Transcription and Recombination: When RNA Meets DNA

Abstract

A particularly relevant phenomenon in cell physiology and proliferation is the fact that spontaneous mitotic recombination is strongly enhanced by transcription. The most accepted view is that transcription increases the occurrence of double-strand breaks and/or single-stranded DNA gaps that are repaired by recombination. Most breaks would arise as a consequence of the impact that transcription has on replication fork progression, provoking its stalling and/or breakage. Here, we discuss the mechanisms responsible for the cross talk between transcription and recombination, with emphasis on (1) the transcription–replication conflicts as the main source of recombinogenic DNA breaks, and (2) the formation of cotranscriptional R-loops as a major cause of such breaks. The new emerging questions and perspectives are discussed on the basis of the interference between transcription and replication, as well as the way RNA influences genome dynamics.

Transcription–replication conflicts may generate DNA breaks or gaps that are repaired by recombination. R-loops, DNA–RNA hybrids formed during transcription, may play a central role in their occurrence.

Homologous recombination (HR) is a conserved pathway responsible for the repair of double-strand breaks (DSBs). In mitotic cells, DSBs may be induced by genotoxic agents, such as γ irradiation or may occur spontaneously, in most of the cases in association with replication. Although HR represents one of the two main mechanisms of DSB repair, the other being nonhomologous end joining (NHEJ), eukaryotic cells favor HR as the preferential DSB repair pathway during the S/G₂ phases of the cell cycle, when the sister chromatid is available as template for error-free repair. Thus, HR events are regulated at different steps along the cell cycle, among others by the cyclin-dependent kinase 1 during 5′-end resection to guarantee its occurrence at S/G₂ (Heyer et al. 2010; Huertas 2010). Consequently, spontaneous mitotic recombination events are generally interpreted as the result of DSB repair during S/G₂, although it cannot be disregarded that recombination could also be initiated by single-stranded DNA (ssDNA) gaps generated during replication.

Spontaneous mitotic recombination might, in principle, take place anywhere in the genome with similar probability. However, as for mutations, both recombination and chromosome breakages occur more frequently in particular regions, referred to as hot spots. Those hot spots might arise from different local features, including the formation of non-B secondary structures, chromatin compaction, DNA–protein barriers to replication, or low-replication initiation density (Aguilera and Garcia-Muse 2013). Nevertheless, the most extended and physiological relevant feature enhancing the probability of recombination is likely to be transcription (Aguilera 2002; Kim and Jinks-Robertson 2012; Gaillard et al. 2013). From bacteria to humans, a large body of evidence has accumulated showing that transcription stimulates spontaneous recombination, a phenomenon referred to as transcription-associated recombination (TAR). As replication failures seem to be the main source of recombinogenic DSBs, our actual view is that the major mechanism by which transcription stimulates recombination is via DSBs or ssDNA gaps potentially generated by the difficulties of the DNA replication fork to progress through transcribed DNA sequences (Aguilera and Gomez-Gonzalez 2008; Bermejo et al. 2012). Here, we review our actual knowledge of the cross talk between transcription and recombination with the aim of providing mechanistic insights into the relevance for TAR as a natural source of genome instability along with discussing emerging questions and perspectives.

TRANSCRIPTION STIMULATES RECOMBINATION FROM BACTERIA TO HUMAN CELLS

The first evidence of TAR comes from λ-phage studies showing that recombination occurs in transcribed DNA regions and depends on Escherichia coli RNA polymerase (RNAP) activity but not on RecA (Ikeda and Kobayashi 1977; Ikeda and Matsumoto 1979). Probably, the most influential reports on TAR were those showing the identification of the recombinant DNA (rDNA) sequence HOT1 as a hot spot of recombination in Saccharomyces cerevisiae (Keil and Roeder 1984), and the following demonstration that the ability of HOT1 to stimulate Rad52-dependent ectopic recombination of non-rDNA sequences depends on the RNAPI activity (Voelkel-Meiman et al. 1987; Stewart and Roeder 1989). Subsequent work on phage transduction (Dul and Drexler 1988), illegitimate recombination in bacterial plasmids (Vilette et al. 1992), RNAPII-mediated recombination between DNA repeats in S. cerevisiae (Thomas and Rothstein 1989), at the ADE6 locus of Schizosaccharomyces pombe (Grimm et al. 1991) and in rodent cells (Nickoloff and Reynolds 1990) showed that TAR is a conserved feature. A survey of these and later examples of TAR from bacteria to mammals can be found in a recent review (Gaillard et al. 2013).

Although the possibility that the impact of transcription on features such as chromatin structure or the channeling of DNA breaks into different types of repair has been considered an explanation of TAR, our current understanding favors TAR being the consequence of an increase in recombinogenic DNA damage caused by transcription. In yeast, DSBs generated by the homothallic switching (HO) endonuclease induce recombination to similar levels regardless of transcription being active or not (Weng et al. 2000; Gonzalez-Barrera et al. 2002). In mammalian cells, transcription did not stimulate DSB-induced recombination (Taghian and Nickoloff 1997). DSB-induced recombination and TAR events share similar features of reciprocal versus nonreciprocal events and are dependent on recombinational DSB repair functions such as Rad52 and Rad51 in yeast (Gonzalez-Barrera et al. 2002) or BRCA2 in mammals (Savolainen and Helleday 2009). Recently, human cell lines defective in XPD, a DNA helicase subunit of transcription factor II H (TFIIH) required for transcription initiation and nucleotide excision repair, revealed that XPD is required for TAR, but not for DSB-induced recombination (Savolainen et al. 2010). Although it is possible that a fraction of TAR events does not initiate via DSBs, we cannot exclude the possibility that some TAR events may also require XPD to be initiated. All data, therefore, strongly support the idea that transcription stimulates recombination by inducing recombinogenic damage at the transcribed DNA sequence.

In support of this idea, it has been shown that genotoxic agents such as 4-nitroquinoline 1-oxide or methyl methanesulfonate induce gene conversion synergistically with transcription (Garcia-Rubio et al. 2003). Although the looser chromatin state of transcribed regions has been proposed to provide a better substrate for the action of genotoxic agents, there is no definitive evidence for this possibility. Also, it does not seem that transcription by itself increases DNA accessibility to endonucleases or other recombination enzymes, as high transcription reduced both HO cleavage efficiency and strand invasion (Saxe et al. 2000; Gonzalez-Barrera et al. 2002). Instead, and according to the “twin-supercoiled domain” model (Liu and Wang 1987) in which transcription creates negative and positive supercoils behind and ahead of the elongating RNAP, respectively, the proneness of the strands of negatively supercoiled double-stranded DNA (dsDNA) to separate would make DNA more accessible to genotoxic agents (Fig. 1A).

Figure 1.

Transcription intermediates can compromise genome integrity in prokaryotes and eukaryotes. (A) In eukaryotes, mRNP biogenesis and export are coupled to transcription. Transcribed genes may also be anchored to the nuclear pore. Negative supercoiling (−) accumulates behind the elongating RNAPII and facilitates the separation of both strands, making the DNA more susceptible to genotoxic agents. Positive supercoiling (+) accumulates ahead of the transcription machinery, in front of a head-on oncoming replication fork. (B) The nascent messenger RNA (mRNA) might hybridize back to its DNA template behind the RNAPII, forming an R-loop. (C) In prokaryotes, cotranscriptional translation of the nascent mRNA may impede the formation of R-loop. (D) Cotranscriptional folding of nonprotein coding RNAs may reduce the probability of R-loop formation in both eukaryotes and prokaryotes.

This model is supported by the reports on transcription-associated mutagenesis. Mutagenicity of a variety of external genotoxic agents is strongly stimulated in actively transcribed DNA sequences (Aguilera 2002; Kim and Jinks-Robertson 2012; Gaillard et al. 2013). Moreover, S. cerevisiae lacking topoisomerase 1 (Top1) activity undergoes a strong increase in recombination leading to the loss of a marker inserted in the rDNA (Christman et al. 1988), which has been recently extended to inverted- and direct-repeat systems under the control of RNAPII (Garcia-Rubio and Aguilera 2012). These observations, as well as the hypermutation engendered by the top1 mutation in yeast (Kim et al. 2011; Lippert et al. 2011; Takahashi et al. 2011), would be consistent with better accessibility of genotoxic agents facilitated by negative DNA supercoiling. Besides, nonrandom mutation clusters probably associated with long ssDNA were found both in yeast grown under chronic aklylation damage and in human malignant tumors (Roberts et al. 2012, 2013; Alexandrov et al. 2013; Burns et al. 2013). These results suggest that ssDNA patches are formed in proliferative cells, possibly during DSB repair, replication, or R-loop formation (see below). Yet, the fact that recombination is tightly regulated during the cell cycle, having its peak during the S/G₂ phases, suggests that TAR, like any other form of mitotic recombination, is associated with replication.

R-LOOPS AS MEDIATORS OF TAR

R-loops are structures in which an RNA segment hybridizes with its DNA template, thereby displacing the complementary strand, which remains as a single-stranded loop (Fig. 1B). Its existence as a physiologically relevant intermediate has been shown in many systems, including origin-independent replication in bacteria (Kogoma 1997), class-switching recombination (CSR) (Yu et al. 2003), or E. coli topA mutants in which transcription-dependent formation of R-loops behind the RNAP have been genetically inferred (Drolet et al. 1995). More recently, R-loops were found at CpG island promoters and G-rich termination elements in human cells (Skourti-Stathaki et al. 2011; Ginno et al. 2012, 2013) and at an Arabidopsis antisense promoter, where it regulates the expression of a long noncoding RNA (Sun et al. 2013).

The first demonstration that R-loops generated during transcription induce genome instability in the form of recombination was provided in S. cerevisiae with mutants of the messenger RNA particle (mRNP) biogenesis and export factor THO (Huertas and Aguilera 2003). The hyper-recombination phenotype of these mutants not only is transcription-dependent and can be suppressed by mutations in the RNAPII machinery that diminish transcription (Santos-Rosa and Aguilera 1995; Fan et al. 1996; Piruat and Aguilera 1996, 1998), but also relies on the nascent RNA molecule and its capacity to form an R-loop behind the elongating DNA polymerase (Huertas and Aguilera 2003). The relevance of R-loops as an intermediate responsible for different forms of instability is, nowadays, supported by an increasing number of reports. These include the demonstration of the formation of R-loops in DT40 chicken and HeLa cells depleted of the splicing factor ASF/SF2 (Li and Manley 2005) or the transcription-dependent instability manifested as gross-chromosomal rearrangements, γ-H2AX foci, or hyper-recombination in yeast, Caenorhabditis elegans, and human cell lines mutated in or depleted of a number of factors involved in RNA metabolism (Gomez-Gonzalez and Aguilera 2007; Gonzalez-Aguilera et al. 2008; Paulsen et al. 2009; El Hage et al. 2010; Dominguez-Sanchez et al. 2011; Wahba et al. 2011; Castellano-Pozo et al. 2012; Stirling et al. 2012; Gavalda et al. 2013; Santos-Pereira et al. 2013). Interestingly, different reports have provided evidence that genome instability associated with DNA repeats with the potential to form non-B DNA structures, such as trinucleotide repeats (Grabczyk et al. 2007; Lin et al. 2010), or common fragile sites (CFSs) in mammalian cells (Helmrich et al. 2011) are also dependent on transcription and associates with the formation of R-loops. Regardless of whether recombination may be triggered in such regions or not, the results confirm that known hot spots of DNA fragility owe their instability to a process dependent on transcription that can be, either partially or completely, related to R-loop accumulation. The RNA, and specifically the R-loop, has thus become a key element in all studies aimed at understanding the cross talk between transcription and recombination (Aguilera and Garcia-Muse 2012).

Top1 restrains R-loop formation by counteracting the transient local accumulation of negative supercoils at transcribed regions that facilitate DNA strand opening (Drolet et al. 1995; Tuduri et al. 2009; El Hage et al. 2010; Marinello et al. 2013; Teves and Henikoff 2014). Cotranscriptional mRNP biogenesis and processing in eukaryotes may also diminish the chances of the nascent RNA to hybridize back with the template DNA behind the RNAP (Fig. 1A,B). The suppression of the transcription-dependent hyper-recombination phenotypes of THO and THSC/TREX-2 mutants by overexpression of heterogeneous nuclear ribonucleoproteins able to bind RNA, such as Tho1, Nab2, or the RNA dependent on ATPase Sub2/UAP56, whose deletion also causes a strong transcription-dependent hyper-recombination (Fan et al. 2001; Jimeno et al. 2002), would support this view. We still know little about the structure of the eukaryotic mRNP, the posttranslational modifications that RNA-binding factors undergo during the RNA cycle, and how they can affect the flexibility or hybridizing properties of the nascent RNA. A speculative view based on the known coupling of transcription with RNA processing could be that improper mRNP biogenesis triggers a transcription checkpoint signal that might affect transcription elongation therefore causing a more stably bound RNAP to the DNA, interfering with other DNA processes such as transcription-coupled repair or replication. In bacteria, it has been postulated that the nascent RNA is prevented from hybridizing with the DNA template by the cotranscriptional nature of translation that disposes ribosomes on the nascent mRNA in protein-encoding genes (Fig. 1C) (Gowrishankar and Harinarayanan 2004). The folding of the nascent RNA into secondary structures in nonprotein-encoding genes may fulfill the same function (Fig. 1D).

It is possible that transcription elongation impairment linked to the formation of R-loops might cause TAR. However, the fact that no transcription defect was observed in ASF/SF2-depleted chicken DT40 cells (Li and Manley 2005), the existence of THO mutants with a transcription defect that does not lead to increased TAR (Huertas et al. 2006), or the lack of correlation between transcription elongation defects and TAR in a number of mRNA processing and export factors (Luna et al. 2005) would argue against such a possibility as a general mechanism. Recently, it has also been concluded that Rad51 is required for DNA–RNA hybrids formed in trans (Wahba et al. 2013), but it is unclear at this point whether this is a mechanism related to TAR.

The increasing evidence accumulated on the relevance of R-loops in different forms of genome instability and existence of proteins with the potential to remove the RNA moiety of R-loops suggest that cotranscriptional R-loops may be formed in wild-type cells proficient in mRNP biogenesis and surveillance. The observation that the sen1-1 mutation of the Sen1/SETX putative RNA–DNA helicase, as reported for the S. pombe Sen1 (Kim et al. 1999), shows a strong TAR phenotype that correlates with transcription and R-loop accumulation argues in favor of this conclusion (Mischo et al. 2011). The questions remaining are whether all TAR events, occurring in both wild-type and mutant cells, are mediated by R-loops, and whether an R-loop by itself can trigger the formation of a DSB and/or a ssDNA gap responsible for the recombination event.

Recently, a link between R-loops and the H3 S10 phosphorylation chromatin condensation mark was uncovered in yeast, nematodes, and human cells (Castellano-Pozo et al. 2013), suggesting an additional possible mechanism by which R-loop can modulate genome dynamics. In the same study, the H3 K9me2 heterochromatin mark has also been associated with R-loops in C. elegans. In addition, DNA–RNA hybrids have been shown to mediate RNA interference–directed heterochromatin formation at pericentromeric regions of S. pombe (Nakama et al. 2012). Hence, chromatin condensation or heterochromatin taking place at or around R-loops could interfere with replication and/or transcription, thus triggering genomic instability (Castellano-Pozo et al. 2013). In agreement with this model, CFSs correlate with slow or incomplete replication at regions of chromatin condensation and cotranscriptional R-loops (Helmrich et al. 2011; Debatisse et al. 2012). It is worth noting, however, that R-loops coincide with increased chromatin decondensation in neurons (Powell et al. 2013). Thus, it seems likely that although the R-loop itself might destabilize nucleosomes, chromatin condensation would occur around and extend from the R-loop. Further work will be required to unravel the mechanism by which R-loops alter chromatin structure and how this chromatin connection modulates TAR.

THE SPECIFIC CASE OF CSR

Whereas TAR represents a global phenomenon of HR, CSR defines a specific type of developmentally controlled recombination occurring by NHEJ, not by HR, in vertebrate B cells, which is worth mentioning here. CSR occurs between the S regions located beside each C segment of the immunoglobulin (Ig) genes. Only when the S region is transcribed does it undergo a DSB event that triggers the NHEJ process responsible for CSR (Fig. 2). The first indications that S regions had the potential to form R-loops were gained using in vitro transcription assays (Reaban and Griffin 1990; Reaban et al. 1994; Daniels and Lieber 1995); definitive evidence of its existence in B cells was provided later by direct mapping (Yu et al. 2003). Significantly, however, CSR differs from TAR in important aspects. CSR is a process catalyzed by the activation-induced cytidine deaminase (AID) enzyme, which deaminates deoxycytidines (dCs) generating high levels of deoxyuracils (dUs) that are processed by multiple base excision and mismatch repair events, resulting in the occurrence of DSBs (Xu et al. 2012b). During transcription-dependent R-loop formation in the S region, the high G content of the displaced DNA strand would favor G-quadruplex (G4) formation, thus stabilizing the R-loop (Duquette et al. 2004). It has also been argued that the high G content of the nascent RNA molecule may confer a structural difficulty on its assembly into a proper mRNP, enhancing their options to back hybridize with the DNA template and form an R-loop (Gomez-Gonzalez and Aguilera 2007). Finally, in contrast to TAR (see below), there is no evidence so far that CSR requires replication. It is also worth noting that AID can induce DSB-mediated translocations like those responsible for Burkitt’s lymphoma, with breakpoints in the Ig S region and a G-rich region of c-MYC (Ramiro et al. 2004; Robbiani et al. 2008; Klein et al. 2011), a phenomenon that has also been reproduced in yeast THO mutants (Ruiz et al. 2011). Therefore, although CSR is a paradigmatic case in which cotranscriptional R-loops become a key substrate for the origin of instability, it is not a model system to understand TAR as a major manifestation of the cross talk between transcription and HR.

Figure 2.

CSR. Transcription of switch (S) regions of the Ig genes generates R-loops, providing ssDNA substrates for the activation-induced cytidine deaminase (AID). Subsequent sequential enzymatic activities of different DNA repair pathways, including base excision and mismatch repair, would lead to DSBs that, by NHEJ, would complete the CSR event.

REPLICATION–TRANSCRIPTION CONFLICTS AS A SOURCE OF TAR

The way replication–transcription conflicts are resolved has been a main question in biology since it became clear that most genes are disposed codirectionally with replication along the E. coli chromosome (Ellwood and Nomura 1982). French (1992) reported that the movement of the replication fork was reduced in the ribosomal operon of E. coli as a result of its high transcriptional activity. Using an in vitro reconstituted T4 phage replication system, the laboratory of B. Alberts showed that, whereas the bacteriophage replication machinery could overtake a codirectionally moving E. coli RNAP without major consequences, a stalled RNAP caused the pausing of a head-on colliding DNA polymerase (Liu et al. 1993; Liu and Alberts 1995). In the Φ29 bacteriophage, the replication fork also stalled when the RNAP was arrested in a head-on orientation (Elias-Arnanz and Salas 1999). In E. coli, the replication machinery was able to displace a codirectionally elongating RNAP, but not an oncoming RNAP (Pomerantz and O’Donnell 2008). In the first case, the transient RNA–DNA hybrid formed during transcription could serve to reinitiate replication (Pomerantz and O’Donnell 2010). The ability of transcription to pause or stall replication in eukaryotes was first shown by 2D-gel electrophoresis at RNAPIII-driven transcription in transfer RNA (tRNA) genes (Deshpande and Newlon 1996). RNAPII transcription of a GAL fusion gene has a stronger effect on head-on replication than codirectional replication encounters (Prado and Aguilera 2005). Genome-wide evidence that transcription supposes an obstacle for the progression of forks, regardless of orientation, has been provided in yeast by determining the distribution of the Rrm3 helicase, which works in association with the replication fork (Azvolinsky et al. 2009). This study does not distinguish whether collisions might be stronger or more persistent under codirectional versus head-on orientation, but it certainly indicates that in vivo transcription in either direction may suppose an obstacle to replication fork progression. These and other observations, such as that E. coli replication fork stop at a dG (deoxiguanosine)-dC tandem repeat in which, normally, the RNAP stalls (Krasilnikova et al. 1998), support the general conclusion that transcription has a negative impact on replication, especially when both processes occur in head-on orientation.

The identification of DNA helicases, such as T4 dda and gp41 (Bedinger et al. 1983; Liu and Alberts 1995), E. coli DinG, Rep, and UvrD (Boubakri et al. 2010), yeast Rrm3 (Azvolinsky et al. 2006, 2009), and human RecQL5 (Li et al. 2011), as well as the E. coli transcription-coupled repair factor Mfd (Pomerantz and O’Donnell 2010), as factors that help bypass head-on stalled RNA polymerases confirms that cells use a battery of proteins to either prevent or resolve putatively harmful collisions. Removal of these proteins has been shown to enhance recombination in a transcription-dependent manner, supporting the idea that at least one major mechanism of TAR is mediated by transcription–replication collisions.

Clearly, in vivo evidence that replication is linked to TAR was provided by the observation that transcription from an S-phase active promoter causes hyper-recombination in yeast, whereas this is not the case of transcription driven from G₁-specific promoters (Prado and Aguilera 2005). In such assays, the effect was stronger in head-on collisions systems as compared with codirectional ones and stimulated in the absence of the Rrm3 helicase. Accordingly, inversion of the E. coli rrn operon from a codirectional to a head-on disposition with respect to replication makes cell viability strictly dependent on the RecBC DSB repair protein (Boubakri et al. 2010). Evidence for the replication dependency of TAR in mammals has been provided in Chinese hamster cells, in which replication stress generated by thymidine, which slows down replication fork progression, enhances TAR at the HPRT gene (Gottipati et al. 2008). These and other data (Gaillard et al. 2013) suggest that recombination stimulated by transcription is a consequence of the difficulties of the replication fork to traverse through transcribed DNA regions. This phenomenon has also been observed in tRNA genes flanked by direct repeats (de la Loza et al. 2009), as well as in TAR mediated by R-loops. Thus, TAR in yeast THO complex mutants is only observed under S-phase transcription, and replication forks pause or stall in the DNA region in which a hot spot of RNA–DNA hybrid formation at the end of a lacZ gene has been identified (Huertas et al. 2006; Wellinger et al. 2006). Indeed, the ability of R-loops to negatively affect progression of replication forks has been confirmed in yeast, C. elegans, and bacteria (Gan et al. 2011; Castellano-Pozo et al. 2012; Santos-Pereira et al. 2013). Therefore, it seems that regardless of the type of TAR detected, whether or not stimulated in mRNP biogenesis and export mutants or mediated by R-loops, its origin is linked to a defective replication progression.

THE rDNA REGION AS A PARADIGM OF TAR EVENTS

TAR takes place in the rDNA region, which is organized in tandem repeats in most organisms to ensure the maintenance of its integrity. Interestingly, many features underlying TAR are recapitulated in the rDNA, such as the dependency on topoisomerase activities to restrain recombination and R-loop formation (Kim and Wang 1989; El Hage et al. 2010). In eukaryotes, DNA elements with replication fork barrier properties (RFBs) are found in rDNA regions, as first shown in yeast (Brewer and Fangman 1988). They associate with specialized proteins to exert a programmed polar pausing of replication, thus avoiding frontal collisions with the transcriptional machinery (Tsang and Carr 2008). Structures with RFB capacity are also found at other genomic locations, including replication termination sites in E. coli (Khatri et al. 1989; Hill and Marians 1990), the mat locus in S. pombe (Dalgaard and Klar 2001), or transposons (Zaratiegui et al. 2011), indicating that programmed fork block to avoid frontal encounters with transcriptional machineries is a broadly used cellular strategy to avoid genomic instability.

In addition to the RFB, another important mechanism contributing to rDNA stability involves the regulated recruitment of cohesin, a multifunctional protein complex that ensures sister chromatid cohesion during replication and at sites of DSB, thus contributing to prevent genomic instability by favoring equal recombination events (Nasmyth 2011; Dorsett and Strom 2012). In the yeast S. cerevisiae, only about half of the rDNA repeats are actively transcribed, and repressed rDNA copies have been shown to be important for sister chromatid cohesion (Ide et al. 2010). rDNA silencing is mediated by the Sir2 protein and appears to control transcription from E-pro, a bidirectional RNAPII promoter located adjacent to the RFB (Santangelo et al. 1988; Ganley et al. 2005). Interestingly, constitutive transcription from E-pro results in dissociation of cohesin from the rDNA spacer, and the proportion of unequal recombination events was shown to increase in a conditional cohesin mutant, leading to an increase in rDNA repeats expansion, marker loss, and extrachromosomal rDNA circle formation (Kobayashi et al. 2004; Kobayashi and Ganley 2005). Hence, the activation of E-pro transcription has been shown to lead to cohesin dissociation at the rDNA. Similarly, transcription activation leads to cohesin delocalization at other loci in yeast (Lengronne et al. 2004; Bausch et al. 2007). It is thus tempting to speculate that transcription-dependent changes of sister chromatid cohesion may also contribute to TAR. Indeed, the hyper-recombination associated with cohesion mutations has been observed at different loci (De Piccoli et al. 2006; Alvaro et al. 2007).

TELOMERES AS A PUTATIVE REGION PRONE TO TERRA-MEDIATED TAR

The ends of eukaryotic chromosomes not only pose the “end-replication” problem, solved in yeast and proliferative tissues by the action of the telomerase, but they also need to be protected by the association with specialized proteins and formation of a lariat structure, the T-loop, created by the invasion of the 3′ ssDNA end into the duplex region to prevent the activation of the DNA damage response (de Lange 2009; O’Sullivan and Karlseder 2010). Replication is challenged at telomeres, which, in this sense, resemble fragile sites (Sfeir et al. 2009). Although telomeres are heterochromatin structures, they are transcribed by RNAPII into telomeric repeat-containing RNA (TERRA) molecules in several eukaryotes from yeast to mammals (Feuerhahn et al. 2010; Bah and Azzalin 2012). TERRA might interfere with replication, either through collision between the transcription and replication machineries or because TERRA might form R-loops with telomeric DNA, which could be further stabilized by G4-DNA structures in the displaced strand or G4-hybrid structures. Evidence in support of this idea includes the observation that a fraction of TERRA remains associated with telomeric chromatin (Azzalin et al. 2007; Schoeftner and Blasco 2008; Deng et al. 2009) and dysfunction of the Rat1 ribonuclease leads to TERRA accumulation and telomere elongation defects in S. cerevisiae, both phenotypes being suppressed by RNase H overexpression (Luke et al. 2008). Furthermore, both telomeric DNA and TERRA adopt G4-DNA structures (Phan 2010; Xu et al. 2010; Lu et al. 2013), and RNA–DNA hybrid G4 structures may represent a telomere component according to recent modeling studies (Xu et al. 2012a). The existence of telomeric R-loops has been recently shown in budding yeast in two independent studies. Interestingly, those telomeric R-loops were significantly increased in THO mutants, which suffer from short telomeres (Pfeiffer et al. 2013), indicating that proper TERRA biogenesis is important for telomere maintenance. Noticeably, THO subunits were identified at human telomeres as well (Grolimund et al. 2013). On the other hand, telomeric R-loop accumulation was shown to promote recombination-mediated elongation events in the absence of telomerase (Balk et al. 2013), thus being an important player in telomere dynamics. Further work will be required to dissect the role of telomeric R-loops in the regulation of telomere function.

FROM TRANSCRIPTION–REPLICATION CONFLICTS TO RECOMBINATION

The concept that the stalling of the replication fork by transcription, whether or not mediated by the formation of R-loops, is a major cause of TAR seems to be generally accepted. However, the mechanism by which a stalled replication fork can lead to the DSBs or ssDNA gaps responsible for recombination remains unclear. In principle, different mutually nonexclusive options could explain this phenomenon, of which we review three.

First, as discussed above, the accumulation of negative supercoiled DNA during transcription may increase its susceptibility to endogenous genotoxic agents. The resulting DNA lesions, unless properly repaired, will constitute physical obstacles for the progression of the replication fork. As a consequence, an ssDNA stretch would be created in front of the lagging or leading DNA polymerases, and replication of the gapped region completed by recombinational template switching (Fig. 3A). In this case, although the recombination event would be replication-dependent, the transcription machinery does not need to be present in the DNA during replication, and a DSB may not necessarily be the initiation intermediate.

Figure 3.

Putative molecular intermediates and mechanisms responsible for TAR. (A) The negative supercoiled DNA accumulating behind the elongating transcription machinery would expose ssDNA segments that would be more susceptible to DNA damage. Unless repaired, such transcription-dependent DNA lesions could block the replicative DNA polymerase. Template switching, DSB repair (in this case, the original damage was a ssDNA nick that, after replication, is converted to a DSB), or fork reversal would be required to complete replication. Postreplicative repair by translesion synthesis polymerases is not drawn for simplification. (B) The transcription machinery can be a direct obstacle for the progression of the replication fork, whether or not mediated by a cotranscriptional R-loop. An R-loop, or any other kind of transcription-dependent obstacle, could block DNA synthesis so that template switching would be needed for replication completion. Similarly, a head-on transcription-replication collision might cause replication fork arrest and reversal. Alternatively, R-loop-primed reinitiation of DNA synthesis might occur. Replication fork arrest could also result in DSBs, whether or not catalyzed by endonucleases, that would use recombination either with the sister chromatid or homologous chromosome or an ectopically located homologous DNA sequence, which can result in a transcription-dependent hyper-recombination phenotype.

Second, during codirectional collisions, the replication fork hitting the back of the RNAP would probably pause or strongly reduce its speed, presumably without arrest in most cases. Once the elongating RNAP finishes its transcription cycle, replication could, in principle, resume normally. Alternatively, it is also possible that the replication fork coming from the other side finishes replication. However, the presence of a DNA lesion with the ability to block RNAP elongation or requiring RNAP backtracking to resume elongation (Dutta et al. 2011), or mutants defective in 3′-end RNA processing or transcription termination (Mischo et al. 2011), may make this codirectional fork pausing permanent, leading to a replication arrest. A similar situation could occur if an R-loop would have formed behind the RNAP, in which case the replication fork would hit the RNA:DNA hybrid directly, presumably causing fork stalling (Fig. 3B). Resumption of replication may then require fork breakage, whether or not mediated by a nuclease, replication restart thus occurring by a recombination-mediated mechanism. Alternatively, the hybrid might be used as primer to reinitiate DNA synthesis.

Third, during head-on collisions, replication arrest likely becomes inevitable. It remains unclear, however, whether or not a physical collision between the two polymerases takes place. The positive supercoiling generated between the two machineries, and its possible impact on nucleosome organization, may produce some signals in response to which cells could block replication fork progression avoiding physical collision (Fig. 3B). The observations that the human RECQL5 helicase interacts with RNAPII and functions in transcription repression (Aygun et al. 2008, 2009; Kassube et al. 2013) suggest that direct inhibition of transcription elongation might provide a means to prevent collisions. RecQL5-depleted mammalian cells suffer from strong increase in sister chromatid exchanges (Hu et al. 2005). Whether this phenotype is caused by the RNAPII-regulating activity of RECQL5, a consequence of a more direct role of RecQL5 in DNA metabolism (Hu et al. 2007), or a combination of both (Islam et al. 2010) remains yet an open question. It is thus possible that transcription may not behave just as a protein–DNA barrier to replication like natural or even artificial recombinogenic RFBs (Horiuchi and Fujimura 1995; Lambert et al. 2005; Jacome and Fernandez-Capetillo 2011). In this sense, genome-wide chromatin immuno-precipitation (ChIP)-chip analyses have revealed that Top2 accumulates in regions in which replication forks encounter highly transcribed genes (Bermejo et al. 2009). Also, a transcription-dependent effect of DNA topoisomerases on recombination between DNA repeats has been described (Garcia-Rubio and Aguilera 2012). Essentially, a DNA region attached to the nuclear periphery would hinder torsional stress spreading and, thus, further enhance the transcriptional barrier to replication. The DNA-damage checkpoint would release the DNA from the nuclear pore, providing the flexibility required to solve the topological constraints generated by the replication-transcription encounter (Bermejo et al. 2011). In this scenario, an R-loop would make the topological constraint more accused, as it would prevent the release of the topological stress generated by the progression of the fork, retaining a structural feature that would contribute further to replication arrest (Alzu et al. 2012).

An additional consequence of a head-on encounter could be the extrusion of the replication fork backward, generating a chicken-foot structure (Fig. 3B) (Postow et al. 2001; Olavarrieta et al. 2002). This structure, by itself, contains a dsDNA end that can initiate recombination or constitute the substrate for structure-selective endonucleases, such as Mus81-Mms4/MUS81-EME1, Slx1-Slx4/SLX1-SLX4, Yen1/GEN1, Rad2/XPG, etc., with the potential to initiate recombination events as potential factors involved in TAR.

CONCLUDING REMARKS

Despite many years of intensive research on TAR using a large battery of different systems, further genetic and molecular analyses are needed to identify the intermediates involved. For example, a detailed and comparative analysis of TAR using different substrates leading to potentially different recombination events would be important to evaluate the prevalence of the kind of initiation event (DSB or ssDNA gap), recombination event (reciprocal or nonreciprocal), and involved pathways (break-induced replication, single-strand annealing, synthesis-dependent strand annealing, or DSB repair) in each case. It is unclear whether spontaneous TAR observed in wild-type cells is similar to that observed in mRNP processing mutants or mediated by R-loops. Although a genome-wide analysis in yeast has shown that the sites of major accumulation of the Rrm3 helicase are the open reading frames of actively transcribed genes, implying that replication has difficulties in traversing such DNA regions (Azvolinsky et al. 2009), a similar study on the distribution of γ-H2AX signals, as a marker of DSBs, has shown that they accumulate in other regions including tRNAs and promoters of protein-encoding genes (Szilard et al. 2010). An explanation for this apparent discrepancy could be that transcription does not necessarily induce DSBs, which would be consistent with recent work performed in mammalian cells (Gottipati et al. 2008) or the proposal that most TAR events induced in yeast THO mutants, although being replication-dependent, would not occur via DSBs (Gomez-Gonzalez et al. 2009). Indeed, hpr1 mutants show a strong transcription-dependent hyper-recombination leading to deletions between direct repeats but a mild increase in other types of events. It has been proposed that a DNA polymerase stalled in front of an R-loop may switch to another homologous region and use it as a template to resume replication, possibly explaining why direct-repeat recombination is the most favored type of event in R-loop accumulating mutants (Fig. 3B) (Gomez-Gonzalez et al. 2009).

It seems, therefore, important to identify the molecular event initiating recombination at transcribed DNA that could explain the different TAR events studied so far—by analogy with CFSs, specific chromosomal sites showing breaks or constrictions in metaphase chromosomes (Durkin and Glover 2007), which have recently been shown to occur in very long genes located in low-replication initiation regions (Helmrich et al. 2011; Le Tallec et al. 2011; Letessier et al. 2011). An interesting and open question is the mechanism by which TAR could be induced in terminally differentiated cells. It could be that cotranscriptionally formed R-loops interfere with DNA-repair activities or transcriptional RNA-binding proteins might have a role in DNA repair itself. However, the emerging field involving small noncoding RNAs in the DNA damage response to DSBs (Wei et al. 2012; d’Adda di Fagagna 2013) points to novel connections between RNA and the maintenance of genome integrity that would need to be further investigated in relation to TAR. Up to which point the nature of the transcriptional intermediates determine the outcome of recombination remains to be seen, but, no doubt, the more we know about the type of event involved, the more we will learn about the molecular nature of the cross talk between transcription and recombination and their connection to replication. This is of major interest in modern genetics and molecular biology not only for the increasing relevance of transcription in genome dynamics and, as recently proposed, accelerated evolution (Paul et al. 2013), but also because of the importance of genome integrity maintenance as a mechanism to prevent cancer and other genetic diseases.