RESEARCH ARTICLE
`
`The Genetic Basis of Natural Variation in
`Kernel Size and Related Traits Using a Four-
`Way Cross Population in Maize
`Jiafa Chen1☯, Luyan Zhang2☯, Songtao Liu3, Zhimin Li1, Rongrong Huang1, Yongming Li1,
`Hongliang Cheng1, Xiantang Li1, Bo Zhou1, Suowei Wu1, Wei Chen1, Jianyu Wu1*,
`Junqiang Ding1*
`
`1 College of Agronomy, Synergetic Innovation Center of Henan Grain Crops and National Key Laboratory of
`Wheat and Maize Crop Science, Henan Agricultural University, Zhengzhou, 450002, China, 2 The National
`Key Facility for Crop Gene Resources and Genetic Improvement, Institute of Crop Science and CIMMYT
`China Office, Chinese Academy of Agricultural Sciences, Beijing 100081, China, 3 Henan Vocational
`College of Agriculture, Zhengzhou, 450002, China
`
`☯ These authors contributed equally to this work.
`* henau1302@126.com (JW); dingjunqiang1203@163.com (JD)
`
`a11111
`
`Abstract
`
`Kernel size is an important component of grain yield in maize breeding programs. To extend
`the understanding on the genetic basis of kernel size traits (i.e., kernel length, kernel width
`and kernel thickness), we developed a set of four-way cross mapping population derived
`from four maize inbred lines with varied kernel sizes. In the present study, we investigated
`the genetic basis of natural variation in seed size and other components of maize yield
`(e.g., hundred kernel weight, number of rows per ear, number of kernels per row). In total,
`ten QTL affecting kernel size were identified, three of which (two for kernel length and one
`for kernel width) had stable expression in other components of maize yield. The possible
`genetic mechanism behind the trade-off of kernel size and yield components was
`discussed.
`
`OPEN ACCESS
`
`Citation: Chen J, Zhang L, Liu S, Li Z, Huang R, Li Y,
`et al. (2016) The Genetic Basis of Natural Variation in
`Kernel Size and Related Traits Using a Four-Way
`Cross Population in Maize. PLoS ONE 11(4):
`e0153428. doi:10.1371/journal.pone.0153428
`
`Editor: Xianlong Zhang, National Key Laboratory of
`Crop Genetic Improvement, CHINA
`
`Received: January 22, 2016
`
`Accepted: March 29, 2016
`
`Published: April 12, 2016
`
`Copyright: © 2016 Chen et al. This is an open
`access article distributed under the terms of the
`Creative Commons Attribution License, which permits
`unrestricted use, distribution, and reproduction in any
`medium, provided the original author and source are
`credited.
`
`Data Availability Statement: All relevant data are
`within the paper and its Supporting Information files.
`
`Funding: This work was supported by Henan Basic
`Research Program of China (142102110051).
`
`Competing Interests: The authors have declared
`that no competing interests exist.
`
`Introduction
`Maize (Zea mays L.) is one of the most important cereal crops in the world, and increasing the
`maize production by selection for the components of grain yield is the main objective in maize
`breeding programs [1]. Maize kernel size, measured by kernel length, width and thickness, is
`an important component of grain yield. Moreover, the characteristic of kernel size is also an
`important factor of appearance quality, which may influence the corn market grades and con-
`sumer preference [2]. Therefore, investigating the genetic basis of kernel size, and discovering
`any possible genetic constraints to optimize it, will facilitate the improvement of grain yield in
`maize breeding programs.
`Quantitative trait locus (QTL) mapping based on molecular markers have been widely used
`in the genetic study for different traits in different crops [3–6]. There were several QTL/genes
`
`PLOS ONE | DOI:10.1371/journal.pone.0153428 April 12, 2016
`
`(cid:44)(cid:51)(cid:53)(cid:21)(cid:19)(cid:20)(cid:28)(cid:16)(cid:19)(cid:20)(cid:19)50
`PRECISION PLANTING v. DEERE
`EXHIBIT 1138
`
`1 / 12
`
`

`

`QTL Mapping for Maize Kernel Size
`
`have been identified in grain crops for kernel size [2,7–9]. Especially, some genes for kernel
`traits have been successfully isolated by map-based cloning strategy in rice, e.g., GS3 [3,10],
`qGL3 [11], GW2 [4], qSW5 [12], GS5 [13], GW8 [14], GL7 [15] and GW7 [16], and these genes
`can be useful targets of molecular-assisted selection for larger seed size in modern rice breeding
`programs. In maize, mutants analysis have identified several genes in key pathways involved in
`seed development, such as Mn1 [17], o2 [18], sh2 [19], gln1-4 [20], o1 [21], and others [22].
`Compared with the effects of mutants in kernel size, which are often dependent on the genetic
`background [23], QTL mapping approach is ideal to identify favorable QTL/genes that can
`contribute to natural variation of kernel size. In recent years, mapping QTL for kernel size is
`attractive in maize, and most studies for kernel size and related traits have been conducted in
`bi-parental populations [7–9,24–28].
`The available QTL generally detected in bi-parental mapping populations have greatly con-
`tributed to the understanding of the genetic basis of kernel size. However, QTL mapping in
`such populations is subject to low allele numbers and limited recombination [29]. In recent
`years, the generation of multi-parent advanced generation integrated cross (MAGIC) popula-
`tions has provided an additional option for QTL mapping. Compared with the bi-parental
`linkage populations, the development of MAGIC populations usually involved inter-crossing
`of multiple parental lines, which may introduce more than two independent alleles at a locus
`and subsequently increased probability of QTL being polymorphic across the multiple parents
`[30]. In addition, the precision and resolution of QTL detection can be increased by the ampli-
`fied number of recombination events [31]. In view of the merits of MAGIC for QTL mapping,
`increasing number of MAGIC populations have been created in model animals and plants
`recently. For example, two MAGIC populations have been developed in mice and used for
`identifying candidate genes for serum cholesterol and coat color traits [32,33]. In plants,
`MAGIC populations were first developed in Arabidopsis and subsequently expanded to crops
`[34]. In recent years, encouraging results have been reported for flowering time, leaf morphol-
`ogy and seed traits of Arabidopsis thaliana [35–37], fruit weight of tomato [38], plant height
`and shoot traits of wheat [39,40], biotic stress and abiotic stress of rice [41] and flowering time
`of barley [42]. Very recently, MAGIC populations were also developed in maize and then used
`for QTL mapping in traits such as flowering time, plant height, ear height and grain yield [43].
`Crop seed is a life-history trait, and the availability of resource pool in seed developmental
`processes drives seed production [2,44,45]. Due to the competing apportionment of resources
`between fitness components (i.e., seed size and seed number), a trade-off between seed number
`and size must occur [46]. A better understanding of natural variation in seed size requires
`simultaneous consideration of trade-off of kernel number related traits to seed development
`[37].
`Given the potential benefits of multi-parental (four-way cross) mapping population, we
`developed a set of multi-parental (four-way cross) mapping population in maize [6]. In the
`present study, we investigated the natural variation in seed size and other seed related traits.
`The objectives of this study were to detect the genetic architecture underlying seed size in
`maize, and specifically we were interested in the genetic mechanism behind trade-off of seed
`traits to better understand the genetic basis of kernel size.
`
`Materials and Methods
`The experiment was conducted in Zhengzhou Experiment Station (34°51'N 113°35'E) and
`Jiyuan Experiment Station (35°4'N 112°36'E) of Henan Agricultural University (HAU). At the
`two experimental locations, HAU has set up experimental field bases for non-profit agricul-
`tural research with a wide array of partners in China. In the present study, the field experiments
`
`PLOS ONE | DOI:10.1371/journal.pone.0153428 April 12, 2016
`
`2 / 12
`
`

`

`QTL Mapping for Maize Kernel Size
`
`in the two stations were approved by HAU. Further, the stations where field studies were con-
`ducted are not protected locations for endangered or protected species.
`
`Plant materials
`The four-way cross mapping population including 305 individuals was developed from the
`four-way cross among D276/D72//A188/Jiao51. The four parental lines were selected based on
`the agronomic performances for a range of traits in maize breeding programs. All 305 individu-
`als were self-crossed to develop progeny families. Twenty eight out of 305 individuals lacked
`enough self-pollinated seeds, and finally, 277 four-way cross F1 individuals were genotyped for
`genetic map construction, and their selfed progeny, known as four-way cross families, were
`used for phenotyping [6].
`
`Field trials and trait evaluation
`In 2010, the 277 four-way cross families, together with their four parents were planted at the
`Jiyuan Experiment Station and Zhengzhou Experiment Station, respectively. Field experiments
`in each location were arranged in a randomized complete block design with three replicates.
`Each plot included one row with 4 m long and 0.67 m wide, and was overplanted and then
`thinned to 15 plants per row at a density of 52,500 plants per hectare.
`To determine whether flowering time (FT) affects the trade-off between kernel size, FT was
`investigated and recorded as the number of days from planting when 50% of the plants in a
`row were shedding pollen. At physiological maturity, eight consecutive plants from the center
`of each plot were harvested by hand for trait measurements. The ear traits were evaluated,
`which included ear row number (ERN) and kernel number per row (KNR). After ears were
`dried down to a constant weight, the kernels at the middle of the ears in each plot were shelled
`and bulked. Four kernel traits were measured, including 100-kernel weight (HKW), kernel
`length (KL), kernel width (KW) and kernel thickness (KT). HKW was estimated from the aver-
`age of three measurements of the weight of 100 randomly selected kernels; KL, KW and KT
`were estimated by the average of three replicated measurements of 50 kernels randomly
`selected from the bulked kernels using electronic digital calipers.
`
`Phenotypic data analysis
`Analysis of variance for phenotype data was performed using the General Line Model (Proc
`GLM) procedure in SAS software [47], and Fisher Least Significant Different (LSD) method
`was used for multiple comparisons. The components of variance were estimated using a ran-
`dom-effect model and broad-sense heritability (H2) for each trait across the two environ-
`ments was calculated as defined by Knapp et al. [48]. Phenotypic correlations among traits
`were calculated by the Pearson correlation method using the mean values of genotypes across
`environments.
`
`Genetic map and QTL mapping
`Genetic linkage map was constructed using the algorithm proposed by Zhang et al. which was
`implemented in software package GACD as functionality CDM [49]. Two hundred and twenty
`one markers were relatively evenly distributed on 10 maize chromosomes and the whole length
`of the genome was 1799.03 cM [6].
`The algorithm of inclusive composite interval mapping (ICIM) for four-way crosses was
`implemented in GACD software (http://www.isbreeding.net) as functionality CDQ [50] and
`used for QTL mapping of six traits, i.e., KL, KW, KT, HKW, KNR and ERN. QTL analysis was
`
`PLOS ONE | DOI:10.1371/journal.pone.0153428 April 12, 2016
`
`3 / 12
`
`

`

`QTL Mapping for Maize Kernel Size
`
`performed on the mean values of each genotype across the two environments. Inclusive linear
`models that includes marker variables and marker interactions so as to completely control
`both additive and dominance effects were built respectively for each trait. Stepwise regression
`was used to select significant marker variables and then used for background control in Inclu-
`sive Composite Interval Mapping (ICIM) of QTL [50]. The two probabilities for entering and
`removing variables were set at 0.001 and 0.002. The scanning step was 1 cM. LOD threshold
`was set at 3.97 by the empirical formula derived from Zhang et al. [50]. The original genotypes,
`phenotypes and linkage maps of the four-way cross population was available in S1 Dataset.
`
`Results
`Phenotypic variation and heritability
`The phenotypic variations of kernel size and related traits among the four parental lines were
`investigated in Jiyuan and Zhengzhou locations in 2010, and significant variations were
`observed for all traits measured in this study, including three kernel size traits (i.e., KL, KW
`and KT), ear traits (i.e., ERN and KNE) and kernel weight (HKW) (Table 1). Among the four-
`way families comprising of 277 entries, extensive phenotypic variation was observed in kernel
`size, HKW, ERN, KNR as well as FT (Table 2, S1 Fig). The heritability (H2) of the traits ranged
`
`Table 1. Summary of phenotype analysis and multiple comparisons of traits evaluated for the four parental lines.
`
`Traitsa
`
`KL (mm)
`
`KW (mm)
`
`KT (mm)
`
`ERN
`
`KNR
`
`HKW (g)
`
`Lines
`
`Jiao51
`A188
`D276
`D72
`D276
`A188
`Jiao51
`D72
`A188
`Jiao51
`D276
`D72
`A188
`D72
`D276
`Jiao51
`D276
`D72
`A188
`Jiao51
`A188
`Jiao51
`D276
`D72
`
`Mean
`
`7.68
`8.70
`9.62
`9.87
`6.93
`7.48
`8.18
`8.42
`3.96
`5.00
`5.31
`5.81
`11.3
`12.5
`13.7
`16.1
`10.5
`12.0
`12.5
`16.8
`9.3
`14.5
`18.3
`23.1
`
`SD
`
`0.56
`0.61
`0.71
`0.66
`0.64
`0.52
`0.59
`0.68
`0.40
`0.51
`0.55
`0.75
`0.76
`0.87
`1.50
`0.62
`3.18
`1.83
`1.93
`0.90
`1.01
`0.41
`3.00
`0.90
`
`Range
`
`6.07–9.44
`7.05–9.78
`8.13–11.34
`8.26–11.08
`5.25–8.96
`5.97–8.74
`6.65–9.58
`6.02–10.03
`3.31–5.75
`3.71–6.26
`3.93–6.94
`6.20–7.39
`10.0–12.0
`11.2–14.0
`11.0–15.7
`15.2–17.0
`7.0–14.0
`9.6–15.0
`10.0–14.5
`15.4–17.9
`7.58–10.5
`13.8–15.2
`13.42–22.5
`21.8–24.15
`
`LSD 0.05
`
`LSD 0.01
`
`a
`b
`c
`d
`a
`b
`c
`d
`a
`b
`c
`d
`a
`ab
`b
`c
`a
`ab
`b
`c
`a
`b
`c
`d
`
`A
`B
`C
`C
`A
`B
`C
`D
`A
`B
`C
`D
`A
`AB
`B
`C
`A
`A
`A
`AB
`A
`B
`C
`D
`
`a: KL: kernel length; KW: kernel width; KT: kernel thickness; ERN: ear row number; KNR: kernel number per row; HKW: hundred kernel weight. LSD:
`Least Significant Difference; The same letters in LSD 0.05 and LSD 0.01 columns indicate that difference is not significant in the same group at P < 0.05
`(LSD 0.05) or P < 0.01 (LSD 0.01) levels.
`
`doi:10.1371/journal.pone.0153428.t001
`
`PLOS ONE | DOI:10.1371/journal.pone.0153428 April 12, 2016
`
`4 / 12
`
`

`

`QTL Mapping for Maize Kernel Size
`
`Table 2. Phenotypic variation among four-way cross families for all traits measured.
`
`Traits
`
`Flowering Time (days)
`Kernel length (mm)
`Kernel width (mm)
`Kernel thickness (mm)
`Ear row number
`Kernel number per row
`100 kernel weight (g)
`
`Min
`
`59.20
`8.95
`7.70
`4.10
`12.30
`28.79
`19.55
`
`Max
`
`67.40
`10.94
`9.28
`5.70
`17.00
`51.80
`29.95
`
`Mean±SD
`63.48±1.45
`9.88±0.34
`8.32±0.29
`4.90±0.02
`14.56±0.93
`41.77±4.08
`23.46±1.85
`
`H2
`
`0.81
`0.77
`0.73
`0.79
`0.75
`0.70
`0.71
`
`Minimum (Min), maximum (Max) phenotypic values for each trait, as well as the phenotypic mean plus or
`minus their standard deviation (SD) and their broad-sense heritability (H2) were shown.
`
`doi:10.1371/journal.pone.0153428.t002
`
`from 0.70 (KNR) to 0.81 (FT) across the two environments, suggesting that genetic factor
`played an important role in the four-way cross population.
`
`Correlation of seed size and other traits
`Of the traits surveyed in this study, a number of significant pairwise correlations were observed
`between kernel size and the other traits (i.e., FT, ERN, KNR and HKW) (Table 3). For example,
`ERN showed significant positive correlation with KL (r = 0.248), while with significant negative
`correlation with KW (r = -0.319) and KT (r = -0.187), suggesting the trade-offs between ERN
`and kernel size. However, KNR was only significantly correlated with one of the three traits of
`kernel size (KT, r = -0.606), which implied the trade-offs between KNR and KT instead of KL
`and KW. Significant positive correlations were also observed between HKW and kernel size (r
`values were 0.462, 0.693 and 0.493 for KL, KW and KT, respectively), which means each of the
`three kernel size components contributed to the weight of the kernel in this population. Corre-
`lation between FT and kernel size was observed, however, FT was significant but weak corre-
`lated with KL (r = 0.156), and not significant with KW and KT.
`Given the extensive correlation among all traits, multiple linear regression model was used
`to estimate the effect of the different life-history traits on kernel size. The best fit model (small-
`est AIC) of KL (F = 54.53, p-value <2.2e-16, r2 = 0.56) included: KW, KT, ERN and HKW,
`which explained 15.9, 9.0, 8.1 and 16.3% of the variation of KL, respectively. For the KW, the
`best model (F = 103.4, p-value <2.2e-16, r2 = 0.65) could explain 65% of the variance. The
`
`Table 3. Pairwise Pearson’s correlations between traits measured.
`
`Traitsa
`KL
`KW
`KT
`ERN
`KNR
`HKW
`
`FT
`0.156*
`-0.051
`-0.036
`0.177**
`-0.069
`-0.107
`
`KL
`
`0.545**
`-0.176*
`0.248**
`0.036
`0.462**
`
`KW
`
`KT
`
`ERN
`
`KNR
`
`0.234**
`-0.319**
`-0.095
`0.693**
`
`-0.187**
`-0.606**
`0.493**
`
`0.043
`-0.196**
`
`-0.260**
`
`a: FT: flowering time; KL: kernel length; KW: kernel width; KT: kernel thickness; ERN: ear row number; KNR: kernel number per row; HKW: hundred kernel
`weight;
`*: Significant at 0.05 level;
`**: Significant at 0.01 level.
`
`doi:10.1371/journal.pone.0153428.t003
`
`PLOS ONE | DOI:10.1371/journal.pone.0153428 April 12, 2016
`
`5 / 12
`
`

`

`QTL Mapping for Maize Kernel Size
`
`model included: KL, ERN and HKW, which explained 19.2, 16.1 and 16.9% of the variation of
`kernel width, respectively. Similarly, KL, KNR and HKW can explain 13.2, 10.9 and 11.8% of
`the variation of KT, respectively (F = 47.0, p-value <2.2e-16, r2 = 0.45). Thus, agronomic traits
`can explain some of the variation in kernel size, but the variance explained is smaller than the
`heritability.
`
`QTL mapping results of kernel size and related traits
`A summary of the QTL detected across environments, including the positions, LOD scores,
`genetic effects (additive effects of aF and aM and dominance effect d), phenotypic variation
`explained (PVE) and the mean values of four different genotypes, were shown in Table 4. A
`
`Table 4. Estimated QTL locations and genetic effects affecting six traits using average data from two environments.
`
`Traitsa
`
`KL
`
`KW
`
`KT
`
`HKW
`
`KNR
`
`ERN
`
`QTL
`qKL3-1
`qKL5-1
`qKL7-1
`qKL7-2
`qKL10-1
`qKW5-1
`qKW6-1
`qKW7-1
`qKT1-1
`qKT5-1
`qHKW1-1
`qHKW1-2
`qHKW3-1
`qHKW5-1
`qHKW7-1
`qKNR1-1
`qKNR3-1
`qKNR5-1
`qKNR5-2
`qKNR5-3
`qKNR5-4
`qERN1-1
`qERN1-2
`qERN4-1
`qERN6-1
`qERN7-1
`qERN9-1
`qERN10-1
`
`Bin
`
`Position (cM)
`
`Left marker
`
`Right marker
`
`3.04/05
`5.06
`7.02/03
`7.03/04
`10.04/05
`5.03/04
`6.00/01
`7.00
`1.07/08
`5.01
`1.06/07
`1.11
`3.04
`5.03/04
`7.05
`1.02
`3.04
`5.01
`5.03/04
`5.04/05
`5.07/08
`1.10/11
`1.11
`4.08/09
`6.02/03
`7.02
`9.03
`10.01/02
`
`72
`129
`78
`121
`50
`61
`0
`42
`131
`7
`114
`233
`70
`61
`157
`33
`67
`0
`54
`70
`179
`206
`235
`121
`36
`76
`79
`13
`
`umc1347
`umc1680
`bnlg1792
`umc1408
`umc1053
`bnlg1700
`phi126
`umc1642
`bnlg1556
`umc1766
`umc1590
`phi227562
`umc1012
`bnlg1700
`umc2222
`bnlg1429
`umc1012
`umc1766
`bnlg1700
`umc1591
`bnlg2305
`bnlg1347
`phi227562
`umc2286
`umc1656
`phi034
`umc1700
`umc1319
`
`bnlg1957
`umc1019
`umc1567
`dupssr13
`umc1506
`umc2298
`umc1018
`bnlg2132
`phi039
`umc1365
`bnlg1556
`bnlg1006
`umc1347
`umc2298
`phi082
`bnlg1007
`umc1347
`umc1365
`umc2298
`umc1348
`zct389
`umc2100
`bnlg1006
`umc1051
`umc1887
`bnlg1792
`umc1691
`umc1576
`
`Genetic effectsb
`aM
`
`d
`
`aF
`
`PVE (%)c
`
`-0.03
`0.01
`0
`-0.09
`-0.07
`0.08
`-0.01
`-0.07
`-6.24
`7.20
`-0.47
`-0.42
`0.30
`0.46
`-0.11
`-0.55
`-0.19
`-0.91
`-0.19
`-0.86
`0.47
`-0.18
`0.19
`-0.02
`-0.04
`0.04
`-0.15
`0.22
`
`-0.06
`0.14
`-0.06
`-0.03
`-0.02
`0
`0.06
`0.04
`-4.22
`0.43
`-0.07
`0.38
`-0.08
`0.13
`-0.36
`0.93
`-1.02
`-0.53
`-0.94
`-0.31
`-0.80
`-0.05
`-0.17
`0.19
`-0.30
`-0.24
`-0.13
`0.01
`
`-0.04
`0.01
`-0.07
`0.01
`-0.02
`0.03
`-0.05
`0
`3.35
`3.19
`0.11
`-0.10
`-0.35
`0.18
`-0.28
`0.03
`0.08
`-0.12
`-0.33
`-0.15
`0.07
`-0.05
`-0.04
`0.11
`-0.01
`0.04
`0.02
`-0.07
`
`7.41
`17.94
`7.41
`8.14
`5.51
`9.15
`7.14
`9.54
`12.81
`11.12
`6.90
`6.72
`6.62
`8.23
`6.72
`6.22
`6.85
`5.83
`6.09
`5.42
`5.13
`5.10
`4.87
`5.19
`11.24
`5.94
`4.74
`5.97
`
`LOD
`
`4.79
`12.13
`4.87
`5.97
`4.04
`4.84
`4.23
`4.77
`6.19
`6.24
`4.09
`4.18
`4.58
`6.15
`4.77
`4.18
`4.43
`5.03
`4.55
`4.27
`4.35
`4.91
`4.06
`4.11
`9.4
`5.63
`4.59
`5.61
`
`a: Trait abbreviation as follow, KL: kernel length; KW: kernel width; KT: kernel thickness; ERN: ear row number; KNR: kernel number per row; HKW:
`hundred kernel weight.
`b: The genetic effects of aF and aM were the additive genetic effects of the two single crosses, D276×D72 and A188×Jiao51, respectively; the genetic
`effect of d was the dominance effect between the two single crosses.
`c: Phenotypic variation explained.
`
`doi:10.1371/journal.pone.0153428.t004
`
`PLOS ONE | DOI:10.1371/journal.pone.0153428 April 12, 2016
`
`6 / 12
`
`

`

`QTL Mapping for Maize Kernel Size
`
`total of 10 QTL were identified for kernel size, including 5 QTL for KL, 3 QTL for KW and 2
`QTL for KT. Single QTL of kernel size explained from 5.51% to 17.94% of the phenotypic vari-
`ation. Five QTL were identified for HKW which located on chromosomes 1, 3, 5 and 7, and
`single QTL explained from 6.62% to 8.23% of the phenotypic variation. Six QTL were identi-
`fied for KNR, which included 4 QTL on chromosome 5 and 1 each on chromosomes 1 and 3.
`Single QTL of KNR can explain from 5.13% to 6.85% of the phenotypic variation. Seven QTL
`were identified for ERN which located on chromosomes 1, 4, 6, 7, 9 and 10, and the largest
`QTL for ERN was located on chromosome 6 and explained 11.24% of the phenotypic
`variation.
`Positions of all detected QTL were marked in the linkage maps, and overlaps between QTL
`of kernel size with other traits were observed (Table 4 and Fig 1). The first overlapped QTL
`
`Fig 1. Genetic linkage maps and QTL identified in the four-way cross population.
`
`doi:10.1371/journal.pone.0153428.g001
`
`PLOS ONE | DOI:10.1371/journal.pone.0153428 April 12, 2016
`
`7 / 12
`
`

`

`QTL Mapping for Maize Kernel Size
`
`located on chromosome 5 (bin 5.03/04). In this region, qKW5-1, which conferred the kernel
`width, shared the same flanking markers with qHKW5-1 and also with qKNR5-2. The second
`region located on chromosome 3 (bin 3.04/05). In this region, qHKW3-1 shared the same
`flanking markers with qKNR3-1, which had the largest effect for KNR. Moreover, one QTL for
`KL, qKL3-1, was also detected, which shared the same flanking marker umc1347 with qHKW3-
`1 and qKNR3-1. Other region with closely linked QTL was also identified on chromosome 7
`(bin 7.02/03). In this region, both QTL for KL (qKL7-1) and ERN (qERN7-1) were identified,
`and they shared the same marker bnlg1792 within the QTL region. Despite the significant cor-
`relations between kernel size traits (i.e., KL, KW and KW), we did not detect any overlapping
`QTL region for the three kernel-size traits.
`
`Discussion
`The trade-off between kernel size traits in maize
`Grain seed is a life-history trait, and the trade-off of grain seed and related traits has widely
`reported in many plant species [2,37,44,45]. However, few studies have addressed the genetic
`mechanism behind trade-off of the factors involved in maize kernel development by taking
`into account life-history traits. In the present study, complex genetic mechanism behind the
`trade-off of seed traits in the four-way cross population was observed. On one hand, over-
`lapped QTL for kernel size (i.e., qKL3-1, qKL7-1 and qKW5-1) and yield-components were
`observed, and most of them had the same direction of additive effects (aF and aM) (Table 4),
`which indicated the allele that increases kernel size is from the same natural accession, indicat-
`ing past occurrence of directional selection for kernel size and yield components. On the other
`hand, kernel size (i.e., KW and KT) showed significant negative correlation with ERN and
`KNR (Table 3), which implied the potential trade-off behind them. However, there is little evi-
`dence for overlap in their genetic architecture since no common QTL between the traits were
`detected.
`
`Comparison with published QTL/gene
`In the present study, we mapped 10 QTL for kernel size, with three of them (qKW5-1, qKL3-1
`and qKL7-1) had consistent co-localization or adjacent to QTL for one of the components of
`maize yield (Table 4). We compared the QTL with published kernel-size QTL, and overlapped
`QTL independent of the genetic background were identified.
`qKW5-1 with flanking markers bnlg1700 and umc2298 in the present study, shared the
`same QTL region with CQTL5-1, an common QTL for kernel width in multiple connected RIL
`populations in maize [8]; In this region, the other QTL for kernel width (i.e., qKW5) was also
`identified in an independent QTL mapping of kernel-size [9] (S1 Table). More importantly, we
`identified qKW5-1 overlapped with the qHKW5-1 (the QTL with the largest effect for 100-ker-
`nel weight in the present study), similar results were also observed by Li et al. [8] and Liu et al.
`[9], who also identified the QTL for kernel-width overlapped with kernel weight in this region.
`Within qKW5-1 region, ZmGW2-Chr5 conferring kernel size in maize has been identified and
`is perhaps one of candidate genes for the QTL [51]. Therefore, it could be concluded that this
`genomic region is very important for grain yield since the QTL has the stable expression across
`different genetic background.
`qKL7-1 located in bin 7.02/03 is another important region for the genetic control of grain
`yield and kernel traits. In this region, cluster QTL for kernel-size and yield-related trait were
`also identified in independent studies. For example, Li et al. [8] found one common QTL
`(CQTL7-1) conferring kernel weight, kernel width and thickness in multiple connected RIL
`populations in maize. Peng et al. [7] identified two QTL, Qqknpp7 and Qqgypp7, conferring
`
`PLOS ONE | DOI:10.1371/journal.pone.0153428 April 12, 2016
`
`8 / 12
`
`

`

`QTL Mapping for Maize Kernel Size
`
`kernel number and grain yield per plant, respectively. Other kernel-size QTL in the present
`study with known QTL/gene included qKL5-1 with gln1-3 [20] and Yqknpp5 [7], qKT1-1 with
`CQTL1-2 [8], and qKT5-1 with qKT5-1 [9]. The consistency of QTL/gene in independent study
`implied the common genetic basis for these traits (S1 Table).
`
`Joint analysis for multiple related traits
`In this study, QTL analysis was performed on the six traits, respectively. In total, 10 QTL for
`kernel size (KL, KW and KT) and 18 QTL for other three traits (ERN, KNR and HKW) were
`detected. In fact, these traits were highly related (Table 3). Multiple-trait analysis took into
`account the correlated structure of multiple traits, and could improve the statistical power of
`the QTL detection and the precision of parameter estimation [52]. However, there were seldom
`studies focused on the QTL mapping methods for jointly analyzing multiple traits in four-way
`crosses till now on. We will try to develop a statistical method for joint analysis on four-way
`cross populations in the future.
`
`Implications for molecular-assisted selection (MAS) breeding
`In the present study, a total of 10 significant QTL for kernel size were identified, which ranged
`from two for KT to five for KL. However, these QTL seemed to be independent genetic regula-
`tion of seed size since no consistent QTL were observed. These QTL could be valuable because
`it means that improvement in one trait can be accomplished without a corresponding decrease
`in the other. Here, we also found that at least three QTL (i.e., qKL3-1, qKL7-1 and qKW5-1)
`with stable expression across kernel size and at least one of the other kernel related traits, and
`they had the same direction of the additive effects. These QTL may imply the genetic regulation
`of seed size and the components of maize yield, and may have high values using MAS to
`improve yield in maize.
`
`Supporting Information
`S1 Fig. The distributions of the seven traits across the two environments. A, kernel length;
`B, kernel width; C, kernel thickness; D, 100 kernel weight; E, number of rows per ear; F, num-
`ber of kernels per row; G, flowing time.
`(TIF)
`
`S1 Dataset. The original genotypes, phenotypes and linkage maps data of the four-way
`cross population used in this study.
`(CDQ)
`
`S1 Table. Comparison of the QTL identified in the present study with previous reported
`QTL/genes from the literature.
`(DOCX)
`
`Acknowledgments
`This work was supported by Henan Basic Research Program of China (142102110051).
`
`Author Contributions
`Conceived and designed the experiments: JW JD. Performed the experiments: JC JD ZL SL RH
`YL XL HC BZ SW WC. Analyzed the data: LZ JC JD. Contributed reagents/materials/analysis
`tools: JD JC JW. Wrote the paper: JD LZ JC JW.
`
`PLOS ONE | DOI:10.1371/journal.pone.0153428 April 12, 2016
`
`9 / 12
`
`

`

`QTL Mapping for Maize Kernel Size
`
`3.
`
`8.
`
`9.
`
`References
`1.
`Tester M, Langridge P. Breeding technologies to increase crop production in a changing world. Sci-
`ence. 2010; 327: 818–822. doi: 10.1126/science.1183700 PMID: 20150489
`2. Gupta PK, Rustgi S, Kumar N. Genetic and molecular basis of grain size and grain number and its rele-
`vance to grain productivity in higher plants. Genome. 2006; 49: 565–571. PMID: 16936836
`Fan C, Xing Y, Mao H, Lu T, Han B, Xu C, et al. GS3, a major QTL for grain length and weight and
`minor QTL for grain width and thickness in rice, encodes a putative transmembrane protein. Theor Appl
`Genet. 2006; 112: 1164–1171. PMID: 16453132
`4. Song XJ, Huang W, Shi M, Zhu MZ, Lin HX. A QTL for rice grain width and weight encodes a previously
`unknown RING-type E3 ubiquitin ligase. Nat Genet. 2007; 39: 623–630. PMID: 17417637
`5. Chen J, Ding J, Li H, Li Z, Sun X, Li J, et al. Detection and verification of quantitative trait loci for resis-
`tance to Fusarium ear rot in maize. Mol Breed. 2012; 30: 1649–1656.
`6. Ding J, Zhang L, Chen J, Li X, Li Y, Cheng H, et al. Genomic dissection of leaf angle in maize (Zea
`mays L.) using a four-way cross mapping population. PLoS One. 2015; 10: e0141619. doi: 10.1371/
`journal.pone.0141619 PMID: 26509792
`7. Peng B, Li Y, Wang Y, Liu C, Liu Z, Tan W, et al. QTL analysis for yield components and kernel-related
`traits in maize across multi-environments. Theor Appl Genet. 2011; 122: 1305–1320. doi: 10.1007/
`s00122-011-1532-9 PMID: 21286680
`Li C, Li Y, Sun B, Peng B, Liu C, Liu Z, et al. Quantitative trait loci mapping for yield components and
`kernel-related traits in multiple connected RIL populations in maize. Euphytica. 2013; 193: 303–316.
`Liu Y, Wang L, Sun C, Zhang Z, Zheng Y, Qiu F. Genetic analysis and major QTL detection for maize
`kernel size and weight in multi-environments. Theor Appl Genet. 2014; 127: 1019–1037. doi: 10.1007/
`s00122-014-2276-0 PMID: 24553962
`Takano-Kai N, Jiang H, Kubo T, Sweeney M, Matsumoto T, Kanamori H, et al. Evolutionary history of
`GS3, a gene conferring grain length in rice. Genetics. 2009; 182: 1323–1334. doi: 10.1534/genetics.
`109.103002 PMID: 19506305
`Zhang X, Wang J, Huang J, Lan H, Wang C, Yin C, et al. Rare allele of OsPPKL1 associated with grain
`length causes extra-large grain and a significant yield increase in rice. Proc Natl Acad Sci. 2012; 109:
`21534–21539. doi: 10.1073/pnas.1219776110 PMID: 23236132
`12. Shomura A, Izawa T, Ebana K, Ebitani T, Kanegae H, Konishi S, et al. Deletion in a gene associated
`with grain size increased yields during rice domestication. Nat Genet. 2008; 40: 1023–1028. doi: 10.
`1038/ng.169 PMID: 18604208
`Li Y, Fan C, Xing Y, Jiang Y, Luo L, Sun L, et al. Natural variation in GS5 plays an important role in regu-
`lating grain size and yield in rice. Nat Genet. 2011; 43: 1266–1269. doi: 10.1038/ng.977 PMID:
`22019783
`14. Wang S, Wu K, Yuan Q, Liu X, Liu Z, Lin X, et al. Control of grain size, shape and quality by OsSPL16
`in rice. Nat Genet. 2012; 44: 950–954. doi: 10.1038/ng.2327 PMID: 22729225
`15. Wang Y, Xiong G, Hu J, Jiang L, Yu H, Xu J, et al. Copy number variation at the GL7 locus contributes
`to grain size diversity in rice. Nat Genet. 2015; 47: 944–948. doi: 10.1038/ng.3346 PMID: 26147619
`16. Wang S, Li S, Liu Q, Wu K, Zhang J, Wang S, et al. The OsSPL16-GW7 regulatory module determines
`grain shape and simultaneously improves rice yield and grain quality. Nat Genet. 2015; 47: 949–954.
`doi: 10.1038/ng.3352 PMID: 26147620
`17. Cheng WH, Taliercio EW, Chourey PS. The miniature1 seed locus of maize encodes a cell wall inver-
`tase required for normal development of endosperm and maternal cells in the pedicel. Plant Cell. 1996;
`8: 971–983. PMID: 12239408
`18. Schmidt RJ, Ketudat M, Aukerman MJ, Hoschek G. Opaque-2 is a transcriptional activator that recog-
`nizes a specific target site in 22-kD zein genes. Plant Cell. 1992; 4: 689–700. PMID: 1392590
`19. Bhave MR, Lawrence S, Barton C, Hannah LC. Identification and molecular characterization of
`shrunken-2 cDNA clones of maize. Plant Cell. 1990; 2: 581–588. PMID: 1967077
`20. Martin A, Lee J, Kichey T, Gerentes D, Zivy M, Tatout C, et al. Two cytosolic glutamine synthetase iso-
`forms of maize are specifically involved in the control of grain production. Plant Cell. 2006; 18: 3252–
`3274. PMID: 17138698
`21. Wang G, Wan