Genome-wide association mapping of stresstolerance traits in cotton

Heng Sun,Minghui Meng,Zhenhu Yn,Zhongxu Lin,Xinhui Nie*,Xiyn Yng,*

aNational Key Laboratory of Crop Genetic Improvement,Huazhong Agricultural University,Wuhan 430070,Hubei,China

bKey Laboratory of Oasis Ecology Agricultural of Xinjiang Bingtuan,Agricultural College,Shihezi University,Shihezi 832003,Xinjiang,China

Keywords:Upland cotton Stress Relative germination rate SSR SNP Candidate gene

ABSTRACT Environmental stresses severely impair cotton production worldwide.To identify the genetic basis of,and molecular markers associated with,environmental stresses such as salt,cold and Verticillium wilt,association mapping of salt-,cold-,and disease-tolerance traits was performed in a population of 503 upland cotton accessions using 179 polymorphic SSR markers and 11,975 array-derived SNP markers.Salt and cold tolerance was evaluated via the relative germination rate(RGR)of the seeds under seven and four environments,respectively.The disease index of Verticillium wilt was investigated for two years in the field in Xinjiang.These three traits showed largevariation across environments.A genome-wide association study revealed that 31,19,and 15 SSR markers were associated with RGR-Salt(the relative germination rates of seeds under salt stress),RGR-Cold(the relative germination rates of seeds under cold stress),and DIV(the disease index of Verticillium wilt),respectively.Six SNPs in seven environments and two SNPs in BLUP(best linear unbiased prediction)results were associated with RGR-Salt,and the phenotypic variance explained ranged from 3.96 to 5.00%.Two SNPs(i02237Gh,i02243Gh)on D01 were concluded to be stable genetic loci associated with RGR-Salt.A total of 223 genes were found in a candidate gene interval(D01,37771-1942912).Of these four genes,GhPIP3A,GhSAG29,GhTZF4,and GhTZF4a,showed expression changes in sensitive and tolerant genotype accessions under salt stress,and were assigned as candidate genes associated with RGR-Salt.This study revealed the genetic basis of stress resistance in upland cotton and will facilitate stress-resistance breeding in cotton.

1.Introduction

Cotton(Gossypium hirsutum L.),one of the major economic crops worldwide,provides the main source of natural fiber in our daily life.Various stress conditions adversely affect cotton growth,severely reducing agricultural productivity and fiber quality[1].Among all abiotic factors,high salinity and low temperature are the major environmental factors that limit cotton growth and development at the germination and seedling stages[2,3].Indeed,soil salinity has become a serious issue for agricultural production in some regions,and will become progressively more severe because of climate change,inefficient irrigation,and excessive fertilization[4].Approximately 20%of irrigated land is affected by soil salinity stress[5].Low temperature is a critical limiting factor for spatial distribution of plants,it reduces seed germination and leaf expansion,results in wilting,and may lead to necrosis[6].Cotton is native to warm habitats and is susceptible to injury when exposed to low temperatures.In addition to abiotic stress,a biotic stress factor,Verticillium wilt(VW),is widely distributed in almost all cotton-growing areas worldwide.This is a destructive fungal disease that is caused by the soilborne fungus Verticillium dahlia[7,8].V.dahlia causes leaf curl,chlorosis,necrosis and defoliation,and vascular discoloration,and caused a direct economic loss of 480 million bales from 1990 to 2014 in the U.S.[9,10].

Plant tolerance is a complex trait influenced by multiple factors including both genetic(genotype)and non-genetic(environmental)factors[11,12].Conventional breeding of stress-tolerant cotton cultivars has been limited by a lack of genetic variation,complex genetics,segregation distortion,linkage drag,and suppression of recombination[13-17].With the rapid development and successful application of molecular markers,quantitative trait locus(QTL)mapping has become an effective tool used by breeders to dissect the genetic control of complex quantitative traits.With the development of high through put technology,genome-wide association studies(GWAS)have become an effective method for analysis of quantitative traits controlled by multiple genes with small effects[18].Compared to conventional gene mapping,GWAS can provide higher QTL mapping resolution with no requirement for linkage group construction[18].Based on these advantages,GWAS has been widely employed in breeding of crops including maize,oilseed rape,barley,potato,and rice[19-21].

Although many QTL have been mapped in cotton,including fiber quality,yield traits,and morphological traits[13,22-24],few studies have reported QTL mapping for stress tolerance.It is desirable to identify stress-responsive genes or QTL for elucidating the molecular basis of stress tolerance.The life cycle of plants includes the germination,seedling, vegetative,and reproductive stages.Germination of cottonseeds is an important phase,because this plant growth stage is the most sensitive to adverse environmental factors,such as salt and cold stresses,which can inhibit seed germination by reducing water uptake and nutrient mobilization[25,26].Some studies[26-30]have focused on salt and cold tolerance during the germination stage in rice,barley,tomato,and maize.Although other studies[17,31,32]have focused on QTL analyses of stress tolerance at the seedling stage,QTL mapping for salt and cold tolerance at the seed-germination stage has rarely been reported.

The accuracy of association mapping depends mainly on marker density and mapping populations[33].When 109 cotton accessions were assessed[34]for marker-trait associations of salinity tolerance with 250 genome-wide SSR markers,BNL3103,NAU478,and BNL3140 were confirmed to be associated with salt treatment.Twenty-six markers distributed on 14 chromosomes were associated with VW resistance in 108 elite cotton lines and 177 SSR markers[35].However,the marker densities and populations used in these previous association studies were relatively low,so that QTL mapping for stress tolerance has remained limited.

A set of 11,975 SNP markers from a high-density genetic map basedona63 Kcottonarray(the Cotton SNP63K Array)[36]and a set of 494 SSR markers were used to genotype an extensive collection of 503 upland cotton accessions in our laboratory.GWAS analysis was performed to identify associations of SSRs and SNPs with yield and fiber quality traits[22,37].This effort provided an opportunity to perform genetic mapping of QTL for stress resistance via GWAS using these SSR markers and SNPs.In the present study,both relative germination rate under salt(RGR-Salt)and cold(RGR-Cold)stress and the disease index of Verticillium wilt(DIV)were studied in this population to(i)evaluate the germination rate under salt and cold stress and DIV in the field,(ii)identify QTL for stress resistance via GWAS using SSRs and SNPs and(iii)to identify candidate QTL regions and genes associated with RGR-Salt.

2.Materials and methods

2.1.Plant materials and growth conditions

The mapping population consisted of 503 upland cotton accessions.Most of these accessions were from diverse regions of China,with the addition of some imported ones,including 225 from the Yellow River Region(YRR),141 from the Yangtze River Region(YtRR),79 from the Northwest Inland Region(NIR),28 from the Northern Specific Early Maturation Region(NSEMR),4 from the South China Region(SCR),20 from the United States(US),and 6 from the former Soviet Union(SU)[22].

Seeds of the population were collected from accessions that were planted in the following environments including Shihezi(SHZ)of north Xinjiang(NIR,N44.27°,E85.94°),Kuerle(KEL)of South Xinjiang(NIR,N41.68°,E86.06°),Yuanyang(YY)of Henan(YRR,N35.05°,E113.97°),and Huanggang(HG)of Hubei(YtRR,N30.44°,E114.87°)in 2012 and 2013 with two replications.

In September of 2012 and 2013,20 bolls were collected from the middle fruiting branches of each line in multiple environments.After delinting,residual fiber was removed with concentrated sulfuric acid.After low-quality seeds were discarded,the seeds were dried to improve seed sprouting.

2.2.Pre-experiment to confirm the best stress treatment

For each treatment,20 healthy,fully mature seeds were placed in a 90-mm-diameter Petri dish on two moistened filter paper disks and covered with another two disks(Fig.S1-A).Three treatments(distilled water control,salt stress,and cold stress)were applied to each accession,with two replications.

Before the formal experiments,some of the accessions,which were identified in this study or had been reported to show different levels of resistance,were selected for a gradient test to confirm the effectiveness of the stress treatments of salt concentration and cold temperature.For the salt-treatment germination test,the mass concentrations of the NaCl gradient were set to 0.4%,0.6%,0.8%,1.0%,and 1.2%,and for the cold treatment germination test,the temperature gradient was set to 10 °C,12 °C,and 15 °C.The Petri dishes were placed in a constant-temperature chamber at 25°C(or low temperature)for seven days without light.

To calculate the RGR,the number of germinating seeds was recorded after seven days.Seeds were considered to have germinated when the radicle was longer than half the length of the seed,and the relative germination rate was calculated as RGR=(germination number under stress treatment/germination number under control treatment)×100%.Treatment conditions were determined from the responses of these accessions under different treatments.

2.3.Germination test for all accessions

After the treatment conditions were established,germination of the seeds in three environments(HG,SHZ,and KEL)in 2012 and four environments(HG,YY,SHZ,and KEL)in 2013 was tested under salt stress.Germination under cold stress was evaluated in seeds under four environments(HG,YY,SHZ,and KEL)in 2013.Each experiment was performed twice.

To estimate phenotypic traits across multiple environments,the phenotypic values of the accessions from multiple environments were estimated via best linear unbiased prediction(BLUP)based on a linear model fitted with the lme4 R package[38].Broad-sense heritability,representing the proportion of total variance accounted for by genetic variance,was calculated for each trait[22].Summary statistics for all phenotypic data were calculated with SPSS 17.0(SPSS Inc.,Chicago,IL,USA).

2.4.Verticillium wilt resistance survey

In Shihezi,the end of August and early September is a peak incidence period of cotton VW every year and the main period for production loss caused by VW[39].Verticillium wilt resistance was investigated at this time in the test field of Shihezi Academy of Agricultural Sciences in 2012 and 2013.Fifty individuals from each line were collected from the field to investigate their resistance to VW according to the infected area of cross-section in the vascular bundle,and each survey was repeated twice.

Usually,there are five levels of VW disease:0(no infection),1(0-25%infection),2(25%-50%infection),3(50%-75%infection),and 4(75%-100%infection)[40].The final DIV of each line was calculated as DIV=∑ (VW disease level)/(4×investigated individuals)×100.

2.5.Association analysis

Marker-trait associations were calculated via the mixed linear model(MLM)association test incorporating Q(structure relatedness)+K(kinship)matrices into the TASSEL(version,5.0)software package[41].For phenotype-genotype association analysis,179 polymorphic SSR markers[22]and 11,975 SNPs[37]were used in a population panel of 503 upland cotton accessions.The P-values of markers associated with QTL were regulated by multiple testing correction via control of the false discovery rate[42].

2.6.Expression analysis of candidate genes

The transcriptome data sets of TM-1(G.hirsutum L.acc.Texas Marker-1)in different tissues were obtained from NCBI(https://www.ncbi.nlm.nih.gov/sra/?term=PRJNA248163)and CottonFGD(https://cottonfgd.org/)[43,44].Genesis software[45]was used to analyze the gene expression patterns.

For qRT-PCR analysis,seeds from four selected cotton(ZY41,ZY402,ZY304,and ZY420)accessions were germinated on solid culture media without or with 1%NaCl.The seeds were collected and immersed in liquid nitrogen at different time points and were frozen at-80°C for subsequent RNA extraction.RNA was extracted using the TIANGEN RNAprep Pure Plant Kit(Cat.#DP441,Beijing,China),following the manufacturer's instructions.High-quality RNA(3 μg)was reverse transcribed to cDNA using SuperScript III Reverse Transcriptase(Cat.No.18080-093,Invitrogen)in accordance with the manufacturer's instructions.qRT-PCR experiments were performed using an ABI Prism 7500 system(Applied Biosystems,USA),and the gene expression levels were calculated via the comparative Ct(2-ΔΔCt)method.GhUBQ7(GenBank accession number DQ116441)served as an internal control.Gene-specific primers were designed according to the cDNA sequences using the Primer Premier 5.0 software and were synthesized commercially(Genscript Bioscience,Nanjing,China).The primers used for qRT-PCR analyses are listed in Table S1.

3.Results

3.1.Optimal salt-and cold-stress treatments

Our preliminary experiments showed that accessions ZY62(Xinluzao 17),ZY35(Jinmian 8),and ZY48(Yumian 3)were respectively high-resistance,medium-resistance and sensitive to salt stress.Fig.S1-B shows that a significant phenotypic difference appeared when the salt concentration was 1.0%and 1.2%and that when the concentration reached 1.2%,the inhibition was too strong for germination of the sensitive accession ZY48.Accordingly,1.0%NaCl treatment was selected as the most suitable treatment concentration.

Accessions ZY163 (Zhongmiansuo 36)and ZY485(Xinluzao 26)were respectively high-resistance and sensitive to cold in a previous study[46].Another eight accessions were randomly selected as controls.Among these accessions,there were almost no germinated seeds at 10°C and 12 °C.When the temperature rose to 15 °C,some accessions began to germinate and showed a difference(Fig.S1-C).Accordingly,15°C was selected as a suitable treatment temperature for cold stress.

3.2.Phenotypic statistical analysis under different stress treatments

Differences in plant growth environments and uneven fertility affecting the vigor of cotton seeds might lead to inconsistent stress-resistance results.The germination rates,under normal conditions,of seeds from seven regions,including 2012HG,2012SHZ,2012KEL,2013HG,2013YY,2013SHZ,and 2013KEL,were evaluated.The mean germination rates ranged from 88%to 96%(Fig.S2).The accessions planted in HG showed lower germination rates than those planted in other regions in both 2012(88%)and 2013(89%),while seeds from SHZ showed the highest germination rate.The germination rates increased by 5%in 2013SHZ,3%in 2013YY,and 2%in 2013KEL,compared with 2013HG.

The RGR of the cotton accessions was severely affected and showed continuous variation under salt and cold treatments(Figs.1A-D,S3).Summary statistics are presented in Table 1.The mean RGR-Salt values varied from 29%to 45%.The mean RGR-Salt values of accessions planted in HG(29%and 33%)were lower than those in other regions in both 2012 and 2013,and the accessions planted in SHZ showed the highest mean RGR-salt(45%and 43%).A similar trend was observed in the germination rate of seeds under normal conditions,indicating that salt resistance was affected by seed vigor in cotton.The correlations between RGR-Salt traits under seven environments are presented in Table S2. Although significant(P≤0.01)correlations between two of the seven environments were found,the correlation coefficients were relatively low and might have been strongly affected by environmental differences.There were higher correlation coefficients between accessions planted in the same region.For RGR-Salt in SHZ as an example,the correlation between 2012SHZ and 2013SHZ was 0.71.

For RGR-Cold,the mean RGR-Cold values varied between 46%(2013HG)and 61%(2013SHZ)(Table 1).The RGR-Cold values for 2013YY and 2013KEL were 53%and 51%,respectively.Consistently with RGR-Salt,the RGR-Cold values in different environments showed a common trend in which those for seeds from HG were significantly lower than for those from other regions,followed by KEL and YY.Seeds from SHZ showed the highest resistance.For RGR-Cold traits,there were significant(P≤0.01)correlations among 2013HG and 2013YY,2013HG and 2013SHZ,and 2013SHZ and 2013KEL(Table S3).

The phenotypic values for RGR-Salt and RGR-Cold in different environments were determined via BLUP.Table S4 describes the statistics of the two sets of data.The broadsense heritability values show that environmental variation had a greater influence on RGR-Cold than RGR-Salt,and the correlation test shows that there was a significant positivecorrelation between RGR-Salt and RGR-Cold.The SD indicates that RGR-Salt showed a larger variance than RGR-Cold.

Table 1-Summary statistics for RGR-Salt,RGR-Cold,and DIV.

The disease index of VW(DIV)in SHZ was measured in 2012 and 2013(Fig.1E-G;Table 1),and the mean disease indexes were 75 and 88,respectively.This result shows that the resistance of the accessions planted in 2012 was greater than those planted in 2013.However,the 2012 disease index showed greater phenotypic variation.

3.3.Genome-wide association mapping with SSR markers

A previous study[22]showed that the population of 503 accessions could be classified into seven subpopulations,and the attenuation distance of LD decreased dramatically to 0-5 cM.A total of 179 SSR markers were used for marker-trait association using the MLM-Q-K model.A total of 54 significant(P＜0.05)markers were associated with these three traits using both single-environment and BLUP estimates of trait values.

Fig.1-Boxplots(A,C,E)and frequency distributions(B,D,F,G)of RGR-Salt(BLUP),RGR-Cold(BLUP),and DIV.RGR-Salt and RGRCold represent relative germination rates of seeds under salt and cold stress,respectively;DIV represents the disease index of Verticillium wilt.

Thirty-one markers were associated with RGR-Salt in seven environments and were unevenly distributed on 14 chromosomes,with a maximum of four markers on Chr19 and Chr25,respectively(Fig.2,Table S5).Eight markers(HAU2770,HAU0590,BNL3790,BNL3635,NAU3084,NAU5480,MONCGR5167,and MON_DC40260)were associated with the BLUP value of RGR-Salt(Table S6).These eight markers were also identified in the 31 markers were associated with RGRSalt in seven environments.HAU2770 showed the lowest P-value(0.001),and was detected in five environments(2012HG,2012SHZ,2012KEL,2013KEL,and 2013SHZ).BNL3790 showed the highest phenotypic variance explained(PVE)both in 2012HG(7.73%)and in 2013HG(8.40%).

There were 15 markers associated with RGR-Cold in four environments,and they were unevenly distributed on 11 chromosomes,with a maximum of three markers on Chr21(Fig.2,Table S7).MON_DC30153 and DPL0461 were detected in two environments.Ten markers were associated with the BLUP value of RGR-Cold,and the mean PVE was 1.85%,ranging from 1.16%to 3.27%;five of them(MON_CGR6869,NAU2985,MON_DC30153,MON_COT125,HAU2770)were detected in a single environment,and MON_CGR6869 showed the highest PVE(Table S8).

Fifteen markers were associated with DIV in two environments(2012SHZ and 2013SHZ),and they were unevenly distributed on 12 chromosomes,with a maximum of two markers on Chr1,Chr3,and Chr10,respectively(Fig.2,Table S9).The PVE ranged from 1.12%(T_CO114183c)to 7.95%(MON_SHIN_0652)at a mean of 3.36%.Of these 15 markers,only BNL3790 was detected in two years(2012SHZand 2013SHZ).

Among these 54 markers,some were associated with multiple traits(Fig.2).HAU0590 was associated with RGRSalt,RGR-Cold and DIV.Four markers(BNL846,BNL3790,NAU5480,and BNL3347)were associated with RGR-Salt and DIV;MON_CGR6869 and MON_SHIN_1343 were associated with RGR-Cold and DIV;and HAU1321,MON_DC30153 and HAU2770 were associated with RGR-Salt and RGR-Cold.

Markers were not evenly distributed on chromosomes.Chr01(A01),Chr10(A10),Chr11(A11),Chr12(A12),Chr15(D01),Chr18(D13),Chr19(D05),Chr21(D11),Chr25(D06),and Chr26(D12)carried more than three markers associated with stress resistance,but there were no associated markers on Chr06(A06),Chr09(A09),Chr13(A13),Chr16(D07)and Chr22 (D04)(Fig.2).The homoeologous chromosomes Chr01(A01)-Chr15(D01),Chr05(A05)-Chr19(D05)and Chr11(A11)-Chr21(D11)carried a total of 25 markers,accounting for 45%of all the detected markers.Compared with the At genome(Chr01-Chr13),the Dt genome(Chr14-Chr26)carried more resistance-associated markers(57.4%).

5. Frog: Frogs symbolize20 new life in many cultures and thus often appear helpful or kind in folklore22. However, frogs also have connections with witchcraft23, often as witches familiars, and are despised by some religious groups for that reason (Philip 1997).

3.4.Genome-wide association analysis of RGR-Salt,RGR-Cold,and DIV using SNP markers

Using 11,975 SNP markers from a high-density genetic map based on a 63 K cotton array,phenotypic data of RGR-Salt,RGR-Cold,and DIV were used to identify association with genome-wide SNPs.Quantile-quantile plots showed that the MLM model(Q+K)was optimal and could be used to identify association signals(Figs.3A-C,S4-A,S5-A,B,S6-A,B).The GWAS results are shown in Manhattan plots,and the significance level was-lg P≥4.078(P≤1/11,975,with 11,975 indicating the number of test SNPs).Six different SNPs(i09541Gh,i19566Gh,i02237Gh,i02243Gh,i45568Gh,and i47111Gh)in seven environments and two SNPs(i02237Gh and i02243Gh)in the BLUP results were associated with RGRSalt,with the PVE ranging from 3.96 to 5%(Fig.3D-F,S4-B;Table 2).It is worth noting that i02237Gh and i02243Gh were detected in multiple environments(2012HG,2012SHZ and 2013SHZ)and were located at respectively 112.5 cM and 112.51 cM on Chr15(D01).SNPs i02237Gh and i02243Gh with the T/C and A/C SNP alleles,respectively.The SNP in i02237Gh was identified in 78 C-type and 393 T-type cotton accessions,and the mean RGRs of accessions with the C and T alleles were 43%and 38%,respectively.There was a significant difference between C-allele accessions and T-allele accessions(t-test,P=5.81546E-07).Accessions carrying the C allele showed higher and those with the T-allele lower RGR than the mean(39%)(Fig.4A).The SNP in i02243Gh SNP was identified in 88C-type and 231 A-type cotton accessions.The mean RGR of the accessions with the C allele(42%)was significantly greater than those of the accessions with the A allele(38%)(t-test,P=1.93E-05)(Fig.4A).i45568Gh(T/C)and i47111Gh(T/G)were identified in 2012 KEL with R2of 5%and 4.85%,respectively,explaining high phenotypic variation(Table 2).The RGR of accessions with the favorable T allele at i45568Gh(41%)and i47111Gh(41%)were both higher than the mean RGR for accessions with the unfavorable C allele at i45568Gh(38%)and the G allele at i47111Gh(37%)(t-test,P=0.003083,P=0.002397,respectively)(Fig.4A).The numbers of favorable SNP alleles affected salt resistance.A higher RGR(44%)was observed in cotton accessions that carried all four favorable SNP alleles than in those either without all four(36%)or with fewer such alleles(Fig.4B).Thus,favorable SNP alleles showed additive effects on salt resistance in cotton.

For RGR-Cold,one SNP(i48098Gh)in one environment(2013YY)and two SNPs(i43138Gh and i18259Gh)in the BLUP results were associated with RGR-Cold,with PVE ranging from 2.19%to 4.00%(Fig.S5-C,D;Table S10).The SNP in i18259Gh(A/G)was identified in 51 G-type and 441 A-type cotton accessions,and the mean RGRs of the accessions with the G and A alleles were 51%and 53%,respectively,and showed a significant difference(t-test,P=3.51E-05)(Fig.S5-E).For DIV,one SNP(i36180Gh)in 2012SHZ and three SNPs(i02907Gh,i52094Gb and i59804Gb)in 2013SHZ were associated with DIV,with PVE ranged from 4.09%to 5.11%(Fig.S6-C,D;Table S11).The SNP i36180Gh(C/T)was identified in 59 T type and 225 C-type cotton accessions,and the mean DIVs of the accessions with T-allele and C-allele were 80 and 72,respectively,and showed a significant difference(t-test,P=1.31E-05).The DIVs of accessions with the T-allele was higher and of those with the C-allele lower than the mean DIV(78)(Fig.S6-E).

3.5.Prediction of candidate genes of seed germination under salt stress

A previous study[37]has shown that the LD decay distance for the same population between all SNP markers was 6.1 cM,when the value of the cutoff for squared correlations of allele frequencies(r2)was 0.1.Accordingly,we determined candidate gene intervals based on this LD decay,within 6.1 cM on either side of each significant SNP.

Fig.2-Distribution of SSR markers associated with stress-resistance traits.Red color indicates that the marker was associated with only one trait,and green or yellow colors indicate that the marker was associated with two or three traits.

Fig.3-Quantile-quantile and Manhattan plots for the GWAS for RGR-Salt.A-C.Quantile-quantile plots for RGR-Salt in 2012HG(A),2012SHZ(B),and 2013SHZ(C).D-F.Manhattan plots for RGR-Salt in 2012HG(D),2012SHZ(E),and 2013SHZ(F).

For the RGR-Salt trait,two SNPs(i02237Gh,i02243Gh)on D01 could represent stable genetic loci responsible for salt resistance in cotton.There were 49 SNPs located in this candidate gene interval. The physical locations of the sequences containing these SNPs were mapped to the TM-1 genome sequence[43],and i02237Gh and i02243Gh were associated with the same candidate gene region (D01,37771-1942912).A total of 223 genes were found in this region.Homology analysis and GO(Gene Ontology)biological process analysis showed that 10 genes(Gh_D01G0008,Gh_D01G0018,Gh_D01G0027,Gh_D01G0057,Gh_D01G0081,Gh_D01G0167,Gh_D01G0180,Gh_D01G0181,Gh_D01G0197,Gh_D01G0202)were involved in stress response,seed germination,or dormancy(Table 3).

Table 2-SNPs associated with RGR-Salt.

Fig.4-Identification of favorable SNP alleles associated with RGR-Salt.A.Relative germination rate of accessions with different SNP alleles at the SNPs i02237Gh,i02243Gh,i45568Gh,and i47111Gh.B.Relative germination rate of accessions with different numbers of favorable alleles(FA).

To investigate the expression patterns of these 10 genes during seed germination,public transcriptome data sets from different tissues,including germinating seed,root,stem,leaf,petal,anther,stigma,ovule,and fiber,were collected.The results showed that all 10 genes were expressed at the seed germination stage(FPKM≥1),except for GhMYB73and GhGER5(Fig.S7).In addition,the tissue expression patterns of these genes were verified using qRT-PCR analysis.In agreement with the transcriptome data sets,three genes(GhPIP3A,GhTZF4,GhTZF4a)showed preferential expression in germinating seeds compared with other tissues(Fig.5A).

To investigate the candidate genes for seed germination under salt stress,the expression patterns of these 10 genes were analyzed in two sensitive(ZY41,ZY402)and two tolerant(ZY304,ZY420)accessions(Fig.5B).All genes showed detectable expression under control and salt stress,except for GhGER5(Fig.5C).GhCSP3,GhSNRK2.6,GhMYB73,and GhERF53 showed upregulated patterns in the sensitive and tolerant accessions,and GhACT7 showed downregulated expression patterns,but there were no obvious differences between sensitive and tolerant accessions.It is noteworthy that GhPIP3A and GhSAG29 and two adjacent genes,GhTZF4 and GhTZF4a,were significantly induced and differently expressed in the sensitive and tolerant accessions under salt stress.The expressions of GhPIP3A and GhSAG29 were significantly induced by salt stress in the tolerant accessions,and their expressions were much higher than those in the sensitive accessions.However,GhTZF4 and GhTZF4a showed higher expression in the sensitive than in the tolerant accessions.GhPIP3A encodes a plasma membrane intrinsic protein(PIP)whose sequence shares 89%amino-acid identity with AtPIP2;7(AT4G35100),which was involved in the adjustment of plant water balance in response to salt stress(Fig.S8-A)[47];GhSAG29 encodes a plasma membrane protein sharing 58%identity with AtSAG29(AT5G13170),which is involved in response to abscisic acid and osmotic stress(Fig.S8-B);GhTZF4 and GhTZF4a showed high similarity(96%identity)with each other and respectively 43%and 42%identity with Arabidopsis AtTZF4(AT1G03790),a CCCH-type zinc finger protein that negatively regulates seed germination(Fig.S8-C)[48,49].These four genes showed expression patterns that differed between sensitive and tolerant accessions,and were predicted to be candidate genes associated with RGR-Salt.

Table 3-Candidate genes in the genomic region of SNP most highly associated with RGR-Salt in cotton.

Fig.5-Expression analysis of 10 candidate genes.A.Tissue expression pattern analysis of candidate genes by qRT-PCR.S,12 h germinated seeds;SR,radicle of 24 h germinated seeds;SC,cotyledon of 24 h germinated seeds;R,root;ST,stem;L,leaf;PE,petal;ST,stigma.Bars represent means±standard error(n=3).B.Relative germination rate of two sensitive(ZY41 and ZY402)and two tolerant(ZY304 and ZY420)accessions.C.Gene expression of candidate genes during the germination stage under salt stress by qRT-PCR.Control represents 12 h germinated seeds without salt treatment.The qRT-PCR results are shown as a heat map.

4.Discussion

High salinity,low temperature,and VW are stress factors that restrict cotton production worldwide.In previous research on cotton breeding,scientists mainly focused on yield and fiber quality traits,while research on resistance to various stresses has progressed at a slow pace.Understanding the mechanisms of stress tolerance and breeding multiple-stress-tolerant cotton varieties are imperative for breeding programs.

Mapping stress-related QTL/genes and cultivating new resistant varieties through marker assisted selection(MAS)are considered effective methods for increasing stress tolerance in cotton[40].However,population representativeness and marker density have become the main limiting factors for QTL mapping in cotton.Previous studies[22,37]have identified 503 accessions that showed high levels of genotypic and genetic variation,and these were selected as an ideal resource for association mapping.They were collected mainly from the five major cotton-producing areas in China,with a small number from other countries and regions.To obtain accurate phenotypic data,the seedgermination experiments were investigated in a multi-plot demonstration for two years and were repeated twice to eliminate environmental influence,and abundant phenotypic variation was obtained in RGR-Salt,RGR-Cold and DIV traits(Figs.1,S3).In addition,179 polymorphic SSR markers and 11,975 high-resolution SNPs for stress-tolerance association analysis.These markers have been successfully used for genome-wide association mapping for agronomic traits in previous studies[22,37].These ideal population representations and high-resolution markers represent useful resources for QTL mapping for stress tolerance in cotton.

In contrast with previous studies,seed germination rates under salt and cold stress were used for QTL mapping in the present study.The RGR-Salt and RGR-Cold of cotton accessions were severely affected by these two stresses.Thus,there is an important significance to investigate the QTL mapping at the seed-germination stage.The identification of RGR-Salt and RGR-Cold in 503 cotton accessions can provide information on excellent resistance accessions for cotton breeding,such as Jinmian 21,Yukang 1,Jinzhong 200,Zhongmiansuo 36,and Binmian 1.These accessions can also be the basis for breeding accessions for cotton-resistance research.In the present study,the RGR-Salt and RGR-Cold were significantly affected by different environments,and they were consistent with the germination rate under normal conditions.This finding indicated that seed vigor played an important role in regulating seed germination under salt and cold stresses.Previous studies have showed that seed vigor was known as a comprehensive trait and was affected by genetic and environmental factors[50].Compared with seed from other regions,the seeds with the highest germination rate came from SHZ.SHZ may provide more suitable conditions during the process of seed development and maturation,and can be considered an ideal seed-production region.

Resistance in plants is generally controlled by multiple genes,and there are some common metabolic pathways involved in response to external stress.In the association results for stress resistance,certain markers were associated with two or three traits,suggesting that one gene may control multiple phenotypes[51].In the present study,we also identified some markers could be simultaneously associated with more traits,such as HAU0590.These results suggest that certain genes are involved in multiple stress resistances,and these markers can provide helpful information about resistance mechanisms.More markers associated with salt and cold resistance were found in the Dt than in the At subgenome.This result was consistent with previous studies[17,34]showing that more stress-associated genes are distributed in the Dt subgenome.

Interaction between genotype and environment(G×E)is an important component of quantitative traits.In the present study,most significant markers were detected in only one environment,suggesting that G×E may be an important factor in QTL mapping of cotton stress.Some significant markers were detected in multiple environments,such as HAU2770,MON_DC40260 and NAU5480 for RGR-Salt,MON_DC30153 and DPL0461 for RGR-Cold,and BNL3790 for DIV.These stable markers associated with stress traits across different environments can be used in molecular marker assisted breeding for cotton stress tolerance.

Salt stress and VW resistance-associated markers have been widely studied in recent years.To identify reliable QTL for those stresses,we compared our results with those of previous studies.Du et al.[52]compiled 145 SSR markers and performed association analysis for ten salt-tolerance-related traits,detecting 95 significant associations.BNL3436 was associated with relative SOD activity,relative POD activity,and relative MDA content(RMDA),NAU6755 was associated with relative chlorophyll content(RCC).These two markers(BNL3436 and NAU6755)were significantly associated with RGR-Salt in the present study.MON_DC20027,HAU2770,and HAU0591 showed the relatively adjacent genomic physical distance(＜1 Mb)with the reported salt-related markers NAU2508,NAU5418,and BNL3594,respectively.For DIV,owing to the highly complex inheritance and external factors such as mapping populations,inoculation methods and environmental influence on disease symptoms,some reports[35,53]indicated that it is difficult to identify reliable QTL for VW resistance.Although none of the DIV-associated markers had been identified in with previous studies,BNL3347,NAU2858,and BNL3790 showed the relatively adjacent genomicphysical distance(＜1 Mb)with the reported VW resistance-associated markers,TMB1295,BNL0663,and NAU2869,respectively[35,54,55].TMB1295 has been identified as a significant marker for increasing VW resistance and is physically close(1.6 kb)to BNL3347 in the genome[35].All of the mentioned markers could be considered to be linked with reliable QTL and be used for MAS in the future.

In comparison with SSR markers,high-density SNP markers can increase the accuracy and resolution of QTL mapping.Thus,11,975 SNP markers and stress-related traits were used for GWAS in this study. Previous study [56] has indicated that marker-based gene pyramiding is an effective strategy for MAS.In the present study,four SNPs with favorable alleles showed additive effects on RGR-Salt,and could be applied for the development of salt tolerant cotton cultivars in future breeding programs. In addition,i18259Gh(A/G)and i36180Gh(C/T)were strongly associated with RGR-Cold and DIV,respectively.These identified genotypes will be beneficial for further breeding strategies.

Based on detailed gene homolog annotation with Arabidopsis and on gene expression patterns,four genes(GhPIP3A,GhSAG29,GhTZF4 and GhTZF4a)were predicted to be candidate genes associated with RGR-Salt.Plasma membrane intrinsic proteins have been reported[47,57]to be involved in adjusting the tissue water balance in plants in response to stress conditions and in facilitating the diffusion of water through the plasma membrane.AtPIP2;7 is a homolog of GhPIP3A,and overexpression of AtPIP2;7 in Arabidopsis plants induced a sixfold increase in root hydraulic conductivity of four-week old seedlings.GhPIP3A was significantly induced at40 h and 48 h at the germination stage and showed higher expression levels in tolerant genotype accessions.We speculate that increased accumulation of GhPIP3A would promote the imbibition of water(Fig.5C).AtSAG29 was induced by salt stress,ABA and senescence in Arabidopsis,and AtSAG29-overexpressing plants showed hypersensitivity to salt stress and reduced cell viability in roots[58].No gene functions of AtSAG29 at the seed germination stage in Arabidopsis have been reported.In our study,GhSAG29 was significantly induced in tolerant cotton accessions at the seed-germination stage.Previous studies[49,59]have shown that tandem CCCH zinc finger proteins(TZFs)are conserved among animals,yeast and plants,and that TZFs are involved in hormone-mediated environmental responses.Arabidopsis AtTZF4was expressed specifically inseeds,and the expression decreased during imbibition[49].In our study,GhTZF4 and GhTZF4a were preferentially expressed in germinating seeds compared with other tissues,and their expression levels were significantly decreased in 40 and 48 h compared with that at 20 h under salt stress(Fig.5C).Contrasting with GhPIP3A and GhSAG29,GhTZF4 and GhTZF4a showed higher expression in sensitive than in tolerant accessions.Combining our results with those from previous researches,the homologous gene Arabidopsis AtTZF4 acts as a negative regulator for seed germination,and we speculate that GhTZF4 and GhTZF4a serve as factors regulating seed germination in cotton[48,49].Future studies may verify the functions of these four genes and explore their application in the breeding of salt-tolerant cotton.

Acknowledgments

This work was supported by the National Key Research and Development Program of China(2016YFD0101006)and the Fundamental Research Funds for the Central Universities(2662016PY001).

Appendix A.Supplementary data

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2018.11.002.