Broad-scale recombination patterns underlying proper disjunction in humans

Abstract

Although recombination is essential to the successful completion of human meiosis, it remains unclear how tightly the process is regulated and over what scale. To assess the nature and stringency of constraints on human recombination, we examined crossover patterns in transmissions to viable, non-trisomic offspring, using dense genotyping data collected in a large set of pedigrees. Our analysis supports a requirement for one chiasma per chromosome rather than per arm to ensure proper disjunction, with additional chiasmata occurring in proportion to physical length. The requirement is not absolute, however, as chromosome 21 seems to be frequently transmitted properly in the absence of a chiasma in females, a finding that raises the possibility of a back-up mechanism aiding in its correct segregation. We also found a set of double crossovers in surprisingly close proximity, as expected from a second pathway that is not subject to crossover interference. These findings point to multiple mechanisms that shape the distribution of crossovers, influencing proper disjunction in humans.

Figure 1. For each chromosomal arm, the estimated fraction of bivalents that segregated properly without a chiasma, in male (blue) and female (red) transmissions.

Shown are maximum likelihood estimates and 95% confidence intervals (see Methods). Estimates significantly above 0 at the 5% level are indicated by a yellow dotted line and at the 1% level by a green dotted line. The data for chromosome 21 combine two sets of pedigrees (see Methods).

Figure 2. For each chromosome, the estimated fraction of bivalents that segregated properly without a chiasma, in male (blue) and female (red).

See the legend of Figure 1 for additional details.

Figure 3. The estimated distribution of chiasmata per bivalent, in female transmissions of chromosome 21.

To the left is the inferred distribution of chiasmata per bivalent, obtained from a maximum-likelihood method. The point estimates are shown as filled red circles and the 95% confidence intervals as orange dotted lines (see Methods). To the right is the observed distribution of crossovers among female transmissions to viable, non-trisomic offspring, in two sets of pedigrees (see Methods). As a test of the goodness of fit of the model, we also show (in gray dotted lines) the prediction intervals obtained from 5000 simulations where we binomially sample the number of crossovers per gamete distribution given the estimated distribution of chiasmata per bivalent; as can be seen, the data fit well within the 95% prediction intervals (see Methods). The data indicate that a substantial and significant (p = 0.0002) fraction of female transmissions involved a nullichiasmatic chromosome 21.

Figure 4. Double crossovers in close proximity.

The number of double crossovers within 5 cM or less versus the number expected under a gamma renewal model, as estimated from the data (distances are based on sex-specific genetic maps; see Supplementary Methods in Text S1). In red are events in female transmissions; in blue are events in male transmissions. A version of this figure with stringently vetted double crossovers is presented in Supplementary Figure 4 in Text S1.

Figure 5. The total recombination rate per chromosome, in males and females, as predicted by an obligate chiasma per bivalent (i.e, an intercept at 0.5 crossovers per gamete) and the physical length of the chromosome.

In red are the observed values for females and in blue the ones for males; circles denote metacentric chromosomes and diamonds acrocentric ones. The regression line is shown for each sex separately. In both cases, there is an excellent fit of this simple model to the data (R² = 0.94 for females; R² = 0.95 for males).