Cr

YRKMa

b

c

d

Ie

f

a

ARRAA

KMRQSG

1

l7[ii[

Mo

U

h0

Plant Science 242 (2016) 131–139

Contents lists available at ScienceDirect

Plant Science

j ourna l ho me pa ge: www.elsev ier .com/ locate /p lantsc i

onstruction of a versatile SNP array for pyramiding useful genes ofice

usuke Kurokawaa,1, Tomonori Nodaa,1, Yoshiyuki Yamagatac,d,e,osalyn Angeles-Shima,b,c, Hidehiko Sunohara f, Kanako Ueharaa, Tomoyuki Furutaa,eisuke Nagaia, Kshirod Kumar Jenab, Hideshi Yasuid,e, Atsushi Yoshimurad,e,otoyuki Ashikari a,c,d, Kazuyuki Doic,d,f,∗

Bioscience and Biotechnology Center, Nagoya University, Furo-cho, Chikusa, Nagoya, Aichi 464-8601, JapanNovel Gene Resources Laboratory, International Rice Research Institute, DAPO Box 7777, Metro Manila, PhilippinesWonder Rice Initiative for Food Security and Health (WISH) project, Japan International Cooperation Agency (JICA), JapanScience and Technology Research Partnership for Sustainable Development (SATREPS) project, Japan Science and Technology Agency (JST) and Japan

nternational Cooperation Agency (JICA), JapanPlant Breeding Laboratory, Kyushu University, 6-10-1Hakozaki, Higashi, Fukuoka 812-8581, JapanGraduate School of Bioagricultural Sciences, Nagoya University, Furo-cho, Chikusa, Nagoya, Aichi 464-8601, Japan

r t i c l e i n f o

rticle history:eceived 4 June 2015eceived in revised form 4 September 2015ccepted 7 September 2015vailable online 10 September 2015

eywords:

a b s t r a c t

DNA marker-assisted selection (MAS) has become an indispensable component of breeding. Singlenucleotide polymorphisms (SNP) are the most frequent polymorphism in the rice genome. However,SNP markers are not readily employed in MAS because of limitations in genotyping platforms. Here theauthors report a Golden Gate SNP array that targets specific genes controlling yield-related traits andbiotic stress resistance in rice. As a first step, the SNP genotypes were surveyed in 31 parental vari-eties using the Affymetrix Rice 44K SNP microarray. The haplotype information for 16 target genes was

ASice breedingTLNPene pyramiding

then converted to the Golden Gate platform with 143-plex markers. Haplotypes for the 14 useful alleleare unique and can discriminate among all other varieties. The genotyping consistency between theAffymetrix microarray and the Golden Gate array was 92.8%, and the accuracy of the Golden Gate arraywas confirmed in 3 F2 segregating populations. The concept of the haplotype-based selection by usingthe constructed SNP array was proofed.

© 2015 The Authors. Published by Elsevier Ireland Ltd. This is an open access article under the CC BY

. Introduction

By 2050, the world population is expected to surpass 9 bil-ion. To address the monumental challenge of feeding 9 billion, a0% increase in global agricultural production has to be reached[1], http://faostat.fao.org/default.aspx]. Rice is one of the most

mportant food crops, providing up to 76% of the caloric intaken Southeast Asia and up to 23% of the caloric needs worldwide2,3]. Increasing rice production, therefore, would play a key role

Abbreviations: QTL, quantitative trait loci; SNP, single nucleotide polymorphism;AS, marker-assisted selection; SSR, simple sequence repeat; BX, BeadXpress; OPA,

ligo pool assay; InDel, insertion/deletion.∗ Corresponding author at: Graduate School of Bioagricultural Sciences, Nagoyaniversity, Furo-cho, Chikusa, Nagoya 464-8601, Japan.

E-mail address: kdoi@agr.nagoya-u.ac.jp (K. Doi).1 These authors equally contributed to this study.

ttp://dx.doi.org/10.1016/j.plantsci.2015.09.008168-9452/© 2015 The Authors. Published by Elsevier Ireland Ltd. This is an open access

license (http://creativecommons.org/licenses/by/4.0/).

in efforts to secure the world food supply. Superior rice varietieswith high yield, good eating quality, pest and disease resistance andother good agronomic characteristics have been developed by tra-ditional breeding. However, there is still a continuing demand fornew varieties that would further improve rice production. To thisend, gene pyramiding has been an efficient breeding strategy thatallows incorporation of multiples genes of agronomic importanceinto a single variety [4].

Currently, a number of useful genes or quantitative trait loci(QTLs) that are related to agronomical traits have been identified.Genes controlling yield-related traits such as panicle or seed shape[5,6], as well as biotic stress resistance genes that can be used torationally manage rice pests and diseases [7–9] have been identi-fied and cloned. DNA markers are now available to precisely select

useful alleles for these traits and pyramid them in rice.

With the development of DNA techniques, marker-assistedselection (MAS) has become an indispensable component of breed-ing. MAS not only eliminated the extensive trait evaluation involved

article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

132 Y. Kurokawa et al. / Plant Science 242 (2016) 131–139

Table 1Rice varieties used in this study.

ID Variety Descriptiona Target gene(s)b

1 ADR52 Brown planthopper resistance bph25, Bph262 ARC10313 Green rice leafhopper resistance Grh2, Grh43 Asominori Green rice leafhopper resistance Ovc4 ASU Cold tolerance/recipient5 Azucena Drought tolerance/recipient6 Basmati217 Good eating quality/recipient7 Basmati370 Good eating quality/recipient8 DV85 Green rice leafhopper resistance, bacterial blight resistance Grh2, Grh4, Xa79 Fukuhibiki High yield/recipient10 IR24 Background of IRBB21 and IRBB4/5/13/2111 IRBB21 Bacterial blight resistance Xa2112 IRBB4/5/13/21 Bacterial blight resistance Xa21, Xa413 Kandang18 Vietnamese cultivar/recipient14 Kinandang Patong Drought tolerance (deep root)/recipient15 Kinmaze Japanese cultivar/recipient16 Koshihikari Japanese cultivar/recipient17 Kuchum Cold tolerance/recipient18 LO1050 Cold tolerance/recipient19 Mizuhochikara High biomass, Japanese cultivar/recipient20 Nipponbare Variety of rice reference genome21 Pakhe Dhan Cold tolerance/recipient22 Sensho Blast resistance pi2123 Silewah Cold tolerance/recipient24 ST12 High yield Gn1ac , WFP, APO1c

25 ST6 High yield, New plant type Gn1ac

26 T65BPH25/26 Brown planthopper resistance bph25, Bph2627 T65GRH2/4/6 Green rice leafhopper resistance Grh2, Grh4, Grh628 Taichung65 Taiwanese cultivar/recipient29 TAL214 Good eating quality/recipient30 TSC3 Brown planthopper resistance (unidentified)31 V103S Vietnamese cultivar/recipient

e 2.r gene

ivppmgogaoom

bacro9hio

eaoGnggwwt

a Only genes targeted in this study were indicated. References are shown in Tablb “/recipient” indicate the varieties which will be potentially used as recipients oc Allele were confirmed by sequencing and QTL analysis (data not shown)

n gene pyramiding but also reduced the breeding duration for aariety. DNA markers such as restriction fragment length polymor-hism (RFLP), simple sequence repeat (SSR), and single nucleotideolymorphism (SNP) have been used in MAS, with SSRs being theost common marker of choice due to their abundance in the

enome, robustness, reproducibility, and low cost. SNPs, on thether hand, have not been used as extensively in breeding pro-rams because of the difficulty in their detection. Thus far, thepplication of SNPs in marker-aided breeding has entailed labori-us and complicated techniques such as allele-specific PCR [10,11]r restriction-enzyme based methods like cleaved amplified poly-orphic sequence (CAPS) [12].In spite of this, the utilization of SNPs as molecular markers for

reeding is becoming a real possibility. SNPs make up the largestmount of DNA polymorphism in the eukaryotic genome and hencean be used more rationally in marker-based breeding [13–15]. Inice, 1.7 million SNPs have been detected by comparative analysisf the draft genomic sequences of cv. Nipponbare (japonica) and3-11 (indica) [16,17]. A vast amount of information on 160,000igh-quality rice SNPs is also now available in OryzaSNP [18] and

n newer databases that serve as repositories for the vast amountf information generated from the 3000 rice genomes project [19].

As for SNP detection, modern SNP genotyping techniques arenabling automated, multi-loci allele calling in many samples at

time [13]. In recent years, there has been a shift from the usef the microarray platform towards the use of Illumina’s Goldenate technology for SNP detection [20–23]. The Golden Gate tech-ology is based on allele-specific extension and ligation, and canenotype 1536 SNPs on 96 samples in its original format [24]. This

ives the advantage of an efficient, high-throughput marker systemith lower cost per data point unlike the microarray-based method,hich might be advantageous in detecting a large number (more

han 40,000) of SNPs, but is too expensive for breeding purposes.

tic backgrounds.

The highly automated Golden Gate-based method also requiresonly a minimum quantity of DNA, allowing selection in the seedlingstage before transplanting. More importantly, the Golden Gate sys-tem offers the advantage of haplotype-based selection because itcan simultaneously detect appropriate numbers of multiple SNPsfor MAS (48-1536 or more SNPs).

In the present study, we developed a MAS system based on SNPs.As a first step, the whole genome SNP information for 31 rice linesthat were selected for the authors’ breeding project was collectedusing the Affymetrix’s SNP microarray [14,25]. The SNPs within thegenomic regions that are linked to 16 target genes were then con-verted to Illumina’s Golden Gate Veracode oligo pool assay (OPA).The quality of SNP detection and the application of the SNP arrayfor haplotype-based MAS were confirmed using 24 parental linesout of the 31 varieties, and early breeding populations.

2. Materials and methods

2.1. Plant materials and target genes

Thirty-one rice varieties that were selected for the authors’breeding project, Wonder Rice Initiative for Food Security andHealth (WISH) were used in this study. WISH is a breeding pro-gram that aims to improve the yield and biotic stress resistanceof existing rice varieties that are preferentially grown by farmersfor their inherent adaptation to a wide range of environments. The31 varieties contain the donors of 16 target genes and potentialrecipient varieties, which are leading varieties, high-biomass vari-eties, or varieties with abiotic stress tolerances, and 1 variety with

unidentified insect resistance gene(s) (Table 1).

A total of 16 target genes controlling yield-related traitsand pest resistance are listed in Table 2. The map positions ofthe cloned genes were assigned using the Nipponbare genomic

Y. Kurokawa et al. / Plant Science 242 (2016) 131–139 133

Table 2Target genes for SNP selection.

Gene Trait ID (IRGSP) ormappedposition

Chromo-some

Position References Number of lociin custom OPA

Gn1a/OsCKX2 High yield Os01g0197700 1 5,270,449–5,275,585 [26] 12GW2 Seed shape Os02g0244100 2 8,120,821–8,121,387 [27] 2Grh4 Green rice

leafhopperResistance

Tightly linkedto XNpb144

3 14,904,428–15,106,953a [28] 10

GS3 Seed shape Os03g0407400 3 16,729,501–16,735,109 [29,30] 1Grh6 Green rice

leafhopperResistance

RM8213-C60248

4 4,290,544–4,647,693a [31] 10

pi21 Blast resistance Os04g0401000 4 19,835,206–19,836,892 [32] 1qSW5 Seed shape Os05g0187500 5 5,360,574–5,360,727 [33] 10bph25 Brown plant

hopperresistance

Tightly linkedto RM6273 andRM6775

6 382,564–1,594,975a [34] 22

Ovc Ovicidal geneto planthoppers

Tightly linkedto R1954

6 4,737,177–5,036,025a [35] 10

APO1 High yield Os06g0665400 6 27,480,082–27,481,450 [36] 10Xa7 Bacterial leaf

blightresistance

RM20576 -RM340

6 28,796,789–29,000,232a [37,38] 7

WFP/OsSPL14 High yield Os08g0509600 8 25,274,541–25,278,696 [39] 10Xa21 Bacterial leaf

blightresistance

Os11g0559200 11 20,802,978–20,806,262 [40] 9

Grh2 Green riceleafhopperResistance

Tightly linkedto G1465

11 23,231,248–23,607,373a [28] 9

Xa4 Bacterial leafblightresistance

R1506 - S12886 11 27,822,819–28,130,150a [41] 8

Bph26 Brown planthopperresistance

Tightly linkedto RM5479

12 22,607,871–23,340,348a [34] 13

own.

sTdbPawa

2

un(r7

A

2

amtdwt

a For the genes that are not cloed, the regions covered by the SNP markers are sh

equence as reference [IRGSP1, http://http://rapdb.dna.affrc.go.jp].he genomic positions of the genes that are not yet cloned wereetermined based on available public information or data obtainedy the authors. Alleles for GW2 and GS3 were discriminated by usingCR-based method. All of the 31 varieties carry small allele of GW2,nd 7 varieties carry long-allele of GS3. Functional alleles of qSW5ere known in not all varieties. For reducing redundancy, GW2, GS3

nd qSW5 are not indicated in Table 1.

.2. DNA microarray analysis (Rice 44k SNP genotyping array)

Rice genomic DNA was extracted from young, green leaf tissuesing the DNeasy Plant Mini Kit (QIAGEN). Hybridization and sig-al generation using the GeneChip Rice 44k SNP Genotyping ArrayAffymetrix) were conducted following the methods of [25]. Fluo-escence intensity was detected using the GeneChip Scanner 3000G. “.CEL” files containing signal intensities were obtained.

The genotypes of the 31 varieties were called using theLCHEMY program [42] following the methods of [25].

.3. Determination of SNP markers for custom Golden Gate assay

To select the SNP markers to be converted into Golden Gatessay, graphical genotypes based on the microarray data were firstade visible using the Flapjack program [43]. The haplotype (pat-

ern of SNPs) information from approximately 200 kb upstream andownstream regions of the target genes was then collected. SNPsith low quality, i.e. low call rate were removed and haplotypes

hat were as unique as possible to the useful alleles of the target

genes were visually selected, taking into consideration the distanceto the gene positions and balance of the minor allele frequencies.Up to a total of 22 SNP loci were chosen for each target gene.

The positions of the selected SNP sites were corrected based onthe reference sequence available at IRGSP1 [http://rapdb.dna.affrc.go.jp/] and flanking sequences on the microarray (17 + 17 bases) foreach SNP sites were verified using the information available fromthe Rice Diversity project website [http://ricediversity.org/data/sets/44kgwas/]. Additional flanking sequences (200 + 200 bases)were obtained from the reference Nipponbare genome and wereused for designing the Golden Gate OPA. Functional nucleotidepolymorphism (FNPs) for the genes, GW2 (2 SNPs), GS3 (1 SNP) andpi21 (1 insersion/deletion (InDel)), were directly used to design theOPA.

The assay design tool (ADT) provided by Illumina was used todesign the custom OPA. A total of 143 SNPs comprise the finalversion of the custom OPA.

2.4. Golden Gate assay

To verify the preciseness of the custom OPA, the 24 varietieswhich covered all of the 16 genes and LG10, the donor of GW2,were genotyped with the Golden Gate assay using the IlluminaBeadXpress instrument following the manufacturer’s instruction.

The same DNA samples used in the microarray analysis were used.Genotypes were called using the Genome Studio Genotyping Mod-ule version 1.8.4 (Illumina). SNPs with low dispersion in genotypeclusters (GC10 less than 0.5) were removed.

1 t Scien

2u

px1esC3-wowstccv

3

3

wcdisbfpahi

3

gt[m1rsdstaf(iGwnO[ucmUw

34 Y. Kurokawa et al. / Plan

.5. Verification of the feasibility of custom Golden Gate OPAsing segregating populations

To confirm the feasibility of the custom Golden Gate OPA, 3 F2opulations from the crosses ST12 x IRBB4/5/13/21, T65BPH25/26

ST12 and ST6 x T65GRH2/4/6 were used. DNAs of 12, 15 and5 plants, respectively, from each of the 3 F2 populations werextracted and genotyped using the custom SNP array. Using theame DNA samples, genotypes at the GN1indel (primer pair: 5′-CTTGTCCCTTCTACAATGG-3′ and 5′- AGTTGAGCATGAGGAGCACT-′) and RM5493 ([44], primer pair: 5′- GCGGTAACAAACCAACCAACC3′ and 5′- AAAGCAGGACACAGTCACACAGG -3′) loci of all samplesere determined by PCR and electrophoresis of PCR amplicons

n 4% agarose gel. GN1indel is located at 5275477- 5275606 bpithin the coding region of Grain Number 1a (Gn1a) on rice chromo-

ome 1, whereas RM5493 is approximately 750 kbp downstream ofhe Wealthy Farmer’s Panicle (WFP) locus on chromosome 8. Con-ordance between the genotypes of samples generated using theustom SNP array and the two PCR markers were compared toerify the accuracy of the SNP array.

. Results

.1. SNP survey using microarray

Whole genome SNP analysis of the 31 parental varieties thatere previously selected for the authors’ breeding program was

onducted using the Rice 44k SNP Genotyping Array which canetect 1 SNP per 5 kb genomic DNA [14,25] (Table 1, Supplementary

nformation 1). Because the Affymetrix’s genuine BRLMM-P (Beye-ian Robust Linear Model with Mahalanobis with perfect match)asecaller program was not applicable for rice and not appropriateor this small population size, the authors employed the ALCHEMYrogram [25,42]. Using the same parameters (call rate >70%, minorllele frequency >1%) for the informatics, a comparable number ofigh-quality SNPs (34,985) were obtained in this study (vs 36,901

n [25]).

.2. Selection of SNPs

Linkage disequilibrium (LD) of SNPs within the subspeciesroups of Oryza sativa has been reported to be greater than 500 kb inemplate japonica, 150 kb in tropical japonica, and 75 kbp in indica45]. The rice genome is approximately 389 Mb [46] and linkage

aps typically contain approximately 1500 cm [47,48]. Therefore, cm (one recombination per 100 meiosis per chromosomes) inice is considered to be approximately 260 kb. Based on this con-ideration, the authors defined the 200 kb region upstream andownstream of the target gene as the haplotype block to allowelection for target alleles with sufficient precision. For example, aotal of 10 SNP markers representing the haplotype of the WFP-ST12llele were selected (Fig. 1). In the same manner, the SNP markersor 4 other cloned genes, Gn1a (12 SNPs), qSW5 (10 SNPs), APO110 SNPs) and Xa21 (9 SNPs) were selected by graphical genotyp-ng. Because the causal mutations (SNPs or insetion/deletions) forW2, GS3, and pi21 are already known [27,29,30,32], these sitesere directly used to design the custom array. For genes that areot yet cloned, linkage map information was used. For example,vc is located near the RFLP marker R1954 on rice chromosome 6

35]. Therefore, the haplotype surrounding R1954 (±200 kb) wassed for the custom array. SNP sites of the remaining genes were

onsidered based on publicly available literature and additionalapping information obtained by the authors (data not shown).sing linkage, the accuracy of SNP selection (markers within 5 cm)as estimated to be higher than 95%.

ce 242 (2016) 131–139

3.3. Custom OPA design

Selected SNPs from the haplotype blocks of each of the 16 tar-get genes were regarded as first step candidates for SNP markers.ADT scoring was conducted using the first step candidate mark-ers and markers with low ADT score (< = 0.5) were replaced withother possible SNPs. A total of 143 sites including 142 SNPs and 1InDel were converted to Golden Gate custom OPA (Illumina OPAID: VC0014297-OPA, Table 2, Supplementary information 2).

3.4. Comparison of microarray and Golden Gate genotypes

In the SNP microarray analysis, genotype calls using ALCHEMYgave an output in an [AA, AB, BB] format. The AA indicates an allelethat is identical with that of the reference Nipponbare genome.In the present study however, the data obtained from Nipponbarecontained BB. This may be due to diversity within the variety prob-ably because the sample was maintained at Nagoya University. Thegenotype calls using the Genome Studio software were convertedto [AA, AB, BB] format (Supplementary information 3) before theresults from microarray and Golden Gate array were compared(Fig. 2 and Supplementary information 4).

The call rate of the parent varieties by the custom OPA was98.5% (Supplementary information 3). The comparison of haplo-types obtained for the WFP gene is shown in Fig. 2. In the 24varieties tested, and when considering 139 SNPs designed fromthe Affymetrix SNP microarray, the rate of genotype concordancebetween SNP microarray and Golden Gate array was 95.2% (Sup-plementary information 4). This reduced rate is because of thepresence of low (or medium) quality SNPs in the custom OPA, 23out of the 139 SNPs contained missing data greater than 10% (>2missing data) among the 24 varieties. This is remarkable in the SNPscovering Xa4 gene, 7 out of the 8 SNPs showed consistency in lessthan 16 out of the 24 varieties. If the SNPs for Xa4 were removed,the consistency was improved to 97.6%. The 2 FNPs for GW2 andone FNP for GS3 was precisely detected and corresponded to geno-types obtained based on PCR (data not shown). The FNP for pi21(21 bp InDel [32]) was not successful in the custom OPA. On thewhole, the results indicate that genotype data obtained from thecustom Golden Gate OPA is consistent with that of SNP microarray.The custom OPA enabled haplotype-based or FNP-based MAS in atotal 14 genes out of the 16 target genes (excepting for pi21 andXa4).

3.5. SNP genotyping using F2 populations

Because Gn1a and WFP significantly change the panicle shape[24,37] and suitable for haplotype-based selection because causalmutations are not known, three F2 populations that segregatedthese 2 genes were picked up and used for the verification ofthe SNP array in actual segregating populations. The genotypesof the 3 F2 populations were summarized in Fig. 3. The 4 SNPmarkers, id1004195 to id1004310, co-segregated with GN1indel.One out of the 15 F2 plants from T65BPH25/26 x ST12 showedrecombination between id1004146 and id1004198. One recombi-nation was observed between id1004230 and id1004310 in theST6 x IRBB4/5/13/21 F2 population. Three out of the 27 plants(ST12 x IRBB4/5/13/21 and T65BPH25/26 x ST12 crosses takentogether) possessed recombination between the likely WFP geno-types and RM5493, and 1 recombination was observed within the

SNP markers id8006925 and id8006944. We thus conclude thatthe SNP markers detected using the custom OPA showed a goodco-segregation to PCR-based markers and sufficient for the use ofMAS.

Y. Kurokawa et al. / Plant Science 242 (2016) 131–139 135

Fig. 1. Haplotypes encompassing WFP gene obtained using Affymetrix microarray and SNPs for designing custom SNP array. Schematic physical map of rice chromosome8 (top) and haplotypes of the region containing WFP locus (bottom) were shown. Coding sequence of WFP is from 25,274,541 bp to 25,278,696 bp. Genotypes from SNPsu The SG nd mid nces t

4

tfeecmTiwadet

acMdigrcnaim

d8001656 to id8006986 were considered as candidates for SNP markers for WFP.enotypes AA, BB, and – indicate the same allele as Nipponbare, alternative allele aonor of WFP, and yellow are alternative genotypes. (For interpretation of the refere

. Discussion

MAS is a common technique used in rice breeding programs. Tohis day, PCR-based marker systems are the most commonly usedor MAS. However, the PCR-based markers, in combination withlectrophoresis have the disadvantage of low throughput. Only sev-ral markers at most, could be detected in one PCR reaction andonventional electrophoresis lane. This limits the application of thearker system for “foreground” selection only of a few target genes.

he practical use of multiplex marker has long been awaited, butts realization has been prevented by the lack of technology that

ould cut both costs and processing time. Recently, “breeders” SNPrrays have been developed for rice [20,21,22,23]. These arrays areesigned to cover a wide range of cross combinations in rice how-ver, they are only suitable for “background” selection and are notargeted for specific genes.

Unlike the previously reported arrays [20,21,22,23], the SNPrray developed in this study contains a SNP set targeting spe-ific genes of interest, and first provide a platform for foregroundAS for multiple genes. The 95.2% match between the genotypic

ata obtained using microarray and the custom Golden Gate arrays comparable with the previously reported 82% match [14]. SNPenotyping using segregating F2 populations confirmed the accu-acy of the custom array. The approach used in the present studyan be used and applied to future breeding projects that require a

ew set of SNPs. The sequence diversity of rice is being intensivelynalyzed by various projects (e.g. [19]). Therefore, the basic SNPnformation that is similar to those generated in this study using

icroarray will be available through public databases in the future.

NP sites with asterisks at the bottom were converted to Golden Gate custom OPA.ssing data, respectively. Genotypes with blue background are identical to ST12, theo color in this figure legend, the reader is referred to the web version of this article.)

Elimination of additional experimentation for SNP discovery willallow faster design and fabrication of custom arrays.

In this study, functional nucleotide polymorphisms (FNPs) forGW2, GS3 and pi21 were directly included in the custom array. TheFNPs for GW2 and GS3 can be detected in the custom array. How-ever, FNPs for most of the target genes, even in the cloned genes,have not been identified as in the case of WFP. WFP was clonedas a gene controlling the number of primary branches per panicle,but the FNP responsible for increased number of primary branch-ing remains unknown [39]. In this case, it is difficult to find a DNAmarker that would segregate with the useful allele. The haplotypeselection approach is useful in such a case because the useful allelecan be monitored using a set of SNP markers encompassing the tar-get loci. Multiplexed SNP markers are considered to be suitable forthis purpose, enabling a versatile SNP array work in various crosscombinations.

The SNP array constructed in the present study is useful in con-structing a set of NILs that are suitable for trait evaluation becausethe SNPs can be used to select useful alleles in a wide range ofgenetic backgrounds. The tightly linked set of SNPs can also be usedfor dissecting “linkage drag”. In the present study, some recom-binants within the haplotype regions were detected (Fig. 3). Thisindicates that the haplotypes around the target genes can be uti-lized for the fine genetic dissection of regions near the target genes.A typical case of a linkage between a useful gene and a harmful

gene was reported in [32]. If a large segregating population is avail-able, the SNP array can be readily used to construct an NIL witha very small introgressed chromosome segment from the donorparent.

136 Y. Kurokawa et al. / Plant Science 242 (2016) 131–139

Fig. 2. Comparison of genotypes obtained from Affymetrix microarray and BeadXpress. Schematic physical map around WFP locus showing the location of the SNP markers( icroarA dataW color i

al(ibltlrcm[igdtwm

tuiwobtboa

top) and comparison of genotypes between those detected with Affymetrix SNP mA, BB, and – indicate the same allele as Nipponbare, alternative allele and missingFP, and yellow are alternative genotypes. (For interpretation of the references to

The simultaneous detection of useful alleles potentially allows dramatic decrease in labor and time required to develop pyramidines. Development of DNA markers has allowed gene pyramidinge.g. [31,34,35,49]). In the initial concept of gene pyramiding, near-sogenic lines (NILs) for each of the target genes must be generatedefore the construction of pyramiding lines [4]. This requires the

aborious, repeated selections and crossings to generate NILs. Addi-ional effort is also necessary to combine multiple genes into oneine. The SNP array in this study has the potential to cut the timeequired to pyramid multiple genes into a single variety by direct-rossing of gene donors and MAS in the progeny. For example, aulti-parent advanced generation inter-cross (MAGIC) population

50,51] can be developed using relatively simple ways of cross-ng. From a MAGIC population, candidate lines with multiple usefulenes can be selected using the SNP array. The low frequency of theesirable genotype can be overcome by the use of a large popula-ion. In addition, this method of direct pyramiding of useful genesill also contribute in increasing the genetic diversity in breedingaterials.However, cost for SNP detection is still expensive for most of

he rice breeders. Currently the BeadXpress platform can processp to 96 samples at a time and the cost is still expensive for use

n a breeding project. These limitations are potentially problematichen handling a large segregating population. The disadvantage

f the GoldenGate system is in its high initial cost and low flexi-ility. Although the cost for 1 genotype per sample is comparable

o PCR-based methods (data not shown), the one purchase of OPAecomes quite expensive (∼$10,000) and not acceptable for mostf the local breeders. However, the potential of SNP markers forutomated and simultaneous detection must be advantageous and

ray (Affy) and Golden Gate with BeadXpress (BX) (bottom) were shown. Genotypes, respectively. Genotypes with blue background are identical to ST12, the donor ofn this figure legend, the reader is referred to the web version of this article.)

will be expanded. It is expected that above limitation will be clearedby the use of next-generation sequencers (NGS). However, the tar-geted SNP detection with NGS still needs a complicated samplepreparation method [52]. On the other hand, one of the NGS-basedgenotyping methods, genotyping by sequencing (GBS) is becomingcommon (e.g. [53]). GBS enables multiplexing in both samples andmarkers (up to 768 samples and hundreds to thousands of markersin 1 NGS run) but is not suitable for targeting useful SNPs. It is stillnecessary for geneticists and breeders to choose a suitable genotyp-ing platform in their breeding project. Wide use of the GoldenGatearray might be limited because of the cost problem, so the SNP arraywill be used for monitoring whether the newly developed lines pos-sesses some of useful allele or not, which is informative for furthergene pyramiding and MAS.

To enhance rice production by new varieties, development ofactual plant materials such as NILs is required to verify the effectsof yield-improving genes. It should be considered that the effectof yield-enhancing genes such as Gn1a [26], WFP [39] or APO1[36], are not confirmed yet in other genetic background and actualfarmer’s field. Therefore, these potential yield-improving genesshould be tested in various genetic backgrounds. The “New PlantType” approach [54] was based on the concept of increasing the“sink size”, and the aforementioned yield enhancing genes areconsidered to be associated with sink size. Furthermore, a genefor increasing rice “source” ability have been recently reported[55]. Currently, scientists have a good technology for MAS, but

efforts on actual material development are limited. The breedingproject Wonder Rice Initiative for Food Security and Health (WISH)was launched by the authors’ group, as an effort to provide pre-varieties to rice scientists and breeders worldwide. In this project,

Y. Kurokawa et al. / Plant Science 242 (2016) 131–139 137

Fig. 3. Confirmation of co-segregation between SNP markers as detected using BeadXpress and PCR-based markers. Genotypes of individual F2 plants obtained fromT65BPH25/26 x ST12, ST6 x T65GRH2/4/6 and ST12 x IRBB4/5/13/21, and parents were shown on the right. Locations of Gn1a and WFP and DNA markers are shown asp nbarea s. (Fort

ttdmsmelbtfoatw

A

on

hysical maps on the left. Genotypes AA, BB, and – inidicate the same allele as Nippore identical to ST12 or ST6, yellow are alternative allele, and green are heterozygouo the web version of this article.)

he authors envision incorporating multiple useful genes (Table 2)o recipient varieties (Table 1) and generating pre-varieties foristribution. The anticipated pre-varieties contain two types ofaterials: (1) breeding materials with uniform backgrounds (i.e.

ets of NILs / pyramid lines with multiple backgrounds) that provideore convenience to rice breeders because they can evaluate the

ffect of incorporated genes on their own breeding sites, (2) singleine with many useful genes (i.e. highly-pyramided lines) that wille conveniently used because multiple genes can be incorporatedo target varieties with a single cross, and (3) the SNP informationor useful genes will be available. The versatile SNP array devel-ped in this study will largely contribute in facilitating the breedingctivity. Both plant materials and SNP arrays will be made availableo help improve varieties that are adapted to various regions in theorld.

cknowledgments

This work was supported in part by the SATREPS projects “Devel-pment of crop genotypes for midlands and mountain areas oforth Vietnam” and “Improvement in productivity and yield stabil-

, alternative allele and missing data, respectively. Genotypes with blue background interpretation of the references to color in this figure legend, the reader is referred

ity of rice under Kenya’s biotic and abiotic stress conditions throughtailor-made breeding and development of cultivation methods”,and National Bio-Resource Project for Rice. The authors expresstheir heartfelt gratitude to Professors Makoto Matsuoka, HidemiKitano, Mr. Nobuhiko Hanazato and JICA staffs for supportingthe WISH project. The authors thank Drs. Chi-Wei Tung, Mark“Koni” Wright and Susan McCouch for their kind instruction in SNPmicroarray analysis.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.plantsci.2015.09.008.

References

[1] H.C.J. Godfray, J.R. Beddington, I.R. Crute, L. Haddad, D. Lawrence, J.F. Muir, J.Pretty, S. Robinson, S.M. Thomas, C. Toulmin, Food security: the challenge offeeding 9 billion people, Science 327 (2010) 812–818.

[2] G.S. Khush, Productivity improvements in rice, Nutr. Rev. 61 (2003) 114–116.

1 t Scien

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

38 Y. Kurokawa et al. / Plan

[3] R. Huang, L. Jiang, J. Zheng, T. Wang, H. Wang, Y. Huang, Z. Hong, Genetic basisof rice grain shape: so many genes, so little known, Trends Plant Sci. 18 (2013)218–226.

[4] M. Ashikari, M. Matsuoka, Identification, isolation and pyramiding ofquantitative trait loci for rice breeding, Trends Plant Sci. 11 (2006) 344–350.

[5] K. Miura, M. Ashikari, M. Matsuoka, The role of QTLs in the breeding ofhigh-yielding rice, Trends Plant Sci. 16 (2011) 319–326.

[6] M. Ikeda, K. Miura, K. Aya, H. Kitano, M. Matsuoka, Genes offering thepotential for designing yield-related traits in rice, Curr. Opin. Plant Biol. 16(2013) 213–220.

[7] X. Wang, S. Lee, J. Wang, J. Ma, T. Bianco, Y. Jia, Current advances on geneticresistance to rice blast disease, in: W. Yan, J. Bao (Eds.), Rice—Germplasm,Genetics and Improvement, InTech, Rijeka, 2014, pp. 195–217.

[8] M.A. Khan, M. Naeem, M. Iqbal, Breeding approached for bacterial leaf blightresistance in rice (Oryza sativa L.), current status and future direction, Eur. J.Plant Pathol. 139 (2014) 27–37.

[9] G. He, B. Du, R. Chen, Insect resistance, in: Q. Zhang, R.A. Wing (Eds.), Geneticsand Genomic of Rice, Springer, New York, 2013, pp. 177–192.

10] C.R. Newton, A. Graham, L.E. Heptinstall, S.J. Powell, C. Summers, N. Kalsheker,J.C. Smith, A.F. Markham, Analysis of any point mutation in DNA. Theamplification refractory mutation system (ARMS), Nucl. Acids Res. 17 (1989)2503–2516.

11] K. Hayashi, N. Hashimoto, M. Daigen, I. Ashikawa, Development of PCR-basedSNP markers for rice blast resistance genes at the Piz locus, Theor. Appl.Genet. 108 (2004) 1212–1220.

12] A. Konieczny, F.M. Ausubel, A procedure for mapping Arabidopsis mutationsusing co-dominant ecotype-specific PCR-based markers, Plant J. 4 (1993)403–410.

13] S.R. McCouch, K. Zhao, M. Wright, C.W. Tung, K. Ebana, M. Thomson, A.Reynolds, D. Wang, G. DeClearck, M.L. Ali, A. McClung, G. Eizenga, C.Bustamante, Development of genome-wide SNP assays for rice, Breed. Sci. 60(2010) 524–535.

14] C.W. Tung, K. Zhao, M.H. Wright, M.L. Ali, J. Jung, J. Kimball, W. Tyagi, M.J.Thomson, K. McNally, H. Leung, H. Kim, S.N. Ahn, A. Reynolds, B. Scheffler, G.Eizenga, A. McClung, C. Bustamante, S.R. McCouch, Development of a researchplatform for dissecting phenotype-genotype associations in rice (Oryza spp.),Rice 3 (2010) 205–217.

15] X. Huang, X. Wei, T. Sang, Q. Zhao, Q. Feng, Y. Zhao, C. Li, C. Zhu, T. Lu, Z. Zhang,M. Li, D. Fan, Y. Guo, A. Wang, L. Wang, L. Deng, W. Li, Y. Lu, Q. Weng, K. Liu, T.Huang, T. Zhou, Y. Jing, W. Li, Z. Lin, E.S. Buckler, Q. Qian, Q.F. Zhang, J. Li, B.Han, Genome-wide association studies of 14 agronomic traits in ricelandraces, Nat. Genet. 42 (2010) 961–967.

16] F.A. Feltus, J. Wan, S.R. Schulze, J.C. Estill, N. Jiang, A.H. Paterson, A SNPresource for rice genetics and breeding based on subspecies indica andjaponica genome alignments, Genome Res. 14 (2004) 1812–1819.

17] Y.J. Shen, H. Jiang, J.P. Jin, Z.B. Zhang, B. Xi, Y.Y. He, G. Wang, C. Wang, L. Qian,X. Li, Q.B. Yu, H.J. Liu, D.H. Chen, J.H. Gao, H. Huang, T.L. Shi, Z.N. Yang,Development of genome-wide DNA polymorphism database for map-basedcloning of rice genes, Plant Physiol. 135 (2004) 1198–1205.

18] K.L. McNally, K.L. Childs, R. Bohnert, R.M. Davidson, K. Zhao, V.J. Ulat, G. Zeller,R.M. Clark, D.R. Hoen, T.E. Hoen, T.E. Bureau, R. Stokowski, D.G. Ballinger, K.A.Frazer, D.R. Cox, B. Padhukasahasram, C.D. Bustamante, D. Weigel, D.J. Mackill,R.M. Bruskiewich, G. Ratsch, C.R. Buell, H. Leung, J.E. Leach, Genomewide SNPvariation reveals relationships among landraces and modern varieties of rice,Proc. Nat. Acad. Sci. U. S. A. 106 (2009) 12273–12278.

19] The 3000 rice genomes project, The 3000 rice genomes project, GigaScience 3(2014) 7.

20] H. He H. Chen, W. Chen Y. Zou, X. Liu R. Yu, Y.M. Gao Y. Yang, L.M. Fan J.L. Xu,Y. Li, X.W. Deng Z.K. Li, Development and application of a set ofbreeder-friendly SNP markers for genetic analyses and molecular breeding ofrice (Oryza sativa L.), Theor. Appl. Genet. 123 (2011) 869–879.

21] M.J. Thomson, K. Zhao, M. Wright, K.L. McNally, J. Rey, C.W. Tung, A. Reynolds,B. Scheffler, G. Eizenga, A. McClung, H. Kim, A.M. Ismail, M.D. Ocampo, C.Mojica, M.Y. Reveche, C.J. Dilla-Ermita, R. Mauleon, H. Leung, C. Bustamante,S.R. McCouch, High-throughput single nucleotide polymorphism genotypingfor breeding applications in rice using the BeadXpress platform, Mol. Breed.29 (2012) 875–886.

22] H. Chen, W. Xie, H. He, H. Yu, W. Chen, J. Li, R. Yu, Y. Yao, W. Zhang, Y. He, X.Tang, F. Zhou, X. Wang Deng, Q. Zhang, A high-density SNP genotyping arrayfor rice biology and molecular breeding, Mol. Plant 7 (2014) 541–553.

23] H. Yu, W. Xie, J. Li, F. Zhou, Q. Zhang, A whole-genome SNP array (RICE6K) forgenomic breeding in rice, Plant Biotech. J. 12 (2014) 28–37.

24] R. Shen, J.B. Fan, D. Campbell, W. Chang, J. Chen, D. Doucet, J. Yeakley, M.Bibikova, G.E. Wickham, C. McBride, F. Steemers, F. Garcia, B.G. Kermani, K.Gunderson, A. Oliphant, high-throughput SNP genotyping on universal beadarrays, Mutat. Res. 573 (2005) 70–82.

25] K. Zhao, C.W. Tung, G.C. Eizenga, M.H. Wright, M.L. Ali, A.H. Price, G.J. Norton,M.R. Islam, A. Reynolds, J. Mezey, A.M. McClung, Genome-wide associationmapping reveals a rich genetic architecture of complex traits in Oryza sativa,Nat. Commun. 2 (2011) 467.

26] M. Ashikari, H. Sakakibara, S. Lin, T. Yamamoto, T. Takashi, A. Nishimura, E.R.

Angeles, Q. Qian, H. Kitano, M. Matsuoka, Cytokinin oxidase regulates ricegrain production, Science 309 (2005) 741–745.

27] X.J. Song, W. Huang, M. Shi, M.Z. Zhu, H.X. Lin, A QTL for rice grain width andweight encodes a previously unknown RING-type E3 ubiquitin ligase, Nat.Genet. 39 (2007) 623–630.

[

ce 242 (2016) 131–139

28] S. Yazawa, H. Yasui, A. Yoshimura, N. Iwata, RFLP mapping of genes forresistance to green rice leafhopper (Nephotettix cincticeps Uhler) in ricecultivar DV85 using near isogenic lines, Sci. Bull. Fac. Agric. Kyushu Univ. 52(1998) 169–175.

29] C. Fan, Y. Xing, H. Mao, T. Lu, B. Han, C. Xu, X. Li, Q. Zhang, GS3, a Major QTL forgrain length and weight and minor QTL for grain width and thickness in rice,encodes a putative transmembrane protein, Theor. Appl. Genet. 112 (2006)1164–1171.

30] N. Takano-Kai, H. Jiang, T. Kubo, M. Sweeney, T. Matsumoto, H. Kanamori, B.Padhukasahasram, C. Bustamante, A. Yoshimura, K. Doi, S. McCouch,Evolutionary history of GS3, a gene conferring grain length in rice, Genetics182 (2009) 1134–1323.

31] D. Fujita, A. Yoshimura, H. Yasui, Development of near-isogenic lines andpyramided lines carrying resistance genes to green rice leafhopper(Nephotettix cincticeps Uhler) with the Taichung 65 genetic background in rice(Oryza sativa L.), Breeding Sci 60 (2010) 18–27.

32] S. Fukuoka, N. Saka, H. Koga, K. Ono, T. Shimizu, K. Ebana, N. Hayashi, A.Takahashi, H. Hirochika, K. Okuno, M. Yano, Loss of function of aproline-containing protein confers durable disease resistance in rice, Science325 (2009) 998–1001.

33] T. Shomura, K. Izawa, T. Ebana, H. Ebitani, S. Kanegae, M.Y. Konishi, ano,Deletion in a gene associated with grain size increased yields during ricedomestication, Nat. Genet. 40 (2008) 1023–1028.

34] D. Myint, M. Fujita, T. Matsumura, A. Sonoda, H. Yoshimura Yasui, Mappingand pyramiding of two major genes for resistance to the brown planthopper(Nilaparvata lugens [Stål]) in the rice cultivar ADR52, Theor. Appl. Genet. 124(2011) 495–504.

35] M. Yamasaki, A. Yoshimura, H. Yasui, Genetic basis of ovicidal response towhitebacked planthopper (Sogatella furcifera Horváth) in rice (Oryza sativa L.),Mol. Breed. 12 (2003) 133–143.

36] T. Terao, K. Nagata, K. Morino, T. Hirose, A gene controlling the number ofprimary rachis branches also controls the vascular bundle formation andhence is responsible to increase the harvest index and grain yield in rice,Theor. Appl. Genet. 120 (2010) 875–893.

37] B.W. Porter, J.M. Chittoor, M. Yano, T. Sasaki, F.F. White, development andmapping of markers linked to the rice bacterial blight resistance gene Xa7,Crop Sci. 43 (2003) 1484–1492.

38] Y. Zhang, J. Wang, J. Pan, Z. Gu, X. Chen, Y. Jin, F. Liu, H. Zhang, B. Ma,Identification and molecular mapping of the rice bacterial blight resistancegene allelic to Xa7 from an elite restorer line Zhenhui 084, Eur. J. Plant Pathol.125 (2009) 235–244.

39] K. Miura, M. Ikeda, A. Matsubara, X.J. Song, M. Ito, K. Asano, M. Matsuoka, H.Kitano, M. Ashikari, OsSPL14 promotes panicle branching and higher grainproductivity in rice, Nat. Genet. 42 (2010) 545–549.

40] W.Y. Song, G.L. Wang, L.L. Chen, H.S. Kim, L.Y. Pi, T. Holsten, J. Gardner, B.Wang, W.X. Zhai, L.H. Zhu, C. Fauquet, P. Ronald, A receptor kinase-likeprotein encoded by the rice disease resistance gene, Xa21, Science 270 (1995)1804–1806.

41] X. Sun, Z. Yang, S. Wang, Q. Zhang, Identification of a 47-kb DNA fragmentcontaining Xa4, a locus for bacterial blight resistance in rice, Theor. Appl.Genet. 106 (2003) 683–687.

42] M.H. Wright, C.W. Tung, K. Zhao, A. Reynolds, S.R. McCouch, C.D. Bustamante,ALCHEMY: a reliable method for automated SNP genotype calling for smallbatch sizes and highly homozygous populations, Bioinformatics 26 (2010)2952–2960.

43] I. Milne, P. Shaw, G. Stephen, M. Bayer, L. Cardle, W.T.B. Thomas, A.J. Flavell, D.Marshall, Flapjack— graphical genotype visualization, Bioinformatics 26(2010) 3133–3134.

44] S.R. McCouch, L. Teytelman, Y. Xu, K.B. Lobos, K. Clare, M. Walton, B. Fu, R.Maghirang, Z. Li, Q. Zhang, I. Kono, M. Yano, R. Fjellstrom, G. DeClerck, D.Schneider, S. Cartinhour, D. Ware, L. Stein, Development and mapping of 2240new SSR markers for rice (Oryza sativa L.), DNA Res. 9 (2002) 199–207.

45] K.A. Mather, A.L. Caicedo, N.R. Polato, K.M. Olsen, S.R. McCouch, M.D.Purugganan, The extent of linkage disequilibrium in rice (Oryza sativa L.),Genetics 177 (2007) 2223–2232.

46] The International Rice Genome Sequencing Project, The map-based sequenceof the rice genome, Nature 436 (2005) 793-800.

47] M.A. Causse, T.M. Fulton, Y.G. Cho, S.N. Ahn, J. Chunwongse, K. Wu, J. Xiao, Z.Yu, P.C. Ronald, S.E. Harrington, Saturated molecular map of the rice genomebased on an interspecific backcross population, Genetics 138 (1994)1251–1274.

48] Y. Harushima, M. Yano, A. Shomura, M. Sato, T. Shimano, Y. Kuboki, T.Yamamoto, S.Y. Lin, B.A. Antonio, A. Parco, H. Kajiya, N. Huang, K. Yamamoto,Y. Nagamura, N. Kurata, G.S. Khush, T. Sasaki, A high-density rice geneticlinkage map with 2275 Markers Using a single F2 population, Genetics 148(1998) 479–494.

49] N. Huang, E.R. Angeles, J. Domingo, G. Magpantay, S. Singh, G. Zhang, N.Kumaravadivel, J. Bennett, G.S. Khush, Pyramiding of bacterial blightresistance genes in rice: marker-assisted selection using RFLP and PCR, Theor.Appl. Genet 95 (1997) 313–320.

50] P.X. Kover, W. Valdar, J. Trakalo, N. Scarcelli, I.M. Ehrenreich, M.D.

Purugganan, C. Durrant, R. Mott, A multiparent advanced generationinter-cross to fine-map quantitative traits in Arabidopsis thaliana, PLoSGenet. 5 (2009) e1000551.

51] N. Bandillo, C. Raghavan, P.A. Muyco, M.A.L. Sevilla, I.T. Lobina, C.J.D. Ermita,C.W. Tung, S. McCouch, M. Thomson, R. Mauleon, R.K. Singh, G. Gregorio, E.

t Scien

[

[

[

[

Y. Kurokawa et al. / Plan

Redona, H. Leung, Multi-parent advanced generation inter-cross (MAGIC)populations in rice: progress and potential for genetics research and breeding,Rice 6 (2013) 11.

52] J.W. Davey, P.A. Hohenlohe, P.D. Etter, J.Q. Boone, J.M. Catchen, M.L. Blaxter,Genome-wide genetic marker discovery and genotyping using

next-generation sequencing, Nat. Rev. Genet. 12 (2011)499–510.

53] J. Spindel, M. Wright, C. Chen, J. Cobb, J. Gage, S. Harrington, M. Lorieux, N.Ahmadi, S. McCouch, Bridging the genotyping gap: using genotyping bysequencing (GBS) to add high-density SNP markers and new value to

ce 242 (2016) 131–139 139

traditional bi-parental mapping and breeding populations, Theor. Appl.Genet. 126 (2013) 2699–2716.

54] G.S. Khush, Green revolution: the way forward, Nat. Rev, Genet. 2 (2001)815–822.

55] T. Takai, S. Adachi, F. Taguchi-Shiobara, Y. Sanoh-Arai, N. Iwasawa, S.

Yoshinaga, S. Hirose, Y. Taniguchi, U. Yamanouchi, J. Wu, T. Matsumoto, K.sugimoto, K. Kondo, T. Ikka, T. Ando, I. Kono, S. Ito, A. Shomura, T. Ookawa, T.Hirasawa, A natural variant of NAL1, selected in high-yield rice breedingprograms, pleiotropically increases photosynthesis rate, Sci. Rep. 3 (2013),2149.