BSA for drought QTL identification
Field Crops Research 134 (2012) 185–192
Contents lists available at SciVerse ScienceDirect
Field Crops Research
journal homepage: www.elsevier.com/locate/fcr
Bulk segregant analysis: “An effective approach for mapping consistent-effect
drought grain yield QTLs in rice”
Prashant Vikram a , B.P. Mallikarjuna Swamy a , Shalabh Dixit a , Helaluddin Ahmed a,1 ,
M.T. Sta Cruz a , Alok K. Singh b , Guoyou Ye c , Arvind Kumar a,∗
a
Plant Breeding Genetics & Biotechnology Division, International Rice Research Institute, DAPO Box 7777, Metro Manila, Philippines
Department of Genetics and Plant Breeding, V.B.S. Purvanchal University, Jaunpur, India
c
Crop Research Informatics Laboratory, International Rice Research Institute, DAPO Box 7777, Metro Manila, Philippines
b
a r t i c l e
i n f o
Article history:
Received 26 February 2012
Received in revised form 30 May 2012
Accepted 30 May 2012
Keywords:
QTLs
Quantitative trait loci
Selective genotyping
Bulk segregant analysis
Grain yield under drought
a b s t r a c t
Mapping QTLs for grain yield under drought in rice involves phenotyping and genotyping of large mapping populations. The huge cost incurred in genotyping could be considered as a bottleneck in the
process. Whole population genotyping (WPG), selective genotyping (SG), and bulk segregant analysis
(BSA) approaches were employed for the identification of major grain-yield QTLs under drought in rice in
the past few years. The efficiency of different QTL mapping approaches in identifying major-effect grainyield QTLs under drought in rice was compared using phenotypic and genotypic data of two recombinant
inbred line populations, Basmati334/Swarna and N22/MTU1010. All three genotyping approaches were
efficient in identifying consistent-effect QTLs with an additive effect of 10% or more. Comparative analysis revealed that SG and BSA required 63.5% and 92.1% fewer data points, respectively, than WPG in
the N22/MTU1010 F3:4 mapping population. The BSA approach successfully detected consistent-effect
drought grain-yield QTLs qDTY1.1 and qDTY8.1 detected by WPG and SG. Unlike SG, BSA did not lead to an
upward estimation of the additive effect and phenotypic variance. The results clearly demonstrate that
BSA is the most efficient and effort saving genotyping approach for identifying major grain yield QTLs
under drought.
© 2012 Elsevier B.V. All rights reserved.
1. Introduction
Advances in molecular marker technology have enabled fasttrack improvement of crop plants in recent years. Marker-based
approaches, including marker-assisted backcrossing and QTL pyramiding, have been applied in cereals for improving the tolerance
of biotic and abiotic stresses (Collard and Mackill, 2008; Ye and
Smith, 2010). Drought is an important abiotic stress causing huge
losses in rice yields. Progress in breeding rice for drought tolerance could be made more efficient by applying marker-assisted
breeding (Bernier et al., 2007). To unveil the genetic basis of a complex trait such as grain yield under drought, it is a prerequisite to
Abbreviations: BSA, bulk segregant analysis; MAS, marker-assisted selection;
QTLs, quantitative trait loci; SG, selective genotyping; WPG, whole population genotyping.
∗ Corresponding author. Tel.: +63 2 580 5600x2586; fax: +63 2 580 5699.
E-mail addresses: p.vikram@irri.org (P. Vikram), m.swamy@irri.org
(B.P.M. Swamy), s.dixit@cgiar.org (S. Dixit), helaluddinahmed@hotmail.com
(H. Ahmed), m.stacruz@cgiar.org (M.T.S. Cruz), alok6619@gmail.com (A.K. Singh),
G.Ye@cgiar.org (G. Ye), akumar@cgiar.org (A. Kumar).
1
Present address: Plant Breeding Division, Bangladesh Rice Research Institute,
Joydebpur, Gazipur 1701, Bangladesh.
0378-4290/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.fcr.2012.05.012
genotype and phenotype mapping populations consisting of a large
number of individuals. The costs associated with large-scale phenotyping and genotyping are a bottleneck to conducting a study on
several populations at any particular time in identifying QTLs with
consistent effects across multiple environments and genetic backgrounds. With little alternative available to phenotype in different
seasons/environments, the costs involved with genotyping could
be effectively reduced by successfully applying different genotyping approaches.
In a model cereal crop such as rice (Oryza sativa) with a genome
size of 389 Mb, it is a common practice to use 200–400 lines (recombinant inbred lines, backcross inbred lines, doubled haploids) as a
mapping population. A dense map covering all 12 chromosomes
with an average genetic distance of 10–15 cM has been developed
to identify the precise QTLs. In most of the QTL mapping experiments, the whole genome is scanned (Gomez et al., 2010; Lanceras
et al., 2004; Xu et al., 2005) to ultimately find only a few significant markers showing association with the trait under study.
To reduce the extra burden and costs associated with genotyping of large mapping populations, alternative strategies have been
proposed. A trait-based genotypic analysis called “selective genotyping” (SG) was suggested by Lebowitz et al. (1987), in which
progenies are categorized into distinct classes based on the trait
186
P. Vikram et al. / Field Crops Research 134 (2012) 185–192
values and marker allele frequencies are compared between the
classes. In this approach, a subset of the population is used for genotyping instead of the whole population (Lander and Botstein, 1989).
This approach was used for mapping major-effect drought grainyield QTL qDTY12.1 using only 169 (38.7%) lines from a population
size of 436 (Bernier et al., 2007); later, this QTL was reconfirmed
with only 4.5% of the lines from the whole population (Navabi et al.,
2009). Another time and effort saving approach is bulk segregant
analysis (BSA) (Michelmoore et al., 1991). In BSA, DNA of progenies
corresponding to the phenotypic extremes is extracted and pooled.
Therefore, only two pools of extreme lines along with the parents
are genotyped for the identification of markers linked with the trait
of interest. BSA was first applied to the identification of markers
linked with disease resistance. Initially, it was applied to the diseases in which resistance was mostly governed by major genes,
usually qualitative in nature (Michelmoore et al., 1991). Recently,
BSA has been applied for quantitative traits also such as QTLs for
heat tolerance in rice (Zhang et al., 2009), salt tolerance in Egyptian cotton (El-Kadi et al., 2006), and drought tolerance in wheat
and maize (Altinkut and Gozukirmizi, 2003; Quarrie et al., 1999;
Kanagaraj et al., 2010; Venuprasad et al., 2009; Vikram et al., 2011),
as well as QTLs for grain yield under drought in maize (Quarrie et al.,
1999). BSA has been used for the identification of QTLs associated
with high grain yield under drought in rice (Venuprasad et al., 2009,
2011; Vikram et al., 2011).
Whole population genotyping involves markers from all the 12
chromosomes so that they represent whole genome and genetic
background is taken in to account while QTL analysis. Contrarily,
BSA and SG (selective genotyping) approaches may not consider
all recombination options during QTL identification due to the
unavailability of additional marker data on the same chromosome
and other chromosomes. However, these approaches have been
proven to be powerful enough to identify major QTLs that are worthy for MAS (Bernier et al., 2007; Venuprasad et al., 2009; Vikram
et al., 2011). BSA and SG might fail to detect QTLs with smaller
effects because only the extreme or tail lines are used for analysis. Also, interactions between different loci are least likely to be
detected through these approaches. QTL analysis through BSA or
SG approaches does not takes in to account the background so that
they might tilt upward the magnitude of phenotypic variance, LOD
(logarithm of odds) score, and additive effects.
The study was undertaken to perform a comparative analysis of
SG and BSA in identifying major-effect QTLs for grain yield under
drought and compare the fluctuations in phenotypic variance, additive effect, LOD, and level of significance values obtained in BSA
and SG compared with WPG (whole population genotyping). WPG,
SG, and BSA approaches were compared in a population derived
from a traditional variety (Basmati334) and a popular lowland rice
variety (Swarna). Comparative analysis of these approaches might
be biased if only one population is taken into account. To validate the results and perform a comprehensive analysis, phenotypic
and genotypic data of a rice population applied to QTL mapping in
earlier studies were used (Vikram et al., 2011).
2. Materials and methods
The study was conducted at the International Rice Research
Institute (IRRI), Los Baños, Laguna, Philippines. An F3:4 Basmati
334/Swarna population was phenotyped for grain yield under
drought and genotyped through WPG, SG, and BSA. Basmati334 is a
drought-tolerant local landrace whereas Swarna is a popular lowland rice cultivar in India (Sivaranjani et al., 2010; Verulkar et al.,
2010). The phenotypic and genotypic data of the N22/MTU1010
population were also used for comparison of WPG, SG, and
BSA (Vikram et al., 2011). Basmati334/Swarna population was
screened under drought stress in wet season 2008 (WS2008) and
N22/MTU1010 in dry season of 2009 (DS2009). WS2008 experiments were shown on June 17, 2008 whereas, DS2010 experiments
on December 9, 2009.
2.1. Phenotyping of Basmati334/Swarna population
The F3:4 Basmati334/Swarna RIL population was phenotyped
for grain yield under drought stress as well as irrigated/non-stress
conditions (Supplementary Table 1). The numbers of lines used
for drought screening were 204 during WS2008 and 367 during
DS2010. The 204 lines used in WS2008 were a subset of 367 lines
that were phenotyped in DS2010. Screening for grain yield under
drought in the lowland rice ecosystem was carried out at IRRI using
a standard phenotyping protocol (Kumar et al., 2008; Venuprasad
et al., 2007). Data for grain yield (g m−2 converted to kg ha−1 ), days
to 50% flowering (DTF), plant height (PH), biomass (BIO), and harvest index (HI) were recorded.
2.2. Genotyping of Basmati334/Swarna population
Leaf samples were collected from a whole F4 plot of the nonstress experiment and bulked. DNA was extracted by the modified
CTAB (cetyl tri-methyl ammonium bromide) method Murray and
Thompson, 1980). DNA was quantified on 0.8% agarose gel to adjust
the concentration to 25 ng L−1 . Quantified DNA was subjected to
PCR amplification with a 15-L reaction mixture involving 50 ng
DNA, 1× PCR buffer, 100 M dNTPs, 250 M primers, and 1 unit
Taq polymerase enzyme. Eight percent non-denaturing PAGE was
used for the resolution of PCR products (Sambrook et al., 1989). A
parental polymorphism survey between Basmati334 and Swarna
was carried out with 880 simple sequence repeat (SSR) markers
(Temnykh et al., 2001; McCouch et al., 2002; IRGSP, 2005). Genotyping of the Basmati334/Swarna population of 367 lines was done
by 71 polymorphic SSR markers spread throughout the 12 chromosomes of the rice genome. BSA was also done for QTL identification
with only 10% of the lines from the population (5% with high yield
and 5% with low yield under drought stress). Grain yield data from
the stress trial of DS2010 were used for selecting lines to constitute BSA. DNA of all the selected lines was quantified and bulked
in equal quantity to prepare high- and low-yield bulks. DNA of
these two bulks was screened with 203 polymorphic SSR markers along with parents – Basmati334 and Swarna (Fig. 1). A similar
procedure for BSA in the N22/MTU1010 population was carried out
(Vikram et al., 2011). Polymorphic markers from all the 12 chromosomes were selected at equal distance so that they represent whole
genome. SG was carried out with the same number of markers used
in WPG. Both RILpulations-Basmati334/Swarna and N22/MTU1010
presented in the study were analyzed through whole genome and
BSA both.
2.3. Statistical analysis
Statistical analysis was performed through SASv9.1.3 (SAS
Institute Inc., 2004). The REML algorithm of PROC MIXED of SAS
v9.1.3 was used for the determination of mean values of entries
using a model in which lines were treated as a fixed effect and
replications and blocks within replications as random. Inbred line
mean-based broad-sense heritability (H) was computed as:
H=
Vg
Vg + Ve /r
where Vg is genetic variance, Ve is error variance, and r is the number of replications.
P. Vikram et al. / Field Crops Research 134 (2012) 185–192
187
Fig. 1. Bulked segregant analysis strategy used for identifying QTLs for grain yield under drought in rice.
2.4. Linkage map construction and QTL analysis
A genetic linkage map of the Basmati334/Swarna population
was constructed through Mapdisto software using the Kosambi
map function and an LOD score of 3.0 as the threshold (Kosambi,
1944; Lorieux, 2007). All the non-linked markers were placed
on their respective chromosomes based on the physical position
(IRGSP, 2005). A similar procedure was followed by earlier workers
(Amrawathi et al., 2008). QTL network software v2.1 was used for
mixed model-based composite interval mapping (Yang et al., 2008).
In this procedure, candidate marker intervals were determined and
selected intervals were used as a co-factor in a one-dimensional
genome scan. The significance level for the determination of candidate intervals, putative QTL detection, and QTL effects was set at a
probability level of 0.05. The threshold was determined using 1000
permutations. For the genome scan, a window size of 10 cM and
walk speed of 1 cM were followed. QTL analysis results were confirmed with QGene software to validate the magnitudes of the QTLs
identified through three different strategies. Phenotypic variance
and LOD of the QTLs were estimated through composite interval
mapping using QGene software (Nelson, 1997). A LOD value of 2.5
was considered as threshold to determine significance of QTLs. Similar threshold value was used because all QTLs used in this analysis
were found significant in analysis through QTL network software.
2.5. Analysis for SG and BSA
QTL analysis with a lesser number of lines in both populations
was carried out to work out the efficiency of SG and BSA. For
analysis through SG, 36.5% of the lines (10% highest yielding, 10%
lowest yielding, and 16.5% random lines) were used (Bernier et al.,
2007). The linkage map constructed using WPG was also used to
locate the QTLs identified by SG analysis. QTL analysis for BSA was
carried out with three markers – RM210, RM223, and RM339 –
in the Basmati334/Swarna population and with eight markers –
RM315, RM431, RM212, RM3825, RM11943, RM12023, RM12091,
and RM12146 – in the N22/MTU1010 population.
To estimate the size of population for QTL analysis for grain
yield under drought, QTL analysis with 90%, 80%, 70%, 60%, and
50% population size was performed in both populations. For this
purpose, a whole population was sorted based on grain yield under
drought. Equal numbers of individuals were omitted from highyielding lines, low-yielding lines, and lines with medium yield from
the stress trials to make sure that the lines selected were random
and distribution remained normal.
3. Results
The results of QTL identification for grain yield, days to 50% flowering, plant height, and harvest index under drought by WPG in
the Basmati334/Swarna and N22/MTU1010 populations are presented in Supplementary Tables 2 and 3. QTLs for grain yield under
drought identified through WPG, SG, and BSA in both populations
are presented in Table 1.
3.1. QTLs identified for grain yield under drought with WPG
A consistent-effect QTL (qDTY8.1 ) for grain yield under drought
located between the marker intervals RM339 and RM210 on chromosome 8 was detected in the Basmati 334/Swarna population.
This QTL explained phenotypic variance (R2 ) of 15.6% and 7.1%,
with an additive effect (additive effect as percent of trial mean)
of 16.8% and 22.6%, during WS2008 (wet season 2008) and DS2010
(dry season 2010), respectively (Table 1). Two QTLs, qDTY1.1 and
qDTY10.1 , were found significant for grain yield under drought in
the N22/MTU1010 population. qDTY1.1 explained phenotypic variance of 11.9% and 4.5%, with an additive effect of 18.1% and 10.0%
during DS2009 and DS2010, respectively. qDTY10.1 explained phenotypic variance of 5.2% and 3.2% and an additive effect of 13.2%
and 12.6% during DS2009 and DS2010, respectively (Table 1).
3.2. QTLs identified for grain yield under drought with SG
qDTY8.1 in Basmati334/Swarna and qDTY1.1 in N22/MTU1010
were detected by the SG approach. The QTL qDTY8.1 explained phenotypic variance of 21.9% and 16.5%, with an additive effect of
25.4% and 39.8% during WS2008 and DS2010, respectively (Table 1).
The QTL qDTY1.1 explained phenotypic variance of 17.7% and 12.9%
188
P. Vikram et al. / Field Crops Research 134 (2012) 185–192
Table 1
QTLs for grain yield identified through SG and BSA and effect of WPG, SG, and BSA approach on F-value, additive effect, LOD, and R2 .
Population
Approach
WPG
Basmati 334/Swarna
SG
BSA
WPG
WPG
N22/MTU1010
SG
SG
BSA
a
b
c
Season
WS2008
DS2010
WS2008
DS2010
WS2008
DS2010
DS2009
DS2010
DS2009
DS2010
DS2009
DS2010
DS2009
DS2010
DS2009
DS2010
QTL
Chr.
Peak intervala
qDTY8.1
8
RM339–RM210
qDTY8.1
8
RM339–RM210
qDTY8.1
8
RM223–RM210
qDTY1.1
1
RM11943–RM12091
qDTY10.1
10
RM216–RM304
qDTY1.1
1
RM11943–RM431
qDTY1.1a
1
RM12091–RM12146
qDTY1.1
1
RM11943–RM431
F-value
14.1
15.4
10.7
15.8
9.5
21.1
36.7
18.5
19.9
15.7
22.8
15.3
–
16.1
34.7
13.9
AE (%)c
LOD
b
−16.8
−22.6
−25.4
−39.8
−12.2
−20.8
18.1
10.0
−13.2
−12.6
28.8
10.3
–
10.8
18.3
10.1
7.1
5.1
4.3
5.6
2.2
4.7
9.3
3.6
3.9
2.6
5.6
3.9
–
2.8
9.3
3.6
R2
15.6
7.1
21.9
16.5
4.9
5.8
11.9
4.5
5.2
3.2
17.7
12.9
–
9.4
11.9
4.5
Peak interval, peak of the QTL interval was between these marker intervals. R2 , phenotypic variance explained by the QTL.
Negative sign indicates that allele contribution is from susceptible parent Swarna.
AE% (additive effect %), additive effect as the percentage of trial mean [additive effect/trial mean × 100].
during DS2009 and DS2010, respectively. Additive effects contributed by qDTY1.1 were 28.8% and 10.3% during DS2009 and
DS2010, respectively.
3.3. QTLs identified for grain yield under drought with BSA
qDTY8.1 in Basmati334/Swarna and qDTY1.1 in N22/MTU1010
were detected by BSA. RM210 was identified to be associated with
grain yield under drought in the Basmati334/Swarna population
via BSA. Three markers were run on the whole population to detect
qDTY8.1 . It explained phenotypic variance of 4.9% and 5.8% and
additive effects of 12.2% and 20.8% during WS2008 and DS2010,
respectively (Table 1). In the case of qDTY1.1 , RM315 and RM431
were detected as linked markers for grain yield under drought in the
N22/MTU1010 population. Eight polymorphic markers were run
on the whole population to determine QTL boundaries (Fig. 1). The
phenotypic variance explained by qDTY1.1 in DS2009 and DS2010
was 11.9% and 4.5%, respectively. Additive effects in DS2009 and
DS2010 were 18.3% and 10.1%, respectively (Table 1).
3.4. Magnitude and efficiency of QTLs in different genotyping
approaches
All three approaches successfully detected two consistenteffect QTLs, qDTY8.1 and qDTY1.1 , in Basmati334/Swarna and
N22/MTU1010 populations, respectively (Table 1). qDTY10.1 , which
was found in the N22/MTU1010 population, was detected only in
WPG. The F-value of qDTY8.1 was lesser in WS2008 and greater
in DS2010 in SG in comparison with WPG. The additive effect of
qDTY8.1 was highest in SG, followed by WPG and BSA in WS2008
and DS2010. The LOD value was highest in WPG and SG in WS2008
and DS2010, respectively. The phenotypic variance was highest in
SG, followed by WPG and BSA in WS2008 and DS2010, respectively
(Table 1). Comparative analysis of the parameters of qDTY1.1 reveals
that the F-value is lower in SG than in WPG (Table 1). The additive
effect in SG was highest, followed by BSA and WPG in DS2009 as
well as DS2010 (Table 1). It is notable that the additive effect was
least in WPG in both years. The LOD value of qDTY1.1 was the same
in WPG and BSA in DS2009 and DS2010. But LOD estimated by SG
was lower during DS2009 and higher in DS2010 than LOD estimated with WPG and BSA. The phenotypic variance of qDTY1.1 was
the same in WPG and BSA, whereas it was greater in SG in DS2009
and DS2010 (Table 1).
3.5. QTL effect analysis with different population sizes
QTL analysis for grain yield under drought was carried out with
different population sizes. qDTY1.1 identified with phenotypic data
(under drought stress) of DS2009 with an original population size
of 362 was also successfully detected in a lesser population size of
218 lines. The same QTL was detected with phenotypic data (under
drought stress) of DS2010 in the original population but failed to be
detected in a reduced population size (Table 2). A similar trend was
observed for qDTY8.1 in the Basmati 334/Swarna population during
DS2008. Additive effect decreased with a decrease in population
size. The trend in DS2010 for this QTL was similar for a decrease
in population size up to 30%, but thereafter it became constant
up to a 50% decrease in population size and then decreased again
(Table 2). The additive effect of qDTY10.1 did not show any pattern
with respect to population size.
4. Discussion
Marker-assisted selection using consistent-effect drought grain
yield QTLs is could be an alternative approach for improving rice
grain yield under drought stress situations (Swamy et al., 2011;
Swamy and Kumar, 2011; Vikram et al., 2012). QTLs have been
identified in the past few years for different drought-related traits
through phenotyping and genotyping of large mapping populations
(Babu et al., 2003; Gomez et al., 2010). Phenotyping of mapping populations under managed drought stress condition that
too at target environments is a difficult task and involves large
expense and efforts and is considered a significant bottleneck.
Over last many years, very limited progress has been made on
the development of high-throughput, cost-effective phenotyping
methodologies. In the process, genotyping of populations with a
large number of markers for each of the 12 rice chromosomes was
followed. Polymorphic markers were selected at equal distances
on all the chromosomes and used for a whole-genome scan. Trait
selection is another important concern in drought molecular breeding. QTLs for a number of physiological and morphological traits
have been identified but they have limited implications in breeding drought-tolerant rice varieties. Large-effect QTLs for grain yield
under drought have been recently identified (Bernier et al., 2007;
Venuprasad et al., 2009) and their introgression is in process at
IRRI (IRRI, unpublished). Genotyping of different populations and
their screening in multiple environments could be made effective
through an effort and time saving genotyping strategy that can be
applied to multiple populations simultaneously. Therefore, three
189
P. Vikram et al. / Field Crops Research 134 (2012) 185–192
Table 2
Comparative analysis of QTL effects in different population sizes with a decrease in population size.
Population
Basmati334/Swarna
QTL
Population size
Year
F-value
AE (%)a
qDTY8.1
183 (∼90%)
331 (∼90%)
162 (∼80%)
295 (∼80%)
141 (∼70%)
259 (∼70%)
120 (∼60%)
223 (∼60%)
99 (∼50%)
187 (∼50%)
78 (∼40%)
151 (∼40%)
WS2008
DS2010
WS2008
DS2010
WS2008
DS2010
WS2008
DS2010
WS2008
DS2010
WS2008
DS2010
26.7
24.1
28.6
18.0
27.7
17.1
17.1
11.5
13.4
11.4
8.2
8.8
−15.8
−23.2
−15.6
−17.2
−14.9
−16.9
−11.3
−17.1
−11.0
−17.1
−9.5
−15.7
Contents lists available at SciVerse ScienceDirect
Field Crops Research
journal homepage: www.elsevier.com/locate/fcr
Bulk segregant analysis: “An effective approach for mapping consistent-effect
drought grain yield QTLs in rice”
Prashant Vikram a , B.P. Mallikarjuna Swamy a , Shalabh Dixit a , Helaluddin Ahmed a,1 ,
M.T. Sta Cruz a , Alok K. Singh b , Guoyou Ye c , Arvind Kumar a,∗
a
Plant Breeding Genetics & Biotechnology Division, International Rice Research Institute, DAPO Box 7777, Metro Manila, Philippines
Department of Genetics and Plant Breeding, V.B.S. Purvanchal University, Jaunpur, India
c
Crop Research Informatics Laboratory, International Rice Research Institute, DAPO Box 7777, Metro Manila, Philippines
b
a r t i c l e
i n f o
Article history:
Received 26 February 2012
Received in revised form 30 May 2012
Accepted 30 May 2012
Keywords:
QTLs
Quantitative trait loci
Selective genotyping
Bulk segregant analysis
Grain yield under drought
a b s t r a c t
Mapping QTLs for grain yield under drought in rice involves phenotyping and genotyping of large mapping populations. The huge cost incurred in genotyping could be considered as a bottleneck in the
process. Whole population genotyping (WPG), selective genotyping (SG), and bulk segregant analysis
(BSA) approaches were employed for the identification of major grain-yield QTLs under drought in rice in
the past few years. The efficiency of different QTL mapping approaches in identifying major-effect grainyield QTLs under drought in rice was compared using phenotypic and genotypic data of two recombinant
inbred line populations, Basmati334/Swarna and N22/MTU1010. All three genotyping approaches were
efficient in identifying consistent-effect QTLs with an additive effect of 10% or more. Comparative analysis revealed that SG and BSA required 63.5% and 92.1% fewer data points, respectively, than WPG in
the N22/MTU1010 F3:4 mapping population. The BSA approach successfully detected consistent-effect
drought grain-yield QTLs qDTY1.1 and qDTY8.1 detected by WPG and SG. Unlike SG, BSA did not lead to an
upward estimation of the additive effect and phenotypic variance. The results clearly demonstrate that
BSA is the most efficient and effort saving genotyping approach for identifying major grain yield QTLs
under drought.
© 2012 Elsevier B.V. All rights reserved.
1. Introduction
Advances in molecular marker technology have enabled fasttrack improvement of crop plants in recent years. Marker-based
approaches, including marker-assisted backcrossing and QTL pyramiding, have been applied in cereals for improving the tolerance
of biotic and abiotic stresses (Collard and Mackill, 2008; Ye and
Smith, 2010). Drought is an important abiotic stress causing huge
losses in rice yields. Progress in breeding rice for drought tolerance could be made more efficient by applying marker-assisted
breeding (Bernier et al., 2007). To unveil the genetic basis of a complex trait such as grain yield under drought, it is a prerequisite to
Abbreviations: BSA, bulk segregant analysis; MAS, marker-assisted selection;
QTLs, quantitative trait loci; SG, selective genotyping; WPG, whole population genotyping.
∗ Corresponding author. Tel.: +63 2 580 5600x2586; fax: +63 2 580 5699.
E-mail addresses: p.vikram@irri.org (P. Vikram), m.swamy@irri.org
(B.P.M. Swamy), s.dixit@cgiar.org (S. Dixit), helaluddinahmed@hotmail.com
(H. Ahmed), m.stacruz@cgiar.org (M.T.S. Cruz), alok6619@gmail.com (A.K. Singh),
G.Ye@cgiar.org (G. Ye), akumar@cgiar.org (A. Kumar).
1
Present address: Plant Breeding Division, Bangladesh Rice Research Institute,
Joydebpur, Gazipur 1701, Bangladesh.
0378-4290/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.fcr.2012.05.012
genotype and phenotype mapping populations consisting of a large
number of individuals. The costs associated with large-scale phenotyping and genotyping are a bottleneck to conducting a study on
several populations at any particular time in identifying QTLs with
consistent effects across multiple environments and genetic backgrounds. With little alternative available to phenotype in different
seasons/environments, the costs involved with genotyping could
be effectively reduced by successfully applying different genotyping approaches.
In a model cereal crop such as rice (Oryza sativa) with a genome
size of 389 Mb, it is a common practice to use 200–400 lines (recombinant inbred lines, backcross inbred lines, doubled haploids) as a
mapping population. A dense map covering all 12 chromosomes
with an average genetic distance of 10–15 cM has been developed
to identify the precise QTLs. In most of the QTL mapping experiments, the whole genome is scanned (Gomez et al., 2010; Lanceras
et al., 2004; Xu et al., 2005) to ultimately find only a few significant markers showing association with the trait under study.
To reduce the extra burden and costs associated with genotyping of large mapping populations, alternative strategies have been
proposed. A trait-based genotypic analysis called “selective genotyping” (SG) was suggested by Lebowitz et al. (1987), in which
progenies are categorized into distinct classes based on the trait
186
P. Vikram et al. / Field Crops Research 134 (2012) 185–192
values and marker allele frequencies are compared between the
classes. In this approach, a subset of the population is used for genotyping instead of the whole population (Lander and Botstein, 1989).
This approach was used for mapping major-effect drought grainyield QTL qDTY12.1 using only 169 (38.7%) lines from a population
size of 436 (Bernier et al., 2007); later, this QTL was reconfirmed
with only 4.5% of the lines from the whole population (Navabi et al.,
2009). Another time and effort saving approach is bulk segregant
analysis (BSA) (Michelmoore et al., 1991). In BSA, DNA of progenies
corresponding to the phenotypic extremes is extracted and pooled.
Therefore, only two pools of extreme lines along with the parents
are genotyped for the identification of markers linked with the trait
of interest. BSA was first applied to the identification of markers
linked with disease resistance. Initially, it was applied to the diseases in which resistance was mostly governed by major genes,
usually qualitative in nature (Michelmoore et al., 1991). Recently,
BSA has been applied for quantitative traits also such as QTLs for
heat tolerance in rice (Zhang et al., 2009), salt tolerance in Egyptian cotton (El-Kadi et al., 2006), and drought tolerance in wheat
and maize (Altinkut and Gozukirmizi, 2003; Quarrie et al., 1999;
Kanagaraj et al., 2010; Venuprasad et al., 2009; Vikram et al., 2011),
as well as QTLs for grain yield under drought in maize (Quarrie et al.,
1999). BSA has been used for the identification of QTLs associated
with high grain yield under drought in rice (Venuprasad et al., 2009,
2011; Vikram et al., 2011).
Whole population genotyping involves markers from all the 12
chromosomes so that they represent whole genome and genetic
background is taken in to account while QTL analysis. Contrarily,
BSA and SG (selective genotyping) approaches may not consider
all recombination options during QTL identification due to the
unavailability of additional marker data on the same chromosome
and other chromosomes. However, these approaches have been
proven to be powerful enough to identify major QTLs that are worthy for MAS (Bernier et al., 2007; Venuprasad et al., 2009; Vikram
et al., 2011). BSA and SG might fail to detect QTLs with smaller
effects because only the extreme or tail lines are used for analysis. Also, interactions between different loci are least likely to be
detected through these approaches. QTL analysis through BSA or
SG approaches does not takes in to account the background so that
they might tilt upward the magnitude of phenotypic variance, LOD
(logarithm of odds) score, and additive effects.
The study was undertaken to perform a comparative analysis of
SG and BSA in identifying major-effect QTLs for grain yield under
drought and compare the fluctuations in phenotypic variance, additive effect, LOD, and level of significance values obtained in BSA
and SG compared with WPG (whole population genotyping). WPG,
SG, and BSA approaches were compared in a population derived
from a traditional variety (Basmati334) and a popular lowland rice
variety (Swarna). Comparative analysis of these approaches might
be biased if only one population is taken into account. To validate the results and perform a comprehensive analysis, phenotypic
and genotypic data of a rice population applied to QTL mapping in
earlier studies were used (Vikram et al., 2011).
2. Materials and methods
The study was conducted at the International Rice Research
Institute (IRRI), Los Baños, Laguna, Philippines. An F3:4 Basmati
334/Swarna population was phenotyped for grain yield under
drought and genotyped through WPG, SG, and BSA. Basmati334 is a
drought-tolerant local landrace whereas Swarna is a popular lowland rice cultivar in India (Sivaranjani et al., 2010; Verulkar et al.,
2010). The phenotypic and genotypic data of the N22/MTU1010
population were also used for comparison of WPG, SG, and
BSA (Vikram et al., 2011). Basmati334/Swarna population was
screened under drought stress in wet season 2008 (WS2008) and
N22/MTU1010 in dry season of 2009 (DS2009). WS2008 experiments were shown on June 17, 2008 whereas, DS2010 experiments
on December 9, 2009.
2.1. Phenotyping of Basmati334/Swarna population
The F3:4 Basmati334/Swarna RIL population was phenotyped
for grain yield under drought stress as well as irrigated/non-stress
conditions (Supplementary Table 1). The numbers of lines used
for drought screening were 204 during WS2008 and 367 during
DS2010. The 204 lines used in WS2008 were a subset of 367 lines
that were phenotyped in DS2010. Screening for grain yield under
drought in the lowland rice ecosystem was carried out at IRRI using
a standard phenotyping protocol (Kumar et al., 2008; Venuprasad
et al., 2007). Data for grain yield (g m−2 converted to kg ha−1 ), days
to 50% flowering (DTF), plant height (PH), biomass (BIO), and harvest index (HI) were recorded.
2.2. Genotyping of Basmati334/Swarna population
Leaf samples were collected from a whole F4 plot of the nonstress experiment and bulked. DNA was extracted by the modified
CTAB (cetyl tri-methyl ammonium bromide) method Murray and
Thompson, 1980). DNA was quantified on 0.8% agarose gel to adjust
the concentration to 25 ng L−1 . Quantified DNA was subjected to
PCR amplification with a 15-L reaction mixture involving 50 ng
DNA, 1× PCR buffer, 100 M dNTPs, 250 M primers, and 1 unit
Taq polymerase enzyme. Eight percent non-denaturing PAGE was
used for the resolution of PCR products (Sambrook et al., 1989). A
parental polymorphism survey between Basmati334 and Swarna
was carried out with 880 simple sequence repeat (SSR) markers
(Temnykh et al., 2001; McCouch et al., 2002; IRGSP, 2005). Genotyping of the Basmati334/Swarna population of 367 lines was done
by 71 polymorphic SSR markers spread throughout the 12 chromosomes of the rice genome. BSA was also done for QTL identification
with only 10% of the lines from the population (5% with high yield
and 5% with low yield under drought stress). Grain yield data from
the stress trial of DS2010 were used for selecting lines to constitute BSA. DNA of all the selected lines was quantified and bulked
in equal quantity to prepare high- and low-yield bulks. DNA of
these two bulks was screened with 203 polymorphic SSR markers along with parents – Basmati334 and Swarna (Fig. 1). A similar
procedure for BSA in the N22/MTU1010 population was carried out
(Vikram et al., 2011). Polymorphic markers from all the 12 chromosomes were selected at equal distance so that they represent whole
genome. SG was carried out with the same number of markers used
in WPG. Both RILpulations-Basmati334/Swarna and N22/MTU1010
presented in the study were analyzed through whole genome and
BSA both.
2.3. Statistical analysis
Statistical analysis was performed through SASv9.1.3 (SAS
Institute Inc., 2004). The REML algorithm of PROC MIXED of SAS
v9.1.3 was used for the determination of mean values of entries
using a model in which lines were treated as a fixed effect and
replications and blocks within replications as random. Inbred line
mean-based broad-sense heritability (H) was computed as:
H=
Vg
Vg + Ve /r
where Vg is genetic variance, Ve is error variance, and r is the number of replications.
P. Vikram et al. / Field Crops Research 134 (2012) 185–192
187
Fig. 1. Bulked segregant analysis strategy used for identifying QTLs for grain yield under drought in rice.
2.4. Linkage map construction and QTL analysis
A genetic linkage map of the Basmati334/Swarna population
was constructed through Mapdisto software using the Kosambi
map function and an LOD score of 3.0 as the threshold (Kosambi,
1944; Lorieux, 2007). All the non-linked markers were placed
on their respective chromosomes based on the physical position
(IRGSP, 2005). A similar procedure was followed by earlier workers
(Amrawathi et al., 2008). QTL network software v2.1 was used for
mixed model-based composite interval mapping (Yang et al., 2008).
In this procedure, candidate marker intervals were determined and
selected intervals were used as a co-factor in a one-dimensional
genome scan. The significance level for the determination of candidate intervals, putative QTL detection, and QTL effects was set at a
probability level of 0.05. The threshold was determined using 1000
permutations. For the genome scan, a window size of 10 cM and
walk speed of 1 cM were followed. QTL analysis results were confirmed with QGene software to validate the magnitudes of the QTLs
identified through three different strategies. Phenotypic variance
and LOD of the QTLs were estimated through composite interval
mapping using QGene software (Nelson, 1997). A LOD value of 2.5
was considered as threshold to determine significance of QTLs. Similar threshold value was used because all QTLs used in this analysis
were found significant in analysis through QTL network software.
2.5. Analysis for SG and BSA
QTL analysis with a lesser number of lines in both populations
was carried out to work out the efficiency of SG and BSA. For
analysis through SG, 36.5% of the lines (10% highest yielding, 10%
lowest yielding, and 16.5% random lines) were used (Bernier et al.,
2007). The linkage map constructed using WPG was also used to
locate the QTLs identified by SG analysis. QTL analysis for BSA was
carried out with three markers – RM210, RM223, and RM339 –
in the Basmati334/Swarna population and with eight markers –
RM315, RM431, RM212, RM3825, RM11943, RM12023, RM12091,
and RM12146 – in the N22/MTU1010 population.
To estimate the size of population for QTL analysis for grain
yield under drought, QTL analysis with 90%, 80%, 70%, 60%, and
50% population size was performed in both populations. For this
purpose, a whole population was sorted based on grain yield under
drought. Equal numbers of individuals were omitted from highyielding lines, low-yielding lines, and lines with medium yield from
the stress trials to make sure that the lines selected were random
and distribution remained normal.
3. Results
The results of QTL identification for grain yield, days to 50% flowering, plant height, and harvest index under drought by WPG in
the Basmati334/Swarna and N22/MTU1010 populations are presented in Supplementary Tables 2 and 3. QTLs for grain yield under
drought identified through WPG, SG, and BSA in both populations
are presented in Table 1.
3.1. QTLs identified for grain yield under drought with WPG
A consistent-effect QTL (qDTY8.1 ) for grain yield under drought
located between the marker intervals RM339 and RM210 on chromosome 8 was detected in the Basmati 334/Swarna population.
This QTL explained phenotypic variance (R2 ) of 15.6% and 7.1%,
with an additive effect (additive effect as percent of trial mean)
of 16.8% and 22.6%, during WS2008 (wet season 2008) and DS2010
(dry season 2010), respectively (Table 1). Two QTLs, qDTY1.1 and
qDTY10.1 , were found significant for grain yield under drought in
the N22/MTU1010 population. qDTY1.1 explained phenotypic variance of 11.9% and 4.5%, with an additive effect of 18.1% and 10.0%
during DS2009 and DS2010, respectively. qDTY10.1 explained phenotypic variance of 5.2% and 3.2% and an additive effect of 13.2%
and 12.6% during DS2009 and DS2010, respectively (Table 1).
3.2. QTLs identified for grain yield under drought with SG
qDTY8.1 in Basmati334/Swarna and qDTY1.1 in N22/MTU1010
were detected by the SG approach. The QTL qDTY8.1 explained phenotypic variance of 21.9% and 16.5%, with an additive effect of
25.4% and 39.8% during WS2008 and DS2010, respectively (Table 1).
The QTL qDTY1.1 explained phenotypic variance of 17.7% and 12.9%
188
P. Vikram et al. / Field Crops Research 134 (2012) 185–192
Table 1
QTLs for grain yield identified through SG and BSA and effect of WPG, SG, and BSA approach on F-value, additive effect, LOD, and R2 .
Population
Approach
WPG
Basmati 334/Swarna
SG
BSA
WPG
WPG
N22/MTU1010
SG
SG
BSA
a
b
c
Season
WS2008
DS2010
WS2008
DS2010
WS2008
DS2010
DS2009
DS2010
DS2009
DS2010
DS2009
DS2010
DS2009
DS2010
DS2009
DS2010
QTL
Chr.
Peak intervala
qDTY8.1
8
RM339–RM210
qDTY8.1
8
RM339–RM210
qDTY8.1
8
RM223–RM210
qDTY1.1
1
RM11943–RM12091
qDTY10.1
10
RM216–RM304
qDTY1.1
1
RM11943–RM431
qDTY1.1a
1
RM12091–RM12146
qDTY1.1
1
RM11943–RM431
F-value
14.1
15.4
10.7
15.8
9.5
21.1
36.7
18.5
19.9
15.7
22.8
15.3
–
16.1
34.7
13.9
AE (%)c
LOD
b
−16.8
−22.6
−25.4
−39.8
−12.2
−20.8
18.1
10.0
−13.2
−12.6
28.8
10.3
–
10.8
18.3
10.1
7.1
5.1
4.3
5.6
2.2
4.7
9.3
3.6
3.9
2.6
5.6
3.9
–
2.8
9.3
3.6
R2
15.6
7.1
21.9
16.5
4.9
5.8
11.9
4.5
5.2
3.2
17.7
12.9
–
9.4
11.9
4.5
Peak interval, peak of the QTL interval was between these marker intervals. R2 , phenotypic variance explained by the QTL.
Negative sign indicates that allele contribution is from susceptible parent Swarna.
AE% (additive effect %), additive effect as the percentage of trial mean [additive effect/trial mean × 100].
during DS2009 and DS2010, respectively. Additive effects contributed by qDTY1.1 were 28.8% and 10.3% during DS2009 and
DS2010, respectively.
3.3. QTLs identified for grain yield under drought with BSA
qDTY8.1 in Basmati334/Swarna and qDTY1.1 in N22/MTU1010
were detected by BSA. RM210 was identified to be associated with
grain yield under drought in the Basmati334/Swarna population
via BSA. Three markers were run on the whole population to detect
qDTY8.1 . It explained phenotypic variance of 4.9% and 5.8% and
additive effects of 12.2% and 20.8% during WS2008 and DS2010,
respectively (Table 1). In the case of qDTY1.1 , RM315 and RM431
were detected as linked markers for grain yield under drought in the
N22/MTU1010 population. Eight polymorphic markers were run
on the whole population to determine QTL boundaries (Fig. 1). The
phenotypic variance explained by qDTY1.1 in DS2009 and DS2010
was 11.9% and 4.5%, respectively. Additive effects in DS2009 and
DS2010 were 18.3% and 10.1%, respectively (Table 1).
3.4. Magnitude and efficiency of QTLs in different genotyping
approaches
All three approaches successfully detected two consistenteffect QTLs, qDTY8.1 and qDTY1.1 , in Basmati334/Swarna and
N22/MTU1010 populations, respectively (Table 1). qDTY10.1 , which
was found in the N22/MTU1010 population, was detected only in
WPG. The F-value of qDTY8.1 was lesser in WS2008 and greater
in DS2010 in SG in comparison with WPG. The additive effect of
qDTY8.1 was highest in SG, followed by WPG and BSA in WS2008
and DS2010. The LOD value was highest in WPG and SG in WS2008
and DS2010, respectively. The phenotypic variance was highest in
SG, followed by WPG and BSA in WS2008 and DS2010, respectively
(Table 1). Comparative analysis of the parameters of qDTY1.1 reveals
that the F-value is lower in SG than in WPG (Table 1). The additive
effect in SG was highest, followed by BSA and WPG in DS2009 as
well as DS2010 (Table 1). It is notable that the additive effect was
least in WPG in both years. The LOD value of qDTY1.1 was the same
in WPG and BSA in DS2009 and DS2010. But LOD estimated by SG
was lower during DS2009 and higher in DS2010 than LOD estimated with WPG and BSA. The phenotypic variance of qDTY1.1 was
the same in WPG and BSA, whereas it was greater in SG in DS2009
and DS2010 (Table 1).
3.5. QTL effect analysis with different population sizes
QTL analysis for grain yield under drought was carried out with
different population sizes. qDTY1.1 identified with phenotypic data
(under drought stress) of DS2009 with an original population size
of 362 was also successfully detected in a lesser population size of
218 lines. The same QTL was detected with phenotypic data (under
drought stress) of DS2010 in the original population but failed to be
detected in a reduced population size (Table 2). A similar trend was
observed for qDTY8.1 in the Basmati 334/Swarna population during
DS2008. Additive effect decreased with a decrease in population
size. The trend in DS2010 for this QTL was similar for a decrease
in population size up to 30%, but thereafter it became constant
up to a 50% decrease in population size and then decreased again
(Table 2). The additive effect of qDTY10.1 did not show any pattern
with respect to population size.
4. Discussion
Marker-assisted selection using consistent-effect drought grain
yield QTLs is could be an alternative approach for improving rice
grain yield under drought stress situations (Swamy et al., 2011;
Swamy and Kumar, 2011; Vikram et al., 2012). QTLs have been
identified in the past few years for different drought-related traits
through phenotyping and genotyping of large mapping populations
(Babu et al., 2003; Gomez et al., 2010). Phenotyping of mapping populations under managed drought stress condition that
too at target environments is a difficult task and involves large
expense and efforts and is considered a significant bottleneck.
Over last many years, very limited progress has been made on
the development of high-throughput, cost-effective phenotyping
methodologies. In the process, genotyping of populations with a
large number of markers for each of the 12 rice chromosomes was
followed. Polymorphic markers were selected at equal distances
on all the chromosomes and used for a whole-genome scan. Trait
selection is another important concern in drought molecular breeding. QTLs for a number of physiological and morphological traits
have been identified but they have limited implications in breeding drought-tolerant rice varieties. Large-effect QTLs for grain yield
under drought have been recently identified (Bernier et al., 2007;
Venuprasad et al., 2009) and their introgression is in process at
IRRI (IRRI, unpublished). Genotyping of different populations and
their screening in multiple environments could be made effective
through an effort and time saving genotyping strategy that can be
applied to multiple populations simultaneously. Therefore, three
189
P. Vikram et al. / Field Crops Research 134 (2012) 185–192
Table 2
Comparative analysis of QTL effects in different population sizes with a decrease in population size.
Population
Basmati334/Swarna
QTL
Population size
Year
F-value
AE (%)a
qDTY8.1
183 (∼90%)
331 (∼90%)
162 (∼80%)
295 (∼80%)
141 (∼70%)
259 (∼70%)
120 (∼60%)
223 (∼60%)
99 (∼50%)
187 (∼50%)
78 (∼40%)
151 (∼40%)
WS2008
DS2010
WS2008
DS2010
WS2008
DS2010
WS2008
DS2010
WS2008
DS2010
WS2008
DS2010
26.7
24.1
28.6
18.0
27.7
17.1
17.1
11.5
13.4
11.4
8.2
8.8
−15.8
−23.2
−15.6
−17.2
−14.9
−16.9
−11.3
−17.1
−11.0
−17.1
−9.5
−15.7