International Union of Basic and Clinical Pharmacology. XCV. Recent Advances in the Understanding of the Pharmacology and Biological Roles of Relaxin Family Peptide Receptors 1–4, the Receptors for Relaxin Family Peptides

1. Relaxin Family Peptide Agonists.

Although the relaxin peptides show considerable variation across species, alignment of amino acid sequences showed that in addition to the cysteines necessary for the two-chain structure there was a conserved RxxxRxxI/V motif in the B-chain. Examination of the X-ray crystal structure of relaxin indicates that the arginines are outward facing and located on the first and second loop of the α-helix (Eigenbrot et al., 1991) where together with an isoleucine or valine residue they form the receptor binding site for relaxin (Schwabe and Bullesbach, 1994; Tan et al., 2002). Replacement of the arginine, isoleucine, or valine residues in this motif (RxxxRxxI/V) markedly reduces or abolishes activity (Schwabe and Bullesbach, 1994; Tan et al., 2002). The relaxin A-chain shows greater sequence variation than the B-chain across species (Sherwood, 1994). Other than the conserved cysteine residues, G14 is also highly conserved and necessary for chain flexibility and structure (Bullesbach and Schwabe, 1994). Truncation of the A-chain of human relaxin produces peptides that progressively lose the ability to bind to and activate RXFP1 and RXFP2 (Hossain et al., 2008) (Fig. 3). Studies using targeted point mutations within the A-chain show an influence of the A-chain on receptor binding and activation. T16A and K17A mutations had little effect on the cAMP response to RXFP1 activation but enhanced cAMP responses after activation of RXFP2; conversely, the R22A and F23A mutations markedly reduced the activity of human relaxin at RXFP2 without reducing binding or activity at RXFP1 (Park et al., 2008); mutations S19A and L20A are also known to reduce peptide binding and activity at RXFP1 (Bullesbach and Schwabe, 1994) (Fig. 3). More recent detailed studies of the key A-chain residues (Chan et al., 2012) show that no single amino acid drives the interactions between RXFP1 and human relaxin, although Y3, L20, and F23 all have some role. These studies also identify differences in the mechanisms by which relaxin binds to and activates RXFP1 compared with RXFP2, leading to the development of a potent and selective peptide agonist for RXFP1, H2:A(4–24) (F23A) that has similar potency and biologic activity to human relaxin but no significant activity at RXFP2 (Chan et al., 2012) (Table 1). The contention that human relaxin binds to RXFP1 and RXFP2 by distinct mechanisms is further supported by the species-specific nature of these structure activity relationships: neither mouse nor rat relaxin is able to bind to or activate RXFP2 (Scott et al., 2005b), and although human relaxin-3 binds with good affinity to RXFP1, it has only a poor affinity for RXFP2 (Bathgate et al., 2006b). The importance of the peptide A-chain is also emphasized by the structure activity relationships displayed by relaxin-3. Replacement of the relaxin-3 A-chain with that of INSL5 produces a chimeric peptide (R3/I5) that no longer binds to or activates RXFP1 but has unchanged activity at RXFP3 (Liu et al., 2005a) (Fig. 3). A similar profile is attained after truncation of the relaxin-3 A-chain (Hossain et al., 2008).

Examination of the role of the B-chain in activation of RXFP1 demonstrates that it is possible to truncate both ends while still retaining binding and functional activity (Hossain et al., 2011). Peptides with a reduced B-chain [H2:(B7–24)] or reduced A- and B-chains [H2:(A4–24) (B7–24)] show some reduction in binding activity and ability to generate cAMP but have similar antifibrotic activity to human relaxin and less activity at RXFP2 (Hossain et al., 2011) (Table 1).

2. C1q-Tumor Necrosis Factor–Related Protein 8: An Additional Relaxin Family Peptide Receptor 1 Ligand?

Short linear peptides derived from a naturally occurring protein containing a collagen-like repeat have been reported to act at RXFP1 (Shemesh et al., 2009). Although the effects produced by the peptides CGEN-25009 and CGEN-25010 in several systems were extremely variable and the effects of human relaxin in these systems unusual (Shemesh et al., 2008, 2009), there is some evidence to suggest relaxin-like activity of these peptides in THP-1 cells and in a fibrosis model (Pini et al., 2010). In the latter study, CGEN-25009 and human relaxin increased cAMP, cGMP, and nitrite and decreased collagen deposition and increased matrix metalloproteinase-2 (MMP-2) activity in human dermal fibroblasts (Pini et al., 2010). More recent studies with these peptides and the precursor protein C1q-tumor necrosis factor–related protein 8 (CTRP8) demonstrated activation of RXFP1 (Glogowska et al., 2013) with cAMP production and a PI3K-mediated promigratory phenotype in glioblastoma cell lines and primary cells. Coimmunoprecipitation studies demonstrated a direct interaction between human CTRP8 and RXFP1. Although these studies suggest that CTRP8 or peptide fragments are able to activate RXFP1, it remains to be seen whether they are native ligands and/or whether activation of RXFP1 by CTRP8 occurs specifically in glioblastomas.

3. Small-Molecular-Weight Agonists.

The recent success of the phase III clinical trial of relaxin for acute heart failure (see below) encouraged a screening campaign to identify small molecule agonists of RXFP1. By using HEK293 cells stably expressing RXFP1, 365,677 compounds were screened for their ability to cause cAMP accumulation. Both of the hits obtained contained a 2-acetamido-N-phenylbenzamide structure that was used as a basis for medicinal chemistry optimization (Chen et al., 2013; http://www.ncbi.nlm.nih.gov/books/NBK153218/). Of these compounds, a number showed reasonable potency for cAMP generation in both HEK293-RXFP1 and THP-1 cells (that endogenously express RXFP1) and had good specificity for RXFP1 versus RXFP2 (Xiao et al., 2013) (Table 1). Several of the compounds increased vascular endothelial growth factor (VEGF) gene expression in THP-1 cells with similar efficacy to relaxin, and compound 8 [ML290; 2-isopropoxy-N-(2-(3-(trifluoromethylsulfonyl)phenylcarbamoyl)phenyl)benzamide] was also shown to increase cellular impedance in a label-free system albeit with at least 500-fold lower potency than relaxin (Xiao et al., 2013). In addition, comparison of the binding mode of action of ML290 with that of relaxin suggests that the small-molecule agonist acts as an allosteric agonist at RXFP1 (Xiao et al., 2013; Christopoulos et al., 2014). Relaxin binds with high affinity to the LRR region of RXFP1 followed by a lower affinity interaction with ECL2 (Sudo et al., 2003; Halls et al., 2005a,b). This triggers an interaction involving the LDLa module that is essential for signaling (Scott et al., 2006; Kern et al., 2007; Kong et al., 2013). Mutations of the calcium binding asparagine within the LDLa module (D58E), produce a receptor where the cAMP response to relaxin is completely abolished but that to ML290 (compound 8) is unaffected (Xiao et al., 2013). Because, unlike the human receptor, mouse RXFP1 does not respond to ML290, chimeric constructs of human and mouse RXFP1 were used to identify the region of RXFP1 that interacts with ML290. Examination of the chimeras narrowed the region of interest down to ECL3, and two pairs of amino acids divergent between human and mouse RXFP1 were identified. The substitution of the N-terminal IL within mouse ECL3 to the VV of human RXFP1 did not rescue function (Xiao et al., 2013). C-terminal substitution of GT within human RXFP1 to the DS of mouse RXFP1 produced a chimeric human RXFP1 that did not respond to ML290. The converse substitution of DS to GT in mouse RXFP1 created a partially active receptor. The mouse receptor with a double substitution of the mouse to the human residues, IL to VV and DS to GT, produced a fully active mouse RXFP1 that responded to ML290 in a similar fashion to the human receptor (Xiao et al., 2013). The 2-acetamido-N-phenylbenzamide derivatives represent the first synthetic small-molecule agonists at RXFP1. They show good potency, selectivity, and a number of relaxin-like functions in cell-based assays. They appear to behave as allosteric agonists acting at the ECL3 loop of RXFP1 (Fig. 1). It remains to be demonstrated whether the profile of activity of ML290 and similar compounds matches that of relaxin in a variety of cellular and animal disease models.

4. Relaxin Family Peptide Receptor 1 Antagonists.

In contrast to the other members of the RXFP receptor family there have been few examples of peptides with antagonist properties at RXFP1. Peptide AT-001 (Neschadim et al., 2014) contains mutations of the B-chain residues R13 and R17 [that are essential for relaxin-like activity (Büllesbach et al., 1992)] to K13 and K17 (B-R13/17K relaxin). Initially, antagonism was demonstrated by lentiviral expression of the B-R13/17K relaxin, which decreased prostate cancer xenograft growth (Silvertown et al., 2007). In addition, medium from cells infected with the lentivirus showed human relaxin immunoreactivity and antagonized cAMP production from THP-1 cells in response to relaxin (Silvertown et al., 2007). However, a subsequent synthesis of the B-R13/17K peptide showed that although it competed for relaxin binding, the peptide was actually a partial agonist for cAMP production in cells overexpressing RXFP1 (Hossain et al., 2010; Neschadim et al., 2014) (Table 1). In cells expressing RXFP1 at lower (more physiologic) levels, B-R13/17K relaxin displayed functional antagonism, inhibited MCF-7 cell invasion, and prevented the inhibitory effect of relaxin on renal myofibroblast differentiation (Hossain et al., 2010). In androgen-independent PC3 prostate cancer xenografts, B-R13/17K suppresses growth and has antiangiogenic effects and synergizes with the chemotherapeutic agent docetaxel (Neschadim et al., 2014). The activity of B-R13/17K at other RXFP receptors has not been examined.

The detailed understanding of the mode of interaction of peptide ligands at RXFP1 has led to the development of potent and selective agonists and also partial agonists that have antagonist properties in systems with low levels of receptor expression. An exciting recent development is the small molecular weight allosteric agonists, typified by ML290, that are currently being characterized in terms of their ability to activate the full repertoire of RXFP1 signaling.

1. Relaxin Family Peptide Agonists.

INSL3 is the cognate ligand for RXFP2 that interacts with RXFP1 only at extremely high concentrations (Kumagai et al., 2002; Halls et al., 2005b), suggesting that it does not bind to or activate RXFP1 in a physiologic setting. In contrast, the cognate ligand for RXFP1, human relaxin, potently activates RXFP2, although the mode of interaction of both human relaxin and INSL3 with RXFP2 shows significant differences compared with the interaction of human relaxin with RXFP1 (see section II.B.1) (Table 2). Mutated INSL3 peptides demonstrate the utilization of distinct amino acids in the B-chain to bind to and activate RXFP2. W27 within INSL3 is critical for RXFP2 binding and activation (Bullesbach and Schwabe, 2004, 2005a), although other single amino acid replacements with either alanine or valine suggest that there are additional influences within the B-chain. Although individual substitutions only slightly reduce INSL3 binding to RXFP2, the combined mutation of H12A, R16A, V19A, R20A, and W27A abolishes binding (Rosengren et al., 2006b).

Truncation of the A-chain of human relaxin results in a peptide that progressively loses the ability to bind to and activate both RXFP1 and RXFP2 (Hossain et al., 2008), possibly because of a loss of structure in the B-chain. However, truncation of the A-chain of INSL3 completely abolishes activation with no effect on binding (Bullesbach and Schwabe, 2005a) (Table 2). Additionally, targeted point mutations within the A-chain selectively alter the receptor binding and activation profile: the T16A and K17A mutations do not markedly alter cAMP responses to RXFP1 activation but enhance cAMP responses at RXFP2; the R22A and F23A mutations reduce human relaxin activity at RXFP2 but not at RXFP1 (Park et al., 2008). The double S19A and L20A mutations reduce peptide binding and activity at RXFP1, again suggesting a role of the A-chain in receptor binding and activation (Bullesbach and Schwabe, 1994) (Table 2). Differences in the mechanism of binding and activation between RXFP1 and RXFP2 are further supported by the species-specific nature of these structure activity relationships; thus neither mouse nor rat relaxin can bind to or activate human, rat, or mouse RXFP2 (Scott et al., 2005b), and human relaxin-3 has only a poor affinity for RXFP2 (Bathgate et al., 2006b). All of these studies highlight the importance of both the A- and B-chains for INSL3 activation of RXFP2 and show that the mechanism of action is different from the relaxin-RXFP1 interaction (Hossain et al., 2008).

2. Relaxin Family Peptide Antagonists.

In contrast to RXFP1, there are numerous peptide antagonists of RXFP2 due to the difference in the mode of activation of RXFP2 by INSL3. Thus deletion of 10 residues from the N terminus of the INSL3 A-chain produces a peptide that still binds to RXFP2 but no longer increases cAMP accumulation (Bullesbach and Schwabe, 2005a). This truncated peptide is a specific competitive inhibitor of INSL3 at RXFP2 (Bullesbach and Schwabe, 2005a) (Table 2). In addition, deletion of eight N-terminal residues of the B-chain produces an INSL3 peptide that retains binding affinity but loses agonist activity (Bullesbach and Schwabe, 2005a). Substitution of eight A- or B-chain residues with alanine does not affect receptor signaling (Bullesbach and Schwabe, 2005a). Disruption of the intra-A-chain disulfide bond (C10S/C15S; or C10del/C15del) produces peptides that retain RXFP2 binding but do not activate cAMP signaling (Zhang et al., 2010) (Table 2). The INSL3 B-chain alone is also a RXFP2 antagonist, although it only has low affinity for the receptor (Del Borgo et al., 2006; Shabanpoor et al., 2007). Modifications of the single-chain structure result in modest gains in affinity, but larger gains are obtained by linking two B-chains using the native cysteine residues (Shabanpoor et al., 2010) (Table 2). Modifications of this B-chain dimer resulted in antagonists with affinities only slightly lower than INSL3 (Shabanpoor et al., 2011).

Despite the similarity between RXFP1 and RXFP2 and their cognate ligands relaxin and INSL3, the precise mode of peptide activation of the receptors is quite different. This has facilitated the successful development of a number of antagonists for RXFP2.

1. Relaxin Family Peptide Agonists.

Human relaxin-3 has a tertiary structure, determined by solution NMR spectroscopy, that is similar to insulin and other relaxin family peptides (Rosengren et al., 2006a). However, unlike the interaction of human relaxin with RXFP1 that requires both A- and B-chains for binding and activation, the human relaxin-3 B-chain alone can bind to and activate both RXFP3 and RXFP4 (Liu et al., 2003b). Replacement of the human relaxin-3 A-chain with the INSL5 A-chain (R3/I5) does not influence RXFP3 binding or activation but markedly reduces the ability of the peptide to bind to and activate RXFP1 (Liu et al., 2005a) (Table 3). Similarly, truncation of the human relaxin-3 A-chain reduces binding to and activation of RXFP1 but retains full activity at RXFP3 (Hossain et al., 2008). The additional deletion of the A-chain disulfide bond results in the complete loss of RXFP1 binding and activation with little effect on RXFP3 binding and activation (Shabanpoor et al., 2012). Further development of this analog with a B23–27 deletion and addition of an arginine at B23 led to a high-affinity, RXFP3-selective, competitive antagonist (analog 3) (Table 3).

Site-directed mutagenesis of B-chain residues reveals that R8, R16, I5, and F20 are important for human relaxin-3 binding to RXFP3 and RXFP4 (Kuei et al., 2007), with R12 also required for binding to RXFP3 but not RXFP4 (Kuei et al., 2007). R26 and W27 toward the C terminus of the B-chain are required for activation of RXFP3 (Kuei et al., 2007). Truncation of the C terminus of the B-chain to C22 and the serendipitous addition of an additional arginine residue at the N terminus in the recombinant production system, when combined with the A-chain of INSL5, led to the discovery of a high-affinity RXFP3-selective antagonist R3(BΔ23–27)R/I5 (Kuei et al., 2007) (see section II.C.4; Table 3). A single-chain antagonist has since been developed based on the B-chain of R3(BΔ23–27)R/I5 with the cysteine residues mutated to serine (H3 B1–22R) (Haugaard-Kedstrom et al., 2011) (Table 3).

2. Biased Agonists Acting at Relaxin Family Peptide Receptor 3.

Ligand-directed signaling bias is increasingly common in GPCR pharmacology and describes stabilization of distinct receptor confirmations by ligands, resulting in selective activation of downstream signal transduction pathways (Baker and Hill, 2007; Evans et al., 2010; Kenakin and Miller, 2010). Although the original studies suggested that only relaxin-3 and its B-chain bound to and activated RXFP3, it is now clear that several relaxin peptides interact with RXFP3 to activate distinct signaling profiles through different, although sometimes overlapping pathways (Fig. 5). Initial studies showed that human relaxin-3 was selective for RXFP3 in both binding and AC inhibition assays, with no receptor activation by human relaxin or INSL3 (Liu et al., 2003b). However, cross-reactivity with other relaxin peptides was not examined over a wider range of signal transduction pathways, and the sensitivity of inhibitory cAMP assays can be influenced by the degree of activation of AC by forskolin and the time of stimulation (for detailed description see section III.C.2).

3. Allosteric Modulators.

A novel selective allosteric modulator of RXFP3 has been described (Alvarez-Jaimes et al., 2012). 135PAM1 (3-[3,5-bis(trifluoromethyl)phenyl]-1-(3,4-dichlorobenzyl)-1-[2-(5-methoxy-1H-indol-3-yl)ethyl]urea) is a positive allosteric modulator (PAM) that curiously only displays activity with C-terminal amidated relaxin-3 or R3/I5 (probe selectivity). Binding studies conducted in HEK293-RXFP3 cell membranes showed that 135PAM1 does not compete with [¹²⁵I]R3/I5(amide) binding at up to 1 μM but at higher concentrations enhances binding consistent with activity as a PAM. The R3/I5(amide) peptide itself competes for [¹²⁵I]R3/I5(amide) binding in a conventional manner (Alvarez-Jaimes et al., 2012). In HEK293-RXFP3 cells coexpressing Gαqi5, 135PAM1 increased calcium responses to EC₂₀ concentrations of relaxin-3(amide) or R3/I5(amide) but not the free-acid (OH) peptides. Concentration-response curves to the amidated but not the free-acid form of the peptides were shifted in a limiting manner by 135PAM1. Similar specificity of 135PAM1 was also displayed in a reporter gene assay that measured inhibition of CRE activity in cells expressing RXFP3 (Table 3). This is the only published example to date of a small molecule compound with activity at RXFP3. The discovery of 135PAM1 identifies an allosteric site on RXFP3 that can be modulated by small molecules.

4. Relaxin Family Peptide Receptor 3 Antagonists.

The synthetic peptide R3(BΔ23–27)R/I5 that was discovered as a RXFP3 antagonist has been instrumental in defining the physiologic functions of the receptor. Specificity for RXFP3 was initially achieved by synthesizing a chimeric peptide featuring the A-chain of INSL5 in combination with the B-chain of relaxin-3. The RXFP3 agonist properties of this peptide chimera were then markedly reduced by truncation of the relaxin-3 B-chain (Kuei et al., 2007; Hossain et al., 2009). Serendipitously, upon recombinant production of R3(BΔ23–27)/I5 an extra arginine remained at the N terminus due to incomplete peptidase “trimming of the RR from the RRRR furin cleavage domain” to produce the peptide R3(BΔ23–27)R/I5. This peptide binds to RXFP3 and RXFP4 with high affinity and is a potent antagonist at RXFP3 in both rats and in human in vitro systems. Synthetic R3(BΔ23–27)/I5 (lacking R23) has lower affinity at RXFP3 and has weak agonist properties (Hossain et al., 2009), highlighting that the extra R23 creates an additional interaction with RXFP3 (Table 3). Administration of R3(BΔ23–27)R/I5 intracerebroventricularly to rats blocks food intake stimulated by the RXFP3 agonist R3/I5. It recently became apparent that R3(BΔ23–27)R/I5 has a complex antagonist profile and blocks some but not all pathways activated by RXFP3. It also has weak biased agonist properties (see section III.C.2). In addition, studies have revealed that the intra-A-chain disulfide bond in relaxin-3 is not important for interaction with RXFP3 but is necessary for the interaction of relaxin-3 with RXFP1. Analogs without this A-chain disulfide bond together with truncation of 10 amino acids from the N terminus of the A-chain remain potent RXFP3 agonists. If in addition, GGSRW is removed from the B-chain and replaced with R, the peptide becomes a high-affinity antagonist with similar biologic activity to R3(BΔ23–27)R/I5 (Shabanpoor et al., 2012) (Table 3).

Studies with R3(BΔ23–27)R/I5 suggested that the additional RXFP3 affinity provided by the R23 residue might result in the B-chain only R3(BΔ23–27)R having significant affinity for RXFP3 (Table 3). This has been proved experimentally and H3 B1–22R is a high-affinity antagonist of RXFP3 that is far easier to produce than the two-chain peptides (Haugaard-Kedstrom et al., 2011). Importantly this peptide is also specific for RXFP3 and has no activity at RXFP4. The peptide was shown to have efficacy in vivo and blocks increases in feeding produced by intracerebroventricular injection of R3/I5 in rats (Haugaard-Kedstrom et al., 2011). It was subsequently used to demonstrate a role for the endogenous relaxin-3 system in addiction in rats (Ryan et al., 2013b) and in motivated food seeking and consumption in mice (Smith et al., 2014).

The structure-activity relationships for the interaction of relaxin peptides with RXFP3 are now well understood. Unlike RXFP1 and RXFP2, RXFP3 can be activated by peptides comprising only the B-chain, and a number of selective agonists and antagonists have been developed. Both relaxin and some RXFP3 antagonists display biased agonist properties at RXFP3. However, the utility of a small molecule allosteric modulator in vivo is currently limited by probe selectivity.

1. Relaxin Family Peptide Agonists.

INSL5 is the cognate ligand for RXFP4 (Liu et al., 2005b). INSL5 has many of the characteristics of an incretin being secreted from enteroendocrine L cells and regulating insulin secretion and glucose homeostasis (Burnicka-Turek et al., 2012). Synthesis of human INSL5 represents considerable challenges so that some of the characterization of the peptide has been performed with mouse INSL5 that has 71% homology (Belgi et al., 2011). Compared with human INSL5, mouse INSL5 displays an 8-fold increase in affinity in binding assays and a 5-fold greater potency in cAMP inhibition assays, possibly related to the overall greater positive charge associated with the nonconserved residues (Belgi et al., 2011) (Table 4). In addition, the amidated peptides show a considerable reduction in potency compared with the free acids (Belgi et al., 2011). Although relaxin-3 binds both RXFP3 and RXFP4 with high affinity, INSL5 is actually a weak antagonist at RXFP3 (Zhu et al., 2008) (Table 4). Furthermore, although the relaxin-3 B-chain alone was sufficient to bind and activate RXFP3, the INSL5 A- and B-chains alone are inactive at RXFP4 (Belgi et al., 2013).

The relaxin-3 B-chain alone is also a weak agonist at RXFP4 (Liu et al., 2003a, 2005a) (Table 4). Up to seven residues of the N terminus of the B-chain of relaxin-3 can be deleted with little effect on binding or activation of RXFP4 (Kuei et al., 2007; Hossain and Wade, 2010), whereas the α-helical region of the B-chain is important for activity at RXFP4 (Hossain and Wade, 2010). R12 and R16 are important for binding to both RXFP1 and RXFP3, whereas only R16 is important for RXFP4 binding (Kuei et al., 2007). A critical residue for RXFP4 and RXFP3 binding appears to be F20, and mutations of this residue are associated with a marked loss of affinity (Kuei et al., 2007; Hossain and Wade, 2010). In addition, the C terminus of the relaxin-3 B-chain appears to be important for activation of RXFP4, and mutation of these residues blocks agonist activity without influencing affinity (Hossain and Wade, 2010).

2. Modified Peptide Agonists.

The potential of RXFP4 as a therapeutic target and the difficulty involved in the synthesis of human INSL5 has led to the search for more easily synthesized and selective analogs. Minimized analogs of relaxin-3 display some interesting properties. Removal of the first eight residues of the A-chain of relaxin-3 produces an analog (analog 2) with similar binding affinity and potency to relaxin-3 at RXFP3 and RXFP4 but no activity at RXFP1 (Shabanpoor et al., 2012) (Table 4). This activity is retained in a peptide where the B-chain is truncated by nine residues (Shabanpoor et al., 2012). These studies indicate that the C terminus of the B-chain and the interchain disulfide bonds are the major structural features required for activity at RXFP4.

3. Relaxin Family Peptide Receptor 4 Antagonists.

Several of the compounds developed as antagonists of RXFP3 also have antagonist properties at RXFP4. These peptides include the minimized relaxin-3 analog 3 (minimized A-chain and truncated B-chain) that has no agonist activity at RXFP4 but only a slightly reduced binding affinity (Shabanpoor et al., 2012) (Table 4). Likewise the R3(BΔ23–27)R/I5 RXFP3 antagonist also is a high-affinity antagonist at RXFP4 (Kuei et al., 2007). Interestingly, the single-chain variant of this peptide H3 B1–22R does not bind to RXFP4 (Haugaard-Kedstrom et al., 2011) (Table 4).

Much less information is available on RXFP4 compared with the other family members, but this has been accelerating due to the identification of INSL5 as a likely incretin. There is now a good understanding of the mode of interaction of INSL5 with RXFP4 and an increasing range of synthetic peptide agonists and antagonists have been developed.

1. Canonical Signaling Pathways.

RXFP1 displays pleiotropic signaling to a variety of pathways including cAMP accumulation, activation of MAP kinases, tyrosine kinases, and NO as well as pathways associated with connective tissue metabolism (Fig. 4). Early studies before receptor identification, showed that treatment with relaxin increased cAMP accumulation in THP-1 cells (Parsell et al., 1996), MCF-7 cells (Bigazzi et al., 1992), the mouse pubic symphysis (Braddon, 1978), uterine strips (Sanborn et al., 1980), uterine longitudinal muscle (Osa et al., 1991) from estrogen-primed rats, and in cultures of human endometrial cells (Fei et al., 1990), human endometrial glandular epithelial cells (Chen et al., 1988), newborn rhesus monkey uterine cells (Kramer et al., 1990), rat myometrial cells (Hsu et al., 1985), and rat anterior pituitary cells (Cronin et al., 1987). The importance of cAMP as a signaling pathway for relaxin was confirmed on RXFP1 deorphanization, because constitutively active mutants of RXFP1 (TM6: D637Y) increased cAMP accumulation in a ligand-independent manner (Hsu et al., 2000, 2002). RXFP1 couples to Gα_s to increase cAMP (Hsu et al., 2000, 2002; Halls et al., 2006), an effect that is negatively modulated by coupling to Gα_oB (Halls et al., 2006) (Fig. 4). RXFP1 also couples to Gα_i3 to activate a delayed surge of cAMP accumulation via a Gβγ-PI3K-PKCζ pathway that activates AC5 (Nguyen et al., 2003; Nguyen and Dessauer, 2005b; Halls et al., 2006, 2009b). The delayed pathway was initially identified in THP-1 cells that endogenously express RXFP1 where relaxin causes a biphasic increase in cAMP accumulation, with the later phase partially blocked by the PI3K inhibitors LY294002 [2-(4-morpholinyl)-8-phenyl-1(4H)-benzopyran-4-one hydrochloride] and Wortmannin. This was confirmed when relaxin stimulation of RXFP1 was also shown to increase PI3K activity (Nguyen et al., 2003). PKCζ was proposed as the candidate protein that linked PI3K activation to cAMP formation and is an atypical isoform that is insensitive to diacylglycerol and calcium but activated by phosphatidylinositol 3,4,5-trisphosphate and other lipids (Ono et al., 1989; Nakanishi et al., 1993; Standaert et al., 2001). Treatment with porcine relaxin causes a concentration-dependent translocation of PKCζ to the cell membrane in MCF-7 (human breast cancer), PHM1-31 (pregnant human myometrial), MMC (mouse mesangial), and THP-1 cells (Nguyen and Dessauer, 2005b), with translocation dependent upon PI3K activation by relaxin but independent of cAMP accumulation. cAMP increases mediated by relaxin were also dependent upon PKCζ expression (Nguyen and Dessauer, 2005b). Although in MCF-7 cells, relaxin stimulation activates PI3K and causes translocation of PKCζ without increasing cAMP accumulation, transfection of AC5, but not AC2 or AC4, into these cells produces a cAMP response to relaxin (Nguyen and Dessauer, 2005a). Thus RXFP1 activates PI3K and causes translocation of PKCζ to the cell membrane, which in turn stimulates AC5 (Fig. 4).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Signal transduction pathways activated by RXFP1. In the resting state at least some RXFP1 receptors are bound to AKAP79 and adenylyl cyclase in the signalosome complex. Low, subpicomolar concentrations of relaxin activate the signalosome to produce tightly controlled cAMP signals. When the relaxin concentration rises to nanomolar the signalosome dissociates and RXFP1 activates canonical pathways to increase cAMP, ERK1/2, and NO and to activate NFκB. In pathologic conditions, the increased expression of the AT₂R promotes formation of RXFP1/AT₂R heterodimers that are essential for the antifibrotic actions of relaxin. The allosteric agonist ML290 interacts with RXFP1 to produce a different signaling pattern to relaxin.

This pathway also occurs in HEK293 cells transiently or stably expressing RXFP1 (Halls et al., 2006, 2009b) and is downstream of Gα and Gβγ subunits, because transfection of cells with Gα_i/o mutants that are insensitive to ADP ribosylation by pertussis toxin (PTX) show coupling of RXFP1 to Gα_i3 (in addition to Gα_s and inhibitory Gα_oB) as the mediator of the Gβγ-PI3K-PKCζ-AC5 pathway. This also occurs in endogenous settings: in rat left atria, relaxin increases cAMP accumulation that is reduced by PTX pretreatment, which also reduced the inotropic and chronotropic responses to relaxin (Kompa et al., 2002); in the failing human heart there is increased Gα_i/o expression (Eschenhagen et al., 1992; Bohm et al., 1994), and the positive inotropic effects of relaxin in humans are preserved in this condition (Dschietzig et al., 2011).

The final 10 amino acids of the RXFP1 C terminus, particularly R752, are essential for activation of the Gα_i3 pathway (Halls et al., 2009b) (Fig. 1), as is the presence of lipid-rich membrane domains, suggesting compartmentalization of the RXFP1-stimulated cAMP response (Halls et al., 2009b). GTPγS-immunoprecipitation studies show that Gα_i3 is activated immediately after RXFP1 stimulation, suggesting that the delay observed is downstream of the G protein and probably involves the translocation of PKCζ (Halls et al., 2009b). In HEK293 cells, only activation of Gα_s- and Gα_oB-dependent cAMP signaling pathways increases CRE-mediated gene transcription, whereas Gα_i3-mediated signaling appears to selectively regulate nuclear factor-κB (NFκB)–dependent gene transcription (Halls et al., 2007a), again suggesting that there is compartmentalization of signaling events and inferring that distinct physiologic outcomes can be anticipated downstream of different cAMP signaling branches (Halls et al., 2007a). cAMP accumulation may also occur in response to relaxin by a G protein–independent mechanism, and in some cells may be downstream of a tyrosine kinase. In THP-1 cells and cultures of primary human myometrial or endometrial stromal cells, porcine relaxin increased cAMP accumulation that was blocked by inhibition of tyrosine kinase activity (Kuznetsova et al., 1999; Bartsch et al., 2001; Anand-Ivell et al., 2007; Heng et al., 2008). This response could be potentiated by the phosphotyrosine phosphatase inhibitors [bpV(phen) and mpV(pic)] that mimic tyrosine kinase activation (Bartsch et al., 2001). In human lower uterine segment fibroblasts, relaxin stimulation caused tyrosine phosphorylation of cellular extracts, with no effect upon cAMP accumulation (Palejwala et al., 1998). The tyrosine kinase–dependent increase in cAMP accumulation may occur by inhibition of a PDE, thereby preventing cAMP hydrolysis and thus reflexively increasing cAMP levels. However, the same tyrosine kinase inhibitors do not affect relaxin-stimulated cAMP accumulation in HEK293 cells expressing RXFP1 (Anand-Ivell et al., 2007), emphasizing the variation in cellular responses between different cell types. In cells using the tyrosine kinase pathway, there was some degree of cAMP inhibition by the PI3K inhibitor LY294002 (Anand-Ivell et al., 2007; Heng et al., 2008) and evidence of a negative feedback loop involving PKA (Anand-Ivell et al., 2007).

The stimulation of NO production, cGMP generation, and PKG activation provides another crucial pathway by which relaxin exerts its effects. Depending on the cell type under investigation, relaxin may activate endothelial nitric oxide synthase (eNOS) (Baccari et al., 2007; Dschietzig et al., 2012) and neuronal NOS (nNOS) (Baccari et al., 2004; Mookerjee et al., 2009) or stimulate the expression of inducible NOS (iNOS) (Bani et al., 1998a; Alexiou et al., 2013) (Fig. 4). In rat isolated lungs, the relaxin-mediated iNOS upregulation depends on a subtle balance between stimulatory ERK1/2 activation and counterregulatory PI3K stimulation (Alexiou et al., 2013).

In addition to cAMP accumulation, many cells that express RXFP1 such as human endometrial stromal cells (Zhang et al., 2002), THP-1 cells and primary cultures of human coronary artery cells, pulmonary artery smooth muscle cells, renal myofibroblasts (Mookerjee et al., 2009), and fibrochondrocytes (Ahmad et al., 2012) respond to relaxin with a rapid activation of ERK1/2 (Fig. 4). In normal human endometrial cells relaxin causes rapid but transient phosphorylation of ERK1/2, with a peak response between 5 and 10 minutes (Zhang et al., 2002). The same time course was observed for phosphorylation of MEK and CREB, but relaxin treatment did not affect Akt or JNK phosphorylation and treatment with a MEK inhibitor blocked phosphorylation of ERK1/2 in response to relaxin, suggesting that MEK is activated upstream of ERK1/2 (Zhang et al., 2002). Increased phosphorylation of ERK1/2 (pERK1/2) in response to relaxin is also observed in THP-1 cells and cultures of human coronary artery and pulmonary artery smooth muscle cells and is associated with increased transcription of VEGF (Zhang et al., 2002).

In HeLa cells and primary human umbilical vein endothelial cells, the pERK1/2 response is more prolonged (Dschietzig et al., 2003). In HeLa, EAhy926 (an endothelial cell line), HT-29 (a colonic cell line), and in primary fibrochondrocyte cells, relaxin increased the phosphorylation of both ERK1/2 and Akt after 30 minutes (Dschietzig et al., 2009a; Ahmad et al., 2012). In primary fibrochondrocytes, treatment with relaxin also activates PI3K, PKCζ, NFκB, c-fos, and Elk-1, all of which influence the expression of MMP-9 (Ahmad et al., 2012). Relaxin also produces a sustained increase in pERK1/2 in rat renal myofibroblasts, potentiated by inhibition of Gα_i/o by PTX, suggesting that phosphorylation of ERK1/2 may be downstream of G protein coupling (Mookerjee et al., 2009). In contrast, in human vascular smooth muscle cells, relaxin stimulation did not affect pERK1/2 levels but instead increased the phosphorylation of p38 mitogen-activated protein kinase (MAPK) (Dschietzig et al., 2003). In rat pulmonary arterial endothelial and smooth muscle cells, relaxin increased iNOS activity and expression dependent on a balance of pERK1/2 and PI3K pathway activation (Alexiou et al., 2013). Thus, although relaxin increases phosphorylation of a number of kinases in multiple cell types, the precise isoform and mode of activation appears to vary, and the physiologic consequences of activation of these pathways are as yet unclear.

Several of the aforementioned canonical pathways are involved in the well established antifibrotic actions of relaxin. Transforming growth factor (TGF)-β₁ is profibrotic, causing Smad2/3 phosphorylation, translocation of the Smad complex to the nucleus, and activation of profibrotic genes. Inhibition of this process by relaxin involves activation of PI3K and NOS-NO-cGMP signaling (Ahmad et al., 2012; Chow et al., 2012). There is also recent evidence that suggests that relaxin influences differentiation of neonatal fibroblasts into myofibroblasts by preventing the inhibition of Notch-1 signaling by TGF-β₁. Notch-1 signaling controls cell differentiation and fate and is involved in fibrosis (Fan et al., 2011). The Notch-1 pathway comprises Notch-1 receptors that, when activated by the TM ligands Jagged1/2, release the Notch intracellular domain that migrates to the nucleus to inhibit transcription (Chillakuri et al., 2012). Downregulation of Notch-1 is a necessary step in the differentiation of rat cardiac fibroblasts into myofibroblasts and is induced by TGF-β₁ (Fan et al., 2011). A recent study demonstrates that relaxin prevents the enhancement by TGF-β₁ of cytoskeletal assembly in 3T3 cells and primary neonatal fibroblasts as measured by F-actin and vinculin staining (Sassoli et al., 2013). In addition, relaxin prevented the TGF-β₁–mediated increases in α-SMA, type 1 collagen, and tissue inhibitor of metalloproteinase (TIMP)-2, while preventing the TGF-β₁–mediated decreases in MMP-2 and MMP-9. Relaxin treatment also prevented the downregulation of Notch-1 expression by TGF-β₁ as well as the inhibition of Notch intracellular domain and Jagged-1 release (Sassoli et al., 2013).

2. Noncanonical Signaling Pathways.

Relaxin activates the GR, a nuclear receptor that acts as a ligand-dependent transcription factor (Dschietzig et al., 2004). The activation, and the subsequent changes in gene transcription, may account for the many effects of relaxin upon the expression levels of a variety of proteins, including those involved in connective tissue metabolism (see above). Relaxin treatment of THP-1 cells that have differentiated into a macrophage phenotype blunts the stimulated production of cytokines including interleukin (IL)-1, IL-6, and tumor necrosis factor (TNF)-α, and this effect is abolished by the GR antagonists RU-486 [mifepristone; 17β-hydroxy-11β-(4-dimethylamino-phenyl)- 17α-(1-propinyl)estra-4,9-dien-3-on] and D06 [bis(4-N,N-dimethylaminophenyl)(2-chloro-5-nitrophenyl) methane] (Dschietzig et al., 2004, 2009a). Relaxin coimmunoprecipitates with the GR, and the amount of the GR within the nucleus increases after 30-minute stimulation with relaxin. In whole cell and in fluorescence polarization assays, relaxin displaces classic glucocorticoids from the GR (Dschietzig et al., 2004, 2009a). The region of relaxin that binds to the GR differs from that involved in relaxin binding to RXFP1, because a modified relaxin that was unable to activate RXFP1 could still interact with the GR, and relaxin was found to cause phosphorylation of S211 of the GR, which is used as a biomarker of agonist-related receptor activation (Dschietzig et al., 2009a). Indeed relaxin, via its interaction with the GR, is also able to autoregulate its own expression by binding to half-sites of glucocorticoid response elements located 160–200 bp upstream of transcription start at the human RLN2 promoter (Dschietzig et al., 2009b). Actions at the GR are also involved in the vasodilator effects of relaxin. In rat aortic rings, relaxation responses to acetylcholine are impaired by TNF-α in a model of endothelial dysfunction, and the effect is reversed by relaxin (Dschietzig et al., 2012). The reversal of the effect of TNF-α by relaxin is blocked by the PI3K inhibitor Wortmannin but also by the GR and progesterone receptor antagonist RU-486. In rat primary aortic endothelial cells TNF-α treatment increased endothelin-1 (ET-1) and arginase II expression, decreased superoxide dismutase (SOD)-1 expression, and stimulated superoxide and nitrotyrosine formation (Dschietzig et al., 2012). All of these effects were restored or attenuated by relaxin treatment acting at the GR (Dschietzig et al., 2012). The findings suggest that relaxin has a protective effect on endothelial dysfunction mediated not only by actions on the canonical PI3K-Akt-eNOS pathway but also by its actions at the GR, with SOD-1 upregulation being dependent on relaxin-GR-c/EBP-β signaling (Dschietzig et al., 2012). Reports that appear to confirm the principal relaxin-GR finding include relaxin-related protection in a rodent model of severe acute pancreatitis that is markedly attenuated after cotreatment with RU-486 (Cosen-Binker et al., 2006), relaxin-induced GR activation in reporter gene experiments (Halls et al., 2007a), and blockade of relaxin-mediated decreases in granulocyte macrophage colony-stimulating factor and IL-8 secretion from primary decidual macrophages by RU-486 (Horton et al., 2011).

Hitherto, the precise mode of interaction between relaxin and GR has been unclear and there are contradictory findings regarding RXFP1 involvement in the relaxin-GR pathway. Although relaxin seems to interact with GR in spleen cells that do not express RXFP1 (Dschietzig et al., 2004) and also after chemical modification to destroy RXFP1 affinity (Dschietzig et al., 2009a), RXFP1 was necessary for relaxin to activate the GR in an in vitro study using reporter genes (Halls et al., 2007a).

3. The Relaxin Family Peptide Receptor 1 Signalosome—A Protein Complex That Facilitates High-Sensitivity Signaling in Response to Relaxin.

Responses to relaxin produced by activation of canonical signaling pathways occur in the nanomolar concentration range (Bathgate et al., 2013a). However, circulating concentrations of relaxin are usually lower than those required to produce cellular responses by canonical mechanisms (Sherwood, 2004). Recently, a constitutive RXFP1-dependent cAMP response has been identified using Förster resonance energy transfer-based cAMP biosensors in single rat cardiac fibroblasts, HeLa cells, and HEK293 cells expressing RXFP1 (Halls and Cooper, 2010). The response is dependent upon a protein complex, or signalosome, linked to the relaxin receptor, and the signalosome is highly sensitive to attomolar concentrations of relaxin. Importantly, this may provide the basis for some of the physiologic responses to relaxin when it is present in the circulation at levels below those necessary to activate canonical signaling pathways. The signalosome consists of RXFP1 that is scaffolded to AC2 by AKAP79, facilitating efficient activation of the AC by Gα_s and Gβγ subunits (Fig. 4). The levels of cAMP produced are tightly regulated by the activity of PKA-activated PDE4D3 that itself is scaffolded to the receptor C terminus (specifically requiring S704) by β-arrestin-2 (Halls and Cooper, 2010). The signalosome possesses stimulatory (AKAP79 and AC2) and regulatory (β-arrestin-2, PKA, and PDE4D3) arms that are both spatially and functionally distinct. Knockdown of AKAP79 has no effect on the regulatory arm, and knockdown of β-arrestin-2 does not influence the stimulatory arm of the signalosome. In addition, there is no effect of inhibitors of classic pathway specific proteins (including Gα_i/o, PI3K, and PKC) on cAMP generated in response to subpicomolar concentrations of relaxin and no effect of inhibition of signalosome-specific proteins (including AC2, AKAP79, and β-arrestin-2) upon classic relaxin cAMP signaling. The pathways also generate cAMP in quite distinct regions of the cell. Thus signalosome-specific AC2 is known to be preferentially excluded from lipid-rich domains (Willoughby et al., 2007), whereas activation of the Gα_i3 pathway with nanomolar relaxin concentrations depends upon lipid-rich domains in HEK293 cells (Halls et al., 2009b). Because AC2 expression occurs predominantly in brain, lung, skeletal muscle, heart, and uterine myometrium (Defer et al., 2000; Willoughby et al., 2007; Sadana and Dessauer, 2009), it is likely that some tissues display RXFP1 signalosome signaling, whereas others do not, which may help to determine the physiologic role of signalosome-localized RXFP1.

Targeted protein knockdown or overexpression of dominant negative mutants (Halls and Cooper, 2010) suggests that the regulatory arm is isoform-specific (i.e., interacts only with PDE4D3) and that assembly of the components depends upon constitutive association between the receptor and β-arrestin-2 (not β-arrestin-1) (Halls and Cooper, 2010). The regulatory arm also has some unusual features that call into question some of the current paradigms surrounding β-arrestins and receptor desensitization and internalization (DeFea, 2011). For example, the β₂-adrenoceptor is phosphorylated by G protein receptor kinase (GRK)-2 after receptor activation, and subsequently becomes a substrate for β-arrestin-2 that triggers formation of clathrin-coated vesicles and internalization (for review, see Luttrell and Gesty-Palmer, 2010). However, and in contrast, the interaction between RXFP1 and β-arrestin-2 is constitutive and does not involve receptor activation or phosphorylation and does not appear to be involved in desensitization (Halls and Cooper, 2010). Although the interaction involves S704 of RXFP1 (Halls and Cooper, 2010), there is no significant increase in RXFP1 phosphorylation after stimulation with high concentrations of relaxin nor any effect of receptor activation on desensitization or internalization (Tan et al., 1998; Callander et al., 2009). Thus β-arrestin-2 appears to act solely as a scaffold for the formation of the regulatory complex and has no role in receptor desensitization or internalization. On the other hand, the stimulatory arm of the signalosome depends upon a specific and constitutive association between helix 8 of the receptor and AKAP79 (Fig. 4). However, AKAP79 is also known to scaffold many other proteins including other GPCRs, PKC, several ion channels, and a number of AC isoforms (see Baillie et al., 2005; Dessauer, 2009; Skroblin et al., 2010; Halls, 2012). The activity of AC isoforms may be variably influenced by interaction with AKAP79 and AC2 activity is inhibited (Efendiev et al., 2010), although this appears to be primarily offset in RXFP1 signalosomes by the scaffolding of RXFP1 and AC2. Signalosome signaling occurs quite separately from canonical signaling, and the signalosome complex dissociates when exposed to nanomolar concentrations of relaxin (Halls and Cooper, 2010).

High-sensitivity signaling such as that exhibited by the RXFP1 signalosome has been previously shown for suppression of proinflammatory cytokine production by IL-15 (Alleva et al., 1997), proliferation of helper T cells by IL-1 (Orencole and Dinarello, 1989), the effects of neuropeptides and neurosteroids in nociception (Sanchez-Blazquez and Garzon, 1995; Ueda et al., 2001), and the long-term effects of TGF-β on basal follicle-stimulating hormone levels (Ying et al., 1986). Thus the RXFP1-signalosome represents a highly sensitive signaling platform that produces tightly regulated cAMP responses over a wide range of subpicomolar relaxin concentrations. Although this represents a mechanism whereby cells and tissues could respond to low circulating levels of relaxin, a precise physiologic role for the mechanism has yet to be determined.

4. Homo- and Hetero-Oligomerization of RXFP1 and Its Functional Consequences.

Like many other class A GPCRs, there is evidence from bioluminescence resonance energy transfer (BRET) studies that RXFP1 receptors form homo- and heterodimers (Kern et al., 2008; Svendsen et al., 2008a,b) in the absence of, and independent of, ligand occupation of the receptor (Svendsen et al., 2008a,b). It has been suggested that dimer formation is necessary for signal transduction, with ligand binding occurring at the LRR region of one dimer partner, followed by interaction of the bound ligand with the ECL2 of the second partner and initiation of signaling (Kong et al., 2010). However, at present there are no experiments with two inactive mutant receptor dimer partners that show complementation and rescue of function by dimerization to support this concept. One functional consequence of dimerization is negative cooperativity (Svendsen et al., 2008a,b), where the two binding sites do not have a fixed affinity and the affinity of each remaining unoccupied receptor binding site decreases as occupancy increases. The evidence for negative cooperativity at RXFP1 comprises receptor kinetic studies where the rate of dissociation after incubation of cells expressing RXFP1 with [¹²⁵I]human relaxin was determined after equilibration followed by infinite dilution in the absence or presence of unlabeled ligand. The presence of unlabeled ligand was associated with modest increases in the rate of dissociation, suggesting negative cooperativity. It should be borne in mind, however, that such behavior does not necessarily reflect the formation of dimers (Chabre et al., 2009), and studies are required using receptors that retain function but are unable to form dimers or on receptors expressed in model phospholipid bilayers that allow examination of their functional characteristics when in monomeric form (Whorton et al., 2007; Velez-Ruiz and Sunahara, 2011). Nevertheless, there are two intriguing, functionally relevant consequences of negative cooperativity: an increased functional range of the ligand over a wider concentration range and a decrease in ligand residence time at the receptor as the free ligand concentration increases, potentially allowing selective activation of different signaling pathways (Shymko et al., 1997). The negative cooperativity concentration-response curve for relaxin binding to RXFP1 is linear, although the absence of a protein structure precludes conclusions regarding the functional consequences of this observation (Svendsen et al., 2008b).

Receptor oligomerization has been suggested to explain one intriguing aspect of relaxin-RXFP1 pharmacology, the phenomenon of bell-shaped concentration-response curves that are observed in many bioassays from studies in recombinant and primary cell systems (Halls et al., 2006; Sarwar et al., 2015), animal studies (Danielson and Conrad, 2003; Debrah et al., 2005), and clinical trials (Teerlink et al., 2009). Negative cooperativity could be a potential explanation, but currently available information does not appear to support this mechanism. Although both RXFP1 and RXFP2 display negative cooperativity, bell-shaped curves for cAMP accumulation are observed only in response to 3- or 30-minute activation of RXFP1 or 30-minute activation of RXFP2, but not 3-minute activation of RXFP2 (Halls et al., 2006). It could be argued that the RXFP2 system at 3 minutes is not in equilibrium and that the effect of increasing receptor occupation balances out the reduced receptor affinity, but this does not convincingly explain the difference between RXFP1 and RXFP2 at the same time point. In studies of human vascular cells that endogenously express RXFP1, venous endothelial and smooth muscle cells displayed pronounced bell-shaped concentration-response curves, yet arterial smooth muscle cells showed conventional sigmoidal concentration-response curves (Sarwar et al., 2015). This would appear to be incompatible with an explanation based on RXFP1 homodimer formation and negative cooperativity.

There is also evidence for dimer formation between the haloreceptor and a number of splice variants (encoding the LDLa module only and up to eight LRR) and dimers are present at all stages of receptor translocation from the endoplasmic reticulum to the plasma membrane (Kern et al., 2008). This suggests an important role of dimerization over the lifetime of the receptor. Dimers were also formed between the haloreceptor and a TM-only domain receptor, although the BRET₂ ratio was decreased compared with that obtained for haloreceptor dimers (Svendsen et al., 2008a,b). This suggests that although the TM domain is sufficient for dimerization, the ectodomains play an important role in stabilizing the oligomer. Examination of three of the splice variants cloned from human fetal membranes showed that although their expression in HEK293 cells produced no response to relaxin (Kern et al., 2008), coexpression with RXFP1 produced a parallel shift to the right of the cAMP concentration-response curve. Because the splice variants also markedly reduced cell surface expression of RXFP1, the reduced responses are likely due to a dominant negative effect of the splice variants (Kern et al., 2008).

Recent studies also suggest that there is heterodimer formation between RXFP1 and other GPCRs. The antifibrotic actions of relaxin are believed to be mediated through a RXFP1-pERK1/2-nNOS-NO-cGMP dependent pathway, leading to regulation of collagen-degrading MMPs (Chow et al., 2012). The effect of relaxin is completely absent in angiotensin type 2 receptor (AT₂R)^−/y mice or in mice treated with the AT₂R antagonist PD123319 [S-(+)-1-[(4-(dimethylamino)-3-methylphenyl)methyl]-5-(diphenylacetyl)-4,5,6,7-tetrahydro-1H-imidazo[4,5-c]pyridine-6-carboxylic acid di(trifluoroacetate)], suggesting that the AT₂R is necessary for the antifibrotic actions of relaxin. Interestingly BRET studies show that RXFP1 and AT₂R form constitutive heterodimers, indicating that the receptor complex is responsible for the novel pharmacology observed (Fig. 4). There is no direct binding of relaxin to the AT₂R, and the peptide does not affect the BRET signal from RXFP1-AT₂R complexes (Chow et al., 2014). Importantly, the findings also explain why the antifibrotic actions of relaxin are only observed in pathologic states: under normal physiologic conditions AT₂R are expressed at low levels but are dramatically increased with injury and disease (Siragy and Carey, 1997; Matsubara, 1998; Carey, 2005; Savoia et al., 2006; Jones et al., 2008).

It is interesting to note that the expression of relaxin-RXFP1 as measured by immunohistochemistry and expression of the renin-angiotensin-aldosterone system may both be increased during pregnancy (Ferreira et al., 2009; Cheung and Lafayette, 2013), suggesting another state where RXFP1-AT₂R heterodimers may have a functional role. However, a study in rat primary mesangial cells did not corroborate this suggestion (Carvalho et al., 2012). Although RXFP1 and AT₂R were increased in cells obtained from pregnant rats compared with virgin animals and the hyporesponsiveness of cells from pregnant animals to angiotensin II could be attributed to increased endogenous relaxin production, the AT₂R antagonist PD123319 had no effect. Thus, the issue of RXFP1-AT₂R interaction during pregnancy remains a hypothesis.

5. Allosteric Agonist Signaling at Relaxin Family Peptide Receptor 1.

The small molecular weight ligand ML290 (see section II.A.3) is an allosteric agonist at RXFP1 (Xiao et al., 2013) that unlike relaxin does not require the LDLa module for signaling (Xiao et al., 2013). ML290 interacts with ECL3 of RXFP1, specifically requiring G659 and T660 within ECL3 for activity (Xiao et al., 2013) (Fig. 1). ML290 does not interact with the classic binding sites on RXFP1 used by relaxin and does not compete for ¹²⁵I relaxin binding (Table 1).

1. Canonical Signaling Pathways.

INSL3 is the cognate ligand at RXFP2, although human relaxin but not rat or mouse relaxin, also activates the receptor. Despite the similar structure of relaxin and INSL3 and their receptors RXFP1 and RXFP2, the mode of interaction of these ligands with RXFP2 is different (see section I.B.1). Signaling pathways that are initiated after stimulation of RXFP2 are simpler than those observed for RXFP1. INSL3 (or relaxin) stimulation of HEK293-RXFP2 cells causes coupling to Gα_s to increase cAMP accumulation and to Gα_oB to negatively modulate this effect (Kumagai et al., 2002; Halls et al., 2006). This stage of signaling therefore closely resembles the first stage of RXFP1 signaling in response to relaxin (Halls et al., 2006). The contribution of Gα_oB to the response is seen by the increase in cAMP after removal of Gβγ using βARK-ct or inhibition of Gα_i/o with PTX (Halls et al., 2006). Interestingly, for RXFP2 there is no evidence for constitutive activity or a high-sensitivity response to INSL3 as seen with the RXFP1 signalosome (Halls and Cooper, 2010); neither is there activation of the Gα_i3-cAMP signaling pathway that is unique to RXFP1 (Halls et al., 2006, 2009b). Activation of RXFP2 (by either relaxin or INSL3) also induces increased CRE-dependent gene transcription (Halls et al., 2007a) as observed with relaxin acting at RXFP1.

Gα_s and Gα_oB are also involved in cAMP accumulation in cells that endogenously express RXFP2. In rat gubernacular cells (Kumagai et al., 2002), in a human osteoblast cell line (MG-63) (Ferlin et al., 2008), and in mouse primary Leydig cells (Pathirana et al., 2012), INSL3 stimulation of RXFP2 leads to increased cAMP accumulation, probably involving Gα_s. However, and in contrast, primary cultures of testicular germ cells and oocytes respond to INSL3 activation of RXFP2 with a PTX-sensitive inhibition of cAMP accumulation (Kawamura et al., 2004), consistent with RXFP2 coupling to Gα_oB. Thus, as for RXFP1, the net signaling outcome of RXFP2 stimulation will depend upon the signaling components (especially G protein isoforms) that are expressed in a particular cell type.

In response to relaxin, RXFP1 activates a number of signaling pathways including cAMP, cGMP, and pERK1/2. Relaxin also activates the GR to alter gene expression. RXFP1 forms protein complexes or signalosomes that facilitate high-sensitivity responses to relaxin. In contrast, INSL3 activates RXFP2 to increase or decrease cAMP in a cell-dependent manner. INSL3 does not activate the GR and RXFP2 does not form signalosomes.

1. Canonical Signaling Pathways.

Activation of RXFP3 by human relaxin-3 causes PTX-sensitive inhibition of forskolin-stimulated cAMP accumulation, suggesting that the receptor is coupled to inhibitory Gα_i/o proteins (Liu et al., 2003b). Activation of Gα_i/o proteins by RXFP3 is also associated with PI3K- and PKC-dependent phosphorylation of ERK1/2 and other MAPKs (van der Westhuizen et al., 2007, 2010) (Fig. 5). Both human relaxin-3 or its B-chain compete for [¹²⁵I]human relaxin-3 binding and inhibit AC activity (Liu et al., 2003b), but more recent studies show that human relaxin also interacts with human RXFP3 to activate a unique signaling pattern (see below) (van der Westhuizen et al., 2010).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Signal transduction pathways activated by RXFP3. The cognate ligand relaxin-3 activates RXFP3 that couples to Gα_i/o proteins to inhibit adenylyl cyclase and activate ERK1/2 and p38MAPK phosphorylation to promote AP1 transcription. Stimulation of RXFP3 by relaxin activates only ERK1/2 and AP1 transcription. There is an allosteric site on RXFP3 that when occupied by a positive allosteric modulator 135PAM1 sensitizes responses to relaxin-3 amide. In tissues where they coexist, activation of RXFP3 may cause transactivation of the EGFR.

Forskolin-stimulated cAMP accumulation in CHO-K1 or HEK293 cells stably expressing RXFP3 or in mouse SN56 cells that endogenously express the receptor (Liu et al., 2003b; van der Westhuizen et al., 2007) is inhibited after stimulation with human relaxin-3, and this response is completely prevented by pretreatment with PTX. In CHO-K1 cells transiently transfected with PTX insensitive (C351I mutation) variants of Gα_i/o proteins and treated with PTX (to remove the influence of endogenous Gα_i/o proteins), Gα_i2 was the major G protein involved in the inhibition of cAMP accumulation, whereas in HEK293 cells, Gα_i3, Gα_oB, and Gα_oA were all involved (van der Westhuizen, 2008). Thus, the G proteins involved, although broadly similar, can vary with the different cell types. Similar effects are observed with human relaxin-3 B-chain peptides or with human relaxin and porcine relaxin in the CHO-K1, HEK293, and SN56 cell backgrounds (Liu et al., 2003b; van der Westhuizen et al., 2007, 2010).

Human relaxin-3 causes a rapid and transient increase in ERK1/2 phosphorylation (peak response 2–5 minutes) in CHO-K1 and HEK293 cells stably expressing human RXFP3 (CHO-RXFP3 or HEK-RXFP3, respectively; van der Westhuizen et al., 2007, 2010) (Fig. 5). The human relaxin-3 B-chain dimer also activates ERK1/2, albeit with low potency and efficacy (van der Westhuizen et al., 2007), and there is also a weak response to human relaxin but not to porcine relaxin or INSL3 (van der Westhuizen et al., 2010) (Fig. 5). PTX pretreatment caused an approximately 90% inhibition of human relaxin-3 stimulated ERK1/2 phosphorylation in CHO-RXFP3 cells, whereas the signal was completely abolished in HEK-RXFP3 or SN56 cells (van der Westhuizen et al., 2007). Thus coupling of RXFP3 to PTX-sensitive Gα_i/o proteins mediates ERK1/2 phosphorylation in both recombinant and endogenous systems (van der Westhuizen et al., 2007). There are two pathways downstream of Gα_i/o involved in RXFP3-mediated ERK1/2 phosphorylation in CHO-K1, HEK293, and SN56 cells. About 50% of the MAPK response is blocked by the PI3K inhibitors LY294002 or Wortmannin, whereas the remainder is inhibited by general and isoform-selective PKC inhibitors (van der Westhuizen et al., 2007). The ERK1/2 response appears to be involved in central feeding responses in rats (Morikawa et al., 2004; Shen et al., 2004; Sasaguri et al., 2005), suggesting that it is physiologically relevant.

Signaling pathway analysis using reporter genes gives a broader view of the signal transduction mechanisms activated by RXFP3. Many MAPK signaling pathways [p38 MAPK (Roux and Blenis, 2004); JNK (Davis, 2000); and ERK1/2 (Price et al., 1996; Whitmarsh and Davis, 1996)] converge on AP-1 elements to increase gene transcription, therefore AP-1-linked reporter genes together with selective inhibitors provide useful information on signaling in different cellular backgrounds. Inhibition of Gα_i/o proteins after pretreatment with PTX blocks RXFP3-mediated AP-1 reporter activation in SN56 cells but not in CHO-RXFP3 and HEK-RXFP3 cells (van der Westhuizen et al., 2010), suggesting that AP-1 reporter gene activation was downstream of Gα_i/o in the mouse-derived cell line but not in the other two cell types (van der Westhuizen et al., 2010). Human relaxin-3–mediated AP-1 reporter gene activation in CHO-RXFP3 and SN56 cells is completely blocked by the p38 MAPK inhibitor (RWJ67657; 4-[4-(4-fluorophenyl)-1-(3-phenylpropyl)-5-(4-pyridinyl)-1H-imidazol-2-yl]-3-butyn-1-ol), whereas MEK (PD98059; 2′-amino-3′-methoxyflavone) or JNK inhibition (SP600125; anthra[1,9-cd]pyrazol-6(2H)-one) only partially blocks AP-1 activation (as shown by a decrease in relaxin-3 potency). In contrast in HEK-RXFP3 cells, JNK inhibition completely blocks human relaxin-3–stimulated AP-1 reporter activation, whereas p38 MAPK or MEK inhibition partially blocked AP-1 activation (van der Westhuizen et al., 2010). This suggests that although all three MAPKs are involved in human relaxin-3–mediated AP-1 activation, the hierarchy of the different signaling pathways varies with the cell background. Several studies show that MAPK signaling is activated in forced swim tests in rats, where there are dramatic increases in pMEK1/2, pERK1/2, and pJNK1/2/3 (Shen et al., 2004). Although this is associated with an increase in relaxin-3 mRNA in the nucleus incertus (NI) (Tanaka et al., 2005), direct links between human relaxin-3, RXFP3, MAPK phosphorylation, and stress responses remain to be demonstrated in brain.

In another reporter gene assay, in CHO-K1 and HEK293 cells transiently expressing human RXFP3 and in SN56 cells endogenously expressing mouse RXFP3, activation of RXFP3 by human relaxin-3 increased NFκB reporter gene activation (van der Westhuizen et al., 2010). Activation of NFκB was blocked by PTX pretreatment (van der Westhuizen et al., 2010), suggesting that the response occurs downstream of Gα_i/o. The physiologic significance of this pathway remains to be determined.

2. Ligand-Directed Signaling Bias at Relaxin Family Peptide Receptor 3.

Ligand-directed signaling bias (as described in section II.C.2) has been described for relaxin-3, relaxin (van der Westhuizen et al., 2005), and the RXFP3 antagonist R3(BΔ23–27)R/I5 (Kocan et al., 2014). The first indication of a functional relaxin-RXFP3 interaction came from examination of metabolic responses recorded by microphysiometry (van der Westhuizen et al., 2005). In this study, human relaxin caused a small change in the extracellular acidification rate in CHO-RXFP3 cells. In CHO-K1, HEK293, and SN56 cell backgrounds, subsequent studies demonstrated that human relaxin, porcine relaxin, and human INSL3 caused weak inhibition of forskolin-stimulated cAMP accumulation compared with the strong inhibition mediated by human relaxin-3 (van der Westhuizen et al., 2010) (Table 3). Interestingly, the ability of INSL3 to activate RXFP3 appears to be specific for the human but not the mouse receptor (van der Westhuizen et al., 2010). Previous studies reported no inhibition of cAMP accumulation by either porcine relaxin or human INSL3 (Liu et al., 2003b), but because sensitivity of inhibitory cAMP assays is highly dependent on both the degree of activation of AC by forskolin and the time of stimulation, the differences observed most likely result from distinct experimental paradigms.

Similar to relaxin-3, relaxin and porcine relaxin also caused AP-1 reporter gene activation (van der Westhuizen et al., 2010) in CHO-K1, HEK293, and SN56 cell backgrounds, with an order of potency relaxin > relaxin-3 > porcine relaxin. Again, some AP-1 activation appears to be independent of Gα_i/o coupling, because pretreatment with PTX failed to block porcine relaxin-stimulated AP-1 reporter gene activation in CHO-RXFP3 cells. Thus, as for relaxin-3, porcine relaxin can activate AP-1 reporter genes by a Gα_i/o-independent mechanism, suggesting ligand-directed signaling bias (van der Westhuizen et al., 2010). In contrast, all of the AP-1 reporter gene responses observed after stimulation of RXFP3 in SN56 cells were blocked by PTX, suggesting that different pathways were involved in mediating the response downstream of the mouse RXFP3 receptor (van der Westhuizen et al., 2010).

In the CHO-K1 cell background, the relaxin-stimulated AP-1 response is strongly inhibited by the p38 MAPK inhibitor (RWJ67657) or the JNK inhibitor (SP600125), whereas the MEK inhibitor (PD98059) only weakly inhibited the AP-1 response, implicating p38 MAPK and JNK as the major MAPKs involved in the response to relaxin. In contrast in the HEK293 cell background, relaxin-stimulated AP-1 reporter gene activation was decreased after p38 MAPK or MEK inhibition but not by JNK inhibition, suggesting that p38 MAPK and ERK were the major MAPKs involved in mediating this response. In SN56 cells, endogenously expressing mouse RXFP3, p38 MAPK, JNK, or MEK inhibition was equally effective in blocking the response to relaxin, suggesting that all three kinases were equally important (van der Westhuizen et al., 2010). Clearly species and cell background are important determinants of the signaling pattern after RXFP3 activation. Direct measurement of pERK1/2, p38 MAPK, and pJNK after addition of relaxin family peptides has confirmed the findings of these inhibitor-based studies (Kocan et al., 2014) and demonstrates signaling bias; however, the Gα_i/o-independent pathway remains to be identified for human relaxin-3 and porcine relaxin. Direct MAPK assays in CHO-RXFP3 cells show that human relaxin activates p38 MAPK and ERK1/2 with lower efficacy than human relaxin-3 but the two peptides have similar efficacy for JNK1/2/3 phosphorylation. Both human relaxin and human relaxin-3 activation of p38 MAPK, JNK1/2/3, or ERK1/2 involved PTX-sensitive G proteins (van der Westhuizen et al., 2007, 2010; Kocan et al., 2014).

Evidence for signaling bias at RXFP3 also comes from studies utilizing an RXFP3 antagonist (Kocan et al., 2014). The RXFP3 antagonist R3(BΔ23–27)R/I5 blocked human relaxin-3 AP-1 reporter gene activation but not human relaxin AP-1 activation or human relaxin-3 NFκB activation. The antagonist itself R3(BΔ23–27)R/I5 activated the SRE reporter but did not inhibit either human relaxin or human relaxin-3 SRE activation. Finally, although the antagonist R3(BΔ23–27)R/I5 blocked human relaxin-3–stimulated p38MAPK and ERK1/2 phosphorylation, it was also a weak partial agonist for p38MAPK and ERK1/2 signaling (Table 3). Interestingly, in findings similar to previous studies, p38MAPK activation by R3(BΔ23–27)R/I5 was found to be G protein independent. Direct examination of interactions between RXFP3 and G proteins using BRET showed that human relaxin-3–activated RXFP3 interacts with Gα_i2, Gα_i3, Gα_oA, and Gα_oB, whereas human relaxin or R3(BΔ23–27)R/I5 can only induce interactions with Gα_i2 or Gα_oB. Interestingly, only human relaxin-3 promoted RXFP3/β-arrestin interactions that were blocked by R3(BΔ23–27)R/I5 (Kocan et al., 2014). This is compelling evidence for ligand-directed signaling bias at RXFP3.

3. Allosteric Modulation of Relaxin Family Peptide Receptor 3.

The RXFP3 allosteric modulator 135PAM1 has been examined only in receptor binding studies and in recombinant systems expressing the chimeric G protein Gα_qI5 or a CRE reporter gene (Alvarez-Jaimes et al., 2012). It is not known what pattern of signaling is observed in systems that naturally express RXFP3, and given that 135PAM1 has poor solubility and displays selectivity for the C-terminal amides that are not naturally occurring, it has limited use experimentally (Alvarez-Jaimes et al., 2012).

1. Canonical Signaling Pathways.

Stimulation of cells expressing human RXFP4 by INSL5 or human relaxin-3 increases GTPγS binding and inhibits forskolin-stimulated cAMP accumulation, suggesting that RXFP4 is Gα_i/o coupled (Liu et al., 2003a, 2005b). Coexpression of RXFP4 with the promiscuous G protein Gα₁₆ in HEK293 cells produces a strong calcium signal in response to INSL5 and human relaxin-3 (Liu et al., 2003a, 2005b) that is not present in normal cells. Although relatively little information exists at present regarding RXFP4 signaling, it is expected that this will rapidly change given the identification of INSL5 as an incretin (Grosse et al., 2014).

Thus both RXFP3 and RXFP4 are Gα_i/o-coupled receptors that show GTPγS binding and inhibition of cAMP accumulation. More extensive studies of RXFP3 signaling have also revealed coupling to ERK1/2, p38MAPK, and JNK as well as ligand-directed signaling bias and allosteric modulation.

1. G Proteins That Couple to Relaxin Family Peptide Receptor 1.

Although it has long been assumed that stimulation of RXFP1 produces an increase in cAMP by Gα_s activation of AC, a direct functional interaction between RXFP1 and Gα_s has been demonstrated only recently. Application of a specific Gα_s inhibitor (NF449; 4,4′,4′′,4′′′-[carbonylbis(imino-5,1,3-benzenetriyl-bis(carbonylimino))]tetrakis-1,3-benzenedisulfonic acid, octasodium salt) to HEK293 cells expressing RXFP1 significantly decreased receptor-stimulated cAMP accumulation, whereas a low concentration of cholera toxin (a Gα_s activator) substantially enhanced the cAMP response to relaxin (Halls and Cooper, 2010). Other supporting evidence includes the increase in cAMP after PTX pretreatment in the early phase (<10 minutes) after stimulation of RXFP1. Furthermore, peptide fragments of ICL3 of RXFP1 (minimum length 615–629) added to rat tissues (striatum, cardiac, and skeletal muscle membranes) increased cAMP and "antagonized" the response to relaxin, whereas addition of peptide fragments of Gα_s “antagonized” both relaxin stimulation and RXFP1-ICL3 peptide stimulation of rat tissues (Shpakov et al., 2007). Many of the responses linked to RXFP1 display sensitivity to PTX (Halls et al., 2006), suggesting coupling to Gα_i/o proteins, and the specificity of the receptor for particular Gα_i/o isoforms was identified using PTX-insensitive G proteins (Halls et al., 2006). In HEK293 cells stably expressing RXFP1, restoration of the expected cAMP signaling profiles (after PTX pretreatment) occurred only with the expression of Gα_oB and Gα_i3. Activation of Gα_oB and Gα_i3 by RXFP1 was subsequently confirmed using [³⁵S]GTPγS immunoprecipitation, and these studies also demonstrated that activation of Gα_i3 occurs within 3 minutes of relaxin stimulation, suggesting that the observed delay in activation of the Gα_i3-cAMP pathway occurs downstream of receptor coupling to this G protein (Halls et al., 2009b). Subsequent studies using a site-directed mutagenesis approach demonstrated that activation of the Gα_i3-cAMP pathway involves the final 10 amino acids of the RXFP1 C terminus and in particular R752 (Halls et al., 2009b). Taken together, this suggests that all G protein coupling occurs rapidly after receptor activation and that Gα_s (and probably Gα_oB) couple to RXFP1 within the ICL3, whereas coupling to Gα_i3 is dependent upon the final 10 amino acids of the receptor C terminus.

2. Interactions between Relaxin Family Peptide Receptor 1 and β-Arrestins.

β-Arrestins were initially identified for their role in receptor desensitization and internalization. Prototypical GPCR signaling is terminated after activation by receptor phosphorylation, β-arrestin binding followed by uncoupling from the G protein, and internalization (Luttrell and Gesty-Palmer, 2010). However β-arrestins are increasingly recognized as scaffold proteins that recruit a variety of catalytically active proteins that influence a wide range of signaling pathways, protein translation, and gene transcription. Unlike many other GPCRs, when expressed in HEK293T or Cos-7 cells, RXFP1 activation does not result in significant receptor phosphorylation, desensitization, or internalization (Callander et al., 2009). Little phosphorylation and internalization could be measured by whole cell radioligand binding, and there was minimal cell surface localization of GFP-tagged β-arrestin visualized by confocal imaging (Callander et al., 2009). In human primary decidual cells or HEK293 cells stably expressing RXFP1, there is weak internalization as measured by cell surface enzyme-linked immunosorbant assay (ELISA) that is enhanced by overexpression of β-arrestin-2 (Kern and Bryant-Greenwood, 2009). Constitutive interaction between RXFP1 and β-arrestin-2 has been identified in the RXFP1 signalosome (Halls and Cooper, 2010), where β-arrestin-2 scaffolds PKA and PDE4D3 to regulate cAMP levels formed in response to subpicomolar concentrations of relaxin (see section III.A.3 for details). The signalosome dissociates in response to nanomolar concentrations of relaxin, and β-arrestin-2 has no role in modulating canonical RXFP1 signaling. Thus, there does appear to be a significant role for β-arrestins in RXFP1 signaling after receptor activation, because there is little β-arrestin recruitment and minimal receptor internalization.

1. G Proteins That Couple to Relaxin Family Peptide Receptor 2.

By using the same approach as that used for RXFP1, the specificity of RXFP2 for particular Gα_i/o isoforms was identified using PTX-insensitive G protein mutants (Halls et al., 2006). After expression of three Gα_i and two Gα_o isoforms in HEK293 cells stably expressing RXFP2, the restoration of the expected cAMP signaling profile (after PTX pretreatment) occurred only with Gα_oB expression. This was subsequently confirmed using [³⁵S]GTPγS immunoprecipitation (Halls et al., 2009b). Unlike RXFP1, RXFP2 does not couple to Gα_i3; this is likely because receptor coupling to Gα_i3 requires the final 10 amino acids of the RXFP1 C terminus (specifically R752) that are missing in RXFP2 (Halls et al., 2009b). It is highly likely that in a manner similar to RXFP1, RXFP2 also couples to Gα_s, because PTX pretreatment increases the cAMP generated after stimulation.

Both RXFP1 and RXFP2 couple to Gα_s and Gα_oB, but only RXFP1 also couples to Gα_i3. There is constitutive interaction between RXFP1 and β-arrestin-2 in the signalosome that dissociates with the advent of canonical signaling. There is only minimal internalization after activation of RXFP1, suggesting little interaction with β-arrestin-2 after signalosome dissociation. RXFP2 does not appear to interact with β-arrestins or internalize (Callander et al., 2009).

1. G Proteins That Couple to Relaxin Family Peptide Receptor 3 after Activation by Human Relaxin-3.

The early deorphanization studies suggested that RXFP3 was a Gα_i/o-coupled receptor that inhibited forskolin-stimulated cAMP accumulation (Liu et al., 2003b; van der Westhuizen et al., 2010). This was supported by studies in the cytosensor microphysiometer, where pretreatment of CHO-RXFP3 cells with PTX strongly inhibited the extracellular acidification rate, suggesting that the signaling pathways activated by RXFP3 were downstream of Gα_i/o (van der Westhuizen et al., 2005). As for RXFP1 and RXFP2, PTX-insensitive G proteins were used to determine the particular G proteins involved (van der Westhuizen, 2008). PTX pretreatment completed abrogated the ERK1/2 response in both CHO-RXFP3 and HEK-RXFP3 cells. In CHO-RXFP3 cells, the response was partially restored by transfection of Gα_i2 or Gα_oB only. Similarly, in HEK-RXFP3 cells, signaling was partially restored by expression of mutant Gα_i2 or Gα_oB but also by mutant Gα_oA. These differences may relate to different colocalization of receptors and G proteins in particular cell types. More recent studies used BRET to investigate ligand-induced interactions between RXFP3-RLuc8 and G proteins (Gγ2-Venus) in live cells. Similar to previous studies, treatment of CHO-RXFP3-RLuc8 cells with human relaxin-3 caused activation of Gα_i2, Gα_oA, and Gα_oB, but also revealed an interaction between RXFP3 and Gα_i3 (Kocan et al., 2014). Thus RXFP3 has the potential to couple to Gα_i2, Gα_i3, Gα_oA, and Gα_oB after activation by human relaxin-3, but the receptor preferentially couples to discrete subsets of these G proteins dependent on cell type.

2. G Proteins That Couple to Relaxin Family Peptide Receptor 3 after Activation by the Biased Ligands Human Relaxin and R3(BΔ23–27)R/I5.

The different signaling patterns observed with the biased ligands human relaxin and R3(BΔ23–27)R/I5 (see section III.C.2 for details) were also reflected in a distinct pattern of G protein coupling in BRET studies (Kocan et al., 2014). The biased ligands coupled only with Gα_i2 or Gα_oB, and the signal was much lower than that observed for human relaxin-3 (Kocan et al., 2014). To determine the potential biologic relevance of these findings, studies will be needed in cells that express RXFP3 at physiologic levels.

3. Interactions between Relaxin Family Peptide Receptor 3 and β-Arrestins.

Interactions between RXFP3 and β-arrestins occur after administration of human relaxin-3 but not relaxin or the RXFP3 antagonist R3(BΔ23–27)R/I5 as measured by real-time kinetic BRET between RXFP3-Rluc8 and a β-arrestin fusion protein (β-arrestin-1-Venus or β-arrestin-2-Venus). Preincubation with R3(BΔ2327)R/I5 completely inhibits the human relaxin-3 stimulated recruitment of β-arrestin-1 and β-arrestin-2 to RXFP3 (Kocan et al., 2014). Furthermore, PTX pretreatment partially blocked β-arrestin recruitment, suggesting that RXFP3 recruits β-arrestins via Gα_i/o-dependent and -independent pathways. The involvement of β-arrestin in ERK1/2 activation downstream of RXFP3 was investigated by comparing cells expressing dominant negative (V53D) versus wild-type (WT) β-arrestin-1. There was no inhibitory effect of the dominant negative β-arrestin-1 on ERK1/2 activation, suggesting that, unlike several other GPCRs (Cottrell et al., 2009; Luttrell and Gesty-Palmer, 2010), the RXFP3/β-arrestin-1 interaction contributes little to ERK1/2 signaling (Kocan et al., 2014).

Previous studies demonstrated internalization of RXFP3 after 10 minutes of stimulation with human relaxin-3 but not human relaxin, porcine relaxin, or INSL3 as assessed by radioligand internalization assays (70–90% of receptors internalized) and confocal microscopy (van der Westhuizen et al., 2010). Taken together, these studies suggest that treatment with only the cognate ligand human relaxin-3 causes RXFP3 to undergo classic β-arrestin–dependent internalization. The detailed mechanisms involved in RXFP3 internalization, phosphorylation, and recycling/degradation all remain to be determined.

1. G Proteins That Couple to Relaxin Family Peptide Receptor 4.

There is currently a paucity of information regarding signaling pathways activated by RXFP4. The receptor is G protein–coupled because activation by INSL5 or human relaxin-3 increases GTPγS binding. cAMP accumulation in response to forskolin is inhibited, suggesting that RXFP4 couples to Gα_i/o proteins (Liu et al., 2003a, 2005b). Calcium signaling is activated only in cells coexpressing RXFP4 and the promiscuous G protein Gα₁₆. Direct evidence for RXFP4-G protein coupling is currently lacking.

In contrast to RXFP1, the information available on proteins interacting with RXFP3 and RXFP4 is largely confined to G proteins and β-arrestins. RXFP3 couples to Gα_oA, Gα_oB Gα_i2, and Gα_i3, although the pattern observed varies with cell type and ligand. There is also good evidence for an interaction of RXFP3 with β-arrestins that modulates receptor internalization. At present, RXFP4 is known to couple only to proteins of the Gα_i/o family.

1. Reproduction.

The discovery of relaxin in 1926 (Hisaw, 1926) and the demonstration of its effects on the pubic ligament led to the peptide being regarded as a hormone of reproduction for many years. Subsequent studies showed that relaxin produced by the corpus luteum and/or placenta has important roles in pregnancy and parturition and is a major circulating hormone during pregnancy in all mammalian species acting on the pubic symphysis, cervix, uterus, vagina, and mammary glands. It also is responsible for many of the cardiovascular changes that occur during pregnancy (Debrah et al., 2006; Conrad, 2011).

Relaxin causes increased flexibility and elasticity of the interpubic ligament during pregnancy in several species (O'Connor et al., 1966; Steinetz et al., 1983), and in the relaxin knockout mouse the interpubic ligament fails to develop (Zhao et al., 1999). The peptide also causes softening and hypertrophy of the cervix during the second half of pregnancy in mammals (Bathgate et al., 2006c), and these effects are impaired in both relaxin-deficient rats (Burger and Sherwood, 1998) and relaxin knockout mice (Zhao et al., 1999; Bathgate et al., 2006c). In humans, although relaxin levels increase during cervical ripening, this still occurs after embryo transfer where circulating relaxin levels are undetectable (Eddie et al., 1990a). In clinical trials, although direct application of porcine relaxin to the cervix appeared to assist in ripening (Evans et al., 1983; MacLennan et al., 1986b), later studies with recombinant human relaxin failed to confirm this finding (Bell et al., 1993; Weiss et al., 2009).

Relaxin reduces uterine contractility (Krantz et al., 1950) and increases uterine growth (Steinetz et al., 1957) in rat, mouse, guinea pig, hamster, and pig (Bathgate et al., 2006c) but in humans, has little effect on uterine tone (MacLennan et al., 1986a, 1995). In humans, relaxin is probably important for implantation rather than the later stages of pregnancy. Relaxin promotes growth of the vagina during pregnancy in mice (Hall, 1960; Schink and Struck, 1968) and rats (Burger and Sherwood, 1995; Zhao et al., 1996, 2001), and consistent with these effects, the growth of the vagina was profoundly decreased in relaxin knockout mice (Zhao et al., 2000). These effects are essential for normal delivery in some species but not humans.

Relaxin has trophic effects on the mammary gland and/or nipple in several species. Nipple development is completely blocked in relaxin knockout mice with limited effects on the mammary gland (Zhao et al., 1999). Pups born to relaxin-knockout mice die within 24 hours unless crossfostered to wild-type mothers, an effect entirely due to poor nipple development, because female knockout mice produce milk normally (Zhao et al., 1999). The same phenotype is displayed by RXFP1 knockout mice (Krajnc-Franken et al., 2004) and is not rescued by transgenic overexpression of INSL3 (Kamat et al., 2004). Relaxin binding sites are present in the mammary glands of pigs, rats, and humans; in humans, RXFP1 receptors are localized to the nipple, epithelial cells (Kohsaka et al., 1998), and stromal tissue (Ivell et al., 2003).

Although relaxin has many well defined roles in reproduction in many species, in humans these effects are often absent or ill defined. Relaxin is produced in the human ovary during the luteal phase of the menstrual cycle and secreted into the circulation (Eddie et al., 1990b; Stewart et al., 1990; Chen et al., 2003). In the clinical trial of relaxin for the treatment of scleroderma (Erikson and Unemori, 2001), women received 24-week subcutaneous infusion of relaxin and reported heavy, irregular, or prolonged menstrual bleeding (Unemori et al., 1999). In humans and other primates, the peak of relaxin secretion is in the first trimester of pregnancy. However relaxin is clearly not mandatory during the induction or maintenance of pregnancy, because women without ovaries can become pregnant by ovum donation even though they have undetectable levels of circulating relaxin (Johnson et al., 1991). Relaxin secretion also correlates with increased expression of RXFP1 mRNA and relaxin binding in the human endometrium in the secretory phase of the menstrual cycle (Bond et al., 2004; Campitiello et al., 2011).

Perhaps somewhat surprisingly given its history as a hormone acting on the female reproductive tract, relaxin is also found in the male reproductive tract in most mammals. Relaxin from human seminal plasma is identical to luteal relaxin (Winslow et al., 1992) and appears to be produced in the prostate (Yki-Järvinen et al., 1983; Sokol et al., 1989). RXFP1 is expressed in sperm in both mice (Krajnc-Franken et al., 2004) and humans (Carrell et al., 1995; Ferlin et al., 2012) and increases sperm motility and facilitates penetration into oocytes (Weiss, 1989). Another potential role for seminal relaxin may be to act on the female reproductive tract to prepare the endometrium for implantation (Telgmann and Gellersen, 1998; Unemori et al., 1999; Anand-Ivell and Ivell, 2014).

2. Central Nervous System.

In mammals, there are high concentrations of RXFP1 in the SFO and OVLT that are activated by relaxin to cause a reduction in plasma osmolality (Sunn et al., 2002). This effect in rats is associated with increased serum relaxin levels during the second half of pregnancy (Sherwood et al., 1980; Lindheimer et al., 1989) and is absent in pregnant rats that have undergone ovariectomy or treatment with relaxin antibodies (Novak et al., 2001). Similarly, the decrease in plasma osmolality with late pregnancy is not observed in relaxin knockout mice (Zhao et al., 1999). In humans, there is a decrease in plasma osmolality with pregnancy, but this may not be completely attributable to relaxin because women who become pregnant after ovum donation (and hence lack circulating relaxin) display attenuated (Smith et al., 2006b) or no decrease in plasma osmolality (Johnson et al., 1991, 1996). However, the apparent lack of relaxin in the absence of a corpus luteum may have to be reassessed using more sensitive methods of detection since the discovery of responses to femtomolar concentrations of relaxin (Halls and Cooper, 2010), well below the level of detection using current ELISA methods; relaxin is also produced locally in many tissues including the uterus. Relaxin administered intracerebroventricularly or intravenously is dipsogenic in nonpregnant rats (Summerlee et al., 1998a), and rats given relaxin monoclonal antibodies in the second half of pregnancy show a reduction in water consumption. Administration of porcine or human relaxin intravenously causes activation (increased c-fos expression) of neurons in the peripheral and dorsal segments of the SFO and in the dorsal cap region of the OVLT as well as in the supraoptic and paraventricular nuclei of the hypothalamus (McKinley et al., 1997, 1998; Sunn et al., 2001, 2002), all sites of localization of RXFP1 and accessible to circulating relaxin (Weisinger et al., 1993).

Relaxin acting on RXFP1 in the circumventricular organs and hypothalamic nuclei may have a role in the timing of parturition, because this is disrupted by central administration of a relaxin monoclonal antibody (Summerlee et al., 1998b). RXFP1 is also highly expressed in the basolateral amygdala, and administration of relaxin to this region impairs fear-related memory consolidation in rats (Ma et al., 2005). However, no specific agonist or antagonist studies have been carried out to determine whether these effects are mediated by RXFP1 or to determine the source of endogenous relaxin that activates these receptors (Callander and Bathgate, 2010). Although RXFP1 is also highly expressed in other regions associated with memory formation, such as the neocortex, thalamic nuclei, hippocampus, and supramammillary nucleus, there are no studies to date that examine effects on memory. RXFP1 is also present in the oxytocin-containing cells of the paraventricular and supraoptic hypothalamic nuclei (Burazin et al., 2005), and intravenous relaxin administration increases oxytocin neuron activity and oxytocin release (Way and Leng, 1992)

3. Blood Vessels.

Relaxin plays an important role in many of the adaptive cardiovascular changes that occur in pregnancy (Conrad, 2011). These include increases in cardiac output, heart rate, and a decrease in vascular resistance (Conrad et al., 2004; Conrad and Novak, 2004; Debrah et al., 2005, 2006). Increases in plasma volume and decreased blood pressure are not consistently observed (Ahokas et al., 1989; Conrad, 2004, 2010, 2011). Ovariectomy or passive immunization with monoclonal antibodies for rat relaxin prevent these adaptive changes (Novak et al., 2001). In both female and male rats, chronic relaxin administration (osmotic minipump) increases renal plasma flow and glomerular filtration rate (Danielson et al., 1999). In humans, a few studies suggest similar effects of relaxin on the cardiovascular system (Conrad and Shroff, 2011). In the clinical trial for scleroderma, long-term (6 months) infusion of relaxin increased creatinine clearance and modestly decreased blood pressure (Erikson and Unemori, 2001; Teichman et al., 2009). In the pilot trial in patients suffering from stable chronic heart failure, a short (24 hours) infusion of relaxin was associated with decreased systemic and pulmonary vascular resistance and pulmonary wedge pressure and elevated cardiac index without significantly affecting systolic or mean blood pressure (Dschietzig et al., 2009c; Teichman et al., 2009).

Vasodilation in arterioles, capillaries, and venules is a common response to relaxin in reproductive tissues (Vasilenko et al., 1986; Bani et al., 1988; Lee et al., 1992), heart (Bani-Sacchi et al., 1995; Masini et al., 1997; Bani et al., 1998b), liver (Bani et al., 2001), and cecum (Bigazzi et al., 1986). Relaxin is a potent vasodilator in arteries (Conrad, 2010; McGuane et al., 2011b), although the effect is vessel specific (McGuane et al., 2011b). Relaxin is a physiologic antagonist of vasoconstrictors in mesenteric arteries (St-Louis and Massicotte, 1985; Massicotte et al., 1989), primary bovine aortic smooth muscle cells (Bani et al., 1998a), and uterine artery (Longo et al., 2003). The vasodilator effects in guinea pig and rat coronary arteries and in bovine cultured smooth muscle cells are associated with increased NO synthesis (Bani-Sacchi et al., 1995; Bani et al., 1998a). The rise in intracellular calcium produced by α-thrombin or angiotensin II is also reduced by relaxin (Bani et al., 1998a; Failli et al., 2002). In humans, there are vasodilator effects in gluteal resistance or subcutaneous arteries but little or no effect in pulmonary, myometrial, or placental vessels (Petersen et al., 1991; Fisher et al., 2002; McGuane et al., 2011b). In gluteal arteries, the vasodilator responses likely involve NO and interestingly were influenced by the medication being administered to patients. Arteries obtained from patients on angiotensin converting enzyme inhibitors showed marked attenuation of the vasodilator response to relaxin, effects that appeared to be further enhanced by inhibition of cyclooxygenase (Fisher et al., 2002). However, in patients not receiving angiotensin converting enzyme inhibitors, inhibition of cyclooxygenase had little effect on the vasodilator response to relaxin (Fisher et al., 2002). The vasodilator mechanisms suggested for relaxin in both humans and in animal models involve activation of NOS (Nistri and Bani, 2003; Conrad and Novak, 2004), VEGF, placental growth factor (PGF), matrix metalloproteinases, ET_B receptors (Novak et al., 2002; Dschietzig et al., 2003), and modification of the extracellular matrix of the vessel walls (Jeyabalan et al., 2003; Lekgabe et al., 2005; Xu et al., 2010; McGuane et al., 2011b). These responses have a distinct temporal hierarchy. The acute (within minutes) responses are endothelium dependent and blocked by NOS inhibitors, the PI3K inhibitors Wortmannin and LY294002, and by PTX pretreatment but not by the VEGF receptor antagonist SU5416 [(3Z)-3-[(3,5-dimethyl-1H-pyrrol-2-yl)methylidene]-1,3-dihydro-2H-indol-2-one] (McGuane et al., 2011b). The effects are consistent with Gβγ activation of PI3K, Akt phosphorylation, and eNOS. Treatment with NOS inhibitors or removal of the endothelium enhances contractions of vascular smooth muscle to agonists in both pregnant rats (Danielson and Conrad, 1995; Gandley et al., 2001) and in rats treated chronically with relaxin (Novak et al., 2002). This does not appear to be attributable to changes in NOS expression, because alterations in renal hemodynamics in pregnancy are not accompanied by large changes in NOS expression. In pregnant rats, eNOS expression in the renal artery fell by 39%, whereas iNOS and nNOS expression increased by 31 and 25%, respectively (Alexander et al., 1999).

Longer-term (within hours) vasodilator responses to relaxin also involve NOS, because blockade with NG-monomethyl-l-arginine prevents the renal hemodynamic and hyperfiltration responses (Danielson et al., 1999), and these effects can be recapitulated in vitro (Novak et al., 2002). The longer-term response involves endothelial ET_B receptors that release NO (Dschietzig et al., 2003; Jeyabalan et al., 2003). Two mechanisms involving the relaxin-ET_B receptor pathway have been found. In rodent and human vessels of ∼100–300 µm diameter, relaxin activates MMP-9 and later also MMP-2 to generate ET-1(1–32), rather than ET-1(1–21), from big ET-1 (Jeyabalan et al., 2003). The effect of MMPs on big ET-1 has been described independently of relaxin (Fernandez-Patron et al., 1999). ET-1(1–32) then preferentially binds to the endothelial ET_B receptor causing NO release. In blood vessels from pregnant or relaxin-treated nonpregnant rats, pro–MMP-2 and MMP-2 activity and pro–MMP-2 protein and mRNA are increased (Jeyabalan et al., 2003, 2006). MMP-9 activity also appears to be increased somewhat, although more recent studies suggest that MMP-9 is more important in the relatively short-term responses to relaxin (4–6 hours) with reversal of the effects being produced by MMP-9 rather than MMP-2–neutralizing antibodies (Jeyabalan et al., 2007). This is supported by studies with the selective MMP-2 inhibitor cyclic CTTHWGFTLC, the MMP inhibitor GM6001 [(2R)-N′-hydroxy-N-[(2S)-3-(5H-indol-3-yl)-1-methylamino-1-oxopropan-2-yl]-2-(2-methylpropyl)butanediamide], or a TIMP-2 or MMP-2 neutralizing antibody, all of which inhibit the vasodilator actions of relaxin in renal arteries (Jeyabalan et al., 2003), whereas inhibition of the formation of ET(1–21) by phosphoramidon has no effect. These studies clearly implicate MMPs in the vasodilator actions of relaxin, although MMPs also cleave calcitonin gene-related peptide to promote vasoconstriction (Fernandez-Patron et al., 2000). These findings correspond well with those obtained with the ET_B antagonist RES-701-1 [cyclic (Gly1-Asp9)(Gly-Asn-Trp-His-Gly-Thr-Ala-Pro-Asp-Trp-Phe-Phe-Asn-Tyr-Tyr-Trp)] that blocks renal hemodynamic changes produced by relaxin (Danielson et al., 2000) in rats and antagonizes inhibition of renal artery smooth muscle produced by relaxin or in pregnancy (Novak et al., 2002). Similar effects were produced in vitro by the ET receptor antagonist SB209670 [(+)-(1S,2R,3S)-3-(2-carboxymethoxy4-methoxyphenyl)-1-(3,4-methylenedioxyphenyl)-5-(prop-1-yloxy)indane-2-carboxylic acid] but not by the ET_A selective BQ123 (2-[(3R,6R,9S,12R,15S)-6-(1H-indol-3-ylmethyl)-9-(2-methylpropyl)-2,5,8,11,14-pentaoxo-12-propan-2-yl-1,4,7,10,13-pentazabicyclo[13.3.0]octadecan-3-yl]acetic acid) (Gandley et al., 2001).

An alternative mechanism that occurs mainly in larger vessels and in the pulmonary circulation (Dschietzig et al., 2001a,b, 2003) involves relaxin-induced upregulation of endothelial ET_B receptor requiring Raf-1, MEK-1, ERK-1/2, and NFκB (Dschietzig et al., 2003). This mechanism spares the smooth muscle ET_B receptor, a contributor to ET-1–related vasoconstriction. Thus, relaxin acts as a functional ET-1 antagonist, because the endothelial ET_B receptor mediates NO production and, equally importantly, ET-1 clearance (Dschietzig et al., 2001b, 2003). Although another study does not support this mechanism (Kerchner et al., 2005), this can be explained by the use of small rather than large arteries, no pulmonary endothelial cells, and significantly lower relaxin concentrations.

Evidence also exists for a role of VEGF in the longer-term vasodilator actions of relaxin. In human endometrial cells, relaxin increases cAMP levels and VEGF expression, and the effects are prevented by AC inhibition and mimicked by forskolin or a PDE inhibitor (Unemori et al., 1999), suggesting that relaxin-stimulated cAMP production mediates increased VEGF transcription and, thus, angiogenesis. However, although treatment of rat or mouse renal arteries or human subcutaneous arteries with the VEGF receptor antagonist SU5416 blocks the vasorelaxant effects of relaxin (McGuane et al., 2011a), the specificity of SU5416 has been questioned (Arora and Scholar, 2005; Loges et al., 2006), and in rat renal arteries, SU5416 potentiates the acute vasodilator effects of relaxin (McGuane et al., 2011b). Furthermore, although VEGF-neutralizing antibodies block the effects of relaxin, this can also be mimicked by placental growth factor antibodies (McGuane et al., 2011a), again making interpretation of the results difficult. It is possible that the antibodies, by blocking the vasodilator effects of VEGF and PGF, may produce physiologic antagonism of the response to relaxin without necessarily being directly related to the primary mechanism of action. Alternatively, these results may reflect MMP-2- or MMP-9-mediated release of extracellular matrix-bound VEGF (and PGF) (Bergers et al., 2000; van Hinsbergh et al., 2006) that could underpin the MMP-ET_B pathway by stimulating expression of prepro-ET-1.

Finally, relaxin promotes angiogenesis (Unemori et al., 1999; Unemori et al., 2000) by inducing VEGF and basic fibroblast growth factor synthesis. These effects are likely important in promotion of wound healing (Bitto et al., 2013) and also in the heavy, irregular, or prolonged menstrual bleeding that accompanies relaxin administration (Unemori et al., 1999). Relaxin also increases arterial compliance by its effects on geometric remodeling (increases in unstressed wall area and wall-to-lumen area ratio) and compositional remodeling (decrease in collagen-to-total protein ratio) in certain arteries (Chan and Cipolla, 2011; Debrah et al., 2011; Gooi et al., 2013). These properties expedite tissue and organ perfusion in the long term.

Compared with arteries, remarkably little is known of the effects of relaxin on veins. It was reported that the vasodilator potency of relaxin is lower in veins than arteries (Li et al., 2005). Rat mesenteric veins, in contrast to arteries, are not remodeled by relaxin (Jelinic et al., 2014). It is important to recognize that venous tone is significantly more dependent on sympathoadrenergic innervation and pacemaker cells (Cajal cells) than arterial tone (Gelman, 2008). Given the interest in the utility of vasodilators in the treatment of cardiac failure (Bhushan et al., 2014), this promises to be an important future area of research.

Relaxin protects against endothelial dysfunction that denotes an inflammatory and oxidative stress-related vascular pathology characterized by compromised vasodilator capacity on the basis of reduced NO bioavailability. In a rat in vitro model (Dschietzig et al., 2012) in which TNF-α impairs endothelial function, relaxin, by a mechanism involving PI3K and GR signaling, attenuates the upregulation of arginase-II that deprives NOS of its substrate l-arginine; antagonizes the increased ET-1 expression that promotes vasoconstriction and inflammation; upregulates SOD-1 expression to combat oxidative/nitrosative stress; and finally, causes partial reversal of dephosphorylation of eNOS, which impairs NO generation. Others have confirmed the SOD-1 finding and also reported elevated SOD-2 expression (Collino et al., 2013). This vascular protection is further complemented by NO-mediated inhibition of platelet and neutrophil activation (Bani et al., 1995a,b; Masini et al., 2004). Relaxin also attenuates VCAM-1 and MCP-1 stimulation that mediate the initial monocyte-endothelium contact and are both key events in early vascular inflammation (Brecht et al., 2011).

4. Heart.

Relaxin acts directly on the heart. The presence of RXFP1 in the heart was first suggested by the demonstration of high-affinity binding sites for relaxin in rat atria (Osheroff et al., 1992; Osheroff and Ho, 1993; Tan et al., 1999). Subsequently, relaxin was shown to be a powerful inotropic and chronotropic agent (Kakouris et al., 1992). Positive chronotropic effects of relaxin occur in perfused intact hearts (Thomas and Vandlen, 1993; Bani-Sacchi et al., 1995; Coulson et al., 1996; Toth et al., 1996) and isolated right atria (Kakouris et al., 1992; Tan et al., 1998; Ward et al., 1992; Wade et al., 1994; Mathieu et al., 2001), and the positive inotropic effects occur in left atria (Kakouris et al., 1992; Ward et al., 1992; Wade et al., 1994; Tan et al., 1998; Mathieu et al., 2001). Chronotropic effects of relaxin are accompanied by the secretion of atrial natriuretic peptide in isolated perfused rat hearts (Toth et al., 1996). In rabbit sinoatrial node cells, relaxin increases the rate of spontaneous action potentials and the L-type calcium current (Han et al., 1994) by a PKA-dependent mechanism. In rat atrial myocytes, relaxin inhibits outward potassium currents, increases action potential duration, and enhances calcium entry; this is the mechanism underlying the positive inotropic effect (Piedras-Renteria et al., 1997a,b). Regarding inotropic effects in the ventricle, there is one report in mice (Shaw et al., 2009) indicating that relaxin increases myofilament calcium sensitivity by promoting phosphorylation of myosin-binding protein C, troponin I, and troponin C. This results in higher force generation at normal calcium levels and is dependent on myofilament-associated PKCδ. This is the only report of inotropic responses to relaxin in ventricles. Although the cardiac actions of relaxin are largely confined to rodents, a recent study (Dschietzig et al., 2011) demonstrates inotropic effects in human atria (but not in ventricular myocardium), which are preserved in failing hearts and involve PKA, outward potassium currents, and PI3K. The effects of relaxin on the cardiovascular system are not gender specific.

Relaxin also protects against myocardial injury caused by ischemia and reperfusion (Bani et al., 1998b). Pretreatment of rats with relaxin 30 minutes before cardiac ischemia produced by ligating the left anterior descending coronary artery markedly reduced the size of the penumbra and reduced cardiac arrhythmias, mortality, myeloperoxidase activity, malonyldialdehyde production, and calcium content as well as causing an improved morphology (Bani et al., 1998b). These findings were confirmed in an in vivo pig model of myocardial infarction in which relaxin, administered during reperfusion of the left anterior descending coronary artery, attenuated leukocyte recruitment and oxidative damage and improved contractile recovery (Perna et al., 2005). In a mouse infarction model, relaxin improves postinfarction remodeling by suppressing reactive fibrosis in vital myocardium, without affecting reparative fibrosis (scarring) within the infarcted area (Samuel et al., 2011). This postinfarction fibrosis-modulating effect could also assist stem cell-based reparative therapies that appear to be promoted by the peptide. Mouse C2C12 skeletal myoblasts engineered to overexpress human relaxin demonstrate growth and functional maturation of neonatal immature cardiomyocytes assessed by the expression of myocardium-specific structural genes (connexin 43, troponin T, and HCN4 ion channel) (Formigli et al., 2009). Coronary infusion of C2C12 skeletal myoblasts overexpressing relaxin in conjunction with administration of exogenous relaxin starting at day 30 after infarction in rats improves myocardial viability in the infarcted area compared with myoblast therapy alone (Bonacchi et al., 2009).

Relaxin also attenuates hypertrophy in rat neonatal cardiomyocytes by inhibiting myofibroblast activation and the subsequent paracrine release of growth factors (Moore et al., 2007), confirming an earlier study in spontaneously hypertensive rats and suggesting that endogenous relaxin has antihypertrophic actions (Dschietzig et al., 2005).

Atrial fibrillation, a common arrhythmia that develops during ageing-related fibrosis, hypertension, or heart failure, is suppressed by relaxin in spontaneously hypertensive rats, an established model of end-organ damage in response to hypertension (Parikh et al., 2013). The aforementioned effects of relaxin on atrial hypertrophy and fibrosis resulted in restoration of normal conduction velocity, a reduced action potential duration, and hence decreased atrial fibrillation.

Relaxin may be a naturally occurring cardioprotective agent, because in chronic heart failure, the expression of human relaxin-1 and human relaxin is increased in both atria and ventricles, and the level of expression correlates with the degree of failure (Dschietzig et al., 2001b; Fisher et al., 2003) but is not a predictor of clinical outcomes (Fisher et al., 2003). Increased circulating relaxin levels in human heart failure were not confirmed in two subsequent clinical studies (Kupari et al., 2005; Heringlake et al., 2009), possibly due to a methodological issue related to the utilization of a new batch of antiserum (Kupari et al., 2005; Heringlake et al., 2009) for the ELISA (Dschietzig et al., 2001b; Fisher et al., 2002). Increased relaxin gene expression in chronic heart failure was confirmed by two independent studies using the rat infarction model (Kompa et al., 2002; Zhang et al., 2005).

After the demonstration of protective actions in the cardiovascular system, relaxin was tested in human heart failure. A hemodynamic pilot study in patients with stable chronic heart failure (Dschietzig et al., 2009c) demonstrated that a 24-hour intravenous infusion of recombinant human relaxin markedly elevated cardiac index without affecting heart rate and decreased pulmonary wedge pressure, systemic, and pulmonary vascular resistance without affecting systolic or mean arterial blood pressure and central venous pressure. Hemodynamic changes occurred within 45 to 60 minutes, and relaxin infusion improved creatinine and blood urea nitrogen clearance and was free of relevant side effects.

The use of relaxin for acute heart failure (defined as new onset or, more frequently, worsening of known heart failure) is promising. In the Pre-RELAX (phase II) and RELAX (phase III) trials (Teerlink et al., 2009, 2013), relaxin moderately improved dyspnea (the primary endpoint), was exceptionally safe, improved renal function, and lowered all-cause as well as cardiovascular mortality at day 180. Because both trials were not powered statistically to assess mortality, another adequately powered mortality trial in acute heart failure is in progress. If successful, relaxin would become the first evidence-based remedy for this highly challenging clinical syndrome.

5. Kidney.

The pronounced effects of relaxin on renal arteries imply a pivotal role in the regulation of renal function. Indeed, relaxin is the major renal vasodilator responsible for the increases in renal plasma flow and glomerular filtration rate during rat pregnancy (Novak et al., 2001). The increase in renal plasma flow and glomerular filtration rate with relaxin also occurs in nonpregnant female rats and male rats (Danielson et al., 1999; Danielson and Conrad, 2003) and is accompanied by natriuresis (Bogzil et al., 2005).

In humans, circumstantial evidence for a role for relaxin comes from a clinical study in which women who were pregnant through egg donation, and therefore relaxin deficient, exhibited a decreased glomerular filtration rate compared with that observed in normal pregnancies (Smith et al., 2006b). In a small study in human volunteers (Smith et al., 2006a), relaxin administered for 5 hours increased renal plasma flow by ∼75% and fractional sodium excretion by ∼25% but did not affect glomerular filtration rate (Smith et al., 2006a). The authors suggest that the volunteers may not have been volume replete, which may offset the effects of relaxin at the afferent arteriole. Alternatively, humans, in contrast to rats, are not typically in glomerular filtration equilibrium, thereby allowing dissociation of the effects of relaxin on renal plasma flow and glomerular filtration rate (Smith et al., 2006a), or the effect may take longer to develop via alteration of the filtration coefficient. For example, during chronic relaxin infusion in the scleroderma trials, the estimated glomerular filtration rate rose significantly in the treated group (Seibold et al., 2000; Khanna et al., 2009). However, the vasodilator effects of relaxin may also influence creatinine handling in the bowels, a process that is usually upregulated in states of renal insufficiency such as scleroderma or congestive heart failure. The renal effects of relaxin in humans therefore remain to be confirmed.

6. Fibrosis.

The antifibrotic effects of relaxin were the first biologic effects to be recorded (Hisaw, 1926), and a number of studies have examined this property as a possible treatment of the connective tissue disease scleroderma. Relaxin was safe and well tolerated in clinical trials and effective in some patients in a phase II trial (Seibold et al., 2000) but failed to show clinical efficacy in a larger scale phase III trial (Erikson and Unemori, 2001). Despite these disappointing findings there are many studies in animals that show a role for relaxin in controlling collagen turnover. In relaxin knockout mice, there is a progressive increase in tissue fibrosis with age in male mice that is prevented (Samuel et al., 2007) or reversed in lung (Samuel et al., 2003b), kidney (Samuel et al., 2004b), and heart (Samuel et al., 2004a) by the administration of relaxin.

In the lung, treatment with relaxin reduces expression of collagen types I and III, increases levels of MMPs, and reduces fibrosis (Unemori et al., 1996). In kidney-derived fibroblasts, relaxin inhibits profibrotic changes induced by TGF-β by a mechanism involving the NO/guanylyl cyclase pathway with associated decreases in Smad2 phosphorylation and nuclear localization (Mookerjee et al., 2009). Relaxin has antifibrotic effects in rat renal fibrosis models, including fibrosis produced by bromoethylamine treatment (Garber et al., 2001), in an antiglomerular basement membrane model (McDonald et al., 2003) and in spontaneously hypertensive rats (Lekgabe et al., 2005). In cardiac fibroblasts, relaxin reduces collagen type I and III expression and increases MMPs (Samuel et al., 2004a). In cardiac fibrosis after chronic stimulation of β-adrenoceptors by isoprenaline (Zhang et al., 2005) or by cardiac-specific transgenic overexpression of β₂-adrenoceptors (Bathgate et al., 2008), relaxin treatment markedly reduced cardiac fibrosis. In streptozotocin-treated mRen-2 rats (a model for diabetic cardiomyopathy), relaxin reduced left ventricular collagen, myocardial stiffness, and diastolic dysfunction (Samuel et al., 2008), associated with a significant decrease in TIMP-1 expression and an increase in extracellular matrix-degrading MMP-13 (Samuel et al., 2008). Relaxin also has antifibrotic actions in a number of models of liver fibrosis (Bennett, 2009). However in a chronic pressure overload model in mice, relaxin was ineffective (Xu et al., 2008), possibly because cardiac RXFP1 may be downregulated in this model or the serum concentrations of relaxin are insufficient to offset the extensive fibrosis.

Relaxin acting via RXFP1 in primary fibrochondrocytes increases the expression of mRNA for MMP-9 and MMP-13 (Ahmad et al., 2012) associated with activation of PI3K, Akt, PKCζ, and ERK1/2; inhibitors of these proteins blocked MMP-9 induction and the antifibrotic effects of relaxin. Relaxin-mediated increases in MMP-9 expression were also blocked by transfection of a dominant negative form of Akt or by small interfering RNA knockdown of ERK1/2, PKCζ, Elk-1, c-fos, and, to a lesser extent, NFκB (Ahmad et al., 2012). This important study connects many of the known pathways of relaxin/RXFP1 signaling to a well recognized physiologic response to relaxin and provides insights that could be useful in translating the antifibrotic effects of relaxin into a clinical setting.

The apparent discrepancy between the highly successful preclinical antifibrosis studies and the failure of relaxin in the phase III human scleroderma trial can be explained by an incomplete understanding of the pathophysiology of scleroderma and by the enrolment of patients in the terminal stages of the disease. It is of interest to note that the dermal fibrosis of the relaxin knockout mouse is rescued by exogenous relaxin at 6 months but not rescued by administration at 12 months of age (Samuel et al., 2005).

One of the interesting aspects that have emerged is that the antifibrotic properties of relaxin are clearly seen in disease conditions associated with excessive collagen deposition. Explanations for this unusual and potentially useful property of relaxin are now starting to emerge. In rat kidney myofibroblasts, it was discovered that the antifibrotic actions of relaxin are completely abolished by the AT₂R antagonist PD123319 (Chow et al., 2014). In an in vivo fibrosis model in mice, produced by unilateral ureteric obstruction, the protective effects of relaxin are lost when the AT₂R is either absent (in AT₂R^−/y mice) or blocked by PD123319, confirming that the AT₂R is obligatory for the antifibrotic actions of relaxin. These effects are not produced by relaxin interacting directly with the AT₂R, because the peptide does not bind to this receptor. However, it was found that RXFP1 can form constitutive heterodimers with AT₂R in cell-based models, and the RXFP1-AT₂R heterodimers may mediate the downstream signaling pathways originally attributed to relaxin (Heeg et al., 2005; Mookerjee et al., 2009; Chow et al., 2012), including inhibition of the TGF-β₁/pSmad2 axis and reduced TGF-β₁–induced collagen deposition. Because AT₂Rs are expressed at low levels in tissues (Matsubara, 1998; Carey, 2005; Jones et al., 2008) and fibroblasts under physiologically quiescent states but are dramatically increased in number and activity under pathologic conditions, their increased availability in injured/diseased tissues increases their functional importance (Siragy and Carey, 1997; Savoia et al., 2006). Hence, pathologic conditions create an environment in which RXFP1-AT₂R heteromerization is more likely to take place and may help explain why relaxin displays its antifibrotic effects to a pronounced extent under pathologic conditions.

Relaxin also has beneficial effects in wound healing (Casten and Boucek, 1958) that may involve the vasodilator effects mentioned above but, in addition, may involve the synthesis of new blood vessels by enhancing the local production of VEGF (Unemori et al., 2000).

7. Organ Protection.

Relaxin not only protects the myocardium from ischemia-reperfusion (IR) injury (see above) but also protects other tissues. Thus, there is protection from IR injury in rat liver (Boehnert et al., 2005, 2008, 2009) by attenuating oxidative stress, leukocyte activation, and improving oxygen supply. In kidney IR in rats, relaxin reduces renal apoptosis and TNF-α levels as well as creatinine and urea levels (Yoshida et al., 2013). In a similar model, relaxin decreases inflammatory cytokines, counteracts oxidative damage by increasing SOD-1 and SOD-2 expression, and ameliorates neutrophil-related injury as assessed by myeloperoxidase levels (Collino et al., 2013). Finally, the beneficial effects of relaxin are also observed in IR experiments in isolated rat lungs (Alexiou et al., 2010, 2013) where there is inhibition of leukocytes and oxidative surge, as well as prevention of ET-1 stimulation, resulting in less edema formation and lower pulmonary vascular pressure.

NO appears to be critically involved in organ protection (Masini et al., 1997; Alexiou et al., 2010, 2013; Collino et al., 2013). In rat lung, relaxin evokes an early and moderate increase in iNOS dependent on ERK1/2 activation that is counterbalanced by PI3K recruitment (Alexiou et al., 2013). In rat kidney (Collino et al., 2013), the peptide upregulates iNOS and activates eNOS via PI3K. It remains to be investigated whether over the long term, additional modes of action, such as the actions of relaxin on the GR pathway, come into play.

8. Formation and Spread of Tumors.

Relaxin is expressed in endometrial (Kamat et al., 2006), mammary (Tashima et al., 1994), thyroid (Hombach-Klonisch et al., 2006), and prostate tumors (Thompson et al., 2006; Feng et al., 2007). The peptide is associated with breast cancer (Bani, 1997; Silvertown et al., 2003), and relaxin treatment of breast cancer cells implanted into nude mice increases their invasive potential (Binder et al., 2002). In contrast, longer-term application of relaxin (up to 8 days) reduces mammary xenograft growth in mice (Radestock et al., 2008). The elevated serum relaxin levels that have been reported in breast cancer patients are associated with metastatic disease (Binder et al., 2004). Relaxin is also associated with prostate cancer progression in the mouse xenograft model (Silvertown et al., 2006), and blocking the actions of relaxin or its receptor in this model decreases cancer growth (Feng et al., 2010). In patients with prostate cancer, elevated relaxin levels are linked to cancer progression, metastasis, and androgen independence (Thompson et al., 2006). Likewise, relaxin heightens the collagenolytic potency of thyroid cancer cells by upregulating MMP-2 that facilitates greater in vitro invasiveness (Bialek et al., 2011). In human osteosarcoma cells, relaxin promotes faster in vitro growth, invasion, and angiogenesis via the Akt and VEGF pathways, and small interfering RNA–mediated knockdown of relaxin greatly mitigates these effects (Ma et al., 2013a,b). The peptide hastens the proliferation of human leiomyoma cells (Suzuki et al., 2012). A clinical study in 108 patients with hepatocellular carcinoma identified expression of relaxin in tumor tissue as an independent predictor of poor prognosis (Pan et al., 2013). Similarly, high plasma relaxin levels were associated with a worse prognosis in 146 patients with esophageal squamous cell carcinoma (Ren et al., 2013). A recent study also suggests that CTRP8 interacts with RXFP1 to facilitate cell migration in brain cancer, dependent on activation of PI3K and PKC (Glogowska et al., 2013). In summary, relaxin does not appear to initiate cancer, but, like other growth and angiogenic hormones, it may promote cancer growth and/or spread by its actions on enhanced matrix degradation and angiogenesis. To date, though lacking definitive proof, this might be clinically relevant for mammary, thyroid, and prostate cancer.

9. Diabetes.

Although relaxin shares many structural similarities with insulin (Schwabe and McDonald, 1977), the peptides interact with different receptors and activate quite different signal transduction mechanisms. RXFP1 is a GPCR, whereas insulin receptors are tyrosine kinases and relaxin and insulin do not display crossreactivity. Although this would appear to argue against any connection between the hormones, there are studies that suggest that relaxin can influence carbohydrate metabolism. For instance, many insulin-dependent diabetic women have hypoglycemic episodes during the first trimester of pregnancy, when circulating levels of relaxin are higher, whereas increased insulin may be required during the third trimester, when relaxin concentration are lowest (O'Byrne et al., 1978). Additionally, intraperitoneal dosing of human relaxin inhibits ad libitum feeding in male rats when given during the early dark phase of the circadian clock (McGowan et al., 2010).

Relaxin concentrations in the serum of diabetic women are higher than those observed in the serum of nondiabetic women during each trimester of pregnancy (Steinetz et al., 1992; Whittaker et al., 2003). One possible explanation is that the increased relaxin is a physiologic response to the insulin-resistance that occurs during pregnancy. More circumstantial evidence comes from studies showing that circulating relaxin levels correlate positively with insulin sensitivity and inversely with β-cell function in women with type-2 diabetes mellitus (T2DM) (Szepietowska et al., 2008) and studies showing an increase in plasma relaxin after successful antidiabetic therapy in men with T2DM (Schondorf et al., 2007). The cellular mechanism is not clear, but it has been reported that relaxin increases the binding of insulin to its receptors on adipocytes (Olefsky et al., 1982; Jarrett et al., 1984) an effect that would enhance glucose uptake. In rat cardiac fibroblasts, high-glucose conditions increase the expression of endogenous relaxin (Wang et al., 2009). In human amniotic epithelial cells, relaxin enhances expression of insulin-like growth factor-2 (Millar et al., 2003). In preclinical studies in lean mice that are insulin-resistant after a high-fat diet, relaxin increases glucose uptake into skeletal muscle by improving hemodynamics and angiogenesis that is known to be an important factor in determining the response to insulin (Bonner et al., 2013). More recently, relaxin applied over 12 days in a genetic mouse model of T2DM lowered blood glucose levels, from an average of 660 mg/dl in vehicle-treated animals to 490 mg/dl (Bitto et al., 2013). In addition, relaxin has also been shown to activate peroxisome proliferator-activated receptor-γ and enhance the actions of the peroxisome proliferator-activated receptor-γ activator rosiglitazone in cells expressing RXFP1 (Singh and Bennett, 2010). In this way relaxin treatment would be expected to influence insulin sensitivity.

Originally identified for its effects on the female reproductive tract, relaxin has now been recognized to have actions in the central nervous system (CNS), cardiovascular system, and kidney, on glucose metabolism and in cancer metastasis in both males and females. Although it has important and well defined roles in pregnancy and reproduction in many species, it is not mandatory for successful pregnancy in humans. The vasodilatory and antifibrotic actions of relaxin on blood vessels are the subject of intense interest, and relaxin was recently shown to produce inotropic responses in human atria. Studies of the antifibrotic actions of relaxin indicate that the effects are dependent on expression of the AT₂R, raising the question whether heterodimer formation between RXFP1 and other GPCRs underpins other physiologic responses to the hormone. Connections have also been established between relaxin and the formation and spread of tumors, although the mechanisms involved are less clear. Relaxin also appears to improve insulin sensitivity and improves glucose tolerance in animal models of T2DM.

1. Reproduction.

INSL3 is primarily a reproductive hormone and was first cloned from testicular cDNA libraries (Adham et al., 1993; Pusch et al., 1996). The peptide is secreted by testicular Leydig cells (Adham et al., 1993; Pusch et al., 1996) but is also found in human cyclic corpora lutea (Tashima et al., 1995); ruminant ovary, uterus, and placenta (Bathgate et al., 1996; Roche et al., 1996); mouse (Zimmermann et al., 1997); and marmoset ovary (Zarreh-Hoshyari-Khah et al., 1999). Importantly INSL3 is a circulating hormone in both males (Foresta et al., 2004) and females (Anand-Ivell et al., 2013), although the levels are much higher in males.

The best characterized physiologic role for INSL3 is in the male where it controls testis descent. Male INSL3 knockout mice show normal development but are infertile and bilaterally cryptorchid, with the testis located high in the abdominal cavity adjacent to the kidney (Nef and Parada, 1999; Zimmermann et al., 1999). A similar phenotype is displayed in RXFP2 knockout mice, confirming the cryptorchid phenotype (Overbeek et al., 2001; Gorlov et al., 2002). Thus, the INSL3/RXFP2 system appears to be essential for the development of the gubernaculum during embryogenesis and for normal transabdominal testicular descent.

Cryptorchidism is the most common birth defect of the male genitalia; affecting 1–4% of live male births, with a greater incidence in premature infants (Ivell and Hartung, 2003). Although the studies in INSL3 or RXFP2 knockout mice suggested that mutations in these genes could account for cryptorchidism, this is unlikely to be the case in all human infants, because few of these mutations result in a functional change in either the peptide or the receptor (Tomboc et al., 2000; Gorlov et al., 2002). Furthermore, because all mutations of INSL3 and RXFP2 identified in cryptorchid patients have been found only as heterozygotes and the expression of INSL3 in testicular Leydig cells is high, it is unlikely that a mutation in a single allele would be sufficient to induce bilateral cryptorchidism.

INSL3 is highly expressed in the testicular Leydig cells in adult mammals and is considered an important marker of Leydig cell function (Ivell and Anand-Ivell, 2009; Ivell et al., 2014). Additionally, INSL3 circulates at nanogram per milliliter levels in males and therefore has potential endocrine functions such as that on bone function. RXFP2 receptors are expressed on Leydig cells (Anand-Ivell et al., 2006) and on germ cells in the seminiferous tubules (Kawamura et al., 2004; Anand-Ivell et al., 2006; Filonzi et al., 2007). A recent study suggested that INSL3 influences steroid production by Leydig cells (Pathirana et al., 2012), and studies in male rats suggest that INSL3 is a paracrine factor involved in male germ cell survival (Kawamura et al., 2004). However, it should be noted that androgen production is normal in INSL3 and RXFP2 knockout mice (Nef and Parada, 1999; Zimmermann et al., 1999; Overbeek et al., 2001; Gorlov et al., 2002). Additionally, conditional deletion of RXFP2 in male germ cells resulted in normal fertility (Huang et al., 2012).

In females, INSL3 is produced in the thecal cells of the ovary and in the corpus luteum (Ivell and Bathgate, 2002). In mouse ovaries, Insl3 expression is higher in the follicular than the luteal phase (Zimmermann et al., 1997), and female INSL3 knockout mice demonstrate impaired fertility associated with longer estrous cycle length, smaller litter sizes, accelerated follicular atresia and luteolysis, and premature loss of corpora lutea (Nef and Parada, 1999; Spanel-Borowski et al., 2001). Further evidence that INSL3 is a circulating hormone of ovarian origin in females comes from the observation that postmenopausal women have undetectable plasma INSL3 levels (Ivell and Anand-Ivell, 2009). Studies in cohorts of adult women suggest that INSL3 plasma levels are indicative of the number of ovarian antral follicles recruited and growing within a follicular wave (Anand-Ivell et al., 2013). As INSL3 is a product of the thecal cells of ovarian follicles, plasma INSL3 levels have been investigated and shown to be elevated in polycystic ovary syndrome, a common female endocrine disorder characterized by excessive production of androgens from ovarian follicles (Gambineri et al., 2007; Szydlarska et al., 2012; Anand-Ivell et al., 2013). Importantly, a recent study in cows demonstrated that INSL3 has a positive autoregulatory role in maintaining thecal androgen production that is essential for normal ovarian follicle development (Glister et al., 2013).

2. Brain.

RXFP2 expression has been found in human brain by RT-PCR (Hsu et al., 2002) and in the rat parafascicular nucleus, dorsolateral, ventrolateral, and posterior thalamic nuclei and medial habenula by in situ hybridization histochemistry (Shen et al., 2005; Sedaghat et al., 2008). These sites correspond to the distribution of RXFP2 determined by [¹²⁵I]INSL3 binding (Sedaghat et al., 2008). Insl3 expression has been demonstrated in the rat brain (Gundlach, personal communication) and also in the bovine hypothalamus by RT-PCR (Bathgate et al., 1996) and Northern blotting (Bathgate et al., 1999). The effects of INSL3 administration to rat brain suggest that RXFP2 influences sensorimotor function (Sedaghat et al., 2008).

3. Bone and Other Tissues.

Physiologic functions of INSL3 in other tissues have been postulated based on INSL3 and RXFP2 expression patterns. In young adults with cryptorchid hypogonadism associated with the T222P mutation of RXFP2, 64% had reduced bone density but normal plasma testosterone levels (Ferlin et al., 2008). RXFP2 mRNA and protein are present in human osteoblasts, and an osteoblast cell line responded to INSL3 with a concentration-dependent increase in cAMP. Mouse osteoblasts also express RXFP2, but not Insl3, and RXFP2 knockout mice are osteopenic and have functional osteoblast impairment (Ferlin et al., 2008). INSL3 regulates expression of genes required for proliferation and differentiation, matrix deposition, and osteoclastogenesis in cultured human osteoblasts (Pepe et al., 2009). Deficits in INSL3/RXFP2 signaling are correlated with reduced bone mass (Ferlin et al., 2008, 2009). Thus INSL3 may have a role in bone physiology, and RXFP2 mutations may be linked with osteoporosis in men (Ferlin et al., 2013). Interestingly, recent studies suggest that RXFP2 is associated with horn growth in sheep (Johnston et al., 2013) and cows (Wiedemar et al., 2014), suggesting actions of INSL3 on bone is a common feature in male mammals.

In tumor biology, there is altered expression of INSL3 and a splice variant in human hyperplastic thyroid adenoma and thyroid cancer (Hombach-Klonisch et al., 2003). INSL3 is also present in human prostate carcinomas and increases motility in the human androgen-insensitive prostate carcinoma cell line PC-3 (Klonisch et al., 2005). All human thyroid adenomas, three types of human thyroid carcinomas, and mouse noncancerous follicular epithelial cells of the thyroid express RXFP2 mRNA (Hombach-Klonisch et al., 2010). Thyroid cancer cells expressing INSL3 demonstrate enhanced motility and increased colony formation in vitro and enhanced tumor growth in vivo. Finally, INSL3 increases levels of the calcium binding protein S100A4, which increases cancer cell mobility and enhances tumor tissue vascularization, suggesting that S100A4 is a downstream target of RXFP2 signaling in thyroid cells (Hombach-Klonisch et al., 2010). Thus INSL3 and RXFP2 appear to play a role in cancer metastasis.

RXFP2 mRNA expression is highest in rat kidney at late stage gestation and decreases dramatically at birth with lowest levels in adulthood (Fu et al., 2006). RXFP2 is expressed in mesangial cells in mature glomeruli and inhibits proliferation of cultured primary glomerular cells, suggesting that INSL3 and RXFP2 influence the genesis or maturation of renal glomeruli and regulate mesangial cell density (Fu et al., 2006). Pod1 is one of the main transcription factors involved in glomerulogenesis (Quaggin et al., 1999), and an E-box consensus sequence capable of binding Pod1 in conjunction with other helix loop helix transcription factors is located upstream of the gene for RXFP2 (Funato et al., 2003). Therefore, Pod1 may regulate expression of RXFP2 in the kidney during development to facilitate glomerular cell proliferation (Familari et al., 2009). RXFP2 expression levels in the embryonic kidney were significantly greater in Pod1 knockout mice than in heterozygous or wild-type controls, indicating that RXFP2 is downstream of Pod1 and Pod1 negatively regulates expression of RXFP2 in the glomeruli (Familari et al., 2009).

Although the INSL3 and RXFP2 system has primarily been studied for its specialized roles in both male and female reproductive physiology, there is emerging evidence to suggest that this system may have other roles in the CNS, in cancer metastasis, in bone physiology and in the kidney.

1. Stress.

Relaxin-3 and RXFP3 are present in hypothalamic and extrahypothalamic regions involved in the hypothalamic-pituitary-adrenal axis (Sutton et al., 2004; Liu et al., 2005a; Ma et al., 2007a; Smith et al., 2010). Corticotropin releasing factor (CRF) is synthesized and released from the paraventricular nucleus of the hypothalamus (PVN) in response to stress. RXFP3 is highly expressed in the PVN, as well as other regions associated with stress and anxiety, such as the bed nucleus of the stria terminalis, lateral septum, periaqueductal gray, and dorsal raphe (Liu et al., 2005a). Neural responses to CRF are mediated by the CRF₁ receptor that is expressed in most relaxin-3–containing neurons in the NI (Sutton et al., 1982; Tanaka et al., 2005). Administration of CRF intracerebroventricularly to rats caused activation of relaxin-3–containing neurons, and neurogenic stressors increased relaxin-3 mRNA in the NI (Tanaka et al., 2005), where relaxin-3 is found in presynaptic vesicles. Relaxin-3 mRNA in the rat NI is increased after a forced swim stress paradigm (Banerjee et al., 2010), and these effects are partly blocked by pretreatment with the CRF₁ antagonist antalarmin. Although the precise CRF neuronal projection(s) to the NI relaxin-3 cells have not yet been described, hypothalamic and preoptic areas containing CRF neurons have been shown to project to the NI (Goto et al., 2001). CRF levels in rat, monkey, or human CSF are in the picomolar range even in pathologic states (Garrick et al., 1987; Arborelius et al., 1999). Altered relaxin-3 expression in the NI due to stress is therefore unlikely to result from CRF in the spinal fluid and most likely comes from CRF neuronal projection(s). Ascending projections from the NI (Olucha-Bordonau et al., 2003) are consistent with RXFP3 receptor autoradiography (Sutton et al., 2004) and with involvement of relaxin-3/RXFP3 in neuropsychiatric disease (Smith et al., 2010).

2. Feeding and Metabolism.

In rodents, RXFP3 is expressed in the PVN and supraoptic nucleus (Smith et al., 2010, 2011), suggesting a role in metabolic control. Injection of relaxin-3 intracerebroventricularly to rats either in the early light or early dark phase, transiently increased food intake (McGowan et al., 2005). This stimulatory effect did not result from increased spontaneous activity or arousal and is likely mediated by RXFP3 and not RXFP1, because human relaxin administration had no effect when given during the early light phase (McGowan et al., 2005) or decreased feeding when given during the early dark phase (McGowan et al., 2010) in ad libitum–fed rats. The inhibitory feeding response to human relaxin when administered in the dark as opposed to the light phase suggests a circadian influence involving RXFP1 that is also expressed in hypothalamic centers associated with feeding (Ma et al., 2006).

Both acute and chronic relaxin-3 infusions into the PVN increased food intake (McGowan et al., 2006). Injection into the supraoptic nucleus, arcuate nucleus, and anterior preoptic area also transiently increased food intake. However, it is unclear how relaxin-3 exerts its effects in the arcuate nucleus or anterior preoptic area, because RXFP3 is not present in these nuclei. Finally, intracerebroventricular infusion of the selective RXFP3 agonist R3/I5 also increased food intake, an effect blocked by preadministration of the antagonist R3(BΔ23–27)R/I5 (Kuei et al., 2007). Administration of R3(BΔ23–27)R/I5 intracerebroventricularly alone has no effect on food intake (Sutton et al., 2009), suggesting that under resting conditions there is minimal tone in the relaxin-3/RXFP3 system.

Although relaxin-3 influences food intake, effects on body weight are equivocal. In two studies, chronic intracerebroventricular relaxin-3 increased body weight (Hida et al., 2006; Sutton et al., 2009), but others show no weight change after intracerebroventricular or intrahypothalamic injection of relaxin-3 (McGowan et al., 2005, 2006). In rats, chronic intracerebroventricular administration of the antagonist R3(BΔ23–27)R/I5 did not affect body weight (Sutton et al., 2009). There is also little evidence to suggest that body weight is influenced by endogenous relaxin-3. Although relaxin-3 knockout mice were initially reported to have lower body weights, the first study (Sutton et al., 2009) was performed on a mixed 129:B6 mouse strain before the mice were back-crossed, and the differential phenotype could not be reproduced on a C57/Bl6N background. This dependence of phenotype on strain suggests there may be compensation for a chronic absence of relaxin-3 by other systems.

Several studies investigating the effects of relaxin-3 infusion on feeding also measured blood hormone levels. In rats, chronic infusion of RXFP3 agonist (relaxin-3 or R3/I5) intracerebroventricularly increased plasma leptin, insulin (Hida et al., 2006), adiponectin, testosterone, angiotensinogen, and decreased growth hormone levels (Sutton et al., 2009). Administration into the PVN caused increases in leptin in ad libitum–fed animals and in thyroid-stimulating hormone in ad libitum– and pair-fed animals (McGowan et al., 2006). Intra-PVN agonist dosing increased feeding but failed to significantly change other behaviors such as drinking, grooming, burrowing, rearing, general locomotion, apparent sleep, head down, or tremor (McGowan et al., 2005). Similarly, chronic R3(BΔ23–27)R/I5 dosing seemed only to decrease plasma levels of growth hormone (Sutton et al., 2009) while not altering the other hormones measured. These effects occurred in the absence of changes in energy expenditure.

Studies of the metabolic role of relaxin-3 have provided only limited supporting mechanisms, although considerable evidence supports the actions of the peptide on feeding behavior. Although there is RXFP3 expression in the pancreas and in vitro relaxin-3 suppression of insulin secretion has been reported in isolated tissue (Yamamoto et al., 2009), human plasma relaxin-3 levels are in the picograms per milliliter (low picomolar) range and do not vary in the diabetic state (Zhang et al., 2013). Considering relaxin-3/RXFP3 expression occurs in CNS areas associated with neuroendocrine and behavioral control (Sutton et al., 2004; Ma et al., 2009a; Ganella et al., 2013b), metabolic effects of an RXFP3 agonist may be largely secondary to increased feeding behavior. As noted above, central injection of an RXFP3 agonist in rats increased feeding both acutely (McGowan et al., 2005) and chronically (Hida et al., 2006; Sutton et al., 2009), suggesting a lack of tolerance. In a second rat model, chronic viral expression of an RXFP3 agonist achieved similar results (Ganella et al., 2013a,b). Acute or chronic dosing of an antagonist alone to rats was ineffective (Sutton et al., 2009), suggesting that relaxin-3 tone is low under resting physiologic conditions. Conversely in mice, RXFP3 agonist dosing does not increase feeding under resting conditions (Ganella et al., 2013b). A recent study showed that an RXFP3 antagonist, although ineffective in altering feeding in baseline/sated mice, blocked motivated feeding behavior (Smith et al., 2014). In this study, R3(B1–22)R/I5 blocked food anticipatory activity after a 4-hour food restriction when administered intracerebroventricularly to C57/Bl6J control mice but not to congenic relaxin-3 knockout mice. Antagonist treatment also significantly reduced feeding in mice trained to expect palatable food at a given time. Although relaxin-3 and RXFP3 neuroanatomy is consistent with effects on feeding and body weight in monkeys (Ma et al., 2009b,c), more studies are needed to predict translation of these effects to humans. In addition, detailed studies are also needed to tease out potential confounding effects on reward, motivation and arousal (see below) and to determine whether there is a true feeding response to relaxin-3 stimulation.

3. Behavioral Activation and Arousal.

Evidence for a role for the relaxin-3/RXFP3 system in behavioral activation and arousal is derived from studies that examined colocalization of projections of relaxin-3 neurons from the NI to septal GABAergic and cholinergic neurons that project to the hippocampus (Ma et al., 2009a). The rodent septohippocampal pathway is heavily innervated by relaxin-3 positive projections from the NI. This pathway generates the hippocampal theta rhythm, with oscillations at 4–12 Hz that is controlled by pacemaker neurons of the medial septum (MS). The hippocampal theta rhythm is involved in behaviors such as vigilance, exploration, orientation, navigation, locomotor control, and working memory. Electrical stimulation of the NI causes theta rhythm in the hippocampus and lesions of the NI disrupt theta rhythm initiated by stimulation of the reticularis pontine oralis (Nunez et al., 2006). In both anesthetized and conscious rats, RXFP3 modulates neuronal activity in the hippocampus and MS to promote hippocampal theta rhythm (Ma et al., 2009a) and blockade of RXFP3 in the MS with the antagonist R3(BΔ23–27)R/I5 dose dependently impairs performance in a paradigm investigating theta rhythm-dependent spatial working memory, effects that are reversed by coadministration of the RXFP3 selective agonist R3/I5 (Ma et al., 2009a).

The relaxin-3/RXFP3 system also influences locomotor activity in rodents. Although chronic intracerebroventricular relaxin-3 had no effect on locomotor activity in male Wistar rats (Hida et al., 2006), acute injection of the selective RXFP3 agonist R3/I5 increased locomotor activity, whereas the antagonist R3(BΔ23–27)R/I5 given alone had no effect (Sutton et al., 2009). Effects of the RXFP3 agonist on general locomotor activity, although statistically significant in one study (Sutton et al., 2009), seem to be small in magnitude and were not reproduced after intra-PVN injections (McGowan et al., 2005). Interestingly, female relaxin-3 knockout mice were hypoactive relative to wild-type littermates in several paradigms including the locomotor cell, large open field, Y-maze, and novel object tests (Smith et al., 2009; Hosken et al., 2014). A similar phenotype is displayed in mice receiving the RXFP3 antagonist R3(B1–22)R/I5 or R3(B1–22)R (Smith et al., 2014).

Because serotonin (5-hydroxytryptamine [5-HT]) has well established roles in cognitive, emotional, and behavioral control (reviewed in Cools et al., 2008) and the NI is located close to the dorsal raphe, a region enriched in 5-HT neurons, studies were carried out on the effects of 5-HT on relaxin-3 expression (Miyamoto et al., 2008); most relaxin-3–containing neurons of the NI were found to coexpress 5-HT_1A receptors. Inhibition of 5-HT synthesis for 3 days increased relaxin-3 mRNA in the NI. The correlation between 5-HT/5-HT_1A inactivity (because of 5-HT depletion) and altered relaxin-3 expression in the NI suggests a role for these systems in anxiety and/or depression. Recent studies show anxiolytic and antidepressant effects of intracerebroventricular RXFP3 agonist dosing in behavioral rat models (Ryan et al., 2013a). These studies emphasize the need to determine the relationship between 5-HT and relaxin-3/RXFP3 systems.

Recent studies in rats link RXFP3 circuits to ethanol self-administration and reinstatement behaviors (Ryan et al., 2013b). The R3(BΔ23–27)R/I5 or R3(B1–22)R RXFP3 antagonists administered intracerebroventricularly reduced alcohol self-administration without concomitant alterations in feeding, locomotor behaviors, or memory. The effects on alcohol self-administration were specific, because the antagonists had no effect on sucrose self-administration. In addition, RXFP3 antagonism blocked both cue and stress-induced reinstatement of alcohol seeking. The reduction of alcohol self-administration/reinstatement was linked to the stress-responsive bed nucleus of the stria terminalis, because direct injection of the RXFP3 agonists into the bed nucleus of the stria terminalis also reduced alcohol self-administration.

Extensive brain mapping and behavioral studies have established that the relaxin-3 and RXFP3 system is involved in stress, metabolic control, and behavioral activation and arousal. Recent studies suggest that modulation of the system could be useful in the treatment of forms of addiction.

1. Metabolism.

There is considerable interest in GPCRs that are located in pancreatic islet β-cells and represent potential targets for the treatment of T2DM (Ahren, 2009). One of the most studied systems is the incretin hormone system: glucagon-like peptide 1 (GLP-1) and its receptor GLP-1R. GLP-1 is secreted from enteroendocrine-L cells that are located in the gastrointestinal tract and concentrated in the colon and rectum. The peptide acts on GLP-1R located in a number of tissues including the GI tract, heart, kidney, lung, pancreatic β-cells, and brain. GLP-1 increases insulin release from the pancreas and improves glucose tolerance (Ahren, 2009). The INSL5 and RXFP4 system displays similarities with the incretins. Like GLP-1, INSL5 is concentrated and released from enteroendocrine-L cells (Grosse et al., 2014) into the circulation to act on RXFP4 located in a number of other tissues. In the mouse, RXFP4 expression has been detected by Western blotting in hypothalamus, pituitary, testis, epididymis, ovary, uterus, pancreas, and liver, with immunohistochemical localization to pancreatic islets, anterior pituitary, and Leydig cells in the testis (Burnicka-Turek et al., 2012). In humans, using a similar approach, RXFP4 was detected in heart, placenta, skeletal muscle, and pancreas (Mashima et al., 2013). Most interestingly, the development of the Insl5 knockout mouse showed that these animals have impaired glucose homeostasis. Glucose levels increase with age, and in Insl5^−/− mice older than 6 months, glucose levels are significantly greater than age-matched control littermates. Glucose tolerance tests revealed impaired glucose tolerance in Insl5^−/− mice that does not appear to be associated with changes in insulin sensitivity. The knockouts also had reduced pancreatic islet area compared with controls, which was associated with a reduced number of β-cells and lower circulating insulin levels. However there was no change in circulating GLP-1 levels (Burnicka-Turek et al., 2012). Insl5^−/− mice also displayed impaired fertility due to a reduction in sperm motility and alterations in the estrus cycle (Burnicka-Turek et al., 2012). Interestingly, recent association studies in humans suggest that polymorphisms of RXFP4 are associated with a high body mass index and show a trend to association with obesity (Munro et al., 2012). Studies in mice show an increase in food intake in wild-type mice after intraperitoneal injection of INSL5 that is not seen in RXFP4 knockout mice (Grosse et al., 2014). Plasma INSL5 levels increase with fasting or calorie restriction but are lowered by feeding adding weight to the suggestion that INSL5 is an orixigenic hormone released from the enteroendocrine-L cells in the gut (Grosse et al., 2014). At present, little is known of the potential mechanisms used by INSL5 and RXFP4 to produce metabolic effects, and a number of apparent anomalies will have to be addressed. For instance, to date all β-islet cell GPCRs that increase insulin secretion are either Gα_s or Gα_q coupled, and Gα_i-coupled receptors actually reduce insulin secretion (Ahren, 2009). This appears to conflict with current evidence suggesting that RXFP4 is Gα_i-coupled (Liu et al., 2003a, 2005a), although very little work has focused on RXFP4 and no studies have directly examined receptor function in pancreatic islets.

The emergence of recent evidence that INSL5 is an incretin secreted from the enteroendocrine-L cells and that both INSL5 and RXFP4 knockout mice display impaired glucose tolerance has awakened considerable interest in this ligand/receptor pair. However, more detailed biologic studies are required to firmly establish its role in metabolic control processes.

1. Heterodimerization of Relaxin Family Peptide Receptor 1: A Possible Explanation for the Selective Actions of Relaxin in Diseased or Pathologic States?

Although there is evidence that RXFP1 and RXFP2 form homo- and heterodimers in cell lines (Kern et al., 2008; Svendsen et al., 2008a,b), a recent study demonstrated that the antifibrotic effects of relaxin in kidney are dependent on heterodimerization between RXFP1 and the angiotensin AT₂R (see section III.A.4; Chow et al., 2014). The inhibitory effects of relaxin on excess collagen deposition were completely lost in mice treated with the AT₂R antagonist PD123319 or in AT₂R knockout mice, and by using BRET in cell studies, the two receptors were found to constitutively associate. These findings are particularly interesting, because they provide an explanation for the established observation that the antifibrotic effects of relaxin are seen only in pathologic conditions, with no effect on healthy tissue. Under normal conditions, AT₂R are typically expressed at very low levels, but their expression dramatically increases after injury or disease (Matsubara, 1998; Carey, 2005). RXFP1 is also expressed at extremely low levels at the cell surface, with much higher levels intracellularly within the ER (Kern et al., 2007). Because many GPCRs require dimerization for trafficking from the ER to the cell surface (Milligan, 2009), the injury-induced increase in AT₂R expression may result in an associated enhanced trafficking of RXFP1 (as an AT₂R-RXFP1 heterodimer) to the cell surface, thus explaining why functional effects of relaxin treatment only occur under specific pathologic conditions.

The relevance of RXFP1 heterodimerization—with the AT₂R and perhaps other GPCRs—in many of the other physiologic roles of relaxin remains to be determined. This may be a general requirement for relaxin/RXFP1 function, given the wide and pleiotropic effects of relaxin compared with the relatively limited and variable phenotype of relaxin and RXFP1 knockout and transgenic mice (Zhao et al., 1999, 2000; Samuel et al., 2003a,b, 2004b; Kamat et al., 2004; Krajnc-Franken et al., 2004; Feng et al., 2006; Ganesan et al., 2009). IUPHAR requires two of the following three conditions to be met to definitively determine the existence of heterodimers in native tissue: 1) evidence for physical association in native tissues or primary cells; 2) a specific functional property for the heterodimeric receptor in native tissue; and 3) use of knockout animals or RNAi to demonstrate heterodimers in vivo (Pin et al., 2007). Although conditions 2 and 3 were demonstrated for the RXFP1-AT₂R interaction, questions still remain regarding the interaction between RXFP1 and AT₂R, including a direct demonstration of receptor interaction in endogenous expression systems, whether RXFP1 usually exists as a monomer and only dimerizes with the AT₂R under pathologic conditions and whether this new knowledge can be exploited to specifically target RXFP1 in disease states.

2. The Development of ML290, a Small-Molecular-Weight Allosteric Agonist at Relaxin Family Peptide Receptor 1.

Relaxin binding to and activation of its receptor RXFP1 involves a number of complex interactions and requires many essential receptor regions from the N-terminal LDLa module to the extracellular LRR and TM loops to the full-length C-terminal tail (see section I.B.1). As such, the development of novel molecules that mimic the actions of relaxin at RXFP1 has been a difficult challenge over a number of decades. However, the recent description of a small molecule agonist for RXFP1, ML290 (Xiao et al., 2013), has reignited the search for small molecule RXFP1 ligands.

Based on the size discrepancy between relaxin and ML290 (6000 versus 500 kDa) and some preliminary studies, it was considered unlikely that ML290 bound to the same regions of RXFP1 as the cognate ligand. ML290 also displayed species-specific actions, and studies of chimeras of mouse and human RXFP1 showed that ML290 binds to a region in ECL3 (Xiao et al., 2013) (see section II.A.3). Binding to a topographically distinct region of the receptor suggests a number of exciting and unique attributes for small molecule activators of RXFP1. First, although there was no evidence for signal bias by native relaxin peptides or peptide-based derivatives at RXFP1 (M. Kocan, C. Siwek, et al., personal communication), small molecule ligands binding to a distinct receptor region are likely to engender unique receptor conformations distinct from those induced by relaxin. A common corollary of this is that the signaling profile of ML290 is likely to be biased relative to relaxin. Second, binding of a small molecule to a site topographically distinct from the orthosteric (relaxin) binding sites, suggests an allosteric mode of action. Allosteric modulation of endogenous ligands has been widely documented for many other GPCRs (reviewed in Christopoulos, 2002, and May et al., 2007). Allosteric ligands are considered highly attractive as therapeutics, because of their ability to fine tune the spatial and temporal subtleties of endogenous ligand-receptor interactions (as both allosteric and orthosteric ligands can be simultaneously bound) and their potential for increased receptor selectivity (as these sites have not typically evolved to accommodate a common endogenous ligand). The potential for fine tuning of the relaxin response by a positive allosteric modulator would be highly beneficial in pathologic fibrosis, whereas a negative allosteric modulator could be beneficial for the prevention of cancer metastases. Finally, although the synthesis of peptide derivatives of relaxin has not yet produced a neutral antagonist, the discovery of a series of small agonist molecules suggests that the development of antagonists for RXFP1 may be possible in the future (because of the complicated activation mechanism of RXFP1, agonists are much more difficult to design than simple blocking/antagonist compounds). Not only are these exciting possibilities likely to result in useful experimental tools to accelerate our understanding of RXFP1 function, they will also provide novel avenues for the future design of disease-specific RXFP1 therapeutics.

3. Novel Activators of RXFP Receptors.

Over the last few years, a number of novel RXFP ligands have been described, ranging from small molecule or novel peptide activators of RXFP1 to a synthetic biased antagonist and positive allosteric modulator for RXFP3.

A recent study identified a novel protein in brain tumors that can interact with RXFP1 and promote cell invasion in the absence of relaxin (Glogowska et al., 2013). By using primary glioblastomas isolated from human patients, tumor cell migration and invasion in response to CTRP8 was found to be dependent on RXFP1 expression and involved activation of PI3K and PKC. Subsequent experiments showed an interaction between CTRP8 and RXFP1 after coimmunoprecipitation of tagged proteins coexpressed in HEK293 cells. Molecular modeling suggested that CTRP8 may interact with LRR7 and LRR8 of the RXFP1 extracellular domain, although this was not validated experimentally. This study raises some intriguing possibilities that yet again expand the repertoire of this pleiotropic receptor. First, the ability of additional peptides to bind to and activate RXFP1 may provide a further explanation (in addition to RXFP1-signalosomes responding to very low concentrations of relaxin, see section III.A.3) for the surprisingly widespread effects of RXFP1 activation relative to the limited sites of relaxin production. If nonrelaxin peptides can activate the receptor in different cells, tissues, or pathologies, this also suggests a potential endogenous signaling bias for RXFP1 and may illuminate the pleiotropic effects of receptor activation. Second, and in contrast to the cognate ligand relaxin that requires binding to both the extracellular and TM domains of RXFP1, CTRP8 and ML290 together may provide examples of alternate mechanisms for RXFP1 activation by interacting with a single receptor domain. Although these observations are inherently intriguing, they must still be considered with caution. In both cases, there has been no direct demonstration of receptor binding using traditional radioligand binding paradigms or demonstration of interactions in isolated protein systems, and in the case of CTRP8, no RXFP1 mutagenesis to determine precise sites of receptor binding. Until these basic parameters of ligand-receptor interactions are demonstrated, it is impossible to exclude that the RXFP1-dependent functional effects might occur via interactions with a third protein or even a higher order protein complex.

A number of peptides with unique properties at RXFP3 and RXFP4 have been synthesized and represent powerful scientific tools for the study of this receptor subgroup. A positive allosteric modulator of RXFP3, 135PAM1, was recently described that showed a probe dependence for the C-terminal amidation state of the peptide agonist (Alvarez-Jaimes et al., 2012). Thus, although 135PAM1 positively modulated the cAMP inhibition stimulated by relaxin-3_NH2 and the selective agonist R3I5_NH2 (as determined using a CRE reporter gene), there was no effect of the modulator in combination with nonamidated peptides. A previously described antagonist for RXFP3, R3(BΔ23–27)R/I5 was recently shown to have agonist properties at some signaling pathways (Kocan et al., 2014). Although the peptide antagonized some pathways downstream of relaxin or relaxin-3 activated RXFP3, it did not block others and activated SRE-mediated gene transcription itself. Such biased efficacy of so-called "antagonists" is an increasingly well-documented phenomenon for many GPCRs (Evans et al., 2010). Finally, a minimized INSL5 variant was recently synthesized (Belgi et al., 2013) that contains only the two interchain disulfide bonds that are characteristic of the relaxin family of peptides, with the intra-A-chain disulfide bond and eight residues from the N terminus of the A-chain removed. The minimized INSL5 analogs exhibited higher affinity and efficacy than native INSL5 and are considerably easier to synthesize. This represents an important progression for the INSL5/RXFP4 pairing and should precipitate more studies on this hitherto neglected RXFP receptor.

There have been significant advances in recent years in both our understanding and in the development of tools for the study of signaling pathways activated by relaxin family peptides. The pleiotropic actions of relaxin are now beginning to be understood thanks to the recognition that RXFP1 is able to promote exquisitely sensitive signaling platforms or signalosomes and that it can form both homo- and heterodimers that have distinct signaling properties. The development of RXFP1 antagonists and small molecular weight allosteric agonists provides novel tools for elucidating the roles of this receptor in physiologic and pathologic conditions. RXFP2 and its cognate ligand INSL3, although structurally closely resembling RXFP1 and relaxin, appear to have relatively few and highly specialized functions. The role of relaxin-3 and RXFP3 as a neuropeptide-receptor pairing is now quite well understood, and there are examples appearing of potent and selective antagonists that promise to be useful tools. The identification of ligand bias at RXFP3 is likely of little physiologic importance owing to the distinct localization of relaxin peptides and receptors in vivo but may be a factor to be borne in mind when potential therapeutics are developed. The recognition of INSL5 as an incretin and the INSL5/RXFP4 system as players in metabolic control has stimulated increased interest in the biology of this, until now, relatively little studied receptor-ligand pair.

1. Is There Therapeutic Potential in Small-Molecule Targeting of Relaxin Family Peptide Receptor 1 Heterodimers?

The recent developments showing the importance of RXFP1 heterodimerization and the identification of the small molecule agonist ML290 (see above) may provide some exciting and unique opportunities to therapeutically target RXFP1 with minimal side effects. The therapeutic potential of many pharmaceutical drugs is often limited by the development of toxic effects that can result from drug binding to nonspecific sites or drug-target binding in nontherapeutically relevant tissues. If heterodimerization of RXFP1 with the AT₂R (and perhaps other GPCRs) occurs exclusively at sites of fibrotic injury, small molecule compounds that specifically target the RXFP1-AT₂R would be expected to exhibit exquisite selectivity, resulting in activation of the therapeutically relevant receptor to produce an antifibrotic effect only at the site of injury. Furthermore, this scenario has the capacity to incorporate endogenous temporal control of therapeutic action—thus reversal of the injury would correspondingly decrease the amount of RXFP1-AT₂R heterodimer and thus the available sites of action for the small molecule. In this way, high-resolution knowledge of a system has the potential to translate into a precise spatial and temporal control of drug action at a site of injury. Whether other GPCRs heterodimerize with RXFP1 under particular injury or disease conditions and at specific locations and whether these dimers can be specifically targeted using small molecules remain to be seen.

2. Relaxin Family Peptide Receptor 3 as a Drug Target.

Studies in rodents suggest that RXFP3 modulators have potential for the treatment of emotional disorders (depression and/or anxiety), social behavior deficits (autism), and inappropriate food/alcohol seeking (Smith et al., 2011). In vivo studies show a link between relaxin-3/RXFP3 signaling and anxiety/depression (Banerjee et al., 2010; Ryan et al., 2013a); alcohol seeking/reinstatement (Ryan et al., 2013b) (the reinstatement portion of which could be partially related to stress; Weiss et al., 2001); and memory and hippocampal theta emissions (Ma et al., 2009a). The neuroanatomy of relaxin-3/RXFP3 expression is consistent with a role in these disease states and is conserved from the mouse (Smith et al., 2010) to the macaque (Ma et al., 2009b), suggesting at least some of the effects demonstrated in lower species disease models may translate to humans.

In vivo studies suggest that relaxin-3 neurons, which largely also express GABA (Ma et al., 2007a), provide predominantly inhibitory inputs. However, the RXFP3 agonist R3/I5 shows diverse effects in cells of the intergeniculate leaflet, with both depolarization and hyperpolarization recorded in different cells from the same preparation (Blasiak et al., 2013). Thus the phenotype of the relaxin-3 recipient neuron and its connections could be crucial to the mode of RXFP3 signaling. Additional work is needed to understand the pharmacology of RXFP3 and its potential therapeutic use.

The in vivo pharmacological studies performed so far have administered peptide modulators or viral constructs intracerebroventricularly or directly into CNS regions. Because this involves a labor-intensive surgical approach, RXFP3 agonists/antagonists with CNS bioavailability would greatly facilitate the extension of these studies. Unfortunately these are currently unavailable, and the positive allosteric modulator (135PAM1) that has been described displays probe selectivity precluding its use in native mammalian systems (Alvarez-Jaimes et al., 2012). Perhaps additional screening will identify useful pharmacological tools, allowing peripheral or enteric dosing of RXFP3 agonists and antagonists.

3. Relaxin Family Peptide Receptors and Metabolism.

Recent evidence supports roles for relaxin family peptides in controlling metabolism. Relaxin infusion improves skeletal muscle perfusion and glucose uptake in lean mice. The effect is absent in mice fed a high-fat diet for 13 weeks (Bonner et al., 2013) but can be rescued by chronic infusion of relaxin for 3 weeks. The effect does not result from a direct action on myocytes but involves improvements in blood flow and antifibrotic actions of relaxin (Bonner et al., 2013). This may represent a novel approach to the treatment of insulin resistance.

Increased serum relaxin-3 levels have been described in women with metabolic syndrome (Ghattas et al., 2013), and it has been suggested that this may be related to the known effects of relaxin-3 on food intake (McGowan et al., 2005) and body weight (Hida et al., 2006). However, it is possible that relaxin-3 could also activate RXFP4, and it is interesting to note that there is also an increased body mass index in individuals with a polymorphism of RXFP4 (rs11264422; Munro et al., 2012). There is now strong evidence emerging that INSL5 is an incretin that acts on RXFP4 to have an orexigenic effect (Grosse et al., 2014) (see section V.D.1). INSL5 mice have impaired glucose tolerance (Burnicka-Turek et al., 2012), and there is a report that RXFP4 knockout mice have similar phenotype (Takeda patent PCT/GB2008/003023). RXFP4 may be a promising target for the treatment of obesity and T2DM.

In conclusion, there have been a number of attempts to establish relaxin as a therapeutic for the promotion of cervical ripening, facilitation of implantation, prevention of preeclampsia, and for the treatment of scleroderma. Only recently has this been successful, and relaxin is now in extended phase III clinical trials with redefined end-points for the treatment of acute heart failure. The relative lack of success of relaxin in previous trials may be illuminated by the recent discovery that the relaxin receptor RXFP1 may in some situations act as a heterodimer partner receptor. As is the case with the RXFP1/AT₂R, the expression pattern and actions of the heterodimer may be highly localized and also susceptible to blockade by antagonists of the partner receptor. This work needs to be validated in models of human disease in human tissues but could explain the variability of response observed throughout the patient cohort in previous trials. Relaxin also consistently reduces blood pressure, and antagonists may have some potential in reducing the invasiveness of cancers. The relaxin-3 and RXFP3 system has been examined as a potential target for the treatment of metabolic diseases such as anorexia, cachexia, and obesity. However, inconsistent evidence from animal studies suggests that the complex regulation of energy homeostasis by the CNS, in addition to the redundancy of these pathways, is difficult to override by manipulation of a single system. Recent studies also imply that RXFP3 may be a target for the treatment of addiction. Although still at an early stage, the studies of the biologic role of the INSL5-RXFP4 system suggest that it could have some potential for the treatment of diabetes. Small molecule agonists and antagonists of both RXFP3 and RXFP4 would facilitate more in-depth studies of the biologic roles and therapeutic potential of these two receptors.