Author response:

The following is the authors’ response to the original reviews.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors investigated how the presence of interspecific introgressions in the genome affects the recombination landscape. This research was intended to inform about genetic phenomena influencing the evolution of introgressed regions, although it should be noted that the research itself is based on examining only one generation, which limits the possibility of drawing far-reaching evolutionary conclusions. In this work, yeast hybrids with large (from several to several dozen percent of the chromosome length) introgressions from another yeast species were crossed. Then, the products of meiosis were isolated and sequenced, and on this basis, the genome-wide distribution of both crossovers (COs) and noncrossovers (NCOs) was examined. Carrying out the analysis at different levels of resolution, it was found that in the regions of introduction, there is a very significant reduction in the frequency of COs and a simultaneous increase in the frequency of NCOs. Moreover, it was confirmed that introgressions significantly limit the local shuffling of genetic information, and NCOs are only able to slightly contribute to the shuffling, thus they do not compensate for the loss of CO recombination.

Strengths:

- Previously, experiments examining the impact of SNP polymorphism on meiotic recombination were conducted either on the scale of single hotspots or the entire hybrid genome, but the impact of large introgressed regions from another species was not examined. Therefore, the strength of this work is its interesting research setup, which allows for providing data from a different perspective.

- Good quality genome-wide data on the distribution of CO and NCO were obtained, which could be related to local changes in the level of polymorphism.

Weaknesses:

(1) The research is based on examining only one generation, which limits the possibility of drawing far-reaching evolutionary conclusions. Moreover, meiosis is stimulated in hybrids in which introgressions occur in a heterozygous state, which is a very unlikely situation in nature. Therefore, I see the main value of the work in providing information on the CO/NCO decision in regions with high sequence diversification, but not in the context of evolution.

While we are indeed only examining recombination in a single generation, we respectfully disagree that our results aren't relevant to evolutionary processes. The broad goals of our study are to compare recombination landscapes between closely related strains, and we highlight dramatic differences between recombination landscapes. These results add to a body of literature that seeks to understand the existence of variation in traits like recombination rate, and how recombination rate can evolve between populations and species. We show here that the presence of introgression can contribute to changes in recombination rate measured in different individuals or populations, which has not been previously appreciated. We furthermore show that introgression can reduce shuffling between alleles on a chromosome, which is recognized as one of the most important determinants for the existence and persistence of sexual reproduction across all organisms. As we describe in our introduction and conclusion, we see our experimental exploration of the impacts of introgression on the recombination landscape as complementary to studies inferring recombination and introgression from population sequencing data and simulations. There are benefits and challenges to each approach, but both can help us better understand these processes. In regards to the utility of exploring heterozygous introgression, we point out that introgression is often found in a heterozygous state (including in modern humans with Neanderthal and/or Denisovan ancestry). Introgression will always be heterozygous immediately after hybridization, and depending on the frequency of gene flow into the population, the level of inbreeding, selection against introgression, etc., introgression will typically be found as heterozygous.

- The work requires greater care in preparing informative figures and, more importantly, re-analysis of some of the data (see comments below).

More specific comments:

(1) The authors themselves admit that the detection of NCO, due to the short size of conversion tracts, depends on the density of SNPs in a given region. Consequently, more NCOs will be detected in introgressed regions with a high density of polymorphisms compared to the rest of the genome. To investigate what impact this has on the analysis, the authors should demonstrate that the efficiency of detecting NCOs in introgressed regions is not significantly higher than the efficiency of detecting NCOs in the rest of the genome. If it turns out that this impact is significant, analyses should be presented proving that it does not entirely explain the increase in the frequency of NCOs in introgressed regions.

We conducted a deeper exploration of the effect of marker resolution on NCO detection by randomly removing different proportions of markers from introgressed regions of the fermentation cross in order to simulate different marker resolutions from non-introgressed regions. We chose proportions of markers that would simulate different quantiles of the resolution of non-introgressed regions and repeated our standard pipeline in order to compare our NCO detection at the chosen marker densities. More details of this analysis have been added to the manuscript (lines 188-199, 525-538). We confirmed the effect of marker resolution on NCO detection (as reported in the updated manuscript and new supplementary figures S2-S10, new Table S10) and decided to repeat our analyses on the original data with a more stringent correction. For this we chose our observed average tract size for NCOs in introgressed regions (550bp), which leads to a far more conservative estimate of NCO counts (As seen in the updated Figure 2 and Table 2). This better accounts for the increased resolution in introgressed regions, and while it's possible to be more stringent with our corrections, we believe that further stringency would be unreasonable. We also see promising signs that the correction is sufficient when counting our CO and NCO events in both crosses, as described in our response to comment 39 (response to reviewer #3).

(2) CO and NCO analyses performed separately for individual regions rarely show statistical significance (Figures 3 and 4). I think that the authors, after dividing the introgressed regions into non-overlapping windows of 100 bp (I suggest also trying 200 bp, 500 bp, and 1kb windows), should combine the data for all regions and perform correlations to SNP density in each window for the whole set of data. Such an analysis has a greater chance of demonstrating statistically significant relationships. This could replace the analysis presented in Figure 3 (which can be moved to Supplement). Moreover, the analysis should also take into account indels.

We're uncertain of what is being requested here. If the comment refers to the effect of marker density on NCO detection, we hope the response to comment 2 will help resolve this comment as well. Otherwise, we ask for some clarification so that we may correct or revise as appropriate.

(3) In Arabidopsis, it has been shown that crossover is stimulated in heterozygous regions that are adjacent to homozygous regions on the same chromosome (http://dx.doi.org/10.7554/eLife.03708.001, https://doi.org/10.1038/s41467-022-35722-3).

This effect applies only to class I crossovers, and is reversed for class II crossovers (https://doi.org/10.15252/embj.2020104858 , https://doi.org/10.1038/s41467-023-42511-z). This research system is very similar to the system used by the authors, although it likely differs in the level of DNA sequence divergence. The authors could discuss their work in this context.

We thank the reviewer for sharing these references. We have added a discussion of our work in the context of these findings in the Discussion, lines 367-376.

Reviewer #2 (Public Review):

Summary:

Schwartzkopf et al characterized the meiotic recombination impact of highly heterozygous introgressed regions within the budding yeast Saccharomyces uvarum, a close relative of the canonical model Saccharomyces cerevisiae. To do so, they took advantage of the naturally occurring Saccharomyces bayanus introgressions specifically within fermentation isolates of S. uvarum and compared their behavior to the syntenic regions of a cross between natural isolates that do not contain such introgressions. Analysis of crossover (CO) and noncrossover (NCO) recombination events shows both a depletion in CO frequency within highly heterozygous introgressed regions and an increase in NCO frequency. These results strongly support the hypothesis that DNA sequence polymorphism inhibits CO formation, and has no or much weaker effects on NCO formation. Eventually, the authors show that the presence of introgressions negatively impacts "r", the parameter that reflects the probability that a randomly chosen pair of loci shuffles their alleles in a gamete.

The authors chose a sound experimental setup that allowed them to directly compare recombination properties of orthologous syntenic regions in an otherwise intra-specific genetic background. The way the analyses have been performed looks right, although this reviewer is unable to judge the relevance of the statistical tests used. Eventually, most of their results which are elegant and of interest to the community are present in Figure 2.

Strengths:

Analysis of crossover (CO) and noncrossover (NCO) recombination events is compelling in showing both a depletion in CO frequency within highly heterozygous introgressed regions and an increase in NCO frequency.

Weaknesses:

The main weaknesses refer to a few text issues and a lack of discussion about the mechanistic implications of the present findings.

- Introduction

(1) The introduction is rather long. | I suggest specifically referring to "meiotic" recombination (line 71) and to "meiotic" DSBs (line 73) since recombination can occur outside of meiosis (ie somatic cells).

We agree and have condensed the introduction to be more focused. We also made the suggested edits to include “meiotic” when referring to recombination and DSBs.

(2) From lines 79 to 87: the description of recombination is unnecessarily complex and confusing. I suggest the authors simply remind that DSB repair through homologous recombination is inherently associated with a gene conversion tract (primarily as a result of the repair of heteroduplex DNA by the mismatch repair (MMR) machinery) that can be associated or not to a crossover. The former recombination product is a crossover (CO), the latter product is a noncrossover (NCO) or gene conversion. Limited markers may prevent the detection of gene conversions, which erase NCO but do not affect CO detection.

We changed the language in this section to reflect the reviewer’s suggestions.

(3) In addition, "resolution" in the recombination field refers to the processing of a double Holliday junction containing intermediates by structure-specific nucleases. To avoid any confusion, I suggest avoiding using "resolution" and simply sticking with "DSB repair" all along the text.

We made the suggested correction throughout the paper.

(4) Note that there are several studies about S. cerevisiae meiotic recombination landscapes using different hybrids that show different CO counts. In the introduction, the authors refer to Mancera et al 2008, a reference paper in the field. In this paper, the hybrid used showed ca. 90 CO per meiosis, while their reference to Liu et al 2018 in Figure 2 shows less than 80 COs per meiosis for S. cerevisiae. This shows that it is not easy to come up with a definitive CO count per meiosis in a given species. This needs to be taken into account for the result section line 315-321.

This is an excellent point. We added this context in the results (lines 180-187).

(5) In line 104, the authors refer to S. paradoxus and mention that its recombination rate is significantly different from that of S. cerevisiae. This is inaccurate since this paper claims that the CO landscape is even more conserved than the DSB landscape between these two species, and they even identify a strong role played by the subtelomeric regions. So, the discussion about this paper cannot stand as it is.

We agree with the reviewer's point. We also found that the entire paragraph was unnecessary, so it and the sentence in question have been removed.

(6) Line 150, when the authors refer to the anti-recombinogenic activity of the MMR, I suggest referring to the published work from Martini et al 2011 rather than the not-yet-published work from Copper et al 2021, or both, if needed.

Added the suggested citation.

Results

(7) The clear depletion in CO and the concomitant increase in NCO within the introgressed regions strongly suggest that DNA sequence polymorphism triggers CO inhibition but does not affect NCO or to a much lower extent. Because most CO likely arises from the ZMM pathway (CO interference pathway mainly relying on Zip1, 2, 3, 4, Spo16, Msh4, 5, and Mer3) in S. uvarum as in S. cerevisiae, and because the effect of sequence polymorphism is likely mediated by the MMR machinery, this would imply that MMR specifically inhibits the ZMM pathway at some point in S. uvarum. The weak effect or potential absence of the effect of sequence polymorphism on NCO formation suggests that heteroduplex DNA tracts, at least the way they form during NCO formation, escape the anti-recombinogenic effect of MMR in S. uvarum. A few comments about this could be added.

We have added discussion and citations regarding the biased repair of DSB to NCO in introgression, lines 380-386.

(8) The same applies to the fact that the CO number is lower in the natural cross compared to the fermentation cross, while the NCO number is the same. This suggests that under similar initiating Spo11-DSB numbers in both crosses, the decrease in CO is likely compensated by a similar increase in inter-sister recombination.

Thank you to the reviewer for this observation. We agree that this could explain some differences between the crosses.

(9) Introgressions represent only 10% of the genome, while the decrease in CO is at least 20%. This is a bit surprising especially in light of CO regulation mechanisms such as CO homeostasis that tends to keep CO constant. Could the authors comment on that?

We interpret these results to reflect two underlying mechanisms. First, the presence of heterozygous introgression does reduce the number of COs. Second, we believe the difference in COs reflects variation in recombination rate between strains. We note that CO homeostasis need not apply across different genetic backgrounds. Indeed, recombination rate is appreciated to significantly differ between strains of S. cerevisiae (Raffoux et al. 2018), and recombination rate variation has been observed between strains/lines/populations in many different species including Drosophila, mice, humans, Arabidopsis, maize, etc. We reference S. cerevisiae strain variability in the Introduction lines 128-130, and have added context in the Results lines 180-187, and Discussion lines 343-350.

(10) Finally, the frequency of NCOs in introgressed regions is about twice the frequency of CO in non-introgressed regions. Both CO and NCO result from Spo11-initiating DSBs.

This suggests that more Spo11-DSBs are formed within introgressed regions and that such DSBs specifically give rise to NCO. Could this be related to the lack of homolog engagement which in turn shuts down Spo11-DSB formation as observed in ZMM mutants by the Keeney lab? Could this simply result from better detection of NCO in introgressed regions related to the increased marker density, although the authors claim that NCO counts are corrected for marker resolution?

The effect noted by the reviewer remains despite the more conservative correction for marker density applied to NCO counts (as described in the response to Reviewer 1, comment #2). Given that CO+NCO counts in introgressed regions are not statistically different between crosses, it is likely that these regions are simply predisposed to a higher rate of DSBs than the rest of the genome. This is an interesting observation, however, and one that we would like to further explore in future work.

(11) What could be the explanation for chromosome 12 to have more shuffling in the natural cross compared to the fermentation cross which is deprived of the introgressed region?

We added this text to the Results, lines 323-327, "While it is unclear what potential mechanism is mediating the difference in shuffling on chromosome 12, we note that the rDNA locus on chromosome 12 is known to differ dramatically in repeat content across strains of S. cerevisiae (22–227 copies) (Sharma et a. 2022), and we speculate that differences in rDNA copy number between strains in our crosses could impact shuffling."

Technical points:

(12) In line 248, the authors removed NCO with fewer than three associated markers.

What is the rationale for this? Is the genotyping strategy not reliable enough to consider events with only one or two markers? NCO events can be rather small and even escape detection due to low local marker density.

We trust the genotyping strategy we used, but chose to be conservative in our detection of NCOs to account for potential sequencing biases.

(13) Line 270: The way homology is calculated looks odd to this reviewer, especially the meaning of 0.5 homology. A site is either identical (1 homology) or not (0 homology).

We've changed the language to better reflect what we are calculating (diploid sequence similarity; see comment #28). Essentially, the metric is a probability that two randomly selected chromatids--one from each parent--will share the same nucleotide at a given locus (akin to calculating the probability of homozygous offspring at a single locus). We average it along a segment of the genome to establish an expected sequence similarity if/when recombination occurs in that segment.

(14) Line 365: beware that the estimates are for mitotic mismatch repair (MMR). Meiotic MMR may work differently.

We removed the citation that refers exclusively to mitotic recombination. The statement regarding meiotic recombination is otherwise still reflective of results from Chen & Jinks-Robertson

(15) Figure 1: there is no mention of potential 4:0 segregations. Did the authors find no such pattern? If not, how did they consider them?

The program we used to call COs and NCOs (ReCombine's CrossOver program) can detect such patterns, but none were detected in our data.

Reviewer #3 (Public Review):

When members of two related but diverged species mate, the resulting hybrids can produce offspring where parts of one species' genome replace those of the other. These "introgressions" often create regions with a much greater density of sequence differences than are normally found between members of the same species. Previous studies have shown that increased sequence differences, when heterozygous, can reduce recombination during meiosis specifically in the region of increased difference. However, most of these studies have focused on crossover recombination, and have not measured noncrossovers. The current study uses a pair of Saccharomyces uvarum crosses: one between two natural isolates that, while exhibiting some divergence, do not contain introgressions; the other is between two fermentation strains that, when combined, are heterozygous for 9 large regions of introgression that have much greater divergence than the rest of the genome. The authors wished to determine if introgressions differently affected crossovers and noncrossovers, and, if so, what impact that would have on the gene shuffling that occurs during meiosis.

(1) While both crossovers and noncrossovers were measured, assessing the true impact of increased heterology (inherent in heterozygous introgressions) is complicated by the fact that the increased marker density in heterozygous introgressions also increases the ability to detect noncrossovers. The authors used a relatively simple correction aimed at compensating for this difference, and based on that correction, conclude that, while as expected crossovers are decreased by increased sequence heterology, counter to expectations noncrossovers are substantially increased. They then show that, despite this, genetic shuffling overall is substantially reduced in regions of heterozygous introgression. However, it is likely that the correction used to compensate for the effect of increased sequence density is defective, and has not fully compensated for the ascertainment bias due to greater marker density. The simplest indication of this potential artifact is that, when crossover frequencies and "corrected" noncrossover frequencies are taken together, regions of introgression often appear to have greater levels of total recombination than flanking regions with much lower levels of heterology. This concern seriously undercuts virtually all of the novel conclusions of the study. Until this methodological concern is addressed, the work will not be a useful contribution to the field.

We appreciate this concern. Please see response to comments #2 and #38. We further note that our results depicted in Figure 3 and 4 are not reliant on any correction or comparison with non-introgressed regions, and thus our results regarding sequence similarity and its effect on the repair of DSBs and the amount of genetic shuffling with/without introgression to be novel and important observations for the field.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) Line 149 - this sentence refers to a mixture of papers reporting somatic or meiotic recombination and as these processes are based on different crossover pathways, this should not be mixed. For example, it is known that in Arabidopsis MSH2 has a pro-crossover function during meiotic recombination.

Corrected

(2) What is unclear to me is how the crosses are planned. Line 308 shows that there were only two crosses (one "natural" and one "fermentation"), but I understand that this is a shorthand and in fact several (four?) different strains were used for the "fermentation cross". At least that's what I concluded from Fig. 1B and its figure caption. This needs to be further explained. Were different strains used for each fermentation cross, or was one strain repeated in several crosses? In Figure 1, it would be worth showing, next to the panel showing "fermentation cross", a diagram of how "natural cross" was performed, because as I understand it, panel A illustrates the procedure common to both types of crosses, and not for "natural cross".

We thank the reviewer for drawing our attention to confusion about how our crosses were created. We performed two crosses, as depicted in Figure 1A. The fermentation cross is a single cross from two strains isolated from fermentation environments. The natural cross is a single cross from two strains isolated from a tree and insect. Table S1 and the methods section "Strain and library construction" describe the strains used in more detail. We modified Figure 1 and the figure legend to help clarify this. See also response to comment #37.

(3) The authors should provide a more detailed characterization of the genetic differences between chromosomes in their hybrids. What is the level of polymorphism along the S. uvarum chromosomes used in the experiments? Is this polymorphism evenly distributed? What are the differences in the level of polymorphism for individual introgressions? Theoretically, this data should be visible in Figure 2D, but this figure is practically illegible in the present form (see next comment).

As suggested, we remade Figure 2D to only include chromosomes with an introgression present, and moved the remaining chromosomes to the supplements (Figure S11). The patterns of markers (which are fixed differences between the strains in the focal cross) should be more clear now. As we detail in the Methods line 507-508, we utilized a total of 24,574 markers for the natural cross and 74,619 markers for the fermentation cross (the higher number in the fermentation cross being due to more fixed differences in regions of introgression).

(4) Figure 2D should be prepared more clearly, I would suggest stretching the chromosomes, otherwise, it is difficult to see what is happening in the introgression regions for CO and NCO (data for SNPs are more readable). Maybe leave only the chromosomes with introgressions and transfer the rest to the supplement?

See previous comment.

(5) How are the Y scales defined for Figure 2D?

Figure 2D now includes units for the y-axis.

(6) Are increases in CO levels in fermentation cross-observed at the border with introgressions? This would indicate local compensation for recombination loss in the introgressed regions, similar to that often observed for chromosomal inversions.

We see no evidence of an increase in CO levels at the borders of introgressions, neither through visual inspection or by comparing the average CO rate in all fermentation windows to that of windows at the edges of introgressions. This is included in the Discussion lines 360-366, "While we are limited in our interpretations by only comparing two crosses (one cross with heterozygous introgression and one without introgression), these results are in line with findings in inversions, where heterozygotes show sharp decreases in COs, but the presence of NCOs in the inverted region (Crown et al., 2018; Korunes & Noor, 2019). However, unlike heterozygous inversions where an increase in COs is observed on freely recombining chromosomes (the inter-chromosomal effect), we do not see an increase in COs on the borders flanking introgression or on chromosomes without introgression."

(7) Line 336 - "We find positive correlations between CO counts..." - you should indicate here that between fermentation and natural crosses, it was quite hard for me to understand what you calculated.

We corrected the language as suggested.

(8) The term "homology" usually means "having a common evolutionary origin" and does not specify the level of similarity between sequences, thus it cannot be measured. It is used incorrectly throughout the manuscript (also in the intro). I would use the term "similarity" to indicate the degree of similarity between two sequences.

We corrected the language as suggested throughout the document.

(9) Paragraph 360 and Figure 3 - was the "sliding window" overlapping or non-overlapping?

We added clarifying language to the text in both places. We use a 101bp sliding window with 50bp overlaps.

(10) Line 369 - what is "...the proportion of bases that are expected to match between the two parent strains..."?

We clarified the language in this location, and hopefully changes associated with the comment about sequence similarity will make the comment even clearer in context.

(11) Line 378 - should it refer to Figure S1 and not Figure 4?

Corrected.

(12) Line 399 - should refer to Figure 4, not Figure 5.

Corrected

(13) Line 444-449 - the analysis of loss of shuffling in the context of the location of introgression on the chromosome should be presented in the result section.

We shifted the core of the analysis to the results, while leaving a brief summary in the discussion.

(14) The authors should also take into account the presence of indels in their analyses, and they should be marked in the figures, if possible.

We filtered out indels in our variant calling. However, we did analyze our crosses for the presence of large insertions and deletions (Table S2), which can obscure true recombination rates, and found that they were not an issue in our dataset.

Reviewer #2 (Recommendations For The Authors):

This reviewer suggests that the authors address the different points raised in the public review.

(1) This reviewer would like to challenge the relevance of the r-parameter in light of chromosome 12 which has no introgression and still a strong depletion in r in the fermentation cross.

We added this text to the Results, lines 377-381, "While it is unclear what potential mechanism is mediating the difference in shuffling on chromosome 12, we note that the rDNA locus on chromosome 12 is known to differ dramatically in repeat content across strains of S. cerevisiae (22–227 copies) (Sharma et a. 2022), and we speculate that differences in rDNA copy number between strains in our crosses could impact shuffling."

(2) This reviewer insists on making sure that NCO detection is unaffected by the marker density, notably in the highly polymorphic regions, to unambiguously support Figure 1C.

We've changed our correction for resolution to be more aggressive (see response to comment #2), and believe we have now adequately adjusted for marker density (see response to comment #38).

Reviewer #3 (Recommendations For The Authors):

I regret using such harsh language in the public review, but in my opinion, there has been a serious error in how marker densities are corrected for, and, since the manuscript is now public, it seems important to make it clear in public that I think that the conclusions of the paper are likely to be incorrect. I regret the distress that the public airing of this may cause. Below are my major concerns:

(1) The paper is written in a way that makes it difficult to figure out just what the sequence differences are within the crosses. Part of this is, to be frank, the unusual way that the crosses were done, between more than one segregant each from two diploids in both natural and fermentation cases. I gather, from the homology calculations description, that each of these four diploids, while largely homozygous, contained a substantial number of heterozygosities, so individual diploids had different patterns of heterology. Is this correct? And if so, why was this strategy chosen? Why not start with a single diploid where all of the heterologies are known? Why choose to insert this additional complication into the mix? It seems to me that this strategy might have the perverse effect of having the heterology due to the polymorphisms present in one diploid affect (by correction) the impact of a noncrossover that occurs in a diploid that lacks the additional heterology. If polymorphic markers are a small fraction of total markers, then this isn't such a great concern, but I could not find the information anywhere in the manuscript. As a courtesy to the reader, please consider providing at the beginning some basic details about the starting strains-what is the average level of heterology between natural A and natural B, and what fraction of markers are polymorphic; what is the average level of heterology between fermentation A and fermentation B in non-introgressed regions, in introgressed regions, and what fraction of markers are polymorphic? How do these levels of heterology compare to what has been examined before in whole-genome hybrid strains? It also might be worth looking at some of the old literature describing S. cerevisiae/S. carlsbergensis hybrids.

We thank the reviewer for drawing our attention to confusion about the cross construction. These crosses were conducted as is typical for yeast genetic crosses: we crossed 2 genetically distinct haploid parents to create a heterozygous diploid, then collected the haploid products of meiosis from the same F1 diploid. Because the crosses were made with haploid parents, it is not possible for other genetic differences to be segregating in the crosses. We have revised Figure 1 and its caption to clarify this. Further details regarding the crosses are in the Methods section "Strain and library construction" and in Supplemental Table S1. We only utilized genetic markers that are fixed differences between our parental strains to call CO and NCO. As we detail in the Methods line 507-508, we utilized a total of 24,574 markers for the natural cross and 74,619 markers for the fermentation cross (the higher number in the fermentation cross being due to more fixed differences in regions of introgression). We additionally revised Figure 2D (and Figure S11) to help readers better visualize differences between the crosses.

(2) There are serious concerns about the methods used to identify noncrossovers and to normalize their levels, which are probably resulting in an artifactually high level of calculated crossovers in Figure 2. As a primary indication of this, it appears in Figure 2 that the total frequency of events (crossovers + noncrossovers) in heterozygous introgressed regions are substantially greater than those in the same region in non-introgressed strains, while just shifting of crossovers to noncrossovers would result in no net increase. The simplest explanation for this is that noncrossovers are being undercounted in non-introgressed relative to introgressed heterozygous regions. There are two possible reasons for this: i. The exclusion of all noncrossover events spanning less than three markers means that many more noncrossovers in introgressed heterozygous regions than in non-introgressed. Assuming that average non-homology is 5% in the former and 1% in the latter, the average 3-marker event will be 60 nt in introgressed regions and 300 nt in non-introgressed regions - so many more noncrossovers will be counted in introgressed regions. A way to check on this - look at the number of crossover-associated markers that undergo gene conversion; use the fraction that involves < 3 markers to adjust noncrossover levels (this is the strategy used by Mancera et al.). ii. The distance used for noncrossover level adjustment (2kb) is considerably greater than the measured average noncrossover lengths in other studies. The effect of using a too-long distance is to differentially under-correct for noncrossovers in non-introgressed regions, while virtually all noncrossovers in heterozygous introgressed regions will be detected. This can be illustrated by simulations that reduce the density of scored markers in heterozygous introgressed regions to the density seen in non-introgressed regions. Because these concerns go to the heart of the conclusions of the paper, they must be addressed quantitatively - if not, the main conclusions of the paper are invalid.

We adjusted the correction factor (See also response to comment #2) and compared the average number of CO and NCO events in introgressed and non-introgressed regions between crosses (two comparisons: introgression CO+NCO in natural cross vs introgression CO+NCO in fermentation cross; non-introgression CO+NCO in natural cross vs non-introgression CO+NCO in fermentation cross). We found no significant differences between the crosses in either of the comparisons. This indicates that the distribution of total events is replicated in both crosses once we correct for resolution.

(3) It is important to distinguish the landscape of double-strand breaks from the landscape of recombination frequencies. Double-strand breaks, as measured by uncalibrated levels of Spo11-linked oligos, is a relative number - not an absolute frequency. So it is possible that two species could have a similar break landscape in terms of topography but have absolute levels higher in one species than in the other.

We agree with this statement, however, we have removed the relevant text to streamline our introduction.

(4) Lines 123-125. Just meiosis will produce mosaic genomes in the progeny of the F1; further backcrossing will reduce mosaicism to the level of isolated regions of introgression.

Adjusted the language to be more specific.

(5) Please provide actual units for the Y axes in Figure 2D.

We have corrected the units on the axes.

(6) Tables (general). Are the significance measures corrected for multiple comparisons?

In Table 3, the cutoff was chosen to be more conservative than a Bonferroni corrected alpha=0.01 with 9 comparisons (0.0011). In text, any result referred to as significant has an associated hypothesis test with a p-value less than its corresponding Bonferroni-corrected alpha of 0.05. This has been clarified in the caption for Table 3 and in the text where relevant.

Figures and data

Coefficients of gaussian generalized linear model modeling CO counts per 20kb window.

Coefficients of gaussian generalized linear model modeling NCO counts per 20kb window.

Spearman’s correlations of NCOs to sequence similarity in introgression in the fermentation cross. A cutoff p-value of 0.001 was selected by calculating the Bonferroni corrected α=0.05 for nine comparisons (0.0011) and rounding down.

Welch two sample t-test results for differences in sequence similarity between CO-adjacent regions and NCO-adjacent regions per introgression.