Reproduction and thyroid function are controlled by the glycoprotein hormones: heterodimeric proteins composed of an α-subunit common to all and a hormone-specific β-subunit (Pierce and Parsons 1981). The free subunits are essentially inactive but can be recombined to form a heterodimer having the activity of the hormone from which the β-subunit is derived (Pierce and Parsons 1981). The tertiary structure of deglycosylated hCG (Lapthorn et al. 1994; Wu et al. 1994) showed that both subunits have a similar topology. Three of the disulfides in each subunit form a cystine knot core that results in an elongated protein containing two paired loops at one end (loops 1 and 3) and a single loop at the other. In the heterodimer, the subunits are oriented in an antiparallel fashion with the single loop of each subunit located adjacent to the paired loops of the other subunit. The β-subunit contains 20 additional amino acids that are wrapped around α-subunit loop 2 like a seatbelt and latched to β-subunit loop 1 by a disulfide (Lapthorn et al. 1994).
The seatbelt makes an obvious contribution to the stability of the heterodimer, yet its location would appear to impede subunit combination. If the seatbelt remained latched, heterodimer formation would occur only if α2 and its associated oligosaccharide are threaded through the small opening between the seatbelt and β1. This requirement would be bypassed if the seatbelt were to open before subunit combination. Protein disulfide isomerase (Huth et al. 1993) and small amounts of reducing agents (Slaughter et al. 1995) accelerate subunit combination in vitro, possibly by breaking the seatbelt latch disulfide. While it might be expected that the seatbelt latch is stable during subunit combination in the absence of these agents, the observation that glycoprotein hormones have an intrinsic thioredoxin activity (Boniface and Reichert 1990; Chew et al. 1995) raised the possibility that docking of their subunits leads to a transient break in the latch disulfide needed for the seatbelt to wrap itself around α2. Following this process, the terminal cysteine of the seatbelt would be in a position to reestablish the seatbelt disulfide, thereby stabilizing the heterodimer. Ruddon et al. (1996) have suggested that the hCG subunits combine before the seatbelt is latched during hCG synthesis in choriocarcinoma and transfected CHO cells. Thus, there is precedent for the possibility that the unlatched seatbelt is wrapped around α2 during the combination process.
Glycoprotein hormone synthesis remains poorly understood in spite of its potential importance as a source of hCG metabolites observed during ectopic pregnancy, Down's syndrome, and even some cancers (Hussa 1987; Cardwell et al. 1997; Cole et al. 1997; Lee et al. 1997). Therefore, we have begun to study this process in greater detail. As described here, we found that Cys 110 was not labeled with alkylating agents during subunit combination, a process that occurs efficiently in oxidizing conditions in vitro. This observation suggested that the unlatching of the seatbelt is not required for subunit combination. Further, none of the cysteines in the β-subunit were alkylated during subunit combination in the presence of a vast molar excess of several alkylating agents. This suggests that breaking of disulfides is not required for subunit combination; this led us to devise a model of heterodimer formation in which movements of the seatbelt and α2 allow this loop and its associated oligosaccharide to pass through the small hole between the seatbelt and the remainder of the β-subunit.
Results and discussion
The seatbelt latch disulfide between Cys 26 and Cys 110 has been shown to be essential for the formation and/or stability of the hCG heterodimer (Suganuma et al. 1989). If the Cys 26–Cys 110 disulfide opened during subunit combination, we anticipated that alkylation of either cysteine would prevent the seatbelt from latching, thereby blocking stable subunit combination. To test this, we monitored subunit combination in the presence of 100 mM VP, 50 mM NEM, 10 mM MB, and 50 and 600 mM IA. We also tested combination in the presence of 1 and 10 mM oxidized glutathione or cystamine. In the absence of alkylating agents, subunit combination was nearly complete within 2 h as analyzed by Western blotting (Fig. 1) The kinetics of subunit combination were similar to those observed by Ingham et al. (Aloj et al. 1973; Ingham et al. 1973) and by Strickland and Puett (1982). With the exception of IA, which inhibited subunit combination somewhat (Table 1), none of the alkylating agents disturbed heterodimer formation, even when present at 10–1000-fold molar excess (Table 1; Fig. 2) Combination was not influenced by glutathione or cystamine, agents that would be expected to trap disulfide bonds that had opened during the combination process (Table 2).
Conceivably, subunit combination requires disruption of disulfides other than the seatbelt latch.
Disruption of the disulfide between the tenth (Cys 93) and eleventh (Cys 100) cysteines would increase the space between the seatbelt and the β subunit core. This disulfide, which stabilizes a small loop within the seatbelt, is also known to be reduced readily (Pierce and Parsons, 1981). We monitored the alkylation of specific cysteines using MALDI-TOF.
Following subunit combination in the presence of IA or VP, we reduced the heterodimer with 100 mM DTT: a treatment sufficient to cause complete reduction of both subunits (Pollak et al. 1990). The free thiols produced were reacted with an alkylating agent that had a different molecular weight than that used during subunit combination, enabling us to distinguish fragments that had been alkylated before and after reduction. The protein was then digested with N-glycanase to remove the oligosaccharides and with trypsin to produce fragments suitable for MALDI-TOF analysis. We were unable to detect all the expected tryptic fragments of hCG using MALDI-TOF, a phenomenon that is not unique to hCG and may have been related to differences in the abilities of the tryptic peptides to be desorbed from the matrix (Landry et al. 2000). In general, hCG peptides containing cysteines that had been alkylated by VP were much more readily detected than those alkylated by IA. As summarized in Figure 4, we were able to characterize at least one cysteine from all but two of the 11 disulfides in hCG following reduction, labeling with either IA or VP, and trypsin digestion. The peptide that contained Cys 100 was observed by MALDI-TOF when this cysteine was alkylated by VP but not by IA. Thus, we were able to monitor alkylation of this cysteine only with VP. Neither Cys 93 nor Cys 100 became alkylated during combination in the presence of 100mM VP (Fig. 3, peak 8). Cys 93 was not alkylated during combination in the presence of 100 mM IA (data not shown). Presumably, Cys 100 was also not alkylated under this condition, but it was not possible to determine this because of the fact that we were unable to observe this peptide when it was labeled with IA. None of the other cysteines detected by mass spectrometry was found to be alkylated by either IA or VP in any trypsin fragment of either subunit (Fig. 3). These observations suggested that within the limits of the technique, no disulfide bonds had been broken during subunit combination in oxidizing conditions.
We had expected that the oligosaccharide on α-subunit loop 2 would prevent subunit combination because of its bulk (Weisshaar et al. 1991) and that dimer formation could not occur without disruption of the seatbelt latch disulfide. The observations just described indicated that this notion was incorrect. We were concerned that the methods used might not have been sufficient to detect transient disruption of the seatbelt latch or other disulfide, and we examined this possibility. First, we tested the abilities of massive concentrations of IA to inhibit subunit combination. At 600 mM, significant amounts of IA were incorporated into the tyrosine residue of Ala 36–Tyr-Pro-Thr-Pro-Leu–Arg 42, a peptide derived from α-subunit loop 2, and the methionine residue of Val 68–Thr-Val-Met-Gly-Gly-Phe–Lys 75, a peptide derived from α-subunit loop 3, even though each lacks cysteine (Fig. 5). IA was not incorporated into Cys 93 or Cys 110, β-subunit residues that should have been alkylated more readily than either of these α-subunit peptides, had either the Cys 93–Cys 100 and Cys 26–Cys 110 disulfides been reduced. Residues in α-subunit loop 2 near its interface with the β-subunit have been shown to reduce subunit combination (Bielinska and Boime 1992a, b, Grossmann et al. 1996). Modification of Tyr 37 would explain the ability of massive amounts of IA to reduce subunit combination without appearing to alter the seatbelt latch or other disulfide (Table 1).
Cys 26 is part of a large tryptic fragment that contains several cysteines (i.e., Cys 23, Cys 26, Cys 34, Cys 38) that was not detected following treatment with IA or VP using the alkylation/MALDI-TOF procedure just described. We were able to study the reactivity of Cys 26 using biotinylated alkylating agents and a monoclonal antibody (i.e., B111) that recognizes a β-subunit epitope that includes the seatbelt latch disulfide (Cosowsky et al. 1995).
Biotinylation of Cys 26 or Cys 110 by these alkylating agents would lead to a molecule that binds streptavidin and monoclonal antibodies to other parts of the β-subunit (e.g., B110), but it also would prevent binding of B111. Alkylation of cysteines or other residues is readily detected in sandwich immunoassays using streptavidin for capture and radioiodinated antibodies for detection or vice versa. Because of the high affinities of streptavidin for biotin and the high specific activities to which the antibodies can be radiolabeled, this assay is much more sensitive than the MALDI-TOF method.
MB failed to inhibit subunit combination (Fig. 6, upper panels), even though it was incorporated into the heterodimer (Fig. 6, lower panels) and both free subunits (Fig. 7). Under these conditions, however, MB was not incorporated into β-subunit Cys 26 or Cys 110 as assessed with a Western blot streptavidin-sandwich assay. In this assay, the proteins were captured by streptavidin that had been adsorbed to the nitrocellulose and detected with anti-β-subunit antibodies B110 and B111. The latter is sensitive to the presence of the seatbelt latch disulfide. As seen (Fig. 6, lower panels), the biotinylated β-subunit and heterodimer were detected nearly equally by B110 and B111. Thus, only trace amounts of biotin, if any, were associated with either Cys 26 or Cys 110. Furthermore, it was possible to drive all of the β-subunit into the heterodimer by increasing the amount of α-subunit in the reaction mixture (Fig. 6, all panels). Alkylation of either Cys 26 or Cys 110 in even a fraction of the β-subunit would have led to detectable amounts of free β-subunit in the streptavidin-B110 assay. We have not attempted to identify residues in either subunit that reacted with MB but have observed that trypsin digestion produced a peptide fragment having a molecular weight of 1340.9 (not shown). This is consistent with biotinylation of Tyr 37 in the sequence Ala 36–Tyr-Pro-Thr-Pro-Leu–Arg 42, a peptide derived from α-subunit loop 2. This peptide had also been alkylated by IA (Fig. 5).
The rate of subunit combination has been shown to be increased by several-fold in mild reducing conditions (Slaughter et al. 1995). This might be explained by opening of the seatbelt latch disulfide, one of the more easily reduced disulfides in the glycoprotein hormones (Pierce and Parsons 1981), a process expected to be highly sensitive to the presence of alkylating agents. In confirmation of earlier studies, we found that subunit combination occurred sixfold more rapidly in the presence of 1 mM β-mercaptoethanol than in its absence (Fig. 8). Higher concentrations of β-mercaptoethanol were less effective in promoting subunit combination (not shown), and when present at 10 mM, β-mercaptoethanol prevented subunit combination, most likely by denaturing the subunits. Even at 1 mM, β-mercaptoethanol caused the reduction of several disulfides, not just the seatbelt latch. We detected alkylation of Cys 72, Cys 93, Cys 100, and Cys 110 by 10 mM VP when subunit combination had been initiated in the presence of 1 mM β-mercaptoethanol (Fig. 9). Thus, in addition to destabilizing the seatbelt latch (Cys 26–Cys 110), mild reduction altered the disulfide between β-subunit loops 1 and 3 (Cys 23–Cys 72) and the disulfide within the small seatbelt loop (Cys 93–Cys 100). Reduction of these disulfides in 1 mM β-mercaptoethanol was only partial, as shown by the finding that complete reduction of the β-subunit with 100 mM DTT and alkylation with IA modified Cys 72, Cys 93, and Cys 110, as well as cysteines that had not been modified by VP (Fig. 10). Unlike subunit combination in oxidizing conditions, which was not blocked by 100–1000-fold molar excess of alkylating agent, subunit combination during mildly reducing conditions was inhibited immediately and quantitatively by a 10-fold molar excess of alkylating agent (Fig. 10). These observations suggested that we would have observed alkylation of the cysteines in the seatbelt latch had this disulfide been broken during subunit combination under oxidizing conditions, particularly as VP and IA modified peptides were readily resolved during MALDI-TOF (Fig. 9).
MALDI-TOF is a powerful analytical technique that can be used to identify peptide fragments, but its use as a quantitative tool is limited. In an effort to determine the relative amounts of each cysteine that had been modified during subunit combination in the presence of mild reducing conditions, we trapped cysteines that had been reduced using a fluorescent alkylating agent (IAEDANS). Following trypsin digestion of hCG, it was possible to quantify the relative amounts of each labeled tryptic fragment by HPLC using a fluorescence detector. As seen (Fig. 11), the AEDANS fluorophore was present in two peptides in nearly equal amounts. MALDI-TOF analyses showed that fluorescent peptides Gly 75–Val-Asn-Pro-Val-Val-Ser-Tyr-Ala-Val-Ala-Leu-Ser-Cys-Gln-Cys-Ala-Leu-Cys–Arg 94 and Ser 96–Thr-Thr-Asp-Cys-Gly-Gly-Pro–Lys 104 each contained one molecule of AEDANS. Though we did not determine the locations of the AEDANS residues in each peptide by tandem mass spectrometry, it is highly probable that these were located on Cys 93, based on the ability of this cysteine in this peptide to be preferentially alkylated following mild reduction (Fig. 9) and on Cys 100. Only trace amounts of AEDANS were found on any other peptide fragment of hCG, including that which contained Cys 110. This showed that even in the presence of mild reducing conditions, the Cys 26–Cys 110 bond does not appear to be disrupted in conditions that enhance subunit combination.
These observations indicate that α-subunit loop 2 and its associated oligosaccharide can pass through the small hole created by latching of the seatbelt to the core of the β-subunit during subunit combination. How can the subunits combine while the seatbelt latch remains intact? We propose that the position of the seatbelt in the free β-subunit differs considerably from that in the heterodimer. Experimental support for this idea comes from the observation that many hCG-specific antibodies (i.e., B107, B109) recognize an epitope composed of β2 and the N-terminal end of the seatbelt (Moyle et al. 1990). These antibodies have very low ability to bind the free β-subunit, indicating that this epitope has a different conformation in hCG and in the free subunit. When present in a position such as that in Figure 12A, the seatbelt would not prohibit docking of the α- and β-subunits. In fact, in this position, the seatbelt may contribute to subunit interaction. Following the reversible interactions between the two subunits (Fig. 12B), movements of the seatbelt relative to α2 and its associated oligosaccharide (Fig. 12C) continue until these portions of each subunit occupy the more stable conformations seen in the heterodimer (Fig. 12D). Loop 2 of the α-subunit is known to be highly mobile (De Beer et al. 1996), a factor that may favor this transition but is not illustrated in Figure 12.
The observation that small quantities of reducing agents appear to disrupt the Cys 93–Cys 100 disulfide suggests that mild reduction (Slaughter et al. 1995) or protein disulfide isomerase treatment (Huth et al. 1993) can facilitate subunit combination by increasing the opening between the seatbelt and the core of the β-subunit. It remains to be determined if this occurs in living cells, however. These studies suggest that the subunits can combine in vitro while the seatbelt is latched and raise the possibility that this can also occur during synthesis in the endoplasmic reticulum.
Materials and methods
Pure recombinant hCG (Ares Serono International) was separated into its constituent subunits by reversed phase HPLC on a Vydac C4 column (22.5 × 250 mm) in 0.1% aqueous TFA with a 15%–40% linear gradient of 0.079% TFA in acetonitrile. The isolated subunits were equilibrated in 10 mM sodium phosphate buffer (pH 7.2) by ultrafiltration on a 5000YM Amicon membrane. The concentration of each subunit was determined in triplicate by automated analysis on an ABI420A amino acid analyzer. Antibodies A113 and B111 were given to us by Glenn Armstrong and Robert Wolfert, Hybritech. B110 was prepared in this laboratory (Moyle et al. 1987). Antibody A113 recognizes an epitope found primarily on the surface of α-subunit loop 1 near its interface with the β-subunit.
Antibody B110 binds an epitope on the convex surface of β-subunit loops 1 and 3 that does not include the seatbelt. Antibody B111 binds an epitope that includes the seatbelt latch disulfide. N-ethylmaleimide (NEM), iodoacetate (IA), iodoacetamide, 4-vinylpyridine (VP), and streptavidin (SA) were purchased from Sigma. 3–(N-maleimidylpropionyl)biocytin (MB), a biotin containing derivative of NEM, and 5–((((iodoacetyl)amino)ethyl)amino)napthalene-1-sulfonic acid (IAEDANS) were purchased from Molecular Probes. Antibodies used for detection of free subunits and heterodimers in Western blotting were radioiodinated using an Iodo-Gen-based technique (Cruz et al. 1987).
Subunit combination and analysis of heterodimer
Equimolar concentrations of subunits (0.08–0.1mM) in 25 mM phosphate, 5 mM KCl (pH 7.5) were incubated at 37°C in the presence or absence of high concentrations of alkylating agents as noted in the figure and table legends. As noted also, lower concentrations of alkylating agents were used to study subunit combination that had been initiated in the presence of 1 mM β-mercaptoethanol. Heterodimer formation was measured using sandwich immunoassays (Moyle et al. 1982) and Western blot analysis (Moyle et al. 1990). Free and combined subunits were separated by electrophoresis on 12% polyacrylamide gels containing 0.1% sodium dodecyl sulfate. Proteins electroeluted from the gel onto nitrocellulose sheets were detected using radiolabeled antibodies to either the α-subunit (A113) or β-subunit (B110, B111). Biotinylation of the subunits and the heterodimer by MB was monitored by electrophoresis and Western blotting. The presence of biotin in these proteins was determined by chemiluminescence using the streptavidin-linked horseradish peroxidase kit obtained from Boehringer Mannheim. Incorporation of MB was also monitored by a Western blot–sandwich procedure in which the biotinylated subunits and dimer were separated by electrophoresis and then transferred to nitrocellulose sheets that had been coated with streptavidin (20 μg/mL, 1 h, 37°C) and blocked with bovine serum albumin (1 mg/mL, 1.5 h, room temperature). Following transfer, only those proteins that contained biotin were retained on the membranes. Detection of biotin labeled heterodimer and β subunit was performed with B110 or B111 radiolabeled to a specific activity of 50 μCi/μg.
Analysis of disulfide opening during subunit combination
We combined the subunits for 2–4 h or overnight at 37°C in the presence of 100 mM IA or 100 mM VP to trap cysteines created by the breaking of disulfide bonds during heterodimer formation. Unreacted alkylating agents were removed from the reaction mixture by reversed-phase HPLC on C18 resin (Phenomenex) or by electrophoresis. Following the latter procedure, the heterodimers and subunits were recovered from gel slices by electroeluting them at 40 V overnight. The samples were denatured and all remaining disulfide bonds were disrupted by incubation in an argon-purged solution containing 10 M urea, 2 mM EDTA, 20 mM Tris (pH 7.5), and 100 mM DTT for 1–2 h at 37°C. Cysteines that had not been alkylated during subunit combination in the presence of IA were alkylated in the presence of 200 mM VP (IA→VP). Conversely, subunits that had been exposed to VP during combination were alkylated with 200 mM IA (VP→IA). These alkylations were performed at room temperature for 2–3 h or overnight at 4°C. The denatured and alkylated proteins were again purified by reversed-phase HPLC or by “zip-tips” containing C18 resin (Millipore) to remove salts, urea, and unreacted VP or IA. After being lyophilized, the dried proteins were dissolved in 50 mM Tris-HCl buffer (pH 7.5) containing 1 mM CaCl2 and digested with TPCK-treated trypsin (Promega) at 37°C overnight.
In some cases, the oligosaccharides were removed before trypsin digestion by treating the samples with N-glycanase (New England Biolabs), as recommended by the supplier except for the omission of NP40. The tryptic digests were analyzed by MALDI/TOF mass spectrometry using the Voyager DE/STR and DE/PRO instruments (PE Biosystems) operated in reflector mode using delayed extraction. Each spectrum was externally calibrated using a mixture of standard peptides. The identities of these peptides were confirmed following electrospray ionization using an LCQ ion trap mass spectrometer (Finnigan–MAT) operated in automated LC/MS/MS mode.
Analysis of disulfide opening using IAEDANS: hCG α- and β-subunits (0.05 mM) were combined in the presence of IAEDANS (5 mM) with 0, 1, and 3 mM β-mercaptoethanol at 37°C for 2 h. The combined and free subunits were separated from unreacted IAEDANS by electrophoresis through 0.1% dodecylsulfate-12% polyacrylamide gels, electroeluted from the gels in running buffer diluted 1 : 10, and mixed.
Following lyophyllization, they were reduced by 100 mM DTT and alkylated with 250 mM iodoacetamide. The alkylated subunits were purified by reversed phase HPLC on a C18 resin and digested with trypsin. The fragments were separated by reversed phase HPLC on C18 using a gradient of 0%–95% acetonitrile in 10 mM ammonium bicarbonate (pH 8). Fluorescence was monitored at 490 nm following 337 nm excitation.