Background

The ABO blood group system is the most clinically significant system in transfusion and transplantation medicine. The A and B genes are co-dominant, so individuals can be phenotyped as A, B, AB or O. A common dimorphism in some populations leads to the division of blood group A into A1 and A2, the latter showing weaker antigenicity. On occasion, A2 activity is weakened further when competition due to a co-dominant B gene occurs (A2B phenotype).

ABO allele nomenclature poses significant problems that are still under consideration by the International Society of Blood Transfusion (ISBT). In the absence of an officially agreed terminology, alleles are referred to here by their serological activity and an alternative allele name1 is given in square brackets.

The ABO gene contains seven exons and a 1062 base pair (bp) sequence codes for a glycosyltransferase, of which the A1 [A101] allele (usually described as the consensus allele) product adds a monosaccharide (N-acetyl-α-D-galactosamine) to a specific acceptor glycoconjugate. The B [B101] allele produces a similar protein, differing only in four amino acids, but these changes result in a modified enzyme specificity [2]. The same acceptor glycoconjugate is utilised, but the B-specific monosaccharide, α-D-galactose, is added (for a review, see [3]). The coding region of the A2 [A201] allele differs from A1 by a seemingly innocuous 467C>T (Pro156Leu) substitution and a 1061delC that causes a frame-shift and extends the reading frame by 64 nucleotides [4]. Many other mutations have been described [5] that weaken these activities and result in weak A or B subgroups. Totally inactivating mutations result in the blood group O phenotype, the most common of which is a deletion at nucleotide (nt.) 261 (261delG) in exon 6 that results in a frame-shift and premature termination of translation [2]. Two major alleles of this type exist. The first allele (O1 [O01]) differs from the consensus A [A101] allele in all seven exons by only this mutation. The second allele (O1v[O02]) [6, 7] has nine further mutations spread across exons 3–7 in addition to 261delG and an additional 13 mutations have been found amongst the 1052 nucleotides in intron 6 [8, 9].

Other alleles have been described that are predicted to lead to amino acid substitutions or frame-shifts [5], or alternative splicing [10, 11] and a dimorphic enhancer region has been reported [12–14]. Several alleles at the ABO locus appear to arise by crossing over between two dissimilar alleles. First described in 1996 [15], a B allele was continued as an O1vallele from downstream of the middle of the last and largest exon (exon 7). The B-enzyme-determining mutations occur after the cross-over point and the O1vsequence after this point only differs from the consensus (A1) sequence by two point mutations, of which only one would lead to an amino acid change (Val277Met). This change presumably weakened the A enzyme activity, as the resulting phenotype was A2, albeit from a B-O1v hybrid allele [A204]. Other examples on the same theme include several Axalleles based on the 5'-ends of A or B alleles in combination with the 3'-end of the O1vallele [9, 16]. However, many of the hybrid alleles described so far are hybrids of O1 and O1v, have 261delG, and hence are O (i.e. inactive) alleles [17].

We describe here a new hybrid A allele having O1v-A2 allele characteristics in five individuals of AweakB phenotype from different parts of the world, but having a common ethnic background. This is the first allele described in which the effect on enzyme activity of the mutations in the O1vallele prior to exon 6 can be shown.

Results

Blood group serology

The five individuals from Portugal (origin: Guinea Bissau), Sweden (n = 2, origins: Senegal and Zimbabwe), Switzerland (origin: the Dominican Republic), and the USA were all Black and of African or Afro-American descent. On routine typing they were found to express the A antigen weakly on their red blood cells (RBC) whilst the B antigen was expressed normally. Fresh blood samples from all five subjects were subjected to an extended serological analysis at the referring transfusion centres. Three of the samples were analysed with a panel of commercially available polyclonal and monoclonal anti-A reagents (described previously in [18]) for further characterisation of the weak A phenotype. One monoclonal reagent (Seraclone) agglutinated the RBC almost completely (3+ reaction with few unagglutinated cells) but the other reagents gave mixed field reactions (macroscopic reading ranging from negative to 1+/2+). All reactions were also read microscopically and weak agglutination with mixed field was verified also in the reagents yielding macroscopically negative readings. Adsorption of red cells with anti-A and subsequent elution yielded an eluate containing anti-A (only performed in one case). Anti-A1 was present in serum from all subjects but one also had a weak anti-A reactive only after incubation at 4°C. Saliva testing was not performed due to lack of saliva samples and/or non-secretor status as judged by Lewis blood group phenotyping.

Blood samples from the parents of the Portuguese individual were also investigated. Whilst the father had the common B phenotype, the phenotype of the mother was Aweak but with a strength of reactions suggesting an intermediate form of A antigen expression, weaker than A2 but stronger than the one observed in the AweakB cases studied here.

Blood group genotyping

All samples were found on initial genotype screening [19] to have the A2B genotype and lacked the majority of previously reported mutations associated with weak A or B expression [16, 20]. Further analysis by A2-specific PCR-ASP [18] established the presence of the 1061delC mutation [4] and B-specific PCR-ASP [16] indicated the presence of a normal B allele [21]. In addition, enhancer mini-satellite PCR analysis indicated that four of these samples were homozygous for four 43 bp-repeats in the CBF-NF/Y-binding domain approximately 4 kbp upstream from the translation start codon, as expected [13]. Interestingly, the individual originating from the Dominican Republic was heterozygous for one and four repeats, thus deviating from the rule [13].

A BstU I-based PCR-RFLP test [7] results in cleavage between 188C and 189G in exon 4 of all known alleles except O1v(and some minor variants of O1 and O1v[5]), which has two mutations at these positions (188G>A and 189C>T). The indication that these five individuals already genotyped as A2B were also heterozygous for an O1vallele-specific mutation led us to sequence exons 1–7 to investigate the reason for this anomaly.

Based on information obtained about the geographic and ethnic origin of the individuals with the anomalous A allele, we also performed genomic typing of the FY and RHD blood group loci according to published methods [22, 23]. The silent Fy allele (based on the Fybsequence with a disrupted GATA-1-binding motif in the promoter region [24]) commonly found in individuals of Black ethnic origin but not in Caucasians was detected in the homozygous or heterozygous state in all five cases whilst the RHD pseudogene [23] exclusively found in some Blacks was detected in the Portuguese sample only.

DNA was also isolated from blood from the parents of the Portuguese AweakB individual. The mother was genotyped as A2O1vand the father as BO1. Further investigation of the mother's A2-like allele showed the same hybrid allele as her son and the other four index cases, as expected.

ABO exon sequences

The complete coding region (all seven exons, comprising 1128 bp) of the A2 allele was sequenced in all five samples.

The nucleotide sequence in exons 6 and 7 was identical to the non-Asian A1 consensus allele except for the characteristic mutations (467C>T and 1061delC in exon 7) found in the common A2 allele.

The sequences in exons 1–5 were similar to the consensus allele, except for 106G>T (exon 3), 188G>A (exon 4) and 220C>T (exon 5), all consistent with the presence of an O1vallele. Surprisingly, the O1v-specific mutation, 189C>T, was absent, whereas a hitherto undescribed mutation in exon 2 (46G>A) was found in all five samples. As opposed to the situation in O1v, the effect of these missense mutations can be seen in the translated transferase, as the O1/O1v-specific 261delG mutation is not present in this allele. The amino acids predicted to occur as well as their relative locations in the translated protein are shown in Figure 1.

Figure 1
figure 1

Depiction of the relative locations of amino acid substitutions caused by the novel allele. A schematic representation of the A transferase resulting from transcription and translation of the novel allele is shown on the left side whilst a computed 3D-model on black background is displayed on the right side. In the former, all amino acid substitutions resulting from the four missense mutations in the O1v-A2 hybrid are shown as coloured circles highlighted by dotted arrows. In the latter, only two of the mutated positions could be displayed in the model since the two N-terminal substitutions were not included in the expressed soluble portion of the crystallized A transferase [32] on which the template structure (molecular coordinate file 1LZ0 in the NCBI structure database) was based. The 3D surface model was created with the DeepView Swiss Pdb Viewer version 3.7, an interactive molecular graphics programme for viewing and analyzing protein structure (http://www.expasy.org/spdbv and [33]) using the molecular surface mode and the 3D rendering display option. No effort was made to calculate the possible structural alterations caused by the amino acid changes. The original sequence expressed for the crystallographic study [32] was used for the analysis. Residue positions 63 and 74 are highlighted in red and green, respectively, to show their relative locations in the model. In addition, the DXD motif (present in almost all glycosyltransferases, here as DVD capable of binding the UDP part of the nucleotide-sugar substrate) is shown in blue in order to highlight the catalytic cleft. The different regions of the glycosyltransferase are shown according to Paulson and Colley [34] and the transmembrane domain (amino acids 17–37) is based on the hydrophobicity plot and amino acid composition originally reported by Yamamoto et al. [35]. The black X represents the approximate location of a proteolytic cleavage site for generation of the soluble glycosyltransferase found in body fluids.

Full size image

A possible hybrid allele breakpoint was suspected somewhere between the end of exon 5 and the beginning of exon 6. We therefore sequenced the intervening intron 5 (see below) in an attempt to localise the cross-over region.

None of the more than 100 samples screened by PCR-ASP (i.e. at least 200 alleles, comprising over 50 O1 and O1valleles and at least 10 A1, A2, B, O2 and O hybrid alleles, as well as several weak A alleles) had the new mutation, 46G>A (Ala16Thr). Amongst all these samples one O1 allele from a Jordanian individual was found to have the O1v-characteristic 188G>A (Arg63His) mutation. One O1vallele from an African individual lacked the O1v-characteristic 189C>T (silent) mutation.

ABO intron sequences

Intron 5

Intron 5 from a number of individuals with the common alleles, A1 (two homozygotes), A2 (two homozygotes), B (one homozygote, one heterozygote), O1 (two homozygotes), O1v(two homozygotes),O2 [O03] (three heterozygotes) and the new suspected hybrid alleles was sequenced. Three A2 and three O1valleles from Black African donors were also sequenced for comparison. Eight polymorphic nucleotide positions were found in the 554 bp sequence and these are shown in Table 1. Intron 5 sequences in the Black African samples were identical to the respective alleles of the other (Caucasian) control material. The sequence in the new allele was consistent with a crossover between an O1vand an A2 allele after nt. 103insCCC but before 306C in intron 5. This is shown schematically in Figure 2.

Table 1 Polymorphic nucleotide positions in introns 2–5 of the ABO gene.
Full size table
Figure 2
figure 2

Representation of the exons and introns in A2 and O1valleles and the new hybrid allele. Exons are represented as thick boxes and introns as thin boxes. They are drawn approximately to scale except for the approx 13 kbp long intron 1. Filled regions indicate an A2 sequence and unfilled regions an O1vsequence. The hatched region in the hybrid allele indicates the cross-over region. The grey area is presumed to have an O1vallele sequence, but lack of known mutations in exons 1 and 2 and the unsequenced intron 1 does not allow proof of this. The asterisks indicate the localisation of the four missense mutations compared to the common A2 allele.

Full size image

Introns 2, 3 and 4

Since the new allele has a novel mutation in exon 2 (46G>A) and an unexpected lack of the O1v-specific mutation in exon 4 (189C>T), other introns in the A2-like allele of the five AweakB samples were also sequenced and all gave identical results. Samples from the same 13 Caucasian and three Black individuals examined for intron 5 above were also analysed (Table 1). Intron 1 was not examined due to its size (approx. 13,000 bp). We found 7, 12 and 17 polymorphic sites (point mutations, deletions and insertions) in introns 2, 3 and 4, respectively, in the common alleles (Table 1).

The intron 2–4 sequences of the novel allele were identical to the O1valleles sequenced except for three adjacent mutations in intron 3. The first, 205T>C, caused a reversion to the consensus from O1v-specific, whereas 354G>A and 399G>A were mutations not previously encountered in any allele. Intron 3 in the O1valleles from the control donors of African origin was sequenced and in each case was identical to Caucasian O1valleles.

Intron 6

Allele-specific variations in intron 6 have already been described [8, 9]. Four new mutations were found in the new hybrid allele when compared to the A consensus sequence (277A>G, 286C>T, 911G>T and 952A>G). However, intron 6 in three normal A2 alleles from Black Africans with the common A2 phenotype also had these mutations.

Discussion

Detailed analysis of the alleles at the blood group ABO locus is shedding light on the effect of polymorphism in different regions of the translated products, the blood group A and B glycosyltransferases, and ultimately the clinically important ABO phenotype of red cells. Factors with the potential to influence the glycosyltransferase activities include base insertions, deletions and substitutions mainly in exons 6 and 7 (for review see [5, 25]), hybrid alleles [17], splice-site mutations [10, 11], variations in enhancer activity [12–14], promoter methylation [26], promoter mutations [27] and alternative promoter regions [28].

The new allele described here is unusual in several respects. The O1v-characteristic mutations, 106C>T, 188G>A and 220C>T in exons 3, 4 and 5, respectively, in combination with the common A2-specific sequence in exon 7 suggested that the allele is an O1v-A2 [O02-A201] hybrid having a crossing-over point after nt. 220 in exon 5 and before nt. 261 in exon 6. A deletion at nt. 261 is the most common inactivating event creating O alleles and is present in both O1 and O1valleles and hybrid variants of these as shown in Figure 3. This mutation is absent from the five hybrid alleles described here and hence the cross-over should occur upstream of this position. In an attempt to determine the cross-over point more precisely and confirm the identity of the two contributing alleles we sequenced the intervening intron (intron 5) in the hope that allele-specific mutations were present in this intron, by analogy with our previous findings in intron 6 [9]. As Table 1 shows, eight of the 554 nucleotides in the intron were polymorphic. In this intron the A1, A2, B and O1 alleles are very similar, differing only at nt. 336 in the O1 allele and at nt. 529 in only one of the three B alleles tested. More pronounced differences were observed in the O2 alleles that differed at each of the first three polymorphic positions (the complex variations in the O2 allele will be presented elsewhere), and in the O1valleles, that differed at four of these sites. This latter information allowed us to determine the crossing-over point to occur in intron 5 between nt. 103 and 306, and that the allele is indeed an O1v-A2 hybrid (Figure 2).

Figure 3
figure 3

Known hybrid alleles at the ABO locus. The common alleles are also shown for comparison. Only changes from the consensus (A1-1) sequence are shown. Mutations causing amino acid changes are shown in bold face. The yellow areas indicate a reading frame shift. The blue areas indicate untranslated regions. The introns are represented by the thick, dark vertical bars. The blue rectangles indicate the region of the allele where a crossing over event occurred. ?, indicates that the nucleotide at this position was not described.

Full size image

This new allele showed some additional interesting characteristics. The mutation, nt. 46G>A (Ala16Thr) in exon 2, has not been described in any other context, nor could we detect this mutation when we analysed 260 alleles from individuals of diverse ethnic background (of whom about 40 were African). Its occurrence in all five index samples, albeit from individuals of Black African descent, collected from diverse parts of the world is surprising.

Few mutations have been found in exon 4. Both 188G>A and 189C>T have hitherto been exclusive characteristics of the O1vallele; 190 G>A has only been found in some Brazilian Black O1 alleles [29]; and 203G>C occurred in some Scandinavian Aweakalleles [16]. The finding of 188G>A without 189C>T in the new allele was also unexpected. When we screened more than 200 alleles at these positions we found one otherwise normal O1vallele (African) having 188G>A but lacking 189C>T and one otherwise normal O1 allele (Jordanian) having 188G>A. Although infrequent, additional genetic diversity obviously exists in individuals with common ABO phenotypes.

Several hybrid alleles at the ABO locus have been reported previously (reviewed in [17] and summarised in Figure 3). About half of these alleles are O alleles containing the common inactivating deletion at nt. 261 contributed by the O1 or O1vpart of the hybrid. The others described so far are A alleles of varying activity, containing initial A or B sequences followed by O1 or O1vsequences, where the critical nt. 261 is supplied by the A or B allele and is thus not deleted. The decreased A antigen expression is attributed to mutations in exon 7 (which encodes the enzyme's active site) contributed by the O1vallele. This novel Aweakallele described here is the first example of a "reverse" hybrid, where the weakening of enzyme activity from A2 to Aweak may be due to the concerted effect of the mutations in exon 2 at 46G>A (Ala16Thr), exon 3 at 106G>T (Val36Phe), exon 4 at 188G>A (Arg63His) and exon 5 at 220C>T (Pro74Ser). However, theoretically the Aweak phenotype may depend on only one or a combination of some of these four mutations.

Based on the linear structure, the former two substitutions presumably occur in or near the predicted transmembrane domain whilst the latter two localise to the postulated stem region of the Golgi-membrane-anchored enzyme. None of them is thought to occur in the proximity of the active site, which would be expected to have the normal A2 enzyme structure based on the sequence of exon 7. In order to visualize this prediction, we created a three-dimensional computer model of the A transferase for mapping of two of the mutated amino acid positions onto the crystal-structure-based molecular surface. The result appears to confirm that residues 63 and 74 are localised far from the enzymatically active site in a stalk-like part of the protein (Figure 1). It should be emphasized that we did not attempt to determine the effect that these mutations would have on enzymatic function or fine structure of the gene product. Expression studies beyond the scope of this study or the future finding of other samples carrying other combinations of the above mutations may be able to address this issue.

The mutations found early on in intron 3 of this new hybrid allele (the consensus 205C rather than the O1v-specific 205C>T, as well as the unique 354G>A and 399G>A) appear to be a specific characteristic of this new allele since intron 3 of the three control O1valleles from Black African individuals did not differ from Caucasian samples. On the other hand, the four new mutations found in intron 6 of the new hybrid allele were also found in the three Black African A2 alleles, which led us to conclude that the alterations observed in intron 6 may simply reflect a common ancestral A2 allele of African evolutionary lineage.

The A and B glycosyltransferases compete for the same acceptor glycoconjugates. However, the weaker A activity due to this new allele was also observed serologically, although to a lesser degree, in the mother of one of these AweakB individuals who had inherited the hybrid allele in combination with an O allele and hence lacked a B gene. This may indeed be the reason why all five index samples studied here were AweakB. Obviously, the weakening effect of this hybrid is relatively mild so that the Aweak phenotype, produced when the A hybrid glycosyltransferase is allowed to convert the available acceptor glycoconjugates to A without any competition from a B transferase, will sometimes escape detection in routine laboratories, especially if automated blood grouping equipment is used.

Conclusions

A new hybrid, Aweakallele with O1vand A2 characteristics with a crossing over point in intron 5 has been found at the blood group ABO locus in five individuals of diverse Black African backgrounds. To our knowledge this is the only defined ABO subgroup allele so far associated with an African ethnic origin.

Sequencing of all seven exons of these Aweakalleles showed two major exons (exons 6 and 7) were identical to the most frequent A2 allele, although the intervening intron 6 had mutations only found so far in Black Africans. Exons 1–5 had sequences consistent with the O1vallele, except for two novel changes, the most important one of which (46G>A) results in an amino acid substitution in the putative trans-membrane region of the translated protein (glycosyltransferase). The other novel change (lack of 189C>T) led further to the identification of two new O alleles.

This new hybrid allele shows how mutations in early exons, far from the enzyme product's active site, can affect expression of the blood group A antigen on the erythrocyte surface.

Methods

Blood samples and blood group serology

Blood samples from individuals living in Portugal, Sweden, Switzerland and the USA were referred to our laboratory for genomic analysis due to unclear phenotyping. Blood samples from first-degree relatives were only obtained in one case.

One hundred and thirty blood samples available at the Blood Centre, University Hospital, Lund from blood donors and other apparently healthy individuals with mixed phenotypes and mixed ethnicity (Africans, Europeans, Jordanians) were used for screening purposes. Their ABO group was determined according to current practice [30].

Routine ABO genotyping

All oligonucleotide primers used were synthesized by DNA Technology ApS (Aarhus, Denmark). DNA was prepared in Lund using a simple salting-out method [31].

The initial ABO genotyping comprised duplex PCR-RFLP and PCR-ASP analysis of exons 6 and 7 across intron 6 and subsequently DNA sequencing of the ABO gene and its regulatory elements was performed [7, 16, 19, 20].

Selected homozygous and heterozygous DNA (samples homozygous for O2 and B allele were not available) from blood donors and other apparently healthy individuals at the reference laboratory in Lund was used for identification of introns 2 to 5 from the common alleles A1, A2, B, O1, O1vand O2.

PCR amplification of the ABO gene for DNA sequencing

Primers used to amplify DNA fragments and for allele-specific direct sequencing of the seven exons and intron 6 are described elsewhere [16].

Alternatively, polymerase chain reaction (PCR) was carried out using Expand High Fidelity PCR system (Roche Molecular Systems, Pleasanton, CA, USA) to amplify different intron fragments and for screening for mutations 188G>A and 46A>G. Amplifications were performed with primer pairs as shown in Table 2. An internal positive control primer pair was used in each PCR reaction. Amplification was performed in a reaction volume of 20 μl with 0.5 μmol/L of each primer, 2 nmol of each dNTP, and 100 ng of genomic DNA. As thermostable enzyme we used 0,5U from the Expand High Fidelity PCR System in the supplied buffer 2 with a final Mg2+ concentration of 1.5 mM according to the manufacturer (Roche Molecular Systems). After an initial denaturation step at 95°C for 2 min followed 10 cycles of denaturation (94°C for 20 s), annealing (65°C for 45 s) and extension (72°C for 1.5 min), then 25 cycles at 94°C for 20 s, 61°C for 30 s and 72°C for 1 min and a final extension for 5 min.

Table 2 Oligonucleotide primers used for PCR amplification, sequencing and screening in this study.
Full size table

PCR products were excised from 3% agarose gels (Seakem, FMC Bioproducts, Rockland, ME, USA) stained with ethidium bromide (0.56 mg/l gel, Sigma Chemicals, St. Louis, MO, USA) following high-voltage electrophoresis and purified using the Qiaquick gel extraction kit (Qiagen GmbH, Hilden, Germany).

The Big Dye Terminator Cycle Sequencing kit (Applied Biosystems, Foster City, CA, USA) and an ABI PRISM 310 Genetic Analyser (Applied Biosystems) were used for direct DNA sequencing with capillary electrophoresis and automated fluorescence-based detection according to the manufacturer's instructions. Sequence analysis was performed with SeqEd software 1.03 (Applied Biosystems).

Note

1Nomenclature used in the Blood Group Antigen Gene Mutation Database http://www.bioc.aecom.yu.edu/bgmut/abo.htm.