Abstract
GPR37L1 is an orphan receptor that couples through heterotrimeric G-proteins to regulate physiological functions. Since its role in humans is not fully defined, we used an unbiased computational approach to assess the clinical significance of rare G-protein-coupled receptor 37-like 1 (GPR37L1) genetic variants found among 51,289 whole-exome sequences from the DiscovEHR cohort. Rare GPR37L1 coding variants were binned according to predicted pathogenicity and analyzed by sequence kernel association testing to reveal significant associations with disease diagnostic codes for epilepsy and migraine, among others. Since associations do not prove causality, rare GPR37L1 variants were functionally analyzed in SK-N-MC cells to evaluate potential signaling differences and pathogenicity. Notably, receptor variants exhibited varying abilities to reduce cAMP levels, activate mitogen-activated protein kinase (MAPK) signaling, and/or upregulate receptor expression in response to the agonist prosaptide (TX14(A)), as compared with the wild-type receptor. In addition to signaling changes, knock-out (KO) of GPR37L1 or expression of certain rare variants altered cellular cholesterol levels, which were also acutely regulated by administration of the agonist TX14(A) via activation of the MAPK pathway. Finally, to simulate the impact of rare nonsense variants found in the large patient cohort, a KO mouse line lacking Gpr37l1 was generated. Although KO animals did not recapitulate an acute migraine phenotype, the loss of this receptor produced sex-specific changes in anxiety-related disorders often seen in chronic migraineurs. Collectively, these observations define the existence of rare GPR37L1 variants associated with neuropsychiatric conditions in the human population and identify the signaling changes contributing to pathological processes.
Significance Statement
G-protein-coupled receptors represent a diverse group of membrane receptors that contribute to a wide range of diseases and serve as effective drug targets. However, a number of these receptors have no identified ligands or functions, that is, orphan receptors. Over the past decade, advances have been made, but there is a need for identifying new strategies to reveal their roles in health and disease. Our results highlight the utility of rare variant analyses of orphan receptors for identifying human disease associations, coupled with functional analyses in relevant cellular and animal systems, to ultimately reveal their roles as novel drug targets for treatment of neurological disorders that lack widespread efficacy.
Introduction
Migraine is a common neurovascular disorder affecting 12–15% of adults with a strong sex bias toward females (Burch et al., 2015; Bonafede et al., 2018). Migraine frequency and symptomology vary widely among individuals. Dissection of the underlying neurobiology has focused on increased neuronal hyperexcitability (Goadsby et al., 2017; Sutherland et al., 2019) and/or changes in the trigeminal–vascular system of the dura (Noseda and Burstein, 2013; Goadsby et al., 2017).
The causal basis for susceptibility to migraine remains poorly understood. Genetic differences account for up to 60% of migraine occurrence (Russell et al., 1995; Mulder et al., 2003; Gormley et al., 2016; Hansen et al., 2017; Sutherland et al., 2019). At one end of the spectrum, genome-wide association studies (GWAS) and subsequent meta-analyses of migraine have identified common markers (single nucleotide polymorphisms, SNPs) in or near genes associated with glutamate signaling (MTDH/AEG-1, LRP1, MEF2D), synaptic development (ASTN2, FHL5), endothelial cell and blood vessel wall function (PHACTR1, MMP16, AJAP1, TSPAN2, TGFBR2), pain pathways (TRPM8), or metabolism (PRDM16, C7orf10) (Gormley et al., 2016). Since GWAS loci are commonly found in intronic or intergenic regions, assigning them to adjacent genes and defining pathologic mechanisms has been challenging. At the other end, familial studies have identified rare mutations in specific genes (NOTCH3, TREX1, CACNA1A, ATP1A2, SCN1A) associated with monogenic hemiplegic migraine (Hansen et al., 2017; Sutherland et al., 2019, 2020). Since rare variants are often present in coding regions, understanding how these mutations directly affect protein function and contribute to migraine is now emerging (Tardiolo et al., 2019). Although both strategies provide important mechanistic insights, the cumulative impact of known genes accounts for only a small fraction of estimated migraine heritability.
To address the “missing” heritability, recent attention has focused on the possible contribution of genes carrying rare variants that are not captured by GWAS genotyping assays and exert moderate effects not easily detected by classical linkage analysis in familial studies (Manolio et al., 2009). As a test of this concept, we selected 85 G-protein-coupled receptors (GPCRs) (Dershem et al., 2019) whose endogenous ligands, biological functions, and disease contributions have not been fully elucidated, that is, orphan GPCRs (Tang et al., 2012; Fricker and Devi, 2018). We examined whole-exome sequences (WES) from 51,289 individuals, predicted likely pathogenic variants, and then binned “likely pathogenic” plus “pathogenic” variants together. Subsequently, sequence kernel association testing (SKAT) was conducted to determine the genes that were significantly associated with the International Classification of Disease Diagnostic (ICD) codes found in the electronic health records (EHRs) of these individuals (Dershem et al., 2019). Among the top hits, the G-protein-coupled receptor 37-like 1 (GPR37L1) gene was significantly associated with generalized epilepsy. Providing proof of concept for this unbiased approach, a point mutation in GPR37L1 has previously been found in a consanguineous family with a progressive form of myoclonus epilepsy (Giddens et al., 2017). Attesting to the discovery potential of this approach, we report here that variation within the GPR37L1 gene was also significantly associated with migraine. The encoded GPR37L1 receptor is highly expressed in astrocytes and microglia (Meyer et al., 2013; Jolly et al., 2018; Liu et al., 2018), as well as in satellite glial cells of the human trigeminal ganglia (Yang et al., 2022). Although postulated to contribute to neuroprotection, the cellular pathways have been poorly defined. In this paper, functional characterization of migraine-associated missense variants in GPR37L1 was conducted revealing agonist-induced alterations in expression, signal amplitude, and pathway bias. In addition, targeted knock-out of GPR37L1 or transient expression of certain rare variants produced alterations in cellular cholesterol homeostasis, suggesting a causal connection (Hanin et al., 2021). Finally, a knock-out (KO) mouse line lacking Gpr37L1 was generated revealing that the loss of this receptor led to sex-specific behavioral changes implicated in migraine, but no effect on cephalic pain in an acute migraine model. Altogether, the human and mouse results focus attention on GPR37L1 as a novel, astrocytic contributor to migraine-related disorders.
Materials and Methods
Genomic analyses
To generate disease associations across the phenotype spectrum represented in the EHR, we used WES from 51,289 participants recruited through the MyCode® Community Health Initiative, which includes Geisinger patients enrolled through primary care and specialty outpatient clinics (Carey et al., 2016). Participants, who consented to broad research use of genomic data in the DiscovEHR database linked to EHR entries, were 59% female, median age of 61 years, predominantly Caucasian (98%), and non-Hispanic/Latino (95%). Sample preparation, sequencing, sequence alignment, variant identification, genotype assignment, and quality control steps were carried out as previously described (Abul-Husn et al., 2016; Carey et al., 2016; Dewey et al., 2016). Rare variants (MAF < 0.01) were sorted into loss-of-function (LOF), nonsense and frameshift, variants and missense (MISS) variants. MISS variants were further triaged by bioinformatics (RMPath score; Dershem et al., 2019) to predict benign, likely benign, likely pathogenic, and pathogenic variants, which were binned for subsequent analyses. Details of the bioinformatics triage are as previously reported (Dershem et al., 2019). Deidentified EHR data obtained from an approved data broker was sorted to identify all unique ICD9 codes with ≥200 patients with at least three independent incidences of the code. Individuals with no calls of each code were set as controls. Non-Europeans and one individual from closely related pairs up to first cousins were excluded from analyses, as non-Europeans were present in numbers too small for statistical analysis as a separate group, and closely related pairs would skew association analyses. Models were adjusted for sex, age, age2, and first four principal components. We used the sequence SKAT with default weights (SKAT R package) to compare the burden of rare variants in cases and controls. To define the directionality of the association, all individuals with rare variants in GPR37L1 in 51,289 WES and with EHR codes (ICD9 or ICD10) for migraine (346.*, G43.*) or epilepsy (345.*, G40.*) were identified and the RMPath scores of the variants determined, as previously described (Dershem et al., 2019). All variants were classified as “variants of unknown/uncertain significance” (VUS), that is, benign (VUSB), likely benign (VUSLB), likely pathogenic (VUSLP), or pathogenic (VUSP), according to the net scores determined by the algorithm. The primary focus of this study was migraine, but both migraine and epilepsy are diseases characterized by hyperexcitability (Rogawski, 2008); therefore, genomic analysis was designed to assess the degree of variant overlap between the two phenotypes.
Generation and culture of SK-N-MC GPR37L1-KO cell line. The SK-N-MC GPR37L1-KO cell line was generated by the CRISPR/Cas9 method (Mashiko et al., 2013). The gRNA was designed using the Optimized CRISPR Design online tool (http://crispr.mit.edu). One highly scored gRNA was found in GPR37L1 exon 1, 5′-GCCCTAACCCCGGCAAGGAT GGG-3′. This gRNA was cloned into pSpCas9(BB)-2A-Puro (PX459) V2.0, (Addgene, #62988). The positive clone was amplified to get the plasmid after sequence analysis. The SK-N-MC cell line (ATCC, #HTB-10™) was cultured in Eagle's minimal essential medium (EMEM) medium (ATCC) with 10% fetal bovine serum (FBS) (Invitrogen). On the day of electroporation, cells were detached by 5 min treatment with StemPro Accutase (Invitrogen) and isolated by centrifugation. 5 × 10⁶ cells were suspended in 100 µl of electroporation buffer with 20 µg plasmid, electroporated (Neon™ electroporation system, manufacturer's suggested conditions), and seeded onto three cell culture dishes (10 cm, Corning). After 24 h, puromycin (Sigma-Aldrich, 1 µg/ml) selection was initiated for 48 h. Puromycin-resistant colonies were picked mechanically using sterile 20 µl tips, and each colony was maintained in a separate well of a 48-well plate. GPR37L1 KO clone(s) were identified by PCR and sequence analysis (forward primer, 5′-GCCCTAACCCCGGCAAGGAT-3′; reverse prime, 5′-AGTAGCTGTGCCACACGATGCACATGAC-3′). Thirteen KO clones were identified by sequence analysis. The KO clones were screened for off-target effects by Sanger sequencing for suspected genic sites that were identified by the Optimized CRISPR Design online tool. In silico prediction determined the resemblance of genomic regions to the gRNA with up to three nucleotide mismatch tolerance; in total, three coding sites were analyzed. No alternate target modifications were observed in any of the GPR37L1-null clones compared with the sequence of the original SK-N-MC clone. Clone #40 was identified without off-target effects and was chosen for use in experimental studies (GPR37L1-KO SK cells). This clone has a 94-base pair deletion starting 3′ of the CRISPR/Cas9 cleavage site. The deletion caused a DNA coding sequence shift, resulting in a premature stop codon (TGA) in the sequence.
Isolation and culture of primary mouse astrocytes
Primary mouse astrocytes were cultured from postnatal Day 1–2 C57BL6 pups. Mice were quickly decapitated, and brains were removed and submerged in an ice-cold dissection medium containing 1 mM sodium pyruvate and 0.1% glucose in HEPES-buffered saline, pH = 7.4. The meninges were removed, and cortices were isolated and finely minced. A single-cell suspension was obtained by enzymatic digestion with 0.25% trypsin at 37°C for 20 min followed by trituration. Cells were seeded in a T75 flask in Dulbecco's minimal essential medium culture medium supplemented with 10% FBS and 1% penicillin–streptomycin antibiotics (Invitrogen). The medium was changed twice weekly. When cells became confluent, microglia and oligodendrocytes were removed by shaking at 260 rpm for 3 h at 37°C in an orbital shaker. Prior to biochemical assays, astrocytes were subcultured into different plates as indicated.
Generation of wild-type (WT) and GPR37L1 variants
The human GPR37L1 cDNA clone (OriGene) was subcloned into pcDNA3.1+ (Invitrogen) and used as the template to insert the Glu-Glu epitope sequence (-EYMPME-) and the GPR37L1 variants used in the experiments by in vitro mutagenesis using Q5 Site-Directed Mutagenesis Kit (BioLabs). cDNA clones were confirmed by Sanger DNA sequencing (GENEWIZ). Primer sequences are available upon request.
Western blotting
WT GPR37L1 and variants (without or with the Glu-Glu tag) were transfected into GPR37L1-KO SK cells (SK-N-MC Transfection Kit, Altogen Biosystems). Cells were lysed after 48 h and adjusted for equivalent amounts of protein using the BCA protein assay (Thermo Scientific), and then 200 µg of lysate was incubated with an anti-Glu-Glu tag antibody (Thermo Scientific) overnight at 4°C. The next day, the immunoprecipitated complexes were pulled down by protein G beads and eluted with 5× SDS sample buffer. Eluted samples were run on 10% TGX gels (Bio-Rad), transferred to PVDF membrane (Bio-Rad), blocked with 5% milk in TBS-T, and exposed overnight at 4°C to an anti-GPR37L1 polyclonal antibody (Abcam #151518). Blots were exposed to a secondary HRP-conjugated antibody (GE HealthCare), developed with SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific), and imaged on an Odyssey Fe imager (LI-COR) and processed with Image Studio. When untagged or natively expressed receptors were studied by Western blot, membrane preparations were prepared according to the Abcam Membrane Preparation Protocol, using 500 µl of cell fractionation buffer per 10 cm dish of confluent cells, scraped into glass tubes on ice (15 min). The cell suspension was passed 10× through a 25 gauge needle and incubated on ice (20 min), and then particulates were pelleted in two rounds of centrifugation (3,000 rpm, 5 min, 4°C), combining the resulting supernatants. Membranes were then pelleted from the combined supernatants by ultracentrifugation (85,000 rpm, 40 min, 4°C), and the pellet was resuspended in the buffer. Membranes were combined with a 5× SDS sample buffer and Western blots performed as described above.
Enzyme-linked immunosorbent assay (ELISA)
WT Glu-Glu-tagged GPR37L1 and Glu-Glu-tagged variants were transfected into GPR37L1-KO SK cells as described above. After 24 h, equivalent numbers of cells were replated to poly-l-lysine-coated 96-well plates (Corning) and further cultured for 24 h. In a subset of experiments, cells were treated for various times with 40 µM TX14(A) (Tocris). Cells were fixed with either MeOH or 4% paraformaldehyde for 15 min on ice; all further steps were carried out at room temperature. Cells were rinsed with TBS-T, blocked in 1% milk/TBS-T (1 h), then incubated with an anti-Glu-Glu tag antibody (1 h), washed, and then incubated with HRP-conjugated antibodies (1 h). Samples were developed with TMB liquid substrate supersensitive solution (Sigma-Aldrich) for 30 min and stopped with 1 M sulfuric acid. Absorbances were read at 450 nm on CLARIOstar (BMG LABTECH). Nontransfected cells were used to determine background, and blanks and background were subtracted from all samples. Three to four replicates for each condition were run in each experiment, which was repeated with at least three independent transfections.
cAMP assay
WT and variants were transfected into GPR37L1 KO SK cells. After 48 h, cells were treated with 10 µM forskolin (FSK; Sigma-Aldrich) in DMSO ± various concentrations (0–40 µM) of TX14(A) (Tocris) in water (1 mM stock, as per manufacturer's instructions), and then cAMP levels were determined with the nonacetylated version of the direct cAMP ELISA kit (ENZO) following manufacturer's guidelines. Controls received DMSO only. Absorbance signals at 405 nm were read on a plate reader (CLARIOstar, BMG LABTECH). Optical densities were fitted to 4PL curve in Excel (Microsoft) and data analyzed with GraphPad Prism (V.8.4.3). Raw data from three independent experiments run in triplicate were combined, averaged, and analyzed.
ERK1/2 phosphorylation (ERK1/2∼P) assay
GPR37-L1-KO SK cells was cultured in an EMEM medium (ATCC) with 10% FBS (Invitrogen). On the day of electroporation, cells were detached by 5 min treatment with StemPro Accutase (Invitrogen) and isolated by centrifugation. 5 × 10⁶ cells were suspended in 100 µl of buffer with 20 µg pcDNA3.1-GPR37L1 plasmid, electroporated (Neon™ electroporation system, manufacturer's suggested conditions), and seeded on 60 mm cell culture dishes (Corning). After 24 h, the cells were detached by 5 min treatment with StemPro Accutase and collected by centrifugation. 30,000 cells were seeded on each well in a 96-well plate. The cells were incubated for another 24 h and then starved in 50 µl EMEM serum-free medium (antibiotic-free) in the presence or absence of a protein kinase inhibitor (PKI) (6–22amide, 50 nM, Gibco-BRL) for 3 h. The culture medium was removed, and cells were rinsed three times with warmed PBS. Cells were treated with various concentrations (0–80 µM) of TX14(A) (Tocris) for 15 min in 50 µl of EMEM in the presence or absence of a PKI at 37°C.
For the ERK1/2∼P assay, phospho- and total ERK1/2 were determined using the two-plate adherent cell protocol (Phosph-ERK1/2 kit, Cisbio), following manufacturer's instructions. Fluorescent emission signals were read at two wavelengths (665 and 620 nm) on a plate reader (CLARIOstar, BMG, LABTECH), the gain of filter setting for 665 nm is at 2,000 and for 620 nm is at 1,700, and HTRF ratios were determined, and the resulting data were analyzed with GraphPad Prism (V.8.4.3). Raw data from three independent experiments run in triplicate were combined, averaged, and analyzed.
Total cellular cholesterol assay
WT SK-N-MC cells, GPR37L1-KO cells, and GPR37L1-KO cells transiently, transfected with WT or variant GPR37L1 cDNAs, were split into 96-well plates and cultured overnight. The total cell cholesterol was measured by cleaving cholesterol esters with esterase and then complexing for 30 min in the dark at 37°C with a 50µl Amplex Red reagent/HRP/cholesterol oxidase working solution (Invitrogen). Fluorescence was measured on a plate reader (CLARIOstar, BMG LABTECH). Values were fitted to an in-plate cholesterol standard curve. Each condition was run in quadruplicate in three independent experiments, and raw data were averaged over all experiments.
Generation and validation of Gpr37l1 knock-out mouse line. To simulate the impact of rare nonsense variants found in the large patient cohort, a KO mouse line lacking Gpr37l1 was generated using the CRISPR/Cas9 method (Mashiko et al., 2013). The online tool (http://crispr.mit.edu) was used to identify a highly scored sgRNA within exon 1 (5′-GGTACAGCACTATGTACCCGAGG-3′), which was subsequently cloned into the pSpCas9(BB)-2A-Puro (PX459) V2.0 vector (Addgene, #62988). After confirmation of positive clones by Sanger sequencing (GENEWIZ), the plasmid was injected into the pronuclei of one-cell-stage C57BL/6J embryos according to standard protocols, and injected embryos were transferred to pseudopregnant mice (Transgenic and Gene Targeting Core, Georgia State University). From the 11 progeny, five mutant mouse pups were confirmed to have a single nucleotide deletion (C) at 3′ of the CRISPR/Cas9 cleavage site that introduced a premature stop codon (TGA), as described below. The Gpr37l1+/− male mice were mated with C57BL/6J female mice (Jackson Laboratory) to produce mice carrying the deleted Gpr37l1 allele (Gpr37L1), which were then intercrossed to produce the WT, heterozygous, and homozygous KO mice on a congenic background. To assess the potential off-target effects of the CRISPR/Cas9 technology, the KO mice were analyzed by direct Sanger sequencing for changes at suspected sites identified by the Optimized CRISPR Design online tool. In silico prediction determined the resemblance of genomic regions to the gRNA with up to three nucleotide mismatch tolerance; in total, three coding sites were analyzed, and no alternate target modifications were observed in Gpr37l1 null mice.
For validation, genotyping was performed on genomic DNA obtained from tail clips, and PCR was performed using Taq DNA polymerase (New England Biolabs). Primers (forward primer 5′-ATCCATGCCAGACGCCTGAC) and reverse primer (5′-ATCCCCGAGTAGCCTTTGCT) were used to amplify a 722 bp fragment flanking the edited Gpr37l1 sequence (Extended Data Figure 7-1). The amplified DNA fragment was digested with Ava1 (underlined) and subjected to agarose gel electrophoresis The PCR product from WT mice was digested into two fragments (322 and 400 bp) with Ava1, whereas the PCR products from the mutant mice could not be digested with Ava1. Subsequently, the loss of the transcript and protein was confirmed by qPCR and immunoblot analysis on mouse brains from WT, heterozygote (het), and KO (hom) mice, using an anti-GPR37L1 polyclonal antibody (GeneTex, #100005).
Animal use
All animal experiments were performed in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee of Florida Atlantic University. For behavioral studies, animals were bred on congenic C57BL/6J line, with Gpr37l1-KO mice designated as Gpr37l1−/− and WT littermates designated as Gpr37l1+/+. All mice were group housed (3–4/cage) and used for experiments after they reached 9 weeks of age. Animals were kept on a 12 h light/dark cycle with lights off at 7:00 P.M. in a quiet temperature-controlled room (24–25°C).
Drugs
CGRP (AnaSpec) was dissolved in PBS, used at the dose of 0.1 mg/kg (Mason et al., 2017; Rea et al., 2018), and intraperitoneally (i.p.) administered in a 5 ml/kg volume injection.
Periorbital mechanical allodynia
Behavioral assessment was conducted by an operator blinded to treatment conditions as we have recently described (Targowska-Duda et al., 2020; Sturaro et al., 2023). In brief, following an acclimation period in which mice were gently handled, habituated to i.p. injection, and held as during the test, two baseline pain measurements were conducted. To reduce variability, only the second baseline was used for data analysis. Female and male mice were tested on separate days by the same operator. von Frey filaments of different forces (0.02, 0.04, 0.07, 0.16, 0.4, 1.0, and 1.4 g) were applied to the periorbital area perpendicular to the skin, with sufficient force to cause slight buckling, and held for ∼2 s to elicit a positive response. Mice were poked five times with the same filament in a uniform manner throughout the periorbital region before changing the filament. The stimulation was initiated with the 0.16 g filament. A response occurred when the mouse stroked the face with its forelimb, withdrew its head from the stimulus, or shook its head. Sensitivity was determined according to the up-down method. The 50% mechanical withdrawal threshold (expressed in g) was then calculated.
Anxiety-like behavior
Anxiety-like behavior was assessed using the elevated plus maze (EPM) test as we described (Targowska-Duda et al., 2020; Sturaro et al., 2023). The EPM apparatus was a black, “plus”-shaped platform equipped with two open arms and two closed arms of the same dimensions (35.6 × 7.6 cm). The platform was at a raised height of 50 cm above the ground and placed under room lighting of 500 lux. The EPM test started when each mouse was placed on the central platform of the EPM apparatus faced to a closed arm and left free to explore the maze for 5 min. Each mouse performance was recorded by a micro camera set in the testing room and examined at a later time by an operator blinded to the treatment schedule. The operator recorded the time spent in the open arms of the maze as well as the entries into the open as measures of anxiety-like activity and the number of closed arm entries as an indicator of the mouse locomotor behavior.
Statistical analyses
TIBCO Statistica (Version 13.5.0.17) was used for data analysis. Assessment of periorbital allodynia was analyzed as area under the curve (AUC) for the various treatment combinations followed by two-way ANOVA. Unpaired t test was used for analysis of anxiety-like behavior in the EPM. Significance was set at p < 0.05.
Results
Disease associations of rare variants in GPR37L1
A computational approach was used to assess the clinical significance of GPR37L1 genetic variants in 51,289 (51 K) WES. Rare [i.e., mean allele frequency (MAF) < 0.001] GPR37L1 variants were present in the heterozygous state and classified as LOF (nonsense or frameshift) and MISS variants (Dershem et al., 2019). Although there were too few LOF variants for computational analysis, missense variants were categorized according to their predicted pathogenicity using a bioinformatics algorithm (i.e., RMPath) (Dershem et al., 2019), and 27 “likely pathogenic” or “pathogenic” variants were binned for further analysis. Subsequently, SKAT was performed by running the binned variants against all ICD9 codes with ≥200 individuals with three or more independent encounters of the code in their EHRs to ensure accuracy of disease diagnoses. As previously reported (Dershem et al., 2019) and shown here (Table 1), this unbiased approach revealed that the top disease association for GPR37L1 variants was generalized convulsive epilepsy (ICD9 code 345.1; Q = 6.29 × 10−17), consistent with a published study linking GPR37L1 with epilepsy in a consanguineous family (Giddens et al., 2017). Although the clinical phenotype of carriers was not reported, family members homozygous for the GPR37L1 variant (p.Lys349Asn) developed a novel, progressive myoclonus epilepsy and early death by 20 years of age (Dershem et al., 2019). Also, pointing to the pleiotropic potential of this approach, epileptic seizures are commonly associated with traumatic mouth injuries (Gawlak et al., 2017), and GPR37L1 variants were strongly linked to traumatic fracture of the tooth (ICD9 code 873,63; Q = 3.07 × 10−12). Likewise, GPR37L1 variants were associated with migraine (ICD9 code 346.1 migraine; Q = 1.13 × 10−8), and epilepsy and migraine are known to share many clinical features and pathological mechanisms (Rogawski, 2012; Zarcone and Corbetta, 2017; Shu et al., 2020). Finally, attesting to the discovery promise of this approach, significant associations with malignant neoplasm of the kidney (ICD9 code 189) and unspecified disorder of kidney and ureter (ICD9 code 593.9) were also identified, underscoring the potential importance of GPR37L1 in vascular and renal function that has been suggested by mouse and human studies (Min et al., 2010; Coleman et al., 2018; Zheng et al., 2019; Mouat et al., 2021a,b).
SKAT analysis of rare variants of GPR37L1
Rare genomic variants in GPR37L1 in individuals with migraine and/or epilepsy
The current study has focused on addressing the impact of genetic variation within the coding sequence of GPR37L1 on the incidence of migraine. Since migraine and epilepsy are both diseases of hyperexcitability and associated with GPR37L1 rare variants, we performed clinical lookups of consented patients carrying GPR37L1 variants with the relevant diagnoses in 51,289 WES, with the goal of assessing variant overlap and/or pleiotropy. First, we assessed the three individuals heterozygous for rare nonsense (LOF) variants and found that none had three validated calls of migraine or epilepsy in their EHRs. Two were males (age range 62–86) and one was a female (age 77). Next, we captured all MISS variants and used the RMPath algorithm to predict their pathogenicity scores (Fig. 1A). Finally, we identified the smaller set of MISS variants found in individuals with at least three encounters for migraine (ICD9 code 346.*) or epilepsy (ICD9 code 345.*). Consistent with the relative prevalence of these disorders in the adult population (Burch et al., 2015), we found a larger number of MISS variants in individuals affected with migraine than epilepsy. Moreover, a smaller number of variants were identified in individuals with both diagnoses with significant overlap of variants but not in the same individuals, suggesting a common genetic basis for these conditions (Fig. 1A, overlap).
Pathogenicity scores of rare GPR37L1 variants. A, Flowchart of GPR37L1 rare variants in 51,289 WES, triaged by MAF, variant type (SYN, synonymous; LOF, loss-of-function nonsense or frameshift; MISS, missense). Bioinformatics triage was done by calculating RMPath scores (Dershem et al., 2019) of all rare variants and then identifying the subset of those found in migraine (346.*) and epilepsy (345.*) patients; range of scores was from −2 to 22; maximum possible score, 24. Scores were divided into seven bins of equal width, B, benign; LB, likely benign; LP, likely pathogenic; and P, pathogenic. Overlap indicates rare variants found in both migraine and epilepsy patients (not necessarily in the same patient). B, Snake plot of GPR37L1 generated with Protter software (Omasits et al., 2014) with locations of rare missense variants chosen for functional studies (p.Arg69Trp, p.Met303Val, p.Val361Met, p.Arg391His, p.Arg391Cys, p.Val414Leu, p.Ser443Leu). The location of the internal Glu-Glu tag is also indicated. Red, signal peptide residues; orange, variants studied; light green, cysteines participating in extracellular disulfide bond.
Since both migraine and epilepsy are complex disorders with genetic and environmental contributors (Ragowski, 2012; Huang et al., 2017; Zarcone and Corbetta, 2017; Shu et al., 2020), the penetrance of one or the other condition in individuals carrying GPR37L1 variants is difficult to estimate. From the EHRs, we extracted both sex and age at the first encounter for migraine and epilepsy. As shown in Table 2, carriers with a migraine diagnosis showed a significant bias toward females (∼79%), while those with an epilepsy diagnosis exhibited comparable incidence among males and females (61%). The DiscovEHR 51,289 WES dataset is slightly biased toward females, that is, 59% female (Dewey et al., 2016). Thus, the relative number of variants and the differential sex bias of the two phenotypes defines sex is an additional contributor. Further, the overall frequency of migraine in individuals heterozygous for GPR37L1 rare MISS variants was 15.5% (males/females combined), which is significantly higher than the frequency of migraine in the general Geisinger population, 10.2% (males/females combined), supporting a contribution of GPR37L1 variant burden to migraine incidence.
Demographics of individuals in 51,289 WES dataset with GPR37L1 variants
SK-N-MC cells as a viable model system to study GPR37L1 expression and function
GPR37L1 is highly expressed in astrocytes (Meyer et al., 2013) and satellite glial cells of sensory ganglia (Yang et al., 2022) and, to a lesser extent, in neurons, microglia, and oligodendrocyte precursors (Jolly et al., 2018; Liu et al., 2018). Accordingly, we identified neuroblastoma-derived SK-N-MC cells, which endogenously express this receptor, as a viable cell model to assess the impact of genetic variation on GPR37L1 expression and function. SK-N-MC cells (hereafter abbreviated SK cells) are easily transfected and used broadly for studies of cellular signaling. To provide a null background, the CRISPR/Cas9 method was used to engineer SK cells lacking GPR37L1 (KO), as confirmed by PCR screening (Fig. 2A) and immunoblotting analyses (Fig. 2B). To assess GPR37L1 signaling in SK cells, we tested fragments of the multidomain-secreted protein prosaposin (PSAP), which has been identified as the endogenous ligand of GPR37L1 in some studies (Meyer et al., 2013; Jolly et al., 2018), but contradicted in others (Coleman et al., 2016; Ngo et al., 2021). The PSAP protein is composed of four structurally similar modules, termed saposins A–D, and saposin C (SapC) is implicated in activation of GPR37L1. Because GPR37L1 is Gi/o coupled (Liu et al., 2018), we determined the efficacies of SapC domain, or a synthetic peptide fragment derived thereof, TX14(A), on inhibition of FSK-stimulated cAMP production. GPR37L1-KO SK cells transfected with WT GPR37L1 were stimulated with FSK (10 µM) in the absence or presence of varying concentrations of either SapC or TX14(A). As shown in Figure 2C, both the endogenous and synthetic agonists inhibited FSK-activated cAMP generation with similar IC50s (12.9 and 7.1 µM, respectively) although SapC produced a more robust inhibition of cAMP generation. The TX14(A)-mediated inhibition of cAMP production is present in WT SK cells but notably absent in GPR37L1-KO SK cells (Fig. 2D). Likewise, TX14(A) showed a greater than 50% reduction in FSK-mediated cAMP production in primary astrocytes (Fig. 2E). Taken together, these results demonstrate that WT SK cells express GPR37L1 as the sole target of TX14(A), and thus GPR37L1-KO SK cells provide the appropriate cellular context for comparison of signaling by WT and variant GPR37L1.
GPR37L1 expression and function in SK-N-MC cells. A, RT-PCR demonstrating endogenous expression of GPR37L1 in SK-N-MC cells (WT) and RT-PCR validation of a CRISPR/Cas9-generated GPR37L1-KO SK cell clone (KO). Final lane shows gel run of mixture of both PCR products (MIX). B, Western blot of membrane preparations from primary mouse astrocytes, WT and GPR37L1-KO SK cells, probed with an anti-GPR37L1 antibody (Abcam). Membranes were isolated as described in Materials and Methods; 30 µg of membranes per lane. Full blot of GPR37L1 in Extended Data Figure 2-1. C, The PSAP module SapC and TX14(A) (Thr-D-Ala-Leu-Asp-Asn-Asn-Ala-Thr-Glu-Glu-Ile-Leu-Tyr) (Meyer et al., 2013) activate GPR37L1. Control SK-N-MC cells were stimulated with 10 µM FSK ± different concentrations of the peptide TX14(A) (IC50 7.2 µM) or SapC (IC50 12.9 µM). cAMP was analyzed as described in Materials and Methods. Curves were generated by fitting all raw data from three independent experiments. D, WT and GPR37L1-KO cells were stimulated with 10 µM FSK ± 40 µM TX14(A), and cAMP levels were assayed as described in Materials and Methods. E, Primary mouse astrocytes were stimulated with 10 µM FSK ± 40 µM TX14(A), and cAMP levels were assayed as described in Materials and Methods. Significances in D,E were determined by two-tailed t test assuming unequal variances, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
Functional analysis of GPR37L1 variants associated with migraine
For functional analyses, we chose GPR37L1 variants across the range of predicted pathogenicity located at sites conserved in the human and mouse to broaden the long-term impact of these studies, that is, VUSB, p.Met303Val and p.Val414Leu, VUSLB, p.Arg69Trp, p.Ser443Leu and p.Arg391His, and VUSLP, p.Val361Met, and p.Arg391Cys. Fortuitously, two variants with different predicted pathogenicities were identified at Position 391, that is, p.Arg391His (likely benign, RMPath score 8) in a single individual with no instances of relevant phenotypes, and p.Arg391Cys (likely pathogenic, RMPath score 15), found in two individuals, one with migraine. Table 3 summarizes the variants chosen and patient demographics, if any. Since bioinformatics predictions cannot account for variant-mediated changes of secondary or tertiary structure or at binding sites, the variant locations are marked in relation to the topology of the receptor (Fig. 1B), that is, in the N-terminal domain (p.Arg69Trp), ECL2 (p.Met303Val), at the junction of the ICL3 and helix 6 (p.Val361Met), ECL3 (p.Arg391His, p.Arg391Cys), toward the cytoplasmic end of helix 7 (p.Val414Leu), and in the C-terminal segment (p.Ser443Leu).
GPR37L1 migraine-associated variants selected for functional studies
Functional analyses of these variants were performed in KO SK cells expressing either WT or variant receptors, and the relative abilities of TX14(A) to attenuate FSK-stimulated cAMP production (Meyer et al., 2013) were measured (Fig. 3, Table 4). We noted four distinct changes in signaling. Two receptor variants [p.Met303Val (VUSB), p.Arg391His (VUSLB)] had signaling properties indistinguishable from the WT receptor in the presence or absence of TX14(A) (Fig. 3Aa). Two other receptor variants [p.Val414Leu (VUSB), p.Ser443Leu (VUSLB)] were more effective at reducing cAMP than WT receptor in the presence of an agonist (Fig. 3Ab). Likewise, two receptor variants [p.Val361Met (VUSLP), p.Arg391Cys (VUSLP)] were more effective at reducing cAMP than the WT receptor in the presence of TX14(A), but also showed significantly reduced cAMP in the absence of the agonist, indicative of basal constitutive activity (Fig. 3Ac). The final receptor variant [p.Arg69Trp (VUSLB)] was less effective at inhibiting cAMP than the WT receptor in the presence of TX14(A) (Fig. 3Ad). Figure 3B summarizes the diverse range of basal and TX14(A)-inhibited cAMP levels as well as apparent alterations relative to WT for each variant. Overall, these results show only a weak correlation between bioinformatics-based predictions of pathogenicity and experimental analysis of receptor variants, with analysis of receptor activity revealing altered signaling properties of most receptor variants.
Migraine-associated variants in transfected GPR37L1-KO cells have differential ability to inhibit cAMP production. A, Migraine-associated variants were assessed for their ability to attenuate FSK-stimulated cAMP production when activated by TX14(A). GPR37L1-KO cells were transiently transfected with WT or variants, and net cellular cAMP was determined at a range of TX14(A) concentrations in cells stimulated with 10 µM FSK, as described in Materials and Methods. Plots present averaged raw data ± SD results of three independent transfections. Variants are plotted with WT (solid line in all graphs), according to their different patterns of response. a, p.Met303Val (dashed line), p.Arg391His (dotted line). b, p.Val414Leu (dashed line), p.Ser443Leu (dotted line). c, p.Val361Met (dashed line), p.Arg391Cys (dotted line). d, p.Arg69Trp (dashed line). B, Quantitation of at least three independent experiments as illustrated in A. Inhibition curves were fitted and the maximal inhibition by TX14(A) (a), the IC50 for TX14(A) (b), and basal cAMP levels (c), zero TX14(A), were determined as described in Materials and Methods. Statistical differences were determined by two-tailed Student's t test, compared with WT responses, with *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
Summary of functional differences between WT and GPR37L1 variants
GPR37L1 also regulates mitogen-activated protein kinase (MAPK) pathways (Meyer et al., 2013); therefore, we next tested the ability of WT and receptor variants to activate ERK1/2∼P (Fig. 4A). We noted three patterns of signaling alterations. Three receptor variants, two with normal cAMP inhibition (p.Met303Val, p.Arg391His) and one with increased potency for cAMP inhibition (p.Ser443Leu), had agonist-induced ERK1/2∼P comparable to the WT receptor (Fig. 4Aa). Three receptor variants (p.Val414Leu, p.Val361Met, p.Arg391Cys) with enhanced cAMP inhibition were found to have reduced ERK1/2∼P relative to WT in response to the agonist (Fig. 4Ab). The last variant (p.Arg69Trp), with lower cAMP inhibition, had an apparent increase in Emax (Fig. 4Ac) which was not well fitted over the accessible TX14(A) concentration range, although there was a significant, reproducible increase in ERK1/2∼P at 60 and 80 µM TX14(A) (Fig. 4Ac). Despite variable changes in Emax, all receptor variants had fitted EC50s comparable to the WT receptor. Finally, we tested whether there was interaction between the cAMP and MAPK pathways by measuring ERK1/2∼P in cells expressing WT GPR37L1 in the absence or presence of 50 nM protein kinase A inhibitor peptide (PKI). As illustrated in Figure 4C, addition of PKI had no impact on ERK1/2∼P in the presence or absence of TX14(A). Taken together, these results demonstrate that a subset of GPR37L1 variants found in patients with a migraine diagnosis showed changes in agonist-induced cAMP inhibition and ERK1/2∼P that may be indicative of changes in signaling bias.
Migraine-associated variants in transfected GPR37L1-KO cells have differential ability to stimulate ERK1/2∼P. A, The ability of TX14(A) to induce ERK1/2∼P in GPR37L1-KO cells transfected with WT or variants was determined in a plate-based assay as described in Materials and Methods. Plots illustrate averages from three independent transfections. Variants were plotted with WT (solid line) and sorted by their response differences. Variants found in patients with migraine are underlined in all plots. a, p.Met303Val (dashed line), p.Arg391His (dotted line), p.Ser443Leu (dash-dotted line). b, p.Val361Met (dashed line), p.Arg391Cys (dotted line), p.Val414Leu (dash-dotted line). c, p.Arg69Trp (dashed line). B, Stimulation curves were fitted and the EC50 estimates (a) and the span (maximum − minimum) (b) were determined as described in Materials and Methods. Statistical differences were determined by two-tailed Student's t test, compared with WT responses, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. C, Interaction of cAMP and MAPK pathways in GPR37L1 responses was tested by treatment of KO cells or cells transfected with WT GPR37L1, with 50 nM PKI, showing its addition had no impact on ERK1/2∼P in the presence or absence of TX14(A), as determined by one-way ANOVA analysis. 
KO cells, 
KO cell + PKI, 
WT cells, 
WT cell + PKI. Three independent experiments were run in triplicate. NS, not significant.
Impact of migraine-associated variants on GPR37L1 expression and localization
To determine whether the signaling differences among GPR37L1 variants could in part be due to altered expression and/or subcellular distribution, we incorporated an epitope tag that allows net expression and plasma membrane targeting to be accurately assessed. Because GPR37L1 can be cleaved by endoproteases at N-terminal sites (Coleman et al., 2016, 2020), the epitope tag was inserted at a site in the amino terminus distal to potential GPR37L1 cleavage sites to ensure accurate assessment of expression levels. The small Glu-Glu epitope tag (-EYMPME) was introduced between residues p.Asn105 and p.Leu106 at the extracellular amino terminus (indicated in Fig. 1B). Tagged WT GPR37L1 (=Glu-GPR37L1) allowed efficient immunoprecipitation (IP) of transfected receptors from KO SK cells. Figure 5A illustrates a Western blot that compares the sizes of untagged WT GPR37L1 (MembWT) and WT Glu-GPR37L1 (IP). The blot was probed with a polyclonal anti-GPR37L1 antibody, and both Glu-GPR37L1 and the untagged WT GPR37L1 had a dominant band consistent with the predicted mass of the full-length receptor. Glu-GPR37L1 had a slightly larger mass than the untagged version, as expected. There was no evidence for proteolytic cleavage of the receptor in this cell line (Fig. 5B).
Variant-associated changes in GPR37L1 expression or localization. A, GPR37L1-KO SK cells were transfected with either untagged WT or Glu-GPR37L1, followed by IP with an anti-Glu antibody or isolation of membranes to assess untagged GPR37L1. A second Glu-GPR37L1-transfected sample was treated ± 40 µM TX14(A) (30 min) prior to IP. Blot was probed with an anti-GPR37L1 polyclonal antibody (Abcam). B, Western blot analysis. Whole-cell extracts (50 µg) were separated by 10% SDS-PAGE, and the membrane was blotted with a rabbit anti-GPR37L1 polyclonal antibody (GeneTex, GTX10005) diluted at 1:1,000. Subsequently, the HRP-conjugated goat anti-rabbit IgG antibody (1:10,000, Jackson ImmunoResearch Laboratories) was used to detect the primary antibody, revealing the absence of a ∼52 kDA band in GPR37L1-KO SK cells (Lane 1) versus its presence in WT SK cells (Lane 2) and in GPR37L1-KO SK cells that were transfected with pcDNA-GPR37L1 in three different experiments (Lanes 3–5). Notably, there was no evidence of smaller bands associated with N-terminal cleavage. C, GPR37L1-KO SK cells transfected with Glu-tagged WT or variant GPR37L1, immunoprecipitated with anti-Glu antibodies, and analyzed by Western blot probed with an anti-GPR37L1 antibody (see Materials and Methods). D, ELISA assays of Glu-tagged WT or variants assessing surface (left panel, fixed with paraformaldehyde) or total (middle panel, fixed with cold methanol) expression in the absence of TX14(A). The right panel illustrates the surface / total distribution of WT and variants. Data presents mean of all points (4 replicates×3 independent experiments) or mean ± SD (right panel). Statistical analysis by two- way ANOVA with multiple comparisons, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. E, Response of GPR37L1-KO SK cells transfected with either WT or Glu-GPR37L1 to TX14(A) inhibition of FSK-mediated cAMP production. F, ELISA assays to assess total expression of GPR37L1 were used to compare basal (no TX14(A), white bars) and 30 min in 40 µM TX14(A) (gray bars) for WT and GPR37L1 variants. Three independent experiments were run in quadruplicate, and mean protein levels for each experiment were averaged and normalized to WT at zero time or after 30 min treatment with TX14(A). Data were analyzed by one-way ANOVA with multiple comparisons, and significance is indicated as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.
Detection of Glu-tagged WT and variant receptor was performed by Western blotting (Fig. 5C), while surface and total expressions were quantified by ELISA (Fig. 5D). Although most variants were comparably expressed to WT (surface expression, Fig. 5Da; total expression, 5 Db), two receptor variants (p.Val361Met and p.Arg391Cys) exhibited a reduced total expression relative to WT and thus higher surface to total expression ratios (Fig. 5Dc). As changes in surface expression of receptors can produce significant alterations in downstream signaling, it is interesting to note that both variant receptors also showed enhanced cAMP inhibition (recall Fig. 3Ac). Surprisingly, we also found that short-term treatment with TX14(A) produced an upregulation of GPR37L1 (Fig. 5A). To explore the nature of this effect and to assess the potential impact of receptor variants, KO SK cells expressing Glu-tagged WT or variant receptors were treated with TX14(A) (30 min, 40 µM). First, we determined that the Glu tag did not affect receptor signaling. Both WT Glu-GPR37L1 and the untagged WT GPR37L1 inhibited FSK-mediated accumulation of cAMP with comparable IC50s (Fig. 5E). Thus, the Glu tag does not alter key features of the WT receptor and was therefore used to assess changes in expression and/or signaling by GPR37L1 variants.
Although the maximal extent varied, agonist treatment caused rapid, time-dependent increases in their expression, with receptor variants p.Val361Met and p.Arg391Cys with the weakest upregulation (Fig. 5F). It is of note that p.Arg391His has signaling equivalent to WT in both cAMP and ERK1/2∼P assays and shows net expression and upregulation comparable to WT, while the p.Arg391Cys variant has enhanced basal activity and decreased EC50 in the cAMP inhibition assay, and reduced Emax in the ERK1/2∼P assay, and significantly less TX14(A)-mediated upregulation. Altogether, these results demonstrate the impact of GPR37L1 genetic variation on signaling efficacy, bias, and expression, with p.Arg391Cys and p.Val361Met showing changes across all three metrics. Results also reveal a rapid upregulation of GPR37L1 expression in response to agonist treatment that is dependent upon receptor occupancy and/or signaling output(s).
GPR37L1 and total cellular cholesterol levels
Perturbations in brain cholesterol homeostasis are linked to neurological disorders including migraine, in which hyperexcitability may play a role (Hanin et al., 2021). To assess the importance of GPR37L1 to cholesterol homeostasis, WT and KO SK cells and KO cells transiently transfected with or without GPR37L1 receptor were used (Fig. 6A). A comparison of total cholesterol content among these cell lines revealed that the loss of endogenous expression of GPR37L1 significantly reduced cholesterol content (Fig. 6B, compare WT and KO cells), while transient transfection of the WT receptor restored cholesterol abundance to a normal level. Of even more interest, acute treatment of cells expressing endogenous or exogenous GPR37L1 with the agonist TX14(A) for 30 min produced statistically significant increases in total cholesterol, which was not observed in KO cells lacking the receptor (Fig. 6B). These results point to a novel role for GPR37L1 in the acute regulation of cholesterol homeostasis in response to the agonist.
GPR37L1 regulation of cellular cholesterol levels. A, Western blot of WT SK cells, GPR37L1-KO cells, KO cells transfected with WT GPR37L1, or KO cells exposed to a transfection reagent (KO-C), blotted with an anti-GPR37L1 antibody (Abcam). B, Total cellular cholesterol levels in WT SK cells, GPR37L1-KO cells transfected with WT GPR37L1, or GPR37L1-KO cells, treated with vehicle or 40 µM TX14(A) for 30 min. C, Total cellular cholesterol was measured in GPR37L1-KO cells transfected with WT or variants and treated without (a) or with 40 µM TX14(A) (b) for 30 min. Analysis of B and C was performed by two-way ANOVA with multiple comparisons, relative to WT under the same condition, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. D, Comparison of total cellular cholesterol changes evoked by 40 µM TX14(A) treatment for 30 min, in WT or variant-transfected GPR37L1-KO cells. White, untreated, black, TX14(A)-treated. Analysis by Student's t test, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. E, Impact of U0126 on total cholesterol levels was assessed by treatment with either U0126 alone or in the presence of TX14(A) (=Combo), using WT SK cells. KO SK cells were used as a control, and treatment with U0126 had no effect on cholesterol levels in the absence of GPR37L1 (data not shown). Significance was determined by one-way ANOVA with multiple comparisons, with *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
Since treatment with TX14(A) enhanced the differences in cellular cholesterol levels between WT and KO SK cells, we used the same approach to compare the total cellular cholesterol levels in KO cells transiently transfected with WT or variant GPR37L1. In the absence of TX14(A), cells expressing WT or receptor variants had equivalent amounts of total cell cholesterol, while there was an ∼20% lower level in the KO SK cell line (Fig. 6Ca). However, agonist treatment revealed differential responses between WT and receptor variants. Two receptor variants (p.Arg69Trp and p.Met303Val) showed weaker responses compared with WT, while two others (p.Val361Met and p.Arg391Cys) failed to increase total cholesterol levels in response to TX14(A) (Fig. 6Cb). Variant responses to TX14(A) are plotted in Figure 6D. Of particular note, two receptor variants (p.Arg391Cys and p.Val361Met) with altered signaling properties in both cAMP and ERK1/2 assays failed to significantly increase total cholesterol levels in response to acute treatment with TX14(A). The different variants at position p.Arg391 again showed differential responses, that is, TX14(A) elicited no cholesterol increase in cells expressing p.Arg391Cys, while cells expressing p.Arg391His had an increase in total cholesterol in TX14(A) similar to WT.
Acute regulation of total cell cholesterol by GPCRs has not been fully characterized, although a recent study has demonstrated neuropeptide Y-mediated upregulation of cholesterol in liver cells, a response that could be blocked by the MEK1/2 inhibitor U0126 (Chen et al., 2020). Here, we determined whether the MAPK pathway contributed to the TX14(A)-mediated upregulation of total cholesterol in WT SK cells. As we have shown (Fig. 4), TX14(A) increases ERK1/2 phosphorylation, and we hypothesized that blocking the upstream activator of ERK1/2 with U0126 should prevent TX14(A)-mediated increases in cellular cholesterol. Figure 6E illustrates that TX14(A) caused a significant increase in total cholesterol in WT SK cells. WT SK cells had higher total cholesterol levels than KO SK cells, suggesting some constitutive activity of the endogenous GPR37L1 in WT SK cells. The U0126 had no effect on unstimulated WT or KO SK cells, but when treated with TX14(A), the U0126 severely attenuated the increase in cholesterol in WT SK cells. Overall, the results demonstrate that acute stimulation of GPR37L1 contributes significantly to total cellular cholesterol homeostasis through activation of MAPK pathways.
Role of GPR37L1 in migraine-related disorders
To simulate the impact of rare nonsense variants found in the large patient cohort, a KO mouse line lacking Gpr37l1 was generated on a congenic background to assess its role in neurological processes and permit dissection of the underlying mechanism(s) (Fig. 7). For the initial in vivo characterization, we focused on two migraine-related behaviors that share similar clinical features and suspected pathophysiology. For the acute migraine phenotype, we relied on a widely used model for migraine pathophysiology, acute CGRP injection, which produces facial pain sensitivity as measured by von Frey monofilaments (Mason et al., 2017; Rea et al., 2018; Sturaro et al., 2023). In this case, KO and WT littermates showed increased pain sensitivity following a single CGRP injection that was indistinguishable between the two genotypes in female and male mice. Two-way ANOVA, conducted to analyze possible differences in allodynic responses calculated as AUC in Gpr37l1−/− versus Gpr37l1+/+ female mice treated with or without CGRP, led to a significant main effect of CGRP treatment [F(1,46) = 4.4; p = 0.04] accompanied by the lack of a genotype effect [F(1,46) = 0.6; p = 0.38] and “treatment × genotype” interaction [F(1,46) = 0.0; p = 0.76; Fig. 8A]. In male mice, there was a significant effect of CGRP treatment [F(1,45) = 18.3; p = 0.0001], which, similarly to the outcome in females, was not accompanied by a significant genotype effect [F(1,45) = 1.0, p = 0.315] or interaction [F(1,45) = 0.4; p = 0.534; Fig. 8B]. One possibility for the lack of difference in the CGRP effects between WT and KO animals is that the CGRP dose was too high. However, the use of lower CGRP dose also revealed no difference (Fig. 8C). Collectively, these results suggest that a single CGRP treatment evoked periorbital mechanical allodynia in male and female mice. However, the loss of Gpr37l1 did not affect the response to acute CGRP treatment.
Generation and validation of Gpr37l1 KO mice. A, Schematic representation of sgRNA targeting sequence and PAM site in the mouse Gpr37l1 exon 1 region. The Ava1 recognition sequence is underlined, and a single base pair “C” deletion in Gpr37l1 KO mice is indicated in sequence. B, Genotyping of WT and Gpr37l1-KO mice. A 722 bp PCR product was amplified with a pair of primers that flanked the gene-edited sequence of Gpr37l1. The PCR product from WT mice was digested into two fragments (322 and 400 bp) with Ava1. However, the PCR products from the mutant mice could not be digested with Ava1. C, RT-PCR shows the relative quantity of Gpr37l1 mRNA extracted from mouse brains in the WT, heterozygote (het), and KO (hom) mice. D, Western blot analysis of protein expression in the brains of WT, het, and KO mice.
Gpr37l1 KO (Gpr37l1−/−) mice show similar responses in an experimental model of migraine. Following acclimation, (A) female and (B) male Gpr37l1−/− and Gpr37l1+/+ mice were tested for two basal thresholds of periorbital mechanical allodynia using the up-down method and then i.p. injected with calcitonin gene-related peptide (CGRP, 0.1 µg/kg) or VEH. In the CGRP model, both Gpr37l1−/− and Gpr37l1+/+ female and male mice showed reduced thresholds of periorbital mechanical allodynia following CGRP treatment. Data are the mean sensitivity thresholds (grams) ± SEM 10–14 (total mice: 99) mice per group. *p < 0.05 difference from Gpr37l1+/+ groups; +p < 0.05, +++p < 0.001 difference from VEH-treated groups. C, Gpr37l1 KO (Gpr37l1−/−) and WT mice (Gpr37l1+/+) respond similarly to the CGRP challenge of 0.05 mg/kg dose. Following acclimation, female and male Gpr37l1−/− and Gpr37l1+/+ mice were tested for two basal thresholds of periorbital mechanical allodynia using the up-down method and then i.p. injected with calcitonin gene-related peptide (CGRP, 0.05 mg/kg) or VEH. Thresholds were then recorded at 30 min and 1 h following CGRP or VEH injections. Data are the mean sensitivity thresholds (grams) ± SEM of 6–8 (total mice: 27) mice per group.
For the anxiety-related characterization, we used an established paradigm, the EPM, which measures time spent in the open versus enclosed arms (Targowska-Duda et al., 2020; Sturaro et al., 2023). To assess the effect of genetic disruption of the Gpr37l1 gene on anxiety phenotype, we compared their behavior in the EPM paradigm. Unpaired t test revealed that female Gpr37l1−/− mice significantly spent less time than WT mice on the open arms of the EPM [t(20) = 4.9; p < 0.0001; Fig. 9A]. In contrast, male Gpr37l1−/− mice showed similar open arm time to WT (Gpr37l1+/+) mice [t(14) = 0.9; p = 0.359; Fig. 9B]. Mutant Gpr37l1−/− female mice also entered the open arms of the maze less than Gpr37l1+/+ mice [t(20) = 3.7; p = 0.0014; Fig. 9C], whereas male Gpr37l1−/− mice performed similarly to WT [t(14) = 1.2; p = 0.238; Fig. 9D]. Gpr37l1−/− mice of both sexes exhibited similar exploration of the maze, as shown by similar entries onto the closed arm of the EPM [females: t(20) = 0.4, p = 0.685; males: t(14) = 2.0, p = 0.0613; Fig. 9E,F]. Altogether, these results suggest that Gpr37l1 deletion leads to sexually dimorphic anxiety-like responses, with female Gpr37l1−/− mice exhibiting increased anxiety-like activity as compared with WT controls.
Sex differences in anxiety-like behavior for Gpr37l1−/− mice. A, Female but not (B) male GPR37L1−/− mice spent less time than Gpr37l1+/+ mice exploring the open arms of the EPM, a behavior consistent with increased anxiety. Similarly, (C) female but not (D) male Gpr37l1−/− mice showed a reduced number of entries onto the open arms of the maze than Gpr37l1+/+ mice, a second measure of anxiety-like behavior. E,F, For both sexes, differences in anxiety parameters were accompanied by a similar number of closed arm entries (similar exploration of the EPM maze) for the two phenotypes. Values are presented as mean percent (%) open arm time and mean number of open and closed arm entries ± SEM of 10–12 (female) and 7–9 (male) Gpr37l1−/− versus Gpr37l1+/+ mice. **p < 0.01, ***p < 0.001 difference between phenotypes.
Discussion
Identifying GPR37L1 as a contributor to migraine and epilepsy risk
Using an unbiased computational approach, we found that GPR37L1 coding variants were significantly associated with both epilepsy and migraine supporting their comorbid status and shared etiologies (Rogawski, 2008, 2012; Zarcone and Corbetta, 2017; Shu et al., 2020). Although previously linked to epilepsy (Giddens et al., 2017), the new linkage to migraine in the patient population was both novel and significant for its strong sex bias. The latter may reflect the higher frequency of this diagnosis in the female portion in the DiscovEHR cohort and the population in general, but additional studies will be needed to resolve the connection between sex and GPR37L1 function. To identify the subset of coding variants driving disease associations, we performed bioinformatics and functional analyses to identify variant receptors with altered functional properties. Notably, some variant receptors were indistinguishable from the WT receptor. However, other variant receptors had significantly impacted properties, including those with MISS mutations in critical domains, that is, p.Arg69Trp in the extended N-terminal domain, p.Val361Met at the junction of ICL3 and helix 6, and p.Arg391Cys in the ECL3 loop. Altogether, these results highlight both the need and value for performing integrated computational, bioinformatics, and functional analyses to reveal novel players and pathways for common diseases with multigenic inheritance (Raraigh et al., 2018; Sun and Yu, 2019; Ishikawa et al., 2021).
Support for PSAP as the endogenous ligand
GPR37L1 has been suggested as the receptor for secreted PSAP (Meyer et al., 2013), a neuroprotective protein that has a complex role in lysosomal function (Carvelli et al., 2015). In the lysosome, PSAP is cleaved into four homologous domains, saposins A–D, with each exhibiting distinct regulatory roles in lipid metabolism (Hiraiwa et al., 1993; Leonova et al., 1996). Mutations in the SapC domain cause lysosomal storage diseases, including Gaucher's disease (Kang et al., 2018). However, controversy remains in the assignment of PSAP as the endogenous ligand of GPR37L1. In certain heterologous expression systems, GPR37L1 has been reported to act independently of the TX14(A) agonist (Ngo et al., 2021). However, in KO mouse models, the loss of Gpr37l1 eliminates prosaptide-evoked signaling in astrocytes (Jolly et al., 2018). Indeed, in SK cells expressing the native GPR37L1, we confirm that WTs respond to both TX14(A) and the native SapC module; KO cells do not respond to TX14(A); and transfection of KO cells with native GPR37L1 restores the ability of TX14(A) to inhibit cAMP production. Altogether, we have demonstrated that the GPR37L1 receptor can be activated by the SapC module of PSAP and the peptidomimetic TX14(A).
Dissection of functional differences among GPR37L1 variants
GPR37L1 variants were studied in the null background of GPR37L1-KO SK cells, which possess all the regulatory and signaling partners for GPR37L1 function. Previously, inhibition of FSK-induced cAMP production and stimulation of ERK1/2∼P were identified as signaling outputs of GPR37L1 (Meyer et al., 2013; Giddens et al., 2017). Using standard functional assays, we identified differential effects of the rare variants on the two pathways (Table 4). Of particular note, short exposures to TX14(A) induced upregulation of GPR37L1 expression. Although considered atypical for most GPCRs, recent studies has reported that a subset of receptors either maintain or increase signaling in the presence of an agonist, and this ability is crucial for invoking distinct cellular events (Rokosh et al., 1996; Breitwieser, 2012; Morton et al., 2015). Previously, we demonstrated a feed-forward mechanism for the calcium sensing receptor that increases trafficking from an endogenous pool to the cell surface (Grant et al., 2011; Breitwieser, 2012). Further experiments will be needed to sort out the exact mechanism for GPR37L1. Likewise, integrating this insight into models of astrocyte regulation will be critical to understanding the importance of GPR37L1 in the underlying pathologies of migraine and epilepsy.
Identifying cellular role(s) for GPR37L1
The astrocytic role(s) for GPR37L is being investigated. An agnostic RNAi screen for genes involved in cholesterol regulation identified PSAP as a regulator of perinuclear cholesterol levels in astrocytes (Bartz et al., 2009). Recent studies have shown that Gpr37l1-KO mice have altered cholesterol homeostasis in astrocytes, which contributes to dysregulation of Patched 1 internalization and trafficking (La Sala et al., 2020). In the current study, KO of GPR37L1 in SK cells has decreased total cellular cholesterol by ∼20%, which was restored upon transfection with WT GPR37L1, suggesting GPR37L1 is a significant contributor for the regulation of total cellular cholesterol. Whether native to the cells or re-expressed in KO SK cells, GPR37L1 increases cellular cholesterol, suggesting the possible presence of an endogenous agonist. PSAP is widely expressed at low levels in most cell types, with low levels of secretion that can be upregulated after stress or injury (Kunihiro et al., 2020). Further, acute stimulation of WT or variant GPR37L1s with TX14(A) differentially upregulated total cellular cholesterol, implicating GPR37L1 in the potential cholesterol cross talk between astrocytes and neurons.
Upregulation of cellular cholesterol levels in response to acute stimulation of GPCRs has not been broadly studied, although a recent report has suggested that NPY stimulates cholesterol synthesis via the SREBP2-HMGCR pathway in mouse hepatocytes (Chen et al., 2020). Specifically, significant upregulation of pathway proteins was observed within an hour, proteins remained elevated for up to 24 h, and cholesterol responses were sensitive to the MEK1/2 inhibitor U0126. Collectively, these results are consistent with GPR37l1 regulation of acute cholesterol changes via a MAPK pathway in SK cells, that is, the ability of GPR37L1 variants to increase ERK1/2 phosphorylation is correlated with upregulation of cell cholesterol in response to TX14(A), and the MEK1/2 inhibitor U0126 significantly attenuates acute cholesterol upregulation by TX14(A).
Assessing in vivo role(s) for GPR37L1
There are both common and rare forms of migraine, and human and mouse studies have begun to reveal the underlying mechanisms that account for the heterogeneity of this disorder (Gormley et al., 2016; Goadsby et al., 2017; Hansen et al., 2017; Sutherland et al., 2019). One specific mechanism involves dysregulation of glutamate homeostasis, including a protective role in astrocytes to counteract the effects of such dysregulation (Capuani et al., 2016; Conti and Pietrobon, 2023). Based on human clinical data, astrocyte localization, and functional studies, we posit a role for GPR37L1 signaling in astrocytes or glial cells. Consistently, dysfunction of brain astrocytes, cells that express the Gpr37l1 receptor, has been shown to trigger migraine-like symptoms in mice (Romanos et al., 2020). In mice, a single systemic administration of CGRP induces periorbital mechanical allodynia and other responses consistent with migraine including spontaneous pain, altered light sensitivity (photophobia), and anxiety-like behavior (Mason et al., 2017; Rea et al., 2018; Sturaro et al., 2023). We examined here the impact of genetic disruption of the Gpr37l1 receptor and found that its deletion did not affect the ability of an acute CGRP injection to elicit cephalic allodynia. Nevertheless, migraine is a complex disorder that extends beyond acute, episodic pain and involves functional and structural brain changes as well as increased responsiveness (sensitization) of central neurons responsible for pain (Brennan and Peitrobon, 2018; Mungoven et al., 2021). The impact of Gpr37l1 may be increased in chronic migraine, as it has been noted that Gpr37l1 is upregulated in microglia and astrocytes after nerve injury, which may be a compensatory mechanism (Kunihiro et al., 2020), and it is conceivable that the lack of GPR37L1 receptor may affect chronic migraine-like states more than episodic migraine evoked by acute challenge of migraine-provoking substances. Additionally, recent evidence, showing high GPR37L1 receptor expression in satellite glial cells of the human trigeminal ganglia (Yang et al., 2022), has suggested possible protection by GPR37L1 from peripheral sensitization mechanisms, which may also be initiated and maintained by the actions of CGRP. In future experiments, we will examine the role of the GPR37L1 receptor in chronic migraine models leading to sensitization of pain responses (Sturaro et al., 2023). Because comorbidities such as anxiety are frequent in chronic migraineurs and anxiety is a predisposing factor for migraine chronification in humans (Eigenbrodt et al., 2021), we also tested the WT and KO mice in the EPM. Intriguingly, female KO mice exhibited a sex-specific change in anxiety-like behavior, suggesting a possible linkage with chronic migraine.
In conclusion, the current studies have begun by using an unbiased approach to identify a strong association between GPR37L1 rare variants and two comorbid conditions, migraine and epilepsy. In vitro studies revealed that GPR37L1 variants exhibited altered signaling outputs, highlighting potential contributions to clinical phenotypes. Subsequent in vivo studies in KO mice simulating LOF showed a sex bias for females in an anxiety-related trait, though no effect in an acute migraine model. Computational methods in human databases demonstrated a clear connection between GPR37L1 and migraine. Although KO mice did not demonstrate an acute migraine phenotype, receptor KO could induce a chronic migraine phenotype, affect the dose of migraine initiator, or only affect certain types of migraine initiation. Future experiments examining these parameters and studying knock-in mice with humanized GPR37L1 mutants should lead to a better understanding of the role of GPR37L1 mutations in migraine and ultimately help patients.
Footnotes
Funding of this work was supported in part by NIH RO1GM111913 awarded to J.D.R. (MPI) and R01DA023281 to L.T.
We thank Kirk Jeffreys for helping with initial experiments, Dr. Diane Smelser for helpful discussions, and Geisinger for support in early stages of these studies including access to the MyCode clinical and genomic data. We acknowledge and appreciate the generosity of the Geisinger patients who agreed to allow their data to be used for research.
- Correspondence should be addressed to Janet Robishaw at robishawj{at}ufl.edu.
