The LHβ Gene of Several Mammals Embeds a Carboxyl-terminal Peptide-like Sequence Revealing a Critical Role for Mucin Oligosaccharides in the Evolution of Lutropin to Chorionic Gonadotropin in the Animal Phyla*

The expression of a previously untranslated carboxylterminal sequence is associated with the ancestral lutropin (LH) β to the β-subunit gene evolution of choriogonadotropins (CG). The peptide extension (denoted as CTP) is rich in mucin-type O-glycans and confers new hormonal properties on CG relative to the LH. Although the LHβ gene is conserved among mammals and only a few frameshift mutations account for the extension, it is merely seen in primates and equids. Bioinformatics identified a CTP-like sequence that is encrypted in the LHβ gene of several mammalian species but not in birds, amphibians, or fish. We then examined whether or not decoding of the cryptic CTP in the bovine LHβ gene (boCTP) would be sufficient to generate the LHβ species of a ruminant with properties typical to the CGβ subunit. The mutated bovine LHβ-boCTP subunit was expressed and N-glycosylated in transfected Chinese hamster ovary cells. However, unlike human (h) CGβ CTP, the cryptic boCTP was devoid of mucin O-glycans. This deficiency was further confirmed when the boCTP domain was substituted for the natural CTP in the human CGβ subunit. Moreover, when expressed in polarized Madin-Darby canine kidney cells, this hCGβ-boCTP chimera was secreted basolaterally rather than from the apical compartment, which is the route of the wild type hCGβ subunit, a sorting function attributed to the O-glycans attached to the CTP. This result shows that the cryptic peptide does not orientate CG to the apical face of the placenta, to the maternal circulation as seen in primates. The absence of this function, which distinguishes CG from LH, provides an explanation as to why the LHβ to CGβ evolution did not occur in ruminants. We propose that in primates and equids, further natural mutations in the progenitor LHβ gene resulted in the efficient O-glycosylation of the CTP, thus favoring the retention of an elongated reading frame.

The family of glycoprotein hormones includes lutropin (LH), 1 follitropin, and thyrotropin, as expressed by the pituitary, and chorionic gonadotropin (CG), which is produced by the placenta of primates and equids (1). Each hormone is a noncovalent heterodimer composed of a common ␣-subunit and a unique ␤-subunit that confers receptor specificity. The ␤-subunits have substantial sequence similarities, including conserved location of cysteine residues that suggests an origin from a common ancestral gene (2). The human (h) CG␤ genes have evolved from an ancestral LH␤ gene by frameshift mutations that resulted in a readthrough into a previously untranslated region and the extension of the reading frame (3,4) (Fig. 1A). As a result, a short carboxyl-terminal sequence of hLH␤ is replaced by a longer carboxyl-terminal peptide (CTP) in hCG␤ (3)(4)(5)(6). In the equine (e), both the placental CG␤ and pituitary LH␤ subunits are products of the same gene (eLH/CG␤) and contain a CTP that is presumed to have evolved by a distinct mechanism (7) (for further explanations see "Results"). Except for primates and equids, no other identified animal species bears this carboxyl-terminal extended version of the LH␤ gene subunit.
Examination of the CTP sequences of CG␤ subunits from human (4), baboon (bab) (8), marmoset (mar) (9), equine (e) (7), donkey (d) (10), and zebra (z) (11) reveals that in addition to containing Ser/Thr-and Pro-rich regions, a cluster of serines located 4 -7 residues downstream from the conserved Cys 110 residue is a sequence signature (Fig. 1B). The attachment of multiple O-glycans is typical of the CTP domains of CG derived from humans and horses.
This carboxyl-terminal extension prolongs the circulatory survival relative to LH (1,(12)(13)(14). Fusing the hCG␤-CTP to the bovine LH␤ subunit resulted in a hormone with a longer halflife and various abnormalities related to the LH excess when this LH␤-CTP chimera was expressed in a transgenic mouse model (15). Although the hydrophobic carboxyl-terminal end of the hLH␤ subunit is involved in the intracellular retention of the protein, the hydrophilic CTP stretch of hCG␤ contains determinants that are favorable for secretion and, when fused to the ␤-subunits of human LH and human follitropin, enhance the release from transfected cells (14,16,17). In vivo, the storage and secretion profiles of the hormones differ; hLH is stored in secretagogue-sensitive granules, and there are data showing it is secreted from the basolateral surface of the pituitary gonadotropes (18 -20). However, hCG is not stored and is constitutively released to the maternal circulation through the villi at the apical side of the trophoblast cells (21)(22)(23).
Recent studies, utilizing the Madin-Darby canine kidney (MDCK) model, identified sorting determinants in the ␤-subunit of the hormones (24,25). Whereas the secretion of the hLH␤ subunit from transfected MDCK cells was basolateral, the hCG␤ subunit was routed to the apical secretion pathway, primarily because of the O-glycosylated CTP. These results suggested that the O-linked glycans are a major factor for the secretion of hCG into the maternal blood, a crucial step in sustaining high circulating levels in pregnancy (24).
It is not known why a carboxyl-terminal extended analog of the LH␤ subunit evolved only in primates and equids. The evolution of the LH␤ to CG␤, and the concomitant expression of the O-glycosylated CTP, resulted in both an enhanced secretion into the maternal serum and prolonged survival in the circulation. These features seem advantageous for the purpose of stimulating the ovary and delaying luteolysis of the corpus luteum of pregnancy, thus maintaining progesterone production. That the CTP sequence is restricted to primates and equids is intriguing because the LH␤ gene is highly conserved among mammalian species (2,26,27), and only a few mutations localized to a small region of this gene account for the extended sequences (4,7).
As a strategy to understand why the CTP extension of the LH␤ subunit is not widespread in the animal phyla, we analyzed the 3Ј region of the locus in different species and examined whether (a) the LH␤ genes in species other than primates or equids contain an untranslated CTP-like sequence and, if an encrypted CTP homolog exists (b), the mutations leading to the incorporation of the sequence in the reading frame would render the protein nonfunctional. Here we show that a CTPlike sequence is embedded in the LH␤ genes in several mammals but is not present in non-mammalian species. The requirement of placentation provides strong evolutionary pressure for the directed development of the elongated LH␤ protein in mammals other than primates and equids. In our search for clues for the restricted CG␤ subunit evolution, we constructed the bovine LH␤ mutant, where readthrough into the 3Ј region of the cDNA generated a carboxyl-terminal extended subunit variant and examined structural and functional properties of the cryptic CTP.

DNA Analyses
DNA sequences for CG␤ and LH␤ genes were downloaded from GenBank TM and are denoted by accession numbers. Because the analyses focus on the portions of the genes that express the carboxylterminal region of CG or LH␤ subunits and associated downstream regions, only sequences between the codon for Cys 110 and the AATAAA polyadenylation signal or 10 nucleotides 3Ј to the polyadenylation signal are illustrated in Figs. 1-3. Sequence comparisons were performed with the BESTFIT portion of the Wisconsin Genetics Computer Group Software (version 8; Madison, WI). The hCG␤ (gene 5) and eLH/CG␤ genes were taken as prototypical for primates and equids, respectively, because they are the most thoroughly characterized.
LH␤111 Subunit Variant-The LH␤ cDNA in pCI-Neo was used as a template for PCR with the forward primer 3 (5Ј-CCAGAACCCAAC-CCTTGGCCTGTGACTAACCCCCG-3Ј) and the reverse oligonucleotide 4 (5Ј-CACATTCCACAGCTGGTTCTTTC-3Ј) in order to engineer a subunit variant in which a termination codon (TAA) follows Asp 111 of the LH␤ subunit in pCI-Neo. We used the unique Van91I (in the bovine LH␤ cDNA) and KspAII (HpaI) (in the plasmid) sites to substitute the PCR product for the corresponding LH␤ cDNA fragment.
LH␤ hCTP Chimera (LH␤CTP)-The human CG␤ subunit was the source for the CTP. Using the plasmid pM 2 HA hCG␤ (28), with primer 5 (forward 5Ј-GCCTGTGACTTCCAGGACTCCTCTTCC-3Ј) and primer 6 (reverse 5Ј-TTAGGGCCCCTCGAGG-3Ј), a fragment corresponding to the CTP of human CG␤ subunit and containing in its 5Ј region 3 codons of the carboxyl-terminal of the bovine LH␤, DNA (codons 109 -111) was amplified. In a second reaction, pCI-Neo LH␤ subunit was used as a template, and the PCR with the forward oligonucleotide 7 (5Ј-GTCCT-GGAAGTCACAGGCCAAGGGTT-3Ј) and the reverse primer 8 (5Ј-GGT-GTCCACTCCCAGTTC-3Ј) amplified a fragment that contained the LH␤ subunit 1-111 and nine nucleotides corresponding to the carboxylterminal region of the hCG␤ subunit (codons 115-117). In an overlapping PCR, the two fragments were used with primers 6 and 7 to generate a DNA encoding the bovine LH␤ subunit, to which the CTP sequence (codons 115-145 of the hCG␤ subunit) was fused following residue Asp 111 and was cloned into pCI-Neo.
LH␤boCTP Variant-The LH␤ cDNA in pFastBac1 was used as a template with primer 9 forward (5Ј-CCAACCCTTGGCCTGTGACCCC-CAGACATCCTCTTCC-3Ј), which deletes 10 bp immediately after the Asp 111 codon, and the reverse primer 10 (5Ј-GCAGTTCAAGAAGTCT-TTATTGG-3Ј), which generates an in-frame stop codon, downstream to the poly(A) motif. This fragment was subcloned in pFastBac1 LH␤, and the modified subunit was used to prime a second PCR along with primers 1 and 2. The product of the second PCR was digested with PpuMI and XbaI and cloned into these sites of the double-digested LH␤ pCI-Neo vector in order to generate the LH␤boCTP variant.

Construction of the hCG␤-boCTP Chimera
For the engineering of the chimeric human CG␤-boCTP subunit variant, we exchanged the hCG␤ CTP with the boCTP, using pairs of mega primers and a bi-directional PCR sequence extension strategy. As a template, we used the hCG␤ subunit gene that includes a SalI site in the 2nd intron (28), and the boCTP sequence was placed following Asp 117 codon of the hCG␤ gene in two rounds of amplification. A forward primer 11 (5Ј-AGCAGACACTCTTCCCCTCCCTTCCCAATAA-AGACTTCTTGATCTAGATCGGACACCCCGATCCTCCCA-3Ј) and a reverse primer 12 (5Ј-GGTTTCCAGAGTTAGGATGGGCATGGGAGG-TTGAAGTGGGGCATCCTTAGAGGAAGAGGAGTCCTGGAAGC-3Ј) were used in the first reaction, and the forward primer 13 (5Ј-CTCCC-TTCCCAATAAAGACTTCTTGAGAATTCTCGGACACCCCGATCCTC-CCA-3Ј) and reverse primer 14 (5Ј-GGGAAGAG TGTCTGCTGGTTTC-CAGAGTTAGGATGGGCATGGGAGGTTG-3Ј) were used in the second round. The hCG␤boCTP variant was cloned in the pM 2 HA vector (28).
To ensure the fidelity of the LH␤ and CG␤ constructs, both DNA strands were sequenced with an ABI prism DNA sequencer (PerkinElmer Life Sciences). A silent mutation in the Val 25 codon was detected in the cDNA encoding the LH␤CTP variant and was not corrected.

Transfection of CHO Cells, Metabolic Labeling, and Immunoprecipitation
Plasmids containing the DNA that encodes the corresponding bovine LH␤ and human CG␤ variants were stably transfected into CHO cells, using the calcium phosphate method or with polyethyleneimine (29). Each transfection was repeated two or three times and either pooled or single clones resistant to G418 (0.250 mg/ml) were analyzed. The transfected CHO cells were maintained in Ham's F-12 medium supplemented with penicillin (100 units/ml), streptomycin (100 g/ml), and glutamine (2 mM), containing 5% fetal calf serum and 0.125 mg/ml of the neomycin analog as described previously (28).
For metabolic labeling experiments, transfected CHO cells were plated in multiwell plates and labeled for 18 (LH␤ variants) or 6 h (hCG␤ variants) with Pro-mix (28 Ci/ml) in cysteine and methionine-free media, as described previously (28). Cell lysate and media samples were immunoprecipitated with antisera raised against the ovine LH␤ or human CG␤ subunits and resolved under reduced conditions on 12.5 or 15% SDS-PAGE, as indicated in the figure legends, and autoradiographed.
To examine the presence of N-linked glycans, the transfected cells were pre-incubated for 6 h with tunicamycin (1 g/ml; Sigma) and labeled with Pro-mix (28 Ci/ml) for an additional 18 (LH␤ subunit variants) or 6 h (unmodified and chimeric CG␤ subunits) in the presence of tunicamycin (1 g/ml) or vehicle alone (0.1% ethanol). The labeled proteins were immunoprecipitated and electrophoresed as described above.

Secretion from MDCK Cells
MDCK cells (strain II) were grown in Dulbecco's modified Eagle's medium/F-12 medium and stably transfected with the plasmid encoding the hCG␤ subunit or the hCG␤boCTP chimera. The cells (maintained for less than 15 passages) were grown on 24-mm Transwell filters (0.4 m pore size, Corning-Costar, Cambridge, MA), permitting separate access of media to the apical and basolateral faces of the membrane. Cells were metabolically labeled with [ 35 S]cysteine (25 Ci/ ml) for 16 h as described (25). Samples of apical and basolateral media were immunoprecipitated with hCG␤ antiserum, and the reduced proteins were resolved on 15% SDS-PAGE and autoradiographed. Several clones were analyzed, and the experiment was repeated three times. The results are expressed as mean (%) ϮS.E.; a p Ͻ 0.05 was considered significant.

RESULTS
A CTP-like Sequence Is Encrypted in Non-primate, Nonequid LH␤ Genes-A comparison of the 3Ј region of the LH␤ and CG␤ genes identified the presumed changes that led to the evolution of the CTP in the reading frame of the extended subunits in primates and equids (4, 7) (Fig. 1). In contrast to humans, no gene duplication occurred in the horse; the LH␤ gene lacking the CTP does not exist in equids, and the developmental pathway of the eLH/CG␤ CTP-bearing gene was proposed, based on sequence alignment with the LH␤ DNA of humans and cattle (7,10). The analyses showed that in humans and horses, frameshift deletion in the region of codons 112-115 of the ancestral LH␤ gene (deletion of one nucleotide in codon 114 and 10 bp subsequent to codon 111 in primates and equids, respectively) accounted for extension of the reading frame (4,7). The mutations share a pivotal 1-bp phase change of the reading frame that skips the LH␤ stop codon, and a new termination codon was created within (as of primates) or downstream (as of equids) to the polyadenylation signal.
The amino acid sequences resulting from one nucleotide deletion in codon 114 in the LH␤ genes from the bovine (30), ovine (31), porcine (32), mouse (33), rat (34), and canine (35) species are illustrated in Fig. 2. The altered reading frames generate Ser/Thr-and Pro-rich sequences characteristic of the CTP sequence in the CG␤ subunit of primates and the LH/CG␤ in equids (Fig. 2). The frameshift in the bovine or ovine LH␤ gene did not generate a termination codon from the AATAAA polyadenylation signal, suggesting that additional modifications would be required to introduce a stop codon. Analogous to the CG␤ subunit, the extended LH␤ variants should be 138 -150 amino acids in length (Table I). Thus, a deletion that shifted the reading frame by 1 bp proximal to codon Cys 110 generated a CTP-like sequence of the LH␤ gene from several mammalian species.
In contrast, similar frameshift mutations in the LH␤ genes of species lacking the placenta, e.g. avian species (GenBank TM accession numbers S70834 for quail and L35519 for turkey) (36,37) or either of the gonadotropin ␤-subunit genes of the toad Bufo japonicus (38) (GenBank TM accession number AB085662) or fishes (GenBank TM accession numbers M27153 and M27154, L35070 and L35096, and D88023 and D88024, for example) (39 -41), did not generate a CTP-like sequence in any of the three reading frames (data not shown). It appeared that DNA sequences that encode CTP-like domains occurred primarily, if not exclusively, in the LH␤ gene of mammals.
Incorporation of the 3Ј-Untranslated Region of the Bovine LH␤ Subunit in the Reading Frame-Considering the high degree of identity in the sequences of the bovine and ovine genes, we decoded the embedded CTP from the bovine LH␤ gene to generate the carboxyl-terminal extended variant of ruminants (Fig. 3). Because the 3Ј region of the equine LH/CG␤ and bovine LH␤ genes was similar, we chose a strategy that mimicked the proposed developmental paradigm of the LH/ CG␤ gene subunit of the horse and donkey (7,10). The bovine LH␤ cDNA was modified by eliminating 10 bp in the 112-115 codons to shift the reading frame. To create a new in-frame stop codon, a dinucleotide mutation (CA to TG) was generated 8 bp FIG. 1. The carboxyl-terminal region and CTP domains of the CG␤ subunit in primates and equids. A, illustration of the LH␤ to CG␤ gene evolution. Diagram of the LH␤-coding sequence is presented by a white box, and the 3Ј-untranslated region (3Ј-UTR) is presented by a line. The location of the Cys 110 residue and termination codons is illustrated by 110 C and STOP, respectively. The deletion mutation in the Cys 110 region, resulting in the frameshift and elongation of the CG␤ gene, is presented by the ⌬. The solid box indicates the CTP domain. B, comparison of CG and CTP amino acid sequences from the human (h; gene 5; GenBank TM accession number X00266), baboon (bab; Gen-Bank TM accession number M14966), marmoset (mar; GenBank TM accession number U04447), equine (e; GenBank TM accession number S41704); donkey (d; GenBank TM accession number X80116), and zebra (z; GenBank TM accession number Y16265). Sequences begin at Cys 110 and have been aligned to illustrate maximum similarity in amino acids.
downstream from the AATAA polyadenylation signal (Fig. 3A). The mutations yielded a 147-amino acid protein, designated as LH␤boCTP, 2 with a putative CTP domain (Fig. 3B). We refer to this homologous domain as cryptic boCTP, because the sequence exists in the DNA but is not translated. For comparison, two additional variants were engineered; one lacks the carboxyl decapeptide by introducing a stop codon following the Asp 111 (designated LH␤111), and in the second, the CTP region of the hCG␤ subunit was fused to the truncated bovine LH␤111 subunit (denoted as LH␤CTP). Schematic diagrams of the subunit variants are depicted in Fig. 3C.
Expression of Carboxyl-terminal Extended Bovine LH␤ Subunit Variants-To examine whether the carboxyl-terminal extended bovine variant can be expressed, we stably transfected the plasmid into CHO cells. The mutagenesis could result in misfolding and degradation of the modified protein, but if the bovine LH␤ subunit tolerates the boCTP domain, we should detect a new protein with a higher molecular weight.
The transfected CHO cells were metabolically labeled for 18 h, and lysate and media samples were immunoprecipitated and resolved on SDS-PAGE (Fig. 4A). The ovine LH␤ antiserum recognized the bovine LH␤ and LH␤111 subunits in the cell lysate (apparent molecular mass of 18 -19 kDa) (Fig. 4A,  lanes 2 and 4). As expected, the LH␤ subunit was not detected with nonimmune rabbit serum (NRS; Fig. 4A, lane 1). The unmodified subunit was not observed in the culture media (Fig.   4A, lane 3), suggesting that it was retained in the cell as shown previously (42,43). The electrophoretic migrations of the intracellular forms of labeled LH␤CTP and LH␤boCTP were similar (apparent molecular mass of 20 -21 kDa, Fig. 4A, lanes 6 and 8) and slower than the LH␤ subunit or the LH␤111 variant (Fig.  4, lanes 2 and 4), reflecting the addition of the heterologous hCG␤ CTP or the homologous boCTP sequence to the protein.
Although the amount of secreted subunit variants was low, that of LH␤CTP was the highest (Fig. 4A, lane 7 versus lanes 5  and 9). Densitometric analysis of bands revealed that 19 Ϯ 3% of the total synthesized LH␤CTP protein was recovered into the media, as opposed to 8 Ϯ 2% for LH␤111 and 7 Ϯ 1% for the LH␤boCTP variant, and was below the detection level for the WT LH␤ subunit (p Ͻ 0.05 for the LH␤CTP as compared with LH␤111 or LH␤boCTP). The data suggest that retention signals are not restricted to the carboxyl terminus of the subunit and imply that the expression of the cryptic boCTP does not result in a quantitative secretion of the monomeric subunit.
The secreted forms of the LH␤CTP variant migrated corresponding to a much higher apparent molecular weight as compared with the lysate sample (Fig. 4A, lane 6 versus 7). The extent of this lysate/media shift was consistent with the processing of the N-linked glycan attached to the LH␤ subunit and the addition of O-linked carbohydrates to the CTP in the Golgi prior to secretion. The data suggest that sequences in the bovine LH␤ subunit do not inhibit O-glycosylation of the CTP. In contrast, there was no significant difference in the mobility of the intracellular and secreted LH␤boCTP forms (Fig. 4A,  lane 8 versus 9). Similar results were observed when the LH␤CTP and LH␤boCTP variants were immunoprecipitated with a different antiserum that was raised against the hCG␤ (Fig. 4B, lane 1 versus 2 as compared with lane 3 versus 4). Thus, despite the sequence similarity with the authentic hCG␤ CTP, the cryptic boCTP sequence did not enhance the secretion of the subunit lacking the carboxyl-terminal decapeptide and appeared in the culture media without the set of O-glycans.
N-Linked Glycan Modification of the LH␤ Subunit Derivates-N-Glycosylation of the gonadotropin subunits had a significant impact on the folding, assembly, secretion, and stability of the gonadotropin subunits. To examine whether the addition of the boCTP sequence inhibited the attachment of the N-linked glycan to the acceptor Asn 13 of the bovine LH␤ subunit, we tested the sensitivity of the intracellular LH␤boCTP form to tunicamycin. The lysate sample of the LH␤ subunit variants expressed in the presence of the inhibitor (Fig. 5, lanes  2, 4, 6, and 8; N-dg) had an increased mobility relative to the corresponding untreated glycosylated proteins (Fig. 5, lanes 1,  3, 5, and 7). This sensitivity was indicative of the attachment of the N-linked carbohydrates to the subunit and suggested that modification of the LH␤ carboxyl-terminal end and expression of the cryptic boCTP or the hCG␤ CTP did not prevent the N-glycosylation of the subunit. Taken together, the experiments showed that the incorporation of the cryptic boCTP sequence in the reading frame did not inhibit the biosynthesis and N-glycosylation of the transformed bovine LH␤ subunit, suggesting the subunit tolerated the mutagenesis. However, the extent of O-glycosylation markedly differed between the natural hCTP and the cryptic boCTP.
Expression of the Human CG␤boCTP Chimera-Unlike the LH␤ subunit, the hCG␤ subunit was quantitatively secreted. This provided a platform to examine further if the cryptic boCTP was O-glycosylated. We therefore substituted the cryptic boCTP for the hCG␤ CTP and engineered a chimera (hCG␤boCTP) that encoded 150 amino acids compared with the 145 residues of the WT hCG␤ subunit (Fig. 6, upper panel, comparison of the sequences). As shown (16), the secreted 2 Because the remainder of this paper deals extensively with the bovine LH␤ subunit, the acronym LH␤ will be used to denote the bovine subunit. Ovine, equine, and human gonadotropins will be denoted by the prefix "o," "e," or "h," respectively.

FIG. 2. Effect of one nucleotide deletion in codon 114 of LH␤ genes from non-primate, non-equid mammalian species.
Each sequence starts at the codon for Cys 110 , and the normal LH␤ stop codon is double underlined. The deduced amino acid sequence is illustrated below the DNA sequence. Note that the frameshifts generate Ser-/Prorich sequences with clustered Ser residues (underlined), characteristic of CTP domains. Species designations are as follows: bovine (b; Gen-Bank TM accession number M11506), ovine (o; GenBank TM accession number S64695), porcine (p; GenBank TM accession number D00579), mouse (m; GenBank TM accession number U25145), rat (r; GenBank TM accession number J00749), and canine (c; GenBank TM accession number Y00518). hCG␤ monomer from CHO cells was resolved into two forms that are due to the presence of one (N1) or two (N2) N-glycans (Fig. 6A, lane 2). The hCG␤boCTP chimera was efficiently secreted and resolved as two forms, and the upper band appeared broader and less distinct on the gels relative to that of the WT subunit (Fig. 6A, lane 4). This may consist of N-linked sugar heterogeneity of the N1 or N2 glycoforms resulting from altered boCTP/hCG␤ interactions, as observed previously in the CG␤ subunit lacking the natural CTP (44). Of significance, whereas the lysate forms of the WT and chimeric subunits co-migrated, the secreted hCG␤boCTP migrated faster on the gels than the WT secreted subunit (Fig. 6A, lanes 1 and 2   versus 3 and 4). These data showed that, compared with the hCG␤-CTP, the cryptic boCTP lacked O-glycans when fused to the core hCG␤ subunit, consistent with the results observed when incorporated into the bovine LH␤ reading frame.
We then examined how the hCG␤boCTP chimera will resolve when the addition of the N-linked glycans was inhibited. If O-linked glycans are attached to the cryptic CTP, differences between the migration rate of the lysate and media forms should be observed following inhibition of the N-glycosylation, and if missing, no shift is expected. Tunicamycin treatment increased the migration rate of the CG␤ subunit and the hCG␤boCTP chimera in the lysate and media samples relative  to untreated cells (Fig. 6B, hCG␤, lanes 1 and 2 versus 3 and 4; hCG␤boCTP, lanes 5 and 6 versus 7 and 8) because of the inhibition of the N-linked glycosylation. It is clear that when N-glycosylation was inhibited, the lysate to medium shift was still observed for the WT CG␤ subunit due to the addition of the O-linked glycans to the CTP prior to secretion (Fig. 6B, lane 3  versus 4). In contrast, the intracellular and the secreted Ndeglycosylated hCG␤boCTP forms co-migrated when the addition of N-glycans was inhibited (Fig. 6B, lanes 7 and 8, arrow). The results suggest that the set of multiple O-glycans attached to the CTP of the hCG␤ subunit was missing in the decoded CTP. This was reproducible in metabolically labeled clones isolated from three different transfections (data not shown). The secretion of the hCG␤boCTP in the absence of O-and N-glycans (Fig. 6B, lanes 6 and 8) suggested that the determinants for secretion of the chimera were established by the peptide backbone. The secreted, tunicamycin-treated hCG␤boCTP did not resolve as a single protein (Fig. 6B, lane 8), and the faint higher molecular bands may have resulted from an incomplete inhibition of the N-glycosylation. That the hCG␤boCTP chimera undergoes N-but not O-glycosylation is further evidence that the decoded domain is not permissive for efficient mucin-type modification.
Secretion of the hCG␤-boCTP Chimera from Polarized MDCK Cells-It has been reported that hCG release to the maternal compartment is through the apical face of the placenta and is attributed to the O-glycans attached to the CTP of the hCG␤ subunit (24). Because of this function, a key question is whether the hCG␤boCTP chimera is sorted to the apical secretory pathway, like the hCG␤ subunit, or is released through the basolateral compartment as the human LH␤ subunit and the O-deglycosylated CG␤ subunit (24). We addressed this point, using the MDCK cell model that was validated previously, to study the secretion patterns of the gonadotropins and their subunits (24,25). The hCG␤boCTP chimera was stably transfected into the cell line, and several clones were selected, grown in Transwells, labeled, and analyzed.
The hCG␤boCTP chimera was secreted from the cells, primarily through the basolateral side (Fig. 7A, lane 1 versus 2). Densitometric analysis of the secreted labeled chimera revealed that 69.6% (Ϯ2.6) was released through the basolateral compartment of the cells and only 30.4% (Ϯ2.6) was secreted through the apical face (Fig. 7B, data pooled from four different clones). This distribution pattern contrasts with that of the WT CG␤ subunit (Fig. 7A, lane 3 versus 4). The majority of the unmodified hCG␤ subunit was sorted toward the apical compartment (Fig. 7B, 65 Ϯ 1.6 and 35 Ϯ 1.6% detected in the apical and basolateral sides, respectively; three different clones), as reported previously (24). That the cryptic boCTP lacked the signals for apical release from polarized cells was consistent with the absence of O-glycosylation. DISCUSSION Here we address the question as to why CG only developed in primates and equids but not in other vertebrates. The restricted evolution of LH␤ to CG␤ in the animal kingdom is curious, because the LH␤ genes from a large variety of mammals share substantial sequence identity. In addition, the evolution of LH␤ genes for the encoding of CG protein results in functions that distinguish CG from LH and appear favorable for successful pregnancy in mammals. We have shown that an untranslated CTP-like sequence is cached in the LH␤ gene of several mammalian species; a 1-bp frameshift would elongate the LH␤ reading frame. This change decodes a carboxyl terminus characteristic to the CG␤ subunit in place of the LH␤ carboxyl-terminal residues.
The existence of a putative CTP in the LH␤ gene of bovine, ovine, and porcine (order Artiodactyla), mouse and rat (Rodentia), and canine (Carnivora) indicates that the LH␤ to CG␤ gene conversion is in principle possible in mammalian species classified into different families and orders. Reading through the 3Ј region of the LH␤ gene in additional species, which naturally do not include the CTP sequence in the reading frame, identified an encrypted CTP-like sequence in the rabbit (GenBank TM accession number AY614703) and white rhinoceros (45) (GenBank TM accession numbers AF024520 and AF024521) genes (data not shown). The 3Ј region of LH␤ gene in the guinea pig (46) (GenBank TM accession numbers AF355775 and AAQ75732) and marsupials (47) (red kangaroo and brushtail possum, GenBank TM accession number AF017450 and AF017448, respectively) has some of the CTP characteristics but deviated from the stretch in most other mammals. In contrast, the 3Ј region of the gonadotropin LH␤ subunit of birds, amphibians, and fishes do not appear to pos-  1% ethanol; lanes 1, 3, 5, and 7). Cell lysates were immunoprecipitated with anti-ovine LH␤ and subjected to SDS-PAGE as in Fig. 4. sess DNA sequences that encode the stretch. Gestation in most mammals is intrauterine; the pregnant female exchanges communication signals and metabolic products with the conceptus by means of a placenta, and the mother produces milk to nourish the young. That the LH␤ gene of non-mammalian species is not suitable to evolve into a CG␤ gene is consistent with the absence of the placenta. These vertebrates usually deposit eggs, lack the placenta, and do not feed the offspring with milk. Thus, the presence of the CTP-like sequence in the LH␤ gene of viviparous organisms and its absence from the locus of oviparous species suggest that diversification of the 3Ј region is correlated to the conceptual nourishing strategy. Taken together, the data imply that the CTP became inscribed in the LH␤ gene after mammals split from other vertebrates.
In species lacking CG, alternative signals evolved to delay the degeneration of the corpus luteum of pregnancy. In ungulate ruminants (e.g. cows and sheep), interferon-is produced by the early embryo and attenuates endometrial prostaglandin F 2␣ , and the down-regulation of the luteolytic prostaglandin prolongs the life span of the corpus luteum (48,49). Because the bovine corpus luteum expresses LH receptors (50, 51), a direct luteotrophic signal secreted from the embryo to the maternal circulation, such as CG, could provide a backup system to reduce early pregnancy loss. Several reasons can account for the lack of emergence of a bovine/ ovine CG␤ subunit during the course of evolution, although a CTP-like sequence exists in the LH␤ gene. These include a placentation mechanism that differs from that of primates and equids, the lack of promoter elements in the LH␤ gene for expression in trophoblast cells, and as proposed here, unfa-vorable structural/functional features of the carboxyl-terminal extending region of the transformed subunit.
The structural requirement for the addition of sugar moieties to proteins is not completely understood. Whereas Asn-X-(Thr/ Ser) is a well characterized N-glycosylation motif, no clear consensus sequence is known for O-glycosylation of hydroxyl amino acids. Screening of data bases and mutagenesis studies showed that there is a high preference for proline, serine, threonine, and alanine residues around the mucin sites (52,53). In many cases, Pro residues occur in the 4 -8-position before or after the acceptor site (52,53). Based on these criteria, the boCTP (that encodes 8 serine, 4 threonine, and nearby proline residues) is a candidate for mucin-type glycosylation; primary sequence analysis using the predictor NetOGlyc3.1 server (54) (www.cbs.dtu.dk) identified all the Ser/Thr residues of the boCTP as potential O-glycosylation sites (data not shown). However, the decoded domain was devoid of detectable O-glycosyl units when fused to the bovine LH␤ subunit or to the human CG␤ subunit. The O-glycan deficiency is a striking difference from the modification of the primate/equid CTP domain, 4 mucin type O-glycans attached to the CTP of the hCG␤ subunit and 12 acceptor sites identified in the eCG␤ CTP (1,55).
The reasons why the multiple potential acceptor sites for O-glycosylation in the boCTP are ignored are currently unknown. This issue is intriguing especially because several mucin sites of the heavily glycosylated equine LH␤/CG␤ CTP are conserved in the cryptic boCTP (Fig. 3B). One of the factors influencing the efficiency of the O-glycosylation initiation is the residues neighboring the acceptor site (56). The absence of  1 and 2) and the hCG␤boCTP chimera. The migration of the CG␤ subunit glycoforms presumably bearing two (N2) or one (N1) N-glycans are indicated by arrowheads. B, N-glycosylation of the hCG␤boCTP chimera. CHO cells expressing the unmodified subunit (lanes 1-4) or the hCG␤boCTP chimera (lanes 5-8) were incubated in the absence or presence (1 g/ml) of tunicamycin for 6 h and labeled for additional 6 h. The gel was electrophoresed for a longer period than in A to allow a magnified resolution of the proteins. glycans may result from amino acids surrounding the potential acceptors in the boCTP, causing misalignment of the primary sequence that prevent the GalNac transferase from recognizing the Ser/Thr targets. Alternatively, regions of the cryptic CTP may inhibit the accessibility of the GalNac transferase(s) to potential acceptor sites, as discussed previously (57,58). The high Pro content of the boCTP (about 25% of the residues) is indicative for a little secondary structure, which is a favorable feature for O-glycosylation. This implies that the inefficient glycosylation could also be related to the lack of recognition of the potential acceptors. Additional mutations to those shifting the frame would also be required in the core bovine LH␤ subunit gene or, more likely, in the cryptic CTP sequence to yield an O-glycosylated domain.
Two of the physiological functions of the O-glycosylated CTP that distinguish CG from LH are the enhancement of circulatory survival and direction of the hormone to the maternal bloodstream at the apical face of the placenta (1,19,22,24). The secretion of the heterodimeric gonadotropins from transfected MDCK cells mirrors that seen for the corresponding ␤-subunit (24,25), and the different in vitro profiles recapitulate the exit compartment of LH and CG from the pituitary and the placenta, respectively (19,22,24). The secretion of the hCG␤boCTP chimera from the polarized MDCK cells was predominantly basolateral. The data imply that if the carboxyl-terminal extended bovine LH␤ subunit had evolved in vivo, the LH-CTP variant would have been secreted through the basolateral compartment. This then will bias the distribution of the hormone between the fetus and maternal sides of the placenta, and it would be difficult to maintain high CG levels in the maternal circulation because of the inefficient supply. That the existing CTP-like sequence in the bovine LH␤ gene is not a major site for the attachment of O-glycans and lacks the signals for the critical physiological function seems disadvantageous for the expression of the bovine CG in the pregnant cow. The absence from the extracellular space of the pregnant organism provides an explanation for the nonappearance of CG in ruminants and other species, which adopted different signals to maintain gestation.
We propose that the CG␤ subunit emerged only in primates and equids following a gradual evolution of the LH␤ DNA in vertebrates. Initially, the primitive LH␤ gene of non-mammalian species was mutated to encrypt a CTP-like sequence in the 3Ј region. As a result, the LH␤ locus in mammals, additional to equids and primates, became compatible for the elongation. Yet, if not for selection pressure, the existence of the CTP-like stretch in the LH␤ DNA is insufficient to drive the evolution of the CG␤ protein. Further mutations in the ancestral LH␤ gene of primates and equids created multiple O-glycan sites in the CTP sequence, and in combination with the frameshift mutations, the resulting elongated subunit acquires new features when expressed, thus motivating its development.