Evolution of Constrained Gonadotropin-releasing Hormone Ligand Conformation and Receptor Selectivity*

Gonadotropin-releasing hormone (GnRH) is the central regulator of reproduction in vertebrates. GnRHs have recently been identified in protochordates and retain the conserved N- and C-terminal domains involved in receptor binding and activation. GnRHs of the jawed vertebrates have a central achiral amino acid (glycine) that favors a type II′ β-turn such that the N- and C-terminal domains are closely apposed in binding the GnRH receptor. However, protochordate GnRHs have a chiral amino acid in this position, suggesting that they bind their receptors in a more extended form. We demonstrate here that a protochordate GnRH receptor does not distinguish GnRHs with achiral or chiral amino acids, whereas GnRH receptors of jawed vertebrates are highly selective for GnRHs with the central achiral glycine. The poor activity of the protochordate GnRH was increased >10-fold at vertebrate receptors by replacement of the chiral amino acid with glycine or a d-amino acid, which favor the type II′ β-turn. Structural analysis of the GnRHs using ion mobility-mass spectrometry and molecular modeling showed a greater propensity for a type II′ β-turn in GnRHs with glycine or a d-amino acid, which correlates with binding affinity at vertebrate receptors. These findings indicate that the substitution of glycine for a chiral amino acid in GnRH during evolution allows a more constrained conformation for receptor binding and that this subtle single amino acid substitution in a site remote from the ligand functional domains has marked effects on its structure and activity.

In vertebrates, gonadotropin-releasing hormone (GnRH) 2 is synthesized in hypothalamic neurones and conducted a few millimeters in the hypophysial portal system to the anterior pituitary, where it binds to high affinity receptors in gonadotrophs to stimulate the release of gonadotropins (1). The gonadotropins in turn stimulate hormone and gamete production by the testes and ovaries. GnRHs have also been isolated from protochordate species (2,3) and are thought to be secreted from neurones to directly regulate the gonads (2)(3)(4) in these representatives of vertebrate progenitors. GnRHs and GnRH receptors have also been found to directly affect vertebrate gonadal function (5), possibly reflecting the earliest role of GnRH as exemplified in protochordates (6,7). It appears, therefore, that GnRHs have an ancient evolutionary role as regulators of reproduction, first through direct neural delivery to the gonads and later as hypothalamic neuroendocrine regulators of the gonads indirectly through gonadal stimulation by gonadotropins.
To date 13 structural variants of the GnRH decapeptide have been identified in vertebrates (8), 9 from protochordates, which are vertebrate progenitors (8,9), and a 12-amino acid homolog from an octopus species (10) (Fig. 1). All of the GnRHs are characterized by the conservation of the N-terminal residues (pGlu-His-Trp-Ser) and the C-terminal residues (Pro-Gly-NH 2 ) with the exception of two conservative substitutions (Fig. 1). In cartilaginous and bony fish, amphibians, reptiles, birds, and mammals, all of the GnRHs are further characterized by the presence of a glycine residue in position six (Fig. 1). Being archiral, the glycine residue allows the peptide to assume a type IIЈ ␤-turn conformation, which is essential for high binding affinity and biological activity in mammals (1,8,11,12). Substitution of glycine with any other L-amino acid decreases biological activity as the type IIЈ ␤-turn conformation is less favored (1,8,11,12). In contrast, substitution of glycine with Damino acids in position six constrains the peptide to the type IIЈ ␤-turn conformation and increases binding affinity and biological activity in mammals, birds, amphibians, and jawed fish species (1, 6 -8, 11, 12).
The GnRHs in the ancient jawless lamprey and protochordate species (with the exception of Ciona VI) are all characterized by the presence of chiral amino acids in position six (Fig. 1). This feature is likely to limit the formation of a type IIЈ ␤-turn conformation, and these GnRHs would be expected to have correspondingly low binding affinities and biological activities at GnRH receptors of higher vertebrates (1,6,7,12).
We, therefore, hypothesized that early in evolution GnRH interacted with the GnRH receptor in a more relaxed (linear) conformation. Coincident with the evolution of the jawed fish, structural changes in GnRH and its receptor required the peptide to interact in the folded type IIЈ ␤-turn conformation. To investigate this notion, we have synthesized and studied the structure and biological activity of three GnRHs from a tunicate protochordate (Ciona intestinalis) at catfish, chicken, human, and protochordate GnRH receptors. We demonstrate that these protochordate GnRHs have low biological activity at jawed vertebrate GnRH receptors but that high biological activity is generated when the natural chiral amino acid in position six is replaced by the achiral glycine, which is characteristic of jawed vertebrate GnRHs. We also studied the structure of these peptides employing a combination of ion mobility-mass spectrometry (IM-MS) and molecular modeling and demonstrated a correlation of the biological activity at vertebrate receptors with the ability of the peptides to assume the folded more compact conforma-tion. These findings reveal a co-evolution of GnRH and its cognate receptor in which the receptors of early evolved organisms bind an extended structure of GnRH, whereas the receptors of higher organisms require a folded type IIЈ ␤-turn conformation of GnRH in which the N and C termini are in close apposition when bound to the receptor.

EXPERIMENTAL PROCEDURES
Peptides-Ciona I, II, and III GnRH, Gly 6 -Ciona I GnRH, and D-Ala 6 -Ciona I GnRH were synthesized by conventional solid phase methodology and purified by high performance liquid chromatography to Ͼ98% purity. Mammalian GnRH was from Peninsula Laboratories (Bachem Ltd., Merseyside, UK).
Cell Culture and Transient Transfection-Plasmid DNA for transient transfection was prepared using Maxi-Prep columns (Qiagen, Chatsworth, CA) according to the manufacturer's instructions. COS-7 and human embryonic kidney 293 cells were cultured as previously described for COS-1 cells (12,13) and transiently transfected with the human (14), chicken (15), and catfish (16) GnRH receptors as well as the recently cloned Ciona receptors (17) 3 using electroporation. These are Ciona A and Ciona B GnRH receptors previously referred to as Ci-GnRHR1 and Ci-GnRHR2.
Receptor Binding Assays-Whole cell receptor binding assays used the 125 I-[His 5 ,D-Tyr 6 ]GnRH analog for the vertebrate receptors and 125 I-GnRH II for the Ciona receptor (13). Transiently transfected COS-7 cells in 12-well culture plates were washed once with ice-cold HEPES, Dulbecco's modified Eagle's medium, 10% fetal calf serum and incubated for 5 h on ice in the same medium with 10 6 cpm/well radiolabeled GnRH analog and varying concentrations of unlabeled GnRH peptides (13). Cell monolayers were then rapidly washed twice in icecold phosphate-buffered saline and solubilized in 0.1 M NaOH, and the lysate radioactivity was counted. Nonspecific binding (consistently less than 10% of total binding) was determined using vector-transfected (pcDNA1/amp) COS-7 cells and subtracted from total binding to give specific binding. Assays were conducted in triplicate and repeated in three separate experiments.
Total Inositol Phosphate Assays-GnRH stimulation of total inositol phosphate production was assayed as previously described (18,19). Briefly, transiently transfected COS-7 cells were incubated with inositol-free Dulbecco's modified Eagle's medium containing 1% dialyzed heat-inactivated fetal calf serum and 0.5 Ci/well myo-[ 3 H]inositol (Amersham Biosciences) for 48 h. Medium was removed, and the cells were washed with 1 ml of buffer (140 mM NaC1, 20 mM HEPES, 4 mM KC1, 8 mM glucose 1 mM MgC1 2 , 1 mM CaC1 2 , and 1 mg/ml bovine serum albumin) containing 10 mM LiC1 and incubated for 1 h at 37°C in 0.5 ml of buffer containing 10 mM LiCl and GnRH agonist at the indicated concentrations. Reactions were terminated by the removal of agonist and the addition of 1 ml of ice-cold 10 mM formic acid, which was incubated for 30 min at 4°C. Total [ 3 H]inositol phosphates were separated from the formic acid cell extracts on AG-X8 anion exchange resin (Bio-Rad) and eluted with a 1 M ammonium format, 0.1 M formic acid solution. The associated radioactivity was determined by liquid scintillation counting. Assays were conducted in triplicate and repeated in three separate experiments.
Ion Mobility-Mass Spectrometry-The gas-phase collision cross-sections of mammalian GnRH, Ciona 1 GnRH, and its two substituted derivates were determined using an ion mobility mass spectrometer, as previously described (20). Briefly, ions created by electrospray ionization are injected into a temperature-regulated drift cell pressurized with helium to 5 torr. The ions drift under the influence of a weak electro-FIGURE 1. Primary amino acid sequences of naturally occurring decapeptide GnRH structural variants spanning ϳ600 million years of evolution. The boxed regions show the conserved Nand C-terminal residues that are involved in receptor binding and activation. The GnRHs are named according to the species in which they were first discovered, and they may be represented in more than one species. For example, mammalian GnRH is widely conserved in amphibians and primitive bony fish, and chicken GnRH II is present in most vertebrate species, including man. An octopus GnRH and an additional Ciona GnRH comprising 12 and 16 amino acids, respectively, but retaining the conserved N-and C-terminal domains are not shown.
static field and are retarded by collisions with the buffer gas. On exiting the cell, ions are selected by a quadrupole mass filter, and their arrival time distributions are recorded. Arrival times are collected at several drift voltages, and the mobility (K) of the ion is obtained from the gradient of a plot of arrival time versus the pressure of helium divided by the drift voltage. This mobility is used to determine the experimental collision cross-section of the ion () according to the equation K ϭ [(3e)/ (16N)][(2)/( B T)] 1/2 [1/] as described (21).
Molecular Structures of GnRH Peptides-Analysis of the structures of mammalian GnRH, Ciona I GnRH, and two substituted Ciona I variants was conducted using molecular mechanics. All calculations were performed using the ff99 AMBER force field (22), employing the Amber7 suite of programs to refine the energy of the peptide models. Residues for the N-terminal pyrolyzed glutamic acid and the amidated glycine present at the C terminus generated using the RESP procedure (23) and the ANTECHAMBER facility within the Amber 7 suite of programs (amber.scripps.edu). IM-MS data were obtained for these peptides both as singly ([MϩH] ϩ ) and doubly ([Mϩ2H] 2ϩ ) protonated ions. From gas-phase basicities (24) it is most probable that Arg 8 in mammalian GnRH will be protonated in [MϩH] ϩ and that the [Mϩ2H] 2ϩ ion will contain protonated His 2 . In the protochordate Ciona I peptides, Arg 8 is substituted by Ser 8 , and it is extremely likely that their [MϩH] ϩ ions contain protonated histidine. The protonation state of His 2 is variable in physiological conditions, and therefore calculations were performed with both protonated and unprotonated histidine for the Ciona I peptides and for the singly and doubly protonated form of mammalian GnRH. At most physiological pH values, Arg 8 of mammalian GnRH will be protonated; hence, calculations on the neutral species have not been performed for this peptide. The Ciona I [Mϩ2H] 2ϩ ions probably arise due to protonation of the amide Gly 10 ; however, since we cannot determine the biological significance of this chemical form and due to some ambiguity in assigning this site of charge, the doubly protonated peptides were not investigated further in this study. For each of these 8 chemically distinct peptides, 300 candidate gas-phase minimized structures were generated using a simulated annealing approach. An initial structure was subjected to dynamics at 800 K for 30 ns and then cooled in a stepwise fashion to 0 K before an energy minimization. This minimized structure was then used as the seed for the next cycle. The two lowest energy structures for each species were then subjected to 1 ns of molecular mechanics at 300 K for the [MϩH] ϩ ions. This was performed in vacuo and employing the Born Solvent distance-dependent dielectric model to simulate their dynamic structure in an aqueous environment (25). For the neutral Ciona I peptides, these extended dynamics calculations were only performed with the use of the solvent model. This procedure generates 1000 co-ordinate files for each 1 ns, i.e. coordinates are saved every ps, with the potential energy surface sampled every fs. No weighting was given to any of the amino acids, and translational energies are removed every ps. Each molecular mechanics data set was analyzed, and conformational variables such as the N-C termini distance and overall peptide flexibility were compared with those found for mammalian GnRH examined under identical conditions. Collision cross-sections for each low energy structure obtained via the simulated annealing procedure were calculated using the projection approximation (26). This was also employed to elucidate conformational changes occurring to these peptides during molecular dynamics.

RESULTS
Receptor Binding-All three Ciona GnRHs had low affinity for the human receptor (IC 50 Ͼ 1 M) in comparison to mammalian GnRH (7.5 nM) ( Fig. 2A, TABLE ONE). Similar results were found in binding of these peptides to the chicken and catfish (data not shown) GnRH receptors. In contrast Ciona I and Ciona II had low affinities (918 and 988 nM) at the Ciona A receptor, which were similar to that of mammalian GnRH (1390 nM) (Fig. 2B). Ciona II had a considerably lower affinity (5600 nM) and may be the cognate ligand for a second Ciona GnRH receptor (17), which expressed poorly and, therefore, did not allow binding studies (data not shown). Because Ciona I was the most active of the Ciona peptides at the human receptor, this was selected for studies to determine whether substitution of the chiral amino acid in position six would increase activity at the human receptor. Replacement of Ala 6 with the achiral Gly 6 or with the type IIЈ ␤-turn-constraining D-Ala 6 resulted in a marked increase (36 -46-fold) in binding affinity (Fig. 2C, TABLE ONE). Similar results were obtained for the chicken (Fig. 2D) and catfish (data not shown) GnRH receptors.
Inositol Phosphate Production-The three Ciona peptides were at least 3 orders of magnitude less potent than mammalian GnRH at stimulating inositol phosphate production (Fig. 3, TABLE ONE). Their relative activities paralleled that of their binding affinities. Substitution of Ala 6 with Gly 6 or D-Ala 6 increased the potency more than 100-fold.
Ion Mobility-Mass Spectrometry and Molecular Modeling of GnRH Peptides-Experimental cross-sections measured for the [MϩH] ϩ ions of mammalian GnRH, Ciona I, and Ciona II GnRH analogs are given in TABLE TWO. Agreement between the measured and calculated values is very good. The gas-phase cross-section of mammalian GnRH is smaller than that obtained for the Ciona I GnRH, and substitution of L-Ala 6 with D-Ala 6 or Gly 6 reduces the cross-section to approach that of mammalian GnRH. Fig. 4 shows representative low energy structures obtained from the in vacuo simulations of the peptide ions. The general conformation of all of the peptides is comparable, as might be expected based on their sequence homologies. All form compact structures with a hairpin shape to the polypeptide backbone, which is sustained through non-covalent interactions across this structure. The radius is larger for the Ciona I form than that found for the Gly 6 -or D-Ala 6 -substituted Ciona I (Fig. 4), in agreement with the IM-MS cross-section data (TABLE TWO). Cross-sections obtained from the geometry calculations with unprotonated histidine are also included.
By performing molecular dynamics on these low energy structures, the conformational space available to them is sampled, thus probing the stability of interactions formed. This was conducted both in vacuo and also with the application of a solvent model. Conformations obtained with the Born Solvent model are presumed to be more representative of biologically active conformations, at least before receptor binding. Several criteria were examined in the room temperature dynamics, and our findings are summarized below and in TABLE THREE.
N-to C-terminal Distance-The proposal that GnRH binds the GnRH mammalian receptor in a type IIЈ ␤-turn conformation suggests non-covalent interactions which position the N-and C-terminal residues closely apposed in space. The low energy gas-phase structures determined here fulfill this criterion (Fig. 4). Over the 1-ns dynamics runs, the persistence of this interaction was explored by plotting the distance between the ␣ carbons of the N-terminal pyrolyzed glutamic acid and the C-terminal glycine amide. The in vacuo simulations for mammalian GnRH and Ciona I GnRH and analogs reveal a continual N-C-terminal proximity, which bears out the analysis made above on the low energy conformation (TABLE THREE). It is apparent that the distance between the ␣ carbons decreases as residue 6 is replaced by D-Ala or achiral glycine. With application of the Born Solvent distance dielectric function, Ciona I GnRH exhibits a somewhat larger average N-C-terminal variation than D-Ala 6 -Ciona and significantly larger than for Gly 6 -Ciona I (TABLE THREE). Mammalian GnRH displays a more rigid conformation, with less deviation from the average N-C distance than the Ciona I peptides. However, all of the forms of GnRH experience an opening of the peptide backbone within the higher dielectric environment, resulting in a larger average N-C-terminal distance as seen in representative structures obtained from the dynamics runs under Born Solvent conditions (Fig. 5).
Trp 3 -Leu 7 Distances-Although it is clear that the GnRH peptides exhibit considerable conformational flexibility, their dominant structural feature is a type IIЈ ␤-turn between residues 5 and 8. The increase in receptor binding affinity when the achiral Gly or D-Ala is substituted for the chiral L-Ala in Ciona I GnRH is attributed to a release in steric hindrance allowing for a tighter hinge to the peptide backbone. We have examined the distance between Trp 3 and Leu 7 as a possible indicator of hinge mobility. For Ciona I and the substituted analogs, the average distance between these residues is remarkably similar (TABLE THREE).     Each is a snapshot from in vacuo dynamics calculations. The amidated N-terminal glycine is shown in each, as is residue six and the proton-carrying residue. As described under "Results" and in TABLE THREE, the cross-peptide radius for Ciona I peptide is larger than for the other peptides (b, c, and d), which possesses a tighter turn around residues 5 and 6. In all of these peptides the proton carrying group (histidine for the Ciona peptides and arginine for mammalian GnRH) "caps" the polypeptide ring, thereby providing points for non-covalent interactions that stabilize these compact geometries of gas-phase ions.

Receptor binding and inositol phosphate production by mammalian and Ciona GnRHs and analogues at the human GnRH receptor expressed in COS-7 cells
It is, however, ϳ8 -16% more than the average value found for the mammalian GnRH, modeled under identical conditions. In the gasphase simulations this effect is most marked. This suggests that the primary sequence and, specifically, the Ser 8 to Arg 8 substitution has a large effect on retaining the rigidity of this type IIЈ ␤-turn. Conformational Flexibility-Two methods were employed to assess conformational flexibility. The first examined the rotationally averaged collision cross-section of each peptide conformation during the course of the 1-ns dynamics run. The second, more conventional approach examined the root mean square deviation of the amino acids from their positions in an averaged structure. Both methods gave comparable results. The in vacuo data essentially confirms the findings discussed above. Over the course of the dynamics the peptides retain their compact conformations, and significantly, the size ordering remains L-Ala 6 Ͼ D-Ala 6 Ͼ Gly 6 . With the application of the solvent model, collision cross-sections increase for all the GnRH peptides, reflecting looser, and at times, elongated structures. The overall conformational flexibility of Ciona I GnRH was greater than that of the mammalian GnRH. Over the course of 1 ns of molecular dynamics at 300 K Ciona I GnRH exhibited more frequent conformational shifts than Gly 6 Ciona I and mammalian GnRH (Fig. 6). In the dynamics runs, the average cross-section of Ciona I [M] (Fig. 6A) was 275.9 Ϯ 12.2 Å 2 , and Gly 6 Ciona I [M] (Fig. 6B) (TABLE THREE). These trends are also seen in the [MϩH] ϩ protonated histidine forms. The D-Ala 6 and Gly 6 forms of Ciona I and mammalian GnRH retain a centrally configured loop in the solvent calculations (Fig. 5, B and C). Comparison with the data from mammalian GnRH (Fig. 5C) shows that this form of the peptide also appears somewhat stabilized during dynamics, although it also exhibits an increase in the N-C-terminal distance. The overall cross-section of mammalian GnRH is larger due to the Arg 8 . Interestingly, this bulky side chain of Arg 8 is rather flexible, which supports data obtained from both the mouse (15) and the human (16) receptors that demonstrate Arg 8 is integral to binding via an acidic residue in extracellular loop 3 (15,16). The jump in cross-section in the earlier part of these dynamics measurements is expected, as the temperature of the low energy starting conformers is brought to thermal equilibration.
Although no attempt was made to model the interaction of the peptide with the receptor, it is apparent that the configurations of the isolated peptides can be related to the biological activity. The correlation between the experimentally measured cross-sections (TABLE TWO) and the stimulation of inositol phosphate data (TABLE ONE) is very good. The small tightly configured mammalian GnRH exhibits the strongest activity at the receptor, whereas the larger, "looser" Ciona I

Table showing the average results of co-ordinate analysis from extended dynamics calculations
Listed are average N-C distances (C␣-C␣), the distance between Trp 3 and Leu 7 , and the collision cross-sections of the confirmation from co-ordinates averaged over 1 ns of dynamics performed at 300 K. The figure given is averaged from those obtained from two molecular dynamic runs. The standard deviation from this value is given in parentheses.  Each is a snapshot from dynamics performed with the Born Solvent model, with the Nand C-terminal residues represented by Corey-Pauling-Koltun structure. The distance between the ␣ carbons of the terminal residues in each snapshot is marked. The D-Ala 6 -Ciona I GnRH structure was similar to Gly 6 -Ciona I (not shown).

FIGURE 6. Cross-section measurements obtained from Born Solvent model simulations for Ciona I [M] (A), Gly 6 -Ciona I [M] (B), and mammalian GnRH [M؉H] ؉ (C).
The dynamics runs were performed for 1 ns at 300 K. The data shown is obtained by determining the collision cross-section of the structures output from these calculations every 1 ps using the projection approximation (26). The initial spike observed here in each plot is an artifact attributable to the effect of rapidly heating the peptide up to 300 K from its 0 K low energy structure. These data correspond to the averaged values shown in TABLE THREE and nicely demonstrate the conformational flexibility of Ciona I compared with mammalian GnRH and to the Gly 6 -substituted form.
L-Ala analog is the weakest binder. The D-Ala 6 -and Gly 6 -Ciona I peptides are intermediate in cross-sections and binding affinity. This coupled with the fact that all the gas-phase structures have proximal N-C termini suggests that the gas-phase conformations are comparable with the receptor-bound form.

DISCUSSION
The importance of having the achiral glycine in position 6 of mammalian GnRH for biological activity was demonstrated in empirical studies three decades ago (1,11,27). Subsequent studies using N-and C-terminal-directed antibodies, fluorescence spectroscopy, molecular modeling, and NMR suggested that Gly 6 was essential to allow GnRH to assume a type IIЈ ␤-turn conformation (1,8,11). Although GnRH receptors had not been cloned when this concept was proposed, these findings together with numerous structure-activity studies indicated that mammalian GnRH interacted with its cognate receptor in this type IIЈ ␤-turn conformation through the amino (pGlu-His-Trp-Ser) and carboxyl (Arg-Pro-Gly-NH 2 ) terminal domains (1, 6 -8). The subsequent elucidation of GnRH structural variants in vertebrates and protochordates reveals the conservation of these domains over more than 500 million years of evolution, thus supporting this conclusion (Fig. 1). Mutation of amino acids in cloned human and rat GnRH receptors has provided direct evidence for the interaction of individual amino acids in these two ligand domains with cognate receptor residues (pGlu with Asn 212 ; His 2 with Lys 121 and Asp 98 ; Trp 3 with Trp 280 ; Arg 8 with Asp 302 ; Gly.NH 2 10 with Asn 102 ) (8, 18, 19, 28 -30). The receptor residues which bind the conserved pGlu, His, Trp, Ser, and Gly-NH 2 of Ciona GnRHs are all conserved in the Ciona receptors as in fish, amphibian, and chicken receptors (8) with the exception of Asn 212 , which is a tyrosine in the Ciona A receptor.
In view of the stringent requirement of jawed vertebrate GnRH receptors for a glycine residue in position six to allow presentation of the N-and C-terminal domains in a folded conformation to the receptor, the presence of chiral amino acids in position six in jawless fish and protochordate GnRHs was unexpected. This finding suggests that the receptors of these species are able to interact with GnRH in an extended conformation. The similar low binding affinities (ϳ1 M) of mammalian, Ciona I, and Ciona III GnRHs at the Ciona A receptor supports this interpretation. The ability of the endogenous protochordate GnRHs and mammalian GnRH to stimulate spawning in protochordate species (2,3,9) also indicates that protochordate GnRH receptors do not distinguish mammalian and protochordate GnRHs. GnRHs with an achiral amino acid in position six were similarly active in mollusk species (31,32). Interestingly, one of the protochordate GnRHs (Ciona VI) has Gly 6 and is active in stimulating spawning in C. intestinalis (9). This finding supports our data demonstrating that the Ciona A GnRH receptor binds GnRHs with chiral (Ciona I and III) and achiral (mammalian GnRH) amino acids in position six equally well. This is consistent with our interpretations, as forms with chiral and achiral amino acids in position six can both assume a relaxed linear extended conformation. In contrast, we have shown here that the types of GnRH with chiral amino acids in position six have very low binding affinity at mammalian, avian, and teleostean GnRH receptors.
Our molecular modeling and IM-MS studies demonstrate a correlation between the propensity for type IIЈ ␤-turn conformation of GnRHs and their biological activity at the vertebrate receptors. Thus, Ciona I GnRH assumes a looser and, hence, less configured structure than mammalian GnRH and has poor binding affinity at vertebrate receptors. By comparing the collision cross-sections from experimental and calculated structures, it is apparent that L-Ala at position six induces steric hindrance to the formation of the more compact folded structure. This is supported by the larger Trp 3 -Leu 7 distances found with Ciona I than for the Gly 6 or D-Ala 6 forms in vacuo and associated larger IM-MS collision cross-sections (TABLE TWO). We have previously investigated the gas-phase conformations of D-Trp-substituted mammalian GnRH by both experiment and calculation (33). Here we find that the D-Trp form adopts a much more compact geometry, with a type IIЈ ␤-turn, whereas the L-Trp variant is more extended due to steric effects caused by the bulky side chain in the naturally occurring form. The solvent calculations we present here on the Ciona I variants show these effects more dramatically. L-Ala at position six allows the backbone to open out in the course of the dynamics simulations, producing a much more elongated and less compact form of the peptide. In contrast, the D-Ala 6 -and Gly 6 -substituted Ciona I have more compact structure throughout the solvent dynamics. These findings concur with NMR studies on mammalian GnRH and D-Trp 6 -GnRH (34), but NMR is unable to distinguish structural differences between GnRHs without D-amino acids in position six. The extensive molecular modeling of Guarnieri and Weinstein (35) revealed that the conformational preference for a type IIЈ ␤-turn in the backbone of mammalian GnRH is significantly diminished by an Arg 8 3 Lys 8 substitution. Our gas-phase results also indicate the functional importance of Arg in the mammalian form, where its guanidinium side chain is flexible and forms favorable interactions across the polypeptide backbone, which assists the type IIЈ ␤-turn around glycine. Glycine residues are often found at turning points in protein structures, as the lack of side chain enables a tight turn to be made, but such turns are only successful if additional non-covalent interactions are present, as promoted here by Arg. This combination of Gly 6 and Arg 8 in mammalian GnRH produces a peptide configured with high affinity for the mammalian receptor. Substitution with D-Ala 6 or Gly 6 in Ciona I increases the affinity for the human receptor, since the peptide can now form a tighter turn. However, the presence of Ser in position eight rather than Arg precludes the very high affinity exhibited by the mammalian form at the mammalian receptor, due in part to the additional stabilization of the compact structure by Arg and the interaction of Arg with an acidic residue in extracellular loop 3 of the GnRH receptor, which contributes to configuration of the ligand at mammalian receptors (18,19).
When the chiral amino acid (Ala) in position six of Ciona I GnRH was substituted with the achiral glycine or with D-Ala, which enhances the type IIЈ ␤-turn conformation, there was a marked increase in the binding affinity of the peptides at the vertebrate GnRH receptors. Moreover, the gas-phase studies indicated that these peptides assume a more compact configuration, whereas in solvent calculations they become more flexible. Thus, there is a correlation between biological activity and the tighter conformation of the peptides at the vertebrate receptors, whereas this situation does not pertain at the protochordate receptor.
These studies suggest that the early-evolved GnRH receptors of protochordates (and probably jawless fish) bind GnRH in a more extended configuration and that the subsequent evolution of the receptors in jawed vertebrates required a more compact configuration of GnRH for binding. The conservation of the N-and C-terminal domains of GnRHs in invertebrates and vertebrates (protochordates, jawless fish, jawed fish, amphibians, birds, and mammals) nevertheless indicates that these domains are functionally important for binding and activating the GnRH receptor. As mentioned earlier, most of the binding sites for the N-and C-terminal GnRH residues in the human and vertebrate receptors are present in the tunicate GnRH receptor. However, the non-requirement for an achiral amino acid in position six in protochordates and jawless fish, but the requirement for an achiral Gly in higher vertebrates indicates that, although the N and C domains of GnRH interact through the same or similar sites in the receptor, these are more closely positioned in the higher vertebrates (which require the folded conformation) than in protochordates and jawless fish (which bind less configured forms of the peptide equally well and with lower affinity). It is possible, therefore, that the protochordate GnRH receptor is also less compact due to fewer transmembrane domain interactions. A major transmembrane domain interaction in GPCRs is between Asp 2.50 in TM2 and Asn 7.50 in TM7. These are reciprocally mutated to Asn and Asp in the human GnRH receptor, and these residues are critical for receptor function (36). In the Ciona GnRH receptors Asn is present in TM2, but His replaces Asp in TM7. In addition to this TM interaction, we recently identified Met 227 (TM5), Phe 272 (TM6), Phe 276 (TM6), and Ile 322 (TM7) as interacting residues in the human GnRH receptor that alter ligand affinity and selectivity when mutated (37). Phe 276 is conserved between the tunicate and human GnRH receptors. Met 132 is Val, Met 227 is Thr, Phe 272 is Ile, and Ile 322 is Val in the Ciona A receptor. These residue differences may, therefore, contribute to reduced ligand affinity and selectivity in the Ciona A receptor and will be the subject of future studies.
In addition to its direct activation of the gonads in Ciona, GnRH appears to serve a role as a pheromone in another protochordate, Saccoglossus (38), and in a mollusk (Chiton) (31). Tunicate spawning is precisely correlated with light cycles, and light appears to stimulate GnRH secretion (39), which is secreted directly from nerve endings onto gonadal cells. In vertebrates, however, GnRH has been co-opted to serve a more complex neuroendocrine role where external environmental factors (e.g. light) stimulate secretion of GnRH into portal vessels to stimulate secretion of pituitary gonadotropins and ultimately the gonads. The dilution of GnRH in the portal vessels may, thus, have driven the evolution of the more compact GnRH with higher affinity binding. However, some of the earlier-evolved functions seen in protochordates appear to have been retained in vertebrates as GnRH and GnRH receptors are present in the gonads of fish, amphibians, and mammals (5-7), suggesting direct effects on gonads (40).
It is, therefore, evident that the GnRH structure existed very early in evolution and was co-opted in diverse ways to regulate reproduction. During at least 600 million years of evolution the N and C termini of GnRH have been conserved as functional domains for binding and activating cognate receptors to accomplish these functions. However, about 400 million years ago a single substitution of the chiral amino acid in position six of GnRH in jawless fish by the achiral glycine facilitated a type IIЈ ␤-turn conformation of GnRH to allow spatially close interaction of these functional domains of GnRH with its receptor, in contrast to the interaction of more extended GnRH structures with receptors in earlier-evolved species. This notion was supported by studies on receptor binding affinities and IM-MS/molecular modeling conformations of GnRHs with Ala, Gly, and D-Ala in position six that showed a close correlation between binding affinity at vertebrate GnRH receptors and their propensity to form a type IIЈ ␤-turn conformation. Thus, an apparently insignificant substitution of a single amino acid at a site remote from the binding and activation domains of GnRH can have a major effect on the conformation of the ligand and affect its interaction with the receptor binding sites. These findings emphasize the importance of subtle changes in three-dimensional structural evolution of a peptide ligand, which is likely to also pertain to other peptide ligands and their receptors.