Extracellular α-synuclein induces sphingosine 1-phosphate receptor subtype 1 uncoupled from inhibitory G-protein leaving β-arrestin signal intact

Parkinson’s disease (PD) is the second most common neurodegenerative disorder. The presence of α-synuclein (α-Syn)-positive intracytoplasmic inclusions, known as Lewy bodies, is the cytopathological hallmark of PD. Increasing bodies of evidence suggest that cell-to-cell transmission of α-Syn plays a role in the progression of PD. Although extracellular α-Syn is known to cause abnormal cell motility, the precise mechanism remains elusive. Here we show that impairment of platelet-derived growth factor-induced cell motility caused by extracellular α-Syn is mainly attributed to selective inhibition of sphingosine 1-phosphate (S1P) signalling. Treatment of human neuroblastoma cells with recombinant α-Syn caused S1P type 1 (S1P1) receptor-selective uncoupling from inhibitory G-protein (Gi) as determined by both functional and fluorescence resonance energy transfer (FRET)-based structural analyses. By contrast, α-Syn caused little or no effect on S1P2 receptor-mediated signalling. Both wild-type and α-Syn(A53T), a mutant found in familiar PD, caused uncoupling of S1P1 receptor, although α-Syn(A53T) showed stronger potency in uncoupling. Moreover, S1P1 receptor-mediated β-arrestin signal was unaltered by α-Syn(A53T). These results suggest that exogenous α-Syn modulates S1P1 receptor-mediated signalling from both Gi and β-arrestin signals into β-arrestin-biased signal. These findings uncovered a novel function of exogenous α-Syn in the cells.


Results
PDGF utilises transactivation of S1P 1 receptor for maximal chemotaxis. It has recently been reported from this laboratory that extracellular α -Syn inhibits PDGF-induced chemotaxis 19 . To identify signalling pathway, which is important in PDGF-induced chemotaxis and that is sensitive to extracellular α -Syn, we first examined the involvement of S1P signalling in this phenomenon. An S1P 1 receptor-specific blocker, W146, inhibited PDGF-induced chemotaxis to an extent similar to α -Syn(A53T) treatment (Fig. 1a). Since S1P 1 receptor is known to couple exclusively with Gi 27 , it is reasonable to assume that pertussis toxin (PTX) inhibits the cell motility. In contrast, a selective S1P 2 receptor antagonist JTE-013 showed no effect on PDGF-induced chemotaxis. Involvement of S1P signalling in PDGF-induced chemotaxis was further confirmed by downregulating one of the subtypes of SphK, SphK1 expression using small interfering RNAs (siRNAs). SphK1-siRNA caused inhibition of PDGF-induced chemotaxis by 30% as compared with the control siRNA (Fig. 1b). Similarly, knockdown of S1P 1 receptor by S1P 1 receptor-siRNA caused 40% inhibition of PDGF-induced chemotaxis. These results indicate that S1P signal is involved in PDGF-induced chemotaxis, consistent with a previous report 20 . As described in a previous study showing a successful detection of conformational changes in α 2A-adrenergic receptor using a fluorescence resonance energy transfer (FRET)-based technique 28 , we have similarly constructed a probe to detect conformational changes in S1P 1 receptor using a FRET technique, where the cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) were separately fused in the same S1P 1 receptor molecule (Fig. 1c). Under a resting state these fluorophores are closely situated, which enables FRET to occur, whereas the specific agonist (S1P) stimulation causes conformational changes in the receptor, resulting in FRET decrease. Upon stimulation by S1P, cells expressing this FRET probe resulted in FRET changes in a W146-sensitive manner (Fig. 1d), validating this probe. Noticeably, when the cells were stimulated with PDGF instead of S1P, there was a rapid change in the CFP/FRET ratio, which was inhibited in SphK1-siRNA-treated cells (Fig. 1e), strengthening that PDGF utilises S1P 1 receptor transactivation for maximal chemotaxis. Our previous findings that extracellular α -Syn suppressed PDGF-induced chemotaxis in SH-SY5Y cells 19 facilitated an investigation of whether α -Syn has any effects on S1P signal. To assess this issue, S1P in place of PDGF was used as an agonist to activate S1P 1 receptor to simplify the system thereafter.
α-Syn causes uncoupling of S1P 1 receptor from Gi. To investigate the effect of extracellular α -Syn on S1P signalling, S1P receptor-mediated downstream events, i.e., G-protein subunit dissociation, were monitored by FRET-based conformational changes. Since SH-SY5Y cells express mainly S1P 1 and S1P 2 receptors as judged by real-time quantitative reverse transcription-PCR (Fig. 2a), we focused on these two subtypes of the S1P receptors for S1P signalling. To see the effect of α -Syn on each S1P receptor-mediated signalling, we carried out a FRET analysis using each S1P receptor-CFP and Gγ -YFP as a FRET pair. Under basal conditions heterotrimeric G-protein subunits are associated (Gα β γ form, low FRET). Upon stimulation by S1P, these subunits dissociate, and S1P receptor-CFP and Gγ -YFP become associated (high FRET) 26 . In control cells expressing S1P 1 receptor-CFP and Gγ -YFP, S1P caused a rapid increase in FRET efficiency, demonstrating a successful detection of S1P 1 receptor-mediated G-protein subunit dissociation (Fig. 2b). Surprisingly, α -Syn treatment made the receptor refractory toward S1P 1 receptor-mediated G-protein subunit dissociation. It is notable that α -Syn(A53T), a mutant α -Syn found in familiar PD, caused more potent effects than a wild-type α -Syn with the same concentration (1 μ M). On the other hand, when the cells expressing S1P 2 receptor-CFP and Gγ -YFP were stimulated with S1P, the agonist-induced G-protein dissociation occurred. However, wild-type α -Syn and α -Syn(A53T) treatment showed no significant effects on S1P 2 receptor-mediated G-protein subunit dissociation (Fig. 2c). These results suggest that α -Syn causes a selective S1P 1 receptor uncoupling from the G-protein but not S1P 2 receptor. α -Syn effects were not from cytotoxic ones since S1P 2 receptor-mediated signalling occurred normally in α -Syn-treated cells (Fig. 2c). Furthermore, two days incubation of the cells with 1 μ M α -Syn(A53T) did not cause apoptosis as judged by chromatin condensation assay (Table I). Since the effect of α -Syn(A53T) is clearer than the wild type with the same concentration, subsequent analysis was mainly carried out using α -Syn(A53T).
To demonstrate that α -Syn(A53T) causes S1P 1 receptor uncoupling from Gi more directly, we have carried out the experiments using another FRET pair with Giα -CFP and Gγ -YFP 26 . Under basal conditions these subunits were associated (high FRET) and S1P 1 receptor stimulation resulted in the dissociation of the Giα from the Gβ γ subunits (low FRET) (Fig. 2d, closed bars). As expected, α -Syn(A53T) treatment abolished the S1P-induced FRET changes (Fig. 2d, hatched bars), indicating that α -Syn(A53T) causes S1P 1 receptor uncoupling from Gi. Next, to address whether agonist stimulation facilitates S1P 1 receptor coupling with Gi, the association of S1P 1 receptor with Giα before and after agonist stimulation was assessed by FRET analysis. Under basal conditions S1P 1 receptor-YFP was already associated with Giα -CFP as detected by a high FRET efficiency (Fig. 2e, closed bar). S1P and PDGF stimulation caused a significant (light grey bar) and a mild decrease (dark grey bar) in FRET efficiency, respectively. These changes may reflect agonist-induced Gi subunit dissociation. Critically, α -Syn(A53T) caused a significant decrease in FRET efficiency (hatched bar). Taken together with the findings that α -Syn(A53T) impairs the S1P-induced Gi subunit dissociation (Fig. 2d), extracellular α -Syn(A53T) may segregate S1P 1 receptor from Gi.

Abrogation of endogenous Gi function by α-Syn(A53T).
The effect of α -Syn(A53T) on endogenous Gi function was assessed next. When cells were treated with S1P, forskolin-stimulated adenylate cyclase activity was potently inhibited (a phenomenon known as a classical Giα function) in a S1P 1 receptor antagonist W146-sensitive manner (Fig. 3a), evaluating the S1P 1 receptor/Gi function in an endogenous cell system. Importantly, α -Syn(A53T) abrogated the S1P 1 receptor/Giα -caused inhibition of forskolin-stimulated adenylate cyclase, supporting the premise that α -Syn(A53T) caused S1P 1 receptor to uncouple from Gi protein. α -Syn(A53T) itself had no inhibitory effect on forskolin-stimulated adenylate cyclase activity. In contrast, S1P-induced increase in cellular Ca 2+ through S1P 2 receptor, which was inhibited by JTE-013 (Fig. 3b), was insensitive to α -Syn(A53T) (Fig. 3c). These results indicate that α -Syn(A53T) causes selective impairment of S1P 1 Values represent means ± s.e.m. of three independent experiments carried out in triplicate. Statistical significance was analysed by Student's t-test (**P < 0.01). (b) SH-SY5Y cells transiently transfected with control, S1P 1 receptor-or SphK1-siRNA were cultured for 24 hr and then plated on the upper chamber in the absence of serum. The cells migrated into the lower chamber in the absence (none) or presence of 20 ng/ml PDGF were counted. Values represent means ± s.e.m. of 3 independent experiments carried out in triplicate. Statistical significance was analysed by Student's t-test (**P < 0.01). (c) A schematic diagram for a FRET-based probe to detect conformational changes in S1P 1 receptor was depicted. CFP and YFP were separately fused to the same receptor molecule. Under non-activated conditions these two fluoroprobes associate closely (high FRET), whereas S1P-induced conformational changes of the receptor cause their dissociation (low FRET). (d) Cells transiently expressing this FRET probe in (c) were serum-starved for 18 hr and stimulated with 100 nM S1P (first arrow) and analysed for FRET in living cells. Two min after S1P stimulation, 10 μ M W146 was added (second arrow). (e) SH-SY5Y cells cotransfected with control or SphK1-siRNA together with vectors encoding this FRET probe in (c) were serum-starved for 18 hr and stimulated with 20 ng/ml PDGF (arrow) and analysed for FRET in living cells. A representative emission ratio of the 2 fluorophores from 5 independent experiments is shown. receptor/Gi signalling leaving S1P 2 receptor signalling intact in an endogenous protein system as well as transient expression system (Figs 2 and 3). Mechanism underlying α-Syn(A53T)-induced uncoupling of S1P 1 receptor from Gi. It is well known that PTX causes G-protein-coupled receptor uncoupling from Gi by ADP-ribosylation of the Giα subunit. The general feature of PTX is to eliminate widely the Gi-dependent phenomena including chemoattractant formyl-Met-Leu-Phe (fMLP)-induced respiratory burst. We therefore undertook a study to compare α -Syn(A53T) with PTX for the ability to cause uncoupling of fMLP receptor. As reported previously 29 , PTX treatment inhibited almost completely fMLP-induced respiratory burst in differentiated HL-60 cells (Fig. 4). In contrast, α -Syn(A53T) had no effects on the phenomena, suggesting that the site of action of α -Syn(A53T) may be distinct from that of PTX. Next, the effect of α -Syn(A53T) on S1P-induced conformational changes in S1P 1 receptor was assessed using an S1P 1 receptor FRET tool. S1P-induced conformational changes in the receptor were not influenced by α -Syn(A53T) treatment ( The effect of α-Syn(A53T) on β-arrestin signal, one of the divergent signals downstream of S1P 1 receptor. Since G-protein-coupled receptors are known to utilise several other downstream signalling molecules such as β -arrestin as well as G-protein 30 , S1P-induced β -arrestin binding to S1P 1 receptor was measured. S1P caused association of β -arrestin with S1P 1 receptor in a time-dependent manner with the maximum level in 30 min (Fig. 6). Importantly, α -Syn(A53T) treatment had little or no effects on the S1P-induced β -arrestin association with the receptor (Fig. 6b, hatched bars). Next, the effect of α -Syn(A53T) on S1P-induced internalisation of the S1P 1 receptor, one of the well known outcomes of β -arrestin signal 31 , was studied. The cell surface proteins were biotinylated from outside the cells expressing S1P 1 receptor-YFP before and after S1P stimulation, and then biotinylated S1P 1 receptor was quantitated. α -Syn(A53T)-treated cells showed a decrease in a cell surface S1P 1 receptor after S1P stimulation to an extent similar to that in the control cells (Fig. 7a, compare hatched bars with closed bars). This observation was further confirmed by detecting cell surface endogenous S1P 1 receptor using an antibody to detect extracellular epitope of the receptor. Stimulation of cells with S1P caused a decrease in endogenous cell surface S1P 1 receptor both in control and α -Syn(A53T)-treated cells to a similar extent ( Fig. 7b and c, hatched bars versus closed bars). Furthermore, S1P 1 receptor internalisation was demonstrated directly in an immunocytochemical study. Upon stimulation of cells with S1P, S1P 1 receptor-YFP became localised in small punctate structures suggesting S1P 1 receptor internalisation (Fig. 7d). It should be noted that α -Syn(A53T) had little or no effects on S1P-induced formation of S1P 1 receptor-positive dot-like structures. Indeed, these results indicate that α -Syn(A53T) causes uncoupling between S1P 1 receptor and Gi protein, leaving β -arrestin signal unchanged.

Discussion
Recent studies have revealed that α -Syn can be released from cultured cells by exocytosis 32 or by exosomes 33 and that α -Syn is detected in cerebrospinal fluid and plasma 16,17 . In addition, cell-to-cell transmission of α -Syn was shown experimentally to induce an inclusion formation and neuronal cell death 34 . From this background we have reasoned that studies on changes in cellular functions induced by exogenous α -Syn may help understand the pathophysiology of α -Syn.
In the present studies we have shown that extracellular α -Syn causes S1P 1 receptor-selective uncoupling from Gi as determined by both functional (Fig. 3) and FRET-based structural analyses (Fig. 2b,c,d). It has previously been reported that in 1-methyl-4-phenylpyridinium (MPP+ ) treatment of SH-SY5Y cells, an in vitro PD model, there was a significant decrease in SphK1 gene expression and that Sphk1 inhibition plays an important role in caspase-dependent apoptotic neuronal death 35 . However, the mRNA levels of S1P signalling molecules such as SphK1, SphK2, S1P 1 and S1P 2 receptors were unchanged under the experimental condition used in the present studies (1 μ M α -Syn treatment for 18 hr) although in the proapoptotic conditions such as a higher dose (10 μ M) of α -Syn or prolonged time (42 hr) treatment the mRNA level of SphK1 decreased, while that of SphK2 increased (data not shown).
As for the site of action of α -Syn(A53T), it may not act on Gi protein but on S1P 1 receptor because fMLP-induced respiratory burst was insensitive to α -Syn(A53T) in differentiated HL-60 cells, although it was completely inhibited by PTX, which acts on Giα subunit directly (Fig. 4). α -Syn(A53T) had little or no effect on S1P 2 receptor as judged by a FRET-based analysis (Fig. 2c) and cellular Ca 2+ rise in an endogenous protein system (Fig. 3c). Taken together, it may be plausible to assume that α -Syn/α -Syn(A53T) works specifically to Scientific RepoRts | 7:44248 | DOI: 10.1038/srep44248 S1P 1 receptor. However, a pull-down of S1P 1 receptor was not able to detect α -Syn(A53T) association (data not shown). This suggests that α -Syn(A53T) may exert its action indirectly or with the aid of other molecules. In this context it has been shown that the translocation of S1P 1 receptor to caveolin-enriched microdomains is necessary for the subsequent efficient signalling 36 . In their studies oxidised 1-palmitoyl-2-arachidonoyl-sn-glycero-3 -phosphocholine-mediated rapid recruitment to caveolin-enriched microdomains of signalling molecules including the S1P 1 receptor and Akt is important in endothelial barrier enhancement in human pulmonary endothelial cells. It has been suggested that α -Syn has an ability to interact with gangliosides in the cholesterol and sphingolipid-rich membrane microdomains known as lipid rafts, and has a potency to alter the functions of several signalling molecules at the raft regions 15 . In these lines it has been shown that disruption of lipid raft by methyl-β -cyclodextrin caused impairment of G-protein effector signalling but not α 1a-adrenergic receptor internalisation 37 . To support this we have recently observed that ganglioside binding-deficient mutant of α -Syn(A53T), α -Syn(A53T)-AAA, lost its ability to suppress PDGF-induced chemotaxis in SH-SY5Y cells 19 . Consequently, it may be possible that α -Syn/α -Syn(A53T) causes changes in the membrane microdomain environment, which in turn alters S1P 1 receptor function.
The present results also show that another downstream signalling of S1P 1 receptor, β -arrestin-involved signalling was insensitive to α -Syn(A53T) (Fig. 6) and its physiological effect-S1P-induced internalisation of the  ) and stimulated with 100 nM S1P for 1 hr as specified. Cells were harvested and incubated with anti-S1P 1 receptor antibody conjugated with Alexa 488, and analysed by flow cytometry. One of the representative histogram plots obtained from three independent experiments is shown. (c) Cell surface S1P 1 receptor was quantitated based on histogram plots in (b) and expressed as % of control. Values represent means ± s.e.m. of 3 independent experiments. Statistical significance was analysed by Student's t-test (**P < 0.01 versus no stimulation). (d) S1P 1 -YFP-transfected SH-SY5Y cells were incubated with or without α -Syn(A53T) as in (a) and stimulated with 100 nM S1P for 1 hr as specified. Cells were fixed with paraformaldehyde. The localisation of S1P 1 -YFP was analysed by confocal microscopy. Bars, 10 μ m. One of the representative results from three independent experiments is shown. S1P 1 receptor was unaffected by α -Syn(A53T) (Fig. 7). It has been shown that β -arrestin binding requires the ligand-induced conformational changes of the G-protein-coupled receptor and the subsequent receptor phosphorylation 38 . We have shown that α -Syn(A53T) caused no effects on the S1P-induced conformational changes in the S1P 1 receptor as judged by FRET-based studies (Fig. 5). These conformational changes may trigger phosphorylation of the receptor necessary for subsequent β -arrestin binding. The demonstration here indicates that exogenous α -Syn modulates S1P 1 receptor-mediated signalling from both Gi and β -arrestin signals into β -arrestin-biased one by uncoupling of the receptor from Gi. Although the discovery of hereditary forms of PD has contributed greatly to understand the pathogenesis of the disease, that of sporadic PD, the majority forms of PD, is still elusive. It has recently been revealed that mutations in the GBA and SMPD1 genes are risk factors for PD. GBA encodes the lysosomal enzyme glucocerebrosidase that catalyses the breakdown of the glycolipid glucosylceramide to ceramide and glucose. GBA mutations, when homozygous, lead to Gaucher's disease, while predispose to PD when heterozygous 39 . SMPD1 mutations that cause Niemann-Pick type A were significantly higher in patients with PD compared to young controls 40 . SMPD1 encodes sphingomyelin phosphodiesterase 1 (acid sphingomyelinase), a lysosomal enzyme that hydrolyzes sphingomyelin to generate phosphorylcholine and ceramide. Interestingly, both gene products share a common feature, i.e., their enzymatic products are ceramide. Ceramide is shown to be involved in exosomal vesicle formation in multivesicular endosomes (MVEs) 41 . Furthermore, we have recently reported that continuous activation of S1P 1 receptor by S1P, a further metabolite of ceramide, on MVEs has an essential role in the cargo sorting into exosomes 26 . Along with the evidence that α -Syn is secreted in the form of exosomes 33 , it is tempting to speculate that uncoupling of S1P 1 receptor from Gi caused by extracellular α -Syn may inhibit exosomal release of α -Syn, which results in the accumulation of α -Syn inside the cells, a pathological hallmark of PD. Further studies on the mechanism underlying extracellular α -Syn-caused uncoupling of S1P 1 receptor from Gi will be the keys to understand the pathogenesis of sporadic PD.

Methods
Reagents. S1P was purchased from Enzo Life Sciences; PTX, forskolin and dibutyryl cAMP from Wako Pure Chemical Industries; fMLP and cytochrome c were from Sigma Aldrich. W146 and JTE-013 were from Cayman Chemical Company. Other reagents and chemicals were of analytical grade.
SH-SY5Y cells were transfected with the siRNAs using Lipofectamine RNAiMAX according to the manufacturer's instructions (Invitrogen, Carlsbad, CA, USA).

Bacterial expression and purification of recombinant α-Syn and α-Syn(A53T). Recombinant
α -Syn and α -Syn(A53T) were expressed in E. Coli and purified as described previously 43 . Briefly, α -Syn or α -Syn(A53T) cDNAs subcloned into pET3a was transformed in E. coli BL21 (DE3) and protein expression was induced by 0.1 mM IPTG for 3 hr. Bacterial pellets were resuspended in TE buffer (10 mM Tris-HCl, pH7.5 and 1 mM EDTA) containing 750 mM NaCl (TE-750 mM NaCl) with protease inhibitors, heated at 100 °C for 10 min, and centrifuged at 70,000 × g for 30 min. The supernatant was dialysed against TE-20 mM NaCl, filtered by 0.22 μ m filter and applied to a Mono S column (GE Healthcare). The unbound fractions were applied to a Mono Q column (GE Healthcare). α -Syn was eluted with a 0-0.5 M NaCl linear gradient. The fractions containing α -Syn were identified by Coomassie Brilliant Blue staining and immunoblot analysis following SDS-PAGE. Protein concentration was determined using Bradford protein assay kit (Bio-Rad).
Values represent means ± s.e.m. of 3 independent experiments carried out in triplicate. To detect endogenous S1P 1 receptor internalisation, SH-SY5Y cells (1 × 10 6 cells) were serum-starved in the absence or presence of 1 μ M α -Syn(A53T) for 18 hr. After cell stimulation with S1P, cells were chilled on ice and harvested by using Cell Dissociation Buffer (Life Technologies, Inc.). Cells were washed with PBS containing 1% sodium azide and 1% BSA and incubated with 1 μ g/ml of anti-S1P 1 receptor antibody conjugated with Alexa 488 (Novus Biologicals), which detects endogenous cell surface S1P 1 receptor from outside of the cells, at 4 °C for 1 hr, followed by washing and analysed by flow cytometry (FACScalibur, BD Biosciences).
Acceptor photobleaching. SH-SY5Y cells were transiently cotransfected with S1P 1 -CFP or S1P 2 -CFP, Gβ and Gγ -YFP 26 , with a donor/acceptor ratio of 1:1:1, with one-molecule cAMP probe, Epac1-camps 42 , with Giα -CFP, Gβ and Gγ -YFP 26 or Giα -CFP and S1P 1 -YFP. Two days after transfection, cells were serum-starved in the absence or presence of 1 μ M α -Syn or α -Syn(A53T) for 18 hr and treated with various reagents. Cells were then fixed and each area of interest was subjected to FRET analysis with acceptor photobleaching method using a LSM 510 META with a 63 x oil plan-apochromat objective. Following excitation at 458 or 514 nm, CFP emission with a 475-525-nm band-pass barrier filter or YFP emission with 530-600-nm band-pass barrier filter, respectively, was collected. An area of interest was selected for photobleaching of YFP. An automated acquisition protocol was then used, which recorded pre-and post-bleaching images using 458 nm excitation at 8% laser power to limit photobleaching, with a bleaching of the selected area with 100%, 514 nm laser power with 50 iterations (acceptor photobleaching). FRET was resolved as an increase in the CFP (donor) signal after photobleaching of YFP (acceptor). FRET efficiency (E) can be determined from the relative fluorescence intensity of the energy donor (CFP) before (Ipre) and after (Ipost) photobleaching of the energy acceptor (YFP): E = 1 − (Ipre/Ipost).
Measurement of superoxide anion production in differentiated HL-60 cells. HL-60 cells were cultured for 48 hr with 0.2 mM dibutyryl cAMP to induce differentiation followed by serum starvation in the absence or presence of 1 μ M α -Syn(A53T) or 100 ng/ml PTX for 18 hr. Differentiated HL-60 cells were suspended in the HEPES-buffered medium consisting of 10 mM HEPES/NaOH (pH 7.4), 130 mM NaCl, 4.7 mM KCl, 1 mM CaCl 2 , 1.2 mM KH 2 PO 4 , 1.2 mM MgSO 4 , 10 mM glucose and 0.2% BSA with or without 1 μ M α -Syn(A53T). Cell suspension (10 6 cells/tube) was treated with 0.1 μ M fMLP and 50 μ M cytochrome c at 37 °C for 10 min followed by rapid centrifugation. The cell superoxide anion production was estimated by measuring the reduction of cytochrome c as the increase in absorbance at 550 nm using a spectrophotometer.  Table 1. SH-SY5Y cells were serum-starved in the absence or presence of 1 μM wild-type α-Syn or α-Syn(A53T) for indicated time periods. Chromatin condensation during apoptosis was detected by 4,6-diamidino-2-phenylindole-2-HCl (DAPI) staining. Data are mean ± s.e.m. of 3 independent experiments. No significant difference was seen between the treatments.