Evolutionary genetics of the human Rh blood group system

Abstract

The evolutionary history of variation in the human Rh blood group system, determined by variants in the RHD and RHCE genes, has long been an unresolved puzzle in human genetics. Prior to medical treatments and interventions developed in the last century, the D-positive (RhD positive) children of D-negative (RhD negative) women were at risk for hemolytic disease of the newborn, if the mother produced anti-D antibodies following sensitization to the blood of a previous D-positive child. Given the deleterious fitness consequences of this disease, the appreciable frequencies in European populations of the responsible RHD gene deletion variant (for example, 0.43 in our study) seem surprising. In this study, we used new molecular and genomic data generated from four HapMap population samples to test the idea that positive selection for an as-of-yet unknown fitness benefit of the RHD deletion may have offset the otherwise negative fitness effects of hemolytic disease of the newborn. We found no evidence that positive natural selection affected the frequency of the RHD deletion. Thus, the initial rise to intermediate frequency of the RHD deletion in European populations may simply be explained by genetic drift/founder effect, or by an older or more complex sweep that we are insufficiently powered to detect. However, our simulations recapitulate previous findings that selection on the RHD deletion is frequency dependent and weak or absent near 0.5. Therefore, once such a frequency was achieved, it could have been maintained by a relatively small amount of genetic drift. We unexpectedly observed evidence for positive selection on the C allele of RHCE in non-African populations (on chromosomes with intact copies of the RHD gene) in the form of an unusually high F( ST ) value and the high frequency of a single haplotype carrying the C allele. RhCE function is not well understood, but the C/c antigenic variant is clinically relevant and can result in hemolytic disease of the newborn, albeit much less commonly and severely than that related to the D-negative blood type. Therefore, the potential fitness benefits of the RHCE C allele are currently unknown but merit further exploration.

Figure 1. RHD deletion genotyping and fiber FISH validation

(a) The paralogous RHD and RHCE genes (98% nucleotide sequence similarity) are located on human chromosome 1p36.11 in inverted orientation. The genes are ~60 kb each in size, separated by ~30 kb. Smaller “Rhesus box” segmental duplications (~9 kb each) with 99% sequence similarity flank RHD and provided the substrate for gene deletion via non-allelic homologous recombination(Wagner and Flegel 2000). We used these features to isolate and distinguish RHD in TaqMan genotyping and fiber FISH validation assays of the deletion. We also resequenced ~12 kb from two unique regions of this locus, as indicated, in a subset of the individuals in our study. (b) TaqMan results for the HapMap population samples and allele frequencies. Note that one YRI and one JPT individual were estimated to have 3 RHD copies. (c-e) Fiber FISH validation for three HapMap trios, with representative images from the two alleles for each individual. The CEU trio depicted in (c) represents the scenario in which the largest proportion of offspring are at for hemolytic disease of the newborn, in which the mother is D-negative and the father has two copies of RHD, such that all of their offspring would be D-positive. The YRI trio depicted in (d) includes one of the individuals estimated to have 3 RHD copies (GM19204). We confirmed the duplication and infer that its origin was also mediated by non-allelic homologous recombination of the Rhesus box segmental duplications; this is the expected reciprocal of the RHD gene deletion. The duplication allele was not transmitted to the child (GM19205). The Yoruba trio depicted in (e) initially seemed a case of non-Mendelian inheritance based on total diploid copy numbers from our TaqMan copy number estimates; however, direct determinations of the copy numbers of each chromosome by fiber FISH show that the father (GM19153) has one copy each of the RHD duplication and deletion alleles, and the child (GM19154) inherited the deletion allele.

Figure 2. Evidence for positive selection on the C allele of RHCE

(a) Cumulative F_ST distributions of HapMap Phase II SNPs for each population comparison (CHB and JPT are considered together), with F_ST values for the functional RHD/CE variants indicated by the circle, triangle, and square symbols. The (CHB+JPT)-YRI F_ST value for the Cc variant is exceptional compared to the distribution of HapMap SNPs, consistent with positive selection. (b) Network analysis of inferred haplotypes from RHD and RHCE functional variants and SNPs within and flanking the locus (from ~12 kb resequencing data), for CEU (n = 44 chromosomes) and CHB (n = 46 chromosomes) together. Circles represent haplotypes with the area of each proportional to frequency; the total number of chromosomes is shown next to each haplotype circle where n > 1. The shortest lines between haplotype circles indicate one mutational step. RHD/CE haplotypes are indicated by circle colors (historical RHD/CE haplotype nomenclature is provided in parentheses). The frequency of the principal DCe haplotype (n = 47 chromosomes) is unexpected given the overall pattern of haplotype diversity at this locus, a signature consistent with positive selection.

Figure 3. Fertility rate and inter-birth interval analysis by D status in the Hutterites

(a) Frequency distributions of the number of children per family, for families with at least one child born prior to 1968 (the first year of RhoGAM availability for this population), separately for families potentially affected by hemolytic disease of the newborn (D-negative mother and D-positive father) and unaffected families (D-positive mother or D-negative father or both). (b) Cumulative distributions of inter-birth intervals (days) for potentially affected (black) and unaffected (grey) families. (c) Frequency distributions of birth years for children in potentially affected and unaffected families.

Figure 4. Influence of parameter variation on frequency-dependent selection at the RHD locus

Simulated change in RHD deletion frequency over one generation (y-axis), given initial RHD deletion frequency (x-axis) and evaluated under varying (a) probability of sensitization if the mother is D-negative and the child is D-positive (ps), (b) probability that an individual child of a sensitized mother would die from hemolytic disease of the newborn (pa), (c) mean family sizes, in families not potentially affected by hemolytic disease of the newborn (fs), and (d) maximum number of attempts to conceive (na). For each round of simulations (a-d), the remaining three of the four variables were fixed at reasonable values (ps = 0.2, pa = 0.9, fs = 3, na = 20).