|
|
||||||||
Minireview |
a Service de Recherches en Hémato-Immunologie, CEA-DSV-DRM, Hôpital St-Louis,> Institut Universitaire d'Hématologie, 75475 Paris, Cedex 10, France
b Fondation Jean Dausset, CEPH, 75010 Paris, France
| ABSTRACT |
|---|
|
|
|---|
embryo, immunology, placenta, pregnancy, trophoblast
A key role in fetal-maternal immune tolerance has been attributed to human leukocyte antigen-G (HLA-G), a nonclassical HLA class I tissue-specific molecule that is highly expressed on extravillous cytotrophoblasts [1, 2]. A striking feature of HLA-G is the alternative splicing of its primary transcript (Fig. 1), giving rise to four membrane-bound (HLA-G1, -G2, -G3, and -G4) and three soluble (HLA-G5, -G6, and -G7) proteins (Fig. 2A). The full-length HLA-G1 mRNA encodes a protein that is associated with ß2-microglobulin and has classical HLA class I structure. The mRNA species encoding HLA-G2, -G3, and -G4 proteins are truncated by deleting exon 3 (
2 domain), exons 3 and 4 (
2 and
3 domains), and exon 4 (
3 domain), respectively. Soluble isoforms of HLA-G molecule are encoded by mRNAs containing introns that generate premature stop codons. Presence of intron 4 in both HLA-G5 (full-length) and HLA-G6 (minus exon 3) mRNAs results in the lack of transmembrane domains of the protein and, thus, permits synthesis of soluble forms of HLA-G1-like and HLA-G2-like molecules, respectively [35]. We recently identified a new alternatively spliced transcript, namely HLA-G7, that possesses a stop codon in intron 2, which gives rise to a soluble isoform with the
1 domain only (soluble HLA-G3-like) [6].
|
|
During the past few years, extensive studies have been carried out on the HLA-G1 protein isoform, for which specific antibodies are available. The HLA-G1 protein is able to inhibit allogeneic responses such as proliferative T lymphocyte cell response, cytotoxic T lymphocytes (CTL)-mediated cytolysis, and both decidua and peripheral blood natural killer (NK) cell-mediated cytolysis [7]. This inhibition is mediated through 1) direct interaction with inhibitory receptors such as ILT-2, KIR2DL4, or p49 [811] and 2) up-regulation of HLA-E cell-surface expression [12, 13], another nonclassical HLA class I molecule that interacts with the CD94NKG2A inhibitory receptor [14]. Thus, HLA-G1 protein expression on trophoblasts was proposed as contributing to immune privilege for the fetus. This biological role is supported by gestational pathologies such as preeclampsia, in which HLA-G expression is down-modulated and associated with placentation defects [15, 16]. The role of both HLA-G and -E molecules (recognized by the W6/32 monoclonal antibody on purified trophoblast cells) in protecting cytotrophoblasts against NK cytolysis has also been demonstrated in semiallogeneic combinations of maternal uterine NK cells toward their own trophoblast counterpart as well as in allogeneic combinations of maternal uterine NK cells toward trophoblast from different mothers [17]. Furthermore, King et al. [18] reported that killing by polyclonal decidual NK cells is inhibited by HLA-G and is reversed by HLA-G-specific antibody, which has no cross-reactivity with HLA-E [18]. Accordingly, using NK cells isolated from maternal decidua, Ponte et al. [19] showed that the lysis of LCL721.221 lymphoblastoid cell line transfected with HLA-G1 and coexpressing HLA-E was partly restored in the presence of anti-CD94 antibody preventing the interaction between HLA-E and its inhibitory receptor (i.e., CD94/NKG2A). However, the maximal restoration of lysis occurs in the presence of the anti-p49 antiserum, which blocks the binding of HLA-G to its specific inhibitory receptor (i.e., p49) [11]. More recently, the work of Fuzzi et al. [19] demonstrated the in vivo biological relevance of HLA-G secretion for successful implantation in humans. Indeed, those authors identified two groups of preimplantation embryos on the basis of soluble HLA-G in culture supernatant before transfer to infertile patients. They showed that after fertilization or intracytoplasmic sperm injection, only transfer of embryos from the group that secretes soluble HLA-G could lead to successful pregnancies [19].
The contribution of HLA-G in fetal-maternal tolerance is debated, however, because of recent reports that described healthy individuals who were homozygous for the HLA-G null allele, HLA-G*0105N, and who had been normally delivered [2022]. In these cases, the full-length HLA-G1 membrane-bound protein cannot be translated, because a single base-pair deletion of a cytosine at position 1597 (1597delC) in exon 3 (corresponding to the
2 domain) causes a frameshift mutation leading to the existence of a premature stop codon at the beginning of exon 4 (corresponding to the
3 domain) (Fig. 1). This deletion has been determined in a variety of ethnic groups and has been found with frequencies of 7.48% in African Americans, 4.8% in Ghanaians, 3% in Spaniards with association to the HLA-A30-B13 haplotype, 2.3% in mixed German-Croatians, 0.6% in Danish populations, and 11% in the Shona ethnic group of Zimbabwe [21, 2327]. The HLA-G*0105N allele was not found in Japanese or American Caucasian populations [21, 23]. Because of the presence of the 1597delC frameshift mutation, the HLA-G*0105N allele was called "null" based on its inability to generate the HLA-G1 protein, which is well known as the functional HLA-G isoform. This raises the question of whether HLA-G function is restricted to the HLA-G1 isoform.
As previously evoked by other authors [21, 22], the HLA-G*0105N allele could still produce the following HLA-G molecules (Fig. 2B): 1) HLA-G2, -G3, and -G6, which are normally generated by alternatively spliced transcripts lacking at least the disrupted exon 3; 2) HLA-G7, which is normally devoid of the exon 3-encoded
2 domain; and 3) atypical HLA-G proteins, which may possess a new amino-acid sequence in the second half of the
2 domain because of the frameshift deletion at codon 130. In this regard, Castro et al. [22] described that in homozygous HLA-G*0105N individuals, at least six splicing mRNAs (i.e., HLA-G1 to -G6) are produced. Furthermore, Ober et al. [21] showed that homozygous HLA-G*0105N placentas are stained by a monoclonal antibody specific for soluble HLA-G protein isoforms, suggesting that the soluble HLA-G2-like protein (i.e., HLA-G6) may be expressed. The possible implication of this soluble isoform in fetal-maternal tolerance is supported by a report proposing that HLA-G6 circulates in maternal blood [28]. Thus, among individuals in whom HLA-G1 expression is absent or altered, any of the other HLA-G isoforms may assume the protective function of HLA-G. Accordingly, we have recently described that HLA-G2 and -G3 proteins are expressed in cytotrophoblasts obtained from first-trimester terminations of normal pregnancies [29]. Although two reports have shown that the short forms of tagged HLA-G molecules do not reach the cell surface and, thus, may have no or limited function [30, 31], we recently described that after blockage of HLA-G1 with a specific antibody, the other HLA-G isoforms that are produced by HLA-G genomic transfected cells can act as inhibitory molecules toward NK function [29]. Furthermore, using untagged vector constructs transfected in an HLA-A-, -B-, -C-, and -E-positive melanoma cell line, we showed that the HLA-G2, -G3, and -G4 truncated isoforms were expressed as nonmature cell-surface glycoproteins [32], a feature that has already been reported for the HLA class I-like molecule CD1d [33]. More importantly, these truncated forms inhibit both innate (NK cells) and acquired (CTL) effectors [32]. Because HLA-G3 does not have
2 or
3 domains, the
1 domain may contain the functional region.
The apparent discrepancies between our results and those of previous studies probably may be explained by differences in the experimental procedures. In particular, the use of tagged transfected constructions may affect protein structure and lead to the elimination of misfolded proteins by cytoplasmic proteases. Another hypothesis is that an unknown chaperone protein that would be absent in the cellular models used by Brainbridge et al. [30] and Mallet et al. [31], namely JAR, J26, and C1R, is required for cell-surface expression of short HLA-G isoforms. In this regard, we propose that HLA-G2, -G3, and -G4 might be coexpressed at the cell surface with other HLA class I molecules. It is noteworthy that the mouse cytomegalovirus glycoprotein gp34 associates with folded class I major histocompatibility complex molecules and exhibits two endo-H-sensitive (immature) oligosaccharides at the cell surface [34].
Altogether, these findings support the hypothesis that at least HLA-G2 and -G3 truncated isoforms may substitute for HLA-G1 in homozygous HLA-G*0105N individuals in protecting the fetus from maternal NK attack. The leader peptide of HLA-G isoforms may also permit stable HLA-E cell-surface expression on trophoblasts [12, 35], constituting an indirect inhibitory pathway by blocking lysis of CD94/NKG2A+ decidua NK cells. In addition, we cannot exclude the possible functionality of the above-mentioned atypical HLA-G molecules exhibiting new amino sequence in the second half of the
2 domain, which, if they are produced, will possess an integral
1 domain. This domain is shared with all HLA-G isoforms. Consequently, in HLA-G*0105N homozygous individuals, the other HLA-G2, -G3, -G6, and -G7 isoforms may be normally produced and should be taken into consideration as contributing to immune privilege for the fetus.
| FOOTNOTES |
|---|
Received: 12 March 2002.
First decision: 1 April 2002.
Accepted: 5 June 2002.
| REFERENCES |
|---|
|
|
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |