AID Induces Double-Strand Breaks at Immunoglobulin Switch Regions and c-MYC Causing Chromosomal Translocations in Yeast THO Mutants

Transcription of the switch (S) regions of immunoglobulin genes in B cells generates stable R-loops that are targeted by Activation Induced Cytidine Deaminase (AID), triggering class switch recombination (CSR), as well as translocations with c-MYC responsible for Burkitt's lymphomas. In Saccharomyces cerevisiae, stable R-loops are formed co-transcriptionally in mutants of THO, a conserved nuclear complex involved in mRNP biogenesis. Such R-loops trigger genome instability and facilitate deamination by human AID. To understand the mechanisms that generate genome instability mediated by mRNP biogenesis impairment and by AID, we devised a yeast chromosomal system based on different segments of mammalian S regions and c-MYC for the analysis of chromosomal rearrangements in both wild-type and THO mutants. We demonstrate that AID acts in yeast at heterologous S and c-MYC transcribed sequences leading to double-strand breaks (DSBs) which in turn cause chromosomal translocations via Non-Homologous End Joining (NHEJ). AID–induced translocations were strongly enhanced in yeast THO null mutants, consistent with the idea that AID–mediated DSBs depend on R-loop formation. Our study not only provides new clues to understand the role of mRNP biogenesis in preventing genome rearrangements and the mechanism of AID-mediated genome instability, but also shows that, once uracil residues are produced by AID–mediated deamination, these are processed into DSBs and chromosomal rearrangements by the general and conserved DNA repair functions present from yeast to human cells.


Introduction
Transcription is a source of genetic instability in eukaryotic cells (reviewed in [1]). High transcription rates can result in increased recombination and mutagenesis [2,3], and a variety of factors have an influence on this relationship, including the collision between replication forks and the transcription machinery [4], and the occurrence of RNA-DNA hybrids (R-loops) [5,6]. In mammalian B cells, class switch recombination (CSR) and somatic hypermutation (SHM) are strongly dependent on transcription. Transcription of the switch (S) regions generates DNA intermediates in which the C-rich template strand forms stable R-loops and the non-template G-rich strand can form secondary structures (reviewed in [7]). Activation Induced Deaminase (AID), an enzyme specifically expressed in mature B cells [8] that is essential in both CSR and SHM, appears to be associated with the transcription machinery [9][10][11], and AID-mediated mutations increase when RNAPII is stalled [12]. Current models of CSR suggest that AID triggers the process by deaminating cytosines in the non-template DNA strand displaced by the R-loop [13][14][15]. In addition, during transcription the DNA sequences upstream of the elongating RNA polymerase are negatively supercoiled, and this transient change in DNA topology may allow AID to access both DNA strands [16]. Cytosine residues deaminated by AID at S regions are converted into uracils, which are processed by base excision repair (BER) and/or mismatch repair (MMR) pathways into DSBs responsible for CSR [17][18][19]. However, the reason why in B cells uracils lead to DSBs instead of being repaired remains unclear, although a breakdown of the cellular protective error-free repair has been suggested [20]. An open question is whether B cells have specialized functions or DNA modifying enzymes not present in other eukaryotic cells to process uracils into DSBs.
The same mechanism leading to CSR seems to be involved in the formation of chromosomal translocations (CTs) between the immunoglobulin S regions and c-MYC (Ig/myc) that are associated with B cell-derived Burkitt's lymphomas, also mediated by AID [21][22][23]. CTs are indeed a hallmark of cancer cells [24,25], and are mediated by DSBs [26]. Repair of AID-mediated DSBs occurs via Non-Homologous End Joining (NHEJ) [27], but in the presence of sequence homology, CTs may also occur by homologous recombination [28][29][30][31][32].
R-loops have been shown to accumulate in yeast THO mutants. The THO complex is functionally involved in transcription elongation and mRNP biogenesis [33], and depletion of some of its components generates strong transcription impairment and hyperrecombination phenotypes, which are linked to the formation of R-loops [5]. These phenotypes depend on gene length, GC content and transcription levels [34,35]. Importantly, THO depletion enhances AID-dependent mutations as a consequence of its ability to accumulate R-loops [36]. Taking advantage of this ability, we developed a yeast genetic system to study the mechanisms of AID-mediated chromosomal rearrangements. With this system we observed that transcription of S and c-MYC sequences in yeast generates AID-mediated DSBs that can be joined with an ectopic HO-induced DSB to yield a reciprocal CT. AID-induced CTs were strongly enhanced in yeast THO null mutants but not in hpr1-101 mutants, which do not form R-loops, consistent with the idea that AID-mediated DSBs depend on Rloop formation. CTs analyzed took place through NHEJ, confirming that the initial event was a DSB. This study not only provides new clues to understanding the role of mRNP biogenesis in preventing genome rearrangements and the mechanism of AIDmediated genome instability, but it shows that yeast cells, as mammalian B cells, have the factors and pathways required to channel AID-mediated DNA lesions into DSBs responsible for NHEJ-dependent genome rearrangements and CSR.

Murine S sequences are hotspots of recombination in yeast THO null mutants overexpressing human AID
We developed different genetic systems to analyze how R-loops initiate genome instability, whether or not via DSBs. As S regions are known to form R-loops in B cells both in vivo and in vitro [6,[37][38][39], which in turn lead to DSBs responsible for CSR, we assayed the ability of these regions to induce recombination between repeats in yeast. We used a direct-repeat recombination assay based on a plasmid with two truncated GFP direct repeats sharing 200-bp of homology under control of the GAL1 promoter [36] ( Figure 1). Different S region fragments, which are characterized by a specific repetitive motif and length, were inserted between the repeats, and recombinants were scored as GFP+ cells by FACS among total cells. We analyzed sequences derived from murine Sm and Sc3 regions, which were 350-bp and 600-bp long, respectively (Sm350 and Sc600) [39], a 1050-bp longer version of Sm carrying three consecutive repeats of Sm350 (Sm1050), and the Sm350 in reversal orientation (Sm350rev). None of these sequences yielded a significant increase of recombination compared with the control carrying no S sequences in wild-type cells ( Figure 1). We next tested whether AID had the ability to stimulate recombination at S regions, given its preference to act on ssDNA within R-loops [13,14]. AID expression did not significantly increase the recombination frequency of the different sequences analyzed in wild-type cells ( Figure 1). We then analyzed the effect of the mft1D THO mutation, which induces R-loop formation but do not compromise viability [35]. Whereas the frequency of GFP+ recombinants in mft1D cells not expressing AID was the same as in the wild-type for the different S regions tested, AID overexpression significantly increased it 3-6-fold above wild-type levels ( Figure 1). These results suggest that the S fragments used here are too short to form R-loops capable of causing detectable hyper-recombination. However, they are able to facilitate the action of AID. As the highest recombination frequency was obtained with the Sm350 sequence (6.1610 23 ), with little differences to those of Sm1050 and Sm350rev (Figure 1), we selected Sm350 for further analyses as the better candidate for the analysis of AID action in vivo.

High levels of reciprocal CTs in yeast THO null mutants overexpressing AID
In order to unequivocally assay the capacity of R-loops to form DSBs, we developed a yeast system for the detection of reciprocal CTs that should occur via DSBs in the absence of DNA homology. The system is based on two non-homologous halves of the LEU2 gene (leu2D59 and leu2D39) integrated at chromosomes XV and III, respectively ( Figure 2A). Upstream to the leu2D59 at chromosome XV, an HO endonuclease cut site fused to the hygromycin gene and the 246-bp 39-end of the ACT1 intron were integrated (HO-

Author Summary
Mammalian B cells have developed complex processes to create genetic diversity from DNA double-strand breaks (DSBs). These must be tightly controlled to avoid harmful chromosomal translocations. Here we report an experimental yeast assay to analyze how the B-cell specific Activation Induced Deaminase (AID) induces transcriptiondependent DSBs in mammalian DNA sequences. Our data suggest that in yeast AID is able to mediate deamination of cytosines in transcribed DNA that are then channeled into DSBs as it occurs in mammalian B cells, leading finally to reciprocal chromosomal translocations. These events are strongly enhanced in THO yeast mutants, which indicates that the impairment of the mRNP biogenesis may generate the appropriate substrates for AID action. Our study demonstrates that the AID-dependent genomic instability mechanisms are mediated by standard DNA repair functions existing from yeast to human cells. The only requirement for these events to occur is the formation of the appropriate substrates for AID action, as they are the transcription-mediated RNA-DNA hybrids known to be accumulated in THO null mutants. This experimental model provides a useful tool for the study of the sequences and the mechanisms leading to genomic instability that are primarily caused by chromosomal rearrangements. ACT1iD5'-leu2D5'). The leu2D39 sequence, under control of the GAL1 promoter, followed by the remaining 62-bp of the ACT1 intron and the Sm350 sequence (leu2D39-ACT1iD39-Sm350) was integrated at the right arm of chromosome III together with the URA3 gene ( Figure 2A). As a control we used a similar system carrying no Sm350 sequence (leu2D39-ACT1iD39). These genetically engineered chromosomes III and XV were made in haploid MATa-inc cells containing the HO endonuclease gene under the GAL1 promoter, and the endogenous LEU2 gene and the ACT1 intron deleted from their chromosomal loci. In galactose, the HO endonuclease is activated causing a DSB at the HO-ACT1iD5'-leu2D5' construct on chromosome XV, whereas the leu2D39-ACT1iD39-Sm350 on chromosome III is heavily transcribed from the GAL1 promoter. Therefore, the HO-induced DSBs on chromosome XV could be sealed with a transcription-mediated DSB on chromosome III to lead a Leu+ translocations ( Figure 2A). These translocations would be expected to contain the breakpoints embedded in a functional ACT1 intron located between the two LEU2 gene halves ( Figure 2A).
After 24h of growth in galactose, Leu+ translocants were obtained at a frequency of 6610 -9 in wild-type cells ( Figure 2B), which indicated that transcription through leu2D39-ACT1iD39-Sm350 was indeed able to generate DSBs that were joined to the HO-induced DSB on chromosome XV. This Leu+ frequency was unaffected by overexpression of AID ( Figure 2B). We then analyzed the effect of a THO null mutation, in this case hpr1D for its known strong effect in transcription [40]. Whereas the frequency of translocations in hpr1D was the same as in wild-type cells not expressing AID (6.1610 29 ), it increased 19-fold when AID was overexpressed ( Figure 2B). This effect was specific for the system containing the S sequence (leu2D39-ACT1iD39-Sm350), as in the absence of S sequences translocations were not detectable (,10 210 ; Figure 2B). As it has been reported that secondary structures do not form in vitro when the sequence is inverted and the G-rich strand is transcribed [39], we also analyzed the effect of the inverted S sequence (Sm350rev) in our translocation assay, for which we constructed the corresponding leu2D39-ACT1iD39-Sm350rev on chromosome III. Interestingly, Sm350rev also increased the frequency of translocations in hpr1D cells overexpressing AID respect to wild-type levels, although it reached lower levels than the forward Sm350 sequence ( Figure S1).
Consistent with previous analyses of direct-repeats recombination (Figure 1), the S sequences used here seem to be too short as to form sufficient R-loops to give a significant increase of CTs in wild-type cells with AID or in hpr1D cells without AID, but they stimulated recombination in hpr1D cells expressing AID. Altogether, these results indicate that transcription through S regions in yeast can lead to DSBs able to generate CTs. The specific increase caused by AID overexpression in THO mutants suggests that at least a proportion of such DSBs are mediated by R-loops.

AID-mediated translocation breakpoints occur at the S regions
To assess whether Leu+ events corresponded to the expected translocations, chromosomes from twenty independent Leu+ recombinants obtained from hpr1D yeast overexpressing AID were analyzed by pulsed-field gel electrophoresis (PFGE). Ethidium bromide staining and Southern analysis with a LEU2 specific probe revealed that all translocated chromosomes showed the pattern expected from the reciprocal joining of the DSBs generated at the S region and HO site in chromosomes III and XV, respectively ( Figure 2C and Figure S2). This reciprocal translocation resulted in two new chromosomes of 596-and 811kb long, respectively, with the LEU2 signal specifically detected in the smaller one, the chromosome in which the two halves of the LEU2 gene are joined ( Figure 2C and Figure S2). Concomitantly, parental chromosomes III and XV disappeared. This indicates that translocations in hpr1D yeast overexpressing AID show a unique and recognizable structure originated by a DSB at the S region.
To map the CT breakpoints we amplified the ACT1 intron reconstituted within the LEU2 gene by PCR using specific primers that anneal at the corresponding sequences involved in the translocations on chromosomes III and XV ( Figure 3A). This amplification generated a variety of DNA fragments longer than the 308-bp of the natural ACT1 intron ( Figure 3A), suggesting a non site-specific distribution of the transcription-mediated DSBs along the Sm350 region. DSBs appeared mainly in the 59-half of the Sm350 region, resulting in 13 to 122-nt long S sequence stretches within the reconstituted ACT1 intron of the translocated chromosomes ( Figure 3B), which was similar regardless of AID expression ( Figure 3B and 3C). Analysis of the junctions by sequencing of the PCR products revealed frequent short microhomologies (1-2-nt) between the Sm and HO breakpoints ( Figure 3C), and the lost of a few nucleotides (0-3-nt) from the HO cut site in the end-joining process ( Figure 3C). These data suggest that DSBs are generated by AID deamination at hpr1D-mediated R-loops in the transcribed Sm350 regions.
AID does not stimulate DSBs in hpr1-101 mutants not forming R-loops Depletion of specific mRNP biogenesis factors, such as THO, generates a transcription elongation impairment that is linked to the accumulation and/or stabilization of R-loops [5]. Accordingly, we found that transcription of the Sm350 sequence in hpr1D yeast reduced the efficiency of the transcription process in our system ( Figure S3). In order to determine whether R-loops formed in hpr1D cells are required for AID-dependent DSBs and CTs, we analyzed CTs in the hpr1-101 mutant strain, which has been shown not to accumulate stable R-loops during transcription [41]. Interestingly, hpr1-101 cells overexpressing AID showed a strong decrease in CT formation after galactose induction (Figure 4), suggesting that the formation of R-loops during transcription of the Sm region contributes to the generation of the DSBs responsible for CTs in this system.

AID-induced translocations in yeast THO null mutants are mediated by Ung1 and NHEJ
The current model proposed for the formation of AIDdependent DSBs suggests that uracils generated after AIDmediated deamination of cytosines are primarily processed by the uracil glycosylase Ung1 leading to abasic sites [19]. Subsequent processing of these abasic sites generates the DSBs that are repaired by NHEJ during CSR in mammalian B cells. To assay whether AID-mediated DSBs generated in our yeast system were processed as in B cells, we determined the effect of both were stained with ethidium bromide (left) and analyzed by Southern using a radiolabeled LEU2 probe (right). The electrophoretic mobility of natural yeast chromosomes is indicated. The LEU2 signal in parental cells corresponds to the two non-homologous halves of LEU2 integrated in chromosomes III and XV (bold). Chromosomes XV and VII have the same electrophoretic mobility in the experimental conditions used here. The translocated chromosomes containing LEU2 (tIII/XV) or not containing it (tXV/III), are marked with black triangles, and their sizes are indicated. doi:10.1371/journal.pgen.1002009.g002 UNG1 and DNL4 deletions on CT formation. As can be seen in Figure 4, both ung1D and dnl4D abolished the capacity of AID to induce DSB-mediated translocations in hpr1D cells. Therefore, as occurs in mammalian B cells, AID-dependent DSBs in yeast are also mediated by Ung1 and repaired by NHEJ. These results validate this yeast chromosomal translocation assay as a model system to study the mechanism by which AID can induce CTs in vivo. It reveals that once cytosine deamination is produced, further steps leading to DSBs and rearrangements in B cells may occur by standard and conserved DNA repair functions present from yeast to human cells, and not by B cell specific DNA modifying enzymes.

The c-MYC oncogene is a recombination hotspot in yeast THO null mutants
The observation that the S regions used in this study become a hotspot for DSBs only in R-loop-forming hpr1D cells, and by the action of AID, suggests that the sequences used are not long enough to harbor the same structural features as full natural S regions in B cells. In order to test this hypothesis, one possibility was to enlarge the size of the S sequence cloned into our yeast systems. The use of such sequences with high abundance of internal repeats, however, makes them particularly unstable, as deletions between the repeats by single-strand annealing occur at high frequency [42,43], a problem accentuated in hpr1D cells. Consequently, we used as an equivalent control the c-MYC oncogene. As the S regions of Ig genes, c-MYC is characterized by its GC-richness, especially in the 59 region that includes the first exon and first intron (up to 60% of GC residues). This region forms co-transcriptional secondary structures during transcription in vitro [44], and is frequently hypermutated and involved in reciprocal translocations with the S regions [22]. The recent identification of c-MYC as a main target of AID-dependent translocations and the demonstration that nuclease-mediated DSBs at c-MYC causes similar translocations [23] suggest that c-MYC carries the same structural features that make S regions highly recombinogenic during transcription in B cells.
Therefore, we cloned a 3.8-kb fragment of the human c-MYC gene that included the first exon and the first intron in our GFP direct-repeat recombination assay, as a functional and structural equivalent of a long S region. As can be seen in Figure 5A, whereas recombination was not affected by the presence of the MYC sequence in wild-type cells, regardless of AID expression, a strong increase was seen in hpr1D cells both without and with AID expression (24-fold and 32-fold increase with respect to the control wild-type, respectively; Figure 5A). To test whether these high recombination levels, notably higher than those obtained with the Sm350 sequence (compare Figure 1 and Figure 5A), were not just  due to the presence of a longer DNA sequence between the GFP repeats, we tested the effect of the long (4.1 kb) and low GC-rich (40%) sequence of the LYS2 gene. Recombination levels in this system containing the LYS2 sequence dropped markedly ( Figure 5A), indicating that the high recombinogenic behavior of c-MYC in this system in hpr1D was due to both its high GC-content (as compared with LYS2) and length (as compared with Sm350). Therefore, c-MYC behaves as a hotspot of recombination in hpr1D cells that is slightly increased by the action of AID.
To determine whether the recombinogenic behavior of the c-MYC sequence in hpr1D was due to an accumulation of DSBs, we next tested its effect on the chromosomal translocation assay ( Figure 5B). For this test, a 2.5-kb c-MYC 59-end fragment, including the first exon and the first intron, was integrated under the inducible GAL1 promoter at chromosome III, instead of the Sm350 sequence (leu2D39-ACT1iD39-cMYC). Interestingly, after galactose induction, Leu+ translocants were recovered in wild-type cells more efficiently in the system with c-MYC than with Sm350 (2- fold and 4-fold increase without and with AID expression, respectively; compare Figure 2B and Figure 5B), suggesting that DSBs are spontaneously induced during transcription of the c-MYC sequences, and slightly enhanced by AID. Slightly higher translocation frequencies were obtained in hpr1D cells (3-fold and 2-fold increase with respect to wild-type cells without or with AID expression, respectively; Figure 5B). This translocation frequency, however, was 3-fold lower than that induced by the Sm350 in hpr1D cells expressing AID (compare Figure 2B and Figure 5B). We mapped the translocation breakpoints by PCR and sequencing as previously described. PCR of the breakpoint fragments rendered heterogeneous sizes that were longer than the 308-bp full-length ACT1 intron (not shown). Breakpoints occurred mainly in the 59-region of c-MYC, resulting in a reconstituted ACT1 intron slightly longer than that obtained with S sequences (from 38 to 241-nt long; Figure 5C). The length of c-MYC stretches within the reconstituted ACT1 intron was the same regardless of whether translocations were obtained with or without AID overexpression ( Figure 5C). Sequencing of the translocation junctions revealed also frequent microhomologies between the c-MYC and the HO cut site ends ( Figure 5C), which were longer (1-5 nt) than those obtained with Sm350, regardless of AID expression. The number of nucleotides lost from the HO cut site in the end-joining event was similar to that of Sm350 ( Figure 5C).
These data suggest that c-MYC behaves as an intrinsic recombination hotspot in yeast THO mutants forming R-loops. Even though AID induce more recombination and DSBs at c-MYC, it only causes a slight enhancement of c-MYC recombination hotspot activity. This suggests that c-MYC can be targeted by recombinogenic DNA breaks in AID-dependent and independent manners in THO mutants, suggesting that a suboptimal mRNP structure is sufficient for c-MYC to behave as a fragile site.

Discussion
We show here that AID induces DSBs in the immunoglobulin S regions and in the c-MYC sequences in yeast THO mutants, which are impaired in mRNP biogenesis. Our data suggest that AIDmediated deamination of cytosines at S sequences in yeast are channeled by the Ung1 activity into DSBs as occurs in mammalian B cells, which are then repaired by NHEJ. The c-MYC oncogene, which shares features of S sequences, also acts as a hotspot of recombination in yeast THO mutants. Using a novel translocation assay we show that translocations involving S or c-MYC sequences occur efficiently in yeast THO mutants, but poorly in wild-type cells. This study demonstrates that the mechanisms of AID-mediated CSR and AID-dependent translocations are mediated by standard DNA repair functions existing in yeast and human cells. The only requirement for these events to occur is the formation of the appropriate substrates for AID action, as they are the R-loops known to be accumulated in THO null mutants.
It is widely accepted that transcription of S regions plays a primary role in CSR in B cells by generating R-loops that are substrates for efficient AID action [7]. However, once cytosines at the single-stranded DNA sequence of the R-loop are converted into uracils by AID deamination, it is unclear whether B cells have specific DNA modification enzymes or pathways to channel such uracils into DSBs responsible for CSR or Ig/myc translocations instead of being efficiently repaired by either BER or MMR, although a breakdown of the protective accurate repair pathways during the process of B cell maturation has been suggested [20]. Our yeast assay for the detection of translocations generated by the NHEJ-mediated processing of DSBs, reveals that yeast cells also have the capability of processing AID-mediated lesions into DSBs. We found that transcription through mammalian Sm regions in yeast generates AID-mediated DSBs that can lead to CTs, similarly to B cells. The breakpoints of these translocations are located at the 5' end of Sm sequences, where most CSR junctions occur in vivo [45], and show short nucleotide microhomologies, consistent with data from mammalian B cells [46]. Notably, AID-dependent CTs in our assay occur at a low frequency in wild-type cells, but are synergistically enhanced in yeast THO null mutants, which are affected in mRNP biogenesis. This is in agreement with previous results showing the mutagenic and recombinogenic phenotype of yeast overexpressing AID [47], and that the transcription impairment in yeast THO null mutants enhances AID-dependent mutation of the non-template DNA strand, consistent with its exposure as a single stranded DNA during the transcription elongation in the form of R-loops [36]. These results, and the fact that AID-dependent CTs are poorly observed in hpr1-101 mutants, which do not form R-loops [41], suggest that the formation of a suboptimal structure of the ribonucleoparticle around the S region RNA fragment may be a major contributor to the formation and/or stabilization of a Rloop as an optimal substrate for AID action. Therefore, our data open the possibility that mRNP biogenesis may have an impact on the generation of genetic variability in mammalian B cells by controlling the formation of R-loops as an efficient substrate for the action of AID.
It has been reported that CSR is orientation-dependent and that inversion of the S sequences decreases the efficiency of CSR [38,45], as it happens with the formation of sequence-dependent secondary structures in DNA during transcription in vitro [39]. We found in our assays a lesser, although substantial, recombinogenic effect of the Sm sequence in the reversal orientation, which suggests that in our yeast system the presence of G-rich non-transcribed sequences is not an unique requisite for recombinogenicity. The size of the S region seems important for the recombination hotspot activity in B cells, as longer S regions could generate more extensive R-loops [48]. A slight increase of the size of the Sm sequence in our yeast assay did not enhance the translocation frequency, because it was still below a minimum size. Notably, the long GC-rich c-MYC oncogene, used in this study as a structural equivalent of the longer S region, was a hotspot for recombination between direct-repeats and NHEJ-mediated CTs in wild-type yeast in our assays. Genetic instability in c-MYC and other oncogenes has been correlated with the GC-richness [44,49], and the identification of c-MYC as a target of AID [23] also suggests that c-MYC shares structural features with S regions that make them highly recombinogenic during transcription in B cells. Interestingly, the recombinogenic potential generated by the transcription of the c-MYC sequence in our assays in hpr1D cells was also observed in the absence of AID expression. This is explained by the direct effect of hpr1D mutation observed in stability of long and GC-rich sequences such as c-MYC, as previously shown for bacterial DNA sequences [40], and by the lower density of AID-recognition motifs along this sequence. Altogether, our results confirm that c-MYC behaves as a natural hotspot of recombination, but this is more pronounced in THO depleted cells. The results suggest that the lower efficiency of mRNP biogenesis caused by THO mutation predispose c-MYC to breakage, leading to translocations. The similarity of phenotypes observed for mutations in the THSC/TREX-2 complex [50], which is located at the nuclear periphery in close contact with the nuclear pore complex and also causes transcription and RNAdependent hyper-recombination, suggests that AID action would enhance DSB formation at S and c-MYC sequences provoking NHEJ-mediated translocations also in other mRNP biogenesis mutants, THO mutants being a paradigmatic example.
In summary, our work points to the idea that the structure of the nascent mRNA molecule may be a major determinant in the formation of R-loops at the S regions in B cells in vivo. Accordingly, previous data demonstrate that the RNA transcript formed during transcription of S sequences in vitro can thread back onto the template DNA strand, leading to an R-loop [51,52]. Our model yeast system provides a validated assay to study the molecular mechanisms mediated by AID leading to DSBs at S regions and the c-MYC oncogene, such as those occurring in B cells responsible for class switching or the Ig/myc translocations that are associated with Burkitt's lymphomas, thus opening new perspectives in understanding AID-mediated rearrangements.

Yeast strains and plasmids
Yeast strains used were isogenic to W303, and are listed in Table S1. GFP direct-repeat system constructs and translocation systems were done as described [32,36]. The hpr1D, dnl4D and ung1D alleles were generated by PCR-based gene replacement and confirmed by PCR and Southern analysis. hpr1-101 and double mutants (hpr1D dnl4D, hpr1D ung1D) with the translocation system were obtained by genetic crosses.

Determination of recombination frequencies
Yeast strains were grown on rich YEPD plates. Recombination tests using the GFP direct-repeat system and translocation assays were done as previously described [32,36].

Physical analysis of translocations
Translocation analysis was performed by Pulsed Field Gel Electrophoresis (PFGE) and Southern analysis as described [32].

Chromosomal breakpoint analysis
Breakpoint regions were obtained by PCR amplification of genomic DNA extracted from yeast translocants using primers p1 (59-GTATGTTCTAGCGCTTGC) and p2 (59-CTAAACATA-TAATATAGCAACA), purified and sequenced. Figure S1 Comparative analysis of Leu+ chromosomal tranlocations in hpr1D cells overexpressing AID in the presence of Sm350 or Sm350 reverse sequences. The control wild-type condition is shown as a reference. Translocation frequencies were obtained as the median value of at least ten independent colonies for each strain and are indicated on top of each bar. Median values and the corresponding standard deviations are shown. Statistical analysis was performed with a Mann-Whitney test compared with hpr1D cells with the Sm350 sequence overexpressing AID. *** p,0.05. (TIF) Figure S2 Molecular karyotype of 12 independent Leu+ translocants from hpr1D cells overexpressing AID analyzed by PFGE and Southern blot. Agarose gel was stained with ethidium bromide (top) and analyzed by Southern using a radiolabeled LEU2 probe (bottom). The electrophoretic mobility of natural yeast chromosomes is indicated. The translocated chromosomes containing LEU2 (tIII/XV) or not containing it (tXV/III) are marked with black triangles, and their sizes are indicated. (TIF) Figure S3 Transcription of the Sm sequence in hpr1D yeast. A diagram of the expected transcriptional products is shown on top. Total RNA was extracted from both wild-type and hpr1D yeast at intervals after induction of the translocation system with galactose. Same quantity of RNA of each sample was used for Northern analysis. Hybridization was performed with a radiolabeled LEU2 probe. Methylen-blue staining of the blot showing the ribosomal RNA levels of the samples is shown at the bottom. (TIF)