Address for correspondence: K. Johan Rosengren, School of Pure and Applied Natural Sciences, University of Kalmar, SE-391 82 Kalmar, Sweden. Voice: +46 480 446152; fax: +46 480 446262. firstname.lastname@example.org
The relaxin peptide hormones are members of the insulin superfamily and share a structural fold that is characterized by two peptide chains which are cross-braced by three disulfide bonds. On this framework, various amino acid side chains are presented, allowing specific interactions with different receptors. The relaxin receptors belong to two unrelated classes of G-protein-coupled receptors, but interestingly they are not selective for a single relaxin peptide. Relaxin-3, which is considered to be an extreme example of the relaxin family, can activate receptors from both classes and in fact interacts to some degree with all four receptors identified to date. To deduce how changes in the primary sequence can fine-tune the overall structure and thus the ability to interact with the various receptors, we have studied a range of relaxin-like peptides using solution nuclear magnetic resonance analysis. Three-dimensional structures of relaxin-3, insulin-like peptide 3 (INSL3), and INSL5 were determined and revealed a number of interesting features. All peptides showed a significant amount of line-broadening in certain regions, in particular around the intra-A-chain disulfide bond, suggesting that despite the disulfide bonds the fold is rather dynamic. Although the peptides share a common structural core there are significant differences, particularly around the termini. The structural data in combination with mutational studies provide valuable insights into the structure–activity relationships of relaxins.
The human relaxin hormone family comprises seven peptides, namely, relaxins 1–3 and insulin-like peptides 3–6 (INSL3 to 6), as shown in Figure 1A. The relaxin peptides are related to insulin and comprise a characteristic structure with two peptide chains, A and B, of about 24 and 27 amino acids, respectively, which are held together by two interchain and one intrachain disulfide bond.1 The crystal structure of human relaxin-2 was resolved in 1991 and revealed a compact fold comprising three helical segments and a short extended region that enclose a hydrophobic core (Fig. 1B).2 Recently a renewed interest in the relaxin structural features and their functional importance was sparked after the identification of the relaxin family peptide receptors 1–4 (RXFP1 to 4), which have been confirmed to be the primary receptors for relaxin-2,3 INSL3,4 relaxin-3,5 and INSL5,6 respectively. Interestingly, while all are G-protein-coupled receptors (GPCRs) they belong to different classes, with RXFP1 and RXFP2 being leucine-rich repeat-containing GPCRs that are characterized by a large N-terminal ligand-binding domain, whereas the smaller receptors RXFP3 and RXFP4 are classic peptide ligand GPCRs and lack this domain. It is also intriguing that despite the differences between the receptors there is no absolute selectivity between the ligands and their receptors (Fig. 1C). The most striking example of this is relaxin-3, which activates with high-potency RXFP1, RXFP3, and RXFP4 and which will also interact with RXFP2, albeit with considerably lower potency.7,8 These observations raise questions regarding what structural features control the receptor–ligand interaction.
Relaxin-2 has long been considered a hormone mainly associated with pregnancy and is responsible for the tissue remodeling needed for pregnancy maintenance and delivery as well as preparing the mammary gland for lactation.9 However, it is becoming clear that relaxin-2 and other relaxins have a wide range of important roles in mammalian biology. Relaxin-2 is involved in collagen regulation and acts as a vasodilator in the heart and kidney.10,11 Furthermore, there is increased evidence for involvement of relaxin-2 in cancer, with overexpression being seen in breast and prostate cancer.12 INSL3 is also important for reproduction and is primarily expressed in Leydig cells of the testis and involved in fetal development of the testis.13 In contrast relaxin-3 is found in high levels in the brain and has been shown to be involved in neurosignaling, modulating stress responses and food intake.14,15 The biological significance of relaxin-1 and INSL4 to 6 is currently not known, and no receptors for INSL4 or INSL6 have been identified.
To further characterize the biological functions of various relaxins and for potential pharmaceutical manipulation of the relaxin-signaling systems, the generation of high-affinity selective agonists and antagonists for each relaxin receptor is highly desirable. An important step towards the development of such analogs is a structural understanding of how the amino acid sequences of the various relaxins influence the overall fold and surface characteristics and thereby give the peptides their abilities to interact with the receptors, and this was the aim of the current study.
Results and Discussion
Over recent years novel insights into the complex biological roles of relaxins and the realization of their therapeutic potential have led to intensified efforts in deducing structure–activity data, and as a result considerable progress in our understanding of their function has been made. Crucial for these studies has been the development of novel protocols for generating relaxins and relaxin analogs by solid-phase peptide synthesis.16,17 The possibility of making synthetic relaxins has allowed efficient probing of important features by generation of chimeric, truncated, and Ala-substituted analogs. Moreover, as described here, over the last few years we have employed nuclear magnetic resonance (NMR) techniques to determine atom resolution three-dimensional solution structures of relaxins and relaxin analogs. Detailed structural data are of great importance for making predictions on the importance of individual amino acids for structure and function and also for analyzing the effects of modifications on the overall structure, thereby significantly improving the interpretation of functional data on mutants.
The A and B chains of relaxin-3, INSL3, and INSL5 were all assembled by fluorenylmethyloxycarbonyl chloride solid-phase peptide synthesis. In most cases a strategy for directed formation of the disulfide bonds was used to ensure a high yield of the correct final product and to minimize purification efforts. This strategy involved using different Cys side-chain-protecting groups that could be selectively removed in a stepwise fashion to allow the formation of one disulfide bond at a time. Oxidized products were purified by reverse-phase high-performance liquid chromatography, and a high purity was confirmed by mass spectrometry.
Recombinant relaxin-2 and synthetic relaxin-3, INSL3, and INSL5 were all subjected to extensive characterization using two-dimensional homonuclear NMR analysis. All peptides gave good spectral data characteristic of folded structures, but significant resonance broadening in some regions was also seen in all peptides, indicating that the fold is flexible. Resonance assignments were achieved by sequential assignment strategies, which allowed close-to-complete assignments of backbone and side chain protons for all peptides. The only exceptions were a few resonances, in particular in INSL3, for which the resonance signals were broadened beyond detection. Figure 2 shows a typical nuclear Overhauser enhancement spectroscopy (NOESY) spectrum recorded for relaxin-2 and illustrates the sequential connectivities used for resonance assignments. Secondary Hα shifts, i.e., the difference between the observed chemical shifts and the chemical shifts expected from a peptide adopting a random coil structure, is a sensitive indicator of secondary structure. Stretches of positive shifts are indicative of β-sheet structure, whereas sequences of negative shifts are consistent with a helical conformation. From the analysis of the relaxin secondary shifts (Fig. 3) it is clear that all peptides have a similar fold, with helical segments as well as short extended regions. Smaller deviations are seen around the termini for most of the peptides, indicating that these regions are less structured.
From the NMR data, structural restraints were derived, including interproton distance restraints based on NOESY cross-peaks, dihedral angle restraints based on coupling constants, and hydrogen bond restraints based on amide exchange and temperature coefficient data. These were combined with knowledge about covalent geometry and used to calculate the 3D structures by using simulated annealing followed by refinement and energy minimization in explicit solvent using the program CNS. For each peptide a family of 50 structures was calculated and the 20 lowest energy models were chosen to represent the solution structure of the hormone.
Structural Features of the Relaxin Framework
As expected from their conserved size, two-chain structure, and disulfide connectivity, the relaxin family peptides share a common overall fold. The solution structures of relaxin-318 and INSL319 and a preliminary structure of INSL5 are shown in Figure 4. The main characteristics of the structure are two helical segments in the A chain that are separated by an extended region, resulting in a U-shaped arrangement, and a helical segment in the B chain that lies across the face of the U. The B chain also contains an extended region forming β-sheet interactions with the extended region on the A chain. It is clear that the fold is mainly stabilized by the three disulfide bonds together with a hydrophobic core that includes primarily the hydrophobic side chains of LeuA3, LeuA6, CysA10, CysA15, IleA20, LeuA23, LeuB9, PheB14, IleB15, and ValB18 in relaxin-3 and the equivalent residues in INSL3 and INSL5.
Interestingly, from the 3D structure it is clear that in all peptides protons experiencing resonance broadening are proximate to the CysA10–CysA15 intrachain disulfide bond. Thus, the likely explanation for this broadening is that rapid rearrangement of the conformation of the disulfide bond results in small structural changes affecting the environment of the protons in this region. In relaxin-3 and INSL3 the aromatic side chain of PheB14 packs against the disulfide bond, which means that a small change in structure could result in a large change in chemical shift of proximal protons due to the ring current effect, which would explain the severe broadening.
A comparison of the available relaxin structures, which is presented in Figure 5, reveals that the main structural differences between the peptides are around the chain termini. The A chain N-terminus in relaxin-2 and relaxin-3 is fully ordered and forms a helix from residue 2 onwards. In contrast, the structures of INSL3 and INSL5 are less ordered in this region. The first five residues of INSL3 are fully flexible and do not interact with any other part of the molecule, which is likely due to a helix-breaking proline at position 5. In contrast, the A chain C-terminus is very similar for all peptides, probably due to the disulfide bond involving CysA24, which locks the conformation in relation to the rest of the molecule. INSL3 has two additional residues at the C-terminus, but none of these appears to have any interactions with the rest of the molecule. The B chain N-terminus appears to be fairly disordered for all peptides. The crystal structure of relaxin-2 indicates a 310 helical turn, but this is likely due to crystal packing.2 The B chain helix represents the main receptor-binding interface and comprises the conserved Arg-x-x-x-Arg-x-x-Ile/Val motif that has been shown to be crucial for activity on RXFP1. Similar motifs with additional residues in close proximity, including IleB15 and PheB20 in relaxin-3 and IleB15 and TyrB20 in INSL5, have been identified as important for interactions with the smaller receptors RXFP3 and RXFP4.20 Interestingly, residues close to the C-terminal tail have also been identified as crucial for receptor interactions in both relaxin-3 and INSL3. ArgB26 and TrpB27 are important for the activation of RXFP3, whereas TrpB27 is crucial for the primary binding of INSL3 to RXFP2. The derived structures reveal differences in the conformation of the C-terminal tail, as the helix in relaxin-2 and INSL5 continues, thus extending away from the core of the peptide. In contrast, in relaxin-3 and INSL3 the tail has contacts with the hydrophobic core. It is possible that the tail may need to be flexible to be able to adopt a correct conformation once it interacts with the receptor.
All relaxins studied so far share a common overall fold but have significant differences, in particular around the termini. Despite the three disulfide bonds the structures appear to be dynamic. Flexibility may be important for function and may allow each ligand to adopt an optimal conformation for interacting with its receptor. It is interesting to note that the B chain C-terminus, which is important for binding of INSL3 to RXFP2 and for activation of RXFP3 and RXFP4 by relaxin-3, appears to have different preferred conformations in different relaxins. It is possible that differences in the degree of flexibility and preferred orientation may contribute to the differences in receptor selectivity.
We thank Dr. E. Rinderknecht for providing a sample of recombinant relaxin-2. KJR thanks the University of Kalmar and Åke Wiberg's Foundation for financial support. RAB and JDW received support from NHMRC Australia. DJC is an ARC Professorial Fellow.