Murine Equivalent of the Human Histo-blood Group ABO Gene Is acis-AB Gene and Encodes a Glycosyltransferase with Both A and B Transferase Activity*

We have cloned murine genomic and complementary DNA that is equivalent to the human ABO gene. The murine gene consists of at least six coding exons and spans at least 11 kilobase pairs. Exon-intron boundaries are similar to those of the human gene. Unlike human A and B genes that encode two distinct glycosyltransferases with different donor nucleotide-sugar specificities, the murine gene is a cis-AB gene that encodes an enzyme with both A and B transferase activities, and thiscis-AB gene prevails in the mouse population. Cloning of the murine AB gene may be helpful in establishing a mouse model system to assess the functionality of the ABO genes in the future.

Histo-blood group A/B antigens are clinically important antigens in blood transfusion and organ transplantation. These antigens are oligosaccharide antigens whose immunodominant structures are defined as GalNAc ␣133 (Fuc ␣132) Gal-and Gal ␣133 (Fuc ␣132) Gal-for A and B antigen, respectively. Functional alleles at the ABO locus encode enzymes that catalyze the final step of synthesis. A alleles encode for A transferase, which transfers the GalNAc residues from the UDP-GalNAc nucleotide-sugar to the galactose residue of the acceptor H substrates defined by Fuc ␣132 Gal-. B alleles encode for B transferase that transfers the galactose residue from UDP-galactose to the same H substrates. O alleles are nonfunctional, null alleles. During the past decade, we have been studying the molecular genetic basis of the histo-blood group ABO system (1). From a human gastric carcinoma cell line cDNA library, we were able to clone human A transferase cDNA (2) based on the partial amino acid sequence of the soluble form of A transferase purified from human lung (3). Using cross-hybridization with A transferase cDNA probes, we then cloned B transferase cDNA and nonfunctional O allelic cDNA from cDNA libraries made with RNA from colon adeno-carcinoma cell lines that exhibited different ABO phenotypes (4). Possible allele-specific mutations were identified. Four amino acid substitutions were discovered between A and B transferases. O alleles were more homologous to A alleles than to B alleles. A single base deletion was found near the N terminus of the coding sequence in most of the O alleles, which caused the codon frame to shift. This resulted in a truncated protein without glycosyltransferase activity. In addition to the three major alleles (A 1 , B, and O), we also identified mutations that modified the enzymatic activity by determination of the partial nucleotide sequences of subgroup alleles (A 2 , A 3 , A x , and B 3 ) (5-7). We also elucidated the molecular mechanisms of two phenomena named cis-AB and B(A) (7,8). Although the incidence was low, another type of O allele was discovered that lacked the single base deletion but contained an amino acid substitution at the residue crucial for nucleotide-sugar recognition/binding (9). Although no functional analyses have been performed to disprove polymorphism, others have reported additional alterations (10 -15). The nucleotide and deduced amino acid sequences of a variety of ABO alleles are posted on the Blood Group Antigen Gene Mutation Database developed by Blumenfeld and colleagues (available on the World Wide Web).
A/B antigens are not restricted to humans but are widely present in nature (16). We therefore investigated the presence/ absence of homologous sequence(s) in the genomes of other species of organisms (17). Hybridization of zoo blots, using the radiolabeled human A transferase cDNA probe, showed weak signals in chicken genomic DNA but strong signals, comparable with the signal detected in human DNA, in genomic DNA from mice and other mammals. No signals were detected in genomic DNA from lower species of organisms in the evolutionary tree. We next determined the partial nucleotide sequences of the primate ABO genes (17). The glycosyltransferases responsible for A or B phenotypes in primates were shown to conserve amino acid substitutions corresponding to codons 266 and 268 in humans. A similar study was also reported by others (18). Through comparative sequence analyses of the ABO genes from humans and apes, we and others proposed a convergent hypothesis of evolution that ABO genes arose from independent mutations after the speciation of humans and apes (19,20). No apparent disadvantages are recognized among any of the phenotypes involving the ABO polymorphism. Hemolytic disease of newborns may be a natural selection against specific combinations of blood groups between the mother and fetus. However, serious incompatibility cases are rare with ABO, since the natural antibodies against A and B antigens are mostly IgM and do not cross the placenta. Although some anti-A, B antibodies are IgG and capable of crossing the pla-* This study was supported in part by funds from The Burnham Institute, the Department of Defense Breast Cancer Research Program (DAMD 17-98-1-8168), and from Kyowa Hakko Kogyo Co. Ltd. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence reported in this paper has been submitted to the DDBJ/GenBank TM  centa, A/B antigens are not well developed in fetuses. Therefore, little damage is done. There should be some reason for the existence of ABO polymorphism in the population. It has been speculated that the possible role of the ABO system is to provide resistance against infection (21). Actually, Le b (Fuc␣132 Gal ␤133 (Fuc ␣134) GlcNAc-), an ABO-related structure, was demonstrated to be the receptor for a Gram-negative bacillus, Helicobacter pylori, a causative agent for gastritis, peptic ulcer, and possibly gastric cancer (22). A and B transferases modify the Le b structure into ALe b and BLe b structures, which H. pylori does not bind to in vitro. This may explain the earlier observation that group O individuals have a higher incidence of stomach ulcer than individuals in any other group (21). Antibodies against the ␣133 Gal epitope (Gal ␣133 Gal-) were demonstrated experimentally to block the interspecies infection of certain retroviruses (23). From this result, anti-A and anti-B antibodies have been suspected to play a role in inhibiting the epidemics of certain infections. Some type of selection based on the advantage/disadvantage of having antigens/antibodies may have been operating at the ABO locus to secure the survival of species from extinction during evolution. The report that an anti-A monoclonal antibody neutralized human immunodeficiency virus particles produced by lymphocytes from group A individuals but not from group B or O individuals (24) may support this hypothesis. To experimentally assess the functionality of the ABO genes, establishing an animal model is critical. As an initial step, we cloned and characterized the murine ABO gene equivalent.

Materials
Mouse genomic DNA library (ML1044j), which was constructed by replacing the internal BamHI-BamHI stuffer fragment of the EMBL3 SP6/T7 vector with MboI-partially cleaved genomic DNA fragments of the BALB/c strain of mouse, was purchased from CLONTECH (Palo Alto, CA). A Marathon cDNA Amplification Kit and an AdvanTage PCR 1 cloning kit were also from CLONTECH. A GeneClean kit was purchased from Bio101 (La Jolla, CA), and pT7T3␣18 plasmid vector, ␣-S-dGTP, ␣-S-dCTP, and S-300 MicroSpin columns were from Amersham Pharmacia Biotech. LipofectAMINE was purchased from Life Technologies, Inc. [ 14 C]UDP-GalNAc, [ 14 C]UDP-galactose, and [␣-32 P]dCTP were from PerkinElmer Life Sciences, and 2Ј-fucosyllactose was from Oxford Glycosystems (Rosedale, NY) and from Calbiochem. Murine anti-A and anti-B monoclonal antibody mixtures were purchased from Ortho Diagnostic Systems (Raritan, NJ). Biotinylated Ulex europaeus agglutinin I, Vectastain Elite ABC kit, and 4-chloro-1naphthol substrate kit were from Vector Laboratories (Burlingame, CA), and EZ-Link Sulfo-NHS-LC-Biotin was purchased from Pierce. Bluescript SKM13ϩ vector, Prime-It II kit, Duralose-UV membranes, and frozen competent XL1-blue strain of Escherichia coli bacteria were purchased from Stratagene (La Jolla, CA). The ULTRAhyb hybridization buffer was obtained from Ambion (Austin, TX). Restriction endonucleases and nucleic acid-modifying enzymes were from Life Technologies, New England Biolabs (Beverly, MA), or Roche Molecular Biochemicals. dRhodamine dye terminator cycle sequencing ready reaction kits and BigDye cycle sequencing kits were purchased from PerkinElmer Life Sciences.

Isolation of Genomic DNA Clones and Construction of Restriction
Endonuclease Cleavage Maps-Under low stringency conditions, ϳ1 million plaques from the mouse genomic DNA library were screened using a human A transferase cDNA probe by the plaque hybridization method (25). Radiolabeled probe was prepared by the random hexamer primer method, using a Prime-It II kit and [␣-32 P]dCTP (26). After four rounds of screening, individual clones were isolated. Phage DNA was prepared, cleaved with restriction endonucleases, gel-electrophoresed, and Southern transferred. Hybridization was then performed to construct restriction enzyme cleavage maps.
Subcloning, Nested Deletion, and DNA Sequencing-DNA from MABO16 phage clone was cleaved with HindIII and SalI and subcloned into the Bluescript SKM13ϩ vector. Nested deletion constructs were prepared by the ExoIII-mung bean nuclease method (27). Where no unique 3Ј-overhang restriction sites were available, the thioderivative fill-in reactions were performed with Klenow enzyme using ␣-S-dGTP and ␣-S-dCTP before ExoIII treatment. After transformation of E. coli XL1-blue strain, plasmid DNA was prepared from individual clones and analyzed for insert size. The nucleotide sequences were determined by Sanger's dideoxy chain termination method using the dRhodamine dye terminator cycle sequencing ready reaction kit (28). Sequences were aligned using Lasergene SeqMan II sequencing project management software.
5Ј-RACE cDNA Cloning-RNA from the CMT-93 rectal carcinoma cell line (ATCC 223-CCL), established from a C57BL strain of mouse, was prepared and used for the 5Ј-RACE experiments (29). We followed the Marathon cDNA amplification protocol provided by the manufacturer. Briefly, the first strand of cDNA was synthesized from RNA using Moloney murine leukemia virus reverse transcriptase and Marathon cDNA synthesis primer. After the second strand was synthesized with RNase H, E. coli DNA polymerase I, and E. coli DNA ligase, the Marathon cDNA adaptor was ligated. Nested PCR was performed, first with AP1 adaptor primer and MY-1 primer and then with AP2 and MY-2 primers. AP1 and AP2 primers were provided in the kit. The nucleotide sequences of MY-1 and MY-2 primers were complementary to the sequences in the coding region of the murine ABO genes. Their sequences were as follows: 5Ј-TTAGTTTCTGATTGCCTGATGGTCCT-TGGGCAC and 5Ј-TCATGCCACACAGGCTCAATGCCGT for MY-1 and 2, respectively. PCR products were electrophoresed through a 3% agarose gel, and the DNA was gel-purified using the GeneClean kit. DNA fragments were then ligated with pT-Adv vector from the Advan-Tage PCR cloning kit by the T-A cloning method. Nucleotide sequences of the inserts were determined.
Preparation of Eukaryotic Expression Constructs-The BamHI-XhoI fragment containing the last coding exon of the murine ABO gene was first subcloned from a murine ABO genomic clone, MABO11, into the pT7T3␣18 plasmid vector. The BamHI site was located in the intron preceding the last coding exon of the murine ABO gene. The XhoI site was in the EMBL3 SP6/T7 vector next to the BamHI site used to accommodate genomic DNA. This construct was then digested with SstI and SnaBI. The SstI site was within the pT7T3␣18 plasmid vector, and the SnaBI site was located downstream of the stop codon of the mouse ABO gene coding sequence. The SstI-SnaBI fragment containing the coding sequence in the last coding exon of the mouse ABO gene was then isolated. The human B transferase expression construct with intron, pBBBB (30), was cleaved with BamHI, blunt-ended by the Klenow filling-in reaction, and then digested with SstI. The SstI site was in the intron preceding coding exon 7 of the human ABO gene; BamHI was in the eukaryotic expression vector (originally pSG-5). The SstI-blunt (BamHI) vector fragment containing the human B transferase cDNA sequence of exons 1-6 was then ligated to the mouse SstI-SnaBI fragment to produce the human-mouse chimeric gene (pHumanmouse chimera). A murine cDNA eukaryotic expression construct was then constructed by replacing the EcoRI-AflII fragment from the chimeric construct with the EcoRI-AflII fragment from a 5Ј-RACE clone in the pT-Adv vector. The EcoRI site of the clone was in the plasmid and located 5Ј upstream of the cDNA end. The AflII was in the last coding exon. The EcoRI site in the pSG-5 vector was located downstream of the SV40 early promoter and upstream of the human cDNA sequence. The resultant construct (pMouse) contained the entire coding sequence of the mouse ABO gene cDNA.
DNA Transfection and Enzymatic Assays-Plasmid DNA was prepared by the SDS-alkaline method (25). The HeLa cell line derived from a human adenocarcinoma of uterus was used as a recipient of transient DNA transfection analyses. The HeLa cells express H antigens on their cell surfaces and have been successfully used in similar transfection experiments of A and B transferase expression constructs (5,30,31). Following the manufacturer's protocol, we used LipofectAMINE for transfection. Seventy-two hours after transfection, the cells were washed and harvested. Cell pellets were then lysed in buffer (0.1 M NaCl, 25 mM sodium cacodylate, 10 mM MnCl 2 , and 0.1% Triton X-100). A/B transferase activity was determined by measuring the transfer of carbon-14 from [ 14 C]UDP-GalNAc or [ 14 C]UDP-galactose to the acceptor substrate 2Ј-fucosyllactose, as described previously (31). After incubation, the reaction products were separated from unincorporated nucleotide-sugars by AG1-X8 anion exchange column chromatography.
The incorporation of radioactivity was determined using a scintillation counter.
PCR Amplification and Nucleotide Sequence Determination of the ABO Gene Fragments from Several Strains of Mouse Species and Subspecies-Murine submaxillary glands were used for the expression analyses of A/B transferases. A and B transferase activities were measured by the incorporation of carbon-14 from [ 14 C]UDP-GalNAc or [ 14 C]UDP-galactose to 2Ј-fucosyllactose. Although the same reaction conditions were used, the reaction product was separated from the precursor substrate by paper chromatography rather than column chromatography. Genomic DNA was prepared by the proteinase K-SDS method and used to amplify a DNA fragment derived from the murine ABO gene. The names and nucleotide sequences of the primers used were as follows: SN-16, 5Ј-GAGACTGCAGAACAACACTT; SN-17, 5Ј-CAATGCCGTTGGCCTTGTC. The PCR-amplified DNA fragments were purified through chromatography using S-300 MicroSpin columns and subjected to direct DNA sequencing reactions with the BigDye cycle sequencing kit. After the sequencing reaction, DNA was purified and then analyzed using an ABI Prism 377 automatic DNA sequencer.
Detection of ABH Antigen Expression-Expression of the ABH antigens in murine submaxillary glands was examined immunologically using extracts spotted on a nitrocellulose membrane. Murine anti-A and anti-B monoclonal antibody mixtures were biotinylated using EZ-Link Sulfo-NHS-LC-Biotin, following the protocol provided by the manufacturer. After biotinylation, the unincorporated biotin was removed using Microcon 30 centrifugal filter devices. The submaxillary glands from C57BL and ICR strains of mice were homogenized in buffer containing 20 mM Tris-HCl (pH 7.5), 0.15 M NaCl, and 1% Triton X-100. After centrifugation, the supernatant was diluted with buffer containing 25 mM Tris-HCl (pH 7.5) and 0.1% SDS. The extract was then spotted onto a Duralose-UV membrane. As controls, the extracts similarly obtained from human colon adenocarcinoma SW48 cells (AB phenotype) and from group A and O porcine submaxillary glands were also spotted on the membrane. After drying for 15 min, the membrane was treated with 0.3% hydrogen peroxide and 0.3% fetal calf serum in phosphate-buffered saline for 5 min to block endogenous peroxidase activity. After washing, the membrane was incubated overnight in phosphate-buffered saline containing 4% bovine serum albumin at 4°C. The membrane was then cut into four pieces, which were individually incubated with either biotinylated murine anti-A monoclonal antibody mixture, biotinylated murine anti-B monoclonal antibody mixture, biotinylated Ulex europaeus agglutinin I, or bovine serum albumin (negative control) for 1 h at room temperature. The filters were washed separately and then incubated collectively with the Elite ABC reagents for 15 min. After washing with phosphate-buffered saline, the membranes were treated with 4-chloro-1-naphthol substrate for color development.

Mouse Genome Contains the Human ABO Gene Equivalent-
Certain mammalian cells exhibit ␣133 Gal epitopes. The cDNA encoding ␣1,3-galactosyltransferase that synthesizes this epitope was cloned from cow (32), mouse (33), and pig (34,35). Humans do not exhibit this epitope but possess the antibody against the epitope in sera (36). Human sequence corresponding to this gene was shown to be a pseudogene due to frameshifts and nonsense mutations (32,37). A/B transferases utilize the galactose substrate with fucose, whereas ␣1,3-galactosyltransferase utilizes the substrate without fucose. ABO genes and ␣1,3-galactosyltransferase genes share significant homology at both the nucleotide and deduced amino acid sequence levels (30). Cloned canine cDNA encoding Forssman glycolipid synthetase (UDP-GalNAc:globoside ␣1,3-N-acetyl-Dgalactosaminyltransferase) also exhibited sequence homology (38). Therefore, these genes are believed to have derived from the same ancestral gene and constitute the ABO gene family. Southern hybridization experiments of murine genomic DNA showed different banding patterns when murine ␣1,3-galactosyltransferase cDNA probe and human A transferase cDNA probe were used (17,33). Accordingly, the ABO gene equivalent was assumed to exist in the mouse genome (17). Results from our cloning experiments of the murine ABO gene described here concluded that mice actually do possess an ABO gene equivalent.
Organization of the Murine Gene Is Similar to the Human Counterpart-We cloned the genomic DNA sequence encompassing most of the murine ABO structural gene. By screening 1 million phage plaques from a murine genomic DNA library, we obtained a total of nine independent clones that hybridized with the human A transferase cDNA probe. A preliminary mapping showed that two phage clones named MABO11 and MABO16 contained the entire coding sequence in the last coding exon. Since the MABO16 clone contained the sequence farther upstream, this clone was used for the nucleotide sequence determination. The MABO11 clone containing the farther downstream sequence was used to construct a humanmouse ABO gene chimeric expression construct as well as a murine gene expression construct. We sequenced the entire insert in the MABO16 clone (ϳ11.2 kilobase pairs) with more than 99.9% accuracy. Almost all of the coding sequence was contained in the sequenced region. The exon-intron boundaries were determined and are shown in Fig. 1. Fig. 1, A and B, represents two probable splicing patterns, although other possibilities still exist because the sequence encoding the first few amino acid residues has not yet been identified. There are six coding exons in Fig. 1A and seven in Fig. 1B. Approximately 4.0 kilobase pairs and 70 base pairs upstream of the splicing acceptor site of coding exon 2 (cEXON 2) in Fig. 1A, there was a CTCAGAG sequence and a TGAATCTCAG sequence, respectively. These sequences may be portions of the coding sequence, since they are found upstream in the cDNA preceding the sequence in cEXON 2. Fig. 1B depicts the case where GAATCT-CAG of the latter TGAATCTCAG sequence represents the sequence in the preceding exon. In that case, the acceptor site of cEXON 3 needs to shift 2 nucleotides upstream, which would break up the GT-AG rule of splice junctions. We determined the entire nucleotide sequence contained in the MABO16 clone, which included ϳ5.0 kilobase pairs of sequence upstream of the splicing acceptor site of cEXON2 in Fig. 1A. No sequence corresponding to the 5Ј-untranslated region was found. Therefore, the promoter region of the murine ABO genes must reside farther upstream. The sequence corresponding to human coding exons 3 and 4 was found in one exon in the mouse gene. However, the number of amino acid residues (19 amino acids) was much smaller than that of human exons 3 and 4 combined (35 amino acids). Further studies are needed, since there may be an alternative splicing that would divide this small exon into two smaller exons with an intron in between. The entire insert sequence in the MABO16 clone and the entire cDNA sequence have been deposited in the DNA Data Bank of Japan (DDBJ) (accession numbers AB041038 and AB041039). The Nucleotide and Deduced Amino Acid Sequences of the Mouse ABO Gene Equivalent-The nucleotide and deduced amino acid sequences in the coding region of the murine cDNA were aligned with those of human A 1 -1 (A101) allele (accession number AF134412 in GenBank TM ) by combining the Clustal method (40) and the J. Hein method (41) using the MegAlign software. Results are shown in Fig. 1. Especially high homology was observed in the coding sequence in the last two coding exons. The percentages of identical nucleotide and amino acid residues in the last two coding exons were 78% (642/822) and 81% (222/273) between the two species, respectively. The amino acid sequence of the murine gene was also aligned with the amino acid sequences of human A and B transferases, mouse ␣1,3-galactosyltransferase, and canine Forssman glycolipid synthetase. Results are shown in Fig. 2A. The percentages of the identical amino acid residues of the coding sequences in the last two coding exons are 47% (127 of 272) between the mouse ABO and ␣1,3-galactosyltransferase genes and 49% (132 of 272) between the mouse ABO gene and the dog Forssman synthetase gene. Fig. 2B highlights the amino acid se-quences of the region important for the recognition/binding of nucleotide-sugars. The phylogenetic tree is shown in Fig. 2C. The cloned mouse gene was evolutionarily mapped closest to the human ABO gene. It was also mapped closer to the canine Forssman gene than the murine ␣1,3-galactosyltransferase gene.
Murine Gene Encodes an Enzyme with both A and B Transferase Activities-We examined whether the isolated mouse ABO gene sequence could encode a functional glycosyltrans-ferase. We first constructed a human-mouse chimeric construct in an eukaryotic expression vector pSG-5. A DNA fragment containing coding sequence in the last coding exon of the mouse genomic sequence was linked downstream of the human cDNA sequence of exons 1-6 in the human B transferase expression construct, pBBBB (30). When DNA from the chimera construct was transiently transfected into HeLa cells, the appearance of both A and B transferase activity was observed (Table I). Because there may be inactivating mutations in the upstream FIG. 1. Comparison of the exon-intron boundaries and the nucleotide and deduced amino acid sequences of the coding region of the murine AB gene with those of the human A 1 -1 (A101) allele. The nucleotide sequence of the mouse gene in the MABO16 phage clone and the sequence of the 5Ј-RACEamplified cDNA were used to determine the exon-intron boundaries. Almost all of the coding sequence was contained in the cloned genomic DNA fragment. Since the sequence encoding the first few amino acid residues is still missing, the earlier exon-intron boundaries have not yet been completely determined. In Fig. 1, A and B, two possible splicing patterns are presented. There are six coding exons in A and seven in B. In A, ϳ70 base pairs upstream of the splicing acceptor site in coding exon 2 (cEXON 2), there was a GAATCTCAG sequence. If the acceptor site is shifted upstream by 2 nucleotides, that sequence may represent the preceding exon sequence as shown in B. This shift, however, breaks up the GT-AG rule of splice junctions. In addition, no 5Ј-untranslated region sequence is found upstream. Therefore, A may represent what most likely occurs. XXXXXX, the nucleotide sequence is unknown. Unlike the human sequence, the sequence in human exons 3 and 4 was contained in a single exon in the mouse gene. The numbers shown in parentheses next to the nucleotide numbers of the coding sequence indicate the nucleotide residue numbers of the MABO16 phage clone insert deposited to DDBJ (accession number AB041038). The nucleotide sequence of the murine AB gene cDNA has also been deposited (accession number AB041039). Also shown is the comparison of the nucleotide and deduced amino acid sequences in the coding region of the murine gene with the human A 1 -1 (A101) gene sequences (accession number AF134412 in GenBank TM ). A combined method incorporating the Clustal method (40) and the J. Hein method (41) was used with the MegAlign program. High homology was observed in the coding sequence in the last two coding exons. The percentages of the identical nucleotide and amino acid residues were 78 and 81% between the two sequences, respectively. sequence, as observed in human O alleles, we constructed a mouse cDNA expression construct. The upstream sequence of the AflII site in the last coding exon of the mouse gene in the chimeric construct was replaced by the cDNA sequence obtained from the 5Ј-RACE experiment. This construction produced a mouse cDNA expression construct containing the entire coding sequence. Results from the enzymatic analysis of the transfected HeLa cells are also shown in Table I. Both A and B transferase activities were detected in the cell extract. Based on these results, we concluded that the murine gene is an AB gene that encodes a protein capable of utilizing both UDP-GalNAc and UDP-galactose donor substrates to synthesize A and B antigens.
AB Gene Is Prevailing in Mice-We have determined the partial nucleotide sequences of the coding region in the last coding exon of the murine ABO gene using genomic DNA from several species and subspecies of mice. The results are summarized in Table II However, none were found at the four positions that would distinguish between the human A and B transferases. Furthermore, both A and B transferase activities were also detected in the submaxillary gland extracts from those strains of mice that exhibited amino acid substitutions. Therefore, there are no mutations that change the donor nucleotide-sugar specificity in FIG. 2. Comparisons of the amino acid sequences of the coding regions among murine AB gene, human A 1 -1 (A101) and B (B101) genes, murine ␣1,3-galactosyltransferase gene, and canine Forssman synthetase gene and an evolutionary tree of the ABO and related genes. A, the amino acid sequence of the mouse AB transferase was compared with those of the human A and B transferases, mouse ␣1,3-galactosyltransferase (mouse GalT), and canine Forssman glycolipid synthetase (dog Forssman). The mouse AB transferase was 47 and 49% identical to the mouse ␣1,3-galactosyltransferase gene and the dog Forssman synthetase gene in the last two coding exons, respectively. B, the coding sequences around the amino acid residues important for the nucleotide-sugar recognition/binding are compared. The sequences from the human O03 allele that contains a missense mutation around this area (9) and the cis-AB01 allele with dual transferase activity (8) are also included. C, a phylogenetic tree of the ABO gene family is shown. The following sequences in GenBank TM were used for the computation of evolutionary distance (46) and neighbor-joining tree (39): marmoset GalT (S71333), bovine GalT (J04989), pig GalT (L36152), mouse GalT (M85153), and dog Forssman synthetase (U66140). Numbers on interior branches denote bootstrap probabilities in percent. Branch lengths are proportional to the numbers of amino acid substitutions, and a scale is given at the bottom of the tree. The dotted line denotes the probable root node of this tree. This node also corresponds to a gene duplication event producing both the ␣1,3-galactosyltransferase gene and ABO-related gene. The latter lineage further experienced another gene duplication, resulting in creation of both Forssman glycolipid synthetase gene and ABO gene. The murine AB gene was mapped closest to the human ABO genes. It also mapped closer to the canine Forssman gene than the murine ␣1,3-galactosyltransferase gene. the sequenced and unsequenced regions of the gene in at least those five strains of mice examined.
A and B Antigens Are Expressed in Mice-We have shown that the mouse equivalent of the human ABO gene encodes a protein capable of transferring both GalNAc and galactose through in vitro enzymatic assays using extracts from HeLa cells transfected with the eukaryotic expression constructs of the murine gene. We have also shown the presence of A and B transferase activity in the murine submaxillary glands by use of in vitro assays of tissue extracts. However, the expression of the protein does not necessarily prove the expression of A and B antigens, since the enzymatic reactions require the appropriate substrates and reaction conditions. Therefore, we next analyzed the expression of A and B antigens. Two laboratory strains of M. musculus domesticus mice were analyzed. We initially examined the agglutination of murine red blood cells using murine monoclonal antibody mixtures against A and B antigens under the regular agglutination conditions used for the ABO blood group typing of human red blood cells. No agglutination was observed (data not shown). Because we observed A and B transferase activity in the submaxillary gland extract, we next performed the inhibition study, using boiled extracts of the murine submaxillary glands. No inhibition of the reference human red blood cell agglutination was observed, although the treatment of murine monoclonal antibody mixtures with the control group A porcine submaxillary gland extract resulted in some inhibition (data not shown). These results suggested that A and B antigens are not expressed in abundance if at all. We therefore examined the expression of these antigens using the more sensitive immunological method of nitrocellulose spotting. As shown in Fig. 3, both A and B antigens were detected in the murine submaxillary glands. Apparently, higher expression was observed with A antigens than with B antigens. This may be attributed to the decreased availability of UDP-galactose substrate, resulting from the competition between B transferase and ␣1,3-galactosyltransferase for the same donor nucleotide-sugar. The low expression of A and B antigens may be similarly explained by the competition for the same acceptor substrate between ␣1,3-galactosyltransferase and ␣1,2-fucosyltransferase. Since ␣133-Gal epitope produced by the ␣1,3-galactosyltransferase is abundantly expressed in murine tissues, it is likely that the ␣1,3galactosyltransferase has higher affinity for the acceptor substrates than the ␣1,2-fucosyltransferase that synthesizes H antigens. DISCUSSION We have cloned murine genomic DNA containing most of the coding sequence that was equivalent to the human ABO gene.

TABLE II Nucleotide and amino acid substitutions in mouse species and subspecies
The nucleotide and the deduced amino acid sequences of the murine equivalents of the human ABO gene were compared among several strains of mouse species and subspecies. The nucleotide and the deduced amino acid sequences in the region between nucleotide 360 and 807 in Fig. 1 were examined. The nucleotide sequence substitutions are shown in boldface type, and the resulting amino acid substitutions are shown in parentheses. A and B transferase activity was examined using extracts from submaxillary glands. ϩ and ND denote the presence of activity and not determined, respectively. In these enzymatic assays, A transferase activity of 0.004 -0.093 nmol/h/mg of protein (1.287 for the positive control human gastric carcinoma cell line MKN 45 cell extract) and B transferase activity of 0.008 -0.030 nmol/h/mg (0.035 for the positive control human gastric carcinoma cell line KATOIII cell extract) were detected among different strains of mice.
The sequence encoding the first few amino acid residues and the promoter region sequence farther upstream remain to be cloned. In humans, ABH antigens are widely expressed on a variety of cell surface molecules in many tissues, depending on the ABO genotypes of the individuals. These include glycoproteins and glycolipids on mucous cells, nerve cells, red cells, epidermis, and vascular endothelium. ABH antigen expression seems to be more highly restricted in lower mammals (42,43). To understand the differential expression mechanism of A/B antigens between humans and mice, cloning of the murine gene promoter region will be necessary. Functional analysis of the murine gene has shown that the cloned murine gene is really an AB gene and encodes an enzyme with both A and B transferase activity. In our prior studies of the human ABO gene, we determined the molecular causes of two interesting phenomena named cis-AB (8) and B(A) (7). Human cis-AB alleles encode a protein with strong A and weak B transferase activity, whereas B(A) alleles encode a protein with strong B and weak A transferase activity. A mutation was found in these alleles at each of the four amino acid substitutions that distinguish human A transferase (AAAA) from B transferase (BBBB). The cis-AB alleles were represented as AAAB, whereas the B(A) alleles were represented as BABB in those four substitution positions. By employing DNA transfection assays of A-B transferase chimeras and in vitro mutagenized expression constructs, we showed that these amino acid substitutions might cause changes in specificity and activity of the enzyme (30,31). As shown in Fig. 2B, the amino acid residues at codons 245 and 247 of the murine gene, which corresponded to human codons 266 and 268 (the third and fourth positions of amino acid substitutions between A and B transferases), were glycine and alanine. The alanine residue is identical to that of human B transferase, but the glycine residue is much smaller than the methionine residue of human B transferase. Therefore, the mouse enzyme is expected to possess a larger space for donor nucleotide-sugar substrate recognition/binding. Since N-acetyl-D-galactosamine is more bulky than galactose, the increase in space may allow for the accommodation of not only the galactose portion of UDP-galactose but also the GalNAc portion of UDP-GalNAc, which may account for the dual transferase activity of the murine enzyme. Dog Forssman glycolipid synthetase contains glycine and alanine, the same amino acid residues as those of the murine ABO gene equivalent, at codons 261 and 263. However, the codon 264 in the Forssman glycolipid synthetase is valine rather than phe-nylalanine in the murine AB gene. This difference in size (valine is smaller than phenylalanine) may possibly render dog Forssman glycolipid synthetase able to catalyze the transfer of only GalNAc residues and not galactose residues.
To understand the meaning and role of ABO polymorphism during evolution, the cloned murine AB gene fragments will be useful in the production of knockout (group O) mice at the ABO locus. Production of group A and B mice may also be possible by knocking in the genes after modifying the specificity of the enzymes so that only A or B transferase activity is retained. Additional manipulation(s) may be a prerequisite to enhance the expression of A and B antigens. These may include the knockout of the ␣1,3-galactosyltransferase gene (44) or the introduction of the ␣1,2-fucosyltransferase gene under strong promoter (45) to either abolish or lessen the expression of the ␣1,3-galactosyltransferase. These mice with different ABO phenotypes in the same genetic background may clarify the functionality of the ABO system in the future.