Functional Contributions of Noncysteine Residues within the Cystine Knots of Human Chorionic Gonadotropin Subunits*

Human chorionic gonadotropin (hCG) is a heterodimeric member of a family of cystine knot-containing proteins that contain the consensus sequences Cys- X 1 -Gly- X 2 -Cys and Cys- X 3 -Cys. Previously, we characterized the contributions that cystine residues of the hCG subunit cystine knots make in folding, assembly, and bioactivity. Here, we determined the contributions that noncysteine residues make in hCG folding, secretion, and assembly. When the X 1 , X 2 , and X 3 residues of hCG- a and - b were substituted by swapping their respective cystine knot motifs, the resulting chimeras appeared to fold correctly and were efficiently secreted. However, assembly of the chimeras with their wild type partner was almost completely abrogated. No single amino acid substitution completely accounted for the assembly inhibition, although the X 2 residue made the greatest indi- vidual contribution. Analysis by tryptic mapping, high performance liquid chromatography, and SDS-polyacrylamide gel electrophoresis revealed that substitution of the central Gly in the Cys- X 1 -Gly- X 2 -Cys sequence of either the a - or b -subunit cystine knot resulted in non-native disulfide bond formationand subunit misfolding. This occurredeven

The cystine knot motif defines a superfamily of dimeric proteins and appears to function as a structural scaffold that stabilizes the 3-loop structures of the individual subunits (1).
As shown in Fig. 1, the cystine knot consists of three disulfide (S-S) 1 bonds; two of these bonds bridge adjacent polypeptide strands, creating a ring that includes the intervening polypeptide backbone, and the third bond penetrates this ring (1,2). The cystine knot is common to a biologically diverse set of dimeric proteins including transforming growth factor-␤, vascular endothelial growth factor, platelet-derived growth factor, and human chorionic gonadotropin (hCG) (1)(2)(3)(4). Additionally, more than 30 other proteins are predicted to contain this motif (2). The functional importance of the cystine knots of hCG (5)(6)(7)(8)(9), transforming growth factor-␤1 (10), and platelet-derived growth factor (11) is evident from studies where cysteine residues within the knot were mutated, thus preventing a particular S-S bond from forming. Disruption of the cystine knot disulfides results in the synthesis of nonfunctional proteins that are usually degraded intracellularly. Thus, a more detailed understanding of the functional components of cystine knots will help to understand how members of this emerging protein family fold and assemble into biologically active molecules.
The subunits of hCG are prototypes for the cystine knot growth factor family (1). Heterodimeric hCG forms when hCG-␤ assembles with the common glycoprotein hormone ␣-subunit (GPH-␣). GPH-␣ also assembles with luteinizing hormone ␤-subunit, thyroid-stimulating hormone ␤-subunit (TSH-␤), and follicle-stimulating hormone ␤-subunit to form luteinizing hormone, TSH, and follicle-stimulating hormone ␤-subunit, respectively (12). Luteinizing hormone-␤, TSH-␤, and follicle-stimulating hormone ␤ also contain the consensus residues required to form a cystine knot, but the actual presence of the knot has not yet been confirmed by structural studies.
Previously, we described the folding pathways of both hCG-␤ (9, 13-15) and GPH-␣ (6, 16) using S-S bond formation as an index of folding. The methods established in our laboratory for the folding of these two subunits now allow us to determine the importance of noncysteine residues of the cystine knot in hCG subunit folding, secretion, and assembly. Fig. 1 illustrates the location of the cystine knot motifs for both subunits of hCG, as well as the noncystine knot disulfides (17,18). The 8-residue ring of the cystine knot consists of four cysteines that form two S-S bridges, a Gly residue common to all 8-membered cystine knot rings, and three nonconserved residues (termed X 1 , X 2 , and X 3 ). Thus, the consensus sequences for this motif can be defined as C-X 1 -G-X 2 -C and C-X 3 -C (3,19).
The residues at the X 1 , X 2 , and X 3 positions vary among cystine knot-containing proteins and their functional importance is largely unknown. The sequence containing X 1 and X 2 in all four glycoprotein ␤-subunits is CAGYC. The equivalent sequence in GPH-␣ is CMGCC, where the Cys at the X 2 position forms a noncystine knot S-S bond with Cys 7 (Fig. 1). In this report, we investigated the contribution of hCG X 1 , X 2 , and X 3 residues in folding, secretion, and assembly by employing a chimeric strategy where the residues within the hCG-␤ and GPH-␣ cystine knots were interchanged individually or collectively. Furthermore, the role of the intervening Gly residue in both cystine knots was determined using various amino acid substitutions. We report that: (i) there is a subunit-specific complement of three "X" residues, all of which are needed for efficient assembly and (ii) the presence of the central invariant Gly is an absolute requirement for efficient folding, cystine knot formation, and hCG assembly.
Site-directed Mutagenesis-Mutations were made using a "megaprimer" polymerase chain reaction methodology (21) with Pfu polymerase (Stratagene). The GPH-␣ polymerase chain reaction products were cloned into pcDNA3 (Invitrogen) and the hCG-␤ polymerase chain reaction products were cloned into pGS (9). DNA sequencing confirmed incorporation of desired mutations. Plasmid DNA was purified using the Maxi Plasmid Kit (Qiagen) according to the manufacturer's protocol and used for transient transfection as described below.
Transient Transfection-293T cells (1.9 ϫ 10 6 ) were plated into 60-mm plastic dishes and grown overnight to 70 -80% confluency. Plasmid DNA was precipitated as described previously (6). For coexpression of hCG-␤ and GPH-␣, both plasmids were included in the precipitation. To obtain comparable expression levels in coexpression studies, a 40:1 GPH-␣ to hCG-␤ ratio of plasmid was used. The resulting precipitate was added dropwise to the dishes and agitated gently to mix. To ensure a uniform precipitate exposure, one large-scale precipitation was distributed equally among all dishes. Cells were incubated for 2 days at 37°C and used for metabolic labeling.
Metabolic Labeling with [ 35 S]Cysteine-Transiently transfected 293T cells were pulse-labeled for the times indicated in the text with L-[ 35 S]cysteine (ϳ1100 Ci/mmol; PerkinElmer Life Sciences), at a concentration of 100 -150 Ci/ml, in serum-free medium lacking cysteine (9). For experiments using dithiothreitol (DTT), the DTT was added with the [ 35 S]cysteine at a final concentration of 2.0 mM. Pulse incubations were carried out as described previously (13) and cells were incubated for the chase times indicated; the chase medium was saved for secretion studies. Cells were harvested by rinsing with cold phosphate-buffered saline and immediately lysed in 2.5 ml of phosphatebuffered saline containing detergents (1.0% Triton X-100, 0.5% sodium deoxycholate, and 0.1% SDS), protease inhibitors (20 mM EDTA and 2 mM phenylmethanesulfonyl fluoride), and 5 mM N-ethylmaleimide (NEM), pH 7.0, or 50 mM iodoacetate, pH 8.0, to trap free thiols in folding intermediates that contained unformed S-S bonds. NEM was used for GPH-␣ because it results in efficient alkylation of GPH-␣ thiols and better separation of folding intermediates by HPLC (6). Similarly, iodoacetate was used for hCG-␤ because it efficiently alkylates ␤-subunit thiols and facilitates mapping of hCG-␤ tryptic peptides (13,14). Cell lysates were incubated for 10 -20 min at room temperature in the dark, followed by disruption through a 22-gauge needle (5 times) and centrifuged for 1 h at 100,000 ϫ g. The collected chase medium was also clarified by centrifugation.
Immunoprecipitation of hCG Subunits from Cell Lysates and Chase Media-Immunoreactive forms of GPH-␣ or hCG-␤ were immunoprecipitated with polyclonal antibodies specific for each respective subunit (6,22). Immunoprecipitations were carried out at 4°C for 16 -20 h with rotation in the dark. Immune complexes were precipitated with protein A-Sepharose (Sigma) and prepared for SDS-polyacrylamide gel electrophoresis (PAGE) or reversed-phase HPLC as described below.

SDS-PAGE and Quantitation of [ 35 S]
Cysteine-labeled Subunits-Radiolabeled folding forms that adsorbed to protein A-Sepharose beads were prepared as described previously (13). Briefly, protein A-Sepharose beads containing immunopurified subunits were resuspended in twice concentrated SDS gel sample buffer (125 mM Tris-HCl, pH 6.8, containing 2% SDS, 20% glycerol, and 40 g/ml bromphenol blue). For reducing conditions, ␤-mercaptoethanol was included at a final concentration of 2%. Samples were boiled for 5 min, loaded on polyacrylamide gradient slab gels (5-20%), and run by the method of Laemmli (23). Gels were dried in vacuo on filter paper and exposed to a phosphorscreen (Molecular Dynamics). The phosphorscreen was scanned on a Molecular Dynamics Storm 860 and bands were quantitated using the Molecular Dynamics ImageQuant (version 5.0) volume report. To determine the percentage of GPH-␣ that had combined with hCG-␤ in Figs. 3 and 7, the following formula was used: [10/12 (amount of hCG-␤ that coimmunoprecipitated with GPH-␣) ϩ (amount of GPH-␣ in anti-␤ immunoprecipitation)]/[total GPH-␣]. For the ␣-C31Y mutant, 8/12 was used instead of 10/12 because this mutant has only 8 cysteine residues.
Reversed-phase HPLC Analysis-Radiolabeled folding intermediates that adsorbed to protein A-Sepharose beads were prepared as described previously (13). Briefly, protein A-Sepharose beads/antibody-antigen immunocomplexes were washed three times with phosphate-buffered saline containing detergents (1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS) followed by four washes with phosphate-buffered saline lacking detergents. Immunocomplexes were pelleted between washes by centrifugation for 1 min at 2000 ϫ g. To dissociate the Sepharose/antibody/antigen interactions, immunocomplexes were treated with 6 M guanidine HCl, pH 3.0 (sequenal grade; Pierce), for 16 -20 h while rotating at room temperature. 100 g of myoglobin was added as a carrier. The guanidine eluates were injected onto a Vydac 300-Å C 4 reversed-phase column equilibrated with 0.1% trifluoroacetic acid and eluted using an acetonitrile gradient as described previously (14). Fractions were collected in 1-min intervals and quantitated by scintillation counting. Samples were stored at Ϫ20°C until further characterization.
Tryptic Digestion and Reversed-phase HPLC Analysis of Tryptic Pep-FIG. 1. Schematic diagram of hCG subunits. GPH-␣ and hCG-␤ each contain 3 loops (labeled L1, L2, and L3) and are arranged in a head-to-tail orientation in the hCG dimer (17,18). Both subunits also contain a cystine knot formed by three S-S bonds (displayed as lines connecting the solid circles; numbers indicate the cysteine residue number). Noncystine knot S-S bonds are represented by the lines connecting open circles. The noncysteine residues of the cystine knots are labeled X 1 , X 2 , and X 3 . A particularly interesting feature of the hCG dimer is the C-terminal region of hCG-␤ that wraps around the GPH-␣ L2, creating the so-called "seat-belt." tides-HPLC fractions from C 4 reversed-phase HPLC representing hCG-␤ folding intermediates p␤1 or p␤2 were pooled, concentrated, and digested for 16 -20 h in silanized polypropylene tubes containing 100 -200 g of myoglobin, 0.03% trypsin (Sigma), 5 mM CaCl 2 , and 50 -150 mM Tris-HCl, pH 8. Digestions were continued for 2 h with two additional aliquots of 25 g of trypsin (0.06% final concentration) (13,14). hCG-␤ tryptic peptides were separated on a Vydac C 18 reversed-phase column as described previously (14). Amino acid sequencing was used previously to identity the peptide(s) in each peak (14).
Amino Acid Analysis Procedure to Determine S-S Bond Content-A modified protocol, similar to the one used in determining the S-S folding pathways for potato carboxypeptidase inhibitor and human epidermal growth factor (24,25) was used. [ 35 S]Cysteine-containing folding forms isolated from reversed-phase HPLC were dried in vacuo and hydrolyzed as described previously (6). Quantitation of [ 35 S]cystine and succinyl-[ 35 S]cysteine (hydrolysis product of NEM-Cys) was performed using a modification of the method described by Cohen and Michaud (26). Hydrolysates were resuspended in 10 l of 10 mM HCl. To this, 70 l of 0.2 N borate buffer, pH 8.8, was added. Derivatization of amino acids was performed by adding 30 l of 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate (8 mg/ml in anhydrous acetonitrile). Samples were dried in vacuo. Before injection, samples were resuspended in 110 l of buffer A (140 mM sodium acetate, 17 mM triethylamine, pH 4.9). Derivatized amino acids were separated by HPLC as described previously (6)

Swapping of hCG-␤ and GPH-␣ Cystine Knot Motifs-
The only conserved sequences in cystine knot-containing proteins that have an 8-membered ring are the C-X 1 -G-X 2 -C and C-X 3 -C sequences (3,19). The contribution that the X residues make toward attaining a native conformation is unknown. Furthermore, it is not known whether cystine knot sequences are protein-specific, or whether the residues of the knot motif are functionally interchangeable. To address this, we used a chimeric strategy wherein cystine knot residues of GPH-␣ and hCG-␤ were swapped singly or collectively.
Swapping of hCG cystine knot motifs was accomplished by site-directed mutagenesis at the X 1 , X 2 , and X 3 positions to match the residue(s) of the other subunit. Three single GPH-␣ mutants (␣-M29A, ␣-C31Y, and ␣-H83Q) and a GPH-␣ mutant containing all three mutations (termed ␣ ␤ knot ) were constructed (Table I). Thus, the amino acid sequence of the ␣ ␤ knot cystine knot matched that of hCG-␤. Cys 7 (which normally pairs with Cys 31 ) was also changed to Ala in the ␣-C31Y mutant to prevent a free thiol at Cys 7 from causing S-S rearrangements or retention of the subunit in the endoplasmic reticulum (27). Importantly, removal of S-S bond 7-31 does not affect GPH-␣ folding, secretion, or hCG biological activity (6,7). In addition, the hCG-␤ cystine knot residues were replaced with those of GPH-␣ to give ␤-A35M, ␤-Y37A, ␤-Q89H single mutants and a ␤ ␣ knot triple mutant (Table I). ␤-Y37 was replaced with Ala instead of Cys to avoid introduction of a free thiol for the reasons alluded to above. Ala was selected because its side chain properties closely resemble those of Cys, and because Ala and Ser substitutions have similar effects on hCG-␤ folding and assembly when they are substituted for Cys (8).
The six single mutants (three GPH-␣ and three hCG-␤), two triple mutants (␣ ␤ knot and ␤ ␣ knot ), and wild type (WT) subunits were analyzed for proper folding as described previously (6,13,14,16). Briefly, GPH-␣ folding was monitored using S-S bond formation and HPLC elution times (6, 16), while hCG-␤ folding was monitored by shifts in migration on nonreducing SDS-PAGE (14,22). No significant differences in folding were observed for any of the mutants (data not shown), suggesting that these cystine knots may be interchangeable. To test this further, we determined the efficiency of subunit secretion. Secretion was assayed by pulse labeling transiently transfected 293T cells for 10 min, followed by a 10-min or 8-h chase. The immunopurified cell lysates and media were analyzed by SDS-PAGE ( Fig. 2A), and the bands were quantitated as described under "Experimental Procedures." The percent secretion for the GPH-␣ and hCG-␤ chimeras are shown in Fig. 2, B and C. Consistent with previous studies, about 80% of WT GPH-␣ and hCG-␤ were secreted by 8 h. Furthermore, swapping of single residues or the entire cystine knot did not significantly affect subunit secretion of GPH-␣ or hCG-␤ (Fig. 2). This efficient secretion is another indicator that these subunits folded to a native or native-like conformation, since misfolded or incompletely folded hCG subunits are generally retained intracellularly and degraded (5,6).
Finally, we assayed for the ability of the GPH-␣ cystine knot chimeras to assemble with ␤-WT and, conversely, for the ability of hCG-␤ cystine knot chimeras to assemble with ␣-WT. To do this, 293T cells coexpressing both subunits were pulse-labeled and chased for 8 h. Unassembled GPH-␣ and intact hCG ␣/␤ dimer were first precipitated from the collected medium with a polyclonal ␣-antiserum. This step was followed by a second precipitation with a polyclonal ␤-antiserum to recover any excess, unassembled hCG-␤. There was a dramatic decrease in the ability of ␣ ␤ knot and ␤ ␣ knot to combine with ␤-WT and ␣-WT, respectively (Fig. 3, A and B). Furthermore, this decreased combination was not due to altering any one residue, as all of the single mutations were significantly less deleterious than the triple mutations. However, the single mutations at the X 2 position (␣-C31Y and ␤-Y37A) did decrease assembly to a level intermediate to that of WT and the triple mutants. Co-transfection of ␣ ␤ knot and ␤ ␣ knot together resulted in Ͻ5% assembly (data not shown). Taken together, these data indicate TABLE I GPH-␣ and hCG-␤ cystine knot chimeras Three single mutants and a triple mutant were constructed for GPH-␣ such that the residues at the X 1 , X 2 , and X 3 positions matched that of hCG-␤. Additionally, three single mutants and a triple mutant were constructed for hCG-␤ such that the residues at the X 1 , X 2 , and X 3 positions matched that of GPH-␣. The residues within the boxes represent those that were changed by site-directed mutagenesis. a Residues 28 -32 in GPH-␣. b Residues 82-84 in GPH-␣. c Residues 34 -38 in hCG-␤. d Residues 88 -90 in hCG-␤. e Cys 7 , which pairs with Cys 31 to form a S-S bond, was also converted to Ala so that no unpaired thiols remained. Removal of S-S bond 7-31 does not affect GPH-␣ folding, secretion, combination with hCG-␤ or hCG bioactivity (6,7). that the noncysteine residues within the GPH-␣ and hCG-␤ cystine knots contribute to an assembly-competent conformation in a subunit-specific manner. Furthermore, this contribution appears to be a function of the set of all noncysteine residues, as opposed to arising from the contribution of a single residue.
The Central Gly Residue in the hCG-␤ Cystine Knot Is Critical for Folding, Secretion, and Assembly-As alluded to above, the central Gly residue in hCG-␤ is conserved among all known cystine knots that contain an 8-membered ring structure (1, 2). A mutation resulting in the conversion of the central Gly residue to Arg in TSH-␤ causes congenital isolated TSH deficiency, wherein the mutant TSH-␤ fails to be secreted and assemble with GPH-␣ (28). This demonstrates that the central Gly of the CAGYC region is critical for TSH activity and suggests that it may also be essential to the function of the other glycoprotein hormones.
Knowledge of the hCG-␤ folding pathway (13, 14) provides a novel system to determine the contribution that residues of the cystine knot make in attaining an assembly-competent conformation. To determine whether substitution of the Gly of the CAGYC region of hCG-␤ alters folding, secretion, and/or assembly, we created and analyzed three Gly mutants: ␤-G36A, ␤-G36N, and ␤-G36R. Mutation to Arg was chosen because this mutation is observed in the naturally occurring TSH-␤ mutant (28), G36N was chosen because Asn has a smaller neutral side chain in comparison with the positively charged Arg, and G36A was chosen because it is the most conservative change possible; however, Ala in most cases can adopt the required positive torsion angle only under conditions of unfavorable steric hindrance (29). 293T cells expressing ␤-WT, ␤-G36N, ␤-G36R, or ␤-G36A were pulse-labeled with [ 35 S]cysteine and chased for 0, 5, 15, 30, 60, 120, or 480 min. Fig. 4A shows the progression of ␤-WT from p␤1 (the earliest detectable folding intermediate) to p␤2 to mature ␤. At chase times Ն60 min, ␤-WT was detectable in the media as mature, secreted ␤. Fig. 4B shows that most of ␤-G36A did not progress beyond p␤1 and was not secreted. However, a p␤2-like species of ␤-G36A was isolated by reversed-phase HPLC (Fig. 4C). This species was termed "p␤ 2-like" since it eluted from HPLC at a similar time to that of WT p␤2 (13,14). Folding and secretion data for ␤-G36N and ␤-G36R subunits yielded similar results to those shown in Fig.  4 (not shown). The amount of radioactivity in this band was used to calculate the percentage of protein that was secreted into the medium (M) after 8 h. By 8 h, Ͻ5% of total protein remaining intracellular (data not shown), indicating that secretion of the radiolabeled protein was essentially complete. SDS-PAGE analysis of the six single mutants was comparable to WT and the triple mutants (gels not shown). The slower mobility of the secreted subunits is due to the addition of O-linked oligosaccharides (38) and/or differential processing of N-linked oligosaccharides (39). B and C, quantitation of the percentage secreted for the respective hCG-␤ and GPH-␣ WT subunits (solid bars), composite chimeras (open bars), and single mutants (shaded bars). The percentage secreted is defined as the amount of 35 S-labeled subunit recovered in the medium after 8 h relative to that present at 10 min intracellular. Each bar represents the mean Ϯ S.D. of at least three experiments.

FIG. 3. Assembly of hCG-␤ and GPH-␣ cystine knot chimeras.
293T cells expressing WT GPH-␣ and hCG-␤ (WT or cystine knot chimeras) (panel A) or WT hCG-␤ and GPH-␣ (WT or cystine knot chimeras) (panel B) were pulse-labeled with [ 35 S]cysteine and chased for 8 h. The medium was collected and immunoprecipitated with polyclonal antibody to GPH-␣. This immunoprecipitation pulls down unassembled and assembled GPH-␣ and co-precipitates hCG-␤ combined with GPH-␣. Shown above each bar is a representative SDS-PAGE gel. The remaining unassembled hCG-␤ was immunoprecipitated with anti-␤ to ensure that there was excess hCG-␤ such that it was not a limiting factor in assembly (data not shown). In some cases, a portion of the total GPH-␣ was present in the anti-␤ immunoprecipitation and was included in calculating the percentage of combined GPH-␣. Quantitation was performed as described under "Experimental Procedures" and is shown in the bar graphs for WT subunits (solid bars), composite chimeras (open bars), and single mutants (shaded bars). In both A and B, the results are displayed as the percentage of total secreted GPH-␣ that had dimerized with hCG-␤. Each bar represents the mean Ϯ S.D. of at least three experiments.
Tryptic digestion of fully folded native hCG-␤ (i.e. all native S-S bonds formed) produces [ 35 S]cysteine-labeled peptides, all of which are linked by S-S bonds (13,14). If a particular S-S bond has not yet formed in a given intermediate, then digestion with trypsin results in the release of a [ 35 S]cysteine-containing peptide from the S-S-linked core material (13,14). Thus, HPLC separation of released peptides from the S-S-linked peptides identifies the S-S bonds of a given hCG-␤ folding intermediate that are unformed. Fig. 5, A and C, show the respective reversed-phase HPLC profiles of trypsin-treated WT hCG-␤ that eluted from HPLC at the positions of p␤1 and p␤2. As defined in previous studies (13,14), the release of peptides 9 -20, 69 -74, 87-94, 96 -104, and 105-114 from G36A p␤1 (Fig. 5B) indicates that Cys 10 , Cys 72 , Cys 90 , Cys 93 , Cys 100 , and Cys 110 , respectively, were not part of a S-S bond. Additionally, unidentified peaks were present (labeled with an asterisk in Fig. 5, B  and D), which suggest the presence of non-native S-S bonds. Peptides 9 -20 and 87-94 were not present in the WT hCG-␤ p␤2 tryptic profile (Fig. 5C), consistent with our previous report that the cysteines in these peptides are involved in S-S bonds in the p␤2 intermediate (14). The tryptic profile of ␤-WT p␤2 clearly differed from that of ␤-G36A p␤2-like (Fig. 5, compare C and D), suggesting that non-native S-S bonds had formed in ␤-G36A. Furthermore, unlike WT-␤, the 87-94 and 9 -20 peptides (Fig. 5D, peaks 3 and 4, respectively) were recovered in G36A p␤2-like digested material, indicating that S-S bonds 38 -90 and 9 -57 had not formed completely. The tryptic map of G36A p␤2-like material was similar to G36A p␤1 (Fig. 5, compare B and D), demonstrating that the material that eluted from HPLC at the p␤2 locus (i.e. p␤2-like) more closely resembled p␤1 than the more folded p␤2. Tryptic maps of ␤-G36N and ␤-G36R revealed that non-native S-S bonds had similarly formed (data not shown).
Taken together, these data indicate that the central Gly of the hCG-␤ CAGYC region is critical for the proper formation of S-S bonds and thus, is important for its folding and secretion. This also implies that the Gly 3 Arg mutation observed in congenital isolated TSH deficiency (28) results from improper folding of TSH-␤, which prevents its assembly with GPH-␣.

Mutation of the Invariant GPH-␣ Cystine Knot Gly Residue-
The previous section demonstrated that the Gly residue in the CAGYC sequence of hCG-␤ is critical for proper folding. This implies that this Gly may be critical for the folding of other cystine knot-containing proteins as well. To test this, we made the equivalent G30A mutation in the CMGCC sequence of GPH-␣.
Unlike hCG-␤ folding intermediates, GPH-␣ intermediates do not migrate differently on SDS-PAGE (6). However, GPH-␣ folding can be monitored by changes in reversed-phase HPLC elution times; unfolded GPH-␣ containing no S-S bonds and being more hydrophobic, elutes later than the native conformation and folding intermediates (6,16). Moreover, HPLC elution position correlates with the formation of S-S bonds as GPH-␣ folds to its less hydrophobic conformation (i.e. earlier eluting species contain more S-S bonds than the later eluting, less folded species) (6,16). Shown in Fig. 6 are the HPLC profiles generated for ␣-G30A after a 10-min pulse in the presence of DTT, followed by 0-, 5-, and 30-min chases after DTT removal. DTT was used in the pulse to delay the formation of S-S bonds until its removal during the chase (6). ␣-G30A folding did not generate a species that eluted at the position of native ␣-WT (Fig. 6, vertical dotted line). Additionally, a late eluting peak Only a small portion of ␤-G36A converted from p␤1 to p␤2. Furthermore, ␤-G36A was not secreted, as evidenced by the lack of bands present in the 30 -480 min secreted material. The lack of any significant radiolabeled intracellular or secreted ␤-G36A at 480 min indicates that ␤-G36A was almost entirely degraded by this time. C, C 4 reversedphase HPLC of ␤-G36A showing the presence of a p␤2-like material that eluted at 30 -40 min. This peak was designated as p␤2-like because it elutes at a similar position to that of WT p␤2 (14). The left-most lane in A and B contain carbonic anhydrase (M r ϭ 29,000).
FIG. 5. Tryptic analysis of WT and ␤-G36A p␤1 and p␤2 folding intermediates. WT and ␤-G36A p␤1 and p␤2 species were isolated from reversed-phase HPLC and fractions representing each species were pooled and concentrated in vacuo. The isolated species were digested with trypsin to release peptides not linked by S-S bonds from the otherwise S-S linked core. The resulting mixture of peptides and S-S linked peptides were separated by C 18 reversed-phase HPLC. A, tryptic map of WT p␤1. B, tryptic map of ␤-G36A p␤1. C, tryptic map of WT p␤2. D, tryptic map of ␤-G36A p␤2-like. The identities of the peaks (13,14) are: peak 1a and 1b, peptide 96 -104; peak 2, peptides 69 -74 and 105-114; peak 3, peptide 87-94; peak 4, peptide 9 -20; peak 5, S-Slinked peptides. Peaks marked with an asterisk are peaks that appear to represent S-S-linked peptides connected by non-native S-S bonds (13,14). (Fig. 6, peak *) was observed at all chase times. This peak migrated at a relative molecular weight of about twice that of ␣-WT, G30A-␣ 1 , and G30A-␣ 2 when analyzed by nonreducing SDS-PAGE (data not shown). This suggests that some of ␣-G30A had formed a homodimer. In contrast, reduction of S-S bonds before SDS-PAGE resulted in migration at the molecular weight of monomeric GPH-␣. These data suggest that a significant proportion of ␣-G30A forms dimers that arise from nonnative, intermolecular S-S bonds.
Amino acid analysis of ␣-G30A ␣ 1 and ␣ 2 was used to quantitate the number of S-S bonds that had formed in each species. Recovery of 97% of the 35 S label as cystine and 3% as succinylcysteine (hydrolysis product of NEM-alkylated cysteine) indicates that ␣ 2 contained five intact S-S bonds (Table II). As expected, the later eluting species, ␣ 1 , contained less than five S-S bonds (3.7 out of 5). Although it is not obvious whether the bonds present in these two species are native, the differences in HPLC elution times compared with native WT GPH-␣ (Fig. 6) nonetheless implies that these folding forms have non-native conformations.
To determine the efficiency of secretion for ␣-G30A, 293T cells expressing ␣-G30A or ␣-WT were pulse-labeled for 10 min with [ 35 S]cysteine and chased for 10 min or 8 h. The immunopurified cell lysates and media containing GPH-␣ were analyzed by reducing SDS-PAGE and bands quantitated as described under "Experimental Procedures." About 80% of ␣-WT present at 10 min was secreted into the medium by 8 h (Fig.  7A). In contrast, only about 40% of ␣-G30A was secreted (Fig.  7A). Furthermore, Ͻ10% of ␣-G30A remained in the cell after 8 h (data not shown), indicating that about 50% had been degraded. This is consistent with our previous reports (6,16) demonstrating that mutant forms of GPH-␣ that do not migrate at the position of ␣-WT on HPLC (i.e. they contain a non-native conformation) are readily degraded.
To determine whether ␣-G30A could assemble with hCG-␤, both subunits were coexpressed in 293T cells, pulse-labeled with [ 35 S]cysteine for 20 min, and chased for 8 h. Immunopurified subunits were analyzed by SDS-PAGE, the bands were quantitated, and the percentage of GPH-␣ that had combined with hCG-␤ was calculated (see "Experimental Procedures"). Results of this analysis are shown in Fig. 7B. Only 30% of the secreted ␣-G30A combined with WT hCG-␤, compared with 80% for ␣-WT. These results further demonstrate the importance of the central cystine knot Gly residue in protein folding and heterodimer formation.

DISCUSSION
The growth factor cystine knot superfamily of dimeric proteins contain similar structural topologies but lack significant sequence similarity other than the spacing of the six cysteine residues that form the three cystine residues of the knot (1). The importance of the cystine residues in producing functional proteins has been well documented (6, 7, 9 -11, 16), but the role of the noncysteine residues located within these cystine knots is largely unknown.
All known growth factors containing a cystine knot motif have an 8-amino acid ring structure, with the exception of nerve growth factor, which contains a 14-membered ring. For the 8-residue rings, such as those found in GPH-␣ and hCG-␤, the two sides of the ring contain 5-and 3-residues and are linked by two S-S bridges. Other than the cysteine residues, the only conserved residue of the ring is a Gly located at the central position of the 5-residue stretch, such that this sequence is termed C-X 1 -G-X 2 -C. Studies showing that this Gly residue is essential for producing functional TSH (28,30) have been repeated in hCG-␤ using identical (Gly 3 Arg) and similar (Gly The immunoprecipitated folding forms were separated by reversedphase HPLC and the amount of [ 35 S]cysteine-labeled ␣-G30A in each 1-min fraction was quantitated. Reducing and nonreducing SDS-PAGE were used to demonstrate that the peak marked with an asterisk (*) represents ␣-G30A aggregates that were mostly S-S linked homodimers with some multimeric species (data not shown). The unfolded monomeric ␣-G30A eluted at 57-58 min (0 min chase, top panel) and converted to two folding forms termed ␣ 1 and ␣ 2 that contained 3.8 and 4.8 S-S bonds, respectively. The vertical dotted line represents the elution position of native WT GPH-␣ (6).

TABLE II
Quantitation of S-S content in ␣-G30A folding intermediates Amino acid analysis was used to determine the relative amount of S-S bond formation as described under "Experimental Procedures." The folding forms listed below correspond to the intermediates referred to on the HPLC chromatograms in Fig. 6. Shown are the calculated percentages of S-S bond formation (S-S), unformed S-S bonds (NEM-Cys), and the calculated number of S-S bonds formed. Native ␣-WT represents the fully folded, assembly-competent conformation (6). c Data from previous report (6).
3 Asn) mutations (31,32). However, a lack of structural data and knowledge of the folding mechanisms for these subunits at the time of these observations failed to define why this Gly is critical. In light of more recent findings, including the hCG crystal structure (17,18) and knowledge of GPH-␣ and hCG-␤ folding (6,9,14,16), we can now address the mechanism by which specific residues within the hCG cystine knots contribute to hormone function. In particular, results from mutational analyses can be more precisely interpreted because we can distinguish between two general consequences of these mutations: (i) the mutation removes a key residue important for a direct subunit interaction; or (ii) the mutation causes global misfolding such that the protein cannot attain native structure. The central Gly residue located between X 1 and X 2 is thought to be necessary because, in contrast with other amino acids, Gly can readily adopt a positive torsion angle, which allows it to avoid steric hindrance with the penetrating S-S bond of the cystine knot (1). The biological importance of this Gly can be inferred from several observations. First, a naturally occurring Gly 3 Arg mutation in TSH-␤ causes congenital isolated TSH deficiency (28). Second, mutation of the equivalent Gly in GPH-␣ prevents the production of functional hCG (31) and TSH (30). Third, mutation of this Gly to Arg or Asp in hCG-␤ results in undetectable levels of heterodimeric hCG being secreted from Xenopus laevis oocytes (32).
In this report, we investigated the role of the invariant cystine knot Gly residue in the folding, secretion, and assembly of GPH-␣ and hCG-␤, two prototypes of the growth factor cystine knot superfamily (1). Even the most conservative substitution possible, Gly 3 Ala, resulted in nearly 100% of hCG-␤ being misfolded and degraded intracellularly. The misfolding was evident from the non-native S-S bonds that had formed (Fig. 5), as well as the failure to efficiently convert to the p␤2 folding intermediate (Fig. 4B). The resulting non-native S-S bond formation and misfolding provides an explanation for why the central Gly residue is essential for hCG function.
Mutation of the equivalent Gly in GPH-␣ (␣-G30A) gave similar results, although, the deleterious effects were less pronounced; 40% of ␣-G30A was secreted, 30% of the which assembled with WT hCG-␤. This result is consistent with a study that detected immunoreactive hCG when ␣-G30A and WT hCG-␤ were coexpressed in X. laevis oocytes (31). However, the mutant hCG heterodimer displayed no bioactivity in a murine testosterone production-based Leydig cells bioassay, suggesting that native hCG conformation was not attained (31).
A notable effect of the G30A mutation on GPH-␣ folding was that it slowed folding significantly. Following a 30-min chase, less than half of the ␣-G30A synthesized converted to the most folded form that contained five S-S bonds. This compares with a t1 ⁄2 of about 90 s for WT GPH-␣ folding. Previously, we reported that disruption of the GPH-␣ cystine knot S-S bonds results in inefficient folding and secretion (6). The observation that ␣-G30A was inefficiently folded and secreted implies that the ␣-G30A mutation also interfered with formation of the GPH-␣ cystine knot.
There are several possible explanations for why mutation of the invariant Gly was more detrimental to hCG-␤ folding than that of GPH-␣. First, the inherent flexibility of the GPH-␣ loop 2 (residues 33-58) (33) may allow for greater perturbation of the GPH-␣ cystine knot and permit closure of the cystine knot ring in a portion of the molecules by adopting the positive torsion angle needed at residue 30. Second, WT hCG-␤ folds at a much slower rate compared with GPH-␣ (t1 ⁄2 ϭ 5 min versus t1 ⁄2 ϭ 90 s, respectively) and, therefore, when the rate of hCG-␤ folding is slowed even further (as noted above for ␣-G30A), ␤-G36A may be more readily degraded before the subunit has time to fold. The latter possibility is based upon a recently proposed model (34,35) that suggests that glycoproteins only have a limited amount of time to fold and exit the endoplasmic reticulum before being degraded.
The other three noncysteine residues of the hCG cystine knots are X 1 , X 2 , and X 3 of the C-X 1 -G-X 2 -C and C-X 3 -C sequences. To understand the role(s) of these residues, we constructed chimeras in which the X residues of GPH-␣ and hCG-␤ were swapped individually or in combination, while leaving the central Gly unchanged. Folding and subunit secretion was not significantly affected in either the single or the triple mutants (␣ ␤ knot and ␤ ␣ knot ) (Fig. 2). However, assembly of ␣ ␤ knot and ␤ ␣ knot with their WT partners was decreased by about 90% (Fig. 3). The decrease in assembly was not due to any one particular substitution at the X 1 , X 2 , or X 3 positions but, rather, was due to the combination of all three substitutions, suggesting that the set of all three X residues act in a subunitspecific manner.
The observation that ␣ ␤ knot and ␤ ␣ knot are efficiently secreted but do not assemble adds to a growing body of evidence that suggests that determinants necessary for assembly and secretion of hCG subunits are different. Recently, we reported two examples of modified GPH-␣ subunits that are efficiently secreted but do not assemble with hCG-␤ (6); one modification changed residues in loop 2 necessary for combination while the other modification simultaneously removed both 7-31 and 59 -87 S-S bonds. In addition, hCG-␤ mutants lacking the 93-100 or 26 -110 S-S bonds are also efficiently secreted but do not assemble with GPH-␣ (5, 9). Thus, structural determinants needed for hCG assembly are not necessarily required for subunit secretion. ]Cysteine-labeled GPH-␣ was immunoprecipitated from the 10-min cell lysate (representing intracellular GPH-␣) and 8 h medium (representing secreted GPH-␣). The immunopurified subunits were analyzed by reducing SDS-PAGE and quantitated. The percentage secreted is defined as the amount of 35 S-labeled subunit recovered in the medium after 8 h relative to that present at 10 min intracellular. Following an 8-h chase, Ͻ5% of the total GPH-␣ remained in the cell lysate (data not shown), indicating that by 8 h the majority of the radiolabeled subunit had either been secreted or degraded intracellularly. B, 293T cells expressing ␣-WT or ␣-G30A and hCG-␤ were pulselabeled with [ 35 S]cysteine and chased for 8 h. The percentage of GPH-␣ that had combined with WT hCG-␤ was determined as described in the legend to Fig. 3. Each bar represents the mean Ϯ S.D. of at least three experiments.
Our data suggest that the noncysteine residues within the hCG-␤ and GPH-␣ cystine knots are critical for intersubunit interactions that are necessary to form a stable dimer. This finding is further supported by an important feature of the hCG structure (17,18). At the core of the dimer interface is a series of interchain ␤-sheets. Intimately involved in this ␤-sheet are the regions encompassing the C-X 1 -G-X 2 -C sequences of both subunits (residues 25-39 of GPH-␣ and 27-40 of hCG-␤), which form a significant number of intersubunit hydrogen bonds. Thus, the simultaneous alteration of multiple X residues, as was done in ␣ ␤ knot and ␤ ␣ knot , could interfere with formation of this intersubunit ␤-sheet, whereas single changes might be less disruptive because most other intersubunit interactions remain intact.
The set of all three X residues within both hCG cystine knots are required for biological activity since they are necessary for dimer formation (Fig. 3) and only the hCG heterodimer is functional (12). Whether or not this region is important for receptor binding and signal transduction is unknown. A singlechain model that tethers assembly-incompetent subunits to their WT partner has been used successfully to address similar questions. GPH-␣ subunits containing cysteine mutations that disrupt the cystine knot, such that the free subunits alone cannot assemble with hCG-␤, maintain in vitro biological activity when tethered to hCG-␤ (36,37). These data suggest that the cystine knot region is necessary for heterodimer formation, but not for receptor binding and signal transduction. Thus, it seems likely that the intervening X residues may also not be directly involved in receptor binding and signal transduction.
A complete understanding of the common cystine knot motif must take into account all residues of the knot. Previous studies have focused primarily on the S-S bonds and some aspects of the central Gly residue (6, 8, 9, 28, 30 -32). The data presented in this report provides evidence that the intervening X residues located between the cystine knot also make important contributions to hCG biosynthesis. Specifically, these residues appear to play a critical role in dimer formation rather than directly influencing individual subunit folding or secretion. Thus, most noncysteine residues within cystine knots may not be interchangeable because they appear to have subunit-specific functions. Future studies aimed at elucidating the role of analogous residues in other cystine knot proteins may determine how universal a role these residues play in the function of other members of the cystine knot superfamily.