Experiments were performed according to National Austrian legal requirements and the recommendations of the IASP to minimize animal suffering (Zimmermann, 1983) with permission of the Austrian Bundesministerium für Wissenschaft, Forschung und Wirtschaft (BMWF-66.011/0102-WF/V/3b/2015).

Male wild-type (wt), S1P₃^−/−, SNS-S1P₁^−/− and S1P₁^fl/fl mice (age between 8 and 12 weeks) were used in all experiments. The mutated alleles were back crossed to C57BL/6J for at least 12 generations and littermates were used for controls (Mair et al., 2011). Animals were housed on a 12 h light/dark cycle and had free access to chow and water.

For induction of nerve crush injury mice were briefly anesthesizsed with Ketasol (Graeub, Innsbruck, Austria) and Xylasol (AniMedica) through intraperitoneal injection. After cutaneous incision, the sciatic nerve was exposed and crushed at mid-thigh level for 1 min by applying a standard force of 19 ± 1.8 N with a semi-electric feedback-controlled forceps as previously published (modified Bioseb Rodent pincher RP-1, Chaville Cedex, France; Quarta et al., 2014).

Animals were maintained in individual cages for the entire duration of the experimental paradigm of 15 days post lesion (dpl). Following nerve lesion, heat sensitivity was quantified by standard testing procedures (Andratsch et al., 2009; Quarta et al., 2014). Measurements were performed two times before injury and repeated at 1, 4, 8, 11, 15 dpl. Mice were allowed to habituate for 1 h before testing. Heat sensitivity was assessed using the Hargreaves test as previously published (Hargreaves et al., 1988), where the withdrawal response latency to an increasing heat stimulus (I.R. intensity = 51) was obtained with an automated algesiometer (Ugo Basile, Comerio, Italy).

Adult mouse dorsal root ganglia (DRG) were dissected as previously published (Obreja et al., 2002; Agarwal et al., 2007). After dissection, ganglia were incubated in Liberase Blendzyme 1 (9 mg/100 ml DMEM, Roche, Vienna, Austria) for 60 min. After washing with PBS, 1× Trypsin-EDTA (Invitrogen, Vienna, Austria) was added for 15 min. Ganglia were washed with supplemented TNB™ medium (Biochrom, Berlin, Germany) containing L-glutamine (Invitrogen), streptomycin sulfate (Invitrogen), penicillin G sodium and Protein-Lipid-Komplex (Biochrom). The cell suspension obtained after mechanical dissociation was centrifuged at 500 rpm through a 3.5% BSA gradient (Sigma Aldrich, Vienna, Austria) for 10 min to remove cell debris. The pellet was resuspended in TNB™ medium and centrifuged for 5 min at 760 rpm. The neurons were plated on cover slips coated with poly-l-lysine (Sigma) and a top layer of laminin (10 μg/ml, Sigma). Neurons were cultured in TNB™ medium at 37°C in a 5% CO₂ atmosphere for 24 h.

After 20 h, culture dishes were transferred to Axio Imager microscope (Carl Zeiss) with 20×, NA 0.30 objective. Images were recorded in phase contrast with a cooled CCD camera (SPOT; Diagnostic Instruments). Neurons were recorded at room temperature (RT) for 60 min with or without S1P or SEW2871 or FTY720-P at a concentration of 1 μM diluted in extracellular solution (ECS). Images at 0, 30 and 60 min were processed and compared as overlapped image for growth cone collapse and neurite retraction signs using MetaMorph image analysis (Meta Imaging Series 7.1, Version 7.1.6.0, Molecular Devices). Neurons were counted as elongating or retracting, if the majority of the neurites/neuron were respectively elongating or retracting, or as no effect, if there was no change in growth. For growth cone collapse assay, the distance between the growth cone position at 0, 30 and 60 min was observed and reported as difference. For inhibitor studies, neuronal cultures were pre-incubated with nontoxic concentrations of the specific inhibitor (C3-toxin) overnight or with Y27632 for 1 h before time lapse recordings. Retracted neurons were counted in one time-lapse recording frame before treatment and in the frame at 60 min after treatment. Control cultures were treated with equal volumes of vehicle which did not affect survival or outgrowth. During time lapse recording neurons were treated with 1 μM S1P for 60 min in the presence of the inhibitor.

Cells were fixed with 4% PFA for 20 min at RT. After permeabilization with 0.01% TX-100 (Pierce, Vienna, Austria) unspecific binding was blocked for 30 min with 10% normal goat serum (Sigma) in PBS. Cells were incubated with the first antibody (α-β-III-tubulin clone TuJ-1, 1:1000; R&D Systems, Vienna, Austria) for 1 h, washed 3× for 10 min with PBS and incubated with the appropriate secondary antibody for 30 min as previously published (Quarta et al., 2014), counterstained with 4′,6-diamidino-2-phenylindole (DAPI; 1:10.000; Sigma) and embedded in Mowiol (Calbiochem, Vienna, Austria). Secondary antibodies used were chicken α-mouse Alexa Fluo^® 594 or donkey α-mouse Alexa Fluo^® 488 (1:1000; Invitrogen) for fluorescence microcopy.

Digitized images of randomly chosen areas of the coverslip were taken using an Axio Imager microscope (Carl Zeiss) with 16×, NA 0.5 or 25×, NA 0.8 oil-immersion objective. Images were recorded with a cooled CCD camera (SPOT; Diagnostic Instruments). Experiments were performed in triplicates as minimum requirement. Only neurons that were clearly separated from neighboring cells were included for quantification. For the analysis of the percentage of neurite bearing cells, neurons with no visible process or with only filopodial formations were counted as negative. Cells showing at least one neurite with a total length of at least one cell diameter were counted as positive. Quantification of the total length was performed as previously (Quarta et al., 2014) using ImageJ software and NeuronJ plugin¹ (Meijering et al., 2004).

Neuroblastoma x spinal cord motor neuron cells (NSC-34) (kindly provided by Dr. Neil Cashman, University of Toronto, ON, Canada), were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% foetal bovine serum (FBS) and 1% penicillin/ streptomycin/glutamine solution (PSG) as previously described (Cashman et al., 1992). Cells were subcultured every 3–4 days. To slow the proliferation and enhance the differentiation, cells were grown to 60% confluence and the maintenance medium (DMEM, 10% FBS, 1% PSG) was exchanged for differentiating medium (1:1 DMEM plus Ham’s F12, 1% FBS, 1% PSG and 1% modified Eagle’s medium non-essential amino acids, NEAA), as previously described by Matusica et al. (2008). After 24–48 h culture, the medium was replaced with fresh differentiating medium and the survived cells were allowed to differentiate for up to ~7 days in the presence of 10 ng/ml BDNF and 1 ng/ml of CNTF. The media was refreshed every 2 days (Matusica et al., 2008).

Motor neuron-like cells (NSC34) were cultured in 24-well glass bottom plates coated with 0.015% poly-L-ornithine (Sigma) and differentiated as described previously (Matusica et al., 2008, 2016). Cells were seeded at a density of 3000 cells per well, differentiated over 48 h as described above, and vehicle or 1 μM of S1P (Sigma) was added to each well at the initiation time of image capture. Neurite outgrowth assays were performed on a FV1000 laser scanning confocal microscope (Olympus) equipped with a 5.1% CO₂ perfused and temperature controlled (36.9°C) environmental control chamber (Solent Scientific). Image acquisition was performed on a 5 min interval over a period of 90 min, not including plate temperature calibration, using the multi-area time lapse (MATL) feature and the DIC filters engaged for the transmitted light image and 0.5 μm 20× lens correction. Four fields of cells were randomly imaged per well and analysis of neurite length and number of neurites/neuronal cluster (cells contained within the field of view with clearly identifiable processes) was performed on images using the NeuronJ plugin (ImageJ). All subsequent analysis was performed with Fiji (ImageJ) and Prism 7 (Graphpad, USA) software.

Adult DRG neurons were cultured at a density of 2000 neurons/cm² for 24 h and medium was replaced by ECS for 3 h before the experimental procedure. Cultures were treated for 10 min with S1P, FCS or SEW2871 as indicated. RhoA-GTP pulldown assays were performed as described previously (Schweigreiter et al., 2004). Briefly, neurons were washed twice with ice-cold TBS and lysed with ice-cold lysis buffer (500 μl; 50 mM Tris pH 7.2, 1% Triton X-100, 150 mM NaCl, 10 mM MgCl₂, plus protease inhibitors). Lysed cells were scraped, collected into a pre-cooled Eppendorf tube and centrifuged at 14,000 rpm at 4°C for 5 min. The major part of the supernatant (about 450 μl) was added to 18 μl Rhotekin agarose beads (Millipore #14-383) and slowly rotated at 4°C for 50 min. A minor portion (50 μl) of the supernatant was used for assessment of total RhoA amount. Beads were washed three times with lysis buffer and elution and protein denaturation was done with 1× Laemmli buffer at 95°C for 5 min. Samples were electrophoresed in 15% polyacrylamide and proteins were transferred onto PVDF (polyvinylidene fluoride) membranes, which were blocked in 1% skim milk solution and incubated with anti-RhoA antibody (Santa Cruz #sc-418; 1:1000).

Lumbar sensory neurons were harvested from 6 wt mice per experiment and pooled for the following procedure. Sensory neurons were cultured for 24 h and growth factor-deprived in ECS (145 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 10 mM HEPES, 10 mM D-Glucose; pH 7.3; Osm 305–308 mOsm/kg) before stimulation with S1P (1 μM) for 15, 30 and 60 min in the presence or not of the ROCK inhibitor Y27632 (10 μM). A pre-treatment of 1 h with Y27632 was performed. Non-treated cultures were used as controls. Neurons were harvested in freshly prepared ice-cold lysis buffer (50 mM Tris-HCl, 200 mM NaCl, 50 mM NaF, 1 mM EDTA, 20 mM β-Glycerophosphate, 1% Triton X-100) enriched with 1:30 phosphatase inhibitor complex (PIC, Sigma) and 1 mM phosphatase inhibitor sodium-ortho-vanadate (Sigma). Lysates were processed as described previously (Quarta et al., 2014) and protein content was measured with BCA™ Protein Assay kit (Thermo Scientific) according to the protocol. SDS-PAGE was conducted under standard denaturing conditions and 25 μg of protein were loaded per lane. For immunodetection, membranes were incubated with the specific antibodies diluted according to the manufacture instructions. Blots were visualized with enhanced chemiluminescence by using the SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific) and membranes scanned with LAS4000 luminescent imager (GE Healthcare). Quantification was performed using ImageJ software and relative values for phosphorylated proteins are related to the non-phosphorylated form. α-phospho-CRMP2 (Thr-555; 1:500; ECM Biosciences), α-CRMP2 (1:1000; Cell Signaling) and α-tubulin (1:2000; Sigma) were used as primary antibodies, and α-rabbit IgG (1:5000; Thermo Scientific), α-mouse IgG (1:10,000; Sigma) secondary antibody for detection.

The protocol was modified from an earlier method from Obernosterer et al. (2007) and Camprubí-Robles et al. (2013). Cryostat sections of mouse spinal cord were fixed with 4% PFA, digested with proteinase K (2 μg/ml in 50 mM Tris buffer containing 5 mM EDTA), fixed again with 4% PFA, acetylated (triethanolamine/HCl/acetic anhydride) and incubated with prehybridization buffer (Roche). The S1P3 target mRNA was hybridized with 0.34 pmol of specific antisense double digoxigenin (DIG) labeled mRNA detection probes (S1P3, 5DigN/ACTGATG AGGAAGGCGATGTAT/3Dig_N) overnight at 55°C (Camprubí-Robles et al., 2013). Scrambled probes (Exiqon, Copenhagen, Denmark) served as negative control. Washing with saline sodium citrate buffer (SSC) was followed by blocking and incubation with alkaline phosphatase coupled anti-DIG antiserum (1:5000, Roche) for 1 h, followed by washing in levamisole buffer. Slides were incubated with BCIP/NBT (Roche) to visualize S1p3 receptor mRNA, washed in DEPC-treated water, coverslipped in buffered glycerol and sealed. Images were taken using a brightfield microscope (BX50, Olympus).

Differentiated NSC-34 cells treated with vehicle (Control) and S1P were harvested and stored in Trizol (Sigma) at −80°C. Samples were mechanically homogenized using a tissue lyser (Qiagen). Total RNA was isolated using a column based method (Zymo-Spin ICC Columns, Zymogen, Irvine, CA, USA). DNA contamination was removed by on-column DNA digestion. The concentration of total RNA was determined using standard photospectrometry (Nanodrop 2000, Thermo Scientific, Australia), quality of RNA was determined using a lab-on-chip system (Bioanalyzer, Agilent). Only samples with RNA integrity numbers (RIN) above seven were used for subsequent analysis. One microgram of total RNA was reverse transcribed into cDNA according to the manufacturer protocol (SuperscriptII, BioRad, Australia).

Quantitative real-time polymerase chain reaction (qRT-PCR) analysis of the relative mRNA expression levels of differentiated NSC-34 cells treated with vehicle (Control) and S1P were performed using the StepOnePlus cycler (Life Technologies). TaqMan primers (Life Technologies) were used for the detection of peptidergic and non-peptidergic markers and molecules characteristic for mature nociceptive neurons. Hypoxanthine-guanine phosphoribosyl transferase (HPRT) was used as a reference gene. The efficiencies of all primer-pairs were determined by 1/5 to 1/625 dilutions in a qPCR and a satisfying efficiency was determined with Q-Gene (Simon, 2003). The primers and efficiencies are listed in Table 1. The final volume for qPCR was 20 μl of which 8 μl were H₂O, 10 μl mastermix (Life Technologies), 1 μl assay-mix (Life Technologies) and 1 μl cDNA. Each q-RT-PCR was done in duplicate. The Ct values were determined for each product and normalized as pairwise comparisons against the Ct value of the reference gene. Subsequently the mean normalized expression (MNE) was calculated and differences in the expression determined (Simon, 2003).

Factors used: 50 μM ABC294640 (Active Biochemicals Co., Limited; Hong Kong); 10 nM, 100 nM or 1 μM S1P (Sigma) dissolved in 0.1% methanol. Inhibitors and agonists: 0.5 μg/ml C3 toxin (Cytoskeleton, Inc.); 10 μM Y27632 (Calbiochem); 10 μM SEW2871 (Calbiochem); 1 μM FTY720-P (Echeion Biosciences).

All experiments were performed in cell cultures/materials from at least three different animals, and numbers of cells are reported under each figure legend. Behavioral assays were performed in 10 or 11 male animals per genotype, as reported under the figure legend. For statistical analysis the GraphPad software package 6 or 7 was used and data are presented as mean ± SEM. Two-way repeated measurements ANOVA followed by Tukey post hoc test or Mann-Whitney U- post hoc test was used for multiple comparisons between groups for behavioral assays. One-way or two-way ANOVA followed by post hoc test was used for multiple comparisons between groups for in vitro experiments. Kruskal-Wallis test or Mann-Whitney U-test was calculated for interindividual comparisons for in vitro experiments. Data were analyzed by χ²-test for comparison of relative group sizes for in vitro experiments. Individual test are reported under each figure legends. Differences were considered statistically significant at p < 0.05.

The involvement of S1P in neuronal outgrowth was first studied in short term cultures of adult DRG neurons as an in vitro model for peripheral nerve injury. Only 20% of adult DRG neurons had visible processes when neurons were grown in culture for 24 h with 1 μM S1P compared to the 43% of neurons which were treated with vehicle (Figures 1A,C; 42.88 ± 1.82% for ctrl vs. 19.33 ± 4.93% for S1P-treated, p = 0.000391, χ²-test). At the same time the total neurite length of neurons in S1P-treated cultures was significantly reduced after 24 h of S1P treatment (Figures 1A,B; 596.4 ± 41.48 μm for ctrl vs. 318.4 ± 35.28 μm for S1P-treated, p < 0.001, U-test). S1P synthesis was previously suggested to act downstream of the NOGO signaling pathway, which has a significant anti-regenerative potential (Kempf et al., 2014) and NGF signaling causes SphK1 translocation to the plasma membrane with subsequent activation of the S1P₁ and S1P₂ receptors (Toman et al., 2004). Thus, we first explored whether SphKs, involved in S1P production, were involved in the intrinsic extension of neuronal processes in adult sensory neurons. Neuron cultures were obtained from mice with a global null mutation of SphK1 (Sphk1^−/−) and grown in the presence or absence of the SphK2 inhibitor ABC294640 (ABC; Gao et al., 2012). The inhibition of the activity of both SphK isoforms by knock-out and ABC treatment for 24 h did not exhibit any significant effect on neurite outgrowth (Figure 1D; p = 0.092, Kruskal-Wallis test), indicating that inhibition of SphK activity did not alter neurite outgrowth of sensory neurons in the absence of exogenous S1P. Together, these results suggest that S1P might serve a critical role in neurite outgrowth of sensory neurons, in vitro, and that autocrine S1P signaling appears not to be critically involved in the outgrowth of sensory neurons.

Since we and others have previously reported S1P effects to occur on a fast time scale (Postma et al., 1996; Toman et al., 2004; Mair et al., 2011; Camprubí-Robles et al., 2013), we explored whether S1P affected neurons with already existing neurites and applied S1P for short durations to cultures in a live imaging set-up (15, 30, 60 and 90 min). In wt neurons, S1P induced a time-dependent retraction of neuronal processes which was visible after 15 min (Figures 2A,B) and increased with prolonged S1P exposure (Figure 2B). The percentage of neurons with retracting neurites was significantly increased in a dose-dependent manner: 10 nM S1P was already sufficient to induce a slight retraction, however more neurons were elongating than retracting their neurites, and nearly 80% of sensory neurons retracted their neurites in response to 1 μM S1P (Figures 2A,C; p = 0.0001, χ²-test). Similarly, the number of neurons which showed elongating neurites was profoundly decreased to 11.8 ± 7.15% upon 1 μM S1P (Figure 2C). The retraction in response to 1 μM S1P was accompanied by growth cone collapse visible at higher magnification (Figures 2B,D). These results revealed that physiologically relevant concentrations of S1P caused fast retraction of already existing neurites of adult sensory neurons.

Of the generally accepted five S1P receptors, S1P₁, S1P₂ and S1P₃ subtypes are expressed in DRG neurons (Zhang et al., 2006; Mair et al., 2011; Camprubí-Robles et al., 2013). Out of those three receptor subtypes, Toman et al. (2004) showed that S1P₁ receptor is involved in neurite extension, whereas S1P₂ and S1P₃ receptors have been associated, respectively, with processes retraction in PC12 cells (Toman et al., 2004) and cell rounding (Van Brocklyn et al., 1999). We first used SEW2871 to activate the S1P₁ receptor, or FTY720-P, a ligand at S1P_1,3,4,5 receptors. In contrast to S1P, both, SEW2871 and FTY720-P induced elongation of neurites (Figure 3A; p < 0.001, U-test), and increased the number of neurons with elongating neurites, rather than neurite retraction (Figure 3B; p < 0.001, χ²-test). It should be noted that DRG neurons do not express S1P₄ or S1P₅ and only a negligible fraction of neurons in our culture conditions express S1P₂ receptor subtype (Camprubí-Robles et al., 2013). Since retraction was observed in more than 80% of DRG neurons upon 1 μM S1P, we hypothesized that S1P-induced neurite retraction could be mediated through activation of S1P₃ receptors. To prove this hypothesis, DRG neuronal cultures from transgenic mouse lines with global or sensory neuron specific deletions of S1P receptors were exposed to S1P. In cultures from adult S1P₃^−/− mice, the average total neurite length was significantly increased in the presence of S1P after 24 h (426.5 ± 69.32 μm vs. 825.4 ± 158.8 μm; p = 0.0432, U-test; Figure 3C). Moreover, in S1P₃^−/− primary DRG cultures, the percentage of neurons that responded to S1P after 60 min with a retraction was significantly reduced compared to wt (74.49 ± 5.89% vs. 38.74 ± 4.98%; p < 0.001, χ²-test; Figure 3D). In contrast, a conditional, sensory neuron specific deletion of the S1P₁ receptor subtype (SNS-S1P₁^−/−) did not produce a significant change in the percentage of retracted neurons compared to control cultures (Figure 3D). SNS-S1P₁^−/− mice lack S1P₁ mRNA expression and immunoreactivity specifically in DRG neurons (Mair et al., 2011). Sensory neuron cultures from SNS-S1P₁^−/− mice showed reduced elongation upon S1P treatment, however, upon exposure to FTY720-P we observed a significant increase in the elongation compared to S1P treated cultures (Figure 3E; p < 0.0001, χ²-test). This puzzling result could be explained with the recently reported antagonistic effect of FTY720-P at S1P receptors (Hla and Brinkmann, 2011; Camm et al., 2014) and possible inhibition at S1P₃ receptor subtype, since sensory neurons in culture express mainly S1P₁ and S1P₃ receptors, as mentioned above.

These observations support the concept that S1P binding to S1P₃ receptor acts to facilitate neurite retraction of primary sensory neurons. In contrast, activation of S1P₁ receptor may rather promote neurite elongation.

S1P₃ receptor activates members of the Rho GTPase family (Pyne and Pyne, 2000a, b; Sanchez and Hla, 2004; Obinata and Hla, 2012) which are known regulators of axon growth and neuronal morphogenesis (Fujita and Yamashita, 2014). We therefore performed pull-down assays and observed that S1P treatment of primary DRG neuron cultures activated RhoA within 10 min in a dose-dependent manner (Figures 4A,E). Furthermore, S1P-induced neurite retraction was significantly reduced in vitro by the specific Rho inhibitor C3-toxin (Figure 4B; retracted p = 0.0005, two sample t-test). In addition, the percentage of neurons responding to S1P with an elongation of neurites increased significantly in the presence of C3-toxin (Figure 4B; elongated p < 0.001, two sample t-test). The same effect was observed with the ROCK inhibitor Y27632, which decreased the percentage of retracted neurons to nearly 30% compared to S1P-treated cultures (Figure 4C; p < 0.001, χ²-test). In contrast, the S1P₁ receptor selective agonist SEW2871 did not activate RhoA even at high doses (Figures 4D,E), which is in line with the findings obtained in S1P₁ deficient neurons. Since the effect of S1P on neurite retraction occurred within the first 15 min, we investigated the ROCK-dependent phosphorylation of one of its downstream targets, the CRMP2. CRMP2 binds to tubulin dimers and enhances microtubule formation (Fukata et al., 2002), however phosphorylation at threonine-555 by ROCK, reduces its ability to bind tubulin, leading to microtubule disorganization and growth cone collapse (Arimura et al., 2000, 2005). We assessed phospho-CRMP2 (pCRMP2) levels before and after 15, 30 and 60 min of S1P treatment in primary DRG neuron cultures from adult wt mice. Western blot analysis revealed that S1P induced rapid phosphorylation of CRMP2 at Thr-555, with a 2.6-fold increase at 30 min (Figures 4F,G; ctrl vs. S1P 15’ p = 0.019, ctrl vs. S1P 30’ p < 0.0001, ctrl vs. S1P 60’ p = 0.027, one-way ANOVA followed by Tukey post hoc test), while total CRMP2 expression was unchanged (Figure 4H, p = 0.2478, one-way ANOVA with multiple comparisons correction). Notably, addition of the ROCK inhibitor, Y27632, inhibited S1P-induced phosphorylation of CRMP2 (Figures 4F,G; p < 0.0001, one-way ANOVA followed by Tukey post hoc test). Altogether, these results support the hypothesis that S1P is a critical inducer of neurite retraction in adult sensory neurons through S1P₃ receptor by activating the downstream Rho/ROCK pathway with subsequent ROCK-dependent CRMP2 phosphorylation.

S1P₃ receptor mRNA is expressed in populations of small- and large-diameter DRG sensory neurons as we previously observed (Camprubí-Robles et al., 2013). Since in situ hybridization of the spinal cord demonstrated the presence of S1P₃ receptor mRNA also in motor neurons in the ventral spinal horn (Figure 5A), we investigated whether S1P could affect neuronal outgrowth of motor neurons as for sensory neurons. We cultured motor neuron-like cells for 24 h and exposed them to S1P in a time-lapse experiment. Akin to sensory neurons, motor neuron-like cells started to retract already after 15 min of S1P stimulation, and severely retracted neurites after 30 min (Figure 5B). At 1 μM, S1P treatment significantly reduced the number of neurites per neuronal cluster (Figure 5C; 0 vs. 15 min p = 0.0039; 0 vs. 30 min p < 0.0001; 0 vs. 60 min p < 0.0001, one-way ANOVA, Bonferroni post hoc test) and the mean neurite length (Figure 5D; 0 vs. 15 min p = 0.0049; 0 vs. 30 min p < 0.0001; 0 vs. 60 min p < 0.0001, one-way ANOVA, Tukey post hoc test) after 30 min stimulation. After 30 min S1P-treated cells had 8 neurites/neuronal cluster compared to 17 neurites/neuronal cluster of the vehicle treated cells. Moreover, S1P treatment decreased the total neurite length of motor neuron-like cells from 79.258 ± 4.202 μm to 59.101 ± 4.192 μm after 30 min (Figure 5D). In line with our ex vivo data, motor neuron-like cells significantly expressed S1P₃ receptor subtype and the normalized mRNA expression did not change after S1P treatment (Figure 5E). Therefore, our data indicate that S1P affects neuronal morphology not only in adult sensory neurons but as well in motor neurons which likewise express the S1P₃ receptor, indicating a wider action of S1P in functionally different neuron types.

To demonstrate the relevance of S1P₃ receptor signaling in peripheral nerve regeneration in vivo, we used a sciatic nerve crush injury model and monitored sensory thresholds as signature of the functional recovery 15 days post-lesion (dpl). S1P₃^−\− mice recovered significantly faster to normal heat paw withdrawal latencies after crush injury compared to wt control group (Figure 6A; day 4 p = 0.016, day 8 p = 0.045, two way RM ANOVA, Mann-Whitney post hoc test with Holm-Bonferroni’s multiple correction): in both wt and S1P₃^−\− mice heat sensitivity decreased after injury, however remained elevated in wt mice and returned to baseline at 15 dpl (day 1: 206 ± 17% vs. 181 ± 15%, and day 15: 125 ± 18% vs. 103 ± 12% for wt and S1P₃^−\−, respectively). In S1P₃^−\− animals the heat sensitivity was significantly improved already from early post-lesion stages (day 1: 181 ± 15%; day 4: 133 ± 10% and day 8: 132 ± 11%). In line with our in vitro data, sensory recovery of conditional knock-out animals for S1P₁ receptor was comparable to wt, indicating that the S1P₁ receptor is not critically involved in the recovery after peripheral lesion (Figure 6B; two way RM ANOVA, genotype: F_(1,131) = 0.371, p = 0.549; time points: F_(6,131) = 16.599, p < 0.001; genotype × time points: F_(6,131) = 0.602, p = 0.728). These in vivo data together with our in vitro results introduce S1P₃ receptor signaling as a general regulator of the outgrowth of peripheral neurons, which appears to be important in the fine-tuning of peripheral nerve regeneration and degeneration.