Meiotic, genomic and evolutionary properties of crossover distribution in Drosophila yakuba

Abstract

The number and location of crossovers across genomes are highly regulated during meiosis, yet the key components controlling them are fast evolving, hindering our understanding of the mechanistic causes and evolutionary consequences of changes in crossover rates. Drosophila melanogaster has been a model species to study meiosis for more than a century, with an available high-resolution crossover map that is, nonetheless, missing for closely related species, thus preventing evolutionary context. Here, we applied a novel and highly efficient approach to generate whole-genome high-resolution crossover maps in D. yakuba to tackle multiple questions that benefit from being addressed collectively within an appropriate phylogenetic framework, in our case the D. melanogaster species subgroup. The genotyping of more than 1,600 individual meiotic events allowed us to identify several key distinct properties relative to D. melanogaster. We show that D. yakuba, in addition to higher crossover rates than D. melanogaster, has a stronger centromere effect and crossover assurance than any Drosophila species analyzed to date. We also report the presence of an active crossover-associated meiotic drive mechanism for the X chromosome that results in the preferential inclusion in oocytes of chromatids with crossovers. Our evolutionary and genomic analyses suggest that the genome-wide landscape of crossover rates in D. yakuba has been fairly stable and captures a significant signal of the ancestral crossover landscape for the whole D. melanogaster subgroup, even informative for the D. melanogaster lineage. Contemporary crossover rates in D. melanogaster, on the other hand, do not recapitulate ancestral crossovers landscapes. As a result, the temporal stability of crossover landscapes observed in D. yakuba makes this species an ideal system for applying population genetic models of selection and linkage, given that these models assume temporal constancy in linkage effects. Our studies emphasize the importance of generating multiple high-resolution crossover rate maps within a coherent phylogenetic context to broaden our understanding of crossover control during meiosis and to improve studies on the evolutionary consequences of variable crossover rates across genomes and time.

Fig 1. Dual-barcoding genotyping method used to obtain crossover rates.

Diagnostic SNPs are used as genetic barcodes and allow the pooling of multiple F₂ individuals from different crosses for a given sequence barcode (left panel). The combination of genetic and sequence barcodes ensures efficient genotyping of multiple individuals and accurate crossover localization along chromosome arms (right panel). Diagnostic, or strain-specific SNPs, are singletons for the complete set of genotypes used in the study, including the tester line.

Fig 2. Number and distribution of crossovers.

A) Observed crossovers per chromatid in D. yakuba (this study) and D. melanogaster (see Materials and Methods for details). Note that studies using visible markers are based on much larger sample sizes but, at the same time, may underestimate crossover events due to limited density of markers along chromosome arms. NCO: non-crossover, 1CO: single crossover, 2CO: double crossover, 3CO: triple crossover, 4CO: quadruple crossovers per chromosome arm. B) Relative location of crossovers along chromosome arms in D. yakuba. Each horizontal line represents a chromosome with one or more crossovers. A change in color represents a crossover event, starting from the telomere (blue). Chromosome arms have been ordered based on crossover class for better visualization, from 1CO (top) to 4 CO (bottom). Number of chromatids analyzed (n): X: n = 1,622; 2L: n = 1,661; 2R: n = 841; 3L: n = 1,701; 3R: n = 1,635.

Fig 3. High-resolution crossover maps for D. yakuba.

Crossover rates (cM/Mb) are shown along chromosome arms from three different crosses as well as the average (thick red line). Maps are shown for overlapping 250-kb windows with increments of 50 kb. Below each chromosome, vertical blue and red bars indicate the presence of transposable and INE-1 elements, respectively. The scale for the different chromosome arms is equivalent and differences in figure size capture differences in chromosome arm length.

Fig 4. Centromere and telomere effects for different chromosome arms of D. yakuba and D. melanogaster.

For D. melanogaster results are shown for genome r5.3 and r6 assemblies. The proportion of the chromosome with significant reduction in crossovers is based on the study of overlapping 1-Mb windows with increments of 100 kb. Two levels of significance are shown.

Fig 5. Maximum Likelihood (ML) estimates of tetrad frequencies.

A) Estimates of E₀ under a ML model with unrestricted E_r. Probability (y-axis) of models with variable E₀ (x-axis), with E_r>0 allowed to vary to provide the best fit for a given E₀. ¹ Estimates of tetrad frequency in D. melanogaster based on the distribution of 196 meiotic products analyzed by Miller et al. (2016) using whole genome sequencing (WGS). ² Estimates of tetrad frequency in D. melanogaster based on visible markers (see text). B) Estimates of tetrad frequencies based on ML models with a biologically relevant range for E_r (E_r ≥ 0). Note that the best model for the D. yakuba X chromosome (E₀ = 0) is, nonetheless, incompatible with the observed data [P(E₀ = 0) = 1.6x10^-13]. E₀: tetrads that do not undergo crossing over, E₁: tetrads with 1 CO, E₂: tetrads with 2 COs, E₃: tetrads with 3 COs, and E₄: tetrads with 4 COs.

Fig 6. Model of tetrad frequencies with crossover-associated meiotic drive (MD_CO).

A) MD_CO model where chromatids with crossovers are preferentially transmitted to the oocyte with a bias b (b > 0.5) when the sister chromatid has no crossovers. B) ML estimates of tetrad frequencies for the D. yakuba X chromosome under a MD_CO model, with a best fit when b = 0.86. C) Observed and predicted frequencies of crossover classes for the chromosome X of D. yakuba under the MD_CO model. D) Joint maximum-likelihood estimates (MLE) of E₀ and b for the X chromosome of D. yakuba (left) and D. melanogaster (WGS; right). The red dot represents point estimates with maximum fit to the data.

Fig 7. Correlation between codon usage bias (CUB) and crossover rates in D. yakuba (Rec_yak) or D. melanogaster (Rec_mel).

A generalized linear model (GLM) was used to estimate the correlation coefficient R between crossover rates (log₁₀), either Rec_yak or Rec_mel, and CUB estimated for each gene in the five species analyzed. Numbers in black indicate significant estimates of R in the direction predicted by models of selection and linkage. Red numbers indicate a significant association in the opposite direction than that predicted by models. *, P < 0.05; **, P < 0.01); n.s., non-significant association (P > 0.05).

Fig 8. Correlation between rates of protein evolution and crossover rates in D. yakuba (Rec_yak) or D. melanogaster (Rec_mel).

A) For each branch across the D. melanogaster subgroup phylogeny, estimates of the efficacy of selection on amino acid changes (ω_R) per gene were compared to crossover rates, either Rec_yak or Rec_mel, with a generalized regression model (GRM) to estimate the correlation coefficient R. B) For each branch across the phylogeny, genes with and without signal of positive selection based on PAML (see text for details) were compared to crossover rates for these same genes. Odds Ratio (OR) from logistic regression analysis was applied to capture variable likelihood of positive selection with crossover rates (see text for details). Numbers in black indicate significant estimates of R in the direction predicted by models of selection and linkage whereas red numbers indicate a significant R in the opposite direction. *, P < 0.05; **, P < 0.01; n.s., non-significant association (P > 0.05).

Fig 9. Relationship between crossover rate (cM/Mb) and levels of neutral nucleotide polymorphism (π_4f) in D. yakuba and D. melanogaster.

π_4f indicates pairwise nucleotide variation (/bp) at four-fold synonymous sites. Autosomal and X-linked regions are indicated as blue and red circles, respectively. Crossover rates are adjusted to allow a direct comparison between X-linked and autosomal regions. Dashed lines indicate linear regressions between π_4f and adjusted crossover rates; solid lines indicate regressions between π_4f and the log₁₀ of crossover rates. Data shown for non-overlapping 100-kb windows.