In a systematic population survey including more than 8,000 antigen D negative blood donations, we identified 14 different RHD positive antigen D negative and 3 different Del haplotypes, the majority of which were novel. The molecular bases were alleles comprising RHD/RHCE hybrids, stop codons, missense mutations and splice site mutations. The cumulative frequency of RHD gene positive antigen D negative haplotypes was about 1:1,500; that of the D
alleles was about 1:3,000. We determined the specificity of antigen D prediction by PCR and devised an optimized RHD PCR strategy with a calculated positive predictive value greater than 0.9999. Five antigen D positive samples missed by routine D typing were uncovered and two anti-D immunizations traced.
For practical purposes, two groups of RHD alleles that do not express antigen D can be distinguished. RHD alleles of the first group lack some or many RHD specific polymorphism and usually represent RHD/CE hybrids. For alleles of this group, a correct antigen D prediction may be accomplished by a prudent selection of the RHD specific polymorphism utilized for RHD genotyping. RHD alleles of the second group carry all RHD specific polymorphism and most often harbor point mutations. For alleles of this group, a correct antigen D prediction necessitates the specific detection of an aberration that is usually unique to the allele. The identification of four new alleles in this group increased the number of known alleles from 3 to 7 and was critical for improving RHD genotyping.
The data of this study allowed for the first time to calculate population frequencies of RHD positive antigen D negative and D
alleles. This information was indispensable to derive rational RHD typing strategies and will be essential for establishing cost-efficient approaches. The majority of samples belonging to the first group of D negative alleles (probable RHD/CE hybrids) was compatible with RHD-CE-D hybrid alleles, in which the DNA segment derived from the RHCE gene encompassed at least exon 4 to exon 7. These samples would be correctly typed, if exon 4/intron 4 and exon 7 were used for RHD genotyping, as proposed previously . With the exception of RHD exon 9, testing additional RHD exons would not have improved the specificity of antigen D prediction. Improving this specificity, however, became possible by the specific detection of frequent alleles of the second group, like RHDψ and RHD(W16X). We demonstrated that testing 5 carefully selected polymorphism would have resulted in an assay yielding false positive results at a rate less than 1:12,000, and hence would have doubled the specificity compared to contemporary approaches testing all informative RHD exons [6,26]. Further improvements may be achieved by the specific detection of additional alleles, that might become practical in massively parallel molecular assays.
The detailed analysis including intron polymorphism revealed that the first group of alleles (probable RHD/CE hybrids) represented at least 9 different molecular events. We proposed that the proximity and inverse orientation of both RH genes favored gene conversions occurring in cis (Fig. 6), which have also been noted in partial D . An exact definition of the molecular bases of the RHD/CE hybrids would allow their specific detection, even if they were positioned in trans to the regular RHD allele. Such a detection would be necessary, if molecular RH zygosity testing is expected to achieve the same specificity as antigen D prediction.
A considerable proportion of seemingly D negative samples carrying the RHD gene presented a Del phenotype. Interestingly, RHD(M295I) coded for weak D, if associated with a ce haplotype , but for Del, if associated with a Ce haplotype; this observation may be explained by the suppressive effect of C in cis .
The nature and frequency of RHD gene positive antigen D negative alleles differ among populations. Apart from a probably lower absolute frequency, we detected in Europeans many parallels to oriental populations: Both populations shared the diverse nature of RHD haplotypes of the first group (probable RHD/CE hybrids) [18,19,20], the preferential occurrence of RHD positive antigen D negative alleles in Cde haplotypes , and the comparatively frequent observation of Del phenotypes . In contrast, RHDψ and Cde
are predominant in African populations . Still another situation may be present in the middle-west USA, where 6 of 26 RHD gene positive antigen D negative samples had aberrations limited to a single exon yet detectable by PCR .
Blood group serologists might note the observation of 5 D positive samples in our study with disturbance. In many centers, donors are checked for antigen D by sensitive methods at first and second donations only. On subsequent donations, carriers of partial D, like DvI or DIM, some weak D and D+/- chimerism may pass unnoticed in tests based on direct agglutination, even with the most avid IgM anti-D. Immunizations caused by units of such donors will generally be missed, because the occurrence of an anti-D in a patient is usually not further investigated . For example, the two anti-D immunizations induced by units of the chimerical donor of this study were found only in a look-back triggered by our molecular screen. Chimeras in the Rh system have repeatedly been observed [36,37] and chimeras may be a more widespread phenomenon than anticipated . A lower antigen density threshold for anti-D immunization has not been established yet, and future studies might indicate a need to exclude even Del donors from transfusion to D negative recipients. A routine investigation of all samples by adsorption and elution is not feasible. However, checking D-negative samples, especially those occurring with a C or E or both, for RHD specific sequences by nucleic amplification techniques may become practical in the near future. The knowledge of the detected alleles is also important for fetal genotyping assays using fetal DNA in maternal plasma, because false positive results will be obtained in mothers harboring RHD positive D negative alleles.