Mutagenesis of GPR139 reveals ways to create gain or loss of function receptors

Abstract GPR139 is a Gq‐coupled receptor activated by the essential amino acids L‐tryptophan (L‐Trp) and L‐phenylalanine (L‐Phe). We carried out mutagenesis studies of the human GPR139 receptor to identify the critical structural motifs required for GPR139 activation. We applied site‐directed and high throughput random mutagenesis approaches using a double addition normalization strategy to identify novel GPR139 sequences coding receptors that have altered sensitivity to endogenous ligands. This approach resulted in GPR139 clones with gain‐of‐function, reduction‐of‐function or loss‐of‐function mutations. The agonist pharmacology of these mutant receptors was characterized and compared to wild‐type receptor using calcium mobilization, radioligand binding, and protein expression assays. The structure‐activity data were incorporated into a homology model which highlights that many of the gain‐of‐function mutations are either in or immediately adjacent to the purported orthosteric ligand binding site, whereas the loss‐of‐function mutations were largely in the intracellular G‐protein binding area or were disrupters of the helix integrity. There were also some reduction‐of‐function mutations in the orthosteric ligand binding site. These findings may not only facilitate the rational design of novel agonists and antagonists of GPR139, but also may guide the design of transgenic animal models to study the physiological function of GPR139.


| INTRODUCTION
, aka GPRg1 or GPCR12, is a highly conserved Gq-coupled receptor that belongs to the rhodopsin-like family of G-protein coupled receptors (GPCR). 1 GPR139 is almost exclusively expressed in central nervous system 2,3 where it is abundantly expressed in the medial habenula, caudate putamen and lateral septum, and is also detected in pituitary, however with much higher levels in rat compared to human. 1 We and others have established that GPR139 is activated by the essential amino acids L-tryptophan (L-Trp) and L-phenylalanine (L-Phe) with EC 50 values in the 30-to 300-μmol L −1 range, which is consistent with the physiologic concentrations of both amino acids. 1,4 Definitive proof that these are the true endogenous ligands is, however, lacking due to the challenges of manipulating the levels of these essential amino acids in vivo without impacting other critical functions such as protein synthesis and neurotransmitter precursors.  [5][6][7][8] We also demonstrated that L-Trp and L-Phe activated GPR139-dependent mitogen activated protein kinase-ERK phosphorylation. 1 Recently, Nøhr et al reported that GPR139 can be activated by adrenocorticotropic hormone (ACTH), α-, and β-melanocyte stimulating hormone (α-MSH, and β-MSH). 9 However, our recent data do not support that GPR139 is activated by ACTH, α-MSH, and β-MSH at physiologically relevant concentrations. But we were able to demonstrate an in vitro interaction between GPR139 and the melanocortin receptors (MCRs, also known as melanocyte stimulating hormone receptors), which may be an artifactual result of recombinant expression. 10 The physiological role of GPR139 is still poorly understood as it not clear yet whether L-Trp and/or L-Phe are the physiologically endogenous ligands nor is it clear whether the receptor exists to detect increasing or decreasing amino acid levels. There are limited in vivo studies utilizing the few tool compounds available. The GPR139 agonist JNJ-63533054 induced a dose-dependent reduction in locomotor activity in rats. 1 Bayer Andersen et al 11 have recently reported that the GPR139 agonist TC-O 9311 protected primary mesencephalic dopamine neurons against 1-methyl-4-phenylpyridinium (MPP+)-mediated degeneration, indicating a potential role of GPR139 in neuroprotection and Parkinson's disease. 11 Another GPR139 agonist, 4-oxo-3,4-dihydro-1,2,3-benzotriazine, has been reported to improve social withdrawal. 12 From a structural perspective, Kaushik and Sahi reported a three dimensional structure prediction and molecular dynamics simulation of GPR139 model. Further structure-based virtual screening was applied and several active site residues of GPR139, including Tyr32, Glu105, Glu108, and Tyr192, have been identified as potential ligand binding sites for inhibition of protein dimerization and receptor activity. 13 Shehata et al 14 also presented a combined structure-activity relationship, and a refined pharmacophore model of GPR139. 14 By applying in vitro and in silico mutagenesis, they demonstrated a common binding site for GPR139 surrogate agonists, including TC-O 9311, JNJ-63533054, L-Trp, and L-Phe, indicating that residues Phe109, His187, and Asn271 are important for the binding of these ligands. 15 The goal of the present study was to further understand the critical structural motifs for GPR139 receptor activity as well as create new tools to further study the physiological role of GPR139. We used both site-directed and random mutagenesis approaches and identified series of human GPR139 mutation clones with gain-offunction, reduction-of-function as well as loss-of-function properties.
The pharmacological profiles of these mutant receptors were characterized by using calcium mobilization assay, radioligand binding assays and receptor protein expression assay. Further, these structure-activity data were incorporated into a homology model. Collectively, the results suggest that the endogenous L-Trp, and L-Phe share a common binding site with the GPR139 agonists, TC-O 9311 and JNJ-63533054. This common binding site consists of a buried hydrophobic pocket defined by residues Ile112, Phe109, Val191, Tyr192, and His187 and a proximal polar region defined by residues Arg244, Glu108, and Asn171. Lastly, the larger GPR139 agonists (TC-O 9311 and JNJ-63533054), also share a less buried hydrophobic pocket defined by residues Val76, Phe79, Ile104, Val83, and Ile80. This common binding site was also previously identified by Norh et al 15 Perhaps more importantly, we were able to identify mutations that rendered the receptor either more sensitive or less sensitive to the purported amino acid ligands. This should allow for the creation of transgenic animals that will require higher or lower levels of endogenous ligand to achieve activation, and enable more insight into the in vivo function of GPR139.

| Molecular cloning of GPR139
GPR139 from human was cloned from respective brain cDNAs as previously described in detail. 1

| Site directed mutation clones of GPR139
Site directed mutagenesis was performed using an overlapping PCR method. Specific primers (forward and reverse primers for each mutant as shown in Table S1) designed specific to the mutations were synthesized by Eton Biosciences. The N-terminal V5-tagged GPR139 DNA plasmid was used as the template for the PCR reactions. The 5′ end primer (5′ ATT CTT CTC GAG GCC ACC ATG GGT AAG CCT ATC-3′) and the reverse primer for the mutant were used to PCR the 5′ end of the mutant gene while the forward primer of the mutant and the 3′ end primer (5′-ATG TCT GCG GCC GCT CAC GGG GAT ACT TTT ATA GGT TTT CCA TTT TTG TCA TAC TG-3′) were used to PCR amplify the 3′ end of the mutant gene. The 5′ end and the 3′ end PCR products were mixed to serve as the template to synthesize the complete mutant gene by PCR using the 5′ end and the 3′ end primers described above. The PCR products were cloned into pCIneo between Xho1 and Not1 sites and the insert regions were sequenced to confirm the identities for each mutant gene.  Briefly, HEK293 cells were grown without antibiotics in black 96well poly-D-lysine coated FLIPR plates in F-12K medium plus 10% FBS and 1 × sodium pyruvate at a density of 20 000 cells/well. One hundred nanograms per microliter of DNA and 3 μL of FUGENE-HD were used for each well transfection. Triplicated cell plates were transfected for each 96-well DNA plate. The wild-type GPR139 DNA was used as the positive control and pCIneo plasmid without an insert was used as the negative control. Two days after transfection, the cell culture medium was replaced with HBSS buffer containing the Ca 2+ dye and then incubated for 45 minutes at 37°C. A two-addition FLIPR assay was used to characterize the mutant receptors with a 1 st addition of low concentration of L-Trp (15 μmol L −1 ) and L-Phe (30 μmol L −1 ) followed 3 minutes later by the addition of a high concentration of L-Trp (1 mmol L −1 ) and L-Phe (2 mmol L −1 ). The Ca 2+ signal from the 1st addition, the 2nd addition, as well as the ratio of the signals from two additions were used to assess the relative response of the mutant receptors.

| Total and surface V5-GPR139 expression
V5-tagged GPR139 protein expression was tested by ELISA assay using HEK293 cells expressing human V5-GPR139 wild-type and the mutant receptors. Mock-transfected HEK293 cells served as the negative control. Transfected cells were plated in 96-well poly-D-lysine plates at a cell density of 20 000 cells/well. For total GPR139 protein expression, the transfected cells were fixed using 10% formaldehyde in Dulbecco's phosphate buffered saline (DPBS, Hyclone, Pittsburgh, PA) and then treated with 1% Triton X-100 (Sigma) in DPBS. Cells were blocked with 3% nonfat milk in 1% Triton X-100 (Sigma) in DPBS and then incubated with 1 μg/ml of mouse monoclonal IgG2a Anti-V5 antibody (Life Technologies, Carlsbad, CA) as the primary antibody. Second detection was carried out using goat anti-mouse IgG, horseradish peroxidase (HRP) conjugate (30 ng/mL, Thermo Scientific, Waltham, MA) and developed with 3,3′,5,5′-Tetramethylbenzidine (TMB) Substrate Reagent (BD Biosciences, San Jose, CA). Cell-surface expression of V5-GPR139 was measured by performing the ELISA assay described above without using Triton X-100 as the penetrating reagent.

| Response normalization strategy
In order to categorize the changes in sensitivity caused by random mutations of GPR139, all candidate clones were first prescreened by a calcium mobilization assay using a double addition normalization strategy ( Figure 1). In essence, we devised this procedure to set each receptor as its own control for normalization in order to attempt to compare across various mutants. Figure 1 represents the theoretical percent (%) response of the random mutation clones in the FLIPR assay. The EC 20 dose of agonist (mixture of 15 μmol L −1 of L-Trp and 30 μmol L −1 of L-Phe) and saturation dose of agonist (mixture of 1 mmol L −1 of L-Trp and 2 mmol L −1 of L-Phe) were defined by a GPR139 wild-type dose response curve ( Figure 1A).
When the candidate clone is stimulated by the EC 20 dose of the agonist, a more sensitive clone will elicit a stronger response when compared to that of the wild type, while a less sensitive clone will elicit a weaker response ( Figure 1B). These changes in response are due to the mutation induced shift in agonist potency as seen in Figure 1A. When the candidate clone is stimulated by a saturating dose of the agonist, all three types of clones will elicit E max responses ( Figure 1C).
When the "two additions" assay is applied, the candidate clone is first stimulated by the EC 20 dose and 3 minutes later by the saturation dose of agonist. The response of EC 20 dose is similar to that of single addition of EC 20 dose illustrated in Figure 1B. This stimulation will lead to receptor desensitization and when the saturation dose addition is applied, a more sensitive clone will elicit a weaker response when compared to the wild type, instead a less sensitive clone will elicit a stronger response ( Figure 1D). Consequently, the normalized percent response ratio of 1st/2nd addition is used to determine the sensitivity change of the candidate clones. The clone with a higher ratio (eg, 40%/60% = 2/3) compared to the wild type (eg, 20%/80% = 1/4) is considered as more sensitive clone and a lower ratio (eg, 10%/90% = 1/9) is considered as less sensitive clone.
With this strategy, since each mutant receptor is self-normalized, the response differences between different mutant clones caused by experimental conditions (eg, difference in cell numbers between wells, transfection efficiencies, protein expression levels, equipment reading differences, etc) will be normalized, and batches of data can be compared in this large-scale screening. All the clones that show changes in sensitivity when compared to wild type are to be further characterized by dose response studies.

| Characterization of human GPR139 receptor gain-of-function random mutations
We identified several human GPR139 receptor gain-of-function random mutation clones by calcium mobilization assay, in terms of their increased sensitivity to the putative endogenous ligands L-Phe and L-Trp (Table 1). The potency of the selective agonists TC-O 9311 and JNJ-63533054 with these mutation clones was also assessed in the assay (Table 1). We subcategorized Table 1  reduction in E max regards to L-Phe and L-Trp response, respectively (Table 1). And a few of the gain-of-function mutations, including Val76Ala, IIe116Thr, Cys23Arg, and Asn93Asp, resulted in an increase of E max to both endogenous ligands. In general, most of these gain-of-function mutations showed similar levels of total and cell-surface protein expression levels (Table 1). Only a few double mutations exhibited moderate reduction (>50%) in both protein expressions (Table 1).
In particular, we show that the single mutation Val191Ala, Ser267Pro, Val76Ala, and IIe116Thr resulted in 5.7 to 6.6-fold increase and 2.4 to 6.3-fold increase in L-Phe and L-Trp potency, respectively ( Table 1). The effect of these mutations on ligand binding affinity were characterized using a saturation binding assay with In general, the K i value of the ligands were not significantly affected by these mutations ( Table 2). The dose response curves to L-Phe and L-Trp of the representative gain-of-function mutants are shown in Figure 2A and B.

| Characterization of human GPR139 receptor reduction-of-function random mutations
We also identified a number of human GPR139 receptor mutations that resulted in reduction-of-function as measured by ligandinduced calcium mobilization (Table 3). Again, we subcategorized Table 3 (Table 3). Also, most of these reduction-of-function mutations exhibited moderate to significant reduction in E max regards to all four ligands responses. (Table 3).
In terms of total and cell-surface protein expression levels, most T A B L E 2 Representative human GPR139 receptor gain-of-function random mutations radioligand binding assays summary  Concentration binding of [ 3 H]-JNJ-63533054 were determined using a saturation binding assay in HEK293 cells transiently transfected with mutated or wild-type human GPR139. K d (nmol L −1 ) and B max (fg/mg protein) values are means ± SD (n = 3). K d and B max fold changes are expressed as a ratio, that EC 50 of agonist of the mutation to the wild type. Competition of [ 3 H]-JNJ63533054 binding by L-Phe, L-Trp, TC-O 9311, and JNJ-63533054 were determined using a competitive inhibition assay in COS-7 cells transiently transfected with mutated or wild-type human GPR139 wild-type human GPR139. Inhibition equilibrium constants (K i ) values are means ± SD (n = 3). K i fold changes are expressed as a ratio, that EC 50 of agonist of the mutation to the wild type. of these mutants showed more than 50% decrease in both protein expression levels (Table 3).
In particular, the single mutation Cys286Arg, Lys225Glu and the double mutations Ile128Asn, Phe237Ser resulted in a significant decrease in L-Phe and L-Trp potency, which was beyond the detection limit of the FLIPR assay (Table 3, Figure 3). All three mutations also exhibited 5.7 to 14.4-fold and 4.4 to 8.0-fold decrease in TCO-9311 and JNJ-63533054 potency, respectively (Table 3). These results confirmed the expression of these mutated receptors on the cell membrane, which was further proven by the cell-surface protein expression assay, though decreased compared to the wild type (Table 3). The binding affinity of selected four single mutants was also characterized by the radioligand binding assays. Leu86Pro resulted in a 16.8-fold increase in K d and a 2.2-fold decrease in B max (Table 4). This mutation also resulted in a significant reduction in the ability of L-Phe, L-Trp, TCO 9311, and JNJ-63533054 to compete for binding, which was revealed by a 42.6 to 363.2-fold decrease in K i value of the ligands in the competitive inhibition assay (Table 4).
In terms of the other three selected mutations, the K d values were not significantly affected. All three mutations showed 1.7 to 6.2-fold decrease in B max (Table 4), which is in consistent with their decreased cell-surface expression ( Table 3). The binding properties of L-Phe, L-Trp, TC-O 9311. and JNJ-63533054 (K i ) were not significantly affected by these mutations ( Table 4). The dose response curves to L-Phe and L-Trp of the representative reductionof-function mutants are shown in Figure 3A and B.

| Characterization of human GPR139 receptor loss-of-function random mutations
Apart from the gain-and reduction-of-function mutations, we also identified a list of human GPR139 receptor loss-of-function random mutations. These mutations resulted in the lack of responses to all four tested ligands at the highest concentrations applied. We further categorized these loss-of-function mutations into two sub groups.
Some of these mutants exhibited proper, though decreased, total and cell-surface expression levels (Table 5) (Table S3).

| Characterization of human GPR139 receptor site directed mutations
As transmembrane domain 3 (TM3) is considered as one of the common core domains involved in ligand binding and activation in most receptors that respond to small organic molecules, 20 we also carried out a screening of human GPR139 receptor TM3 site directed mutations, from Ile103 to Val130. All residues were mutated to Ala individually and the potency to all four ligands was assessed in the calcium mobilization assay (Table S4). We report that TM3 site-directed mutations did not result in any gain- another common binding site shared by these ligands and a mutation from a hydrophilic residue Asn to a hydrophobic residue Ala also resulted in a complete loss of function. In our case, we identified Asn271Ser change through random mutation studies, which mutated between two hydrophilic residues, and did not cause significant functional changes (Table S3). Ile112 forms the bottom of the orthosteric site and mutation to an Ala would impact ligand binding. And lastly Asp125 and Arg126 are a part of the DRY motif at the end of TM3 that is involved in G-protein binding, mutations to these residues disrupt the ability of the GPCR to bind with the G-protein.

| Homology model of human GPR139 receptor
Finally, all the structure-activity data described above are incorporated into a homology model of hGPR139 with all four ligands pre- sented. An overview of hGPR139 ligand binding site with surface representation is shown in Figure 4.  (Tables 1 and 2), reduction-of-function (Table 3 and 4) and loss-of-function with protein expression (Table 5) are shown in Figure S5A, S5B and S5C, respectively.

| DISCUSSION
In the present study, we identified a series of human GPR139 mutations which result in a gain, reduction, or loss of function. To further characterize these single and multiresidue mutations, we mapped their positions onto a homology model of GPR139 in order to interpret their impact on both ligand binding and receptor activation.
Lys225Glu (Table 3) which reside on helix 5 and 6, two helices that are known to structurally shift upon binding of G-proteins. 22 Additionally, Phe287Tyr, and Cys286Arg, a gain-of-function (Table 1) and a reduction-of-function (Table 3) mutation located at the junction of helix 7 with the terminal helix 8 also residing in an area in close proximity to G-protein binding.
A third set of mutations ( Figure 5C) were identified which impact the transmembrane helices structure which in turn could change the shape of the orthosteric pocket or the ability for the G-protein to bind. These include reduction-of-function mutations Leu249Pro, Ser62Pro, and Leu86Pro (Table 3) and gain-of-function mutations Ile116Thr and Ser267Pro (Table 1). All but one of these mutations  Figure 4) docked into the orthosteric pocket (transparent gray surface). Gain-of-function mutations are shown with carbon atoms as green spheres, and reduction-of-function mutations in with carbon atoms as pink spheres, nitrogen atoms are shown as blue and oxygen atoms as red spheres, respectively. Identified mutations are grouped by their impact on (A) orthosteric ligand binding site, (B) G-protein binding area, and (C) GPR139 secondary or tertiary structure are to a proline residue. Proline residues are known to break helical secondary structure thus altering the helical nature of the transmembrane helices. Ile116Thr is the only exception, it is found to be both the gain-of-function mutation in the calcium mobilization (Table 1) and the radioligand binding (Table 2) assays. It sits near the orthosteric pocket but is not involved in forming the ligand binding site.
However, it sits just under Val191Ala (identified as residue that impacts ligand binding site, Figure 5A). A mutation of a hydrophobic isoleucine to a polar threonine, likely impacts the relative placement of Val191 and thus impacts the structure of the orthosteric site.
There are two additional sets of mutations identified which could have a structural impact on the function of GPR139. The first set of mutations, Thr278Ile and Leu41Ser ( Figure S5B), were identified as reduction-of-function in both calcium mobilization and radioligand binding assays (Tables 3 and 4). These residues reside in a region that is known to form a water channel important for GPCR activation. 23 In particular, the Thr278Ile mutation would impact the water channel itself by changing from a polar threonine residue to a hydrophobic isoleucine residue. Leu41Ser sits just proximal to Thr278Ile but would not be involved in formation of the channel.
However, a mutation from a large hydrophobic leucine to a small polar serine likely impacts the placement of the Thr278 residue changing the shape of the water channel. The second set of mutations, Asn93Asp and Cys23Arg are two extracellular loop mutations ( Figure S5A) that were identified as gain-of-function in the calcium mobilization assay (Table 1). At first glance these residues would appear to be in a location that would not impact the structure of the GPCR, the ability of the ligands to bind orthostatic pocket, nor the ability of G-proteins to bind. However, movement of the transmembrane helices is known to be required for GPCR activation and given the connectedness of the extracellular region to the transmembrane helices, mutations to this region could influence the orientation of the transmembrane helices and thus modulate receptor activation. 24 Further, we identified several multipoint loss-of-function mutations with proper protein expression (Table 5, Figure S5C). Although it is harder to pinpoint the exact structural changes resulting from multipoint mutations, three of the loss-of-function mutations are either proline mutations to nonproline residues or vice versa, including Pro142Gln at the end of TM2, Pro238Thr in the middle of TM6, and Ser197Pro in the middle of TM5. As discussed previously, proline residues are known to break helical structures and any mutations involving prolines could impact secondary structure.
Additionally, two loss-of-function multipoint mutations involving Tyr285 were identified. Tyr285 is located at the bottom of TM7 in a region where G-protein binding occurs and where other mutations (Phe287Tyr and Cys286Arg) were found to also impact GPR139 function.
In conclusion, our study demonstrated several critical structural motifs in human GPR139 receptor activity. These lines of information not only facilitate the rational design of novel potent and selective agonists and antagonists of GPR139, but also guide the design of transgenic animal models (carrying the gain-of-function or reduction-of-function point mutations) for better understanding the physiological function of GPR139 in the future. In particular, traditional knockout models may lead to a compensatory phenotype by the system. While in comparison, the gain-in function transgenic animal may provide a different angle exploring the biological function of receptors, which are activated by their natural ligands with relatively high EC 50 values under normal conditions.

DISCLOSURE
None declared.