Three Molecular Structures Cause Rhesus D Category VI Phenotypes With Distinct Immunohematologic Features

THE RHESUS BLOOD GROUP system is of great importance for transfusion medicine because of the high immunogenicity of its antigens. Rhesus antigens are carried by two highly homologous proteins, the RhD and RhCE proteins.1-6 The D antigen (ISBT 004.001; RH1) determined by the RhD protein is the most important Rhesus antigen and the leading cause for hemolytic disease of the newborn.7 About 17% of Caucasians lack the expression of the D antigen.8 The transfusion of a single unit of D positive red blood cells to a D negative patient is associated with an immunization rate of greater than 80%.9 

The D antigen comprises several different antigenic epitopes. Rare individuals carry a partial D antigen10 and may produce alloantibodies directed against D epitopes that are lacking in their RhD protein. Initially, these individuals have been classified into six distinct categories (D^II to D^VII, D^I being obsolete) based on the mutual reactivity with polyclonal anti-D sera from immunized partial D carriers.11Today, characterization of partial D is performed by differential reactivity with monoclonal anti-D antibodies.12,13 D category VI (D^VI) is the clinically most important partial D. Severe cases of hemolytic disease of the newborn have occurred in RhD positive babies born to D^VI mothers with anti-D.14 D^VI is the most abundant serologically defined partial D occurring among weak D samples. D^VI is reported to comprise about 6% to 10% of weak D samples8,15 and has a phenotype frequency of 1:6,200 in Germany (range, 0.02% to 0.05% in Caucasians).8,15,16 The majority of RhD positive individuals with allo-anti–D were D^VI.17 

D^VI occurs in CDe, cDE, and cDe haplotypes.17The CD^VIe haplotype is due to an RHD-RHCEhybrid molecule in which exons 4 to 6 of RHD were substituted by the respective exons of RHCE.18Samples with a cD^VIE haplotype were initially assumed to carry a deletion of exons 4 to 6,18 but in fact, are due to an exon 4 to 5 hybrid.19,20 Most CD^VIe haplotypes carry the low frequency Rhesus antigen BARC (ISBT 004.052; RH52).21 No other consistent serologic differences in D^VI have been described.21 

We recently completed a serologic random survey for partial D in southwestern Germany.8 22 Here we report the results of the molecular and immunohematologic work-up of D^VI samples. We describe a novel molecular event that caused a D^VIphenotype carrying a normal number of RhD proteins accessible on the red blood cells' surface. We show that the three D^VItypes may be readily discriminated by flow cytometry based on distinct immunohematologic features. We demonstrate considerable differences in the distribution of D^VI types within German-speaking populations showing the importance of a full molecular description for Rhesus genotyping purposes, eg, in prenatal testing.

EDTA- or citrate-anticoagulated blood samples came from southwestern Germany (DRK-Blutspendedienst Baden-Württemberg, Ulm, Germany), northern Germany (DRK-Blutspendedienst Niedersachsen, Oldenburg) and Tyrol, Austria (Zentralinstitut für Bluttransfusion und Immunologische Abteilung Innsbruck, Innsbruck, Austria). As described previously,8 blood samples in Ulm screened for differential reactivity with a monoclonal IgM anti-D (BS226; Biotest, Dreieich, Germany; not reactive with D^VI) and with polyclonal anti-D in antiglobulin technique were checked for the D^VIphenotype by a panel of monoclonal anti-D (D-Screen; Diagast, Lille, France). The D^VI phenotype was further confirmed by reactivity with monoclonal anti-D BS221, H41 (Biotest), and BRAD-2 (International Blood Group Reference Laboratory, Bristol, UK), as well as absence of reactivity with BS227, BS229, BS231, BS232 (Biotest) and RUM-1 (Bio Products Laboratory, Elstree, UK). Similarly, blood donors in Oldenburg and Innsbruck with weak D phenotype were screened by theRHD exon-specific polymerase chain reaction with sequence-specific primers (PCR-SSP; see below).

DNA was prepared using a modified salting out procedure23or QIAAmp Blood DNA isolation kit (Qiagen, Hilden, Germany). RNA was isolated using RNeasy kit (Qiagen). Reverse transcription was performed with oligo-dT-priming and Moloney murine leukemia virus (MMLV) reverse transcriptase (Sigma, Deisenhofen, Germany). RHD exon-specific PCR-SSP was performed as previously described.24 cDNA was amplified in a nested PCR-reaction (High Fidelity PCR system, Boehringer Mannheim, Mannheim, Germany) with external primers RR1 and RR3 and internal primers Rh5 and Rh7. The 5′ part of intron 1 was amplified with primers RB13 and RB45. Restriction fragment length polymorphism-PCR (RFLP-PCR) of intron 2 was performed using the primers of Poulter et al.25 The 3′ region of intron 3 was amplified with primers RB46 and RB5, RB12 and RI4R2. The intron 3 length polymorphism data were based on seven RhD-negative and 20 RhD-positive samples. Intron 5 was amplified using primers RA9B and Rh2 and intron 6 using primers RB25, RB7, and RB27.

Nucleotide sequencing was performed with a DNA sequencing unit (Prism dye terminator cycle-sequencing kit with AmpliTaq FS DNA polymerase; ABI 373A, Applied Biosystems, Weiterstadt, Germany). We subcloned the PCR product into pMos-T-vectors (pMos-T-kit, United States Biochemicals, Cleveland, OH). Three independent cDNA clones were sequenced using T7 promoter primers, U19 reverse plasmid primers, and internal Rhesus primers. Genomic sequences were established from cloned PCR fragments using both primer walking and nested deletion strategies (Nested deletion kit, Pharmacia, Freiburg, Germany) and verified by sequencing of PCR products using internal Rhesusprimers. Intron 1 sequences were based on six RhD-negative (three ccee, two ccEE, one CCee), eleven RhD-positive samples and one sample of each D^VI type, intron 3 sequences on six RHCE (two ce, Ce, cE each) and four RHD alleles, intron 5 and 6 sequences on at least two RHCE (ce) and RHD alleles. Intron DNA sequences were analyzed for the presence of repetitive sequences by the CENSOR program26 (censor@charon.lpi.org).

RB13, ctagagccaaacccacatctcctt (promoter,27 position -675 to -653 relative to the A of the start codon of the cDNA); RR1, tgttggagagaggggtgatg (5′ untranslated, -60 to -41); Rh5,28 gcacagagacggacacag (5′ untranslated, -19 to -2); Rh1,29 tatctagagacggacacaggATGAGC (5′ untranslated to exon 1, -17 to 6); RB45, acactgttgrctgaatttcggtgc (intron 1, antisense); RA21,25 gtgccacttgacttgggact (intron 2, sense); RA22,25 gtggacccaatgcctctg (intron 2, antisense); RB46, tggcaagaacctggaccttgacttt (intron 3, sense); RA9B, GGTGCCTGCCAAAGCCTCTACCC (exon 4, 554 to 576); RB5, GGCAGACAAACTGGGTATCGTTGC (exon 4, 627 to 604); RB12, tcctgaacctgctctgtgaagtgc (intron 4, antisense, RHD-specific); RI4R2, ttggctcactgcaacctccaccac (intron 4, antisense,RHCE-specific); RB25, agcagggaggatgttacag (intron 4, sense); Rh2,29 AGAAGGGATCAGGTGACACG (exon 5, 900 to 881); RB7, ATCTCTCCAAGCAGACCCAGCAAGC (exon 7, 1022 − 998); RB27, AGCCCAgtgacccacatg (exon7/intron 7); Rh7, acgtacaaatgcaggcaa (3′ untranslated, 1330 − 1313); RR3, cagtctgttgtttaccagatg (3′ untranslated, 1512 − 1431, RHD-specific).

Monoclonal anti-D were provided by the Workshop on Monoclonal Antibodies against Human Red Blood Cells and Related Antigens.30 All monoclonal anti-D were tested for agglutination in a gel matrix test (LISS-Coombs 37°C, DiaMed-ID Micro Typing System, DiaMed, Cressier sur Morat, Switzerland). As detailed in the Results, positive reactivities were obtained with BIRMA-D6; BTSN6; BTSN10; HIRO-3; HIRO-4; HIRO-7; HIRO-8; H41; LHM76/55; LHM76/59; LHM-77/64; LOR11-2D9; LOR17-6C7; LOR29-F7; MCAD-6; MS26; NAU3-2E8; NAU6-4D5; P3G6; P3x21223B10; P3x290; 822; negative with AUB-2F7/Fiss; BIRMA-D56; BRAD-3; BRAD-5; BS229; BS231; BS232; B9A4B2; CAZ7-4C5; CLAS1-126; C205-29; D6D02; D10; D89/47; D90/12; D90/17; F5S; HeM-92; HG/92; HIRO-1; HIRO-2; HIRO-6; HM10; HS114; H2D5D2F5; ID6-H8; LHM50/2B; LHM50/3.5; LHM59/19; LHM59/20; LHM59/25; LHM70/45; LHM-76/58; LHM169/80; LHM 169/81; LHM174/102; LORA; LOR12-E2; LOR17-8D3; LOR28-7E6; LOR28-21D3; L87.1G7; MAR-1F8; MS201; NaTH28-3C11; NaTH53-2A7; NaTH87-4A5; NAU6-1G6; NOI; NOU; P3AF6; P3F17; P3F20; P3x35; P3x61; P3187; RAB.B15; RUM-1; SALSA-12; SAL17-4E8; SAL20-12D5; T3D2F7; VOL-3F6; ZIG-189; 17010C9; 175-2; 819; and weak positive or variable with BIRMA-DG3; BTSN4; D90/7; LORE. Furthermore, reactivity with two polyclonal anti-D produced by carriers of the D^VI phenotype (CcD^VIee and ccD^VIEe), as well as with anti-BARC (ISBT 004.052; RH52) serum and eluate (kindly provided by Drs Geoff Daniels and Carole A. Green, Bristol, UK) was checked.

Determination of RhD antigen density was performed by indirect immune fluorescence as described previously.31 32 All 22 IgG monoclonal anti-D reactive to RhD epitopes present in D^{VI 30} were used (BIRMA-D6, BTSN6, BTSN10, HIRO-3, HIRO-4, HIRO-7, HIRO-8, H41, H41.11B7, LHM76/55, LHM76/59, LHM 77/64, LOR11-2D9, LOR17-6C7, LOR29-F7, MCAD-6, MS26, NAU3-2E8, NAU6-4D5, P3G6, P3x290, and 822). The secondary antibody was goat antihuman immunoglobulins, fluorescein isothiocyanate (FITC)-conjugated (Sigma).

All blood samples were stored on fluid nitrogen. The fluorescence was compared with that of a standard CDe/cde red blood cell (13,000 RhD antigens per cell). Background fluorescence was determined with RhD-negative samples. The number of RhD epitopes detected on the sample cells was calculated as [median fluorescence of sample—background fluorescence] / [median fluorescence of standard cell—background fluorescence] × RhD antigen density of standard cell.

Markers were set to count all red blood cells, even if a fraction of red blood cells appeared unstained. To account for a log-normal distribution, we based the parametric statistical analysis on the logarithms of the RhD antigen densities.

Twenty-six D^VI samples were examined usingRHD-specific PCR-SSP for exons 2, 3, 4, 5, 6, 7, 9, and 10 (3′ untranslated).24 All D^VI samples differed in their PCR-SSP pattern from the wild-type RHD allele and showed one of three PCR-SSP patterns (Fig 1). Two PCR-SSP patterns were compatible with the previously described genomic rearrangements associated with D^VItype I (lack of RHDexons 4 and 5) and D^VItype II (lack of RHDexons 4 to 6).18,19 33 A third pattern could not be explained by any known RHD/RHCE variation and is hereby calledD^VItype III.

Because the PCR-SSP pattern of D^VI type IIIwas novel, we determined the full-length coding sequence of its cDNA (EMBL/GenBank/DDBJ nucleotide sequence database accession numberZ97026). The D^VI type III cDNA comprising all 10Rhesus gene exons represented a RHD-CE-D cDNA, in which the complete exons 3, 4, 5, and 6 of the RHD gene were replaced by the corresponding exons of the RHCE gene.

We applied a PCR-RFLP method for the characterization of theRhesus genes' intron 2.25 A length polymorphism discriminates between the RHC and RHc/RHD alleles of the two Rhesus genes (Fig 2A). An RFLP allows the further separation of the RHC, RHc andRHD alleles (Fig 2B). We excluded the presence ofRHD-specific sequences in D^VI type III at the position of this polymorphism in intron 2 (Fig 2B). The discrimination between an RHC- or RHc-origin of theD^VI type III intron 2 was achieved by the length polymorphism. The D^VI type III sample showed an enhanced band of 1,177 bp size (RHC) compared with that of 1,068 bp size (RHc) (Fig 2A). This indicated that two copies ofRHC-like intron 2 sequences were present in theCD^VIe/ce genotype, one from theD^VI type III allele, the other from the Ceallele of the CD^VIe haplotype. We concluded that the RHCE-derived genomic sequences of the D^VItype III allele originated from the RHCe allele and extended 5′ of this polymorphism, which is located in the middle of intron 2.

The guanosine at nucleotide position 48 relative to the A of the translation start codon in the D^VI type III cDNA was compatible with both an RHD or an RHcorigin.3 28 To prove the RHD derivation of exon 1, we characterized the 5′ portion of intron 1 for bothRhesus genes (EMBL/GenBank/DDBJ nucleotide sequence database accession number Z97362 and Z97363). D^VI type IIIpresented all three nucleotide substitutions and the insertion characteristic for the RHD allele (Fig 3). This observation indicated that the genomic sequences of the D^VI type III allele 5′ of this part of intron 1 were derived from the RHDgene. The molecular characteristics of D^VI type III were summarized and compared with other published alleles (Fig 4).

To define the 3′ limits of the conversion regions of the threeD^VI types, we established the complete nucleotide sequence ranging from exon 5 to exon 7 including both introns 5 and 6. We found the breakpoint of D^VI typeI to be located at the border of intron 5 and exon 6 in a nucleotide range of 215 bp between -100 bp and +115 bp relative to the first nucleotide of exon 6 (Fig 5). The breakpoint ofD^VI type III was located in intron 6 between +360 bp and +963 bp (range of 603 bp) relative to the first nucleotide of intron 6. Finally, the breakpoint of D^VI type II³⁴ was also located in intron 6 between +1,781 and +1,821 bp (range of 40 bp) relative to the first nucleotide of intron 6.

We found a 288-bp deletion in intron 3 of the RHD gene, when compared with the RHCE gene. Based on this deletion, a PCR typing method for RHD was devised. In this intron 3 PCR,D^VI type I and D^VI type IIIsamples reacted like RHD negative controls, whileD^VI type II samples displayed the shorter,RHD-specific band (Fig 6A). This indicated that the conversion point of D^VI type Ihad to be 5′ of the intron 3 deletion. We confirmed the conversion point of D^VI type II adjacent to an Alu repeat34 (data not shown, see Z97030 and Z97031) and 5′ of the conversion point were two additional Alu repeats with inverse orientation, one of which was partly deleted in RHD(Fig 6B).

We observed the three D^VI types associated with specific Rhesus haplotypes: all D^VI type Isamples (n = 14) were found in the cD^VIE haplotype, all D^VI type II (n = 9), and D^VItype III (n = 3) in the CD^VIe haplotype. Because the genomic structure of D^VI type III isD-Ce(3-6)-D, a conversion event among the two Rhesusgenes in cis-position may be the cause of this hybrid allele.

The distribution of the different D^VI types varied depending on the regional origin of the samples (Table 1). In Tyrol (Austria), all samples were D^VI type I, while in southwestern Germany,D^VI type I and D^VI type II were observed about equally frequently. In northern Germany, the only D^VI samples that we found so far were D^VItype II.

One sample of each D^VI type was tested with two polyclonal anti-D and anti-BARC (Table 2).D^VI type III qualified as a D category VI, because it was nonreactive with anti-D produced by probands ofD^VI type I and D^VI type II. Further, D^VI type III carried the BARC antigen (ISBT 004.052; RH52). Anti-BARC did not differentiateD^VI type II and D^VI type III.

One sample of each D^VI type was tested with the full panel of monoclonal anti-D provided in the recent Workshop on Monoclonal Antibodies against Human Red Blood Cells and Related Antigens.35 The three D^VI types did not differ in their reaction pattern (Table 3, upper panel). All positive and most negative reactivities reported by the Workshop coordinator30 were confirmed. Four anti-D (BIRMA-DG3; BTSN4; D90/7; LORE), reported to be nonreactive,30 showed discrepant results and were tested with additional D^VI samples (Table 3, lower panel). We found variable, ie, negative or weak positive, reactivity. This reactivity would have been considered negative by the Workshop criteria30 and thus our observations were in full agreement with the Workshop results.

Fifteen D^VI samples and three control samples were tested with the 22 IgG monoclonal anti-D of the Workshop30 that bind the RhD epitopes of D category VI. In contrast to the control samples, the number of RhD epitopes per cell detected on theD^VIsamples varied considerably depending on the monoclonal antibody used (Fig 7). This variation in the number of epitopes detected did not correlate with the epitope specificity30 of the anti-D (data not shown,P = .23 in the analysis of variance). D^VItypeI and D^VItype II presented consistently low numbers of RhD epitopes per cell with all anti-D. Interestingly, many monoclonal anti-D detected normal, if not enhanced, numbers of RhD epitopes per cell in D^VItype III.

Using the results of all 22 anti-D, we calculated the RhD antigen densities as correlates of the number of RhD proteins accessible on the red blood cells' surface (Table 4). The RhD antigen density of D^VI type III was similar to the CcDee control and several fold higher than that ofD^VI typeI and D^VI type II. Still, the RhD antigen densities of D^VItypeI and D^VI type II differed significantly.

Two of four monoclonal anti-D that are binding to epD37³⁰detected fair numbers of RhD epitopes per cell for D^VItype I, but rather low numbers for D^VItype IIand D^VItype III. This deviation from the actual RhD antigen densities (Table 4) was neither observed with the two other monoclonal anti-D binding to epD37 nor any other anti-D binding to the remaining RhD epitopes present in D^VIsamples. This heterogeneity of anti-D's binding to epD37 may represent a flow cytometric split: epD37a (BTSN10 and HIRO-3) was detected equally well in all D^VI types, epD37b (MCAD-6 and 822) was reduced in D^VI type II and D^VI type III. The binding characteristic of MCAD-6 was used to discriminate the threeD^VItypes by immunohematologic methods, which also allowed separation from normal controls (Fig 8).

The population-based study showed that the variability of the D^VI phenotype is greater than previously reported for the underlying molecular structures18-20 and the RhD antigen densities.15,36-39 We characterized a D-Ce(3-6)-Dhybrid allele of the RHD gene. In accordance with the previous nomenclature,18 this new allele was dubbedD^VItype III. Its D^VI type III phenotype is associated with an almost normal number of RhD proteins accessible on the red blood cells' surface. All three D^VItypes and RhD controls showed distinct immunohematologic features in flow cytometry. The distribution of the D^VItypesvaried significantly even within German-speaking populations.

The observation of a D-Ce(3-6)-D hybrid allele, which represented a D^VI phenotype, contributed considerably to the understanding of the immunoreactivity in partial D. The D^VI phenotype is caused by several different genotypes that are strictly associated with specific Rhesus haplotypes. As a common feature, all known D^VIgenotypes sharedRHCE exons 4 and 5 and RHD exon 7. Substitutions ofRHD exon 4 or exon 5 alone by the corresponding exon ofRHCE result in different partial D (exon 4: DFR, exon 5: D^Va),40 the additional substitution of exon 7 results in the loss of most41 or all42 RhD immunoreactivity. Our report of a D-Ce(3-6)-D allele proved that in contrast to exon 7, RHD exon 3 is not necessary for a D^VI phenotype. This observation supported the current RhD loop model.43,44 All polymorphic sites of exon 3 and exon 6 are believed to occur in the transmembrane and intracellular portions, and hence, may not be expected to influence RhD immunoreactivity very much. In contrast, the polymorphic amino acids of the extracellular loops 3, 4, and 6 are determined by exons 4, 5, and 7, respectively. In concordance with several recent reports,19,20,45 we were unable to find the “deletion type”¹⁸ that has been proposed for the cDE haplotype of D^VI. However, the ccD^VIee phenotype observed in one individual37likely represented a fourth D category VI genotype (proband lost to follow-up; J.W. Jones, personal communication, 1996). Interestingly, the D-Ce(3-6)-D hybrid protein (D^VItype III) is complementary to the Ce-D(2-6)-Ce hybrid protein. The latter hybrid protein causes some Evans (D··) phenotypes,46 47 encodes several RhD epitopes, and lacks all CcEe antigens.

An unexpected feature of D^VI type III was its almost normal number of RhD proteins per cell. The determination of epitope density profiles in D^VI samples gave unequivocal evidence that the lack of certain RhD epitopes need not correlate with the loss of RhD proteins per cell. The observation of the D^VI type III phenotype provided a formal proof that the limited RhD immunoreactivity detected with polyclonal anti-Ds in D^VI type I and D^VI type II³⁷ cannot be explained by the lack of these RhD epitopes only, but must be due to a reduced number of RhD proteins accessible on the red blood cells' surface. The Rhesus protein conformation is likely to influence its red blood cell membrane integration. However, there is currently no conclusive model to predict the effect of any substitution on RhD protein expression. This is exemplified by the different D^VItypes showing a paradoxical, inverse correlation between the size of the substituted protein segments and the RhD antigen density. Substitution of exon 3⁴⁸ increases RhD antigen density37 in D^IIIc, also. Furthermore, single residue substitutions as occurring in D^VII49,50 and DNU⁵¹ may have considerable effects on RhD antigen density.32 37 

It is intriguing to note that all three D^VI typesmay be explained by gene conversion events occurring among bothRhesus genes in cis position. The molecular structures of mostRhesus hybrids (D^IIIb,52D^Va,40 hybrid-VS,42DBT⁵³) are also compatible with this proposed mechanism. Only one RHD hybrid characterized so far (Rh D-E variant ISBT4954) seemed to be caused by a gene conversion in trans position. The impression that conversions in trans position were predominant in RHCE hybrids (R^N,55R₀^Har,56 and r^G57) is likely due to an observation bias because RHCE hybrids will almost exclusively be detected in RhD negative samples.

We referred to D^VI type III as aD-Ce(3-6)-D hybrid, but a D-Ce(2-6)-D hybrid could not formally be excluded. The approximately 4,500-bp region encompassing exon 2 was reported to be identical between the RHD andRHCe alleles and to contain many repetitive elements.58 The 5′ conversion point ofD^VI type III resided in the region between the polymorphisms in intron 1 and intron 2 (Figs 2 and 3). A further characterization did not seem worthwhile because of the long stretch of identical sequences and repetitive elements in that region. The 5′ conversion points of several independent gene conversion events with substitutions in the RHCe allele by RHDsequences in D−− probands were shown by Kemp et al58 to occur also in this stretch of identical sequences. It is tempting to speculate that the sequence identity over more than 4,000 bp including many repetitive elements facilitated conversion events. A similar accumulation of repetitive Alu and LINE elements (Fig 3) occurred adjacent to the breakpoint region of D^VI type II in intron 3, which hosted the conversion points of fourRHD/RHCE hybrids.34 

Characterization of the 3′ breakpoint regions of the threeD^VI types (Fig 5) showed that their breakpoints were not clustered. The breakpoint of D^VI typeI occurred in a stretch of 195 bp covering parts of intron 5 and exon 6, that of D^VI type III in a stretch of identity between RHD and RHCE over 605 bp including an Alu repeat. The extent of the whole gene conversion sequence thus varied between about 4,800 bp (D^VI type II) and > 19,500 bp (D^VI type III). The breakpoint region ofD^VI type II in intron 6 was identical to that recently described by Matassi et al.34 These findings are compatible with a common origin (identity by descent) of allD^VI type II samples described so far in France, the Netherlands, and Germany.

Our quantitative RhD epitope analysis of molecularly characterized samples clarified several previously controversial issues of D^VI immunohematology. First, the use of epitope density profiles addressed the problem of variable antibody affinities in partial D. Studies based on single or few monoclonal antibodies15,36 were likely to underestimate the true number of RhD proteins accessible on the red blood cells' surface, in particular, if the anti-D used36 happened to lack affinity for D^VI.12 Even with conditions believed to be saturating, variable epitope densities were obtained with different anti-D.37 We established epitope density profiles using a panel of monoclonal IgG directed to different RhD epitopes present in the partial D tested.32 Such epitope density profiles in D^VI showed the variability of anti-D affinities. In difference to other partial D like D^VII and DNU,32 there was no narrow antigen density peak (Fig 7), and therefore, the median of the results of all antibodies was used. This robust approach may slightly underestimate the RhD antigen density as correlate of the true number of RhD proteins accessible on the red blood cells' surface because antibodies of marginal affinities to D^VI were not excluded. However, our results forD^VItypeI and D^VItype II were in good agreement with previous reports.37 

Second, the immunohematologic features were correlated with molecular structures instead of serologic haplotypes. Previously, the influence of the molecular structures were not checked. Controversial results indicating low36-38 or variable15,39 RhD antigen densities may simply reflect the absence or presence ofD^VI type III samples in the CcD^VIee group tested. The close linkage of Rhesus haplotype and molecular structure also explains the observation37 that the presence of C suppresses RhD antigen density in normal RhD samples,59,60 but not in D^VIsamples.37 In D^VI, the slight suppressive effect of antigen C was overwhelmed by the effects of the molecular structures as the principal determinants of RhD protein expression.

Third, the quantitative analysis by flow cytometry separated overall RhD antigen density caused by variations in the number of RhD proteins accessible on the red blood cells' surface from variable expression of certain RhD epitopes. Flow cytometry allowed differentiation of theD^VI types from one another and from normal RhD. Furthermore, we could demonstrate a flow cytometric split of RhD epitope epD37. However, the only qualitative serologic difference that we could correlate with the molecular structures was a paucity, but not lack, of epD37b on D^VI type II andD^VI type III. We suspect that some previously reported serologic splits8,16,18,21,61,62 that were mainly observed with weak overall antibody reactivity8,16,18,62may be due to quantitative differences in RhD epitope expression rather than lack of certain RhD epitopes. We propose that a meaningful report of a serologic split in partial D should exclude the confounding effect of low antigen densities. This exclusion may be achieved by inverse reaction patterns of different monoclonal antibodies,12 by quantitative methods like flow cytometry,31,32 and enzyme-linked immunosorbent assay (ELISA)37 or by the demonstration of different underlying molecular structures.

Our findings have several practical implications for RhD phenotyping and RHD genotyping. Comprehensive RHD genotyping is a complex task24 because of many Rhesus hybrid genes18,19,33,40,42,52,53,55-57,63 and of RhD-negative phenotypes still harboring RHD-specific sequences.6,42,64,65,D^VI type III adds to this complexity. It would type RHD negative in a standard intron 2-based PCR method25 previously believed to type D^VI samples reliably as RHD positive. The population study (Table 1) provided further evidence for the allelic variation between closely related populations, which influences the specificity and sensitivity of Rhesus genotyping. An absolute match of phenotype and genotype is unlikely to be achieved by current technology because sporadic nonfunctional alleles occur rather frequently in genes66 includingRhesus.67 Hence, the expense of a genotyping system must be weighed against its residual failure rate in phenotype prediction. D^VI is the clinically most importantRHD variant and it might be advantageous to dissociate this variant both from RHD positive and RHD negative. To this end, a simple system testing intron 4 and exon 7 may suffice because any D category VI genotype is likely to lack bothRHD exon 4 and 5 and to retain RHD exon 7.

D^VI recipients should be transfused with RhD negative blood to limit anti-D immunization,17 a rationale that prompted RhD negative transfusion in patients carrying weak D. This essentially RhD antigen density-based transfusion strategy is today considered wasteful, as it became apparent that most weak D patients may be safely transfused RhD positive. The wastage might be reduced by lowering of the weak D threshold for RhD negative transfusion. However, this measure would trigger RhD positive transfusion in partial D likeD^VItype III, while still many RhD negative units would be transfused to weak D patients not requiring RhD negative transfusion. In this context, a strategy based on two monoclonal anti-D that do not react with D^VI is advantageous.8,17This RhD epitope-based transfusion strategy abandons RhD antigen density as the trigger for RhD negative transfusions and became mandatory in Germany in 1996.68 It should be advocated in all regions where D^VI is the single clinically important partial D. For donor typing, weak D is considered Rhesus positive.69 D^VItype III proved that D^VI erythrocytes may carry rather high RhD antigen densities. The threshold of RhD antigen density and the RhD epitopes that most likely cause anti-D immunization are not fully established. We think the transfusion of D^VI red blood cells should be restricted to RhD positive individuals, until further evidence for lack of immunogenicity is established.

We thank Drs Hans-Hermann Sonneborn and Manfred Ernst, Dreieich, Germany, for generously supplying us with their monoclonal anti-Ds; Drs Geoff Daniels and Carole A. Green, Bristol, England, for providing anti-BARC serum and eluate; Dr Jeff W. Jones, Liverpool, England, for the determination of absolute RhD antigen density on red blood cell samples that were used as standards; Elisabeth Hörner, Olga Zarupski, and Esther Rainer for expert technical assistance; and Bryan Hillesheim for preparing red blood cell and DNA samples.