Figure 1.
New sex-specific pedigree-based genetic maps (10 kb scale).
The male map (blue) shows intense crossover activity throughout PAR1, with particularly high rates towards the telomeric end. Fine-scale variation in rates could not be estimated in two regions (1–1.4 Mb and
2–2.25 Mb, build 36) due to large unmapped and repetitive sequences and lack of genotyped SNPs. The female map (red) has a low rate through much of PAR1, and a trend of increasing rate towards the pseudoautosomal boundary. Vertical black tick marks show marker positions.
Figure 2.
Concordance between pedigree-based and HapMap2 population-averaged LD-based estimates of recombination.
(A) Comparison of sex-averaged pedigree rates and LD-based rates at the 50 kb scale shows high correlation (Spearman's ,
). Recombination in PAR1 is dominated by crossovers in males, and the LD-based map is informative about male recombination (Spearman's
,
). (B) Rates in the LD-based map (5 kb scale, at 500 bp intervals) averaged over the best-resolved 10% of paternal crossovers (n = 12, resolution 13 kb–45 kb, maximum extent shown by vertical black dotted lines), centred such that they all have their midpoint at 0 (red dotted line). PAR1-wide average LD-based rate of 9.06 cM/Mb is shown with the horizontal black dashed line. The LD-based map has a rate significantly elevated above the average rate at crossover midpoints (
).
Figure 3.
(A) Rate in the HapMap2 population-averaged LD-based map around instances of the 13-bp motif CCnCCnTnnCCnC in PAR1. Motifs in regions with an average SNP density of greater than one SNP per 2 kb in the surrounding 50 kb were included; clusters of motifs within 150 bp of one another were thinned to the most central motif. The plot shows 2 kb averaging, at 100 bp intervals. Motif positions show a strong local increase in recombination rate in the LD-based map. (B) As (A), but showing rates around ChIP-seq binding locations of the PRDM9 reference allele (B allele). In the event of more than one peak within 10 kb, only the most strongly signalled peak was included. ChIP-seq binding peaks of PRDM9 show a stronger rate increase in the LD-based map above local background rate than 13-bp motif sites.
Figure 4.
Separate LD-based recombination rates in PAR1 in three human continental groups, around the binding sites of the PRDM9 B allele.
The B and other alleles predicted to bind similar motifs predominate in Europe and East Asia (91% frequency), but not in Africa (58% frequency). In PAR1, the recombination pattern is consistent with being activated by PRDM9, as both Asian and European populations show a much stronger increase in rate at these binding sites than Africans (P = 0.002 African/Asian, P = 0.02 African/European).
Figure 5.
(A) A comparison of allele frequency spectra of different mutation types segregating in humans in PAR1. A significantly greater fraction of GC-increasing mutations have allele frequencies >90%, while a significantly greater fraction of GC-reducing mutations have allele frequencies <10%. Differences with are marked (3 asterisks indicate
while four indicate
). (B) Quantile-quantile plots show the difference between the allele frequency spectra of GC
AT mutations (x-axis) and AT
GC mutations (y-axis) in PAR1 and Chr 20. If both types of mutation had the same allele frequency spectrum, we would expect to see a straight line. Points above the diagonal indicate that AT
GC mutations are at higher frequencies than GC
AT mutations, while points below the diagonal show the opposite trend. For example, the green dashed guide lines show that, in PAR1 (black) the same proportion of AT
GC sites has allele frequencies
as GC
AT sites with allele frequencies
. This bias towards higher allele frequencies for AT
GC mutations is significant (
). It is comparable to the bias in the hottest 15% of loci, 1 kb in size, in Chr 20 (red), which have a sex-averaged rate comparable to that of PAR1 as a whole (8.2 cM/Mb). The coldest 15% of loci in Chr 20 (average rate
cM/Mb) do not show a significant elevation of GC allele frequencies (
).
Figure 6.
Allele frequencies near the PRDM9 binding motif show the strong influence of recombination.
(A) Quantile-quantile plots comparing ATGC and GC
AT allele frequencies in the vicinity of the motif CCnCCnTnnCCnC relative to the rest of PAR1 in human populations. The elevation of GC allele frequencies and suppression of AT alleles is extremely powerful closest to the motif, and drops off rapidly with distance away from it. (B) A comparison of the allele frequency spectra of different mutation types within 25 bp of copies of the motif CCnCCnTnnCCnC in PAR1 and Chr 20 shows an extreme skew towards GC bases in PAR1.
Table 1.
Results of predicting divergence rate of different types of substitutions from human recombination rate, after regressing out effects of GC content, CpG content and the divergence rate of other types of substitutions in unique DNA.
Figure 7.
The rate of GC-biased substitutions in human and chimpanzee, in human hotspots, suggests no sharing of human and chimpanzee hotspots in PAR1.
These plots show histograms for the estimated increase in the rate of GC-biased transition substitutions in regions overlapping human hotspots (red) and coldspots (blue), relative to the rest of PAR1 in: (A) Human: High recombination rate regions show a significant excess of GC-biased transitions in humans (+0.21% per base) while low recombination rate regions show a lower substitution bias towards GC transitions (−0.05% per base) relative to regions with intermediate rates. The difference between hotspots and coldspots is significant (). (B) Chimpanzee: There is no systematic increase in the rate of GC-biased transitions in chimpanzee (−0.01% per base) in regions containing human hotspots. The difference between hotspots and coldspots is not significant (
).