The relaxin-like peptide family includes: relaxin-1, relaxin-2, relaxin-3, and the insulin-like (INSL) peptides, INSL3, INSL4, INSL5 and INSL6. All share high structural similarity with insulin due to the presence of six cysteine residues, which confer two inter-chain and one intra-chain disulfide bonds. Thus, it was postulated that relaxin and insulin had derived from a common ancestral gene and were therefore grouped as the insulin superfamily, which later included insulin-like growth factors I and II (IGF-1 and -2) (reviewed in 1).

Despite less than 50% predicted sequence similarity across members of the insulin superfamily, primary structural determinants are retained, resulting in the similar tertiary structures of relaxin and insulin (reviewed in 1). The structures of insulin, relaxin 1, relaxin-3 2 and INSL3 3 are formed by the cleavage of the pro-hormone peptide into three chains (A, B and C), removal of the C chain and the formation of three disulfide bridges between six invariant cysteine residues found on the A and B chains, to produce an active protein. Based on primary sequence similarity the native structures of human relaxin-1 (H1 relaxin), INSL4, 5 and 6 should be similar, but to date this has not been confirmed.

Various studies have highlighted the importance of a [Arg-X-X-X-Arg-X-X-Ile] motif in the relaxin B chain for interaction with the relaxin receptor and biological activity, which is dependent upon the presence of this motif 4. Coupled with the insulin-like cysteine bond pattern, the presence of this motif is used to distinguish relaxin sequences. Hence, three relaxin genes are present in humans: relaxin-1 is found only in humans and the great apes, its expression is limited to the decidua, placenta and prostate 5; relaxin-2 is the major circulating form of relaxin in the human 6 and the functional equivalent to the relaxin-1 in all non-primates; while relaxin-3 was only recently discovered and shows brain specific expression 2. Throughout this paper relaxin will be used to refer to relaxin-2 in humans and great apes, and the equivalent relaxin-1 in all other mammals (table 1).

Relaxin has been well characterized in a reproductive context.. It is a product of the ovary and/or placenta in most species studied, and has various roles in pregnancy and parturition, which differ between species 7. Recent advances have revealed relaxin to be a multifunctional hormone, with numerous non-reproductive roles (reviewed in 1). Of the other relaxin-like peptides, INSL3, or relaxin-like factor (RLF), is closely related to relaxin and is critical for testis descent 89. Early placenta insulin-like peptide (EPIL), placentin or insulin-like 4 (INSL4) is primate specific and likely to have diverged from a common relaxin ancestor before the duplication that led to the two relaxin genes seen in humans and great apes 10. Insulin-like 5 and 6 (INSL5 and INSL6) were identified by searching EST databases with the cysteine motif conserved in all insulin-like peptides 1112. Both INSL5 and 6 peptides show higher sequence similarity to relaxin rather than insulin and are functionally uncharacterised.

Until recently, receptors for all peptides of the relaxin family were unknown. Given the high degree of structural similarity between relaxin and insulin, it had been believed that the relaxin receptor would also be a tyrosine kinase receptor, similar to the insulin receptor. However, it was finally demonstrated in 2002, that relaxin activated two previously orphan leucine rich repeat containing heterotrimeric guanine nucleotide binding protein-coupled receptors (GPCR), LGR7 and LGR8 13. Further studies have shown that LGR7 is the relaxin receptor 13, although it also specifically interacts with relaxin-3 14, and that LGR8 is the INSL3 receptor 1516. Even more recently, another two GPCRs which are activated by relaxin-3, the somatostatin- and angiotensin- like peptide receptor (SALPR or GPCR135) and GPCR142 have been identified 1718. More recently GPCR142 has been shown to be the receptor for INSL5, 19 while GPCR135 appears to be the specific receptor for relaxin-3. Surprisingly, while LGR7, LGR8, GPCR135 and GPCR142 are all Type I GPCRs, they are from different branches within this family and are only very distantly related 171820. No other relaxin-like peptides have been shown to interact with these receptors 1314161718.

Relaxin and INSL3 have primarily been of interest as hormones of pregnancy and reproduction; therefore it was assumed that the largely uncharacterised INSL4, 5 and 6 would share similar functions. However the discovery of the brain specific relaxin-3, and the widespread expression of INSL5 and its receptor GPCR142 have resulted in a re-evaluation of these assumptions and raised the need for a new approach to investigating these peptides.

Relaxin evolution has confounded researchers for decades. High sequence variability in relaxins across closely related species is a well-known feature of this peptide, however startling similarities have been observed between very distant species such as pigs and whales 21. Other studies have reported the existence of an invertebrate relaxin, a hormone with "relaxin-like" properties has been described in protozoa (T. pyriformis) 22, ascidians (H. momus) 23 and tunicates (C. intestinalis) 23. A cDNA and peptide sequence with almost 100% similarity to porcine relaxin was isolated from C. intestinalis 24. Contrastingly, numerous efforts to identify and sequence a relaxin ortholog in bovine and other ruminants have been unsuccessful. A likely non-functional single copy relaxin-like gene has been found in the ovine 25. Despite this controversy very few phylogenetic studies of relaxin, or the other relaxin-like peptides, have been reported. An in-depth analysis analysing only relaxin-1, -2 and INSL3 in several primates has been performed 26 and a more recent paper has discussed the evolution of the family without including a detailed phylogenetic analysis 27.

The increasing availability of genomic data has provided an opportunity to clarify the area of relaxin evolution using phylogenetic analysis. The relaxin-like peptide family phylogeny shows relaxin-3 is likely to be the ancestral relaxin, emerging prior to the divergence of fish, presumably with a function in the brain. Multiple relaxin-3 sequences are present in the fish and frogs, while the phylogeny also suggests that a relaxin with reproductive functions is present in these lineages. However, relaxin has been lost in the chicken and its genome instead contains two relaxin-3-like sequences. Evolutionary rate analysis shows positive Darwinian selection to be driving the most recent members of the family, INSL6, INSL4 and relaxin-1.

While relaxin evolution has been the centre of much controversy (relaxin is often cited as a gene that conflicts with the Darwinian theory of evolution 24323334), this report is the first attempt to describe the evolutionary history of the whole relaxin-like peptide family from a phylogenetic perspective. Previous studies have concentrated on the primate relaxins and relaxin-like factors 26, or not included detailed phylogenetic analyses 27. We have sought to expand upon these by incorporating sequences identified in all the available completed genomes with a subset of cloned relaxin-like sequences, particularly those from non-mammalian species.

None of the phylogenetic tree construction programs used was able to completely resolve the evolution of the relaxin-like peptide family. This is likely due to variable divergence across the family and the short sequence length 35. Incorporating results from the MP and NJ methods suggested positions for several branches that were unresolved after ML analysis. Minimizing incongruence between the gene and species trees by reducing the number of assumed duplications in the reconciled tree also provided a method to infer the evolutionary history of this family.

Similarly to previously published results, searches of available genomic and EST data failed to identify any novel members of the relaxin-like peptide family 28. Given the stringent and well-described insulin family signature that revolves around the invariant cysteine residues that confer the insulin-like structure seen across the superfamily, we find it improbable that any novel relaxin or insulin-like sequences will be identified.

The presence of an invertebrate relaxin has been of speculation since 1983 when relaxin-like activity was first detected in the protozoa, T. pyriformis 22. Similar activity was reported in H. momus 23 and in C. intestinalis, where a cDNA sequence almost identical to pig relaxin was found 24. However, our searches of all completed invertebrate genomes (including C. intestinalis) failed to identify any relaxin-like sequences, including the published sequence. Multiple insulin-like peptides have been found in several invertebrates, including: Bombyxi mori (silkworm) 36, D. melanogaster 37 and C. elegans 38. As these sequences lack the relaxin-specific motif, and show no homology to other relaxin family peptides, they are not considered part of the relaxin subfamily and therefore have not been included in these analyses. Much of the controversy surrounding relaxin evolution concerns the identification of an invertebrate relaxin sequence (a cDNA sequence from Ciona intestinalis) almost identical to pig relaxin (Georges and Schwabe, 1999). Completion of the C. intestinalis and other invertebrate genomes has allowed us to conclude that there is not a relaxin-like sequence in any invertebrate sequenced to date. If an invertebrate relaxin does exist, it does not contain the relaxin-specific motif characterized in vertebrates.

A hallmark of relaxin sequences is their high variability, even amongst closely related species. Relaxin-like peptide sequences isolated from two whales are almost identical to porcine relaxin 21, however as these sequences were derived from amino acid sequencing, without nucleotide or and genomic sequence available, they have not been able to be included in these phylogenetic analyses.

The presence of a functional relaxin in the ruminant lineage has yet to be confirmed 25. More genomic data is required to confirm the presence of a non-functional relaxin gene sequence in the bovine, similar to that observed in the ovine 25. Searches of the preliminary bovine genome assembly have failed to find a relaxin gene. Interestingly, a relaxin sequence has been identified in the camel 39 and relaxin expression found in the closely related llama and alpaca 40. While classified as a ruminant, Camelidae have a unique reproductive anatomy and physiology 41. A bovine EST (BI682322) with high similarity to exon 2 of human relaxin-3 was identified. Confirmation of the presence of relaxin and relaxin-3 orthologs in ruminants awaits further sequencing of the bovine and ovine genomes.

The presence of an avian relaxin has also been of speculation. While relaxin-like activity has been reported in the chicken 42, an avian relaxin-like peptide or gene has not been identified. While two relaxin-3-like genes were identified in the nearly completed chicken genome, no avian relaxin gene was found. As no other relaxin-like genes were found, the reported relaxin activity may be due to one of the relaxin-3-like genes.

The phylogeny of the relaxin-like peptide family indicates relaxin-3 is the ancestral relaxin, appearing prior to the divergence of teleosts. The finding of multiple relaxin-3-like genes in the fugu fish and zebrafish suggests multiple lineage-specific duplications of a single relaxin-3-like ancestor have occurred in fish 27. However, the possibility the other mammal specific relaxin-like peptides emerged earlier before being lost in the teleost can not be excluded 27. We find it more likely that these duplications, and the resulting multiple relaxin-3-like genes, are fish specific and due to genome wide duplications hypothesized to have occurred during fish evolution 43. Phylogenetic analyses show multiple fish homologs of both the mammalian relaxin-3 and INSL5 genes, meaning that two relaxin-3-like genes existed prior to the genome duplication event proposed to have occurred in the teleost ancestor. The putative fish relaxin homolog was either, present in the teleost ancestor, duplicated and the second copy lost or emerged shortly after or, as a result of, the genome-wide duplication event.

While termed relaxin-3-like based on sequence similarity, phylogenetic analysis indicates that several non-mammalian sequences (OmRLX3, DrRLX3b, DrRLX3d, TrRLX3d, TrRLX3e and GgRLX3b) could be INSL5 homologs. None of the sequences found in the complete X. tropicalis genome were placed in this group, while there are members present in the more ancient fish lineage and the younger avian lineage. It is possible that this gene has either been lost, or remains unidentified, in the X. tropicalis genome. A sequence with similarity only to the B chain of relaxin-3 was also found, but a corresponding A chain match was not, however, there is a gap in the genome assembly upstream which might contain the missing domain. Future assemblies of the Xenopus genome should resolve this issue. These results suggest that INSL5 could have emerged during teleost evolution, far earlier than previously believed. Unlike the mammal-specific relaxin-like genes, which are clustered together (on chromosome 9 in the human and chromosome 19 in the mouse), INSL5 is localized independently (chromosome 1 in the human and chromosome 4 in the mouse). These findings are of particular interest in the analysis of INSL5, which is still functionally uncharacterised.

All the potential non-mammalian INSL5 homologs retain the relaxin-specific B chain [RxxxRxxI/V] motif, hence would be capable of interacting with the relaxin receptor, LGR7, and thus functionally classified as a relaxin. Recent studies have shown INSL5 is a high affinity ligand for GPCR142 but not GPCR135, LGR7 or LGR8 19. As the residues required for interaction with GPCR135 and GPCR142 are not known, it is unknown whether the non-mammalian INSL5 homologs would interact with GPCR142, GPCR135 and/or LGR7.

Phylogenetic results from this study suggest the presence of a relaxin homolog in fish and frogs, although not in the chicken. Relaxin sequences have previously been isolated and peptide sequenced from either the ovaries or testes of the edible frog 30, little skate (Raja erinacea) 44, spiny dogfish (Squalus acanthias) 45, Atlantic stingray (Dasyatis sabina) 46 and the sand tiger shark (Odontaspis taurus) 47. While having high similarity with relaxin-3, these sequences are not relaxin-3 homologs (as the B chain of the stingray sequence is lacking the relaxin-specific motif, it is not a functional relaxin 46 and has not been considered further). Based on the expression of all these genes in reproductive organs such as the testes and ovaries, and the failure to find the R. esculenta gene expressed in the brain using northern blot analysis 30, we believe these to be among the first relaxin peptides with a reproductive function. Based on the similarity with relaxin-3 observed in these sequences, the ancestral relaxin homolog, and its new reproductive function, is likely to have emerged prior to the divergence of teleosts. A complete picture of relaxin-like peptides present in non-mammalian genomes will be invaluable in understanding the evolution of relaxin from neuropeptide to reproductive hormone.

The ancestral RLN3 gene is under very strong purifying selection, highlighting the importance of its highly conserved function, likely to be in the brain 2. As high divergence is a hallmark of relaxin sequences, it is somewhat unsurprising that RLN2 is under only weak purifying selection. We suggest that this lack of selective pressure has contributed to the high sequence divergence seen between many relaxins (e.g. human and mouse) and the differences in relaxin's functions observed across mammals.

Information about the selective constraints placed upon these peptides, can provide valuable insight into the nature of interactions with their receptors. Based on selection pressures we can conclude that the interactions between relaxin-3 and GPCR135, INSL5 and GPCR142 are very specific, while the binding of relaxin to LGR7 is much looser. In this context the cross-reactivity seen between LGR7 and INSL3 or H1 relaxin, which are both similar to relaxin in sequence but especially in structure, is understandable, as is the lack of binding between GPCR135 and GPCR142 with any other relaxin-like peptide. Unexpectedly, synonymous and nonsynonymous substitution rate estimates for RLN1 and INSL6 show these to be under positive selection. Positive selection is often difficult to observe using pairwise comparisons that average over the whole length of a sequence, making these results even more striking. While pairwise comparisons failed to confirm positive selection was acting on INSL4, further statistical tests suggested that positive Darwinian selection acted on several sites in the INSL4 sequence after its emergence. Further analysis will be required to confirm these sites as important in the acquisition of a new receptor and a new function by INSL4, particularly in light of recent studies that question the reliability of ML methods to accurately detect positive selection acting on single sites 484950. We are encouraged that both branch-specific and site-specific ML models find positive selection to be acting on INSL4.

When the B and A domains of each gene were analyzed separately, further differences in selection pressures became apparent. The interaction between relaxin and its receptor has been thought to be primarily mediated through the B chain of the peptide 4, so the finding that selection pressures are stronger on the A chain of relaxin-1, INSL4 and INSL6 was unexpected. We also find it noteworthy that INSL4, INSL6 and relaxin-1 are the most recent members of the family to emerge and all appear to be under the effects of positive Darwinian selection. INSL6 emerged during mammalian development, INSL4 and RLN1 during primate evolution, they remain functionally uncharacterized and INSL4 and INSL6 are without known receptors. The low selection pressure on the B domain and the strong constraints placed on the A domain of INSL4 and INSL6 suggests that, unlike the B chain mediated interaction of relaxin and INSL3 with their receptors, the interaction of these peptides with their receptors could be dependant on the A chain instead. The low d_N/d_S rate observed for INSL5 indicates this peptide to be evolutionary stable and of functional importance. In particular the constraints placed on both A and B chains of INSL5 suggest a well-defined receptor interaction system, while the total absence of these constraints on either chain within relaxin-1 suggests the opposite, that perhaps this peptide is still evolving its function.

We present here a phylogeny for the relaxin-like peptide family. Relaxin has long been used as an example of a gene that conflicts with the Darwinian theory of evolution 24323334. However, we have shown that these can issues can be resolved when studied in the context of the rest of the relaxin-like peptide family, in particular the new, but likely ancestral relaxin, relaxin-3.

We have demonstrated that positive selection has been a driving force in the recent expansion of the relaxin-like peptide family during mammalian evolution. While strong purifying selection has maintained the structural core of these peptides by constraining the insulin superfamily cysteine motif, outside these residues, positive selection has acted after at least three gene duplication events (which generated INSL6, INSL4 and relaxin-1) to allow these new genes to acquire a new receptor and novel functions. Given the known roles of relaxin and INSL3 in reproduction (and the likely similar roles of INSL4 and INSL6 given the specificity of their expression in reproductive tissues) these findings correlate with a general trend towards rapid evolution in several reproduction associated genes 51525354. We anticipate that further analysis of the coevolution of the relaxin-like peptides with their receptors will contribute much towards our understanding of the pleiotropic actions of this family as well as mechanisms involved in the evolution of peptide hormone systems.

TW performed all sequence and phylogenetic analysis and drafted the manuscript, TS participated in phylogenetic analysis, design and coordination of study, GWT participated in the design of the study, and RADB participated in phylogenetic analysis, conceived of the study and participated in its design and coordination.