CD3- Leukocytes Present in the Human Uterus During Early Placentation: Phenotypic and Morphologic Characterization of the CD56++ Population

In this study, the CD3- LGL/NK cells present in the pregnant human uterus have been characterized. Phenotypic and morphologic analyses of decidual LGL revealed many similarities to the minor CD56bright+, CD16- subset in peripheral blood, but there were some important differences. The relative surface density of CD56+ is greatly increased on decidual LGL to 22 x that found on the majority of CD56 peripheral blood NK cells. The CD56bright cells in decidua show LGL morphology, whereas in peripheral blood, they are .mainly agranular. Proliferation of CD56+ cells occurs predominantly during the nonpregnant secretory (luteal) phase, indicating these CD56+ uterine LGL do not migrate as terminally differentiated cells. The appearance of CD56 cells was examined at the ultrastructural level using immunoelectron microscopy. Cells with phenotypic characteristics of decidual LGL occur in a higher percentage (1.11%) in the peripheral blood of women of reproductive age than in men (0.66%). On the basis of these results, it is proposed that the CD56bright+ uterine leukocytes represent a distinctive, hormonally regulated subset possibly adapted to control human placentation.


INTRODUCTION
At the time the human placenta is established, the predominant leukocytes (>70%) present in the maternal uterine mucosa (decidua) are large granular lymphocytes (LGL) with a distinctive phenotype (CD3-, CD16-, CD56/) and natural killer (NK) function (Starkey et al., 1988;Bulmer, 1989;King et al., 1989aKing et al., , 1989bManaseki and Searle, 1989). There is a significant variation in the numbers of these cells during the menstrual cycle, indicating that their recruitment to the uterus is probably under hormonal control. However, it has not yet been established to what lineage these decidual LGL belong. In peripheral blood, a very small subset (< 1%) of LGL appears to have similar phenotypic and functional characteristics (Lanier et al., 1986), but it is still unclear how they relate to decidual LGL. Recently, *Corresponding author.
Lanier and his colleagues have made a comprehensive analysis of peripheral blood LGL/NK cells on the basis of both qualitative and quantitative expression of their cell-surface markers, together with morphological features such as granularity and the degree of NK cytolytic activity (Nagler et al., 1989). We now report a similar analysis of decidual LGL. In addition, we have studied these cells at the ultrastructural level using immunoelectron microscopy and have also compared the numbers of CD3-, CD56 bright+ LGL in peripheral blood of males and females to see whether there are any sex differences. To assess whether CD56 / LGL migrate to the uterus as terminally differentiated cells or whether they can proliferate in situ, we have used double immunohistological staining for the proliferation marker, Ki-67 combined with CD56 on frozen sections of pregnant and nonpregnant endometrium. Our findings lead us to conclude that decidual LGL may represent a unique uterine-specific lymphoid population.

170
A. KING et al.

RESULTS
Characterization of CD56 bright+ P.eripheral Blood Lymphocytes in Males and Females Two-colour immunofluorescence of PBL with conjugated-Leu-19 (CD56) plus  (CD16) and Leu-4 (CD3) followed by FITC. GAM revealed three distinct subsets (Fig. 1) with a small population of CD3-, CD16-, CD56 bright+ cells ( Fig. 1; Box R1), as previously described (Lanier et al., 1986). We also found that this subset comprised < 2% of peripheral blood lymphocytes. The proportion of cells with this phenotype was compared in eight females aged 18-25 years and seven similarly aged males ( Table  1). Histograms of a representative male and female are shown in Figs. 2a and 2b. It was found that the mean percentage of these cells n males (0.66% of total PBL) was significantly lower than in females (1.11%) (Student's test,t=3.43, p=<.005). The proportion of CD3-, CD16-, CD56 bright+ cells was analyzed in three females weekly throughout the menstrual cycle. Although there was a slightly lower proportion of these cells in the secretory (luteal) phase in two women, these differences were not significant (data not shown). Mean 0.66 + 0.03 Mean 1.11 + 0.11 aAnalysis was performed randomly throughout the menstrual cycle in females. The time in the cycle is given from the first day of the first menstrual period.
Characterization of Decidual Leukocytes LCA. Six preparations-of decidual leukocytes, pooled from three to four pregnancies and isolated by mechanical sieving followed by separation over Lymphoprep, were stained for liukocyte common antigen (CD45) to assess the success of our extraction method for bone-marrow-deriVed cells. In these FIGURE 1. PBL were purified over Lymphoprep and stained with PEconjugated anti-CD56 (Leu-19) followed by anti-CD3 (Leu-4) and anti-CD16 (Leu-11b) with a secondary FITC-conjugated antibody. A representative two-color contour map displaying PE (CD56) and FITC (CD3 and CD16) fluorescence. Three (Fig. 2c) on the basis of fluore-LCA / at the appropriate gate setting (Fig. 3). scence intensity. The majority (80%) are CD56 bright+. CD56. The proportion of CD56 + cells in 12 Lanier has shown that the CD56 bright+ PBL populadecidual cell preparations was analyzed. 70% + 12% tion expresses 3.5 more Leu-19 antigen than the of cells were CD56 /. As in peripheral blood, the CD56 dim+ population (Lanier et al., 1986) CD56 + antigenic density levels to that of peripheral that in CD56 dim+ PBL and 4 x brighter than blood (Fig. 4). It was found that CD56 is '5x CD56 bright+ PBL. In decidual LGL, the CD56 bright+ brighter on CD56 bright+ PBL than CD56 dim+ PBL in population is 14x brighter than CD56dim++. After males and 6.4 x brighter in females. In CD56 / culturing in rlL-2 for 1-4 days, there was a further decidual cells, the antigenic density of CD56 is 22 x slight increase in the density of CD56.   Morphology of CD56 Cells. The morphology of CD56 / cells was analyzed by forward and side scatter in both peripheral blood and decidua. The CD56 bright+ decidual cells appear larger and more granular than their peripheral blood counterparts (Figs 5a and 5b). The decidual CD56 dim+ subset was also granular (results not shown). One decidual cell preparation was sorted into CD3-and CD56 bright+ cells and cytospin smears of CD56 bright+ cells were stained with Giemsa. More than 75% of CD56 bright+ cells were observed to be granular ( Fig. 6a) as opposed to 0.5-25% of CD56 bright + cells in peripheral blood (Lanier et al., 1986).
Phenotype of CD56 bright+ and CD56 aim+ cells A detailed phenotypic characterization of CD56 / decidual leukocytes was performed using two-color immunofluorecence. Each Mab was tested on at least three separate occasions. Live cells were analyzed and markers defining quadrants were positionedto exclude >99% of unstained cells in the lower left quadrant of a negative control sample, which included a minor autofluorescent population. Because there was a biphasic expression of the CD56 antigen, the data were also analyzed by setting regions around each distinct population. All the data shown were analyzed using the same fixed regions. This also allowed the fraction of autofluorescent cells, which may overlap with the weakly staining cells, to be subtracted from the results. These results are summarized as a histogram ( Fig. 7) and contour diagrams (Fig. 8).
T-Cell Markers. It can be seen that the majority of CD56 / decidual cells co-express CD2, CD7, CD45RA, CD11a, and CD18 like their peripheral blood counterparts, but not mature T-cell markers such as CD3 and CD5. We find there is a differential expression of CD2 and CD7 surface antigens on CD56 bright+ and CD56 dim+ decidual subsets, being stronger in the CD56 bright+ cells (results not shown). However, while most decidual CD56 bright+ cells are CD2 / and CD7 /, there is a small population (5-10%) that is either CD2-or CD7-. We have also observed an additional small subpopulation (' 3%) of decidual cells that is CD56bright+, and co-expresses the TCR-associated CD3 molecule. However, their size and granularity show they are large granulated cells unlike classical T cells (Fig. 9a). It is apparent that in decidua, the CD56 bright+ cells are larger and more granular.

Decidua
CD56bright cells. CD3-, CD56 HUMAN UTERINE LEUKOCYTES 175 CD57 was not expressed on PBL CD56bright/, populations K 1%) of CD56 bright / cells, and CD16dim+/CD16populations. CD16 dim+ or CD16-; and a major (10-15%) classical NK population CD56 dim+ CD16 bright+. We have identib) CD16. Three distinct NK cell subsets have fled both CD16and CD16 dim+ in the CD56 / been identified in peripheral blood on the basis of decidual lymphocytes. The CD56 bright+ were almost CD16 expression (Nagler et al., 1989): two sinai1 entirely CD16-with only a very few cells expressing ;ery low levels of CD16 (Fig. 10c). The small CD56 dim+ population was also CD16 dim+ (Fig. 10b). However, as this population overlaps with the minor autofluorescent population, some of the CD16 / cells may be autofluorescent cells. We have allowed for this by gating round the regions and subtracting the autofluorescent population on the negative control and we still find that 13% CD56 / cells express low levels of CD16. CD16 is also expressed at low density on some CD56-cells (Fig.  10a). There appear, therefore, to be subsets of CD56 cells in decidua in relation to the level of expression of both CD56 and CD16. Activation Markers. CD69 is detectable on the surface on NK cells within a few hours after exposure to IL-2, but reacts only weakly with resting peripheral blood NK cells (Lanier et al., 1988). We found 40% of freshly isolated CD56 / decidual cells expressed this antigen, yet it was reported to be absent from CD56 / PBL (Nagler et al., 1989).
HML-1 is an antigen detected on more than 90% of human intestinal intraepithelial lymphocytes and 40% of intestinal lamina propria .lymphocytes, whereas in other tissues, it is rarely or not expressed (Bensussan et al., 1987). We confirm that this antigen is not confined to the unique environment of the gut, but is also detectable on 20% of both bright and dim subsets of CD56 / decidual leukocyteso Other Leukocytes Present 1. CD3 / Cells. A population of CD56-, CD3 / cells was always present in the decidual preparations amounting to "10%. The morphology of these cells as assessed by foreward/side scatter (Fig.  9a) and Giemsa staining of sorted CD3 /, CD56-cells was of small agranular lymphocytes (Fig. 6b). They are likely to be mature T cells. 2. Macrophages. HLA DR + Leu M3 + macrophages CD3 decidual cells stained with Giemsa. Numerous leukocytes with typical large granular lymphocyte morphology are seen in are a characteristic feature of early decidua (Bulmer,CD56 smears. The CD3 cells appear to be small agranular 1989), where they account for '20% of cells in lymphocytes. (See Colour Plate VII at the back of this publiimmunohistological studies (Bulmer et al., 1988;cation Percentage positive cells FIGURE 7. Leukocyte differentiation antigens expressed on decidual CD56 cells. The results were analyzed using two-color fluorescence as described in Methods. The (*) indicates antigens expressed at very low density on CD56 cells. King et al., 1989a). These cells comprised 5-10% of our decidual preparations as assessed by HLA DR + cells. The lower numbers detected may be due to the mechanical sieving procedure used to isolate LGL.
LGL are probably more loosely attached to the stroma than macrophages and are, therefore, more easily extracted.

Proliferation of CD56 + Lymphocytes In Vivo
Proliferating cells were identified using Ki-67, a Mab detecting a nuclear proliferation antigen present on all cells not in GO (Gerdes et al., 1984). The sections were subsequently stained with Leu-19 to determine whether proliferation of CD56 / cells occurred in vivo. The proportion of at least 200 CD56 / cells, which were also Ki-67 /, was counted in endometrium at all stages of the menstrual cycle and in decidua from 6-16 weeks sestational age.
In proliferative-phase endometrium, numerous proliferating cells were found in both glands and stroma (Fig. 11a). CD56 / cells were sparse and Ki-67 + were also very few (Fig. 12). However, in the secretory phase, Ki-67 / cells were only found in the endometrial stroma apart from endothelial cells, which were often Ki-67 /. Most Ki-67 / cells were also CD56 /, indicating that it is the LGL and not stromal and glandular cells that were proliferating at this stage of the cycle. Specimens of decidua showed a similar picture, although, in general, there appeared to be fewer Ki-67 + cells in the stroma (Fig.  11b). The proportion of CD56 / cells that were also Ki-67 / is given in Fig. 12. It is apparent that the maximal proliferation occurs in the non-pregnant secretory-phase endometrium. Although some Ki-67+/CD56/ cells are seen in decidua, there were less proliferating CD56 / cells with a slightly downward trend as gestation proceeded. Decidua from a molar pregnancy also contained numerous CD56 / cells, of which a proportion was Ki-67 / similar to a normal pregnancy.

Electron Microscopy
To compare the ultrastructural appearances of decidual LGL with that reported for both endometrial granulocytes (Cardell et al., 1969;Loubet et al., 1989) and NK/LGL in the peripheral blood (Kang et al., 1987;Polli et al., 1987), electron microscopy was performed together with immunoelectron microscopy for CD56.

Morphology
Sections contained glands, uterine stromal cells, blood vessels, and leukocytes. Stromal cells were easily recognized by their large size, pale nucleus, and cytoplasm with evenly distributed organelles. Of the leukocytes, the most frequent were round 10-15/zm diameter cells containing a round or kidney-shaped nucleus with dense chromatin, small mitochondria, and a few elongated cisternae of rough endoplasmic reticulum (Fig. 13A). Their cytoplasm was much denser than that of the stromal cells and fairly short cytoplasmic projections were often seen. The most characteristic feature of these cells was the presence of 1-10 membrane-bound granules (0.2-0.4/m) per section of cell. The granules mostly had a homogeneous dense content, but sometimes had a dense core surrounded by a variable-width light zone containing small vesicles, granular material and irregular membranous inclusions (Fig. 13B). No "paracrystalline" or "parallel tubular" arrays were found in the granule periphery. The Golgi body varied considerably in size; the large FIGURE 11. Frozen sections of (a) proliferative endometrium and (b) decidua (7 weeks gestation) double stained with the nuclear proliferation Mab,    (Kang et al., 1988). and intraepithelial leukocytes containing the charac-Pre-embedding Localization on Isolated Cells. Incuteristic granules (Fig. 13D) were also observed. A bating with Leu-19 prior to fixation and embedding significant number of decidual leukocytes appeared resulted in labeling of the leukocyte plasmalemma. similar ultrastructurally but contain no granules. The Using the single-sandwich technique resulted in a macrophage was the only other leukocyte seen in plasmalemma label of 2-3 gold particles per cell significant numbers. Ultrastructurally, macrophages perimeter, which was barely above the background had a far more complex and divided outline. They of 1-2 particles. However, the double-sandwich contained granules, usually far smaller than those technique produced a higher, if somewhat variable, present in granulated leukocytes, but some were of label between 30-100 particles per cell perimeter similar dimensions and appearance. There were, (Fig. 14). No significant labeling was seen in the therefore, cells present that were not classifiable as cytoplasm. Decidual stromal cells, glandular cells, granulated lymphocytes or macrophages on ultra-and endothelial cells never labeled above backstructural grounds alone. The nature of the granu-ground levels. All cells showing the morphological lated leukocytes was therefore confirmed using attributes of uterine granulated lymphocytes labeled immunoelectron microscopy to CD56.
with Leu-19. The few unlabeled granulated leukocytes could morphologically equally well be classified as macrophages. A variety of leukocytes Immmoelectron microscopy without granules labeled with Leu-19. Some were small and round with little cytoplasm; others were Post-embedding Localization on Whole Tissue. None of elongated. Since serial sections were not taken, it is the techniques tried was successful on either Aral-possible that some of these cells did have granules dite or K4M-embedded sections of decidua. Neither and cytoplasmic inclusions out of the plane of the Leu-19 or Leu-M3 produced any significant level of few sections examined per cell. LGL labeled prior to embedding with Leu-19 antibody followed by gold colloid. There are 69 gold particles associated with the plasmalemma. Granules, asterisks; mitochondria, m; nucleus, n. x22,000.

DISCUSSION
In this study, we have characterized the phenotype and morphology of decidual LGL. Our immunoelectron microscopic (EM) findings have now pro-vided definitive proof that cells with the morphology of what early European histopathologists called endometrial granulocytes or "K" cells (Weill, 1921;von Numers, 1953;Hamperl and Hellweg, 1958;Kazzaz, 1972) in the Rhesus monkey FIGURE 13. (A) Electron micrograph of a characteristic large granular lymphocyte (LGL) in first-trimester decidua. There is a kidneyshaped-nucleus (n) short stubby cytoplasmic projections, long cisternae of rough endoplasmic reticulum (large arrowheads) mitochondria (M), and membrane-bounded granules (arrows). There is a portion of macrophage cytoplasm in the top left; compare the much smaller granules (small arrowheads) with those in the LGL. X8,800. (B) Electron micrograph of an LGL in first-trimester decidua. Note the granules of various sizes with (large arrowheads) and without (small arrowheads) a peripheral pale rim. Nucleus (n), mitochondria (m), golgi body (asterisk), glycogen (arrows). x20,000. (C) Electron micrograph of a dividing LGL in first-trimester decidua. The cell type is identified by the three characteristic membrane-bounded granules (arrowheads). Chromosomes, c; endoplasmic reticulum, asterisks, x12,000. (D) Electron micrograph of part of a gland (G) in first-trimester decidua which includes an intraepithelial LGL (asterisk) identified as such by its size, granules (arrows), and endoplasmic reticulum profiles (small arrowheads). There are numerous mitochondria (m) and lipid droplets (L) in the gland cells, which are separated from the connective tissue (t) by a basement membrane (large arrowheads), xl0,000.

184
A. KING et al. (Cardell et al., 1969) and man (Jaeger and Dalleneffects of IL-2 (100 U/ml) and are transformed by bach- Hellweg, 1969;Sengel and Stoebner, 1972; this lymphokine into potent lymphokine-activated Loubet et al., 1989) are indeed CD56 / LGL. killer (LAK) cells with marked NK cytolytic activity Although this relationship was appreciated for some against susceptible targets, including normal and time (Bulmer and Sunderland, 1983;Bulmer et al., malignant trophoblast (King and Loke, 1990a). In 1987), it had been difficult to confirm by light contrast, naive decidual LGL show only moderate microscopy because the requirement of cryostat seccytolytic activity (King et al., 1989b;Manaseki and tions for immunostaining with Mabs, and in parti- Searle, 1989;Ferry et al., 1990). These functional cular Leu-19 (CD56), also results in disruption of the properties are very similar to those reported for granules by freezing. We find the EM morphology peripheral blood CD56 bright+ cells and are different of decidual LGL very similar to that described for from classical CD16 bright+ NK cells with respect to LGL in peripheral blood, although we have not both the levels of cytotoxicity and the proliferative observed the parallel tubular arrays, which are a response to IL-2 (Nagler et al., 1989). consistent feature of the latter. These unique The phenotypic and functional characteristics of organelles are considered to be a typical structural decidual LGL, therefore, correspond closely to those rcharacteristic of human NK cells (Huhn et al., 1982; described by Lanier and his colleagues for a tiny Kang et al., 1988), so their absence in decidual subset of NK cells identified in peripheral blood. CD56 / LGL argues against these cells being classical These investigators suggested recently that these NK cells, small CD56 bright+ subsets may represent immature We have analyzed the phenotypic characteristics stages in the classical NK cell differentiation of mechanically disaggregated decidual LGL both pathway, although they have not dismissed entirely qualitatively and quant!.tatively using flow the possibility that they represent distinctive NK/ cytometry. Our qualitative observations using LGL subsets with varied functions (Nagler et al., double immunostaining have confirmed that CD56 / 1989). We favor he second of these hypotheses and decidual LGL are CD3 /, CD16dim+/-, and CD2 /. A propose that the CD56bright+, CD16 -/dim+ decidual previous flow cytometry study using enzymic dis-LGL arise from a different, but Closely related, LGL aggregation of decidual LGL noted an appreciable lineage rather than representing an early differentia-CD2population (Starkey et al., 1988), whereas our tion stage of classical NK cells. We have observed present study, using mechanical disaggregation, virtually no CD16 bright + cells among decidual LGL, so finds only 5-10% of CD56 / decidual LGL are CD2-. there is no evidence of progressive acquisition of It is possible that enzymic extraction could have this surface molecule in decidua as there is thought resulted in the loss of certain surface antigens, partito be in peripheral blood. Previous flow cytometric cularly CD2, which is known to be vulnerable to studies did report on the finding of up to 10% proteolytic treatments (Ritson and Bulmer, 1987a). CD16 / cells in decidua (Starkey et al., 1988; Mana-Our quantitative analyses have revealed that the seki and Searle 1989), but the fact that these cells density of CD56 on decidual LGL is very high, were NKH-1 (CD56) negative and that the Mab used reaching a level 22 x that found in the majority of (Leu-11c) is not specific for the NK-cell Fc receptor CD56 / PBL and 5 x the level of the brightest CD56 / would suggest that these cells are probably neutro-PBL population. In contrast, the density of CD16 on phil granulocytes or possibly even trophoblast. decidual LGL is low or absent. Microscopic exami-Immunohistological staining of decidual sections nation of Geimsa-stained preparations after cell also failed to reveal any CD16 + cells (Ritson and sorting show that the majority of CD56 bright+ Bulmer, 1987b;King et al., 1989a), reflecting our decidual LGL (75%) are granulated with a small flow cytometric results of the small numbers of cells number of agranular cells. This is in contrast to the expressing CD16 at low density. Furthermore, albeit CD56 brigl't+ subset in peripheral blood, which are indirect support for our hypothesis comes from the mainly agranular (Lanier et al., 1986). Thus, decidual observation that the staining of CD56 on decidual LGL are granulated cells with the phenotype LGL is so much more intense than any of the peri-CD56bright+, CD3-, and CD16-. They differ from the pheral blood CD56 / populations and such CD56 bright+ PBL in being mainly granulated and CD56 bright+ cells have not been observed in any expressing an increased density of CD56. organ, or at any other mucosal surface ( (unpublished). This suggests that the identified the granulated lymphoid cells in pregnant granules are not present in dividing cells and may mouse uterusthe granulated metrial gland (GMG) appear in utero only .as the LGL differentiate. Precellsas perforin-containing NK-like cells that can vious reports have stressed that because the numbe seen to proliferate in utero from small agranular bers of endometrial granulocytes/LGL vary during precursors to large granular lymphocytes (Parr et al., the menstrual cycle, it seemed likely that they were 1990). Depletion of NK cells with asialo-GM1 subjected to hormonal control. Our present finding antibody results in ablation of NK activity from the of a sex difference in the numbers of these LGL in spleen, but NK function in decidua is only partially peripheral blood of male and female individuals reduced (Croy et al., 1985). It has proved difficult to supports a role for hormones. It is of interest to note identify the cells responsible for NK activity in that LGL are found only in areas where stromal murine decidua because their isolation is difficult cells show evidence of decidual change either in the technically, but the putative candidates are thought uterus or in ectopic foci of decidualization in the to be the GMG cells (Croy, 1990). From her exten-tube, cervix, or ovary (Hamperl and Hellweg,. 1958).
sive studies, Croy has also been cautious in classi-We have observed that decidual cells secrete signififying GMG cells as classical NK cells. Recently, she cant amounts of laminin (Loke et al., 1989), the procrossed mice with no mature T or B cells (scid/scid) duction of which is enhanced by progesterone (unwith those with defective NK function (bg/bg) and published). IL-2-stimulated decidual LGL also profound that the resulting progeny reproduced nor-duces laminin (unpublished). Perhaps homotypic mally, implying that neither T, B, nor NK cells were interaction between LGL and decidual cells via this essential for reproduction. However, the typical adhesion molecule can provide one mechanism for giant lyzozomes found in all granulated cells of bg/ LGL homing and retention in the uterus. bg mice, including NK cells were not identified in Our present study also indicates that decidual the uterine GMG cells, whose number and mor-LGL are in a state of activation different from phology were identical to GMG cells in normal mice CD56 / LGL in peripheral blood. CD69 is absent (Croy and Chapeau, 1990). This finding, therefore, from CD56 bright+ PBL, yet is present in decidual implies murine uterine granulated lymphocytes are CD56 / cells (Nagler et al., 1989). This antigen is distinct from the classical NKor T-cell lineages, induced rapidly on peripheral blood NK cells after Whichever hypothesis regarding the lineage of IL-2 stimulation (Lanier et al., 1988). In addition, decidual LGL should eventually prove to be correct, HML-1, originally thought to be confined to intesthe fact that these cells are so numerous (> 70% of tinal lymphocytes, is now thought to be an actibone-marrow-derived cells) in decidua compared to vation antigen as it can be induced on Con-A blasts peripheral blood (K 1%) implies that there is some from peripheral blood (Shieferdecker et al., 1990). mechanism for selective recruitment to or inductive HML-1 is also present on decidual LGL. proliferation within the uterus. That the latter does The CD56 molecule is identical to the embryonic occur in vivoisconfirmed by our present observation form of the neural-cell adhesion molecule (Eof the nuclear proliferation marker Ki-67 in many NCAM) ), so the very high level CD56 / decidual LGL. Previous studie have noted of its expression on decidual LGL could function as that cell proliferation in the endometrium continues an adhesion molecule regulating binding to other throughout the menstrual cycle (Ferenczy et al., decidual elements, including the fetally derived 1979). Recently, CD45 / lymphoid cells were hown placental trophoblast cells. We could demonstrate no to be almost entirely responsible for this proliferaconjugate formation between trophoblast and tive activity (Pace et al., 1989;Tabibzadeh 1990). We decidual LGL (King et al., 1990), which could be due now demonstrate that this is attributable to the to the highly sialylated molecules on both LGL CD56 / uterine LGL and that proliferation of LGL (CD56/E-NCAM) and trophoblast (Whyte and Loke, continues in early pregnancy. This finding may 1978). Other adhesion molecules also expressed at appear to contradict the old reports that mitotic high density on decidual CD56 / cells include the figures were rarely, if ever, found in endometrial integrins CD18, CD11a, and CD11c, and these might granulocytes (Hamperl and Hellweg, 1958). Howbe important in migration and in determining the ever, although we have noted mitoses in Giemsacharacteristic tissue distribution of LGL. stained smears of isolated decidual and endometrial The functions of uterine LGL are not known and, 186 A. KING et al. in particular, the question of whether they are essential for pregnancy remains unanswered. It would be reasonable to suggest that an appropriate target for interaction is with the invading extravillous trophoblast cells of the placenta. NK cells produce many cytokines (Cuturi et al., 1989) and those produced are different in the various CD56 / subsets (Nagler et al., 1989). Although there is no information on cytokine production by human decidual LGL, in the mouse, CSF-1, IL-1, and LIF have been identified in GMG-cell conditioned media (Croy et al., 1989). These cytokines may influence trophoblast growth, differentiation, and migration. From our observation that decidual CD56 bright+ cells are activated by IL-2 to become LAK cells capable of killing normal trophoblast and to an even greater extent malignant choriocarcinoma cells, we have proposed that decidual LGL could also control against undue invasion or malignant transformation (King and Loke, 1990b). A similar proposal has been made on the basis of murine studies (Head, 1989). The nature of these cellular interactions has not been established. There is now a great deal of interest in the nonclassical HLA-G antigen whose expression appears to be restricted only to extravillous trophoblast (Ellis et al., 1990;Kovats et al., 1990). It is tempting to suggest that this could be the putative target molecule recognized by decidual LGL. The expression of a non-polymorphic Class molecule "may explain the survival of trophoblast in the uterus. NK cells reject cells with absent or low Class expression, whereas allotypic CTL will reject cells bearing classical Class molecules (Storkus et al., 1989). This phenomenon has led to the "missing self" hypothesis, which states that the absence of self-HLA on target cells is seen by NK cells, rather than the presence of some other distinct NK receptor (Ljunggren and Karre, 1990). Trophoblast may have evolved a mechanism to escape attack by both allospecific T cell and NK cells by the expression of Class HLA-G, which may be seen by the mother as self or neutral MHC. Interestingly, HLA-G is a relatively primitive antigen in phylogenetic terms (Watkins et al., 1990). We suggest that decidual LGL are also primitive lymphocytes that differentiate in the uterus along a separate pathway to that of classical NK cells, presumably as an adaptation to accommodate placentation in mammals. The idea that lymphocytes may differentiate extrathymically at mucosal surfaces is not new (Fichtelius, 1968;Campana et al., 1989). Analogous granulated CD3-, CD7 /, but CD56-cells are found in the intraepithelial component of the gut (Jarry et al., 1990). Reproductive immunologists have traditionally considered the fetomaternal relationship primarily in terms of transplantation immunology, which is an artificial model in a highly evolved species. In the broader context of evolution, placentation in mammals must have developed a separate mechanism of immunological recognition based perhaps on some preexisting primitive defense system.

Peripheral Blood Lymphocytes
Peripheral blood lymphocytes (PBL) from men and women aged 18-25 years sampled at the same time each day were isolated by centrifugation over Lymphoprep (Flow).

Decidual Leukocytes
Decidual cell preparations were prepared as described previously (King et al., 1989a(King et al., , 1989b. Briefly, pieces of decidua were obtained from first-trimester elective vaginal termination of pregnancy (6-12 weeks). The fragments of decidua were identified macroscopically and washed in RPMI. They were then minced between two scalpel blades and pushed through a 53m sieve (Gallenkamp). The cell suspension was washed, layered onto Lymphoprep, and spun at 400g for 20min. The cell.s were removed from the interface, washed, and resuspended at 3 X107 cells/ml in phosphate buffered saline/0.1% bovine serum albumin (PBS/0.1% BSA).

Culture of Decidual Leukocytes
Decidual leukocytes were cultured as previously described (King and Loke, 1990). Briefly, they were resuspended in RPMI and 10% fetal calf serum (FCS) supplenented with antibiotics and 2mM L-Glutamine and 100-U/ml recombinant IL-2 (rIL-2) (Sigma). After incubation in 25-mm flasks, the nonadherent cells were transferred to a new flask and cultured upright in 5% CO 2, 95% air at 37C.

Monodonal Antibodies
The source and specificities of the monoclonal antibodies (Mabs) used in this study are shown in Table   2.  (a) Direct Staining.
1.5 X106 cells were incubated with 50/1 of the first fluorescein-(FITC) or phycoerythrin-(PE) conjugated antibody for 30 min and washed twice in PBS. In single direct staining, the cells were then fixed with an equal volume of 2% formaldehyde (100/zl) and PBS. Samples were kept at 4C in the dark until analyzed.
Following incubation with an unconjugated antibody, a secondary conjugated antibody, either goat antimouse FITC (GAM.FITC, Sigma) or rabbit antirat FITC (RAR.FITC, Sigma) was diluted in RPMI and 20% human AB serum (Sigma), incubated at 4C for 30 min, and then spun 10 min at 200 g. 100/1 of this secondary FITC-conjugated antibody was added to the cells (30 min, 4C), which were then washed and fixed as before.
(c) Double Staining. An indirect immunofluorescence stain was followed by a direct stain. Thus, the cells were further incu-(d) Controls.
Fluorochrome-conjugated isotype-matched irrelevant Mab IgG1 +IgG2 (Simultest; Becton Dickinson) was used as a negative control. A positive control--Leu-4mwas always used to confirm the effectiveness of GAM-FITC Ab and to quantify the accuracy of selective scatter gating. A positive dual fluorescence control was also used (Leukogate; B.ecton Dickinson). In addition, a negative control of GAM-FITC Ab alone was essential to set the threshold on the immunofluorescence histogram, above which the cells are considered positive for the Ab binding. In double staining, GAM-FITC Ab followed by an irrelevant directly conjugated Ab was used as the negative control.
The samples were analyzed on a FACStar Plus (Becton Dickinson).

Histology
Samples of endometrium were taken from routine hysterectomy specimens as previously described (King et al., 1989a). Decidual and endometrial fragments (1 cm2) were frozen in liquid nitrogen for immunohistology. A piece was fixed in formalin for routine histology. The stage of the menstrual cycle ethanol and embedded either in Araldite at room was assessed as previously described (King et al., (Wells, 1989a). Samples (2 mm) for electron microscopy 1985). These sections were incubated with Leu-19, were fixed for l hr in either 0.01M perio-or an irrelevant Mab, Leu-M3, which identifies date-0.075 M lysine-2% paraformaldehyde (PLP), or macrophages. Both antibodies were diluted 1/10 in 1% glutaraldehyde with 3% paraformaldehyde in PBS +1% BSA+0.05% Thimerosal (Sigma). The sec-0.01 M cacodylate buffer, then sorted in 0.1 M tions were then treated with either (i) goat anticacodylate buffer at 4C until required. The tissues mouse gold colloid (Janssen Pharmaceutica) (1/30); were postfixed for I hr in 2% osmium tetroxide with (ii) rabbit antimouse IgG (Dako) (1/100), followed by 1.5% potassium ferrocyanide buffered with veronal goat antirabbit gold colloid (Janssen Pharmaceutica) acetate at pH 7.2, dehydrated in increasing concen-(1/30); (iii) rabbit antimouse IgG, followed by Protrations of ethanol followed by propylene oxide and tein A-coated gold colloid (Janssen Pharmaceutica). embedded in Araldite. 75 m sections were stained All incubations were carried out at room temperawith uranylacetate and lead citrate prior to exami-ture for 40min with two PBS washes between nation in a JEOL 100C electron microscope, incubations.

Immunohistology
(b) Pre-embedding Localization on Isolated Cells. A double immunohistological method was per-Isolated decidual cells were incubated with either formed using a Mab to a nuclear proliferation  or Leu-M3, cells were then fixed in suspenmarker (Ki-67) (Gerdes et al., 1984) and CD56 to sion with PLP for 15 rain, 4C, washed twice, and identify uterine LGL. All incubations were per-resuspended in PBS. For the single-sandwich techformed in a wet box with two washes with tris bufnique, the cells were resuspended, in 80 1 of goat fered saline (TBS) in between incubations. 5/m antimouse IgG adsorbed to 15-nm gold particles for frozen sections were air dried, fixed in acetone (4C, 40 min at room temperature and then washed twice 5 min), rehydrated in TBS, and Ki-67, (1/10) applied in PBS. For the double-sandwich methods, the cells to the section. A 30-min incubation with biotinywere suspended in 80/1 of rabbit antimouse IgG (1/ lated antimouse antibody (1/100) (Dako) was fol-100), 60min, and washed twice in PBS. They were lowed by peroxidase-conjugated avidin (Dako) (1/ then resuspended in 80/1 of goat antirabbit IgG 400). This reaction was visualized using adsorbed to 15-nm gold particles (1/25) for 40 min at 3,3'diaminobenzidine (DAB) with 0.01% hydrogen room temperature and washed twice in PBS.
peroxide. The sections were then incubated with normal mouse serum (1/10) (Sigma) for 10min. Excess serum was wiped off the sections and Leu-19 ACKNOWLEDGMENTS (1/50) was applied for 30min. Rabbit antimouse immunoglobulins (Dako) (1/20) in 10% human AB We acknowledge financial support from the Medical serum was then incubated for 30 min. After further Research Council, East Anglian Regional Health Authority, washing, APAAP complexes (Dako) (1/50) were the Wellcome Trust, and the Special Programme of added for 30 min. The reaction was developed using Research, Development and Research Training in Human a red substrate (Napthol AS-MX phosphate, Reproduction, World Health Organization. We would also like to thank our obstetric colleagues and staff at Addendimethyl formamide, 200 ml, 9.8 ml of 0.1 M Trisbrooke's Hospital for collecting the placental material. HC1 buffer, pH 8.2, levamisole 3 mg, and fast Red TR sale 10 mg), which was stopped with tap water. (Received November15, 1990) After counterstaining in Carazzi's haemotoxylin for 5 min, the slides were mounted in glycerol gelatin (Accepted November21, 1990) (Sigma).