- Split View
-
Views
-
Cite
Cite
A. van der Meer, H.G.M. Lukassen, B. van Cranenbroek, E.H. Weiss, D.D.M. Braat, M.J. van Lierop, I. Joosten, Soluble HLA-G promotes Th1-type cytokine production by cytokine-activated uterine and peripheral natural killer cells, Molecular Human Reproduction, Volume 13, Issue 2, February 2007, Pages 123–133, https://doi.org/10.1093/molehr/gal100
- Share Icon Share
Abstract
Soluble forms of HLA-G (sHLA-G) have been implicated in immune regulation. Fetal trophoblast cells are a prime source of HLA-G. Hence, an interaction between sHLA-G and uterine lymphocytes in the decidual tissues can easily be envisaged. These lymphocytes, when properly activated, are implicated in successful trophoblast invasion, placental maturation and maintenance of pregnancy. However, so far, no data are available on the effect of sHLA-G on the function and phenotype of these cells. Herein, we used a recombinant sHLA-G construct to determine the effect of sHLA-G on uterine lymphocyte cells present in endometrium at the time that it is optimally receptive to trophoblast invasion. In addition, we ascertained the effect of sHLA-G on peripheral lymphocytes. We found that upon co-culture with sHLA-G, proliferation of unfractionated IL-15-stimulated uterine mononuclear cells (UMCs) was inhibited. However, sHLA-G increased both interferon (IFN)-γ and tumour necrosis factor (TNF)-α production by these cells. Vascular endothelial growth factor (VEGF) production was reduced. Notably, in contrast to membrane-bound HLA-G, sHLA-G did not affect the natural cytolytic activity of UMCs. Similarly, sHLA-G inhibited proliferation but stimulated pro-inflammatory cytokine production by cytokine-activated, unfractionated peripheral blood mononuclear cells (PBMCs). In addition, we showed that the overall inhibitory effect of sHLA-G on proliferation of the whole cell population could be ascribed to selective inhibition of CD4+ T cells. In contrast, sHLA-G induced proliferation and IFN-γ production by both uterine and peripheral natural killer (NK) cells. In conclusion, our data show that the sHLA-G modulates both UMC and PBMC function. sHLA-G, by promoting IFN-γ production by uterine NK cells, may contribute to vascular remodelling of spiral arteries to allow for successful embryo implantation.
Introduction
The non-classical HLA class I molecule HLA-G has gained in interest because of its restricted tissue distribution at sites where immunoregulation is needed. HLA-G is expressed at the fetal–maternal interphase only by the semi-allogeneic trophoblast cells that invade the maternal decidua, and in the thymus by epithelial cells. Furthermore, HLA-G expression can be induced under pathological conditions. HLA-G expression was found in psoriasis (Aractingi et al., 2001), in heart transplants (Lila et al., 2000) where expression was associated with a decreased rejection rate and on different tumours, although the latter is not always found (reviewed in Seliger et al., 2003). It has been hypothesized that HLA-G might play a role here in dampening the inflammatory response or providing a mechanism for tumour escape. All these findings are suggestive of an immunoregulatory role for HLA-G in the placenta as well as in peripheral tissues and organs.
HLA-G occurs in a membrane-bound and a soluble form, and functional data suggest that the soluble forms of HLA-G (sHLA-G) may have a different function than its membrane-bound counterpart and that it even may counteract effects of mHLA-G (Kanai et al., 2001a). sHLA-G can be generated by at least two different processes. First, because of alternative splicing, three different soluble isoforms occur that have a stop codon in intron 4, thus preventing translation of the transmembrane domain and cytoplasmic tail. There still is a lot of debate on the expression and function of these different isoforms (Bainbridge et al., 2000; Mallet et al., 2000; Riteau et al., 2001; Ulbrecht et al., 2004; Blaschitz et al., 2005). Second, sHLA-G can be produced by cleavage of the membrane-bound form from the cell surface by metalloproteinases (Dong et al., 2003; Park et al., 2004). At present, it is unclear to what extent either of these processes contribute to the levels of sHLA-G found in pregnancy or under pathological conditions (Dong et al., 2003; Blaschitz et al., 2005; Hunt and Geraghty, 2005; Le Bouteiller, 2005; Sargent, 2005).
Three different receptors have been described that can bind HLA-G, i.e. KIR2DL4 (Ponte et al., 1999; Rajagopalan and Long, 1999), ILT2 (Colonna et al., 1997; Allan et al., 1999) and ILT4 (Colonna et al., 1997; Allan et al., 1999). Of these three, only KIR2DL4 appears to selectively interact with sHLA-G (Rajagopalan et al., 2006). Functional signalling through KIR2DL4 is dependent on internalization of both ligand and receptor into early endosomes. This does not occur with membrane-bound mHLA-G. Expression of KIR2DL4, ILT2 and ILT4 occurs in both uterine and peripheral natural killer (NK) cells (KIR2DL4 and ILT2), T cells (ILT2), B cells (ILT2) and monomyelocytic cells (ILT2 and ILT4) (Colonna et al., 1997; Allan et al., 1999; Ponte et al., 1999; Rajagopalan and Long, 1999; Lukassen et al., 2004).
In pregnancy, sHLA-G is thought to play a role in implantation and maintenance of the semi-allogeneic embryo. sHLA-G is secreted by different trophoblast populations (Solier et al., 2002) and is present at high levels in amniotic fluid (Rebmann et al., 1999; Hunt et al., 2000). More recently, variations in amniotic fluid sHLA-G levels were correlated with embryo gender and pathological pregnancy (Emmer et al., 2002, 2003). There are also indications that sHLA-G is present in embryo culture supernatant, and levels are positively correlated with the likelihood of pregnancy (Fuzzi et al., 2002; Sher et al., 2005; Yie et al., 2005), but using different detection reagents, data to the contrary have been reported (van Lierop et al., 2002). Although it is well known that the different lymphocyte populations in the uterus express receptors for HLA-G (Verma et al., 2000), so far no studies have been performed to analyse the direct effect of sHLA-G on uterine mononuclear cells (UMCs). The mechanism by which sHLA-G affects implantation and maintenance of the embryo therefore remains speculative.
Several in vitro as well as in vivo data point towards a role for HLA-G in modulating peripheral lymphocytes. sHLA-G was found in biopsies and serum of heart transplant recipients, and its presence was associated with a decreased rejection rate (Lila et al., 2002). In vitro data show that sHLA-G could be produced by alloreactive CD4+ T cells in mixed lymphocyte cultures where it suppressed alloproliferation (Lila et al., 2001; Le Rond et al., 2004). Furthermore, sHLA-G was shown to induce apoptosis of activated CD8+ T cells (Fournel et al., 2000) and NK cells (Spaggiari et al., 2002) and to prevent alloreactive cytotoxic T-cell responses (Kapasi et al., 2000) and T-cell proliferation (Le Friec et al., 2003). More recent data suggest that sHLA-G may induce regulatory T cells (Le Rond et al., 2006). The modulatory effect of sHLA-G on cytokine production by peripheral blood lymphocytes is still debated. Three studies have shown that sHLA-G induces a shift towards a Th1-type cytokine pattern (Kapasi et al., 2000; Kanai et al., 2001a; Le Friec et al., 2003). In the latter study, however, this appeared to be a concentration-dependent effect, whereby at high concentrations of sHLA-G a shift towards a Th2-type pattern occurred. For peripheral NK cells, it has been shown that sHLA-G induces the production of pro-inflammatory and pro-angiogenic cytokines and chemokines (Rajagopalan et al., 2006).
In this study, we have analysed the effect of sHLA-G on the functional capacity of both uterine and peripheral cytokine-stimulated lymphocytes. UMCs were derived from the non-pregnant endometrium during the so-called window of implantation. At this late secretory phase of the menstrual cycle, the endometrium is at its most receptive and contains high numbers of uterine NK cells that have receptors for all three class I molecules that are present on the extravillous trophoblast (HLA-G, HLA-E and HLA-Cw). These uterine NK cells have a phenotype that differs from that of peripheral blood NK cells (Lukassen et al., 2004) and are imputed to play a role in the implantation process by producing a variety of immunoregulatory cytokines and angiogenic growth factors that contribute to endometrial angiogenesis (Li et al., 2001). We measured production of interferon (IFN)-γ, tumour necrosis factor (TNF)-α and vascular endothelial growth factor (VEGF). Furthermore, the effect of sHLA-G on NK-cell cytotoxicity was determined. As a source of sHLA-G, we used a recombinant sHLA-G construct consisting of the full-length HLA-G5 sequence, an HLA-G-binding peptide (RLPKDFVDL) and β2-microglobulin (β2-M), coupled by linkers, as described previously (van Lierop et al., 2002).
Materials and methods
Cells and cell lines
Endometrial tissue was obtained from healthy non-pregnant women (n = 5) or women participating in an IVF programme who experienced total fertilization failure (n = 10). Indications for IVF were male factor infertility (5), idiopathic infertility (3), tubal factor infertility (1) and polycystic ovary syndrome (1). The healthy non-pregnant women had a regular cycle, and four of them were nulliparous. They had a partner with severe male infertility and were recruited from the waiting list for ICSI. Endometrial biopsy was performed by a microcurettage using a pipelle de cornier (Prodimed, Reilly-en-Thelle, France) on day 7 after the LH surge in a natural cycle or 6 days after oocyte retrieval in the case of IVF treatment. Of these women, peripheral blood samples were drawn before the endometrial biopsy was taken. Characteristics of the sample group as well as the isolation procedure have been described previously (Lukassen et al., 2004). The strength of the cellular response was quite variable between different women within either group. So, we did not observe a significant difference in the responses to sHLA-G between the hormone-treated women and the women with a regular cycle. As we were limited in the available material, we have therefore decided not to subdivide these two groups.
To obtain mononuclear cells, we mechanically disrupted endometrial tissue by mincing between two scalpels and then filtered through a sieve. Live mononuclear cells were isolated from the cell suspension by density centrifugation (Lymphoprep, Nycomed, Oslo, Norway), yielding a median of 1.0 × 106 mononuclear cells per biopsy in the IVF group (mean of 1.3 × 106 ± 1.05) and 0.5 × 106 in the control group (mean of 0.9 × 106 ± 0.8). Median weight of the biopsies was 1.1 g in the IVF group and 0.7 g in the control group. UMCs were used immediately upon isolation.
Peripheral blood mononuclear cells (PBMCs) were isolated by density centrifugation (Lymphoprep) from peripheral blood drawn just before the endometrial biopsy. For additional experiments [carboxyfluorescein succinimyl ester (CFSE) staining and isolation of T or NK cells], PBMCs were isolated from peripheral blood of healthy volunteers. In this case, the cells were frozen and stored in liquid nitrogen until use. All women gave informed consent according to the Medical Ethical Review Committee of the Radboud University Nijmegen Medical Center.
The HLA class I- and II-negative erythroleukaemia cell line K562 was obtained from the American Type Culture Collection (ATCC, Rockville, MD, USA). K562 cells were stably transfected by electroporation with the full-length transcript of HLA-G1 inserted in the pNGV1-vector (in-house-generated vector). K562-HLA-G(Eneg) cells have been described previously (Schleypen et al., 2003). K562-HLA-G(Eneg) transfectant cells were obtained with an HLA-G1 cDNA subcloned into pcDNA3.1(–). The P4 residue methionine of the signal sequence of HLA-G was changed by PCR-mediated mutagenesis to threonine that disables the signal sequence-derived ligand for HLA-E. Thus, no functional up-regulation of HLA-E is possible using this HLA-G expression vector.
HLA-G expression was confirmed by flow cytometry using the HLA-G-specific monoclonal antibody MEM-G/9 (Exbio, Prague, Czech Republic) or 56B (van Lierop et al., 2002).
Recombinant sHLA-G
Recombinant sHLA-G (G5) was purified from culture supernatant of Chinese hamster ovary (CHO) cells transfected with a construct containing the α1, α2 and α3 extracellular domains of HLA-G, an HLA-G-binding peptide (RLPKDFVDL) and β2-M, coupled to each other by linkers, as described previously (van Lierop et al., 2002). Briefly, a single-chain construct was made and cloned into the pNGV1 expression vector. CHO cells were transfected with this vector, and those cells expressing HLA-G were selected by staining the cell culture supernatant in a western blot with HCA2 (kindly provided by Prof. Dr H. Ploegh, Boston, MA, USA). Recombinant sHLA-G was purified in two steps: first, by ion-exchange chromatography using a Q-Sepharose column (Streamline Q XL, pH 8.0) with 0.8 M sodium chloride as elution buffer, and second, by immuno-affinity chromatography using a 56B-coupled NHS-sepharose column with 0.1 M glycine, pH 2.7, as elution buffer. The final eluate was neutralized and desalted. The concentration and purity of the recombinant protein was determined using densitometry from a Coomassie Brilliant Blue-stained protein gel containing both the recombinant protein as well as a protein standard range. The purity of the recombinant product was 85%; the concentration was 8.4 µg/ml. The 48-kDa recombinant protein and degradation products were specifically stained by 56B on a western blot as previously described (van Lierop et al., 2002). The recombinant sHLA-G was recognized by different HLA-G-specific (G233, MEM-G/9, 56B) and HLA class I-specific (W6/32) antibodies, in different enzyme-linked immunosorbent assays (ELISAs) as described previously (van Lierop et al., 2002). The functional activity of the recombinant product was confirmed using flow cytometric analysis of Annexin V expression on activated CD8+ cells. In brief, PBMCs (1 × 106 cells/ml) were cultured in the presence of phyto haemagglutinin (PHA) (5 µg/ml). After 3 days of culture at 37°C, CD4+ cells were removed by flow cytometric sorting, and the CD4-depleted fraction was cultured with or without sHLA-G (1 and 10 µg/ml). After 24 h, cells were stained by fluorescein isothiocyanate (FITC)-conjugated Annexin. In concert with the data of Fournel et al. (2000), 7% specific apoptosis was achieved.
Monoclonal antibodies and flow cytometry
Cells were phenotypically analysed by a direct one-step triple-labelling procedure. The following FITC-, phycoerythrin- (PE) or Cy5-labelled monoclonal antibodies were used: CD3, CD4, CD8, CD16, CD45 (all from Dako, Glostrup, Denmark) and CD56 (Coulter Immunotech, Marseille, France). The samples were run on a Coulter Epics XL Flowcytometer (Beckman Coulter, Fullerton, CA, USA), and 10000 events were collected based on live lymphocyte cell gating. This gate was set based on the staining of single separate sample with propidium iodide (PI, 5 µg/ml). Isotype-matched antibodies, usually below background staining, were used to define marker settings. Analysis of the data was performed using Coulter Epics Expo 32 software (Beckman Coulter).
Isolation of NK and T cells and T-cell enrichment
NK cells were purified from PBMCs by negative selection [a cocktail of anti-CD3, CD4, CD19 and CD33 antibodies using magnetic cell sorting (MACS)]. The purity of the NK-cell fraction was tested using flow cytometry and was not <95%. T cells were purified by negative selection (a cocktail of anti-CD14, CD19, CD33, CD16 and CD56). The purity of the T-cell fraction was tested using flow cytometry and was not <95%. To study IFN-γ production by T cells, we depleted CD56+ cells (NK cells) from PBMCs using MACS according to the manufacturer’s protocol (Miltenyi Biotech, Bergisch Gladbach, Germany). The NK-cell–depleted fraction always contained <1% NK cells as shown using flow cytometric analysis.
Cell culture
PBMCs or UMCs (5 × 104) were cultured in RPMI 1640 with glutamax supplemented with pyruvate containing 100 U/ml penicillin, 100 µg/ml streptomycine (all from Gibco, Paisley, UK) and 10% heat-inactivated pooled human serum (culture medium) in the presence of IL-2 (10 U/ml, Proleukine, Eurocetus, The Netherlands) or IL-15 (10 ng/ml, Biosource International, Nivelles, Belgium) and in the presence or absence of sHLA-G (1 µg/ml) or as a control; cells were cultured in culture medium alone. Phosphate-buffered saline (PBS) was used as a control because sHLA-G was dissolved in PBS. An equal volume of PBS was added to the cultures without sHLA-G. Furthermore, to confirm HLA-G specificity, we added the HLA-G-specific monoclonal antibody 56B (van Lierop et al., 2002) at the beginning of the culture in several experiments at 1 µg/ml.
After 3, 5, 7 or 9 days, half of the supernatant was harvested for cytokine measurements. Subsequently, 0.5 µCi [3H]thymidine was added to each well. The plates were harvested the following day using a Micromate 196 harvesting device (Canberra, Meriden, CT, USA) and counted in the Packard Matrix 96 Direct Beta counter (Canberra). All tests were performed in triplicate. Results are presented as the mean ± SD from triplicate wells.
Cell division kinetics according to CFSE dilution
The cell division rate of cells was studied by labelling viable responder cells with carboxyfluorescein diacetate succinimidyl ester (CFDA-SE, Molecular Probe, Eugene, OR, USA) just before stimulation. In the cell, esterases cleave the acetyl group of CFDA-SE, leading to the fluorescent diacetylated CFSE. At each cell division, the mean CFSE fluorescence halves. Simultaneous labelling of membrane markers enabled the study of division kinetics of lymphocyte subsets over time. Samples were analysed using flow cytometry at different time points. Data, preferentially 100000 live gate events, were analysed using Coulter Epics Expo 32 software.
Cytokine enzyme-linked immunosorbent assay
IFN-γ, TNF-α, (both from Pelikine Compact human ELISA kit, CLB, Amsterdam, The Netherlands) and VEGF (Biosource International, Camarillo, CA, USA) were measured in cell culture supernatants of UMCs and PBMCs by sandwich ELISA. The assays were performed according to the supplier’s manual. The results were measured photometrically at 450 nm using an ELISA plate reader (Titertek Multiskan MCC/340). Data are mean of a duplicate. VEGF data were corrected for background production of VEGF by irradiated 721.221 cells or 721.221 cells transfected with mHLA-G.
Cytotoxicity assay
The cytotoxic capacity of UMCs, PBMCs or purified peripheral NK cells was examined by 51Cr-release of 51Cr-labelled K562 cells or K562 cells transfected with mHLA-G or mHLA-G(Eneg). K562 cells (1 × 106) were labelled with 100 µCi 51Cr (Amersham, UK) and used as target cells at 1000 cells/well. Different effector/target (E/T) ratios were tested in triplicate. PBMCs or purified peripheral NK cells were pre-activated by overnight incubation in the presence of IL‐2 (100 U/ml IL-2). sHLA-G (1 µg/ml) or PBS (negative control) was added to the culture just before the addition of target cells. Culture supernatants were examined for released 51Cr on a γ-irradiation counter (Wallac 1470 γ-counter, Turku, Finland). Cytotoxic capacity is shown as percentage lysis of the indicated target, calculated according to the following equation:?
Statistical analysis
Differences between groups were analysed for significance (P < 0.05) by a two-sided Student’s t-test (Figures 1A,D, 3 and 5), by a paired two-sided Student’s t-test (Figure 1B) or, in case of three or four sample groups, by a one-way analysis of variance (ANOVA) (Figures 1C, 2 and 6). Post-testing was performed by Bonferroni’s multiple comparison test.
Results
sHLA-G inhibits IL-15-induced proliferation of UMCs but promotes IFN-γ and TNF-α production
To study the effect of sHLA-G on functional activity of UMCs, we cultured IL-15-stimulated UMCs in the presence or absence of sHLA-G. We used IL-15 as a growth factor because, in contrast to IL-2, this cytokine is abundantly produced in the uterus under healthy conditions (Verma et al., 2000). UMCs strongly proliferated in response to IL-15 with a peak response on day 6 of culture (Figure 1A). Co-culture with sHLA-G (1 µg/ml) resulted in an inhibition of the response varying between 20 and 65% inhibition (Figure 1B). HLA-G specificity of this inhibitory effect was confirmed by the addition of the anti-HLA-G monoclonal antibody 56B to the cultures, leading to a complete restoration of the proliferative response (Figure 1C).
Next, we determined levels of IFN-γ, TNF-α and vascular endothelial growth factor (VEGF) produced by UMCs in response to sHLA‐G. Upon co-culture with sHLA-G, IL-15-stimulated UMCs produced up to 7-fold increased levels of IFN-γ compared with UMCs cultured with IL-15 alone (Figure 1D). In addition, production of TNF-α was increased upon co-culture with sHLA-G (Figure 1D). Levels were up to 3-fold higher compared with UMCs cultured with IL-15 alone. In the absence of IL-15 and sHLA-G, UMCs did not produce IFN-γ or TNF-α. Although the difference was not statistically significant, IL-15 treatment caused a trend of enhanced background levels of VEGF produced by cultured UMCs up to 30% (Figure 1D). The addition of 1 µg/ml of sHLA-G to the culture reduced VEGF production to levels comparable to background levels in the absence of IL-15.
Summarizing, our data show that sHLA-G can directly affect the function of leukocytes that are in the uterus at the time of implantation. sHLA-G inhibits IL-15-induced proliferation and VEGF production and, notably, stimulates the production of pro-inflammatory cytokines.
Because UMCs from endometrial tissues are very difficult to obtain in sufficient numbers, we could not perform the analyses necessary to distinguish between the effects of sHLA-G on T and NK cells, separately. These studies were therefore performed with peripheral cells, although the relative amounts of NK cells and T cells differ between UMCs and PBMCs (Lukassen et al., 2004). Furthermore, the uterus contains a unique NK subset that cannot be found in peripheral blood and that has a very high expression of CD56.
sHLA-G does not affect natural cytolytic activity of UMCs
An important feature of the membrane-bound form of HLA-G is the inhibition of NK-mediated cytotoxicity (Rouas Freiss et al., 1997; Khalil-Daher et al., 1999). To address the question whether sHLA-G directly affects the natural cytolytic activity of UMCs, we performed a standard NK cytotoxicity assay using the HLA class I-negative cell line K562 as target in a 4-h 51Cr-release assay. To avoid the induction of lymphokine-activated NK-cell activity, we did not pre-activate UMC with IL-15. The data show that killing of K562 cells by UMCs was not affected by the addition of sHLA-G (Figure 2). According to expectation, the addition of HLA-G1-transfected K562 cells did inhibit the natural cytolytic activity of UMCs. The inhibitory effect of mHLA-G could not be reversed by the addition of sHLA-G. This might indicate that sHLA-G preferentially binds to a receptor that does not affect cytolytic activity of NK cells. So, in contrast to membrane-bound HLA-G, the sHLA-G does not affect UMC-mediated natural cytolytic activity.
sHLA-G differentially affects peripheral T and NK cells
Having established a regulatory effect of sHLA-G on UMCs, we extended our study to the effect of sHLA-G on peripheral lymphocyte subsets.
First, we determined the effect of sHLA-G on whole PBMCs. Similar to UMCs, PBMCs were stimulated with IL-15. In addition, we also studied the effect of sHLA-G on IL-2-activated PBMCs. Both IL-15 and IL-2 bind to the β- and γ-subunits of the IL-2 receptor. sHLA-G inhibited the proliferative response of both IL-15- and IL-2-stimulated PBMCs (Figure 3A,D), but, as with uterine cells, the production of IFN-γ and TNF-α was increased (Figure 3B,C,E,F). To differentiate between the effects of sHLA-G on distinct lymphocyte populations, we labelled PBMCs with CFSE, stimulated with IL-2 and cultured in the presence or absence of sHLA-G and analysed using flow cytometry at several time points during culture. On the basis of double staining of CD3 and CD56, we made a distinction between T cells (CD3+/CD56−), NK cells (CD3−/CD56+) and NK-T cells (CD3+/CD56+). CFSE analysis revealed that sHLA-G clearly stimulated NK-cell proliferation (Figure 4). In the presence of sHLA-G, 87% of the NK cells went through up to nine proliferation cycles in 7 days, whereas in the absence of sHLA-G, only 47% of the NK cells divided, with a maximum of seven proliferation cycles. Regarding CD4+ and CD8+ T-cell proliferation, we found that sHLA-G inhibited CD4+ T cells, but hardly affected CD8+ T cells (Figure 4). To confirm this differential effect of sHLA-G on NK- and T-cell proliferation, and to analyse IFN-γ production by these cells, we purified NK cells and T cells from peripheral blood and cultured them separately in the presence of sHLA-G and IL-2. The data confirm an increased NK-cell proliferation and a decreased T-cell proliferation in the presence of sHLA-G (Figure 5A,C). Notably, IFN-γ production by purified NK cells is increased in the presence of sHLA-G, as observed after 7 days of culture (Figure 5B). For T cells, the effect was more difficult to determine because in our hands IL-2 alone appeared insufficient to induce IFN-γ production in purified T cells (Figure 5D). Therefore, as an indirect approach, we depleted the NK cells from PBMCs and stimulated the remaining T cells, monocytes and B cells with IL-2. In this case, the presence of sHLA-G led to a reduction of IFN-γ production by the remaining T cells (Figure 5E). These data indicate that T cells produce less IFN-γ in the presence of sHLA-G, although it is not yet clear whether this is a direct effect on T cells or an indirect effect of sHLA-G via monocytes or B cells. Taken together, the data show that the increased production of IFN-γ by whole PBMCs upon co-culture with sHLA-G should be ascribed to an increased IFN-γ production by NK cells.
Similar to UMCs, the natural cytolytic activity of PBMCs (Figure 6A) as well as purified NK cells (Figure 6B) was not inhibited by sHLA-G but was inhibited by membrane-bound HLA-G. To exclude a possible effect of HLA-E, we used in these experiments K562 cells transfected with an HLA-G gene construct that carried a mutated signal sequence [K562-HLA-G(Eneg)]. Expression of HLA-G may lead to concomitant expression of HLA-E through binding of this molecule to the HLA-G-derived signal peptide. The mutation precluded binding of the resultant peptide to HLA-E and the resultant cell surface up-regulation of this complex. These data confirm the lack of the effect of sHLA-G on the lytic activity of NK cells. Again, sHLA-G also did not reverse the inhibitory effect of mHLA-G.
Summarizing, our data show that sHLA-G inhibits (cytokine induced) proliferation but promotes IFN-γ production of whole PBMCs. Moreover, analysing purified subsets, it appears that sHLA-G acts differently on T and NK cells. T-cell proliferation and IFN-γ production are inhibited by sHLA-G, whereas NK-cell proliferation and IFN-γ production are stimulated. The cytotoxic capacity of NK cells remains unaffected.
Discussion
The location and timing of sHLA-G expression during the first stages of implantation are suggestive of a role in modulating the local uterine environment in such a way as to promote a successful pregnancy. Although several studies have shown different effects of sHLA-G on peripheral lymphocytes, its direct effect on lymphocytes that are present in the uterus during the window of implantation was not studied so far. In this study, we used a recombinant form of sHLA-G to address the question whether sHLA-G affected proliferation and cytokine production of uterine lymphocytes. Our data show that sHLA-G inhibits proliferation and VEGF production by IL-15-stimulated UMCs, while stimulating IFN-γ and TNF-α production by these cells.
The effects of sHLA-G on cytotoxicity and cytokine production by UMCs are different from the effects observed for membrane-bound HLA-G. First, sHLA-G does not affect NK-mediated cytolysis, in contrast to mHLA-G. Second, the presence of sHLA-G leads to increased IFN-γ production by UMCs, whereas the presence of mHLA-G has been shown to result in a decrease in IFN-γ production by UMCs (Kanai et al., 2001b; van der Meer et al. 2004). Recently, in a collaborative study, we have shown that sHLA-G can activate peripheral NK cells by binding to KIR2DL4 (Rajagopalan et al., 2006). This activation was specific to sHLA-G and not to mHLA-G, because signalling of the receptor only occurred after the whole complex of sHLA-G and KIR2DL4 had been internalized into the cell in endosomes. Altogether, these data support the idea that sHLA-G has a specific role in human pregnancy that differs from that of membrane-bound HLA-G.
Furthermore, because sHLA-G is released from the cell, this molecule can perform its action at places further away from the invading trophoblast with an altogether different effect. In a mouse model, it has been shown that the combination of IL-15 and interferon regulatory factors are necessary for further differentiation of the uterine NK cells (Ashkar et al., 2003). The IFN-γ produced by uterine NK cells appears essential for vascular remodelling and necessary for successful pregnancy (Ashkar et al., 2000). This is achieved by up-regulation of genes in the decidua, the arterial smooth muscle cells and endothelial cells, leading to dilatation, elongation and branching (Croy et al., 2003). Also, in a pig model, it has now been shown that uterine lymphocytes at the implantation site produce IFN-γ (Tayade et al., 2006). Data on human peripheral NK cells show that the receptor that binds HLA-G, KIR2DL4 (Rajagopalan and Long, 1999), triggers strong IFN-γ production upon cross-linking (Rajagopalan et al., 2001; Kikuchi-Maki et al., 2003) or sHLA-G binding (Rajagopalan et al., 2006). sHLA-G binding to KIR2DL4 on peripheral NK cells has been shown to induce the production of a whole array of pro-angiogenic factors by peripheral NK cells (Rajagopalan et al., 2006). Our data support the idea that in the decidual tissue surrounding the implantation site, sHLA-G may contribute to the differentiation of uterine NK cells and vascular remodelling by locally increasing the IFN-γ production.
Although it is clear that sHLA-G are produced by fetal trophoblasts at the fetal–maternal interface, it is debated whether these proteins are a product of the alternative splice form HLA-G5 or rather a result of membrane shedding through metalloproteinase activity of membrane-bound HLA-G (sHLA-G1) (Blaschitz et al., 2005; Hunt and Geraghty, 2005; Le Bouteiller, 2005; Sargent, 2005). Regardless of their primary source, both soluble forms are equal in their α1, α2 and α3 domains, and it is thus not likely that they would have different functional effects. Binding studies show that both forms bind to KIR2DL4 (Rajagopalan et al., 2006). The sHLA-G that we used is a recombinant molecule consisting of sHLA-G (α1, α2 and α3), an HLA-G-binding peptide and β2-M coupled to each other by linkers, leading to a stable sHLA-G product (van Lierop et al., 2002). It was produced in mammalian cells (CHO cells), recognized by different HLA-G antibodies in sandwich ELISAs, identified in a western blot and shown to be functional in inducing apoptosis in CD8+ T cells. Nevertheless, it cannot be fully excluded that this protein functions distinctly from the naturally occurring sHLA-G because of differences in glycosylation or stability of the protein, as shown by McMaster et al. (1998).
Our data show similar effects of sHLA-G on UMCs and PBMCs, i.e. inhibition of proliferation and increased production of IFN-γ and TNF-α. It should be emphasized, however, that the composition of these two populations is quite different. In our endometrial biopsies, the leukocyte fraction contained ∼35% T cells and 35% NK cells, and the NK fraction consists largely of CD56bright NK cells. Furthermore, it contains a fraction of NK cells (CD56superbright) that is not present in peripheral blood (Lukassen et al., 2004). In peripheral blood, 70% of the leukocytes are T cells versus 10% NK cells and most of the NK cells are of CD56dim phenotype. Also, functional differences exist between uterine and peripheral blood NK cells (reviewed in Moffett-King, 2002). Although, in our experimental set-up, peripheral and uterine leukocytes respond similarly to sHLA-G with respect to proliferation and IFN-γ production, it is conceivable that functional differences exist between peripheral and uterine NK cells in response to sHLA-G, e.g. in the production of other cytokines as well as the level of cytokine production.
The finding of sHLA-G in biopsies and serum of transplant recipients (heart and liver/kidney) and its association with a decreased rejection rate has raised interest in the use of sHLA-G as a modulator of peripheral immune responses (Lila et al., 2002; Creput et al., 2003). There are several mechanisms by which sHLA-G could dampen such a response. Regarding the T-cell response, we show in concert with others (Lila et al., 2001) the inhibition of CD4+ T-cell proliferation. Recent data have shown that sHLA-G may induce regulatory T cells and thereby also indirectly affect T-cell proliferation (Le Rond et al., 2006). sHLA-G has been shown to induce apoptosis in activated CD8+ T cells (Fournel et al., 2000), although this is not always found (Bahri et al., 2006). The question is how the activation of the NK cells by sHLA-G and the concurrent IFN-γ production would fit this picture. Two other studies, using different forms of sHLA-G and different stimulation methods, also show that sHLA-G can enhance IFN-γ production by PBMCs (Kapasi et al., 2000; Kanai et al., 2001a). Our data suggest that this increase is due to IFN-γ production by activated NK cells. In general, IFN-γ is associated with a Th1 response, which is thought to be harmful in pregnancy as well as in transplantation. This might be a too simplistic view and be more dependent on timing, location and the type of cells involved. In a mouse model, it was shown that IFN-γ can activate autoantigen-specific regulatory T cells, which in turn results in increased production of immunosuppressive nitric oxide by APCs (Chen et al., 2006). Also, several other data point towards a more diverse role for IFN-γ in immunomodulation (Rosloniec et al., 2002).
On the basis of the role of IFN-γ in decidual vascularization and the pro-angiogenic cytokines and chemokines induced by sHLA-G in NK cells (Rajagopalan et al., 2006), one may also speculate that sHLA-G might affect vascularization of transplanted organs. In that case, the relation between sHLA-G levels and lower rejection rates in transplant recipient is not necessarily due to suppression of the alloresponse by sHLA-G but might be due to the contribution of sHLA-G to the neovascularization of the transplanted organ. Also, it can be envisaged that sHLA-G in combination with other cytokines or cells induces a distinct local cytokine microenvironment, e.g. via modulating antigen-presenting cells (APCs) (LeMaoult et al., 2004). The group of Horuzsko, who generated HLA-G transgenic mice, found that the maturation of myelomonocytic cells into functional APCs was compromised in these mice, resulting in a diminished cellular immunity (Horuzsko et al., 2001). In these mice, the (human-specific) HLA-G molecule appeared to bind and trigger the murine homologue of human ILT4 (the PIR-B inhibitory receptor) (Liang et al., 2002).
The precise role of sHLA-G in immune modulation will probably depend on location, cell type, receptor binding, concentration and form of sHLA-G. Of the three receptors that are known for HLA-G, one has a stimulatory function [KIR2DL4 (Rajagopalan et al., 2006)] and the two others appear to have mainly inhibitory functions [ILT2 and ILT4 (Le Rond et al., 2006)]. Expression of these receptors is different for different leukocyte populations. The form of sHLA-G may determine which receptor is bound. Monomeric sHLA-G has been shown to bind and trigger KIR2DL4 (Rajagopalan et al., 2006). In contrast, only sHLA-G coated to beads is able to inhibit proliferation via ILT2 and ILT4 (Le Rond et al., 2006). In the uterus, it is more likely that the main role is to trigger NK cells via KIR2DL4 to facilitate implantation and vascularization. In the periphery, T-cell responses play a more important role; there, the inhibitory role of sHLA-G via ILT2 and ILT4 (APCs) may prevail.
In conclusion, our data show that sHLA-G directly affects proliferation and cytokine production by lymphocytes that are present in the uterus at the time of implantation and of peripheral blood lymphocytes. The effects partly differ from the effects induced by membrane-bound HLA-G and may be indicative of a different functional role for sHLA-G. This protein may act further away from the invading trophoblast to trigger functional maturation of the uterine NK cells and thereby contribute to vascular remodelling and decidualization. The peripheral expression of sHLA-G may affect allograft acceptance by enhancing vascularization of a transplanted organ.
References
Author notes
1Department of Bloodtransfusion and Transplantation Immunology, 2Department of Gynaecology and Obstetrics, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands, 3Biocentrum University Munich, Planegg-Martinsried, Germany and 4Department of Pharmacology, NV Organon, Oss, The Netherlands