After exposure to DNA-damaging agents, both prokaryotic and eukaryotic cells activate stress responses that result in specific alterations in patterns of gene expression. Bacteria such as Escherichia coli possess both lesion-specific responses as well as an SOS response to general DNA damage, and the molecular mechanisms of these responses are well studied. Mechanisms of DNA damage response in lower eukaryotes such as Saccharomyces cerevisiae are apparently different from those in bacteria. It becomes clear that many DNA damage-inducible genes are coregulated by the cell-cycle checkpoint, a signal transduction cascade that coordinates replication, repair, transcription and cell-cycle progression. On the other hand, among several well-characterized yeast DNA damage-inducible genes, their effectors and mechanisms of transcriptional regulation are rather different. This review attempts to summarize the current state of knowledge on the molecular mechanisms of DNA damage-induced transcriptional regulation in this model lower eukaryotic microorganism.
DNA is the carrier of genetic information in most organisms. Any damage to the molecular structure of DNA has the potential to cause genomic instability, mutagenesis or even cell death. Unfortunately, DNA damage is unavoidable; DNA is continually exposed to insults resulting from exogenous and endogenous DNA-damaging agents as well as challenges posed by DNA replication. Therefore, it is not surprising that living organisms have developed numerous pathways to deal with DNA damage. In response to DNA damage, cells can alter many of their intracellular processes, including DNA repair, metabolism and cell-cycle progression, for cell survival and maintenance of genomic stability (Friedberg et al., 2006). Such dynamic responses can be achieved through a variety of molecular mechanisms such as protein localization, protein degradation, posttranslational modification and genetic regulation. Generally speaking, altering gene expression is perhaps one of the most fundamental responses and it is reasonable to suggest that most, if not all, cellular processes are affected at the level of transcriptional regulation. In the field of DNA repair and mutagenesis, the notion that DNA damage causes an alteration in the expression profile of damage responsive genes has been an important area of research for many years. It is expected that the expression profile of a cell following DNA damage will provide us with critical information on how a cell protects itself from such stress.
Transcriptional regulation in response to DNA damage has been studied extensively in model bacteria such as Escherichia coli. In contrast, such information is relatively scarce in eukaryotic organisms, from simple lower unicellular eukaryotes such as Saccharomyces cerevisiae to humans. In this review, we attempt to summarize recent findings related to gene regulation in response to DNA damage in lower eukaryotes, particularly the budding yeast S. cerevisiae, and to compare this response with the well-studied bacterial SOS response.
Bacterial transcriptional responses to DNA damage
The SOS response in E. coli
When E. coli cells are subjected to DNA damage, about 48 unlinked genes are coordinately induced through a complex SOS regulatory network (Courcelle et al., 2001). The increased expression of these SOS regulon genes results in the elaboration of a set of physiological responses such as an enhanced capacity for recombination repair and excision repair, enhanced mutagenesis (due to error-prone translesion DNA synthesis mediated by PolIV and PolV) and inhibition of cell division. These responses have been collectively termed the SOS response (Radman et al., 1975; Witkin et al., 1976; Gottesman et al., 1981; Little & Mount, 1982).
Current model for transcriptional control of the SOS response
The SOS regulatory network is mainly controlled by two proteins: RecA and LexA (Radman, 1975; Little & Mount, 1982; Walker, 1984). LexA is a transcriptional repressor that binds to SOS boxes located near or inside the operator site of the SOS-induced genes. SOS boxes are often palindromic structures with a high degree of homology in nucleotide sequence. An ideal symmetrical consensus sequence 5′-TACTGTATATATATACAGTA-3′ was derived from the analysis of a pool of SOS boxes (Berg, 1988). The sequence distinction in SOS boxes allows the LexA repressor to bind to operators with different strengths (Lewis et al., 1994). The LexA occupancy prevents accessibility to RNA polymerase so as to inhibit the initiation of transcription. LexA is likely to interact with DNA via its N-terminal domain, and the C-terminus of LexA is required for its dimerization. The dimerization of LexA is essential for its ability to repress SOS-regulated genes in vivo. Meanwhile, LexA is able to undergo a slow intramolecular self-cleavage termed autodigestion, whose rate significantly increases upon interaction with RecA (Little et al., 1984, 1991, 1993).
The RecA protein of E. coli has at least three functions in the SOS response. It not only plays a role in the transcriptional regulation in response to DNA damage but also directly participates in translesion synthesis and homologous recombination. Following DNA damage, DNA synthesis becomes discontinuous and single-strand DNA (ssDNA) is produced by failed attempts to replicate damaged DNA. In the presence of ATP, RecA binds to the ssDNA region and forms helical RecA-ssDNA nucleoprotein filaments. LexA then diffuses to deep grooves in the RecA-ssDNA filaments and interacts with them in a manner that results in autocatalytic cleavage of LexA at a scissile peptide bond located between Ala84 and Gly85. Cleavage of LexA inactivates its ability as a repressor, thus releasing the SOS regulon genes from transcriptional repression. As cells begin to recover from the inducing treatment by various DNA repair and tolerance processes, the regions of ssDNA disappear, and thus the inducing signal is diminished. Without the cleavage stimulated by RecA-ssDNA filaments, the pool of LexA is boosted, which leads to repression of the transcription of SOS regulon genes and a return to the uninduced state (Walker, 1984; Friedberg et al., 2006).
Fine tuning in the induction of the SOS response
The SOS regulatory system provides E. coli with a rapid transcriptional response to the presence of DNA damage. Furthermore, during SOS induction, the timing, the duration and the level of induction are diverse for different LexA-regulated genes, suggesting a fine-tuning mechanism in the SOS induction. The fine tuning is possibly determined by at least four parameters: (1) the binding affinity of LexA for the SOS box in the operator region; (2) the number of SOS boxes in the operator region; (3) the location of the SOS box relative to the promoter; and (4) the strength of the promoter.
After exposing E. coli cells to DNA-damaging agents, genes with operators that bind LexA relatively weakly are the first to turn on fully. For example, uvrA, uvrB, ruvA, ruvB, recN and sulA are induced within 5 min after 40 J m−2 UV radiation (Courcelle et al., 2001). uvrA and uvrB encode proteins involved in nucleotide excision repair (NER); recN, ruvA and ruvB encode proteins used for recombination repair; while the protein product of the sulA gene can temporarily arrest cell division to allow bacteria time to complete the repair of damaged DNA. After the activation of loosely controlled SOS genes, if the damage cannot be fully repaired by NER and homologous recombination, genes with operators that are tightly controlled by LexA will be turned on. For example, the full induction of umuC and umuD is not observed until 20 min after 40 J m−2 UV irradiation (Courcelle et al., 2001). Similar to LexA, the protein encoded by umuD also has a latent ability to auto-digest, and the auto-digestion is strongly stimulated by the interaction between the RecA/ssDNA nucleoprotein filament and UmuD (Nohmi et al., 1988). UmuC and a posttranslationally processed form of UmuD (UmuD′) serve as a mutagenic lesion-bypass DNA polymerase (PolV). This last response allows the survival of E. coli after severe DNA damage, but at the expense of introducing errors into the genome.
In the SOS regulatory network, the transcript level of key regulatory genes recA and lexA is also regulated by LexA, thus forming a delicately controlled circuit (Brent & Ptashne, 1980; Little et al., 1981; Brent, 1982). recA has one SOS box positioned between the −35 and −10 regions of the promoter, and a tandem pair of SOS boxes is located in the lexA promoter region (Brent & Ptashne, 1981; Little et al., 1981; Schnarr et al., 1991). LexA can bind to these SOS boxes to prevent the initiation of transcription. Relative binding affinity experiments reveal that LexA binds to the recA operator more strongly than to the lexA operator and many other operators of SOS regulon genes such as uvrA, uvrB and uvrD (Brent & Ptashne, 1981; Peterson & Mount, 1987; Schnarr et al., 1991). Accordingly, it allows an intermediate inducible state, bridging an uninduced state and a fully induced state (Little, 1983). A low amount of inducing signal can thus lead to the activation of some of the SOS functions, such as uvr+-dependent NER, without substantial amplification of the RecA protein. When the inducing signal continues to accumulate, recA will be induced, resulting in a full induction of SOS regulon genes. Meanwhile, due to the repression of lexA by LexA itself, the SOS response system is robust to prevent substantial induction caused by very small amounts of inducing signal. Furthermore, the strong repression of recA by LexA and the constant expression of LexA in SOS response ensure a fast return to the uninduced state once the level of the inducing signal begins to decrease (Walker, 1984).
In addition to the two central regulators RecA and LexA, recent studies reveal that some other proteins are also involved in the subtleties of SOS induction. All these proteins appear to affect SOS regulation by modulating RecA-ssDNA stability. RecX can block the assembly of the RecA-ssDNA filament while not affecting the disassembly through capping the assembly ends (Drees et al., 2004). As a result, it strongly inhibits RecA-mediated DNA strand exchange, ATPase and coprotease activities. The recX gene is located 76-bp downstream of recA, and these two genes belong to the same operon. Although recX is cotranscribed with recA, recX transcription is downregulated with respect to recA by an intrinsic transcription terminator that is located between the recA and recX coding sequences. Despite the presence of this terminator, a recA–recX message resulting from transcriptional read-through is detected at a level of 5–10% of the recA message (Pages et al., 2003). RecX is barely detected during vegetative growth, but robust expression of recX is observed after treating cells with DNA-damaging agents (Stohl et al., 2003). The maximal recX expression is observed at a later time than maximal expression of RecA after UV irradiation (Courcelle et al., 2001). All these observations suggest that RecX is likely involved in the subtle regulation that helps shut off the SOS response.
Early studies have reported that DinI can destabilize the RecA-ssDNA filament when its concentration is 50–100-fold above the natural RecA concentration, and it inhibits all activities of RecA (Yasuda et al., 1998; Voloshin et al., 2001). Therefore, DinI was initially thought to aid the return of SOS-induced cells to a steady state. In contrast, recent research has shown that DinI is also able to stabilize the RecA-ssDNA filament to prevent its disassembly when present at concentrations that are stoichiometric with or somewhat greater than those of RecA (Lusetti et al., 2004). Furthermore, DinI-mediated stabilization affects RecA-mediated UmuD cleavage rather than RecA-mediated ATP hydrolysis and LexA coprotease activities (Lusetti et al., 2004), indicating that DinI could have a biological role in fine-tuning the activity of UmuD in order to limit SOS mutagenesis.
Role of DNA helicases and nucleases in SOS induction
A critical step in the SOS response is the production of a RecA-ssDNA filament, and this step requires DNA helicases and nucleases. Either the RecFOR or the RecBCD pathway is necessary for the SOS response after UV irradiation (Ivancic-Bace et al., 2006). In E. coli, SOS induction immediately after UV irradiation is dependent on the RecFOR pathway. The RecFOR pathway includes the DNA helicase RecQ, the nuclease RecJ and the RecFOR complex, which facilitates RecA loading (Lusetti et al., 2006). It has been suggested that the pathway may aid RecA binding to ssDNA gaps. RecQ appears to be needed for fast degradation of the LexA repressor (Hishida et al., 2004). This observation leads to a model in which RecQ unwinds the template duplex in front of a stalled fork on the leading strand, and then switches over to the lagging strand to generate ssDNA on the leading strand template, allowing formation of the RecA filament in the 5′–3′ direction for SOS induction (Heyer et al., 2004; Hishida et al., 2004).
Bacteria also possess damage-inducible systems that are independent of the SOS response. For example, alkylating or oxidative agents induce not only the SOS response but also other more specific DNA damage responses.
Following exposure to a sublethal dose of an alkylating agent (e.g. N-methyl-N′-nitro-N-nitrosoguanidine, MNNG), bacteria such as E. coli manifest a pronounced resistance to both lethal and mutagenic effects caused by a much higher dose of the same or a similar alkylating agent (Jeggo et al., 1977). The resistance is due to the induction of one or more genes in response to low levels of the alkylating agents; the protein products of these genes can directly remove the alkylated adducts in DNA (Samson & Cairns, 1977). Therefore, this regulatory pathway is called the adaptive response to alkylation damage.
In E. coli, the Ada protein possesses dual functions and plays a pivotal role in the adaptive response: it is a methyltransferase that removes alkyl groups from the methylated base and a strong transcriptional activator of several genes. The 39-kDa Ada protein is comprised of two functional domains linked by a hinge region. The C-terminal 19-kDa domain is capable of acquiring the methyl group from a mutagenic O6-methylguanine or O4-methylthymine to its Cys321 residue. The N-terminal 20-kDa domain specifically catalyzes the transfer of methyl groups from methyl phosphotriester lesions to its Cys38 residue in a direct and irreversible way (Demple et al., 1985; Nakabeppu & Sekiguchi, 1986; Teo et al., 1986; Lindahl et al., 1988; Myers et al., 1993b; Takinowaki et al., 2006). Methylation of the Cys38 triggers a conformational change in the Ada protein, thereby dramatically (up to 1000-fold) enhancing the sequence-specific DNA-binding affinity of the Ada protein to the promoter regions of its own gene, ada, and other alkylation resistance genes including alkA, aidB and alkB (Teo et al., 1986; Sakumi & Sekiguchi, 1989; Akimaru et al., 1990; Myers et al., 1993a; Sakashita et al., 1993; Takinowaki et al., 2006). The transcriptional regulatory element (Ada box) in the ada gene to which methylated Ada binds has been defined precisely, with the sequence 5′-AAAGCGCA-3′. Methylated Ada also displays binding affinity for the sequences 5′-AAANNAAAGCGCA-3′ and 5′-AAT(N)6GCAA-3′ in the alkA and aidB promoter regions, respectively (Teo et al., 1986; Nakamura et al., 1988; Landini & Volkert, 1995).
Because the alkylation of Ada is irreversible, how is the adaptive response switched off? Several mechanisms have been proposed. One suggests that in the absence of alkylation adducts, the activated Ada is simply diluted by cell divisions (Lindahl et al., 1988). Research has revealed that physiologically relevant high concentrations of unmethylated Ada are able to inhibit the activation of ada transcription by methylated Ada, both in vitro and in vivo (Saget & Walker, 1994). It is also possible that activated Ada is proteolytically degraded. The proteolytic cleavage of activated Ada in the hinge region linking the N-terminal to C-terminal domain of the protein might downregulate the expression of ada. Consistent with this model, a methylated 20 kDa Ada can bind to the ada promoter; however, it does not facilitate further binding of RNA polymerase to the promoter nor does it promote ada transcription in vitro (Akimaru et al., 1990).
The expression of genes responding to oxidative DNA damage can be induced by the presence of reactive oxygen species (ROS). Two regulatory responses induced by ROS have been identified in E. coli– one controlled by soxRS and the other by oxyR. In the soxRS regulatory system, SoxR serves both as a sensor and as an activator. In the presence of ROS, SoxR forms 2Fe–2S centers, which convert SoxR to an active state. The activated SoxR induces transcription of soxS, a positive regulator that stimulates transcription of superoxide-responsive genes. Upon relief of oxidative stress, SoxR is rapidly converted to its transcriptionally inactive form, thus turning off the response (Wu & Weiss, 1992; Hidalgo & Demple, 1994; Hidalgo et al., 1995; Ding et al., 1996; Gaudu et al., 1997; Volkert & Landini, 2001). The oxyR regulatory system is in response to hydrogen peroxide (H2O2). H2O2 activates the transcriptional activity of OxyR through formation of a disulfide bond between two of its cysteine residues. Activated OxyR induces the transcription of several oxidative stress genes (Zheng et al., 1998; Aslund et al., 1999; Volkert & Landini, 2001).
Transcriptional responses to DNA damage in S. cerevisiae
From the above analyses, it appears that bacteria possess transcriptional responses to specific types of DNA lesions as well as general DNA damage, and that these regulatory pathways are fairly well understood. In contrast, DNA damage response in eukaryotes is apparently more complex than in prokaryotes. Rather than direct cleavage of transcription repressors, eukaryotes most likely use posttranslational modifications, such as phosphorylation and ubiquitination, and a sophisticated signal transduction cascade to achieve gene regulation in response to DNA damage. Here, we attempt to summarize the current knowledge of mechanisms of DNA damage response in S. cerevisiae, a unicellular lower eukaryotic model organism. Following a historic overview of identification and characterization of genes whose transcript levels are altered following treatment of cells with DNA-damaging agents, this review focuses on regulatory responses that are either common or unique to some well-studied genes.
Genome-wide DNA damage-induced expression studies in yeast
Following work in prokaryotes revealing that DNA damage causes alterations in the transcriptional program of a cell, an important initiative was to identify a comparable response in the lower model eukaryote, S. cerevisiae. Although some of these experiments were performed one gene at a time, a precedent for systemic studies of transcription in bacteria (Kenyon & Walker, 1980) and the emergence of high-throughput technologies spurred the use of large-scale screens for such studies in yeast. To originally identify some of the genes that might be induced in response to DNA damage in S. cerevisiae, two early studies utilizing different genomic screens for DNA damage-induced genes are most notable. The first was based on a transcriptional fusion technique that is still widely used today for routine expression analysis. Using an experimental approach used previously in E. coli to screen for DNA damage-inducible genes (Kenyon & Walker, 1980), Szostak and colleagues tested yeast genes as a random pool of lacZ fusions and screened for coordinately regulated DNA damage-inducible genes over various DNA damage treatments. From c. 8000 independent clones, a total of six damage-inducible (DIN) genes were identified, including one gene from a previous screen (Ruby et al., 1983; Ruby & Szostak, 1985). Based on different expression profiles of DIN genes to specific DNA-damaging agents, the authors proposed at least two DIN classes. Such grouping of inducible genes is analogous to the clustering of large-scale data sets that is often necessary to manage and interpret modern DNA microarray experiments. Meanwhile, McClanahan & McEntee (1984) used a differential plaque hybridization strategy to identify genes whose transcript levels are altered after DNA damage treatment. Radiolabeled cDNA probes were generated from poly(A) mRNA harvested from control and treated cells, and subsequently used for differential plaque hybridization. By comparing the intensity of labeled probe bound between the control and the treated samples, differential expression of genes could be evaluated. Altogether, c. 9000 genomic clones were screened and four DNA damage-responsive (DDR) genes were identified. The experimental approach also enabled the identification of another set of genes, with decreased expression in response to DNA damage treatment. This latter finding testifies to the complexity of the DNA damage response in eukaryotes, because no such deactivation of genes was ever found in the E. coli SOS response (Walker et al., 1984).
These two early screens were significant to the field of transcriptional response to DNA damage in yeast. Not only were they among the first to demonstrate altered levels of transcripts in S. cerevisiae after DNA damage insult but also they were carried out in a genome-wide manner and foreshadowed some of the benefits and complications associated with analogous DNA microarray experiments that would follow. The powerful yet straightforward experimental global approaches also paid dividends by considerably expanding the number of genes known to be induced by DNA damage at the time. Together, 10 new DNA damage-inducible (DIN/DDR) and four DNA damage-repressive genes were identified. These studies allowed estimation of total DNA damage-responsive genes. Because the genomic library of Ruby and Szostak likely contained only c. 500 yeast genes representing about 8% of the genome, extrapolation led to an estimation of roughly 80 DNA damage-inducible genes in yeast. Significantly, the two independent screens yielded nonoverlapping data sets, indicating that neither screen was saturated and that the number of DNA damage-inducible genes was probably underestimated. Thus, even in the earliest stages of the studies of yeast DNA damage-induced gene expression, it was obvious that S. cerevisiae would have a much larger repertoire of DNA damage-induced genes than E. coli.
Despite the above large-scale experimental approaches, a plausible high-throughput and truly global approach was not available until the emergence of genomic technologies. In particular, DNA microarrays have provided the majority of damage-induced gene expression data in budding yeast to date. This wealth of expression data provided by DNA microarrays has revealed a new realm for conceptualizing regulatory networks in a global manner. In the following paragraphs, we highlight only the first DNA microarray studies of DNA damage-induced expression in yeast, and note that there are numerous other significant studies that cannot be feasibly covered in this review. Table 1 is provided for an overview of genomic expression studies in S. cerevisiae focusing on DNA damage treatments or related gene mutants. We refer the reader to the Yeast Functional Genomics Database (http://yfgdb.princeton.edu/) for a navigable compilation of these and related studies, 93 of which are currently listed as genomic expression studies involving stress.
The premiere microarray study on DNA damage-induced expression in S. cerevisiae was published in 1999 (Jelinsky & Samson, 1999). Representative sequences of 6218 yeast ORFs were arrayed and cRNA probes were generated from untreated cultures or cultures exposed to the DNA-alkylating agent methyl methanesulfonate (MMS). As evidenced by the overlap with the previous independent data mentioned above, the results validated the use of DNA microarrays for damage-induced expression analysis. Of the genes already known to be induced by DNA-damaging agents by all other methods combined, 86% (18 out of 21) were identified in the DNA microarray experiment. The use of a single MMS dose and a fourfold minimum cut-off likely explain the omission of the three other previously identified inducible genes. Altogether, an astonishing 325 yeast genes (c. 5% of entire yeast genes) were found to be induced by MMS treatment, with a significant portion (112) representing uncharacterized ORFs. In addition, 76 ORFs had reduced expression upon MMS exposure. The reduced expression of some genes after DNA damage treatment is reminiscent of the early DDR screen (McClanahan & McEntee, 1984). Altogether, a single DNA microarray experiment augmented the complement of known DNA damage-induced genes in S. cerevisiae by over 20-fold. Despite this wealth of new data, however, meaningful conclusions regarding transcriptional regulation were not possible with such an abundance of raw, unorganized data.
A follow-up study (Jelinsky et al., 2000) provided a more comprehensive analysis of transcriptional response to DNA damage and demonstrated how large amounts of microarray data can be manageably interpreted. This enabled the use of a large dataset to address a specific hypothesis, namely that MAG1 belongs to a novel regulatory network. The DNA microarray data were computationally assembled into self-organizing maps (SOMs) (Tamayo et al., 1999) that group genes with similar expression profiles. Using SOMs, 18 different groups of genes were coregulated across various experimental conditions. Of the genes in the SOM containing MAG1, it was noted that most contained an upstream repressor sequence 2 element (URS2) previously identified upstream of several genes involved in DNA repair and metabolism (Xiao et al., 1993; Singh & Samson, 1995). It was thus hypothesized and resolved that the genes within the MAG1 SOM were regulated at the transcriptional level by the same repressor element and constitute a regulatory network. This study identified Rpn4 as a putative transcriptional regulator of a large number of genes in response to DNA damage. The fact that Rpn4 serves as a 26S proteasome-associated protein as well as a transcriptional factor that regulates at least 26 out of 32 proteasomal genes (Mannhaupt et al., 1999) points to a possibility that ubiquitin-mediated proteolysis may play a role in the transcriptional response to DNA damage. Other studies on the transcriptional response to DNA damage in yeast using DNA microarray experiments have successfully applied similar methods of data management to test certain hypotheses (see Table 1).
DNA damage checkpoint pathways and transcriptional regulation
Both individual gene analyses and global transcriptional studies in budding yeast indicate that many genes are coordinately regulated in response to DNA damage. Although transcriptional responses to specific damaging agents result in distinct and signature profiles, the microarray data point to a general stress response pathway that controls transcription. These observations are consistent with previous reports based on individual gene analyses that DNA damage and replication checkpoints are involved in a transcriptional response to DNA damage.
The DNA damage checkpoint mutants were first isolated by their failure to delay cell-cycle progression into mitosis after irradiation with X-rays (Weinert & Hartwell, 1988; Rowley et al., 1992; Weinert et al., 1994). These mutants displayed increased radiation sensitivity, which could be largely negated through reimposing an artificial arrest before the M phase by an antitubulin agent. Hence, DNA damage checkpoint was initially defined as a nonessential and reversible response that slows down or arrests cell-cycle progression in response to DNA damage, allowing time for DNA repair (Hartwell & Weinert, 1989). At present, about 20 genes in budding yeast have been identified or anticipated to be involved in DNA damage checkpoints (Table 2) (Elledge, 1996; Nyberg et al., 2002; Friedberg et al., 2006). These genes are required for different phases of the cell cycle and for different types of lesions. Because inactivation of damage checkpoint genes affects the transcriptional response to DNA damage as well as other cell fates, it has been well accepted that checkpoint pathways play critical roles in addition to cell-cycle arrest (Zhou & Elledge, 2000).
DNA damage sensors and signal transducers in Saccharomyces cerevisiae and their Schizosaccharomyces pombe and mammalian orthologs
RFC-like clamp loader
WRN, BLM, RTS
BRCA1, MDC1, 53BP1
It is now clear that the checkpoint pathway comprises a subroutine integrated into the larger DNA damage response that regulates multifaceted responses (Zhou & Elledge, 2000). Besides arresting cell-cycle progression, the damage checkpoint pathway has been shown to promote DNA repair (Mills et al., 1999), control telomere length (Ritchie et al., 1999), activate transcription (Elledge, 1996) and trigger apoptosis in metazoan cells (Roos & Kaina, 2006). In a broad sense, the damage checkpoint coordinately regulates DNA damage responses including transcriptional regulation.
Based on the data from previous research, a putative model is proposed for the signal transduction pathways in response to DNA damage, consisting of sensors, transducers (including mediators) and effectors (Bachant & Elledge, 1998). The sensors are proteins that initially sense the damaged DNA and initiate the signaling response. Transducers can be activated by the DNA damage signal passed down from the sensor's and then amplify and relay the signal to the downstream effectors. The ultimate effectors are defined as proteins that execute the cellular response. In the case of transcriptional regulation, the effectors are likely to be transcription factors that recognize cis-acting elements in the promoter of DNA damage-inducible genes.
DNA damage sensors in S. cerevisiae
Because of the complexity of the source and type of DNA damage, as well as the different cell-cycle stages at which damage may occur, DNA damage sensors and their intrinsic interactions have not been well established. Much of our knowledge comes from studies with damage and replication checkpoints.
The budding yeast proliferating cell nuclear antigen (PCNA)-like 9-1-1 complex composed of Rad17, Ddc1 and Mec3 is loaded on the DNA damage site with the assistance of the clamp loader Rad24–Rfc complex, which leads to activation of the main damage response kinase Mec1 (Kondo et al., 1999; Rouse & Jackson, 2002; Majka & Burgers, 2003; Majka et al., 2006). Therefore, Rad24–Rfc and the 9-1-1 complexes together have been speculated to be sensors. Mec1 forms a stable complex with Lcd1/Ddc2, and this complex itself has been anticipated to be a sensor due to the ability of Ddc2 to bind to ends of single- and double-stranded oligonucleotides in vitro and to recruit Mec1 to double-strand break (DSB) sites in vivo in a manner independent of the 9-1-1 complex (Kondo et al., 2001; Melo et al., 2001; Rouse & Jackson, 2002; Niida & Nakanishi, 2006). Another sensor candidate is Rad9. Rad9 is required for the activation of DNA damage checkpoint pathways in budding yeast, is phosphorylated after DNA damage in a Mec1- and Tel1-dependent manner and subsequently interacts with the downstream kinase Rad53 (Naiki et al., 2004). Rad9 displays the ability to associate with DSBs and controls the DNA damage-specific induction of some repair, replication and recombination genes (Aboussekhra et al., 1996; Naiki et al., 2004). Furthermore, Rad9 is thought to act in a pathway distinct from that of Rad24/9-1-1 for damage checkpoint (Lydall & Weinert, 1995) and transcriptional response (de la Torre-Ruiz et al., 1998), suggesting that Rad9 may serve as an alternative DNA damage sensor. The above candidate sensors primarily respond to DNA damage caused by UV, ionizing radiations and endogenous DSBs that induce G1/S and G2/M checkpoints. They also respond to excessive telomere ssDNA in the cdc13-1 mutant (Lydall & Weinert, 1995) and the accumulation of DSBs due to a defective Cdc9 DNA ligase (Weinert & Hartwell, 1993), both inducing the G2/M checkpoint.
The intra-S checkpoint is experimentally activated by modest doses of alkylating agents, such as MMS, that cause replication–blocking lesions. In addition, inhibiting replication by means other than DNA damage, such as treatment with the ribonucleotide reductase (Rnr) inhibitor hydroxyurea, can activate similar cellular responses. Allele-specific mutants in the catalytic subunit of DNA polymerase ɛ (Pol2) are defective in MMS- and hydroxyurea-induced RNR3 (Navas et al., 1995) and MAG1 (Zhu & Xiao, 1998) expression. Hence, Pol2 and its interacting partners Dpb11 and Drc1 may serve as S-phase sensors. It is of great interest to note that Dpb11 has been reported to interact with the Ddc1 subunit of 9-1-1 (Wang & Elledge, 2002), and that physcial interactions between Polɛ and the human homologs of Rad24/9-1-1 were also reported (Makiniemi et al., 2001; Post et al., 2003), indicating that the damage signal may be relayed between sensors. Recently, the helicase Sgs1, which is a member of the E. coli RecQ helicase subfamily that includes mammalian homologs of human BLM, WRN and RTS, responsible for the Bloom, Werner and a subset of Rothmund–Thomson syndromes, respectively, was proposed as a candidate sensor. Sgs1 possesses ATPase activity and preferentially binds to branched DNA substrates (Bennett et al., 1998, 1999). Sgs1 might be involved in the creation of the signal for checkpoint activation, perhaps by resolving aberrantly paired double helices (Cobb et al., 2002). It functions in the same epistasis group as Polɛ to activate Rad53 in the presence of hydroxyurea, and this signaling pathway acts in parallel to that of Rad17 and Rad24 (Navas et al., 1995; Frei & Gasser, 2000). However, it remains unclear whether putative S-phase-specific sensors like Polɛ and Sgs1 recognize a subset of DNA damage or their activities lead to common DNA damage intermediates such as ssDNA (like RecA in the SOS response).
The above checkpoint proteins are placed in the sensor class primarily based on genetic and biochemical properties or inference, as little direct biochemical evidence yet exists to support their sensor role. In principle, checkpoint sensors are capable of directly recognizing DNA lesions or interacting with pre-existing DNA damage repair proteins specific for individual lesions. In this context, DNA repair proteins may serve as damage sensors for both checkpoint and/or transcriptional responses. A good example is the interaction between NER and DNA damage checkpoint in response to UV treatment. A genetic screen for mutants defective in the activation of checkpoint following UV lesions but not other types of damage resulted in the identification of RAD14, which is absolutely required for the UV-induced damage checkpoint. Interestingly, the checkpoint activation requires processing of UV lesions by the NER complex instead of mere lesion recognition by Rad14, and Rad14 physically interacts with Ddc1 (Giannattasio et al., 2004). Similarly, the Mre11/Rad50/Nbs1 (MRN) DSB-binding complex is required for the activation of ATM kinase in mammalian cells (Lee & Paull, 2005), although a similar function has not been reported for the corresponding Mre11/Rad50/Xrs2 (MRX) complex in budding yeast. In both the above cases, processing of UV-induced lesions by NER and DSBs by MRN/MRX results in ssDNA regions that may serve as an ultimate damage signal.
DNA damage transducers in S. cerevisiae
Considerable efforts have been made to understand signal transducers in the damage and replication checkpoint pathways, and the same signal transduction cascade likely functions in the transcriptional response to DNA damage. Most of the transducers are protein kinases that, once activated by the DNA damage signal passed down from the sensors, amplify and relay the signal to the downstream effectors. In S. cerevisiae, checkpoint protein kinases Mec1, Rad53 and Dun1 are necessary for the transcriptional response of most, if not all, genes to DNA damage, and they appear to be central transducers in the regulation network (Zhou & Elledge, 1993; Allen et al., 1994; Kiser & Weinert, 1996; Gasch et al., 2001). The initiator of this signal transduction kinase cascade in budding yeast appears to be Mec1 and its regulatory subunit Lcd1/Ddc2, which is a serine/threonine protein kinase belonging to the phosphatidylinositol-3-kinase (PI3K) family. Mec1 is required for the activation of Rad53 through phosphorylation of consensus PI3K sites within Rad53 (Pellicioli & Foiani, 2005; Ma et al., 2006). However, efficient and direct phosphorylation of Rad53 by Mec1 is only observed in the presence of Rad9 in vitro, and the stimulatory activity of Rad9 requires both phospho- and FHA-dependent interaction with Rad53, which allows Rad53 to be recognized as a substrate for Mec1 (Sweeney et al., 2005). Hence, Rad9 is defined as a checkpoint mediator (or adaptor) that serves to facilitate and amplify signals. Rad9 functions primarily at G1/S and G2/M, while Mrc1 and its mammalian homolog claspin has been proposed to function as an S-phase or a replication checkpoint mediator. Mrc1 and Tof1 form a stable replication-pausing complex at the stalled replication fork, which is required to anchor subsequent DNA repair events (Katou et al., 2003).
Activation of Rad53 is an essential intermediary step in yeast DNA damage responses that include delaying cell-cycle progression, promoting repair processes, stabilizing stalled replication forks and regulating transcription (Branzei & Foiani, 2006). The kinase Dun1 is one of the identified targets of Rad53, and the activation of Dun1 requires phosphorylation by Rad53 (Chen et al., 2007). Dun1 was originally identified as a DNA damage-uninducible (dun) mutant defective in the transcriptional activation of genes encoding Rnr in response to DNA damage (Zhou & Elledge, 1993), and was subsequently shown to be required for the induction of other DNA damage-inducible genes (Basrai et al., 1999; Zhu & Xiao, 2001; Fu & Xiao, 2006). Genomic expression research showed that deletion of DUN1 affected the expression of >1000 genes in response to MMS, and the response in the dun1Δ mutant is largely the same as the response seen in the mec1 mutant (Gasch et al., 2001), suggesting that most of the Mec1-dependent effects on genomic expression are mediated by the downstream Dun1 kinase. The mechanism used by Dun1 for DNA damage induction appears to be rather diverse and requires further investigation.
In response to DNA damage, the primary mechanism for transcriptional regulation is likely to be at the stage of transcription initiation. The downstream effectors are therefore expected to be transcription factors that directly influence transcription initiation. However, unlike damage checkpoint transducers that appear to be aligned in a linear cascade and control most if not all DNA damage-inducible genes, the downstream effectors are clearly diverse and dedicated to a particular gene or a group of genes (Fig. 1). In the next section, selected examples are presented to provide insights into the molecular mechanisms of DNA damage-induced transcriptional response and their interaction with the checkpoint cascade.
A working model of DNA damage-induced transcription in budding yeast. In budding yeast, DNA damage-induced transcription is controlled by signal transduction pathways including sensors, transducers and effectors. Mec1, Rad53 and Dun1 form the central kinase cascade to regulate the DNA damage-induced transcription of most genes examined. The phosphorylation of trans-acting factors (effectors) may change their affinity for the corresponding cis-acting elements in the promoter region of target genes, thus activating their transcription. Only three sets of well-characterized genes are depicted. Solid ovals indicate protein kinases. Dotted lines and question mark indicate that the molecular mechanism of this step(s) remains unknown. Note that the signal-sensor-transducer cascades are based on studies with cell-cycle checkpoints and have not been extensively examined with transcriptional regulation.
Mechanisms of transcriptional regulation of yeast genes in response to DNA damage
Regulation of RNR genes is the best-known example of a eukaryotic transcriptional response to DNA damage to date. Rnr is an enzyme that converts nucleoside diphosphates (NDPs) into deoxynucleoside diphosphates (dNDPs), which represents the rate-limiting step in the production of the four dNTPs for DNA synthesis and repair (Reichard, 1988; Elledge et al., 1993; Jordan & Reichard, 1998). Altered levels or an imbalance of dNTP pools can lead to a higher rate of spontaneous mutagenesis or cell death. Rnr is an α2β2 tetramer, and four genes, RNR1-4, encode the subunits of budding yeast Rnr: RNR1 and RNR3 encode α (large) subunits, while RNR2 and RNR4 encode β (small) subunits. Expression of all four genes is inducible at the transcriptional level by a variety of DNA-damaging agents (Elledge & Davis, 1987, 1989, 1990; Huang & Elledge, 1997), among which RNR3 exhibits the highest level of induction, and its transcriptional regulation has been examined most extensively.
The transcriptional level of the RNR3 gene is very low under normal conditions. However, when treating yeast cells with DNA-damaging agents such as UV, MMS or 4-nitroquinoline-1-oxide (4-NQO), the transcript level of RNR3 can be increased 100–500-fold (Elledge & Davis, 1990). In order to determine the mechanisms of induction, a series of mutants have been isolated that cause constitutive expression of RNR3 (crt mutants) (Zhou & Elledge, 1992). These negative regulators of RNR3 expression are divided into two groups: indirect regulators that result in endogenous DNA damage or a state of metabolic stress such as nucleotide depletion that results in the upregulation of RNR3, and direct regulators involved in the regulatory pathway, including transcription factors. The CRT1, TUP1 (CRT4) and SSN6 (CRT8) genes encode negative regulators that bind to the RNR3 promoter. A second screen was carried out for the mutations that disrupt the ability of DNA damage to induce transcription of RNR3, and the corresponding genes were designated DUN for DNA damage uninducible (Zhou & Elledge, 1993). The nonessential serine/threonine protein kinase Dun1 was isolated through this screen. Genetic analysis of crt1, tup1 and ssn6 showed that these mutations were epistatic to dun1, providing a strong genetic verification that CRT1, TUP1 and SSN6 function downstream of DUN1 (Huang et al., 1998). Combined with observations that the Dun1 upstream kinases Mec1 and Rad53 are also essential for the DNA damage-induced transcription of RNR3 (Huang et al., 1998), the signal transduction pathway for RNR3 becomes conceivable (Fig. 2).
Summary of the transcriptional regulation of RNR3. In response to DNA damage or stalled replication, activation of the Mec1-Rad53-Dun1 kinase cascade results in phosphorylation of the repressor protein Crt1, and phosphorylated Crt1 loses its binding affinity for the RNR3 promoter. With the assistance of Wtm1, Rpd3 and Hos2, the transcription of RNR3 is highly activated. The transcription of CRT1 is negatively regulated by itself and positively regulated by Crt10. Ccr4 influences Crt1 protein abundance by controlling the CRT1 mRNA stability.
It is now clear that in response to DNA damage or replication blocks, the Rad53 protein kinase is activated in a Mec1-dependent manner, and activated Rad53 further phosphorylates the protein kinase Dun1. The Mec1-Rad53-Dun1 kinase cascade culminates in the phosphorylation of Crt1. Crt1 is a DNA-binding protein that recognizes a 13-bp consensus sequence termed the X box. Multiple X box-related sequences of different strength can be found in the promoter region of all four RNR genes. Crt1 is able to recruit the Tup1–Ssn6 general repressor complex to suppress the transcription of RNR genes. In the presence of DNA damage or replication blocks, Crt1 is phosphorylated in a Dun1-dependent manner and this phosphorylation abolishes its ability to bind X boxes, leading to the transcriptional induction of RNR genes (Huang et al., 1998).
Surprisingly, multiple X boxes were also identified in the CRT1 promoter. Data from mutated X boxes showed that they confer CRT1-dependent repression on CRT1 itself (Huang et al., 1998). The expression level of CRT1 is very low under normal growth conditions but is inducible by DNA-damaging agents. Interestingly, the X boxes in the CRT1 promoter consist of one with a weak affinity for Crt1 and one with a strong affinity. Thus, it is speculated that a weak induction of CRT1 occurs immediately following DNA damage, which provides a buffer against spurious transcriptional activation of the pathway. Delaying full activation may ensure a rapid restoration of the basal repressed state when the DNA damage is repaired (Huang et al., 1998), which may serve as an autonomous regulatory circuit (Fig. 2). This mechanism is reminiscent of the autonomous regulation of the LexA repressor in the E. coli SOS response.
Crt1 mediates repression by recruiting the general repressor Tup1–Ssn6 complex to RNR promoters via its N-terminus (Huang et al., 1998; Li & Reese, 2000). Tup1-Ssn6 recruitment establishes a nucleosomal array over the promoter of RNR3 with a positioned nucleosome occupying the TATA box to block access by the general transcriptional machinery (Li & Reese, 2000, 2001; Sharma et al., 2003). The derepression of RNR3 correlates with the disruption of the nucleosome position. In response to DNA damage signals, hyperphosphorylated Crt1 loses the ability to bind to the RNR3 promoter, and the Tup1–Ssn6 complex is not recruited to the promoter region. Consequently, the chromatin structure is remodeled and thus increases the accessibility of DNA to transcription factors. The chromatin remodeling at the RNR3 promoter requires a number of general transcriptional factors, such as TBP-associated factors (TAFIIS), and RNA polymerase II. Furthermore, the remodeling is also dependent on the SWI/SNF complex, which possesses a DNA-stimulated ATPase activity and can destabilize histone–DNA interactions in an ATP-dependent manner (Sharma et al., 2003). The preinitiation complex components TFIID and RNA polymerase II aid in recruitment and retention of the SWI/SNF complex to the RNR3 promoter.
Recently, it was reported that the expression of CRT1 is also controlled by the novel regulator Crt10 (Fu & Xiao, 2006) (Fig. 2). CRT10 was identified through screening of the S. cerevisiae deletion strains for hydroxyurea resistance. Epistatic analysis indicates that CRT10 belongs to the CRT1 pathway. Deletion of CRT10 does not affect the transcript level of TUP1 or SSN6 regardless of hydroxyurea treatment, but significantly reduces the basal level as well as hydroxyurea-induced expression of CRT1; thus, CRT10 appears to be a positive regulator of CRT1 transcription (Fu & Xiao, 2006). Furthermore, the dun1 mutation is epistatic to crt10 with respect to both hydroxyurea sensitivity and RNR gene expression. Interestingly, the expression of CRT10 itself is induced by DNA-damaging agents and this induction requires DUN1 (Fu & Xiao, 2006). The increased Crt10 activity may be required to bring Rnr activity back to a normal level through increasing the expression of repressor Crt1 once the DNA damage is repaired. Like CRT1, the induction of CRT10 itself depends on DUN1, suggesting that Crt10 functions downstream of Dun1 and forms another component of the autoregulatory circuit to regulate the expression of RNR genes (Fu & Xiao, 2006).
CCR4 encodes a component of the major cytoplasmic deadenylase, which is involved in mRNA poly(A) tail shortening (Tucker et al., 2001). Cells defective in CCR4 display particular sensitivity to the Rnr inhibiter hydroxyurea (Woolstencroft et al., 2006). The ccr4 dun1 double mutants exhibit synergistic sensitivity to hydroxyurea, and simultaneous overexpression of RNR2, RNR3 and RNR4 partially rescues the hydroxyurea hypersensitivity of a ccr4 dun1 strain, implying that CCR4 and DUN1 function in different pathways to regulate the activity of Rnr. Deletion of CRT1 suppresses hydroxyurea sensitivity of the ccr4 mutant, and overexpression of CRT1 hypersensitizes ccr4 to hydroxyurea. These observations lead to the conclusion that Ccr4 regulates CRT1 mRNA poly(A) tail length and may thus subtly influence the Crt1 protein abundance (Woolstencroft et al., 2006) (Fig. 2).
More regulatory factors were found to be involved in the transcriptional regulation of RNR genes. Wtm1 and Wtm2 have been reported to modulate the expression of RNR3 (Tringe et al., 2006). Moderate overexpression of both genes or high-level expression of WTM2 alone upregulates RNR3-lacZ in the absence of DNA damage. In response to hydroxyurea and γ-rays, the expression level of RNR3 attenuated 45% in wtm2Δ mutants, but not in wtm1 mutants. Wtm2 was found to associate directly with the RNR3 promoter, and the association correlates with its ability to increase constitutive RNR3 expression. It remains unknown how Wtm2 increases RNR3 transcription, although some observations hint that Wtm2 might enhance RNR3 transcription by participating in chromatin remodeling (Tringe et al., 2006). Furthermore, Wtm1 associates with and anchors the Rnr2/Rnr4 complex in the nucleus in untreated cells. In response to DNA damage or replication inhibition, the interaction between Wtm1 and Rnr2/Rnr4 is disrupted and this complex is released from the nucleus to the cytoplasm (Lee & Elledge, 2006; Zhang et al., 2006). A recent report shows that two histone deacetylases (HDAC), Rpd3 and Hos2, are required for the transcriptional activation of RNR3 in response to DNA damage (Sharma et al., 2007) (Fig. 2). Although a direct association between Hos2 and the RNR3 promoter has not been observed, Rpd3 was found to specifically bind to X boxes in the RNR3 promoter region in a Tup1- or a Crt1-independent manner. The HDAC activity of Rpd3 and Hos2 is essential for the activation of RNR3. The observation that the recruitment of RNA polymerase II is dramatically reduced in the rpd3 hos2 mutant indicates that Rpd3 and Hos2 activate the transcription by regulating the assembly of the preinitiation complex or inducing multiple rounds of RNA polymerase recruitment (Sharma et al., 2007).
PHR1 encodes a photolyase that specifically and exclusively repairs pyrimidine dimers, which are the most abundant lesions found in DNA following UV irradiation (Sancar, 1985). Interestingly, the transcription of PHR1 is induced by various DNA-damaging agents such as MMS, MNNG, UV irradiation, 4-NQO or γ-ray (Robinson et al., 1986; Sebastian et al., 1990) despite the fact that Phr1 only reverses UV-induced pyrimidine dimers but not lesions induced by other DNA-damaging agents. Three transcriptional regulatory elements have been defined within the PHR1 promoter region: an upstream activation sequence (UAS), an upstream repression sequence (URS) and an upstream essential sequence (UES) (Sancar et al., 1995). A 22-bp interrupted palindrome comprises UASPHRI, and it is responsible for 80–90% of basal and induced expression. It alone can activate transcription of a CYC1 minimal promoter but does not confer damage responsiveness (Sancar et al., 1995). URSPHR1 is defined to a 39-bp region that includes a 22-bp palindrome. Deletions or specific mutations within URSPHR1 increase basal-level expression but decrease the induction ratio. It functions as a strong URS and confers a low-level damage inducibility when placed in the context of a heterologous gene (Sancar et al., 1995). The UESPHR1 is required for efficient derepression when URSPHR1 is present. Deletion of URSPHR1 also eliminates the requirement for UESPHR1 for transcriptional activation (Sancar et al., 1995).
Three proteins have been identified that regulate the expression of PHR1 by binding to the upstream regulatory elements. Ume6 is a bifunctional transcriptional regulator involved in several metabolic pathways (Strich et al., 1994). It is a positive regulator of PHR1 transcription and binds specifically to the UASPHR1 (Sweet et al., 1997). Multiple copies of Ume6 enhance the expression of PHR1; however, deletion of UME6 reduces the expression of PHR1 during vegetative growth, but only at a distinct cell-cycle phase (Sweet et al., 1997; Sancar et al., 2000). Rph1 and Gis1, which are 35% identical to each other at the amino acid sequence level, are two DNA damage-responsive repressors of PHR1 transcription (Jang et al., 1999; Sancar et al., 2000). Both Rph1 and Gis1 contain two putative zinc fingers that are >90% identical overall and identical in the DNA-binding loop, and they regulate the transcriptional response of PHR1 through binding to URSPHR1. Deletion of both RPH1 and GIS1 is required to fully derepress PHR1 in the absence of damage, suggesting that they are functionally redundant. In vitro footprinting and binding competition studies indicate that the sequence AG4 (C4T) within the URSPHR1 is the binding site for Rph1 and Gis1 (Jang et al., 1999).
Induction of PHR1 is controlled by the DNA damage signal transduction pathway. The serine and threonine residues of Rph1 can be phosphorylated, and the phosphorylation of Rph1 is increased in response to DNA damage. The DNA damage-induced Rph1 phosphorylation requires DNA damage checkpoint proteins Rad9, Rad17, Mec1 and Rad53, indicating that the phosphorylation of Rph1 is under the control of the Mec1-Rad53 DNA damage checkpoint pathway (Kim et al., 2002). On the other hand, deletion of DUN1, TEL1 or CHK1 does not affect the phosphorylation of Rph1, indicating that PHR1 is regulated by a potentially novel damage checkpoint that is distinct from the Mec1-Rad53-Dun1 protein kinase cascade. Based on the results of a coimmunoprecipitation assay, Rad53 does not appear to interact physically with Rph1, indicating the existence of a yet unknown kinase(s) in the Mec1-Rad53-Rph1 pathway (Kim et al., 2002).
MAG1 and DDI1
MAG1 encodes a 3-methyladenine DNA glycosylase that initiates the base excision repair pathway by removing lethal lesions such as 3-methyladenine (Chen et al., 1989). MAG1 is not only induced by DNA-alkylating agents such as MMS but also by UV, 4-NQO or hydroxyurea (Chen et al., 1990; Chen & Samson, 1991; Xiao et al., 1993). By analyzing the DNA sequence immediately upstream of MAG1, another DNA damage-inducible gene, named DDI1 for DNA Damage Inducible, was identified (Liu & Xiao, 1997). Similar to MAG1, DDI1 is also induced by MMS, UV, 4-NQO and hydroxyurea. Furthermore, both genes require a similar dosage for maximum induction, and the induction profile is similar (Liu & Xiao, 1997).
MAG1 and DDI1 lie in a head-to-head configuration and are transcribed divergently. The expression of both MAG1 and DDI1 is controlled by common as well as distinct UAS and URS elements, possibly through antagonistic mechanisms (Liu & Xiao, 1997). A UASMAG1 and a URSMAG1 have been identified in the promoter region of MAG1 (Xiao et al., 1993). Interestingly, fine mapping of the UASMAG1 sequence reveals that it is located within the DDI1 protein coding region, and the electrophoretic mobility shift assay using labeled UASMAG1 as a probe detected sequence-specific binding protein(s) (Liu & Xiao, 1997), although the identity of the UASMAG1-binding protein has not been reported. The expression of DDI1 is negatively regulated by a URSDDI1 in its promoter region (Liu & Xiao, 1997). The intergenic region between MAG1 and DDI1 also contains a cis-acting element that coregulates the expression of both genes. In this shared promoter region, UASDM, which contains two 8-bp tandem repeat sequences 5′-GGTGGCGA-3′, is required for the bidirectional expression of MAG1 and DDI1; deletion or point mutations of this tandem repeat result in a reduced basal level expression and significant reduction of damage-induced expression (Liu & Xiao, 1997). Furthermore, UASDM alone is able to confer a DNA damage response when fused to a CYC1 promoter (Liu & Xiao, 1997), indicating that this cis-acting element independently interacts with a transcriptional factor(s). With a yeast one-hybrid screen using UASDM as a bait, a transcriptional activator called Pdr3 was isolated (Zhu & Xiao, 2004). Pdr3 binds to UASDMin vivo and in vitro, and deletion of PDR3 reduced both the basal level and the DNA damage-induced expression of MAG1 and DDI1. In addition, deletion of PDR3 does not further affect MAG1 and DDI1 expression if UASDM is deleted, indicating that UASDM is indeed the target for Pdr3 activation (Zhu & Xiao, 2004). Another transcriptional activator, Rpn4, was shown to be required for MAG1 (Jelinsky et al., 2000) and DDI1 (Zhu & Xiao, 2004) expression; however, Rpn4 does not appear to bind UASDM. Moreover, deletion of RPN4 does not alter MAG1 and DDI1 expression in the pdr3 mutant cells, suggesting that RPN4 acts upstream of PDR3 (Zhu & Xiao, 2004). Deletion of PDR3 and RPN4 has no effect on the basal level or the DNA damage-induced expression of PHR1, RNR2 or RNR3. Meanwhile, Crt1 and Tup1/Ssn6, the repressors of the RNR genes, and Rph1 and Gis1, the repressors of PHR1, are not involved in the control of MAG1 expression (Zhu & Xiao, 2001). Hence, it becomes apparent that all three sets of well-studied yeast damage-inducible genes (RNR, PHR1 and MAG1-DDI1) have distinct regulators and that the regulatory mechanisms are also different from each other.
The expression of MAG1 and DDI1 is also controlled by the DNA damage checkpoint. Mutation of POL2, MEC1 or DUN1 reduces the DNA damage response of MAG1 (Zhu & Xiao, 1998, 2001), suggesting that MAG1 is regulated by the POL2-MEC1-RAD53-DUN1 checkpoint pathway. In contrast, DDI1 remains inducible in sad1-1 (rad53), dun1 or mec1 mutants, but its induction is diminished in pds1 sad1-1 (rad53) or pds1 dun1 double mutants (Zhu & Xiao, 2001). PDS1 encodes an anaphase inhibitor and functions downstream of CHK1; CHK1-PDS1 and RAD53-DUN1 may form two parallel branches in the DNA damage checkpoints (Fig. 1) (Gardner et al., 1999; Schollaert et al., 2004). This suggests that the CHK1-PDS1 and MEC1-RAD53-DUN1 checkpoint pathways may function redundantly in the control of DDI1 expression (Zhu & Xiao, 2001).
A putative eukaryotic SOS response in budding yeast?
Although cell-cycle checkpoints have been regarded as a eukaryotic SOS response, their similarity to the bacterial SOS response is limited. In budding yeast, the Rad51 protein is a sequence homolog of E. coli RecA and possesses ssDNA-binding and ATPase activities (Shinohara et al., 1992). However, the only Rad51 enzymatic activity reported to date is ATP- and ssDNA-dependent recombinase, and it is involved in homologous recombination and homology-mediated DNA repair and damage tolerance pathways (Sung et al., 1994; Paques & Haber, 1999; Gangavarapu et al., 2007). Although RAD51 itself is DNA damage inducible (Shinohara et al., 1992), deletion of RAD51 does not affect the expression of the DNA damage-inducible genes examined (Y. Fu & W. Xiao, unpublished results). Nevertheless, it remains possible that there is a functional homolog of RecA in yeast with respect to a transcriptional response but it shares no sequence homology and enzymatic activity with RecA.
In a broad sense, E. coli RecA controls three important cellular responses to DNA damage, namely homologous recombination via RecBCD and RecFOR, TLS by working with PolIV and PolV and the SOS response, which collectively provides a survival mechanism when cells encounter replication blocks and/or contain excessive ssDNA. The above functions are reminiscent of the Rad6–Rad18 complex in budding yeast. Rad6–Rad18 forms a stable complex that is essential for the DNA postreplication repair (PRR) pathway (Broomfield et al., 2001). The Rad6–Rad18 complex possesses ssDNA-binding and ATPase activities (Bailly et al., 1997) and assumes most if not all RecA functions to coordinate broad cellular responses to DNA damage. Firstly, Rad6–Rad18 as an E2–E3 ubiquitination complex mono-ubiquitinates PCNA to promote Polζ- and Polη-mediated TLS (Hoege et al., 2002; Stelter & Ulrich, 2003). Secondly, this mono-Ub-PCNA is required for PCNA poly-ubiquitination via a Lys63 chain linkage for an error-free mode of DNA damage tolerance (Hoege et al., 2002) reminiscent of the RecFOR activity in E. coli (Chow & Courcelle, 2004). Finally, although Rad6–Rad18 does not have homologous recombinase activity like RecA, it may compete with the Ubc9–Siz1 complex that sumorylates PCNA at the same K164 residue; sumorylated PCNA recruits the DNA helicase Srs2, which inhibits the recombinase activity of the yeast RecA homolog Rad51 (Papouli et al., 2005; Pfander et al., 2005). Hence, it would be of great interest to investigate whether the Rad6–Rad18 complex also plays a role in an SOS response and, if so, how it interacts with the damage checkpoints.
In summary, transcriptional response to DNA damage is an important process for cell survival of microorganisms under such life-threatening stress. In bacteria such as E. coli, the DNA damage-induced transcriptional regulation is centrally controlled by an SOS response. All SOS-inducible genes contain SOS boxes in their operator regions, and their expression is repressed by the interaction between the transcriptional repressor LexA and SOS boxes under noninduced conditions. The DNA damage induction is achieved through enhanced self proteolysis of LexA assisted by ssDNA-activated RecA. In addition, bacteria also possess regulatory mechanisms in response to specific DNA damage.
DNA damage-induced transcriptional response in eukaryotic microorganisms is apparently more complicated and less understood. Nevertheless, several conclusions can be drawn from data analyses as presented in this review. Firstly, the idea that eukaryotic microorganisms possess an SOS response structurally conserved from bacteria has been effectively ruled out. Secondly, in budding yeast, a large number of yeast genes (more than 5%) are induced after DNA damage. Almost all DNA damage-induced genes individually investigated to date respond to a broad range of DNA-damaging agents regardless of whether the gene function provides a survival value to the particular damage (Birrell et al., 2002). This observation supports the hypothesis that there may exist a coordinated signal transduction pathway to regulate transcriptional response to DNA damage. Thirdly, DNA damage checkpoint pathways variably affect transcriptional response to DNA damage in budding yeast and have been regarded as a functional analog of the bacterial SOS response. Fourthly, unlike the SOS-inducible genes in bacteria, few DNA damage-inducible genes in budding yeast share a common promoter sequence, suggesting that the downstream effectors are diversified. This is consistent with reports that all three well-characterized sets of damage-inducible genes in budding yeast are controlled by different transcriptional regulators and furthermore, their modes of regulation are different: while RNRs and PHR1 are induced by derepression, MAG1 and DDI1 are induced by activation. Fifthly, the search for a novel mechanism(s) in eukaryotic microorganisms that may better resume the RecA-mediated SOS response is underway and may result in exciting insights into the eukaryotic SOS response. Finally, although transcriptional response to DNA damage is a convenient and useful indicator for the gene function, transcriptional levels do not necessarily reflect cellular protein levels, and this appears to be the case for DNA damage response in budding yeast (Lee et al., 2007). Hence, other posttranscriptional regulatory mechanisms have to be taken into account when examining a particular gene function and its product activity.
The authors wish to thank the laboratory members for helpful discussion and Michelle Hanna for proofreading the manuscript. This work is supported by the Natural Sciences and Engineering Research Council of Canada discovery grant OGP0138338 to W.X. Y.F. is a recipient of the Arthur Smyth Memorial Scholarship from College of Medicine, University of Saskatchewan.
(1998) Regulatory networks that control DNA damage-inducible gene in Saccharomyces cerevisiae. DNA Damage and Repair, Vol. 1: DNA Repair in Prokaryotes and Lower Eukaryotes (NickoloffJA, HoekstraMF, eds), pp. 383–410. Humana Press, Totowa, NJ.
(1987) Identification and isolation of the gene encoding the small subunit of ribonucleotide reductase from Saccharomyces cerevisiae: DNA damage-inducible gene required for mitotic viability. Mol Cell Biol 7: 2783–2793.
(1995) Transcriptional activation of the Escherichia coli adaptive response gene aidB is mediated by binding of methylated Ada protein. Evidence for a new consensus sequence for Ada-binding sites. J Biol Chem 270: 8285–8289.
(2006) The Pep4p vacuolar proteinase contributes to the turnover of oxidized proteins but PEP4 overexpression is not sufficient to increase chronological lifespan in Saccharomyces cerevisiae. Microbiology 152: 3595–3605.
(1997) The Saccharomyces cerevisiae RAD30 gene, a homologue of Escherichia coli dinB and umuC, is DNA damage inducible and functions in a novel error-free postreplication repair mechanism. Genetics 147: 1557–1568.
(2003) The human checkpoint Rad protein Rad17 is chromatin-associated throughout the cell cycle, localizes to DNA replication sites, and interacts with DNA polymerase ɛ. Nucleic Acids Res 31: 5568–5575.