Individual variation of human S1P₁ coding sequence leads to heterogeneity in receptor function and drug interactions.

Sphingosine 1-phosphate receptor 1 (S1P₁), an abundantly-expressed G protein-coupled receptor which regulates key vascular and immune responses, is a therapeutic target in autoimmune diseases. Fingolimod/Gilenya (FTY720), an oral medication for relapsing-remitting multiple sclerosis, targets S1P₁ receptors on immune and neural cells to suppress neuroinflammation. However, suppression of endothelial S1P₁ receptors is associated with cardiac and vascular adverse effects. Here we report the genetic variations of the S1P₁ coding region from exon sequencing of >12,000 individuals and their functional consequences. We conducted functional analyses of 14 nonsynonymous single nucleotide polymorphisms (SNPs) of the S1PR1 gene. One SNP mutant (Arg¹²⁰ to Pro) failed to transmit sphingosine 1-phosphate (S1P)-induced intracellular signals such as calcium increase and activation of p44/42 MAPK and Akt. Two other mutants (Ile⁴⁵ to Thr and Gly³⁰⁵ to Cys) showed normal intracellular signals but impaired S1P-induced endocytosis, which made the receptor resistant to FTY720-induced degradation. Another SNP mutant (Arg¹³ to Gly) demonstrated protection from coronary artery disease in a high cardiovascular risk population. Individuals with this mutation showed a significantly lower percentage of multi-vessel coronary obstruction in a risk factor-matched case-control study. This study suggests that individual genetic variations of S1P₁ can influence receptor function and, therefore, infer differential disease risks and interaction with S1P₁-targeted therapeutics.


Data collection from NHLBI GO ESP Exome Variant Server
Data for genetic variations of the S1PR1 gene was collected from NHLBI GO ESP Exome Variant Server ( 32 ) according to the guidelines of the data usage . The data set was not linked to clinical phenotypes.

Site-directed mutagenesis
cDNAs for S1P 1 mutants were generated using a QuikChange site-directed mutagenesis kit (Agilent). Primer information used for the mutagenesis is available in the supplementary material.

Lentivirus vectors and cell transduction
cDNAs for green fl uorescent protein (GFP)-tagged WT and mutant S1P 1 were cloned in a pCDH-CMV-MCS-EF1-Puro vector (System Biosciences). A lentivirus vector for shRNA-mediated S1P 1 suppression was purchased from Sigma-Aldrich (the RNA Interference Consortium clone TRCN0000011359). Pseudoviral particles were made by transfecting these vectors together with packaging (pMDLg/pRRE and pRSV-Rev) and envelope (pMD2.G) plasmids into 293T producer cells. The culture supernatant containing pseudoviral particles was recovered and used to transduce cells. Transduced cells were selected by treatment with puromycin overnight.

Receptor endocytosis assay
Endocytosis of S1P 1 upon stimulation with S1P or FTY720-P was observed as described ( 33 ).

Cell-surface protein biotinylation assay
Cells were treated with a cell nonpermeable biotinylation reagent, EZ-link Sulfo-NHS-SS-Biotin (Thermo Scientifi c), and lysed in the lysis buffer. Biotinylated proteins were purifi ed by using NeutrAvidin agarose beads (Thermo Scientifi c).

Measurement of intracellular calcium concentration
Changes in intracellular calcium concentrations upon ligand stimulation were monitored with a FLEX-station scanning fl uorometer system (Molecular Devices) as described ( 34 ). recent crystal structure determination of S1P 1 has also enhanced our understanding of ligand interactions with the receptor ( 17,18 ). Upon S1P binding, S1P 1 activates heterotrimeric G ␣ i protein and transmits signals via the p44/42 MAPK, PI3K/Akt, and Rac pathways (19)(20)(21). Subsequently, S1P 1 gets phosphorylated by GRK2 protein in the carboxyl terminal region ( 22 ) and undergoes clathrin-dependent endocytosis, then recycles back to the plasma membrane ( 23 ). This endocytosis/ recycling pathway is essential for the regulation of S1P 1 activity. Impairment of the receptor endocytosis results in enhanced signaling and hyper-immune consequences, whereas vascular endothelial function is enhanced (24)(25)(26).
The study of the S1P-S1P 1 signaling system has entered the therapeutic era with the successful launch of fi ngolimod/ Gilenya (FTY720) in the treatment of relapsing-remitting multiple sclerosis ( 27,28 ). Indeed, many novel agonists and antagonists for S1P 1 have been developed and some of them are undergoing clinical trials to test their efficacy in the control of many diseases, including relapsingremitting multiple sclerosis, the secondary progressive form of multiple sclerosis, psoriasis, and amyotrophic lateral sclerosis ( 29 ). Among them, FTY720, which gets phosphorylated into FTY720-P by sphingosine kinase-2 ( 10 ), induces irreversible ubiquitin-dependent degradation of S1P 1 and blocks lymphocyte egress from secondary lymphoid organs ( 25,30 ). Genetic variations represented by single nucleotide polymorphisms (SNPs) confer individual variation in terms of susceptibility to certain diseases, development and severity of diseases, and effi cacy and adverse effects of therapeutic drugs. However, our knowledge is quite limited on the genetic variations of S1P 1 and their effects on the receptor functions and disease development ( 31 ). Individual genetic variations of the S1PR1 gene in various disease cohorts and their functional consequences may lead to a better understanding of S1P 1 biology in disease outcomes and in targeted therapeutics.
The National Heart, Lung, and Blood Institute Grand Opportunity Exome Sequencing Project (NHLBI GO ESP) sequenced protein-coding regions of the human genome across diverse, richly phenotyped populations with the ultimate aim to discover novel genes and mechanisms contributing to heart, lung, and blood disorders ( 32 ). The project has identifi ed 14 nonsynonymous SNPs in the S1PR1 gene out of more than 10,000 patients across many different clinical trials, with variations of the SNP prevalence from 1 to 0.001%. In this study, we determined the functional consequences of such nonsynonymous S1P 1 mutations. We also conducted a sequencing analysis of the S1PR1 gene in a cardiovascular disease cohort with more than 1,800 patients in which we had access to clinical data including coronary angiograms. We report the identifi cation of one loss-of-function mutation and two mutations that result in impaired receptor endocytosis. We also report one mutation that is associated with a signifi cantly higher incidence in the cardiovascular disease cohort.

Collection of patient samples and sequencing analysis of the S1PR1 gene
Patients were recruited from Dartmouth-Hitchcock Medical Center under a protocol approved by the Institutional Review Board ( 35 ). Genomic DNA was extracted from blood using a Pure-geneH system (Gentra Systems). The coding sequence of the human S1PR1 gene was amplifi ed by PCR using the human genomic DNAs as templates and the following primers: 5 ′ -tctccagccaaggaaaagc-3 ′ and 5 ′ -agtaaagagcgcttccgg-3 ′ . The amplifi ed DNAs were purifi ed using a QIAquick PCR purifi cation kit (Qiagen) and subjected to Sanger DNA sequencing analyses at the Genomics Facility of the Biotechnology Resource Center, Cornell University.

Human HDL preparation
Lipoprotein particles with the density ranging from 1.066 to 1.21 were collected from human plasma by repeated ultracentrifugation as described ( 36 ), and dialyzed against PBS. S1P concentration contained in the HDL fraction was determined by a

Measurement of trans -endothelial electric resistance
HUVECs were seeded onto gelatin-coated gold electrodes (8W10E plates, Applied BioPhysics), serum starved in EBM2 medium (Lonza) for 1 h, and stimulated with S1P. Endothelial barrier function was assessed by monitoring electric resistance of the electrodes with an ECIS system (Applied BioPhysics).

Genetic variations of the S1PR1 gene found in the NHLBI GO ESP database
To examine the genetic variations of the S1PR1 gene, we performed a database search in the NHLBI GO ESP database ( 32 ). As of June 2012, we found 33 SNPs in the S1PR1 gene, consisting of 14 nonsynonymous mutations ( Table 1 ), 14 synonymous mutations, 4 mutations in the 5 ′ -untranslated region (UTR), and 1 mutation in the 3 ′ -UTR. The positions of the amino acids changed in the 14 nonsynonymous mutations are illustrated in the secondary structure of S1P 1 ( Fig.  1 ). Those include Ala 11 to Asp (A11D), R13G, A39T, I45T, P79S, R120P, Y198C, R238C, S245N, G305C, P332R, R342P, G347S, and S351G. Among them, A11D and R13G show relatively higher prevalence (>1%) in the African American and European American populations, respectively ( Table 1 ).

Functional analysis of the S1P 1 SNP mutants
We expressed S1P 1 harboring each mutation in CHO-K1 cells, and stimulated the cells with the ligand S1P to test processes ( 23 ). All of the mutants showed cell surface localization after serum starvation in the same way as WT S1P 1 ( Fig. 2B , supplementary Fig. I), which indicates that none of the mutations affect receptor translation and traffi cking to the plasma membrane in the basal condition. Upon S1P stimulation, R120P mutant showed no receptor endocytosis ( Fig. 2B ), which confi rms that activation of the receptor by S1P is impaired in the R120P mutant. Surprisingly, I45T showed diminished endocytosis and G305C failed to undergo endocytosis upon S1P stimulation ( Fig. 2B ). All other mutants responded to S1P and endocytosed in an indistinguishable manner from the WT (supplementary Fig. I).
We further explored the time course and ligand-dose dependency of p44/42 and Akt activation in CHO-K1 cells the ability of the SNP mutations to infl uence the receptormediated activation of p44/42 MAPK and Akt signaling proteins. As shown in Fig. 2A , R120P mutant failed to mediate the activation of p44/42 MAPK and Akt in response to 100 nM S1P, while the other mutants did not show any change compared with WT S1P 1 . Arg 120 is a critical residue for the binding of S1P through the interaction of its basic side chain with the phosphate group of S1P ( 17,38 ). S1P binding ability of the receptor was largely impaired by R120A mutation ( 38 ). Thus, the defect of the R120P mutant in S1P-induced signaling is likely due to impaired S1P binding ability.
We also examined ligand-induced receptor endocytosis that is an important step for the receptor desensitization   for 1 h (B). A: Total cell lysates were prepared and subjected to Western blot analyses for the degradation of S1P 1 . B: Cell-surface proteins were biotinylated, purifi ed, and subjected to Western blot analyses to check the amount of S1P 1 remaining on the cell surface after FTY720-P treatment. C: Densitometric analysis of (B) using the National Institutes of Health ImageJ program is shown. Data represent the mean ± SD (n = 3). overexpressing WT S1P 1 , I45T, R120P, and G305C mutants. We confi rmed a similar level of expression of each receptor (supplementary Fig. II). As shown in Fig. 3A , I45T did not show any differences compared with WT S1P 1 both in the time course and dose dependency. G305C had overall similar, but slightly enhanced, signaling compared with WT S1P 1 , which could be due to the defect in endocytosis-mediated desensitization processes ( Fig. 2B ). A higher dose of S1P (>100 nM) did induce modest activation of p44/42 MAPK in cells expressing the R120P mutant. This indicates that the R120P mutant is not completely insensitive to S1P, suggesting that several other residues, such as Tyr 29 , Lys 34 , Asn 101 , and Glu 121 , also contribute to the recognition of the head group of S1P ( 17 ). We also measured the increase in intracellular calcium concentration that is a proximal downstream event of the receptor activation. I45T and G305C mutants did not show any differences from WT S1P 1 , but the R120P mutant failed to mediate the calcium response ( Fig. 3B ). These results show that I45T and G305C mutants maintain the normal properties in the receptor activation but the receptor desensitization processes are impaired, while the R120P mutant shows largely impaired receptor activation.

Diminished effi cacy of FTY720 on I45T and G305C mutants
Endocytosis of S1P 1 plays an important role in the receptor desensitization. It is a crucial process in the regulation of lymphocyte traffi cking. In circulating lymphocytes, S1P 1 is internalized in response to high concentrations of S1P in blood and lymph. In secondary lymphoid organs, S1P concentrations are maintained low, thus S1P 1 recovers its cell surface localization ( 3 ). This is the basis for enabling lymphocytes to egress from lymphoid organs toward circulation along the gradient of S1P concentration. FTY720 inhibits the egress of lymphocytes from secondary lymphoid organs by inducing internalization and irreversible S1P 1 degradation in the proteasome ( 25,30 ). Recently, our group showed that mutant mice harboring the S1PR1 gene encoding phosphorylation-defi cient, and as such endocytosis-defi cient, receptors [S1P 1 (S5A)] developed severe experimental autoimmune encephalomyelitis due to T helper 17 cell-mediated autoimmunity, both in the peripheral immune system and the nervous system ( 26 ). This S1P 1 (S5A) receptor shows resistance to FTY720induced degradation both in vivo and in vitro ( 25,30 ).
The defect in the endocytosis response found in I45T and G305C mutants phenocopies the S1P 1 (S5A) receptor and may also lead to the resistance to FTY720-induced endocytosis and degradation. We examined the effi cacy of FTY720-P, a phosphorylated form of FTY720, on I45T and G305C mutants, and found that these mutants showed much weaker endocytosis in response to FTY720-P treatment compared with the strong internalization of WT S1P 1 (supplementary Fig. III). As a consequence of diminished endocytosis, I45T and G305C mutants showed much less FTY720-induced degradation than WT S1P 1 ( Fig. 4A ). To confi rm quantitatively that the endocytosis response is impaired in I45T and G305C mutants, we measured cell surface S1P 1 before and after FTY720-P treatment. As shown in Fig. 4B, C , WT S1P 1 showed a 56% decrease in the cell surface amount after 1 h of FTY720-P treatment, while I45T and G305C showed only 16 and 13% decreases, respectively. These results suggest that the effi cacy of FTY720 treatment could be impaired in individuals who carry these genetic variations in the S1PR1 gene.

Sequencing analysis and phenotype correlation of the S1PR1 gene in a cardiovascular cohort
The S1P-S1P 1 signaling system also plays an important role in vascular development and homeostasis. Dysfunction of this system leads to increased vascular permeability and hypersensitivity to anaphylactic stimulus ( 2,(9)(10)(11). Antiinfl ammatory effects of HDL on the endothelium are partially attributed to its cargo S1P (12)(13)(14)(15). Moreover, one of the adverse effects of FTY720 in therapeutics for multiple sclerosis is bradycardia, which is associated with the agonistic effects of FTY720 on S1P 1 and subsequent activation of the G protein-coupled inwardly rectifying potassium channel in cardiomyocytes ( 39 ). Therefore, it is of clinical importance to examine the prevalence of nonsynonymous mutations in the S1PR1 gene in a cardiovascular disease cohort and its relationship with disease severity or with sensitivity to FTY720. Toward this end, we performed a sequencing analysis of the S1PR1 gene in 1,814 patients in a high cardiovascular risk population ( 35 ). As summarized in Table 2 , we found fi ve nonsynonymous mutations that resulted in amino acid changes of A11D, R13G, V20I, R238C, and S333R. The mutations for V20I and S333R have not been reported before. R13G showed a signifi cantly higher incidence than that observed in the NHLBI GO ESP (3.58% versus 1.24%). The other detected mutations were rare (0.0055%). Because coronary angiograms (for detection of coronary obstruction) were available on all the patients, we compared the R13G patients (n = 59) to age-and risk factor-matched control patients (n = 236 in a 4:1 match). R13G appeared to protect from coronary artery disease (CAD); signifi cantly more patients had no CAD and signifi cantly fewer patients had multiple vessel obstruction compared with the control patients ( Table 3 , Fig. 5 ).

Functional analysis of the R13G mutation of S1P 1 in endothelial cells
R13G mutation was enriched in the high cardiovascular risk population, but appeared to protect from CAD. To understand this apparent discrepancy, we examined functional differences of the R13G mutant in endothelial cells. GFPtagged WT S1P 1 or R13G mutant were overexpressed in HUVECs with endogenous S1P 1 knocked-down by specifi c shRNA against 3 ′ -UTR of S1P 1 ( Fig. 6A ). These cells were stimulated with an increasing dose of HDL prepared from human plasma (S1P concentration from 1 to 100 nM) to check activation of p44/42 MAPK and Akt. As shown in Fig. 6B, C , R13G mutant did not show any signifi cant differences from WT S1P 1 . We also measured changes in trans -endothelial electric resistance (TEER) upon S1P stimulation in these HUVECs, because one of the important functions of S1P 1 in endothelial cells is strengthening adherens junctions to maintain vascular barrier integrity ( 9,40 ). HDL-bound S1P increased the TEER in a dosedependent manner, and the R13G mutant did not show any difference from WT S1P 1 ( Fig. 6D , upper). The same tendency was observed when these cells were stimulated with BSA-bound S1P ( Fig. 6D , lower). Thus, we did not fi nd detectable difference of the R13G mutant in endothelial cells, at least in the assay systems we performed, though the cohort analysis of cardiovascular patients showed that patients with the R13G mutant had a signifi cantly lower incidence of multi-vessel obstructions. We also examined the effects of I45T and G305C mutations on the S1P-induced increases in TEER, but did not fi nd any differences compared with the WT receptor (supplementary Fig. IV).   a Age, gender and risk factors were matched between the case (R13G mutant, n = 59) and the control (WT S1P 1 , n = 236).
b Numbers of patients with no CAD and with multivessel disease (obstruction in >1 major coronary artery) are shown.

Fig. 5.
Lower percentage of multi-vessel obstructions in the patients with R13G mutation. A: A pie chart shows the percentages of the patients with zero or one (black) or with two or more (gray) obstructed vessels observed in coronary angiography in the casecontrol study between the R13G mutant and WT S1P 1 . B: Distribution of the obstructed vessel number is shown.

DISCUSSION
In this study, we characterized 14 nonsynonymous mutations of S1P 1 found in the NHLBI GO ESP database. We found three mutations that showed functional differences; R120P showed a defect in both ligand-induced activation and endocytosis, while I45T and G305C only showed a defect in ligand-induced endocytosis. These three residues are all located in the transmembrane helixes in sharp contrast to the other 11 mutations in the extra-or intracellular region ( Fig. 1 , and Fig. 7 ), which suggests that nonsynonymous mutations in transmembrane helixes of GPCRs tend to result in functional alterations. R120P is a loss-of-function mutation and can be a risk factor for various diseases, considering a pivotal role of the S1P-S1P 1 signaling system in cardiovascular and immune systems. Because the S1P 1 knockout mouse is embryonic lethal at around E12.5 due to defects in vascular development ( 41,42 ), homozygotes for the R120P mutation are most likely embryonic lethal. Effects of heterozygous loss-offunction of S1P 1 should be explored in mouse disease models and/or in various cohort studies.
I45T and G305C showed attenuated ligand-induced endocytosis/degradation, though they mediate intracellular signaling normally. I45T and G305C mutants phenocopy a S1P 1 (S5A) mutant in which fi ve serine residues in the carboxyl-terminal region are substituted with alanine. Recently, our group reported that S1P 1 (S5A) mutant mice showed enhanced autoimmunity and developed severe experimental autoimmune encephalomyelitis ( 26 ). Thus, impairment of the receptor internalization in I45T and G305C mutants may result in increased autoimmunity as observed in S1P 1 (S5A) mice, raising a possibility that these SNPs can be potential risk factors for autoimmune diseases. In multiple sclerosis patients, these two SNPs would hamper the ability of FTY720 treatment to induce degradation of S1P 1 . It is of clinical importance to explore the prevalence of I45T and G305C mutations and their involvement in disease development in several autoimmune disease cohorts.
In contrast to the established importance of Arg 120 in S1P binding ( 17,38 ), no clear function has been attributed to Ile 45 and Gly 305 so far. Substitution of Ile 45 to Thr results in an amino acid sequence of Asn 43 Ser 44 Thr 45 , which makes a potential N-linked glycosylation site (N-X-S/T). N-linked glycosylation of transmembrane proteins, including GPCRs, occurs in the endoplasmic reticulum and plays roles in proper folding and traffi cking to the cell surface ( 43 ). Most GPCRs have one or more N-linked glycosylation sites in the extracellular domain ( 44 ), though the functions of N-linked glycosylation have not been wellclarifi ed. The I45T mutant of S1P 1 showed normal localization on the cell surface and ligand-induced activation of intracellular signaling pathways, which means this mutation does not affect proper folding of the receptor and traffi cking to the cell surface. In mouse S1P 1 , which has one N-linked glycosylation site in the N-terminal region, mutation of this Asn resulted in poor association with caveolae and impaired ligand-induced endocytosis ( 45 ). A similar defect in ligand-induced endocytosis by a nonglycosylated mutant was observed in protease-activated receptor 1 ( 46 ). Thus, glycosylation of the extracellular region of GPCRs is somehow involved in the regulation of ligandinduced endocytosis in some cases. Human S1P 1 has two putative N-linked glycosylation sites (Asp 30 and Asp 36 ) in the close vicinity of Asp 43 , though glycosylation of these residues has not been confi rmed. Additional glycosylation of Asp 43 might interfere with functions of these putative glycosylated residues. Because Ile 45 is located in the beginning of transmembrane helix I and faces the ligand-binding pocket, as illustrated in Figs. 1 and 7 , another possibility is that the I45T mutation and/or glycosylation of Asp 43 interferes with ligand-induced conformational changes of the receptor necessary for internalization responses.
G305C shows an even more severe defect in the endocytosis than I45T. Gly 305 is located in the middle of transmembrane helix VII ( Figs. 1, 7 ). A stretch of polar amino acids with a glycine in the middle (Asn 303 Ser  304 Gly  305 Thr   306 Asn 307 ) precedes a highly conserved NPXXY motif and makes the helix VII loose, allowing some fl exibility of this transmembrane region ( Fig. 7 ). Glycine is energetically Fig. 6. Functional comparison between WT S1P 1 and R13G mutant in HUVECs. HUVECs were transduced with lentivirus vectors for the expression of GFP-tagged S1P 1 or R13G mutant together with a vector for endogenous S1P 1 -specifc shRNA. A: Western blot analysis was performed to check the overexpression of WT S1P 1 and R13G and the suppression of endogenous S1P 1 by shRNA. B: The cells were starved for 3 h and stimulated with the indicated concentrations of HDL-bound S1P for 5 min. Total cell lysates were prepared and subjected to Western blot analyses for the activation of p44/42 MAPK and Akt. C: Densitometric analysis of (B) by the National Institutes of Health ImageJ program is shown. Data represent the mean + SD (n = 3). D: The cells were starved for 3 h and stimulated with the indicated concentrations of HDLbound (upper panel) or BSA-bound (lower panel) S1P. The changes in TEER were monitored in an ECIS system (Applied Biophysics). Data are the mean ± SEM (n = 3) and are representative of three independent experiments with the same tendency . unfavorable for the stability of ␣ -helixes and has the second lowest helix propensity next to proline ( 47 ). Gly 305 to Cys mutation may constrain this fl exible loop and interfere with ligand-induced conformational changes required for endocytosis. Another possibility is that Cys 305 might create an additional disulfi de bond between another Cys residue. There are four cysteine residues within the transmembrane helixes of S1P 1 [Cys 56 , Cys 57 (helix I), Cys 206 (helix V), and Cys 268 (helix VI)]. These cysteine residues could be potential intra-molecule bonding partners. It is also possible that Cys 305 could make an inter-molecule disulfi de bond with a cysteine residue of a transmembrane domain of another molecule, because many GPCRs are shown to homo/hetero-dimerize ( 48 ).
In addition to the functional analysis of 14 nonsynonymous mutations of S1P 1 found in the NHLBI GO ESP database, we performed a sequencing analysis of the S1PR1 gene in the cardiovascular disease cohort and found that R13G shows a signifi cantly higher incidence than that observed in the NHLBI GO ESP (3.58% versus 1.24%). Moreover, the number of patients with multi-vessel obstructions was signifi cantly lower in the R13G group in comparison with the age-, gender-, and risk factor-matched control group, which suggests that the R13G mutation might reduce the severity of coronary atherosclerotic obstructions. Although we tried to clarify functional differences of the R13G mutant in an in vitro study using HUVECs, we did not fi nd any differences compared with WT S1P 1 , at least in the assay systems we performed. However, atherosclerotic arterial obstructions develop through complex interactions between endothelial cells, smooth muscle cells, macrophages, and lymphocytes. Therefore, it is possible that the R13G mutant might be functionally different in some of these cell types or might affect interaction between these cell types. To establish physiological/clinical signifi cance of the R13G mutation, further case-control studies should be conducted in a cardiovascular disease cohort, as well as other disease cohorts. In vitro characterization of the R13G mutant in various cell types will also be necessary.
In summary, fi ndings in this study suggest that individual variations in S1P 1 may infl uence cardiovascular/immune functions as well as drug sensitivity and, therefore, infer differential risks for cardiovascular and autoimmune diseases as well as response to S1P receptor modulatory drugs. Suffi ciently powered prospective cohort studies should be conducted to validate possible relationships between individual variations in the S1P 1 receptor and specifi c diseases/drug sensitivity.