HLA-G is a nonclassical class I MHC molecule of unknown function expressed on human trophoblast. The level of polymorphism at the HLA-G locus is of considerable importance, since the paternally inherited gene product is exposed to the maternal immune system during pregnancy. However, previous studies of HLA-G polymorphism using genomic DNA samples have produced conflicting results. Our aim was to investigate polymorphism in trophoblast HLA-G mRNA from pregnancies in ten Caucasian and twelve Afro-Caribbean women by RT-PCR. A similar PCR protocol was also applied to umbilical cord blood genomic DNA from two Caucasian and two Afro-Caribbean neonates. Caucasian cDNA yielded only two different sequences: G*01011, and one containing a previously reported synonymous substitution. Afro-Caribbean samples yielded these sequences as well as one previously reported conservative (leucine-to-isoleucine) substitution. PCR amplification from genomic DNA samples from both populations using previously published primer pairs generated sequences containing multiple substitutions, many of which were nonsynonymous. More than two sequences were produced from genomic DNA from each individual. In contrast, amplification from the same genomic DNA using new primers complementary to exons of the HLA-G gene yielded the same few sequences generated from cDNA. These results suggest that polymorphism at the HLA-G locus is extremely limited in Caucasian and Afro-Caribbean populations. This suggests that spurious polymorphism has been reported in African Americans due to the use of intron-complementary PCR primers on genomic DNA samples. The monomorphic nature of HLA-G may allow trophoblast to carry out the immunological functions of class I-bearing tissues without compromising successful pregnancy.

Human histocompatibility leukocyte Ag (HLA)-G is a nonclassical MHC class I molecule expressed on fetal trophoblast cells in the human placenta. It has been suggested that HLA-G plays a role in protection of the developing human fetus from maternal immune attack, possibly by modulating maternal NK cell or T lymphocyte activity (1, 2). The possible role of HLA-G in pregnancy may be related to the nonclassical features of the molecule: restricted tissue expression, structural features, and limited polymorphism.

Unlike classical class I molecules, HLA-G is thought to be expressed on a restricted set of different cell types. It is present on subsets of fetal cells in the placenta, including cell types in direct contact with maternal tissues, the extravillous and chorionic cytotrophoblasts (3, 4, 5, 6, 7, 8). Notably, these cells are unusual among human nucleated cells in not expressing HLA-A and –B, although they do express HLA-C and nonclassical HLA–E (9, 10).

Although HLA-G has a truncated cytoplasmic region (4, 11, 12), the structure of the rest of the molecule is thought to be similar to that of classical class I molecules. Not only are the deduced amino acid sequences of the HLA-G and other HLA molecules similar, but HLA-G also complexes with β2-microglobulin and peptides, binds to CD8, and is recognized by some Abs that also bind to classical class I molecules (7, 13, 14, 15).

HLA-G is thought to be less polymorphic than classical class I molecules. It has been suggested that limited HLA-G polymorphism ensures that the paternally derived HLA-G gene product present on fetal trophoblast cells is similar to those of the mother, so that trophoblast expression of HLA-G does not induce a maternal alloresponse. Thus, lack of HLA-G polymorphism is central to theories concerning its function, and, as a result, a number of studies of HLA-G polymorphism have been conducted (16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28). Initial studies (3, 4) of HLA-G protein expressed on the placenta in British and American populations failed to detect significant variation in the protein between individuals, suggesting that HLA-G is effectively monomorphic. As molecular, and especially PCR-based, techniques have become more widely used, several research groups have conducted searches for different HLA-G gene variants in diverse human populations. Initially, these results confirmed the protein studies and demonstrated only limited coding changes in the HLA-G gene. For example, polymorphisms in the 3′-untranslated region of the HLA-G gene have been reported in Japanese (17), Australian (18), and French (19) populations by RFLP or PCR amplification of genomic DNA. PCR-based studies of the coding region of HLA-G gene in Spanish (16), Japanese (22), Finnish (23), and French (14, 28) populations have demonstrated a limited number of variable nucleotides within the HLA-G gene, and most of these induce either no change in the deduced amino acid sequence or conservative amino acid substitutions. There are few published data on the sequence of HLA-G cDNA, presumably because cDNA is more difficult to obtain than genomic DNA, but a study of HLA-G cDNA in Danes (27) has also detected few coding substitutions in different HLA-G alleles.

In contrast to these studies, the data of van der Ven and Ober (20) appear to demonstrate considerable polymorphism in HLA-G in African Americans. This study identified multiple nucleotide substitutions in the HLA-G gene, the majority of which (thirty-five) were clustered in exons 2 and 3 encoding the α1 and α2 domains, which contain the peptide binding groove. Twenty-six of these reported substitutions are nonsynonymous, and many of these would involve nonconservative amino acid substitutions. One substitution would lead to a stop codon in exon 3 leading to a truncated, and presumably nonfunctional, molecule.

Confirmation of these apparently remarkable differences between HLA-G in African-origin humans and other human populations is essential, since they have considerable ramifications for both the evolution and the potential function of HLA-G. Evidence of HLA-G polymorphism in one human population would conflict with the theory that nonpolymorphism of HLA-G is essential to prevent fetal allorejection. Like most other studies of HLA-G polymorphism, the study of van der Ven and Ober involved the application of a PCR technique to genomic DNA samples using PCR primers designed to be specific for HLA-G. The human genome contains multiple class I-like sequences, many of which are not expressed and some of which are poorly characterized (29). Thus, care must be taken to ensure that PCR primers are specific for HLA-G to prevent accidental amplification of other sequences similar to HLA-G. In addition, the presence of multiple DNA regions with sequences similar to HLA-G may affect the fidelity of replication of HLA-G sequences by the polymerases used for PCR.

The aim of this study was to overcome some of the problems inherent in detecting HLA-G polymorphism in genomic DNA samples by sequencing RT-PCR products generated from mRNA from fetal chorionic cytotrophoblast cells derived from pregnancies in Caucasian and Afro-Caribbean women. These sequences were compared with those generated from PCR products produced from genomic DNA isolated from cord blood from the same pregnancies.

Placental membranes were collected at term from normal pregnancies in ten Caucasian women and twelve Afro-Caribbean women admitted to the maternity departments of the John Radcliffe Hospital, Oxford, and at St. Thomas’ Hospital, London. The fetal membranes of humans consist of amniotic and chorionic membranes that are mutually adherent. The two membranes were gently peeled apart, and 2-mm2 pieces of chorion were excised with a scalpel, snap-frozen in microcentrifuge tubes on dry ice, and stored at −80°C until use.

From two of the above pregnancies in Caucasians and two of the pregnancies in Afro-Caribbean women, umbilical cord blood samples were also collected.

mRNA was prepared from chorion samples by cell lysis and extraction on oligo(dT)-coated beads (Dynabeads mRNA Direct Kit; Dynal, Oslo, Norway), and cDNA was generated by reverse transcription (cDNA Synthesis System; Life Technologies, Paisley, U.K.). Genomic DNA was prepared from 200-μl aliquots of heparinized whole umbilical cord blood using the QIAamp Blood Kit (Qiagen, Hilden, Germany).

A 426-nucleotide fragment of exons 2 and 3 of HLA-G cDNA was amplified by PCR using primers DB-G1 and DB-G2 (Table I) designed according to a published HLA-G sequence, G*01011 (11). The primers were designed to amplify the region containing most reported HLA-G polymorphisms and not to be complementary to regions of HLA-C and -E cDNAs. Thirty-five PCR cycles (94°C for 1 min., 55°C for 2 min., and 72°C for 2 min.) were conducted. The PCR volume was 100 μl containing 200 mmol/L Tris-HCl (pH 8.4), 50 mmol/L KCl, 5 mmol/L MgCl2, 25 μmol/L dATP, dCTP, dGTP, and dTTP, 1 μmol/L of each primer, 2.5 U Taq polymerase (Life Technologies), and one-tenth of the cDNA produced from a single chorion sample. PCR products were separated by agarose gel electrophoresis, extracted, and ligated into SmaI-digested M13 mp18 vector (Pharmacia, Uppsala, Sweden) and then used to transform TG1 strain Escherichia coli. Multiple clones from each chorion sample were sequenced by the dideoxy chain termination method (Sequenase; USB, Cleveland, OH). At least nine clones were fully sequenced for each sample. PCR amplification was repeated with cDNA samples that had yielded sequences differing from the first two HLA-G sequences to be published, G*01011 and G*01012 (11, 12) to exclude the possibility of PCR error. Sequences of an HLA-G cDNA lacking exon 3 transcripts were frequently generated and were excluded from the analysis (a truncated HLA-G mRNA transcript lacking exon 3 is produced by trophoblast cells; Ref. 30).

Table I.

Primer pairs used to amplify portions of HLA-G from complementary and genomic DNA

NamePrimer SequenceComplementary Region of HLA-G gene/cDNAaTemplate DNA UsedAmplicon Length
DB-G1 5′-CCGGAGTATTGGGAAG-3′ 370-385 (exon 2) cDNA 426 
DB-G2 5′-CTCATAGTCAAAGACAG-3′ 1618-1602 (exon 4)   
GCS6/5 5′-GACCCTCTACCTGGGAGAACCCCA-3′ 591-614 (intron 2) Genomic DNA 347 
G3i3 5′-TCTGTGGAGCCACTCCACGCACGT-3′ 937-914 (exon 3)   
G3i5 5′-CAGATCTCCAAGCGCAAGTGTGAG-3′ 848-871 (exon 3) Genomic DNA 352b 
GCS4/3 5′-CCTCCACTCCCTCAGAGACTTCAT-3′ 1199-1176 (intron 3)   
DB-G1 5′-CCGGAGTATTGGGAAG-3′ 370-385 (exon 2) Genomic DNA 568 
G3i3 5′-TCTGTGGAGCCACTCCACGCACGT-3′ 937-914 (exon 3)   
G3i5 5′-CAGATCTCCAAGCGCAAGTGTGAG-3′ 848-871 (exon 3) Genomic DNA 771 
DB-G2 5′-CTCATAGTCAAAGACAG-3′ 1618-1602 (exon 4)   
NamePrimer SequenceComplementary Region of HLA-G gene/cDNAaTemplate DNA UsedAmplicon Length
DB-G1 5′-CCGGAGTATTGGGAAG-3′ 370-385 (exon 2) cDNA 426 
DB-G2 5′-CTCATAGTCAAAGACAG-3′ 1618-1602 (exon 4)   
GCS6/5 5′-GACCCTCTACCTGGGAGAACCCCA-3′ 591-614 (intron 2) Genomic DNA 347 
G3i3 5′-TCTGTGGAGCCACTCCACGCACGT-3′ 937-914 (exon 3)   
G3i5 5′-CAGATCTCCAAGCGCAAGTGTGAG-3′ 848-871 (exon 3) Genomic DNA 352b 
GCS4/3 5′-CCTCCACTCCCTCAGAGACTTCAT-3′ 1199-1176 (intron 3)   
DB-G1 5′-CCGGAGTATTGGGAAG-3′ 370-385 (exon 2) Genomic DNA 568 
G3i3 5′-TCTGTGGAGCCACTCCACGCACGT-3′ 937-914 (exon 3)   
G3i5 5′-CAGATCTCCAAGCGCAAGTGTGAG-3′ 848-871 (exon 3) Genomic DNA 771 
DB-G2 5′-CTCATAGTCAAAGACAG-3′ 1618-1602 (exon 4)   
a

Nucleotide numbering refers to position within the HLA-G gene sequence, including introns (11), with the first nucleotide of exon 1 as nucleotide 1.

b

The previously reported (20) amplicon size is incorrect.

Fragments were also amplified from the genomic DNA samples using the two primer pairs (GCS6/5, G3i3 and G3i5, GCS4/3, see Table I) used in a previous study of African Americans (20) under the same conditions as above. Each of these primer pairs consists of one primer complementary to a region within an exon of the HLA-G gene and one primer complementary to a region within an intron. PCR products were isolated and sequenced as above.

Finally, fragments were amplified from the same genomic DNA samples using the two primer pairs (DB-G1, G3i3 and G3i5, DB-G2, see Table I). All of these primers are complementary to a region within an exon of the HLA-G gene. PCR products were isolated and sequenced as above. These sequences were then compared with those derived previously from chorion cDNA and genomic DNA from the same pregnancies.

Analysis of the sequences generated by RT-PCR was conducted as follows. The number of heterozygous individuals and the number of examined HLA-G loci differing from the G*01011 sequence in the two populations were compared using the V2 modification of the χ2 test.

The HLA-G cDNA sequences derived from ten Caucasian and ten Afro-Caribbean fetuses are presented in Table II. Each sample yielded a single PCR product band on gel electrophoresis. No sample from any individual generated more than two different sequences. The Caucasian material gave rise to only two sequences: G*01011 (11) and a sequence containing a single synonymous substitution that has been described (G*01012) in other human populations (20, 22) as well as in the BeWo choriocarcinoma cell line (12). The Afro-Caribbean material generated the above two sequences as well as one with a previously reported (20, 22) nonsynonymous substitution that would be expected to alter a leucine residue at the N-terminal end of the α2 domain to an isoleucine (G*01041). The number of heterozygous individuals in the Caucasian group was not different from that in the Afro-Caribbean group (p > 0.05). The numbers of examined HLA-G loci differing from the G*01011 sequence in the two populations were also not significantly different (p > 0.05).

Table II.

Chorionic cytotrophoblast HLA-G cDNA sequences derived from Caucasian and Afro-Caribbean pregnancies by PCRa

PopulationNumber of IndividualsSequences Generated
Caucasian G*01011 only 
 G*01011 and C706 → T 
 C706 → T 
Afro-Caribbean G*01011 only 
 G*01011 and C706 → T 
 G*01011 and C755 → A (Leu134 → Ile) 
 C706 → T only 
 C706 → T and C755 → A (Leu134 → Ile) 
PopulationNumber of IndividualsSequences Generated
Caucasian G*01011 only 
 G*01011 and C706 → T 
 C706 → T 
Afro-Caribbean G*01011 only 
 G*01011 and C706 → T 
 G*01011 and C755 → A (Leu134 → Ile) 
 C706 → T only 
 C706 → T and C755 → A (Leu134 → Ile) 
a

Sequences are referred to as G*01011 where they correspond to this previously published HLA-G gene sequence (11). Other sequences are referred to by substitutions they contain with respect to G*01011 (DNA sequence positions are described as in Table I). Predicted nonsynonymous substitutions are described in parentheses (amino acids are numbered from the first amino acid encoded by exon 1).

The genomic DNA sequences derived from two Caucasian and two Afro-Caribbean fetuses by PCR amplification using mixed intron/exon primers are presented in Table III. Each sample yielded a single PCR product band on gel electrophoresis. PCR amplification of genomic DNA from the two Caucasian samples generated five sequences from each neonate. Both samples generated the sequences previously derived from cDNA as well as others, some predicted to encode multiple amino acid substitutions, one deletion that would be expected to disrupt the reading frame of the gene and one substitution that would insert a stop codon within exon 3. PCR amplification of genomic DNA from the two Afro-Caribbean samples generated five and three different sequences respectively (Fig. 1). Both samples generated the sequences previously derived from cDNA as well as others, some predicted to encode multiple amino acid substitutions.

Table III.

Comparison of HLA-G sequences generated from two Caucasian and two Afro-Caribbean babiesa

IndividualSequences Generated from Chorionic Cytotrophoblast cDNAAdditional Sequences Generated from Cord Blood Genomic DNA Using the Primer Pairs (GCS6/5, G3i3) and (G3i5, GCS4/3)
Caucasian A 1. G*01011 1. AA878&879→CG(Asn180→Arg), ACAAAGC889-895→CGAGGAC (GluGlnArg178-180→AspGluAsp), A933→G(His193→Arg), G958→C(Glu201→Asp), T960→C(Met202→Thr), C965→G(Gln204→Glu) 
  2. A889→G, G916→C, G920→A(Val189→Met), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
  3. T960→C(Met202→Thr), A966→G(Gln204→Arg) 
  4. A966→G(Gln204→Arg) 
Caucasian B 1. G*01011 1. A748→T, C829→deleted (disrupts reading frame) 
  2. A889→G, G916→C, G920→A(Val189→Met), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
  3. A889→G, G916→C, C919→A(Cys188→STOP), GT920&921→AA(Val189→Lys), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
  4. A966→G(Gln204→Arg) 
Afro-Caribbean A 1. G*01011 1. A889→G, G916→C, G920→A(Val189→Met), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
  2. G916→C, G920→A(Val189→Met), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
 2. C755→A(Leu134→Ile) 3. A966→G(Gln204→Arg) 
Afro-Caribbean B 1. C706→T 1. A889→G, G916→C, G920→A(Val189→Met), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
  2. G916→C, G920→A(Val189→Met), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
IndividualSequences Generated from Chorionic Cytotrophoblast cDNAAdditional Sequences Generated from Cord Blood Genomic DNA Using the Primer Pairs (GCS6/5, G3i3) and (G3i5, GCS4/3)
Caucasian A 1. G*01011 1. AA878&879→CG(Asn180→Arg), ACAAAGC889-895→CGAGGAC (GluGlnArg178-180→AspGluAsp), A933→G(His193→Arg), G958→C(Glu201→Asp), T960→C(Met202→Thr), C965→G(Gln204→Glu) 
  2. A889→G, G916→C, G920→A(Val189→Met), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
  3. T960→C(Met202→Thr), A966→G(Gln204→Arg) 
  4. A966→G(Gln204→Arg) 
Caucasian B 1. G*01011 1. A748→T, C829→deleted (disrupts reading frame) 
  2. A889→G, G916→C, G920→A(Val189→Met), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
  3. A889→G, G916→C, C919→A(Cys188→STOP), GT920&921→AA(Val189→Lys), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
  4. A966→G(Gln204→Arg) 
Afro-Caribbean A 1. G*01011 1. A889→G, G916→C, G920→A(Val189→Met), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
  2. G916→C, G920→A(Val189→Met), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
 2. C755→A(Leu134→Ile) 3. A966→G(Gln204→Arg) 
Afro-Caribbean B 1. C706→T 1. A889→G, G916→C, G920→A(Val189→Met), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
  2. G916→C, G920→A(Val189→Met), A933→G(His193→Arg), T938→C(Tyr195→His), T960→C(Met202→Thr) 
a

Sequences were obtained from PCR products amplified from chorionic cytotrophoblast cDNA (all primers complementary to exonic regions) and from fetal cord blood genomic DNA (each primer pair consisting of one primer complementary to a region within an exon and one primer complementary to a region within an intron). DNA sequence positions and amino acid substitutions are described as in Tables I and II.

FIGURE 1.

Three genomic DNA sequences derived from cord blood from a baby born to an Afro-Caribbean woman. The sequences were generated by PCR using the primer pairs (GCS6/5, G3i3) and (G3i5, GCS4/3). Each primer pair consists of one primer complementary to a portion of an exon of HLA-G and one primer complementary to an intron. Sequence I corresponds to G*01011 (11 ) and has been numbered according to that sequence (as in Table II). Nucleotides where adjacent sequences differ are marked by an asterisk.

FIGURE 1.

Three genomic DNA sequences derived from cord blood from a baby born to an Afro-Caribbean woman. The sequences were generated by PCR using the primer pairs (GCS6/5, G3i3) and (G3i5, GCS4/3). Each primer pair consists of one primer complementary to a portion of an exon of HLA-G and one primer complementary to an intron. Sequence I corresponds to G*01011 (11 ) and has been numbered according to that sequence (as in Table II). Nucleotides where adjacent sequences differ are marked by an asterisk.

Close modal

Replacement of PCR primers complementary to HLA-G introns with those complementary to exons considerably reduced the number of sequences amplified from genomic DNA. Multiple PCR product bands were usually generated on gel electrophoresis when these exonic primers were used, and fragments of the correct size were isolated and purified. When this was done, none of the Caucasian or Afro-Caribbean genomic DNA samples yielded any sequences other than those previously amplified from cDNA from the same individuals.

Our RT-PCR-based system has detected little polymorphism in fetal HLA-G mRNA in pregnancies in Caucasian and Afro-Caribbean women. The limited HLA-G polymorphism demonstrated in the present study is similar to that previously detected in Caucasian and Asian populations but is in marked contrast to the variation in the HLA-G gene detected by van der Ven and Ober (20) in African-Americans by PCR amplification of genomic DNA. In the present study, RT-PCR of cDNA from each individual yielded only one or two different sequences, and these presumably represent the maternally and paternally inherited HLA-G alleles. This confirms recent reports that both parental alleles of HLA-G are expressed by fetal trophoblast cells (27, 31) and since HLA-G heterozygote samples yield both allele sequences with approximately equal frequency, it appears that genetic imprinting of HLA-G does not occur in trophoblast. Thus, our results strongly suggest that, although paternally inherited HLA-G is present on trophoblast cells, it is unlikely to be recognized as foreign by the maternal immune system because limited HLA-G polymorphism ensures that paternal and maternal HLA-G are extremely similar or identical.

The use of RT-PCR allowed specific amplification of HLA-G sequences even though the approach to PCR primer design was simple: the primers were designed to amplify the region containing most of the reported HLA-G polymorphisms and not to anneal to cDNAs of other class I molecules present in trophoblast (HLA-C and –E). In contrast, PCR amplification of genomic DNA from both Caucasian and Afro-Caribbean samples usually generated more than two sequences per individual, suggesting that this technique amplifies DNA sequences other than that of HLA-G, despite the suggestion of van der Ven and Ober that the GCS6/5, G3i3, G3i5, and GCS4/3 PCR primers used are HLA-G specific (20). This raises the possibility that some previously reported HLA-G polymorphisms may be spurious, reflecting the lack of specificity of the methods used rather than true HLA-G variation in the human population. The generation of additional sequences does not appear to be dependent on the ethnic origin of the samples since not only were extra sequences generated from both Caucasian and African DNA, but there was considerable similarity between the additional sequences generated from the two populations.

It is unlikely that the use of genomic DNA as a PCR template is the only reason why sequences other than HLA-G were amplified. Laboratories around the world routinely and successfully use well-validated PCR amplification techniques on genomic DNA to determine the classical class I genes present in patient tissue samples. However, extreme caution must be exercised in the design of PCR primers for use with genomic samples to prevent the primers annealing to unknown HLA-like sequences elsewhere in the genome. In particular, the use of primers complementary to regions within the introns of the HLA-G gene should be avoided since the introns of both functional and nonfunctional human class I genes are rather poorly characterized. This may explain why, in the present study, the replacement of intronic primers by exonic ones prevented the amplification of additional sequences from genomic DNA.

A simple RT-PCR-sequencing protocol was used in the present study to allow PCR fidelity errors to be detected easily. Error is a problem inherent in PCR, and the simplest way to reduce it is to conduct a single PCR reaction before sequence analysis and to repeat PCR amplification of any samples yielding nonconsensus sequences. For example, in the study in African-Americans (20), thirty-five PCR cycles were used to generate fragments for single-stranded conformational polymorphism analysis, and these products were then used in a second PCR with an unspecified number of cycles to generate template for sequencing reactions. Any novel polymorphisms detected by such a method should be confirmed by repetition, or better, by a different method such as PCR amplification using sequence-specific primers. In addition, it would have been instructive if the method employed in the above study had also been used to generate sequence data from Caucasian or Japanese populations to allow comparison with other studies of these populations.

The African-origin study group used in our study is broadly comparable with that used in the study of van der Ven and Ober (20). Ten of the Afro-Caribbean mothers believed their families to originate in West Africa and two believed they had Ethiopian or Kenyan origins. Hence, if it is assumed that most African Americans (20) are of West African descent, then comparisons between the two studies are reasonable. However, the paternity of the pregnancies was not recorded, and it is possible that Caucasian women were carrying the children of Afro-Caribbean fathers and vice versa.

The present study was not designed to be an exhaustive survey of the HLA-G alleles present in the human population, since RT-PCR is a rather laborious method for screening large numbers of individuals. For example, placental tissue samples are required for RT-PCR amplification of HLA-G. Instead, the aim was to use RT-PCR to clarify the discrepancy between the results of van der Ven and Ober (20) and many other studies of HLA-G polymorphism. The use of RT-PCR on sample groups of limited size allowed many of the potential shortcomings of PCR of genomic DNA to be highlighted. The present study does not preclude future use of PCR of genomic DNA for HLA-G typing: indeed, such a system is routinely used to characterize the far more polymorphic HLA-A, -B and -C genes. Instead, these findings demonstrate how any PCR-based HLA typing protocol should be carefully validated before it is applied to large population samples. Once such a protocol is established, it would then be extremely interesting to reexamine the HLA-G polymorphisms present in African Americans.

The results of the present experiments demonstrate little evidence of HLA-G polymorphism in either Afro-Caribbean or Caucasian populations. The data strongly suggest that, during pregnancy in both Caucasian and Afro-Caribbean women, fetal trophoblast expresses both maternally and paternally inherited HLA-G genes and that the mother does not mount an alloresponse to the fetus because HLA-G is almost monomorphic at the protein level in human populations. Thus, expression of HLA-G on trophoblast with concurrent down-regulation of classical class I expression may allow trophoblast to carry out some or all of the immunological functions of a class I-bearing tissue without compromising successful pregnancy.

We thank Dr. Lorin Lakasing and Prof. Lucilla Poston for their assistance with sample collection in London as well as Dr. Michael Bunce for his helpful comments during the planning of these experiments.

1

This study was funded by the Wellcome Trust.

1
Sanders, S. K., P. A. Giblin, P. Kavathas.
1991
. Cell-cell adhesion mediated by CD8 and human histocompatibility complex class 1 molecule on cytotrophoblasts.
J. Exp. Med.
174
:
737
2
Pazmany, L., O. Mandelboim, M. Vales-Gomez, D. M. David, H. T. Reyburn, J. L. Strominger.
1996
. Protection from natural killer cell-mediated lysis by HLA-G expression on target cells.
Science
274
:
792
3
Ellis, S. A., I. L. Sargent, C. W. G. Redman, A. J. McMichael.
1986
. Evidence for a novel HLA antigen found on human extravillous trophoblast and a choriocarcinoma cell line.
Immunology
59
:
595
4
Kovats, S., E. K. Main, C. Librach, M. Stubblebine, S. J. Fisher, R. DeMars.
1990
. A class I antigen, HLA-G, expressed in human trophoblasts.
Science
248
:
220
5
Yelavarthi, K. K., J. L. Fishback, J. S. Hunt.
1991
. Analysis of HLA-G mRNA in human placental and extraplacental membrane cells by in situ hybridization.
J. Immunol.
146
:
2847
6
Chumbley, G., A. King, N. Holmes, Y. W. Loke.
1993
. In situ hybridization and Northern blot demonstration of HLA-G mRNA in human trophoblast populations by locus-specific oligonucleotide.
Hum. Immunol.
37
:
17
7
Schmidt, C. M., H. T. Orr.
1993
. Maternal/fetal interactions: the role of the MHC class I molecule HLA-G.
Crit. Rev. Immunol.
3
:
207
8
Loke, Y. W., A. King, T. Burrows, L. Gardner, M. Bowen, S. Hiby, S. Howlett, N. Holmes, D. Jacobs.
1997
. Evaluation of trophoblast HLA-G antigen with a specific monoclonal antibody.
Tissue Antigens
50
:
135
9
King, A., C. Boocock, A. M. Sharkey, L. Gardner, A. Beretta, A. G. Siccardi, Y. W. Loke.
1996
. Evidence for the expression of HLA-C class I mRNA and protein by human first trimester trophoblast.
J. Immunol.
156
:
2068
10
Wei, X. H., H. T. Orr.
1990
. Differential expression of HLA-E, HLA-F, and HLA-G transcripts in human tissue.
Hum. Immunol.
29
:
131
11
Geraghty, D. E., B. H. Koller, H. T. Orr.
1987
. A human major histocompatibility complex class I gene that encodes a protein with a shortened cytoplasmic segment.
Proc. Natl. Acad. Sci. USA
84
:
9145
12
Ellis, S. A., M. S. Palmer, A. J. McMichael.
1990
. Human trophoblast and the choriocarcinoma cell line BeWo express a truncated HLA class I molecule.
J. Immunol
144
:
731
13
Lee, N., A. R. Malacko, A. Ishitani, M. C. Chen, J. Bajorath, H. Marquardt, D. E. Geraghty.
1995
. The membrane-bound and soluble forms of HLA-G bind identical sets of endogenous peptides but differ with respect to TAP association.
Immunity
3
:
591
14
Diehl, M., C. Munz, W. Keilholz, S. Stevanovic, N. Holmes, Y. W. Loke, H. G. Rammensee.
1996
. Nonclassical HLA-G molecules are classical peptide presenters.
Curr. Biol.
6
:
305
15
Sanders, S. K., P. A. Giblin, P. Kavathas.
1991
. Cell-cell adhesion mediated by CD8 and human histocompatibility leukocyte antigen G, a nonclassical major histocompatibility complex class 1 molecule on cytotrophoblasts.
J. Exp. Med.
174
:
737
16
Morales, P., A. Corell, J. Martinez-Laso, J. M. Martin-Villa, P. Varela, E. Paz-Artal, L.-M. Allende, A. Arnaiz-Villena.
1993
. Three new HLA-G alleles and their linkage disequilibria with HLA-A.
Immunogenetics
38
:
323
17
Tamaki, J., Y. Arimura, T. Koda, S. Fujimoto, T. Fujino, A. Wakisaka, M. Kakinuma.
1993
. Heterogeneity of HLA-G genes identified by polymerase chain reaction/single strand conformational polymorphism (PCR/SSCP).
Microbiol. Immunol.
37
:
633
18
Harrison, G. A., K. E. Humphrey, I. B. Jakobsen, D. W. Cooper.
1993
. A 14 bp deletion polymorphism in the HLA-G gene.
Hum. Mol. Genet.
2
:
2200
19
Alizadeh, M., C. Legras, G. Semana, P. Le Bouteiller, B. Genetet, R. Fauchet.
1993
. Evidence for a polymorphism of HLA-G gene.
Hum. Immunol.
38
:
206
20
van der Ven, K., C. Ober.
1994
. HLA-G polymorphisms in African Americans.
J. Immunol.
153
:
5628
21
Ober, C., B. Rosinsky, C. Grimsley, K. van der Ven, A. Robertson, A. J. Runge.
1996
. Population genetic studies of HLA-G: allele frequencies and linkage disequilibrium with HLA-A1.
Reprod. Immunol.
32
:
111
22
Yamashita, T., F. Tomoyuki, Y. Watanabe, K. Tokunaga, K. Tadokoro, T. Juji, Y. Taketani.
1996
. HLA-G gene polymorphism in a Japanese population.
Immunogenetics
44
:
186
23
Karhukorpi, J., I. Ikaheimo, S. Silvennoinen-Kassinen, A. Tiilikainen.
1996
. HLA-G polymorphism and allelic association with HLA-A in a Finnish population.
Eur. J. Immunogenet.
23
:
153
24
Suarez, M. B., P. Morales, M. J. Castro, V. Fernandez, P. Varela, M. Alvarez, J. Martinez-Laso, and A. Arnaiz-Villena. A new HLA-G allele (HLA-G*0105N) and its distribution in the Spanish population. Immunogenetics 45:464.
25
Kirszenbaum, M., S. Djoulah, J. Hors, I. Le Gall, E. B. de Oliveira, S. Prost, J. Dausset, E. D. Carosella.
1997
. HLA-G gene polymorphism segregation within CEPH reference families.
Hum. Immunol.
53
:
140
26
van der Ven, K., S. Skrabin, G. Engels, D. Krebs.
1998
. HLA-G polymorphisms and allele frequencies in Caucasians.
Hum. Immunol.
59
:
302
27
Hviid, T. V., M. Meldgaard, S. Sorensen, N. Morling.
1997
. Polymorphism of exon 3 of the HLA-G gene.
J. Reprod. Immunol.
35
:
31
28
Moreau, P., P. Paul, N. Rouas-Freiss, M. Kirszenbaum, J. Dausset, E. D. Carosella.
1998
. Molecular and immunologic aspects of the nonclassical HLA class I antigen HLA-G: evidence for an important role in the maternal tolerance of the fetal allograft.
Am. J. Reprod. Immunol.
40
:
136
29
Le Bouteiller, P..
1994
. HLA class I chromosomal region, genes, and products; facts and questions.
Crit. Rev. Immunol.
14
:
89
30
Ishitani, A., D. E. Geraghty.
1992
. Alternative splicing of HLA-G transcripts yields proteins with primary structures resembling both class I and class II antigens.
Proc. Natl. Acad. Sci. USA
89
:
3947
31
Hashimoto, K., C. Azuma, M. Koyama, T. Nobunaga, T. Kimura, K. Shimoya, Y. Kubota, F. Saji, Y. Murata.
1997
. Biparental alleles of HLA-G are co-dominantly expressed in the placenta.
Jpn. J. Hum. Genet.
42
:
181