Patterns of crossover distribution in Drosophila mauritiana necessitate a re-thinking of the centromere effect on crossing over

Abstract

We present an SNP-based crossover map for Drosophila mauritiana. Using females derived by crossing 2 different strains of D. mauritiana, we analyzed crossing over on all 5 major chromosome arms. Analysis of 105 male progeny allowed us to identify 327 crossover chromatids bearing single, double, or triple crossover events, representing 398 crossover events. We mapped the crossovers along these 5 chromosome arms using a genome sequence map that includes the euchromatin-heterochromatin boundary. Confirming previous studies, we show that the overall crossover frequency in D. mauritiana is higher than is seen in Drosophila melanogaster. Much of the increase in exchange frequency in D. mauritiana is due to a greatly diminished centromere effect. Using larval neuroblast metaphases from D. mauritiana—D. melanogaster hybrids we show that the lengths of the pericentromeric heterochromatin do not differ substantially between the species, and thus cannot explain the observed differences in crossover distribution. Using a new and robust maximum likelihood estimation tool for obtaining Weinstein tetrad distributions, we observed an increase in bivalents with 2 or more crossovers when compared with D. melanogaster. This increase in crossing over along the arms of D. mauritiana likely reflects an expansion of the crossover-available euchromatin caused by a difference in the strength of the centromere effect. The crossover pattern in D. mauritiana conflicts with the commonly accepted view of centromeres as strong polar suppressors of exchange (whose intensity is buffered by sequence nonspecific heterochromatin) and demonstrates the importance of expanding such studies into other species of Drosophila.

In meiosis, homolog segregation is usually ensured by crossovers. The number and distribution of crossovers is in part regulated by cis-acting factors such as the cis-acting centromere effect, a polar suppression of exchange emanating from the vicinity of the centromere. Hawley et al. use SNP-based crossover mapping to show that in Drosophila mauritiana, the centromere effect is greatly reduced on four of the five major chromosome arms. The authors conclude that the centromere effect differs between Drosophila mauritiana and Drosophila melanogaster and that the ability to attenuate the centromere effect is not a general property of heterochromatin.

Introduction

Genetics functions best as a comparative science—Kenneth W. Cooper

Since its discovery over a century ago in Drosophila melanogaster, crossing over has been the subject of extensive study in both meiotic biology and Drosophila genetics. One of the most active areas of that research focuses on determining the factors and mechanisms that control the number of crossover events and their distribution (Lindsley and Sandler 1977). We seek to understand the cis- and trans-acting elements that control the number and distribution of crossovers, as well as those mechanisms that mediate such processes as crossover assurance (Pazhayam et al. 2021) and crossover interference (Sturtevant 1913, 1915; Berchowitz and Copenhaver 2010). Our goal is to use this knowledge to elucidate the genetic basis for the regional variation in crossover frequencies in the euchromatin (Lindsley and Sandler 1977; Szauter 1984) both within a given genome and between species. Long-standing dogma suggests that such control could be exerted both by the existence of trans-acting factors that act on a genome-wide basis as well as cis-acting processes, such as polar exchange suppression exhibited by centromeres and telomeres.

Several approaches have been employed to understand variation in crossover rates between D. melanogaster and closely related species. First, mapping studies in Drosophila simulans that used visible markers suggested that the genetic map of D. simulans is approximately 30% longer than that of D. melanogaster (Sturtevant 1913, 1915; Barker and Moth 2001). Second, True et al. (1996) measured crossover rates in Drosophila mauritiana by using a large collection of marked transposon insertions in D. mauritiana. The position of these insertions was precisely determined by in situ hybridization to salivary gland polytene chromosomes to create crossover maps. The crossover frequencies between pairs of these insertions were determined with 2-factor crosses and sets of adjacent intervals were summed to determine the distance from any given insertion to the most proximal insertion. For D. melanogaster and D. simulans, the recombination frequency between a given site and the base of the arm was obtained from standard reference maps. This effort allowed comparisons of the maps of these 3 species. Together, these studies demonstrated that the total map length of D. mauritiana was approximately 1.8 times that of D. melanogaster, while the total map length of D. simulans was estimated to be 1.3 times that of D. melanogaster. Surprisingly, the centromere-proximal regions of the euchromatin in D. simulans and D. mauritiana showed greater rates of crossing over when compared with D. melanogaster. This difference was not observed in more distal regions. Third, SNP-based maps of crossover distribution in D. melanogaster (Miller et al. 2016) and Drosophia yakuba (Pettie et al. 2022) have allowed more precise measurements of both crossover density along the 5 autosomal arms and the type of crossover recovered (single, double, or triple crossovers). Presgraves and collaborators have used a variety of tools to decipher the genetic basis for the observed differences in crossover rates in several species within the melanogaster species group (Cattani et al. 2012; Brand et al. 2018, 2019). They concluded that interspecies variation reflects the action of both cis- and trans-acting elements that determine crossover number and position.

The centromere effect: The suggestion by True et al. (1996) and Brand et al. (2018) that cis-acting elements, such as the centromeres, play a major role in creating interspecies variation in crossover number and distribution is intriguing. In D. melanogaster, the term “centromere effect” is used to describe a strong polar suppression of exchange generated by centromeres. (A brief history of the centromere effect is provided in the legend to Fig. 1.) Our current understanding of the centromere effect has been recently reviewed by the Sekelsky lab (Hartmann, Kohl, et al. 2019; Hartmann, Umbanhowar, et al. 2019; Pazhayam et al. 2021; Pazhayam, Frazier, et al. 2024), and so we present only a brief overview here. The centromere effect was first observed by Beadle (1932) following his analysis of exchange in translocation heterozygotes and then by Sturtevant and Beadle (1936) as part of their analysis of crossing over in X chromosome inversion homozygotes. Several subsequent studies both by Yamamoto and Miklos (1977, 1978) and by Hawley (1980) have supported the prevailing view that the centromere effect is generated by the centromeres themselves with its impact on euchromatin attenuated by the expanse of pericentromeric heterochromatin between the centromere and the euchromatin (see Fig. 1a).

Fig. 1.

The canonical view of the centromere effect. A schematic depiction of the canonical view of the centromere effect as defined by studies in D. melanogaster. Here the centromere effect is depicted as a polar suppression of recombination (orange triangles) emanating out from the centromeres (gray circles). Black lines represent the X chromosome euchromatin and dark gray boxes represent the X chromosome heterochromatin. The genetic locus forked (f) is shown for reference. a) The strength of the centromere effect on exchange in the euchromatin is thought to be attenuated by blocks of pericentromeric heterochromatin. b) As shown by Sturtevant and Beadle (1936), the movement of a large portion of the heterochromatin to a distal site by homozygous inversions allows the centromere effect to penetrate more deeply into the proximal euchromatin. c) Deleting proximal heterochromatin allows the centromere effect to spread further into the euchromatin (Schalet and LeFevre 1976; Yamamoto and Miklos 1977, 1978). d) T(1;4)s show much stronger centromere effects on exchange when heterozygous with a normal sequence X chromosome than do T(1;Y)s with similar breakpoints on the X chromosome. This was presumed to be a consequence of the ability of the larger blocks of heterochromatin on the Y chromosome to attenuate the centromere effect more effectively than the much smaller amount of heterochromatin on the 4th chromosome (Hawley 1980). The Y chromosome heterochromatic sequences are represented in teal and the 4th heterochromatic sequences are shown in purple.

Following Pazhayam, Frazier, et al. (2024), we should note that the absence of exchange within the satellite-rich heterochromatin itself is an intrinsic feature of centric heterochromatin presumably due to the fact that programed double-strand breaks (DSBs) in meiosis are precluded from these regions (Mehrotra and McKim 2006). Indeed, the current consensus is that the centromere effect reflects a polar suppression of crossing over in the centromere-proximal euchromatin, such that if these regions of euchromatin are moved distally from centric regions, they will regain the capacity to recombine at expected levels. It is proposed that the attenuation of recombination by pericentric heterochromatin is not by a special property of heterochromatin but is merely due to the increased physical distance between the centromere and euchromatin where crossing over occurs.

Several recent studies suggest that the centromere effect is partially under the genetic control of trans-acting factors. For example, the centromere effect is abrogated in Drosophila lacking the Bloom syndrome helicase (Blm) (Hatkevich et al. 2016). This has been proposed to arise from the induction of DSBs in regions proximal to the centromere being shunted from the Class I meiotic pathway, which would resolve as non-crossovers in proximal regions due to the centromere effect, into an alternate pathway of crossover resolution designated the Class II pathway that is insensitive to the centromere effect (Hatkevich et al. 2016). Variants of C(3)G, a transverse-filament protein of the synaptonemal complex (SC), also lead to increased amounts of recombination in proximal regions near the centromere (Billmyre et al. 2019). This suggests that the SC plays a role in the transmission of the centromere effect along the chromosome arm. Finally, mutations in mei-218, which has a crucial function in crossing over, lead to a change in the distribution of crossovers along the chromosome, with a greater proportion (though not greater number) of crossovers in centromere-proximal regions (Carpenter and Sandler 1974; Hartmann, Kohl, et al. 2019; Hartmann, Umbanhowar, et al. 2019; Pazhayam 2025). Modifying the extent and severity of the centromere effect via these trans-acting factors might be a strong driver of interspecies variation. This was shown by the work of Brand and colleagues (Brand et al. 2018, 2019) who provided evidence that the rapid evolution of the trans-acting factor MEI-218 might be a component of intraspecific variation in the recombination rate by modulating the severity of the centromere effect. These authors showed that replacing the D. melanogaster mei-218 gene with a transgenic copy of its D. mauritiana ortholog resulted in an increase in crossing over, most notably in the proximal and distal euchromatin. Indeed, the authors suggest the different mei-218 genes in D. melanogaster and D. mauritiana control crossover distribution by “modulating the intensity of centromeric and telomeric suppression of crossing over.” In other words, species-level differences arise from both the actions of cis-acting mechanisms, such as the centromere and telomere effects, and trans-acting factors such as the MEI-218 protein, which is thought to be a pro-crossover factor (Kohl et al. 2012; Brand et al. 2019; Hartmann, Kohl, et al. 2019; Hartmann, Umbanhowar, et al. 2019).

Here we extend this body of knowledge by providing an SNP-based map of crossing over in D. mauritiana. We confirm that the frequency of crossing over is increased in D. mauritiana compared with D. melanogaster. Our data confirm and extend the suggestion of True et al. (1996) that much of this increase can be ascribed to a greatly weakened centromere effect on all arms except 2L. This diminishment of the centromere effect increases the area in which crossovers can occur for 4 of the 5 chromosome arms in D. mauritiana. The diminishment of crossover suppression in proximal euchromatin, in the absence of obvious changes in the amount of centromeric heterochromatin, requires a reconsideration of our current understanding of the centromere effect. Instead of being fixed between species, the strength of the centromere effect may vary and this is likely to impact the evolution of global recombination rates.

Materials and methods

Fly stocks and husbandry

The following D. mauritiana wild-type lines were obtained from the National Drosophila Species Stock Center at Cornell University: Le Reduit, Mauritiana—14021-0241.150 (Mau_LR), Mauritiana 2—14021-0241.05 (Mau_2), Mauritiana 3—14021-0241.08 (Mau_3), and Mauritiana genome line—14021-0241.151 (Mau_G). D. mauritiana stocks were maintained on a 12-h light/dark cycle on standard laboratory molasses food at 25°C with ∼40% humidity.

Husbandry for the D. mauritiana—D. melanogaster hybrid: All fly stocks were raised on standard Bloomington medium at 25°C. D. melanogaster y w and D. mauritiana w¹ (DSSC#14021-0241.60) were used for the mitotic chromosome spreads.

DNA preparation and sequencing of parental lines

Genomic DNA from several strains of D. mauritiana, including the 2 strains used for the mapping studies, was isolated from 20 females (2 × 10 flies/prep) from each line using the Maxwell 16 Tissue DNA purification kit on the Maxwell 16 instrument (Promega). Samples were eluted in 300 µl water containing RNase A at 20 mg/ml. Like samples were combined, extracted once with phenol/chloroform, and precipitated at −20°C. Samples were resuspended in 22 µl of ultrapure water and quantitated on a Qubit fluorometer (Thermo Fisher Scientific). DNA-seq libraries were generated from 500 ng of DNA. The libraries were made according to the manufacturer's directions for the Illumina DNA Library Prep, (M) Tagmentation (Illumina, Cat. No. 20060060) and Nextera DNA UD Indexes (set A) (Illumina, Cat. No. 20018704) with 5 cycles of PCR. The resulting short fragment libraries were checked for quality and quantity using the Bioanalyzer (Agilent) and Qubit 2.0 Fluorometer (Thermo Fisher Scientific). The libraries were pooled and sequenced as 100 bp paired reads on a P2 flow cell using the Illumina NextSeq 2000 instrument. Following sequencing, Illumina Primary Analysis version NextSeq2K RTA 3.9.2 and bcl-convert-3.10.5 were run to demultiplex reads for all libraries and generate FASTQ files.

After sequencing, raw fastq files were adapter trimmed using Trim Galore (v0.6.10) and aligned to the D. mauritiana reference genome (D_mauritiana_ASM438214v1; RefSeq) using bwa mem (v0.7.17). Aligned reads were then deduplicated by processing them through samtools (v1.18) fixmate, sort, and markdup with the -r argument. Variant calling was run on these deduplicated reads using deepvariant (v1.5.0) with the Whole Genome Sequence model type argument. SNPs detected by deepvariant were filtered by vcftools (v0.0.16) to only retain biallelic, substitution point mutations with a minimum genotype quality of 20. A list of line-specific parental mutations was generated by comparing the filtered variant calls from the 3 parental samples and dropping identical shared mutations.

DNA preparation, sequencing, alignment, and SNP calling of individual flies

Among pairwise comparisons, the Mauritiana 2 (Mau_2) and the Mauritiana genome line (Mau_G) were determined to have the largest number of unshared single-nucleotide variants (i.e. found in Mau_2 but not Mau-G and vice versa) across any pair of parental lines and the broadest distribution across the genome (Supplementary Fig. 1). We then used Mau_G and Mau_2 heterozygotes for the recombination studies presented below. Mau_2 virgin females were mated to Mau_G males to produce heterozygous female progeny (denoted as 2-G in the sequence file name, see below). Heterozygous virgin female progeny were then mated individually to Mau_LR, Mau_G, or Mau_2 males to produce male progeny bearing 1 copy of each maternal, potentially recombinant, chromosome. We used different male genotypes depending on their experimental availability as some cross combinations were poorly fertile (Garrigan et al. 2014). Use of different male genotypes for backcrossing is not expected to influence results since crossing over occurs in females and the bioinformatics pipeline for crossover detection was easily adapted to each male genotype. Crosses between F1 females and Mau_LR males were designated as outcrosses. Crosses between F1 females and either Mau_2 or Mau_G males were designated as backcrosses.

A total of 108 individual male progeny were obtained from all the crosses: 69 of those progeny were obtained from the outcross to Mau_LR males, 25 from the backcross to Mau_2 males, and 14 from the backcross to Mau_G males. These males were starved for 4–6 h and frozen for downstream analysis. Sequence was successfully obtained from 105 males.

Genomic DNA was extracted from individual frozen males using the Qiagen DNeasy Blood and Tissue Kit (Qiagen #69506) with the following modifications. Single flies were homogenized using a pestle (Kimble #749520-0090 or similar) with a handheld homogenizer (VWR 47747-370) in 180 µl of ATL buffer, after which 4 µl of RNase A [100 mg/ml] was added. All other steps followed the standard manufacturer’s protocol, except DNA was eluted in 2 × 50 µl of 10 mM Tris (pH 8.0). The DNA eluate was then concentrated to ∼50 µl by centrifugation in Microcon DNA Fast Flow centrifugal filters (Merck Millipore, MRCFOR100) to achieve an average yield of 112 ng/fly. DNA was quantified using a Qubit fluorometer. Individual males were labeled to identify the father, the heterozygous mother, and sibling number. Thus, xG_13.2 was the second male isolated from female 13 mated to a Mau_G father. At least 3 sibling males were collected from each cross. Numbers may not be sequential because not all preps yielded sufficient DNA for sequencing.

Initially, DNA-Seq libraries were generated for 12 samples from 20 ng of genomic DNA, as assessed using the Qubit 2.0 Fluorometer, according to the manufacturer's directions with 15 minutes of enzymatic fragmentation at 37°C, targeting 200–450 bp, followed by a 10-fold dilution of the universal adaptor and 6 cycles of PCR with the NEBNext Ultra II FS DNA Library Prep Kit for Illumina (NEB, Cat. No. E7805S), and the NEBNext Multiplex Oligos for Illumina (96 Unique Dual Index Primer Pairs) (NEB, Cat. No. E6440S) and purified using the SPRIselect bead-based reagent (Beckman Coulter, Cat. No. B23318). Resulting short fragment libraries were checked for quality and quantity using the Bioanalyzer (Agilent) and Qubit 2.0 Fluorometer. Equal molar libraries were pooled, quantified, and sequenced on a Mid-Output flow cell of an Illumina NextSeq 500 instrument using NextSeq Control Software 4.0.1 with the following read length: 8 bp Index1, 151 bp Read1, 151 Read2, and 8 bp Index2. Following sequencing, Illumina Primary Analysis version NextSeq RTA 2.11.3.0 and bcl-convert-3.10.5 were run to demultiplex reads for all libraries and generate FASTQ files. Subsequently, additional DNA from the same 12 samples was prepared in duplicate using 1/10th reaction volumes, 1:30 fold adapter dilutions, and 8 cycles of PCR with the same reagents and conditions.

The additional 93 samples used in the analysis were prepared using the Seqwell plexWell 96 library prep. DNA-seq libraries were generated from 6.5–16 ng of DNA, as assessed using the Qubit Fluorometer. The libraries were made according to the manufacturer's directions for the seqWell purePlex DNA Library Prep Kit for Illumina Sequencing Platforms (seqWell, Cat. No. 301067) with kit supplied Indexes (seqWell, Cat. No. 301067) with 8 cycles of PCR. The resulting short fragment libraries were checked for quality and quantity using the Bioanalyzer and Qubit 2.0 Fluorometer. Equal molar libraries were pooled, quantified, and converted to process on the Element Biosciences AVITI with the Element Adept Library Compatibility Workflow, following the Adept Rapid PCR-Free Protocol. The converted pool was sequenced on a 2 × 150 Cloudbreak High Output flow cell (Cat. No. 860-00003), using AVITI OS 2.3.0 with the following read length: 10 bp Index1, 10 bp Index2, 151 bp Read1, and 151 bp Read2. Following sequencing, Bases2Fastq was run to demultiplex reads for all libraries and generate FASTQ files. The sequencing depth of the offspring from both the original 12 and subsequent 93 offspring samples is represented in Supplementary Fig. 2.

Detection and validation of crossovers

A schematic of the crossover detection workflow is described in Supplementary Figs. 3 and 4. R (R 4.2.3) was used to process VCF files from individual offspring and identify candidate crossover events using parent-specific variants. The first step filters out the offspring vcf file to only retain variants that match the line-specific parental variant (private variants) list. Supplementary Table 1 provides the variants identified in the sequenced offspring that did not match the parental lines. These were apparently segregating in the parental lines and were not used to identify the location of crossovers. Each chromosome was then binned into non-overlapping 50 kb bins, summing the number of private variants (scored as either present or absent, irrespective of zygosity) belonging to each of the 3 parental lines in each bin. For each 50 kb bin, this provided a count for variants that are private to each of the 3 lines. Using these counts, 2 approaches were taken, depending on whether the cross was a backcross or an outcross. For offspring derived from the F1 by Mau_LR outcrosses, bins in the bottom 25 percentile of the total number of matched variants were dropped to decrease noise caused by low levels of variation. Crossovers were then detected by identifying bins where the proportion of total variants derived from 1 line (either Mau_2 or Mau_G) flipped compared with the previous bin (Outcross detection method 1). For example, if 1 50-kb bin identified the presence of 195 variants from Mau_2 and 2 variants from Mau_G, and the subsequent bin identified the presence of only 3 variants from Mau_2, but 249 variants from Mau_G, this would indicate a crossover.

For the offspring of backcrosses to either Mau_G and Mau_2 (Backcross detection method 2), a different approach was taken. This is because, in contrast to the outcross detection method, a crossover is not simply identified when, across non-overlapping windows, there is a balanced “flip” in the counts of variants between the Mau_G and Mau_2 lines. Instead, a crossover is identified when the proportion of private variants identified from 1 line switches between approximately 50% (the heterozygous state of the 2 haplotypes) and either 0 or 100% (the homozygous state for one of the lines). For this procedure, bins with variant counts above the 95th percentile or below the 10th percentile of total SNPs matching the male parental line were removed to limit the impact of mapping artifacts. In addition to counting the number of private variants belonging to each parental line identified in each 50 kb bin, we also counted the number of Mau_2 and Mau_G private variants found in the 3 upstream and downstream bins. k-means clustering was applied on those 14 total features (current, upstream, and downstream Mau-G and Mau2 counts) to cluster each bin into 1 of 2 clusters. Crossovers were detected by finding bins where the 3 previous bins belong to 1 cluster and the 3 following bins belong to the other. Since all sequenced offspring were males, the X chromosome crossovers were found using outcross detection method 1 for all samples. All crossovers were validated visually in IGV (Robinson et al. 2011). Only 1 crossover event (a possible double crossover on 3L) failed to validate in IGV. Crossover events were categorized into classes based on the number of crossovers observed per chromatid: zero or non-crossovers (NCOs), single crossovers (SCOs), double crossovers (DCOs), triple crossovers (TCOs), and quadruple crossovers (4COs).

The number of SNPs within the heterochromatic regions for the offspring is described in Supplementary Table 2.

Availability of sequence data

All sequences used in this study are publicly available at https://www.ncbi.nlm.nih.gov/sra/PRJNA1168080. The sample names for sequences posted identify the paternal strain used. For example, sample name 2-Gx2 refers to heterozygous female (Mau_2/Mau_G) crossed to Mau_2 males, sample name 2-GxG refers to heterozygous female (Mau_2/Mau_G) crossed to Mau_G males, and sample name 2-GxLR refers to heterozygous female (Mau_2/Mau_G) crossed to Mau_LR males. VCF files are provided in GSA FigShare https://doi.org/10.25386/genetics.28506611.

DNA fluorescence in situ hybridization

For mitotic chromosome spreads, larval 3rd instar brains were squashed according to previously described methods (Larracuente and Ferree 2015). Briefly, tissue was dissected into 0.5% sodium citrate for 5–10 min and fixed in 45% acetic acid/2.2% formaldehyde for 4–5 min. Fixed tissues were firmly squashed with a cover slip and slides were submerged in liquid nitrogen until bubbling ceased. Coverslips were then removed with a razor blade and slides were dehydrated in 100% ethanol for at least 5 min. After drying, hybridization mix (50% formamide, 2× SSC, 10% dextran sulfate, 100 ng of each probe) was applied directly to the slide, samples were heat denatured at 95°C for 5 min and allowed to hybridize overnight at room temperature. Following hybridization, slides were washed 3 times for 15 min in 0.2× SSC and mounted with VECTASHIELD with DAPI (Vector Labs). The following satellite DNA probes were used for in situ hybridization: (AAGAG)₆, (AATAACATAG)₃ and AGGATTTAGGGAAATTAATTTTTGGATCAATTTTCGCATTTTTTGTAAG (359 bp repeat) and have been previously described (Jagannathan et al. 2017). Fluorescent images were taken using a Leica TCS SP8 confocal microscope with 63× oil-immersion objectives (NA = 1.4). Brightfield images were acquired using a Keyence microscope. Images were processed using Adobe Photoshop software.

Maximum likelihood estimates of tetrad classes

The frequencies of tetrad classes, denoted as Er, estimate the number of tetrads with r crossovers, where E₀ corresponds to tetrads with no crossovers, E₁ to those with 1 crossover, E₂ to those with 2, and so on. Crossover events can be categorized based on the number of crossovers per chromatid. Let $X=(X_{nCO},X_{sCO},X_{dCO},X_{tCO},X_{qCO})$ represent the counts for zero crossover (nCO), single crossover (sCO), double crossovers (dCO), triple crossovers (tCO), and quadruple crossovers (qCO), respectively. Zwick et al. (1999) demonstrated that the distribution of crossover counts follows a multinominal distribution

$$\begin{array}{rl}& \left({\mathrm{X}}_{nCO},{\mathrm{X}}_{sCO},{\mathrm{X}}_{dCO},{\mathrm{X}}_{tCO},{\mathrm{X}}_{qCO}\right)\\ & \sim Multinomial\left(N,\left(p_{nCO},p_{sCO},p_{dCO},p_{tCO},p_{qCO}\right)\right)\end{array}$$

Where p_nco, psc_o, p_dco, p_tco, and p_qco denote the probabilities associated with each crossover category, and N represents the total number of observed crossovers. Assuming that crossovers are randomly distributed along chromatids without interference, Weinstein (1936) demonstrated that each crossover class probability is a linear combination of the frequencies of tetrad classes, E_r. The likelihood function for the observed crossover events is given by:

$$L(X|p)={\displaystyle \frac{N!}{x_{nCO}!*x_{sCO}!*x_{dCO}!*x_{tCO}!*x_{qCO}!}(p_{nCO}^{nCO}{p}_{sCO}^{sCO}{p}_{dCO}^{dCO}{p}_{tCO}^{tCO}{p}_{qCO}^{qCO})}$$

where

$$p_{nCO}=E_0+{\displaystyle \frac{E_0}{2}+{\displaystyle \frac{E_1}{4}+{\displaystyle \frac{E_2}{8}+{\displaystyle \frac{E_3}{16}}}}}$$

$$p_{sCO}={\displaystyle \frac{E_1}{2}+{\displaystyle \frac{E_2}{2}+E_3*{\displaystyle \frac{3}{8}+E_4*{\displaystyle \frac{4}{16}}}}}$$

$$p_{dCO}={\displaystyle \frac{E_2}{4}+E_3*{\displaystyle \frac{3}{8}+E_4*{\displaystyle \frac{6}{16}}}}$$

$$p_{tCO}={\displaystyle \frac{E_3}{8}+E_4*{\displaystyle \frac{4}{16}}}$$

$$p_{qCO}={\displaystyle \frac{E_4}{16}}$$

We employed optimization techniques to estimate the maximum likelihood estimates (MLEs) of E_rs. Specifically, the Nelder-Mead algorithm was utilized via the optim() function in R, with Weinstein's algebraic solutions serving as the initial parameter values. Additionally, the probabilities were constrained to be positive throughout the optimization process.

A public-facing web app for applying our MLE method on user-submitted crossover count data is available at https://simrcompbio.shinyapps.io/CrossoverMLE/. The app was created using R (v4.3.1) and shiny (v1.7.5) (https://CRAN.R-project.org/package=shiny) and hosted on shinyapps.io.

Results

Generation of the Mau_2/Mau_G heterozygous females: In order to choose the 2 D. mauritiana strains used to create the heterozygous females that we used for mapping crossovers, we sequenced 4 strains of D. mauritiana (Mau_G, Mau_2, Mau_3, and Mau_LR) and identified SNPs present in each line (Supplementary Fig. 1). SNPs were filtered using vcftools to only retain biallelic, substitution point mutations with a minimum genotype quality of 20. A list of line-specific parental mutations was generated by comparing the filtered variant calls from the parental samples and removing any identical shared mutations.

We chose to use the Mau_G and Mau_2 lines for our crosses. Mau_G and Mau_2 were chosen because they had the largest number of unshared variants (i.e. found in Mau_2 but not Mau_G and vice versa) across any pair of parental lines. The original analysis found 1,467,376 SNPs unique to 1 of the 2 lines, the most out of all possible pairwise comparisons of sequenced strains. The rationale being the more unique SNPs that exist between the parents, the more accurate our ability to position crossover sites.

We then performed single-fly sequencing of 105 individual male progeny from Mau_2/Mau_G heterozygous females. The average depth of sequencing for the initial 12 test samples was 21.42×, while the average depth of the subsequent 93 samples was 15.81×. Analysis of these males allowed us to evaluate crossover events on 5 chromosome arms (X, 2L, 2R, 3L, and 3R) in each male, obtaining the CO positions for 525 arms. From this data, we identified 198 NCO arms, and 327 arms with 1 or more crossover events: 262 SCO-bearing arms, 59 DCO arms, and 6 TCO arms. Together, we mapped 398 crossover events (Table 1) and determined their precise position (Fig. 2). Because our data was relatively low coverage, we did not search for gene conversion events or assay for crossing over on the 4th chromosome. With 15× coverage, a single heterozygous SNP from a given haplotype has about a 3 in 100,000 (0.5¹⁵) chance of being sequenced with zero depth, which is expected to yield a significant challenge for detecting rare gene conversion events across the entire genome.

Table 1.

Crossover progeny type recovered.

Crossover type	Chr 2La	Chr 2R	Chr 3L	Chr 3R	Chr X
D. mauritiana (this paper N = 105)
NCO	47.6 (50)	36.2 (38)	35.2 (37)	28.6 (30)	40.9 (43)
SCO	47.6 (50)	51.4 (54)	55.2 (58)	50.0 (53)	44.8 (47)
DCO	4.8 (5)	12.4 (13)	7.6 (8)	17.1 (18)	14.3 (15)
TCO	0.0 (0)	0.0 (0)	1.9 (2)	3.8 (4)	0.0 (0)
D. melanogaster (Miller et al. 2016. N = 196)
NCO	52.5 (101)	51.5 (101)	49.4 (97)	50.5 (99)	49.5 (97)
SCO	44.9 (88)	45.9 (90)	42.8 (84)	43.9 (86)	43.9 (86)
DCO	2.5 (5)	2.5 (5)	7.6 (15)	5.1 (10)	6.6 (13)
TCO	0.0 (0)	0.0 (0)	0.0 (0)	0.5 (1)	0.0 (0)

^aNumber is percent of total flies of that class. Number in parentheses is the total number of flies of a given class.

Fig. 2.

Crossover positions for the 398 recombinants. Position of the 398 exchange events observed from 105 individual male progeny, where each panel represents 1 of the 5 major D. mauritiana chromosome arms (the small 4th chromosome was not examined). The top track in each panel shows the 262 SCO events. The DCO tracks illustrate the locations and spans of 59 DCOs. Six TCO events were identified, 2 on chromosome 3L and 4 on chromosome 3R. That the SCO events making up a DCO or TCO event are well-spaced suggests that interference in D. mauritiana is similar to that seen in D. melanogaster. Indeed, the average distance between the 2 crossovers that comprise doubles in D. mauritiana is 10.6 Mb, which compares well with the value of 10 MB obtained by Miller et al. (2016) and suggests that the 2 species do not differ greatly, if at all, in terms of interference.

Species-relatedness: a comparison of the genomes; D. melanogaster and D. mauritiana are both members of the melanogaster species complex. (Supplementary Fig. 5). Their karyotypes and the maps of their salivary gland polytene chromosomes are identical, except for a large paracentric and euchromatic inversion on 3R. Full genome sequence studies by Chakraborty et al. (2021) confirm the close relatedness adding only several small inversions on the X chromosome to the difference in euchromatic sequence. According to the canonical understanding of the centromere effect, if the pericentromeric regions of heterochromatin in D. mauritiana were substantially longer than those of D. melanogaster, that should decrease centromere-induced suppression of euchromatic crossing over in D. mauritiana compared with D. melanogaster. However, as shown below, there is no evidence for a substantial difference in the lengths of these regions between the 2 species.

Total map length comparison: Table 2 displays the total map length of each arm in D. mauritiana compared with the values obtained by Miller et al. (2016) for D. melanogaster, which used a similar sequencing approach for identifying crossovers. In sum, the total map length of D. mauritiana is 1.36 times greater than that observed for D. melanogaster. This increase is substantial, albeit lower than the 1.8-fold increase reported by True et al. (1996). This increase in map length is not uniform across the 5 arms. The map lengths of the 3R and 3L arms in D. mauritiana are increased 1.3 and 1.8 times, respectively, compared with their map lengths in D. melanogaster, while the map length of 2L is increased by only a factor of 1.05. True et al. (1996) also found that the greatest increase in total map length was on the 3rd chromosome (2.1 times) compared with D. melanogaster than on the X (1.8 times) or 2nd chromosome (1.6 times) (True et al. 1996).

Table 2.

Total map lengths in cM (cM per Mb in parentheses).

	Chr 2L	Chr 2R	Chr 3L	Chr 3R	Chr X	Total
Species
D. mauritiana	57.1 (2.36)	76.2 (3.24)	76.2 (2.91)	96.2 (3.33)	73.3 (3.16)	379.0 (3.01)
D. melanogaster	54.1 (2.30)	51.0 (2.02)	58.1 (2.07)	55.6 (1.73)	57.1 (2.43)	275.9 (2.08)

Distribution of all crossover events: Fig. 3 presents the crossover density plots for each of the 5 arms in D. mauritiana. Similar plots based on the data from Miller et al. (2016) for D. melanogaster are presented for comparison. These plots also show the euchromatin-heterochromatin boundary and the position of a fixed euchromatic inversion on 3R in D. mauritiana. The distribution of medial and distal crossover events in D. mauritiana does not appear different from that of D. melanogaster (see Fig. 3), even on 3R. However, there is an obvious increased frequency of exchange in the proximal euchromatin of D. mauritiana when compared with D. melanogaster. To evaluate whether the increased map length of D. mauritiana could be attributed primarily to differences in proximal regions, we performed a statistical comparison with D. melanogaster across arm segments divided into thirds: distal, medial, and proximal (Table 3). We considered the total number of crossovers per segment irrespective of the physical distance of each segment, and also the number of crossovers in each segment normalized by physical distance.

Fig. 3.

CO density plots for each of the 5 arms in both D. mauritiana and D. melanogaster. The euchromatin:heterochromatin boundaries are denoted by dashed yellow lines and the breakpoints of the euchromatic inversion on 3R are indicated by black dotted lines. The chromosome illustrations indicate the direction of the centromeres in the above plots. Brackets on the chromosomes represent the heterochromatin not included in the plots. D. melanogaster data is from Miller et al. (2016). Chromosome arms were divided equally into thirds for the proximal, medial, and distal classifications.

Table 3.

Segment CO status.

	Distal 2L			Medial 2L			Proximal 2L			Proximal 2R			Medial 2R			Distal 2R
	mau	mel	P	mau	mel	P	mau	mel	P	mau	mel	P	mau	mel	P	mau	mel	P
NCO	82	147		75	155		98	181		87	194		75	153		75	142
SCO	23	48		30	41		7	15		18	2		28	43		30	53
DCO	0	1		0	0		0	0		0	0		2	0		0	1
RF	0.219	0.255	0.760	0.286	0.209	0.073	0.067	0.077	0.632	0.171	0.010	<0.001	0.305	0.219	0.066	0.286	0.281	0.468
			(0.800)			(0.093)			(0.658)			(<0.001)			(0.032)			(0.315)

	Distal 3L			Medial 3L			Proximal 3L			Proximal 3R			Medial 3R			Distal 3R
	mau	mel	Pa	mau	mel	P	mau	mel	P	mau	mel	P	mau	mel	P	mau	mel	P
NCO	83	134		67	149		88	192		83	194		74	146		61	140
SCO	22	62		35	46		17	4		22	2		29	50		42	55
DCO	0	0		3	1		0	0		0	0		2	0		2	1
RF	0.210	0.316	0.02	0.390	0.245	0 .008	0.162	0.020	<0.001	0.210	0.010	<0.001	0.314	0.255	0.156	0.438	0.291	0.008
			(0.050)			(0 .003)			(<0.001)			(<0.001)			(0.067)			(0.001)

	Distal X			Medial X			Proximal X
	mau	mel	P	mau	mel	P	mau	mel	P
NCO	84	163		74	146		80	169
SCO	21	33		31	48		25	27
DCO	0	0		0	2		0	0
RF	0.200	0.168	0.255	0.295	0.265	0.298	0.238	0.138	0.018
			(0.238)			(0.275)			(0.015)

1/3 segment (Distal, Medial, Proximal) crossover counts for each chromosome arm (2L, 2R, 3L, 3R, and X). P-values were calculated by bootstrapping the segment status 1 million times. P equals the probability, among 1 million random samples, that the recombination frequency in that segment of D. melanogaster is higher than the segment in D. mauritiana. P-values below in parentheses indicates the same, but normalized for physical distance.

NCO, non-crossover segment; SCO, single crossover segment; DCO, double crossover segment; RF, Recombination frequency.

^aIndicates a P-value with the hypothesis that the recombination frequency in that segment of D. mauritiana is higher than the segment in D. melanogaster. Significant P-values are bolded.

Not accounting for physical distance, 4 (2R, 3L, 3R, and X) out of the 5 proximal regions showed significantly greater amounts of crossing over in D. mauritiana. Only 1 medial region, on 3L, showed a significant difference, with D. mauritiana also displaying a higher crossover rate. The recombination frequency of 2 distal regions was significantly different, with one showing a higher recombination rate in D. mauritiana (3R) and the other showing the opposite pattern (3L). Accounting for physical distance, the results were essentially the same, with minor differences. In particular, the medial region of 2R now showed a significantly higher per bp recombination rate in D. mauritiana and the distal region of 3L no longer showed a significantly greater amount of recombination in D. melanogaster. Together, these results are most easily interpreted as a decrease in the ability of the centromere effect to reduce the frequency of crossing over in the proximal euchromatin of D. mauritiana present on 4 arms of D. mauritiana.

Dissimilarities between the crossover distributions in D. mauritiana and D. melanogaster cannot be explained by differences in the amount of proximal heterochromatin. One possible explanation for the reduced centromere effect on these 4 arms might be a substantial increase in the amount of pericentric heterochromatin in D. mauritiana that could attenuate the centromere effect relative to D. melanogaster by increasing the physical distance between the centromere and the proximal euchromatin. However, careful measures of the genome sizes of these 2 species show that D. mauritiana is, in fact, slightly smaller than D. melanogaster in terms of genome size (Gregory and Johnston 2008), despite the observation by Chakraborty et al. (2021) that the length of euchromatic sequence is greater in D. mauritiana than it is in D. melanogaster. A greater amount of euchromatic sequence in D. mauritiana in the face of a smaller genome size leads to the conclusion that there is less heterochromatin compared with D. melanogaster.

To directly test whether a greater amount of pericentric heterochromatin could be the cause of the reduced centromere effect in D. mauritiana, we performed DNA fluorescence in situ hybridization (DNA FISH) on mitotic chromosome spreads from D. melanogaster and D. mauritiana hybrid females. Specifically, we used probes that hybridize with pericentromeric satellite DNA repeats that are abundant in D. melanogaster that allow us to identify the species-origin for each mitotic chromosome. Our data clearly show that the D. melanogaster chromosomes (arrowheads, Fig. 4, a and b) contain larger DAPI-dense blocks of pericentromeric heterochromatin in comparison to the D. mauritiana chromosomes (Fig. 4, a and b). These data suggest that differences in the amount of pericentromeric heterochromatin are unlikely to account for the weakened centromere effect in D. mauritiana.

Fig. 4.

Cytogenetic comparison of the relative lengths of pericentromeric heterochromatin in D. melanogaster and D. mauritiana hybrids. a and b) FISH against the (AATAACATAG)_n repeat (blue), the (AAGAG)_n repeat (green), and the 359 bp repeat (magenta) on larval neuroblast mitotic chromosomes from female D. melanogaster and D. mauritiana hybrids and costained with DAPI (gray). Arrowheads point to D. melanogaster chromosomes.

A consideration of exchange frequencies at the level of individual bivalents. The crossover density plots shown in Fig. 3 can only be used to identify where the crossovers occurred, not the arrangement of crossovers on individual bivalents. However, we can obtain such information using the data in Table 1, which presents the primary crossover data in terms of crossover chromatid types (i.e. NCO, SCO, DCO, and TCOs). Because each of these types of chromatids can be obtained from bivalents with a higher number of exchanges (Miller et al. 2016), Weinstein (1936) created an algebraic method for characterizing a given population of bivalents in terms of bivalents with no crossovers (E0), 1 crossover event (E1), 2 crossover events (E2), 3 crossover events (E3), and so on. These values are referred to as Exchange Ranks. However, this method can be computationally difficult, especially with small numbers of progeny (Miller et al. 2016, 2018). We used a maximum likelihood estimate to derive Weinstein's values. The method is available online at https://simrcompbio.shinyapps.io/CrossoverMLE/.

Results for this analysis are shown in Table 4. We first consider the 3rd chromosome where in D. mauritiana, 3R displays E2 and E3 frequencies of 0.33 and 0.28 respectively, substantially higher than the values of 0.14 and 0.04, respectively, seen in D. melanogaster. A similar high frequency of E3 events can be seen for 3L. Both 2R and the X display an increased frequency of E2 bivalents in D. mauritiana. These increases in bivalents with 1 or more exchange events came at the expense of non-crossover (E0) bivalents. We admit to being impressed by E0 values of zero for 2R, 3L, and 3R. Only on 2L, the arm on which the centromere effect was unaltered, did D. mauritiana show an exchange rank pattern similar to that seen in D. melanogaster. Although more complex interpretations are possible, such as arm-specific alterations in the strength of interference, it seems most parsimonious to explain these data as being primarily driven by a reduction in the centromere effect on each arm. This leads to more of each chromosome arm being used for crossovers, resulting in more double, triple, and quadruple exchange events. In support of this hypothesis, we note that our average distance between crossovers on DCO-bearing chromatids is 10.6 MB. This compares well with the value of 10 MB obtained by Miller et al. (2016) and suggests that the 2 species do not differ greatly, if at all, in terms of interference.

Table 4.

MLEs of Weinstein tetrad frequencies.

Exchange rank frequencies	Chr 2L	Chr 2R	Chr 3L	Chr 3R	Chr X
D. mauritiana (this paper)
E0	0.0476	0.0000	0.0000	0.0000	0.1048
E1	0.7619	0.4903	0.7054	0.3903	0.3238
E2	0.1905	0.5097	0.1607	0.3293	0.5714
E3	0.0000	0.0000	0.1339	0.2804	0.0000
D. melanogaster (Miller et al. 2016.)
E0	0.1021	0.0815	0.1428	0.1121	0.1224
E1	0.7145	0.8165	0.5511	0.7042	0.6122
E2	0.1835	0.102	0.3061	0.1429	0.2654
E3	0.0000	0.0000	0.0000	0.0408	0.0000

Discussion

We report both a SNP-based map for crossover distribution in D. mauritiana and the development of a new and freely available algorithm for determining Weinstein Exchange Ranks using maximum likelihood estimation. At a minimum, our data confirms and extends the work of True et al. (1996). However, our ability to position each crossover event along a genome sequence denoting the heterochromatin:euchromatin boundaries of each arm allows the additional conclusions regarding the nature of the centromere effect that are presented below. Specifically, we show that despite reduced amounts of pericentromeric heterochromatin on 4 of the 5 arms of D. mauritiana when compared with D. melanogaster, the centromere effect on those arms in D. mauritiana is much weaker than it is in D. melanogaster.

In addition, our ability to detect chromatids bearing double and triple crossovers, and thus determine exchange ranks, allows us to include that the amelioration of the centromere effect on 4 of the 5 chromosome arms increase the euchromatin region available for crossing over and thus provide greater opportunity for double and triple crossovers that might otherwise be precluded by interference.

A new model for the centromere effect: Our data show clearly that reduced heterochromatin on the arms of D. mauritiana does not enhance the centromere effect but, paradoxically, is associated with a decrease. In conjunction with prior studies (discussed above) that show a heightened centromere effect when the adjacent pericentromeric heterochromatin is diminished, these data also suggest that the strength of the centromere effect is not related to the simple amount of pericentromeric heterochromatin. We can thus speculate on different models for how the centromere effect is mediated and how this might explain interspecific variation in the genome-wide recombination rate. In 1 model, which could be designated the Inert Spacer Model, the chromatin state and sequence of DNA residing between the centromere and regions where recombination is observed have no influence on the centromere effect in any species. Instead, this DNA functions solely as a spacer and DNA of any kind would have the same impact on the centromere effect. Under this model, the centromere effect is primarily mediated by the centromere and genetic factors acting in trans with respect to physical distance along the chromosome arm. In a second model, which may be designated the Epigenetic Model, heterochromatin is imbued with a capacity to both limit crossing over within and, more importantly, beyond, in flanking euchromatin. Finally, a third combined model may explain the centromere effect. In this case, the centromere effect is mediated by the centromere and pure physical distance that influence the action of trans-acting factors, but the chromatin state and/or sequence composition of DNA between the centromere and heterochromatin modulates the centromere effect.

Studies suggest that a strict epigenetic model for the centromere effect is unlikely. If the mass action of heterochromatin was the sole cause of the centromere effect, loss of pericentric heterochromatin would lead to either increased recombination in regions flanking the deleted heterochromatin (if there was a cumulative repressive effect of bulk heterochromatin) or no impact (if a domain of heterochromatin could induce a similar amount of suppression as long as it was sufficient in size). However, neither of these were observed when blocks of pericentric heterochromatin were removed. Rather, deletions of pericentric heterochromatin that bring euchromatin closer to the centromere lead to lower rates of recombination in those euchromatic regions (Yamamoto and Miklos 1978). This is consistent with the Inert Spacer Model. But, it does not exclude the possibility that the chromatin state or the DNA sequence composition of pericentric heterochromatin modulates the manner in which the centromere effect is transduced along the chromosome arm.

What explains the ameliorated centromere effect in D. mauritiana relative to D. melanogaster? Under the Inert Spacer Model, pericentric heterochromatin has no impact on the centromere effect, independent of its functioning as a spacer. Under this model, increased proximal recombination would be explained by increased amounts of spacer pericentric heterochromatin. However, this does not appear to be the case. Since there does not appear to be vastly different amounts of pericentric heterochromatin, a difference in the centromere effect must be explained by differences in centromere structure, differences in trans-acting factors that shape the recombination landscape with respect to physical distance from the centromere, differences in the nature of pericentric heterochromatin, which modulates the centromere effect, or a combination of these differences. Trans-acting factors may include Mei-352, for which mutations in D. melanogaster appear to make crossover frequencies proportional to physical distance, and C(3)G, for which variants increase pericentric recombination (Baker and Carpenter 1972; Page and Hawley 2005; Billmyre et al. 2019).

Several observations support the possibility that the evolution of trans-acting factors or the centromere itself might contribute to differences in the centromere effect. First, factors that influence meiotic recombination are known to evolve rapidly (Brand et al. 2019). Thus, it is likely that factors governing the placement of crossovers with respect to physical distance from the centromere can change in the ways they are influenced by the centromere effect. This is supported by evolutionary observations of mei-218. A previous study demonstrated that the replacement of the D. melanogaster mei-218 gene with its D. mauritiana counterpart results in an increase in exchange in the proximal euchromatin in D. melanogaster (Brand et al. 2018). Mei-218 is very rapidly evolving, with a signature of positive selection, and this rapid evolution appears to have caused a change in the way the centromere impacts recombination. However, as demonstrated by the clear centromere effect observed on 2L in D. mauritiana (see Fig. 3), it cannot be the case that the MEI-218 protein found in D. mauritiana cannot impose a centromere effect. In addition to the evolution of these trans-acting factors, the evolution of the centromere itself might have an impact. A recent study showed that there has been significant turnover in the sequences that comprise the centromere since the divergence of D. melanogaster and D. mauritiana (Courret et al. 2024). This turnover may lead to a change in the way the centromere shapes proximal meiotic recombination.

Alternatively, while the evolution of trans-acting factors and the centromere likely contribute to the amelioration of the D. mauritiana centromere effect, we propose that only some types of heterochromatic regions can buffer the centromere effect, and thus, repeat blocks in the heterochromatin may differ in their ability to attenuate the centromere effect. This is based on several observations. First, heterochromatin is known to modulate flanking recombination. For example, homozygosity for an inserted fragment of heterochromatin that gives rise to the brown^Dominant allele suppresses recombination (Slatis 1955), though heterozygosity does not appear to have this effect (Hartmann, Kohl, et al. 2019; Hartmann, Umbanhowar, et al. 2019). Furthermore, Chakraborty and colleagues demonstrated that one of the most striking differences between the D. mauritiana and D. melanogaster genomes is the existence of numerous rearrangements within the proximal heterochromatin of the arms (Chakraborty et al. 2021). We believe it is likely that these rearrangements influence the impact of the centromere effect. Secondly, Carpenter (1975b, 1975a) demonstrated with serial section electron microscopy that in D. melanogaster, the SC has a different morphology in the heterochromatin compared with euchromatin. Since C(3)G can modulate the centromere effect, changes in heterochromatin are likely to influence the centromere effect through the SC (Billmyre et al. 2019). Less obvious perturbations in SC structure might also be found in the proximal euchromatin. It will be interesting to see comparative high-resolution studies of SC structure in multiple species within the melanogaster complex.

Finally, the exception in proximal region 2L should be noted. Unlike other proximal regions, the recombination frequency in this region in D. mauritiana is actually less than what is observed in D. melanogaster. If differences in the recombination rate in proximal regions were solely explained by differences in the centromeres or trans-acting factors, this exception is difficult to explain. This is especially the case since proximal region 2R shares a centromere with 2L, and the recombination rate in 2R, on the other side of the same centromere, is about 17 times greater in D. mauritiana. Thus, if the centromere and trans-acting factors explain the difference in 2R, something must be different about the span of pericentric heterochromatin that maintains significant suppression in 2L of D. mauritiana.

Concluding thoughts: It is perhaps most surprising to us that D. mauritiana does not show a centromere effect on 4 of the 5 major arms, but it does show both a level of interference comparable to that observed in D. melanogaster and exchange suppression in the telomere adjacent regions. This suggests that changes in the mechanisms that control crossover patterning within a genome (for example, affecting only 4 of the 5 chromosome arms) can alter 1 process, such as the centromere effect, without affecting other processes, such as interference and telomere adjacent suppression. We note in that respect a recent paper showing that the high levels of crossing over and interference observed in budding yeast are not observed in many other species of yeast (Dutta et al. 2024). It occurs to us that a proper understanding of the mechanisms that control crossover placement may well require being willing and able to think outside the well-studied model organism box.

Data availability

Original data underlying this manuscript can be accessed from the Stowers Original Data Repository at http://www.stowers.org/research/publications/libpb-2507. All sequences used in this study are publicly available at https://www.ncbi.nlm.nih.gov/sra/PRJNA1168080. VCF files are provided in GSA FigShare https://doi.org/10.25386/genetics.28506611.

Supplemental material available at GENETICS online.

Funding

Funding for this work came from the Stowers Institute for Medical Research (RSH), Swiss National Science Foundation (Project#: 310030_189131) (MJ.), National Institute of Health award DP5OD033357 (DEM), and National Science Foundation award 2025197 (JB). RSH is an American Cancer Society Research Professor.