Meiotic recombination in Saccharomyces cerevisiae is initiated by the creation of DNA double strand breaks (DSBs), an event requiring 10 recombination initiation proteins. Published data indicate that these 10 proteins form three main interaction subgroups [(Spo11-Rec102-Rec104-Ski8), (Rec114-Rec107-Mei4), and (Mre11-Rad50-Xrs2)], but certain components from each subgroup may also interact. Although several of the protein–protein interactions have been defined, the mechanism for DSB formation has been challenging to define. Using a variation of the approach pioneered by others, we have tethered 8 of the 10 initiation proteins to a recombination coldspot and discovered that in addition to Spo11, 6 others (Rec102, Rec104, Ski8, Rec114, Rec107, and Mei4) promote DSB formation at the coldspot, albeit with different frequencies. Of the 8 proteins tested, only Mre11 was unable to cause DSBs even though it binds to UASGAL at GAL2. Our results suggest there may be several ways that the recombination initiation proteins can associate to form a functional initiation complex that can create DSBs.
IN Saccharomyces cerevisiae, meiotic recombination is initiated by the formation of DNA double strand breaks (DSBs) (e.g., Keeney 2001). These initiation events do not occur randomly along the length of each chromosome; rather, some regions (hotspots) have a higher probability of DSB formation (Sun et al. 1989; Gerton et al. 2000). Hotspots are often found in promoter regions and other areas of open chromatin and are usually absent around centromeres and telomeres (Baudat and Nicolas 1997; Petes 2001; Blat et al. 2002). A genomewide microarray analysis suggested that hotspots may be preferentially located in large (tens of kilobases) chromosomal domains of higher GC content (Gerton et al. 2000); however, there appears to be no obvious hotspot consensus sequence (Haring et al. 2004).
Ten proteins [Spo11, Rec102, Rec104, Ski8/Rec103 (hereafter referred to as Ski8), Rec114, Mei4, Rec107/Mer2 (hereafter referred to as Rec107), Mre11, Rad50, and Xrs2] are essential to produce the DSBs that initiate meiotic recombination (reviewed in Keeney 2001). The meiotic phenotypes conferred by null mutations of the initiation genes are indistinguishable: reduced sporulation, no DSBs or recombination, disrupted synaptonemal complex, and inviable spores (e.g., Malone et al. 1991). Of the 10 recombination initiation proteins, only Spo11 has been assigned a biochemical function; it covalently attaches to the ends of the DSBs (Bergerat et al. 1997; Keeney et al. 1997).
The requirement for 10 proteins to start recombination suggests that at least some of them form a complex to make DSBs. Several analyses have demonstrated that three subcomplexes exist among the 10 recombination initiation proteins: the Spo11-subcomplex: Spo11, Ski8, Rec102, and Rec104 (Salem et al. 1999; Uetz et al. 2000; Kee and Keeney 2002; Jiao et al. 2003; Arora et al. 2004; Cheng et al. 2004; Kee et al. 2004); the Rec114-subcomplex: Rec114, Rec107, and Mei4 (Arora et al. 2004; Henderson et al. 2006; Li et al. 2006; Maleki et al. 2007; Sasanuma et al. 2008); and the MRX-subcomplex: Mre11, Rad50, and Xrs2 (Borde 2007). Some data suggest that components from different subcomplexes also interact (Arora et al. 2004); one interpretation is that the 10 proteins form a single holocomplex to create DSBs. However, the exact mechanism for how these 10 proteins lead to DSB formation has been challenging to define; the literature describing their interactions contains a number of results that appear inconsistent. One example of differing observations is the relationship between the Spo11-subcomplex and the Rec114-subcomplex (Prieler et al. 2005; Sasanuma et al. 2008; Wan et al. 2008). Chromatin immunoprecipitation indicated that Spo11 was not present at the ARE1 hotspot in the absence of REC114 (Prieler et al. 2005; Sasanuma et al. 2008) or phosphorylatable REC107 (Sasanuma et al. 2008). Consistent with this, Rec114 and Mei4 were present on chromatin (Maleki et al. 2007), and at hotspots (ChIP) (Sasanuma et al. 2008) in the absence of SPO11 (Sasanuma et al. 2008). However, Prieler et al. (2005) found that Spo11 binds hotspot DNA in both a rec107Δ and a mei4Δ strain. These articles come to different conclusions about the interactions of Spo11 and the Rec114-subcomplex.
Naturally occurring recombination hotspots [e.g., ARE1 (formerly YCR048W), ARG4, CYS3, and HIS2] have been useful to study the interactions of recombination initiation proteins with S. cerevisiae chromosomes (e.g., Lichten and Goldman 1995; Prieler et al. 2005; Sasanuma et al. 2007) and with each other. A complementary approach to the analysis of natural hotspots, attracting recombination proteins to a site that was previously cold, was devised by Peciña et al. (2002). They demonstrated that the Gal4 DNA binding (DB) domain fused to Spo11 efficiently promoted recombination initiation at UASGAL sites normally cold for recombination. For example, DSB formation at GAL2 increased from undetectable levels to ∼12% of total DNA (Peciña et al. 2002). The new Gal4DB-Spo11-dependent hotspot was also completely dependent on the presence of all nine of the other initiation proteins. Although Spo11 is the putative DSB catalyst, it alone was incapable of producing DSBs, even when brought to the DNA by the Gal4 DB domain. These results suggest that Spo11 was able to recruit all the other required proteins to the coldspot. Given the complex network of interactions among the recombination initiation proteins, we thought it was probable that some initiation proteins would be capable of inducing DSBs when tethered to a coldspot, while others would not. By comparing which proteins could make DSBs vs. those which could not, our goal was to gain a better understanding of how a functional initiation complex was assembled on the DNA.
We tethered 8 of the 10 recombination initiation proteins to UASGAL sites by using a slight modification of the Peciña et al. (2002) approach. We demonstrate that, in addition to Spo11, six other Gal4DB fusions (DB-Rec102, DB-Rec104, DB-Ski8, DB-Rec114, DB-Rec107, and DB-Mei4) are able to cause DSBs upstream of the naturally cold GAL2 gene. Proteins capable of stimulating initiation include components from the Spo11-subcomplex and the Rec114-subcomplex. We are unable to detect DSB formation when a protein from the MRX-subcomplex (Mre11) is tethered.
MATERIALS AND METHODS
The yeast strains used in these experiments were derived from the diploid RM96 (S288C background), whose genotype is listed in Table 1. The relevant genetic markers of all other strains derived from it are listed (Table 1). Recombination mutations were constructed by gene replacement using standard one-step gene replacement with PCR fragments amplified from the Research Genetics strain collection (Giaever et al. 2002). Diploids containing the RAD50 wild-type allele were used to test fusion constructs by complementation tests; diploids containing the rad50S mutation were used for the analysis of DSBs (see below).
All GAL4DB-HA-REC fusion constructs, where “REC” represents any of eight recombination initiation genes, were constructed with the YEplac195 2μ-vector (Gietz and Sugino 1988). These plasmids were constructed as follows: A REC104-containing fragment was removed from pAMG406 (Salem et al. 1999) using PvuII and religated to make pSJH17. A ScaI–PvuII fragment containing pADH1-GAL4DB-HA-REC104-tADH1 from pYM9D-1 was then ligated into the PvuII site of pSJH17 to make pSJH30 (Figure 1). pDRH2 was made by removing the NheI–SalI fragment containing REC104 from pSJH30. pDRH1, pDRH3, pDRH4, pDRH5, pDRH7, pDRH8, and pSJH23 were made by inserting a PCR fragment containing the coding region of the appropriate REC gene into the NheI and SalI sites pSJH30. The PCR primers were engineered such that the final PCR product contained an upstream NheI site and a downstream SalI site. pSJH22 was constructed similarly, except the final PCR product contained NheI and PstI restriction enzyme sites. pDRH6 was made by subcloning a KpnI–XbaI MEI4-containing fragment from pBluescript SK+ (Menees et al. 1992) into pRS316 (Sikorski and Hieter 1989). All fusion constructs were sequenced to verify that the construction was correct. All plasmids used in this study are listed in Table 2.
Complementation of fusion plasmids:
For each GAL4DB-HA-REC fusion, the appropriate Rec− diploid was separately transformed with one of three plasmids: the GAL4DB-HA vector (pDRH2), the wild-type (WT) REC plasmid, or the appropriate GAL4DB-HA-REC construct. Transformed diploids were grown on SC −Ura plates, replica plated to 2% acetate sporulation plates, and incubated at 30° for 4–5 days. Sporulation plates were then replicated to media diagnostic for gene conversion (SC −Leu, SC −Lys, and SC −Trp) and crossing over (SC +Can/Cyh). Both can1r and cyh2r are recessive drug resistant markers. In a normal meiosis, 25% of spores are resistant to both drugs because of random segregation in the reductional division. A proper reductional division is dependent on proper amounts of crossing over. Rec− initiation mutants have no crossing over so the frequency of doubly drug-resistant products is equal to the background mitotic recombination frequency (∼10−8). Complementation of the mutant Rec− defect restores meiotic recombination and proper meiotic segregation, resulting in increases of haploid products expressing both drug resistances. Cells from the sporulation plates were dissected to determine viability of the meiotic products and sporulation was counted (≥200 cells were counted for each of 2–4 independent diploids).
Analysis of DSBs:
To determine the frequency of DSB formation at ARE1 and GAL2, the rad50KI81 (Cao et al. 1990) (hereafter referred to as rad50S) mutation was used. rad50S∷ura3Δ strains were made by one-step gene replacement of haploid strains using a rad50S∷URA3 PCR fragment. Removal of the URA3 selectable marker was done by cotransforming a PCR fragment of wild-type DNA downstream of RAD50 with YEp426 (Ma et al. 1987). Transformants were selected by using SC −Lys media and rad50S strains lacking the URA3 selectable marker were identified by growth on 5-FOA. All rad50S transformants were verified by MMS testing, PCR testing, and Southern analysis as previously described (Mao-Draayer et al. 1996). Yeast meiotic DNA for analysis of DSBs was isolated as in Bullard et al. (1996) except with the following change: DNA back extractions were performed on all samples after the two phenol chloroform isoamyl (PCI) purifications. For the Southern analyses at GAL2, a ∼1-kb upstream probe was used (see results). As a control, Southerns were repeated for four DB-REC fusions using a smaller probe (∼0.4 kb) and the same results were obtained (data not shown). Imaging was performed using a Molecular Dynamics and/or Quantity One phosphorimager and all quantifications were performed on the 15-hr meiotic samples using Quantity One software. The 15-hr time point was used because DSBs plateau at that time; the amount of DSBs falls slightly at later times, presumably due to degradation. For GAL2 blots, the close proximity of A with B and C with D prevented quantification of individual bands, so we therefore quantified bands A and B together, and C and D together. All DSB frequencies are expressed as percentages relative to the total band intensity and were corrected for background. DSB percentages were also corrected for plasmid loss. For statistical analyses, the Student t-test was used.
Chromatin immunoprecipitation was performed utilizing a protocol modified slightly from (Goldfarb and Alani 2004; Prieler et al. 2005); all solutions were as described in these articles except for the protease inhibitor cocktail (1 mm ε-aminocaproic acid, 1.5 μm aprotinin, 2 mm benzamidine, 1 μm leupeptin, 2 μm pepstatin, and 1 mm PMSF). Samples were treated with a Fisher Scientific pulsing vortex mixer at 3000 rpm for 15 min at 4°. Lysates were sonicated with a model 150 Artek sonic dismembrator sonicator (50% output; 4 × 10-sec pulses) to fragment the chromatin. To 50 μl of whole cell extract (WCE), 200 μl of 1× TE, 1% SDS was added and samples stored at 4° until reversal of the crosslinks. For immunoprecipitation (IP), α-Mre11 antibody (Rockland no. 401-872) was added to WCE to a final concentration of 2.5 mg/ml and incubated overnight at 4°. IPs were incubated with Protein A magnetic beads (Dynal, Invitrogen) for 2 hr at 4°. Washes were as described in Prieler et al. (2005) and elutions were as in Goldfarb and Alani (2004). After crosslinks were reversed, DNA was purified with PCI extraction and ethanol precipitation. The final DNA pellets (WCE and IP DNA) were resuspended in 50 μl 1× TE.
Real-time PCR was performed using PP1 [216-bp product upstream of the GAL2 UASGAL sites (targeted hotspot)] and PP2 [159-bp product, ∼16 kb downstream of GAL2 at CSF1 (a coldspot)] and carried out on a Roche LightCycler480 using SYBR Green I Master mix (Roche). PCR primers were used at a final concentration of 20 pmol per primer. Primer sequences are: PP1F-ACGCGACAGTAAAAGCAGCA; PP1R-CAAGTTGAGTGCGGCTACCT; PP2F-ACGTGGTGTTCCATTGACACAC; PP2R-CGCTCGTTTCCATAGTAGCCAT. For immunoprecipitated DNAs, 1/50 of the final volume was used in the PCR reactions. For WCE DNAs, 1 μl from 1:100, 1:500, and 1:2500 dilutions was used to create standard curves. All final melting curves showed a single DNA product. The amount of precipitated (IP) DNA was normalized to WCE and the results are expressed as a ratio (Borde et al. 2004) of GAL2 UASGAL hotspot vs. CSF1 coldspot.
GAL4DB fusion constructs complement recΔ diploids:
To examine the ability of different initiation proteins to attract the others to a coldspot, we made fusion proteins with Gal4DB. All fusions placed the GAL4 DNA binding (DB) domain at the amino-terminal end of the eight recombination initiation genes (SPO11, REC102, REC104, SKI8, REC114, MEI4, REC107, and MRE11) (Figure 1). We tested the ability of GAL4DB-recombination initiation protein fusion constructs (generally referred to below as DB-REC) to complement the corresponding recombination gene deletion (recΔ) using five metrics: spore viability, sporulation percentage, heteroallelic recombination (gene conversion), crossing over, and DSB formation at a natural hotspot. Each construct was transformed into diploids carrying null mutations (deletions of the open reading frames) of the WT gene (e.g., the DB-REC104 construct was transformed into rec104Δ/rec104Δ cells). Each Rec− diploid was also independently transformed with a plasmid containing only GAL4DB (pDRH2 vector) as a negative control or a REC+ WT copy (i.e., no fusion) of the corresponding recombination gene as a positive control.
The RM96 WT diploid (Rec+ for all initiation genes) sporulated at 88% and produced spores that were 96% viable (Table 3). Rec− diploids with the corresponding WT gene on a plasmid sporulated almost as well as the RM96 WT diploid and produced spores with ∼85% viability (Table 3). In contrast, cells with only the vector present had much lower sporulation and produced almost no viable spores. Sporulation and spore viability increased in every DB-Rec fusion construct compared to the vector control and the values were similar to those of cells containing the WT recombination genes. This implies that all DB-Rec fusions restore overall meiotic recombination to normal levels.
To verify that the DB-REC fusion constructs restored meiotic recombination, a replica plating assay diagnostic for gene conversion and, indirectly, for crossing over, was used (Figure 2). Only background (i.e., mitotic) levels of recombination were observed in all Rec− diploids (e.g., Figure 2, top row). All WT REC plasmids conferred levels of recombination similar to the wild-type RM96 diploid. As predicted by the sporulation and viability values, all fusion constructs promoted wild-type levels of meiotic gene conversion (at LEU1, LYS2, and TRP5) and crossing over (monitored by can1r and cyh2r) (Figure 2).
Although it was clear from the above that overall meiotic recombination occurred at high levels, we also examined DSB formation at the ARE1 hotspot (Figure 3A). The DB-REC fusion constructs were each transformed into recΔ rad50S diploids otherwise isogenic to the diploids used for the other complementation tests. Only background levels of DSBs were detected in the absence of any recombination initiation protein (e.g., see rec107Δ and mei4Δ, Figure 3B). In a Rec+ control (DRH1-90 containing pDRH2 vector), 12.3% DSBs were observed at ARE1 (Figure 3B). DSBs were detected with all fusion proteins; however, the DSB frequency varied among them (Figure 3B). In recΔ diploids, high DSB percentages were observed at ARE1 with DB-Ski8, DB-Spo11, DB-Rec102, and DB-Rec104, consistent with previously published ARE1 values (Baudat and Nicolas 1997; Peciña et al. 2002). Lower levels were observed for DB-Mei4, DB-Rec107, DB-Rec114, and DB-Mre11, but lower percentages of DSBs at ARE1 do not necessarily imply that the fusion protein is less functional overall. Robine et al. (2007) demonstrated that the addition of the Gal4DB moiety to Spo11 altered the relative affinity of DB-Spo11 for different natural hotspots; overall meiotic recombination was high, but the relative frequencies of hotspots changed. This observation is likely to be true for the other DB-Rec fusion proteins we examined, since we know that overall recombination was high, as was spore viability. Taking all the data together, we conclude that the fusion proteins are capable of interacting properly with the other recombination initiation proteins to initiate meiotic recombination.
Several recombination initiation proteins are able to promote DSBs at GAL2:
As a control, we examined DSBs produced at GAL2 in a spo11Δ rad50S diploid containing the GAL4DB vector, the WT SPO11 plasmid, or the DB-SPO11 fusion plasmid. Levels of DSBs significantly above background were not observed with either of the control plasmids (Figure 4B), but the DB-fusion protein caused DSBs upstream of GAL2 at a total frequency of 15% (± 3%) (Figure 4B and later figures). This result is similar to the 12% ± 2% detected by Peciña et al. (2002). Figure 4 demonstrates that there are four DSB bands (Figure 4B). This is suggestive, since there are several UASGAL consensus sequences upstream of GAL2 (Figure 4A) (Vashee et al. 1993) (see discussion). Our experiments utilized a probe upstream of GAL2 (Figure 4A), whereas Peciña et al. (2002) used a downstream probe and observed only one DSB (Peciña et al. 2002). We repeated the experiment using a probe at the 3′ end of the GAL2 gene (Figure 4A). The results in Figure 4C demonstrate the same four bands, except that the banding pattern, as expected, is in the opposite orientation.
To determine if other initiation proteins were able to promote DSB formation at the GAL2 coldspot, we measured DSBs at GAL2 for seven DB-Rec fusion proteins (DB-Rec102, DB-Rec104, DB-Ski8, DB-Rec114, DB-Rec107, DB-Mei4, and DB-Mre11). In every experiment, neither the WT REC control plasmid nor the negative control (vector) displayed DSBs above background at GAL2 (e.g., Figure 5A and data not shown). This confirms that the GAL2 region is naturally a coldspot for recombination, and that DSB formation is dependent on tethering a recombination initiation protein to GAL2. Six initiation proteins (DB-Rec102, DB-Rec104, DB-Ski8, DB-Rec114, DB-Rec107, or DB-Mei4), in addition to DB-Spo11, could induce DSBs upstream of the GAL2 coldspot (Figure 5, B and C). Only DB-Mre11 produced insignificant levels (compared to background, P = 0.18) of DSB formation at GAL2 (Figure 5B; Table 4). These data indicate that seven of the eight fusion proteins tested can recruit all other required initiation proteins to the region where they are tethered. Although seven fusions can create DSBs, they do so with different frequencies (Table 4).
DSBs at GAL2 in gal4Δ diploids:
GAL4 is a positive transcriptional activator of the galactose-metabolizing genes. It is constitutively expressed in cells with and without galactose, and binds constitutively to UASGAL sites upstream of the regulated genes (Giniger et al. 1985; Huibregtse et al. 1993). It was therefore possible that the different frequencies of DSBs observed in cells with the various fusion constructs reflected, at least partially, varying abilities of the fusion proteins to compete with native Gal4 for binding to the UASGAL sequences at GAL2. This possibility was addressed by removing cellular Gal4 protein by a gal4Δ mutation and measuring DSBs at GAL2 in gal4Δ diploids containing DB-Spo11 (with high levels of DSBs) or DB-Rec107 (with low levels of DSBs) fusions. In these gal4Δ experiments, the same DSB bands (A, B, C, D) were observed as in previous experiments, but the amount of DSBs increased similarly for both DB-Spo11 and DB-Rec107 (1.6- and 1.4-fold, respectively; Figure 6; Table 5). These data indicate that there is some competition between the DB-Rec and Gal4 proteins, but suggests that the competition is similar for all fusion proteins. Thus, failure to compete with Gal4 protein is not the reason that the DB-Rec107 fusion creates fewer DSBs than the DB-Spo11 fusion at GAL2.
DSB formation at GAL2 in Rec+ cells containing DB-Rec fusion proteins:
For recombination initiation components suggested to normally function as multimers (Arora et al. 2004; Sasanuma et al. 2007; Fukuda et al. 2008), we reasoned that addition of the wild-type initiation protein might help to better assemble the complex. To provide the WT protein, we performed the experiments in a Rec+ strain (Figure 7) so that both DB-Rec protein and wild-type Rec protein were present in the cell. Four DSB bands (Figure 7) were detected at the same positions as observed in the previous experiments. As in Figure 5, there were differences in the frequency of DSBs detected with the various fusion proteins (Table 6). The ratio of DSBs observed upstream of GAL2 in the Rec− vs. the Rec+ background is close to 1.0 for DB-Ski8, DB-Rec107, and DB-Mre11, indicating that the presence of the WT protein did not affect the ability of initiation to occur (Table 6). In contrast, the WT protein significantly increased DSB formation for DB-Rec102 and DB-Rec114 (P = 0.02 and P < 0.01, respectively). We observed significant decreases in DSBs for DB-Mei4, DB-Spo11, and DB-Rec104 (P = 0.03, P = 0.04, and P < 0.01, respectively). The presence of WT protein did not help the DB-Mre11 construct. One interpretation for how the WT protein might decrease DSBs at GAL2 is if the WT protein preferentially sequesters the DB-Rec fusion or other initiation proteins at natural hotspots.
DB-Mre11 binds GAL2 UASGAL sites:
The only fusion protein which could not make DSBs at GAL2 was DB-Mre11. A possible explanation for this is that the fusion protein does not effectively bind to the UASGAL at GAL2, either because it cannot compete with native Gal4 for binding or because the Gal4DB moiety can not function. Since the data from Table 6 indicate that the Gal4 protein competes with DB-Spo11 and DB-Rec107, we examined the effect of removing Gal4 on the ability of DB-Mre11 to make DSBs at GAL2. We found that DSBs still do not form when native Gal4 is eliminated (Figure 6). To determine if DB-Mre11 is capable of binding at GAL2, we performed chromatin immunoprecipitation (ChIP) on mre11Δ diploids containing either the GAL4DB vector or the DB-MRE11 fusion construct. Two regions were monitored by PCR (Figure 8): one near the UASGAL region of GAL2, and one ∼16 kb downstream of GAL2. The latter region has no consensus UASGAL sites, is void of DSBs (Gerton et al. 2000), and has a physical map: genetic map ratio (Saccharomyces Genome Database) indicative of low levels of recombination. Our results indicate that DB-Mre11 does bind GAL2 UASGAL DNA during meiosis (Figure 8). At the onset of meiosis, enrichment at the targeted hotspot is the highest, is followed by an initial drop at 4 hr into sporulation, and then a steady increase as meiosis proceeds. Since DB-Mre11 is expressed in both mitosis and meiosis, it seems likely that much of the DB-Mre11 present in the cells at 0-hr time binds to UASGAL sites, but that after meiosis has commenced, some of the fusion protein becomes recruited to other hotspots for meiotic recombination. The data confirm that DB-Mre11 does associate with GAL2 promoter DNA, indicating the lack of DSBs observed with this fusion is not due to an inability to bind.
Seven DB-Rec fusion proteins create DSBs at GAL2 at different frequencies:
Ten proteins initiate meiotic recombination, but how these 10 proteins interact with each other, or with the DNA, to initiate DSB formation is still not known. It is not clear if all 10 proteins simultaneously form the initiation complex on the DNA, if a sequential assembly occurs on the DNA, or if assembly occurs off the DNA. The results here indicate that tethering 7 different initiation proteins to a coldspot causes initiation to occur; even the least effective fusion created DSBs at frequencies greater than most natural recombination hotspots (Sun et al. 1991; De Massy et al. 1994; Haring et al. 2003). Of the 8 initiation proteins tested, only the DB-Mre11 fusion was unable to induce significant DSBs at the coldspot.
Since Borde et al. (2004) has shown that Mre11 associates with hotspot chromatin in meiosis prior to the formation of DSBs, it was surprising to us that DB-Mre11 cannot target DSBs at GAL2. The complementation tests indicate that the DB-Mre11 fusion protein is fully functional, restoring wild-type levels of meiotic recombination and viability in an mre11Δ mutant. One possibility is that the binding domain of DB-Mre11 is occluded and cannot bind to UAS. The ChIP data indicate that this is not true. An alternative possibility is that once tethered, the Mre11 protein is in an orientation that prohibits an appropriate association with other required components. A third explanation is that the MRX-subcomplex normally only interacts with the other initiation proteins after components of the other two subcomplexes are associated with the DNA. This possibility is supported by data indicating that the association of Mre11 with chromatin is dependent upon the presence of all initiation proteins except Rad50 protein (Borde et al. 2004). If this explanation is correct, it suggests that DB-Xrs2 and DB-Rad50 would also be unable to recruit the necessary components to make breaks in this assay system.
Our data indicate the eight fusion proteins examined cause different amounts of DSB formation. The frequency of DSB formation (from highest to lowest) at GAL2 was: Spo11, Rec104, Rec102, Mei4, Rec103, Rec114, Rec107, Mre11. One explanation for the DSB differences is that the DB domain fused with the Rec proteins might be differentially occluded from GAL2 UAS sites. We find no clear correlation, however, between the size of the recombination protein in the fusion and the amount of DSBs [e.g., Spo11 is 45 kDa (15.4% DSBs); Rec107 is 36 kDa (2.8% DSBs)]. Alternatively, if an initiation complex forms off the DNA prior to binding, then the DB moiety in each fusion would be located in a different environment. For example, fusion proteins normally situated toward the center of the complex could have the DB occluded by other proteins, thereby reducing DSBs at GAL2. A second explanation for different DSB frequencies is that some DB-Rec fusions cannot compete with native Gal4 as effectively. Differential Gal4 competition could occur whether the DB is binding as a single fusion protein or in a complex of initiation proteins. We analyzed the effect of removing Gal4 on the two fusions (DB-Spo11 and DB-Rec107) with the highest and lowest levels of DSBs. After removing Gal4 protein, DSBs at GAL2 were increased by ∼50% in each strain, and the relative difference between DB-SPO11 and DB-REC107 strains was not significantly different (P = 0.52). Thus, variations in DSB frequency in strains with different fusion proteins are not easily explained by differential competition with Gal4 protein for binding to UASGAL sites.
A third possibility is that the DSB frequency reflects a sequential assembly of subcomplexes. Three subcomplexes (the Spo11-subcomplex, the Rec114-subcomplex, and the MRX-subcomplex) exist among the initiation proteins (see Introduction). Our data indicate that the Spo11-subcomplex components are, on average, most effective at creating DSBs at GAL2 (average DSB value per protein = 9.9%). The Rec114-subcomplex components (average DSB value per protein = 4.0%) are generally less effective at creating DSBs. The MRX-subcomplex (as indicated only by Mre11) was ineffective at creating DSBs at the coldspot. Therefore, differences in the amount of DSBs could reflect the normal order of assembly; a fusion protein in a subcomplex that normally associates with DNA first would be better at attracting the other components.
Of the four Spo11-subcomplex members, the tethered Ski8 protein is the least effective at inducing DSBs. Arora et al. (2004) showed that Ski8 associates with both chromatin loops and axes; in contrast, Rec102 and Spo11 were reported to be primarily on chromatin loops (Blat et al. 2002; Kee et al. 2004). It was proposed that Ski8 acts as a scaffold for the proper assembly of Spo11-Rec104-Rec102 (Arora et al. 2004; Li et al. 2006). If the main role of Ski8 were to assemble the other three components, then Ski8 may not be present at hotspots when DSBs are actually made. Perhaps the protein normally dissociates after bringing the other three together, and when continually present, as with DB-Ski8, actually reduces the efficiency of DSB formation. This view would predict that Rec103 should not be as effective in pulling down hotspots in ChIP experiments where DSB progression was blocked by rad50S.
Double strand break pattern upstream of GAL2:
Previous studies of tethered Spo11 showed only one DSB at GAL2 (Peciña et al. 2002). Our data indicate four DSBs. One possible explanation for the difference may be that Peciña et al. (2002) used strains of the SK1 background (Peciña et al. 2002), while we used the S288C background. It is possible that the GAL2 promoter regions of the two strains have sequence polymorphisms, resulting in an altered pattern of breaks. Alternatively, there may be some technical difference in how we monitor DSBs. However, we observe the same four breaks (and the same patterns of breaks) when we probe the chromosome using the approach as Peciña et al. (2002) (Figure 4C).
We detected doublet DSBs in the coding region of the ARE1 hotspot at the locations of two consensus UASGAL sites. Peciña et al. (2002) also observed that a DB-Spo11 fusion made breaks at these two sites. Gal4DB binds UASGAL as a dimer (Carey et al. 1989), allowing the initiation complex to potentially form on either side of it, which would result in doublets. This observation raises the possibility that the four DSB bands at GAL2 are actually two doublets produced from the fusions binding to two of the four UASGAL sites (Figure 4A). At this time we know of no reason why only two sites should be favored.
The recombination initiation complex might be able to assemble in alternative ways:
Meiotic recombination initiation requires the presence of multiple proteins; the simplest idea would be that initiation depends on these proteins interacting with a strictly defined spatial orientation. However, our results suggest that the idea of a complex, forming in a static, fixed way, is not sufficient. On average the Spo11-subcomplex members appear to recruit a functional complex most effectively, and the Rec114-subcomplex members cause DSBs at a lower frequency. We propose that the functional initiation complex would form differently when different proteins are tethered. The result that seven different tethered proteins make breaks suggests there is flexibility in the ways the proteins assemble, all of which still place Spo11 in a proper position to create DSBs. Such flexibility in the assembly of a functional complex might be useful to ensure recombination initiation occurs if different proteins (or subcomplexes) interact with the DNA first; it would also explain, in part, why different experimental approaches have produced different conclusions about the order of assembly (see Introduction).
We thank Jan Fassler, Tom Petes, Scott Keeney, and Karry Jannie for comments on the manuscript. We thank Kelley Foreman for laboratory assistance, Doug Houston for generously providing qPCR materials and expertise, Amnon Kohen for use of his Molecular Dynamics phosphorimager, and Josh Weiner for use of his Invitrogen magnetic separator. We also thank Eric Alani who provided a protocol for ChIP. D.R.K. was supported during summers by the Evelyn Hart-Watson and Avis Cone graduate fellowships and in the last stages of the work by National Science Foundation grant MCB-0743983. J.M.W. was supported by a University of Iowa summer honors undergraduate fellowship. The initial work was supported by National Science Foundation grant MCB-0083816. The majority of the work was supported by the Department of Biology (University of Iowa). The final experiments were supported by National Science Foundation grant MCB-0743983.
↵1 Present address: Department of Chemistry, Biochemistry, and Molecular Biology, North Dakota State University, Fargo, ND 58105.
Communicating editor: S. Keeney
- Received March 10, 2009.
- Accepted March 26, 2009.
- Copyright © 2009 by the Genetics Society of America