Qing Li, Lunjing Du, Dongju Feng, Yun Ren, Zhexin Li, Fnlei Kong,Jicho Yun,*

aChongqing Key Laboratory of Economic Plant Biotechnology, Collaborative Innovation Center of Special Plant Industry in Chongqing, Institute of Special Plants, Chongqing University of Arts and Sciences, Yongchuan, 402160 Chongqing, China

bKey Laboratory of Crop Ecophysiology and Farming System in Southwest China, Ministry of Agriculture, Sichuan Agricultural University,Chengdu 611130, Sichuan, China

Keywords:Maize Grain filling Nitrogen management Yield Population filling rate

A B S T R A C T To investigate the effect of nitrogen management on the grain-filling characteristics and yield formation of maize cultivars with contrasting nitrogen efficiencies, and to identify differences in grain-filling characteristics and yield of maize cultivars in response to nitrogen management, a two-year field experiment was conducted in southwest China in 2015–2016. The grain-filling rate and duration of the N-inefficient cultivar XY 508 were higher than those of the N-efficient cultivar ZH 311. The 100-kernel weight of XY 508 was significantly higher than that of ZH 311. The kernel number per ear of ZH 311 was significantly higher than that of XY 508, making the population filling rate of ZH 311 significantly higher than that of XY 508.The higher population filling rate of the N-efficient maize cultivar led to a significant yield advantage over the N-inefficient maize cultivar.Nitrogen management effectively improved maize grain yield, but the response of maize cultivars with contrasting nitrogen efficiencies to nitrogen management was inconsistent.A basal fertilizer ratio 60.43% with a topdressing ratio 39.57% effectively increased grainfilling rate,delayed the time to maximum filling rate,prolonged the active filling period and effective grain-filling time,increased the 100-kernel weight,and maintained higher kernels per ear, thereby improving the population filling rate and maximizing the yield advantage of the N-efficient cultivar. A 100% basal fertilizer ratio not only increased the number of kernels per ear,but also maintained high grain filling characteristics to obtain a higher 100-kernel weight and increased the population filling rate, leading to a high grain yield in the N-inefficient cultivar. Thus, the 100% basal fertilizer ratio partially compensated for the deficient grain yield of the N-inefficient cultivar.

1.Introduction

Maize (Zea mays L.) is the world’s largest food crop. Total maize production exceeds 1 billion t,accounting for 41%of the world’s total grain production [1,2]. Maize production is vital to global food security [3]. Nitrogen fertilizer input is the simplest and most effective measure to increase maize yield[4,5]. However, the detrimental effect of excessive nitrogen application in maize production in China is severe[6],and not only leads to a decline in yield quality (owing to increased levels of pests and diseases) but increases nitrogen loss,production costs, and environmental pollution [7,8]. Increasing maize yield while reducing nitrogen fertilizer application will help to solve the food crisis and ensure environmental safety [9,10]. Nitrogen-efficient maize cultivar selection is a path to increasing maize yield and reducing nitrogen fertilizer overuse[11,12].

Grain filling is an important period in maize growth and development [13–15]. Grain filling rate and grain filling time affect the filling degree of grain storage capacity, which determines the quality and yield of maize [16–18]. Li et al.[16] found that integration of water and fertilizer application prolonged the filling period by 4.4 days and increased the mean filling rate by 19.5%,and that filling time and filling rate were significantly positively correlated with maize grain yield.Wang et al. [19] showed that the maximum grain-filling rate and grain filling duration of the maize grain-filling process increased significantly under the combined treatment of controlled-release urea and common urea compared to that of common urea alone. Wang et al. [20] found that mediummaturity cultivars quickly started their grain-filling period and had a short active grain-filling period and effective grainfilling time. However, mid-late maturity, late maturity, and super-late maturity cultivars started their grain-filling period slowly and had a long active grain-filling period and effective grain-filling time. Prolonging the active grain-filling period,effective grain-filling time, and grain-filling duration of middle and late stages and enhancing the mean early stage grain-filling rate may increase the grain yield of maize hybrids with differing maturities.

Physiological and biochemical characteristics, biomass accumulation characteristics, and yield formation of maize cultivars with contrasting nitrogen efficiencies have been reported [12,21]. However, there are few studies of the grainfilling characteristics of maize cultivars with contrasting nitrogen efficiencies, and the effect of nitrogen fertilizer management on these cultivars and its mechanism remain unclear.

In the present study, an N-efficient and an N-inefficient maize cultivar screened in a previous experiment[11,22]were used to compare the effects of nitrogen management on grain filling characteristics and yield of maize cultivars with contrasting nitrogen efficiencies to provide a theoretical basis for the rational and efficient application of nitrogen fertilizer in maize production.

2. Materials and methods

2.1. Experimental site and materials

The experiment was conducted in Sichuan in 2015–2016.Meteorological data from maize sowing to harvest are shown in Fig. 1. The soil at the experimental site is purple clay, and the 0–30 cm soil layer had the following chemical compositions in 2015 (2016): 13.83 (12.05) g kg?1organic matter, 1.87(1.21)g kg?1total N,0.42(0.81)g kg?1total P,7.58(16.70)g kg?1total K, 55.65 (61.83) mg kg?1alkali-hydrolyzable N, 4.75 (4.35)mg kg?1Olsen-P, 160.33 (146.13) mg kg?1exchangeable K, and 8.11(7.81)pH.

The N-efficient maize cultivar ZH 311 and N-inefficient maize cultivar XY 508 served as experimental materials. Seeds of the cultivars were obtained from respectively Sichuan Nongda Zhenghong Seed Co. Ltd., Chengdu, Sichuan province, China and Pioneer Technology Co., Tieling, Heilongjiang province,China. They are the main cultivars in southwest China and have consistent growth periods(approximately 120 days).

Fig.1– Meteorological conditions during maize growth.

2.2. Experimental design

The experiment was designed as a two-factor randomized block experiment.The two factors were maize cultivar(ZH 311 and XY 508)and nitrogen management,with five treatments:B1, 0 kg ha?1; B2, 225 kg ha?1(100% base fertilizer); B3,225 kg ha?1(75% base fertilizer + 25% topdressing); B4,225 kg ha?1(50% base fertilizer + 50% topdressing); B5,225 kg ha?1(25% base fertilizer + 75% topdressing). Ten treatments with three replicates were performed. Plots were 6 m × 7.5 m (45 m2total). Cultivars were planted in wide and narrow rows, 1.4 m and 0.6 m, and 50,000 plants ha?1were used. In all treatments, 600 kg ha?1of superphosphate and 150 kg ha?1of potassium chloride were applied as base fertilizer. Other management measures were the same as used in local maize production.

2.3. Sampling and determination of grain filling

Ears with relatively uniform silking time and size were chosen during silking and tagged. A total of 120 ears were tagged in each plot, and five samples were collected from each tagged plant at 7-day intervals from silking to maturation stages.During sampling, the top, middle, and bottom kernels were obtained. Kernels were heated at 105 °C for 60 min, dried at 80 °C to constant weight, and weighed. The top, middle, and bottom kernels were obtained as follows: maize ears were divided into three equal parts (top, middle, and bottom) and the top, middle, and bottom kernels were sampled from the middle of each section.

A logistic equation was fitted to the grain-filling process to obtain characteristic grain-filling parameters. The number of days after silking (t) was treated as the independent variable and the dry weight of 100 kernels per sample as the dependent variable (y). The results were fitted with the following equation following Wang et al.[20]and Li et al.[23]:

where y is the dry weight of 100 kernels(g), A is the theoretical maximum dry weight of 100 kernels,t is the time after flowering(days), B is the initial value parameter, k is the growth rate parameter,and e is the base of natural logarithms.

Based on the above parameters,the following values were calculated:

t1(starting date of grain-filling peak period)=(ln B ?1.317)k?1, corresponding to 100-kernel dry weight (W1) at this time;

t2(end date of grain-filling peak period)=(ln B+1.317)k?1,corresponding to 100-kernel dry weight(W2)at this time;

t3(when the dry weight of 100 kernels reaches 99% of final weight,the effective grain-filling period)=(ln B+4.59512)k?1,corresponding to the 100-kernel dry weight(W3)at this time;

Wmax(100-kernel dry weight at the maximum grain-filling rate)=A 2?1;

Rmax(maximum grain-filling rate)=(k Wmax)×(1 ?WmaxA?1);

Tmax(time to the maximum grain-filling rate)= ln B k?1;

Rmean(mean grain-filling rate)=W3;

P(active grain-filling period)= 6 k?1;

T(effective grain-filling time)= t3

Population grain-filling rate (kg day?1ha?1) = (top grainfilling rate+middle grain-filling rate+bottom grain-filling rate)×kernels per ear× plant density×3?1[24,25].

Yield and 100-kernel weight were determined using all but the border plants in a 10-m2area within each plot. Moisture content was adjusted to 14.0%. Yield components (row numbers, kernels per row, and kernels per ear) were determined from 20 sequential plants per plot.

2.4. Data analysis

Statistical analysis was performed using Microsoft Excel and IBM SPSS Statistics 20.0 (IBM Corporation, Armonk,NY, USA),and significance testing was performed using the least significant difference method. The significance level was set to P < 0.05, and GraphPad Prism 5.0 software (GraphPad Software,San Diego,CA,USA)was used for drawing.

3. Results

3.1. Grain-filling characteristics of different maize cultivars

Table 1 shows that the decision coefficients of the grain filling process equation for top, middle, and bottom maize grains were above 0.98,indicating that the logistic equation fitted the maize grain-filling process. In terms of the final 100-kernel weight, bottom grain > middle grain > top grain; the middle and bottom grain values of XY 508 were higher than those of ZH 311 by 9.75%and 10.18%in 2015,and by 11.54%and 12.97%in 2016, respectively. The difference in top grain dry weight between the two cultivars was not significant in either year,indicating that the grain-weight difference between the two cultivars gradually increased from top to bottom, and the difference between the two cultivars was accounted for mainly by the middle and bottom kernels. Nitrogen fertilizer management and its interaction with maize cultivar showed significant effects on the grain-filling processes in different maize parts. In 2015, the top grain final 100-kernel weight(FKW) of ZH 311 was highest in the B5 treatment, and the middle and bottom grain FKW was highest in the B2 treatment. In 2016, all FKW values were highest in the B4 treatment. XY 508 values were highest in the B5 treatment in all parts in 2015, the top grain FGW was highest in the B4 treatment, the middle grain FGW was highest in the B5 treatment, and the bottom grain FGW was highest in the B2 treatment. Increasing the topdressing ratio promoted grain filling in the top grain and increased its 100-kernel weight but showed smaller effects on the 100-kernel weights of middle and bottom grain, especially for the N-efficient cultivar ZH 311. The difference in FGW between the two cultivars in different parts showed the same increasing trend for topdressing ratio over both years. The top grain FGW of ZH 311 was higher than that of XY 508 (except for the B2 treatment),and the middle and bottom grain FGW of XY 508 were higher than those of ZH 311.The FGW difference between ZH 311 and XY 508 was smallest in the B4 treatment. Thus, the grain weight of the N-inefficient cultivar XY 508 was significantly higher than that of the N-efficient cultivar ZH 311. This difference was accounted for mainly by the differences in middle and bottom grain. Balanced nitrogen application during the growth period fully exploited the grain filling potential of ZH 311, resulting in the highest grain weight in the B4 treatment group.The high base fertilizer ratio(B2)fully exploited the top and bottom growing potential of XY 508 while the high topdressing ratio(B5)exploited the potential of middle grain. Accordingly, the B2 and B5 treatment groups had higher grain weights than the other treatment groups.

The grain-filling characteristic parameters of maize cultivars with contrasting nitrogen efficiencies under different nitrogen fertilizer management modes were markedly different (Table 2). The Rmaxand Rmeanof ZH 311 and XY 508 gradually increased from top to bottom (except for ZH 311 in 2015);the Tmaxof the middle grain appeared earliest,while the Wmax, P, and T of the two cultivars increased gradually from top to bottom in both years. The bottom grain Wmaxwas highest, and P and T were longest. In terms of cultivar, the Rmaxand Rmeanof ZH 311 top grain were higher than those of XY 508,while the Rmaxand Rmeanof ZH 311 bottom grain were lower than those of XY 508. The difference between the middle grain of the two cultivars was not significant.The Tmax(except for bottom grain in 2015) and Wmaxof XY 508 were higher than those of ZH 311, and the P and T of XY 508 were longer than those of ZH 311 in both years.Thus,the top grainfilling rates of N-efficient cultivar ZH 311 were superior to those of N-inefficient cultivar XY 508, while the 100-kernel weight of XY 508 was higher than that of ZH 311 because of its higher middle and bottom grain filling rates and longer effective filling period.For ZH 311,balanced nitrogen application during the growth period effectively increased the grain filling rate, delayed Tmax, and prolonged P and T so that the 100-kernel weight in B4 was highest in both years.For XY 508,higher basal fertilizer ratio (B2) and topdressing ratio (B5)increased the Rmaxand Rmean, delayed Tmax, prolonged P and T, and resulted in significantly higher 100-kernel weights under B2 and B5 treatments than those of other treatments.

According to the logistic curve, the maize grain filling process is divided into early, middle, and late stages. Table 3 shows that the early-stage duration was greatest in the top grain of both cultivars,while the mean grain filling rate in the early stage was the highest in the bottom grain in both years.The middle stage and late-stage duration of the two cultivars was gradually prolonged from top to bottom and the mean filling rate gradually increased from top to bottom so that the middle-stage duration of the bottom grain was greatest and the mean filling rate was highest. Greater middle stage and late stage durations and higher filling rates were the major contributors to the higher final grain weight of bottom grain than of top and middle grain. In terms of cultivar, the earlystage duration in each part of ZH 311 was longer than that of XY 508 in 2015, while the middle- and late-stage duration in each part of XY 508 were higher than those of ZH 311 in both years. Thus, the effective filling time of XY 508 was longer than that of ZH 311,and the difference in filling time between the two cultivars was accounted for mainly by the middle and late stages.The mean grain filling rate of the top grains of ZH 311 was higher than those of XY 508 in middle and late stages in both years, while the mean filling rate of the middle and bottom grain of XY 508 was higher than that of ZH 311 in both years.This result indicated that ZH 311 had an advantage over XY 508 in top grain filling rate but that the middle and bottom grain-filling rate of XY 508 was significantly higher. Nitrogen management affected the duration and filling rate of the both cultivars. The grain duration and filling rate of XY 508 were superior to those of ZH 311. The difference between the two cultivars was smallest under the B4 treatment and largest under the B2 treatment, especially in the middle and bottom grain.

3.2. Difference in population grain-filling rate between maize cultivars

Nitrogen application increased the population grain-filling rate of the two cultivars at different stages, but the effect on XY 508 was greater than that of ZH 311 (Fig. 2). At the peak filling stage(28 days after silking),the mean population grainfilling rate of ZH 311 nitrogen application treatments was higher than that of B1 by 8.60% in 2015 and 17.42% in 2016,while those of XY 508 increased by 14.90% and 36.08%,respectively. In addition, there were clear differences in population grain-filling rate between the two cultivars in response to nitrogen management. At the peak filling period(28 days after silking), the population grain-filling rate of ZH 311 was highest in the B4-treated plants in both years(17.41%and 26.72% higher than those of B1 in 2015 and 2016,respectively). The population grain-filling rate of XY 508 was highest in the B2 treated plants in both years (respectively 20.17%and 40.91%higher).

Nitrogen management, cultivar type, and their interactions showed significant effects on the mean population grain-filling rate of maize (Fig. 3). Nitrogen application significantly increased the mean population grain-filling rates of the two cultivars. The mean nitrogen application treatment of ZH 311 resulted in a 5.76% and 19.66% higher mean population grain-filling rate than that of B1 in 2015 and 2016, respectively, while the mean nitrogen application treatment of XY 508 resulted in respectively 6.16% and 22.09% higher mean population grain-filling rates than that of B1. The increasing effect of nitrogen application on the mean population grain-filling rate of N-inefficient cultivar XY 508 was greater than that of N-efficient cultivar ZH 311. The mean population grain-filling rate of ZH 311 was higher than that of XY 508.In 2015,B1–B5 treated plants were 7.27%,1.18%,10.78%, 14.11%, and 2.16% higher, respectively. In 2016, they were 8.07%, 1.27% and 11.63%, 23.27%, and 6.02% higher,respectively.The differences between the two cultivars in the mean population grain-filling rate were highest in B4 and lowest in B2 treatments for both years. A high basal fertilizer ratio (B2) was conducive to increasing the population grain filling of the N-inefficient cultivar, while balanced nitrogen application (B4) fully exploited the population grain-filling potential of the N-efficient cultivar.

3.3. Grain yield and component differences between maize cultivars with contrasting nitrogen efficiencies

Fig. 2 – Difference in population grain-filling rate of maize cultivars with contrasting nitrogen efficiencies.

Cultivar type,nitrogen management method,and their interaction significantly affected the number of maize kernels per ear,100-kernel weight,and grain yield(P<0.01,Table 4).Kernels per ear and grain yield of ZH 311 were significantly higher than those of XY 508 in 2015 (9.47% and 11.06%, respectively), and in 2016(13.32% and 11.22%, respectively); while the 100-kernel weight and harvest index of XY 508 were higher than those of ZH 311 in 2015 by 2.73% and 10.20%, and in 2016 by 2.38% and 6.12%,respectively. Nitrogen management significantly affected kernels per ear, 100-kernel weight, grain yield, and harvest index.Nitrogen application improved ear characteristics,and increased kernels per ear and 100-kernel weight,thereby increasing maize yield and decreasing harvest index. The difference in the cultivars’ grain yield response to nitrogen management was significantly different (P < 0.01). Grain yield of the N-efficient cultivar ZH 311 was highest under B4 treatment for both years,while that of the N-inefficient cultivar XY 508 was highest under B5 treatment in 2015 and under B2 treatment in 2016.The grain yield of B2 and B5 were markedly higher than that of other treatments for both years.The difference in grain yield between the two cultivars tended to increase first, then decrease with increasing topdressing ratio.The regression equation for ZH 311 compared to XY 508(yield difference between the two cultivars,y) and topdressing ratio (x) was y = ?0.0007x2+ 0.0554x + 0.4088(R2= 0.9575). When the topdressing ratio was 39.57%, the yield difference between the two cultivars was largest.That is,when the basal fertilizer was 60.43% and topdressing was 39.57%, the N-efficient cultivar ZH 311 had the higher yield advantage(1.51 t)compared to the N-inefficient cultivar, XY 508. The 100-kernel weight of maize was positively correlated with grain filling rate(R2> 0.89), but there was a low correlation between 100-kernel weight and effective filling time(R2<0.20)(Fig.4).The correlation between 100-kernel weight and grain yield was low (R2< 0.50),while the grain yield was significantly positively correlated with the population grain-filling rate(R2>0.85)(Fig.5).

Fig.3–Difference in the mean population grain-filling rate of maize cultivars with different nitrogen efficiencies.Values with different lowercase letters are significantly different at P <0.05 according to the least significant difference test.

4.Discussion

4.1. Effects of nitrogen management on grain-filling characteristics of maize cultivars with contrasting nitrogen efficiencies

The accumulation of grain dry matter in the grain-filling stage affects maize grain weight and yield; grain dry matter accumulation is determined by both grain filling rate and duration time. Li et al. [26] showed that the grain weight of different cultivars was affected mainly by grain-filling duration time,while that of the same cultivar was affected mainly by grain-filling rate. Our results showed that the grain dry weight of the top,middle and bottom grain increased in an‘S’curve; that is, with a “slow-fast-slow” growth trend, consistent with the results of Ren et al.[27]The grain-filling rate and duration time of middle and late stages of both cultivars showed that the grain duration time from top to bottom grain was gradually extended, and that the grain filling rate gradually increased. The highest grain filling rate and the longest middle stage were found in the bottom grain. Longer middle and late stages,and a higher grain filling rate were the most important reasons for the higher FGW values of bottom grains compared to those of top and middle grain.The P and T of the N-inefficient cultivar XY 508 were longer than those of the N-efficient cultivar ZH 311,indicating that the FGW of XY 508 was superior to that of ZH 311.The FGW advantages of XY 508 from bottom to top grain gradually decreased, even becoming disadvantages. This finding showed that the top grain of ZH 311 had a higher filling potential than that of XY 508, which reduced the risk of top grain abortion and was more conducive to a high and stable yield. This result was consistent with previous findings by Lv et al.[17].

Nitrogen application improved the grain-filling characteristics of maize,and the grain dry matter of maize in nitrogen application was significantly higher than that without nitrogen application in all stages [12,28]. This results shows that nitrogen management significantly affected the grain-filling characteristics of maize,while the grain-filling characteristics of maize cultivars with contrasting nitrogen efficiencies in response to N management differed sharply. Nitrogen application significantly improved the grain-filling characteristics of both cultivars, but the effect of N management on grain filling in the N-inefficient cultivar XY 508 was significantly higher than that in the N-efficient cultivar ZH 311, especially for the top and middle grains,indicating that the demand for nitrogen fertilizer in the grain-filling process of an Ninefficient cultivar was greater than that of an N-efficient cultivar. The 100-kernel weight of maize was positively correlated with grain filling rate, but there was a low correlation between 100-kernel weight and effective filling time. These results were consistent with those of other authors [29,30]. Grain-filling rate is under mainly genetic control and is positively correlated with grain weight, while grain filling duration is strongly affected by environmental factors and less well correlated with grain weight[29].

4.2. Effects of nitrogen management on grain yield and its components in maize cultivars with contrasting nitrogen efficiencies

Maize grain yield is composed of three elements:ear number,kernels per ear, and grain weight. When a certain number of ears are present,the relationship between kernels per ear and grain weight must be well coordinated for full development of maize yield potential [19,31]. The grain-filling rate and duration of the N-inefficient cultivar XY 508 were higher than that of the N-efficient cultivar ZH 311, resulting in a significantly higher 100-kernel weight than that of ZH 311 in both years,while the kernels per ear of the N-efficient cultivar ZH 311 was significantly higher than that of the N-inefficient cultivar XY 508. Thus, the population grain-filling rate of ZH 311 was significantly higher than that of XY 508.Furthermore,the high population grain-filling rate of the N-efficient cultivar ZH 311 resulted in a significantly higher grain yield than that of the N-inefficient cultivar XY 508. This result is consistent with findings of Chen et al. [12], who showed that an N-efficient maize cultivar had a significant grain yield advantage over an N-inefficient cultivar, and that its yield advantage came from the population grain-filling advantage conferred by the higher number of kernels per ear.

Nitrogen management showed a significant effect on the grain yield differences between maize cultivars with contrasting nitrogen efficiencies.In comparison with the N-inefficient cultivar XY 508, the grain yield advantage of the N-efficient cultivar ZH 311 first increased and then decreased as the topdressing ratio increased, and the grain yield advantage was largest under B4 treatment (50% base fertilizer + 50%topdressing)in both years.The grain yield difference between the cultivars was the highest when the topdressing ratio was 39.57%. When the basal fertilizer ratio was 60.43% and the topdressing ratio was 39.57%,the grain yield advantage of the N-efficient cultivar ZH 311 over that of the N-inefficient cultivar XY 508 was 1.51 t ha?1.

5. Conclusions

Grain-filling rate and duration of the N-inefficient cultivar XY 508 were higher than those of the N-efficient cultivar ZH 311.Accordingly,the 100-kernel weight of XY 508 was significantly higher than that of ZH 311. The kernels per ear of the Nefficient cultivar ZH 311 was significantly higher than that of the N-inefficient cultivar XY 508, making the population filling rate of ZH 311 significantly higher than that of XY 508.The higher population filling rate of the N-efficient maize cultivar resulted in a significant yield advantage over the Ninefficient maize cultivar. Although nitrogen management increased maize grain yield, the response of maize cultivars with contrasting nitrogen efficiencies to nitrogen management was inconsistent. A basal fertilizer ratio of 60.43% and topdressing ratio of 39.57% effectively increased grain-filling rate, delayed Tmax, prolonged P and T, increased 100-kernel weight, and maintained higher kernels per ear, thereby improving the population filling rate and maximizing yield advantage in the N-efficient cultivar.The 100%basal fertilizer ratio not only increased the number of kernels per ear, but also maintained high grain-filling characteristics to produce a higher 100-kernel weight and increased the population filling rate, thus producing a high grain yield in the N-inefficient cultivar. Thus, the 100% basal fertilizer ratio partially compensated for the deficient grain yield in the N-inefficient cultivar.

Declaration of competing interest

The authors declare no conflicts of interest.

Acknowledgments

This work was supported by the National Key Research and Development Program of China(2016YFD0300307 and 2016YFD0300209), the Special Fund for Agro-scientific Research in the Public Interest of China (20150312705), the Sichuan Agriculture Research System of Maize Industry.

Author contributions

Qiang Li and Jichao Yuan designed the study;Qiang Li,Lunjing Du, and Dongju Feng performed the experiments; Yun Ren and Zhexin Li analyzed the data; Fanlei Kong developed methods;and Qiang Li and Lunjing Du wrote the paper.