Unlocking wheat genetic resources for the molecularidentification of previously undescribed functionalalleles at the Pm3 resistance locusNavreet K. Bhullara, Kenneth Streetb, Michael Mackayc, Nabila Yahiaouia,1, and Beat Kellera,2

aInstitute of Plant Biology, University of Zurich, 8008 Zurich, Switzerland; bInternational Center for Agricultural Research in the Dry Areas, Aleppo, Syria;and cBioversity International, 00057 Maccarese, Rome, Italy

Communicated by Jeffrey L. Bennetzen, University of Georgia, Athens, Georgia, April 16, 2009 (received for review December 27, 2008)

The continuous improvement of crop plants is essential for agriculturein the coming decades and relies on the use of genetic variabilitythrough breeding. However, domestication and modern breedinghave reduced diversity in the crop germplasm. Global gene banksconserve diversity, but these resources remain underexplored owingto a lack of efficient strategies to isolate important alleles. Here wedescribe a large-scale allele-mining project at the molecular level. Wefirst selected a set of 1,320 bread wheat landraces from a database of16,089 accessions, using the focused identification of germplasmstrategy. On the basis of a hierarchical selection procedure on this set,we then isolated 7 resistance alleles of the powdery mildew resis-tance gene Pm3, doubling the known functional allelic diversity at thislocus. This targeted approach for molecular utilization of gene bankaccessions reveals landraces as a rich resource of new functionalalleles. This strategy can be implemented for other studies on themolecular diversity of agriculturally important genes, as well as formolecular breeding.

allele mining � powdery mildew � gene banks � wheat landraces

Wheat is one of the most important human food crops, andproduction has to be increased significantly in the next

decades (1). This has to be done in a sustainable way, with lessagricultural input but increased yield. Natural biodiversity has beenused for enriching diversity of cultivated plants with novel alleles toimprove productivity by breeding (2). For example, tomato lineswith introgression of wild alleles from Lycopersicon hirsutum out-performed the original variety by 48% and 33% for yield and fruitcolor, respectively (3). In barley, the mlo-11 gene providing resis-tance against barley powdery mildew originated from Ethiopianlandraces (4). During agricultural development, early domesticateswere gradually replaced first by landraces and traditional varieties,and later by genetically less-diverse modern cultivars. This hasresulted in genetic bottlenecks and loss of diversity in the breedinggermplasm (5). Therefore, gene bank collections are essential toconserve biodiversity and thus pay big dividends to agriculturewhen used efficiently (6).

Despite many studies illustrating the utilization of genetic re-sources in plant breeding (7, 8), the global germplasm collectionsare underutilized for many crop species. One important reason isthe sheer number of accessions stored, and these large collectionscannot be phenotyped or genotyped by an average laboratory orbreeding program. Thus, the major challenge to identify rare allelesfrom large collections is to identify a subset of accessions that iseconomically feasible to screen while maximizing the probability offinding the desired trait. Core collections have been widely pro-moted as a means of approaching large collections by definingsmaller subsets that represent maximum diversity. As an alternativeapproach, the focused identification of germplasm strategy (FIGS)was recently suggested as a rational method that uses informationabout the environment from which accessions with specific traitshave been collected to predict where selection pressures for adap-tive traits may occur. On the basis of this information, trait-specificsets can then be assembled from large collections (9).

Genetic variation is caused by allelic diversity at the genetic locicontributing to a particular trait. Allele mining is a relativelyunderexplored method to identify new alleles at a known locus.However, it is being used in important plant species, such as maizeand barley (ref. 10; N. Stein, et al., personal communication).Because the first wheat disease-resistance genes have been cloned(11–16), the sequence information of these genes should allow theanalysis of genetic diversity at these loci and the identification ofnew alleles through allele mining. Pm3, existing in 7 functionallydistinct alleles (Pm3a to Pm3g), is the only wheat powdery mildewresistance gene cloned to date (13, 14, 16). In addition to the allelesfrom the bread wheat gene pool, a new functional allele recently hasbeen described in a wild tetraploid wheat accession (17). All Pm3alleles encode coiled-coil (CC), nucleotide binding site (NBS), andleucine-rich repeat (LRR) proteins. The high sequence conserva-tion of the Pm3 alleles suggested their recent evolution from theancestral sequence Pm3CS that is a susceptible Pm3 allele (16).

Here we describe the successful and efficient screening of genebank accessions for the molecular identification of allelic variants atthe Pm3 locus. We report the cloning of 7 previously undescribedfunctional Pm3 alleles from a targeted subset of wheat landracesthat was established by FIGS, demonstrating its successful use incombination with allele mining. We also found that at least 2 ofthese Pm3 alleles confer slow-acting resistance. The strategy de-scribed here can be implemented for other diversity and molecularbreeding studies involving agriculturally important traits.

ResultsEstablishment and Screening of a Focused Set of Wheat Landraces forPm3-Based Powdery Mildew Resistance. To maximize the chances offinding functional diversity of powdery mildew resistance whilelimiting the number of accessions to a workable size, we used FIGSto define a subset of accessions (FIGS powdery mildew set). Froma virtual collection of 16,089 accessions, we identified 1,320 acces-sions drawn from 323 geographic sites with potentially high selec-tion pressure for powdery mildew resistance. To select the resistant

Author contributions: N.K.B., K.S., M.M., N.Y., and B.K. designed research; N.K.B., K.S., andM.M. performed research; N.K.B., N.Y., and B.K. analyzed data; and N.K.B. and B.K. wrote thepaper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

Data deposition: The previously undescribed Pm3 sequences have been deposited in theGenbank database (accession nos. Pm3�9939 � FJ212300, Pm3�42255 � FJ212301,Pm3�42868 � FJ212302, Pm3�31594 � FJ212303, Pm3�13636 � FJ212304, Pm3�41606 �

FJ212305, Pm3�42281 � FJ212306, Pm3�14475 � FJ212307, Pm3�10963 � FJ212308,Pm3�42469 � FJ212309, Pm3�23728 � FJ212310, Pm3�42277 � FJ212311, Pm3�42416 �

FJ212312, Pm3�42525 � FJ212313, Pm3�42920 � FJ212314, and Pm3�14442 � FJ212315).

1Present address: Centre de Cooperation Internationale en Recherche Agronomique pour leDeveloppement, Unite Mixte de Recherche Developpement et Amelioration des Plantes,F-34398 Montpellier, France.

2To whom correspondence should be addressed at: Institute of Plant Biology, University ofZurich, Zollikerstrasse 107, 8008 Zurich, Switzerland. E-mail: bkeller@botinst.uzh.ch.

This article contains supporting information online at www.pnas.org/cgi/content/full/0904152106/DCSupplemental.

www.pnas.org�cgi�doi�10.1073�pnas.0904152106 PNAS � June 9, 2009 � vol. 106 � no. 23 � 9519–9524

PLA

NT

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

July

13,

202

0

accessions, the complete set of 1,320 landraces was screened withdifferent powdery mildew isolates (18). A total of 211 accessionsthat showed complete or intermediate resistance against at leastone powdery mildew race were further analyzed at the molecularlevel. They were first screened for the presence of a Pm3-like genewith a diagnostic sequence-tagged site marker and second for thepresence of the already-known Pm3 alleles (18, 19). This led to theidentification of 111 landraces as candidates for the isolation of Pm3alleles that were positive for the Pm3 diagnostic fragment but lackedany of the known Pm3 alleles.

Cloning of Pm3 Alleles from Selected Wheat Landraces. The isolationof Pm3 alleles was performed on 56 landraces completely resistantto at least one powdery mildew isolate, whereas lines with inter-mediate resistance were not considered further. The Pm3 codingsequences were successfully amplified from 45 landraces, cloned,and sequenced. In the remaining landraces, amplification of a Pm3sequence was not possible, perhaps due to the absence of a codinggene or low sequence homology at the primer binding sites. Theanalysis of sequence diversity led to the identification of 16 previ-ously unknown Pm3 allelic sequences, because several landracespossessed identical alleles (Fig. 1 and supporting information TableS1). Among the 45 sequences, 9 were identical to the susceptiblePm3CS (16), suggesting that the observed resistance is not due toa Pm3 type of gene but is caused by other known or still unchar-acterized Pm genes. Among the remaining 36 landraces from whichthe previously undescribed Pm3 alleles were isolated, 24 accessionsoriginated from Turkey, with the majority from Eastern parts of thecountry; the remainder originated from Afghanistan (mostly local-ized in the Bamian Province), Pakistan, Azerbaijan, and Turkmeni-stan (9, 1, 1, and 1, respectively; Table S1). In good correspondencewith this distribution, 8 of the 16 previously undescribed allelesoriginated from landraces collected in Turkey and 4 alleles fromAfghanistan (Table S1). Different landraces with identical Pm3alleles always originated from the same country, except Pm3�13636,which was found in 2 landraces originating from Afghanistan andTurkey. However, for alleles found in multiple accessions, the

collection sites were not necessarily geographically close within thecountry (e.g., the accessions from Turkey; Table S1).

Sequence Diversity of the Previously Unknown Pm3 Alleles. The DNAsequence comparison of the 16 previously undescribed Pm3 se-quences with the known alleles Pm3a to Pm3g and Pm3CS showedan overall high similarity (Fig. 1). The previously undescribed allelesalso consist of 2 exons separated by an intron of 200 bp and encoderesistance proteins with an NBS and an LRR domain associated atthe N-terminus with a CC domain. Nine of the previously unde-scribed alleles are 4442 bp long, corresponding to the size of theancestral reference sequence Pm3CS, whereas 7 alleles bear inser-tions and deletions (InDels) (Pm3�42868, Pm3�42255, Pm3�42416,Pm3�42920, Pm3�31594, Pm3�41606, and Pm3�14442) making themvariable in size. None of these InDels altered the ORFs exceptfor Pm3�14442, which is a pseudogene missing 294 bp, resultingin a frame shift. Illegitimate recombination seems to be the causeof deletions: we have identified 4-bp imperfect repeat motifs atthe breakpoint of the deletions in Pm3�42416 and Pm3�42920that could have served as template sequence for illegitimaterecombination (20). In Pm3�31594, a perfect 3-bp repeat wasfound at the breakpoint of the deletion, also suggesting illegit-imate recombination.

The CC-NBS encoding region of the new alleles is highlyconserved, with the exception of a 3-bp deletion and a single basechange in the NBS regions of Pm3�41606 and Pm3�42281, respec-tively. The intron sequence was identical among the previouslyundescribed alleles and the known Pm3 alleles, except for 5previously uncharacterized alleles (Pm3�42416, Pm3�42277,Pm3�42525, Pm3�42920, and Pm3�23728) that differed by maxi-mally 6 bp. The comparison of allelic sequences showed 2 groupsof sequences. The first group of 4 sequences (Pm3�42469,Pm3�41606, Pm3�42281, and Pm3�13636) has very few or only asingle polymorphic residue(s) compared with the reference se-quence Pm3CS. In the second group, polymorphic residues arepresent as sequence blocks (blocks A–F, Fig. 1) completely orpartially shared between alleles. This existence of defined sequenceblocks indicates the frequent sequence exchange between alleles,

Fig. 1. Schematic representation of sequence alignment (exons) of previously undescribed Pm3 alleles with the known alleles. This figure presents the Pm3 genestructure and the alignment of previously undescribed Pm3 alleles with the known alleles Pm3CS and Pm3a. The domains encoded by the Pm3 alleles are depicted atthe top [CC, NBS, Interspacer (IS), and LRR]. Red bars in the Pm3 alleles and numbers in red indicate the polymorphic nucleotides as compared with Pm3CS, leading tonon-synonymous changes in the protein. Black bars and numbers in black represent the polymorphic nucleotides leading to synonymous mutations. Boxes A, B, C, D,E, and F indicate the putative gene conversion tracts among these alleles. Numbers in parentheses correspond to the number of landraces possessing that particularPm3 allele. Golden bars indicate the regions used for VIGS constructs. The functional alleles Pm3�42416, Pm3�42525, Pm3�23728, Pm3�42255, Pm3�10963, Pm3�42277,and Pm3�42868 are labeled in blue.

9520 � www.pnas.org�cgi�doi�10.1073�pnas.0904152106 Bhullar et al.

Dow

nloa

ded

by g

uest

on

July

13,

202

0

possibly by gene conversion. The differences in proteins encoded bythe previously undescribed Pm3 alleles lead to a total of 87 aachanges compared with PM3CS, in addition to the InDels. At amajority of sites, different alleles have a single residue change (72of 87 positions), with 13 sites having 2 alternative residues and 1 sitehaving 3 alternative residues (R/D/Y at position 1332 instead of Win PM3CS).

Functional Analysis of Pm3 Candidate Resistance Genes by Virus-Induced Gene Silencing. To determine whether the resistance ob-served in landraces with previously undescribed alleles is basedspecifically and solely on the Pm3 alleles, we used virus-inducedgene silencing (VIGS) to suppress the Pm3 gene expression. Weused 2 silencing constructs carrying fragments of the CC(BSMV.Pm3�CC) and the LRR domain (BSMV.Pm3�LRR), re-spectively (golden bars in Fig. 1), and a control construct(BSMV.Lr10�CC) silencing the Lr10 leaf rust resistance gene by afragment of its CC region. The landraces carrying previouslyundescribed Pm3 allelic sequences (Pm3�42416, Pm3�10963,Pm3�42525, Pm3�23728, Pm3�42469, and Pm3�42255) were infectedwith these constructs and further challenged with avirulent Bgtisolate 98275. The leaves of landraces carrying the allelesPm3�42416 and Pm3�42255 (IG42416 and IG42255) lost resistanceto Bgt isolate 98275 after infection with BSMV.Pm3�CC andBSMV.Pm3�LRR (Fig. 2) but remained resistant when inoculatedwith the control BSMV.Lr10�CC. This demonstrates thatPm3�42416 and Pm3�42255 confer the observed powdery mildewresistance in these lines. In contrast, resistance was not altered in thelandraces with the alleles Pm3�42469, Pm3�42525, and Pm3�10963,indicating that here the resistance might result either from a genedifferent from Pm3 or from a combination of Pm3 and additionalgenes. In the case of landrace IG23728 (Pm3�23728), results werenot conclusive, possibly owing to a heterogenous seed mixture forthis accession. Because of this relatively high number of inconclu-sive results (4 of 6 genes), we used transient transformation forfurther functional studies on the candidate alleles.

Identification of Functional Pm3 Alleles by Transient Transformation.In total, 13 previously undescribed alleles were tested by transienttransformation: 8 alleles against isolate Bgt 98275, 4 with Bgt 97011,and 1 with Bgt 96224. It was not possible to identify an appropriateisolate for Pm3�14475, Pm3�13636, and Pm3�42281 because severalindependent landraces with these alleles behaved differently for thetested isolates, suggesting that the observed resistance was not dueto the Pm3 allele. We used the nonfunctional Pm3CS allele (16) asa control. Seven alleles (Pm3�42416, Pm3�42525, Pm3�23728,Pm3�42255, Pm3�10963, Pm3�42277, and Pm3�42868) showed asignificant reduction in the haustorium index in comparison withPm3CS (Fig. 3). Transformation with the remaining 6 alleles

A

B

C

D

Fig. 2. BSMV-mediated virus induced silencing ofPm3�42416 and Pm3�42255. (A) The landrace containingPm3�42416 infected with powdery mildew but not withthe virus served as resistant control. (B) Infection withBSMV.Lr10�CC (control viral construct) did not alter theobserved resistance against mildew. (C and D) The resis-tant landrace turned susceptible to powdery mildewwhen infected with BSMV.Pm3�CC and BSMV.Pm3�LRR,respectively. Similar observations were made forPm3�42255. (Scale bars, 0.5 cm.)

Fig. 3. Functional analysis of the previously undescribed Pm3 alleles in thetransient transformation assay. The graph shows the transient assay results forthe 7 functional Pm3 alleles tested with corresponding avirulent Bgt isolates, incomparison with the susceptible control Pm3CS. The haustorium index (percent-age of cells with haustoria) is indicated by the mean � SD of 3 independentexperiments, each contributing at least 50 interactions. *Significant differencesat P � 0.05; **significant differences at P � 0.001. Transient assay results forPm3�42416 and Pm3�42255, upon infection with the virulent isolate, are repre-sented by dark gray bars.

Bhullar et al. PNAS � June 9, 2009 � vol. 106 � no. 23 � 9521

PLA

NT

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

July

13,

202

0

(Pm3�42920, Pm3�9939, Pm3�42469, Pm3�31594, Pm3�14442, andPm3�41606) did not result in a reduction of the haustorium index(Fig. S1). To check for race specificity of the resistance conferredby the previously undescribed alleles, the 2 alleles showing the mostsignificant reduction in haustorium index (Pm3�42416 andPm3�42255) were also tested against the virulent isolate 97019. Noreduction of haustorium indices was observed compared withPm3CS (Fig. 3), demonstrating that the observed activity was notdue to overexpression but to race specificity of gene action.

In conclusion, the alleles Pm3�42416, Pm3�42255, Pm3�23728,Pm3�10963, Pm3�42525, Pm3�42277, and Pm3�42868 are previouslyuncharacterized, functionally active forms of Pm3 that are calledPm3l, Pm3m, Pm3n, Pm3o, Pm3p, Pm3q, and Pm3r, respectively.Five of these 7 alleles were isolated from landraces originating fromTurkey, whereas Pm3�10963 and Pm3�23728 were from Afghani-stan and Turkmenistan, respectively (Table S1).

Are the Previously Uncharacterized Pm3 Alleles Late-Acting Resis-tance Genes? To determine whether the relatively high haustoriumindices of previously undescribed alleles in comparison with theknown Pm3 alleles (13, 14) are due to late action of the previouslyundescribed alleles, we performed lactophenol trypan blue stain-ing. The resistance in landraces with alleles Pm3�42416 andPm3�42255 was silenced by Pm3-specific VIGS, thereby allowingdetailed study of functional activity and specificity of these 2 genesin planta. Therefore, we specifically monitored pathogen growthand cell death [hypersensitive response (HR)] in these landraces at6 different time points [2, 3, 4, 5, 6, and 7 days after inoculation(dpi)]. Cultivar Chul carrying Pm3b was included as a comparisonwith a known allele.

Formation of secondary hyphae and haustoria in the attackedhost epidermal cells was observed at a frequency of 46% and 30%at 2 dpi for Pm3�42416 and Pm3�42255, respectively. Thus, therelatively low rate of resistance observed in the transient assay at 2dpi is in agreement with the low degree of resistance observed inplanta at this time point. For IG42416 and IG42255, almost all cellshaving a haustorium were associated with hypersensitive cell death

at 5 dpi and 3 dpi, respectively (Fig. 4). Resistance triggered by the2 studied genes was thus mainly based on a late hypersensitive celldeath occurring after fungal penetration. For Chul, no susceptibleinteractions were observed at any time point, indicating a rapidPm3b-mediated resistance already at 2 dpi. These experimentsdemonstrated that, although the resistance induction of these 2alleles is slower than the activity of the known alleles, they conferfull resistance and are agronomically useful genes.

One possible reason for the slow response could be low ordelayed expression of these genes. Therefore, semiquantitativeRT-PCR was carried out to compare levels of differentially ex-pressed mRNAs in the landraces possessing the 2 previouslyundescribed alleles (Pm3�42416 and Pm3�42255) and some of theknown rapidly acting Pm3 alleles (Pm3b, Pm3e, and Pm3g or thesusceptible allele Pm3CS). No difference in the expression levels ofthe Pm3 alleles was observed (Fig. S2).

To determine whether the previously undescribed Pm3 alleleshave unique specificities compared with known Pm3 alleles, wecharacterized the landraces carrying Pm3�42416 and Pm3�42255with a set of 6 additional (total 10) powdery mildew isolates (TableS2). On the basis of the data, we conclude that these 2 previouslyundescribed alleles encode new Pm3 specificities and thus broadenthe spectrum of available resistance genes for breeding.

DiscussionPm3 Allele Mining in a Subset of Landraces Specifically Selected forthe Isolation of Powdery Mildew Resistance Genes. With more than560,000 wheat accessions held in nearly 40 gene banks globally(The Global Crop Diversity Trust, 2007; http://www.croptrus-t.org/documents/web/Wheat-Strategy-FINAL-20Sep07.pdf), itis not possible to screen the entire collection for specific traits.Thus, the problem is to find the most effective way to choose asubset of lines to screen. The set should be of a manageable sizefor handling and molecular analysis, with a reasonable proba-bility for containing the relevant trait. Core collections havebeen proposed as a strategy to provide representation of thegenetic diversity of a crop in a collection (21). However, core

A B

C D

E F

Fig. 4. Time course analysis of pathogen growth on thelandraces with the Pm3�42416 or Pm3�42255 allele. (Aand B) Percentages of cells with haustoria in the epider-malcellsof landraceswithPm3�42416 (A)andPm3�42255(B). Blue lines indicate the haustorium index calculatedfromthecellswithhaustoriumbutnoHR;pink linesmarkthe cells with haustoria that also exhibit HR. These ob-servations were recorded at different time points (2, 3, 4,5, 6, and 7 dpi) indicated at the bottom. (C) Microscopicview of cells from the landrace with Pm3�42416 at 2 dpi.This shows the occurrence of interactions with and with-out haustoria formation, whereas none of these leads tocell death at this stage. Arrows indicate spore (Sp), sec-ondary hyphae (SH), and appressorium (App). (D) Thecells with haustoria are accompanied by HR at 5 dpi. Ablue-stained cell indicates cell death resulting from HR.Arrow indicates haustorium (Ha). (E and F) Macroscopicview of the infected leaf segments from landraces withPm3�42416 and Pm3�42255 at 8 dpi, respectively. (Scalebars, 100 �m in C and D, 0.5 cm in E and F.)

9522 � www.pnas.org�cgi�doi�10.1073�pnas.0904152106 Bhullar et al.

Dow

nloa

ded

by g

uest

on

July

13,

202

0

collections aim to maximize genetic diversity, whereas breedersand biologists are usually interested only in one or a few traits ata time from a genetic resource collection (22). This was also thecase in this study: we were specifically interested in previouslyundescribed functional alleles of the gene Pm3.

Our large-scale allele mining for powdery mildew resistance onthe FIGS set of landraces demonstrated the effectiveness of FIGSto identify a manageable and diverse set of material for screening.Forty percent of the collection sites chosen in the process yieldedresistant accessions, and almost 16% of the accessions carriedresistance with a high frequency of variation for the Pm3 loci. Thisis an excellent result from a subset of approximately 8% of thestarting collection of more than 16,000 accessions. In the search formore variation in the powdery mildew resistance loci, these resultscan be used in an iterative process to further refine the FIGSselection process by feeding the data of the resistant accessions intothe next round of selection. For example, matching the geo-coordinates of the accessions originating from eastern Turkey orBamian province in Afghanistan to that of collection sites of genebank accessions may help to identify additional wheat lines withpowdery mildew resistance.

The 7 previously undescribed, functionally active Pm3 allelesextend the previously known Pm3 allelic series and in fact doublethe number of functionally active Pm3 alleles in bread wheat to 14.Thus the Pm3 alleles now represent one of the largest allelic seriesof a resistance gene in plants. Pm3 alleles exhibit a high level(�97%) of sequence identity among themselves (ref. 16 and thisstudy), which is in contrast to the RPP13 locus of downy mildewresistance in Arabidopsis, which shows high polymorphism betweenalleles (23). The allelic series of Mla locus of barley and the flax rustresistance L locus also show a lower level of homology amongalleles. Sequence identity of 94% was reported between the 6cloned Mla alleles (16, 24–26), and 8 of 13 L alleles were found tobe �90% identical (27).

A Combined Strategy for Functional Analysis of the Previously Unde-scribed Pm3 Alleles. We have used VIGS as well as a transienttransformation assay for functional analysis of previously unde-scribed Pm3 alleles. Although VIGS has proven to be an effectivemethod for several dicots (28), it is only recently that its successfuluse in monocots such as barley and wheat has been demonstrated.VIGS was shown to silence the barley phytoene desaturase gene(29) and was successfully used for silencing of leaf rust resistancegenes Lr21 (30) and Lr1 (15). In our study, we found that VIGS wasan effective strategy to assign function to the previously unde-scribed alleles. However, VIGS will only give conclusive results ifthe resistance in a line of interest is based specifically on the silencedgene. According to our data, in a majority of lines resistant topowdery mildew the resistance was not due to the Pm3 allele, orthere was more than 1 resistance gene present. As we learnt fromtransient assays, 3 of the Pm3 alleles present in the 4 lines thatshowed no silencing by VIGS turned out to be active resistancealleles. In these 3 lines, at least 1 additional Pm gene must bepresent, masking the effect of Pm3 silencing. Because of thefrequency in the gene pool of such lines with multiple Pm genes,transient or stable transformation seems to be best suited forfunctional analysis of a candidate resistance gene. Nevertheless, ifa gene can be silenced by VIGS, the resistance in the donor landracecan immediately be attributed to the gene. In such plant accessions,the particular gene can then be studied in detail, allowing assess-ment of resistance activity and specificity, as was demonstrated for2 of our previously undescribed Pm3 alleles.

Transient transformation has previously been found to be aneffective method to study gene function in wheat (13, 14, 31). Whentested with this method, 7 previously undescribed Pm3 allelesshowed significant reduction in the haustorium index as comparedwith the susceptible Pm3CS. Among the tested genes were also the2 genes Pm3�42416 and Pm3�42255, which according to VIGS are

the only mildew-resistance genes in the accessions from which theywere isolated. These 2 genes were found to be late acting, whereasthe known Pm3 alleles provide a rapid resistance response withoutformation of haustoria (ref. 13 and this study). Complete resistanceby these 2 alleles in fact occurred later than 2 dpi when observationwas made in the transient assay. This explains the relatively lowreduction of the haustorium indices by these alleles in thetransient assay. Thus, these genes are not weak alleles but simplyhave a different time-course of activity, making them interestingfor future functional studies at the molecular level. A latehypersensitive cell death–associated resistance has also beenreported for some barley powdery mildew resistance genes, forexample Mla3 and Mla7 (25, 32), and might be a widespreadphenomenon in allelic series of R genes.

Sequence Diversity Lies in the LRR Region of Pm3 Alleles. Theisolation of 16 previously undescribed alleles was possible becauseof the conservation of Pm3 gene structure and the high level ofsequence identity among the Pm3 alleles. The 100% conservationof the N-terminal region encoding a CC domain among thepreviously undescribed as well as the known Pm3 alleles suggests ahighly conserved function of this domain in Pm3 resistance. Thealleles have a mosaic pattern of sequence blocks, which probably arethe result of rearrangement of variation present in the ancestralalleles or which might have arisen from gene conversion events. Thefinding of variability mainly in the LRR domain of the previouslyundescribed alleles is in agreement with the suggested role of thisdomain in recognition specificity (33). In the case of flax rustresistance locus L, the major sequence variation among the L alleleswas also found in the LRR domain (27). The previously unde-scribed Pm3 alleles also show differences that originated fromInDels and point mutations. InDels were also found in the func-tional flax L alleles (27) and among the various members of the Mlafamily (34). On the basis of the identification of conserved smallrepeats flanking the InDels in Pm3 alleles, we propose that theseInDels have originated from illegitimate recombination. Illegiti-mate recombination has been suggested to be a major evolutionarymechanism that is at the basis of the size variability of the LRRdomain of R proteins (35).

Application of Allele Mining in FIGS Sets for Breeding and BasicResearch. Plant breeding relies on the identification of geneticdiversity present in the crop gene pools that can provide functionalvariants of genes. Our data suggest that FIGS is an effectivesampling strategy for the efficient and targeted establishment ofsubsets from gene bank collections. Allele mining on such subsetsthen offers a rapid approach to identify new alleles as comparedwith traditional ways of identifying new sources of resistance.However, the absence of molecular sequence information for themajority of agronomically important genes is a major limitingfactor. For accessing the diversity stored in gene banks, a concen-trated effort is essential to identify the genes involved in agronom-ically relevant traits. Only then will allele mining in FIGS sets fullydevelop its potential.

The previously undescribed genes double the number of theknown functional Pm3 alleles in bread wheat to 14. The previouslyundescribed alleles provide additional diversity for studies on themolecular function and specificity of the Pm3 alleles while enrichingthe genetic basis for resistance breeding in wheat. From results ofthis pilot project, we conclude that FIGS combined with an allelemining approach bears a high potential to be applied to otherimportant crop plants for targeting important traits or genes.

Materials and MethodsPlant Material and FIGS. A FIGS approach was used to assemble a set of 1,320bread wheat landraces from a total of 16,089 accessions present in 3 differentgene banks of the Australian Winter Cereal Collection, the International Centerfor Agriculture Research in the Dry Areas, and the N.I. Vavilov Institute of Plant

Bhullar et al. PNAS � June 9, 2009 � vol. 106 � no. 23 � 9523

PLA

NT

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

July

13,

202

0

Industry (Russia). The 1,320 accessions originated from Turkey (420), Iran (393),Afghanistan (292), Pakistan (131), Armenia (34), Turkmenistan (16), Russia (9),India (6),Azerbaijan (10),Uzbekistan (1), SerbiaandMontenegro (3),Bulgaria (3),Macedonia (1), and Romania (1).

The eco-geographic profile of 400 accessions with known powdery mildewresistance was used as a template to identify environmentally similar collectionsites from the FIGS database of 16,089 landraces (refs. 9 and 18; Street K, et al.,unpublished data). Individual accessions were selected using multivariate statis-tical procedures that determined how eco-geographically similar the collectionsite of a given accession was to the resistant set template.

Powdery Mildew Infections and Isolates. The powdery mildew infections and thescoring were performed as described by Kaur, et al. (18). The choice of isolatesused for screening and characterization was made on the basis of their virulence/avirulence patterns on the lines carrying known Pm3 alleles (Table S2).

Isolation of Pm3 Alleles. Alleles were amplified by using Pm3 locus-specific,long-range PCR amplification, followed by nested long-range PCR (13, 14) usingPfuUltra high-fidelity DNA polymerase (Stratagene). Amplified fragments werecloned into the multiple cloning site of expression vector PGY1 (31). DNA se-quencing was performed with an Applied Biosystems Capillary Sequencer 3730.

Sequence Analysis. Sequence assembly was performed using the Gap4 programof the Staden Package. The ClustalX software (36) was used for sequence align-ments, which were further analyzed in the program Genedoc (http://www.nrb-sc.org/gfx/genedoc/index.html). The different R protein domains (CC, NBS, andLRR) were determined according to Meyers, et al. (37) and Yahiaoui, et al. (13).

Virus-Induced Gene Silencing. PCR-amplified fragments from the Pm3 gene wereinserted into the �-subfragment of the viral genome (29, 30). The infectiousconstructs of barley stripe mosaic virus RNAs were prepared by in vitro transcrip-

tion using T7 DNA-dependent RNA polymerase (mMESSAGE mMACHINE T7 Kit;Ambion).

Single-Cell Transient Transformation Assay and Microscopy. Transient geneexpression assays (13, 31) were performed as described by Yahiaoui, et al. (16).Leavesofthepowderymildew–susceptible lineChancellorwerebombardedwitha 1:1 (wt/wt) mixture of pUbiGUS containing the GUS reporter and the PGY1control vector containing the Pm3CS gene or the previously undescribed Pm3alleles. Eight alleles (Pm3�42416, Pm3�42920, Pm3�9939, Pm3�10963, Pm3�42525,Pm3�23728, Pm3�42255, and Pm3�42469) were tested against isolate Bgt 98275,4 alleles (Pm3�31594, Pm3�42277, Pm3�42868, and Pm3�14442) with Bgt 97011,and 1 allele (Pm3�41606) with Bgt 96224. For selection of the powdery mildewisolate for the transient assay, we took advantage of the fact that the same allelewas identified in several landraces. Therefore, the isolate to which all indepen-dent landraces with identical alleles showed resistance was used to infect thebombarded leaves.

Staining for hypersensitive response was performed on leaf segments ofresistant landraces using lactophenol trypan blue.

Semiquantitative RT-PCR. Total RNA was extracted from leaves of 10-day-oldseedlingsusingTRIzol reagent (InvitrogenLifeTechnologies). ForRT,2 �gof totalRNA was denatured at 70 °C for 5 min in the presence of 0.06 �g of oligo(dT)20

primers. RT was done with the M-MLV reverse transcriptase kit (Promega).Aliquots of the reverse transcripts (2 �L from 1:8 dilution) were then amplified ina 25-�L PCR containing 0.2 mM specific oligonucleotide primers.

ACKNOWLEDGMENTS. We thank the gene banks of Australian Winter CerealCollection, the International Center for Agriculture Research in the Dry Areas(Syria), and the N.I. Vavilov Institute of Plant Industry (Russia) for providing theseed material; Dr. S. Scofield for help with setting up the VIGS system; Drs. S. Bieriand C. Loutre for the modified barley stripe mosaic virus RNA with Pm3 and Lr10fragments, respectively;andDr.S.TravellaforhelpwithsemiquantitativeRT-PCR.This project was funded by FP6 EU Grant ‘‘Bioexploit’’ and by Swiss NationalScience Foundation Grant 3100–105620.

1. Hoisington D, et al. (1999) Plant genetic resources: What can they contribute towardincreased crop productivity? Proc Natl Acad Sci USA 96:5937–5943.

2. Gur A, Zamir D (2004) Unused natural variation can lift yield barriers in plant breeding.PLoS Biol 2:e245.

3. Bernacchi D, et al. (1998) Advanced backcross QTL analysis of tomato. II. Evaluation of near-isogenic lines carrying single-donor introgressions for desirable wild QTL-alleles derived fromLycopersicon hirsutum and L pimpinellifolium. Theor Appl Genet 97:170–180.

4. Piffanelli P, et al. (2004) A barley cultivation associated polymorphism conveys resis-tance to powdery mildew. Nature 430:887–891.

5. Tanksley SD, McCouch SR (1997) Seed banks and molecular maps: Unlocking geneticpotential from the wild. Science 277:1063–1066.

6. Johnson RC (2008) Gene banks pay big dividends to agriculture, the environment andhuman welfare. PLoS Biol 6:e148.

7. Yu J, Bai G, Cai S, Dong Y, Ban T (2008) New Fusarium head blight resistant sources fromAsian wheat germplasm. Crop Sci 48:1090–1097.

8. Singrun C, Rauch P, Morgounov A, Hsam S, Zeller F (2004) Identification of powderymildew and leaf rust resistance genes in common wheat (Triticum aestivum L.). Wheatvarieties from the Caucasus, Central and Inner Asia. Genet Resources Crop Evol51:355–370.

9. Mackay MC, et al. (2004) Focused identification of germplasm strategy—FIGS. Cereals2004. Proceedings of the 54th Australian Cereal Chemistry Conference and the 11thWheat Breeders’ Assembly, 21–24 September 2004, Canberra, Australian CapitalTerritory. eds Black CK, Panozzo JF, Rebetzke GJ (Royal Australian Chemical Institute,Melbourne), pp 138–141.

10. Harjes CE, et al. (2008) Natural genetic variation in lycopene epsilon cyclase tapped formaize biofortification. Science 319:330–333.

11. Huang L, et al. (2003) Map-based cloning of leaf rust resistance gene Lr21 from thelarge and polyploid genome of bread wheat. Genetics 164:655–664.

12. Feuillet C, et al. (2003) Map-based isolation of the leaf rust disease resistance gene Lr10from the hexaploid wheat (Triticum aestivum L.) genome. Proc Natl Acad Sci USA100:15253–15258.

13. Yahiaoui N, Srichumpa P, Dudler R, Keller B (2004) Genome analysis at different ploidylevels allows cloning of the powdery mildew resistance gene Pm3b from hexaploidwheat. Plant J 37:528–538.

14. Srichumpa P, Brunner S, Keller B, Yahiaoui N (2005) Allelic series of four powderymildew resistance genes at the Pm3 locus in hexaploid bread wheat. Plant Physiol139:885–895.

15. Cloutier S, et al. (2007) Leaf rust resistance gene Lr1, isolated from bread wheat(Triticum aestivum L.) is a member of the large psr567 gene family. Plant Mol Biol65:93–106.

16. Yahiaoui N, Brunner S, Keller B (2006) Rapid generation of new powdery mildewresistance genes after wheat domestication. Plant J 47:85–98.

17. Yahiaoui N, Kaur N, Keller B (2009) Independent evolution of functional Pm3 resistancegenes in wild tetraploid wheat and domesticated bread wheat. Plant J 57:846–856.

18. Kaur N, Street K, Mackay M, Yahiaoui N, Keller B (2008) Molecular approaches forcharacterization and use of natural disease resistance in wheat. Eur J Plant Pathol121:387–397.

19. Tommasini L, Yahiaoui N, Srichumpa P, Keller B (2006) Development of functionalmarkers specific for seven Pm3 resistance alleles and their validation in the breadwheat gene pool. Theor Appl Genet 114:165–175.

20. Devos KM, Brown JKM, Bennetzen JL (2002) Genome size reduction through illegiti-mate recombination counteracts genome expansion in Arabidopsis. Genome Res12:1075–1079.

21. Ortiz R (2002) No just seed repositories: A more pro-active role for gene banks. GeneConserve 1:21–24.

22. Mackay MC (1995) One core collection or many? Core Collections of Plant GeneticResources, eds Hodgkin T, Brown AHD, van Hintum ThJL, Morales AAV (John Wiley &Sons, Oxford), pp 199–210.

23. Rose LE, et al. (2004) The maintenance of extreme amino acid diversity at the diseaseresistance gene, RPP13 in Arabidopsis thaliana. Genetics 166:1517–1527.

24. Zhou FS, et al. (2001) Cell-autonomous expression of barley Mla1 confers race-specificresistance to the powdery mildew fungus via a Rar1-independent signaling pathway.Plant Cell 13:337–350.

25. Shen QH, et al. (2003) Recognition specificity and RAR1/SGT1 dependence in barley Mladisease resistance genes to the powdery mildew fungus. Plant Cell 15:732–744.

26. Halterman DA, Wise RP (2004) A single-amino acid substitution in the sixth leucine richrepeat of barley MLA6 and MLA13 alleviates dependence on RAR1 for disease resis-tance signaling. Plant J 38:215–226.

27. Ellis JG, Lawrence GJ, Luck JE, Dodds PN (1999) Identification of regions in alleles of theflax rust resistance gene L that determine differences in gene-for-gene specificity.Plant Cell 11:495–506.

28. Burch-Smith TM, Anderson JC, Martin GB, Dinesh-Kumar SP (2004) Applications andadvantages of virus-induced gene silencing for gene function studies in plants. PlantJ 39:734–746.

29. Holzberg S, Brosio P, Gross C, Pogue GP (2002) Barley stripe mosaic virus-induced genesilencing in a monocot plant. Plant J 30:315–327.

30. Scofield SR, Huang L, Brandt AS, Gill BS (2005) Development of a virus-inducedgene-silencing system for hexaploid wheat and its use in functional analysis of theLr21-mediated leaf rust resistance pathway. Plant Physiol 138:2165–2173.

31. Schweizer P, Christoffel A, Dudler R (1999) Transient expression of members of thegermin-like gene family in epidermal cells of wheat confers disease resistance. Plant J20:541–552.

32. Boyd LA, Smith PH, Foster EM, Brown JKM (1995) The effects of allelic variation at theMla resistance locus in barley on the early development of Erysiphe graminis f.sp.hordei and host responses. Plant J 7:959–968.

33. DeYoung BJ, Innes RW (2006) Plant NBS-LRR proteins in pathogen sensing and host-defense. Nat Immunol 7:1243–1249.

34. Wei F, Wing RA, Wise RP (2002) Genome dynamics and evolution of the Mla (powderymildew) resistance locus in barley. Plant Cell 14:1903–1917.

35. Wicker T, Yahiaoui N, Keller B (2007) Illegitimate recombination is a major evolu-tionary mechanism for initiating size variation in plant resistance genes. Plant J51:631– 641.

36. Thompson JD, Gibson TJ, Plewniak F, Jeanmougin F, Higgins DG (1997) The CLUSTAL�Xwindows interface: Flexible strategies for multiple sequence alignment aided byquality analysis tools. Nucl Acids Res 25:4876–4882.

37. Meyers BC, Kozik A, Griego A, Kuang H, Michelmore RW (2003) Genome-wide analysisof NBS-LRR-encoding genes in Arabidopsis. Plant Cell 15:809–834.

9524 � www.pnas.org�cgi�doi�10.1073�pnas.0904152106 Bhullar et al.

Dow

nloa

ded

by g

uest

on

July

13,

202

0