Human Reproduction Update Advance Access originally published online on November 9, 2005
Human Reproduction Update 2006 12(3):209-232; doi:10.1093/humupd/dmi048
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
HLA-G in human reproduction: aspects of genetics, function and pregnancy complications
Department of Clinical Biochemistry, Copenhagen University Hospital, Rigshospitalet and H:S Hvidovre Hospital, Copenhagen, Denmark
To whom correspondence should be addressed at: Department of Clinical Biochemistry, Roskilde University Hospital, 713 Køgevej, DK-4000 Roskilde, Denmark. E-mail: hviid{at}dadlnet.dk
Submitted on July 29, 2005; revised on September 27, 2005; accepted on October 4, 2005
| Abstract |
|---|
The non-classical human leukocyte antigen (HLA) class Ib genes, HLA-E, -G and -F, are located on chromosome 6 in the human major histocompatibility complex (MHC). HLA class Ib antigens resemble the HLA class Ia antigens in many ways, but several major differences have been described. This review will, in particular, discuss HLA-G and its role in human reproduction and in the human MHC. HLA-G seems to be important in the modulation of the maternal immune system during pregnancy and thereby the maternal acceptance of the semiallogenic fetus. Recent findings regarding aspects of HLA-G polymorphism, the possible significance of this polymorphism in respect to HLA-G function and certain complications of pregnancy (such as pre-eclampsia and recurrent spontaneous abortions (RSA)) are discussed together with possible importance to IVF. Finally, aspects of a possible role of HLA-G in organ transplantation and in inflammatory or autoimmune disease, and of HLA-G in an evolutionary context, are also briefly examined.
Key words: gene expression / immunology / implantation / pregnancy / trophoblasts
| Introduction |
|---|
During pregnancy, the maternal immune system is in close contact with cells and tissue from the semiallogenic fetus. Therefore, specific mechanisms must exist to modulate and moderate the maternal immune system, so that the pregnant woman does not reject her own fetus. Additionally, aberrations in these mechanisms may in theory lead to complications in pregnancy. The so-called non-classical human leukocyte antigen (HLA) class Ib molecules may be involved in these mechanisms, as they are expressed on trophoblast cells in the placenta (Kovats et al., 1990
This review presents and critically discusses recent findings in primary studies of HLA-G polymorphism, genetics, expression and function. Furthermore, the review will focus on studies that examine a possible role for HLA-G genetics and HLA-G expression in the pathogenesis of pre-eclampsia and certain cases of RSA. Evidence for the importance of HLA-G in blastocyst implantation and the success of IVF will also be discussed. Finally, aspects of HLA-G in organ transplantation and in autoimmune disease will briefly be presented, together with a brief discussion of broader perspectives on HLA-G and the major histocompatibility complex (MHC) in reproduction.
| The MHC in humans |
|---|
The human MHC is located on the short arm of chromosome 6. It spans
4 Mb and encodes at least
130 functional genes. The best described are the classical HLA class Ia and II genes (HLA-A, -B, -C, -DR, -DQ and -DP) (The MHC Sequencing Consortium, 1999). These are well known for their role in organ transplantation and antigen-peptide presentation and for their association with a range of autoimmune diseases (Doherty and Zinkernagel, 1975| The non-classical HLA class Ib genes |
|---|
The HLA class Ib antigens, HLA-E, -F and -G, share some characteristics with the class Ia antigens, but also differ from them in a range of ways. There seems to be growing evidence to support the viewpoint that HLA class Ib molecules, at least in respect to HLA-G, may play a role in the suppression of immune responses and contribute to long-term immune escape or tolerance (Carosella et al., 1999
The HLA-G gene is located on chromosome 6 close to HLA-A, the classical HLA class Ia gene with which it seems to be in closest homology (Messer et al., 1992
) (Figure 1). The HLA-E gene is localized between HLA-C and -A, whereas HLA-F is located close to HLA-G and -A (The MHC Sequencing Consortium, 1999
) (Figure 1). The gene structure of the HLA class Ib genes is very similar to that of HLA class Ia genes. However, differences exist, especially regarding the 3'-end (the cytoplasmic tail) (Geraghty et al., 1987
; Heinrichs and Orr, 1990
). The short cytoplasmic tail of HLA-G seems to be important for the highly reduced spontaneous endocytosis of HLA-G (Davis et al., 1997
). HLA-E is expressed on many different types of cells and tissues as the class Ia antigens (Lee et al., 1998
). The HLA-E-binding groove has a great affinity for the HLA-G signal peptide, and this binding is important for the expression of HLA-E on the trophoblast cell surface. The intracellular transport and expression on the cell surface of HLA class I proteins are dependent on peptides within the endoplasmic reticulum. It seems that most peptides bound to HLA-E are nonamers derived from HLA class I signal sequences and that the binding of heavy chains to these peptides is dependent on a functional transporter associated with antigen processing. The HLA-E molecule has unique amino acid substitutions in the peptide-binding cleft, probably facilitating the binding of class I signal sequence-derived peptides. HLA-E interacts with the CD94/NKG2C receptor, and HLA-E is probably the most important ligand for the inhibition of NK cells (Llano et al., 1998
). It is possible that HLA-G interacts with a killer inhibitory receptor (KIR) and balances the activation signal from HLA-E/the G-nonamer, resulting in the inhibition of lysis by a possible activation of other pathways (e.g. secretion of cytokines) (Ishitani et al., 2003
).
|
The function of HLA-F is not known at present. HLA-F expression has only been detected on invasive cytotrophoblast cells in the placenta, and these are the only cells described so far that express all three HLA class Ib molecules. In this regard, the three class Ib molecules may act in synergy in the placenta and may to some extent functionally be able to substitute for each other (Ishitani et al., 2003
).
The gene polymorphism of the HLA class Ib genes is very sparse in contrast to the highly polymorphic HLA class Ia and II genes. Regarding HLA-E, there is only consensus for three alleles at the protein level and for HLA-F only two; however, the difference is related to a shortened cytoplasmic tail (He et al., 2004
) (HLA Informatics Group, http://www.anthonynolan.org.uk/HIG/lists).
| The HLA-G gene |
|---|
Figure 2 shows the gene structure of HLA-G. The external part of the HLA-G molecule consists of three parts: the
1,
2 and
3 domains (exons 24) (Figure 2);
1 and
2 contribute to the peptide-binding cleft. The crystal structure of HLA-G has recently been determined (Clements et al., 2005
3 domain of HLA-G, and this domain is structurally distinct from the
3 domain of classical MHC class I molecules. HLA-G can bind peptides. In HLA-G transfected cells (B lymphoblastoid cell line 721.221), HLA-G molecules are associated with peptides derived from a variety of intracellular proteins. However, the complexity of the bound peptides may be lower than that of some HLA class Ia molecules (Lee et al., 1995
|
| HLA-G polymorphism |
|---|
The HLA-G proteins are nearly monomorphic, with only four single amino acid polymorphisms described in the literature so far. This stands in marked contrast to the very polymorphic HLA class Ia and II antigens, which exhibit some of the largest genetic polymorphism in the human genome. Regarding HLA-G, polymorphism has also been reported in the 5'-upstream regulatory region (5'URR)/the promotor region and in the 3'-untranslated region (3'UTR) of the gene (Harrison et al., 1993
HLA-G polymorphism in coding regions
HLA-G polymorphism has by now been studied extensively. The first reports on Caucasian populations showed a very limited polymorphism (Morales et al., 1993
; Ober et al., 1996
; Yamashita et al., 1996
; Hviid et al., 1997
); in marked contrast to this, an early study of African Americans showed a rather high polymorphism in exon 3 (the
2 domain) (van der Ven and Ober, 1994
). However, these findings have never been reproduced, and different African populations have shown the same limited polymorphism as that described in Caucasian and Japanese populations (Bainbridge et al., 1999
; Ishitani et al., 1999
; Matte et al., 2000
, 2002
). Fifteen HLA-G alleles at the nucleotide level have been acknowledged by the WHO Nomenclature Committee for Factors of the HLA System (Marsh et al., 2002
; http://www.anthonynolan.org.uk/HIG/). However, more HLA-G alleles at the DNA level have been described in several publications (Ober et al., 1996
; Hviid et al., 1997
, 2002; Kirszenbaum et al., 1999
) (Table I). Regarding the HLA-G protein, only five HLA-G alleles with (single) amino acid substitutions have been described in the literature (Table I). These polymorphisms are located outside the binding groove. Two amino acid substitutions have been found in exon 2 (defining the alleles G*0102 and G*0103), one in exon 3 (defining the G*01040x alleles) and one in exon 4 (the G*0106 allele) (Hviid et al., 2001
, 2002). So single-nucleotide polymorphisms (SNPs) that change the amino acid sequence of the HLA-G protein define the major groups of HLA-G alleles; silent nucleotide variation within these groups defines further specific allelic variants. However, polymorphism in the non-coding regions of the HLA-G gene (see section on HLA-G polymorphism in non-coding regions) is not included in the definition of the WHO-acknowledged HLA-G alleles. HLA-G*0105N is a null allele; it includes a deletion of the first base of codon 130 or the third of codon 129, which results in a frameshift (Ober et al., 1996
; Hviid et al., 1997
; Suarez et al., 1997
). Table II lists the HLA-G allele distributions in different ethnic populations. An interesting observation is that the G*01010x group of alleles, except G*010102, are the predominant alleles, having a frequency up to approximately 80% in certain African populations, whereas these alleles only constitute approximately 5060% in Caucasian and Japanese populations. The G*010102 allele, however, which includes a 14 bp sequence in the 3'UTR of the gene, is sparsely represented in the African populations, but has a frequency of approximately 30% in the other populations (Table II). This is curious, because the G*010102 allele might be an older allele than the others; all studied MHC-G genes in primates (chimpanzee, gorilla, orangutan) include the 14 bp sequence (Castro et al., 2000a
).
|
|
HLA-G polymorphism in non-coding regions
Several studies have identified DNA sequence variation in the 5'URR in introns and in the 3'UTR of the HLA-G gene (Harrison et al., 1993
; Tamaki et al., 1993
; Hiby et al., 1999
; Hviid et al., 1999
, 2004b; Yamashita et al., 1999
; Ober et al., 2003
). It is beyond the scope of this review to list and discuss sequence variation in introns of the HLA-G gene. Efforts have been made to elucidate the polymorphism in the 5'URR and in the 3'UTR parts of the HLA-G gene. Table I lists some of these polymorphisms. In brief, sequences and polymorphism in both the 5'URR and 3'UTR parts of the HLA-G gene may be of importance in the regulation of HLA-G expression. Based on studies of both HLA-G transgenic mice and the binding of nuclear factors, a region between 1.1 and 1.4 kb from the start site of transcription of the HLA-G gene has been proposed to contain an important regulatory element (Schmidt et al., 1993
; Moreau et al., 1997
). Sequence polymorphism has also been reported in this region (Hviid et al., 1999
, 2004b; Ober et al., 2003
). However, the functional links and importance of this polymorphism are still uncertain. A polymorphism at position 725 might be associated with SA (Ober et al., 2003
) (Table I), and there might be an association with the status of the methylation and expression of sHLA-G; however, further studies are clearly needed (Ober et al., 2003
; Hviid et al., 2004
b). In the promotor region of HLA-G, an interesting polymorphism at position 201 has been reported (Hviid et al., 1999
) (Table I). A G in G*010101 is changed to an A in G*010102, G*010103, G*0104 and G*0105N. The polymorphism is located in the NF-
B2 element (enhancer A) and reverts the regulatory sequence to that of HLA-A2. Still, further studies are needed to elucidate whether this promotor polymorphism has any effect on the regulation of expression of certain HLA-G alleles by allowing the binding of NF-
B2 factors to the HLA-G enhancer A (Hviid et al., 1999
; Solier et al., 2001a
). Harrison et al. (1993
) was the first to describe a 14 bp deletion/insertion polymorphism (5'-ATTTGTTCATGCCT-3') in the 3'UTR of the HLA-G gene located at position 3741 in exon 8 (according to the reference sequence of Geraghty et al., 1987
) (Figure 2). In Caucasian populations, the frequencies of these two HLA-G polymorphisms are nearly equal (frequency of alleles including the 14 bp sequence:
45%; alleles with the 14 bp sequence deleted:
55%); however, in African populations the allele with the deleted 14 bp sequence may dominate when the linkage disequilibrium between this polymorphism and the WHO-acknowledged HLA-G alleles is considered (Table II), although no direct study of this matter has yet been published.
In conclusion, there are no convincing data that the few polymorphisms in the coding regions of the HLA-G gene affect HLA-G function, except for the deletion in the HLA-G-null allele G*0105N. Polymorphism in the 3'UTR and 5'URR parts of the HLA-G gene has been found to be associated with differences in HLA-G expression, which may affect HLA-G function and may be of importance in certain complications of pregnancy. These aspects will be outlined in detail in later sections of this review.
| HLA-G expression and function |
|---|
HLA-G expression
The expression of HLA-G is restrictive. Although HLA-G mRNA has been detected in many different tissues, HLA-G protein expression has been described repeatedly only on and by the trophoblast cells in placenta, on and by certain immune cells (in most cases monocytes) and in the thymus (Kovats et al., 1990
; Crisa et al., 1997
; Lila et al., 2001
; Ishitani et al., 2003
; Rebmann et al., 2003
). However, HLA-G protein expression can sometimes be observed in muscle fibres and in liver biliary and renal tubular epithelial cells (Wiendl et al., 2000
; Lila et al., 2002
; Creput et al., 2003a
). HLA-G can be detected in serum/plasma from both women and men, but there is controversy in various published studies to whether sHLA-G can be detected in the blood of all men and non-pregnant women under normal conditions (Rebmann et al., 2001
; Lila et al., 2002
; Hviid et al., 2004
b). Reviewing the literature, it seems that sHLA-G can be detected in all plasma samples from pregnant and non-pregnant women, whereas sHLA-G can only be detected in some serum samples, at least from non-pregnant women (and from men); however, one study using a unique sHLA-G assay seemed to detect sHLA-G in all serum samples from pregnant women (Yie et al., 2004
). It can be speculated that in blood containing small amounts of sHLA-G, the sHLA-G can be lost in some way in the subsequent generation of the serum samples. In future sHLA-G studies, plasma samples may be preferred or serum samples should only be compared with other serum samples in, e.g., casecontrol studies. Furthermore, no consensus exists regarding the sHLA-G assays used in various publications. So far, they have all been in-house assays with different antibodies and sHLA-G standards, so evaluation and standardization of the different assays are thus greatly needed. The main source of HLA-G5 (sHLA-G) in the blood of non-pregnant women and in men is most likely monocytes; though CD4+ and CD8+ T cells and B cells seem to be able to secrete HLA-G5 as well, though to a lesser extent (Rebmann et al., 2003
). Studies so far indicate that the presence, or the level, of sHLA-G in serum/plasma samples is associated with HLA-G polymorphism (Rebmann et al., 2001
; Hviid et al., 2004
b; Rizzo et al., 2005a
).
HLA-G mRNA alternative splicingmembrane-bound and soluble isoforms
HLA-G potentially exists in four membrane-bound isoforms, HLA-G1 to -G4, and three soluble isoforms, HLA-G5 to -G7 (Ishitani and Geraghty, 1992
; Fujii et al., 1994
; Kirszenbaum et al., 1994
; Hviid et al., 1998
; Paul et al., 2000
; LeMaoult et al., 2003
) (Figure 2). The possible expression on the cell surface of the shortened isoforms HLA-G2G4 is controversial (Menier et al., 2000
; Bainbridge et al., 2000
; Mallet et al., 2000
; Riteau et al., 2001b
; Morales et al., 2003
; Ulbrecht et al., 2004
). The sHLA-G isoforms are generated by alternative splicing of the HLA-G transcript. Intron 4 includes a stop codon, and this intron is retained in the HLA-G mRNAs which are translated into the sHLA-G isoforms (Fujii et al., 1994
; Hviid et al., 1998
). sHLA-G, like HLA class Ia soluble molecules, is probably also generated by the shedding of membrane-bound HLA-G molecules (Park et al., 2004
). The expression of soluble HLA-G6 with exon 3 spliced out has been reported; however, more independent studies are needed to establish the secretion and functionality of this HLA-G isoform beyond any doubt.
Associations between HLA-G polymorphisms and HLA-G expression
The 14 bp deletion/insertion polymorphism in exon 8 of the 3'UTR is present in both the HLA-G gene and transcript (Harrison et al., 1993
) (Figure 2). HLA-G was previously found to be co-dominantly expressed (Hviid et al., 1998
); however, HLA-G alleles containing the 14 bp sequence have been found to be associated with a lower HLA-G mRNA level for most isoforms in heterozygous (first trimester) trophoblast samples (Fujii et al., 1994
; Hiby et al., 1999
; Hviid et al., 2003
) (Figure 3). Hviid et al. (2003
) have investigated villous and extravillous trophoblast cell fractions as well as whole biopsy samples from late first-trimester placentas for HLA-G mRNA expression. The 14 bp sequence polymorphism in the 3'UTR of the HLA-G gene made it possible to study the two allelic HLA-G mRNA expression levels and alternative splicing patterns in heterozygous trophoblast samples. Overall, when the expression of different HLA-G*010102 mRNA isoforms including the 14 bp sequence were compared with that of G*010101, where the 14 bp sequence is deleted, the expression levels were significantly reduced. These more detailed results were in accordance with the data of OBrien et al. (2001
). In the cases of transcription and alternative splicing that involved the G*010102 allele, additional mRNA isoforms were observed to be lacking the first 92 bp of exon 8, these HLA-G mRNAs were variants of HLA-G1 and HLA-G5/G6. These HLA-G mRNA isoforms were also found in the G*010103 allele, and here unique G2 (92 bp), and possibly G4 (92 bp), isoforms were observed. On the other hand, regarding the levels of HLA-G mRNA expression, the G*010103 allele more closely resembled the G*010101 allele. In conclusion, it can be speculated whether the presence of the 14 bp sequence at the beginning of exon 8 functions as a cryptic branch-point sequence for HLA-G mRNA splicing. Interestingly, HLA-G transcripts with the 92 bp deletion have been reported to be more stable than the complete mRNA forms (Rousseau et al., 2003
). However, at the moment, it has not been determined whether the 14 bp polymorphism is directly involved in the observed differences in HLA-G mRNA expression or whether it is polymorphism in the HLA-G 5'URR in linkage disequilibrium with the 14 bp polymorphism. Furthermore, several studies have been unable to detect sHLA-G in serum samples from individuals homozygous for the presence of the 14 bp sequence. These samples originated from both donors and couples attending IVF clinics (Hviid et al., 2004
b; Rizzo et al., 2005a
). sHLA-G was measured in serum by enzyme-linked immunosorbent assay (ELISA), which detects the HLA-G molecule in a ß2-microglobulin-associated form, so the assay in theory detects both sHLA-G1 that is shed by membrane-bound HLA-G1 and the soluble HLA-G5 isoform. In only 15.4% (23/149) of all serum samples could HLA-G5/sHLA-G1 be detected with a mean concentration of 11.7 ± 2.5 ng/ml (±SEM) for the positive samples; no differences being observed between males and females. No HLA-G5/sHLA-G1 was detected in any of the +14/+14 bp samples (P = 0.011). These findings are somewhat in agreement with a study of serum samples from heart transplant patients, 18% of whom were sHLA-G positive; but this is, naturally, an abnormal condition (Lila et al., 2002
). Rebmann et al. (2001
) have also studied associations between HLA-G alleles and sHLA-G plasma levels; however, no HLA-G-specific monoclonal antibodies (mAbs) were included in the ELISA used. The main results were a split between low- and high-secretor HLA-G alleles, G*01013 and G*0105N being low secretors, G*0104 a high secretor and G*01011 and G*01012 being between the two levels. When families were studied, the low and high sHLA-G plasma levels did not segregate with HLA haplotypes, so it was concluded that additional mechanisms might be involved in the regulation of the individual sHLA-G levels. These results are only partly in agreement with those of studies of the 14 bp polymorphism and HLA-G5/sHLA-G1 serum levels. Further study is clearly needed in this area. Discrepancies between results may well depend on the specific anti-HLA-G antibodies used and whether they bind to ß2-microglobulin-associated HLA-G or intron 4-retaining sHLA-G isoforms (Fournel et al., 2000
b). There might even be some differences related to whether the analysis is of serum or plasma samples. Shed HLA-G molecules may also blur associations between HLA-G polymorphism and sHLA-G protein expression. Finally, preparations of HLA-G protein standards may also vary and influence comparisons between different sHLA-G assays.
|
The first and also the most comprehensive studies of the functions of HLA-G molecules relate to a range of in vitro experiments, which have shown that HLA-G can inhibit NK and T-cell-mediated cell lysis, both through direct interaction with the receptors ILT2 and ILT4 and with the killer Ig-like receptor 2 DL4 (KIR2DL4 receptor) (Navarro et al., 1999
; Ponte et al., 1999
; Rajagopalan and Long, 1999
; Riteau et al., 2001a
; Menier et al., 2002
) (Figure 4). Regarding the KIR2DL4 receptor, different responses are elicited depending on the activation of the effector cells (Rajagopalan and Long, 1999
; Hofmeister and Weiss, 2003
). The HLA-G
1 domain may be important in the inhibition of NK cell activity, and Met76 and Gln79 residues in the
1 domain may be involved in KIR2DL4 recognition (Rouas-Freiss et al., 1997
; Yan and Fan, 2005
). This is interesting in relation to preserved functions of HLA-G isoforms lacking exon 3 (the
2 domain).
|
On the trophoblast cells of the placenta, which originate from the fetus, no classical HLA class Ia and II antigens are expressed, except for a possible weak expression of HLA-C (Redman et al., 1984
; Hunt et al., 1987
; Lata et al., 1990
; King et al., 1996
, 2000). In this way, the HLA-semiallogenic fetal cells will not come in direct contact with the maternal immune system. However, cells that do not express MHC/HLA on the surface will undergo NK-mediated cell lysis. The strong expression of the nearly monomorphic HLA-G molecule on and by the invasive cytotrophoblast cells, together with the expression of HLA-E (and -F) in the placenta, will prevent this. As mentioned earlier, the function of HLA-F is not known. However, HLA-F expression has so far only been detected on invasive trophoblast cells. Actually, these cells are the only cells that express all three HLA class Ib molecules, which is an interesting observation regarding these non-classical HLA antigens probable importance in the pregnant womans acceptance of the semiallogenic fetus. It can be hypothesized that the products of the three genes may act in synergy and may even be able to substitute for each other (Ishitani et al., 2003
).
A so-called T-helper 1 (Th1) response is mediated by a subset of CD4+ T cells that are primarily characterized by the cytokines [e.g. interferon-
(IFN-
), interleukin-2 (IL-2) and tumour necrosis factor-
(TNF-
)] they produce. A Th1 response is also called a (pro)inflammatory response. A Th2 response is mediated by another subset of CD4+ T cells that produce cytokines, such as IL-10, IL-3 and IL-4. Th2 cells are mainly involved in stimulating B cells to produce antibodies (humoral immune response). A successful pregnancy has been called a Th2 phenomenon characterized by a Th2 cytokine profile, whereas certain complications of pregnancy, such as RSA and pre-eclampsia, have been associated with a Th1 response (Chaouat et al., 2004
).
IL-10 has been shown to be able to activate HLA-G expression (Moreau et al., 1999
). On the other hand, the presence of both membrane-bound and sHLA-G seems to induce changes in the secretion of cytokines from allo-CTL-activated peripheral blood mononuclear cells (PBMCs) (Maejima et al., 1997
; Kapasi et al., 2000
). However, the exact nature of the induced changes in cytokine profiles is controversial. Comparisons of various in vitro studies are complicated by differences in the sources of HLA-G used and in the cytokine-secreting cells studied; therefore, it is also difficult to judge which results may be the most credible. The studies can be divided into four categories.
Cell lines transfected with membrane-bound HLA-G1 in co-culture with PBMC
A range of studies in which cell lines (the B-lymphoblast cell line 721.221 or K-562 leukaemia cells) have been transfected with the membrane-bound HLA-G1 (mHLA-G1) gene and co-cultured with PBMC or uterine mononuclear cells (UMC) or perhaps specific subsets of these, has been performed (Maejima et al., 1997
; Kanai et al., 2001a
,b, 2003; Rieger et al., 2002
; van der Meer et al., 2004
). Regarding the PBMC response to mHLA-G1-transfected cells, there is consensus concerning the four most-studied cytokines. The secretion of TNF-
and IFN-
, both considered Th1 cytokines, decreased, whereas IL-10 and IL-4, both Th2 cytokines, increased (Maejima et al., 1997
; Kanai et al., 2001a
,b). IL-10 is an important anti-inflammatory cytokine, and elevated levels of IL-10 in pregnancy may be a mechanism for inducing tolerance by generating tolerance-inducing T cells (Huizinga et al., 1999
). However, this interpretation may have to be modified and will be discussed later. Maejima et al. (1997
) found a decrease in IL-3, while Kanai et al. (2001a
) found an increase in this cytokine.
Cell lines transfected with membrane-bound HLA-G1 in co-culture with uterine or decidual mononuclear cells
Regarding the responses of UMC from non-pregnant women on day 7 after the LH surge, large granular lymphocytes (LGL) and decidual mononuclear cells (DMC), both from human deciduas to mHLA-G1 transfected cells, there is almost consensus (Kanai et al., 2001b
, 2003; Rieger et al., 2002
; van der Meer et al., 2004
). Again, TNF-
and IFN-
decreased; however, the level of IL-10 decreased or was unchanged compared with that of controls, and IL-4 was unchanged. One study has investigated the response of uterine NK cells from non-pregnant women (van der Meer et al., 2004
), finding that IFN-
increased in these cells. Interestingly, in the same study, vascular endothelial growth factor (VEGF) was found to increase in both UMC and uterine NK cells in comparison with the controls.
In conclusion, these studies of the effect of mHLA-G1-transfected cells show a diminishing of Th1-like cytokines, and especially in the studies with PBMC, a Th2-like response. A Th2 response in mononuclear cells from the uterus is, however, not that evident. Importantly, the effect of mHLA-G1 on IL-10 secretion seems modest (Kanai et al., 2001a
; Rieger et al., 2002
). These in vitro experiments can be difficult to interpret in a physiological context. The most relevant studies in relation to pregnancy are the ones that include UMC. Here, the HLA-G transfected cells can be viewed as the trophoblast part. The traditional concept of an up-regulation of Th2 and down-regulation of Th1 in an uncomplicated pregnancy is somewhat challenged in relation to the expression of membrane-bound HLA-G in these experiments; this will be further addressed later in this review. In relation to the studies with PBMC, it is important to note that these cells originated from males or nulligravidae women. As a whole, the studies indicate that the cytokine response to mHLA-G-expressing cells may differ between a non-pregnant and a pregnant state, or at least the response may differ between cells from peripheral blood and local uterine cells. However, it must be remembered that the content of specific mononuclear cells in PBMC and UMC populations is different, and the UMC samples may be contaminated with other cell populations.
Cytokine secretion from model systems with immune cells and addition of sHLA-G
A range of studies of the effect of sHLA-G on cytokine secretion from immune cells has been conducted. In mixed lymphocyte cultures (MLC), Kapasi et al. (2000
) found biphasic responses regarding TNF-
, IFN-
and IL-10 when HLA-G from placentas was added in the 01000 ng/ml range. At HLA-G concentrations of
50 ng/ml, secretion of TNF-
and IFN-
increased and that of IL-10 was decreased compared with that of control cultures; that is, the IL-10 concentrations rose as sHLA-G levels diminished to zero. At an HLA-G concentration of 1000 ng/ml, the cytokine levels were inverted, as in a Th2 response. These results are partly in accordance with a study of recombinant sHLA-G1 (rsHLA-G1) added to DMC, in which rsHLA-G1 concentrations of 250 and 500 ng/ml were found to decrease levels of TNF-
and IFN-
; however, when rsHLA-G1 was added to PBMC at a concentration of 1000 ng/ml, the levels of these two cytokines increased (Kanai et al., 2001a
, 2003). Two more studies have measured IL-10 levels after the addition of sHLA-G. At a concentration of 1000 ng/ml, rsHLA-G1 significantly increased IL-10 levels in PBMC/.221 cell cultures, quite in accordance with studies by Kapasi et al. (2000
) and McIntire et al. (2004
) in which phorbol 12-myristate 13-acetate (PMA)/IFN-
-treated U937 myelomonocytic cells increased IL-10 levels after the addition of 10 nM (
250 ng/ml, according to McIntire et al., 2004
) of rHLA-G5 (Kanai et al., 2001a
; McIntire et al., 2004
). However, in this last study, rHLA-G5 and -G6 at concentrations of 1000 nM (
25 000 ng/ml) were found to suppress IL-10 secretion, while rHLA-G5/-G6 at concentrations of only 1 nM (
25 ng/ml) did not affect IL-10 levels. IL-10 has been shown to stimulate HLA-G expression (Moreau et al., 1999
). In in vitro time-course experiments, IL-10 secretion preceded HLAG5/sHLA-G1 expression in lipopolysaccharide (LPS)-activated PBMC cultures. Addition of anti-IL-10 mAb to the LPS-activated cultures also blocked the HLA-G5/sHLA-G1 expression (Rizzo et al., 2005a
). In PBMC cultures with the +14/+14 bp HLA-G genotype, however, a significantly higher concentration of IL-10 was observed together with the same basic level of HLA-G5/sHLA-G1, indicating that in some individual PBMC samples a higher level of IL-10 is needed to induce sHLA-G expression. Furthermore, in a study of donor serum samples, a trend towards an inverse relationship between HLA-G5/sHLA-G1 and IL-10 levels was observed (Rizzo et al., 2005a
). Furthermore, LPS activation of PBMCs resulted in a rough trend towards decreasing levels of IL-10 with increasing concentrations of HLA-G5/sHLA-G1 in the cultures (Rizzo et al., 2005a
). These results are somewhat in accordance with those of other studies of IL-10 and sHLA-G (Kapasi et al., 2000
; McIntire et al., 2004
), although not with those of Kanai et al. (2001a
), which examined unstimulated PBMCs. Finally, McIntire et al. (2004
) found that rHLA-G5 at 10 nM (
250 ng/ml) and 1000 nM (
25 000 ng/ml) augments transforming growth factor (TGF)-ß1 secretion by the human myelomonocytic cell line. TGF-ß1 may promote placental cell growth and differentiation (Ingman and Robertson, 2002
). It seems that rHLA-G5 and -G6 mediate their effects through interaction with ILT4 and ILT2 receptors. IL-4 and IL-6 secretion seem not to be affected by rsHLA-G1, rHLA-G5 or rHLA-G6 (Kanai et al., 2001a
, 2003; McIntire et al., 2004
). In conclusion, it is rather difficult fully to compare the cited studies of sHLA-G added to or spontaneously secreted by different mononuclear cells. Some studies investigate MLC, others PBMC/DMC or cell lines stimulated in different ways. It seems that for IL-10, TNF-
and IFN-
secretion, the levels secreted are related to sHLA-G concentration. A trend found is that at increasing levels of sHLA-G, both TNF-
and IFN-
secretion seems to decrease, at least in the MLC and DMC experiments. For IL-10, decreasing amounts of sHLA-G may result in increased secretion of IL-10 from activated mononuclear cells.
Cytokine secretion in a model system with both mHLA-G1-expressing cells and addition of sHLA-G1
One unique study examined the effect of rsHLA-G1 added to PBMC/HLA-G1.221 cells at concentrations of 2501000 ng/ml (Kanai et al., 2001a
). The results, in comparison with control cultures, showed a significant increase in TNF-
, IFN-
and IL-10 and unchanged levels of IL-4. However, a similar study of DMC/HLA-G1.221 cells found a significant decrease in TNF-
and IFN-
levels, whereas IL-4 levels were unaltered, at rsHLA-G1 concentrations of both 250 and 500 ng/ml; IL-10 was not considered in this study (Kanai et al., 2003
). This model system with both mHLA-G and sHLA-G in the cultures is very relevant to the in vivo situation in the placenta.
Several of the studies discussed above point out important differences between the cytokine response in PBMC and in uterine/DMC populations to HLA-G expression. Also, the expression levels of HLA-G in the placenta are probably of critical importance for the exact cytokine response. The sHLA-G concentrations in plasma/sera of pregnant women are approximately 5100 ng/ml (Pfeiffer et al., 2000
; Steinborn et al., 2003
; Yie et al., 2004
), although one study has reported finding several serum concentrations in the third trimester as high as 30004000 ng/ml; however, local concentrations of sHLA-G may be much higher. Unfortunately, not many data regarding HLA-G local concentrations in the placenta are available. Yie et al. (2004
) have studied the HLA-G levels in 14 placental lysates (gestational weeks 3842) and found a median concentration of 88 ng/mg protein (range 6.5200 ng/mg protein). However, the most interesting finding in relation to the in vitro experiments would be sHLA-G concentrations in uterine blood samples, which to the best of the authors knowledge have not been published. Altogether, the in vivo data might indicate whether the sHLA-G concentrations used in the in vitro experiments are physiologically relevant. One problem in drawing a general conclusion from these studies is that both the sHLA-G preparations and sHLA-G assays used need to be standardized, so direct comparisons might be problematic.
So, in conclusion, a possible function of HLA-G in uncomplicated pregnancy could be to shift a proinflammatory Th1 cell-mediated response towards a Th2 response, although this concept might have to be revised, especially as regards the local situation at the fetomaternal interface. It is important to note that the concept of Th1/Th2 responses may be far too simplistic as discussed (Chaouat et al., 2004
). Also, HLA-G seems able to inhibit an allo-CTL response in vitro, but it seems that the response is biphasic, so a moderate amount of HLA-G added to the cell culture system increases the response (Kapasi et al., 2000
). Further investigation is needed to elucidate the relationships between HLA-G expression and cytokine secretion.
Regarding allograft tolerance, this can be transferred between recipients by regulatory T cells: suppression can be actively mediated by regulatory T cells, or tolerance can be maintained by the deletion or functional non-responsiveness (anergy) of alloreactive T cells (Wekerle et al., 2002
). In primary mixed lymphocyte reactions (MLR), responder T cells (both CD4+ and CD8+) could express both sHLA-G1 and HLA-G5, and this suppressed the alloproliferative response of T cells (Lila et al., 2001
; Le Rond et al., 2004
). No HLA-G expression could be detected in autologous MLR combinations. However, HLA-G expression was not detected in all allogenic combinations, and differences existed in the HLA-G isoform profile. Interestingly, HLA-G expression was influenced by the specific MLR combinations.
A recent study found that HLA-G1 transfected antigen-presenting cells can inhibit the proliferation of CD4+ T cells, shed HLA-G1 molecules, induce CD4+ T cell anergy or at least long-term unresponsiveness and cause the differentiation of CD4+ T cells into suppressive cells (LeMaoult et al., 2004
). So HLA-G-positive antigen-presenting cells may be directly involved in the suppression of immune responses and contribute to long-term immune escape or tolerance (LeMaoult et al., 2004
).
HLA-G can bind peptides, as noted above, and a single study of HLA-G transgenetic mice has rendered it probable that HLA-G may be able to present peptides to the T-cell receptor, though this finding is controversial (Lenfant et al., 2003
). Finally, HLA-G may also be implicated in cellular aggregation and cell adhesion (Odum et al., 1991
).
| HLA-G in human reproduction |
|---|
HLA-G expression and function during implantation and pregnancy
Some studies have found HLA-G expression in blastocysts and maybe even in preimplantation blastocysts; however, this issue is still rather controversial (Jurisicova et al., 1996
; Hiby et al., 1999
; Fuzzi et al., 2002
; van Lierop et al., 2002
; Sher et al., 2004
; Noci et al., 2005
; Yie et al., 2005
). One study detected HLA-G mRNA and protein expression in 40% of 2- to 16-cell stage preimplantation embryos with a total of 148 blastocysts investigated (Jurisicova et al., 1996
). Expression of the HLA-G transcript was associated with an increased cleavage rate when compared with embryos lacking HLA-G mRNA. Another study of HLA-G mRNA in 108, day-3 preimplantation embryos from 25 couples, found that 44% were positive (Cao et al., 1999
). On the other hand, Hiby et al. (1999
) could not detect HLA-G mRNA in embryos from the 2-cell to the blastocyst stage; however, only 11 embryos were investigated. These findings should be considered in relation to more general studies, which find that the embryonic genome becomes activated by the 2-cell stage in mammals, and perhaps by the 4-cell stage in humans (Braude et al., 1988
; Zeng and Schultz, 2005
). In another study of human embryo culture supernatants from embryos at the 8-cell morula or beyond the blastocyst stage, no sHLA-G could be detected using three different ELISA set ups (van Lierop et al., 2002
). However, it is not quite clear how many embryos were tested, and only pooled supernatants were studied. In line with the finding that only a fraction of preimplantation embryos may express HLA-G, three independent groups have detected sHLA-G expression in the culture media of these in relation to IVF procedures after 4672 h of culture before transfer. Noci et al. (2005
) recorded the presence of sHLA-G in 36.2% of single-embryo cultures (N = 318) after 72 h of culture, while only 8.9% of the cultures were sHLA-G-positive after 48 h. In another study of 594 embryos from 201 women, the exact frequency of sHLA-G-positive single-embryo cultures was not specified; however, in 65% of these women at least one of the transferred embryos (mean transferred 3.0) was positive (Sher et al., 2004
). Caution should be exercised, because in the sHLA-G ELISA used in this study, an mAb MEM-G1 was directed against denaturated HLA-G. Noci et al. (2005
) used an mAb against HLA-G in a ß2-microglobulin-associated form (HLA-G5/sHLA-G1). The frequencies of HLA-G-positive preimplantation embryos found by the two independent studies of Jurisicova et al. (1996
) and Noci et al. (2005
) are in agreement. Interestingly, the detection of sHLA-G in the culture media is associated with a higher success in IVF treatment, defined as the obtaining of a clinical pregnancy. But the expression of HLA-G is no guarantee of implantation, in that a large fraction of sHLA-G-positive IVF treatments fail. On the other hand, these studies found that all IVF treatments with sHLA-G-negative preimplantation embryo cultures ended unsuccessfully with no signs of implantation (Fuzzi et al., 2002
; Sher et al., 2004
; Noci et al., 2005
). A recent study by Yie et al. (2005
) further supports these findings. Of 386 embryo culture supernatants (with 2.9 embryos per culture), 69.9% were positive for sHLA-G after 72 h of culture. The embryos originated from 137 IVF cycles (corresponding to the same number of couples). Again, the live birth rate in women who had HLA-G-positive embryos transferred was significantly higher than that in women who had only HLA-G-negative embryos transferred (48.4 versus 17.1%; P = 0.0026). Furthermore, a significantly higher mean embryo cleavage rat



