Developmental regulation of a pregnancy-specific oligosaccharide structure, NeuAcalpha2,6GalNAcbeta1,4GlcNAc, on select members of the rat placental prolactin family.

Successful pregnancy is dependent upon an array of signaling proteins secreted by the trophoblast cells of the placenta. Among these is a group of proteins related to pituitary prolactin, known as the prolactin/growth hormone family. These proteins are expressed at specific times during gestation and synthesized in distinct trophoblast cell types in the rat placenta. We report here that select members of this family, prolactin-like protein (PLP-A), PLP-B, PLP-C, decidual/trophoblast PRP, and placental lactogen I variant, only which are expressed in the spongiotrophoblast, late in rat placental development bear Asn-linked oligosaccharides terminating with NeuAcα2,6GalNAcβ1,4GlcNAcβ-R. This reflects the concurrent expression of these prolactin/growth hormone family members with the peptide-specific β1,4GalNAc-transferase and an α2,6-sialyltransferase, which can add sialic acid to terminal β1,4-linked GalNAc. We have determined that at least one of the prolactin-like proteins, PLP-A, is recognized by the protein-specific GalNAc-transferase. The presence of NeuAcα2,6GalNAcβ1,4GlcNAcβ-R on only a limited number of glycoproteins synthesized by the spongiotrophoblasts between mid gestation and birth reflects the need for both the GalNAc-transferase and the peptide recognition determinant for efficient addition of GalNAc. Thus, expression of the GalNAc-transferase and specific members of the prolactin/growth hormone family is developmentally regulated in the rat placenta, suggesting a physiological role for the terminal NeuAcα2,6GalNAcβ1,4GlcNAcβ-R sequence on Asn-linked oligosaccharides of these proteins.

the result of high expression levels of a protein-specific Gal-NAc-transferase in the rat placenta. Levels of the transferase increase from undetectable at mid gestation to maximal at day 18 of gestation. This corresponds precisely with the expression of the prolactin/growth hormone family members PLP 1 -A, PLP-B, PLP-C, d/t PRP, and PL-Iv by spongiotrophoblasts from the junctional zone of the placenta. At least one of these family members, PLP-A, contains a recognition determinant that can be utilized by this protein-specific GalNAc-transferase, which we have previously shown (2)(3)(4)(5) to mediate ␤1,4-linked GalNAc addition to Asn-linked oligosaccharides on the pituitary glycoprotein hormones lutropin and thyrotropin.
We have shown previously that unique Asn-linked oligosaccharides terminating with the sequence SO 4 -4-GalNAc␤1, 4GlcNAc␤1,2Man, which is closely analogous to the oligosaccharide structure on the placental prolactin family members, are present on the pituitary glycoprotein hormones lutropin and thyrotropin and are critical for the expression of full biological activity in vivo (6 -8). The sulfated oligosaccharides determine the circulatory half-life of the glycoproteins bearing them because they are recognized by a receptor, the GalNAc-4-SO 4 receptor, in hepatic endothelial cells, which rapidly removes these glycoproteins from circulation (9 -12). In the case of lutropin, the short circulatory half-life is essential for producing the episodic rise and fall of hormone in the blood throughout the ovulatory cycle (13)(14)(15). The presence of terminal GalNAc-4-SO 4 on lutropin and thyrotropin reflects the action of a protein-specific GalNAc-transferase and a GalNAc-4-sulfotransferase expressed in the gonadotrophs and thyrotrophs of the pituitary (16). The peptide sequences recognized by the GalNAc-transferase consist of a cluster of basic amino acids in the ␣ subunit and a similar cluster of basic residues in the ␤ subunit of lutropin (4,5). The recognition determinant on the ␣ subunit, as well as GalNAc-4-SO 4 , is found on pituitary glycoprotein hormones of all classes of vertebrates, indicating that these are highly conserved and biologically important structures (17). In addition, the levels of the GalNAc-transferase in the pituitary gonadotroph rise and fall in parallel to the expression of lutropin in response to changing estrogen levels thus assuring that the oligosaccharides on lutropin are always fully modified with GalNAc-4-SO 4 (18).
A rapidly increasing number of glycoproteins that bear Asnlinked oligosaccharides with ␤1,4-linked GalNAc, but which are unrelated to the glycoprotein hormones, have subsequently been described. Like the glycoprotein hormones, the ␤1,4linked GalNAc on a number of these glycoproteins has been shown to be modified with sulfate at the 4-hydroxyl (19 -25). In other instances oligosaccharides have been shown to bear terminal GalNAc (23,25,26), NeuAc␣2,6GalNAc (25,(27)(28)(29)(30)(31)(32), and/or GalNAc␤1,4(Fuc␣1,3)GlcNAc (25,29,32,33). Each of these structures is confined to a limited number of glycoproteins because of the peptide specificity of the GalNAc-transferase. In contrast to the sulfated oligosaccharides on lutropin and thyrotropin, biological significance of these structures remains to be established; however, the presence of such unique structures on limited numbers of glycoproteins makes it highly probable that they too will be recognized by specific receptors and play critical biological roles.
The concurrent expression in the rat placenta of the proteinspecific GalNAc-transferase and specific members of the prolactin/growth hormone family which contain a peptide recognition determinant results in nearly quantitative modification of their Asn-linked oligosaccharides with the termini NeuAc␣2,6GalNAc␤1,4GlcNAc␤. Taken together this suggests that these structures will play a role in the expression of the biological activity of prolactin/growth hormone family members during pregnancy. Animals, Tissue Preparation, and Tissue Culture-Timed pregnant female Sprague-Dawley rats were obtained from Harlan Sprague-Dawley (Indianapolis). Animals were sacrificed by CO 2 inhalation on days 9 -21 of pregnancy (day 0 being the day rats were found to be spermpositive). The placentas were removed and washed with ice-cold sterile 20 mM phosphate buffer containing 0.15 M NaCl (pH 7.4). Placentas were homogenized and used for glycosyltransferase assays, or tissue slices were made for isolation of secreted glycoproteins. Undifferentiated Rcho-1 trophoblast cells were obtained from subconfluent cultures in fetal bovine serum, and differentiated Rcho-1 trophoblast cells were obtained from confluent cultures maintained in donor horse serum as described in (35,36).

Materials
Glycosyltransferase Assays-Extracts of whole rat placentas, isolated junctional and labyrinth zones, or Rcho-1 trophoblast cells were made as described previously (16). The protein concentration of the combined postnuclear supernatants was determined by the Bradford dye binding assay (Bio-Rad) using bovine serum albumin as a standard.
Transfer of GalNAc or Gal by the glycoprotein hormone ␤1,4GalNActransferase or the ␤1,4Gal-transferase, respectively, to Asn-linked oligosaccharides acceptors on hCG was compared using the assay described previously (18).
Rat placenta postnuclear supernatants were assayed for ␣2,6-sialyltransferase activity using GalNAc-Trf (39), a modification of a method described previously (40). Reactions consisted of 100 g of postnuclear supernatant protein, 50 mM sodium cacodylate (pH 6.0), 0.5% (w/v) Triton X-100, 50 g of bovine serum albumin, protease inhibitors described above for the GalNAc-transferase assay, 2 M CMP-[ 3 H]NeuAc (2 ϫ 10 5 cpm), and 20 M GalNAc-Trf in a total volume of 60 l. The enzyme reaction was terminated by the addition of 40 l of 5 mg/ml bovine serum albumin and 100 l of ice-cold 10% (w/v) trichloroacetic acid, 4% (w/v) phosphotungstic acid. The precipitated protein was pelleted by brief centrifugation, the supernatant containing unincorporated CMP-[ 3 H]NeuAc was discarded, and the pellet was resuspended and washed three times with 5% trichloroacetic acid. The final precipitated protein was solubilized with 250 l of 1 N NaOH, neutralized with 250 l of 1 N HCl, and incorporated [ 3 H]NeuAc was determined by liquid scintillation counting.
Isolation of ␤1,4GalNAc-containing Secreted Placental Glycoproteins by Lectin Affinity Chromatography-Individual rat chorioallantoic placentas from mid to late pregnancy were sliced using a Stadie-Riggs slicer (A. H. Thomas Co., Philadelphia) and incubated in a 35-mm dish containing 2 ml of minimum essential medium/Earle's medium with penicillin (100 units/ml) and streptomycin (100 g/ml). Alternatively, sliced placentas were incubated with sulfate-, cysteine-, and methionine-free minimum essential medium/Earle's containing 100 Ci/ml Trans 35 S-label (ICN Inc., Irvine, CA; 70% [ 35 S]methionine, 1,173 Ci/ mmol). In either case, medium was collected after 16 h of incubation at 37°C under an atmosphere of 95% air, 5% CO 2 . Medium was separated from tissue debris by centrifugation at 10,000 ϫ g for 20 min. Saturated ammonium sulfate was added to give a final concentration of 70% (v/v) and stirring for 1 h at 4°C. Precipitated proteins were isolated by centrifugation at 10,000 ϫ g for 20 min.
[ 35 S]Cysteine/methionine-labeled ammonium sulfate pellets were resuspended in TBS (20 mM Tris-HCl (pH 7.5), 150 mM NaCl), and unincorporated 35 S label was separated from the protein in 2-ml aliquots by gel filtration on a 45-ml bed volume column of Sephadex G-25 (Pharmacia Biotech Inc.) equilibrated in water. Protein-containing fractions were pooled and lyophilized. The proteins were resuspended in 200 l of 20 mM sodium cacodylate (pH 6.0) and incubated with either C. perfringens or Newcastle disease virus neuraminidase as described below. The protein mixture was diluted to 1 ml with TBS and applied to a 1.5-ml bed volume column of WFA-agarose (EY-laboratories, Inc., San Mateo, CA) equilibrated in TBS. Following adsorption, the column was washed with TBS (ϳ50-column volumes) until radioactivity was no longer detectable, and bound proteins were specifically eluted with TBS containing 50 mM GalNAc.
Nonradiolabeled ammonium sulfate precipitates were resuspended in TBS and dialyzed against 3 ϫ 2 liters of TBS and applied to 5-ml columns of WGA-Sepharose equilibrated in TBS. The unbound fraction of proteins was brought to 1 mM MnCl 2 and 1 mM CaCl 2 and adsorbed to 5-ml columns of ConA-Sepharose (Pharmacia) equilibrated in TBS containing 1 mM MnCl 2 and 1 mM CaCl 2 . After washing with 25 column volumes of the equilibration buffer, bound proteins were specifically eluted with TBS containing 0.5 M ␣-methyl mannoside.
WGA-Sepharose columns were washed with 25 column volumes of TBS, and bound proteins were specifically eluted with TBS containing 0.5 M GlcNAc. Following removal of GlcNAc by dialysis and lyophilization, the bound glycoproteins were resuspended in 200 l of 20 mM sodium cacodylate (pH 6.0), treated with neuraminidase, and subjected to WFA affinity chromatography as described above.
Glycosidase Digestions-Peptide:N-glycanase F digestions were carried out as described previously (10). Digestions with C. perfringens neuraminidase, Newcastle disease virus neuraminidase, or diplococcal ␤-galactosidase were performed in 20 mM sodium cacodylate (pH 6.0). Digestion with jack bean ␤-hexosaminidase was carried out in 50 mM sodium citrate (pH 4.5). All buffers contained protease inhibitors described above in the GalNAc-transferase assay.

RESULTS
Expression of the Glycoprotein Hormone-specific GalNActransferase, but Not ␤1,4Gal-transferase, Increases during Late Gestational Development of the Rat Placenta-Extracts of placenta were prepared at various times during gestation and tested for the presence of the glycoprotein hormone-specific GalNAc-transferase and GalNAc-4-sulfotransferase. GalNActransferase activity with the expected properties was detected in placental extracts ( Fig. 1), commencing at day 9 of gestation and increased 150-fold by day 18 before falling to lower levels just before parturition at day 21. The specific activity of the GalNAc-transferase on day 18 is 8-fold higher than that typical of rat pituitary extracts, making late gestational rat placenta the richest source of glycoprotein hormone GalNAc-transferase found to date. The placental GalNAc-transferase displays the same protein specificity as the pituitary enzyme transferring 78 pmol of GalNAc to the oligosaccharide acceptor on agal-hCG compared with 1 pmol to the oligosaccharides on agal-Trf at an acceptor concentration of 4 M. GalNAc-4-sulfotransferase was not detected in placental extracts at any point during gestation (not shown). Furthermore, the specific activity for ␤1,4Galtransferase, which is not protein-specific (2), does not change over the same time frame (Fig. 1). Thus, a GalNAc-transferase with the same peptide and oligosaccharide specificity as the pituitary GalNAc-transferase is expressed at high levels in the rat placenta during pregnancy between days 9 and 21 of gestation.
A Limited Number of Late Gestational Rat Placenta Glycoproteins Bear Asn-linked Oligosaccharides Terminating with NeuAc␣2,6GalNAc␤1,4GlcNAc␤-R-To identify glycoproteins bearing oligosaccharides with ␤1,4-linked GalNAc, tissue slices from day 18 rat placenta were incubated with [ 35 S]cysteine/ methionine for 16 h so as to incorporate 35 S label into the peptide backbone of newly made glycoproteins. The medium containing secreted glycoproteins was collected, gel filtered over Sephadex G-25 to remove unincorporated 35 S label, and then incubated overnight in the presence or absence of neuraminidase. These samples were then fractionated by lectin affinity chromatography using immobilized WFA, which is specific for terminal ␤1,4-linked GalNAc (45). The bound 35 Slabeled proteins were eluted and detected by fluorography following separation by 10% SDS-PAGE (Fig. 3). Secreted placenta glycoproteins were not bound by WFA prior to treatment with C. perfringens neuraminidase (Fig. 3, lane 3), whereas large amounts of a limited number of glycoproteins were bound following digestion (Fig. 3, lane 1). Digestion with Newcastle disease virus neuraminidase did not result in binding to the WFA column (Fig. 3, lane 2), indicating that a major fraction of ␤1,4-linked GalNAc is substituted with ␣2,6-linked sialic acid.
The secreted placenta glycoproteins recognized by WFA after neuraminidase digestion are bound quantitatively by immobi- lized WGA. Secreted, placental glycoproteins that are bound by immobilized WGA represent 15% of the protein applied. The glycoproteins bound by WGA were digested with neuraminidase, applied to WFA-agarose, and equal amounts of protein from the unbound fraction (lane 1) and bound fraction (lane 2) were separated by 12.5% SDS-PAGE and transferred to polyvinylidene difluoride (Fig. 4A). Glycoproteins bearing terminal ␤1,4-linked GalNAc were visualized using biotinylated WFA. The absence of glycoproteins reactive with biotinylated WFA in the unbound fraction (Fig. 4A, lane 1) and their presence in the bound fraction indicated quantitative binding by WFA-agarose (Fig. 4A, lane 2). Digestion of glycoproteins eluted from WFAagarose with diplococcal ␤-galactosidase (Fig. 4A, lane 3) did not reduce their reactivity with biotinylated WFA, whereas digestion with jack bean ␤-hexosaminidase (Fig. 4A, lane 4), reduced the reactivity with biotinylated WFA. Taken together these results indicate that glycoproteins bearing terminal ␤1,4linked GalNAc are efficiently bound by immobilized WFA and that virtually all of the ␤1,4-linked GalNAc is capped by ␣2,6sialic acid.
The GalNAc-4-sulfotransferase is highly specific for the terminal sequence GalNAc␤1,4GlcNAc␤-R (46). Incubation of the neuraminidase-digested, WFA-bound fraction of glycoproteins from day 18 placenta with [ 35 S]PAPS does not result in 35 SO 4 incorporation (not shown), confirming that no endogenous Gal-NAc-4-sulfotransferase is present. The addition of exogenous partially purified bovine pituitary GalNAc-4-sulfotransferase results in 35 SO 4 incorporation into the oligosaccharides of multiple glycoproteins (Fig. 4B, lane 1) with mobilities similar to those identified above by blotting with WFA (Fig. 4A, lane 2) and to those identified by affinity chromatography of metabolically labeled glycoproteins on immobilized WFA (Fig. 3). The labeled glycoproteins are of placental origin since only two bands are radiolabeled if the partially purified GalNAc-4-sulfotransferase is incubated with [ 35 S]PAPS in the absence of added placental proteins (Fig. 4B, lane 3). The incorporated 35  15, 18, and 21 rat placental tissue slices by sequential lectin affinity chromatography as described in Fig. 4. Glycoproteins bound by WGA-Sepharose were digested with neuraminidase and separated into fractions that were not bound (Ϫ) or bound (ϩ) by WFA. These fractions were separated by 12.5% SDS-PAGE, transferred to polyvinylidene difluoride, and probed with biotinylated WFA to determine their relative levels of ␤1,4GalNAc-containing oligosaccharides (Fig. 6). At day 12, when the transferase levels are low, no glycoprotein binding to WFA-agarose is detected even though d/t PRP and PLP-B are expressed at this time. At day 15, there is a 10-fold increase in GalNAc-transferase activity (see Fig. 1) over day 12, and there is a corresponding increase in binding of glycoproteins to WFAagarose (Fig. 6). Maximal levels of ␤1,4GalNAc-bearing glycoproteins migrating with apparent molecular masses of 29 and 33 kDa occurs on day 18 when the GalNAc-transferase levels are also maximal (see Fig. 1). As the transferase levels decline on day 21, so does the amount of WFA reactive material. Thus, the addition of ␤1,4-linked GalNAc to rat placenta glycoproteins in the range of 29 and 33 kDa is proportionate to the levels of GalNAc-transferase during late gestation.
The forms of PLP-A, PLP-B, PLP-C, d/t PRP, and PL-Iv which bear NeuAc␣2,6GalNAc␤1,4GlcNAc␤ are bound quantitatively by WGA and WFA. In contrast, those glycoproteins not bound to WGA-Sepharose in the isolation procedure do not bear ␤1,4-linked GalNAc since digestion with neuraminidase does not result in binding to WFA-agarose or incorporation of 35 SO 4 by exogenous GalNAc-4-sulfotransferase (not shown).
The forms of PLP-A, PLP-B, PLP-C, d/t PRP, and PL-Iv present in the unbound fraction from WGA-Sepharose were bound by ConA-Sepharose, indicating they are glycosylated (Fig. 7, ConA-Sepharose). The amount of each placental pro-lactin family member in the WGA(Ϫ)/ConA(ϩ) fraction was compared with that in the WGA(ϩ)/WFA(ϩ) fraction (Fig. 7) to determine the proportion of each glycoprotein which is modified with NeuAc␣2,6GalNAc␤1,4GlcNAc␤. PLP-A is fully modified with GalNAc as it is found almost exclusively in the WFA-bound fraction on days 15, 18, and 21 of gestation. In contrast, PLP-B, PLP-C, d/t PRP, and PL-Iv are not fully modified with ␤1,4-linked GalNAc as they are present in both the WGA(Ϫ)/ConA(ϩ) and the WGA(ϩ)/WFA(ϩ) fractions. Only the 33 kDa form of PL-Iv is present in the WGA(ϩ)/WFA(ϩ) fraction, whereas both the 33-and 29-kDa forms are present in the WGA(Ϫ)/ConA(ϩ) fraction. PL-Iv has two potential Asn glycosylation sites, and the 29-and 33-kDa forms are thought to arise from differential glycosylation at one of these two sites. The presence of GalNAc exclusively on the 33-kDa form suggests that the less efficiently utilized glycosylation site on the 33-kDa form is a significantly better substrate for GalNAc addition than the more efficiently modified glycosylation site found on both the 33-and 29-kDa forms.
The results presented in Figs. 6 and 7 demonstrate that the proportion of modification with terminal ␤1,4-linked GalNAc reflects the level of GalNAc-transferase expressed; however, not all of the placental prolactin family members are modified with the same efficiency (see Fig. 7). Addition of ␤1,4-linked GalNAc to glycoproteins by the peptide-specific GalNAc-transferase requires that the transferase and target protein be expressed in the same cell type and that the target protein con- tains a peptide determinant that is recognized by the transferase. Differences in either the site of glycosylation and/or the recognition determinant could account for this.
The Peptide-specific GalNAc-transferase Is Expressed in Spongiotrophoblasts during Late Development of the Rat Placenta-At mid gestation, dramatic changes occur in cell types that make up the rat uteroplacental unit (49). At day 9, decidualized uterus is abundant, and the developing placenta is made up of stem cells and TGC. By day 12 the decidua begins to regress, and spongiotrophoblasts develop in the junctional zone of the placenta. Expression of d/t PRP and PLP-B switches from anti-mesometrial cells of the decidua to spongiotrophoblasts of the placenta. 2 Spongiotrophoblasts account for the bulk of PLP-A, PLP-C, and PL-Iv expression. The increase in number of spongiotrophoblasts that express placental prolactin family members coincides with the increase in expression of GalNActransferase activity, suggesting that GalNAc-transferase is expressed selectively by spongiotrophoblasts.
Homogenates were made from isolated day 10 decidua and either isolated junctional or labyrinth zones from day 13, 16, and 19 of gestation. ␤1,4Gal-transferase activity, but no Gal-NAc-transferase activity, was detected in decidua (not shown). In contrast, GalNAc-transferase was present in both junctional (Fig. 8A) and labyrinth (Fig. 8B) zones and showed the same pattern of increasing and decreasing expression with time as whole placenta homogenates. This is consistent with the absence of ␤1,4-linked GalNAc on glycoproteins of decidual origin and their presence on glycoproteins from the junctional zone. It is not clear which cell types in the junctional zone express GalNAc-transferase, since at mid gestation the junctional zone is made up of both spongiotrophoblasts and TGC.
Although TGC cannot be separated from spongiotrophoblasts, the Rcho-1 trophoblast cell line can be induced to differentiate into TGC (35,36). When cultured in the presence of fetal bovine serum under subconfluent conditions, Rcho-1 trophoblast cells have the properties of trophoblast progenitor cells, whereas Rcho-1 trophoblast cells allowed to grow to confluence in the presence of horse serum differentiate into TGC over a period of 12 days. Undifferentiated Rcho-1 trophoblast cells contained 215 pmol/mg/h GalNAc-transferase activity. Following induction to TGC, no GalNAc-transferase activity was detected even though ␤1,4Gal-transferase activity was readily detectable (not shown). These results suggest that Gal-NAc-transferase is expressed exclusively in spongiotrophoblasts of the junctional zone along with PLP-A, PLP-B, PLP-C, d/t PRP, and PL-Iv. Since these placental prolactin family members are all expressed in spongiotrophoblasts, the more efficient modification of PLP-A with ␤1,4-linked GalNAc than the other family members may reflect features of the glycoproteins themselves, most likely recognition of the peptide determinant.
PLP-A Serves as an in Vitro Target for the Glycoprotein Hormone GalNAc-transferase-We have shown that the glycoprotein hormone GalNAc-transferase recognizes a peptide determinant as well as the oligosaccharide acceptor (2)(3)(4)(5). In the presence of this determinant the catalytic efficiency for GalNAc addition to the same acceptor oligosaccharide is increased by as much as 500-fold. The high levels of transferase seen in near term rat placenta extracts could result in modification of glycoproteins devoid of the recognition determinant. This seems unlikely since even among the placental prolactin family members examined (Fig. 7), significant amounts of PLP-B, PLP-C, d/t PRP, and PL-Iv are not modified. Transfer of GalNAc to oligosaccharide acceptors on PLP-A and PLP-B were compared with transferrin, which does not have a recognition determinant, and hCG, which has recognition determinants on its ␣ and ␤ subunits (Table I). PLP-A and hCG are both modified efficiently with GalNAc. PLP-B is modified more efficiently than transferrin, 10-fold more GalNAc incorporated, but is a poor substrate when compared with PLP-A (Table I)

DISCUSSION
The mammalian placenta is a complex tissue that performs a number of essential functions, among them: 1) nutrient and waste exchange between the fetal and maternal circulations; 2) modulation of the maternal immune response to prevent rejection of the embryo; and 3) transduction of fetal and maternal signals (1). The rat has proved to be a useful model to study placental development (48,49). Between implantation and mid gestation the choriovitelline placenta, which contains a single differentiated trophoblast cell type, the TGC, plays the dominant role. From mid gestation onward the chorioallantoic placenta, which consists of a junctional zone and labyrinth zone, predominates. Of four differentiated trophoblastic cell types, TGC, glycogen cells, and spongiotrophoblasts are present in the junctional zone, whereas TGC and syncytial trophoblasts are present in the labyrinth zone. Individual PLPs that are structurally and functionally related to the prolactin and growth hormone produced by the pituitary are expressed at characteristic times during gestation by cells within the decidualized uterus, choriovitelline placenta, and chorioallantoic placenta (see Fig. 5). The initial appearance and rapid increase in glycoprotein hormone-specific GalNAc-transferase after day 13 of gestation in extracts from both the junctional and labyrinth zones of the chorioallantoic placenta suggested that specific trophoblast lineages of the chorioallantoic placenta are induced to express high levels of the GalNAc-transferase as they differentiate. The absence of either GalNAc-transferase or glycoproteins bearing ␤1,4-linked GalNAc on Asn-linked oligosaccharides in extracts from either decidualized uterus or the placenta prior to day 13 and the direct correlation of late developmental increases in GalNAc-transferase activity with spongiotrophoblast development support this conclusion.
PLP-A, PLP-B, PLP-C, d/t PRP, and PL-Iv are among the glycoproteins that we have shown bear Asn-linked oligosaccharides terminating with NeuAc␣2,6GalNAc␤GlcNAc␤. PLP-A, PLP-C, and PL-Iv are expressed predominantly in spongiotrophoblasts from mid gestation to term. The glycoforms of PLP-A, PLP-C, PL-Iv bearing ␤1,4-linked GalNAc are virtually quantitatively modified with sialic acid since little if any of these glycoproteins is bound by WFA prior to neuraminidase digestion, whereas each is essentially quantitatively bound following enzymatic removal of sialic acid. Thus, it is the concurrent expression of both the ␣2,6-sialyltransferase and the GalNActransferase in spongiotrophoblasts over the same time frame as their glycoprotein substrates which accounts for the highly efficient modification of their Asn-linked oligosaccharides with both GalNAc and sialic acid.
PLP-B and d/t PRP are also expressed in spongiotrophoblasts; however, prior to day 13 of gestation they are expressed in cells of the decidualized uterus (47). PLP-B and d/t PRP synthesized prior to mid gestation are devoid of NeuAc␣2, 6GalNAc␤GlcNAc␤, whereas after day 12 significant amounts of PLP-B and d/t PRP contain NeuAc␣2,6GalNAc␤GlcNAc␤, consistent with their synthesis in spongiotrophoblasts. As a result of the change in the cell type synthesizing these hormones, their glycoforms are altered.
We have shown previously that the glycoprotein-specific Gal-NAc-transferase recognizes peptide as well as oligosaccharide determinants (2)(3)(4)(5). In the case of the glycoprotein hormone ␣ subunit, the peptide determinant consists of a cluster of basic amino acids present in two turns of an ␣-helix. The crystal structure of hCG led us to the conclusion that it is the proximity of these residues in three-dimensional space to the oligosaccharides rather than their relationship within the linear amino acid sequence which is critical for efficient transfer of GalNAc. PLP-A also contains a peptide recognition determinant that is recognized by the GalNAc-transferase from bovine pituitary and is modified with nearly the same efficiency as hCG at a concentration of 7.5 M. PLP-B, in contrast, is modified almost 10-fold more efficiently than transferrin but at 1% of the rate of hCG. Thus, even though PLP-B is expressed in spongiotrophoblasts after day 12 of gestation, it is not modified to the same extent as PLP-A. PLP-C and PL-Iv are also not completely modified with GalNAc, suggesting that their peptide recognition determinants are not as effective as that on PLP-A.
The rodent prolactin/growth hormone family members are homologous with 30 -80% similarity in their amino acid residues (48). The crystal structure of growth hormone is known; and based on similarities to the placental prolactin family members, it has been predicted that each prolactin family member consists of four tightly packed ␣-helical domains (51). There is little homology in the first three domains of these proteins, but the fourth domain at the carboxyl terminus is highly conserved. This domain contains a high number of basic amino acids that may take on an ␣-helical conformation. Even though this region is carboxyl-terminal to the Asn-linked oligosaccharides on each of the placental prolactin family members, it is a good candidate region for the peptide determinant recognized by the GalNAc-transferase.
The terminal sequence NeuAc␣2,6GalNAc␤1,4GlcNAc␤ has to date been described on a number of glycoproteins. However, only a limited number of the glycoproteins synthesized by the rat placenta between mid gestation and birth contain this structure. Efficient modification of these glycoproteins with ␤1,4-linked GalNAc requires that they contain a peptide recognition determinant and be synthesized in cells expressing the GalNAc-transferase. In the case of PLP-B and d/t PRP, the mid gestational cell type switch from anti-mesometrial cells of the decidua to spongiotrophoblasts of the placenta is accompanied by a major change in their glycoforms.
What is the biological significance of oligosaccharides terminating with NeuAc␣2,6GalNAc␤1,4GlcNAc␤ rather than the more commonly encountered NeuAc␣2,6Gal␤1,4GlcNAc␤? Our results demonstrate that the synthesis of the former structure is developmentally regulated and that it is confined to a limited number of rat placenta glycoproteins. We are able to find PLP-A and other glycoproteins with the NeuAc␣2,6GalNAc␤1, 4GlcNAc␤ structure in the maternal circulation between mid gestation and birth, but not at other times. Glycodelin (placental protein 14), a human decidual and placental protein expressed in high levels early in pregnancy, is reported to have immunomodulatory effects (52,53). Glycodelin has recently been shown to bear Asn-linked oligosaccharides terminating with NeuAc␣2,6GalNAc␤1,4GlcNAc␤ (32). The authors suggest that this oligosaccharide structure may have a role in immunomodulation; however, direct evidence for this remains to be obtained.
The oligosaccharide structure terminating with NeuAc␣2, 6GalNAc␤1,4GlcNAc␤ may be highly characteristic of pregnancy in a number of species including humans. There are a number of potentially important roles for this structure. Receptors specific for NeuAc␣2,6GalNAc␤1,4GlcNAc␤ could regulate circulating half-life or direct glycoproteins bearing it to specific sites in the mother or fetus. This structure could have an immunomodulatory role by interacting with carbohydratespecific receptors related to CD22 or the selectins (50,54). Receptors for many of the placental prolactin family members are likely to be present in both the fetus and mother. Specific glycoforms could direct the different members of this family to specific sites and/or modulate the activation of their receptors.
Clearly many potential biological roles for NeuAc␣2, 6GalNAc␤1,4GlcNAc␤ are possible. The highly regulated ex-pression of the GalNAc-transferase and ␣2,6-sialyltransferase at the same time as select members of the placental prolactin family strongly supports the view that this structure will prove biologically important for these hormones.