Fig. 4: Non-Mendelian cosegregation of crossover products maintains heterozygosity.

a, Cartoon depicting the four possible segregation patterns following a crossover. Shown are the stage following meiosis II, at which there are four haploid pronuclei and the stage following central fusion, when the diploid zygote is formed. For each pattern, the expected and observed numbers of crossovers detected by linked-read sequencing are shown. Crossovers cannot be detected if neither crossover product was inherited, so the numbers for that outcome are listed as not available (NA). Under Mendelian segregation, the probability of loss of heterozygosity is 0.5 but such values are incompatible with our results. If crossover products cosegregate (Cosegregation bias) such that they are always inherited together, heterozygosity would rarely be lost. b,c, Probability mass functions of binomial models depicting the number of chromosomes expected to incur a loss of heterozygosity per meiosis, assuming the adjusted empirical crossover rate of 5.5 crossovers per meiosis. b, The model assumes Mendelian segregation. c, The model assumes a cosegregation probability of 0.9 (chosen because, at this probability, it is more likely for heterozygosity to be lost for zero chromosomes than for one or more chromosomes). Above the plots are the probabilities that losses of heterozygosity occur on zero chromosomes or on one or more chromosomes. d, Developmental mortality, shown as the proportion of individuals that died between the early egg stage and pupation in seven replicate colonies (magenta dots); mean (dotted line) and 95% CI (black bar). e, Theoretical model depicting the ability of homozygous lethality to maintain heterozygosity if segregation is Mendelian. Shown in green is the expected proportion of offspring with a loss of heterozygosity on one or more chromosomes (from the binomial model depicted in b) minus the homozygous lethality. Values based on the mean empirical crossover rate are shown as a dark line and values based on the 95% CI are shown in green shading. The empirical developmental mortality rate (which is the hypothetical maximum homozygous lethality) is shown in magenta (mean depicted by dark line, 95% CI depicted by shading), revealing that at least 86% of offspring should bear a loss of heterozygosity. The empirical proportion of offspring with at least one loss of heterozygosity is shown in orange (mean depicted by dark line, 95% CI depicted by shading), revealing that developmental mortality rates would need to be at least 0.87 for homozygous lethality to produce proportions that fall within the observed range. Thus, homozygous lethality can have at most a small effect on heterozygosity maintenance. f, Heatmap depicting the proportion of offspring expected to bear a loss of heterozygosity on one or more chromosomes. ‘Reasonable’ parameter combinations can be found above the black line, which depicts proportions of 0.2, the upper bound of our 95% CI for proportions of offspring that bear one or more loss of heterozygosity. The empirical developmental mortality rate (the hypothetical maximum homozygous lethality) is shown in magenta. Therefore, the only parameter combinations that can explain the degree of heterozygosity maintenance observed in this study are found within the blue dashed outline, revealing that a cosegregation probability >0.91 occurs in O. biroi. g, Heterozygosity is maintained via a programmed violation of Mendelian segregation, whereby crossover products cosegregate such that both are inherited.

Source data