Cubilin, a binding partner for galectin-3 in the murine utero-placental complex.

Galectin-3 is a lectin important in animal development and regulatory processes and is found selectively localized at the implantation site of the mouse embryo. To better understand the role of galectin-3 at the maternal-fetal interface, a binding partner was isolated and characterized. Homogenates of uteroplacental tissue were incubated with immobilized recombinant galectin-3, and specifically bound proteins were eluted using lactose. The principal protein, p400, had an M(r) of 400,000 in SDS-PAGE. Physical properties of p400 and amino acid sequences of seven tryptic peptides were similar to cubilin from rats, humans, and dogs, identifying p400 as the murine ortholog of cubilin. This was further supported by the tissue distribution observed only in yolk sac, kidney, and ileum with monospecific antiserum for p400. Cubilin occurred in yolk sac epithelium throughout pregnancy, but galectin-3 was there only during the last week. Unexpectedly, cubilin was found only in perforin-containing granules of uterine natural killer (uNK) cells, although galectin-3 occurred throughout the cell cytoplasm. In situ hybridization revealed cubilin mRNA in yolk sac epithelium but not uNK cells, implying that yolk sac-derived cubilin is endocytosed by uNK cells via galectin-3. This is consistent with cubilin being an endogenous partner of galectin-3 at the maternal-fetal interface and suggests an important role for cubilin in uNK cell function.

Galectin-3 (gal-3) 1 is one of at least 10 members of a lectin family, which share a conserved carbohydrate recognition domain and which exert their varied biological effects through interaction with complementary ␤-galactoside ligands of glycoprotein or glycolipid "counterreceptors." Since galectins are divalent and are able to form multivalent aggregates, they can cross-link counterreceptors, effecting cell-cell, cell-matrix, or matrix-matrix interactions, thus possibly initiating signal transduction. Isolation of several membrane or matrix counterreceptors by affinity chromatography has revealed that galectins are linked to a variety of important processes including neoplastic transformation, cell adhesion, tumor invasiveness and metastasis, cellular proliferation, and localized immunomodulation (see Ref. 1). gal-3 and its mRNA are expressed in trophoblast cells of placenta as well as in the granular uNK cells of the metrial gland and decidualized endometrium of the murine implantation site (2)(3)(4). That finding, along with the additional observation that gal-3 is absent from nondecidualized endometrium between implantation sites and is not present in uteri of nonpregnant females (3), strongly suggests that this lectin has a pregnancy-related function at the maternalfetal interface. Although a role for gal-3 at the maternal-fetal interface is not yet known, it may contribute in some fashion to cell adhesion, proliferation, or immunomodulation, all of which occur in other contexts. Therefore, to gain more insight into a role for this lectin at the maternal-fetal interface, we isolated and characterized a counterreceptor for gal-3 from tissues of the murine utero-placental complex that was identified as cubilin. This large protein is produced by epithelial cells of the kidney, small intestine, and yolk sac, and it can function as coreceptor for endocytosis of many important biological molecules in these tissues. Consequently, given the proximity of yolk sac to the maternal-fetal interface, the association of cubilin with gal-3 is especially intriguing, as is its association with the granules of uNK cells in the context of cell adhesion, proliferation, or immunomodulation.

EXPERIMENTAL PROCEDURES
Animals-Outbred Swiss-Webster mice were used for all studies. Virgin females (10 -12 weeks of age; Charles River, Wilmington, MA) were selected at random stages of the estrous cycle, paired with fertile males, and checked daily for the presence of a vaginal plug (designated as day 1 of pregnancy).
Preparation of Lactosyl-Sepharose Affinity Matrix-Sepharose 4B was suspended in an equal volume of 500 mM Na 2 CO 3 , filtered in a Buchner funnel, and washed with 5 volumes of 500 mM Na 2 CO 3 . This was resuspended in an equal volume of 500 mM Na 2 CO 3 ; divinylsulfone (10% of the gel volume) was added to activate the matrix; and the slurry was stirred for 70 min gently at 20°C, washed as above, resuspended in an equal volume of 10% lactose in 500 mM Na 2 CO 3 , and stirred for 15 h gently at 20°C. The lactosyl-Sepharose matrix was washed as above, followed by 5 volumes of distilled H 2 O and 5 volumes of PBS (pH 7.2). The gel was resuspended in an equal volume of PBS, packed in a 200-ml column (5 ϫ 30 cm), and stored at 4°C. All subsequent procedures were done at 4°C unless otherwise indicated.
Preparation of rgal-3 Affinity Matrix-The affinity resin was prepared by hydrating cyanogen bromide-Sepharose with 0.001 N HCl (15 min), washed with coupling buffer (100 mM NaHCO 3 , pH 8.3, 500 mM NaCl), and centrifuged (750 ϫ g for 10 min). Supernatant was discarded, and rgal-3 was added to the activated Sepharose at 8.4 mg to 1 g of gel and mixed for 6 h at 20°C. After 10 min of centrifugation at 750 ϫ g, the pellet was mixed overnight in coupling buffer with 200 mM glycine to block any reactive sites remaining. The slurry was placed in a 1-ml column and washed with coupling buffer, followed by 100 mM sodium acetate, pH 4.0, 500 mM NaCl, and coupling buffer. It was equilibrated with 5 volumes of TET buffer (50 mM Tris-HCl, pH 7.8, 1 mM EDTA, 1% Triton X-100) for isolation of gal-3-binding proteins.
Preparation of Placental-Fetal Membrane Homogenate-Animals were killed by CO 2 asphyxiation followed by cervical dislocation at various times of pregnancy, and uteri were removed and opened along the antimesometrial margin. Placentas and associated fetal membranes were separated from uterine implantation sites, and fetuses were discarded; maternal and fetal components were frozen separately in liquid nitrogen and stored at Ϫ80°C. Tissues were homogenized (PT 10/35; Brinkman, Westbury, NY) at 1.0 g of tissue/2 ml of buffer A (PBS with 4.0 mM 2-ME, 1 mM PMSF) with Complete Mini Protease Inhibitor Mixture (Roche Molecular Biochemicals) at 1 tablet/10 ml of solution with 1 mM EDTA and 300 mM lactose. The homogenate was centrifuged (10,000 ϫ g for 20 min at 4°C), and the supernatant was discarded. The resulting pellet was homogenized with the equivalent of 1.5 g of tissue/1 ml of buffer B (20 mM Na 2 PO 4 , pH 7.2, 1.0 M NaCl, 1 mM PMSF, and Complete Mini Protease Inhibitor Mixture) and centrifuged (10,000 ϫ g, 20 min). This pellet was homogenized in buffer C (50 mM Tris-HCl, pH 7.8, 1 mM PMSF, and Complete Mini Protease Inhibitor Mixture) again at the equivalent of 1.5 g of tissue/1 ml of buffer, and centrifuged (10,000 ϫ g, 20 min). This pellet was homogenized in buffer C with 1% Triton X-100 at the equivalent of 1.5 g of tissue/1 ml and centrifuged (10,000 ϫ g, for 20 min). Protein content of the final supernatant (i.e. the detergent extract of placental-fetal membranes) was determined by the BCA method (Pierce). In a tissue survey experiment, extracts were similarly prepared from fetal membranes, placentas with and without fetal membranes, decidualized and nondecidualized endometrium, artificially induced deciduomata, kidney, ileum, smooth muscle (large bowel), heart, lung, spleen, brain, and liver.
Isolation of gal-3-binding Proteins from Placental-Fetal Membranes-gal-3-binding proteins were isolated by affinity chromatography on rgal-3-Sepharose and eluted with lactose. In initial experiments, placental-fetal membrane proteins were 125 I-labeled (IODO-BEADs; Pierce) and applied to an rgal-3 matrix. Nonbinding material was eluted by extensive washing with TET buffer containing 300 mM sucrose, specifically bound proteins were eluted using TET in 300 mM lactose, and 200-l fractions were collected. gal-3-binding proteins were run on SDS-PAGE (4 -15%) under reducing conditions (Ready Gels; Bio-Rad) and autoradiographed. Preparative amounts of p400 were isolated batchwise, using a slurry of placental-fetal membrane and rgal-3/Sepharose (10 mg of placental-fetal membrane protein/mg of rgal-3), mixed for 4 h, placed in a chromatography column (0.7 ϫ 10 cm) and washed with 100 column volumes of TET. The matrix was resuspended in an equal volume of TET containing 300 mM lactose and rotated in a 12 ϫ 75-mm Falcon tube overnight. The slurry was put in a 500-l microcentrifuge tube with a pinhole in the bottom and containing a glass wool plug, fitted in a 12 ϫ 75-mm Falcon tube and centrifuged for 1 min at 188 ϫ g. The quality of each filtrate was checked by SDS-PAGE on a 4 -15% gel under reducing conditions and visualized with Coomassie Blue stain. Typically, 50 g of gal-3-binding proteins were recovered from placental-fetal membrane extracts per 50 placental units.
Production of Anti-p400 -gal-3-binding proteins (137 g), eluted from a rgal-3 column, were electrophoresed on a 5% acrylamide gel in nonreducing conditions and stained with Coomassie Blue. The M r 400,000 band was excised, frozen, homogenized in PBS, and emulsified in an equal volume of Freund's complete adjuvant. Half of the emulsion was injected intradermally at multiple sites in a rabbit, and the remainder was injected intramuscularly in the quadriceps muscle. Test bleeds were done at 6, 7, and 11 weeks; a booster (30 g of homogenized p400, but without adjuvant emulsification) was given intramuscularly at 8 weeks. The rabbit was exsanguinated by cardiac puncture at 12 weeks, blood was allowed to clot for 1 h at 20°C, and the clot contracted overnight. Antiserum was collected, frozen in aliquots, and stored at Ϫ80°C. Specificity was demonstrated by Western blotting of gal-3binding proteins (1 g) and placental-fetal membrane proteins (20 g). Samples were run on 4 -15% polyacrylamide gels (Ready gels) under nonreducing conditions and transferred to nitrocellulose. Blots were blocked in 3% nonfat dry milk in TBS (10 mM Tris-HCl, pH 7.5, 200 mM NaCl) for 1-2 h at 20°C, followed by incubation overnight in primary antiserum at 1:25,000 (i.e. p400 antiserum, preimmune serum, or a nonrelevant antiserum (a rabbit antiserum (6) to the zonadhesion holoprotein purified from pig spermatozoa)). Blots were washed (4 ϫ 5 min) in TBS, incubated in secondary antibody (1:20,000 (ImmunoPure goat anti-rabbit IgG horseradish peroxidase-conjugated; Pierce) in 3% milk in TBS for 30 min, washed in TBS, incubated for 5 min in substrate (Super Signal West Pico chemiluminescence substrate; Pierce), and exposed to film. Some blots were incubated in secondary antibody alone as a negative control. For use in immunohistochemistry, anti-p400 was made monospecific by absorption with acetone mouse liver powder (i.e. liver absorbed anti-p400) to remove the M r 70,000 immunoreactive species found in most tissues. 300 l of anti-p400 (1:50 in TBST plus Complete Mini Protease Inhibitor Mixture) and 5 mg of liver powder were mixed for 30 min and centrifuged at 20,000 ϫ g for 15 min. Supernatant was then collected, and the process was repeated twice. This treatment effectively removed anti-p70 from antiserum; aliquots of liver absorbed, monospecific anti-p400 were stored at Ϫ20°C.
Deglycosylation of p400 -The extent of glycosylation of p400 was examined with endoglycosidases using a protocol modified from manufacturer's instructions (Glycofree Deglycosylation Kit; Prozyme, San Leandro, CA). Briefly, 3 g of gal-3-binding proteins isolated from placental-fetal membrane were dialyzed against 250 mM sodium phosphate buffer, pH 7.0, SDS was added to 0.1% final concentration, and proteins were denatured at 65°C for 5 min. The sample was cooled to 20°C; Triton X-100 was added to 0.75% final concentration; 1 l of peptide N-glycosidase F (500 units/ml), endo-O-glycosidase (1.25 units/ ml), or sialidase A (5 units/ml) was added; and samples were incubated at 37°C. Aliquots were removed at various times up to 72 h and subjected to nonreducing SDS-PAGE (5%) and Western analysis with anti-p400.
Assay for Intermolecular and Intramolecular Disulfide Bonds-gal-3-binding proteins were trichloroacetic acid-precipitated, washed with ice-cold 90% acetone, and dried. The pellet was dissolved in 0.01 N NaOH, denatured at 65°C for 10 min in nonreducing Laemmli buffer, loaded onto a 4% acrylamide tube gel, and electrophoresed at 100 V for 3 h. The tube gel was extruded, incubated in reducing buffer (loading buffer plus 20 mM dithiothreitol for 15 min), and loaded onto a 4% preparative slab gel run at 100 V for 1 h in a second dimension. The gel was either silver-stained or transferred to nitrocellulose and probed with anti-p400 for Western analysis.
Amino Acid Sequence Analysis of p400 -Amino acid sequencing of p400 was performed by Prof. C. Slaughter (Howard Hughes Medical Institute, University of Texas Medical School, Dallas, TX). Peptides, separated by reverse phase high pressure liquid chromatography on a 150-mm RP300 column (PerkinElmer Life Sciences), were subjected to automated Edman degradation using a model 477A amino acid sequencer (Applied Biosystems, Foster City, CA). Sequences of p400 peptides were compared for alignment with proteins of the National Center Biotechnology Information data bank.
Tissue Collection for Immunohistochemistry and in Situ Hybridization-Tissue was prepared from uterine horns of various days of pregnancy, fixed in 4% paraformaldehyde in PBS overnight, and embedded (Paraplast Plus, Sherwood Medical Laboratories, St. Louis, MO), and 5-m serial sections were mounted on Superfrost Plus slides (Fisher).
In some experiments, a ligature was made at one or both utero-tubal junctions on day 1 of pregnancy to create a unilateral or bilateral sterile pseudopregnant horn (3), using a standardized stimulus (i.e. an 11-mm cut along the antimesometrial margin of the ovarian end of the uterine horn) delivered on day 4; the resulting deciduomata were harvested on days 8 -14 and processed as for normal implantation sites.
Dual Label Fluorescence Immunochemistry-Slides were treated as above, through the primary antibody incubation step (anti-p400 1:4,000, anti-perforin 1:400, or anti-gal-3 1:800), washed in PBS, and incubated with secondary antibodies (Alexa Fluor 488 goat anti-rat IgG and Alexa Fluor 594 goat anti-rabbit IgG; Molecular Probes, Inc., Eugene, OR) for 45 min at 37°C. Slides were washed in PBS three times and then in PBS at pH 8.5 mounted in MOWIOL (Calbiochem) solution under glass coverslips and viewed with a Zeiss Axiovert microscope using standard epifluorescence.
In Situ Hybridization-Mouse cubilin (GenBank TM accession number AF 197159) PCR primers were used to prepare a 394-bp cDNA with T7 promoter (see Table I (4). Sections were deparaffinized, rehydrated, and treated with proteinase K and with acetic anhydride. Slides were prehybridized for 4 h at 20°C; hybridized Ͼ16 h at 50°C with probes (1 ϫ 10 6 cpm sense or antisense/slide); treated with RNase A; washed in 15 mM NaCl, 1.5 mM sodium citrate for 2 h at 65°C; and dehydrated in ethanol with 300 mM ammonium acetate. Slides were dried and exposed 3 days on film, and hybridization signal intensity was estimated. Slides were dipped in NTB-2 emulsion (Eastman Kodak Co.), exposed 7-14 days at 4°C, developed in Kodak D19 developer, counterstained with Mayer's hematoxylin, and viewed on an Olympus BX50 microscope (Leeds Instruments Inc., Irving, TX) by light and dark field optics.

Isolation of gal-3-binding Proteins-Since gal-3 may have an
important physiological role at the maternal-fetal interface, gal-3-binding proteins were isolated from placental-fetal membrane extracts using immobilized rgal-3. Labeled, nonbinding protein (ϳ99%) passed through the column (Fig. 1), and no more was released by a sucrose wash. In contrast, a sharp peak of radioactivity (ϳ1%) eluted with lactose. After lactose removal, virtually all rebound to the rgal-3 column and could be re-eluted with lactose but not sucrose, confirming specific binding. 2 gal-3-binding proteins were run on SDS-PAGE under reducing conditions and visualized autoradiographically. The major labeled gal-3-binding protein is larger than 200 kDa, extrapolating to about 400 kDa (p400) with several minor bands 2 around 120 kDa. Larger amounts of gal-3-binding proteins isolated by a batch method were detected with Coomassie Blue after SDS-PAGE (Fig. 2); again, the major protein is about M r 400,000, with much fainter ones around M r 120,000. p400 was selected for further study, because as the major gal-3binding protein, it may be important for gal-3 function in the implantation complex.
Specificity of Anti-p400 -A batch method provided sufficient p400 to produce an antiserum for studies of its chemical and immunohistochemical properties. The specificity of anti-p400 was examined as follows: gal-3-binding proteins and placentalfetal membrane extracts were subjected to PAGE, blotted to nitrocellulose, and probed with anti-p400 (Fig. 3); an immunoreactive band (lane 1) extrapolating to about M r 400,000 is found, with a slower moving band extrapolating to about M r 800,000. Immunoreactive proteins in placental-fetal membranes are found (lane 2) at an M r of 400,000 and an M r of 70,000. These did not react with preimmune serum (lanes 3 and  4), a nonrelevant antiserum, or secondary antibody alone. After liver powder adsorption, the antiserum was monospecific 2 for p400.
Chemical Characteristics of p400 -Electrophoretic mobility of p400 was run under nonreducing and reducing conditions to see if it contains intermolecular or intramolecular disulfide bonds. One-dimensional analysis, visualized with silver staining (Fig. 4A), shows reduced p400 runs slower than the nonreduced form; Western blot analysis (Fig. 4C) confirms both are p400. This is supported by two-dimensional gel analyses (Fig.  4, B and D), in which the p400 is above a theoretical diagonal 2 S. Crider-Pirkle, unpublished observations.  on the gel, showing that migration was slower under nonreducing conditions. These results are only compatible with the presence of intramolecular disulfide bonds in a p400 molecule, so p400 is a monomer whose tertiary structure is stabilized by disulfide bonds. Finally, anti-p400 always reacted much more intensely with nonreduced p400 than an equal amount of reduced protein. Since nonreduced p400 was used as immunogen to make this antiserum, loss in Western sensitivity to p400 after reduction probably corresponds to epitope loss from p400.
The p400 was exposed to glycosidases and subjected to Western analysis to determine whether it is glycosylated. After 18 h of incubation (Fig. 5), the electrophoretic mobility of p400 increased about 10% (lane 6) compared with untreated (lanes 1 and 8) and mock-treated samples (lane 7). Thus, p400 is a glycoprotein with significant amounts of N-linked carbohydrate. Incubations of 48 and 72 h or treatment with O-glycosidases or sialidases, produced no additional detectable changes. 2 Amino acid sequences of seven peptides of p400 (p400-1 through p400-7) are shown in Fig. 6A. The p400 peptides bear a close resemblance to segments of a full-length protein of rat (AAC 71661), human (AAC 82612), and dog (AAF 14258); i.e. of the 73 amino acids from p400, 72.6% are identical versus those in rat, 69.9% versus those in human, and 67% versus those in dog. 12 of the remaining 21 nonidentical p400 rat-cubilin amino acid pairs are conservative substitutions. In addition, for the only overlap with the recently published fragment of mouse cubilin (AAF 61487), all nine amino acids of p400-7 are identical. Alignment of these seven peptide fragments with rat cubilin is also shown to map them relative to the full-length rat sequence (Fig. 6B). Because these highly similar peptides are  6. p400 is the mouse ortholog of cubilin. A, sequence comparison of p400 peptides versus cubilin of other species. B, map of mouse cubilin peptides aligned against the 3623 amino acid sequence of rat cubilin, predicted from its cDNA (12). dispersed across the rat sequence, we conclude that p400 is indeed most likely cubilin.
Tissue Distribution of p400 -Based on known tissue distributions of cubilin, a survey for p400 was done in further support of this conclusion. Western blot analyses of tissues from pregnant females were done with unadsorbed anti-p400 to survey p400 distribution (Fig. 7). Immunoreactive p400 is seen at M r 400,000 in 10 g of protein isolated from fetal membranes  13) were also negative; increasing the total protein of all tissues to 30 g again yielded negative results. 2 Immunoreactive material was observed (Fig. 7) at M r 70,000 in liver (lane 9), placenta (lane 3), fetal membranes (lane 1), and placental-fetal membrane (lane 2); increasing total protein to 30 g revealed p70 in all tissues except kidney. 2 In conclusion, p400 is only in yolk sac, kidney, and small intestine, confirming that it is murine cubilin.
Spatiotemporal Pattern of Expression of p400 -The pattern of its expression in implantation sites was determined by probing histologic sections with liver-absorbed anti-p400. Immunoreactive protein was detected in yolk sac endoderm as early as day 6, and it was present at the apical surface of those cells throughout the remainder of pregnancy, as shown specifically for day 15 (Fig. 8, A1 and A4); it was also clear that no immunoreactive material occurs in the placenta per se. In situ hybridization studies with antisense cubilin cRNA revealed, as shown for day 15 (Fig. 8, A2, A3, A5, and A6), that the gene is transcribed in yolk sac endoderm during the entire course of pregnancy. Unexpectedly, anti-p400 reacted with the cytoplasmic granules of the large granular cells from day 10 through day 21, as shown for day 16 (Fig. 9, B1 and C1). These are characterized as uNK cells, because they also label with antiperforin (Fig. 9, B2) (9)(10)(11). Indeed, dual labeling with anti-p400 and anti-perforin demonstrates that both proteins are localized to the same granules (Fig. 9, B3). Although perforinpositive uNK cells are detected in decidua as early as day 5, p400 is not detected in these cells until day 10 and thereafter. Furthermore, no cubilin mRNA was detectable in uNK cells at any time, as shown for day 14 (Fig. 10).

Comparison of Spatiotemporal Patterns of Expression of gal-3 and Cubilin-
The cellular sites of synthesis of these proteins and their mRNAs were examined on adjacent histologic sections, using both monoclonal antibody against gal-3 and radiolabeled gal-3 antisense cRNA as probes. On day 5 of pregnancy, decidualizing endometrium and uNK cells were labeled with both the antibody and the cRNA, and by day 12 trophoblast as well as decidual and uNK cells were observed to be expressing both the gal-3 mRNA and protein, as previously reported (4). However, while there was no evidence of labeling of the visceral yolk sac epithelium with either the gal-3 antibody or cRNA up to day 12 of pregnancy, patchy labeling was observed with both probes on day 12. Throughout the remainder of pregnancy, an increasing proportion of epithelial cells of the visceral yolk sac were labeled with both gal-3 antibody and cRNA probes, as shown for day 15 (Fig. 8, B1 and B3).
Results of probing sections of day 16 yolk sac simultaneously with anti-p400 and anti-gal-3 are shown in Fig. 9, A1-A3. The p400 was observed mainly in the apical portion of the cells (Fig.  9, A1), while gal-3 appeared to be distributed throughout the cytoplasm (Fig. 9, A2); superimposition of these images suggests an overlap of these proteins within the apical cytoplasm ( Fig. 9, A3). Although anti-gal-3 labels uNK cells (Fig. 9, C2), it is localized mainly in the cytoplasm and possibly the plasma membrane; it does not appear to be directly associated with p400 within perforin-positive granules (Fig. 9, C1-C3).
Finally, to determine if p400 is also present in uNK cells associated with artificially stimulated deciduomata, which do not have an adjacent yolk sac, deciduomata sections were probed with anti-perforin and anti-p400. Granules were labeled with anti-perforin as well as anti-p400. In situ hybridization work on adjacent sections demonstrated that the cells did not contain cubilin mRNA. 2

DISCUSSION
This study demonstrates that a monomeric protein, p400, ϳ400 kDa, can be isolated from homogenates of murine uteroplacental complex by affinity chromatography with immobilized rgal-3. Peptide fragments of p400 have amino acid sequences similar to those in cubilin of other species. Although the complete amino acid sequence of murine cubilin is unknown, its sequence has been deduced from full-length cDNA clones of several other species (rat, AAC 71661; human, AAC 82612; and dog, AAF 14258). Also, p400 is similar in size to cubilin, which is a 460-kDa peripheral membrane protein with 13% of its mass accounted for by carbohydrate. Like cubilin, p400 is heavily glycosylated and has intramolecular disulfide bonds. Finally, p400 is found only in the yolk sac, kidney, and small intestine, and thus its tissue distribution is identical to that of cubilin (12). Consequently, we conclude that p400 is the mouse ortholog of cubilin.
Cubilin has eight N-terminal epidermal growth factor-like domains followed by 27 "CUB" domains (12), characterized as conserved stretches of amino acids separated by nonconserved regions of variable length. CUB domains are predicted to result from a barrel-like structure with two layers of five-stranded ␤-sheets, stabilized by two disulfides from four conserved cysteines, and with ␤-turns in a surface-exposed position, similar to antigen binding regions of IgG (13). CUB domains 5-8 are the binding site for intrinsic factor-vitamin B 12 complex, and domains 13 and 14 bind the "receptor-associated protein," a chaperone-like protein that protects multiple ligand binding sites of processed low density lipoprotein receptor family proteins (15,16). Other proteins known to bind cubilin with high affinity include megalin, a 600-kDa intrinsic membrane protein, which functions as a multiligand receptor and is localized in clatherin-coated pits of absorptive epithelium in kidney and yolk sac (12, 16 -19); IgG light chain (20); holoparticle high density lipoprotein (21); transferrin (22); and apolipoprotein A-I (23). That cubilin comprises several "specific" binding sites for these and presumably other molecules accounts for its ca-pacity as a multiligand receptor or carrier and explains its ability, possibly in association with megalin, to facilitate endocytosis of a variety of proteins (7,12,14).
With the identification of p400 as cubilin, it was expected that anti-p400 would label yolk sac epithelium (16 -18). Consistent with observations on the ontogeny of cubilin in rats (17), our in situ hybridization studies in mice show that the gene is transcribed in extraembryonic endoderm of visceral yolk sac as early as day 6 of pregnancy through term. The importance of cubilin in the yolk sac has become clear with a demonstration that it was the antigenic target of "teratogenic antibodies" in previous studies (16, 24 -26). In this earlier work, it was found that antibodies raised against yolk sac or kidney and administered to pregnant rats (27) or anti-p400 against mouse cubilin administered to pregnant mice 2 during the period of organogenesis resulted in severe fetal abnormalities and death. Although the mechanism responsible for the pathological effect of the antibodies is unknown, they localized in yolk sac epithelium, possibly inhibiting the endocytic apparatus in those cells (25)(26)(27). Since rodent yolk sac is the major portal for maternalfetal exchange prior to development of chorioallantoic circulation (28), it is presumed that disruption of transport of nutrients, such as maternal serum proteins, vitamin B 12 , and cholesterol-containing lipoproteins, during the critical phase of organogenesis could be involved (12,25,29).
Having isolated cubilin as a gal-3 binding partner and having demonstrated its expression in yolk sac epithelium, it became important to determine whether there was overlap in the tissue distributions of these two molecules and thus the potential for them to interact in vivo. Probing sections of the implantation site with anti-gal-3 and with antisense gal-3 cRNA confirmed earlier observations of its expression in trophoblast cells, decidualized endometrial cells, and uNK cells throughout the course of pregnancy (4). Although neither the gal-3 protein nor its mRNA were detected in yolk sac epithelium before day 12, both were present in increasing amounts throughout the last week of pregnancy, and thus it is possible that during that time cubilin and gal-3 do interact in this region. The prospect of a mechanism by which gal-3 might alter the cubilin-dependent uptake of specific ligands by yolk sac during the last week of pregnancy is potentially very important. However, that interesting question remains to be answered by future experiments.
The observation that cubilin accumulates in perforin-positive granules of the uNK cells was completely unexpected. These cells, considered to be a subset of NK lymphocytes, accumulate in large numbers in the decidua basalis and the metrial gland of the rodent implantation site (30,31). Their cytoplasmic granules contain perforin and granzyme, typical of cytolytic lymphocytes (32). However, compared with peripheral NK cells, uNK cells are notoriously poor killers of YAC cell targets in vitro, especially after recovery from the uterus in the latter half of pregnancy (33). While the uNK cells are more potent when stimulated with interleukin-2 (34), normally they do not mount a damaging immune reaction against the semiallogeneic products of conception. Although the importance of uNK cells to pregnancy seems to be that they influence developing placental vasculature by production of proteases, matrix, or biologically active factors (30,35,36), the question of why they do not mount an effective immune response against placenta or extraembryonic membranes may be just as important perforin expression. A4 -A7, controls for cubilin and gal-3 in yolk sac; A4, rabbit preimune serum as cubilin negative control in the presence of anti-gal-3 (A5). A6, anti-cubilin in the presence of the nonrelevant monoclonal antibody 53-6.72 as gal-3 negative control (A7). B4 -B7, controls for cubilin and perforin in metrial gland cells; B4, rabbit preimmune serum as cubilin negative control in the presence of anti-perforin (B5). B6, anti-cubilin in the presence of nonrelevant monoclonal antibody 53-6.72 as perforin negative control (B7). C4 -C7, controls for cubilin and gal-3 in metrial gland cells; C4, rabbit preimmune serum as cubilin negative control in the presence of anti-gal-3 (C5). C6, anti-cubilin in the presence of nonrelevant monoclonal antibody 53-6.72 as gal-3 negative control (C7). Bar (A1), 20 m (applies to all panels). to our ultimate understanding of the biology of the implantation site.
Since the lytic granules of NK cells are dual function organelles, which combine the function of secretory and prelysosomal compartments (37), the finding of immunoreactive cubilin in the cytoplasmic granules of uNK cells raises several interesting questions. First, what is the source of cubilin in uNK cells? Since its mRNA was not detected in these cells, the gene appears not to be transcribed. Although adjacent yolk sac is a likely source, the observation that it is present in uNK cells in artificially induced deciduomata, which are not associated with the developing yolk sac, is compatible with cubilin also being taken up from the systemic circulation. Additionally important and interesting questions include the following. What is the mechanism of cubilin uptake by the uNK cells? Once cubilin is in uNK cells, what is its vectoring pathway to the granules? What is the final molecular form of immunoreactive cubilin in the granules? What is the nature of its association with perforin or granzyme in the granules? What is the function of cubilin in uNK granules? Finally, what are the details of the relationship of gal-3 to cubilin within uNK cells; i.e. does gal-3 facilitate a cubilin function or vice versa? Cubilin uptake by uNK cells might be of major significance if it reflects some mechanism for sampling or processing fetal antigens or for altering immune function by modifying the ability of granules to lyse potential targets.
The original objective of this work, to isolate a binding partner for gal-3 in the uteroplacental complex as a step toward elucidating the role of that lectin in pregnancy, was achieved with the identification of cubilin. The finding that it co-localized with the lectin in yolk sac epithelium in the last week of pregnancy is intriguing and may suggest the existence of unappreciated mechanisms governing maternal-fetal exchanges. The additional finding that cubilin accumulates in uNK cells may also have far reaching implications for the localized modulation of the immune system that occurs at the implantation site. These observations provide important new directions for research of the maternal-fetal interface.