| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
* University of Kansas Medical Center, Kansas City, Kansas, USA; and
The University of Chicago, Chicago, Illinois, USA
1 Correspondence: Department of Anatomy and Cell Biology, University of Kansas Medical Center, Kansas City, Kansas, 66160-7400 USA. E-mail: jhunt{at}kumc.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT BACKGROUNDHLA-G: A NOVEL GENE...FUNCTIONS OF HLA-GPROSPECTS AND PREDICTIONSREFERENCES |
|---|
Key Words: human • placenta • immune privilege
|
BACKGROUND |
|---|
|
TOPABSTRACTBACKGROUNDHLA-G: A NOVEL GENE...FUNCTIONS OF HLA-GPROSPECTS AND PREDICTIONSREFERENCES |
|---|
. The proposed mechanisms included physical separation of maternal and fetal tissues, lack of fetal antigens that could stimulate graft rejection, and development of tolerance. Today certain aspects of each of these strategies have been identified: the blood circulations of the fetus and the mother are entirely separate; in the fetus, transplantation antigens are late appearing; tolerance and immune privilege at the maternal-fetal interface are readily identified. Three major principles emerging from these studies are that 1) multiple mechanisms provide protection, 2) both the fetus and the mother contribute to development and maintenance of the pregnant uterus as an immune privileged site, and 3) fetal factors drive changes in maternal immune responses. In this article, we first briefly discuss a number of conditions responsible for immune privilege and maternal tolerance for which scientific evidence is strong, then focus on a central feature—a unique capacity of placental cells to select specific genes within the major histocompatibility complex (MHC) for expression. We present evidence that this unique capacity for selection of specific MHC antigens, which in humans are called human leukocyte antigens (HLA), may be responsible in large part for the reprogramming of local maternal immune responses that characterize successful semiallogeneic pregnancy.
Strategies for protecting the semiallogeneic fetus from maternal graft rejection responses
During pregnancy, the maternal immune system is clearly active, and under certain conditions may contribute to fetal damage/death. Well-defined pathological processes include destruction of fetal erythrocytes (Rh antigen, erythroblastosisis) and platelets (HPA-1 and -2, alloimmune thrombocytopenia) by maternal antibodies and infections of pregnancy, where activated macrophages secreting high levels of Th1-type cytokines alter the delicate cytokine balance at the maternal-fetal interface (2
, 3)
. Yet even with a demonstrably active maternal immune system, mothers usually seem to tolerate rather than reject their genetically disparate fetuses. Ordinarily, the mother would be expected to generate graft-attacking antibodies and cytotoxic T lymphocytes (CTL) to foreign (paternal) HLA or other antigens expressed by fetal cells. HLA antigens are called "transplantation" antigens because they comprise the most powerful stimulators of graft rejection. Thus, in organ transplantation the matching of certain donor and patient alleles is an absolute requirement for successful grafting.
Even though mothers and fathers are almost invariably disparate at multiple HLA loci, rejection of the fetus as a consequence of maternal recognition of paternal HLA as foreign is essentially undocumented. Although anti-paternal HLA antibodies are common in pregnant women, they do no damage. Even novel HLA antigens expressed in the fetal membranes are tolerogenic rather than immunogenic (4)
. Attack by maternal CTL is also effectively thwarted, as described in detail below. Thus, mechanisms underlying maternal tolerance are unusually effective and raise the critical question of how immune privilege might be established in the natural situation of pregnancy so as to assure viability of the embryo/fetus.
Figure 1
illustrates the principle that maternal and fetal processes contribute to the generation of a safe environment. Within the uterus, a dramatic change in endometrial leukocyte subpopulations occurs as a consequence of implantation. After a brief inflammatory reaction caused by blastocyst breaching of the uterine epithelium, a reaction best documented in rodents, the altered endometrium (now termed decidua) settles into a pattern where local protection is provided by the innate immune system. From this point onward, the major players in acquired immunity, T and B lymphocytes, are identified mainly in the myometrium distal to fetal tissues whereas members of the innate immune system, natural killer (NK) cells and macrophages, predominate in the decidua (5
, 6)
. An exception is the Treg subset of CD4+/CD25+ cells, which produce interleukin-10 (IL-10) and transforming growth factor-ß1 (TGF-ß1), and are believed to be critical to maintenance of tolerance (7
, 8)
. These cells, whose proliferation is stimulated by estrogen (9)
comprise 
14% of CD4+ cells in early decidua (10)
. Uterine cytokine networks, while of great complexity and containing many examples of Th1-type reactivities (11
, 12)
, were first shown in the mouse to be dominated by these and other Th2-type cytokines (13)
. Similar reports have appeared in human pregnancy (14)
. Immunomodulatory hormones such as prolactin, chorionic gonadotropin, and progesterone are abundant; chemokines operate to control immune cell numbers and types.
|
The fetal contributions are unique. Figure 2
shows that the fetus, which is derived from the inner cell mass of the blastocyst, is secluded within a protective shell composed of trophoblast cells that arise from the trophectoderm layer of the blastocyst. It is therefore the heavy responsibility of the trophoblast cells to interact appropriately with the maternal environment and protect the fetus from maternal immune attack. As illustrated in Fig. 1
, the trophoblast cells circumvent antibody-mediated damage by exhibiting high levels of the complement regulatory proteins (15)
, and reduce cell-mediated immunity by expressing inhibitory members of the B7 family (16)
, and apoptosis-inducing members of the tumor necrosis factor (TNF) family of ligands (17)
. As with cells in the decidua, fetal cells produce immunosuppressive cytokines, chemokines, and prostaglandins that dampen T lymphocyte proliferation and export high levels of immune suppressive hormones such as progesterone.
|
Most important, trophoblast cells strictly regulate their expression of HLA genes and the production of their proteins. It is these proteins that, if recognized as foreign by maternal immune cells, would stimulate maternal anti-fetal CTL capable of destroying HLA-expressing fetal cells (18)
. Instead, the antigens expressed in trophoblast cells program maternal leukocytes into pathways consistent with tolerance.
|
HLA-G: A NOVEL GENE EXPRESSED IN PLACENTAS |
|---|
|
TOPABSTRACTBACKGROUNDHLA-G: A NOVEL GENE...FUNCTIONS OF HLA-GPROSPECTS AND PREDICTIONSREFERENCES |
|---|
20 to 25 HLA class I genes, relatively few are transcribed or translated; most are pseudogenes or gene fragments. The expressed class I genes are subdivided into class Ia, which includes HLA-A, -B, and -C, and class Ib, which includes HLA-E, -F, and -G. HLA class II (HLA-D) genes, if transcribed, are not translated in human trophoblast cells even under inducing conditions (19)
and will not be discussed further in this review. One remarkable difference between the HLA class Ia and Ib genes is that the former are highly polymorphic, with many alleles, and the latter have few variants. Major differences have been observed in glycoproteins associated with these two subsets of class I antigens. In general, class Ia antigens are membrane bound. By contrast, one member of the class Ib group, HLA-G, is alternatively spliced. Seven alternatively spliced transcripts have been identified, of which four are predicted to encode membrane bound and three are predicted to encode soluble proteins. A final difference is that expression of class Ia antigens is ubiquitous whereas expression of class Ib antigens may be tissue/organ-specific and/or conditional.
Human trophoblast cells express one class Ia molecule (HLA-C) and all three class Ib molecules. The HLA-C gene is moderately polymorphic, and could stimulate maternal anti-fetal acquired immunity if paternal alleles differed from maternal. Yet allelic disparity at the HLA-C locus does not seem to be a causal factor in infertility or termination of pregnancy. The other HLA class I antigens expressed by trophoblast cells are HLA-E, -F, and -G. These class Ib antigens are distinguished by low numbers of alleles that differ at the protein level. For example HLA-E has 2 alleles (20)
and HLA-G has five alleles (reviewed in ref 21
). There are no allelic variants of HLA-F in the published literature, although two nonsynonymous (amino acid) substitutions are reported in online databases (http://genome.ucsc.edu), which are likely rare in the general population. Moreover, most polymorphisms in the HLA-G gene do not alter the amino acid sequence; the few that do are not predicted to change secondary structures of the heavy chains (4
, 21)
.
Of the HLA class Ib molecules expressed by trophoblast cells, HLA-G was the first to be identified and remains an antigen of great interest and a focus of experimental evaluation (reviewed in refs 22
23
24
).
Structural features of HLA-G
Although the genomic structure of HLA-G is similar to other class I genes, it is unique in most other respects. The HLA-G gene has eight exons encoding a signal peptide (exon 1), the 
1, 2, and 3 domains (exons 2, 3, and 4, respectively), the transmembrane domain (exon 5), and the intracellular domain (exons 6 and 7), similar to other class I genes (Fig. 3
). However, a premature stop codon in exon 6 results in a truncated cytoplasmic tail that reveals a cryptic retrieval motif (25)
. This results in the slower turnover and prolonged expression of HLA-G at the cell surface, and possibly the inefficient presentation of exogenous peptides. Park and colleagues interpret this as evidence that the primary function of HLA-G is not antigen presentation but as an inhibitory ligand for NK cells (25)
.
|
A second unique feature of HLA-G is that it encodes multiple isoforms as a result of alternative splicing. Five of the seven transcripts that result from alternative splicing are shown in Fig. 3
. The full-length isoform HLA-G1 is structurally similar to other class I genes, except for the truncated cytoplasmic tail. The G2 isoform results from the removal of exon 3 and homodimerizes to form an HLA class II-like structure (26
, 27)
. These two isoforms are expressed as soluble proteins (HLA-G5 and -G6, respectively) due to the inclusion of intron 4 sequences in the mature mRNA, resulting in secreted proteins with an additional 21 amino acids (encoded by intron 4 sequences) following the 3 domain (28)
. HLA-G3 results from the removal of exons 3 and 4. HLA-G4 and -G7 (not shown in Fig. 3
) mRNAs are not abundant in placentas. Exon 4 (encoding the 3 domain) is spliced out of the HLA-G4 transcript; the HLA-G7 transcript includes exon 2 and part of intron 2 and is predicted to encode a small soluble isoform.
Compared with the classical class I genes, the most polymorphic genes in the human genome, HLA-G has relatively little polymorphism in its coding region. Figure 3
shows the location of the 13 polymorphisms in exons 1–4 that have been identified to date and one in the 3'UTR. Polymorphisms at codon 31 in the 1 domain (Thr
Ser), at codon 110 in the 2 domain (LeuIle), and at codon 258 in the 3 domain (ThrMet) result in an amino acid substitution; polymorphisms at nucleotide +15 and +36 in exon 1, at codons 35, 57, and 69 in exon 2, at codons 93, 100, and 107 in exon 3, and at codon 188 in exon 4 do not alter the amino acid sequence of the protein. A single base pair (bp) deletion at nucleotide 1597 causes a frameshift at amino acid 130 (29)
, resulting in nonfunctional HLA-G1 and -G5 proteins (30)
. The polymorphisms that alter the protein sequence define five alleles, called G*0101, G*0103, G*0104, G*0105N, and G*0106. Silent variation within these allelic classes defines subtypes, referred to as G*010101, G*010102, etc. A 14 bp insertion/deletion polymorphism is present in the untranslated exon 8.
In contrast to the class Ia HLA loci, amino acid substitutions at codons 31 and 110 in the 1 and 2 domains, respectively, are conservative changes that occur in residues that are not predicted to interact with bound peptide or T cell receptor (21)
. The third amino acid polymorphism at codon 258 is a nonconservative substitution in the 3 domain that is highly conserved in the class Ia genes (31)
. It is located in the pleated sheet structure of the 3 domain, where it might affect recognition of CD8 in the HLA-G1 and -G5 isoforms, or binding to CD4 in the HLA-G2 and -G6 isoforms (4)
. Last, a polymorphic 1 bp deletion of a cytosine (C) residue at codon 130 results in a null allele (called G*0105N), which does not encode functional HLA-G1 or -G5 protein isoforms (30)
. This mutation, called 1597
C, occurs in the homozygous form in ostensibly healthy individuals, indicating that HLA-G1 and -G5 isoforms are not essential for fetal survival (30
, 32)
. In these situations other isoforms presumably suffice (33)
. However, this null allele has been associated with increased risk for recurrent miscarriage (34
, 35)
, suggesting that HLA-G1 and/or -G5 proteins do indeed play an important role in the maintenance of pregnancy and that reduced levels of one or both is a risk factor for recurrent miscarriage.
A 14 bp insertion/deletion polymorphism in the untranslated exon 8 was first described by Harrison and colleagues (36)
, but has recently been shown to influence mRNA transcript size and stability. The presence of the 14 bp insertion allele generates a 92 bp deletion in the 3'UTR of the G*01012 and G*01013 mRNAs, possibly because it acts as a cryptic splice site (37)
. Transcripts with the 92 bp deletion were associated with more stable mRNA in JEG-3 cells (which are homozygous for G*01013) and in an M8 cell line transfected with G*01012, perhaps because they were less susceptible to degradation (38)
. Further, the relative abundance of the alternatively spliced transcripts may be influenced by polymorphisms in HLA-G (39)
. For example, studying mRNA from term trophoblast cells, Hviid et al. (39)
showed that heterozygotes for the G*01012 allele (but not the G*01013 allele) had reduced levels of transcripts encoding membrane-bound isoforms (G1, G2, G3) with the 14 bp insertion, whereas heterozygotes for the G*01013 allele (but not for the G*01012 allele) had higher levels of the G2/G4 transcripts than the G*01011 allele (39)
. Although the significance of these findings is not clear, it is noteworthy that a number of studies have demonstrated reduced expression of HLA-G mRNA or protein in term placentas of pre-eclamptic pregnancies (40
41
42
43)
. The lack of association between the G*0105N null allele and pre-eclampsia in one study suggested that deficiencies of the HLA-G1 and/or -G5 isoforms were not a risk factor for pre-eclampsia (44)
, but reduced expression of other isoforms, such as the short G3 transcript, may influence risk (45)
. Indeed, the relative abundance of transcripts and noncoding polymorphisms in HLA-G have been associated with pre-eclampsia (45)
, suggesting that the regulation of expression of HLA-G is influenced by genetic variation.
This suggestion is further supported by the recent discovery of variation in the 5'-upstream regulatory region of HLA-G (46
, 47)
. In contrast to the limited polymorphism in the exons, the upstream region containing all of the known promoter and regulatory elements is extraordinarily polymorphic (Fig. 4
). To date there have been 18 polymorphisms identified in the 1300 bp upstream from exon 1. Most are very common polymorphisms with minor allele frequencies (>20%), and many reside within or close to known transcription factor binding sites (Fig. 4)
. These polymorphisms define at least eight unique haplotypes (47)
. Some promoter region haplotypes are shared among alleles that differ in their coding regions (for example, G*01012, G*0105N, and G*01061 have identical promoter region sequences), while some alleles with identical coding regions have different promoter region haplotypes (e.g., the common G*01011 allele is associated with at least three different promoter region sequences). One variant, –725G, changes the methylation status of a CpG dinucleotide in the promoter region and has been associated with an increased risk for sporadic miscarriage in an unselected sample of healthy women, which suggest at least some of these polymorphisms may influence transcription and mRNA abundance (47)
.
|
Patterns of expression of HLA-G
Figure 2
illustrates that placentas contain several distinct subpopulations of trophoblast cells that arise from progenitor cells within the trophectoderm layer of the blastocyst. These progenitor cells may merge to form a syncytialized single cell layer that interfaces directly with maternal blood or proliferate to form columns of extravillous cytotrophoblast cells that contact and infiltrate the decidua (upper insert, Fig. 2
). The migrating cells anchor the placenta to the decidua and infiltrate maternal spiral arteries, thus facilitating blood flow to the placenta. The extravillous cytotrophoblast cells ultimately regress to form the chorion membrane (lower insert,Fig. 2
).
HLA class I antigen expression was first identified (and usually is described) as being restricted to the extravillous trophoblast cell population, with proteins particularly prominent in cells immediately adjacent to the decidua in both early and late gestation placentas (reviewed in refs 22
23
24
). However, the reagent used most frequently to identify these antigens in tissue sections or isolated cells, the mouse monoclonal antibody W6/32, requires light chain (ß2m)/heavy chain association, so neither free heavy chains nor isoforms such as HLA-G2 and -G6 that do not associate with ß2-microglobulin (ß2m) were detected. More recent studies using monoclonal antibodies that identify the amino acid sequence derived from intron 4 nucleotides—16G1 generated by D. Geraghty being a good example—have supplied localization data on HLA-G5 and -G6 heavy chains. New isoform-specific antibodies show that HLA-G5 is present throughout the placenta and within the chorion membrane, decidua, and maternal blood (48)
. In villous CTB cells, HLA-G5 is likely to be mainly free heavy chain (J. S. Hunt and P. Morales, unpublished results) as ß2m is missing. An antibody recognizing HLA-G2 and -G6 shows that one or another of these isoforms is prominent in/on extravillous cytotrophoblast cells distal to the placental villi, cytotrophoblast cells infiltrating the decidua, and some chorion membrane cytotrophoblast cells (48)
(Fig. 5
). HLA-G2/G6 is located in the same cells as HLA-G1 (43)
, suggesting that in women carrying the null allele (G*0105N), which does not encode functional HLA-G1 or -G5 protein isoforms, HLA-G2/G6 may comprise an adequate substitute (30)
.
|
Receptors for HLA-G
The original view of how regulation of the HLA class I antigens in fetal trophoblast cells might contribute to maternal tolerance during pregnancy was restricted to the negative: a failure to express these antigens meant that T lymphocytes would not recognize trophoblast cells as foreign, thus allowing them to survive (49)
. The finding that subpopulations of trophoblast cells invading the decidua express HLA class I antigens (50)
drove investigators to revise their thinking and explore new explanations for placental immune privilege.
Binding of trophoblast cell HLA-G to receptors that inhibit activating signals on decidual leukocytes may be the answer (Fig. 6
). HLA-G appears to be recognized mainly by immunoglobulin-like transcript (ILT) receptors, which are expressed by T and B lymphocytes, as well as by NK cells and mononuclear phagocytes (51)
, and abrogate activating signals received by these cells. Whether these ILT receptors or others, such as TcR and the CD94/NKG2A HLA-E receptors on uterine CD56bright NK cells, are more important to lymphocyte recognition remains to be established. Early studies suggested that ILT4 may be the main receptor for HLA-G exhibited by monocyte/macrophages, the second most populous leukocyte population in the human decidua (6)
. However, ILT2 has not been formally excluded as a mononuclear phagocyte recognition entity, and recent experiments in a macrophage cell line suggest these two receptors may exhibit isoform-specific binding (52)
. This novel finding illustrates the point that little is known of how HLA-G-activated signaling pathways, which for ILT2/ILT4 include tyrosine phosphorylation, SHP-1 association, and calcium regulation, may be translated into expression of specific genes in decidual leukocytes as required for programming the cells into pregnancy-appropriate behavior.
|
|
FUNCTIONS OF HLA-G |
|---|
|
TOPABSTRACTBACKGROUNDHLA-G: A NOVEL GENE...FUNCTIONS OF HLA-GPROSPECTS AND PREDICTIONSREFERENCES |
|---|
, recent studies using HLA-G proteins from transfected cells indicate that these proteins may regulate immune cells and thus be integral to immune privilege in pregnancy. Figure 6
shows that HLA-G proteins probably target all of the major immune cell subsets. In the following paragraphs we describe activities related to T and B lymphocytes, NK cells and antigen-presenting cells (APC).
HLA-G interactions with T lymphocytes
Wegmann and co-workers were the first to report that in pregnant mice, Th2 cytokine-producing lymphocytes flourish in preference to those producing Th1 (13)
. Subsequent research in women has shown that failure to achieve this preference for Th2-type, anti-inflammatory cytokines may lead to unsuccessful pregnancy (54)
. The idea that trophoblast cell signals drive T cells into this anti-inflammatory profile has emerged as a popular explanation for maternal tolerance to the semiallogeneic fetus.
Convincing evidence for an ability of HLA-G to influence T cells was first presented by Sanders et al. (55)
, who showed that HLA-G-expressing cells bind to CD8-expressing cells. This finding has recently been confirmed by Shiroishi et al. (56)
. In both studies, HLA-G binding to the CD8 homodimer, the molecular form expressed by a subset of T cells in the intestine and by NK cells (57)
, was evaluated. Yet most T cells express the CD8ß heterodimer, which acts as a coreceptor to the T cell receptor (TcR) and an essential signal transduction molecule during T cell activation. Interaction between the TcR and HLA-G has not been demonstrated. Many investigators question whether binding to the TcR would be a primary function of HLA-G because of its limited polymorphism (58)
. However, it was recently shown that cytomegalovirus-derived peptides can stabilize surface HLA-G expression, leading to HLA-G-restricted, cytomegalovirus-specific cytotoxic T cell response in transgenic mice (59)
. This strongly suggests that HLA-G can indeed act as a classical presenter of foreign peptide. In solving the debate of whether there is any role for cooperative physiological binding between HLA-G and the ßTcR, it will be important to analyze HLA-G-derived peptides from women with known human cytomegalovirus infection.
Functional studies have yielded even stronger evidence. In vitro studies have clearly demonstrated an ability of soluble and membrane-associated HLA-G to modulate cytokine release from human allogeneic peripheral blood mononuclear cells (60)
and to have a concentration-dependent effect on generation of an allogeneic CTL response (61)
. Regarding initiation of the acquired immune response, Le Maoult et al. (62)
have shown that APC transfected with HLA-G1 prevent proliferation of CD4+ T cells and direct them toward an immunosuppressive phenotype. Although Le Maoult et al. did not test CD4+ cells to ascertain whether they might be related to the CD4+CD25bright T regulatory phenotype apparently essential for pregnancy in mice (63)
and possibly in women (64)
, it is tempting to speculate this might be the case. The idea that this phenotype might be the consequence of exposure to a murine Qa-2, which bears structural similarities to HLA-G (65)
, is intriguing.
Membrane-associated and/or soluble HLA-G may play a critical role in regulating CD8+ T cells during pregnancy by eliminating alloreactive (antipaternal) T cells. Two independent groups have reported that HLA-G induces CD8+ T cell apoptosis (66
67
68)
. In both models, exposure of CD8+ T cells to soluble HLA-G was shown to trigger surface expression and secretion of Fas ligand, resulting in death of the activated T cells through the Fas/FasL pathway. These experiments were based on the well-documented ability of other HLA antigens to induce Fas/FasL-related cell death, a phenomenon that has been identified in patients undergoing transplantation and is believed to dispose of alloreactive CD8+ cells (69)
. In pregnancy, soluble HLA-G might induce apoptosis in CD8+ cells that react to paternal antigens; soluble HLA-G has been reported in maternal sera by several groups (70
, 71)
.
A second mechanism by which HLA-G may induce maternal tolerance to fetal antigens is to reduce or prevent cytotoxic activity of CD8+ T cells against target cells, possibly independent of inducing T cell apoptosis. The specific receptor capable of binding HLA-G is not clearly defined, but could be either the CD8 molecule itself, as proposed by Contini et al. (66)
and Shiroishi et al. (56)
, or the immunoglobulin-like transcript receptor 2 (ILT2), which transduces an inhibitory signal (61)
(Fig. 6)
. HLA-G1 and its soluble counterpart HLA-G5 protect potential target cells from lysis by antigen-specific cytotoxic T cells (61
, 66
, 72)
. The same protection by the smaller isoforms of membrane HLA-G, HLA-G2, -G3, and -G4 has now been shown (73)
, which is consistent with the idea that smaller isoforms compensate when HLA-G1 and -G5 are absent in mothers who are homozygous for the G*0105N allele (74)
.
Intriguingly, HLA-G down-regulates expression of CD8 mRNA and protein in IFN-
-treated blood mononuclear cells without inducing apoptosis or altering CD3 expression (48)
. Although T cells and NK cells express CD8/ß, signals for CD8 specific message and protein in Northern blots and immunoblots were strong, suggesting it is the more numerous T cell affected. Because the effects of HLA-G are highly concentration dependent and HLA-G-producing cells are in the placental bed, these results imply that soluble isoforms of placental HLA-G might reduce the ability of T cells to function effectively in the pregnant uterus but be less potent in the periphery (48
, 71)
.
When considering the impact of HLA-G on T cells, it is critical to bear in mind two facts: 1) T cells are not numerous in the decidua (6)
, and 2) trophoblast cells in normal, healthy placentas are unlikely to elicit a CTL cell response. The inability of trophoblast cells to stimulate CTL is due to the fact that polymorphic HLA class I antigens are absent on syncytiotrophoblast and are scarce on other subpopulations. However, an exception might occur in the case of virally infected extravillous trophoblast cells, where trophoblast cell-associated HLAs might present antigen to CD8+ cells, thus becoming potential targets themselves.
HLA-G interactions with B lymphocytes
ILTs appear to be major receptors on T lymphocytes and antigen-presenting cells that interact with HLA-G, as described above. They are also present on B lymphocytes, which express ILT2 inhibitory receptors (75)
. So far there is no evidence for binding of HLA-G to these receptors on B cells or of B cells responding directly to HLA-G; yet the possibility exists that this occurs, because production of antibodies to placental HLA-G occasionally occurs in pregnant women.
A recent study in our laboratories demonstrated that maternal tolerance to HLA-G in terms of antibody production is the usual condition (4)
. Ninety-one percent of mothers, as well as all women who had never been pregnant and all men, lack antibodies to HLA-G in their sera. Yet tolerance is not absolute: 9% of women who had undergone at least one pregnancy generated anti-HLA-G antibodies that were readily identified in maternal sera by ELISA and immunoblotting. Whether a "mimic" antigen on microbes or other environmental molecules might stimulate these antibodies is not known, but since the anti-HLA-G antibodies were found only in women who had been pregnant, some association with the state of pregnancy would be required. Not unexpectedly, maternal anti-HLA-G antibodies have no deleterious effect; all the women who developed these antibodies had multiple successful pregnancies. In this respect, circulating anti-HLA-G antibodies resemble other anti-HLA antibodies, which are common in pregnancy and do not damage the developing fetus. Because several HLA-G alleles have been identified, we tested the hypothesis that mothers would generate antibodies to foreign (paternal) alleles, but found no relationship between exposure to a foreign (paternal) HLA-G protein on placental cells and maternal production of anti-HLA-G (4)
.
How tolerance is achieved remains to be elucidated. Potential routes include 1) activation of B cell ILT2 receptors by circulating HLA-G5 or -G6, an idea that remains untested, and 2) activation of ILT2 receptors on Th lymphocytes, as suggested by Le Maoult and colleagues (62)
. The potential role of APC driven into a B lymphocyte-inhibiting immunosuppressive profile should not be overlooked, as these cells are abundant at the maternal-fetal interface where HLA-G isoforms are prominent. Alternatively, HLA-G produced in the placenta during fetal life might well induce lifelong tolerance in the adult.
In any event, data accumulated to date, though scarce, strongly support the idea that inhibition is specific to HLA-G and that the tolerogenic mechanisms underlying specific repression of production of maternal antibodies are immensely effective. The deficiency of anti-HLA-G antibodies is not due to maternal, pregnancy-associated reduction in B cell function. Mothers produce high levels of other types of antibodies in order to provide defense against pathogens during pregnancy and assure transfer of protective antibodies to the fetus. The placenta supports this hyperproduction. A recent study from our laboratory shows that human placentas produce two non-apoptosis-inducing TNF superfamily ligands: BAFF (also known as BlyS, or B lymphocyte stimulator, as well as TALL-1, THANK, and zTNF4) and APRIL (a proliferation inducing ligand), both of which sustain B lymphocytes (76)
. Placental BAFF and APRIL therefore may bear a degree of responsibility for maternal host defense and provision of antibodies to the fetus required in the postpartum period.
HLA-G interactions with NK cells
A complete absence of HLA class I molecules on trophoblast cells could provide protection against CTL but could target the trophoblast subpopulation migrating into the decidua (Fig. 2
, upper insert) for destruction by resident NK cells, a leukocyte programmed to recognize and destroy HLA null cells. Expression of HLA class Ib genes HLA-E, -F and -G may be an evolutionary move to avoid this risk. As noted above, these genes have few alleles and their products are unlikely to be recognized as foreign by the mother.
NK cells of an unusual phenotype, CD16–CD56bright, are abundant in first and second trimester decidua but decline in number thereafter (6)
. These cells are poor killers of the usual NK targets, suggesting that environmental conditions in the pregnant uterus affect their functions. Decidual NK cell cytotoxicity could be bypassed by ligation of HLA class I antigens to one or more NK cell surface inhibitory receptor. Transfection-based assays initially suggested that interaction between HLA-G and the CD94/NKG2A heterodimer on the surface of NK cells prevented cytolysis induced by NK cells (77
78
79)
. Further investigation, however, led to the realization that class Ib gene HLA-E utilizes a leader peptide derived from other class I proteins, including HLA-G, to stabilize its own surface expression. Thus, initial findings led to the erroneous conclusion that HLA-G was responsible for reducing NK-mediated cytolysis via CD94/NKG2A, when in fact HLA-G expression only permitted expression of HLA-E, the true ligand for this inhibitory receptor (80
, 81)
.
Nonetheless, the idea that HLA-G serves as the major inhibitor of NK toxicity at the maternal fetal interface remains popular because experiments using HLA-G-specific antibodies show that trophoblast cell HLA-G inhibits NK-mediated cell death in the absence of HLA-E (82
, 83)
. These effects could be mediated through the killer inhibitory receptor (KIR) 2DL4, through ILT2, or both (75
, 84
, 85)
. ILT2 has been shown to bind HLA-G and is expressed by decidual NK cells, albeit the quantity of ILT2 on decidual NK cells may be low (86
87
88
89)
. More controversial is a role for KIR2DL4, which in soluble form binds to HLA-G-transfected cells (84
, 85)
. More recently, however, biochemical analysis of KIR2DL4 binding to HLA-G monomers and dimers yielded only negative results, raising the question of validity of HLA-G-KIR2DL4 interaction (90)
. Furthermore, decidual NK cells appear to express an abundance of CD94/NKG2A, the receptor for HLA-E, suggesting this interaction may supercede that of HLA-G and its NK-expressed receptors (91)
. A final confounding observation against a role for HLA-G-mediated inhibition of NK cytolytic activity is that immortalized and primary trophoblast cells are resistant to NK cell-mediated lysis independent of HLA class I expression (91
92
93
94)
, suggesting these cells use other strategies for self protection.
New information has recently surfaced regarding a potential role for HLA-G/NK cell interactions at the maternal-fetal interface. KIR2DL4 ligation and treatment of NK cells with HLA-G result in production of the cytokine interferon- (IFN-) (95
96
97)
but, as suggested above, it remains to be determined whether this commonality is causally linked. Even so, stimulation of NK cell production of IFN- at the maternal-fetal interface by HLA-G is an intriguing possibility. IFN-, usually considered a proinflammatory cytokine, can drive cells into immunosuppressive profiles when linked with other modulators, and could therefore paradoxically serve as an anti-inflammatory cytokine. Whether HLA-G might stimulate IFN--associated pathways of vascular and decidual remodeling (98
99
100)
remains an intriguing possibility.
Interactions with antigen-presenting cells
Two populations of APC, macrophages and dendritic cells, reside in the human decidualized endometrium throughout pregnancy. These powerful, multifunctional leukocytes are located in close proximity to invasive cytotrophoblast cells, uterine glandular epithelium, and uterine blood vessels (6
, 101
102
103
104)
, and are proposed to play central roles in uterine and placental homeostasis as well as immune modulation (2
, 6
, 101
102
103
104
105)
.
APC in pregnant endometrium include 1) CD14+ macrophages (106
, 107)
, 2) CD83+ mature dendritic cells (103
, 104
, 108)
, and 3) CD83– immature macrophage/dendritic cells. The immature cell type appears to be a family of related cell types, some of which are dendritic cell (DC)-SIGN+CD14+ (109)
and others are DC-SIGN–CD14–DEC-205+ (104)
. Alternatively, the immature phenotype cluster may be composed of similar cells in different transitional stages. It has been suggested that DC-SIGN+ cells are a precursor population capable of differentiating into macrophages or dendritic cells (110)
, an idea supported by the finding that macrophages and immature dendritic cells have been shown to trans-differentiate in vitro under the influence of various cytokines (111)
.
Decidual macrophages, which are activated in the pregnant uterus as evidenced by expression of HLA class II, CD11c, and CD86 antigens (112)
, appear to be programmed for immunosuppression. These cells produce the lymphocyte inhibitory molecule, prostaglandin E2 (113
, 114)
, elicit reduced allogeneic and autologous T cell responses in comparison with monocytes (105
, 115
, 116)
, and spontaneously secrete anti-inflammatory cytokines such as interleukin (IL)-10 and transforming growth factor (TGF)-ß1 (52
, 114
, 117)
. More recently, decidual macrophages have been reported to express B7-H1 (16)
, ILT3 (117)
, DC-SIGN (104
, 110
, 118)
, MS-1, and factor 13 (119)
, all of which are markers associated with immune evasion and activation of macrophages into a suppressive profile. Mature CD83+ decidual dendritic cells appear to exhibit an immune cell inhibitory profile. These cells secrete less IL-12 than monocyte-derived dendritic cells and induce Th2 cells when cocultured with naïve CD4+ T cells. The C-type lectin, DC-SIGN, is used for immune evasion by several viral and bacterial pathogens, with ligand binding to DC-SIGN altering cytokine production and antigen presentation to the benefit of the pathogen (120)
. Thus, this molecule may be of importance to immunosuppression by the decidual CD83–/DC-SIGN+ macrophage/dendritic cell-like cells.
The question then arises as to how decidual APC are driven into immune inhibitory profiles. Placental HLA-G is a reasonable possibility. Although decidual dendritic cells have not been tested, mononuclear phagocytes and decidual macrophages express the two well-described true receptors for HLA-G: ILT2 and ILT4 (75
, 121
, 122)
. The decidual dendritic cells are likely to exhibit these receptors; studies in mice show that HLA-G tetramers bind only to dendritic cells (123
, 124)
. The ability of the mice to reject grafts is strikingly diminished in the presence of HLA-G tetramers. In humans, an immune suppressive effect of HLA-G on APC or other immune cell function is implied by the finding that high levels of HLA-G proteins in the sera of heart transplant recipients is positively associated with prolonged graft survival (125)
.
The development of recombinant HLA-G5 (rG5) and rG6 from eukaryotic cells (48)
has permitted us to test the hypothesis that HLA-G drives mononuclear phagocytes into an immune cell inhibitory profile. In contrast to results of another group who reported that G5 killed T cells via a Fas/FasL pathway (68)
, we found that our rG5 and rG6 did not negatively affect blood mononuclear cell viability (48)
or the viability of APC (52)
. This result is in accord with a study of monocyte-derived dendritic cells indicating that HLA-G does not decrease the viability of immature or mature dendritic cells, alter differentiation of dendritic cells from blood monocytes, or influence maturation of dendritic cells (126)
. Recombinant G5 and rG6 drive mononuclear phagocytes into suppressive pathways. In particular, both strongly stimulate TGF-ß1 production by activated APC. The condition of activation is extremely important, as clearly illustrated in the study by Morales et al. (48)
, where rG5 and rG6 reduced CD8 mRNA and protein in cells treated with the monocyte-activating factor IFN- to levels the same as, not below, those of CD8 in unstimulated cells.
Do APC themselves, when activated, express HLA-G? This possibility was first suggested in a report from our laboratory indicating that IFN--activated, but not resting, mononuclear phagocytes contain HLA-G mRNA and protein (127)
. The idea that inflammatory conditions stimulate APC production of HLA-G has been supported by numerous reports of HLA-G expression in macrophages and dendritic cells during cytomegalovirus infection, lung carcinoma, nontumoral pulmonary disease, and HIV infection (128
129
130)
. Although immunohistochemical studies may be compromised by the use of antibodies that lack specificity, soluble isoforms of HLA-G have been identified by Western blot in the sera of male heart transplant patients (125)
. There are indications that APC-derived HLA-G is functional. LeMaoult et al. (62)
showed that HLA-G1-transfected myelomonocytic cell lines (KG1a and U937) suppress CD4+ T cell proliferation in a mixed lymphocyte reaction, with the resultant T cells driven into an immune-suppressive phenotype. The converse idea that resting APC are not producers of HLA-G is supported by a report from Laupeze et al., who were unable to detect HLA-G expression in/on resting mononuclear phagocytes or dendritic cells (131)
. It is therefore puzzling that activated APC in the decidua reportedly fail to exhibit membrane or soluble HLA-G and do not contain a specific message (132)
. More work is needed to understand local tissue conditions that might direct APC into production of HLA-G, as activation is clearly not the only important factor.
|
PROSPECTS AND PREDICTIONS |
|---|
|
TOPABSTRACTBACKGROUNDHLA-G: A NOVEL GENE...FUNCTIONS OF HLA-GPROSPECTS AND PREDICTIONSREFERENCES |
|---|
Studying HLA-G structure-function relationships reminds us of several unifying principles of immunity, including 1) the genetic variation provided by alleles, 2) the potential of polymorphisms outside the coding region to influence rates of transcription and regulate production consistent with varying conditions and specific needs, 3) the power of alternative splicing, which has been revealed in other systems to be directly related to functional diversity (133)
, 4) the central relationship between the concentration of an immunomodulatory molecule and its final effects, and 5) the precise, possibly cell type-specific conditions such as activation and inflammation that may be required for stimulating production of the molecule and achieving desired effects.
The major HLA-G function that has emerged to date is its ability to program cells into immunosuppressive phenotypes. Since HLA-G is produced in great amounts at the maternal-fetal interface, the implication is that the immunosuppressive function must be important. Yet definitive proof that HLA-G is required remains elusive, as in vivo experiments are difficult to design. Mice do not demonstrate a clear correlate of HLA-G, although Qa-2 appears to have some of the same structural characteristics as HLA-G and is strongly expressed in placentas (65)
. Nonhuman primate placental cells transcribe and translate proteins from HLA-G-like genes, but the products are not identical to those described above for HLA-G. For example, in baboons, no HLA-G6-like isoform is detectable (134)
. Thus, although all indications are that HLA-G serves as a key effector of the changes that occur in maternal immune cells with the onset of pregnancy, definitive in vivo studies remain to be done.
Assessment of levels of HLA-G isoforms may well be useful in predicting the ability of mothers to establish and maintain pregnancy (71
, 135
, 136)
as well as determining the likelihood that patients will accept allografts (125)
. If such assessments are ultimately found to be reliable, recombinant HLA-G might be a powerful tool for inducing a desirable state of local immune suppression so as to facilitate pregnancy or graft acceptance, as has been shown in the porcine model (137)
.
|
FOOTNOTES |
|---|
Received for publication September 7, 2004. Accepted for publication December 22, 2004.
|
REFERENCES |
|---|
|
TOPABSTRACTBACKGROUNDHLA-G: A NOVEL GENE...FUNCTIONS OF HLA-GPROSPECTS AND PREDICTIONSREFERENCES |
|---|
