Alternative Pre-mRNA Splicing of the Mu Opioid Receptor Gene, OPRM1: Insight into Complex Mu Opioid Actions

Most opioid analgesics used clinically, including morphine and fentanyl, as well as the recreational drug heroin, act primarily through the mu opioid receptor, a class A Rhodopsin-like G protein-coupled receptor (GPCR). The single-copy mu opioid receptor gene, OPRM1, undergoes extensive alternative splicing, creating multiple splice variants or isoforms via a variety of alternative splicing events. These OPRM1 splice variants can be categorized into three major types based on the receptor structure: (1) full-length 7 transmembrane (TM) C-terminal variants; (2) truncated 6TM variants; and (3) single TM variants. Increasing evidence suggests that these OPRM1 splice variants are pharmacologically important in mediating the distinct actions of various mu opioids. More importantly, the OPRM1 variants can be targeted for development of novel opioid analgesics that are potent against multiple types of pain, but devoid of many side-effects associated with traditional opiates. In this review, we provide an overview of OPRM1 alternative splicing and its functional relevance in opioid pharmacology.


Introduction
The actions of most clinically used opioid analgesics, such as morphine and fentanyl, are complex. Although they provide potent analgesia, they produce many undesirable side-effects such as respiratory depression, constipation, pruritus, physical dependence, and addiction. Misusing these opioids often leads to development of opioid use disorder, contributing significantly to the worldwide opioid epidemic. These opioid drugs act primarily on the mu opioid receptor (MOR) that was initially proposed by Martin in 1967 [1], established by pharmacological studies in 1973 [2][3][4], and verified by molecular cloning in 1993 [5][6][7]. Clinically, patients often showed divergent responses to mu opioids not only in analgesia but also in side-effects, and incomplete cross-tolerance among mu opioids led to the clinical practice of opioid rotation. These observations suggest the presence of multiple mu opioid receptors, a concept that was originally proposed by early pharmacological studies [8][9][10] and evolved by recent molecular studies of the mu opioid receptor gene, OPRM1 [11,12].
Only a single-copy OPRM1 gene has been identified, raising the hypothesis that alternative pre-mRNA splicing of the OPRM1 gene generates multiple mu opioid receptors, contributing to the diversity and complexity of opioid responses seen in humans and THC 2-AG ACEA TM38837 VD60 Mainly brain, DRG, and kidney [65][66][67] CNR2 CB2 N-term variant (2) THC 2-AG Nabilone SR144528 AM-630 Mainly testis and spleen [68] Corticotropin-Releasing Hormone Receptor

Concept of Multiple Mu Opioid Receptors
The concept of multiple mu opioid receptors was originally suggested based on clinical observations that different patients displayed divergent responses to mu opioids not only in analgesia, but also in side-effects and that incomplete cross-tolerance among mu opioids led to the clinical practice of opioid rotation. Similar observations were seen in animal models. Different inbred mouse strains showed various sensitivities to mu opioids. For example, the 129/P3 inbred mouse strain is resistant to produce morphine tolerance and naloxone-precipitated withdrawal, sharply contrasting with the robust morphine tolerance and physical dependence seen in C56BL/6J and SWR/J inbred mouse strains [100,101]. Multiple mu binding sites were first demonstrated with opioid receptor binding assays using [ 3 H]-labeled opioids in 1981 [9], subsequently validated by computer modeling approaches [102,103] and visualized by autoradiography [104][105][106]. Further pharmacological studies using two antagonists, naloxazone and naloxonazine, defined two mu subtypes, mu 1 and mu 2 [10]. A third mu subtype, morphine-6β-glucuronide (M6G) receptor, was proposed based on differences in analgesia potency, binding affinity of morphine, morphine analgesia in CXBK mice, and sensitivity toward this.
3-O-methylnaltrexone, a mu antagonist, had high affinity toward M6G and blocked analgesia of M6G, but not morphine [10,[107][108][109]. Subsequent antisense oligonucleotide and Oprm1 knockout studies provided molecular insights into the M6G site (see below) [107,[110][111][112]. The identification of the MOR-1 sequence and its gene, OPRM1, allows us to explore multiple mu opioid receptors at the molecular level. The concept of multiple mu opioid receptors has been evolved by isolation and characterization of an array of alternatively spliced variants or isoforms from the single-copy OPRM1 gene. The correlation between mu subtypes such as mu 1 and mu 2 defined by the early pharmacological studies and the OPRM1 splice variants remains unknown, although several hypotheses such as heterodimerization were speculated.

Evolution of OPRM1 Gene
All four opioid receptor genes, including mu receptor (OPRM1), delta receptor (OPRD1), kappa receptor (OPRK1), and opioid-receptor like (OPRL1), are found only in vertebrates. As suggested by phylogenetic studies, the opioid receptor genes were generated through a two-round genomic duplication (paleoploidization) from a single ancestral opioid receptor gene (unireceptor) [113][114][115]. The first-round duplication produced OPRD1/OPRM1 and OPRL1/OPRK1 genes, which were subsequently segregated into individual OPRD1, OPRM1, OPRL1, and OPRK1 [116][117][118]. Like the evolutionary tree of life, analyzing MOR-1 protein sequences predicted from 27 species define four major clades as follows: (1) fish, (2) amphibians, (3) birds, and (4) mammals [119]. Sequence alignments of MOR-1 proteins from different species reveal that the high-homology regions are located in the transmembrane domains and the three intracellular loops, suggesting the importance of these conserved structures in mu ligand binding and G protein coupling.
The genomic structure of the OPRM1 gene appears to have evolved along with the four polygenetic clades [12,120]. In fish (teleosts), the OPRM1 has five exons and each exon encodes regions with related TM domains: exon 1 (TM1), exon 2 (TM2-TM4), exon 3 (TM5), exon 4 (TM6), and exon 5 (TM7). In amphibians, the last three exons were merged into a single third exon encoding TM5-TM7. This structure with three exons encoding 7TMs has been maintained throughout birds and mammals. Starting in the chicken, exon 4 that encods 12 amino acids at the intracellular C-terminal tail evolved, which is maintained in all the mammals. Additional 5 and/or 3 exons such as exons 7 and 11 were integrated in mannals as well, providing a genomic structure for complex alternative splicing of the OPRM1 gene. The functional importance of exon 7-and exon 11-associated splice variants in opioid pharmacology is discussed below.

OPRM1 Gene Structure
The chromosonal location of the mouse and human OPRM1 gene was initially mapped to chromosomes 10 (10 cM from the centromere) and 6 (6q24-25), respectively, using traditional chromosomal mapping approaches [121][122][123][124], soon after MOR-1 cDNA was identified. Genomic sequencing initiated by the Human Genome Project confirmed the early OPRM1 mapping results, and revealed the precise chromosomal locations of the OPRM1 gene not only in mice and humans, but also in over 30 species. All the species contain a single copy of OPRM1 gene. Genomic sequencing also showed that genes adjacent to OPRM1 are conserved [11,120]. For example, the regulator of G protein signaling 17 (RGS17) gene is located at the upstream of the OPRM1 gene, while the interaction protein for cytohesin exchanges factors 1 (IPCEF1) gene and subunit 5 of the splicing factor 3b (SF3b5) gene are found downstream of the OPRM1 gene. Interestingly, a 70 bp sequence of exon 7 or exon O in the OPRM1 gene overlaps with the IPCEF1 gene in the reversed orientation, raising questions about their interaction and regulation at transcription and posttranscription levels [11].
The OPRM1 gene was initially identified as a structure with four exons and three introns using the original MOR-1 cDNA probe [125]. Mapping of additional exons at the 5 and 3 of the OPRM1 genes identified in alternatively spliced transcripts further extends the OPRM1 genes. Currently, the mouse, rat and human OPRM1 genes have approximate 270, 264, and 210 kb, respectively [11,12].
The OPRM1 gene has two independent promoters, exon 1 (E1) and exon 11 (E11) promoters, which are conserved from rodents to humans. E1 promoter was identified as a dual-promoter model that contains a proximal promoter and distal promoter [126][127][128]. E1 promoter lacks a TATA box but has several cis-acting elements such as GC-rich elements with multiple transcription start points, suggesting a housekeeping gene mode. E11 promoter, however, has a TATA box using a single transcription start point, favoring an eukaryote class II promoter mode. Each promoter controls expression of a unique set of splice variants. Differential expressions of the two promoters in brain regions such as hippocampus and substantia nigra in a transgenic mouse model suggest region-specific promoter activity [129].

Alternatively Spliced OPRM1 Variants
The OPRM1 splice variants can be categorized into three major types based on receptor structure: (1) full-length 7TM C-terminal variants; (2) truncated 6TM Variants; and (3) single TM variants [11,12] (Figures 1-3).     Full-length C-terminal variants are generated through OPRM1 3′ splicing, which mainly refers to splicing from exon 3 to its downstream exons. All of the full-length 7TM C-terminal variants share the same exons 1/2/3 that encode the majority of the receptor, including the N-terminus, 7TM domains, three intracellular and three extracellular loops, and part of intracellular C-terminus, except for a distinct C-terminal tail encoded by an alternative exon(s) produced from the 3′ splicing.

Full-Length 7TM C-Terminal Splice Variants
Full-length C-terminal variants are generated through OPRM1 3 splicing, which mainly refers to splicing from exon 3 to its downstream exons. All of the full-length 7TM C-terminal variants share the same exons 1/2/3 that encode the majority of the receptor, including the N-terminus, 7TM domains, three intracellular and three extracellular loops, and part of intracellular C-terminus, except for a distinct C-terminal tail encoded by an alternative exon(s) produced from the 3 splicing.
The first two full-length 7TM C-terminal variants, hMOR-1A [130] and rMOR-1B [131], were identified in human and rat, respectively. Subsequent 3 RACE using anchor primers from exon 3 isolated many alternative 3 exons and further PCR cloning generated fulllength C-terminal variant sequences. To date, 22 full-length 7TM C-terminal variants in mice, 13 in rats, and 12 in humans have been isolated (Figures 1-3) [132][133][134][135][136][137][138][139][140]. The predicted amino sequences from each of the 7TM C-terminal variants are unique and the lengths of the C-terminal tail sequences vary from the shortest with one amino acid in hMOR-1B4 [136] and rMOR-1D [137] to the longest with 88 amino acids in mMOR-1U [138] ( Table 2). Moreover, several potential phosphorylation sites, such as protein kinase C, GRKs, cAMP-and cGMP-dependent protein kinase, and Casein kinase II, were identified in these C-terminal tails. The functional importance of these 7TM C-terminal variants in opioid pharmacology has been demonstrated using in vitro cell models showing differences in mu agonist-induced G protein coupling, phosphorylation, internalization and post-endocytic sorting and by differential region-and cell-specific expression, as well as suggested by findings using C-terminal truncation mouse models (see below). Table 2. Amino acid sequences of encoded by alternative exons downstream of exon 3 in 7TM C-terminal variants.

Species Variant Exons Amino Acid Sequence Downstream of Exon3
Mouse The amino acid sequences resulting from 3 splicing are presented. S, T, and Y are predicted phosphorylation sites, labeled with red and italic letters, and highlighted with yellow. Underlined and yellow-highlighted sequences are predicted phosphorylation codes, PxPxxE/D or PxxPxxE/D, for β-arrestin binding based on crystal G protein coupled receptors (GPCR) structures [141]. Adapted from Pasternak and Pan 2013 [12].
Identification of the mouse exon 11-associated variants quickly led to isolation of the mouse homologs of exon 11-associated variants in rats and humans (Figures 1 and 3) [144,145]. Based on the genomic databases, rat and human exon 11 are found at~21 and~28 kb upstream of their exon 1, respectively, a genomic location similar to that of the mouse exon 11 (~30 kb). PCR cloning using primers anchoring the respective exon 11 homologs and downstream exons identified four exon 11-associated variants in humans and seven in rats [144,145]. Like mMOR-1G, splicing from exon 11 to exon 2 generates two 6TM variants, rMOR-1G1, and rMOR-1G2 or hMOR-1G1 and hMOR-1G2, in both rats and humans. However, both rat and human exon 11 have an alternative 5 splice site, which splits exon 11 as exons 11a and 11b. Splicing from exon 11a to exon 2 produces rMOR-1G2 or hMOR-1G2. The rat and human exon 11a predict 7 and 16 amino acids, respectively, which are in frame with exon 2. Thus, both rMOR-1G2 and hMOR-1G2 encode a 6TM receptor translated from the exon 11a AUG without the first TM. On the other hand, splicing from exon 11b to exon 2 yields rMOR-1G1 and hMOR-1G1. Similar to mMOR-1K and mMOR-1L, both rMOR-1G1 and hMOR-1G1 predict a 6TM receptor when the AUG in exon 2 is used, and produce a short peptide sequence if exon 11a AUG is used. Splicing from exon 11 to exon 1 produces an additional five variants in rats and two variants in humans, all of which encode a 7TM receptor when translated from exon 1 AUG. The functional relevance of the 6TM variants has been implicated in mediating the actions of a new type of opioid analgesics that are potent against a spectrum of pain models without many side-effects associated with traditional opiates [12] (see below).
An additional human 6TM variant hMOR-1K was isolated based on the finding of a mouse exon 13 homolog in the human OPRM1 gene [146]. Unlike mMOR-1K that is associated with exon 11, hMOR-1K did not have exon 11 and its expression appeared to be controlled by its own promoter located at upstream of the human exon 13. Another human 6TM variant was identified as mu3 receptor with an exon composition of 2/3/new exon [147]. Both hMOR-1K and mu3 variants predict a 6TM receptor when translated using the exon 2 AUG, a scenario similar to rMOR-1G1, hMOR-1G1, mMOR-1K, and mMOR-1L ( Figure 1).

Single TM Variants
All of the single TM variants contain exon 1 that encodes the entire N-terminus and the first TM. Alternative splicing from exon 1 to downstream exons by exon skipping or insertion causes a reading-frame shift to terminate translation early, predicting a truncated receptor with a single TM. Currently, there are five single TM variants in mice, two in rats, and four in humans (Figures 1-3). The first two single TM variants, hMOR-1S and hMOR-1Z, were isolated from the human OPRM1 gene [11,85,148]. Splicing from exon 1 to exon 4 by skipping exons 2 and 3 produces hMOR-1S, while hMOR-1Z has exons 1/3/4 by skipping exon 2. Both MOR-1S and MOR-1Z were subsequently identified in mice and rats with the same splicing mechanisms [85]. An additional two human single TM variants were isolated as SV1 and SV2 in which an exon A or B was inserted between exons 1 and 2, respectively [149]. Early translation termination in exon A or B results in a truncated single TM receptor.
The mouse Oprm1 gene has another three single TM variants, mMOR-1Q, mMOR-1R, and mMOR-1T [85]. Both mMOR-1Q and mMOR-1R have exon 2 skipping, but contain a different exon composition downstream of exon 3. In mMOR-1T, there is an exon 16 insertion between exons 1 and 2, predicting a single TM variant. Although the single TM variants cannot bind any opioids, their roles were shown in modulating the expression and function of the full-length 7-TM mu opioid receptor (MOR-1) (see below) [85]. Northern blot analysis was initially employed to determine mRNA expression of OPRM1 variants in the brain using individual exon probes or combined exon probes. Multiple bands with different sizes and intensities were observed [6,123,132,[134][135][136][137]143,150], suggesting OPRM1 alternative splicing. However, it is difficult to examine mRNA expression of individual splice variants using exon probes in Northern blot analyses because some exons are shared among several splice variants. Semi-quantitative RT-PCR assays were designed to investigate the expressions of the OPRM1 variant mRNAs in a number of dissected brain regions, including the cortex, thalamus, striatum, hypothalamus, hippocampus, brain stem, cerebellum, and spinal cord [132,133,143,144]. For example, high level of mMOR-1C mRNA but low levels of mMOR1D and mMOR-1E mRNAs were expressed in the mouse thalamus [132]. In human brain, hMOR-1G2 was highly expressed in the prefrontal cortex, piriform cortex, nucleus accumbens, and pons, but not in the temporal cortex and spinal cord [144].

Expression and Function of the
Using real-time quantitative PCR (qPCR), a more accurate method measuring mRNA levels, the expression profiles of thirty Oprm1 splice variant mRNAs in ten different brain regions of four inbred mouse strains, C56BL/6J, 129P3/J, SJL/J, and SWR/J [151], and CD-1 mice [152] were examined. The results showed marked differences of the variant mRNA expression among the brain regions or among the mouse strains, suggesting region-specific or strain-specific alternative splicing of the OPRM1 gene. For example, 7TM C-terminal variant mMOR-1A was expressed highly in the hypothalamus and thalamus but lowly in the striatum and prefrontal cortex in SJL/J mice. In the same PAG region, mMOR-1G, a 6TM variant, was expressed much higher in C57Bl/6J mouse strain than in the other three mouse strains.
Among the three types of Oprm1 variants, the overall expressions of all 7TM variants were dominant throughout the brain regions and mouse strains, while the overall expressions of all 6TM and 1TM variants were lower than those of all 7TM variants (Table 3) [153]. However, all 6TM and 1TM variants were expressed at relatively high levels in certain regions or mouse strains. For example, in the PAG region of C57BL/6J mice, the expression level of all 6TM variants was 20.4% of all 7TM variants, which was the highest among the brain regions and among the mouse strains. The expression level of all 1TM variants in the prefrontal cortex of SJL/J mice was higher than those in the other three mouse strains. These results further demonstrated region-specific and strain-specific alternative splicing of the Oprm1 gene. Table 3. mRNA expression levels of all 7TM, 6TM, and 1TM splice variants in brain regions of four inbred mouse strains.

Strain
Brain Region Altered expression of the OPRM1 splice variants were seen in multiple brain regions in a morphine tolerance mouse model [152], the human brain of HIV-infected individuals [153,154], and the medial prefrontal cortex of human heroin abusers and heroin self-administering rats [155]. Sex difference in expression of the Oprm1 variant mRNAs among the mouse brain regions was also observed [156].

Region-Specific and Cell-Specific Expression of the OPRM1 Variant Proteins
The distribution of OPRM1 variant proteins was investigated using several antisera generated against individual exon coding sequences [157][158][159][160]. These studies revealed dramatic differences in the regional distribution of the exon-containing OPRM1 variant proteins. For example, the distribution of exons 7/8-like immunoreactivity (LI) detected with an antisera against a peptide from mouse exons 7/8 clearly differed from that of exon 4-LI using an antisera against a peptide from exon 4 in several regions, such as the medial eminence, thalamic nuclei, and nucleus ambiguous [158]. The two antisera also labeled different cell populations, as demonstrated in the laminae layer I-II of the spinal cord [158]. An antisera raised against an exon 8 peptide labeled uniquely in the dentate gurus, the mossy fibers of the hippocampal formation and the nucleus of the solitary tract [157], while an antisera raised against exon 3/5a hybridized predominantly in the olfactory bulb [161]. The localization of variant proteins was also examined at the ultrastructural level [162]. For example, in the superficial laminae of the spinal cord, the exon 7/8-LI was predominantly distributed in presynaptic cells and co-localized with calcitonin gene-related peptide (CGPR), whose antibody labels unmyelinated primary afferents, in contrast to the exon 4-LI, which displayed an equal distribution between presynaptic and postsynaptic membranes and did not co-localize with CGPR [162].
It should be noted that several variant proteins encoded by the same exon(s) commonly share identical epitope(s), thus it is challenging to distinguish each individual variant at the protein level using exon-specific antisera by immunohistochemistry. For example, antisera against a peptide from mouse exon 4 would detect at least five exon 4-containing variants in mice including mMOR-1, mMOR-1H, mMOR-1I, mMOR-1J, and mMOR-1G. Similarly, exon 8-LI would correspond to mMOR-1D and mMOR-1N, whereas exon 7/8-LI would represent two variants, mMOR-1C and mMOR-1M. These studies also revealed the discrepancies between mRNA and protein levels for some of the variants. For example, in the mouse brain, the protein level of mMOR-1 detected by the exon 4-specific antisera was found to be comparable to that of mMOR-1C and mMOR-1M as determined by the exon 7/8-specific antisera [158], while the expression level of mMOR-1 mRNA was much higher than that of both mMOR-1C and mMOR-1M mRNA [132,143]. Although different antisera may differ in sensitivity, these studies reiterated the poor correlation between mRNA and protein levels in the brain and highlighted the importance of examining the expression of variants at both the mRNA and proteins expression levels.

Opioid Receptor Binding Affinity
To determine the binding profiles of full-length 7TM C-terminal variants, opioid receptor binding assays using [ 3 H] DAMGO, a full mu agonist, were performed with membranes from CHO cell lines that stably express each individual full-length C-terminal variant [132][133][134][135][136][137]142]. The results from saturation studies indicated that all the 7TM Cterminal variants had high affinity toward [ 3 H] DAMGO, which was anticipated since they all share the same binding pocket. Competition studies showed mu selectivity with a number of mu opioids such as morphine, M6G, and naloxone for all the 7TM C-terminal variants. However, several endogenous opioid peptides revealed different binding affinities among the 7TM C-terminal variants. For example, both β-endorphin and dynorphin A were over four times more potent in mMOR-1D than in mMOR-1 [132]. These results raise intriguing questions of how different C-terminal tails impact on the binding pocket, specifically toward the endogenous opioid peptides.

Mu Agonist-Induced G Protein Coupling, Internalization, Phosphorylation, and Post-Endocytic Sorting
The intracellular C-termini of GPCRs play an important role in the signaling transduction system, especially the G-protein activation. Mu agonist-induced G protein activation in the OPRM1 7TM C-terminal variants were extensively studied using membranes from CHO cells stably expressing the individual variant in [ 35 S] GTPγS binding assay, a commonly used method to assess the overall agonist-induced receptor-G-protein coupling. The results revealed marked differences in mu agonist-induced G-protein activation in both potency, measured by the EC 50 values, and efficacy, determined by the maximal stimulation, among the 7TM C-terminal variants in mice, rats, and humans [135][136][137]163]. Several mu agonists displayed varied efficacies among various 7TM C-terminal variants. For example, morphine and M6G were more efficacious in hMOR-1A than hMOR-1B1, hMOR-1B3, and hMOR-1B5, while β-endorphin was a full agonist for hMOR-1B5 and a partial agonist for hMOR-1A [136]. There was lack of correlation between the binding affinity (K i ) and potency (EC 50 ). β-endorphin bound to both rMOR-1 and rMOR-1C1 with similar affinity. However, β-endorphin was over 20 times more potent in rMOR-1 than in rMOR-1C1 in [ 35 S] GTPγS binding. Also, there was little correlation between potency (EC 50 ) and efficacy. Together, these results demonstrated that different intracellular C-terminal tails in the 7TM full-length variants have significant influence on mu agonist-induced G protein activation.
Mu agonist-induced MOR-1 phosphorylation involves receptor desensitization and internalization, contributing to the development of mu opioid tolerance and physical dependence [164][165][166][167]. Several mu agonist-induced phosphorylation sites in the intracellular C-terminus encoded by exon 3 in MOR-1, such as 354 TSST 357 and 375 STANT 379 motifs, were identified [168][169][170]. Involvement of these phosphorylation sites in mu opioid actions such as analgesia, tolerance, respiratory depression, and constipation was demonstrated using several phosphorylation-deficient mouse models [171]. However, these phosphorylation sites are shared by all the 7TM C-terminal variants. The potential phosphorylation sites for several kinases such as GRKs, protein kinase C, casein kinase, and tyrosine kinase predicted from alternatively spliced C-terminal tails raise questions regarding their responses to mu agonists (Table 2). Koch et al. showed that morphine-induced receptor phosphorylation in mMOR-1D and mMOR-1E was robust, sharply contrasting with the nonresponsiveness in mMOR-1 and mMOR-1C [172], suggesting additional morphine-induced phosphorylation sites in the C-terminal tails encoded by the respective C-terminal tails or modulation of overall receptor phosphorylation by the C-terminal tails. The differential phosphorylation profiles among these variants were correlated with morphine-induced internalization and resensitization [172]. Additionally, threonine 394 in the exon 4-encoded C-terminal tail of rMOR-1 was phosphorylated in response to DAMGO [173]. Interestingly, a mutant mouse with the single point mutation T394A showed loss of etorphine-induced MOR-1 internalization in the spinal dorsal horn neurons and diminished tolerance by morphine and etorphine, as well as increased vulnerability to heroin self-administration [174]. Furthermore, morphine induced robust internalization of exon 7-associated mMOR-1C, but not exon 4-associated mMOR-1, in the mouse lateral septum [175].
After being internalized by mu agonists, mMOR-1 is either migrated to the lysosome for degradation or recycled back to the plasma membrane. Exon 4-encoded C-terminal sequences contain a motif termed the MOR1-derived recycling sequence (MRS) that allows efficient post-endocytic sorting of mMOR-1 to the plasma membrane [176]. Absence of the exon 4-encoded MRS in mMOR-1B, mMOR-1D, and mMOR-1E led to rapid lysosomal degradation due to inefficient receptor recycling following DAMGO-induced internalization, indicating the importance of the C-terminal sequences in determining the post-endocytic sorting direction of GPCRs [177].

Mu Agonist-Induced Biased Signaling at 7TM C-Terminal Variants
Biased signaling of GPCRs is defined based on observations that different agonists can produce divergent signaling transduction pathways through a single receptor. Biased signaling through G protein and β-arrestin 2 has been extensively studied in the original MOR-1 [168,[178][179][180][181][182]. Identification of multiple OPRM1 7TM C-terminal variants opened questions about their roles in mu agonist-induced biased signaling. The effect of different 7TM full-length C-terminal variants on mu agonist-induced signaling bias between G protein coupling and β-arrestin 2 recruitment was studied using [ 35 S] GTPγS binding and PathHunter β-arrestin 2 recruitment assay in the same CHO cells stably expressing individual 7TM C-terminal variants [183]. The results showed marked differences in mu agonist-induced G protein activation and β-arrestin 2 recruitment among the 7TM variants. For example, exon 7-associated 7TM variants, particularly mMOR-1O, displayed greater β-arrestin 2 bias for most mu agonists than exon 4-associated 7TM variant mMOR-1, suggesting that exon 7-associated variants favor mu agonist-induced β-arrestin 2 binding. This hypothesis was supported by the identification of a consensus phosphorylation code, PxPxxE/D or PxxPxxE/D, in exon 7-encoded sequences for high affinity arrestin binding proposed by homology modeling with the crystal structures of several GPCRs (Table 2) [141]. When this predicted phosphorylation code was mutated in mMOR-1O, DAMGO-induced β-arrestin 2 recruitment was greatly attenuated (our unpublished data) [183]. These results also suggested a mechanism of the in vivo function of exon 7-associated variants in several morphine actions observed in an exon 7 C-terminal truncation mouse model (mE7M-B6), which showed similar phenotypes in a β-arrestin 2 knockout (KO) mouse (see below) [83]. The finding that a single mu agonist can produce differential signaling through multiple 7TM C-terminal variants from in vitro cell line studies offers new insights into our understanding of biased signaling in vivo where all the 7TM C-terminal variants are co-expressed.

In Vivo Function of 7TM C-Terminal Variants Morphine-Induced Itch (Pruritus)
Opioid-induced itch (pruritus) is one of the most prevalent acute side effects of the spinal or epidural use of opioids in patients [184][185][186], which has limited the clinical use of opioids. However, the molecular mechanisms underlying opioid-induced itch remained largely unknown. Liu et al. reported that an Oprm1 7TM C-terminal splice variant mMOR-1D was the major player in spinal morphine-induced scratching in mice [187]. Downregulating the spinal mMOR-1D with siRNAs reduced morphine-induced scratching without affecting morphine analgesia. They further demonstrated that morphine-induced scratching was mediated through heterodimerization of mMOR-1D and gastrin-releasing peptide receptor (GRPR) in the spinal cord. Disrupting mMOR-1D/GRPR heterodimerization attenuated morphine-induced scratching, but not morphine-induced analgesia [187]. The same group further found that similar to mMOR-1D, one human 7TM C-terminal variant hMOR-1Y can physically associate with GRPR to function as an itch receptor [188]. These studies further established the pharmacological relevance of OPRM1 7TM C-terminal variants and also provided a potential therapeutic strategy to specifically alleviate itch without attenuation of opioid analgesia.

Morphine Tolerance, Dependence, and Reward
To explore the in vivo functions of alternatively spliced C-terminal tails, three Cterminal truncation mouse models were established in two inbred mouse strains, C57BL/6J (B6) and 129/SvEv (129), using a stop codon targeting strategy that truncates a designated exon at the translational level with limited effects on overall Oprm1 transcription and alternative splicing (Figure 4) [83]. In the first model (mE3M), the C-terminal tails of all the variants are truncated at the end of E3 by a stop codon introduced at the end of E3. In the second (mE4M) and third models (mE7M), the E4-or E7-encoded C-terminal tails are truncated by inserting a stop codon at the beginning of E4 or E7, respectively [83]. Studies using these mice show divergent roles of the C-termini in various morphineinduced behaviors. For example, the E7-encoded C-terminal truncation in mE7M-B6 mice decreases morphine tolerance and reward without altering physical dependence, whereas the E4-encoded C-terminal truncation in mE4M-B6 mice accelerates morphine tolerance and reduces morphine dependence without affecting morphine reward (Figure 4) [83]. Moreover, in mE7M-B6 mutant mice, morphine-induced receptor desensitization in the hypothalamus and brain stem was lost, suggesting an E7-mediated mechanism underlying morphine tolerance [83].
The similarities found in several morphine-related behaviors and receptor desensitization between mE7M-B6 mutant and β-arrestin 2 KO mice suggest a physical and functional association of E7-encoded C-terminal tails with β-arrestin 2 [83,[189][190][191][192]. This supposition was supported by our in vitro cell studies showing greater β-arrestin 2 bias for several mu opioids in the E7-associated variant mMOR-1O than in the E4-associated variant mMOR-1 [83,183]. However, several mu opioid-induced actions, including morphine analgesia, locomotion, and reward [193,194], and fentanyl and methadone tolerance (unpublished observation), observed in mE7M-B6 mice were unlike those detected in β-arrestin 2 KO mice, suggesting a β-arrestin 2-independent mechanism. Together, these results further highlight the functional relevance of the 7TM C-terminal variants, particularly exon 7-associated 7TM variants, in mu opioid pharmacology. ver, in mE7M-B6 mutant mice, morphine-induced receptor desensitization in the hypothalamus and brain stem was lost, suggesting an E7-mediated mechanism underlying morphine tolerance [83].  [83]. (A). Schematic of the strategy generating C-terminal truncation mouse models. Three targeted mouse models, mE3M, mE4M, and mE7M, were generated by inserting a stop codon at an appropriate site within indicated exons shown by colored boxes. Inserted and original stop codons are indicated by yellow and pink bars, respectively. In mE3M, a stop codon was inserted at the end of exon 3 to eliminate every C-terminal tails of all 7TM and 6TM variants, as well as 1TM mMOR-1S. In mE4M and mE7M, a stop codon was created at the beginning of exon 4 or exon 7 to eliminate individual C-terminal tails encoded by exon 4 or exon 7 of indicated variants, respectively. (B). Morphine tolerance in the mutant mice on C57BL/6J (B6) background was induced by twice daily injections with morphine (10 mg/kg, s.c.) for 5 days. Morphine analgesia was determined 30 min after s.c. injection using a radiant-heat tail-flick assay. Results are shown as the percentage of maximum possible effect (% MPE). WT, wildtype mice; Mut, homozygous mice. ***: compared to WT, p < 0.001, 2-way ANOVA with Bonferroni's post hoc test. (C). Morphine physical dependence was assessed on day 5 of chronic morphine treatment with naloxone (s.c.,1.0 mg/kg) injection 3 h after the last morphine treatment to precipitate withdrawal. The number of jumps within 15 min was used for the measurement of withdrawal. *: p < 0.05; #: p < 0.0001, compared with WT, 1-way ANOVA with Bonferroni's post hoc test. (D). Morphine reward was assessed using a 6-day condi-

Involvement of 6TM Variants in Heroin and M6G Analgesia
Although the truncated 6TM variants did not bind any available radiolabeled opioids when expressed in cell lines, their functional relevance was mainly revealed using gene targeting mouse models ( Table 4). The pharmacological function of 6TM variants was first shown to involve the analgesic actions of heroin and M6G, but not morphine and methadone, by using two Oprm1 KO mouse models. In an exon 1 knockout (E1 KO) mouse, morphine and methadone analgesia was completely lost [111]. However, heroin and M6G analgesia still was active despite a slightly reduced potency. Since knocking out E1 abolished all 7TM C-terminal variants and did not disrupt 6TM variants, these results suggested that 6TM variants involve heroin and M6G analgesia, while 7TM variants were responsible for morphine and methadone analgesia. This suggestion was further supported by the opposite phenotypes seen in E11 KO mice in which heroin and M6G analgesia were greatly decreased, while morphine and methadone analgesia were normal [112]. Furthermore, the analgesic actions of all mu opioids including morphine, M6G, and heroin were lost in a exon 2 deficient mouse in which both 7TM and 6TM variants were disrupted [111], consistent with the results from E1 KO and E11 KO. The truncated 6TM variants were further discovered to mediate the actions of a novel opioid analgesic drug, 3 -iodobenzoylnaltrexamide (IBNtxA), which was derived from naltrexone [195,196]. Although the N-cyclopropyl group on the morphinan template leads to an antagonist pharmacophore, an example being naltrexone, incorporating an aryl amido group at the 6-position of the morphinan template leads to agonism at 6TM sites in the case of IBNtxA [195]. Similarly, such an incorporation in MP1104 confers its characterization as a kappa agonist at KOR-1 site [197], while MP1207 and MP1208 with the incorporation become mixed mu/kappa agonists [182]. Therefore, the 6-position arylamido group on the morphinan template puts the receptor in an agonist conformation. IBNtxA analgesia was lost in E11 KO mice, suggesting the role of exon 11-associated 6TM variants in IBNtxA analgesia. On the other hand, IBNtxA analgesia was fully active in a triple KO mouse in which only Oprm1 6TM variants remained, while all other opioid receptors including MOR 7TM variants, kappa opioid receptor (KOR-1), and delta opioid receptor (DOR-1) were knocked out, indicating that IBNtxA analgesia was independent of 7TM MOR variants, DOR-1, and KOR-1 [195]. Furthermore, intrathecal administration of lentivirus expressing individual 6TM variants, including mMOR-1G, mMOR-1M, mMOR-1N, mMOR-1K, and mMOR-1L, restored IBNtxA analgesia in a complete Oprm1 E1/E11 KO mouse in which all 7TM, 6TM, and single TM variants were disrupted [84,198]. Together, these results demonstrate that 6TM variants are essential for IBNtxA analgesia. Additionally, the opioid receptor binding studies with [ 125 I]BNtxA revealed a unique binding site in the triple KO mouse that was lost in both E11-KO and E2 KO mice [195,196,199,200]. Intriguingly, kappa-like drugs, such as ketocyclazocine and kappa 3 -like drug NalBzoH, competed for this site with high affinity, while many traditional mu, delta, and kappa opioids had very poor affinity for the site.
IBNtxA represents a novel class of opioid analgesics that are potent against multiple types of pain including thermal, inflammatory, and neuropathic pain, but do not produce many sides effects of traditional opioids, such as respiratory depression, physical dependence, and reward [195,201]. Therefore, the 6TM variants represent promising therapeutic targets for development of the novel analgesics with more favorable side-effect profiles.

Function of 6TM Variants in Delta/Kappa Opioids and Non-Opioids
Using E11 KO and E1/E11 KO mice in combination with lentivirus rescue and antisense studies, Marrone et al. revealed the involvement of truncated 6TM variants in the analgesic actions of delta and kappa opioids, as well as α 2 -adrenergic drug [202]. The analgesia of DPDPE, a delta agonist, U50,488H, a kappa agonist and clonidine, a α 2 -adrenergic agonist, was lost in the E11 KO mice, consistent with the results from the antisense studies using an antisense oligo targeting E11 [202]. Re-expression of mMOR-1G, a 6TM variant, through lentivirus in E11 KO mice rescued the analgesia of DPDPE, U50,488H and clonidine [202], further confirming that the analgesia of delta, kappa and α 2 -adrenergic agonists is dependent on the 6TM variants. However, 6TM variants did not involve delta agonist-induced seizure, kappa agonist-induced aversion, and α 2 agonist-induced hypolocomotion since these activities were still active in E11 KO mice [202]. Additionally, DPDPE analgesia was also dependent on Oprm1 7TM variants based on the observation that lentivirus-expressing mMOR-1G can only restore DPDPE analgeia in E11 KO mice, but not in the E1/E11 KO mice that lacked both 7TM and 6TM variants [202]. The 6TM variants demonstrated the ability to heterodimerize with β 2 -adrenergic receptors, providing a molecular mechanism of the analgesic synergy between the 6TM variants and β 2 -adrenergic drugs [203]. These studies expended the known functions of the 6TM variants beyond mu opioid receptors.

A Chaperon-Like Function of 6TM Variants in Enhancing Expression of the 7TM MOR-1 at Protein Level
The function of the 6TM variants was also demonstrated in enhancing expression of the 7TM MOR-1 at the protein level through a chaperon-like function [204], a similar scenario seen in the single TM variants [85] (see below). When co-expressed with mMOR-1 in a Tet-Off inducible CHO cell line, mMOR-1G dose-dependently promoted the expression of the protein but not the mRNA of mMOR-1, as determined by opioid receptor binding and [ 35 S]GTPγS binding. Using co-immunopreciptation and subcellular fractionation analysis, the study showed that the increased mMOR-1 receptor protein was mainly found in the plasma membrane, which was mediated through heterodimerization of mMOR-1G with mMOR-1 in the endoplasmic reticulum, suggesting a chaperone-like function [204].

Role of 6TM Variants in Morphine Hyperalgesia
Several studies indicated that the 6TM variants were involved in morphine hyperalgesia. In the E11-KO mouse without any 6TM variants, morphine analgesia was fully retained, but morphine-induced hyperalgesia was diminished [112,205]. However, the triple KO mouse, which retained only the truncated 6TM variants and disrupted KOR1, DOR-1, and all the E1-associated 7TM MOR variants, lost morphine analgesia but retained morphine hyperalgesia [206]. Downregulating the 6TM variant mMOR-1K with intrathecal administration of a siRNA in CXB7/ByJ mice diminished morphine-induced hyperalgesia [207], suggesting involvement of mMOR-1K in morphine hyperalgesia. Together, these findings suggest that morphine analgesia is independent of truncated 6TM variants, whereas morphine hyperalgesia is dependent on truncated 6TM variants. In addition, truncation of 6TM variants in E11-KO mice led to reduction of morphine-induced hyperlocomotion and tolerance with no effect on morphine reward and respiratory depression [205].

The Function of Single TM Variants
It is intriguing that all the single TM variants contain the first TM encoded by exon 1, which could complement the 6TM variants that lack the exon 1. Although our unpublished data showed that the single TM variants can physically associate with the 6TM variants when co-transfected into CHO cells, the functional relevance of this physical association remains unknown. Similar to the 6TM variants, the functions of the single TM variants were indicated to increase expression of 7TM MOR-1 at the protein level through a chaperonelike mechanism [85]. Using a Tet-Off CHO system, Xu et al. showed that the single TM variants can dimerize with the 7TM MOR-1 in the endoplasmic reticulum to facilitate the proper folding of MOR-1 and reduce MOR-1 degradation through endoplamic reticulumassociated degradation (ERAD), leading to increased expression of MOR-1 protein. In vivo antisense studies further suggested the role of the single TM variants in morphine analgesia through modulation of 7TM MOR-1 expression [85].
Interestingly, the first TM domain of MOR-1 was also implicated in heterodimerization with DOR-1 to disrupt MOR-1/DOR-1 dimer using an engineered protein, MOR TM1 -TAT (transactivator of transcription), resulting in enhanced morphine analgesia and reduced morphine tolerance [208]. This study raises questions on whether endogenously expressed single TM variants containing the first TM have a similar function.

Molecular Mechanisms Underlying OPRM1 Alternative Splicing
Identification of multiple OPRM1 splice variants raises questions regarding the molecular mechanisms underlying OPRM1 alternative splicing. For example, what are cis-acting elements and trans-acting factors that control the OPRM1 alternative splicing? Xu et al. initially identified an intronic single nucleotide polymorphism (SNP)(rs9479757) located near the 3 of OPRM1 exon 2 that associated with heroin addiction severity in a cohort of Chinese male heroin addicts [209]. Further characterizing this SNP using electrophoretic mobility shift assay (EMSA), minigene constructs, and RNA affinity purification coupled with proteomics led to the identification of a SNP-containing cis-acting element as an intronic splicing enhancer (ISE) that can interact with a splicing factor hnRNPH to regulate the exon 2 exclusion (skipping) or inclusion, and modulate expression of the full-length 7TM MOR-1 and 1TM variants hMOR-1S and hMOR-1Z [209]. The G allele of rs9479757 enhanced the binding of hnRNPH to this ISE, facilitating exon 2 inclusion to increase the expression of the 7TM hMOR-1 and reduce the expression of the 1TM hMOR-1S and hMOR-1Z, while the A allele had opposite effects. These results not only revealed the molecular mechanisms underlying OPRM1 exon 2 exclusion/inclusion, but also demonstrated the function of rs9479757 in regulating OPRM1 alternative splicing and contributing to heroin addiction severity. OPRM1 alternative splicing is complex. Further deciphering the molecular mechanisms of OPRM1 alternative splicing will advance our understanding of OPRM1 gene regulation and function.

Conclusions
The early concept of multiple mu opioid receptors suggested by the pharmacological studies over 40 years ago has now been revolutionized by identifying an array of OPRM1 alternatively spliced variants. These splice variants are generated through various splicing mechanisms such as alternative 3 or 5 splicing, intron retention, alternative promoters, or combinations of two or three mechanisms in a region-specific, cell-specific, or strain-specific fashion and conserved from rodents to humans. Based on the predicted transmembrane domains, these splice variants can be categorized mainly into three types: (1) full-length 7TM C-terminal variants, (2) truncated 6TM variants; and (3) truncated single TM variants. The functional relevance of these variants demonstrated by both in vitro cell lines and in vivo gene targeting animal models provides new insights into our understanding of complex mu opioid actions. More importantly, these OPRM1 splice variants are potential therapeutic targets for developing novel analgesic drugs with more favorable side-effect profiles. Many GPCRs have predicted splice variants like those of the OPRM1 gene. However, the functions of these variants remain largely unknown. Revealing the distinct functions of various OPRM1 splice variants opens questions of whether alternative splicing may extend the pharmacological repertoire of GPCRs in general.