Zi Shi,Wen Ren,Yanxin Zhao,Xiaqing Wang,Ruyang Zhang,Aiguo Su,Shuai Wang,Chunhui Li,Jiarong Wang,Shuaishuai Wang,Yunxia Zhang,Yulong Ji,Wei Song*,Jiuran Zhao*

Beijing Key Laboratory of Maize DNA Fingerprinting and Molecular Breeding,Maize Research Center,Beijing Academy of Agriculture & Forestry Sciences,Beijing 100097,China

Keywords:

A B S T R A C T Genic male sterility(GMS)is one of the most important resources for exploiting heterosis in crop breeding,so that identifying genomic loci regulating GMS is desirable.However,many regulatory genes controlling GMS have not yet been characterized in maize,owing partly to a lack of genetic materials.We generated a recessive male-sterile maize mutant in the Jing 724 genetic background via ethyl methanesulfonate treatment,and found the male sterility to be due to a single gene mutation.Bulk-segregant RNA sequencing of three replicates indicated that one genomic region located at the end of chromosome 4 was associated with the observed mutant phenotype.Among genes with nonsynonymous mutations,Zm00001d053895(bHLH51)showed abolished expression in the sterile bulks and was annotated as a bHLH transcription factor orthologous to Arabidopsis AMS,suggesting an association with the male sterility of the mutant.Kompetitive Allele-Specific PCR assays further validated the exclusive correlation of male sterility with the single C-to-T mutation in the fifth exon.The new maize mutant and the potential SNP locus provide novel genetic material for investigating the molecular mechanism underlying tapetal development and may facilitate the improvement of hybrid production systems.

1.Introduction

Because maize is one of the crops in which heterosis has been most successfully exploited,its yield has increased worldwide[1].The self pollination of maternal plants is conventionally avoided via detasseling,but this approach is time-and/or laborconsuming,as well as cost-inefficient[2,3].More importantly,detasseling may lead to yield compensation in hybrids because of the damage to the uppermost leaves.The employment of malesterile lines is one of the most feasible alternatives for the production of maize hybrids.

Cytoplasmic male sterility(CMS),which is caused by nuclear–mitochondrial interactions[4,5],has been widely applied for hybrid crop production[6].However,its application is somewhat restricted by the fertility restoration[7–9],the poor genetic diversity[6],and increased susceptibility to certain diseases associated with T-type CMS[10–12].Thus,more practical methods in the hybrid production are needed.

Genic male sterility(GMS),controlled by nuclear genes,is able to rectify all of the disadvantages associated with CMS,and tremendous efforts have been made to identify genes conditioning GMS in maize.To date,hundreds of maize GMS mutants have been analyzed[13],but only 17 GMS genes have been cloned and functionally characterized[3].These genes play roles in various developmental stages from the premeiotic to postmeiotic morphogenesis of the maize anther,including archesporial cell specification,anther somatic cell division,tapetum development,pollen mother cell meiosis,pollen formation,and anther dehiscence.Most of the identified GMS genes are involved in tapetum or microspore development[14],serving as transcription factors or contributing to lipid and polysaccharide metabolism and other processes[3].InArabidopsis,a genetic pathway ofDYT1-TDF1-AMS-MS188-MS1and the associated regulatory network mediating tapetum development have been investigated[15–23].AMS,encoding a basic helix-loop-helix(bHLH)transcription factor[24],directly regulates the expression of several target genes conditioning pollen wall formation,including MS188[17],EXL5[19],and TEK[25].Theamsmutant exhibits defects in phenolic compound accumulation,sporopollenin deposition,and microspore release,indicative of the central role for AMS in the sophisticated and precise transcriptional network regulating pollen wall architecture[19].Despite our substantial understanding of the genetic basis underlying the postmeiotic male reproductive development in model plant species,the corresponding transcriptional pathway in maize remains unclear.

Although GMS is stable and universally present in various germplasm,it is difficult to self-produce a large amount of male sterile lines in a commercial setting.Accordingly,several biotechnological strategies have been developed to deploy GMS in maize hybrid production[3].For example,seed production technology(SPT)and multi-control sterility(MCS)systems have been applied in the maize production by employing several GMS genes,includingMs45[2],ZmMs7[26],Ms30[1],andZmMs33[27].These approaches can overcome the difficulties associated with using GMS genes.Therefore,the investigation of new GMS genes in elite Chinese germplasm will thus advance male sterility-dependent maize seed production in China.

Jingke 968 is one of the most widely cultivated maize varieties in China,with the accumulative planting area exceeding 6 million hectares[28],largely because of its high yield,broad-spectrum resistance,and robust adaptability to various ecological regions.The S-type CMS-based three-line system has already been successfully applied for the commercial production of Jingke 968,but lines showing unstable fertility were detected in a backcross population[7,11],which might potentially jeopardize seed purity.Furthermore,to date,no GMS mutant has been identified in the maternal line,Jing 724.Thus,the theoretical basis for developing a biotechnology-based male sterility system for the commercial production of Jingke 968 remains limited.

The objective of this study was to generate a recessive malesterile mutant in Jing 724 and map a candidate gene.The mutant and potential single nucleotide polymorphism(SNP)mutation were expected to provide novel genetic resources and tools for elucidating the molecular mechanism underlying tapetum development in maize and for the application of GMS in the seed production of Jingke 968 and other varieties.

2.Materials and methods

2.1.Plant materials

Maize inbred line Jing 724 was used to generate GMS mutants.Specifically,Jing 724 pollen was treated with EMS as previously described[29,30]and then used to produce M1seeds.The M1plants were self-pollinated to produce M2seeds.M2plants were grown and their fertility was evaluated.Among the male-sterile M2plants,two mutants,named 4624 and 4611,were selected to cross to the inbred line B73,and then selfed to obtain the F2generation,which was subjected to phenotypic characterization and further analyses.

2.2.Phenotypic analysis

Two F2populations were planted on an experimental site at the Beijing Academy of Agriculture and Forestry Sciences in May 2019.During the tasseling stage,tassel fertility was evaluated by anther exsertion and dehiscence.Plants with a slender primary branch and a lack of anther exsertion were designated as sterile.The segregation ratios of fertile to sterile individuals were calculated and a chi-square analysis was performed to determine whether the sterility was controlled by a single gene.

Spikelets were collected from both fertile and sterile plants and placed on slides.Pollen grains were squeezed out using tweezers and then stained with 1% I2-KI solution containing 0.5%(w/v)iodine and 1%(w/v)KI.The stained samples were examined under a light microscope(Olympus IX73,Shinjuku,Japan)at 40×magnification.

Unopened spikelets were collected from fertile and sterile plants after tasseling.Anther samples were isolated and fixed in FAA fixative(Coolaber,Beijing,China).The samples were then dehydrated,sliced,stained,and observed as previously described[7].

2.3.Transcriptome analysis and gene mapping via bulk-segregant RNA-sequencing(BSR-seq)

The F2population generated from M2plant 4624 was analyzed by BSR-seq,as the male sterility of 4624 was inferred to be controlled by a single gene.Spikelets of the F2population were collected at the very beginning of the tasseling stage,immediately frozen in liquid nitrogen,and stored at-80 °C.Collected spikelets were designated as sterile or fertile according to the subsequent phenotypic analysis during later tassel development.Total RNA was extracted individually from 90 fertile and 75 sterile spikelets with TRIzol(Fisher Scientific,Hampton,NH,USA),after which the RNA concentrations and integrity were evaluated by Agilent2010(Agilent,Santa Clara,CA,USA).Only RNA samples with a RIN value greater than 6.8 were used for subsequent analysis.RNAs extracted from sterile spikelets were grouped into three bulks,each containing 30 samples with five samples randomly shared with other sterile bulks,and designated as S1,S2,and S3.Similarly,the RNA samples extracted from fertile spikelets were grouped into bulks F1,F2,and F3,each representing 30 fertile individual plants.Three BSR-seq replicates were constructed,each containing one sterile and one fertile bulk(S1 vs.F1,S2 vs.F2,and S3 vs.F3).Transcriptome sequencing was performed for each bulk with the HiSeq 4000 system(Illumina,San Diego,CA,USA)at Biomarker Technologies Co.,Ltd.(Beijing,China).

Clean reads were aligned to the maize reference genome Zea_-mays.AGPv4(ftp://ftp.ensemblgenomes.org/pub/plants/release-33/fasta/zea_mays/dna/)to obtain mapped data.FPKM[31]was used to determine the gene expression level,and differentially expressed genes(DEGs)were identified as previously described[7].Briefly,genes with an expression fold change≥2 and a false discovery rate＜0.01 were recorded as DEGs.The Euclidean Distance(ED)algorithm was employed to assess the genetic regions associated with male sterility.Before association analysis,all SNPs were filtered to eliminate SNPs with multiple genotypes,SNPs supported by fewer than four reads,and SNPs that were not polymorphic between the sterile and fertile bulks.The ED values were calculated as described by Su et al.[6],and then fitted according to the SNPNUM method.The median+3SD of the fitted value at all loci was set as the association threshold[32],which was calculated as 0.45,0.20,and 0.28 for the three BSR-seq replicates,respectively.Candidate regions were then determined.

A BLAST search of multiple databases,including the NR,Swiss-Prot,GO,KEGG,and COG databases,was performed for the indepth annotation of DEGs and genes with nonsynonymous SNPs in the candidate regions.

2.4.Identification of the candidate gene and the functional mutation

The full-length genomic DNA sequence of the candidate gene,Zm00001d053895,was amplified from wild-type Jing 724 and sequenced to identify natural variations between B73 and Jing 724.The amplification and sequencing primers are presented in Table S1.The potential functional mutation in the gene was identified by alignment of the sequences of the sterile plants with those of wild-type Jing 724 plants.The predicted gene structure and functional domains were obtained from MaizeGDB(https://www.maizegdb.org)and the conserved domains function of the NCBI protein database(https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi?RID=DSSYFCWH01R&mode=all),respectively.

2.5.Kompetitive Allele-Specific PCR(KASP)assays

Genomic DNA was extracted from individual plants of the F2population by the CTAB method[33]with minor modifications.KASP assays were developed for the potentially functional SNP inZm00001d053895,with SNPs in the other three nonsynonymous genes in the candidate region used as negative controls.The sequences of KASP primers are presented in Table S1.The assays were carried out using a 1536-well plate and a 1 μL system containing 30 ng dried DNA,0.5 μL 2×KASP MasterMix(LGC Inc.,Middlesex,UK),0.486 μL sterile ddH2O,and 0.014 μL primer working solution(final concentration of 12 μm L-1for the two fluorescence binding forward primers and 30 μm L-1for the common reverse primer).The PCR conditions were as follows:preheat at 95 °C for 15 min;10 cycles of touchdown PCR from 61 °C to 55 °C with a 0.6 °C decrease per cycle;32 cycles of 94 °C for 20 s and 55 °C for 1 min.The PCR endpoint fluorescence was measured with the BMG Pherastar software(LGC Inc.),and the SNP allele clusters was visualized with the Kluster Caller software(LGC Inc.).

2.6.qRT-PCR analysis

The previously mentioned RNAs of five homozygous fertile plants and five sterile plants were used for qRT-PCR analysis.Reverse transcription was performed using the PrimeScript II 1st Strand cDNA Synthesis Kit(Takara Bio Inc.,Shiga,Japan).The PCR reaction was conducted as previously described[34]with the maizeGpn1gene as internal control.The gene-specific primers are described in Table S1.The relative gene expression level was determined by the 2-ΔΔCtmethod.

3.Results

3.1.Male-sterile mutant due to a single recessive allele

Two of the male-sterile M2plants,4624 and 4611,were crossed with B73 to generate F2mapping populations comprising 288 and 184 individual plants,respectively.The male-sterility phenotype was evaluated in both F2populations during the tasseling stage.Plants with anther exsertion and dehiscence were designated as fertile,whereas plants showing a slender central spike and no anther exsertion were assigned as sterile(Fig.1A).Respectively 75 and 94 sterile plants were identified in the F2populations derived from 4624 and 4611.The respective chi-square(χ2)values were calculated as 0.17(P＞0.2)and 66.78(P＜0.01)(Table 1),indicating that the segregation of the 4624 population followed a 3:1 ratio and its sterility was due to a single recessive mutation,while this was not the case for 4611.

To further assess the sterility caused by a single mutation in 4624,unopened spikelets were collected from the F2population plants before anther exsertion.The sterile spikelets and anthers inside were pale and withered,markedly smaller,and more shrunken than those of the fertile plants(Fig.1B–D).To investigate the pollen viability,we collected pollen grains from fertile and sterile plants and stained with the I2-KI solution.The pollen grains from fertile plants were round and thoroughly stained(Fig.2A),whereas no pollen grains were detected in sterile plants(Fig.2B).In addition,transverse anther sections revealed structural differences between fertile and sterile anthers.At stage 13 of anther development(based on An et al.[1]),pollen grains of fertile plants had developed normally and the tapetum had degenerated(Fig.2C).In contrast,only a collapsed locule and endothecium with complete absence of mature pollen grains were observed in sterile plants(Fig.2D),indicating that tapetal development and microspore initiation were substantially compromised in the GMS mutant.Consistent with other GMS mutants,the observed mutant phenotype was stable across different planting seasons(2017–2019)and locations(Beijing and Hainan,China).

Fig.1.Phenotypic comparison of fertile and sterile plants in the segregating population.(A)Representative image of the major tassel branch collected from fertile and sterile plants 7 days after tasseling.(B)Representative image of 15 unopened spikelets from fertile and sterile plants.(C)Anatomical structure of the spikelets.(D)Six anthers harvested from the spikelets in panel(C).The lengths of the scale bars are as indicated in each panel.

3.2.Identification of the candidate gene associated with male fertility

To map the genetic loci conditioning the male sterility of the mutant,we extracted RNA samples from the immature tassels of all 75 sterile and 90 fertile plants in the F2population derived from 4624 to construct three replicates of sterile bulks(S1,S2,and S3)and fertile bulks(F1,F2,and F3).Transcriptome analysis revealed that an average of 4042 and 837 genes were up-and downregulated,respectively,in fertile bulks compared to the sterile bulks for all three comparisons(Table 2).The number and expression patterns of DEGs were similar for the three comparisons,reflecting the relatively robust reproducibility of three RNA-seq replicates.In addition,the total numbers of high-quality SNPs used for further BSR-seq analyses were also comparable among the replicates.

Using the ED algorithm,the genomic locus linked to GMS was identified based on the SNPs between the sterile and fertile bulks.One significant ED peak at the end of chromosome(Chr.)4 was detected in all three BSR-seq replicates(Fig.3),and despite varying in region sizes,only one genomic region was mapped using the median+3SD of the fitted value at all loci as the significant cutoff for each BSR-seq analysis.The region detected for S2 vs.F2 which harbored 107 nonsynonymous SNPs in 47 genes and 31 DEGs,was the smallest and may represent the overlapping region of threeBSR-seq analyses(Table 3).Among the overlapping nonsynonymous genes from three replicates,five exhibited differential expression between the sterile and fertile bulks,with four upregulated and one down-regulated.The coding sequence of the down-regulated gene,Zm00001d053882,revealed that the nonsynonymous SNP was a natural variation between B73 and Jing 724 and not due to the EMS treatment.The annotation of the nonsynonymous genes and the expression patterns of the DEGs in the mapped region revealed that one of the genes,Zm00001d053895,encodes a bHLH transcription factor,and its expression level was approximately 80-fold higher in fertile than in sterile plants(Fig.4A).In addition,it was the maize ortholog of theArabidopsisAtAMSgene and the riceOsTDRgene,both of which are critical for tapetum development and male fertility[24,35,36],suggesting thatZm00001d053895is the candidate gene controlling male fertility in maize and that its sequence alteration introduced by EMS is potentially responsible for the sterile mutant phenotype.

Table 1Segregation of fertility in the F2 populations derived from mutants 4624 and 4611.

Table 2Number of DEGs and high-quality SNPs from three sets of BSR-seq data used for gene mapping.

Table 3Numbers of significant genes and SNPs in candidate regions identified from three sets of BSR-seq data.

Fig.2.Pollen grains stained with I2-KI solution as well as the transverse section of anthers.(A)Representative image of pollen grains harvested from fertile plants.Round pollen grains with black staining were considered to be normal.(B)Pollen grains were undetectable in sterile plants.(C)Paraffin microscopic analysis of the transverse anther section from a fertile plant.(D)Transverse anther section from a sterile plant.Scale bars represent 200 μm.

3.3.Validation of the candidate gene and the functional mutation

The BSR-seq data showed a total of 5 nonsynonymous SNPs anchored onZm00001d053895between the two bulks, but the polymorphism was determined based on the B73 reference genome. To determine whether any of the nonsynonymous mutations were resulted from the natural variation, we cloned the full-length Jing 724 genomic sequence. Sequencing revealed that the first four nonsynonymous SNPs were indeed natural variations between B73 and Jing 724 (Fig. 4B). In addition, a 6-bp insertion encoding two extra amino acids was detected in the fifth exon of the wild-type Jing 724 sequence. However, these variations did not attribute to the male sterility of the mutant. In contrast, a nonsynonymous Cto-T alteration at position 243,231,250 in the sterile bulk was not a natural variation between the two inbred lines, and the resulting proline-to-serine conversion occurred in the HLH domain of the predicted protein (Fig. 4B), making it a highly plausible functional point mutation due to the EMS treatment.

Fig.3.Chromosomal distribution of ED association values.(A)Representative image of the ED distribution over the whole genome based on the BSR-seq analysis of S1 vs.F1.The chromosomes are indicated on the X-axis.(B–D)Enlarged ED distributions on chromosome 4 obtained from the bulk analysis of S1 vs.F1,S2 vs.F2,and S3 vs.F3,respectively.Colored dots represent the ED values of each SNP.The black line is the fitted ED value.The red dashed line represents the threshold for significant association.The X-axis indicates the physical location on chromosome 4.S1,S2,and S3 represent sterile bulks 1,2,and 3;F1,F2,and F3 represent fertile bulks 1,2,and 3.

To further validate the relationship between the point mutation and fertility,we performed KASP assays on this SNP,along with other nonsynonymous SNPs anchored on the other three upregulated genes,Zm00001d053952(Chr.4:244,167,795),Zm00001d053856(Chr.4:242,288,691),andZm00001d053875(Chr.4:242,703,540)in the candidate region as negative controls for the whole mapping population.Among the 276 available plant samples,all 72 sterile plants carried a homozygous T allele at position 243,231,250,whereas homozygous C or the heterozygous C/T allele were observed in all 204 fertile plants,without exception(Fig.5A;Table S2).However,for the KASP analysis of the other three genes the SNP calls were random and showed no correspondence between genotype and fertility(Fig.5B,S1;Table S2).Therefore,such a complete association between the nonsynonymous SNP inZm00001d053895and the sterility phenotype implies that this SNP is possibly the causal mutation for the recessive male sterility in the 4624 mutant.

4.Discussion

Because male sterility is particularly important for maize hybrid production,identifying the underlying genes is of great significance in maize breeding.In this study,we generated a malesterile mutant via an EMS treatment(Figs.1 and 2),mapped one genomic region at the end of Chr.4 via BSR-seq(Fig.3),identifiedZm00001d053895(bHLH51)as a candidate gene(Fig.4),and validated the correlation between the SNP mutation and male fertility(Fig.5).Despite the complete association between a mutation inbHLH51and male sterility,we cannot rule out the slim possibility that the mutation is not the functional one but is tightly linked to the causal locus.Further functional analyses ofbHLH51,such as knockout via the CRISPR/Cas9 system and complementation assays,may confirm its role in male fertility.

Male sterility is caused by multiple factors and may occur at any stage during anther development[3].Tapetum development,which contributes greatly to pollen production and fertility,is regulated by a complex transcriptional network,consisting of multiple tapetum-specific transcription factors[16].The candidate gene identified in this study,bHLH51,is an ortholog of theArabidopsis AMSgene[17,18,25]and the riceTDRgene[35,36],which encode master mediators of tapetal differentiation and function.TheArabidopsis amsmutant exhibits a sporophytic recessive male-sterile phenotype,with an underdeveloped tapetum and degenerated microspores[19],similar to our EMS-induced mutant(Figs.1 and 2),suggesting thatbHLH51shares a similar function in maize.In a recent study[37],maize bHLH51 was involved in promoting tapetal maturation as part of a heterodimer under the regulation of Ms23.In fact,according to the comparison of the orthologous GMS genes between maize andArabidopsis,the maize genetic counterparts of theArabidopsistapetal pathway genes,DYT1-TDF1-AMS-MS188-MS1,have been predicted[3]to be Ms32-Ms9-bHLH51-ZmMYB80-ZmMs7.Of these genes,Ms32,Ms9,andZmMs7have already been characterized[26,38,39].Unfortunately,the complete maize pathway remains unclear,owing to a lack of mutants forbHLH51andZmMYB80[3].To the best of our knowledge,this study is the first to identify abHLH51mutant allele in maize.Thus,the generated mutant provides a novel genetic resource that may facilitate the functional analysis of the bHLH51 transcription factor and the further elucidation of the complex genetic pathway controlling tapetal development in maize.

Fig.4.Expression and mutation of the candidate gene Zm0001d053895 in the sterile plants.(A)Average FPKM value of Zm0001d053895 in the immature tassels of the sterile and fertile bulks.The histogram represents the mean±SD for the three sequenced bulks.*indicates the statistical significance at the 0.01 probability level.(B)Variations in the Zm0001d053895 sequence between the Jing 724 mutant and B73.Blue boxes and black bent lines represent exons and introns,respectively.Black vertical lines and triangle indicate SNPs/InDels due to natural variation,whereas the red vertical line represents the position of the SNP introduced by EMS treatment.Pink boxes in the protein sequence indicate predicted functional domains.The red diamond on the HLH motif represents the position of the potentially functional amino acid change.

Fig.5.Validation of the potential SNP mutation based on KASP assays of the whole mapping population.KASP markers were developed for the nonsynonymous SNPs identified in the BSR-seq.(A)KASP analysis of the potential causal mutation in Zm00001d053895.The genotypes for the observed phenotypes are as follows:sterile:TT and fertile:CC or CT.(B)KASP assay of another SNP on Zm00001d053952 in the candidate region,which served as the negative control.The genotypes for the observed phenotypes are as follows:sterile:AA or CA and fertile:CC or CA or AA.Black dots represent the no-template control.Pink dots indicate ambiguous data.

The C-to-T point mutation results in the substitution of a proline with a serine in the encoded HLH domain(Fig.4B).Given that this domain usually mediates interaction with other proteins to form dimeric complexes during transcriptional regulation[40,41],it is possible that the amino-acid change due to the point mutation disrupts the binding of the transcription factor to its interacting partners,thereby altering the expression of downstream target genes and thus impeding tapetal development.Although the full genetic pathway of tapetal development in maize is still unclear and the structural genes have not been well characterized,the relative expressions of two potential targeted transcription factors,MYB80andZmMs7,were substantially reduced in our sterile mutants(Fig.S2).In addition,the lower expression level ofZm00001d053895in sterile than in fertile plants(Fig.4A)and the absence of mutations in the 1.6-kb region upstream of the ATG start codon suggest that the mutation in the gene may reduce its own expression via a regulatory feedback mechanism.Additional investigations are needed to comprehensively explore the regulatory mechanism of this genetic pathway in the maize mutant.

In conclusion,we identified a recessive nuclear allele controlling male sterility in maize lines developed at our research facility.Thus,this allele may serve as a novel genetic material to be integrated into Jing 724 and/or our other core inbred lines,and along with the SPT[2]or MCS[1,26,27]strategies,it may provide new breeding resources with independent intellectual property rights to substantially advance hybrid production.In this case,the KASP marker developed in this study(Fig.5A)may be applicable as a robust molecular marker to expedite the breeding process accurately and efficiently.

CRediT authorship contribution statement

Zi Shidesigned the experiment,performed the phenotypic and genetic analysis of the mutant in the field,analyzed the BSR-Seq data,supervised the cloning of the candidate gene,designed the KASP assays,and drafted the manuscript;Wen Renanalyzed the transcriptome data and Sanger sequencing data and revised the manuscript;Yanxin Zhao,Ruyang Zhang,and Shuai Wangidentified the initial mutant and constructed the F2mapping population;Xiaqing Wang,Aiguo Su,Chunhui Li,and Yulong Jicollected field samples for RNA-Seq;Jiarong Wangprovided technical assistance toZi Shi;Shuaishuai Wang and Yunxia Zhangperformed the KASP assays;Wei Songsupervised the experiments and revised the manuscript;Jiuran Zhaosupervised the experiments.All authors read and approved the final manuscript.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

We appreciate the project funding supported by the Beijing Nova Program(Z171100001117033),the Beijing Scholars Program(BSP041),and the Youth Research Fund of BAAFS(QNJJ201931).

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2020.09.007.