Genqio Li,Xingyng Xu*,Chengheng TnBrett F.Crver,Guihu Bi,

Xuewen Wangd,J.Michael Bonmane,Yanqi Wub,Robert Hungerf,Christina Cowgerg

a Wheat,Peanut and Other Field Crops Research Unit,USDA-ARS,Stillwater,OK 74075,USA

b Plant and Soil Science Department,Oklahoma State University,Stillwater,OK 74078,USA

c Hard Winter Wheat Genetics Research Unit,USDA-ARS,Manhattan,KS 66506,USA

d Department of Genetics,University of Georgia,Athens,GA 30602,USA

e Small Grains and Potato Germplasm Research Unit,USDA-ARS,Aberdeen,ID 83210,USA

f Entomology and Pathology Department,Oklahoma State University,Stillwater,OK 74078,USA

g Plant Science Research Unit,USDA-ARS,Raleigh,NC 27695,USA

Keywords:Wheat Pow dery m ildew GWAS Linkage m apping Genom e-w ide linkage disequilibrium

A B S T R A C T Wheat pow dery m ildew(Blumeria graminis f.sp.tritici,Bgt)is a disease of increasing im portance globally due to the adoption of high yielding varieties and m odern sustainable farming technologies.Growing resistant cultivars is a preferred approach to managing this disease,and novel pow dery m ildew resistance genes are urgently needed for new cultivar developm ent.A genom e-w ide association study w as perform ed on a panel of 1292 w heat landraces and historical cultivars using 5011 single nucleotide polym orphism(SNP)m arkers.The association panel was evaluated for reactions to three Bgt inoculants,OKS(14)-B-3-1,OKS(14)-C-2-1,and Bgt15.Linkage disequilibrum(LD)analysis indicated that genom e-w ide LD decayed to 0.1 at 23 Mb,and population structure analysis revealed seven subgroups in the panel.Association analysis using a m ixed linear m odel(MLM)identified three loci for powdery mildew resistance on chromosome 2B,designated QPm.stars-2BL1,QPm.stars-2BL2,and QPm.stars-2BL3.To evaluate the efficacy of GWAS in gene discovery,QPm.stars-2BL2 w as validated using F2 and F2:3 populations derived from PI 420646×OK1059060-126135-3.Linkage analysis delimited the powdery mildew resistance gene in PI 420646 to an interval w here QPm.stars-2BL2 w as located,lending credence to the GWAS results.QPm.stars-2BL1 and QPm.stars-2BL3,which w ere associated w ith four SNPs located at 457.7-461.7 Mb and tw o SNPs located at 696.6-715.9 Mb in the Chinese Spring reference IWGSC Ref Seq v1.0,respectively,are likely novel loci for pow dery m ildew resistance and can be used in w heat breeding to im prove pow dery m ildew resistance.

1.Introduction

Wheat(Triticum aestivum L.),unrivalled in its range of cultivation,is a major staple crop that provides about 16%of total dietary calories consum ed by hum ans[1].Adoption of sem i-dw arf varieties and high-input cultivation technologies such as increasing fertilizer input and irrigation has revolutionized w heat production w orldw ide.As yield increases,pow dery m ildew,caused by Blumeria graminis f.sp.tritici(Bgt),has becom e m ore and m ore im portant in m any w heatgrow ing regions.Pow dery m ildew has been one of the dominant diseases in Europe,China and m any other w heat grow ing regions w orldw ide[2].With a rapid increase of the hum an population,m ore intensive farm ing practices are required to m eet challenges,m ore likely resulting in m ore favorable environm ents for pow dery m ildew epidem ics.

Grow ing resistant cultivars is an effective approach to minimizing pow dery mildew dam age.More than 50 powdery mildew resistance genes(Pm1-Pm62;Pm19=Pm1c;Pm22=Pm1e;Pm23=Pm4c;Pm31=Pm21)have been officially named,and multiple functional alleles have been identified at several loci including five at Pm1,three at Pm2,18 at Pm3,five at Pm4,five at Pm5,and two at Pm24.Tw enty-five additional resistance genes have been temporarily named,and a number of QTLfor pow dery mildew resistance have been m apped in different w heat cultivars[3-8].How ever,most pow dery m ildew resistance genes originated from wheat wild relatives and many have not been widely used in wheat breeding because of linkage drag.Also,the evolution of new Bgt pathotypes that overcome pow dery m ildew resistance genes is common even to the extent that some genes have lost their effectiveness before they w ere w idely utilized in commercial cultivars[9].

In m ost w heat breeding program s,only a lim ited num ber of pow dery m ildew resistance genes are used.The narrow genetic basis of pow dery mildew resistance is m ore pronounced in China.A recent study revealed that 580 out of 908 breeding lines in China carried the Pm8 gene,w hich has lost effectiveness to Bgt isolates[10].Sim ilar results w ere reported in the U.S.A,w here a considerable percentage(37.5%)of recently released pow dery m ildew resistant breeding lines in the hard red w inter w heat grow ing regions carried the Pm3a gene[11].Given that it m ay take years of breeding to elim inate linkage drag from alien chrom osome fragments harboring pow dery m ildew resistance genes,identification of novel pow dery m ildew resistance genes in bread w heat is of great urgency and agronom ic im portance.

Genetic m apping has been w idely used to characterize new pow dery m ildew resistance genes.Most pow dery m ildew resistance genes or QTLhave been located to specific genom ic regions using traditional linkage mapping.More recently,genome-wide association studies(GWAS)provide an alternative approach to quickly screen a germ plasm collection for genes of interest w ith relatively low cost,and a gene-level high resolution has been achieved in crops w ith abundant genom ic resources[12-17].

GWAShas also been w idely used in w heat gene discovery,leading to the identification of novel loci for yield[18],quality[19],coleoptile length[20],leaf tip necrosis and pseudo black chaff[21],as w ell as resistance to leaf rust[22,24,25],stripe rust[24,26],stem rust[27],and wheat soil-borne mosaic virus[28].A grow ing body of literature confirms GWAS as an efficient tool for identifying genes of agronom ic im portance in w heat,but not for novel pow dery m ildew resistance loci yet.

The USDA-ARSNational Sm all Grain Collection(NSGC)has＞40,000 bread w heat accessions,from w hich an inform ative core subset of 3230 accessions w as assem bled[29].A recent study suggested that the NSGC core set has valuable genetic variation associated w ith powdery mildew resistance[10].The objective of this study w as to explore novel loci for pow dery m ildew resistance in the w inter habit accessions from the NSGC core subset using GWASand linkage m apping.

2.Materials and m ethods

2.1.Plant materials

A panel of 1292 w inter/facultative w heat accessions w as random ly selected from NSGC inform ative core set of germplasm,which includes 1674 winter-facultative accessions and 1556 spring accessions[29].These selected accessions w ere from 75 countries including 636 landraces,110 breeding lines,385 registered cultivars,10 genetic stocks,and 151 accessions with uncertain improvement status(Table 1).All seeds w ere provided by the NSGC.A set of 28 w heat differential lines w ith know n pow dery m ildew resistance genes w ere used to characterize the Bgt isolates used in this study.

For linkage m apping,an F2population and a set of 177 F2:3lines derived from PI 420646×OK1059060-126135-3 w ere used to confirm one of the pow dery m ildew resistance loci identified in GWAS analysis.PI 420646 is a Kazakhstani cultivar in the association panel,and carries the resistance allele at the target locus.In addition,Chinese Spring and its deletion lines,2BL-3,2BL-4,2BL-5,2BL-1,and 2BL-6,w ere used to determine the physical location of the powdery m ildew resistance gene in PI 420646.

2.2.Evaluation of powdery mildew resistance

The 1292 accessions w ere evaluated for their responses to the Bgt inoculants Bgt15,OKS(14)-B-3-1 and OKS(14)-C-2-1 at seedling stage under the controlled greenhouse conditions in 2015 and 2016.Bgt15 w as collected from Oklahom a w heat fields in 2015,representing a bulk of com m on Bgt pathotypes in Oklahom a[10].Bgt isolates OKS(14)-B-3-1 and OKS(14)-C-2-1 w ere collected from w heat fields in Oklahom a in 2014,and maintained as single-pustuled pure isolates on detached leaves by the USDA-ARS Plant Science Research Unit at Raleigh,North Carolina.A random ized com plete block design with tw o replicates w as used in this study.Ten seeds of each entry w ere planted in tw o cells in 73-cell grow ing trays containing Sunshine Redi-earth grow ing m ix(SubGro Horticulture Cabada Ltd.).The resistant control(TAM 110),w hich confers high resistance to current Bgt isolates in Oklahoma,w as seeded in one cell,and the susceptible check(Jagalene)in two cells of each tray.In the spring of 2015,seedlings at the tw o-leaf stage were inoculated w ith Bgt15 conidia using a method described previously[10],and then placed in a greenhouse at 20±2°C with natural light.Infection types(ITs)were scored using a 0-4 scale at 14 days after inoculation,w ith 0 and 1 as highly resistant,2 moderately resistant,3moderately susceptible,and 4 highly susceptible[30].All tested plants were re-exam ined in tw o days.Wheat accessions with an IT score≤3 w ere subjected to a confirmation test using the same experimental design.The association panel was evaluated for reactions to Bgt isolates OKS(14)-B-3-1 and OKS(14)-C-2-1 in the spring of 2016 using the sam e procedures.

Table 1-Distribution of 1292 w heat accessions used in this study.The num ber of countries represented by this panel in each continent is given in p arentheses.

In addition,an F2population derived from a cross PI 420646×OK1059060-126135-3 w as evaluated for resistance to Bgt isolate OKS(14)-B-3-1 in 2016,and transferred to a greenhouse after being vernalized at 4°C for six w eeks.The F2:3lines w ere evaluated for resistance in 2017 using the same isolate.The experim ent used a random ized com plete block design w ith tw o replicates.In each replicate,16 plants w ere grow n and tested for each line.

2.3.GWASanalysis

A previous study suggested that square root transformation can significantly improve the normality of infection type(IT)data[26].Given that the IT distribution severely skewed to susceptibility(Fig.1),we used square root transformed IT data[ITtransf=]to perform GWAS analysis.The NSGC core subset w as previously genotyped w ith 5011 SNPm arkers,and the m inor allele frequencies of these markers are at least 5%[29].SNP data are available at http://w w w.triticeaecap.org/,and 4716 of them w ere previously m apped to the 90 K SNPconsensus m ap[31].Given that the new version of Chinese Spring reference sequence has been released recently,we further mapped 3544 of them to the Chinese Spring reference assembly IWGSC RefSeq v1.0(http://urgi.versailles.inra.fr/).These SNPs w ere used to estimate genome-w ide linkage disequilibrium(LD)using the R package“adegenet”[32].The STRUCTURE2.3.4 program[33]was used to determ ine the subpopulation number and estimate the Q matrix that corrects effects of population structure.The mixed linear model(MLM)approach[34]implemented in TASSEL 5[35]was run with the optim um level of com pression and the P3D approach to identify loci governing Bgt responses in the panel.SNPmarkers and the Q m atrix w ere used as fixed effects,and a kinship matrix calculated w ith the scaled Identity-By-State(IBS)method[36]was used as a random effect component in the MLM model.A false discovery rate(FDR)of 0.01 w as set as a threshold for claiming significant marker-trait associations[37].

Fig.1-Decay in genom e-w id e linkage d isequilibrium(LD),rep resented as the squared allele correlation coefficient(r2)betw een intra-chrom osom al SNPs,over p hysical distance.The blue curve show s the fitted m odel,and red-dashed lines suggest that LD declines to 0.1 at about 23.0 Mb.

2.4.Linkage analysis

Genom ic DNA w as extracted from tw o-w eek-old leaves of each F2plant using a protocol described by Dubcovsky et al.[38].The bulked segregant analysis(BSA)[39]was used to map the pow dery m ildew resistance gene in PI 420646.The resistant and susceptible bulks w ere constructed by pooling equal amount of DNA from 10 highly resistant plants(IT=0)and 10 highly susceptible plants(IT=4),respectively.The tw o contrasting bulks and parental DNA w ere used to screen a set of over 600 preselected SSR m arkers that are evenly distributed over all 21 chromosom es using a protocol described by Xu et al.[40].The polym orphic SSRm arkers betw een the bulks w ere subsequently used to genotype the F2population.Linkage analysis w as used to confirm their linkage w ith the pow dery mildew resistance gene in PI 420646.Then,SSRand STSm arkers that w ere previously m apped in the target region w ere used to genotype the F2population.GMATA softw are[41]w as used to identify SSR loci in the target region of the Chinese Spring reference sequence(IWGSC Ref Seq v1.0.https://urgi.versailles.inra.fr/),and prim ers w ere designed for those in non-transposon regions.

The genotype of each F2plant w as inferred from the phenotype of the corresponding F2:3line.Linkage analysis w as conducted using Mapm aker 3.0b[42]and the Kosam bi function[43]w as used to convert recom bination fraction to genetic distance.A logarithm of the odds score of 3.0 w as set as a threshold to declare linkage betw een resistance gene and m arkers.

2.5.Identification of the T.timopheevii chromosome segment harboring the Pm6 gene

The STSm arker NAU/STSBCD135-2(hereafter BCD135-2),w hich cosegregated with the Pm6 gene[44],was used to identify the T.timopheevii-derived Pm6 from a set of 75 accessions w ith high resistance to OKS(14)-B-3-1.Primer sequences and PCRconditions were described by Jiet al.[44],and PCRproducts wereseparated in 6%polyacrylamide gel stained with ethidium bromide.

3.Results

3.1.LD analysis and population structure

A total of 3544 SNPs w ith 1656 SNPs m apped on the A genom e,1679 mapped on the B genome,and 209 mapped on the D genom e based on the Chinese Spring reference assem bly IWGSC Ref Seq v1.0 w ere em ployed to calculate LD values.A scatter plot of LD values w as plotted against the physical distances to estimate the genome-w ide LD(Fig.1).The fitted m odel show ed that LD decayed to 0.1 at 23 Mb.

The population structure analysis divided the association panel into seven subgroups(S1-S7)w ith 151,343,137,253,56,95,and 257 accessions and fixation indexes(Fst)of 0.4912,0.3843,0.3809,0.6694,0.8510,0.5346,and 0.5206 from S1 to S7,respectively(Fig.2).About 64.2%accessions in the S1 w ere Iranian landraces(Iranian landrace I subgroup),and the rest of 35.8%accessions in S1 m ainly cam e from Turkey,Afghanistan,Azerbaijan,Georgia and Arm enia.In S2,273 accessions w ere European cultivars,breeding lines,and landraces(European subgroup)and the others include 50 accessions from the USA,14 from Asia,three from Southern Am erica,and one each from Africa,Australia,and the Caribbean regions.Accessions in S3 m ainly originated from southeastern Europe,including Albania,Bulgaria,Romania,Serbia,Croatia,Greece,Macedonia,Bosnia and Herzegovina,Montenegro,Turkey,and Italy(southeastern European subgroup)w ith som e accessions collected from Spain,Czech Republic,USA,and Germany.Iranian landraces w ere dom inant(85%)in S4(Iranian landrace II subgroup),w ith the other landrace accessions m ainly from Afghanistan,Pakistan,and Uzbekistan.Iran is am ong the centers of w heat origin and diversity,and Iranian landraces are valuable resources for w heat im provement.The 325 Iranian landraces in this association panel w ere m ainly included in S1(29.8%)and S4(66.5%).Accessions in S5 w ere m ainly cultivars or breeding lines released in Russia,Rom ania,and Bulgaria,and landraces collected in Georgia(Russian and East European subgroup).The m ajority of accessions(87.4%)in S6 came from East and southern Asia,including China,Japan,South Korea,India,Bhutan,Nepal,and Pakistan(Asian subgroup).Subgroup S7 consisted predom inantly of cultivars or breeding lines released from the USA and w estern Europe,including Germ an,France,Belgium,Sw eden,Sw itzerland,Netherland,and Italy(USand Western European subgroup).

Fig.2-Seven w inter/facultative w heat pop ulations inferred from structure analysis w ith a m ix ture m od el.Subgroup s S1 to S7 rep resent Iranian landrace I,Europ ean accessions,southeast Europ ean accessions,Iranian landrace II,Russian and East European accessions,East Asian accessions,and US and Western Europ ean accessions.Vertical lines show m em bership coefficients of accessions corresp onding to the seven subgroups.

3.2.Pow dery mildew resistance in the association panel

The association panel w as evaluated for reactions to Bgt15.The IT distribution of 1292 accessions w as skew ed toward susceptibility(Fig.3).Am ong them,4.0%,4.5%,11.2%,and 80.3%accessions w ere highly resistant(HR)w ith ITs of 0 and 1,m oderately resistant(MR)w ith a IT of 2,m oderately susceptible(MS)w ith a IT of 3,and highly susceptible(HS)w ith a IT of 4,respectively.

OKS(14)-B-3-1 w as a less virulent isolate than Bgt15,evidenced by the fact that m ost differential lines carrying known powdery mildew resistance gene(s)had the same or a m ore resistant reaction to OKS(14)-B-3-1 than to Bgt15,except for a line carrying Pm2(Table 1).About 4.0%,2.7%,7.8%,and 85.5%of accessions in the association panel w ere HR,MR,MS,and HSto OKS(14)-B-3-1,respectively(Fig.3).

Com pared w ith Bgt15 and OKS(14)-B-3-1,OKS(14)-C-2-1 w as virulent to m ore know n pow dery m ildew resistance genes.Lines w ith Pm1c,Pm2,Pm3a,Pm3b,Pm3c,Pm3e,Pm3f,Pm3g,Pm4a,Pm4b,Pm5a,Pm5b,Pm6,Pm7,Pm8,Pm9,Pm25,Pm34,and Pm35 show ed MS or HS reaction to OKS(14)-C-2-1,but Pm1a,Pm3d,Pm5d,Pm13,Pm16,Pm17,Pm20,Pm21,and Pm37 conferred high or m oderate resistance to OKS(14)-C-2-1(Table 2).Am ong the nine resistance genes,only Pm1a,Pm3d,and Pm5d derived from com m on w heat.Pm3a,the m ajor resistance gene located on the short arm of chrom osom e 1A and w idely deployed in US w inter w heat[10],w as overcom e by OKS(14)-C-2-1,suggesting urgent need of novel resistance sources in w inter w heat breeding program s.In agreem ent w ith the responses of differential lines to OKS(14)-C-2-1,few er HR or MR accessions w ere found in the association panel.About 1.3%,1.5%,7.2%,and 90.0%of accessions w ere rated HR(IT=0,1),MR(IT=2),MS(IT=3),and HS(IT=4)(Fig.3),respectively.

A total of 12 accessions,including tw o breeding lines,seven cultivars,and three landraces,exhibited high resistance to all three inoculants(Table S1).They are valuable pow dery m ildew resistance sources for w heat breeding.

3.3.Pow dery mildew resistance loci revealed by GWAS

Seven SNPs,IWA4096,IWA1707,IWA5077,IWA4517,IWA5794,IWA3840,and IWA5850,on chromosome 2Bw ere significantly associated w ith resistance to OKS(14)-B-3-1(P=5.4E-25 to 2.64E-5)at FDR0.01in the MLM analysis(Figs.4,5).Am ong them,IWA5850 w as also associated w ith resistance to Bgt15 w ith a P-value of 6.8E-9(Table 3),but no SNPs was significant for resistance to OKS(14)-C-2-1.

SNPs IWA5077,IWA4517,IWA5794,and IWA3840 w ere m apped to an interval betw een 457.68 and 461.74 Mb on the long arm of chromosom e 2B using Chinese Spring Ref Seq 1.0(Table 3).The LD values(r2)am ong these SNPs ranged from 0.993 to 0.997,confirm ing that they w ere tightly linked on the consensus map[31],w hile LD values betw een these four SNPs and the other three SNPs ranges from 0.023 to 0.048.Thus,they represent an independent locus designated QPm.stars-2BL1.

Fig.3-Distribution of m ean p ow d ery m ildew infection types of 1292 accessions challenged w ith inoculants OKS(14)-B-3-1,OKS(14)-C-2-1,and Bgt15.

Table 2 - Reactions of 28 known powdery mildew resistance genes to Bgt isolates OKS(14)-C-2-1 and OKS(14)-B-3-1.

IWA1707,IWA4096,and IWA5850 w ere located at 690.0,696.6,and 715.9 MB on the Chinese Spring reference sequence Ref Seq 1.0,respectively.LD analysis indicated that IWA4096 and IWA5850 are linked w ith an LDvalue larger than 0.1(r2=0.153),w hile IWA1707 had low LD values(＜0.1)w ith both IWA4096(r2=0.046)and IWA5850(r2=0.093).These results suggested that IWA1707 likely represented the second locus designated QPm.stars-2BL2,w hile IWA4096 and IWA5850 w ere associated w ith another locus d esignated QPm.stars-2BL3.

QPm.stars-2B3 conferred resistance to OKS(14)-B-3-1 and Bgt15,and the representative SNP,IWA5850,explained 8.64%and 2.64%of the total phenotypic variance for pow dery m ildew resistance in the tw o experim ents,respectively.Its allele substitution effect w as 1.76 for IT in the OKS(14)-B-3-1 experim ent(Table 3).Most of the accessions(69.3%)that carry the m arker allele(G)associated w ith resistance w ere either highly(46.2%)or moderately(23.1%)resistant to OKS(14)-B-3-1.The allele substitution effect of QPm.stars-2B3 w as 0.96 for IT in the Bgt15 experim ent(Table 3),and 42.5%of w heat accessions that carry the resistance m arker allele w ere highly(30%)and m oderately resistant(12.5%),respectively.Eleven accessions w ere identified to carry the resistance m arker allele at this locus and also exhibit high resistance to both OKS(14)-B3-1 and Bgt15(Table S2).These accessions,including cultivars or breeding lines from the United States,United Kingdom,Crotia,Kazakhstan,and Bulgaria,as w ell as landraces from Iran,m ay carry QPm.stars-2B3.

QPm.stars-2BL1 and QPm.stars-2BL2 w ere significantly associated w ith resistance to OKS(14)-B-3-1 only,and each locus explained 1.55%-1.85%of the total phenotypic variance w ith allele substitution effects ranging from 0.35 to 0.60 for IT.Of the 205 accessions that carry the resistance m arker alleles linked to QPm.stars-2BL1,15.1%and 7.8%exhibited high and m oderate resistance to OKS(14)-B-3-1,respectively.In contrast,only 2.7%and 4.5%of w heat accessions that carry the susceptible m arker alleles at this locus show ed high and moderate resistance to OKS(14)-B-3-1,respectively.Similar trends w ere observed for QPm.stars-2BL2.Table S2 listed all w heat accessions that carry resistance allele(s)at QPm.stars-2BL1 or/and QPm.stars-2BL2 and exhibit high resistance to OKS(14)-B-3-1,which are mainly cultivars and breeding lines from Europe and landraces from Iran.

3.4.Validation of QPm.stars-2BL2 using linkage mapping

To validate the efficacy of GWASin gene discovery,w e further used linkage m apping to m ap the pow dery m ildew resistance gene in PI 420646,intending to confirm QPm.stars-2BL2.PI 420646 is a Kazakhstan cultivar w ith im m une reaction(IT=0)to Bgt isolate OKS(14)-B-3-1 and HRreaction(IT=1)to Bgt2015[10],and PI 420646 carries the resistant SNP alleles at the IWA1707,IWA4096,and IWA5850 loci.

Fig.4-Quantile-Quantile plot of P-values com p aring the uniform d istribution of the expected-lg(P)to the observed-lg(P)for pow d ery m ildew resistance to Bgt isolate OKS(14)-B3-1(a)and Bgt2015(b)in MLM analyses.

A set of 177 F2plants derived from PI 420646×OK1059060-126135-3 w ere evaluated for responses to Bgt isolate OKS(14)-B-3-1,and 131 and 46 plants w ere resistant and susceptible,respectively,indicating that a dom inant gene conferred pow dery m ildew resistance in PI 420646(χ23:1=0.092279,df=1,P=0.76).BSA result indicated that SSR m arkers WMC332 and WMC441 w ere polym orphic betw een the resistant and the susceptible bulks,therefore,they w ere used to genotype the F2population.Linkage analysis show ed that the pow dery mildew resistance gene in PI 420646 was 21.4 c M distal to WMC441,and 6.4 c M proxim al to WMC332(Fig.6).Given that WMC441 and WMC332 w ere previously m apped to chrom osom e 2BL,a set of 48 SSRm arkers located in the target region w ere developed using the Chinese Spring reference sequence(Table 4).These new ly developed m arkers,as w ell as other m arkers previously m apped to the target region,w ere screened for polymorphism between the tw o parents,leading to the assignm ent of the resistance gene to an interval betw een 693.1 and 723.4 Mb in the Chinese Spring reference assem bly,flanked by SSR m arker STARS412 and STS m arker BCD135-2.The resistance gene was 4.5 c M distal to STARS412,and 2.9 c M proxim al to BCD135-2.Given that IWA1707,IWA4096,and IWA5850 w ere m apped at 690.0,696.6,and 715.9 Mb of the Chinese Spring reference sequence,respectively(Table 3),the resistance gene in PI 420646 may be either QPm.stars-2BL2 or QPm.stars-2BL3.

Of the know n genes on the long arm of chrom osom e 2B,Pm6,introgressed from Triticum timopheevii into bread w heat in 1960s[44,45]and has been w idely used in w heat breeding,w as 1.6 c M distal to STS m arker BCD135-2[46].We used an STS m arker developed to tag the T.timopheevii-derived Pm6[47]to screen w heat accessions w ith high resistance to OKS(14)-B3-1,and the target band w as am plified from 13 accessions.Of these,10 accessions,including PI 420646,show ed the sam e band pattern as CItr 15888(Michigan Amber/8*Chancellor),a carrier of Pm6.An additional band w as am plified from another three accessions.Thus,w e excluded these 13 accessions from the dataset,and reanalyzed the data.We found that IWA4096(P=7.71E-6)and IWA5850(P=1.07E-17)w ere still significantly associated w ith resistance to OKS(14)-B3-1 at FDR0.05and FDR0.01,respectively,w hile IWA1707(P=2.59E-4)w as not significant at all.These results indicated that QPm.stars-2BL2,as w ell as the gene in PI 420646,is likely Pm6,and further confirm ed that QPm.stars-2BL3 is independent of QPm.stars-2BL2.

Fig.5-Manhattan plot of-lg(P)by chrom osom es for p ow dery m ild ew resistance to Bgt isolate OKS(14)-B3-1(a)and Bgt2015(b)in MLM analyses.

Table 3-Designated nam e,related SNP(s),p hysical location,P-value,R2-value,SNP alleles,and allelic effect at three resistance loci id entified at a false d iscovery rate of 0.01 w hen the association p anel w as inoculated w ith Bgt isolates OKS(14)-B3-1 and Bgt15.

Fig.6-Linkage(left)and physical bin(right)m aps of QPm.stars-2BL2,w hich w as confirm ed in F2 and F2:3 p opulations d erived from PI 420646×OK1059060-126135-3.Marker nam es are show n at the right of the linkage m ap,and genetic d istances in c M on the left.The p hysical positions of som e markers on the Chinese Spring reference assem bly are given in the follow ing p arentheses.Molecular m arkers flanking QPm.stars-2BL2 are connected to their approp riate p hysical bins.The breakpoint of each Chinese Spring d eletion line is show n w ith an arrow,and the corresp onding fraction length(FL)value is given in the follow ing p arentheses.

Four m arkers flanking QPm.stars-2BL2(Pm6),STARS411,STARS412,BCD135-2,and WMC332,w ere used to genotype Chinese Spring and its five deletion lines,including 2BL-1,2BL-3,2BL-4,2BL-5,and 2BL-6.All but STARS411 failed to am plify the target bands from any of these deletion lines,indicating that the resistance gene resides in the term inal bin 2BL-0.89-1.0(Fig.6).STARS411 am plified the target band from 2BL-6,but not from the other four lines.

4.Discussion

4.1.The relationship of pow dery mildew resistance loci identified in this study with known powdery mildew resistance genes

Chrom osom e 2B harbors the m ost know n genes or QTL for pow dery m ildew resistance,including PmL962[48],MIW170[49],Pm49[50],Pm26[51],pm42[52],and QPm.ipk-2B[53]on the short arm,and Pm6[46,47],Pm33[54],Pm51[55],Pm52[56],Pm62[57],PmJM22[58],MlAB10[59],and Mlzec1[60]on the long arm.Coincidently,all three pow dery m ildew resistance loci identified in this study reside on chromosome 2B,reflecting the abundance of pow dery m ildew resistance genes on this chrom osom e.

The genom ic locations of SNPs underlying QPm.stars-2BL1 range from 457.7 to 461.7 Mb on the Chinese Spring reference assembly(Table 3).Alm ost all genes on chrom osome 2BLwere m apped to the term inal bin 2BL6-0.89-1.0[56],w hich are far from QPm.stars-2BL1.Only Pm52 w as flanked by SSR m arker WMC441 and STS m arker BF292219 in bin 2BL2-0.36-0.50[56].Sequence alignm ent located WMC441 and BF292219 to approxim ately 598.1 and 594.8 Mb on the Chinese Spring reference assem bly,respectively.Thus,Pm52 is at least 135 Mb distal to QPm.stars-2BL1.Given that LD decayed to 0.1 at 23 Mb,and all other pow dery m ildew resistance genes on chromosome 2BL were distal to Pm52,QPm.stars-2BL1 is likely a novel locus for pow dery m ildew resistance.

We dem onstrated that QPm.stars-2BL2 is likely Pm6,and QPm.stars-2BL3 is distal to Pm6 in the term inal bin 2BL-0.89-1.0,w here Pm33,Pm51,Pm62,PmJM22,MlZec1,and MlAB10 reside[54-60].PmJM22,identified in w heat cultivar Jim ai 22,w as m apped 7.7 c M distal to Xw mc149[58],positioned at 779.1 Mb in the Chinese Spring reference sequence IWGSC Ref Seq v1.0.Given that the tw o SNPs underlying QPm.stars-2BL3,IWA4096 and IWA5850,w ere m apped at 696.6 and 715.9 Mb,respectively,QPm.stars-2BL3 is at least 63.2 Mb proxim al to PmJM22.The other genes in this region,including Pm33 from T.carthlicum[54],Pm51 from Thinopyrum ponticum[55],Pm62 from D.villosum[57],and MlZec1 and MlAB10 from T.dicoccoides[59,60],w ere recently introgressed into w heat from w ild species,and w heat germplasm carrying these genes were not included in the association panel.Therefore,QPm.stars-2BL3 is a new locus for pow dery m ildew resistance gene.A couple of projects aim ed at validating QPm.stars-2BL1 and QPm.stars-2BL3 are currently underw ay at the USDA-ARS w heat,peanut,and other field crop research unit.

In this study,w e found only seven SNPs to be significantly associated w ith pow dery m ildew resistance at three loci.The high virulence of Bgt pathotypes used in this study is likely the m ajor reason.As discussed above and reported previously[10],it appears that the current Bgt population in Oklahom a has evolved to be virulent against many resistance genes,evidenced by the fact that m ost know n pow dery m ildew resistance genes derived from bread w heat have lost their effectiveness to Bgt15 and OKS(14)-C-2-1.The high virulence of Bgt15 and OKS(14)-C-2-1 m ay prevent the expression of som e pow dery m ildew resistance genes present in the association panel in the experim ents.In addition,the estim ated genom ewide LD w as about 23 Mb,and SNPm arkers used in this study w ere several tim es m ore than the m inim um m arkers needed for a GWAS study.How ever,the LD values m ay be dram atically different in different genom ic regions,and m ore SNPs may be needed in the genom ic regions w here recombination occurs m ore frequently and LD declines m ore quickly.Given that SNPs are unevenly distributed in the w heat genom e and m uch less SNPs have been discovered in the D genom e[20],genotyping the association panel w ith additional SNPs m ay help to identify pow dery m ildew resistance loci present in the panel.

Table 4-Prim er sequences,genom ic locations,PCRprod uct sizes,and SSRm otifs of 48 new ly d eveloped SSRm arkers.The m elting tem p eratures(T m)of all SSRm arkers are 60°C.

4.2.Mining w heat landraces and historical cultivars for pow dery mildew resistance genes

Genetic erosion caused by early dom estication and m odern breeding has seriously narrow ed the genetic base of w heat.Wheat w ild relatives,landraces,and historical cultivars are essential resources that conserve genetic diversity and provide the necessary foundation for introgression of valuable genes into elite cultivars.Most catalogued pow dery mildew resistance genes originated from w heat relatives,including Pm1b,Pm1d,Pm2,Pm3b,Pm4a,Pm4b,Pm4d,Pm5a,Pm5c,Pm6,Pm7,Pm8,Pm12,Pm13,Pm16,Pm17,Pm19,Pm20,Pm21,Pm25,Pm26,Pm27,Pm29,Pm32,Pm33,Pm34,Pm35,Pm36,Pm37,Pm40,Pm41,Pm42,Pm43,Pm49,Pm50,Pm51,Pm53,and Pm55[3,4].The m ajority of these genes cannot be directly used in w heat im provement because they often carry alien chrom osom al fragments w ith many undesirable genes or do not adequately com pensate for the w heat genes they replace,resulting in linkage drag that m ay require m any years of breeding to elim inate.For exam ple,Pm21,transferred from Haynaldia villosa into w heat in early 1990s[61],show s broad-spectrum resistance to Bgt isolates in China.How ever,w heat cultivars w ith Pm21 w ere not developed for the Huang-Huai River Valley,the m ajor w heat production region of China,until recently[9].The difficulty in using these alien resistance genes also contributes to the narrow genetic base of pow dery m ildew resistance in the U.S.w heat breeding program s[10].

Wheat landraces represent another im portant source for pow dery m ildew resistance.Pow dery m ildew resistance genes derived from w heat landraces,such as Pm5e,Pm47,MlHLT,MlXBD,and Pm61 have been characterized in recent years[61-65].These landrace-derived genes are much easier to use in w heat breeding than those identified from w heat relatives.In addition,historical cultivars that w ere replaced by m odern sem i-dw arf varieties have conserved pow dery m ildew resistance genes[10].These early cultivars are rarely used in m odern w heat breeding because of inferior agronom ic traits,and pow dery m ildew resistance genes they carry are therefore w arehoused in seed rep ositories.Although landraces and historical cultivars are perceived to be inferior for the im provem ent of com plex traits such as yield,the pow dery m ildew resistance genes present am ong thousands of accessions at the USDA-ARSNSGC are valuable for w heat im provem ent.In the present study,w e identified three loci for pow dery m ildew resistance,and one of them,QPm.stars-2BL2,w as confirm ed by linkage analysis,suggesting that GWAS is a reliable approach for gene discovery.Other tw o loci,QPm.stars-2BL1 and QPm.stars-2BL3,are clearly novel because of their unique chrom osom e locations.We have selected a set of accessions that m ay carry these novel genes for further characterization,intending to precisely locate these genes and introgress them into adapted germ p lasm.

5.Conclusions

Pow dery m ildew is a disease of increasing im portance globally,and GWAS is a reliable approach to quickly identify pow dery m ildew resistance genes.Linkage disequilibrium(LD)analysis indicated that genom e-w ide LD decayed to 0.1 at 23 Mb in w heat,and GWAS analysis revealed three pow dery m ildew resistance loci on chrom osom e 2B,including QPm.stars-2BL1,QPm.stars-2BL2,and QPm.stars-2BL3.QPm.stars-2BL2 coincided w ith a pow dery m ildew resistance gene in PI 420646,lending credence to the GWAS results.QPm.stars-2BL1 and QPm.stars-2BL3 are likely novel loci for pow dery m ildew resistance and can be used in w heat breeding.

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2019.01.005.

Acknow ledgm ents

We thank Dr.Harold Bockelm an of the USDA-ARS National Sm all Grain Collection for providing som e of the germ plasm used in this study.Mention of trade nam es or com m ercial products in this publication is solely for the purpose of providing specific information and does not im ply recomm endation or endorsem ent by the USDA.The USDA is an equal opportunity provider and em ployer.