The phylogenetic distribution of ABO phenotypes and genotypes. Shown is a phylogenetic tree of primate species, with a summary of phenotypic/genotypic information given in the first column, and the genetic basis for the A versus B phenotype provided in the second column (functionally important codons at positions 266 and 268 are in uppercase letters). See Dataset S1 for the source of information about phenotypes/genotypes. Only species with available divergence times are represented here (34 of 40). The phylogenetic tree is drawn to scale, with divergence times (on the x axis) in millions of years taken from ref. 29. OWM, Old World monkeys; NWM, New World monkeys. Under a model of convergent evolution, these data suggest that A is the ancestral allele, and a turnover (e.g., a neutral substitution) occurred on the branch leading to Old World monkeys. If instead, B were ancestral, all Old World monkeys would have had to serendipitously converge from ATG to TTG to encode a leucine, whereas all New World monkeys and hominoids would have had to converge to the CTG codon.

Expectations under the two possible evolutionary hypotheses. Schematic of expectations for (A) the trans-species polymorphism and (B) the convergent evolution hypotheses. We illustrate the latter case assuming that A is ancestral. The population tree is outlined in gray, and the A and B lineages sampled from three species are shown in blue and red, respectively. Under the trans-species polymorphism hypothesis, the divergence between A and B is deeper than the species split and species share a short ancestral segment with shared polymorphisms; neither of these patterns is expected under a model of convergent evolution. (C) Plot of the expected length of the segment in which a signal of a trans-species polymorphism should be detectable. Specifically, we present the expected (two sided) segment contiguous to the selected site in which the divergence between A and B lineages from different species should exceed the divergence between A (or B) lineages from different species (Methods and SI Methods, SI Note S4). We considered a recombination rate of 1 cM/Mb, because the recombination rate in this exon is estimated to be ∼1 cM/Mb in humans (54) and higher in chimpanzees (55), and varied the generation time between 10 and 20. Higher recombination rates lead to shorter segment lengths (Fig. S2). OWM, Old World monkeys.

Tree of ABO exon 7 alleles in (A) hominoids and (B) Old World monkeys. In A, the tree is based on 300 bp; in B, the tree is based on a smaller window of 200 bp, because the shorter generation time in Old World monkeys should lead to a smaller segment with a signal of a trans-species polymorphism (2) (Fig. 2C and Methods). The tree is centered on the two functional sites. We excluded rare recombinant haplotypes between functional classes (SI Methods, SI Note S3). In red is the median clade credibility (based on three runs that yielded identical consensus trees), i.e., the proportion of trees sampled from the posterior distribution that had this clade (56).

Evidence for a trans-species polymorphism in Old World monkeys and hominoids. Shown in A for Old World monkeys and B for hominoids are the synonymous pairwise differences (dS) among ABO haplotypic classes in 201-bp (i.e., 67 codons) sliding windows, as well as the shared SNPs between the species compared. For details about how dS and the 95% confidence interval were estimated, see Methods. Genome-wide mean synonymous diversity estimates within species were taken from ref. 45. When multiple species were available per genera, we chose one representative with the largest sample size, namely Macaca mulatta, Colobus angolensis, and Hylobates lar. Only the most informative comparisons are presented here, with the rest shown in Figs. S3 and S4 (which also includes similar figures using a larger sliding window choice of 300 bp). Because of recombination, the segment carrying the footprint of a trans-species polymorphism will not necessarily be contiguous and hence, even if there were no stochasticity in the mutation process, diversity levels may be jagged and may only be unusually deep in (at most) small windows (2). In C, a depiction is shown of the synonymous substitutions inferred to have occurred in the hominoid phylogeny (represented as ticks, in pink for monomorphic lineages with only one allelic class and in purple for polymorphic lineages with multiple allelic classes). Numbers along the lineages represent millions of years. A and B lineages are shown in blue and red, respectively. Orangutan is not shown because it appears that a recent turnover occurred in this species, so the lineage is not informative for our test; similarly, the branch from the common ancestor with chimpanzee to bonobo is too short to be informative (Methods). Parsimony was used to assign synonymous changes to hominoid lineages. Gorilla is shown closest to human because a substitution (at a non-CpG site) is inferred to have occurred in the ancestor of humans and gorillas but is not found in chimpanzees, suggesting incomplete lineage sorting in this region of the genome (49) (Dataset S3); treating it instead as a multiple hit in humans and gorillas only strengthens our conclusions.