miR-221 and miR-222 promote Schwann cell proliferation and migration by targeting LASS2 after sciatic nerve injury

The peripheral nervous system (PNS), unlike the central nervous system (CNS), has the intrinsic capacity to regenerate. Severed peripheral nerves are able to regrow and reconnect to their targets, even if their previous functions were seriously compromised (Gruart et al., 2003). The sciatic nerve, which comprises a mixed population of motor and sensory axons, is a commonly used model for nerve regeneration studies. Nerve regeneration is a complex biological phenomenon that involves many cell types, growth factors and extracellular matrices (Gu et al., 2011; Rishal and Fainzilber, 2010). There is a growing consensus that the distinct ability of peripheral nerves to regrow back to their targets hinges on the regenerative properties of the glia or Schwann cells (SCs). When peripheral nerve injury occurs in adult animals, mature differentiated SCs undergo profound phenotypic modulation, shedding their myelin sheaths and dedifferentiating to a progenitor or stem-cell-like state (Freidin et al., 2009). Dedifferentiated SCs can replenish lost or damaged tissues by proliferation, and produce a favorable environment for axonal outgrowth both by helping to clear myelin debris and forming cellular conduits or corridors that guide axons through the degenerated nerve stump and back to their targets (Parrinello et al., 2010). Although in the past few decades many studies have described the process whereby peripheral nerves repair after injuries, the precise mechanisms, which account for the phenotypic modulation in mature SCs, such as SC dedifferentiation, proliferation, migration and redifferentiation to a myelinating phenotype responsible for formation of new nerve tissue to rejoin the distal stump, remain largely unclear.

microRNAs (miRNAs) are a class of approximately 22 nucleotide non-coding RNA molecules that negatively regulate the expression of a wide variety of genes mainly through direct interaction with the 3′-untranslated regions (3′-UTR) of their corresponding mRNA targets (Bartel, 2009). It has been estimated that miRNAs regulate up to 60% of the total human genes at the post-transcriptional level (Friedman et al., 2009), indicating that miRNAs have pivotal roles in physiological and pathological processes. The importance of miRNA in neural development and neurodegeneration is starting to be recognized (Eacker et al., 2009; Fineberg et al., 2009), but their roles in nerve injury and repair currently remain largely unknown. It was reported that miRNA expression profiles were significantly altered in the spinal cord injury model of adult rats (Liu et al., 2009a; Strickland et al., 2011). Using microarray analysis, we found that abnormal expression of miRNA in dorsal root ganglia might illustrate the molecular mechanisms of nerve regeneration during the early phase after sciatic nerve transection (Zhou et al., 2011). The role of miRNA in nerve myelination was recently shown by independent studies of Dicer1-deficient oligodendrocytes and SCs (Dugas et al., 2010; Emery, 2010; Pereira et al., 2010; Yun et al., 2010). Despite the different molecular approaches for gene silencing in these studies, the ablation of Dicer1 (a key molecule in biogenesis of miRNA), as identified by each study, led to the common conclusion about glial overproliferation and aberrant myelination. To date, however, no reports are available on the vital roles of miRNAs in controlling the remodeling of SCs during nerve injury and repair.

In this study we investigated the role of miRNAs in regulating SC gene expression and functions during nerve regeneration. Our results indicate that the miR-221 and miR-222 cluster plays an important regulatory role in the promotion of SC proliferation and migration by targeting to longevity assurance homologue 2 (LASS2).

To examine the involvement of miRNA in nerve regeneration, we studied the expression profile of miRNA in proximal stump of the nerve after sciatic nerve transection with Agilent miRNA microarray. A total of 77 miRNAs showed dynamic alteration between the serial time points (at 0, 1, 4, 7 and 14 days) with random variation model screening. The hierarchical cluster analysis was conducted, and the miRNA transcriptome show an extraordinary resemblance between 4 and 14 days compared with the control group and 1 day after nerve injury (Fig. 1). To screen some key miRNAs, expression pattern analysis was performed and four significant patterns (profile 6, 26, 76, 9) were found (supplementary material Fig. S1). Notably, the expression of five miRNAs in profile 76 (miR-21, miR-31, miR-221, miR-222, miR-132) were promptly elevated at 1 day after sciatic nerve injury, and then leveled off with significantly higher values, compared with that for the control group, over a period of 14 days.

We searched for the putative targets of these five miRNAs in profile 76 using the miRBase database, then integrated putative miRNA targets with differentially expressed mRNA yielding 274 potential targets. To get credible biological functions, we conducted Gene ontology (GO) (Table 1) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway (supplementary material Table S1) enrichment analyses for the intersected genes. Table 1 lists the GO terms that have the highest enriched score and most significant P value. The most significant GO functions were related to cell phenotype modulation, including cellular homeostasis, negative regulation of cell proliferation, taxis and ensheathment of neurons. Functional classification by the KEGG pathway [cytokine–cytokine receptor interaction; ABC transporters; cell adhesion molecules (CAMs) and MAPK signaling pathway] was significantly enriched by the targets. In addition, we constructed a network for 274 target genes and 5 miRNAs in profile 76 (supplementary material Fig. S2).

Of these five miRNAs, the miR-221/222 cluster was specifically selected for functional studies in light of its known pro-proliferative and pro-migratory roles in different cell types (Galardi et al., 2007; Liu et al., 2009b; Wong et al., 2010), which is consistent with our findings that cellular-modulation-related processes are high in the list of GO terms of miRNA targets. We demonstrated the expression of miR-221 and miR-222 by quantitative real-time polymerase chain reaction (qRT-PCR) assay in proximal stumps of the nerve following nerve injury at different time points, and the tendency of miR-221/222 alteration was in agreement with the microarray data (Fig. 2A,B). In situ hybridization studies confirmed that the upregulation of miR-221/222 occurred in SCs at 4 days after injury (Fig. 2C).

To test whether the dysregulation of miR-221/222 is sufficient to impact cellular function, SCs were transfected with miR-221/222 mimics, miR-221/222 inhibitors or non-targeting negative controls. Incorporation of 5-ethynyl-2′-deoxyuridine (EdU), which was used to assay the cell proliferation rate of SCs transfected with miR-221 mimic or miR-222 mimic was significantly increased by approximately 2.5-fold when compared with that of control (Fig. 3A). However, miR-221/222 inhibitors blocked the proliferation rate of SCs (Fig. 3B). Transwell migration assay showed that transfection of miR-221/222 mimics significantly increased the migratory ability of SCs when compared with control cells (Fig. 4A). Furthermore, silencing of miR-221/222 with their inhibitors in SCs markedly impaired cell migration (Fig. 4B). Thus, miR-221/222 upregulation is sufficient to stimulate proliferation and migration of SCs.

In summary, SCs undergo profound phenotypic modulation, adopting more proliferative and/or migratory phenotypes after peripheral nerve injury, and the response of SCs under nerve regeneration is partially regulated by mechanisms mediated by the endogenous miR-221/222 cluster.

Because miRNAs exert their biological functions through the suppression of target genes, it is important to identify miRNA and gene target pairs. To explore the underlying molecular mechanism through which miR-221/222 initiate SC proliferation and migration, we only selected the inversely correlated miRNA target pairs that were expressed in the proximal nerve stump, and thereby identified the significantly correlated pairs (supplementary material Table S2). Among them, LASS2 was involved in the suppression of cell growth or metastasis (Fei et al., 2004; Pan et al., 2001), which was shown to be potentially downregulated both by miR-221 and miR-222. We found, as assessed by qRT-PCR and western blotting, a reduction in LASS2 mRNA and protein expression after nerve injury compared with the alteration of miR-221/222 (Fig. 5A). To verify whether these miRNAs affected the endogenous LASS2 level, we analyzed the effects of ectopic alteration of miR-221 and miR-222 in primary SCs. qRT-PCR and western blot analysis further demonstrated that overexpression of miR-221/222 dramatically suppressed the endogenous mRNA and protein level of LASS2 by more than 50%, whereas silencing of miR-221/222 significantly increased the expression of LASS2 (Fig. 5B). In an effort to determine whether LASS2 is regulated by miR-221 and miR-222 through direct binding to its 3′-UTR, the wild-type or mutant 3′-UTR of LASS2 was constructed and inserted into the region immediately downstream of the luciferase reporter gene (Fig. 5C). For luciferase activity assays, miR-221/222 mimics were co-transfected with p-Luc-UTR construct into HEK293T cells. The relative luciferase activity was significantly reduced by nearly 70% by miR-221 and miR-222 when the wild-type 3′-UTR of LASS2 was present. This reduction was sequence specific, because the relative luciferase activity did not drop as sharply in UTRs that contained a mutant binding site compared with a wild-type binding site (Fig. 5D). These results suggest that high expression levels of miR-221/222 might be one mechanism that negatively regulates LASS2 in nerve regeneration.

To explore the function of LASS2, two specific small interfering RNAs (siRNAs) against LASS2 were synthesized. As shown in Fig. 6A, both siRNA-1 and siRNA-2 greatly reduced the expression of LASS2 mRNA and protein. Cell proliferation assays showed that both siRNAs significantly facilitated the proliferation of SCs (Fig. 6B). Furthermore, we investigated the effect of siRNAs targeting LASS2 on the migration of SCs. Notably, Transwell assay showed that siRNA-2 elicited an facilitative effect on SC migration compared with the control group, and siRNA-1 showed a more obvious effect than siRNA-2 (Fig. 6C), mainly because of a more effective knockdown of LASS2 mRNA and protein for siRNA-1. Thus, our results indicated that knockdown of LASS2 protein by siRNAs had similar effects on the SCs to those induced by miR-221/222.

Because LASS2 protein is strongly downregulated after sciatic nerve transection and inhibits SC proliferation and migration, and miR-221/222 can post-transcriptionally regulate the expression of LASS2 by directly binding to its 3′-UTR, we hypothesized that downregulation of LASS2 directly mediated miR-221/222-initiated SC proliferation and migration. To further address this issue, we downregulated miR-221/222 by transfection with miR-221/222 inhibitors with or without siRNA-1 against LASS2. We observed a significant decrease in cell proliferation and migration in groups transfected with miR-221/222 inhibitors, and, by contrast, a remarkable increase in cell proliferation (Fig. 7A) and migration (Fig. 7B) in groups co-transfected with miR-221/222 inhibitors and siRNA-1. Notably, inhibited expression of LASS2 rescued the proliferation and migration suppression induced by miR-221/222 inhibitors. Taken together, these findings show that inhibitory action of LASS2 could abrogate anti-miR-221/222-induced cell proliferation and migration suppression, suggesting that LASS2 is a functional mediator for miR-221/222 in SCs.

Peripheral nerves have a remarkable regeneration potential even after severe injury, with the possibility of some functional recovery, and the regeneration partially recapitulates developmental processes (Bosse et al., 2006). SCs are the myelinating glial cells of PNS, which support and ensheath axons with myelin to enable rapid saltatory action potential propagation in the axon. In injured nerves, however, SCs undergo dramatic phenotypic modulation, dedifferentiation and re-establishment of the proliferative and migratory capacity, and they then secrete numerous factors that control Wallerian degeneration and nerve regeneration (Jessen and Mirsky, 2005; Webber and Zochodne, 2010). There have been many studies on gene expression profiles of SCs in injured nerves to describe the molecular mechanisms underlying phenotypic modulation (Barrette et al., 2010; Bosse et al., 2006; Zickler et al., 2010). Following the recent discovery that miRNAs provide a powerful mechanism for post-transcriptional control of gene expression and participate in many biological processes, it is necessary to investigate whether injury-induced miRNAs can regulate phenotypic modulation of SCs. Using a microarray screen and a systematic analysis, we identified miRNAs with significant expression variance in the proximal stumps of nerve after sciatic nerve resection. More importantly, these miRNAs might be involved in many aspects of nerve repair and they provide an opportunity to decipher which molecules they target to regulate nerve regeneration. We further showed that miR-221/222 promote SC proliferation and migration by targeting LASS2. To our knowledge, this is the first report on a regenerative role for miRNAs in SCs.

Most investigations addressing the regenerative milieu of the injured peripheral nerves focused on the distal nerve stump, where metabolic and structural changes occur rapidly after nerve injury. Proximal stumps of transected nerves, which act as the locus for the first regenerative events, might create an important microenvironment; this has not received much attention. Thus, examining the impact of miRNAs during axonal sprouts emerging from the severed axons is a valuable goal. Here, our data show that miR-221/222 was a major mediator of this process. In situ hybridization confirmed that the upregulation of miR-221/222 primarily occurred in SCs at 4 days after injury in the proximal stumps of nerve. miR-221/222 upregulation is in agreement with the process of nerve regeneration in that proliferative activity reaches a peak in 4–11 days in the proximal stump of transected sciatic nerves (Cheng and Zochodne, 2002). Primary culture of SCs is thought to be the most accurate model for the physiology of SCs in severely injured and/or denervated peripheral nerves (Woodhoo et al., 2009). By recapitulating miRNA roles in vitro by the culture of SCs, we showed that miR-221/222 trigger a highly efficient switch in SC proliferation and migration. This switch is induced by post-transcriptionally downregulating the LASS2 protein in SCs. These findings suggest that, through LASS2, miR-221/222 induce the SC proliferation and migration responsible for formation of new nerve tissue to rejoin the distal stump.

miR-221 and miR-222 genes are clustered on chromosome Xp11.3. They have been reported to function as an oncogene by targeting the cell cycle inhibitor p27Kip1, thereby controlling cell proliferation (Galardi et al., 2007; Liu et al., 2009b; Medina et al., 2008). In addition, miR-221/222, by targeting PTEN and TIMP3 tumor suppressors, induce TNF-related apoptosis-inducing ligand resistance and enhance cellular migration through activation of the AKT pathway and metallopeptidases (Garofalo et al., 2009; Wong et al., 2010). These studies suggest that the promoting effect of cell growth or cell metastasis through miR-221/222 signaling is a general function in different tissues. In this study, a direct and functional new target of miR-221/222, LASS2, was identified. The longevity assurance homologue family members are highly conserved from yeasts to mammals (Mizutani et al., 2006). Interestingly, LASS2 is a suppressor of cell growth or metastasis and might increase intracellular H⁺ through the interaction with subunit c of vacuolar type H⁺-ATPase (V-ATPase) (Fei et al., 2004; Pan et al., 2001). The pH of intracellular compartments is a carefully controlled parameter that affects many cellular processes. The transporters responsible for controlling this crucial parameter in many intracellular compartments are the V-ATPases, which promote cell metastasis in that extracellular H⁺ involves in matrix metalloproteinase (MMP)-mediated proteolysis of extracellular matrices for cell migration and invasion (Gatenby and Gillies, 2008; Hinton et al., 2009). Importantly, MMPs, such as MMP-9, control crucial trophic signal transduction pathways and phenotypic remodeling of SCs (Chattopadhyay and Shubayev, 2009; Mantuano et al., 2008). Consequently, after nerve injury, SCs might upregulate miR-221/222, and then inhibit production of LASS2 protein to promote the discharge of H⁺ through V-ATPase to initiate injury-induced cellular modulation. Additional functional assays mediated by LASS2 in SCs will contribute towards a better understanding of miR-221/222 signaling pathway in sciatic nerve repair.

Several independent studies have established that miRNAs play a key role in SC differentiation, and each study indicates that the ablation of Dicer1 leads to SC overproliferation and aberrant myelination (Pereira et al., 2010; Yun et al., 2010). We now identify the role of miRNAs in wound-repair-initiating nerve reconstruction by orchestrating directed SC proliferation and migration. More importantly, the regulation of SC proliferation and migration is crucial for myelination. The major players that regulate SC dedifferentiation or proliferation and remyelination during peripheral regeneration are extracellular matrix proteins, neurotrophic factors, and hormones (Chen et al., 2007). In injury to the PNS, SCs migrate toward axonal signals, where they differentiate by ensheathing and/or myelinating axons and form a framework for nerve regeneration (Mantuano et al., 2008). Thus, understanding the function of SC miRNAs, as well as the major players discussed above is an important goal for studying nerve regeneration. Here, we show that miR-221/222 is, at least, one of major contributors to SC proliferation and migration and subsequent myelination for nerve reconstruction.

In conclusion, our findings indicated that trauma significantly and promptly increase miR-221/222 expression, which decreases the level of LASS2 protein and causes SC phenotypic modulation, suggesting that miRNAs might act as a master switch to modulate various post-transcriptional processes in SCs. These results demonstrate that the injury-induced miRNA signaling cascades are important for understanding the molecular mechanisms responsible for peripheral nerve regeneration.

Thirty adult, male Sprague-Dawley (SD) rats (180–220 g) were randomly divided into five groups of six rats each. Each animal was anaesthetized by an intraperitoneal injection of complex narcotics, and the sciatic nerve was exposed and lifted through an incision on the lateral aspect of the mid-thigh of the left hind limb. A 1-cm-long segment of sciatic nerve was then resected at the site just proximal to the division of tibial and common peroneal nerves, and the incision sites were then closed. To minimize the discomfort and possible painful mechanical stimulation, the rats were housed in large cages with sawdust bedding after the surgery. The proximal stumps of the sciatic nerve (0.5 cm) were collected at 0, 1, 4, 7 and 14 days after injury. The experiment was repeated three times. All the experimental procedures involving animals were conducted in accordance with Institutional Animal Care guidelines and approved ethically by the Administration Committee of Experimental Animals, Jiangsu Province, China.

Total RNA was extracted using the mirVana miRNA Isolation Kit (Ambion, Austin, TX) according to the manufacturer's instructions. The quality of the purified RNA was assessed using a BioAnalyzer 2100 (Agilent Technologies, Santa Clara, CA). The purified RNA was quantified by determining the absorbance at 260 nm using a Nanodrop ND-1000 spectrophotometer (Infinigen Biotechnology Inc., City of Industry, CA). A miRNA microarray (Agilent Technology), containing probes for the complete Sanger miRBase 10.0, was used to screen RNA from rat sciatic nerve of different groups. The labeling and hybridization were performed at the Shanghai Biochip Company (Shanghai, China), according to the protocols in the Agilent miRNA microarray system. Agilent Scan Control software was used for scanning the microarray slides, and Agilent Feature Extraction software version 9.5.3 was used for image analysis. Microarray data were analyzed using GeneSpring GX v11.0 software (Agilent Technology).

miRNA and mRNA expression profiles were scanned by microarray after segmentation of the peripheral nerve. Two groups of data were compared, the significance and false discovery rate were calculated using the adjusted F-test with the Random Variation Model (Wright and Simon, 2003), and the differentially expressed genes in these serial time points were obtained via screening.

Using the miRNAs with significant expression variance, we conducted the following analysis: (1) Hierarchical clustering with the expression of these miRNAs. We calculated the Z-score from the expression of miRNA. The euclidean distance measure was used to compute the distance (dissimilarity) in two ways (miRNA and time). (2) miRNA expression pattern clustering. The significance analysis of expression tendency was screened for differentially expressed miRNAs (Ramoni et al., 2002). (3) Search for putative targets of miRNA in specific profile using miRBase database, and integrating putative miRNA targets with differentially expressed mRNA yielding potential targets. (4) GO and KEGG enrichment analyses for the integrated targets. (5) Network construction for miRNAs and the targets.

We then integrated our specific miRNAs and serial analysis of gene expression data and selected only the miRNA–target pairs that were inversely correlated. We calculated the Pearson correlation coefficient and P value between miRNA and mRNA. The mRNA was thought to be downregulated by the miRNA when P<0.1.

SCs were isolated from the sciatic nerves of 1-day-old SD rats and further selected from fibroblasts using anti-Thy1.1 antibody, as previously described (Mantuano et al., 2008). The final preparations consisted of 98% SCs, as determined by immunofluorescence for S100, which is a specific SC marker. Primary cultures of SCs were maintained in DMEM containing 10% fetal bovine serum (complete medium) at 37°C under humidified 5% CO₂. SC cultures were passaged no more than three times before conducting experiments. SCs were transfected with miRNA mimics, miRNA inhibitors or siRNAs (Ribobio, Guangzhou, China) respectively using Lipofectamine RNAiMAX transfection reagent (Invitrogen, Carlsbad, CA), according to the manufacturer's instructions. The sequences of siRNA duplexes are listed in supplementary material Table S3.

Reverse-transcribed complementary DNA was synthesized with the Prime-Script RT reagent Kit (TaKaRa, Dalian, China). PCR was performed with SYBR Premix Ex Taq (TaKaRa). For miRNA detection, mature miR-221/222 was reverse-transcribed with specific RT primers, quantified with a TaqMan probe, and normalized by RNU6B mature miRNA using TaqMan miRNA assays (Applied Biosystems, Foster City, CA). The relative expression level was calculated using the comparative 2^−ΔCt method.

In situ hybridization was performed using the miRCURY LNA microRNA ISH Optimization Kit (Exiqon, VedBaek, Denmark) according to the manufacturer's instructions. Each section was treated with 3 μg/ml proteinase K for 20 minutes at 37°C. After treatment with 0.2% glycine-PBS for 5 minutes, sections were washed twice in PBS for 5 minutes each and acetylated with 0.25% acetic anhydride in 0.1 M triethanolamine hydrochloride for 10 minutes. Hybridization with DIG-labeled probes was carried out for 2 hours at 55°C in hybridization buffer. After hybridization, sections were washed in 5× SSC for 5 minutes at 55°C, 1× SSC twice for 5 minutes at 55°C, 0.2×SSC twice for 5 minutes at 55°C, and 0.2×SSC for 5 minutes at room temperature. Blocking was performed for 2 hours at room temperature with alkaline phosphatase-conjugated Fab anti-DIG antibody (Roche, Mannheim, Germany) in 2% sheep serum. The slides were stained using 5-bromo-4-chloro-3-indolyl-phosphate and Nitroblue tetrazolium (Roche), and then counter stained with Nuclear Fast Red (Vector Labs, Burlingame, CA).

Protein extracts were prepared from cell cultures or proximal stumps of the nerve. Equal amounts of protein were subjected to SDS-PAGE and electrotransferred to nitrocellulose membrane (Bio-Rad, Hercules, CA). The membrane was blocked with 5% non-fat dry milk in Tris-HCl buffered saline, pH 7.4 with Tween-20 and incubated with the primary antibody to detect LASS2 (Santa Cruz Biotechnology, Santa Cruz, CA), according to the manufacturer's recommendations. Antibody binding was detected by HRP-conjugated species-specific secondary antibody followed by enhanced chemiluminescence (Pierce Chemical Company, Rockford, IL).

SCs were resuspended in fresh pre-warmed (37°C) complete medium, counted and plated at a density of 2×10⁵ cells/ml on 0.01% poly-L-lysine-coated 96-well plates. At the indicated time point after cell transfection, 50 μM EdU was applied to the cultures and the cells were grown for an additional 2 hours. Finally, the cells were fixed with 4% formaldehyde in PBS for 30 minutes. After labeling, the SCs were assayed using Cell-Light EdU DNA Cell Proliferation Kit (Ribobio) according to the manufacturer's protocol. Analysis of SC proliferation (ratio of EdU⁺ to all SCs) was performed using images of randomly selected fields obtained on a DMR fluorescence microscope (Leica Microsystems, Bensheim, Germany). Assays were performed three times using triplicate wells.

Migration of SCs was studied using 6.5 mm Transwell chambers with 8 μm pores (Costar, Cambridge, MA) as described previously (Mantuano et al., 2008). The bottom surface of each membrane was coated with 10 μg/ml fibronectin. 100 μL SCs (10⁶ cells/ml) resuspended in DMEM were transferred to the top chambers of each Transwell and allowed to migrate at 37°C in 5% CO₂, and 600 μl of complete medium was injected into the lower chambers. The upper surface of each membrane was cleaned with a cotton swab at the indicated time point. Cells adhering to the bottom surface of each membrane were stained with 0.1% Crystal Violet, imaged, and counted using a DMR inverted microscope (Leica Microsystems). Assays were performed three times using triplicate wells.

The 3′-UTR sequence of LASS2 was amplified from the genomic DNA and subcloned into the region directly downstream of the stop codon in the luciferase gene in the luciferase reporter vector. With appropriate primers, PCR amplification of the 3′-UTR sequence of LASS2 generated different p-Luc-UTR luciferase reporter vectors. The sequences of wild-type and mutant 3′-UTRs were confirmed by sequencing. HEK293T cells were seeded in 96-well plates and transfected with a mixture of 30 ng p-Luc-UTR, 5 pmol miRNA mimics, and 5 ng Renilla following the recommended protocol for the Lipofectamine 2000 transfection system (Invitrogen). After 48 hours of incubation, firefly and Renilla luciferase activities were measured using the dual-luciferase reporter assay system (Promega, Madison, WI) from the cell lysates.

Statistical analyses were carried out using SPSS 15.0 for windows (SPSS, Chicago, IL). The Student's t-test was used for comparison between groups. P<0.05 was considered statistically significant. All data are expressed as means ± s.d.

We thank Jie Liu for his help in revision of the manuscript.