Novel Nucleotide Sequence Motifs That Produce Hotspots of Meiotic Recombination in Schizosaccharomyces pombe

Abstract

In many organisms, including yeasts and humans, meiotic recombination is initiated preferentially at a limited number of sites in the genome referred to as recombination hotspots. Predicting precisely the location of most hotspots has remained elusive. In this study, we tested the hypothesis that hotspots can result from multiple different sequence motifs. We devised a method to rapidly screen many short random oligonucleotide sequences for hotspot activity in the fission yeast Schizosaccharomyces pombe and produced a library of ∼500 unique 15- and 30-bp sequences containing hotspots. The frequency of hotspots found suggests that there may be a relatively large number of different sequence motifs that produce hotspots. Within our sequence library, we found many shorter 6- to 10-bp motifs that occurred multiple times, many of which produced hotspots when reconstructed in vivo. On the basis of sequence similarity, we were able to group those hotspots into five different sequence families. At least one of the novel hotspots we found appears to be a target for a transcription factor, as it requires that factor for its hotspot activity. We propose that many hotspots in S. pombe, and perhaps other organisms, result from simple sequence motifs, some of which are identified here.

MEIOSIS is a form of cell division common to all sexually reproducing organisms. It differs from mitosis in that two cell divisions follow a single round of DNA replication, resulting in four haploid cells known as gametes or spores. The first division of meiosis differs from mitosis in that paternal and maternal homologs segregate to opposite poles, reducing by half the number of chromosomes in the resulting gametes. Prior to the first division, homologous chromosomes recombine with each other at a greatly elevated frequency compared to mitosis (Esposito and Wagstaff 1981). This recombination serves at least two important functions. First, it forms crossovers (chiasmata) between chromosomes, which are required in most organisms for the proper segregation of homologous chromosomes (Baker et al. 1976). Second, the random shuffling of maternal and paternal alleles at each generation increases genetic diversity, which enhances the ability of a species to adapt to its environment through natural selection.

Meiotic recombination events are not distributed evenly throughout the genome of most organisms. Rather, they occur at high frequency at some sites and low frequency at others. Sites that recombine at a frequency significantly higher than the genomic average are known as hotspots. These hotspots coincide with the formation of DNA double-strand breaks (DSBs) in both the fission and budding yeasts and likely many other organisms (Sun et al. 1989; Cao et al. 1990; Fan et al. 1995; Cervantes et al. 2000; Mahadevaiah et al. 2001; Petes 2001; Steiner et al. 2002; Young et al. 2002; Cromie et al. 2007). Formation of DSBs requires a number of different proteins, including Spo11 (Rec12 in Schizosaccharomyces pombe), a widely conserved protein among eukaryotes that has the active site for cleaving the phosphodiester backbone (Keeney et al. 1997; Cervantes et al. 2000; Malik et al. 2007). After the formation of DSBs, the two broken ends of the DNA initiate recombination by invading intact homologous DNA to form joint molecules, which can be resolved to produce both crossover and noncrossover exchanges (Pâques and Haber 1999).

DNA breaks occur at preferred genomic positions during meiosis, but the factors determining the positions of most break sites are not clearly understood. A global analysis of the distribution of DSB sites in the budding yeast Saccharomyces cerevisiae showed that most breaks occur in 5-kb regions where the GC content exceeds the average GC content for the genome (Gerton et al. 2000), but the causal relation, if any, between elevated GC content and DSB formation is unknown. In addition, DSBs occur primarily in intergenic regions (IGRs) (Baudat and Nicolas 1997; Gerton et al. 2000), which supports the possibility that many hotspots are associated with the binding of transcription factors. Indeed, several well-characterized hotspots are known to require transcription factor binding (White et al. 1993; Kon et al. 1997; Petes 2001; Mieczkowski et al. 2006).

Like S. cerevisiae, the fission yeast S. pombe shows a nonrandom distribution of meiotic DSBs (Cervantes et al. 2000; Young et al. 2002). A genomewide analysis of the distribution of DSBs in S. pombe revealed that DSBs occur primarily in intergenic regions, particularly in exceptionally large intergenic regions (Cromie et al. 2007). In fact, it was noted that large IGRs are strongly predictive of DSBs. The median IGR size in S. pombe is 0.7 kb (Wood et al. 2002), but approximately half of all prominent DSBs occur in IGRs >3 kb, and 44% of IGRs >3 kb contain prominent DSBs (Cromie et al. 2007). However, the basis for the association of DSBs and IGRs remains unclear. One possibility is that some IGRs contain specific sequence motifs that create recombination hotspots whereas others do not.

One simple sequence motif, ATGACGTCA, known as M26 or CRE (cyclic-AMP response element), acts as a recombination hotspot at multiple sites in the S. pombe genome. These sites include both IGRs and protein-coding regions, for example, the ade6 gene where this hotspot motif was originally discovered (Gutz 1971; Fox et al. 1997; Steiner and Smith 2005a). The CRE sequence is a binding site for a heterodimeric transcription factor, Atf1-Pcr1, which is required for activity of the hotspot (Kon et al. 1997). Since the CRE hotspot can predict only a small fraction of the DSBs in the S. pombe genome (Cromie et al. 2007 and W. Steiner, unpublished observation), we hypothesized that there may be other simple sequence motifs that contribute to recombination hotspots. The results shown in this study confirm that hypothesis.

MATERIALS AND METHODS

Strains and genetic procedures:

All strains used in this study with assigned ade6 allele numbers are shown in Table 1. The ade6-4001 allele was generated by amplification of the kanMX6-ura4⁺ construct contained in the plasmid pura4-kanMX6 (Steiner and Smith 2005b). This construct was amplified by PCR using oligonucleotides with 80 base 5′ extensions homologous to the ade6 insertion site (Table 1). The resulting PCR product was used for linear transformation of strain WS121 selecting for uracil prototrophy. Transformants were confirmed by resistance to G418, sequencing, and Southern blot hybridization.

TABLE 1.

Strains

ade6 allele	Mating type	Description or sequencea	Base pairsb	Motifc	Straind
+	h90			NA	WS121
52	h−	G796Ae		NA	WS3
M375	h90	TGAGGACGTGAGf	133–144	NA	WS135
M26	h90	GGATGACGTGAGf	133–144	CRE	WS136
M26	h+	his7-366			WS322
M26	h+	his7-366 php2Δ∷kanMX6g			WS310
M26	h+	his7-366 php3Δ∷kanMX6g			WS324
M26	h+	php5Δ∷kanMX6g			WS295
469	h−	C1468Tf		NA	WS315
469	h−	php2Δ∷kanMX6			WS313
469	h−	php3Δ∷kanMX6			WS317
469	h−	php5Δ∷kanMX6			WS311
3074	h90	GGATGACGTCAGh	133–144	CRE	WS137
4001	h90	Δ(-162-451)∷kanMX6-ura4+	—	NA	WS129
4002	h90	CCAATCAi	129–134	7-7	WS224
4002	h+	his7-366			WS326
4002	h+	his7-366 php2Δ∷kanMX6			WS328
4002	h+	his7-366 php3Δ∷kanMX6			WS320
4002	h+	his7-366 php5Δ∷kanMX6			WS293
4003	h90	A__CATGACATCAT	148–160	6-6/8-1	WS330
4003	h−				WS182
4005	h90	ACGTAA_T	138–145	7-1	WS149
4006	h90	CGTCATAi	149–155	7-15	WS150
4008	h90	TATTACGTAAT	160–168	6-1 (Con)	WS331
4008	h−				WS184
4009	h90	ATAGCGTCATATACT	152–165	7-15 (Con)	WS332
4009	h−				WS186
4010	h90	GATGACATAA	151–159	6-21 (Con)	WS365
4010	h−				WS188
4071	h90	ATGATGTCACi	152–161	7-2 (Con)	WS237
4072	h90	ACCCCGCACGCA	167–177	7-4 (Con)	WS240
4073	h90	ACGGCCCCCA_CAATTi	126–141	7-31 (Con)	WS241
4094	h90	GGATGTAAGTj	130–139	10-1	WS374
4095	h90	GGTCTGGACCj	130–139	10-2	WS376
4096	h90	GATGACATCAj	130–139	8-1	WS378
4099	h90	TGAACCCCGCACTGAj	129–143	7-4 (Con)	WS382
4100	h90	GCCCCCACAj	132–140	7-31 (Con)	WS384
4101	h90	ACCCCGCACGTAAT	167–179	7-4	WS386
4102	h90	ATGGCCCCCA_CTATTi	126–141	7-31	WS394
4103	h90	TGACCCCGCACGTj	130–142	7-4	WS409
4104	h90	TGGCCCCCACTATj	130–142	7-31	WS410

^aBases are numbered starting at the first nucleotide of the ade6 open reading frame. Bases in regular type indicate wild-type sequence. Boldface type indicates base substitutions. Insertions or deletions are underlined. For any given ade6 allele, the relevant sequence is shown only once. Additional genotypic information for strains containing the same allele of ade6 is also indicated in this column.

^bThe nucleotide positions of the ade6 sequence shown in the third column. Numbering indicates wild-type sequence before insertions or deletions.

^cIndicates the motif (Table 2, Table S2a, and Table S3) on which the allele is based. NA, not applicable. Con, consensus sequence (Figure S1).

^dIn addition to the ade6 mutations shown in the third column, all strains also contain ura4-D18 and leu1-32, except for WS3 (ura4⁺ leu1⁺). All homothallic strains (h⁹⁰) also contain the plasmid pWS35 (Figure 1).

^eM. Fox and G. Smith, personal communication.

^fSzankasi et al. (1988).

^gMercier et al. (2006).

^hSteiner and Smith (2005b).

ⁱComplement strand shown.

^jThese strains also contain a closely linked nonsense mutation, A121T.

The plasmid pWS35 was constructed by PCR amplification of a 613-bp fragment of ade6 with primers 5′-NNNNNCTCGAGCTTGGAAATGTAACGATGAC-3′ and 5′-NNNNNCTCGAGTAAGCCAATGTTTTACTTTTCAG-3′, containing XhoI restriction sites near their 5′ ends. (The additional random bases at the 5′ ends were added to permit efficient cleavage by XhoI.) The resulting PCR product was digested with XhoI and ligated into the XhoI site of pSP1 (Cottarel et al. 1993), producing pWS35. This plasmid contains the S. cerevisiae LEU2 gene as a selectable marker. The 613-bp ade6 fragment of pWS35 is the same fragment deleted in the ade6-4001 allele (Table 1).

Growth and sporulation medium have been previously described (Gutz et al. 1974; Steiner and Smith 2005a,b). Heterothallic strains were grown in rich medium supplemented with adenine (A), uracil (U), leucine, histidine, and lysine (YEL-5S). Strains containing the plasmid pWS35 were grown in minimal medium supplemented with adenine and uracil (NBL-AU). Crosses between heterothallic strains were performed on SPA-5S, and homothallic crosses were performed on SPA-AU. Crosses were incubated for 2 days at 25° before harvesting and analysis as described (Steiner and Smith 2005b). Any given cross was performed a minimum of three times for determination of recombination frequencies.

Screen for hotspots:

WS129 was grown overnight in NBL-A to 0.5-1 × 10⁷ cells/ml and transformed using a lithium acetate-mediated transformation procedure (Bähler et al. 1998). Transforming DNA consisted of a 1.6-kb ade6 fragment containing a 15- or 30-bp random sequence substitution from nucleotides 125–139 or 125–154, respectively (nucleotides from start of ade6 open reading frame). Both substitutions also include an A→T substitution at nucleotide 121, which produces a stop codon and insures that all transformants are adenine auxotrophs. The linear DNA used for transformation was generated by overlap extension PCR (Vallejo et al. 1995). The primers used were oWS202: ACGAACATCATTAAGCGCGAAGCG, oWS203: ACGCATGAGTTGTGGAAGTCGAGA, oWS208: GTTAGGCAGGAGAATTTGCTGCA, oWS209: TGCAGCAAATTCTCCTGCCTAACNNNNNNNNNNNNNNNNNNNNNNNNNNNNNNCATCATTTACTGACCCCGATGCAATTG, and oWS241: TGCAGCAAATTCTCCTGCCTAACNNNNNNNNNNNNNNNGTGAGCACATTGATGCATCATTTAC. The product of oWS202 and oWS208 was combined with the product of oWS209 (or 241) and oWS203 for the overlap extension reaction. All PCR products were performed with the high-fidelity polymerases Vent (New England Biolabs) or Pfx (Invitrogen), using conditions recommended by the manufacturers. A plasmid clone of the ade6 gene (pAS1; Szankasi et al. 1988) was used as the starting template for PCR reactions. Approximately 5 μg of linear ade6 DNA was used per 10⁸ cells transformed.

Following transformation, cells were plated on NBA-AU (2–4 × 10⁷ cells/plate) and incubated at 32° for 2 days. Those plates were replica plated to NBA-AU containing 1 mg/ml 5-FOA (Toronto Research Chemicals) to select for cells with ade6 replacements. After 3 days growth, individual colonies became visible. These plates were replica plated a second time to NBA-AU + 5-FOA to reduce background growth. After 4 days growth at 32°, colonies were replica plated to SPA-AU and incubated 2 days at 25° to induce meiosis. Those colonies were then exposed to acetone vapor (Egel 1977) to kill remaining vegetative cells, replica plated to YEA-4S and incubated 3 days at 32°. Red colonies showing many white papillae were picked and streaked for single colonies to NBA containing uracil and limiting adenine (10 μg/ml) to distinguish ade6⁻ colonies from ade6⁺ recombinants. Red colonies from each streak were patched in a grid pattern to NBA-AU along with 5 control strains: WS129, WS3, WS135, WS136, and WS137 (Table 1). After 2 days growth at 32°, those patches were replica plated to three different media:

1. SPA-AU to confirm the original meiotic hotspot phenotype as described above and qualitatively estimate the activity of each hotspot relative to ade6-M375, ade6-M26, and ade6-3074 by comparing the density of white papillae. The heterothallic strain, WS3, which cannot sporulate, served as a control for the effectiveness of the subsequent acetone treatment for this step.

2. YEA-4S to test for a possible mitotic hotspot phenotype. Strains containing a mitotic hotspot should show many white papillae prior to undergoing meiosis (only one strain with a potential mitotic hotspot was found).

3. YEA-5S containing 100 μg/ml G418 to confirm homologous replacement of the ade6 gene (Storici et al. 2001). Homologous gene replacement results in the simultaneous loss of both the ura4⁺ and kanMX6 markers found in the ade6-4001 allele, resulting in both resistance to 5-FOA and sensitivity to G418.

Sequence analysis:

G418-sensitive strains that showed an obviously higher density of white papillae than an ade6-M375 control strain were allowed to lose the plasmid, pWS35, by two successive nonselective streaks onto YEA-4S. Plasmid-free derivatives were identified by the inability to grow in the absence of leucine and used for preparation of genomic DNA, which was used as a template for PCR amplification of ade6 using primers oWS202 and oWS203 (above). PCR products were sequenced by the High Throughput Genomics Unit (University of Washington, Seattle).

The sequence substitutions in ade6 were analyzed for common 6- to 10-bp motifs using YMF3.0 (Sinha and Tompa 2002, 2003) (http://wingless.cs.washington.edu/YMF/YMFWeb/YMFInput.pl) and MEME (Bailey et al. 1996) (http://meme.nbcr.net/meme/intro.html).

Sequence randomization was accomplished by using an algorithm (Pearson and Lipman 1988) available through the San Diego Super Computer Center (http://workbench.sdsc.edu/).

Reconstruction of specific motifs:

Specific motifs were reconstructed in the ade6 gene by overlap-extension PCR (Vallejo et al. 1995) using inner primers containing the desired mutations and outside primers oWS202 and oWS203 (above). All reconstructed motifs listed in Table 1 were confirmed by sequencing and Southern blot hybridization.

RESULTS

The screen for recombination hotspots:

Screening for sequence-dependent recombination hotspots from a large pool of random sequences required a means of rapidly identifying potential candidates. We therefore took advantage of the fact that strains containing mutations in the ade6 gene produce red-colored colonies on medium containing limiting quantities of adenine (Gutz 1971), whereas ade6⁺ strains form white colonies on the same medium. As shown in Figure 1, strain WS129 was transformed with ade6 DNA containing either a 15- or 30-bp random nucleotide sequence substitution. Transformed cells were selected by loss of the ura4⁺ gene inserted into ade6 (resistance to 5-FOA) and homologous gene replacement was later confirmed by the simultaneous loss of the kanMX6 gene (G418 sensitivity; Storici et al. 2001). After transformed cells formed colonies on 5-FOA medium, they were replica plated to sporulation medium, producing colonies containing mostly spores. Following treatment with acetone vapor to kill remaining unsporulated cells (Egel 1977), these colonies were replica plated to growth medium containing limiting quantities of adenine. (See materials and methods for full details of the screen.) Since the cells carry a plasmid containing a fragment of the ade6 gene, recombination between chromosome and plasmid can produce ade6⁺ spores during meiosis. Plasmid × chromosome recombination has previously been shown to accurately reflect chromosome × chromosome recombination in S. pombe (Ponticelli and Smith 1989). These recombination events can be seen at this stage as white papillae in otherwise red colonies. Strains with no recombination hotspot in ade6, for example ade6-M375 (Gutz 1971), produce red colonies with few white papillae, whereas strains containing a recombination hotspot, such as ade6-M26 or ade6-3074 (Steiner and Smith 2005b), produce red colonies with a large number of papillae (Figure 2). Those colonies are easily distinguished from neighboring colonies, most of which resembled ade6-M375 in our screen.

Figure 1.—

Method for screening a large number of unique 15- or 30-bp sequences for hotspot activity. (A) A strain (WS129) with a partial ade6 deletion and ura4⁺-kanMX6 insertion and plasmid pWS35 (not shown) is transformed with linear ade6 DNA containing a 15- or 30-bp random sequence substitution (shaded bar). (B) Homologous recombinants lose the ura4⁺-kanMX6 marker and now share homology to a fragment of ade6 carried on the plasmid pWS35. After transfer to sporulation medium, these cells will self-mate and form spores. Strains containing a hotspot in the random sequence region will recombine with the plasmid at high frequency (open arrow) to form ade6⁺ spores, which can be identified as white papillae on red colonies after replica plating to the appropriate medium (Figure 2). Although this figure implies noncrossover conversion to ade6⁺, crossover recombinants would also likely produce ade6⁺ spores, because the plasmid carries 164 bp of the ade6 promoter, which is adequate for ade6 expression (Zahn-Zabal et al. 1995). ade6 genes and fragments are drawn to scale; other genes and constructs are not. Open rectangles represent open reading frames.

Figure 2.—

Visual assay for hotspots. Strains containing known hotspot alleles (ade6-M26 or -3074) and one control allele (ade6-M375) were put through the steps of the screen described in the text. (A) ade6-M375, (B) ade6-M26, (C) ade6-3074, (D) ade6-M375 (left), and ade6-3074 (right). A larger version of this figure is shown in Figure S2.

We screened ∼27,600 colonies each containing a 30-bp random sequence and 18,300 colonies each containing a 15-bp random sequence. Among the 30-bp and 15-bp random sequences, we identified 393 and 102 strains, respectively, showing recombination frequencies obviously higher than ade6-M375. This number includes only strains capable of plasmid loss (see materials and methods), since plasmid integration, particularly if it occurs within the ade6 gene itself, might create a recombination hotspot unrelated to any specific sequence (Virgin et al. 1995). The higher frequency of hotspots among the 30-bp random sequences (1.4%) than the 15-bp sequences (0.6%) is consistent with the expectation that longer sequences have a greater chance of containing any given sequence motif. The complete list of hotspot-containing sequences and their relative activities is shown in supporting information, Table S1.

To confirm that the hyper-rec phenotype observed in our transformed strains was due to the sequence substitution in ade6, we crossed plasmid-free derivatives of the hyper-rec strains with the parent strain WS129. Spores from the cross that were 5-FOA resistant were tested for hotspot activity (papillation) by repeating the steps of the screen described above. In 36 randomly chosen strains, the hyper-rec phenotype showed complete linkage to ade6 among the 50–100 spores tested from each cross. That is, every spore containing the sequence substitution from each of the 36 crosses showed the hyper-rec phenotype. This result is strong evidence that the observed hotspot activity in those strains, and probably all of our hotspot strains, is almost certainly due to the sequence substitution within ade6 and not some other cause.

Common sequence motifs appear frequently:

Since it is unlikely that the entire 15- or 30-bp sequence substitution is required for hotspot activity in any of our hotspot-containing strains, we analyzed those sequences for common 6- to 10-bp motifs as described in materials and methods. The most common motifs found contained the 6-bp sequence TGACGT, referred to here as the CRE-core sequence (Table S2). This result demonstrates that our screen had, in fact, identified sequence-dependent recombination hotspots, as expected. However, the high frequency of the CRE motif was surprising since there was no prior reason to expect that the CRE hotspot would appear more frequently than other potential hotspot motifs. A likely interpretation of this result is that the CRE hotspot has the shortest sequence capable of creating a recombination hotspot, <7 bp in the region of ade6 utilized in our screen (Fox et al. 2000). Hotspots requiring a longer sequence than this would occur at lower frequency.

Since the goal of our screen was to identify novel hotspot motifs, we eliminated 97 sequences from our pool that contained the 6-bp CRE-core sequence, TGACGT, which is required for all known Atf1-Pcr1-dependent hotspots in S. pombe (Fox et al. 2000; Steiner and Smith 2005a). These strains accounted for almost 20% of our total sequence pool. We analyzed the remaining 398 sequences for common motifs as described above (Table 2 and Table S3). To determine the likelihood that any of the observed high-frequency motifs formed hotspots, we compared our results to an analysis of the same set of sequences following sequence randomization. Among random sequences, multiple occurrences of any motif would result from chance and not from any property of the motif, e.g., hotspot activity. For any given motif size, we found a significantly higher number of motifs among the actual sequences, i.e., those containing hotspots, compared to randomized sequences. For example, we found 17 different eight-base motifs that occurred four or more times among our pool of 398 sequences. Among random sequences, there were only four eight-base motifs that occurred at the same frequency and none that occurred more frequently (Table S3b). This result suggests that the majority of those motifs are hotspots, or perhaps form part of a larger hotspot.

TABLE 2.

Common motifs among hotspot sequences lacking CRE hotspot

10-base motif number and sequence		Counta	8-base motif number and sequence		Count	7-base motif number and sequence		Count	6-base motif number and sequence		Count
1	GGATGTAAGT	3	1	TGACATCA	6	1	ACGTAAT	14	1	ACGTAA	22
2	GGTCTGGACC	3	4	CCAATGAG	4	2	ATGTCAC	9	6	ACATGA	16
			5	AGAGCTCTb	4	4	CCCCGCA	8	9	AAAGAT	15
			6	TCGGCCGAb	4	7	CCAATCA	7	10	GTATGA	15
			7	AGACGCAG	4	8	CCCACCC	7	45	CTATTA	11
			8	GTCTAGACb	4	15	CGTCATA	6	57	CATCCC	10
			9	ATAATTGG	4	31	CCCCCAC	5	21c	GATGAC	18
			10	AACAGGCG	4
			11	ATTGGCGG	4
			12	AAGCATGA	4
			13	CGCAGTAA	4
			21	AATGGATA	3
			22	CCATTACG	3
			41c	AGGGATGA	4

Some of the most common 6- to 10-base motifs among 398 sequences lacking the CRE core sequence, TGACGT. Only two motifs longer than 8-bp that occurred more than twice were found, 10-1 and 10-2. Only sequences tested for hotspot activity (Figure 3 and Table S4) are shown. Motif numbers correspond to those shown in Table S3a.

^aNumber of times that a given motif is found among the 398 sequences lacking the CRE-core sequence, TGACGT.

^bA palindrome. Palindromes are counted twice, because they occur on both strands.

^cThese motifs are found only among our total pool of sequences and are shown in Table S2a.

Confirmation of hotspot activity by motif reconstruction:

To test whether any of the high-frequency sequence motifs shown in Table 2 were hotspots, we reconstructed some of them in the ade6 gene by specific base-pair changes to the wild-type sequence, for example motifs 6-6, 8-1, 7-1, and 7-15. (The first number of each motif name refers to the length of the motif and the second number refers to the number in that series; Table 2 and Table S3). These experiments generated hotspots in some cases (ade6-4002, -4003, and -4096; Figure 3), but not in others (ade6-4005, 4006; Table 1 and data not shown). We also tested two 10-base motifs that occurred three times each among our pool of sequences (ade6-4094 and -4095; Figure 3), one of which, ade6-4095, produced a hotspot. A hotspot in this case is considered to be any allele that produces a significantly greater frequency of recombinants than ade6-M375 (P < 0.01; Student's t-test), a common control allele for the ade6-M26 hotspot (Gutz 1971).

Figure 3.—

Reconstructed sequence motifs produce hotspots. High-frequency sequence motifs (Table 2 and Figure S1) were reconstructed in the ade6 gene and tested for hotspot activity in homothallic crosses (plasmid pWS35 × chromosome; solid bars) or heterothallic crosses (chromosome × chromosome; open bars). Heterothallic strains were crossed with WS315 (ade6-469; Table 1). Each bar represents the average of at least three crosses ±1 SEM. Sequences show the original motif (Table 2; boldface type) with some flanking nucleotides (regular type). Consensus sequence nucleotides (Figure S1) are underlined when applicable. In this table, only ade6-4009 and ade6-4094 are not significantly more active than ade6-M375 (P > 0.01, Student's t-test). Notes: ^aThe motif (Table 2) on which the allele is based. Con, consensus sequence. NA, not applicable. Motifs with three numbers indicate reconstruction of a particular allele from a subscreen (Figure S1), e.g., 7-31-16 = sequence number 16 from subscreen of motif 7-31. ^bThese motifs are inverted relative to ade6-4100 and -4104. ^cThese motifs are located 36 bp downstream relative to ade6-4099 and -4103.

Motifs that occurred at high frequency but failed to produce hotspots in our reconstruction experiments (ade6-4005, -4006, and -4094) could be explained by either of two possibilities: (1) those motifs are simply not hotspots, that is, they occurred by chance (for example, see Table S3b), or (2) those motifs form only part of a larger sequence necessary for hotspot activity. Reasoning that the second possibility may be true for many motifs, we streamlined our tests for hotspot activity by first testing most of the motifs shown in Table 2 in a “subscreen.” Each subscreen involved repeating the original screen with a given motif flanked by several random nucleotides on one or both sides of the motif (Table S4). Transformants producing hotspots were sequenced and aligned to identify potential consensus sequences (Figure S1). Thus, many of the reconstructed motifs shown in Figure 3 represent consensus sequences identified in those subscreen experiments. Five of the six consensus sequences tested produced hotspots with activity significantly greater than ade6-M375 (P < 0.01, Student's t-test; consensus sequences for motifs 6-1, 6-21, 7-2, 7-4, 7-31; Figure 3), suggesting that the consensus sequence is sufficient for hotspot activity, at least within this narrow region of the genome. However, we cannot infer that the entire consensus sequence is necessary for activity, which can be determined only by systematic mutagenesis.

For some hotspots, we observed an additional level of complexity. For example, the consensus sequence for motif 7-4 found by subscreen was DACCCCGCACD (Figure S1; D = A, G, or T). When the consensus sequence was reconstructed at its original location (bp 131–141), it produced a strong hotspot (ade6-4099, Figure 3). However, when moved only 36 bp away, the same motif produced less than one-tenth as many Ade⁺ recombinants (ade6-4072). This reduced activity suggests either (1) that additional nucleotides outside of the consensus sequence are required for full activity or (2) that there is a position-dependent effect on hotspot activity. Since each hotspot sequence found by subscreen of motif 7-4 was 13 bp long (Figure S1), we tested whether one of these slightly longer sequences would create a more active hotspot at the new location. One of those motifs (7-4-11, Figure S1) is identical to the relatively weak ade6-4072 hotspot over its 13-bp length except that the last base is a T rather than a C. Making this single base substitution more than doubled activity of the hotspot at that position (ade6-4072 vs. -4101, Figure 3). However, hotspot activity of the ade6-4101 allele still remains significantly lower than the identical 13-bp sequence at its original position (ade6-4103). Thus, neither a position-effect nor an effect of more distant nucleotides can be excluded.

Similar complexity was observed for the 7-31 motif. In that case, the 9-bp consensus sequence, RCCCCCACA, was reconstructed at approximately the same position as in the subscreen, but in the inverted orientation. Activity of that hotspot (ade6-4073) was significantly lower than the same sequence in the forward orientation (ade6-4100). However, two additional base changes (ade6-4102) more than tripled the number of Ade⁺ recombinants. The ade6-4102 and -4104 alleles have the same 13-bp sequence as motif 7-31-16 found in a subscreen (Figure S1), but are inverted relative to each other. Both alleles also include the additional K (G or T) consensus nucleotide at the first position (Figure S1), which may increase their activities relative to the weaker ade6-4073 hotspot. The similar activities of the ade6-4102 and -4104 alleles suggests that the nucleotides flanking the hotspot motif have a greater influence on hotspot activity than the motif orientation per se.

Potential hotspot families:

Upon close inspection of some of the sequences in Figure 3, it became apparent that several of them looked quite similar to the previously characterized CRE hotspot. The most active form of this hotspot is a 10-bp palindrome, ATGACGTCAT (Steiner and Smith 2005b). Figure 3 shows that seven other hotspot motifs differ from the 10-bp CRE palindrome at only one or two positions. Given the previously demonstrated flexibility of the CRE hotspot sequence (Fox et al. 2000), it is conceivable that these related sequences could also be Atf1-Pcr1-dependent hotspots.

Since several of the hotspots we identified could be grouped into one family (CRE) on the basis of sequence, we also compared other hotspots (Figure 3 or consensus sequence hotspots from Figure S1) to see whether similar groupings were possible. Five motifs were grouped on the basis of their common CCAAT sequence. We speculated that this group of hotspots could be recognized by the CCAAT-binding factor. The CCAAT-binding factor is an evolutionarily conserved heteromeric transcription factor encoded by the php2, php3, and php5 genes and is required for expression of many eukaryotic genes (McNabb et al. 1997). Hotspot activity of one member of this group, ade6-4002, was reduced significantly by deletion of each gene encoding a subunit of the CCAAT-binding factor, while ade6-M26 activity was largely unaffected (Figure 4). In fact, the very low level of recombination observed in these experiments suggests that the ade6-4002 allele is completely dependent on the CCAAT-binding factor for hotspot activity, much as ade6-M26 hotspot activity is completely dependent on the Atf1-Pcr1 transcription factor (Kon et al. 1997).

Figure 4.—

The CCAAT-binding factor is required specifically for activity of the ade6-4002 hotspot, but not ade6-M26. Crosses were performed between heterothallic strains containing the indicated ade6 alleles and mutations in the php2, php3, or php5 genes, which encode subunits of the CCAAT-binding factor. Bars represent the average Ade⁺ recombinant frequencies ± SEM from a minimum of three experiments for each cross.

A third potential family of hotspots is based on the set of alleles derived from the 7-4 and 7-31 motifs, referred to here as the oligo-C family of hotspots. A search of the TransFac database (http://www.gene-regulation.com/pub/databases.html) suggests that these sequences may be targets for the MIG1 DNA-binding factor in S. cerevisiae, which has several orthologs in S. pombe including Scr1, Rsv1, Hsr1, and Rst2. Yet a fourth family contains two closely related sequences, ade6-4095 and the consensus sequence of motif 8-8. The last potential family has only a single representative, motif 8-6, which shows no obvious similarity to any other identified motif (Figure 5). The very high frequency of hotspots observed in a subscreen of this motif (61%; Table S4), suggests that virtually all of the essential bases are contained within the eight-base motif tested.

Figure 5.—

Potential hotspot families. Hotspot motifs (Figure 3) and hotspot consensus sequences (Figure S1) are aligned to show similarities. In the CRE family of hotspots, bases differing from the 10-bp CRE palindrome (ade6-3083, Steiner and Smith 2005b) are underlined. Comp, complement of consensus sequence (Figure S1).

DISCUSSION

In this study, we devised a method to rapidly screen short nucleotide sequences for hotspot activity and produced a sizable library of 15 and 30 bp sequences containing recombination hotspots. Within those sequences, we identified many shorter motifs ≥6 bp in length that occurred multiple times and hence may form all or part of recombination hotspot sequences. On the basis of our results, we can conclude that the previously characterized CRE hotspot is clearly not the only sequence motif capable of creating a recombination hotspot. However, it was surprising to us how frequently that previously identified hotspot occurred in our library of sequences. The 6-base sequence TGACGT (the CRE core sequence) common to all known Atf1-Pcr1-dependent hotspots (Schuchert et al. 1991; Fox et al. 2000; Steiner and Smith 2005b) was found 75 times, or in 15% of our total pool of hotspot sequences (Table S1 and Table S2a). This frequency increases to almost 20% if one includes sequences from our pool in which the CRE core is formed at the junction between random and nonrandom nucleotides. For example, if the first 4 bases of our 15- or 30-base random region reads GTCA, the 6-base sequence ACGTCA is formed at the junction, which is the complement of the CRE core sequence shown above (see sequence of oligonucleotides oWS209 and oWS241 in materials and methods and Table S1). However, the frequency of CRE-like hotspots may be even greater than the observed frequency of the traditional CRE-core sequence, as several hotspots lacking that core still showed obvious sequence similarity to CRE (Figure 5).

Since our data indicate that other sequence motifs unrelated to CRE also create hotspots, it is possible that the overrepresentation of CRE is due to it having the shortest sequence necessary for observable hotspot activity, which may be fewer than 7 bp in some locations (Fox et al. 2000). Consistent with this, we have observed that the ade6-4002 hotspot requires not only the CCAATCA sequence shown in Figure 3, but also three partially degenerate bases to the left (C. Kalinowski and W. Steiner, unpublished observation). The results from subscreens of other motifs (Table S4) also suggests that hotspots other than CRE may require more than 7 bp for activity. In those experiments, none of the seven or eight base motifs tested produced hotspots in 100% of transformed cells, suggesting that one or more specific nucleotides are required in the random regions flanking each motif.

The number of different hotspot motifs:

What does the frequency of hotspots we observed say about the potential number of different hotspot motifs? We observed that ∼0.6% of random 15mers produced an observable hotspot. In a random 15-bp sequence, there are as many as 9 unique 7-bp sequences when viewed in 7-bp windows moving in steps of 1 bp. Therefore, the probability of finding any unique 7-bp sequence in a random 15mer is (0.25)⁷ × 9 = 0.055%. The observed frequency of hotspots was ∼10-fold higher than this, suggesting that a minimum of 10 unique 7-bp motifs are required to account for the observed frequency of hotspots. By the same reasoning, ∼50 8-bp motifs, or 225 9-bp motifs would be required. On the basis of the frequency of hotspots among random 30mers, the same calculations produce slightly lower estimates of the number of different hotspots. In either case, however, our data suggest that the number of different hotspot motifs is potentially large.

Without knowing the precise nucleotide sequence required for any given hotspot, it is not possible to determine precisely which of the sequences in our library can have their hotspot activity attributed to a particular motif, with the exception of many of the well-characterized CRE-like hotspots. However, even under generous assumptions about which motifs may create hotspots (for example, that all CRE-core sequences and all CCAAT sequences are hot; see highlighted motifs in Table S1), 197 sequences remain with no identified hotspot motif. Thus, additional hotspot motifs remain to be identified.

The basis of recombination hotspots:

Meiotic recombination hotspots have been most thoroughly analyzed in the budding and fission yeasts S. cerevisiae and S. pombe, respectively. In both organisms, hotspots of recombination are sites of DSBs, which occur preferentially at a limited number of positions throughout their respective genomes. What determines the location of hotspots is incompletely understood, though some correlations have been made. For example, in S. cerevisiae DSBs occur predominantly in intergenic regions where the GC content modestly exceeds the genomic average GC content (Baudat and Nicolas 1997; Gerton et al. 2000; Mieczkowski et al. 2006). In S. pombe DSBs also occur predominantly in intergenic regions, particularly in large intergenic regions, but unlike S. cerevisiae the correlation between GC content and DSBs is very weak (Cromie et al. 2007). In both yeasts, tested DSB hotspots also colocalize to sites of “open” chromatin, i.e., sites that are sensitive to cleavage with S1 or micrococcal nucleases (Wu and Lichten 1994; Mizuno et al. 1997; Hirota et al. 2007).

Beyond the aforementioned genomic features, predicting the precise location of recombination hotspots in most cases remains elusive. For example, the well-characterized mbs1 hotspot of S. pombe occurs in an unusually large 7-kb intergenic region (Young et al. 2002; Cromie et al. 2005). Nevertheless, the breaks within that large region are distributed over a much narrower ∼2-kb region. And within that smaller region, the breaks are further focused to four discrete clusters (Cromie et al. 2005). What directs DSBs to these particular sites is unknown. However, a 9-bp CRE motif does predict the location of multiple DSB sites scattered across the S. pombe genome (Steiner and Smith 2005a). This sequence is a binding site for the Atf1-Pcr1 transcription factor, which is required for its hotspot activity (Kon et al. 1997). We propose that there may be many other unrelated short sequence motifs that are responsible for many, and potentially most or all, of the DSB hotspots in S. pombe and perhaps other organisms. However, such hotspot motifs alone are unlikely to be sufficient for hotspot activity at all sites in the genome, since DSBs are not observed at all CRE sites in the genome (Steiner and Smith 2005a). Thus, other factors, such as chromatin structure, also play a role in promoting or permitting hotspots at particular sites. In Figure 6, we propose a model that recombination hotspots are found at positions where these two factors, a hotspot sequence motif and permissive chromatin structure, intersect.

Figure 6.—

Model to explain the location of recombination hotspots. The large rectangle indicates a portion of a chromosome. Shaded regions indicate regions of the genome that are permissive for recombination hotspots. In S. pombe, these regions coincide primarily with large intergenic regions (Cromie et al. 2007). Solid lines indicate hotspot sequence motifs. Hotspots occur where these two chromosomal features coincide.

How do sequence motifs create recombination hotspots? One possibility is that the nucleotide sequence itself possesses some property that makes it unusually susceptible to cleavage during meiosis. For example, tandem repeats of a pentanucleotide sequence reported to inhibit nucleosome formation can create a recombination hotspot in S. cerevisiae (Kirkpatrick et al. 1999). It has also been reported that polypurine/polypyrimidine tracts (PPTs) of ≥12 bp are associated with hotspots in S. cerevisiae (Bagshaw et al. 2006). However, we found no extensive tandem repeats and only a handful of PPTs ≥12 bp among our hotspot sequences (Table S1). Further, a direct test of a 30-bp random PPT in our experimental system did not produce hotspots in any of eight independent transformants (W. Steiner, unpublished observation). Thus, we favor instead the model that most of the sequences in our hotspot library contain target sequences for DNA binding proteins, for example transcription factors, that promote DSBs when bound to DNA. This model is also consistent with current data. First, we are aware of only two previously reported examples of defined sequence motifs creating recombination hotspots, the CRE hotspot of S. pombe (Schuchert et al. 1991) and the Bas1 target sequence, TGACTC, of S. cerevisiae (Mieczkowski et al. 2006). Both of these motifs require the binding of a transcription factor for their hotspot activity (Kon et al. 1997; Mieczkowski et al. 2006). Second, it has been recently observed in S. pombe that hotspots of recombination show significant colocalization to sites expressing noncoding RNAs (Wahls et al. 2008), suggesting that these sites are bound by transcription factors.

How might transcription factors promote recombination when bound to their target sequence? Yamada et al. (2004) showed that the ade6-M26 hotspot was dependent on both a histone acetyl transferase (Gcn5) and an ATP-dependent chromatin remodeling factor (Snf22). These researchers suggested that binding of the Atf1-Pcr1 transcription factor to the M26 motif recruits these chromatin modifying enzymes, resulting in localized chromatin remodeling and making the site accessible to the DSB machinery. It would be interesting to determine whether other hotspot-associated transcription factors operate by a similar mechanism, and what such factors might have in common that results in the recruitment of chromatin modifying enzymes.

It is possible that simple sequence motifs produce hotspots in many different organisms. For example, several sequence motifs have been reported as potential hotspots in humans (Zhang et al. 2004; Myers et al. 2005, 2008). Since recombination hotspots disrupt gene linkages, they complicate efforts to find human disease genes by linkage analysis (Hey 2004; Nishant and Rao 2005). Thus, the ability to identify potential hotspots solely on the basis of sequence is of practical significance. It is possible that some hotspots we have identified in our analysis are also active in other organisms. Two of the motifs reported as potential human hotspots, CGCCCCCGC and CCCCACCCC, show strong similarity to motifs found in our screen, motifs 7-4 and 7-8, respectively (Table 2), at least one of which (7-4) was confirmed to be a hotspot (Figure 3 and Table S4). This result suggests the intriguing possibility that some of the motifs identified here may also be active in other organisms.