Evolution of Lutropin to Chorionic Gonadotropin Generates a Specific Routing Signal for Apical Release in Vivo

One of the fundamental differences among mammals is the mechanism of maintaining the corpus luteum of pregnancy. Placentation in primates is associated with the production of the glycoprotein hormone chorionic gonadotropin (CG), which is secreted into the maternal serum and stimulates progesterone synthesis from the corpus luteum, which is essential for early development of the embryo. CG together with the pituitary hormones lutropin (LH), follitropin, and thyrotropin constitute the family of glycoprotein hormones comprised of a common alpha subunit and a hormone-specific beta subunit. The LHbeta and CGbeta subunits share 85% amino acid sequence identity, and functionally LH and CG are interchangeable. CGbeta evolved by a recent gene duplication event from the LHbeta locus, and despite the close relationship between them, their modes of secretion are quite different. CG release from the placenta is apically directed, whereas LH is released from the basal side of the cell, and the determinant(s) for this redirected trafficking are unknown. Here, using the polarized Madin-Darby canine kidney (MDCK) cell line, we provide evidence for the molecular basis of the different secretory patterns of LH and CG in vivo. The apical targeting of CG is programmed by a carboxyl-terminal sequence, which encodes a novel sorting signal. It is also apparent that the presence of the O-linked oligosaccharides in the CTP sequence contributes to this apical routing. The CTP, which is absent in LH, redirects CG to the maternal serum and permits the unique arrangement for primate placentation. Our data also show that the MDCK cells can distinguish the different secretory pathways for the gonadotropins and will be a valuable model for elucidating the determinants associated with the unique sorting of these functionally related hormones.

The evolution of primates is associated with a fundamental change in the morphogenesis of the placenta (1,2). Placentation in primates is coupled to the synthesis of glycoprotein hormone chorionic gonadotropin (CG), 1 which is secreted into the maternal serum and stimulates progesterone synthesis from the corpus luteum, thereby allowing early implantation and development of the embryo (3,4). CG and its pituitary counterpart, lutropin (LH), comprise a family of heterodimeric glycoprotein hormones, including follitropin and thyrotropin, that share a common ␣ subunit, but differ in their hormonespecific ␤ subunits (5). Although the ␤ subunits determine biological specificity of each hormone, there is significant structural similarity between them. This is most evident for the LH␤ and CG␤ subunits, which share 85% amino acid identity in the first 114 amino acids (6). A major difference between the two subunits is the presence in the CG␤ subunit of a 31-amino acid carboxyl-terminal extension (CTP) compared with a shorter 7-amino acid stretch in LH␤ (see "Results").
The CG␤ subunit is specific to primates and evolved by a gene duplication event from the LH␤ locus (7); functionally the two are interchangeable. Similar to other mammals, primates express LH during pregnancy, raising the question as to why primates require CG. When comparing the biosynthesis of LH and CG, two important differences emerge: the pathways of secretion and polarity of their release. LH is packaged into storage granules (8), subject to regulated exocytosis by secretagogues (9), and released via the basolateral surface of pituitary gonadotrope cells (10,11). In contrast, CG is secreted by an apical route through the villous and directly into the intervillous space created by the implanted placenta (1,(12)(13)(14). The molecular basis by which the intracellular trafficking of CG is directed to exit the chorionic villi into the maternal lumen is unknown. To examine whether the unique circulation profiles exhibited by LH and CG in vivo are reflected in differences in polarized secretion, and to identify potential targeting sequences, we co-transfected the common ␣ subunit and LH␤/ CG␤ genes into Madin-Darby canine kidney (MDCK) cells. This cell line is an excellent model to study polarized secretion of endogenous and exogenous proteins in culture (15,16). The plasma membrane of these epithelial cells is divided into apical and basolateral domains by tight junctions (17). The apical side faces the luminal or exterior milieu, which is covered by microvilli and is involved in absorptive or secretory processes, whereas the basolateral side faces the serosal environment (16,17). Here we examined the polarized secretion of LH and CG in transfected MDCK cells and show that apical targeting of CG is programmed by the unique CTP, a novel sorting signal that is generated by a single frameshift mutation in the ancestral LH␤ gene.

MATERIALS AND METHODS
Polyclonal antisera directed against the ␣ and CG␤ subunit were prepared in this laboratory. Purified hCG (CR-127; 14,900 IU/mg) was provided by Dr. A. F. Parlow (National Hormone and Pituitary Program, National Institutes of Health, Torrance, CA).
Cell Culture-MDCK cells (strain II) were a gift of Dr. Sharon Milgram (University of North Carolina). These cells were grown in DMEM/ F-12 medium (for no more than 10 -15 passages) supplemented with 2 * This work was supported by a grant from Organon. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin and containing 5% FBS (v/v) at 37°C in a humidified 5% CO 2 incubator (18). The cells (3.5 ϫ 10 5 cells/ml) were seeded on 12-mm or 24-mm Transwell filters (0.4 m pore size, Corning-Costar, Cambridge, MA) allowing separate access of media to the apical and basolateral faces of the membrane. Filter-grown cells were cultured for 3-4 days with changes of medium daily.
To measure the integrity of epithelial monolayer and formation of tight junctions (as an indicator of achieving polarization), electrical resistance was measured across the cell monolayer using the Millicell Electrical Resistance System (Millipore Corp., Bedford, MA) (18).
Transfection and Clone Isolation-Transfection was performed using the DNA-calcium-phosphate precipitation method as described previously (19). Cells grown on plastic dishes were stably transfected with pM 2 HA (20, 21) containing either the ␣, LH␤, CG␤, or CG␤114 subunit genes.
Metabolic Labeling-MDCK cells were labeled overnight with 25 Ci/ml [ 35 S]cysteine in DMEM/F-12 medium lacking cysteine and supplemented with 5% dialyzed fetal bovine serum, L-glutamine, penicillin, and streptomycin (20). Labeling medium was added to the apical and basolateral compartments, and media from them were collected into separate tubes and were immunoprecipitated with antisera directed against ␣ and CG␤ subunits using Pansorbin. The reduced (with 10 mM ␤-mercaptoethanol) and heated (4 min) proteins were resolved as on 15% SDS-polyacrylamide gels described previously (19).

RESULTS AND DISCUSSION
We first examined the secretion patterns of the LH and CG dimers from transfected MDCK cells grown in Transwells and labeled overnight with [ 35 S]cysteine. The media were immunoprecipitated with CG␤ antiserum, which reacts equally well with either the LH␤ or CG␤ subunits; co-precipitation of the ␣ subunit is indicative of heterodimer formation. LH dimer released from the cells was primarily associated with the basolateral side, indicating that it is secreted in a polarized manner (Fig. 1A, lanes 1 and 2). The basolateral to apical ratio for LH was 3:1 (76.2 Ϯ 1.6% and 23.8 Ϯ 1.6%, respectively: n ϭ 6; p Ͻ 0.05) (Fig. 1B). Control experiments demonstrated that when the integrity of the monolayer was disrupted with a cell scraper, determined by reduction in electrical resistance, LH dimer traversed the filter and distributed equally to both compartments (Fig. 1A, lanes 3 and 4). The preferential basolateral routing of LH is consistent with its known distribution and release from the pituitary (10,11). LH-containing secretory granules redistribute subcellularly and become polarized to the side of the gonadotrope nearest to the vascular sinusoid during the preovulatory stage, and release of the granule content occurs in the basal region of the cell (10,11).
In contrast to LH, CG was secreted preferentially into the apical compartment ( Fig. 2A, lanes 1 and 2). Densitometric analysis of labeled bands demonstrated an apical to basolateral ratio of 2.7:1 (73 Ϯ 1.5% versus 27 Ϯ 1.5%, respectively: n ϭ 6; p Ͻ 0.05) (Fig. 2B). These data reflect differences in the secretion behavior of CG compared with its pituitary homologue LH. The apical secretion of CG is consistent with its in vivo release, since CG enters the maternal blood lakes of the uterus via an apical route through the villous (1,(12)(13)(14).
To examine the possibility of nonspecific transfer of CG and LH across the MDCK cell monolayer, conditioned media containing 35 S-labeled CG or LH from the MDCK cell-line was added to nontransfected MDCK cells grown on Transwell filters. After overnight incubation lysates and media from the apical and basolateral compartments were immunoprecipitated. No transfer of protein from one compartment to the other was detected (data not shown), indicating direct and specific sorting of CG and LH into the apical and basolateral compartments, respectively. These data demonstrate that the MDCK model recapitulates the secretion polarity for pituitary LH and placental CG as seen in vivo.
Numerous studies have shown that the uncombined ␣ sub-unit is primarily secreted constitutively (22,23). To assess the specificity of the MDCK model, cells expressing only the ␣ subunit were examined (Fig. 3). In this case the secretion pattern was random, i.e. the ␣ subunit was observed equally in both compartments (Fig. 3A, lanes 1 and 2). Presumably, the differential sorting exhibited by LH and CG is due to the ␤ subunit, since the ␣ subunit is the same for both hormones. To test this prediction, and to address the question if dimer formation is a prerequisite for sorting, MDCK cells expressing either the LH␤ or CG␤ subunits were grown on Transwells (Fig. 3, lanes 3-6). The ␤ subunits exhibited the same polarity as the corresponding dimers, indicating that the sorting signals are encoded in the hormone-specific ␤ subunit, and heterodimer formation is not required for the secretion polarity.
Based on their extensive sequence identity, it was suggested that CG␤ evolved from an ancestral LH␤ gene (7). The sequences of the carboxyl ends of the LH␤ and CG␤ termini are shown in Scheme 1.  1) and basolateral (BL, lane 2) compartments were immunoprecipitated with antiserum directed against the CG␤ subunit, which reacts equally with either the LH␤ or CG␤ subunits. Precipitation of the ␣ subunit with the CG␤ antiserum indicates dimer formation. The proteins were analyzed by SDS-polyacrylamide gels and autoradiography. Lanes 3 and 4 represent the control experiment and correspond to media collected from Transwell under the same conditions except that the monolayer was disrupted with a cell scraper, resulting in a low resistance (below 150 ohms/cm 2 ). B, the relative percent secretion of LH into the apical (stippled bars) and basolateral (solid bars) compartments was quantitated by densitometry from autoradiographs. The total combined secretion of protein into the medium of apical and basolateral compartments was taken as 100%. The apical and basolateral ratio for LH was determined from the percent of protein present in each compartment. Results are shown as the mean Ϯ S.E. (n ϭ 6). The asterisk indicates significant difference, p Ͻ 0.05.

Evolution of Lutropin to Chorionic Gonadotropin 880
pared with the 121 amino acids of the LH␤ subunit, and thus the two proteins have different COOH-terminal sequences (7). The seven amino acids at the carboxyl terminus of LH␤ are very hydrophobic; in contrast, the CG␤ carboxyl end is hydrophilic containing several serine residues, of which four are O-glycosylated. Thus, the apparent major evolutionary change from LH␤ to the CG␤ subunit was the appearance of the CTP. CG serum levels rise rapidly in the first trimester of pregnancy and decline thereafter (4,24). It is widely accepted that the main role for the CTP sequence is to enhance the circulatory half-life of CG during peak synthesis (21,25,26), maintaining a bolus of activity in early gestation to stimulate maternal progesterone production in the corpus luteum. However, our data implicate a novel role for this sequence as a routing signal. To examine whether the CTP is essential for polarity, a stop codon was introduced at residue 115 of the CG␤ subunit (21), resulting in a truncated protein of 114 amino acids (CG␤114) (Fig. 2A, lanes 3 and 4). In contrast to wild type CG dimer, the dimer bearing the CG␤114 mutant (CG114) was secreted basolaterally (68.9% Ϯ 1.7: n ϭ 6; p Ͻ 0.05) (Fig. 2B). These results confirm that the carboxyl-terminal sequence of the CG␤ subunit is a major routing signal for the apical secretion of CG dimer.
It has been reported that N-and O-linked oligosaccharides are associated with apical sorting signals for some glycoproteins. The O-glycosylated stalked region of the neurotrophin receptor has been implicated in its sorting to the apical membrane, since deletion of this region lead to a mistargeting of the mutant protein to the basolateral membrane in MDCK cells (27). Moreover, eliminating or reducing the number of O-glycans affects the polarity of intestinal brush border enzymes released from MDCK cells (28,29). However, deleting the Olinked oligosaccharides from aminopeptidase N had no effect on its apical secretion pattern from MDCK cells (28). Because there are four O-linked oligosaccharides in the CTP sequence, it is unclear whether the targeting function of CTP is encoded in the protein sequence and/or the carbohydrate. To address whether the O-linked carbohydrate in the CTP contributes to the apical secretion of CG, we constructed a CG␤ mutant (CG␤-Odg) (Fig. 4A) in which the four serine acceptor sites at residues 121, 127, 132, and 138 in the CTP were changed to alanine. (The Ser at position 130 was also mutated in this variant because we observed that some alternative O-glycosylation occurred at this site when the mutant was expressed in CHO cells.) 2 Synthesis of CG dimer bearing this mutation (Fig.  4A, lanes 1 and 2) or the uncombined CG␤ mutant (Fig. 4A,  lanes 3 and 4) displayed a basolateral preference (68.3% Ϯ 2.8 and 79% Ϯ 2.2, respectively; p Ͻ 0.05) (Fig. 4B). These results 2 V. Garcia-Campayo and I. Boime, manuscript in preparation.

FIG. 2. Secretion of CG and CG114 dimers.
A, MDCK cells expressing wild type CG or the truncated carboxyl-terminal mutant CG114 dimer were metabolically labeled overnight with [ 35 S]cysteine. The media from the apical (Ap) and basolateral (BL) compartments were immunoprecipitated with CG␤ antiserum and analyzed as described for Fig. 1. The migration of CG␤114 species bearing two (N2) or one (N1) asparagine-linked oligosaccharides and the ␣ subunit are indicated by arrows. N2 indicates the presence of both of the CG␤ Asn-linked oligosaccharides and N1 is a minor form bearing only one Asn-linked carbohydrate unit. CG␤114 co-migrates with the ␣ subunit on SDS gels. That the N2 form of CG␤114 and ␣ subunit reflect heterodimer formation was confirmed by nondenatured Western blots, which demonstrated a molecular mass of 35 kDa, corresponding to the intact dimer (data not shown). B, the relative percent secretion of CG and CG114 dimers was quantitated by densitometry. Results are shown as the mean Ϯ S.E. (n ϭ 6 -8). Asterisks indicate significant difference, p Ͻ 0.05. suggest that the apical sorting signaling of the CTP is mediated by the O-linked oligosaccharides.
A distinguishing structural feature between the LH␤ and CG␤ subunits is that the LH␤ subunit contains one N-linked oligosaccharide, whereas CG␤ contains two units. Thus, while the CTP is sufficient to route CG apically, we cannot exclude that the observed sorting is achieved in concert with the additional N-linked oligosaccharide on the CG␤ subunit. There is evidence that N-glycans can act as apical signals for some secretory and integral membrane proteins (30), but their exact role is unclear (31), since many glycoproteins do not display glycan-dependent apical targeting (32). This indicates that the contribution of N-glycans to apical targeting is protein specific. It has been suggested that N-glycans do not function as sorting signals directly but rather play an accessory role necessary for the function of an apical determinant (31,33).
There are several reports showing that motifs consisting of a tyrosine residue near at least one hydrophobic amino acid, or a motif bearing a leucine/leucine or a leucine/isoleucine pair are involved in basolateral sorting of membrane proteins (34 -36). It is intriguing that the unique carboxyl-terminal heptapeptide in the LH␤ subunit contains the dileucine motif.
Why only the CG␤ subunit among the glycoprotein hormone family contains CTP is unclear. Expression of CG in the placenta suggests that this extension is a critical factor in the gestational role of CG. A major function of the placenta during pregnancy is the exchange of metabolic products between the fetal and maternal blood streams (1). Thus, the CTP directs CG apically from the fetally derived placenta into the maternal intervillus spaces; these are luminal compartments filled with maternal blood derived from pressure gradients generated in the spiral arteries. Our data reveal that LH, which is elaborated by the pituitary directly into a single vascular compartment, is distinct from the production of humoral agents in the feto-placental unit. The use of the MDCK model has permitted the identification of a novel targeting domain that reversed the polarity of two evolutionarily related heterodimers, which activate the identical receptor. The appearance of CTP represents not only an adaptive response to maintain a high level of circulating gonadotropin, but also a targeting signal redirecting a pituitary function to the placenta, thus ensuring a successful implantation and development of the primate embryo.