Tree Genetics & Genomes

, 14:24 | Cite as

Genome-wide association mapping of fruit-quality traits using genotyping-by-sequencing approach in citrus landraces, modern cultivars, and breeding lines in Japan

  • A. Imai
  • K. Nonaka
  • T. Kuniga
  • T. Yoshioka
  • T. Hayashi
Original Article
  • 116 Downloads
Part of the following topical collections:
  1. Complex Traits

Abstract

Association mapping is an attractive method to identify QTLs in perennial horticultural crops such as citrus, as it does not need a designed cross between parental genotypes and can save time and labor to construct a segregating population. It usually requires more genetic markers than linkage-based QTL mapping owing to a lower degree of linkage disequilibrium (LD). However, recent advances in next-generation sequencing offer high-throughput, cost-effective methods, including genotyping-by-sequencing (GBS), for genotyping massive amounts of single nucleotide polymorphisms (SNPs). In this study, we performed a genome-wide association study (GWAS) of fruit-quality traits in citrus using SNPs obtained by GBS. We evaluated 110 citrus accessions, including landraces, modern cultivars, and breeding lines, for eight fruit-quality traits (fruit weight, fruit skin color, fruit surface texture, peelability, pulp firmness, segment firmness, sugar content, and acid content) during 2005 to 2012 (except 2007). GBS found 2309 SNPs, which we anchored to the clementine reference genome. We evaluated LD in the 110 accessions and confirmed that GBS gave enough SNPs to conduct GWAS. We identified seven QTLs, including four novel ones, comprising four significant QTLs for fruit weight and one QTL each for fruit skin color, pulp firmness, and segment firmness. These QTLs offer promise for use in citrus crossbreeding.

Keywords

Citrus breeding Fruit quality Single nucleotide polymorphism (SNP) Marker-assisted selection (MAS) 

Introduction

Citrus (Citrus spp.) is commercially grown in over 100 countries between approximately 40° north and south of the equator, with global production exceeding 136 million tonnes in 2013 (FAO 2016 The most common citrus species is sweet orange (Citrus sinensis (L.) Osbeck), including many cultivars derived from mutations selected over decades (Hearn et al. 2014). Mandarins, with wider variations in several agronomically important characters than sweet orange, are another widely grown citrus species, and new cultivars are developed mainly as hybrids between mandarins or between mandarins and other species such as sweet orange, grapefruit (Citrus paradisi Macfad.), and modern hybrid cultivars and breeding lines (Deng and Xu 2011).

Target traits for mandarin crossbreeding range widely and include tree performance, disease resistance, fruit characteristics, and postharvest quality (Khan 2007). However, the improvement of fruit-quality traits is one of the most important objectives for mandarin breeding (Omura and Shimada 2016). Mandarin fruits are usually consumed fresh, and therefore, the improvement of both external and internal fruit qualities is important (Deng and Xu 2011). Especially, high sugar content, ease of peeling, seedlessness, pulp softness, and segment softness have been the main breeding targets since the 1970s in Japan (Imai et al. 2017a).

Although fruit quality is an important breeding target, many agronomically important fruit-quality traits are controlled by multiple genes, and F1 progeny are commonly highly variable in citrus (Furr 1969; Soost 1987). To support citrus breeding programs including mandarin breeding, several genetic mapping studies of fruit-quality traits had been conducted (reviewed by Gmitter et al. 2007). Recently, we have conducted QTL analysis in a ‘Harehime’ × ‘Yoshida’ ponkan mapping population and identified nine QTLs associated with fruit weight, sugar content, peel puffing, and water rot (Imai et al. 2017b). However, QTL analysis in an F1 population cannot detect QTLs at which both parents are homozygous and no segregation occurs in F1 generation. Therefore, it would be desirable to explore the large genetic diversity in breeding materials to detect as many important QTLs as possible.

In contrast to QTL analysis in F1 populations, genome-wide association study (GWAS) can explore QTLs in more diverse germplasms, where construction of a designed segregating population is not required leading to the save of time and labor for cultivating F1 seedlings. It therefore offers advantages in fruit crop breeding programs, because breeding materials are usually derived from multiple founders (e.g., Imai et al. 2017a; Kunihisa et al. 2014).

Despite its advantages in fruit crops, GWAS relies on linkage disequilibrium (LD) between genetic markers and genes behind target traits, and generally requires enough genetic markers to capture the effects of relevant genes through LD between markers and the genes in target populations. Consequently, the uses of GWAS in fruit crops have so far been limited (e.g., Kumar et al. 2013; Moriya et al. 2017). However, recent advances in next-generation sequencing offer high-throughput, cost-effective genotyping methods, including genotyping-by-sequencing (GBS) (Elshire et al. 2011). GBS can provide a large enough number of single nucleotide polymorphisms (SNPs) to allow GWAS in many plant species (He et al. 2014), including citrus, with reduced genotyping costs.

Therefore, our aims were (1) to confirm whether GBS is practical for GWAS in citrus and (2) to identify markers significantly associated with fruit-quality traits that are useful for marker-assisted selection (MAS) in Japanese citrus breeding with GWAS using SNPs obtained by GBS.

Materials and methods

Plant materials and trait evaluation

We used 110 citrus accessions, composed of landraces, modern cultivars, and breeding lines (Table 1 and Fig. 1). Pedigree chart of these accessions were drawn using Pedimap software (Voorrips et al. 2012). Accessions were grafted onto trifoliate orange (Poncirus trifoliata L.) or satsuma mandarin (Citrus unshiu Marcow.) interstocks grafted onto trifoliate orange, and their trees were grown in the experimental orchard of the Kuchinotsu Citrus Research Station, NARO (Nagasaki, Japan). All trees were maintained in accordance with the standard management protocol in Japan, namely, four applications of fertilizer and 10 to 20 applications of agrichemicals a year.
Table 1

Citrus landraces, modern cultivars, and breeding lines used in this study

No.

Name

Species or parentage

Accessiona

Rootstockb

1

satsuma mandarin

Citrus unshiu Marcow.

‘Okitsu wase’ (170630)

trifoliate orange

2

clementine

Citrus clementina hort. ex Tanaka

Stock strain (113161)

trifoliate orange

3

sweet orange

Citrus sinensis (L.) Osbeck

‘Trovita’ (172154)

trifoliate orange

4

iyo

Citrus iyo hort. ex Tanaka

‘Miyauchi’ (117374)

trifoliate orange

5

hyuga-natsu

Citrus tamurana hort. ex Tanaka

Stock strain (117317)

trifoliate orange

6

grapefruit

Citrus paradisi Macfad.

‘Duncan’ (168864)

7

dancy tangerine

Citrus tangerine hort. ex Tanaka

Stock strain (117396)

trifoliate orange

8

king mandarin

Citrus nobilis Lour.

Stock strain (117386)

trifoliate orange

9

mediterranean mandarin

Citrus deliciosa Ten.

Stock strain (117393)

trifoliate orange

10

ponkan

Citrus reticulata Blanco

‘Yoshida’ (113178)

trifoliate orange

11

kishu mandarin

Citrus kinokuni hort. ex Tanaka

‘Mukaku-kishu’ (171490)

trifoliate orange

12

hassaku

Citrus hassaku hort. ex Tanaka

Stock strain (115524)

trifoliate orange

13

Tanikawa-buntan

Unknown

Stock strain (117433)

14

kimikan

Citrus flavicarpa hort. ex Tanaka

Stock strain (172148)

trifoliate orange

15

Kawachi-bankan

Unknown

Stock strain (117412)

trifoliate orange

16

San-jacinto

Unknown

Stock strain (113350)

trifoliate orange

17

Soren-tangelo

Unknown

Stock strain (113347)

Trifoliate orange

18

tankan

Citrus tankan Hayata

‘Tarumizu 1 Gou’ (113508)

satsuma mandarin interstocks

19

Haruka

Unknown

Stock strain (237315)

satsuma mandarin interstocks

20

yuge-hyokan

Citrus yuge-hyokan hort. ex Yu. Tanaka

Stock strain (117507)

trifoliate orange

21

Nankou

Miho wase × clementine

Stock strain (117160)

satsuma mandarin interstocks

22

HF15

Hayashi unshiu × Fukuhara sweet orange

Stock strain (n.a.)

trifoliate orange

23

HF17

Hayashi unshiu × Fukuhara sweet orange

Stock strain (n.a.)

satsuma mandarin interstocks

24

HF20

Hayashi unshiu × Fukuhara sweet orange

Stock strain (n.a.)

trifoliate orange

25

HF21

Hayashi unshiu × Fukuhara sweet orange

Stock strain (n.a.)

trifoliate orange

26

HF9

Hayashi unshiu × Fukuhara sweet orange

Stock strain (n.a.)

trifoliate orange

27

Aki Tangor

Okitsu wase × Trovita

Stock strain (n.a.)

satsuma mandarin interstocks

28

Kiyomi

Miyagawa wase × Trovita

Stock strain (115521)

satsuma mandarin interstocks

29

Minneola

Duncan grapefruit × dancy tangerine

Stock strain (113377)

30

Orlando

Duncan grapefruit × dancy tangerine

Stock strain (113327)

trifoliate orange

31

Encore

king × mediterranean mandarin

Stock strain (117421)

trifoliate orange

32

Wilking

king × mediterranean mandarin

Stock strain (117425)

trifoliate orange

33

Ariake

Seike navel orange × clementine

Stock strain (117158)

satsuma mandarin interstocks

34

Sweet Spring

Ueda unshiu × hassaku

Stock strain (168866)

trifoliate orange

35

No. 1087

Miyauchi iyo × Trovita

Stock strain (n.a.)

36

Awa orange

hyuga-natsu × Trovita

Stock strain (115757)

37

JHG

Jutarou unshiu × hyuga-natsu

Stock strain (n.a.)

satsuma mandarin interstocks

38

Kara

Owari unshiu × king

Stock strain (113158)

satsuma mandarin interstocks

39

Hayaka

Imamura unshiu × Nakano 3 Gou ponkan

Stock strain (117154)

satsuma mandarin interstocks

40

Kankitsu Chukanbohon Nou 6 Gou

king × Mukaku-kishu

Stock strain (223661)

satsuma mandarin interstocks

41

Southern Yellow

Tanikawa-buntan × Mukaku-kishu

Stock strain (117470)

satsuma mandarin interstocks

42

Nishinokaori

Kiyomi × Trovita

Stock strain (118841)

satsuma mandarin interstocks

43

Osceola

clementine × Orlando

Stock strain (113329)

trifoliate orange

44

Lee

clementine × Orlando

Stock strain (113385)

trifoliate orange

45

Robinson

clementine × Orlando

Stock strain (113386)

trifoliate orange

46

KyOw14

Kiyomi × Okitsu wase

Stock strain (n.a.)

trifoliate orange

47

KyOw21

Kiyomi × Okitsu wase

Stock strain (n.a.)

trifoliate orange

48

EnOw20

Encore × Okitsu wase

Stock strain (n.a.)

trifoliate orange

49

EnOw21

Encore × Okitsu wase

Stock strain (n.a.)

trifoliate orange

50

EnOw7

Encore × Okitsu wase

Stock strain (n.a.)

trifoliate orange

51

EnOw8

Encore × Okitsu wase

Stock strain (n.a.)

trifoliate orange

52

Shiranuhi

Kiyomi × Nakano 3 Gou ponkan

Stock strain (117159)

satsuma mandarin interstocks

53

Setomi

Kiyomi × Yoshiura ponkan

Stock strain (223670)

satsuma mandarin interstocks

54

Harumi

Kiyomi × F2432 ponkan

Stock strain (117468)

satsuma mandarin interstocks

55

Youkou

Kiyomi × Nakano 3 Gou ponkan

Stock strain (n.a.)

satsuma mandarin interstocks

56

Page

clementine × Minneola

Stock strain (113370)

trifoliate orange

57

No. 2681

Kiyomi × Miyauchi iyo

Stock strain (n.a.)

trifoliate orange

58

Tsunonozomi

Kiyomi × Encore

Stock strain (n.a.)

satsuma mandarin interstocks

59

KyEn4

Kiyomi × Encore

Stock strain (n.a.)

trifoliate orange

60

Amaka

Kiyomi × Encore

Stock strain (118843)

satsuma mandarin interstocks

61

Tamami

Kiyomi × Wilking

Stock strain (223657)

satsuma mandarin interstocks

62

F118

Kiyomi × Wilking

Stock strain (n.a.)

63

Okitsu 45 Gou

Kiyomi × Wilking

Stock strain (n.a.)

satsuma mandarin interstocks

64

Benibae

HF9 × Encore

Stock strain (n.a.)

satsuma mandarin interstocks

65

HF9·En29

HF9 × Encore

Stock strain (n.a.)

trifoliate orange

66

A7

Sweet Spring × Trovita

Stock strain (n.a.)

67

Okitsu 46 Gou

Sweet Spring × Trovita

Stock strain (n.a.)

satsuma mandarin interstocks

68

Murcott

Unknown

Stock strain (113374)

trifoliate orange

69

Hareyaka

Encore × Nakano 3 Gou ponkan

Stock strain (117510)

satsuma mandarin interstocks

70

2700·OIy25

Nishinokaori × Ohtani iyo

Stock strain (n.a.)

trifoliate orange

71

E647

Kiyomi × Osceola

Stock strain (n.a.)

satsuma mandarin interstocks

72

KyOw21·D4

KyOw21 × dancy tangerine

Stock strain (n.a.)

trifoliate orange

73

KyOw21·D49

KyOw21 × dancy tangerine

Stock strain (n.a.)

trifoliate orange

74

Kuchinotsu 28 Gou

KyOw21 × dancy tangerine

Stock strain (n.a.)

trifoliate orange

75

Kuchinotsu 27 Gou

EnOw21 × Youkou

Stock strain (n.a.)

satsuma mandarin interstocks

76

Seinannohikari

EnOw21 × Youkou

Stock strain (n.a.)

satsuma mandarin interstocks

77

Amakusa

KyOw14 × Page

Stock strain (117161)

satsuma mandarin interstocks

78

Kuchinotsu 33 Gou

KyOw14 × Encore

Stock strain (n.a.)

satsuma mandarin interstocks

79

Tsunokagayaki

KyOw14 × Encore

Stock strain (n.a.)

satsuma mandarin interstocks

80

Kuchinotsu 18 Gou

KyOw21 × Encore

Stock strain (n.a.)

trifoliate orange

81

Kuchinotsu 35 Gou

KyOw21 × Encore

Stock strain (n.a.)

satsuma mandarin interstocks

82

Kuchinotsu 38 Gou

KyOw21 × Robinson

Stock strain (n.a.)

trifoliate orange

83

No. 1408

EnOw21 × No. 2681

Stock strain (n.a.)

trifoliate orange

84

Kankitsu Chukanbohon Nou 5 Gou

Lee × Mukaku-kishu

Stock strain (223660)

trifoliate orange

85

Kanpei

Nishinokaori × Shiranuhi

Stock strain (n.a.)

satsuma mandarin interstocks

86

Okitsu 57 Gou

Okitsu 46 Gou × Harumi

Stock strain (n.a.)

satsuma mandarin interstocks

87

Asumi

Okitsu 46 Gou × Harumi

Stock strain (n.a.)

satsuma mandarin interstocks

88

Okitsu 59 Gou

A7 × F118

Stock strain (n.a.)

satsuma mandarin interstocks

89

Setoka

Tsunonozomi × Murcott

Stock strain (118842)

satsuma mandarin interstocks

90

KyOw21·CC33

KyOw21 × Clementine Caffin

Stock strain (n.a.)

trifoliate orange

91

KyOw21·Ariake22

KyOw21 × Ariake

Stock strain (n.a.)

trifoliate orange

92

Kuchinotsu 40 gou

KyOw21 × No. 1087

Stock strain (n.a.)

satsuma mandarin interstocks

93

Haruhi

Okitsu 46 Gou × Awa orange

Stock strain (n.a.)

satsuma mandarin interstocks

94

LeeAo25

Lee × Aoshima unshiu

Stock strain (n.a.)

trifoliate orange

95

LeeAo35

Lee × Aoshima unshiu

Stock strain (n.a.)

trifoliate orange

96

LeeAo9

Lee × Aoshima unshiu

Stock strain (n.a.)

trifoliate orange

97

No. 1010

Nankou × 2700·OIy25

Stock strain (n.a.)

satsuma mandarin interstocks

98

No. 1011

Nankou × 2700·OIy25

Stock strain (n.a.)

trifoliate orange

99

Harehime

E647 × Miyagawa wase

Stock strain (n.a.)

satsuma mandarin interstocks

100

Kuchinotsu 51 Gou

KyOw21·D4 × Kuchinotsu 27 Gou

Stock strain (n.a.)

satsuma mandarin interstocks

101

No. 1051

HF9 × 2700·OIy25

Stock strain (n.a.)

trifoliate orange

102

Ehime Kashi No. 28

Nankou × Amakusa

Stock strain (n.a.)

satsuma mandarin interstocks

103

Kuchinotsu 52 Gou

Tsunokagayaki × Ariake

Stock strain (n.a.)

satsuma mandarin interstocks

104

960203

Kuchinotsu 18 Gou × Youkou

Stock strain (n.a.)

satsuma mandarin interstocks

105

Kuchinotsu 49 Gou

Kuchinotsu 38 Gou × No. 1408

Stock strain (n.a.)

satsuma mandarin interstocks

106

Mihaya

Tsunonozomi × No. 1408

Stock strain (n.a.)

satsuma mandarin interstocks

107

980389

HF15 × No. 1408

Stock strain (n.a.)

satsuma mandarin interstocks

108

Okitsu 56 Gou

Okitsu 45 Gou × Kankitsu Chukanbohon Nou 5 Gou

Stock strain (n.a.)

satsuma mandarin interstocks

109

040381

No. 1011 × hyuga-natsu

Stock strain (n.a.)

trifoliate orange

110

031045

Harehime × Clementine Caffin

Stock strain (n.a.)

satsuma mandarin interstocks

Accessions with “n.a.” have not been registered in the collection

Accessions with “–” were not used for phenotypic evaluations

aNumbers in parentheses represent the accession ID (JP number) in the NIAS germplasm collection

bAccessions were grafted onto trifoliate orange (Poncirus trifoliata L.) or satsuma mandarin (Citrus unshiu Marcow.) interstocks grafted onto trifoliate orange

Fig. 1

Pedigree chart of citrus landraces, modern cultivars, and breeding lines used in this study. Pedigrees generations run from left to right. Red lines represent seed parents and blue lines represent pollen parents. Cross symbols connect parents to offspring

Five fruits were randomly collected for immediate trait evaluation from a tree of each accession in mid-December of 2005, 2006, and 2008–2012. We considered colored fruits to be mature, and evaluated eight traits (Table 2). Fruit weight (FW), sugar content (SC), and acid content (AC) were instrumentally measured, and the averages of the measurements taken over the five fruits were adopted as the phenotypes of these traits for each genotype. Fruit skin color (FSC), fruit surface texture (FST), peelability (PE), pulp firmness (PF), and segment firmness (SF) were visually or sensorily measured on an ordinal scale from 1 to 5. This scale was readjusted to range from 1 to 3 (Table 2) to avoid different ratings by different evaluators. Alternate bearing caused variations in the number of accessions for phenotypic evaluations from year to year, but most accessions were evaluated on all traits at least once during 2005–2012.
Table 2

Eight fruit-quality traits evaluated in this study

Trait

Abbreviations

Data type

Measurement unit

Fruit weight

FW

Continuous

Mean weight of mature fruits (g)

Fruit skin color

FSC

Ordinal

1:yellow, 2:orange, 3:red

Fruit surface texture

FST

Ordinal

1:rough, 2:medium, 3:smooth

Peelability

PE

Ordinal

1:difficult, 2:medium, 3:easy

Pulp firmness

PF

Ordinal

1:firm, 2:medium, 3:soft

Segment firmness

SF

Ordinal

1:firm, 2:medium, 3:soft

Sugar content

SC

Continuous

Mean Brix of juice (Brix%)

Acid content

AC

Continuous

Mean citric acid concentration of juice (%)

SNP genotyping and quality control of SNP data

Genomic DNA of each accession was isolated from young leaves with a DNeasy mini kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. SNP genotyping was conducted following the GBS protocol of Elshire et al. (2011) in combination with other samples unrelated with this experimental population. The 96-plex GBS libraries were constructed at Genomic Diversity Facility, Cornell University (Ithaca, NY, USA) using the barcode adapter set. Genomic DNA was digested with the restriction enzyme PstI (CTGCAG), and barcoded adapters were ligated to each sample. Samples were pooled depending on the 96-plex level into a single library and amplified by PCR. Each library was sequenced in reads of 100 bp on a single lane of an Illumina HiSeq 2000 sequencer (San Diego, CA, USA). SNP calling was performed with the Tassel-GBS pipeline (Glaubitz et al. 2014). Trimmed and cleaned sequence tags were aligned to the clementine reference genome v. 1.0 ( http://www.phytozome.net ) (Ollitrault et al. 2012; Wu et al. 2014) using Bowtie2 software (Langmead et al. 2009). The minor allele frequency (MAF) cutoff was set at 0.01. Raw SNPs were called, and the resulting files were input into TASSEL software (Bradbury et al. 2007) for further filtering. SNPs missing > 20% of data were removed.

The SNPs obtained by GBS were subjected to quality control. Firstly, Mendelian errors of SNP genotypes in trios consisting of parental pairs and their F1 genotypes were detected in Merlin software (Abecasis et al. 2002), where inconsistency in the inheritance of alleles between parents and offspring was investigated (Fig. 2). We assumed that the trios with > 10% Mendelian errors had pedigree errors and corrected their parentages according to Ninomiya et al. (2015) or the records of field notes. Remaining Mendelian errors following correction of parentages were assumed to be genotyping errors of the GBS approach. These errors were replaced by missing genotypes. Then, missing genotypes were imputed in Fimpute v. 2.2 software (Sargolzaei et al. 2014). Finally, all pairs of SNP loci that were all the same or all opposite were eliminated according to the Wiggans et al. (2009), along with SNP loci with MAF ≤ 0.05. All these procedures for SNP genotyping and their quality control were summarized in Fig. 3.
Fig. 2

Detection of Mendelian errors using Merlin software and their classification. Mendelian errors were classified as parental errors or genotyping errors. If a trio (i.e., parental pair and their F1 genotype) had > 10% Mendelian errors, we considered that the parentage was incorrect. Otherwise, all Mendelian errors were assumed to be genotyping errors of the GBS approach

Fig. 3

Procedures of genotyping-by-sequencing (GBS) analysis and quality control of single nucleotide polymorphism (SNP) data conducted in this study. MAF denotes minor allele frequency. High correlation with other SNP means that SNP genotypes were all the same or all opposite in all samples

Population structure and kinship matrix

To avoid spurious associations, we estimated both the population structure and kinship relatedness among the 110 citrus accessions and incorporated these estimates into the mixed model for GWAS (Yu et al. 2006) described below. The population structure was estimated by Bayesian model-based clustering in STRUCTURE software (Pritchard et al. 2000) with a subset of 270 SNPs (i.e., 30 SNPs per linkage group) obtained by GBS and selected to be at least 500 kbp apart. To calculate the most likely number of subpopulations (K), we conducted 10 independent runs of the admixture models ranging from K = 1 to 10, by Markov chain Monte Carlo (MCMC) simulation using 100,000 burn-in cycles followed by 1,000,000 run cycles. The optimal K was determined by the value of Delta K proposed by Evanno et al. (2005). The kinship relatedness was estimated from an additive genetic relationship matrix calculated from the pedigree information of the 110 citrus accessions using the R package “nadiv” (Wolak 2012).

Linkage disequilibrium (LD)

The physical positions of the SNPs obtained by GBS were determined on the clementine reference genome and were used to calculate the extent of the genome-wide LD and the LD decay pattern for nine linkage groups (LGs corresponding to the nine scaffolds comprising chromosome pseudo-molecules of the clementine reference genome) in the 110 accessions. The LD between pairs of SNPs in the same LG was evaluated from pairwise r2 values in Plink software (Purcell et al. 2007). Pairwise LD was plotted against physical distance as a two-dimensional histogram in the “plot3D” R package (Soetaert 2013). The LD decay pattern was estimated by fitting a trend line based on a nonlinear regression of r2 on physical distance using “smooth.spline” R function. We calculated the 95th percentile of the r2 distribution between pairs of SNPs in the different LGs to determine the threshold of r2. Two SNPs on the same chromosome were considered in a state of significant LD when r2 between them exceeded the threshold.

Genome-wide association study (GWAS)

We predicted the genotypic values of FW, SC, and AC showing continuous variations by using the following linear mixed model:
$$ {Y}_{ij}=\mu +{g}_i+{b}_j+{e}_{ij}, $$
where Y ij is the observation of the ith genotype in the jth year; μ is the intercept; g i is the random effect of the ith genotype; b j is the fixed effect of the jth year; e ij is the residual associated with each observation. We assumed that g i are independently distributed as \( N\left(0,{\sigma}_g^2\right) \) for all i, and e ij are independently distributed as \( N\left(0,{\sigma}_e^2\right) \) for all i and j, where \( {\sigma}_g^2 \) is the variance of the genotypic effect and \( {\sigma}_e^2 \) is the residual variance.
We predicted the genotypic values of FSC, FST, PE, PF, and SF which were measured on ordinal scales by using the following generalized linear mixed model:
$$ {L}_{ij}=\mu +{g}_i+{b}_j+{e}_{ij} $$
with
$$ {Y}_{ij}=\left\{\begin{array}{c}1\left({L}_{ij}\le {t}_1\right)\\ {}2\left({t}_1<{L}_{ij}\le {t}_2\right)\\ {}3\left({L}_{ij}>{t}_2\right)\end{array}\right., $$
where L ij is the latent continuous variable underlining the observed scale, referred to as liability, and t1 and t2 are the cut-points relating liability with phenotypes. The other model parameters, μ, g i , and b j are those as described above in the continuous traits. We used the probit link function to associate L ij with Y ij under the assumption of the residual e ∼ N(0, 1) (Gianola and Foulley 1983; Harville and Mee 1984).

For all traits, we tested the statistical significance of the year effect b j by using the Wald statistic (P < 0.05) and applied a best linear unbiased prediction method to obtain the predictions of g i , denoted as \( {\widehat{g}}_i \), for all individuals.

As a second step of GWAS, we analyzed the association between the predicted genotypic values and a SNP, by using the following linear mixed model:
$$ {\widehat{g}}_i=c+{\mathbf{x}}_{il}^{\prime }{\mathbf{s}}_l+\sum \limits_{k=1}^K{\nu}_k{q}_{ik}+{u}_i+{\varepsilon}_i $$
where c is the intercept of the model; \( {\mathbf{x}}_{il}^{\prime } \) is a vector indicating the genotype at the lth SNP for the ith individual with \( {\mathbf{x}}_{il}^{\prime } \) = (0,0), (1,0), and (0,1) for three SNP genotypes, i.e., homozygote with a reference allele of the clementine reference genome, homozygote with an alternative allele and heterozygote, respectively; s l  = (α l , δ l ) is a vector of regression coefficients for \( {\mathbf{x}}_{il}^{\prime } \); v k is the population effect associated with subpopulation K; q ik is the corresponding element of a Q matrix of the ith genotype in subpopulation K; u i represents the polygenic effects of the ith genotype, distributed as \( N\left(0,\mathbf{A}{\sigma}_u^2\right) \), where A is the additive genetic relationship matrix described above and \( {\sigma}_u^2 \) is polygenic variance; and ε i is the residual associated with \( {\widehat{g}}_i \). The SNP effect and the corresponding Wald P value were calculated in ASReml software (Gilmour et al. 2015). The significance threshold of the SNP effect was determined from Bonferroni’s corrected P value (P < 0.05) adjusted with the number of SNPs tested. All significant SNPs located within the LD threshold determined by the above procedure were considered to belong to the same QTL region.

To illustrate the localization of the significant SNPs, we created Manhattan plots of all tested SNPs for each trait, and to evaluate the absence of confounders in the GWAS, we created quantile–quantile (QQ) plots for each trait. Manhattan plots and QQ plots were drawn in the “qqman” R package (Turner 2014).

We calculated the genetic variance explained by each significant SNP (i.e., variance of SNP effect associated with QTL: \( {\sigma}_{QTL}^2 \)) as the sum of additive QTL variance (\( {\sigma}_{QTL\_ add}^2 \)) and dominant QTL variance (\( {\sigma}_{QTL\_\mathit{\operatorname{dom}}}^2 \)), which were defined for a significant SNP, say the lth SNP, as follows:
$$ {\displaystyle \begin{array}{l}{a}_l={\alpha}_l/2\\ {}{d}_l={\delta}_l\\ {}{\sigma}_{\mathrm{SNP}\_\mathrm{add}}^2=2{p}_l{q}_l{\left\{{a}_l+{d}_l\left({q}_l-{p}_l\right)\right\}}^2\\ {}{\sigma}_{\mathrm{SNP}\_\operatorname{dom}}^2={\left(2{p}_l{q}_l{d}_l\right)}^2\\ {}{\sigma}_{\mathrm{SNP}}^2={\sigma}_{\mathrm{SNP}\_\mathrm{add}}^2+{\sigma}_{\mathrm{SNP}\_\operatorname{dom}}^2\\ {}{Q}_p={\sigma}_{\mathrm{SNP}}^2/\left({\sigma}_g^2+{\sigma}_e^2\right),\end{array}} $$
where a l and d l are the additive and dominant effects calculated from the significant SNP genotype effects; p l and q l are the frequencies of reference and alternative alleles, respectively; and Q p is the percentage of phenotypic variance explained by a SNP. When there were < 5 homozygous SNP genotypes, we assumed that the SNP had only additive effect without dominance effect and reconfirmed the significance of the corresponding SNP effect.

Results

Phenotypic characterization and prediction of genotypic values

Each year, we evaluated 70 to 99 accessions for each trait (Table 3). Only FW, SC, and AC were evaluated in 2005, and PF and SF were not evaluated in 2006. Most accessions were evaluated at least once for each trait, but seven accessions (‘Duncan’ grapefruit, ‘Tanikawa-buntan,’ ‘Minneola,’ ‘No. 1087,’ ‘Awa Orange,’ ‘F118,’ and ‘A7’) were never evaluated for any traits owing to poor growth or unexpected death. These accessions without phenotypic records were only used in the estimation of population structure and the additive genetic relationship matrix, which were considered in the GWAS model. For the other accessions, the Wald test showed a significant effect of year in all traits (P < 0.05), and the predictions of genotypic values adjusted for year effect were used in the subsequent GWAS.
Table 3

Summary statistics of the eight fruit-quality traits evaluated in this study

Year

Descriptive traits

FW (g)

FSC

FST

PE

PF

SF

SC (Brix%)

AC (%)

2005

Mean

187.4

10.5

1.28

S.D.

100.9

1.2

0.64

Min

23.1

8.4

0.59

Max

714.8

13.5

3.32

Records

74

0

0

0

0

0

73

74

2006

Mean

182.8

2.00

2.16

2.39

11.9

1.40

S.D.

85.1

0.47

0.62

0.73

1.3

0.55

Min

29.7

1

1

1

8.3

0.57

Max

626.0

3

3

3

15.8

3.37

Records

75

72

75

75

0

0

75

75

2008

Mean

192.0

2.04

2.50

2.32

2.78

2.50

11.0

1.27

S.D.

91.1

0.53

0.61

0.77

0.47

0.72

1.1

0.55

Min

30.6

1

1

1

1

1

8.2

0.56

Max

621.5

3

3

3

3

3

13.7

3.00

Records

82

82

82

82

82

82

82

82

2009

Mean

241.5

1.94

2.62

2.24

2.73

2.39

10.5

1.37

S.D.

121.0

0.48

0.62

0.82

0.56

0.69

1.2

0.60

Min

38.2

1

1

1

1

1

7.4

0.93

Max

949.5

3

3

3

3

3

13.2

3.52

Records

71

70

71

71

71

71

71

71

2010

Mean

184.3

2.03

2.39

2.15

2.82

2.30

11.6

1.45

S.D.

76.5

0.46

0.64

0.83

0.41

0.60

1.4

0.64

Min

32.8

1

1

1

1

1

8.7

0.61

Max

541.0

3

3

3

3

3

14.6

3.49

Records

99

99

99

99

99

99

98

99

2011

Mean

214.3

2.10

2.24

2.18

2.84

2.51

10.6

1.32

S.D.

96.8

0.54

0.52

0.78

0.42

0.60

1.1

0.53

Min

34.6

1

1

1

1

1

8.0

0.59

Max

746.1

3

3

3

3

3

13.2

3.16

Records

91

91

91

91

90

90

89

90

2012

Mean

180.2

2.09

2.27

2.23

2.74

2.49

10.5

1.50

S.D.

71.9

0.50

0.51

0.70

0.51

0.65

1.1

0.66

Min

29.8

1

1

1

1

1

7.8

0.67

Max

552.0

3

3

3

3

3

13.5

3.82

Records

96

96

96

96

96

96

95

96

FW fruit weight, FSC fruit skin color, FST fruit surface texture, PE peelability, PF pulp firmness, SF segment firmness, SC sugar content, AC acid content

SNP data obtained by genotyping-by sequencing approach

Sequencing generated about 2,000,000 reads per sample at the 96-plex level. The Tassel-GBS pipeline initially identified 156,275 SNP loci with an average allelic read depth of 24.1. Further filtering (SNP call rate ≥ 0.8; MAF ≥ 0.01) returned 5134 SNP loci distributed across nine major scaffolds of the clementine reference genome.

Further quality control of the SNP data of the 5134 loci detected Mendelian errors in all trios, most at a rate of 2 to 6% of Mendelian errors, although error rates attained ≥ 10% in some trios (data not shown). After correcting the parentages of the latter group, we considered remaining errors to be genotyping errors of the GBS method and imputed all erroneous and missing genotypes in all 5134 loci. Eliminating all pairs of SNP loci that had all the same or all opposite, and SNP loci with MAF ≤ 0.05, we obtained the 2309 SNPs distributed across the nine scaffolds of the clementine reference genome (186–353 SNPs per scaffold with an average marker interval of 0.12 Mbp, Supplementary Table S1).

Population structure

Bayesian model-based clustering analyses of the population structure of the 110 accessions deduced K = 2 as the most likely number of subpopulation (Fig. 4). Thus, we used the proportions of each accession for K = 2 in subsequent association analyses, as well as the additive genetic relationship matrix estimated from pedigree information.
Fig. 4

Evaluation of the population structure of citrus landraces, modern cultivars, and breeding lines in Japan. a Plot of Delta K vs. K. b Estimated population structure at K = 2. See Table 1 for the corresponding genotypes. Population structure was estimated using STRUCTURE v. 2.2 software with 30 SNPs per each of the nine linkage groups

Estimation of linkage disequilibrium decay

The two-dimensional histogram of pairwise LD between SNPs concisely showed the distribution of LD (r 2 ) against physical distance (Fig. 5a). The threshold value for significant LD was r 2  = 0.093. The significant LD ranged 4 to 5 Mbp at the genome-wide level (Fig. 5b). Therefore, our accessions showed relatively high LD.
Fig. 5

Evaluation of linkage disequilibrium (LD) of citrus landraces, modern cultivars, and breeding lines in Japan. a Two-dimensional histogram of pairwise LD between markers measured as r2 and physical distance. b Genome-wide LD decay patterns. LD decay was estimated using a total of 2309 single nucleotide polymorphisms along each of nine linkage groups, and then all estimates were combined. The curve is the trend line based on nonlinear regression of r2 on physical distance. Horizontal dashed line represents the r2 value of the top 5% of the distribution of the r2 value between pairs of unlinked markers

QTL identification and estimation of QTL effect

GWAS identified four significant QTL regions associated with FW, and one associated with FSC, PF, and SF, respectively (Table 4). The Manhattan plot and QQ plot for FW are shown in Fig. 6, and those for the other seven fruit-quality traits, including those with no significant QTL regions detected (i.e., FST, PE, SC, and AC), are shown in Supplementary Fig. S1.
Table 4

Significant SNPs associated with fruit-quality traits identified in the genome-wide association study

Trait

Significant SNPs

Scaffolda

Positionb (bp)

Major allele

Minor allele

Minor allele frequency

Proportion (%)c

P value

Additive

Dominance

Total

Fruit weight

5

34,619,963

C

T

0.150

4.72

4.72

2.2 × 10−10

2

11,899,695

G

C

0.145

0.80

0.80

6.3 × 10−9

3

48,681,070

G

A

0.077

0.48

0.48

8.1 × 10−9

7

3,228,616

A

G

0.109

0.38

0.38

8.1 × 10−9

Fruit skin color

3

1,745,277

T

G

0.468

0.00

3.62

3.62

6.0 × 10−6

Pulp firmness

3

4,027,526

C

T

0.500

3.18

2.43

5.61

1.5 × 10−5

Segment firmness

7

2,890,075

A

T

0.123

7.55

7.55

2.7 × 10−6

aScaffold number of the clementine reference genome v. 1.0 (http://www.phytozome.net) (Ollitrault et al. 2012; Wu et al. 2014)

bPositons on the clementine reference genome

cThe proportion of phenotypic variance explained by the SNP. Variance of the dominance effect was not estimated when fewer than five homozygotes were found for the minor allele

Fig. 6

Manhattan plots (left) and quantile–quantile plots (right) for fruit weight. The y-axis of Manhattan plots shows the –log10 (P values) of SNP association. The horizontal line shows the threshold value of significant association. The threshold value was calculated by the Wald test with Bonferroni correction based on the tested number of SNP markers (P < 0.05/2309)

In the seven significant QTL regions, the proportion of phenotypic variance explained by the corresponding SNPs ranged from 7.55 to 0.38%, and SNPs associated with FSC and PF showed dominance. The boxplot distribution of FW at ccsc5_34619963 (SNP identified at the 34619963rd bp on the 5th scaffold of the clementine reference genome) is shown in Fig. 7. Boxplot distributions of the other significant SNPs for FW are shown in Supplementary Fig. S2, and stacked bar graphs of each significant SNP for FSC, PF, and SF are shown in Supplementary Fig. S3. With respect to FW, the accessions carrying favorable alleles in the significant SNPs—that is, T allele at ccsc5_34619963 (Fig. 7), G allele at ccsc2_11899695 (Supplementary Fig. S2a), and G allele at ccsc7_3228616 (Supplementary Fig. S2c)—tended to have larger fruits than those with other alleles, although the effect of SNP ccsc3_48681070 (Supplementary Fig. S2b) is not clear. With respect to FSC, PF, and SF, the accessions carrying favorable alleles with additive effect or favorable genotypes with dominance effect—that is, genotype TG at ccsc3_1745277 (Supplementary Fig. S3a), C allele or genotype CT at ccsc3_4027526 (Supplementary Fig. S3b), and T allele at ccsc7_2890075 (Supplementary Fig. S3c)—tended to have red fruit skin, soft pulp, and soft segments, respectively. However, we could not narrow down the candidates of causal genes through annotation of genes within the LD regions containing significant SNPs in CitrusCyc Pathways v. 4.0 databases (http://pathways.citrusgenomedb.org/) owing to the high LD of our materials, as described above. The locations of all seven identified QTL regions for FW, FSC, PF, and SF on the physical map are shown in Fig. 8.
Fig. 7

Distribution of fruit weight according to genotypes of significant single nucleotide polymorphism (ccsc5_34619963). Box edges represent the upper and lower quantiles, with the median value shown as a bold line in the middle of the box

Fig. 8

Physical map of citrus comprising nine linkage groups with 2309 single nucleotide polymorphisms and QTLs identified in this study. Each number of linkage groups corresponds to the scaffold number of the clementine reference genome v. 1.0 (http://www.phytozome.net) (Ollitrault et al. 2012; Wu et al. 2014)

Discussion

GBS in genetic studies of citrus

In the present study, GBS and subsequent quality control provided 2309 SNPs across the genomes of the materials. With recent advances of next-generation sequencing (NGS) technologies, they offered cost and time-effective platforms for SNP discovery and genotyping, such as restriction-site associated DNA sequencing (RAD-seq; Baird et al. 2008), double-digest RAD-seq (Peterson et al. 2012), and GBS.

Using RAD-seq, Guo et al. (2015) constructed a high-density genetic map in a biparental population of pummelo (Citrus maxima). This genetic map consists of 1543 SNP and 20 simple sequence repeat (SSR) markers covering the whole genome, measuring 976.58 cM with an average distance of 0.62 cM. Oueslati et al. (2017) applied GBS to 55 modern citrus cultivars and revealed 30,289 SNPs and 8794 Indels covering the nine chromosomes at high density. Among these markers, they selected 11,133 polymorphic markers which can characterize the genomic differences between Citrus reticulata and Citrus maxima and revealed genomic regions introgressed from Citrus maxima into major cultivars of mandarin, mandarin hybrids, tangelo, and tangor. These studies and our results demonstrate the potential of GBS and RAD-seq in citrus genomics, including the construction of genetic maps, QTL analysis, and GWAS, estimation of genetic diversity, and evaluation of genetic admixture and structures in citrus population.

In both RAD-seq and GBS, restriction enzymes are used to reduce genome complexity and avoid the sequencing of repetitive genomic regions. We used PstI to digest DNA and construct GBS libraries. However, in our preliminary studies using mixed genomic DNA of satsuma mandarin and sweet orange, ApeKI (GCWGC) and EcoT22I (ATGCAT) could also be used for GBS analysis in citrus (Supplementary Fig. S4). Indeed, Oueslati et al. (2017) used ApeKI and successfully genotyped more SNPs and Indels in more diverse citrus cultivars than in our study. Therefore, combinations of these three enzymes for GBS could enable more detailed genome scanning and SNP discovery in diverse citrus germplasms.

Quality control for GBS in citrus

The capability of GWAS to identify true genetic associations between SNPs and target traits depends greatly on the overall quality of the data, including the accuracy of SNP genotyping, and thus quality control of SNP data is an essential step for GWAS. We identified potentially mistyped SNP genotypes (and also potentially incorrect parentages) by detecting Mendelian conflict in trios. This method can be applied to pedigreed populations, as are common in advanced fruit breeding materials (e.g., Imai et al. 2017a; Kouassi et al. 2009; Nishio et al. 2016). However, population-based association study is typically used with populations composed of unrelated genotypes to prevent the inflation of test statistics and a high false positive rate (Marchini et al. 2004). In fruit crops including citrus, this situation may correspond to the analysis of diverse germplasms. In this case, checking for Hardy–Weinberg equilibrium at each SNP locus could detect potential SNP genotyping errors, and is a well-established procedure, at least in human GWAS (Turner et al. 2011).

Sample relatedness and linkage disequilibrium

In GWAS, population structure and kinship relatedness can result in spurious associations (Yu et al. 2006). To avoid this problem, we inferred the population structure among the 110 accessions and incorporated the inferred structure as a covariate in the statistical model for GWAS. Bayesian model-based clustering in STRUCTURE software deduced K = 2 as the most likely number of subpopulations. However, as the population structure is not clear, it might not be necessary to incorporate it in the model. In addition to population structure, we used a genetic relationship matrix to detect kinship relatedness, because our samples have an obvious pedigree structure. The model taking the kinship relatedness into consideration, as used in this study, is appropriate for GWAS in the actual population including related genotypes.

Population structure and kinship relatedness also affect the extent of LD, which is a key factor in GWAS because it determines the number of genetic markers required to cover a genome and mapping resolution (Flint-Garcia et al. 2003). Since LD can vary greatly even within species depending on the individuals in a target population (Rafalski and Morgante et al. 2004), it is important to estimate the extent of LD in a target population before performing a GWAS. Using the 2309 genome-wide SNP markers, we evaluated the extent of significant LD with r2 > 0.093 in the 110 accessions as 4–5 Mbp. A similar high LD was observed in breeding cultivars or breeding progeny of Japanese pear (Iwata et al. 2013) and apple (Kumar et al. 2013; Moriya et al. 2017). In contrast, GBS and the following quality control gave an average interval of 0.12 Mbp between the 2309 SNPs with no Mendelian errors. Therefore, these procedures provided enough high-quality SNPs to perform GWAS in Japanese citrus breeding materials. However, the cause of the high LD observed in our materials is the close relationships among the materials, and thus, 2309 SNPs might fall short of covering the extent of LD in analysis of more diverse germplasms, which have no clear family relationships and thus are expected to have less LD. In that case, GBS with more read depth or combinations of other restriction enzymes could give enough SNPs for GWAS.

QTLs for fruit quality and implications for marker-assisted selection

We identified seven significant QTL regions associated with four fruit-quality traits: four QTLs for FW and one each for FSC, PF, and SF. These traits are important to citrus fruit quality, especially in mandarin (Goldenberg et al. 2018): fruit skin color and fruit size crucially influence consumers’ attention and appreciation, and softer pulp and segment firmness are preferred. Therefore, there are several previous studies for QTL analysis of these traits in citrus. For example, using 116 F1 hybrids from a cross between ‘Fortune’ and ‘Murcott,’ Yu et al. (2016) identified 48 QTL regions for eight important fruit-quality traits, including fruit size or weight and flavedo color. Curtolo et al. (2017) identified 19 QTL regions for 12 fruit characteristics, including fruit diameter using 278 F1 hybrids from a cross between ‘Murcott’ and ‘Pera’ sweet orange. In these previously identified QTLs, two regions for FW (ccsc3_48681070 and ccsc7_3228616) that we identified in this study are common with our recent report (Imai et al. 2017b), and one region for PF (ccsc3_4027526) is common with another recent report (Minamikawa et al. 2017). These results show that the GBS and SNP quality control described in this paper are suitable for GWAS in citrus and suggest the possibility of using the three QTLs for MAS in citrus breeding, because the reproducibility of QTLs identified by GWAS should be confirmed in independent experiments (Chanock et al. 2007). We consider the remaining four QTLs to be novel; they should be confirmed in additional experiments.

Although we identified seven QTL regions, the genetic variance explained by each SNP was not large, and therefore, there may not be enough selection effect for the use of these SNPs alone in MAS. In addition, no QTL regions were identified for FST, PE, SC, or AC. This suggests that the variations of all eight fruit-quality traits evaluated in this study can be attributed to polygenes rather than a few major genes. From the perspective of breeding for complex traits regulated by polygenes, MAS could limit the power for prediction of genotypic performance of selected individuals (Heffner et al. 2011). Instead, genomic selection (GS) is considered to be an efficient method for the selection of complex traits because it uses high-density genotype data sets which can capture polygenic effects and simultaneously models all additive genetic variance at the relevant loci (Meuwissen et al. 2001). Although introduced for animal breeding initially, GS has since shown potential in fruit crops such as Japanese pear (Iwata et al. 2013), apple (Kumar et al. 2012), grape (Viana et al. 2016), and citrus (Gois et al. 2016; Minamikawa et al. 2017). In the same way as GWAS does, GS uses LD between markers and causal genes or QTLs. Therefore, GBS and our SNP quality control procedures could be used not only for GWAS, but also for GS in citrus, and offer potential to improve the efficiency of cost- and labor-intensive citrus crossbreeding.

Conclusions

The cost-effective, high-throughput GBS approach followed by SNP quality control provided 2309 SNPs across the whole genome of the Japanese citrus breeding materials, which were enough for GWAS, as confirmed by our identification of seven QTL regions for MAS. These findings could facilitate efficient citrus breeding using whole-genome information.

Notes

Acknowledgements

We would like to thank Y. Yamamura, H. Hira, and other staff members of agricultural field in Kuchinotsu Citrus Research Station, NARO for careful management of plant materials used in this study.

Data Archiving Statement

All relevant data are presented in the main paper and in the Supplementary Materials.

Compliance with ethical standards

Conflict of interest

The authors declare that they have no conflict of interest.

Supplementary material

11295_2018_1238_MOESM1_ESM.pptx (513 kb)
Supplementary Fig. S1 Manhattan plot (left) and quantile–quantile plot (right) for (a) fruit skin color, (b) fruit surface texture, (c) peelability, (d) pulp firmness, (e) segment firmness, (f) sugar content, and (g) acid content. The y-axes of Manhattan plots show the –log10 (P-values) of SNP association. The horizontal lines show the threshold value of significant association. The threshold value was calculated by the Wald test with Bonferroni correction based on the tested number of SNP markers (P < 0.05/2309). (PPTX 512 kb)
11295_2018_1238_MOESM2_ESM.pptx (303 kb)
Supplementary Fig. S2 Distribution of fruit weight according to genotypes of significant SNPs. Box edges represent the upper and lower quantiles, with the median value shown as a bold line in the middle of the box. (a) ccsc2_11899695, (b) ccsc3_48681070, and (c) ccsc7_3228616. (PPTX 302 kb)
11295_2018_1238_MOESM3_ESM.pptx (150 kb)
Supplementary Fig. S3 Stacked bar graph of fruit skin color (a), pulp firmness (b), and segment firmness (c) according to genotypes of significant SNPs. The y-axes show the number of observations for each SNP genotype: (a) ccsc3_1745277, (b) ccsc3_4027526, and (c) ccsc7_2890075. (PPTX 149 kb)
11295_2018_1238_MOESM4_ESM.pptx (276 kb)
Supplementary Fig. S4 Gel images (left) and electropherogram traces (right) of GBS libraries made by digesting mixed genomic DNA of satsuma mandarin and sweet orange with methylation-sensitive restriction enzymes (a) ApeKI, (b) EcoT22I, and (c) PstI. Discrete peaks or bands (15 bp and 1500 bp) indicate the presence of repetitive DNAs. Most fragments in all three GBS libraries were <500 bp, and were thus appropriate for Illumina short-read sequencing. ApeKI (essentially a 4-base cutter) produced a larger fragment pool than did EcoT22I and PstI (6-base cutters). Therefore, an ApeKI GBS library could produce more SNPs than the other two restriction enzymes. In contrast, EcoT22I and PstI GBS libraries could increase sequence coverage per locus relative to an ApeKI GBS library. (PPTX 275 kb)
11295_2018_1238_MOESM5_ESM.xlsx (114 kb)
Supplementary Table S1 (XLSX 113 kb)

References

  1. Abecasis GR, Cherny SS, Cookson WO, Cardon LR (2002) Merlin–rapid analysis of dense genetic maps using sparse gene flow trees. Nat Genet 30:97–101CrossRefPubMedGoogle Scholar
  2. Baird NA, Etter PD, Atwood TS, Currey MC, Shiver AL, Lewis ZA, Selker EU, Cresko WA, Johnson EA (2008) Rapid SNP discovery and genetic mapping using sequenced RAD markers. PLoS One 3:e3376CrossRefPubMedPubMedCentralGoogle Scholar
  3. Bradbury PJ, Zhang Z, Kroon DE, Casstevens TM, Ramdoss Y, Buckler ES (2007) TASSEL: software for association mapping of complex traits in diverse samples. Bioinformatics 23:2633–2635CrossRefPubMedGoogle Scholar
  4. Chanock SJ, Manolio T, Boehnke M, Boerwinkle E, Hunter DJ, Thomas G, Hirschhorn JN, Abecasis G, Altshuler D, Bailey-Wilson JE, Brooks LD, Cardon LR, Daly M, Donnelly P, Fraumeni JF, Freimer NB, Gerhard DS, Gunter C, Guttmacher AE, Guyer MS, Harris EL, Hoh J, Hoover R, Kong CA, Merikangas KA, Morton CC, Palmer LJ, Phimister EG, Rice JP, Roberts J, Rotimi C, Tucker MA, Vogan KJ, Wacholder S, Wijsman EM, Winn DM, Collins FS (2007) Replicating genotype–phenotype associations. Nature 447:655–660CrossRefPubMedGoogle Scholar
  5. Curtolo M, Cristofani-Yaly M, Gazaffi R, Takita MA, Figueira A, Machado MA (2017) QTL mapping for fruit quality in Citrus using DArTseq markers. BMC Genomics 18:289CrossRefPubMedPubMedCentralGoogle Scholar
  6. Deng Z, Xu J (2011) Breeding for fruit quality in citrus. In: Jenks MA, Bebeli PJ (eds) Breeding for fruit quality. Wiley, New York, pp 349–371CrossRefGoogle Scholar
  7. Elshire RJ, Glaubitz JC, Sun Q, Poland JA, Kawamoto K, Buckler ES, Mitchell SE (2011) A robust, simple genotyping-by-sequencing (GBS) approach for high diversity species. PLoS One 6:e19379CrossRefPubMedPubMedCentralGoogle Scholar
  8. Evanno G, Regnaut S, Goudet J (2005) Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Mol Ecol 14:2611–2620CrossRefPubMedGoogle Scholar
  9. FAO (2016) FAOSTAT. http://www.fao.org/economic/ess/en/. Accessed 29 Nov 2016
  10. Flint-Garcia SA, Thornsberry JM, Buckler ES IV (2003) Structure of linkage disequilibrium in plants. Annu Rev Plant Biol 54:357–374CrossRefPubMedGoogle Scholar
  11. Furr JR (1969) Citrus breeding for the arid southwestern United States. In: Chapman HD (ed) Proceedings of 1st international citrus symposium, vol 1. University of California, Riverside, CA, USA, pp 191–197Google Scholar
  12. Gianola D, Foulley JL (1983) Sire evaluation for ordered categorical data with a threshold model. Genet Sel Evol 15:201–224CrossRefPubMedPubMedCentralGoogle Scholar
  13. Gilmour AR, Gogel BJ, Cullis BR, Thompson R (2015) ASReml user guide release 4.1. VSN International Ltd., Hemel HempsteadGoogle Scholar
  14. Glaubitz JC, Casstevens TM, Lu F, Harriman J, Elshire RJ, Sun Q, Buckler ES (2014) TASSEL-GBS: a high capacity genotyping by sequencing analysis pipeline. PLoS One 9:e90346CrossRefPubMedPubMedCentralGoogle Scholar
  15. Gmitter FG, Chen C, Nagesware R, Soneji JR (2007) 14 Citrus fruits. In: Kole C (ed) Genome mapping and molecular breeding in plants, volume 4: Fruits and nuts. Springer-Verlag, Berlin Heidelberg, pp 265–279Google Scholar
  16. Gois IB, Borém A, Cristofani-Yaly M, de Resende MDV, Azevedo CF, Bastianel M, Novelli VM, Machado MA (2016) Genome wide selection in Citrus breeding. Genet Mol Res 15:gmr15048863CrossRefGoogle Scholar
  17. Goldenberg L, Yaniv Y, Porat R, Carmi N (2018) Mandarin fruit quality: a review. J Sci Food Agric 98:18–26CrossRefPubMedGoogle Scholar
  18. Guo F, Yu H, Tang Z, Jiang X, Wang L, Wang X, Xu Q, Deng X (2015) Construction of a SNP-based high-density genetic map for pummelo using RAD sequencing. Tree Genet Genomes 11:1–11CrossRefGoogle Scholar
  19. Harville DA, Mee RW (1984) A mixed model procedure for analyzing ordered categorical data. Biometrics 40:393–408CrossRefGoogle Scholar
  20. He J, Zhao X, Laroche A, Lu ZX, Liu HK, Li Z (2014) Genotyping-by-sequencing (GBS), an ultimate marker-assisted selection (MAS) tool to accelarate plant breeding. Front Plant Sci 5:484CrossRefPubMedPubMedCentralGoogle Scholar
  21. Hearn CJ, Bai J, Baldwin E, McCollum TG, Hall DG, Stover E, Driggers R (2014) Breeding “sweet oranges” at the USDA US Horticultural Research Laboratory. In XXIX International Horticultural Congress on Horticulture: sustaining lives, livelihoods and landscapes (IHC2014): 1127 (pp. 41–44)Google Scholar
  22. Heffner EL, Jannink JL, Sorrells ME (2011) Genomic selection accuracy using multifamily prediction models in a wheat breeding program. Plant Genome 4:65–75CrossRefGoogle Scholar
  23. Imai A, Kuniga T, Yoshioka T, Nonaka K, Mitani N, Fukamachi H, Hiehata N, Yamamoto M, Hayashi T (2017a) Genetic background, inbreeding, and genetic uniformity in the national citrus breeding program, Japan. Hortic J 86:200–207CrossRefGoogle Scholar
  24. Imai A, Yoshioka T, Hayashi T (2017b) Quantitative trait locus (QTL) analysis of fruit-quality traits for mandarin breeding in Japan. Tree Genet Genomes 13:79CrossRefGoogle Scholar
  25. Iwata H, Hayashi T, Terakami S, Takada N, Sawamura Y, Yamamoto T (2013) Potential assessment of genome-wide association study and genomic selection in Japanese pear Pyrus pyrifolia. Breed Sci 63:125–140CrossRefPubMedPubMedCentralGoogle Scholar
  26. Khan IA, Kender WJ (2007) Citrus breeding: introduction and objectives. In: Khan IA (ed) Citrus genetics, breeding and biotechnology. CAB International, Wallingford, pp 1–8CrossRefGoogle Scholar
  27. Kouassi AB, Durel CE, Costa F, Tartarini S, van de Weg E, Evans K, Fernandez-Fernandez F, Govan C, Boudichevskaja A, Dunemann F, Antofie A, Lateur M, Stankiewicz-Kosyl M, Soska A, Tomala K, Lewandowski M, Rutkovski K, Zurawicz E, Guerra W, Laurens F (2009) Estimation of genetic parameters and prediction of breeding values for apple fruit-quality traits using pedigreed plant material in Europe. Tree Genet Genomes 5:659–672CrossRefGoogle Scholar
  28. Kumar S, Chagné D, Bink MCAM, Volz RK, Whitworth C, Carlisle C (2012) Genomic selection for fruit quality traits in apple (Malus× domestica Borkh.) PLoS One 7:e36674CrossRefPubMedPubMedCentralGoogle Scholar
  29. Kumar S, Garrick DJ, Bink M, Whitworth C, Chagne D, Volz RK (2013) Novel genomic approaches 145 unravel genetic architecture of complex traits in apple. BMC Genomics 14:393CrossRefPubMedPubMedCentralGoogle Scholar
  30. Kunihisa M, Moriya S, Abe K, Okada K, Haji T, Hayashi T, Kim H, Nishitani C, Terakami S, Yamamoto T (2014) Identification of QTLs for fruit quality traits in Japanese apples: QTLs for early ripening are tightly related to preharvest fruit drop. Breeding Sci 64:240–251CrossRefGoogle Scholar
  31. Langmead B, Trapnell C, Pop M, Salzberg SL (2009) Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Bio l10:R25CrossRefGoogle Scholar
  32. Marchini J, Cardon LR, Phillips MS, Donnelly P (2004) The effects of human population structure on large genetic association studies. Nat Genet 36(5):512–517CrossRefPubMedGoogle Scholar
  33. Meuwissen TH, Hayes B, Goddard M (2001) Prediction of total genetic value using genome-wide dense marker maps. Genetics 157:1819–1829PubMedPubMedCentralGoogle Scholar
  34. Minamikawa MF, Nonaka K, Kaminuma E, Kajiya-Kanegae H, Onogi A, Goto S, Yoshioka T, Imai A, Hamada H, Hayashi T, Matsumoto S, Katayose Y, Toyoda A, Fujiyama A, Nakamura Y, Shimizu T, Iwata H (2017) Genome-wide association study and genomic prediction in citrus: potential of genomics-assisted breeding for fruit quality traits. Sci Rep 7:4721CrossRefPubMedPubMedCentralGoogle Scholar
  35. Moriya S, Kunihisa M, Okada K, Iwanami H, Iwata H, Minamikawa M, Katayose Y, Matsumoto T, Mori S, Sasaki H, Matsumoto T, Nishitani C, Terakami S, Yamamoto T, Abe K (2017) Identification of QTLs for flesh mealiness in apple (Malus × domestica Borkh.) Hortic J 86:159–170CrossRefGoogle Scholar
  36. Ninomiya T, Shimada T, Endo T, Nonaka K, Omura M, Fujii H (2015) Development of citrus cultivar identification by CAPS markers and parentage analysis. Hort Res 14:127–133CrossRefGoogle Scholar
  37. Nishio S, Norio T, Saito T, Yamamoto T, Iketani H (2016) Estimation of loss of genetic diversity in modern Japanese cultivars by comparison of diverse genetic resources in Asian pear (Pyrus spp.) BMC Genet 17:81CrossRefPubMedPubMedCentralGoogle Scholar
  38. Ollitrault P, Terol J, Chen CX, Federici CT, Lotfy S, Hippolyte I, Ollitrault F, Bérard A, Chauveau A, Cuenca J, Costantino G, Kacar Y, Mu L, Garcia-Lor A, Froelicher Y, Aleza P, Boland A, Billot C, Navarro L, Luro F, Roose ML, Gmitter FG, Talon M, Brunel D (2012) A reference genetic map of C. clementina hort. ex Tan.; citrus evolution inferences from comparative mapping. BMC Genomics 13:593CrossRefPubMedPubMedCentralGoogle Scholar
  39. Omura M, Shimada T (2016) Citrus breeding, genetics and genomics in Japan. Breeding Sci 66:3–17CrossRefGoogle Scholar
  40. Oueslati A, Salhi-Hannachi A, Luro F, Vignes H, Mournet P, Ollitrault P (2017) Genotyping by sequencing reveals the interspecific C. maxima / C. reticulata admixture along the genomes of modern citrus varieties of mandarins, tangors, tangelos, orangelos and grapefruits. PLoS One 12:e0185618CrossRefPubMedPubMedCentralGoogle Scholar
  41. Peterson BK, Weber JN, Kay EH, Fisher HS, Hoekstra HE (2012) Double digest RADseq: an inexpensive method for de novo SNP discovery and genotyping in model and non-model species. PLoS One 7:e37135CrossRefPubMedPubMedCentralGoogle Scholar
  42. Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genotype data. Genetics 155:945–959PubMedPubMedCentralGoogle Scholar
  43. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira M, Bender D, Maller J, Sklar P, de Bakker P, Daly M, Sham P (2007) PLINK: a toolset for whole-genome association and population-based linkage analysis. Am J Hum Genet 81:559–575CrossRefPubMedPubMedCentralGoogle Scholar
  44. Rafalski A, Morgante M (2004) Corn and humans: recombination and linkage disequilibrium in two genomes of similar size. Trends Genet 20:103–111CrossRefPubMedGoogle Scholar
  45. Sargolzaei M, Chesnais JP, Schenkel FS (2014) A new approach for efficient genotype imputation using information from relatives. BMC Genomics 15:478CrossRefPubMedPubMedCentralGoogle Scholar
  46. Soetaert K (2013) plot3D: plotting multi-dimensional data. R package version 1.0Google Scholar
  47. Soost RK (1987) Breeding citrus-genetics and nucellar embryony. Improving vegetatively propagated crops. Academic Press, London, pp 83–110Google Scholar
  48. Turner SD (2014) qqman: an R package for visualizing GWAS results using Q-Q and manhattan plots. bioRxiv (2014): 005165Google Scholar
  49. Turner S, Armstrong LL, Bradford Y, Carlson CS, Crawford DC, Crenshaw AT, de Andrade M, Doheny KF, Haines JL, Hayes G, Jarvik G, Jiang L, Kullo IJ, Li R, Ling H, Manolio TA, Matsumoto M, McCarty CA, McDavid AN, Mirel DB, Paschall JE, Pugh EW, Rasmussen LV, Wilke RA, Zuvich RL, Ritchie MD (2011) Quality control procedures for genome-wide association studies. Curr Protoc Hum Genet Chapter 1: Unit1 19Google Scholar
  50. Viana AP, Resende MDV, Riaz S, Walker MA (2016) Genome selection in fruit breeding: application to table grapes. Sci Agric 73:142–149CrossRefGoogle Scholar
  51. Voorrips RE, Bink MCAM, van de Weg WE (2012) Pedimap: software for the visualization of genetic and phenotypic data in pedigrees. J Hered 103:903–907CrossRefPubMedPubMedCentralGoogle Scholar
  52. Wiggans GR, Sonstegard TS, Vanraden PM, Matukumalli LK, Schnabel RD, Taylor JF, Schenkel FS, Van Tassell CP (2009) Selection of single-nucleotide polymorphisms and quality of genotypes used in genomic evaluation of dairy cattle in the United States and Canada. J Dairy Sci 92:3431–3436CrossRefPubMedGoogle Scholar
  53. Wolak ME (2012) nadiv: an R package to create relatedness matrices for estimating non-additive genetic variances in animal models. Methods Ecol Evol 3(5):792–796CrossRefGoogle Scholar
  54. Wu GA, Prochnik S, Jenkins J, Salse J, Hellsten U, Murat F, Perrier X, Ruiz M, Scalabrin S, Terol J, Takita MA, Labadie K, Poulain J, Couloux A, Jabbari K, Cattonaro F, Del Fabbro C, Pinosio S, Zuccolo A, Chapman J, Grimwood J, Tadeo FR, Estornell LH, Muñoz-Sanz JV, Ibanez V, Herrero-Ortega A, Aleza P, Pérez-Pérez J, Ramón D, Brunel D, Luro F, Chen C, Farmerie WG, Desany B, Kodira C, Mohiuddin M, Harkins T, Fredrikson K, Burns P, Lomsadze A, Borodovsky M, Reforgiato G, Freitas-Astúa J, Quetier F, Navarro L, Roose M, Wincker P, Schmutz J, Morgante M, Machado MA, Talon M, Jaillon O, Ollitrault P, Gmitter F, Rokhsar D (2014) Sequencing of diverse mandarin, pummelo and orange genomes reveals complex history of admixture during citrus domestication. Nat Biotechnol 32:656–662CrossRefPubMedPubMedCentralGoogle Scholar
  55. Yu J, Pressoir G, Briggs WH, Bi IV, Yamsaki M, Doebley JF, McMullen MD, Gaut BS, Nielsen DM, Holland JB, Kresovich S, Buckler ES (2006) A unified mixed-model method for association mapping that accounts for multiple levels of relatedness. Nat Genet 38:203–208CrossRefPubMedGoogle Scholar
  56. Yu Y, Chen C, Gmitter FG Jr (2016) QTL mapping of mandarin (Citrus reticulata) fruit characters using high-throughput SNP markers. Tree Genet Genomes 12:77CrossRefGoogle Scholar

Copyright information

© Springer-Verlag GmbH Germany, part of Springer Nature 2018

Authors and Affiliations

  1. 1.NARO Institute of Fruit Tree and Tea ScienceTsukubaJapan
  2. 2.Graduate School of Life and Environmental SciencesTsukuba UniversityTsukubaJapan
  3. 3.NARO Institute of Fruit Tree and Tea ScienceShimizuJapan
  4. 4.NARO Western Region Agricultural Research CenterZentsujiJapan
  5. 5.NARO Institute of Crop ScienceTsukubaJapan

Personalised recommendations