Motor function recovery during peripheral nerve multiple regeneration

Authors


Abstract

Neuronal functional compensation and multiple regenerating axon sprouting occur during peripheral nerve regeneration. Sprouting nerve buds were quantitatively maintained and had matured when multiple injured distal nerves were anastomosed to smaller number of proximal nerve stumps; this has positive clinical significance for proximal stump damage. This study investigated whether sprouting axon buds would reinnervate the distal neuromuscular junction and maintain the function of the target organ under compensation conditions. The results showed that the sprouting axon buds maintained the numbers and morphology of motor end plates repaired by a smaller number of proximal nerve stumps, and recovered 80.0% tetanic muscle force compared with the normal side. Meanwhile, nerve conduction velocity, compound muscle action potential and diameter of muscular fibres declined 72.7%, 73.2% and 61.8%, respectively, compared with normal. This observation indicates the potential functional reserve of neurons and that it is feasible to repair nerve fibre injury through anastomosis of multiple distal nerve stumps with a smaller number of proximal nerve stumps, within the limits of compensation. Copyright © 2013 John Wiley & Sons, Ltd.

1 Introduction

Peripheral nerve injury is common in clinical practice. Mangled nerve injury and avulsion often result in significant damage to proximal nerves, which render the nerve repair difficult (Lundborg et al., 1994; Allodi et al., 2012). Current repair methods, such as nerve transfer or nerve implantation, usually need sacrifice of another normal nerve as the donor nerve. Previous studies have revealed that regeneration and compensation occurs in the regeneration of peripheral nerves (Redett et al., 2005; Jiang et al., 2007; Wang et al., 2009). Furthermore, these studies have provided preliminary evidence for the maturation of sprouting nerve fibres (Redett et al., 2005; Yin et al., 2011). These findings provide a possible novel strategy for repairing serious damage to proximal nerve stumps through direct anastomosis to smaller number of proximal nerve stumps. This would avoid injury to donor nerve and not have the disadvantages of allograft transplant reactions. However, the recovery outcome has many contributing factors. Correlations have been observed between the functionality of repaired nerves and the maturity of nerve fibres, the quantity of motor end plates and the thickness of muscle fibres (Wood et al., 2011).

In order to explore the possibilities of this repair method as a strategy, it should be considered from several aspects, such as nerve fibres, neuromuscular junctions and muscle fibres. A rat tibial nerve model was adopted in this study to further examine whether collateral axon buds could completely re-establish their control of motor end-plates and maintain the diameter and function of the original muscle fibres when multiple injured distal nerves were repaired through anastomosis to a smaller number of proximal nerve stumps.

2 Materials and methods

2.1 Animals

Experiments were performed using 36 specific pathogen-free (SPF) female Sprague–Dawley rats with a body weight of 200 g. The animals were randomly divided into three experimental groups, groups A, B and C, for the subsequent surgical treatments. Experimental procedures were reviewed and approved by the Ethics Committee of the People's Hospital, Peking University.

2.2 Materials

Biodegradable chitin conduits (People's Hospital of Peking University and Chinese Textile Academy, Patent Number 01136314.2) with a length of 8 mm, an inner diameter of 1.5 mm and a wall thickness of 0.1 mm were used.

2.3 Animal model preparation

The surgery was performed under a microscope. The rats were anaesthetized through intraperitoneal (i.p.) injection of sodium pentobarbital (30 mg/kg). Following complete anaesthesia, skin preparation and disinfection were carried out in the right hind limb. The right sciatic nerve and its two main branches (common peroneal nerve and tibial nerve) were isolated until fully exposed. The tibia nerve was severed at 5 mm below the bifurcation site of the sciatic nerve. The animals in each group (n = 12) were treated with the following procedures (Figure 1):

Figure 1.

(a) Proximal tibial nerve and distal tibial nerve at a nerve fibre ratio of 1:1; repairing the nerve through anastomosis with small-gap conduits. (b) Proximal common peroneal nerve and distal tibial nerve at a nerve fibre ratio of about 1:3 to 1:2; repairing the nerve through anastomosis with small-gap conduits. (c) The two stumps of the severed tibial nerve were sutured to the muscle in opposite directions to avoid self-repair, which led to denervation of the gastrocnemius. Tp, proximal tibial nerve stump; Td, distal tibial nerve stump; CPp, proximal common peroneal nerve stump

Group A: The severed tibial nerve was repaired through anastomosis with the small-gap conduits. The distance between stumps was 2 mm.

Group B: The common peroneal nerve was severed at the same level, and the proximal stump was anastomosed to the distal stump of the severed tibial nerve through a conduit (with a gap of 2 mm). The other two stumps were sutured to the muscle in opposite directions to avoid self-repair.

Group C: The tibial nerve was severed, and the resulting two stumps were sutured to the muscle in opposite directions to avoid self-repair.

Nerve anastomosis through conduits was carried out with 10–0 nylon suture, and the incision was subsequently closed with 4–0 suture.

2.4 Observed parameters

2.4.1 Gross morphology and behavioural observation

The animals (n = 36) were observed to evaluate wound healing, muscle morphology of the hind limb, and behavioural changes at different time-points (weeks 4, 6, 8 and 12).

2.4.2 Neuroelectrophysiological examination

A Medlec Synergy electrophysiological system (Oxford Instrument Inc., Oxford, UK) was used for the examination. The repaired sciatic nerve was exposed at week 12 after the surgery. The stimulating electrodes were placed on the distal and proximal nerve trunks, on the anastomotic plane in the sciatic nerve, while the recording electrode was inserted into the middle of gastrocnemius; the reference electrode was placed in the thigh muscle on the same side. Paraffin oil was applied around the nerve trunk to reduce bypass conduction through the liquid. The stimulation signal was a square wave, with an intensity of 0.9 mA, a wave width of 0.1 ms and a frequency of 1 Hz. The conduction velocity of the regenerated nerve fibres was recorded by measuring the latent period. The stimulation intensity was gradually strengthened until the amplitude of the compound muscle action potential (CMAP) wave ceased to progressively increase and a generally identical shape for the CMAP wave was formed from the stimulation at both the distal and proximal stumps. The amplitude of the distal CMAP was recorded, which was the distance from the initiation point to the negative peak of the wave.

2.4.3 Tetanic muscle contraction strength

The rats were fully anaesthetized before sample collection at week 12 after surgery; this involved dissection and isolation of the gastrocnemius. The hind limb was fixed on a specially made holding frame, with the distal end of the gastrocnemius connected to a tension sensor and then using the holding frame kept the gastrocnemius and the tension sensor aligned. The initial tension was maintained at a chosen level (0 < F < 0.1 N). An electrophysiological system was used to generate an initial electric stimulation with an intensity of 0.9 mA, a wavelength of 0.1 ms and a frequency of 1 Hz. The stimulation electric current was subsequently strengthened until the waveform of the tetanic contraction induced stopped increasing. A PCLAB-UE biomedical signal acquisition and processing system (Beijing Microsignal star Inc., Beijing, China) was used to record the waveform of the tetanic contraction of the gastrocnemius on both sides. The amplitudes of the waves were measured and the ratio of the wave amplitudes of experimental side to the untreated normal control side was used as the overall recovery rate of muscle strength (Shin et al., 2008).

2.4.4 Wet muscle weight measurement and diameter of muscle fibres by haematoxylin and eosin (H&E) staining

The gastrocnemius was isolated by severing it at its starting and ending point immediately after the aforementioned parameters were measured, and the weight of the muscle was measured. Transverse sectioning of the muscle samples was performed for H&E staining after fixing with paraformaldehyde, dehydrating with graded ethanol and embedding in paraffin wax. The cross-sections of the muscle fibres were photographed under a magnification of 10 × 20 and five fields were selected in the upper left, lower left, upper right, lower right and centre of the cross-section of the muscle fibres for quantification of the diameter of muscular fibres in each field and for measurements using image pro plus 6.0 software (Media Cybernetics Inc., Rockville, MD, USA).

2.4.5 Osmium tetroxide staining of the tibial nerve and quantification of nerve fibres

Twelve samples of each group were post-fixed in 1% osmium tetroxide for 1 day, after which the specimen was sliced into 2 µm cross-sections. The cross-section of the nerve was photographed under a magnification of 10 × 20 and five fields were selected in the upper left, lower left, upper right, lower right and centre of the nerve for quantification of the myelinated nerve fibres in each field and the measurement of the area of each field using a combination of manual measurements and measurements using image pro plus 6.0 software. The numbers of myelinated nerve fibres in each field were calculated manually and the area of the field and total area were measured using image pro plus. The average number of myelinated nerve fibres per unit area were then calculated. The total number of myelinated nerve fibres (N) = the number of myelinated nerve fibres per unit area (n/ds) × area of the cross-section(s). The total number of myelinated nerve fibres was calculated using this above equation.

2.4.6 Immunohistochemical staining of motor end plates

Using the cupric-ferricyanide staining method developed by Kamovsky and Roots, acetylthiocholine iodide was added to freshly prepared incubation buffer as the enzyme's substrate, which was hydrolysed into thiocholine by acetylcholinesterase (AChE) in tissue (Karnovsky and Roots 1964). Thiocholine reduced the ferricyanide in the incubation buffer into ferrocyanide and this reacted with copper ions to form cupric ferrocyanide, which was deposited as a brown precipitate at sites with AChE activity. The entire muscle was divided into three equal parts along its longitudinal axis, and consecutive sagittal cryosectioning (10 µm) was performed. One section was collected every 100 sections for staining. The selected sections were washed with phosphate-buffered saline (PBS) three times for 20 min each time, followed by incubation in Kamovsky–Roots (KR) solution for 6 h. The sections were then washed with PBS three more times for 20 min each (Ma et al., 2002). The sections were finally mounted and observed for quantification of the motor end plates under a 10 × 20 light microscope. The total number of motor end-plates (N) = the sum of motor end-plates per section (n1 + n2 + …… + n) × 100.

2.4.7 Immunofluorescence staining of acetylcholine receptors

Sagittal cryosectioning was performed with the muscle specimen, and one section was collected every 100 sections. The sections were 20 µm thick. After being air dried, the sections were washed with PBS three times for 20 min each time and blocked with serum. The sections then underwent specific staining of the acetylcholine receptor using tetramethylrhodamine-labelled α-bungarotoxin (T-BTX). The T-BTX stock solution (1 mg/ml in PBS) was diluted 1:400. The sections were incubated with diluted T-BTX at 4°C overnight (12 h) and then were washed with PBS three times for 20 min each time. The sections were mounted with anti-quenching mounting media. The morphology of the motor end plates was observed and recorded under a magnification of 10 × 40 with fluorescence excitation at 620 nm (Ma et al., 2002; Magill et al., 2007). Five fields were observed in the upper left, lower left, upper right, lower right and centre areas for each section, and the perimeter and area of a motor end plate were measured by drawing the maximal smooth perimeter around the image of each endplate according to their grey level and then calculated under image scale using image-pro plus automatically.

2.5 Statistical analysis

Statistical analyses was performed using SPSS 11.0 (SPSS Inc., Chicago, IL, USA). The measurement data are expressed as mean ± SD. An independent t-test was adopted for two-group comparison, and analysis of variance ( anova) was used for multi-group comparison. Comparison between the groups was made by analysing data with a post-hoc method Student-Newman-Keuls (S-N-K). The S-N-K test is a pairwise comparison of every combination of group pairs. This test calculates a q test statistic for each pair, and displays the P value for that comparison. Enumeration data were analysed using a chi-square test. Statistical significance was established as p < 0.05.

3 Results

3.1 General observation of the animals

Claudication of the right hind limb, and gradual muscular atrophy in the lower leg were observed in the right hind limb in each group after the surgery. At 10 days after repair, the rats in all groups exhibited discoloration of the toenails, sparse and dull hair coat on the lower leg and foot and an insensitivity to pain stimuli on the experimental side. Foot ulcers were observed at 6 weeks postoperatively in the untreated group (n = 2), without significant improvement of gait, while the other two groups displayed nearly normal gait and gradual improvement coordination.

3.2 Quantification of nerve fibres and electrophysiological examination

The number of normal tibial nerve fibres was 3313 ± 204 at 12 weeks after the surgery. The numbers of proximal and distal nerve fibres were 3384 ± 273 and 3197 ± 242, respectively, in group A and 1159 ± 151 and 2909 ± 189, respectively, in group B. The distal nerve fibres in group C showed evident degeneration, and some even disappeared (Table 1). No statistically significant difference was observed in the number of distal nerve fibres between group A and group B.

Table 1. Comparison of Myelinated axon numbers, MCV, Peak of CAMP, muscle force and diameter of muscular fibers for all group
 Myelinated axon numbers    
GroupProximalDistalMCV (m/s)Peak of CAMP (%)Muscle force (%)Diameter of muscular fibers (µm)
  • * p<0.05, with significant difference.
Normal3313±20449.2±7.1NANA31.9±5.6
Group A3384±2733197±24243.4±11.788.4±5.693.7±21.421.2±6.9*
Group B1159±151*2909±18936.3±8.1*73.2±18.980.0±10.119.7±6.4*
Group CNANANANANA13.6±4.8*

Electrophysiological examination was performed with fully anaesthetized rats. The nerve conduction velocity of a normal tibial nerve was 49.9 ± 7.1 m/s. The nerve conduction velocity was 43.4 ± 11.7 m/s and 36.3 ± 8.1 m/s in group A and group B, respectively. The nerve conduction velocity in group B was lower than that of the intact nerve; the difference was statistically significant. The compound muscle action potential indicated that the wave amplitude of the experimental side was 88.4 ± 5.6% and 73.2 ± 18.9% of the contralateral side in group A and group B, respectively, but this was not statistically significant.

3.3 Muscle morphology and strength

Haematoxylin and eosin staining showed that the cross-sectional diameter of a normal muscle fibre was approximately 31.9 ± 5.6 µm, with distinct borders and uniform staining (Figure 2). The muscle fibres of group C displayed marked atrophy caused by prolonged denervation, with indistinct borders and uneven staining. The muscle fibre diameters of the three groups were 21.2 ± 6.9 µm, 19.7 ± 6.4 µm and 13.6 ± 4.8 µm for groups A, B and C, respectively. The muscle fibre diameter of the untreated normal control group was greate than those of the experimental three groups; that of group C was significantly lower than those of the other two groups.

Figure 2.

(a) Cross-section of normal muscle, showing distinct borders, uniform staining and large diameter. (b,c) Distinct borders and uniform staining, but with shortened diameter, after anastomosis with small-gap conduits and nerve repair in group B. (d) Indistinct borders, uneven staining and reduced diameter in group C

There were no significant differences among the body weights of rats of all groups before the surgery and there were also no significant differences in the body weights of the rats among all groups 4 weeks after the surgery. At 12 weeks after the surgery, the body weights of the rats in group C were slightly greater than those of the other two groups, but this was not significantly different. The gastrocnemius was collected for wet weight measurement after the rats were euthanized. The wet weight ratios of the experimental side to the normal side were 64.75 ± 13.5%, 54.66 ± 12.5% and 28.27 ± 10.9% in groups A, B, and C, respectively; the ratio of group C was significantly lower than those of the other two groups.

Under anaesthesia, the ratio of the tetanic contractility of the experimental side to the contralateral side was 93.7 ± 21.4% and 80.0 ± 10.1% for groups A and B, respectively. The muscles from group C did not show significant tetanic contraction when the reverse sutured distal nerve was stimulated.

3.4 Morphology and quantity of neuromuscular junctions

After the staining of AChE, the normal neuromuscular junction appeared as spherical or clostridial form with a brown colour and the centre was shallow while the peripheral region was saturated (Figure 3). The numbers of motor end-plates in the gastrocnemius nerve fibres of the normal control and the three experimental groups (A, B and C) were 27400 ± 6698, 22950 ± 8817, 20433 ± 7187 and 9283 ± 1653, respectively; the numbers in group C were significantly lower than those of the other groups.

Figure 3.

The staining of acetylcholinesterase (AChE) 12 weeks postoperatively (×20 objective lens): (a) normal group, (b) group A, (c) group B, (d) group C. The normal end plate appeared as spherical or clostridial form with a brown colour and the centre of the end plate was shallow while the peripheral region was saturated. End plates in group B were observed with nearly normal shape and numbers, which is better than the condition of end plates in group C (denervated group) (p < 0.001) and similar to the end plates in group A (p = 0.600)

To further evaluate morphometry of the motor end plates in the postsynaptic membrane, specific staining of the acetylcholine receptor was performed, and the perimeter and area of the staining parts were measured (Figure 4). No significant differences were observed in the area and perimeter of the stained motor end plates between the normal control and group A with group B. In contrast, group C showed a significantly reduced perimeter compared with the other groups (Figure 5).

Figure 4.

Immunofluorescence staining of motor end plates: (a) normal group, (b) group A, (c) group B, (d) group C, (e) single end plate, (f) the outline of end plate according to grey level. The perimeter and area of a motor end plate were measured by drawing the maximal smooth perimeter around the image of each endplate according to their grey level and the area was calculated from the image scale automatically using image-pro plus software

Figure 5.

(a) Comparison of the number of motor end plates. (b) Comparison of the area of motor end plates. (c) Comparison of the perimeter of motor end plates. The denervated group exhibited a significantly reduced number and perimeter of motor end plates than the other two experimental groups and the normal control. Asterisk indicates a significant difference at p < 0.05

4 Discussion

Injuries to the peripheral nerves result in partial or total loss of motor, sensory functions in the denervated segments of the body because of the interruption of axons, degeneration of distal nerve fibres, and eventual death of axotomized neurons. Functional deficits caused by nerve injuries can be compensated by reinnervation of denervated targets by regenerating injured axons buds. This phenomenon utilizes the functional reserve of peripheral nerve, the essence of which is the result of extensive reinnervation by functional nerve fibres after the cross-mixing of polyneurons in the proximal sites of peripheral nerve. Thus, exploring the potential of axon buds will be one of the essential points in the process of peripheral regeneration.

The multiple regenerating axons sprouting exist in the process of peripheral nerve regeneration. Previous studies indicated that the amount of regenerating axon buds will obviously be greater than the amount of proximal nerve fibres after transected peripheral nerve injuries. It is broadly accepted that most of these regenerating axon buds could not extend into distal tubules of the basement membrane and instead turn into neuromas or ultimately degenerate. The possible reasons for the failed extension include the mechanical blockade of scar tissue, interference from soft tissues and other factors, and incorrect growth direction of the axon buds (Morris et al., 1972; Ito and Kudo, 1994; Skouras et al., 2011). Recently, some studies have shown that the number of regenerating axon buds grown into distal stumps could maintain about two to three times the number of the proximal fibre. Furthermore, these axon buds could become mature sprouting nerve fibres, which are considered as multiple regenerations of axon sprouting (Redett et al., 2005; Jiang et al., 2007). Therefore, for the proximal damaged nerve, it is possible to repair multiple injured distal nerves through anastomosis to smaller number proximal nerve stumps. However, under these conditions, whether collateral axon buds could utilize their potential functional reserve and compensate for the original function of muscle fibres is unclear.

This study investigated the changes in motor end plates and muscle function after the repair of multi-nerve injuries through anastomosis with a smaller number of nerve stumps. According to the experimental results, at a certain ratio (35.0%) of distal and proximal nerve stumps in anastomosis repair, the number of distal axon buds could reach 87.8% of that of normal tibial fibres, the amount and morphology of motor end plates could be basically maintained and the tetanic force of muscle fibres could be 80.0% of the force of the normal untreated side. It is concluded that the sprouting axon buds of peripheral nerves can extend into distal stumps, grow to the neuromuscular junction and control the connected muscles to achieve the recovery of motor function.

In the current study, more than one lateral bud sprouted from a single neuron and was retained to reach the neuromuscular junction and play the role of compensation under the condition of less proximal fibre amount. From the results, the amount of proximal nerve fibres in group B is obviously less than that of group A and the untreated normal control group, but the amount of distal stumps in group B is close to the latter. This phenomenon is likely related to the number of the tubules of the basement membrane at the distal nerve stump (Jiang et al., 2007; Moradzadeh et al., 2008; Nichols et al., 2004). As the tubules of the basement membrane are more than merely proximal nerve fibres, the multiple regenerating axon buds could grow into distal tubules and be surrounded by proliferative Schwann cells in the process of bidirectional myelinisation. In addition, nerve regeneration chambers formed in the biological conduits provide spaces for selective growth. In contrast, traditional epineurium anastomosis lacks space for selective growth and, considering directional changes of nerve stumps during surgery, regenerated axons often grow into soft tissue between nerve bundles instead of the end organ, which might be partly responsible for poor functional recovery.

In the present study, the motor end plates were counted using AChE staining. A neuromuscular junction is composed of three components: nerve endings, muscular fibres and Schwann cells. A motor end plate is actually referred to as the postsynaptic membrane of a neuromuscular junction, which is located on a muscular fibre and faces the nerve ending (Ma et al., 2002). Acetylcholinesterase staining is an effective method for locating motor end plates. The numbers of motor end plates in group B and group A were compared with that of normal rats, and the differences were not statistically significant. However, the number of motor end plates in group C was significantly reduced 12 weeks after surgery. These results indicate that the two treatment methods maintained the number of motor end plates. In previous studies, reduced activity of motor end plates was determined based on lighter AChE staining, the absence of staining at the centre of the motor end plates and vacuolar changes. However, both the morphology and the intensity of AChE staining are affected by a variety of factors such as staining time and temperature and the storage time of the staining solution. Thus, AChE staining can only be used for qualitative analysis. Accordingly, this study adopted specific fluorescence staining of the acetylcholine receptor for the semi-quantification of morphological changes in motor end plates (Deschenes et al., 2003, 2006, 2010). After measuring the perimeter and area of motor end plates, no significant changes were found in the motor end plate areas of the three groups compared with motor end plates in normal muscles, suggesting that the activity of the motor end plates was maintained at a normal level. However, with prolonged denervation, the perimeter of the motor end plates in group C showed significant reduction, indicating a trend towards gradual atrophy.

In addition, considering the slight reduction in the numbers of nerve fibres and motor end plates, as well as in muscle size and strength, the electrophysiological parameters and muscle cross-sectional diameter were significantly lower in group B than in those of normal muscle; this may be related to the decreased activity of the axon buds. Previous studies have shown that during the process of motor end plate reinnervation, neurons could give rise to collaterals, resulting in a single motor end plate being controlled by multiple nerve endings (Keller-Peck et al., 2001; Buffelli et al., 2003; Livet et al., 2007; Magill et al., 2007). In the process of multiple collaterals competing for a motor end plate, the collaterals exhibited varied activity and competitiveness. In general, with fewer collateral branches from same neuron and a larger diameter, the collateral will show higher activity for material transport and neurotransmitter release (Kasthuri and Lichtman, 2003). In this study, the use of a smaller number of nerve stumps to repair a larger number of injured nerve fibres was evaluated. The potential functional reserve of neurons was explored and the activity of each axon bud was found to significantly reduce. Therefore, although the number and morphology of motor end plates were maintained, the electrophysiological parameters and muscle fibre diameter were decreased.

This phenomenon of compensation and multiple regeneration was also reported in previous studies of end-to-side neurorrhaphy and nerve transposition. Studies have proven that donor nerves could grow into the recipient nerve through end-to-side neurorrhaphy, suggesting that the nerve axon buds have the capability of growth and maturation (Kostakoglu, 1999; Okajima and Terzis, 2000; Al-Qattan, 2001; Lowe et al., 2002; Geuna et al., 2006). Similar reports of using thin nerves to repair some essential thick nerves were also found in studies of nerve transposition in which the function of the nerves for muscle control was greatly recovered (Rohde and Wolfe, 2007, Schalow et al., 1992). Although these evaluations use clinical functional scores, their results are fundamentally consistent with our findings. The results of this study demonstrate that when the proportion of anastomosis is in the range one-half to one-third of the nerve fibres at the proximal stumps and at the distal stumps in nerve anastomosis, the repair might achieve a satisfactory functional recovery. Therefore, it is suggested that a small number of nerve fibres can be separated from the adjacent stump of nerve trunk and used as donor nerve fibres. This will not only maintain the original function of the adjacent nerve to a certain degree, but it will also restore the function of the region controlled by the injured nerve. This method may become a new strategy for the repair of high-level nerve injuries.

Some limitations of this study should be noted. In this study, the observations are in specific special animal models, and therefore considering topographic specificity and species differences during peripheral nerve regeneration, the results should be further confirmed in other models. Further, owing to the differences and variations in nerve position and distribution of regeneration, although the operation and observation was completed according to unified standard methods, the result was inevitably influenced by subjective cognizance and system bias. Whether muscle function can be further restored through neuronal compensation, or whether muscle function will decline with long-term compensation, requires further study.

Conflict of interest

The authors declare that there are no conflicts of interest.

Acknowledgements

The authors thank Zhang Hong-Bo, Li Qing-Tian, Xu Chun-Gui and Chen Xu-Hong for their kind help and Dong Jian-Qiang for guidance. This research project was funded by the Chinese National Natural Science Fund for Outstanding Youth (30625036), the Chinese 973 Project Planning (2005CB522604), the Chinese National Natural Science Youth Fund (30801169), the Beijing City Science & Technology New Star Classification A-2008-10, the Chinese National Natural Science Fund (31171150, 81171146, 30971526, 31040043) and the Chinese Educational Ministry New Century Excellent Talent Support Project (2011).

Ancillary