as MCM185 rnr1Δ::kanMX rnr3Δ::his5++ pSC2/rnr1C883A
as MCM185 rnr1Δ::kanMX rnr3Δ::his5++ pRS424/rnr1C883A-GFP
as YPH499 bar1Δ::hphMX rnr1Δ::kanMX rnr3Δ::his5+leu2-Δ1::p405-BrdU-Inc + pSC2/rnr1C883A
MATa ade2-1 trp1-1 can1-100 leu2-3,112 his3-11,15 URA3::GPD-TK7xAUR1c::ADH-hENT1 bar1Δ::LEU2 cdc21Δ::kanMX
Unless indicated otherwise, cells were grown in YPD medium (1% yeast extract, 2% bacto-peptone, 2% glucose). Synthetic media SM and SMCA correspond to the synthetic minimal medium described in (Sherman 1991), supplemented with the bases and amino acids corresponding to the strains’ auxotrophic markers, and with casamino acids (2%), respectively.
Detection and quantification of BrdU incorporation
Equal amounts of genomic DNA (100 ng) were denatured in 0.4 M NaOH, spotted onto a nitrocellulose membrane and UV-crosslinked. Immunodetection was performed using a monoclonal antibody against BrdU (MD5300, Caltag Laboratories) and a secondary IgG coupled to horseradish peroxidase (W402B, Promega). Signals were detected by electrochemiluminescence (ECL kit, Amersham) and quantified with the ImageJ software. Serial dilutions of BrdU-marked DNA preparation were systematically performed so as to check the linearity of the signal. The fraction of replicated DNA y was deduced from the fraction of incorporated BrdU x (normalized by the maximal amount of incorporated BrdU) using the equation: y = x/(2 − x).
Yeast cells were fixed in 1% glutaraldehyde, washed and embedded either in Epon for ultrastructural studies (Fig. 4a–c) or in Unicryl for immunogold labeling (Fig. 4d–e) as described previously (Ni et al. 2006). For immunogold labeling, 90 nm-thick sections were labeled with purified anti-GFP antibodies (Ab290, Abcam) followed by 10-nm protein A-gold, and stained in uranyl-acetate and lead citrate as described (Ni et al. 2006).
Quantification of plasmid copy number
About 30 ng of DNA was digested with EcoRV, separated on a 0.8% agarose gel for 15 h at 70 volts and the resulting Southern blot was hybridized with a probe corresponding to the EcoRV-XcmI fragment of the TRP1 gene, which is present both at the TRP1 chromosomal locus and as a marker in the pSC2/rnr1C883A plasmid. The signal corresponding to the pSC2/rnr1C883A plasmid was normalized using the signal corresponding to the TRP1 locus. As a control, we found a normalized copy number close to 1 for the centromeric vector pRS314 (data not shown).
Quantification of rnr1C883A ARNm by RT-PCR
Reactions were performed with the Eurogentec kit MESA GREEN qPCR MasterMix Plus for SYBR Assay. For reverse transcription reactions, 200 ng of RNAs was used, and 0.5 ng of cDNAs was used for quantitative PCR. Signals corresponding to the RNR1 and rnr1C883A mRNAs were normalized using signals corresponding to the actin ACT1 mRNAs.
Determination of cell volume by FACS analysis
rnr1C883Arnr3Δ cells exhibit an extended S phase
In an attempt to analyze the resistance of S. cerevisiae cells to a constitutive replicational stress, we sought to construct a nonrecyclable version of Rnr1 by preventing the reduction of the catalytic cysteines. This was done with the idea that dNTP synthesis in cells containing only nonrecyclable forms of R1 would be strongly slowed down, which would cause a dNTP depletion during the S phase. The carboxyl end of Rnr1 is supposed to act as a flexible shuttle of reducing equivalents from the surface of Rnr1 to the active site, which suggests that the regenerating step of Rnr1 catalytic cycle can be easily disrupted without affecting NDP reduction, by mutating one of the C-terminal cysteines, C883 or C886. We had shown, indeed, that in vivo, a large part of mutant Rnr1C883A proteins (whose cysteine C883 is mutated into alanine) contains a disulfide bond between the two cysteines of the catalytic site (Camier et al. 2007). This result indicates that an Rnr1C883A protein becomes definitively inactivated by the reduction of one NDP, which acts like a suicide substrate.
We constructed a mutant rnr1C883Arnr3ΔS. cerevisiae strain by disrupting RNR1 and RNR3 chromosomal alleles and introducing the multicopy plasmid pSC2/rnr1C883A harboring the rnr1C883A allele under the control of its own promoter. In this strain, the nonrecyclable Rnr1C883A proteins are the only catalytic subunits of RNR. The rnr1C883Arnr3Δ mutants grew slowly, indicating either low viability or extended generation time.
Low, constitutive activation of the replication checkpoint in rnr1C883Arnr3Δ mutants
Replication stress can also be assessed via the activation of the S-phase checkpoint, a signal transduction pathway triggered by perturbations of DNA replication resulting from the presence of DNA lesions, defective replication proteins or dNTP depletion (Osborn et al. 2002; Longhese et al. 2003). dNTP depletion induced by treatment with hydroxyurea [HU, a potent inhibitor of RNR that quenches R2 radical (Ehrenberg and Reichard 1972)] has been intensively studied. In the presence of HU, the activation of the S-phase checkpoint, which can be readily monitored by the phosphorylation of Rad53 (one of its key components), prevents the irreversible dissociation of the replication machinery from DNA, restrains the activity of recombination enzymes at stalled forks, inhibits the elongation of the spindle and delays the firing of replication origins (Branzei and Foiani 2005, 2006).
We conclude from these two analyses that rnr1C883Arnr3Δ cells are subjected to a strong replication stress, although the activation of Rad53 remains limited. However, in spite of their constitutive replicational stress and extended S phase, rnr1C883Arnr3Δ mutants (MCM824) grew robustly in rich medium (YPD), with an apparent generation time (determined in triplicate as the time required for a doubling of the optical density of cell cultures at 660 nm) of 6.2 ± 0.4 h compared to 1.5 ± 0.1 h for the wild-type cells (MCM185). The plating efficiency of rnr1C883Arnr3Δ mutants (percentage of cells forming viable colonies, determined in triplicate for three clones) was 74 ± 10 in YPD, 71 ± 8 in synthetic minimum medium supplemented with casaminoacids (SMCA) and 55 ± 7 in synthetic minimum medium supplemented with the amino acids and bases required by the strain auxotrophic markers (SM), compared to 100 for the wild type (MCM185) in all cases. These results illustrate the robustness of rnr1C883Arnr3Δ cells to a severe, permanent replication defect, which is not their only anomaly, as detailed below.
The median volume of rnr1C883Arnr3Δ cells is increased by a factor of 8 compared to the wild type
Rnr1C883A proteins accumulate in massive inclusion bodies
Rnr1C883A proteins represent on average half the protein amount of rnr1C883Arnr3Δ cells
Analysis of the amplification steps in the production of Rnr1C883A proteins
With ~300,000 copies/cell, Rnr1 is one of the 25 most abundant proteins in exponentially growing yeast cells, as determined by Ghaemmaghami et al. (2003). Neglecting Rnr1 degradation, and considering that wild-type cells produce 300,000 Rnr1 proteins per cell cycle, the production of ~26 million Rnr1C883A proteins per cell cycle corresponds to an increase by a factor of ~90. To understand the main steps allowing this increase, we analyzed the copy number of the rnr1C883A allele and the corresponding mRNAs.
The copy number of the pSC2/rnr1C883A plasmid was determined by Southern blotting. We found that rnr1Δ rnr3Δ cells complemented with pSC2/rnr1C883A (the rnr1C883Arnr3Δ mutants) contained on average 21 ± 3 plasmid copies, whereas rnr1Δ rnr3Δ cells complemented with pSC2/RNR1 (the same vector harboring the wild-type RNR1 allele) contained only 4 ± 1 plasmid copies. The accumulation of a high number of plasmids containing a defective gene has already been observed for 2 μ-derived vectors harboring the leu2-d allele (Erhart and Hollenberg 1983). In that case, the low level of leu2-d expression leads to the predominance of cells with a high copy number of leu2-d-containing plasmids under conditions selective for leucine. Similarly, the higher copy number of the pSC2/rnr1C883A plasmid compared to pSC2/RNR1 in rnr1Δ rnr3Δ cells can be explained by the low activity of Rnr1C883A and the reduced generation time of cells accumulating pSC2/rnr1C883A.
Using quantitative PCR analysis, we found that the abundance of rnr1C883A mRNAs in asynchronously growing rnr1C883Arnr3Δ cells was equal to 470 ± 80 times the abundance of RNR1 mRNAs in asynchronously growing wild-type cells. The transcription of RNR1 is strongly induced in the S phase (Elledge and Davis 1990). Considering that the S phase occupies the largest part of the cell cycle in rnr1C883Arnr3Δ mutants, but only about one-third of it in wild-type cells, we can estimate that, during the S phase, the amount of rnr1C883A mRNAs in rnr1C883Arnr3Δ mutants is about 150 (470/3) times the amount of RNR1 mRNAs in wild-type cells. The 21-fold increase in copy number of the rnr1C883A allele harbored by the pSC2/rnr1C883A plasmid compared to the chromosomal RNR1 allele in wild-type cells cannot account for that difference. This observation could therefore be explained in at least two ways: (1) the transcriptional activity of the rnr1C883A allele harbored by the pSC2/rnr1C883A plasmid could be higher than that of the chromosomal RNR1 allele during the S phase (because: the replicational stress experienced by the rnr1C883Arnr3Δ mutants leads to an increase in the transcription of rnr1C883A; the plasmid contains only part of the RNR1 promoter; or the plasmidic chromatin context of rnr1C883A is different from the chromosomal chromatin context of RNR1) or (2) the replicational stress experienced by the rnr1C883Arnr3Δ cells could induce an increase in the half-life of rnr1C883A mRNAs.
In summary, both a high copy number of the rnr1C883A allele and either a higher transcriptional activity of the rnr1C883A allele or a higher stability of the rnr1C883A transcripts cause a strong increase of rnr1C883A mRNAs in rnr1C883Arnr3Δ cells compared to RNR1 mRNAs in wild-type cells. It has to be noted also that the increase of rnr1C883A mRNAs, compared to RNR1 mRNAs, seems higher than the increase of Rnr1C883A proteins, compared to Rnr1 proteins, which suggests that rnr1C883A mRNAs could be less efficiently translated than RNR1 mRNAs.
We have constructed a mutant strain containing only nonrecyclable RNR catalytic subunits, in which the production of each dNTP requires the synthesis of one 888 amino acid long Rnr1C883A protein. rnr1C883Arnr3Δ cells grow under constant replicational stress and their S phase is largely extended compared to the wild type. However, we found that rnr1C883Arnr3Δ mutants exhibit a good plating efficiency and can be propagated indefinitely, although they grow more slowly than the wild type. Yeast cells thus prove resistant to a constitutive depletion of dNTPs and to a large extension of their S-phase length.
Somewhat paradoxically, the dNTP depletion experienced by rnr1C883Arnr3Δ cells, although more severe than the one induced by a 20 mM HU-treatment in wild-type cells (as inferred from their impact on S-phase length), led to a weaker activation of Rad53. This low activation of Rad53 in rnr1C883Arnr3Δ cells could be explained by the dependency of Rad53 phosphorylation on the number of active replication forks, as suggested by the fact that mutants partially defective in initiation of DNA replication, such as clb5Δ, cdc7 and orc2-1 cells, were defective in activating the DNA checkpoint after genotoxic treatments (Shimada et al. 2002; Tercero et al. 2003). When α-factor-arrested wild-type cells are released into HU-containing medium, they start with normal dNTP pools; hence early origins are fired, giving numerous replication forks able to sense the ensuing dNTP depletion [Kitada et al. found that 116 replication origins out of a total of 260 were fired 90 min after α-factor-blocked cells were released in the presence of 0.2 M HU (Yabuki et al. 2002)]. In contrast, rnr1C883Arnr3Δ cells probably start their S phase with low dNTP levels. It is conceivable that only few origins can be fired before a signal from the first replication forks activates Rad53 and blocks further origin firing. This would explain that, in rnr1C883Arnr3Δ cells, Rad53 activation, triggered by a few replication forks at the beginning of the S phase, is low but sufficient to control further firing. Therefore, at any given time during the S phase, only few replication forks operate simultaneously, maintaining a limited activation of Rad53. Alternative explanations are also plausible, including a decrease in checkpoint activation due to adaptation or to abnormally low amounts of proteins involved in checkpoint signaling caused by the overexpression of Rnr1C883A.
Besides dNTP depletion, rnr1C883Arnr3Δ mutants also exhibit a large increase in median cell volume. For a given species in a given environment, cell size is generally held constant by processes coordinating cell growth and division [for recent reviews see (Jorgensen and Tyers 2004; Cook and Tyers 2007)]. However, the physical or physiological selective forces determining cell size are far from clear. Constraints induced by diffusion processes (which affect the import of nutrients and the “mixing times” of intracellular components) probably play a role in favoring small-sized cells, but only a partial one (Young 2006). The rnr1C883Arnr3Δ mutants illustrate yeast robustness to size variations.
Another particular aspect of rnr1C883Arnr3Δ mutants is the metabolic cost of dNTP production. Using the costs of amino acids and nucleotides, previously calculated by Wagner (2005) for S. cerevisiae, and considering growth on glucose under respiratory conditions, we found that the additional cost of one Rnr1C883A protein per dNTP increases the cost of a dNTP in rnr1C883Arnr3Δ cells compared to the wild type by a factor of 70 in rich medium (YPD or SMCA, containing all amino acids) and by a factor of 500 in minimal medium (SM, without amino acids). This observation is reminiscent of the large variations in DNA content found in related species of higher eukaryotes, since the metabolic cost of DNA should to a first approximation be proportional to the genome size. Variations of 340-, 240-, 190- and 130-fold have been measured in the genome sizes of flatworms, crustaceans, insects and amphibians, respectively (Gregory 2005). The fact that yeast cells can survive a 500-fold increase in DNA cost suggests that large increases in DNA content can take place in a small number of generations. In contrast with the large genomes of vertebrates and land plants that contain 90-98% of non-coding DNA and up to 120,000 megabases (Mb), the genomes of unicellular eukaryotes, including S. cerevisiae, are usually much smaller (3–100 Mb) and more compact as they contain 80–95% of coding DNA (Lynch 2006; Gregory et al. 2007). The fact that yeast cells can accommodate large increases in DNA cost also suggests that there is no stringent metabolic barrier opposing the expansion of DNA content by orders of magnitude in yeast, and that its streamlined genome can rather be attributed to the power of natural selection acting on its large populations to eliminate the metabolic and/or mutational burdens associated with increased DNA content (Lynch 2006).
Finally, another characteristic of rnr1C883Arnr3Δ mutants is the production of a quantity of Rnr1C883A proteins equivalent to half the protein complement of a wild-type cell, each time they replicate their DNA, and the accumulation of these proteins. Yeast cells can thus be propagated indefinitely, despite dramatic changes in their transcriptional and translational programs and the formation of massive inclusion bodies. This observation indicates that a protein of interest could be overproduced using fusion constructs including the rnr1C883A sequence, as exemplified by the Rnr1C883A-GFP fusion. This strategy could be used for continuous production since the synthesis of the Rnr1C883A-fusion protein is permanently required for cell division. No induction is necessary and cultures can be maintained under stable, optimized conditions.
We thank Jean Daraspe for expert technical assistance in electron microscopy, Jean-Yves Thuret for support in video-microscopy, Jean-Louis Sikorav for insightful comments, Sylvie Camier for discussions on RNR at the outset of this work, Etienne Schwob and Christelle de Renty for discussions on BrdU incorporation and detection and Sophie Chéruel for discussions on RNR constructs. This study was partly financed by grants from the Association pour la Recherche sur le Cancer (ARC) and from the Agence Nationale de la Recherche (ANR).
- Cousens LS, Shuster JR, Gallegos C, Ku LL, Stempien MM, Urdea MS, Sanchez-Pescador R, Taylor A, Tekamp-Olson P (1987) High level expression of proinsulin in the yeast, Saccharomyces cerevisiae. Gene 61:265–275Google Scholar
- Ghaemmaghami S, Huh WK, Bower K, Howson RW, Belle A, Dephoure N, O’Shea EK, Weissman JS (2003) Global analysis of protein expression in yeast. Nature 425:737–741Google Scholar
- Gregory TR, Nicol JA, Tamm H, Kullman B, Kullman K, Leitch IJ, Murray BG, Kapraun DF, Greilhuber J, Bennett MD (2007) Eukaryotic genome size databases. Nucleic Acids Res 35:D332–D338Google Scholar
- Mao SS, Holler TP, Yu GX, Bollinger JM, Jr., Booker S, Johnston MI, Stubbe J (1992) A model for the role of multiple cysteine residues involved in ribonucleotide reduction: amazing and still confusing. Biochemistry 31:9733–9743Google Scholar
- Nasmyth K (1999) Control of S Phase. In: DePamphilis ML (ed) Concepts in Eukaryotic DNA Replication. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, pp 331–386Google Scholar
- Ni J, Verbavatz JM, Rippe A, Boisde I, Moulin P, Rippe B, Verkman AS, Devuyst O (2006) Aquaporin-1 plays an essential role in water permeability and ultrafiltration during peritoneal dialysis. Kidney Int 69:1518–1525Google Scholar
- Pellicioli A, Lucca C, Liberi G, Marini F, Lopes M, Plevani P, Romano A, Di Fiore PP, Foiani M (1999) Activation of Rad53 kinase in response to DNA damage and its effect in modulating phosphorylation of the lagging strand DNA polymerase. EMBO J 18:6561–6572Google Scholar
- Zhang J, Schneider C, Ottmers L, Rodriguez R, Day A, Markwardt J, Schneider BL (2002) Genomic scale mutant hunt identifies cell size homeostasis genes in S. cerevisiae. Curr Biol 12:1992–2001Google Scholar