PRDM9 variation strongly influences recombination hot-spot activity and meiotic instability in humans

Abstract

PRDM9 has recently been identified as a likely trans regulator of meiotic recombination hot spots in humans and mice. PRDM9 contains a zinc finger array that, in humans, can recognize a short sequence motif associated with hot spots, with binding to this motif possibly triggering hot-spot activity via chromatin remodeling. We now report that human genetic variation at the PRDM9 locus has a strong effect on sperm hot-spot activity, even at hot spots lacking the sequence motif. Subtle changes within the zinc finger array can create hot-spot nonactivating or enhancing variants and can even trigger the appearance of a new hot spot, suggesting that PRDM9 is a major global regulator of hot spots in humans. Variation at the PRDM9 locus also influences aspects of genome instability-specifically, a megabase-scale rearrangement underlying two genomic disorders as well as minisatellite instability-implicating PRDM9 as a risk factor for some pathological genome rearrangements.

Figure 1

PRDM9 ZnF variants and crossover hot-spot activity in sperm. (a) Examples of tandem repeats encoding the ZnF array, with variant repeat units coloured differently. The predicted DNA binding sequence is shown below, with dots indicating weakly and uppercase the most strongly predicted bases, and aligned with the hot-spot motif CCNCCNTNNCCNC (ref. 4). The binding sequence for allele A matches all 8 specified bases in the motif, while allele L1 matches at best only 6 of the 8 bases. (b) PRDM9 ZnF array diversity in Europeans and Africans, with alleles classified either by structure or by the strength of the predicted match with the motif (details of alleles provided in Supplementary Fig. 1). (c) Variation between men in sperm crossover activity in hot spots, named above each panel, containing a central hot-spot motif. Different sets of men informative for SNPs required for crossover detection were analysed at each hot spot (Supplementary Table 1). Men carrying two PRDM9 A alleles (A/A, black), one A allele (A/N, blue) or two non-A alleles (N/N, red) were grouped separately in ascending order. Confidence intervals for each estimate of RF are shown, and median RFs within each group are indicated by dotted lines. Mann-Whitney test results for the significance of differences between the A/A group and the A/N or N/N groups are given at top right (ns P > 0.05; * P < 0.05; ** P < 0.01; *** P < 0.001). (d) Corresponding analyses of hot spots lacking an obvious hot-spot motif (Supplementary Fig. 2).

Figure 2

Activating and non-activating PRDM9 alleles. Non-activating alleles at each hot spot were defined as alleles present in N/N men who showed <5% of the median RF seen in A/A men. Allele structures are coded as in Supplementary Fig. 1, with predicted DNA binding sequences and best motif matches shown as in Fig. 1. Data for each hot spot give the number of specific N alleles detected in suppressed men; for example, 5 such men typed at hot-spot F carried the C allele. Evidence that the B allele is active is based on a B/L6 heterozygote assayable only at hot-spot CG who showed crossovers at 40% of the median frequency seen in A/A homozygotes. Since allele L6 is a non-activator at CG, this implies that B is similar in activity to A.

Figure 3

PRDM9 variation influences crossover activity at hot-spots MSTM1a and MSTM1b. (a) Variant alleles present in men previously typed for sperm crossovers at these hot spots; all other men were A/A homozygotes. RFs at each hot spot are shown, together with 95% CIs. Note that alleles L9/L24 associate not only with MSTM1a activity but also apparently with elevated RF at MSTM1b (P = 0.035). (b) DNA sequence structures of these alleles, coloured as in Fig. 2, plus amino acid sequence changes relative to allele A, with locations given with respect to the main ZnF DNA-contact residues (−1, 2, 3 and 6).

Figure 4

Influence of PRDM9 variation on minisatellite instability in sperm. (a) Minisatellite repeat units, with the best matches to the hot-spot motif indicated. Minisatellites MS1 and CEB1 both contain perfect matches. Two contiguous repeats are shown for MS1. (b) Representative small-pool PCR results for minisatellite CEB1, for a PRDM9 A/A homozygote and C/L16 heterozygote. (c) Mutation frequencies in A/A (black), A/N (blue) and N/N (red) men, with CIs plus median values per group indicated as in Fig. 1.

Figure 5

PRDM9 variation and de novo HNPP/CMT1A rearrangements in sperm DNA. (a) Detection of HNPP deletion junctions in sperm DNA from a PRDM9 A/A homozygote and a C/L14 heterozygote, with 40 ng DNA input per PCR reaction. (b) Rearrangement frequencies in A/A (black), A/N (blue) and N/N (red) men for HNPP deletions and CMT1A duplications, with CIs and median values per group shown as in Fig. 1. The same men were analysed for both rearrangements, but not all men were typed for duplications, which arise at a lower frequency than deletions. Rearrangements in blood were rare (<2.2×10⁻⁶ for deletions, <1.3×10⁻⁶ for duplications, P > 0.95).

Figure 6

PRDM9 variation and t(11;22) translocation frequencies. (a) Detection of de novo der(22) translocation junctions by nested PCR amplification of multiple 150 ng aliquots of blood and sperm DNA. Minor variation in junction size results from differences in translocation breakpoint locations within the PATRR. (b) der(22) translocation frequencies, with CIs, in sperm DNA in PRDM9 A/A, A/N and N/N men, with identities of N alleles shown above. The median frequency per group, similar to previously-reported translocation frequencies, is indicated by a dotted line. There is no significant difference in translocation frequency between groups (Kruskal-Wallis test, P = 0.98). No de novo translocations were seen in blood DNA tested from four different men (frequency < 5×10^–7, P > 0.95).