Sphingosine 1-phosphate receptor 2 (S1P2) attenuates reactive oxygen species formation and inhibits cell death: implications for otoprotective therapy

Ototoxic drugs, such as platinum-based chemotherapeutics, often lead to permanent hearing loss through apoptosis of neuroepithelial hair cells and afferent neurons of the cochlea. There is no approved therapy for preventing or reversing this process. Our previous studies identified a G protein-coupled receptor (GPCR), S1P2, as a potential mediator of otoprotection. We therefore sought to identify a pharmacological approach to prevent cochlear degeneration via activation of S1P2. The cochleae of S1pr2−/− knockout mice were evaluated for accumulation of reactive oxygen species (ROS) with a nitro blue tetrazolium (NBT) assay. This showed that loss of S1P2 results in accumulation of ROS that precedes progressive cochlear degeneration as previously reported. These findings were supported by in vitro cell-based assays to evaluate cell viability, induction of apoptosis, and accumulation of ROS following activation of S1P2 in the presence of cisplatin. We show for the first time, that activation of S1P2 with a selective receptor agonist increases cell viability and reduces cisplatin-mediated cell death by reducing ROS. Cumulatively, these results suggest that S1P2 may serve as a therapeutic target for attenuating cisplatin-mediated ototoxicity.

S1P is a bioactive lipid signalling molecule that is known to act as a potent extracellular ligand for a family of five cognate GPCRs, S1P 1 -S1P 5 4 . These receptors have distinct but overlapping patterns of expression, and are known to be important activators of many cellular processes, such as cell proliferation, cell death, cytoskeletal rearrangement, migration/motility, and differentiation 5,6 . Notably, receptor-mediated S1P signalling has been shown to affect the production of ROS in the heart 7 , blood vessels 8 , fibroblasts 9 , and hematopoietic progenitor cells 10 .
Many of the biomedically relevant roles of S1P receptors have been elucidated with the study of genetically engineered knockout mice. These studies have shown that S1P signalling is essential for a number of processes including vascular maturation 11 , lymphocyte trafficking 12 , epithelial sheet migration 13 , B cell regulation 14 , egress of natural killer cells 15 , and mechanisms underlying the multiple sclerosis drug known as fingolimod (Gilenya) [16][17][18] . Recently, it was shown that S1P 2 knockout mice uniformly exhibit a progressive loss of inner ear function, resulting in profound deafness and vestibular dysfunction, demonstrating that S1P 2 activity is necessary for cochlear viability [19][20][21] that was a ligand-dependent process since loss of S1P transporter gene Spns2 phenocopies S1P 2 loss 22 . Cochlear degeneration was associated with early vascular defects that likely alter cochlear perfusion pressure and disrupt electrochemical gradients required for hair cell function 20 . While there is strong evidence that this mechanism contributes to loss of cochlear integrity, we note that S1P 2 is expressed in the hair cells and supporting cells of the cochlea, with expression increasing over time, coincident with the progression of the cochlear degeneration 19 . Therefore, we examined the possibility that additional, cell-intrinsic functions of S1P 2 promote viability of cochlear structures.
The recent approval of fingolimod, a non-selective functional antagonist of four S1P receptors, for the treatment of multiple sclerosis 23 demonstrates the feasibility of developing small molecule drugs that target S1P receptor signalling. Considering the pleotropic functions of the S1P receptor family, it would be a significant advantage to develop subtype-selective ligands as drug candidates. In this study, we demonstrate that activation of S1P 2 is associated with reduction of ROS accumulation by a specific S1P 2 agonist and provide proof-of-concept for its use as an otoprotective agent.

Results
Loss of S1P 2 results in ROS generation and cochlear degeneration. We previously reported that S1pr2 −/− knockout mice exhibit progressive degeneration of the sensory structures of the cochlea and vestibular end organs 19 . This is particularly evident in the spiral ganglia that innervate the organ of Corti. At 2 weeks of age, the neurons of the spiral ganglia are intact and indistinguishable from those of wild-type mice (Fig. 1A,B), but as previously described 19 , by 8 weeks of age there is marked degeneration of the ganglia at the basal turn of the cochlea, characterized by pronounced neuronal loss (Fig. 1C,D). To determine whether accumulation of ROS may contribute to this process, we evaluated wild-type and knockout cochlea for ROS content at 3 weeks of age prior to the onset of frank degeneration. Cochlea derived from S1pr2 −/− mice were characterized by the formation of dense blue labeling in the area of the spiral ganglia following incubation with NBT, indicating accumulation of ROS in the afferent nerve fibers. By contrast, specific labeling of the spiral ganglia was absent in S1pr2 +/− littermates ( Fig. 1E-H).
To explain how loss of S1pr2 may lead to ROS accumulation, we sought to determine whether S1P 2 could regulate the activity of NOX3 24,25 . Transfection of HEK293 cells with the NOX3 complex resulted in a marked increase in ROS generation (Fig. 1I). This could be significantly attenuated by co-transfection with S1P 2 , and further attenuated by activation of S1P 2 with S1P (1 μM). S1P-dependent reduction of ROS was not observed upon co-transfection with either of two other S1P receptor subtypes, S1P 1 and S1P 3 . NOX3 activity was also similarly inhibited by co-transfection with constitutively-active RhoA (RhoA-CA), which is known to act downstream of S1P 2 26 . The use of RhoA-CA provides a useful control condition in that it 1) confirms that S1P 2 -regulated pathways inhibit NOX3, and 2) demonstrates the maximum inhibitory response that would be expected from activation of the RhoA pathway.
CYM-5478 is a potent, selective agonist for S1P 2 . To determine whether activation of S1P 2 may be cytoprotective, we sought an S1P 2 -selective agonist. CYM-5478 was identified as a candidate S1P 2 agonist in a high throughput screen by The Scripps Research Institute Molecular Screening Center (http://pubchem.ncbi.nlm. nih.gov/assay/assay.cgi?aid= 872). This compound (PubChem CID: 7802604) had a reported EC 50 of 723 nM in the original screen, and an EC 50 of 780 nM in subsequent validation studies 27 . We performed several cell-based assays to confirm this result, and to evaluate the selectivity of CYM-5478. Use of a TGFα -shedding assay 28 demonstrated that CYM-5478 activates S1P 2 with an EC 50 of 119 nM, but had less than 25% efficacy and showed 10-fold lower potency against the other S1P receptor subtypes ( Fig. 2A-F). Control cells transfected with empty vector did not exhibit a measurable response when stimulated with either S1P or CYM-5478 (data not shown).
To confirm that the activation of S1P 2 was not an artefact of receptor overexpression, we evaluated the ability of CYM-5478 to activate endogenously expressed S1P 2 receptors in MDA-MB-231 breast cancer cells. Upon stimulation with S1P (1 μM), serum-starved MDA-MB-231 cells displayed pronounced cytoskeletal rearrangement characterized by process retraction and cell rounding (Fig. 2G,H). This response was completely abrogated by pre-treatment with S1P 2 -selective antagonist JTE-013 29 . CYM-5478 elicited an identical rounding response in MDA-MB-231 cells, which was similarly abrogated by JTE-013 (Fig. 2I).
Activation of S1P receptor-EGFP constructs results in their internalization into cytoplasmic vesicles that can be visualized by fluorescence microscopy 18 . We exploited this effect to further validate the use of CYM-5478 as an S1P 2 -specific agonist. Both S1P (1 μM) and CYM-5478 (1 μM) were able to induce the translocation of S1P 2 from the plasma membrane to cytoplasmic vesicles, but only S1P was effective against S1P 1 and S1P 3 (Fig. 2J-R). Hematoxylin and eosin staining of cochlear tissue sections demonstrates structurally intact neurons in the spiral ganglia of S1pr2 +/− (A) and S1pr2 −/− (B) mice at 2 weeks of age. By 8 weeks of age, S1pr2 +/− cochlea remain intact (C), but spiral ganglia of S1pr2 −/− mice (D) demonstrate marked, progressive degeneration. (Results previously described 19 ). The NBT assay demonstrates little staining in S1pr2 +/− cochlea (E), but a consistent banding pattern in S1pr2 −/− littermate cochlea (F), indicative of ROS accumulation. (G,H) Higher magnification reveals that the most intense staining is localized to the spiral ganglia. Images are representative of 7-9 mice per genotype. (I) In vitro assay for recombinant NOX3 activity reveals that S1P 2 , but not S1P 1 or S1P 3 , can inhibit NOX3 activity in a ligand dependent manner. Co-transfection with constitutively active Rho, a known downstream mediator of S1P 2 signalling, demonstrates a similar, but ligand-independent, inhibitory effect on NOX3. Scale bars represent 50 μm (A-D), 2 mm (E,F), and 1 mm (G,H), respectively. (*p < 0.05, **p < 0.01, ***p < 0.001).

CYM-5478 promotes viability and inhibits cell death in neural cells in vitro.
Since loss of S1P 2 results in progressive degeneration of sensory epithelial hair cells, supporting cells, and afferent neurons of the cochlea 19 , we sought to determine whether activation of S1P 2 could promote viability of a neural-derived cell line. C6 cells rat glioma cells 30 were evaluated by RT-PCR and shown to express S1P 2 as its predominant S1P receptor subtype (Fig. 3A). Under nutrient-deprivation stress produced by serum-starvation, CYM-5478 induced a statistically significant increase in the viability of C6 cells in a dose dependent manner at concentrations above 100 nM (Fig. 3B). This effect was absent in the presence of 10% fetal bovine serum (data not shown) suggesting that the increase in viability was a result of decreased starvation-induced cell death, rather than an increase in proliferation.
Since cisplatin is a known ototoxic compound that exerts its action, at least in part, by increased NOX3 activity 3 , we sought to determine whether activation of S1P 2 could protect cells from cisplatin-mediated death. In the presence of CYM-5478 (10 μM) there was a statistically significant, 3-fold increase in the EC 50 of cisplatin-mediated reduction in the viability of C6 glioma cells (Fig. 3C), consistent with pronounced cytoprotection produced by Scientific RepoRts | 6:24541 | DOI: 10.1038/srep24541 S1P 2 activation. Similarly, CYM-5478 treatment was able to abrogate cisplatin-induced cell death, as evaluated by propidium iodide dye exclusion assay (Fig. 3D). We further found that this effect was coincident with the reduction of apoptosis. Treatment with CYM-5478 also attenuated cisplatin-induced caspase 3/7 activity in C6 cells (Fig. 3E). Taken together, these data demonstrate a significant cytoprotective effect through the activation of S1P 2 by CYM-5478. S1P 2 activation inhibits the generation of ROS. To determine whether the cytoprotective effect of CYM-5478 was the result of decreased ROS production, we evaluated the sensitivity of C6 cells to cisplatin in the presence of a potent antioxidant, N-acetylcysteine 31 . When 1 mM N-acetylcysteine was added to culture media, there was no significant difference in the EC 50 of cisplatin toward vehicle-treated or CYM-5478-treated cells (Fig. 4A), thus implicating ROS in the previously observed effect (Fig. 3C). To confirm that CYM-5478 is inhibiting the production of endogenous ROS, and is not directly acting as an antioxidant, we evaluated the ability of CYM-5478 to protect C6 cells from exogenously administered ROS. C6 cells were equally sensitive to hydrogen peroxide in the presence and absence of CYM-5478 (Fig. 4B), indicating that CYM-5478 does not have significant antioxidant activity, consistent with its actions via S1P 2 . Using CellROX ® reagent, we confirm that there is an increase in ROS in C6 cells that are exposed to cisplatin (20 μM) for 24 hours (Fig. 4C), and that this increase can be significantly attenuated by co-administration of CYM-5478 (10 μM) (Fig. 4D). Treatment of C6 cells with either JTE-013 (S1P 2 antagonist) (Fig. 4E) or Y-27632 (Rho-associated protein kinase (ROCK) inhibitor) (Fig. 4F) had no effect on cisplatin-induced ROS, but resulted in a complete attenuation of CYM-5478 activity. This demonstrates that the protective effect of CYM-5478 is mediated by S1P 2 and activation of Rho signaling. Furthermore, we show that the effect of CYM-5478 can be phenocopied by treatment with either NSC23766 (Rac1 inhibitor) (Fig. 4G) or diphenyleneiodonium (DPI, NADPH oxidase inhibitor) (Fig. 4H). This implicates Rac1 and NOX as likely targets that are inhibited by CYM-5478.

Discussion
The present study sought to identify cell-intrinsic phenomena that contribute to the cochlear degeneration that occurs in S1pr2 −/− mice. We found that loss of S1P 2 results in the accumulation of ROS. The Nox family of NADPH oxidases are unusual in that they produce ROS as their primary function rather than as a byproduct 32 , and are thus a major source of signalling ROS. It is not surprising, then, that the activity of these enzymes requires strict control by a regulatory complex, with subunits that include known second-messengers such as Rac1. This provides a mechanism by which receptor-mediated signal transduction can limit ROS generation. Our current study demonstrates for the first time that S1P 2 activity inhibits NOX3, resulting in the reduction of cytotoxic ROS accumulation. It is likely that this occurs via activation of Rho 26 , and subsequent inhibition of Rac 33 , which is an obligate member of the NOX complex 2 .
Interestingly, several previous studies have found that S1P signalling can regulate NOX activity, but possible mechanisms are poorly characterized and often paradoxical. For example, while S1P signalling has been shown to increase ROS in fibroblasts 9 , vascular endothelial cells 34 , and isolated arteries 8 , S1P signalling has been shown to decrease ROS accumulation in vascular smooth muscle cells 35 . This apparent inconsistency in the literature is likely due to the heterogeneity of S1P signaling, much of which stems from differential receptor expression coupled with the fact that different S1P receptor subtypes activate different downstream signalling pathways. It is particularly notable that S1P 1 exclusively activates G αi 26 , which is a known activator of Rac1, whereas S1P 2 is a strong inducer of G α12/13 , which activates Rho 26 . Therefore, S1P 1 would be expected to activate NOX, whereas S1P 2 should inhibit it (Fig. 4I). Furthermore, it was recently demonstrated that S1P complexed with high-density lipoprotein (ApoM) acts as a biased agonist, inducing distinctly different S1P 1 -mediated effects compared to albumin-bound S1P 36 . This implies that ligand presentation can affect S1P signaling, and identifies an additional variable that may contribute to S1P otoprotective effects.
To underscore the translational relevance of this study, we have validated the activity of a recently identified S1P 2 agonist, CYM-5478. This compound (PubChem SID #46371153, CID #7802604) was identified in a high throughput screen for S1P 2 agonism and was used for SAR studies 27 , but was not systematically evaluated for receptor selectivity. In the current study, we provide the first demonstration that CYM-5478 is a potent and highly selective agonist for S1P 2 . In addition, we have used CYM-5478 to demonstrate that S1P 2 mediates pro-migratory responses in oral squamous cell carcinoma (Patmanathan, manuscript under review), further demonstrating that CYM-5478 is a valuable tool for identifying biological functions of S1P 2 .
Interestingly, a recent study reports the identification of an autosomal-recessive nonsyndromic hearing impairment (ARNSHI) locus that contains the S1P 2 gene (S1PR2) 37 . The authors further identify two mis-sense mutations that co-segregate with profound hearing loss in consanguineous families. This provides compelling evidence that the S1pr2 −/− mouse is a faithful model for the role of S1P signalling in hearing loss.
These combined results suggest that S1P 2 represents a potential drug target for the prevention of the ototoxicity caused by cisplatin. Despite the 75-100% occurrence of hearing loss associated with cisplatin therapy 38 , platinum-based chemotherapeutics remain first-line treatments for lung cancer and other tumor types. It is well-established that toxic accumulation of ROS is at least partly responsible for the effect of cisplatin on hearing loss 39 . Interestingly, the use of antioxidant therapy has shown some otoprotective potential, but has had limited success in actual clinical studies 40 . There are a number of issues that may be complicating this approach. For example, antioxidant therapy is non-selective, and typically has a protective effect on tumor cells, thus interfering with the desired effect of cisplatin treatment. This can be partially ameliorated by careful adjustment of the dosing schedule 39 , or by transtympanic administration of the antioxidant 41 . As a targeted potential therapy, administration of an S1P 2 agonist would be predicted to have increased selectivity for otoprotection, without known mechanisms for increasing tumor resistance to cisplatin. Furthermore, since S1P 2 agonists would initially act extracellularly via cell-surface receptors to minimize endogenous ROS production, they should have higher potency and more favorable pharmacokinetics than antioxidants that must enter the cell and act on ROS that have already accumulated.
Further validation of the value of S1P 2 as a therapeutic target was recently provided using an ex vivo culture model of gentamycin ototoxicity 42 . The authors demonstrated that administration of an antagonist for S1P 2 , but not S1P 1 and S1P 3 , results in increased gentamycin ototoxicity 43 . They further show that inhibition of sphingosine kinase potentiates cisplatin ototoxicity and itself promotes apoptosis and hair cell loss 44 , which confirms the importance of endogenous S1P signalling in cochlear integrity. Additionally, S1P signalling has been implicated in other cellular and physiological functions in the inner ear 45 . Notably, S1P signalling has been shown to regulate cochlear blood flow and affect the integrity of the stria vascularis 20,46 . Future in vivo studies are required to further characterize S1P 2 as a bona fide drug target for otoprotective therapy.
Animal Welfare and Ethical statement. Mice were housed in ventilated cages in the vivarium at The Scripps Research Institute (TSRI) on a 12 hour light/12 hour dark cycle, with ad libitum access to water and standard chow. S1pr2 −/− mice were generated and maintained as described 47 in a 129/SvJ, C57BL/6N mixed background. No procedures were performed on live animals for this study. All procedures were in compliance with state and federal regulations regarding animal welfare, followed the ARRIVE guidelines of the National Centre for the Replacement Refinement and Reduction of Animals in Research, and were performed as humanely as possible. All procedures were approved the Institutional Animal Care and Use Committee at TSRI and complied with the US National Research Council's "Guide for the Care and Use of Laboratory Animals, " and the US Public Health Service's "Policy on Humane Care and Use of Laboratory Animals" and "Guide for the Care and Use of Laboratory Animals".
Nitro blue tetrazolium (NBT) assay. Accumulation of ROS ex vivo was determined using a modification of a NBT assay previously described for visualization of ROS in cultured cells in situ 48 , in which the presence of ·O 2 − is detected by the conversion of NBT to a blue, insoluble formazan precipitate 49 . Cochlea were rapidly isolated and exposed by perforating the boney wall, washed in HBSS, incubated with NBT (1.6 mg/ml) in HBSS at 37 °C for precisely 45 min, then fixed in 100% methanol and photographed with a stereo microscope equipped with a Nikon Coolpix 950 camera.
TGFα-shedding assay. The TGFα -shedding assay was performed essentially as described 28 . Briefly, HEK293 cells were co-transfected with the indicated receptor expression construct and TGFα -alkaline phosphatase using Lipofectamine 2000 reagent (cat #11668019, Thermo Fisher), collected by trypsinization, washed with phosphate-buffered saline, and seeded into 96-well plates in Hank's Balanced Saline Solution (HBSS). To improve assay sensitivity, S1P 1 and S1P 2 cells were co-transfected with G αq/i1 chimeric protein, and S1P 4 and S1P 5 were co-transfected with G αq/16 , as previously described 28 . Cells were stimulated with ligand for 1 hour, then alkaline phosphatase activity was detected in cells and in the supernatant. Receptor activity (% shedding) was defined as alkaline phosphatase activity of the supernatant/total alkaline phosphatase activity (cells + supernatant). Data processing and statistical analyses were performed with GraphPad Prism 6.
Cell rounding assay. Cell rounding was performed essentially as described 51 . MDA-MB-231 breast cancer cells were seeded on collagen-coated coverslips at 30-50% confluence, incubated overnight, and serum-starved for 4 hours. They were then pretreated with JTE-013 or vehicle for 15 minutes, and treated with S1P, CYM-5478, or vehicle. After 15 min, the cells were fixed and stained with fluorescein phalloidin and DAPI for cell morphology. The number of cells with retracted neurites and the number of total cells were counted in three separate fields for each sample, and the percentage of cells with retracted neurites was calculated. Reverse transcriptase polymerase chain reaction (RT-PCR). Total RNA was isolated from C6 glioma cells using Trizol reagent (Life Technologies) per manufacturer's instructions. Approximately 2 μg of each sample was primed with oligo-dT and random hexamer primers prior using Thermo Scientific Maxima First Strand cDNA synthesis kit (Life Technologies). For quantitative real-time RT-PCR, targets were amplified with Maxima SYBR Green/ROX qPCR Master Mix (Life Technologies) on an Applied Biosystems ViiA 7 Real-Time PCR System (Life Technologies) using gene-specific primer pairs (Table 1). Quantitation was determined with a standard curve analysis as described 52 . Viability assay. Cell viability was determined by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay as described 13 . Briefly, C6 cells were seeded into 96-well plates at 20,000 cells/well, incubated overnight, serum-starved for 4 h, and treated for 72 hours with cisplatin in the presence of CYM-5478 or vehicle.
Cell death assay. Cell death was evaluated by propidium iodide exclusion assay. Cells were plated in 96-well plates at ~50% confluence, incubated overnight, then treated with cisplatin (67 μM) for 24 hours in the presence of vehicle or 10 μM CYM-5478. Cell were treated with Hoechst 33342 (1 mg/ml) and propidium iodide (0.3 mg/ml) for 20 minutes, washed with phosphate-buffered saline, then rapidly photographed with a Cytation 3 automated imager (Biotek Instruments, Inc.). % cell death reflects a ratio of propidium iodide-positive and Hoechst 33342-positive cells.
Caspase assay. Cells were plated in clear-bottom, black-walled 96-well plates at ~50% confluence, incubated overnight, then treated overnight with cisplatin in the presence of vehicle or 10 μM CYM-5478. Caspase activation was evaluated by Caspase-Glo 3/7 Assay System (Promega Corporation) per manufacturer's instructions. ROS assay. Cells were plated in clear-bottom, black-walled 96-well plates at ~50% confluence, incubated overnight, then treated overnight with cisplatin in the presence of the indicated compounds. All conditions were vehicle-controlled (0.1% DMSO). Cells were treated with CellROX ® Orange reagent (ThermoFisher #C10443) per manufacturer's protocol and counter stained with Hoechst 33342 (ThermoFisher #62249). Cells were photographed with a 20X objective lens, 3 fields per well, 3 wells per condition. Quantification was performed using ImageJ software, but dividing the total integrated density of CellROX ® labelling by the number of cells per Hoechst staining.