Skip to content

Interference Genetics Example Essay

Genetic linkage mapping

Genetic mapping or linkage mapping in plants is commonly based on the measurement of co-segregation patterns of hundreds to millions of polymorphic molecular markers evaluated for hundreds of individuals from a biparental cross (a mapping population). The process of constructing a genetic map is based on statistical algorithms involving a stepwise approach: first grouping of markers on linkage groups, then ordered of markers within these groups and finally estimating the distance between the markers (Cheema and Dicks, 2009). Genetic map distances are usually expressed as recombination frequency in centimorgans (cM). One application of genetic maps is the identification of genomic regions linked to quantitative traits termed quantitative trait locus/loci (QTL) mapping. Genetic maps can also provide the basis for map-based cloning of major genes involved in important agronomic traits and the development of markers for MAS. Other applications include the dissection of genome and trait organization, the understanding of evolutionary relationships between and within species and the assistance in genome sequence assembly.

The first genetic map was constructed in 1913 using phenotyping markers applied to a segregating population of the fruit fly Drosophila melanogaster. The rapid evolution of DNA manipulation technology since the early 1980s led to a steady increase of the application of DNA-based molecular markers in diversity and genetic mapping studies in many crop species (Schlötterer, 2004). In the beginning of genetic mapping studies since 1988, RFLP markers have been intensively used to analyse the genome structure and evolutionary relationship between Brassica species (Figdore et al., 1988). In B. napus RFLP markers were applied for the first time by Landry et al. (1991) to construct a genetic linkage map from an F2 segregating mapping population. A few years later, a landmark paper by Parkin et al. (1995) described the comparative mapping of RFLP markers between different Brassica species including A. thaliana.

Genetic linkage mapping requires the production of segregating mapping populations by crossing two parents with phenotypic difference(s) in at least one trait of interest. Common types of mapping populations in B. napusgenetic linkage mapping consist of F2, DH, recombinant inbred line (RIL) and backcross (BC) populations (Gali and Sharpe, 2012). A list of B. napus mapping populations used for genetic linkage mapping in the past has been made available by the MBGP at A summary on genetic linkage maps of Brassica species published from 2006 to 2012 is listed in Table 5–1 in Gali and Sharpe (2012). Genetic linkage maps published since 2012 for B. napus are listed in Table 16.3.

Applications of genetic linkage maps in B. napus include biparental QTL analyses, multiple population QTL analyses and integration, comparative mapping with related Brassica species and map-based cloning. In recent years a strong increase in the number of published biparental QTL and association studies can be observed for B. napus. Major traits which have been studied in B. napus in the last two years by QTL mapping and by association mapping are listed in Tables 16.3 and 16.5. B. napus traits studied within the last two decades using marker-based approaches include yield-related traits and heterosis, quality-related traits (seed oil, glucosinolates, fatty acid composition), flowering time and disease resistance traits (fungus and virus resistance). For a review also see Table 16.1 in Snowdon and Friedt (2004).

In recent years, the increasing availability of molecular markers of different types allowed map alignments for multiple biparental crosses within B. napus and other Brassica species using bridge or anchor markers for comparative genome and trait analyses. QTL analyses in multiple populations and integrative QTL or MetaQTL analysis for the production of consensus maps across multiple populations and comparison of QTL regions recently has become more and more common in B. napus (e.g. Raman et al., 2013; Wang et al., 2013; Zhou et al., 2014). QTL mapping in multiple populations directly targeted to breeding approaches has also been undertaken (Würschum et al., 2012). Comparative genetic linkage mapping using genetic and genomic resources from related species such as B. rapa and A. thaliana has shown to be very useful in the genetic and physical fine mapping of a number of agronomically important gene loci, e.g. for disease resistance (Mayerhofer et al., 2005), male sterility (Xie et al., 2012; Lu et al., 2013) and seed oil content (Zhao et al., 2012). For the study of yield-related genes in B. napus a comparative genomics approach using information from the rice genome has been applied (Li et al., 2012). Genetic linkage mapping combined with comparative genomics has been suggested to be very useful for map-based cloning and marker development from functional genes involved in the trait(s) of interest.

Citation: Anderson CM, Oke A, Yam P, Zhuge T, Fung JC (2015) Reduced Crossover Interference and Increased ZMM-Independent Recombination in the Absence of Tel1/ATM. PLoS Genet 11(8): e1005478.

Editor: Michael Lichten, National Cancer Intitute, UNITED STATES

Received: January 27, 2015; Accepted: July 31, 2015; Published: August 25, 2015

Copyright: © 2015 Anderson et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited

Data Availability: Besides the data contained in the paper, we have submitted all sequences already to the Sequence Read Archive under accession number SRP044001. We have uploaded the rest of the analysis files in Dryad under accession numbers SRP028549 (wild type) and SRP041214 (all other strains).

Funding: This work was supported by National Institute of Health grant: 5R01GM0972 and The Alfred P. Sloan Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.


Sexual reproduction depends on meiosis, a specialized type of cell division that produces haploid gametes from diploid cells. Recombination between homologous chromosomes is a key feature of the first meiotic division. A subset of recombination events creates reciprocal exchanges known as crossovers (COs), which help ensure that homologs segregate properly in meiosis I. Recombination also includes non-reciprocal events called noncrossovers (NCOs). The number and distribution of COs are highly regulated to ensure proper chromosome segregation. A striking feature of the CO landscape is the non-random spacing of COs, a phenomenon known as interference (reviewed in [1]). As a result of interference, COs tend to be relatively evenly spaced along chromosomes. Although interference was first reported over a century ago as the decreased probability that a CO would occur if another CO occurred nearby [2], its mechanistic underpinnings are still not well understood.

Both COs and NCOs arise from double-strand DNA breaks (DSBs) induced by the Spo11 enzyme [3]. How each DSB’s fate is determined is poorly understood, but several findings indicate that a decision is made prior to formation of stable strand invasion intermediates [4,5,6]. Formation of both COs and NCOs begins with resection of DSBs to expose 3’ single-stranded tails that can invade homologous duplex DNA (Fig 1A). At sites of future COs, initial strand invasion is followed by formation of stable intermediates known as single-end invasions and double Holliday junctions (dHJs) [4,6]. Normal timing and levels of these CO-specific intermediates require the ZMM proteins (Zip2-Zip3-Zip4-Spo16, Msh4-Msh5, Mer3) [5]. Upon pachytene exit, dHJ-containing intermediates are resolved to form COs. In contrast, NCOs appear prior to pachytene exit, without formation of stable intermediates, and without the need for ZMMs [4,5,6]. Thus COs and NCOs show distinct timing, intermediates, and genetic dependencies, but how the repair pathway is initially chosen at each DSB is unknown.

Fig 1. Overview of meiotic recombination.

A) Major recombination pathways. A Spo11-induced DSB is resected to expose single-stranded tails. A 3’ tail invades a homologous duplex and is extended using the homolog as a template. Displacement of the invading strand leads to NCO formation by synthesis-dependent strand annealing (SDSA). Alternatively, capture of the second DSB end leads to formation of a dHJ. In wild type, dHJs are typically resolved as COs, but NCO formation is also possible. B) CO patterning. During or soon after DSB formation, a subset of DSBs becomes committed to the CO fate. These sites are marked by SICs and show interference. Subsequent steps convert CO-committed sites into COs. The majority of non-SIC-marked sites become NCOs, but some of them may also become COs.

In budding yeast, a subset of COs is associated with cytologically observed foci known as synapsis-initiation complexes (SICs) [7,8]. SICs contain the ZMM proteins and appear to promote polymerization of the synaptonemal complex (SC). Multiple lines of evidence indicate that SICs form at CO-committed sites. [9,10,11,12]. SICs, like COs, show interference [9,13,14,15,16]. Strikingly, however, in certain deletion mutants the distribution of SICs (cytological interference) is normal even though CO interference as assessed genetically is defective (e.g. zip1Δ, msh4Δ, and sgs1Δ) [9]. Based on these findings a two-phase model for establishment of CO interference has been proposed (Fig 1B) [5,9]. First, DSBs are formed and designated as future sites of COs or NCOs, with SICs marking CO-committed sites. Second, these sites are processed into their respective products. According to this model zip1Δ, msh4Δ, and sgs1Δ cause defects in the implementation phase without disrupting the initial CO/NCO decision. SICs thus provide a readout of repair pathway choice.

Formation of SICs requires the presence of Spo11-induced DSBs [8,10]. SICs are seen in the processing-defective rad50S strain, in the recombination-defective dmc1Δ strain, and in haploid cells, indicating that normal DSB processing and interhomolog recombination are not required for SIC formation [7,8,17,18], thus prompting us to ask whether recombination pathway choice hinges on events immediately after break induction.

In mitotic cells, where the response to DSBs has been extensively characterized, the earliest known events after DSB formation are the binding and activation of proteins involved in the DNA damage response, including Mre11-Rad50-Xrs2 (MRX), Tel1, Mec1, and the 9-1-1 complex (Ddc1-Mec3-Rad17 in budding yeast) [19]. MRX and Tel1 are recruited to unresected DSBs, while Mec1 and 9-1-1 respond to single-stranded DNA (ssDNA). Since SICs are seen in the processing-defective rad50S mutant, we reasoned that Tel1, which responds to unprocessed DSBs, might play a role in SIC formation.

Tel1/ATM is known to control meiotic DSB levels. In mice, loss of ATM causes a dramatic increase in DSB frequency [20]. In flies, mutation of the ATM ortholog tefu causes an increase in foci of phosphorylated H2AV, suggesting an increase in meiotic DSBs [21]. Measurements of DSB frequency in tel1Δ yeast have given conflicting results, with three studies showing an increase [22,23,24] and two showing a decrease [25,26]. All but one of these studies relied on mutations that prevent DSB repair (rad50S or sae2Δ) to enhance detection of DSBs. These mutations may themselves influence the number and distribution of DSBs, confounding interpretation of the results. The one study that examined DSB levels in tel1Δ single mutants found a convincing increase in DSBs [23].

Tel1/ATM also influences the outcome of recombination. In mice, loss of ATM causes meiotic arrest due to unrepaired DSBs [27,28,29]. Infertility due to a failure to produce mature gametes is a feature of the human disease ataxia telangiectasia, suggesting that ATM is also required for meiotic DSB repair in humans. Meiotic progression in Atm−/− mice can be partially rescued by heterozygosity for Spo11 [30,31]. Compared to Spo11+/− alone, Spo11+/−Atm−/− spermatocytes show synapsis defects and higher levels of MLH1 foci, a cytological marker for COs [30]. In these spermatocytes the spacing of MLH1 foci is less regular and the sex chromosomes often fail to form a CO in spite of greater overall CO frequency. These results point to a role for ATM in regulating the distribution of COs. In yeast, examination of recombination intermediates at the HIS4LEU2 hotspot found that Tel1 is required for efficient resection of DSBs when the overall number of DSBs genome wide is low [32]. Under these conditions, the preference for using the homolog as a repair template was decreased in the absence of Tel1.

Tel1 also regulates DSB distribution (reviewed in [33]). In budding yeast DSBs are distributed non-uniformly throughout the genome, falling into large “hot” and “cold” domains spanning tens of kb, as well as smaller hotspots of a few hundred bp or less [3]. DSBs, like COs, are thought to show interference. Direct measurement of DSBs at closely spaced hotspots found that the frequency of double cuts on the same chromatid was lower than expected under a random distribution [23]. These calculations could only be done in repair-defective mutants due to detection issues, but nevertheless provide the most compelling evidence to date of DSB interference. This study found that DSB interference in yeast depends on TEL1. The existence of DSB interference was originally proposed based on the observation that introduction of a new hotspot greatly reduces DSB frequency in nearby areas [34,35,36,37]. It remains unknown whether this hotspot-hotspot competition and DSB interference represent the same phenomenon. A careful examination of recombination products at the HIS4LEU2 hotspot found evidence that DSBs also inhibit each other in trans, i.e. between chromatids, and that trans inhibition depends on Tel1 [24]. The authors proposed that spreading of trans inhibition along chromosomes could contribute to even spacing of DSBs.

Several proteins with key meiotic roles are subject to Tel1/Mec1-dependent phosphorylation, although in many cases the individual contribution of Tel1 (separate from Mec1) has not been tested. These include the axial protein Hop1, the Spo11 accessory factor Rec114, histone H2A, Sae2, and Zip1 [22,38,39,40]. Tel1-dependent phosphorylation of Rec114 may at least partially account for Tel1 regulation of DSB levels, although this has yet to be definitively tested [22]. Loss of Tel1 causes only a mild defect in spore viability and little or no delay in meiotic progression [39,41].

Multiple lines of evidence indicate that interactions between homologs influence DSB formation (reviewed in [42]). Experiments in worms first led to the proposal that nascent COs inhibit additional DSBs on the same chromosome [43,44]. This mechanism would allow DSB formation to continue until each chromosome has achieved a CO. Studies of worms, mice, and yeast indicate that some aspect of homolog engagement, possibly SC formation, leads to inhibition of DSBs [45,46,47,48]. High-resolution mapping of DSBs in synapsis-defective yeast found a change in the genome-wide distribution of DSBs in populations of cells [47]. To our knowledge, no previous studies have assessed whether regular spacing of DSBs along individual chromosomes is dependent on synapsis or other interhomolog interactions.

Our lab and others have developed techniques for mapping recombination products genome-wide in budding yeast [49,50,51,52]. We mate two yeast strains, S96 and YJM789, with sequence differences at about 65,000 sites. After recovery of the four progeny of a single meiosis, we use next-generation sequencing or microarrays to genotype progeny. The resulting map allows us to deduce the locations of all COs and nearly all NCOs with a median resolution of 81 bp.

Using this technique, we show here that loss of Tel1 causes an increase in recombination along with decreases in CO interference and the CO/NCO ratio. Yet the number of SICs in tel1Δ cells is similar to wild type, and these SICs show normal interference. These results suggest that in the absence of tel1Δ, a substantial number of COs arises from a ZMM-independent pathway. Our analysis of recombination in tel1Δ zip3Δ confirms this conclusion. Furthermore, we also see a change in the distribution of all recombination products in tel1Δ, zip3Δ, msh4Δ and sgs1Δ, which we infer indicates a change in DSB distribution. Since SIC distribution is normal in these strains (except zip3Δ, which lacks SICs) this result implies that DSB interference is not required for proper patterning of CO precursors. We argue that the opposite is true: the CO patterning process contributes to DSB interference, as CO-designated sites repress formation of additional DSBs in surrounding areas.


Loss of Tel1 increases recombination and alters its outcome

To investigate the role of Tel1 in meiotic recombination, we identified recombination products genome-wide in the progeny of 14 tel1Δ hybrid diploids. Eight tetrads were genotyped at high resolution by next-generation sequencing and used for analysis of both NCOs and COs, while six were genotyped at lower resolution and used for analysis of COs only. As wild-type controls, we used data from 46 tetrads genotyped by high-density tiling array [51] and six wild-type tetrads sequenced in our lab [53]. As expected based on analysis of recombination at a single hotspot [24], deletion of TEL1 significantly increases the overall rate of recombination (Fig 2A). This finding is also consistent with reports that DSB levels are increased in tel1Δ [22,23,24], although our data should be taken only as a rough estimate of DSB levels, since not all DSBs produce detectable products. In addition, there is potential for selection bias in our results since we are only able to assay cells that complete meiosis and produce viable spores. In the case of tel1Δ this bias is expected to be mild since the defects in sporulation and spore viability are quite modest (Fig 3E). We find that NCOs are increased disproportionately: the mean number of NCOs per tetrad increases by 60%, while COs increase by only 23%, resulting in a significantly lower CO/NCO ratio in tel1Δ compared to wild type (Fig 2B; p = 0.009; Student’s t-test).

Fig 2. Absence of Tel1 alters the outcome of recombination.

A) The average number of COs, NCOs, and all events (COs + NCOs) per tetrad is shown. COs include event types E2, E3, E5, E6, and E7 as defined in [53] and Fig 3. NCOs include E1 and E4. B) The average ratio of COs to NCOs is shown for wt and tel1Δ. C) Histogram of distances between pairs of adjacent COs. D) Interference (1 –CoC) for COs in wild-type and tel1Δ tetrads. For each inter-interval distance, the CoC was calculated individually for all possible interval pairs genome-wide, and the average is plotted. For all plots, analysis of COs used data from 52 wild-type and 14 tel1Δ tetrads; analysis of NCOs and all events used data from 52 wild-type and eight tel1Δ tetrads. Error bars: standard error (SE).

Fig 3. tel1Δ and sgs1Δ show distinct recombination phenotypes.

A) The average number of each product type is shown. Event types are as defined in [53]. “Disc” = discontinuous. B) The average number of COs, NCOs, and all events is shown. COs include E2, E3, E5, E6, and E7. NCOs include E1 and E4. Plots of all contributing event types are in S2 Fig. C) The average length of GC tracts at simple COs (E2) is shown. D) Histogram of the lengths of simple NCOs (E1). Error bars in all plots: SE. For all plots except analysis of COs in part B, data were derived from 52 wild-type, eight tel1Δ, nine sgs1Δ, seven zip3Δ, six zip3Δ tel1Δ, and six zip3Δ sgs1Δ tetrads. Analysis of CO frequency in part B used an additional set of six tel1Δ, four sgs1Δ, and 23 zip3Δ tetrads genotyped at lower resolution. Calculations of E8s in wild type used only the six wild type tetrads sequenced in our lab (see Materials and Methods). E) Sporulation frequency was measured in three independent cultures of each genotype, with the exception of sgs1Δ for which only two cultures were used. At least 300 cells per culture were counted. Average and SE are shown. The distribution of spores per ascus is shown in S3D Fig. Viability was measured for at least 200 tetrads per genotype.

The lower CO/NCO ratio in tel1Δ suggests that loss of Tel1 alters repair pathway choice. Another readout of pathway choice is CO interference, which refers to the relatively even spacing of COs in wild type. One way to assess interference is to analyze distances between adjacent COs (Fig 2C). In wild-type cells, inter-CO distances are well fit by a gamma distribution [50]. The value of the shape parameter γ of the best-fit distribution indicates the strength of interference, with γ > 1 indicating positive interference and γ = 1 indicating random distribution. γ is reduced from 2.0 in wild type to 1.6 in tel1Δ (Fig 2C), revealing a decrease in interference. Since γ is sensitive to changes in CO density, we also analyzed interference using the coefficient of coincidence (CoC) method in which the frequency of COs in two intervals is compared with the expected frequency of double COs under an assumption of no interference. Interference expressed as 1 –CoC also shows a significant decrease in tel1Δ (Fig 2D; p < 0.0001, chi-square test).

Higher frequency of complex products in tel1Δ

tel1Δ cells show a striking increase in complex products containing discontinuous gene conversion (GC) tracts or genotype changes on multiple chromatids (Fig 3A). To classify recombination products, we merge changes within 5 kb of each other into a single “event” that is assigned to one of eight event types (E1-E8). Except for E8, all of the types were previously defined [53]. 5 kb was chosen based on prior analysis of wild-type tetrads showing that events within 5 kb have distinct properties suggesting they arise from a single DSB [49]. All figures use a 5 kb threshold for merging unless otherwise specified. Results without merging are qualitatively similar and are shown in S1, S2B, S3, S6 and S8A Figs. In our classification system, “simple NCOs” (E1) and “simple COs” (E2) are products without any other genotype switches within 5 kb. A “CO with discontinuous GC” (E3) is a CO with a nearby GC tract (within 5 kb). A “discontinuous NCO” (E4) contains two GC tracts within 5 kb of each other. We also identify three categories of “minority” events (E5-E7), which are ambiguous products that could arise more than one way. For example, a minority event on three chromatids (E6) could be two closely spaced COs or a CO with a nearby NCO. In the current study we add a new category, E8, containing 4:0 tracts. These may represent cases of two overlapping NCOs, or may arise from pre-meiotic recombination. In wild type, complex events (categories E3-E7) account for about 14% of all meiotic recombination products; in tel1Δ they represent 22% of products, a statistically significant difference (p < 0.0001, Student’s t test). We see a similar increase in complex events in sgs1Δ ([53] and Fig 3A). The phenotypes of tel1Δ and sgs1Δ show several other similarities. Both mutants have higher recombination frequency, a decrease in the CO/NCO ratio, and a moderate decrease in CO interference [53,54,55,56]. These similarities suggested that Tel1 and Sgs1 might act together in regulating recombination.

Sgs1 and Tel1 have distinct roles in regulating recombination

Sgs1 is thought to control recombination pathway choice by unwinding nascent strand invasion intermediates unless they are protected by ZMMs [54]. Deletion of SGS1 rescues CO levels in zmm mutants [54,55]. We find that tel1Δ and sgs1Δ rescue crossing over in zip3Δ to similar extents (Fig 3B). However, in other ways tel1Δ and sgs1Δ show dramatically different phenotypes. First, loss of Zip3 causes a striking increase in NCOs. This increase is largely suppressed by sgs1Δ but not by tel1Δ (Fig 3B). Second, zip3Δ displays abnormally long GC tracts associated with COs (Fig 3C and [53]). This tract lengthening is suppressed by sgs1Δ [53] but only partially suppressed by tel1Δ. Third, a notable feature of recombination in sgs1Δ is the presence of a population of very short NCOs that we propose arise from aberrant SDSA [53]. This cohort of short NCOs is not seen in tel1Δ (Fig 3D). Together these results indicate that Sgs1 and Tel1 have distinct roles in regulating recombination.

SIC abundance and interference are similar in wild type and tel1Δ

To determine whether Tel1 acts upstream or downstream of SIC formation we measured the number and positions of Zip3 foci on chromosome IV or on all chromosomes in pachytene spreads of wild type and tel1Δ (Fig 4A). We find that tel1Δ cells show no increase in Zip3 foci compared to wild type in spite of greater numbers of COs and DSBs (Fig 4B and 4C). Since the number of foci in tel1Δ could be underestimated if foci are less intense and thus more difficult to detect, we determined whether the intensity of foci is similar in wild type and tel1Δ. By mixing both strains on a single slide, we control for slide-to-slide variation in staining. The two strains were labeled with arrays of tet operators on chromosomes of dramatically different size, allowing the genotype of individual cells to be identified after imaging. We find that the intensity of Zip3 foci in tel1Δ is slightly higher than in wild type (Fig 4D), indicating that the lack of increase in focus abundance is not caused by detection problems. Detection of Zip3 foci could also be impaired if foci are closer together in tel1Δ, causing adjacent foci to appear as a single merged focus. However, we find that the median distance between pairs of adjacent foci is similar in the two strains (0.42 μm in wild type vs. 0.44 μm in tel1Δ, a difference that is not statistically significant (S4A Fig)). We would also expect an increase in focus size if many more adjacent foci were unresolvable in tel1. This is not the case since the size of individual foci is the same in the two strains (S4B Fig). Together these results indicate that tel1Δ does not cause an increase in Zip3 foci. Zip3 foci in tel1Δ also show normal interference as determined by CoC analysis (Fig 4E).

Fig 4. SIC abundance and interference in tel1Δ are similar to wild type.

A) Meiotic chromosomes from wild type and tel1Δ were spread and labeled with antibodies to Zip1 (red) and Zip3-GFP (green). An array of tet operators on the right arm of chromosome IV was identified by coexpression of a tetR-mCherry fusion; the native mCherry signal was used for visualization. Scale bar: 2 μm. B) The number of Zip3-GFP foci on chromosome IV in 204 wild-type and 202 tel1Δ nuclei with full synapsis. Data are pooled from five independent experiments using two independent isolates of each strain. Small but significant decreases in the average number of foci on chromosome IV (7%), and in the total number of foci per cell (3%) were observed for tel1Δ. Individual experiments are shown in S5 Fig. Significance is lost if the most striking experiment is removed. Bars: mean and standard deviation (SD). C) Number of Zip3 foci per cell determined by automated focus finding in ImageJ, using the same images scored in B. Three of five contributing experiments showed a decrease in Zip3 foci, and the difference is statistically significant in two of the three; the other two experiments showed a slight increase in tel1Δ, one of which is statistically significant. Bars: mean and SD. D) Intensity of Zip3 foci. Wild-type and tel1Δ cells marked with arrays of tet operators on either chromosome IV or XIV were mixed on the same slide. The genotype of each cell was identified based on the size of the labeled chromosome. Four different slides were analyzed with equal numbers of wild-type and tel1Δ images from each slide (25 cells of each genotype total). Both labeling schemes (wild type marked on chromosome XIV and tel1Δ marked on chromosome IV, and vice-versa) gave similar results; figure includes data from both sets of strains. Total intensity of each focus is plotted. E) Interference calculated as 1-CoC for Zip3-GFP foci on chromosome IV from 72 wild-type and 76 tel1Δ cells. Error bars: SE. F) Chromosome IV Zip1-stained SC lengths in 209 wild-type and 212 tel1Δ nuclei. Data are pooled from five independent experiments using two independent isolates of each strain; each of the five individual experiments shows a decrease in SC length in tel1Δ, and the difference is statistically significant in two of the five. Individual experiments are shown in S5 Fig. Bars: mean and SD.

SC length has been shown to correlate with the number of cytologically distinguishable CO-committed sites in worms and mammals [57,58] and not necessarily with the total CO number [59,60,61]. We find that the mean length of chromosome IV SC is 6% shorter in tel1Δ than in wild type (Fig 4F; p = 0.0004, Student’s t test). Thus in yeast, SC length parallels the number of SICs and not the overall number of COs.

The lack of increase in SIC abundance in tel1Δ is unexpected because three previously tested mutants with higher levels of COs (sgs1Δ, pch2Δ, and ndj1) had more SICs, while mutants with fewer COs (msh4Δ and zip1Δ) had fewer SICs [9,17]. By comparing the number of COs on chromosome IV in our recombination mapping experiments with the number of Zip3 foci on chromosome IV, we calculated a ratio of SICs to COs (Fig 5A). This ratio should be viewed as a rough estimate, since the measurements of SICs and COs were performed in different strains (isogenic and hybrid diploids, respectively). In wild type, the SIC/CO ratio is 0.63, implying that the majority of COs occur at SICs. In tel1Δ this ratio is reduced to 0.40, suggesting that non-SIC-associated COs are the major class. For comparison we determined the SIC/CO ratio in two other mutants with increased CO levels, sgs1Δ and ndj1Δ. For this analysis we compared the number of Zip2 foci on chromosome XV with the number of COs on that chromosome, both from published studies [9,50]. We find no significant change in the SIC/CO ratio in these mutants compared to wild type (Fig 5B). These results reveal a specific role for Tel1 in regulating the fraction of SIC-associated COs.

Fig 5. COs are less Zip3 dependent in tel1Δ.

A) The average number of Zip3-GFP foci on chromosome IV detected on spreads (as in Fig 4) divided by the average number of COs on chromosome IV in genotyped tetrads (as in Fig 2A). B) The average number of Zip2 foci on chromosome XV detected on spreads [9] divided by the average number of COs on chromosome XV in genotyped tetrads (this study and [50].) C) Meiotic chromosomes from rad50S cells prepared as in Fig 4A. D) The average number of COs genome wide expressed as a percent of all interhomolog events. Per-tetrad averages are shown. E) Pachytene spreads stained with anti-Red1 antibody to detect axes. Three examples are shown for each genotype. Error bars: SE.

We considered the possibility that the failure of tel1Δ cells to make more Zip3 foci than wild type might be caused by DSB processing defects. A role for Tel1 in resection of meiotic DSBs has been suggested [32,39,62] Yet high levels of Zip3 foci are seen in the resection-defective rad50S strain (Fig 5C and [7]). These results indicate that resected ends are not required for formation of SICs.

A larger share of COs in tel1Δ is ZMM-independent

Non-ZMM associated COs, often called Class II COs, are assumed to lack interference [63,64,65]. A possible reason for decreased CO interference in tel1Δ is that non-ZMM-associated COs, which represent a minority of events in wild-type cells, make up a larger share of events in tel1Δ. To further test this we compared the effect of deleting ZIP3 on CO abundance in wild type and tel1Δ (Fig 5D). To adjust for different DSB frequencies, we normalized CO numbers by expressing them as a percent of all interhomolog events. The percent of events resolved as COs drops from 72% in wild type to 39% in zip3Δ. As predicted, the decrease in COs between tel1Δ (67%) and tel1Δ zip3Δ (49%) is more modest. Thus COs in tel1Δ show less ZMM dependence than in wild type. An even more dramatic decrease in ZMM dependence is seen in sgs1Δ: CO frequency is similar in sgs1Δ (67%) and sgs1Δ zip3Δ (61%). We conclude that in tel1Δ, SICs are still at least partially functional in terms of promoting the CO fate, since loss of Zip3 in tel1Δ causes a decrease in COs. The opposite is true in sgs1Δ: SICs are either not fully functional or not functionally relevant in terms of promoting COs, since very little effect was seen upon deleting ZIP3.

tel1Δ does not cause pseudosynapsis in zip1Δ

In cells lacking the SC central element Zip1, synapsis is lost and axes are held together at a few sites per chromosome, termed axial associations. The exact nature of these links is unknown, but they are thought to correspond to SIC-marked sites [8]. In the zip1Δ sgs1Δ double mutant, axes are held closely together by a dramatic increase in the number of axial associations, a phenomenon referred to as pseudosynapsis [56]. Given the similar numbers of recombination products in tel1Δ and sgs1Δ (Fig 3A), we tested whether pseudosynapsis also occurs in zip1Δ tel1Δ. We find strikingly distinct phenotypes in zip1Δ sgs1Δ and zip1Δ tel1Δ (Fig 5E). In zip1Δ sgs1Δ, virtually no regions of axial separation are seen, whereas many sites of axis separation are visible in zip1Δ tel1Δ, similar to zip1Δ alone. This is consistent with the finding that SICs are increased in sgs1Δ but not in tel1Δ, and supports the idea that axial associations occur at SICs. Alternatively, the close association of axes in zip1Δ sgs1Δ may arise from aberrant structures, such as trapped recombination intermediates, found only in zip1Δ sgs1Δ and not in zip1Δ tel1Δ.

Analysis of all detectable recombination products suggests that DSB interference depends on Tel1, ZMMs, and Sgs1

To test whether Tel1 mediates DSB interference we examined the distribution of all recombination products in our tel1Δ tetrads, using all interhomolog events as a proxy for DSBs. A potential concern relating to this analysis is that we are unable to detect some recombination events. These include intersister events, estimated to arise from 15–30% of all DSBs [66], and NCOs falling between markers or in which mismatch repair restored the original genotype, together estimated to include 30% of interhomolog NCOs [51]. However, failure to detect a percentage of the DSB population per se should not affect the calculated strength of interference since CoC does not vary significantly with event density [15], a fact that we verified by randomly removing events from a wild-type data set to simulate loss of detection (S7 Fig). The inability to detect some events would only be problematic if the undetected events were distributed non-uniformly throughout the genome. Previous analysis of the genome-wide distribution of COs and NCOs found good agreement between recombination frequencies in wild type and DSB frequencies in dmc1Δ [51], indicating that the distribution of detectable interhomolog events reflects the underlying DSB distribution.

We find that the distribution of all interhomolog events in wild type displays interference, and this interference is decreased (from 0.37 to 0.21) in tel1Δ (Fig 6A; p = 0.0007; chi-square test). We infer that Tel1 mediates DSB interference, in agreement with physical assays [23].

Fig 6. The distribution of recombination events is altered in tel1Δ, sgs1Δ, and zmmΔ.

A) Interference calculated as 1-CoC for a bin size and inter-interval distance of 25 kb is shown for COs only, NCOs only, or all events from whole-genome recombination data. msh4Δ data comprise seven tetrads sequenced in our lab and five tetrads genotyped by Mancera et al. [51]. B) Simulations were performed in which an interfering population of DSBs was first created, and then COs were selected from the DSBs. COs were selected either with or without additional interference. Remaining DSBs were considered NCOs. Failure to detect some events was simulated by removing 20% of all events and 30% of the remaining NCOs. Interference between all simulated DSBs or between “detectable” products is shown. Left: the strength of DSB interference was varied, and the strength of CO interference was selected to recapitulate observed interference between COs in wild type. Right: conditions were the same as on the left except no CO interference was incorporated. C) “Complex” events include the event types shown, and are events that could arise from more than one DSB. Randomized data consist of at least 10000 simulated tetrads per genotype in which the CO and GC tract positions in real tetrads were randomized. “With DSB landscape” indicates that event positions take into account DSB frequencies (see Materials and Methods). D) As in C, but includes only events involving four chromatids. Error bars: SE.

Unexpectedly, we find that the combination of all interhomolog products in zip3Δ, msh4Δ, and sgs1Δ also shows reduced interference (from 0.37 in wild type to 0.14, 0.11, and 0.21, respectively; p = 0.0003, 0.004, and 0.002 respectively). These results suggest that DSB interference is defective in these mutants. These three mutants are known to disrupt CO interference, but to our knowledge they have not been proposed to affect DSB-DSB spacing. Based on these results, we hypothesize that CO designation and/or formation of a SIC suppresses formation of DSBs nearby. Several previous studies point towards the existence of feedback between interhomolog interactions and DSB formation [43,44,45,46,47,48] and indicate that there is considerable temporal overlap between DSB and SIC formation [47,67,68]. We suggest that, beyond controlling the levels of DSBs, some aspect of CO designation also shapes the pattern of DSBs along individual chromosomes.

One potential question in interpreting these results is whether reduced interference among COs would automatically be expected to cause reduced interference among all detectable products, even without an underlying change in DSB interference. To test this we performed a simulation in which DSB interference was established entirely independently of CO interference. All DSB positions were first selected (with interference), and then CO positions were selected (with additional interference) from the DSBs, with the remaining DSBs becoming NCOs. We then randomly removed 20% of all events to simulate intersister repair, and 30% of the remaining NCOs to simulate loss of detection due to restoration and lack of markers. Results are shown for a wild-type level of CO interference with various levels of DSB interference (Fig 6B, left), and for the same conditions without CO interference (Fig 6B, right). These simulations illustrate several points. First, in the presence of CO interference, the strength of interference between all detectable recombination products is slightly higher than the true DSB interference among all four chromatids. This is due to preferential detection of COs (i.e., we detect essentially all COs, which strongly interfere, but we fail to detect some NCOs, which do not). Second, the level of interference between NCOs varies with the strength of DSB and CO interference. At low levels of DSB interference, selection of strongly interfering COs from an almost randomly spaced pool of DSBs results in NCOs that show negative interference, i.e. a tendency to cluster. At high levels of DSB interference, imposition of CO interference enhances the regular spacing of both COs and NCOs. In this model, to achieve a level of interference between all products equivalent to what is observed in wild type, it is necessary to impose strong DSB interference (1-CoC = 0.32). At this level of DSB interference, NCOs show strong interference. In contrast, NCOs in wild type do not show significant interference (Fig 6A). In wild type, interference for NCOs alone is 0.1, which does not differ significantly from no interference (p = 0.18). In addition, there are no statistically significant differences between wild type and any of the mutants in the strength of interference between NCOs. This lack of interference among NCOs lends support to the notion that DSB interference is at least partially driven by DSB suppression near COs. If DSB interference arose entirely independently of COs, we would expect NCOs to show interference.

Third, these simulations show that complete loss of CO interference only slightly reduces the interference among all detectable events (Fig 6B, compare left and right panels). This reduction is too small to account for the observed reductions in tel1Δ, zip3Δ, msh4Δ, and sgs1Δ.

It should be noted that in these simulations, DSB interference was applied to all four chromatids equally; i.e., a DSB on one chromatid suppressed DSBs equally along the same chromatid and along the three other chromatids, a situation that might not occur in vivo. We have separately simulated situations where DSB interference exclusively affects DSBs on the same chromatid or on the same pair of sister chromatids (S8B Fig). We found that it was not possible to recapitulate the observed strength of DSB interference among all four chromatids when the simulated DSB interference only affected DSBs on the same chromatid. Simulations in which DSB interference acted on a chromatid and its sister were capable of recapitulating the wild-type level of interference among all events on all chromatids, but this simulation again predicted much stronger interference among NCOs than is actually observed. In reality, DSB interference may arise from a combination of same-chromatid, intersister, and interhomolog effects, but our simulations suggest that none of these scenarios can account for the observation of very weak interference among NCOs if we assume DSB interference is entirely independent of CO designation. These results do not rule out that DSB interference may be partially created upstream of CO designation, but they suggest that such a mechanism does not solely account for the observed distribution of events.

Multi-chromatid recombination products in tel1Δ likely result from decreased DSB interference along with increased DSB frequency

A previous study of the HIS4LEU2 hotspot found many tetrads with multiple COs and/or GC tracts in both wild type and tel1Δ (20% and 36% of detectable recombination products, respectively) interpreted as arising from multiple DSBs [24]. To test whether the complex recombination events we observed in tel1Δ could be caused by closely spaced DSBs, we modeled a total loss of DSB interference by randomizing the positions of COs and GC tracts in our unmerged tel1Δ or wild-type data. GC tracts falling within the boundaries of a CO were not randomized since they are assumed to arise from the same DSB as the CO.

In the simulation, we incorporated the DSB landscape, such that the probability of an event falling in a particular area was determined by the frequency of DSBs in that region [69]. We then merged genotype changes within 5 kb into a single event and classified them as event types E1-E8. Zhang et al. [24] classified recombination products as T0, T1, or T2 based on the inferred number of initiating DSBs. We consider our event types E3-E8 as equivalent to T2 events (inferred to arise from two DSBs). Some of these event types could not be detected by Zhang et al. due to the limited number of markers available at HIS4LEU2. Surprisingly, we find that events inferred to arise from two DSBs occur more frequently in wild type than expected based on random chance (Fig 6C). If a specific mechanism existed to prevent these events, we would expect the opposite: these events should be more frequent in randomized data than in real tetrads. The high number of these events may reflect the fact that such events could arise from a single DSB; for example, three-chromatid events could result from two ends of a DSB invading different chromatids. Such multi-chromatid events were proposed to underlie the high number of complex products potentially arising from two DSBs in the sgs1-ΔC795 mutant [24]. Alternatively, DSBs in both wild type and tel1Δ might show negative interference, i.e. a tendency to cluster. If so, this effect would presumably operate only over short distances (less than 5 kb), since we see positive interference when genotype changes within 5 kb are treated as a single event (Fig 6A). In accordance with this, concerted formation of DSBs on the same chromatid within an approximately 8 kb range was observed in tel1Δ cells by a physical assay [23].

Due to the ambiguous origins of two- and three-chromatid events, we separately analyzed four-chromatid events (E7). We consider these more likely to be cases of more than one DSB occurring in trans (i.e. on different chromatids), since only a very aberrant recombination event could produce genotype switches on all four chromatids from a single DSB. We find that the frequency of four-chromatid events in wild type is significantly lower than the frequency expected due to random chance (Fig 6D; p = 0.0007; Student’s t test). In contrast, the frequency of these events in tel1Δ is statistically indistinguishable from the frequency expected due to random chance (Fig 6D; p = 0.78) These results support the conclusions of Zhang et al. that a Tel1-dependent mechanism suppresses the occurrence of more than one DSB per quartet of chromatids. As noted by Zhang et al. and Garcia et al. [23,24], trans inhibition could operate either between sister chromatids or between homologs. Our analysis of E7 products cannot distinguish between these two models, since we are unable to determine whether the initiating DSBs occurred on homologs or sisters. In theory, E8 products (4:0 tracts), which are increased in tel1Δ, may represent cases where DSBs occurred on both sisters. However, such products can also arise from premeiotic gene conversions. We find that the majority of E8 events have perfectly overlapping endpoints (i.e., gene conversion tracts beginning and ending at the same markers on both chromatids). Of the 4:0 tracts that are not part of a complex event, 72% (in wild type) or 74% (in tel1Δ) have perfect overlap. Such a high degree of overlap would not be expected if the majority of these events represented independent NCOs. Therefore we suspect that the tel1Δ-dependent increase in these events may arise from an increase in premeiotic recombination. Some, but not all, previous studies of recombination in vegetatively growing tel1Δ cells have found an increase [70,71,72].

Our simulations show that complex products arising from multiple DSBs are expected to occur more often in hot genome regions compared to cold regions (S8C and S8D Fig). This trend may explain the unusually high number of complex events seen by Zhang et al. at HIS4LEU2, an artificial hotspot with higher DSB frequency than natural hotspots.


Tel1 is involved in an early step in recombination pathway choice

Our data indicate that Tel1 is required for an early step in recombination pathway choice (Fig 7). In the absence of Tel1, the ratio of COs to NCOs, CO interference, and the dependence of COs on ZIP3 are all decreased, indicating that a greater proportion of recombination events occurs via non-ZMM-dependent mechanisms. The abundance of SICs is also similar to wild type, which is surprising given the higher levels of DSBs and COs in tel1Δ. Zhang et al [16] found modestly increased numbers of SICs in tel1Δ in the SK1 strain background, (11% increase on chromosome XV). Given the differences in strain backgrounds and chromosomes analyzed, these may represent essentially the same result. In SK1, the increase in SICs was smaller than the increase in DSBs (50% increase at HIS4LEU2 in a rad50S background) and COs (23% increase at HIS4LEU2) previously reported in SK1 [24]. Thus both studies point to the conclusion that the number of SICs per CO is reduced in tel1Δ.

Fig 7. Model for recombination pathway choice with and without Tel1.

A) In contrast to Fig 1 where DSB formation and CO designation were depicted as independent processes, we propose that formation of a SIC suppresses DSB formation nearby, so that later DSBs tend not to occur near a SIC. Early forming DSBs thus have a greater tendency to become interference-capable CO-designated sites and later DSBs tend to become NCOs or “non-interfering” COs. B) In tel1Δ, the number of DSBs is higher than in wild type and DSB distribution is less regular. A smaller fraction of DSBs becomes committed to the CO fate and marked by SICs; SICs still show an orderly distribution, as in wild type. DSBs not marked by SICs become NCOs or “non-interfering” COs, leading to decreased CO interference.

Taken together, our results suggest two non-mutually-exclusive mechanisms for the modulation of recombination by Tel1. One possibility is that in tel1Δ there are two distinct populations of DSBs: a normal cohort of DSBs repaired as in wild type, and a population of “excess” DSBs repaired via non-ZMM-dependent pathway(s). Another model consistent with our results is that tel1Δ causes a general defect in commitment of DSBs to the ZMM-dependent CO pathway. The wild-type-like number of foci in tel1Δ may be the net result of a decrease in SIC-forming ability partially offset by an increase in the abundance of DSBs. If Tel1 does promote SIC formation, other factors must have functional overlap with Tel1 in this role, since SICs show normal abundance in tel1Δ. We speculate that Tel1 phosphorylation of ZMMs may promote their recruitment to specific DSBs. All of the ZMM proteins contain multiple SQ/TQ sites, the consensus sequence for Tel1/Mec1 phosphorylation. Mutation of the four SQ/TQ sites in Zip3 reduces its association with DSB hotspots and reduces CO frequency in some intervals, suggesting its ability to form a SIC is impaired [11]. However, zip3-4AQ causes only a mild decrease in COs and no loss of spore viability, indicating that other relevant Tel1 targets in addition to Zip3 must exist.

Our results confirm that interference among CO-committed sites is not defective in tel1Δ, as previously reported [16]; instead, poor CO interference arises from the fact that many COs in tel1Δ occur via a non-ZMM pathway. Our analysis of recombination outcomes in tel1Δ zip3Δ provides experimental evidence for the prediction that in mutants with higher levels of DSBs without an increase in SICs, “extra” DSBs would be channeled into ZMM-independent repair pathways [15].

In previous observations of Atm−/− Spo11 +/− mouse spermatocytes [30], MLH1 served as a cytological marker for CO positions. Loss of ATM caused a decrease in interference between MLH1 foci, whereas Zip3 foci in yeast show normal interference (this study and [16]). MLH1 foci are often assumed to mark all COs rather than only ZMM-associated COs [73], although this view is not universally accepted (for example, [74,75,76].) If the view that MLH1 foci mark all COs is correct, the decreased interference between MLH1 foci would be consistent with our genome-wide mapping of tel1Δ recombination products, which showed decreased overall CO interference. Alternatively, ATM may play distinct roles in CO patterning in mammals and yeast.

COs are often categorized as Class I (ZMM-dependent) or Class II (Mus81-Mms4 dependent), with only Class I COs participating fully in CO interference. In tel1Δ the additional non-ZMM COs may be typical Class II COs dependent on Mus81-Mms4, or may form by another mechanism. Regardless of the mechanism, due to not participating in ZMM-dependent CO patterning, they would be expected to show decreased CO interference. Class II COs are often described as “non-interfering”, but as noted by Zhang et al. this terminology is probably inaccurate [16]. Since all sites of recombination are influenced by DSB interference, even Class II COs are expected to show weak interference.

Evidence for Tel1-mediated DSB interference

The distribution of all events in tel1Δ is consistent with a decrease in interference between DSBs. Effects of tel1Δ on DSB spacing have been previously reported [23,24], but it was not necessarily obvious that this would be detectable at the level of all recombination products genome wide. Garcia et al. observed a defect in DSB interference along single chromatids, but could not assay interference among all four chromatids in a homolog pair [23]. The genetic analysis by Zhang et al. observed trans inhibition among all four chromatids at a particular hotspot, but could not determine whether such inhibition extends laterally along chromosomes [24]. It is thus striking that a defect in interference among all recombination products is detectable in our data among all four chromatids and at distances of tens of kb. This supports the proposal of crosstalk between homologs in determining DSB positions [24].

Crossover designation may regulate DSB positioning

The distribution of all events in zip3Δ and msh4Δ also implies a decrease in interference between DSBs. The inferred decrease in DSB interference in zip3Δ and msh4Δ suggests that CO designation and/or formation of a SIC suppresses formation of DSBs nearby (Fig 7a). Consistent with this model, recent analysis of the genome-wide DSB distribution in a population of zip3Δ cells found that regions with the greatest change in DSB frequency in zip3Δ were enriched for Zip3 binding in wild type [47]. This strongly suggests that the influence of Zip3 on DSBs is at least partially a local effect, and is not solely attributable to chromosome-wide or nucleus-wide effects such as altering the timing of synapsis. Importantly, this model explains why CO-NCO pairs show interference while NCO-NCO pairs do not [51]. One implication of this model is that earlier-forming DSBs would have a greater tendency to become CO-designated sites compared to later-forming DSBs. In support of this, Zip3 localization is reduced at hotspots believed to represent late-forming DSBs [11]. A prediction of the model is that any mutation causing changes in SIC distribution or defects in SIC formation will also cause changes in DSB distribution. This may explain a recent observation in hed1Δ dmc1Δ cells, which have a reduced number of SICs. In this mutant CO distribution is altered such that the difference in recombination rates between adjacent hot and cold regions is diminished [