2023 Volume 98 Issue 3 Pages 103-119
Organisms have evolved elaborate mechanisms that maintain genome stability. Deficiencies in these mechanisms result in changes to the nucleotide sequence as well as copy number and structural variations in the genome. Genome instability has been implicated in numerous human diseases. However, genomic alterations can also be beneficial as they are an essential part of the evolutionary process. Organisms sometimes program genomic changes that drive genetic and phenotypic diversity. Therefore, genome alterations can have both positive and negative impacts on cellular growth and functions, which underscores the need to control the processes that restrict or induce such changes to the genome. The ribosomal RNA gene (rDNA) is highly abundant in eukaryotic genomes, forming a cluster where numerous rDNA copies are tandemly arrayed. Budding yeast can alter the stability of its rDNA cluster by changing the rDNA copy number within the cluster or by producing extrachromosomal rDNA circles. Here, we review the mechanisms that regulate the stability of the budding yeast rDNA cluster during repair of DNA double-strand breaks that are formed in response to programmed DNA replication fork arrest.
The genome contains all the information an organism requires to function. Therefore, organisms have evolved elaborate mechanisms to maintain genome stability (Hoeijmakers, 2001). These mechanisms include faithful DNA replication during S phase of the cell cycle, accurate chromosome segregation during M phase, repair of DNA damage, and checkpoints that delay the cell cycle according to the cellular status of DNA damage/repair. A deficiency in these DNA maintenance processes may induce genome alterations from small-scale changes, such as variations in the nucleotide sequence, up to large-scale changes, such as differences in copy number and overall genome structure. These genomic changes have been implicated in numerous human diseases (Zhang et al., 2009; Sasaki et al., 2010).
Although the changes to the genome are often considered to have a negative impact on the cell, they also act as the driving force of genetic diversity, phenotypic diversity and evolution. In sexually reproducing organisms, meiosis generates gametes by reducing the genomic content by half, which is achieved by conducting one round of DNA replication followed by two consecutive rounds of chromosome segregation (reviewed by Lam and Keeney, 2014). During meiosis, numerous DNA double-strand breaks (DSBs) are introduced in the genome that promote homologous recombination between homologous chromosomes. Meiotic recombination results in shuffling of genetic information between the parental chromosomes, driving genetic diversity among haploid gametes. Furthermore, antibody- and immunoglobulin-producing B and T cells induce genomic alterations at the immunoglobulin gene loci by carrying out V(D)J recombination, class switch recombination and somatic hypermutation (reviewed by Chi et al., 2020). Through these processes, cells generate a highly diverse repertoire of immunoglobulins and T cell receptors that facilitate an effective defense against numerous pathogens and viruses. In this way, genomic alterations can be programmed to have beneficial influences on cellular integrity.
Because genomic alterations can elicit both positive and negative impacts on the organism, it is critical to properly control the processes that lead to such changes. This review focuses on the ribosomal RNA gene (rDNA) locus in budding yeast that undergoes genomic alterations in response to programmed DNA replication fork arrest and DSB formation. We first present an overview of the rDNA sequence and its unique organization in the genome. Because an essential function of rDNA is to transcribe ribosomal RNA (rRNA), we briefly summarize the transcriptional regulation of rRNA. We then review insights into mechanisms that regulate the stability of the rDNA locus. Although this review primarily focuses on the budding yeast rDNA locus, we occasionally refer to studies involving other organisms. Finally, we give a perspective on the unresolved questions that need to be addressed in future studies.
In all living organisms, proteins are essential for cellular viability. Therefore, substantial cellular resources are devoted to the biogenesis of the protein synthesis machinery, the ribosome. Ribosomes are ribonucleoproteins comprising rRNAs and ribosomal proteins (Fig. 1A, 1B). The biosynthesis of rRNAs and ribosome biogenesis are carried out inside a specialized membraneless nuclear compartment known as the nucleolus (Fig. 1A), through a complex process that is beyond the scope of this review. Only a brief summary of ribosome biogenesis is given, but readers are encouraged to refer to other reviews available in the literature (Warner, 1989; Gourse et al., 1996; Paul et al., 2004; Bassler and Hurt, 2019). Initially, cells produce individual components of the ribosome. A long precursor rRNA (35S pre-rRNA in yeast and 45S pre-rRNA in mammals) is transcribed from the rDNA by RNA polymerase I, and is then processed into mature 18S, 5.8S and 25S (28S in mammals) rRNA (Fig. 1B). 5S rRNA is transcribed from the 5S rRNA gene by RNA polymerase III. In addition, approximately 80 ribosomal proteins are produced via transcription of mRNAs from their respective genes by RNA polymerase II and subsequent translation by pre-existing ribosomes. These components are assembled into two ribosomal subunits, a small 40S subunit (18S rRNA and 33 ribosomal proteins) and a large 60S subunit (25S, 5.8S and 5S rRNA, and 46 ribosomal proteins in budding yeast or ~50 in mammals), with the assistance of ribosome assembly factors and small nucleolar RNAs (Fig. 1B).
rRNA transcription and ribosome assembly in the nucleolus of the budding yeast cell. (A) The budding yeast rDNA cluster (green) is localized inside the nucleolus. (B) rRNA transcription and ribosome assembly. A long precursor RNA is transcribed by RNA polymerase I and processed into 18S, 5.8S and 25S rRNA. 5S rRNA is transcribed by RNA polymerase III. 18S rRNA forms part of the 40S ribosome subunit and 5.8S, 25S and 5S rRNA form part of the 60S ribosome subunit. These are transported as ribonucleoproteins out of the nucleolus into the cytoplasm where the ribosome is assembled. (C) Christmas tree structure of a transcriptionally active rDNA copy. The nucleolar spread (also known as a Miller spread) was prepared from a mouse cell and detected by electron microscopy. The transcriptionally active rDNA copies resemble a Christmas tree, in which the rDNA coding sequence forms the trunk of the tree, which is packed with branches of increasing length of nascent precursor rRNAs with a terminal knob structure at the end of each transcript. Scale bar, 0.5 µm. The image is reprinted from Scheer and Benavente (1990) (copyright permission of Wiley-Liss). (D) Miller spread of three consecutive rDNA copies. The Miller spread was prepared from oocytes of the amphibian Triturus alpestris. The rDNA copies numbered 2 and 3 are being transcribed, while copy number 1 is devoid of transcripts. Scale bar, 1 µm. The image is reprinted from Scheer et al. (1976) (copyright permission of Rockefeller University Press).
Ribosome biogenesis is the most energy-consuming process inside the cell. Therefore, ribosome components need to be synthesized in a coordinated manner in accordance with cellular proliferation status, metabolic activity and environmental signals. Ribosome biogenesis is restricted under conditions of stress such as nutrient or amino acid starvation, or in non-dividing cells, to save energy (Conconi et al., 1989; Sandmeier et al., 2002). In contrast, rapidly growing cells demand a high-level constant supply of ribosomes (Warner, 1999; von der Haar, 2008). Defects in ribosome biosynthesis can slow down or even cease cellular growth (Pederson, 2011). Furthermore, defects in ribosome biogenesis that do not cause cell death can lead to diseases referred to as ribosomopathies, such as Treacher Collins syndrome and Diamond–Blackfan anemia (Aspesi and Ellis, 2019; Farley-Barnes et al., 2019), or can predispose cells to tumorigenesis (Ruggero and Pandolfi, 2003).
The nucleolus, where rRNA transcription and ribosome assembly take place, was first observed under the microscope almost 200 years ago (reviewed by Raska et al., 2004; Boisvert et al., 2007; Pederson, 2011; Lafontaine et al., 2021). Budding yeast normally forms a single, crescent-shaped nucleolus that is associated with the nuclear envelope and that occupies about one-third of the nuclear volume (Fig. 1A). rRNA transcription is inhibited during anaphase by the Cdc14 phosphatase, which is critical for condensin binding to rDNA and disjunction of rDNA-bearing chr XII (Torres-Rosell et al., 2007; Clemente-Blanco et al., 2009). In contrast, human cells contain several nucleoli (i.e., ~2–5 in HeLa cells on average) and rRNA synthesis is cell cycle-regulated, reaching its peak during S and G2 and being suppressed during metaphase of mitosis (Klein and Grummt, 1999; Lafontaine et al., 2021). Accordingly, the nucleoli are assembled in G1 and disassembled when cells enter mitosis in human cells (Boisvert et al., 2007). The nucleolus has recently received considerable attention due to its biophysical properties as a biomolecular condensate involving assembly by liquid–liquid phase separation (Lafontaine et al., 2021). In addition to rDNA transcription and ribosome biogenesis, the nucleolus also performs a host of other diverse functions such as cell cycle regulation, sequestration of specific proteins, regulation of mitosis, and stress sensing (Boisvert et al., 2007; Pederson, 2011).
rDNA is abundant in eukaryotic genomesWhile ribosomal proteins are encoded by either a single-copy gene or duplicated genes, the copy number of the rDNA that encodes rRNA transcription units is much higher in eukaryotic genomes; it varies substantially among different organisms, ranging from less than 100 to more than 10,000 (Long and Dawid, 1980). The budding yeast genome normally contains ~150 copies of ~9.1-kb rDNA sequence, which are tandemly arrayed at a single locus on chr XII (Fig. 2A) (Petes and Botstein, 1977; Petes, 1979; Johnston et al., 1997; Kobayashi et al., 1998). Each rDNA copy contains ~6.9-kb large precursor rRNA and ~120-bp 5S rRNA transcription units separated by two intergenic spacers (IGSs). IGS1 contains a replication fork barrier (RFB) sequence that blocks the replication fork from entering the 35S rDNA from its 3′ end and a bidirectional, RNA polymerase II-dependent promoter, E-pro, that synthesizes non-coding RNA (Santangelo et al., 1988; Kobayashi and Ganley, 2005). IGS2 contains a precursor rRNA promoter, an origin of DNA replication (ribosomal autonomous replicating sequence, rARS) and a cohesin-associated region. As will be discussed below, these sequences are important regulatory elements not only for rDNA transcription but also for maintaining the stability and variability of rDNA copy number.
Organization of the rDNA cluster(s) in the budding yeast and human genomes. (A) The budding yeast rDNA cluster on chr XII. The budding yeast rDNA region as indicated in green contains ~150 copies of rDNA sequence, which are tandemly arrayed as a single cluster. Each rDNA copy contains 35S and 5S rDNA coding units, which are separated by two intergenic spacers (IGSs). IGS1 contains a replication fork barrier (RFB) and an RNA polymerase II-dependent, bi-directional promoter (E-pro). IGS2 contains an origin of DNA replication (rARS) and cohesin-associated region. (B) Human rDNA clusters. The human long precursor rRNA coding unit, 45S rDNA, is tandemly arrayed and forms clusters on the short arms of the acrocentric chromosomes, chr 13, 14, 15, 21 and 22, as indicated in green. Note that the 5S rDNA forms a separate cluster on chr 1, which is not indicated. (C) Human rDNA sequence. The 43-kb human rDNA consists of a ~13-kb precursor rRNA coding unit and ~31-kb IGS, which contains several repetitive sequences. The 2–3 copies of the R repeat are located downstream of the coding sequence and contain multiple Sal boxes (red lines). There are also 0–4 copies of Butterfly/Long repeats.
The large precursor rDNA and 5S rDNA are not linked in most eukaryotic genomes (Long and Dawid, 1980; Garcia et al., 2010; Wicke et al., 2011). The 45S rDNA is repeated ~100–600 times in the diploid human genome and forms ten clusters on the short arms of the acrocentric chromosomes, chr 13, 14, 15, 21 and 22 (Fig. 2B), each of which forms a nucleolar organizer region (Prokopowich et al., 2003; Stults et al., 2008; Gibbons et al., 2015; Parks et al., 2018). The 5S rDNA forms a separate, single cluster on chr 1 and its copy number varies within a range similar to that of 45S rDNA (Gibbons et al., 2015). The large rDNA sequence in human is ~43 kb in length and consists of a ~13-kb precursor rRNA coding region and ~31-kb IGS (Fig. 2C) (Gonzalez and Sylvester, 1995). The IGS is highly complex and contains numerous repetitive sequences (Gonzalez and Sylvester, 1995; Agrawal and Ganley, 2018; Smirnov et al., 2021). The R repeat is located downstream of the coding sequence and contains multiple Sal boxes that act not only as the transcription terminator sequence bound by TTF-I but also as the replication fork blocking site (Grummt et al., 1985; Kuhn et al., 1990; Gerber et al., 1997; Akamatsu and Kobayashi, 2015). The IGS contains the Butterfly/Long repeats and other repetitive sequences, although their functions remain unknown (Gonzalez and Sylvester, 1995; Agrawal and Ganley, 2018).
It is often assumed that a high rDNA copy number is necessary for rRNA production to sustain cellular growth. Indeed, when the rDNA copy number is reduced from the normal average of ~150 to two copies in budding yeast, cells do not survive unless rRNA is supplied by other means (i.e., cells harboring a multicopy plasmid expressing rRNA) (Kobayashi et al., 2001). However, the rate of rRNA synthesis and cellular growth are comparable between strains with ~20 rDNA copies and strains with ~150 rDNA copies in unperturbed growth conditions (Kobayashi et al., 2001; French et al., 2003; Ide et al., 2010; Iida and Kobayashi, 2019b). The situation is similar in plants. For example, in Arabidopsis thaliana, rDNA copy number can be reduced to ~10% of the normal average using CRISPR-Cas9-mediated genome editing without affecting rRNA transcription rates (Lopez et al., 2021). Thus, provided there are a minimum number of rDNA copies, cells can produce enough rRNAs for normal cellular functions. Budding yeast cells show increased sensitivity to DNA damage as the rDNA copy number decreases (Ide et al., 2010). Furthermore, rDNA low-copy strains show enhanced sister chromatid separation, reduced condensin association, and accumulation of recombination intermediates in the rDNA region, compared to the normal-copy strain. Therefore, one reason for the presence of the excess copies is to provide the intact template for sister chromatid recombination upon DNA damage. In addition to this function, organisms may have other reasons to possess a vast excess of rDNA, which remain unexplored.
Electron microscopic analyses of nucleolar chromatin, using the technique referred to as Miller spreads, allow direct visualization of the transcription of rDNA coding units by RNA polymerase I (Miller and Beatty, 1969a, 1969b). Early EM analyses were performed on rDNA prepared from amphibian oocytes. The transcriptionally active rDNA resembles a Christmas tree, in which the rDNA coding sequence comprising the trunk of the tree is packed with branches of increasing length of nascent precursor rRNAs (Fig. 1C, 1D). These observations suggest that numerous rRNA molecules are simultaneously synthesized on the actively transcribed rDNA copies. However, rDNA copies that are devoid of nascent transcripts have also been observed (Fig. 1D) (McKnight and Miller, 1976; Scheer et al., 1976; Trendelenburg and McKinnell, 1979; Morgan et al., 1983). These findings once led to the proposal that rDNA copies are a binary unit, and are either “on” or “off” in terms of rDNA transcription (Grummt and Pikaard, 2003; Raska et al., 2004; Russell and Zomerdijk, 2005).
Chromatin structures of rDNA copies can be analyzed by treatment of cells with psoralen, which intercalates between the two strands of double-stranded DNA upon exposure to UV light, followed by monitoring the mobility of DNA fragments by agarose gel electrophoresis and Southern blotting (Conconi et al., 1989). Psoralen crosslinking studies in yeast, mouse and human have demonstrated the existence of two classes of rDNA: (i) psoralen-inaccessible copies of ‘closed’ chromatin conformation and (ii) psoralen-accessible copies of ‘open’ chromatin conformation (Conconi et al., 1989; Dammann et al., 1993, 1995). In budding yeast cells with a normal number of rDNA copies (~150), approximately half of these copies are considered open (Dammann et al., 1993; Ide et al., 2010). However, the proportion of open copies is reduced to ~25% when cells enter stationary phase and rRNA synthesis levels are reduced (Sandmeier et al., 2002). When rDNA copy numbers are decreased, rRNA synthesis rates are unaltered but the proportion of psoralen-accessible copies relative to psoralen-inaccessible copies is increased (Ide et al., 2010). In bone marrow cells isolated from 4-week-old mice, ~40% of rDNA copies are psoralen-accessible, but this is reduced by ~20% in bone marrow cells from 2-year-old mice (Watada et al., 2020). Therefore, the ratio of rDNA copies of open to closed chromatin structures can be influenced by various factors such as the rDNA copy number, growth conditions and the aging process.
Do rDNA copies of psoralen-accessible (open) and psoralen-inaccessible (closed) chromatin structures represent transcriptionally active and inactive copies, respectively? French et al. (2003) used EM analysis to show that strains carrying ~40 and ~140 rDNA copies show similar levels of rRNA synthesis, but the average number of rDNA copies loaded with RNA polymerase I is ~40 in the ~40-copy strain and ~75 in the ~140-copy strain. This finding challenges the binary switch model, which would predict that a similar number of rDNA copies should undergo transcription to sustain an equivalent level of rRNA. Closer examination of the active genes showed that the rDNA low-copy strains increase the rate of transcription from the active copies by loading more RNA polymerase molecules per gene, compared to the rDNA higher-copy strain. In human cells, rRNA synthesis levels are lowered by ~6-fold during stationary phase and metaphase compared to exponentially growing cells, without changing the ratios of psoralen-accessible to psoralen-inaccessible copies (Conconi et al., 1989). Therefore, although psoralen crosslinking studies can distinguish between open and closed chromatin structures, it is difficult to definitively demonstrate that open rDNA copies are always transcribed or that closed rDNA copies are never transcribed (Banditt et al., 1999; French et al., 2003). Nonetheless, cells modulate rRNA synthesis by regulating the rate of transcription at individual active genes.
Homologous recombination (HR) is a well-known pathway that repairs DSBs by copying sequence information from a homologous donor sequence to restore genetic information that is lost at the break site (Fig. 3) (reviewed by Pâques and Haber, 1999; Symington et al., 2014). Because this process requires the presence of homologous template that is normally the identical sequence on the sister chromatid, HR is used for DSB repair in S phase after completion of DNA replication and G2/M phase of the mitotic cell cycle (Fig. 3). During meiosis, DSBs are introduced in the genome by the topoisomerase-like Spo11 protein and these DSBs are repaired by HR, preferentially using a template on homologous chromosomes (Game et al., 1989; Sun et al., 1989; Cao et al., 1990; Keeney et al., 1997; Lam and Keeney, 2014). HR is often considered to be an error-free DSB repair pathway. However, eukaryotic genomes are replete with repeated sequences, including low-copy repeats such as duplicated genes as well as high-copy repetitive elements such as rDNA, subtelomeric repeats, long interspersed element-1 and Alu elements. When DSBs are formed in or near these repeated sequences, they can be repaired using a homologous sequence present elsewhere at non-allelic positions in the genome, and such HR events can markedly alter genome architecture by generating chromosome rearrangements (Sasaki et al., 2010). These chromosome rearrangements have been implicated in numerous human genomic disorders. Thus, organisms must prevent non-allelic HR, for example by suppressing the formation of DSBs in or near repeated sequences in the first place and by promoting the use of the homologous template at the allelic position.
The structure of one-ended and two-ended DSBs. When replication forks are arrested at the RFB site in the rDNA, a single-stranded DNA (ssDNA) nick is formed on the leading strand template, resulting in the formation of a DSB with one double-stranded end. By contrast, meiotic DSBs and DSBs formed after DNA replication is completed in S phase or G2/M phase of the mitotic cell cycle have two double-stranded ends. To direct the repair of two-ended DSBs into HR, 5’ ends of the DSB need to be resected. The initial, short-range resection is carried out by the MRX complex and Sae2 while subsequent long-range resection is catalyzed by Exo1 or Sgs1–Dna2. The ssDNA is first bound by a ssDNA binding protein, RPA, which is replaced by Rad52. When Rad51 is recruited to DSB ends by Rad52, Rad51 catalyzes strand invasion of the DSB end into the homologous duplex. The DSB is resolved by synthesis-dependent strand annealing or double-strand break repair pathways. Because Rad52 has DNA annealing activity, it can catalyze annealing of the two complementary repeated sequences exposed after extensive DSB end resection, which facilitates DSB repair by single-stranded annealing.
Meiotic inter-homolog recombination within the budding yeast rDNA cluster is strongly repressed by > 70-fold, compared to other genomic regions (Petes and Botstein, 1977). This repression of recombination is primarily caused by suppressing the formation of Spo11-dependent meiotic DSBs (Pan et al., 2011). In striking contrast, the rDNA sequence induces HR in vegetatively growing budding yeast cells when inserted near a HR reporter gene positioned in a non-rDNA region (Keil and Roeder, 1984). These findings demonstrate that the budding yeast rDNA displays a hotspot activity of mitotic recombination. The hotspot activity is located in a region, referred to as HOT1, overlapping the enhancer and initiator of 35S rRNA transcription (Voelkel-Meiman et al., 1987). DNA damage within the rDNA cluster is highly prone to induce non-allelic HR due to the cluster’s tandemly repeated organization. Accordingly, this can result in rDNA instability that involves expansion and contraction of the rDNA cluster (Kobayashi et al., 1998, 2004).
Budding yeast cells are inviable when transcription of 35S rRNA by RNA polymerase I is abolished by mutations in a subunit of RNA polymerase I (e.g., RPA135, encoding the second largest subunit) (Mémet et al., 1988; Nogi et al., 1991). These cells only survive when rRNA production is supplied by other means, for example by introducing a multi-copy plasmid that allows rRNA transcription by RNA polymerase II (Nogi et al., 1991). However, the size of chr XII in this strain gradually decreases, leading to a reduction in rDNA copies from the normal average of ~150 to almost half that number (Kobayashi et al., 1998). When RPA135 is complemented back to the mutant, cells gradually increase the rDNA copy number. Intriguingly, however, the rDNA copy number appears to plateau at the normal average (Fig. 4) (Kobayashi et al., 1998). This observation indicates that cells undergo rDNA amplification to restore rDNA copies to the normal level. Similar rDNA amplification is observed when rDNA copies are reduced to two copies (Kobayashi et al., 2001). These findings indicate that budding yeast has a preferred number of rDNA copies and has evolved mechanisms to restore the deficit when the rDNA copy number is reduced.
rDNA amplification to restore the normal rDNA copy number. Budding yeast cells normally contain ~150 rDNA copies. When the rDNA copy number is reduced to less than half, budding yeast cells gradually increase the rDNA copy number in the rDNA cluster via the rDNA amplification pathway (Kobayashi et al., 1998). However, once the copy number reaches a normal average, the copy number plateaus. Therefore, budding yeast cells have the ability to induce or restrict rDNA copy number changes, depending on their rDNA copy numbers.
Sir2 is a NAD+-dependent histone deacetylase that plays an important role in gene silencing at silent mating type loci, telomeres and rDNA in budding yeast. In the absence of Sir2, cells exhibit continuous expansion and contraction of the rDNA cluster, suggesting that Sir2 is a key player in suppressing rDNA instability (Kaeberlein et al., 1999; Kobayashi et al., 2004). To identify other factors responsible for regulating rDNA copy numbers, we previously conducted a genome-wide screen in which the degree of rDNA copy number changes was examined for ~4,800 mutants lacking non-essential genes in the Yeast Knockout Collection (Saka et al., 2016). In this screen, genomic DNA was isolated from each mutant, chromosomes were separated by pulsed field gel electrophoresis, and the size and size heterogeneity of rDNA-bearing chr XII were examined. Due to the labor-intensive nature of this analysis, each mutant was examined only once for its rDNA stability. The study revealed that nearly 10% of the budding yeast genes may contribute to regulation of rDNA copy numbers. It should be noted that we have verified the rDNA instability phenotype in many of the candidate mutants. However, the number of mutants that reproducibly display rDNA instability is lower than that originally revealed by the screen (Kobayashi and Sasaki, 2017). Nonetheless, many mutants reproducibly display frequent rDNA copy number changes. We have identified the molecular mechanisms of how the stability of rDNA is regulated by the genes mutated in some of these mutants: RTT109 that encodes a histone acetyltransferase (Ide et al., 2013); DPB3 and DPB4 that encode subunits of DNA polymerase ε and MRC1 that encodes a factor associated with DNA polymerase ε (Saka et al., 2016); CTF4 that encodes a component of the replisome at the replication fork (Sasaki and Kobayashi, 2017); TEL1 that is an ortholog of the human ataxia–telangiectasia mutated gene and functions in response to DNA damage (Horigome et al., 2019); POP2 and CCR4 that encode subunits of the CCR4-NOT deadenylase complex (Hosoyamada et al., 2019); EAF3 that encodes a component of the NuA4 histone acetyltransferase complex and of the Rpd3S histone deacetylase complex (Wakatsuki et al., 2019); and CLB5 that encodes an S-phase cyclin (Goto et al., 2021).
Enhanced size changes of the rDNA cluster are evident in human cells deficient for the Bloom syndrome patient protein and ataxia–telangiectasia mutated protein, which are involved in the maintenance of genome stability (Killen et al., 2009). HR that involves gene conversions between homologous but not identical copies can transfer sequence information from one copy to the other, leading to sequence homogenization (Ganley and Kobayashi, 2007). The degree of sequence variation among different human rDNA copies remains poorly characterized. This paucity of information largely stems from the fact that the rDNA is large in size, highly abundant and tandemly repeated, rendering it extremely difficult to map any sequenced rDNA reads to individual copies in the rDNA cluster during genome-wide sequencing studies. Long-read sequencing such as Oxford Nanopore Technologies and single-molecule real-time sequencing technology from Pacific Biosciences should overcome the difficulties of assembling repetitive regions in the genome (Pollard et al., 2018). Indeed, by utilizing Oxford Nanopore Technologies, a recent study reveals variability in the copy number and sizes of the R and Butterfly/Long repeats between different rDNA copies (i.e., the copy number of the R repeat is 2–3 on average but can vary from 0 to 4) (Fig. 2C) (Hori et al., 2021). Importantly, this study also demonstrated that contiguous rDNA copies display similar arrangements of the R and Butterfly/Long repeats, indicating that the human rDNA cluster displays some HR activities and undergoes sequence homogenization.
Production of extrachromosomal rDNA circles (ERCs)rDNA copies can also exist as an ERC, which also contributes to the total number of rDNA copies inside the cell. Amphibia normally contain a rDNA cluster at a single chromosomal locus and somatic cells thus contain the number of nucleoli that match the ploidy of the species. In almost all animal species, oocytes are arrested in prophase of the first meiotic division, containing a large nucleus called the germinal vesicle. While the germinal vesicle of tetraploid amphibia is expected to have four nucleoli, it instead has hundreds of extrachromosomal nucleoli that contain ERCs (reviewed by Brown and Dawid, 1968; Miller, 1981). Such immature oocytes exhibit active rRNA synthesis, which is accompanied by a huge increase in rDNA copy number (Hourcade et al., 1973). These extrachromosomal nucleoli disappear when the germinal vesicle breaks down and the nuclear membrane reforms around the gametes. In contrast to somatic cells, which produce rRNA as required for their survival, oocytes must produce not only ribosomes that are used during oogenesis but also a huge number of ribosomes needed later during embryogenesis. It has been proposed that oocytes generate sufficient ERCs to provide excess templates to produce a stockpile of rRNA (Gall and Rochaix, 1974; Gall et al., 2004).
Budding yeast cells undergo asymmetric cell divisions, in which a mother cell produces a daughter cell as a small bud during each cell division. Like other organisms, budding yeast cells divide a finite number of times before ceasing cell division and thus have a replicative lifespan. Microscopic observations of cells that are nearly at the end of their lifespan demonstrate that the old mother cells show enlarged or fragmented nucleoli (Sinclair et al., 1997). These abnormal nucleolus structures are correlated with the accumulation of ERCs in old cells (Sinclair and Guarente, 1997). As described below, ERCs are produced during the repair of DSBs formed in response to DNA replication fork arrest in the rDNA. Although ERCs can be detected in normally growing budding yeast cell cultures that are mostly enriched for cells that have not divided or have done so just once, ERCs show a strong bias toward being segregated into the mother cells during cell divisions (Sinclair et al., 1997). Intriguingly, cells lacking Sir2 display an unstable rDNA cluster, elevated ERC production and a shortened lifespan, while over-production of Sir2 leads to an extension of lifespan (Kaeberlein et al., 1999; Kobayashi et al., 2004). In contrast, cells lacking Fob1 display a stable rDNA cluster, suppression of ERC production and a prolonged lifespan (Kobayashi et al., 1998; Defossez et al., 1999; Takeuchi et al., 2003). Thus, there is a close relationship between the stability of the rDNA cluster, ERC production and cellular longevity. Readers are encouraged to refer to other reviews summarizing research findings that suggest how rDNA stability influences cellular senescence (Sinclair et al., 1998; Kobayashi, 2011; He et al., 2018).
Programmed DNA replication fork arrestOur understanding of patterns of DNA replication within a given locus has been substantially advanced by developments in two-dimensional agarose gel electrophoresis (Bell et al., 1977). This technique allows various patterns of DNA replication in a given DNA locus to be examined, such as the presence of a DNA replication origin, replication fork blocking sites, the extent of DNA replication, direction of replication fork movement, and the occurrence of homologous recombination (Brewer and Fangman, 1987; Huberman et al., 1987). Two-dimensional agarose gel analyses of DNA isolated from S-phase cells of budding yeast provided physical evidence that IGS2 contains an origin of DNA replication (Fig. 2A), which becomes activated during S phase in approximately one in every three to ten rDNA units (Brewer and Fangman, 1988; Linskens and Huberman, 1988). After bi-directional DNA replication is initiated from an active origin of DNA replication, the leftward moving replication fork is free to progress, whereas > 90% of the rightward forks progressing in the direction opposite to the 35S rDNA halt at the RFB site located in the IGS1 region (Fig. 5) (Brewer and Fangman, 1988; Linskens and Huberman, 1988; Brewer et al., 1992; Kobayashi et al., 1992). Thus, due to orientation-dependent DNA replication fork arrest, the rDNA region appears to be replicated unidirectionally in the same orientation relative to 35S rDNA.
Proposed model of the mechanisms that maintain or change rDNA copy numbers during DSB repair. When DNA replication is initiated, the replication fork moving in the direction opposite to the 35S rDNA is arrested by Fob1 protein bound to the RFB site, leading to the formation of a one-ended DSB. The first important decision point as to whether cells induce or restrict rDNA copy number changes is whether or not cells initiate DSB end resection. (Left-hand side) When cells choose “NO” to DSB end resection, the DSB is repaired by an as-yet-unidentified MRX-dependent, HR-independent pathway, which does not induce rDNA copy number changes. Ctf4 protein, a component of the replication fork, is involved in suppression of DSB end resection. (Right-hand side) When cells choose “YES” to DSB end resection, for example when they lack Ctf4 or carry a low rDNA copy number, DSBs undergo end resection, initiating the Rad52-dependent pathway. Furthermore, absence of Sir2 derepresses E-pro transcription, leading to cohesin dissociation. (i) When DSB repair engages a misaligned rDNA copy on the sister chromatid, DNA synthesis takes place from a DSB end to the RFB site, inducing rDNA amplification. (ii) When DSB repair engages another copy on the same chromosome, the rDNA copy is lost as an ERC from the rDNA cluster. (iii) When DNA replication is initiated from DNA replication origins in two consecutive rDNA copies, two one-ended DSBs are formed. This fragment can be released and circularized to form an ERC, possibly by Rad52-dependent single-strand annealing activity, as proposed by Mansisidor et al. (2018). The remaining region is re-replicated by the replication fork that arrives at the permissible side of the RFB from a downstream origin. In this pathway, the rDNA copy number within the chromosomal rDNA cluster is unchanged but ERCs are produced, leading to a net increase in the rDNA copy numbers.
Fob1 protein is responsible for replication fork arrest at the RFB site, inducing expansion and contraction of the rDNA cluster as well as ERC production (Kobayashi and Horiuchi, 1996; Kobayashi et al., 1998; Defossez et al., 1999). Fob1 protein purified from yeast cells binds to the RFB sequence in vitro (Kobayashi, 2003). Chromatin immunoprecipitation experiments demonstrate that Fob1 is associated with the RFB site through a putative zinc finger motif in vivo (Kobayashi, 2003). Interestingly, association of Fob1 with the RFB sequence is observed even from cells untreated with crosslinking reagent, indicating that the interaction has a high level of affinity (Kobayashi, 2003). Furthermore, structural analysis of the DNA sequence in complex with the Fob1 protein by atomic force microscopy demonstrates that the DNA appears wrapped around the Fob1 protein, reminiscent of a nucleosome structure (Kobayashi, 2003). Thus, by forming a tight complex with the RFB sequence, Fob1 blocks progression of the replication fork.
Analysis of forks arrested at the RFB site showed that the leading and lagging strands are arrested at similar positions, the leading strand being three bases behind the lagging strand (Gruber et al., 2000). These findings indicate that single-stranded DNA (ssDNA) is minimally exposed at most arrested forks (Fig. 5). EM studies of chromatin at the arrested forks also demonstrate that the DNA immediately behind the branch point of the arrested fork appears mostly double-stranded (Lucchini and Sogo, 1994). When the RFB site is inserted in an ectopic non-rDNA region of the budding yeast genome, Fob1 bound to this ectopic RFB site pauses the replisome specifically by blocking movement of the Cdc45–MCM–GINS replicative helicase complex (Calzada et al., 2005) that translocates 3′ to 5′ on the leading strand template (Li and O’Donnell, 2018).
Tof1 and Csm3 in budding yeast and their homologs (Swi1 and Swi3 in fission yeast and TIMELESS and TIPIN in humans) are involved in checkpoint activation, along with Mrc1, upon replication stress. Tof1 and Csm3, but not Mrc1, are required for replication fork pausing at the RFB (Calzada et al., 2005; Tourrière et al., 2005; Mohanty et al., 2006). Rrm3 is a 5′ to 3′ DNA helicase that acts as an accessory helicase at the replisome (Ivessa et al., 2003). Absence of Rrm3 causes an elevation in the level of arrested forks at the RFB (Ivessa et al., 2000; Mohanty et al., 2006). Intriguingly, this elevated replication fork arrest in the rrm3∆ mutant appeared to be suppressed to wild-type levels by deletion of TOF1 or CSM3 (Mohanty et al., 2006), which led to the original model that the Tof1–Csm3 complex promotes replication fork arrest by counteracting the activity of Rrm3 to remove Fob1 from the RFB site (Mohanty et al., 2006). However, a recent study demonstrated that the level of arrested forks is reduced by deletion of TOF1 or CSM3 in rrm3∆ cells to a level below that seen in WT cells (Shyian et al., 2020). Furthermore, the replication fork-pausing reactions at the RFB have recently been reconstituted in vitro using purified proteins. In this study, pausing of the replisome at the RFB requires both Fob1 and the Tof1–Csm3 complex but the efficiency of fork pausing is unaffected by the presence of Rrm3 (Hizume et al., 2018). Thus, the Tof1–Csm3 complex promotes fork arrest even in the absence of Rrm3 and these factors act independently as a regulator of replication fork pausing. The list of regulators of Fob1-dependent replication fork arrest continues to expand and includes the Smc5/Smc6 complex, Dbf1-dependent kinase, Pif1 helicase and topoisomerases. We refer readers to other reviews on this subject (Hizume and Araki, 2019; Shyian and Shore, 2021).
It is often assumed that replication fork arrest is required to prevent collision between the DNA replication and transcription machineries. However, replication fork arrest at the RFB does not rely on rDNA transcription (Brewer et al., 1992; Kobayashi et al., 1992). Furthermore, strains lacking Fob but with normal rDNA copy numbers, where replication forks can enter the 35S rDNA sequence from the 3′ end, display neither any sign of replication fork slowdown nor any growth defects (Takeuchi et al., 2003; Ide et al., 2010). When Fob1 is absent from strains carrying ~20 rDNA copies, slowdown of the replication fork is evident within the 35S rDNA coding region, presumably due to collision between the replication and transcription machineries (Takeuchi et al., 2003). This rDNA low-copy fob1 mutant shows increased sensitivity to DNA damaging agents but grows at the same rate as normal-copy strains in unperturbed conditions (Ide et al., 2010). Therefore, cells are not necessarily adversely affected when replication fork-blocking activity is removed.
DSB formation and its end resectionReplication fork arrest at the RFB leads to DSB formation (Fig. 5) (Weitao et al., 2003; Burkhalter and Sogo, 2004; Kobayashi et al., 2004). This DSB has only a single DSB end (Burkhalter and Sogo, 2004; Sasaki and Kobayashi, 2017), which is structurally different from the two-ended DSBs that are formed during meiosis, S phase after DNA replication completion and G2/M phase of the mitotic cycle (Fig. 3). High-resolution mapping of the broken site for DSBs formed at the RFB failed to detect any breakage on the parental lagging strand in an S-phase-specific and Fob1-dependent manner, leading to the proposal that the parental leading strand is cleaved to generate a one-ended DSB (Burkhalter and Sogo, 2004).
Extensive studies conducted over the past several decades have identified two major repair pathways for two-ended DSBs: non-homologous end-joining (NHEJ) and HR. Only a brief summary is provided here and readers are encouraged to refer to several excellent reviews on this topic (Pâques and Haber, 1999; Symington et al., 2014; Chang et al., 2017; Scully et al., 2019; Zhao et al., 2020). In NHEJ, the Ku70–Ku80 heterodimer binds to the DSB ends and then ligases are recruited to re-join the DNA (reviewed by Chang et al., 2017; Scully et al., 2019; Zhao et al., 2020). Because nucleotides can be removed or added to the DSB ends in a template-independent manner, NHEJ may facilitate sequence changes at the broken site. HR is usually considered to be an error-free pathway (reviewed by Pâques and Haber, 1999; Symington et al., 2014). To direct DSB repair into HR, 5’-ended strands of the DSB ends are degraded by a process called DNA end resection (Fig. 3). The Mre11–Rad50–Xrs2 complex (the human homolog of Xrs2 is Nbs1) and Sae2 (CtIP in human) carry out an initial, short-range resection, which is taken over by the redundant action of Exo1 and Sgs1–Dna2 (the human homologs of Sgs1 are BLM and WRN) for long-range resection. The resulting 3’-ended ssDNA is first bound by a ssDNA binding protein, RPA, which is subsequently replaced by Rad52. Rad52 recruits Rad51 that carries out strand invasion of the 3’ DSB end into a homologous donor duplex and generates a D-loop, which is resolved by synthesis-dependent strand annealing or double-strand break repair pathways (Fig. 3, Rad51/52-dependent pathway). Alternatively, Rad52 possesses ssDNA annealing activity and can anneal ssDNA of two complementary, ssDNA overhangs that are exposed by extensive DSB end resection (Fig. 3, Rad52-dependent pathway). The ssDNA annealing-mediated DSB repair can lead to deletion of DNA between two repeated sequences.
rDNA amplification from low-copy strains is abolished when cells lack Rad52 (Kobayashi et al., 2004). Furthermore, ERC formation is also suppressed by deletion of RAD52 (Park et al., 1999). Considering that HR is critical for repairing two-ended DSBs, these findings have been interpreted as meaning that HR is required for repairing one-ended DSBs formed at the RFB. Depending on which rDNA sequence is used as the repair template, DSB repair results in changes or no changes to the number of rDNA copies (Kobayashi, 2006, 2011).
Fritsch et al. (2010) determined the level of DSBs formed at the RFB in various mutants by isolating genomic DNA from logarithmically growing cells and detecting arrested forks and DSBs by two-dimensional agarose gel electrophoresis and Southern blotting. If cells are defective in DSB repair, unrepaired DSBs should accumulate, leading to a higher level of DSBs than in DSB repair-proficient cells. Cells lacking either Rad52 or the NHEJ factor Dnl4 did not show elevated DSBs.
We previously conducted detailed, quantitative analyses by monitoring formation and repair of DSBs during synchronous S phase progression using Southern blotting (Sasaki and Kobayashi, 2017). This study offered several important findings. First, we estimate that at least one DSB is generated at the RFB during S phase of every cell cycle (Sasaki and Kobayashi, 2017). Second, DSB end resection, which is important for initiating HR, is suppressed in WT cells carrying normal rDNA copy numbers (Fig. 5). In the canonical HR-mediated repair, the MRX complex is central to DSB repair either through its structural role of maintaining DSB ends or sister chromatids in close proximity to one another, or through its catalytic role along with Sae2, which initiates DSB end resection via the nuclease activity of Mre11 (Fig. 3) (Stracker and Petrini, 2011; Oh and Symington, 2018; Casari et al., 2019). The MRX complex is important for DSB repair at the RFB but neither the Mre11 nuclease activity nor Sae2 is required, suggesting that the MRX complex facilitates DSB repair through its structural role (Sasaki and Kobayashi, 2017). Moreover, although Rad52 is essential for all HR events during repair of two-ended DSBs (Pâques and Haber, 1999), it is dispensable for repair of one-ended DSBs formed at the RFB (Sasaki and Kobayashi, 2017). Consistent with these findings, genetic analyses demonstrate that cells lacking the MRX complex but not the nuclease activity of Mre11 or Rad52 are sensitive to FOB1 expression (Calzada et al., 2005; Bentsen et al., 2013; Sasaki and Kobayashi, 2017). DSB repair does not require Ku70 or the DNA ligase Dnl4, which are required for NHEJ (unpublished data). In summary, end resection is normally suppressed for DSBs formed at arrested forks and these DSBs are repaired by the HR- and NHEJ-independent pathway to keep the rDNA copy number unchanged (Fig. 5, left-hand side).
We uncovered two situations where suppression of DSB end resection is relieved (Sasaki and Kobayashi, 2017). First, when cells lack Ctf4, a component of the replication fork, DSBs undergo end resection and Rad52-mediated repair that induces rDNA hyper-amplification. This finding is also supported by genetic analyses that show that deletion of CTF4 results in synthetic sickness/lethality with deletion of RAD52, which was suppressed by further deletion of FOB1 (Fumasoni et al., 2015; Sasaki and Kobayashi, 2017). Thus, Ctf4 is critical for preventing rDNA instability by suppressing DSB end resection. Second, cells carrying a low rDNA copy number undergo DSB end resection and Rad52-dependent rDNA amplification (Kobayashi et al., 2004; Sasaki and Kobayashi, 2017). It remains to be determined whether Rad52-mediated reactions are an absolute requirement for DSB repair in the rDNA low-copy strains. Therefore, the decision of whether or not cells allow DSB end resection is a critical step that dictates the outcome of DSB repair as to whether it maintains or compromises rDNA stability.
Transcription of non-coding RNA from E-pro and cohesin associationsCohesin complexes establish sister chromatid cohesion, which facilitates faithful chromosome segregation and accurate DSB repair. Cohesin complexes are associated with the cohesin-associated region in IGS2 (Fig. 2A and Fig. 5) (Laloraya et al., 2000). Chromatin immunoprecipitation experiments demonstrate that the level of cohesin complexes localized to the cohesin-associated region is reduced in the absence of Sir2. Based on these findings, a model was proposed in which Sir2-mediated association of cohesin is crucial for positioning sister chromatids in close proximity, thereby facilitating equal sister chromatid recombination and maintaining rDNA stability (Fig. 5) (Kobayashi et al., 2004).
In the IGS1 region, there is a bi-directional RNA polymerase II-dependent promoter, named E-pro, that is normally repressed by Sir2 (Fig. 2A and Fig. 5) (Santangelo et al., 1988; Ganley et al., 2005). To investigate the importance of E-pro regulation for rDNA stability, E-pro was replaced with a galactose-regulatable, bi-directional GAL1/10 promoter in the rDNA 2-copy strain (Kobayashi and Ganley, 2005). While rDNA amplification is inhibited when transcription from the GAL1/10 promoter is repressed, transcription activation upon addition of galactose to the medium was found to induce cohesin dissociation, rDNA amplification and ERC production (Kobayashi and Ganley, 2005; Saka et al., 2013). Therefore, transcription repression from E-pro is critical for ensuring stable association of cohesins and rDNA stability (Kobayashi, 2011).
Budding yeast cells undergo rDNA amplification and produce more ERCs when the rDNA copy number is lowered, highlighting the existence of mechanisms that monitor the cells’ own rDNA copy numbers and alter rDNA stability according to their rDNA copy number (Fig. 4) (Kobayashi et al., 1998; Kobayashi et al., 2004; Mansisidor et al., 2018). Intriguingly, gene silencing at telomeres and silent mating-type loci is enhanced in cells that have undergone a large rDNA deletion, while gene silencing at the rDNA is normal in these cells (Michel et al., 2005). SIR2 expression is reduced in the rDNA low-copy cells, compared to normal-copy cells (Michel et al., 2005; Iida and Kobayashi, 2019b). Thus, the rDNA low-copy cells down-regulate SIR2 expression and release nucleolar Sir2 into the nuclear space to be used for gene silencing at non-rDNA loci. A recent study reveals that this rDNA copy number monitoring system depends on Upstream Activating Factor (UAF), which is normally bound to the rRNA promoter to enhance rRNA transcription (Iida and Kobayashi, 2019a, 2019b). When the rDNA copy number is reduced, UAF is released from the rRNA promoter and redistributed to the SIR2 promoter to act as a repressor of SIR2 transcription. Thus, in rDNA low-copy cells, repression of SIR2 expression results in activation of transcription from E-pro and cohesin dissociation, and induces rDNA instability, specifically rDNA amplification.
A model of rDNA copy number changes during DSB repairRecent findings have advanced our understanding of the regulatory processes that influence cellular responses to either maintain or change rDNA stability during DSB repair (Fig. 5). The first important decision point is whether or not cells initiate DSB end resection. If the answer is “NO” to DSB end resection, DSBs are repaired by a novel, as-yet-unidentified, pathway that is dependent on the MRX complex but not on HR factors (Fig. 5, left-hand side), which maintains the rDNA copy number (Sasaki and Kobayashi, 2017). It is possible that the arrival of the replication fork from the other side of the RFB facilitates resealing of broken DNA strands. The detailed mechanism of how DSBs are repaired without HR remains to be determined.
Cells carrying a temperature-sensitive allele of SMC1 have been shown to display an increased frequency in the loss of a marker gene inserted into the rDNA cluster (Kobayashi et al., 2004). Deficiency in cohesin association may induce DSB end resection. Alternatively, cohesin dissociation may promote HR-independent DSB repair and induce rDNA copy number changes (Fig. 5, dashed arrow).
If cells choose “YES” to undergo DSB end resection, DSB repair is directed into the Rad52-dependent pathway but this pathway is highly prone to unequal recombination in regions such as an rDNA cluster where the same copies are tandemly repeated (Fig. 5). Cohesion between sister chromatids is an important regulatory process that influences template choice during DSB repair. Sir2-mediated repression of E-pro facilitates cohesin association, promoting the engagement of the aligned copy on the sister chromatid for DSB repair and leading to DSB repair without causing rDNA copy number changes. When these processes are compromised, for example by the absence of Sir2, the DSB end is highly prone to engage the misaligned copy on the sister chromatid and DNA synthesis occurs; DSB repair is accompanied by expansion of the rDNA cluster (Fig. 5, (i)). If the broken end engages the misaligned copy on the same chromosome, DSB repair can result in release of ERCs from the rDNA cluster at the cost of loss of the rDNA copy from the rDNA cluster (Fig. 5, (ii)). It is important to note that the absence of Rad52 causes more severe defects in DSB repair in the absence of Ctf4 as well as ERC production in wild-type cells, as compared with cells lacking Rad51 (Park et al., 1999; Sasaki and Kobayashi, 2017). Therefore, Rad52-dependent reactions that do not involve the strand invasion step may contribute more to DSB repair that leads to rDNA copy number changes.
The group led by Hochwagen has examined the relationship between ERC formation and changes to the rDNA cluster (Mansisidor et al., 2018). Based on the results of this study, a model of ERC formation was proposed in which ERCs are generated without causing contraction of the rDNA cluster (Fig. 5, (iii)) (Mansisidor et al., 2018). When DNA replication is initiated from two adjacent replication origins, replication forks are arrested at adjacent RFB sites. It is proposed that formation of DSBs at these arrested forks can induce Rad52-dependent single-strand annealing, generating an ERC. Moreover, the results of a study using DNA combing and single-molecule imaging suggest that active replication origins form clusters of two to three consecutive rDNA copies (Pasero et al., 2002). Thus, DSB formation at neighboring rDNA copies is highly likely in some rDNA copies. We propose to modify the model of Mansisidor et al. by adding additional requirements to ERC formation. Specifically, we propose that a critical initiator of ERC formation is not just DSB formation but that these DSBs also undergo DSB end resection. Because cells lacking Sir2 accumulate ERCs at a substantially higher level than wild-type cells (Kaeberlein et al., 1999), Sir2 may promote DSB end resection and single-strand annealing-dependent ERC formation. The precise role of Sir2 will need to be examined in future studies. Furthermore, sister chromatid cohesion between the excised fragment and the unbroken sister chromatid must prevent the excised fragment from diffusing away from the chromosome. Thus, cohesin dissociation may also drive this pathway. After ERCs are formed the looped-out region can be filled by DNA synthesis, which generates ERCs without any associated reduction in chromosomal rDNA copies (Fig. 5, right-hand side, (iii)) (Mansisidor et al., 2018). ERCs can also re-integrate into the chromosomal rDNA cluster, leading to rDNA expansion (Ide et al., 2013; Mansisidor et al., 2018).
A low number of rDNA copies is unfavorable to budding yeast cells because this generates a phenotype with increased sensitivity to DNA damage (Ide et al., 2010). In such a scenario, cells undergo rDNA amplification to restore the normal rDNA copy number (Fig. 4) (Kobayashi et al., 1998). However, once the cells harbor the preferred number of rDNA copies, constraints are imposed to prevent further changes to the rDNA copy number (Kobayashi, 2011). These findings demonstrate that rDNA instability can have both positive and negative influences on cellular life. However, the details of how cells switch between maintaining rDNA stability and inducing rDNA instability remain to be resolved. To begin to answer this question, it is important to advance our understanding of how changes in rDNA copy number impact cellular growth and phenotypic variation. To this end, comprehensive characterization of phenotypes induced in cells with low, normal or high rDNA copy numbers is required. Furthermore, because the rate of rRNA synthesis changes according to cell proliferation status, metabolic activity and environmental signals, it is also important to gain insights into how demands for ribosome production in various cellular contexts influence rDNA copy number.
Mounting evidence supports the notion that HR factors play key roles in the maintenance of genome stability in the response to DNA replication stress. Addition of DNA replication inhibitors results in an accumulation of nuclear foci of HR factors, enhancement of HR frequency in the genome, and an increase in growth defects in HR-deficient cells (Arnaudeau et al., 2000; Saintigny et al., 2001; Lundin et al., 2002). When DSBs are induced such that one of the two DSB ends has homology elsewhere in the genome, these DSBs are repaired by break-induced replication, which depends on HR factors (Anand et al., 2013). These findings have been interpreted as meaning that HR is required for repairing DSBs generated during replication stress. However, we emphasize that there is not a single, common DSB repair pathway to repair DSBs formed as a result of DNA replication stress.
First, DNA replication is inhibited by various factors, including pre-existing DNA damage to the template DNA, DNA secondary structure, nucleotide depletion, proteins tightly bound to the DNA, and the presence of transcription factors. When genomic DNA isolated from human cells that have been treated with various types of DNA-damaging and replication-inhibiting agents is separated by pulsed-field gel electrophoresis, broken chromosomal fragments are evident (Zellweger et al., 2015). However, because these clastogens induce replication inhibition at essentially random places, it is difficult to analyze the structures of DSBs induced by such agents. Specifically, it is technically challenging to establish whether the DSBs are one-ended or two-ended, comprise fully replicated double-stranded DNA, or have replication proteins at the DSB ends. Furthermore, it is difficult to analyze the DSB repair intermediates by physical assays such as Southern blotting. Thus, we cannot definitively demonstrate how these DSBs are repaired, or which factors are involved in the DSB repair. Furthermore, previous studies have demonstrated the DSB-independent function of HR factors during responses to replication stress. For example, HR factors are required to protect arrested forks from degradation (Schlacher et al., 2011, 2012). HR can induce chromosome rearrangements during the re-start of DNA synthesis, independently of DSB formation (Lambert et al., 2005). Therefore, without directly monitoring DSB repair processes, we can distinguish neither whether HR is involved in DSB repair or DSB-independent processes nor whether HR is the main or pathological DSB repair pathway, if indeed HR is involved in DSB repair.
The budding yeast rDNA region is a unique region in the eukaryotic genome where DSBs are formed upon DNA replication inhibition at a frequency high enough for their detection by Southern blotting. Our recent work has uncovered the unexpected finding that DSB end resection, an initiating event for HR, is normally suppressed to maintain the stability of the rDNA during DSB repair (Sasaki and Kobayashi, 2017). Relief of this suppression, for example in cells lacking Ctf4 or with a low rDNA copy number, induces Rad52-dependent reactions that lead to rDNA copy number changes as a consequence of DSB repair. These findings demonstrate that Rad52-mediated HR is the pathological DSB repair process that is used to induce genome instability. Future study is required to understand the regulation of DSB end resection in great detail. Moreover, it is important to identify how DSBs are repaired in the HR-independent repair pathway. We believe that insights gained from the budding yeast rDNA region will serve as the starting point to understanding the similarities and differences in the mechanisms involved in repairing DSBs that are formed by various causes during DNA replication.
Transcription of non-coding RNA from E-pro causes cohesin dissociation, which yields an outcome that leads to alterations in rDNA copy number (Fig. 5, right) (Kobayashi and Ganley, 2005; Saka et al., 2013). Although E-pro activation and cohesin dissociation induce rDNA instability, the outcome of the rDNA copy number changes is biased toward rDNA expansion in rDNA low-copy cells, while rDNA normal-copy cells lacking Sir2 undergo continuous expansion and contraction of the rDNA cluster. Thus, as-yet-uncovered processes influence how rDNA copies change during DSB repair. Characterization of mutants that undergo either expansion or contraction will provide insights into the regulatory processes that dictate how DSB ends are repaired. When transcription of 35S rRNA by RNA polymerase I is abolished by a mutation in RPA135, encoding a subunit of RNA polymerase I, but rRNA production is complemented by other means, cells gradually decrease the size of the rDNA cluster but they start to increase the rDNA copy number once RPA135 is complemented back to the mutant cells (Fig. 4) (Kobayashi et al., 1998). From our genome-wide screening, we have identified several mutants that undergo rDNA amplification, such as a mutant lacking Ctf4, the histone acetylase Rtt109, and the Mms22 subunit of an Rtt101-based E3 ubiquitin ligase (Ide et al., 2013; Saka et al., 2016). Furthermore, previous studies have demonstrated that the rDNA cluster is shortened when DNA replication is compromised, for example in the temperature-sensitive orc2-1 and orc1-4 mutants (Ide et al., 2007; Salim et al., 2017; Sanchez et al., 2019). Characterization of the defects in these mutants may reveal as-yet-uncovered processes of how cells choose to expand or contract the rDNA cluster during DSB repair.
We would like to thank Junko Kanoh for insightful comments on the presentation of the data. Work from the authors’ laboratory is supported in part by Grants-in-Aid for Scientific Research (20H05382, 20K06597, 18H04709, 17K15160 and 15K18581 to M. S., and 17H01443 and 21H04761 to T. K.), the Uehara Memorial Foundation, Takeda Science Foundation and Naito Foundation to M. S., JST FOREST Program (Grant Number JPMJFR214P to M. S.) and JST CREST (Grant Number JPMJCR19S3 to T. K.).
