A Soluble Isoform of the Rhesus Monkey Nonclassical MHC Class I Molecule Mamu-AG Is Expressed in the Placenta and the Testis¹

The nonclassical MHC class I molecule HLA-G has been shown to be expressed on the surface of invasive extravillous trophoblasts of the human placenta, with potential significance for regulation of the maternal immune response to pregnancy (1, 2, 3, 4, 5, 6). As with other MHC class I molecules, HLA-G is expressed on the cell surface complexed with β₂-microglobulin and interacts with a variety of receptors on cells of the innate immune system, including NK cells and monocytes (7, 8, 9, 10, 11). One of the intriguing characteristics of the molecular biology of HLA-G is the expression of an alternatively spliced transcript coding for a truncated, soluble molecule (12, 13), now designated HLA-G5. These studies have also identified intron 4-retaining isoforms lacking the α2 domain (HLA-G6). In these transcripts, the fourth intron is retained, and the reading frame of the intron results in a unique 21-aa carboxyl terminus following the fourth exon. A stop codon in the intron 4 sequence deletes the 41 aa of the fifth and sixth exons, including the transmembrane domain. Initial studies with an Ab (16G1) raised against the unique C-terminal extension detected the protein within the placenta (14), but not in peripheral lymphocytes. In addition, several groups have now reported assays for detecting soluble HLA-G in peripheral blood and other biological fluids (15, 16, 17, 18). Unexpectedly, soluble HLA-G was detected not only in pregnancy, but also in nonpregnant women as well as males. Recent studies suggested that HLA-G5, one isoform of soluble HLA-G, can induce apoptosis in peripheral CD8⁺ T cells via a Fas-Fas ligand mechanism (19), can suppress an allogeneic proliferative T cell response in mixed lymphocyte cultures (20), and can inhibit peripheral blood NK cell-mediated cytotoxicity (21). These results collectively suggest that while membrane-bound HLA-G may interact with resident leukocytes within the decidua, soluble HLA-G may act on blood cells in the peripheral circulation in both pregnant and nonpregnant individuals.

The consistent observation of soluble HLA-G in nonpregnant women as well as men raises a number of provocative questions, including the source(s) of this molecule as well as its function both during and outside the setting of pregnancy. To address these questions in an experimental animal model, we have characterized the expression of a nonclasical MHC class I molecule, Mamu-AG, a putative HLA-G homolog, in the rhesus monkey placenta. Mamu-AG shares a number of biochemical and molecular features of HLA-G, including a relatively low level of polymorphism, the presence of alternatively spliced mRNAs and multiple isoelectric isoforms, and a high level of expression at the maternal-fetal interface (22, 23, 24). We have also previously shown that, as with HLA-G, the mRNA for Mamu-AG has restricted expression in extraplacental tissues (25). In the current study we have shown that the placenta as well as the testis in the rhesus monkey expresses a transcript that could encode a soluble Mamu-AG retaining the fourth intron, which contains a premature stop codon. Both placenta and testis demonstrate reproducible immunostaining with mAb 16G1, raised against the fourth intron-encoded C-terminal peptide of HLA-G. Expression in the testis was strikingly cell specific, suggesting that the rhesus monkey will be an excellent model to study the physiology of a soluble nonclasical MHC class I molecule.

Adult female rhesus monkeys (Macaca mulatta) used for timed matings were from the colony maintained at Wisconsin Regional Primate Research Center. Rhesus monkey placental tissues were obtained by cesarean section as we have previously described (26). Other tissues were obtained from healthy animals euthanized in other studies. All surgical procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and under approval of the University of Wisconsin Graduate School animal care and use committee.

In RT-PCR experiments, 1 μg total RNA was reverse transcribed to cDNA using an oligo(dT) primer and Moloney murine leukemia virus reverse transcriptase (PerkinElmer, Foster City, CA). The cDNA was amplified for 34 cycles, with each cycle consisting of 94°C for 30 s, 55°C for 30 s, and 72°C at 1 min, in a reaction mixture containing 10 mM Tris-HCl, 1.5 mM MgCl₂, 50 mM KCl, 0.25 mM of each dNTP, and 1.25 U AmpliTaq DNA polymerase (PerkinElmer). Primers used for the amplification and sequencing of rhesus MHC class I cDNAs are depicted in Table I. Amplification reactions using as template RT reactions to which no RNA or no reverse transcriptase had been added served as negative controls. The PCR products were separated on 3% agarose gels, isolated with the GeneClean II kit (BIO 101, Vista, CA), and subcloned using a TA cloning kit (Invitrogen, Carlsbad, CA). Subcloned fragments were sequenced using an ABI 377 automated sequencing machine and the BigDye Terminator mix (PE Applied Biosystems, Foster City, CA).

Tissues collected at surgery or necropsy were immediately prepared for frozen sections. After fixing in 2% paraformaldehyde for 4 h, the tissues were washed with PBS twice for 20 min each time. They were then dehydrated in 9% sucrose for 4 h and 20% sucrose overnight. The tissues were then embedded in OCT mounting medium (Sakura Finetek U.S.A., Torrance, CA) and frozen in isopentane cooled with dry ice and ethanol. The mouse mAbs 25D3, 16G1, and goat anti-β₂-microglobulin (DAKO, Carpinteria, CA) were used at a concentration of 5 μg/ml. W6/32 was used at 1 μg/ml. A mouse IgG1 κ Ab (Sigma-Aldrich, St. Louis, MO) or goat polyclonal IgG antiserum (Santa Cruz Biotechnology, Santa Cruz, CA) were used as negative controls at the same concentrations as the primary Abs. Immunostaining was performed as previously reported (23, 24). Peptide blocking experiments were performed on placental and testicular sections with 16G1 by incubating the Ab with its immunizing peptide or a nonspecific peptide, GnRH, as a control. The Ab was mixed with either peptide at molar concentrations of 4/1 and 10/1 (peptide/Ab) and incubated for 30 min at 0°C. During this time, the sections were blocked with animal serum. After Ab neutralization, the mix was added to the sections for 1 h. 16G1 alone and mouse IgG were used as additional controls. The remainder of the procedure was performed as described above. Peroxidase and alkaline phosphatase kits were used along with diaminobenzidene, Vector NovaRed, Vector Red, and Vector Blue substrate kits (Vector Laboratories, Burlingame, CA). Some slides were counterstained with hematoxylin.

For ISH, placental and testicular tissues were prepared for paraffin sections. Tissues were fixed in 2% paraformaldehyde for 4 h and then embedded in paraffin. Ten-micrometer sections were cut and processed as previously described (23). A digoxigenin-labeled 141-bp RNA probe derived from exon 2 of Mamu-AG, as described previously (23), was generated for ISH (digoxigenin labeling kit, Roche, Indianapolis, IN). Transcription of sense and antisense RNA was confirmed by dot blot and gel electrophoresis. Hybridization buffer containing 400 μg/ml probe was applied to sections, which were then covered with a glass coverslip to prevent drying, and incubated at 55°C for up to 40 h. Slides were washed extensively to ensure specific hybridization (23). The slides were washed in 2× SSC with 0.1% SDS (0.3 M sodium chloride and 0.03 M sodium citrate) for 30 min, placed in 2× SSC at 60°C for 1 h, 0.2× SSC at room temperature for 15 min, and 0.1× SSC at 60°C for 1 h. Sections were blocked with 25% goat serum in PBS for 30 min and then incubated with alkaline phosphatase conjugated anti-digoxigenin (1/2000; Roche) for 2 h. Sections were washed in TBS for 2, 10, and 10 min. Slides were incubated in detection buffer for 5 min and then covered with the substrate BM-purple (Roche) overnight.

The RNA for Mamu-AG5 was amplified using RT-PCR and primers in exons 2 and 6 containing the start codon and the stop codon, respectively (Table I, primers 6 and 7). The sequence for an enterokinase cleavage site (GACGATGACGATAAG) and hemagglutinin (HA) protein tag (TACCCATACGATGTTCCGGATTACGCTAGCCTC) were then added to the 3′ region of the transcript by a second round of PCR. The resulting cDNA was directly sequenced and subcloned into pGEM vector (Promega, Madison, WI) for subsequent transfection into 293 cells. 293 cells were transiently transfected according to a standard calcium phosphate transfection procedure, and cells were harvested after 48 h of culture, lysed, and prepared for Western blotting. Protein samples were prepared from cultured transfected 293 cells by lysis in 25 mM Tris-phosphate (pH 7.8), 2 mM DTT, 2 mM 1,2-diaminocyclohexane-N,N,N′,N′-tetraacetic acid, 10% glycerol, and 1% Triton X-100 , (E1531; Promega). Samples were quantitated using the Bradford assay. Discontinuous SDS-PAGE was performed using a 10% acrylamide resolving gel and a 3.9% stacking gel. Fifty to 100 μg cell extract was fractionated along with Rainbow markers (RPN 800; Amersham, Arlington Heights, IL) to estimate m.w. After electrophoresis, proteins were electroblotted onto a polyvinylidene difluoride membrane (Bio-Rad, Hercules, CA). The membrane was washed twice with TBS and blocked with a 0.2% nonfat dry milk/TBS solution overnight at 4°C. Following blocking, the membrane was washed three times with a 0.1% Tween 20/TBS solution, then incubated with the anti-HA Ab (0.5 μg/ml; Roche) or 16G1 (5 μg/ml) for 2 h at room temperature. The membranes were washed three times for 10 min each time and then incubated with alkaline phosphatase-conjugated goat anti-mouse secondary Ab (1/3000; Bio-Rad) for 1.5 h. The membrane was washed three times and incubated with Immun-Star chemiluminescent substrate solution (Bio-Rad) for 5 min, and immunoreactive bands were visualized by autoradiography.

We investigated whether the placenta expresses a transcript for Mamu-AG retaining intron 4, as with the soluble HLA-G transcripts previously reported in the human placenta and other tissues. First, we amplified reversed transcribed cDNA with primers 2 and 5 located in the fourth and sixth exons, respectively, reasoning that since the mRNA was only detectable by RT-PCR in human placenta, we might find the same situation with the rhesus placenta. Fig. 1,B demonstrates that several amplicons dependent on reverse transcription were obtained from rhesus placental RNA. Whereas the major band seen in Fig. 1,B primarily contained the expected sequences of the fourth, fifth, and sixth exons, cloning and sequencing of the band indicated by an arrow revealed the presence of an intron highly homologous to the human HLA-G intron (see below). The intron sequence from this mRNA was confirmed to represent intron 4 in its entirety by sequencing PCR product amplified from rhesus genomic DNA (not shown). Most significantly, a stop codon was noted at the same location as that for soluble HLA-G (Fig. 1 C). Since the intron 4-containing PCR amplicon might be derived from a splicing intermediate, we sought to confirm the splicing at exon-intron junctions 5′ from the fourth intron region in putative soluble Mamu-AG transcripts. For this experiment we designed downstream primer 4 within the fourth intron, and upstream primer 1 within the second exon (α1 domain) to determine whether transcripts containing the fourth intron also contained other upstream introns. We cloned and sequenced seven placental cDNAs amplified by these primers from three different placentas, and none was found to contain intron 2 or 3 (not shown). We concluded that the soluble Mamu-AG cDNAs represent mature mRNAs retaining only the fourth intron. In addition, these clones all contained exons 2, 3, and 4, demonstrating that in our hands only a Mamu-AG mRNA homologous to HLA-G5 is detected in the placenta.

The predicted amino acid sequences for the fourth intron of Mamu-AG and the fifth intron carboxyl terminus were determined. Alignment of Mamu-AG intron 4 peptide with that of HLA-G revealed that the predicted amino acid sequences were 86% identical between soluble MamuAG and soluble HLA-G intron 4-derived carboxyl termini (Fig. 1,D). The amino acid substitutions between rhesus and human intron 4 peptides were mostly conservative. Sequencing of placenta-derived amplicons also resulted in the identification of an additional novel cDNA, which retained 18 nucleotides derived from the 3′ end of the fifth intron (Fig. 1,E, Mamu-AGv5). All of these clones derived from placental mRNA lacked intron 4. The entire intron 5 was also cloned and sequenced by PCR from rhesus genomic DNA, and was 442 bp in length (not shown). The inclusion of the intron 5 fragment resulted in an insertion of a novel six-amino acid carboxyl-terminal extension N-terminal to the two amino acids encoded by exon 6 (Fig. 1 F). We are not aware of any reports of a similar splice variant in human MHC class I mRNAs.

Our previous studies of the expression of Mamu-AG demonstrated that the mRNA is expressed in a variety of tissues, with the highest nonplacental level of Mamu-AG mRNA in the testes (25). Those studies were conducted with a ribonuclease protection assay selective for the Mamu-AG α 1 domain (exon 2). We re-evaluated a number of rhesus nonplacental tissues by RT-PCR with primers 3 and 5 that would selectively amplify mRNAs containing the exon 4/intron 4 junction. Fig. 1,G presents representative results from several tissues. As expected, placental tissues had abundant intron 4-containing Mamu-AG mRNA, while in most tissues this isoform was undetectable (eye, adrenal, lymph nodes, skeletal muscle, skin, heart, lung, liver, and pancreas; not shown) or barely detectable (ovary, Fig. 1,G; kidney and intestine, not shown). Several tissues had a moderate, but consistent, expression of intron 4-containing mRNA, including amniotic membranes, testes, spleen (Fig. 1 G) and thyroid (data not shown).

Although to our knowledge the retention of intron 4 is unique in humans to HLA-G and is not known in classical loci, the question arose of whether the PCR fragments amplified represented rhesus classical loci. Sequencing of multiple PCR-generated clones from both testicular and placental RNA indicated that all transcripts from these tissues contained an entire intron 4, and all contained the diagnostic TAA stop codon in exon 6. Four of the six testicular and one of three spleen (but none of 19 placental) clones sequenced that retained intron 4 also contained 18 nt of the 3′ end of intron 5 discussed above. Since the stop codon was located within intron 4, the intron 5 sequences do not contribute to the coding sequence for these nonplacental transcripts. Several amplicons from thyroid and spleen also were found to contain the stop codon diagnostic for Mamu-AG (Fig. 2) (22, 27, 28). Surprisingly, one transcript from spleen and two from thyroid were apparently derived from classical MHC class I molecule(s), since they did not contain the diagnostic stop codon (Fig. 2). Further work will be required to understand the significance of this observation.

We wished to explore the expression of soluble Mamu-AG isoforms in rhesus monkey tissues with the Ab 16G1 raised against the carboxyl-terminal region of soluble HLA-G encoded by intron 4 sequences. To evaluate recognition of Mamu-AG5 by 16G1, a Mamu-AG5 cDNA was cloned into an expression vector containing sequences encoding an in-frame HA tag at the 3′ end of the Mamu-AG5 cDNA. 293 cells were transiently transfected with the construct, and cell extracts of transfected and naive cells were analyzed by Western blot. Parallel analysis was performed with 16G1 and an Ab against the HA epitope (Fig. 3). Fig. 3,A demonstrates that extracts of transfected, but not naïve, cells expressed three immunoreactive bands of ∼35–39 kDa as identified by the anti-HA Ab. Fig. 3 B demonstrates that an identical trio of bands was identified by Ab 16G1 in transfected, but not naïve, cells.

We next conducted immunocytochemical analysis for Mamu-AG in these tissues, evaluating three placentas and testes and epididymi from five different animals. We used an mAb (25D3) against Mamu-AG we have previously described (24), as well as an mAb against the intron 4 peptide of HLA-G (16G1) (12). Consistent with previous reports, 25D3 identified Mamu-AG on the villous syncytial apical membranes (Fig. 4, A and H), in the proximal columns (Fig. 4,F), and in the cytotrophoblastic shell delineating the maternal-fetal interface (Fig. 4,J). Mesenchymal cells in the core of the villi (Fig. 4 H) as well as the maternal decidua and uterus were consistently negative.

By contrast, immunostaining with the 16G1 Ab was localized to the syncytia, villous cytotrophoblasts, and some villous mesenchymal cells of day 36 placentas (Fig. 4, B, D, G, and I). Staining was often pronounced on the syncytial border toward the intervillous space and the cytotrophoblastic edge facing the core of the villi (Fig. 4,I), suggesting cytoplasmic localization. Diffuse staining was also observed throughout both villous syncytial and cytotrophoblastic cell layers (Fig. 4, B, D, and I). Large cells with abundant cytoplasm within the villi, presumably Hofbauer cells, frequently demonstrated positive staining with 16G1 (Fig. 4,I). Blood vessels within the villous stroma were also positively stained (Fig. 4,B). Extravillous cytotrophoblasts, notably in the cytotrophoblastic shell, were not stained by 16G1, unlike the results with 25D3 (compare serial sections Fig. 4, F and G; also Fig. 4,J). The specificity of immunostaining was demonstrated by the absence of immunostaining when the 16G1 primary Ab was preincubated with the immunizing peptide (Fig. 4,C), but not with an unrelated peptide (Fig. 4,D). A nonspecific primary antiserum likewise gave no positive staining (Fig. 4 E).

We next conducted similar immunostaining experiments with rhesus testes. 25D3 staining in the testis was minimal and was localized exclusively to occasional cells in the interstitium surrounding the seminiferous tubules (Fig. 5,G). Germ cells and Sertoli cells within the tubule were consistently negative. Although only occasional interstitial cells were 25D3 positive, these cells were found to be consistently positive on serial sections, occasionally appearing in small aggregates. Other sporadic cells were typically within peritubular regions, and immunostaining of serial sections suggested that at least some of these cells were positive for CD68, indicating their likely identity as macrophages (Fig. 5,H). Serial sections (Fig. 5, F–H) demonstrate interstitial macrophages that appear to coexpress Mamu-AG, soluble Mamu-AG, and CD68. Epidydimal structures and mature sperm were not positive for Mamu-AG with 25D3 (not shown).

Analysis of rhesus testes with 16G1 revealed prominent immunostaining in the seminiferous tubules, with only occasional positive cells in the interstitium (Fig. 5, A, C, and F). Staining was notably found in the mural aspect of the tubular compartment, through the entire cross-section of the seminiferous tubules, suggesting staining of Sertoli cells (e.g., Fig. 5,F; see also Fig. 6). Individual Sertoli cells were identified by their position passing between germ cells and their pointed, irregular nuclei (Figs. 5,J and 6), with cytoplasmic processes extending toward the lumen and surrounding the nuclei of advanced germ cells. Staining was not noted in primary spermatocytes in early metaphase during meiosis, identified by their distinct speckled chromatin structure (S1, Fig. 6). However, 16G1 immunostaining was abundant in late stage primary and secondary spermatocytes and spermatids, with distinctly condensed nuclei (Sd, Fig. 6). Variable staining was observed in spermatozoa and residual bodies. Spermatozoa undergoing spermiogenesis were positively stained, while those that appeared to have completed the process and were free in the lumen were generally not stained (Fig. 5F). Residual bodies free in the lumen were sometimes stained by 16G1. Staining was not observed in the epidydimi or in mature sperm. A summary of immunostaining results is presented in Table II.

The discrepancy between 25D3 and 16G1 immunostaining in the testis was both intriguing and puzzling. In particular, it is surprising that 25D3 did not detect Mamu-AG within the seminiferous tubules. Since the 25D3 Ab only recognizes surface-bound Mamu-AG complexed with β₂-microglobulin, it seemed possible that 16G1 immunostaining is detecting soluble Mamu-AG H chain that is not complexed with β₂-microglobulin. Immunostaining confirmed this hypothesis (Fig. 5, D and E), since while abundant β₂-microglobulin was identified within the interstitial tissue, the seminiferous tubules were devoid of immunostaining. Interestingly, Sertoli cells do not express β₂-microglobulin.

We also evaluated several tissues previously shown to express low levels of Mamu-AG mRNA (25). Adrenal, spleen, ovary, kidney, heart, and pituitary samples did not reveal any Mamu-AG expression by immunostaining with 25D3 or 16G1 (data not shown).

Immunocytochemical analyses indicated that the Mamu-AG transcripts we detected by RT-PCR were expressed in the seminiferous tubules. To confirm that the tubules contain Mamu-AG mRNA, we conducted ISH with a probe for the α1 domain, with which we have previously defined conditions for locus-specific ISH (23). As we have previously reported, the villous syncytiotrophoblasts and trophoblastic shell consistently expressed Mamu-AG mRNA (Fig. 7,A). In addition, villous cytotrophoblasts and occasional cells within the core of the villi were positive (Fig. 7,B). Maternal decidua and uterine tissue were always negative, and no hybridization was observed with the sense probe (Fig. 7 C).

ISH with rhesus testes localized Mamu-AG mRNA mainly in the seminiferous tubules (Fig. 7,D). Most cells in the process of spermatogenesis hybridized with the antisense probe. Cells occupying the mural compartment of the seminiferous tubule, primarily spermatogonia and Sertoli cells, had relatively lower hybridization signal compared with spermatocytes (Fig. 7,D). Higher magnification (Fig. 7,E) revealed that spermatogonia contained less Mamu-AG mRNA than the germ cells closer to the lumen, i.e., primary spermatocytes and spermatids. Sertoli cells were difficult to identify without counterstaining during ISH. Hybridization with spermatozoa was uncommon, but was occasionally noted. In general, all the seminiferous tubules in the testis sections demonstrated some degree of mRNA expression. A few of the surrounding interstitial cells of the testis, possibly macrophages or Leydig cells, were positive for Mamu-AG mRNA (Fig. 7,E). However, their staining intensity was generally observed to be lower than that in germ cells. Sections incubated with the sense probe revealed no staining in testes (Fig. 7 F).

Finally, we conducted RT-PCR for Mamu-AG mRNA exons 4–6 with samples of ejaculated rhesus sperm. Cloning and sequencing of amplified cDNAs revealed that both whole semen and “swim-up” fraction sperm contained Mamu-AG mRNA; however no transcripts containing intron 4 were detected.

We have demonstrated the expression of a mRNA encoding a soluble form of Mamu-AG, a nonclassical MHC class I molecule expressed at high levels in the rhesus monkey placenta. The mRNA is expressed in the testes and several other nonplacental tissues at moderate levels. Immunostaining with an Ab recognizing the soluble HLA-G carboxyl-terminal peptide showed that the expression of soluble Mamu-AG in the placenta overlapped but was not identical with that of cell surface Mamu-AG. While we found only occasional cells expressing cell surface Mamu-AG in the testes, we also demonstrated widespread, although apparently cell-selective, expression of soluble Mamu-AG in the seminiferous tubules of the testes. Only a few other tissues had detectable soluble Mamu-AG mRNA, and no immunostaining was noted outside the placenta and testes in the tissues evaluated.

Soluble MHC class I molecules can be generated by several mechanisms. Molecules can be shed from the cell surface, or surface-bound molecules may be liberated by extracellular proteases (29). In addition, alternative splicing can generate MHC class I isoforms that are released from the cell. Classical MHC transcripts lacking exon 5, which encodes the transmembrane domain and therefore will be released from expressing cells, have been identified in the human and the mouse (30, 31, 32). Although the role(s) of soluble MHC is not completely defined, it may be involved in immunoregulation and tolerance in the setting of organ transplantation (33).

A novel mechanism for generating a soluble MHC class I molecule was revealed in studies of HLA-G, a nonclassical molecule expressed at high levels in the human placenta. Exon 5 encodes the membrane-spanning domain, and exons 6–8 encode the short cytoplasmic domain, which contains a stop codon 18 aa upstream of the conventional MHC class I molecule stop codon. An HLA-G transcript that included the fourth intron was first identified in placental RNA (12, 13). Translation of the intron results in an in-frame stop codon that terminates the protein upstream of the transmembrane domain. The resulting protein has a novel carboxyl terminus and is able to complex with β₂-microglobulin, but would be released from the cell due to the lack of a transmembrane domain (12, 13). In addition to the placenta (14, 34), soluble HLA-G has also been reported in CD4⁺ T cells (20), activated monocytes (35), thymic epithelium (36), and lung tumor cells (37). It is intriguing that like HLA-G, the Mamu-AG gene also is alternatively spliced to give rise to a soluble molecule. The retention of this novel splicing pattern is an additional facet of the molecular biology of the Mamu-AG locus that suggests a close functional homology with HLA-G. The splice patterns of HLA-G are complex, with various combinations of exons identified in human normal and transformed cells (12, 13, 38, 39, 40). However, the mechanisms that control splicing patterns in different cell types are not understood. Whether the mechanisms that direct the retention of intron 4 in Mamu-AG are related to those controlling HLA-G gene expression remains to be determined. A difference noted between the soluble HLA-G and soluble Mamu-AG isoforms is that whereas the predominant soluble HLA-G isoform may be HLA-G6 (18), we did not detect Mamu-AG6 mRNA in the rhesus placenta, although the rhesus clearly expresses membrane-bound Mamu-AG2 mRNA (22, 25). Alternatively, perhaps the placenta is not the primary source of circulating HLA-G.

An mAb (16G1) directed at the soluble HLA-G carboxyl-terminus tail (12) has been used to study the expression of soluble HLA-G isoforms in the human placenta. Using this Ab, soluble HLA-G has been localized within the extravillous trophoblasts and has been reported in association with human villous cytotrophoblasts as well as in some cases with the syncytiotrophoblasts (5, 14) and villous endothelial cells (34). Finally, some macrophages within the placental villi were also seen to express soluble HLA-G (14). It is intriguing that our current results generally parallel these observations in the human placenta. Soluble Mamu-AG is also noted not only in syncytiotrophoblasts, but within cells of the villous stroma, apparently including Hofbauer cells, and villous cytotrophoblasts of the rhesus placenta. Our previous studies have shown that β₂-microglobulin is readily detected within rhesus syncytiotrophoblasts and in villous stromal cells, but very low staining is found within villous cytotrophoblasts (23). The immunostaining pattern obtained with 16G1 would suggest that villous cytotrophoblasts, but not syncytiotrophoblasts or Hofbauer cells, are likely to produce free H chain Mamu-AG. Recently, several groups have sought to establish whether soluble HLA-G is present in the serum of pregnant women (15, 16, 17, 18). ELISA-based assays using various Ab capture and detection approaches have indicated the presence of soluble HLA-G not only in the serum of pregnant women, but also in nonpregnant women and in men as well. Whether soluble Mamu-AG is detectable in the serum of the rhesus monkey remains to be investigated.

The pattern of expression of soluble Mamu-AG expression in the placenta overlaps but is not identical with that of the membrane-bound molecule. This may reflect the putative targets for these two isoforms. Soluble Mamu-AG immunoactivity was relatively low in rhesus extravillous cytotrophoblasts, whereas membrane-bound Mamu-AG was readily detected. One might speculate that membrane-bound HLA-G and Mamu-AG on extravillous trophoblasts will ligate receptors on adjacent decidual leukocytes to regulate their activity, whereas a soluble circulating molecule may modulate maternal immune cell activity both locally, within the decidua, as well as systemically. In this situation the syncytiotrophoblasts are a more effective way to release a soluble molecule directly into the maternal circulation. The expression of soluble HLA-G by the syncytiotrophoblasts remains somewhat controversial (5, 14, 41), and detection may be dependent on the conditions used for fixation, embedding and processing as well as the Abs used. Indeed, conflicting results on placental and extraplacental localization of HLA-G support both further study as well as the development of appropriate animal models.

The pattern of soluble MHC class I expression in the rhesus testis was unexpected. HLA-G protein was not detected in human testes in previous studies (5, 41), although mRNA expression has been reported in some, but not all, studies (42, 43). In our studies with the rhesus, whereas there was only very rarely detectable membrane-bound Mamu-AG within the seminiferous tubules (as determined by 25D3 staining), it was clear that abundant soluble protein was present in selected germ cell populations, colocalized with Mamu-AG mRNA. The lack of β₂-microglobulin within the rhesus seminiferous tubules sufficiently explains the lack of cell surface protein, an observation made previously with the human testis and the pan-MHC class I Ab W6/32 (43). Free class I H chain was detected in some spermatocyte populations in humans (43). However, it seems unlikely that testicular soluble HLA-G may be the source of circulating serum HLA-G in men, since the patency of the blood-testis barrier would preclude this route of trafficking of the molecule. In addition, the assays used to detect soluble HLA-G typically detect the class I-β₂-microglobulin complex (15, 16, 17, 18), and immunostaining of human testes with 16G1 did not detect soluble HLA-G protein (41). Thus, another site of expression of human soluble HLA-G seems likely.

It is intriguing to speculate on a function for seminal soluble Mamu-AG in the female reproductive tract, perhaps suppressing an innate immune response within the vagina or the uterus to male leukocytes within the ejaculate. A role in controlling T cell expansion, blocking CTL activity, and inducing apoptosis in alloreactive T cells has been suggested for soluble HLA-G (19, 20, 44). Alternatively, soluble Mamu-AG might also help prevent anti-sperm immune responses following minor breaks in the blood-testis barrier of the seminiferous basement membrane. In one tubule we noted high expression of membrane-bound Mamu-AG on a cellular aggregate (not shown) that also stained positively for the macrophage marker CD68. The expression of HLA-G has been noted in macrophages invading psoriatic skin lesions (45) and in activated macrophages and dendritic cells in lung tumors (34, 46). Although the expression of soluble or membrane-bound HLA-G in placental macrophages is not fully resolved (5, 14, 41), the rhesus monkey may be an experimental model in which to explore the functional significance of soluble nonclassical MHC class I molecules in vivo.

We thank Stephen G. Eisele and the Reproductive Services Unit for timed matings and rhesus semen samples, the veterinary staff of the Wisconsin Regional Primate Research Center for surgical assistance, members of the Golos laboratory for critical reading of the manuscript, Carlos Quian-Suarez (Georgetown University, Washington, D.C.) for advice on histological interpretation, and Steven Busch for editorial assistance. We acknowledge the assistance of George Flentke and the Microanatomy Core of the University of Wisconsin Developmental Toxicology Center for helpful training and advice for riboprobe preparation.