In a bivalent ligand binding model, the receptor N-terminus serves as a high affinity bait for the carboxy terminus of the ligand. This then brings the amino terminus of the ligand in close proximity with the ECLs of the receptor where it is able to stabilize the receptor active state. In this model AS of the N-terminus of the receptor would enable new receptor variants to be generated. These could recognize alternative ligands that differed in their carboxy but not amino terminal end, alter the rank order of affinities a particular receptor had for multiple endogenous ligands or even generate alternative signalling bias from multiple endogenous ligands. Alternative N-termini have been reported for CRF1, CRF2, PAC1 and the Secretin receptor.
Splice variants in which the third coding exon is skipped (Δ3e) have been reported for CRF1[CRF1c (Ross et al., 1994)], PAC1[PAC1vs (Dautzenberg et al., 1999)] and Secretin (Figure 3 and Table 1) (Ding et al., 2002b). The third coding exon contains the 2nd, 3rd and 4th conserved cysteines and its deletion may lead to incorrect folding (see Figure 2). Consistent with a poorly functional N-terminus, CRF has very low potency at CRF1c compared with CRF1α and CRF1c did not bind CRF at concentrations tested (Ross et al., 1994), this is in spite of the fact that CRF1c does appear to traffic correctly to the cell surface (Zmijewski and Slominski, 2009b). The function of CRF1c has not been established although its expression is regulated by cell density (Zmijewski and Slominski, 2009a). CRF1 has been reported to form proximers1 (Kraetke et al., 2005) and also displays biphasic responses attributable to coupling to two different G protein pools (Wietfeld et al., 2004). This being the case, co-expression of CRF1c with CRF1α could alter the signalling profile, but this remains untested. In the case of the Secretin receptor, endogenous co-expression of full-length and Δ3e mRNA in pancreatic carcinoma cell lines results in at least a three order of magnitude reduction in Secretin potency (Ding et al., 2002a). This can be recovered by transfection of increasing amounts of wild-type receptor (Ding et al., 2002a). This pseudo dominant negative effect of the Δ3e Secretin receptor is very difficult to explain; Ding et al. (2002a) provide BRET data showing the wild-type and Δ3e receptors form a proximer alluding that interaction between the two receptor forms causes the observed loss of affinity (Ding et al., 2002b) and potency (Ding et al., 2002a). Elsewhere homo-dimerization of the Secretin receptor has been validated through combined disruption of BRET signal through mutation of the dimerization interface, disruption of BRET signal through the use of peptides that mimic the dimerization interface and cysteine disulphide bonding across TM regions of the receptor (Harikumar et al., 2007; 2008; Gao et al., 2009). Nonetheless, in the absence of direct data on the relative expression of the mature proteins at the cell surface as well as composition and stoichiometry of receptor proximers, interpretation of the severe loss of binding/signalling is extremely difficult. If we accept that endogenously all proximers contain both receptor variants, the implication is that the N-terminus of wild-type receptor provides strong cooperativity for ligand binding. Currently, the level of cooperativity reported across the Secretin dimer (Gao et al., 2009) appears insufficient to explain the magnitude of affinity and potency loss engendered by co-expression of the Δ3e variant. The PAC1 receptor is unusual in this family as the variant usually referred to as ‘normal’ contains six coding exons for the N-terminal domain (Figures 2 and 4). This is the result of the inclusion of two small exons between those corresponding to the 3rd and 4th of other family members (Figure 2 and PAC1n Figure 4). PAC1vs has exon 3 as well as the extra exons, with no equivalent in other family members, deleted (Figure 4) and is thus equivalent CRF1c and Δ3e Secretin receptors above. This variant displays a two order of magnitude reduction in affinity for its ligands pituitary adenylate cyclase activating peptide (PACAP) 38 (comprising 38 amino acids) and 27 (the same peptide with a C-terminal truncation) relative to the normal receptor variant (Dautzenberg et al., 1999), as well as a 20- to 100-fold decrease in potency for cAMP production (Dautzenberg et al., 1999; Lutz et al., 2006). The mRNA encoding this variant is co-expressed with full-length and short (see later, exon 3 present, unique exons absent) PAC1 in neuroblastoma cell lines SH-SY-5Y and Kelly cells (Lutz et al., 2006). The potency of PACAP38 to elicit a cAMP response in these cell lines is consistent with the potencies observed in heterologous systems for PAC1n and PAC1s (Lutz et al., 2006), suggesting that PAC1vs does not exert a dominant effect in this pathway.
In the above cases of CRF1c, Δ3e Secretin receptor and PAC1vs it is not clear what physiological relevance would be engendered by their pharmacology. In each case the N-terminus would be predicted to fold incorrectly and the effect on wild-type receptor signalling (if any) could be achieved simply by altering wild-type receptor expression. In comparison, the AS that gives rise to CRF2α, β and γ and PAC1n and PAC1s variants results in alternative N-termini that are all fully functional and elicit altered pharmacology. For CRF1, AS arises through the use of different promoters that drive expression of alternative first exons, with CRF2α containing a single first exon that is 5′ proximal to the first common exon and CRF2β containing two further upstream exons before splicing to the common first exon, CRF2γ contains a 3rd alternative first exon located between 5′αβ exons and again splices to the same common exon. The consequence of this AS is that the N-terminal sequence containing the first conserved cysteine differs between the three forms with protein lengths of 411, 438 and 397 amino acids respectively (Figure 3, N-terminal exchange). The crystal structure of CRF2α shows an alpha helix extending either side of the first conserved cysteine before the loop that connects it with the first beta strand (Pal et al., 2010). This contrasts with the NMR solution structure of CRF2β in which the corresponding structural element forms a disordered loop that is constrained by the disulphide bond and its link with the first beta strand (Grace et al., 2007). The N-terminus of CRF2γ has not been solved; however, the existing structural studies demonstrate that the interaction between the ligand and N-terminal domain occurs via the part of this fold opposite the AS sequence. In this light, the observation that these CRF2 variants have identical binding properties is unsurprising (Kostich et al., 1998; Ardati et al., 1999). These two studies have examined different downstream outputs from receptor activation. Kostich et al. (1998) measured cAMP accumulation and reported pEC50 values for CRF, sauvagine, urotensin and urocortin, which did not differ between CRF2α and γ but showed 10-fold higher potency at CRF2β. This is in contrast to the report of Ardati et al. (1999), who used a cAMP response element driven reporter assay and showed identical potencies of CRF2α and β with respect to CRF, sauvagine, urotensin and urocortin with pEC50 values consistent with the higher potency values established for CRF2β by Kostich et al. (1998). This difference may be due to differences in receptor reserve in the two assays. The β and γ variants show a more restricted pattern of tissue expression compared with the α variant; however, with the existing molecular pharmacology, it is unclear why the different isoforms exist. One possibility may be that CRF2 can form a complex with one or more RAMP isoforms. The crystal structure of the N-terminal domains of CLR in complex with RAMP1 shows extensive contacts between RAMP1 and the alpha helix of CLR (Haar et al., 2010), which corresponds to the AS region of the CRF2α, β and γ variants. AS splicing could, therefore, regulate the interaction between CRF2 and RAMPs thereby regulating ligand selectivity. As mentioned above, PAC1 is unusual among Secretin family members in that there is a common receptor variant (PAC1n) for which the N-terminus is encoded by six exons (Figure 3, C-terminal of the two N-terminal inserts and Figure 4). AS of this region to remove coding exons 4 and 5 results in a receptor, PAC1s (Figure 4), with exon organization that mimics the remainder of the family. PAC1, VPAC1 and VPAC2 all respond to physiologically relevant concentrations of PACAP; however, PAC1 is normally considered a type I PACAP receptor due to the low affinity and potency that VIP (vasoactive intestinal peptide) displays at this receptor. The first molecular pharmacological description comparing PAC1n and PAC1s was performed in HEK293 cells and indicated that VIP displays low affinity and potency for cAMP production only at the PAC1n variant and PAC1s does not show the same ligand selectivity (Dautzenberg et al., 1999). In contrast to these results a follow-up study performed in CHO cells showed that VIP had low affinity and potency for cAMP production at both PAC1n and PAC1s (Ushiyama et al., 2007). This, combined with the observation that RAMP2 selectively interacts with VPAC1 and alters its pharmacology (Christopoulos et al., 2003), suggests that AS of the N-terminal domain of PAC1 could yield receptors with significantly altered pharmacology by regulating the interaction with RAMP proteins. In rat testis an additional functional N-terminally spliced variant of PAC1 has been reported, PAC13a (Daniel et al., 2001). This variant has, in addition to the six exons that encode the N-terminus of PAC1n, an additional exon between coding exons 3 and 4 adding another 24 amino acids to the N-terminal domain (Figure 3, N-terminal of the two N-terminal inserts). At PAC13a PACAP27 displays equivalent affinity but slightly reduced potency for cAMP and IP3 production when compared with PAC1n (Daniel et al., 2001). PACAP38 displays higher affinity but significantly lower efficacy at PAC13a compared with PAC1n (Daniel et al., 2001) suggesting the higher affinity has been achieved in part by stronger G protein coupling. This variant has not been reported in humans.