Published March 3, 2017
RESEARCH
Construction of a High-Density Genetic Map
and Its Application to QTL Identification
for Fiber Strength in Upland Cotton
Zhen Zhang, Qun Ge, Aiying Liu, Junwen Li, Juwu Gong, Haihong Shang, Yuzhen Shi, Tingting Chen,
Yanling Wang, Koffi Kibalou Palanga, Jamshed Muhammad, Quanwei Lu, Xiaoying Deng, Yunna Tan,
Ruixian Liu, Xianyan Zou, Harun Rashid, Muhammad Sajid Iqbal, Wankui Gong,* and Youlu Yuan*
ABSTRACT
Cotton (Gossypium sp.) is an important
worldwide cash crop that provides a competitive
renewable natural fiber supply for the demands
of textile industry. The development of new textile
technologies and the improvement of living
standards increase the demands for both fiber
quantity and fiber quality. 0153 is an upland
cotton cultivar with excellent fiber quality derived
from Asiatic cotton sources, especially with
regards to fiber strength. To identify quantitative
trait loci (QTLs) for fiber strength in this line, a
recombinant inbred line population consisting of
196 lines was developed from a cross between
it and sGK9708. A genetic linkage map
consisting of 2393 loci was constructed using
this recombinant inbred line population, with
single nucleotide polymorphism (SNP) markers
from the IntlCottonSNPConsortium_70k chip.
Quantitative trait loci for fiber strength were
detected across 11 environments using both
single-environment and combined multiple-
environment models. A total of 63 QTLs
controlling fiber strength were detected by
the single-environment model. Sixteen QTLs
were identified by the combined multiple-
environment model. These QTLs could make a
contribution to the improvement of fiber quality
via marker-assisted selection and provide useful
information for QTL fine mapping and functional
gene research activities as well.
774 www.crops.org crop science, vol. 57, march april 2017
Z. Zhang, Q. Ge, A. Liu, J. Li, J. Gong, H. Shang, Y. Shi, T. Chen,
Y. Wang, K.K. Palanga, J. Muhammad, X. Deng, Y. Tan, R. Liu,
X. Zou, H. Rashid, M.S. Iqbal, W. Gong, and Y. Yuan, State Key
Laboratory of Cotton Biology; Key Laboratory of Biological and
Genetic Breeding of Cotton, The Ministry of Agriculture; Institute of
Cotton Research, Chinese Academy of Agricultural Sciences; Anyang,
455000 Henan, China; Q. Lu, Anyang Institute of Technology, 455000
Henan, China. Z. Zhang, Q. Ge, and A. Liu contributed equally to this
work. Received 23 June 2016. Accepted 8 Nov. 2016. *Corresponding
authors (wkgong@aliyun.com, youluyuan@hotmail.com).
Abbreviations: CIM, composite interval mapping; FS, fiber strength;
ICIM, inclusive composite interval mapping; LOD, logarithm of
odds; MAS, marker-assisted selection; PIN, probability in stepwise
regression; PV, phenotypic variation; QTL, quantitative trait locus; RIL,
recombinant inbred line; SDM, segregation distortion marker; SDR,
segregation distortion region; SNP, single-nucleotide polymorphism;
SSR, simple sequence repeat.
C otton (Gossypium sp.) is a worldwide cash crop with high
value that is grown as a renewable natural fiber source for the
textile industry. Upland cotton (Gossypium hirsutum L., 2n = 52)
is the most widely grown cotton species because of its high yield,
wide adaptability, and acceptable fiber quality. The improvement
of living standards inevitably demands an increased fiber supply,
and the development of new textile technologies requires a fiber
supply with improved quality. Since both of the yield and fiber
quality traits of upland cotton are quantitative and negatively
correlated, traditional breeding methods to improve both at the
same time is a challenge for breeders (Park et al., 2005; Shen
et al., 2005; Lacape et al., 2009; Zhang et al., 2016). Based on
the development of the molecular marker technologies, marker-
assisted selection (MAS) may make it possible to improve both
quality and agronomic traits simultaneously. As MAS is based
on direct genotype selection, when applied in quantitative trait
Published in Crop Sci. 57:774788 (2017).
doi: 10.2135/cropsci2016.06.0544
Crop Science Society of America | 5585 Guilford Rd., Madison, WI 53711 USA
This is an open access article distributed under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
locus (QTL) selection it could increase the selection effi-
ciency by eliminating environmental variance. Presently,
there are numerous genetic maps in cotton with an array
of fiber quality QTLs identified. Shen et al. (2007) identi-
fied seven QTLs for fiber strength (FS), which explained
a range of observed phenotypic variations (PVs) from 4.31
to 16.15%. Sun et al. (2012) identified of 40 QTLs for fiber
quality traits in recombinant inbred lines (RILs) across
four environments, which explained 1.60 to 27.86% of the
total observed PVs. Ning et al. (2014) identified 23 QTLs
for FS, which were distributed on 10 chromosomes and
explained 3.73 to 17.55% of the observed PVs. Fang et
al. (2014) identified 20 QTLs for FS on 14 chromosomes
using a random-mated recombinant inbred population in
upland cotton. Tan et al. (2015) identified 59 QTLs con-
trolling five fiber quality traits across five environments.
Despite these mapping successes, the practical applica-
tion of QTLs in MAS is still rarely reported. Most of the
genetic maps have been constructed using simple sequence
repeat (SSR) markers, but the low abundance rate of SSR
polymorphism in cotton (especially in intraspecific pop-
ulations) results in SSR-based maps with large marker
intervals, unexpected gaps, or even two or more linkage
groups constructed within one chromosome (Zhang et al.,
2016), thus yielding maps with poor coverage of the whole
genome. This phenomenon impedes MAS application in
breeding programs. Another reason for the lack of applica-
tion is related to marker transportability; it is not easy to
apply the QTLs identified from one population to another.
As a result, it has been difficult to apply map-based stud-
ies such as functional gene identification or the pyramiding
breeding of QTLs through MAS in cotton (Wei et al.,
2014; Guo et al., 2015; Ma et al., 2015; Zhang et al., 2016).
Developing genome-wide markers to construct satu-
rated linkage maps with high marker density is the key
step in the application of map-based research activities
(Timmerman-Vaughan et al., 2004; Burstin et al., 2007;
Lejeune-henaut et al., 2008). Single-nucleotide polymor-
phism (SNP) markers provide a good opportunity due to
their abundance and stability in most genomes (Wang et
al., 2014, 2015; Maccaferri et al., 2015; Tian et al., 2015).
Single-nucleotide polymorphism markers are increas-
ingly being applied to genetic map construction and MAS
with the advent of next generation sequencing and high-
throughput genotyping technologies, which make the
development of SNP markers possible in large through-
put scales. Our objective was to develop a high-density,
intraspecific genetic map of upland cotton using an
IntlCottonSNPConsortium_70k chip developed through
the cooperation of the International Cotton SNP con-
sortium and manufactured by Illumina Inc. (San Diego,
CA) (Hulse-Kemp et al., 2015). The mapped RIL popu-
lation, consisting of 196 RILs, was from a cross between
upland cotton strain 0153 with excellent fiber quality
crop science, vol. 57, march april 2017 www.crops.org 775
and sGK9708 with high yield and wide adaptability. To
ensure robustness, six locations and multiple years were
involved for a total of 11 environments where pheno-
typic data was collected and analyzed. Based on these
11 environments of the 196-RIL population, a linkage
map consisting of 793 SSR markers ( Jamshed et al., 2016;
Sun et al., 2012) and a linkage map of 5521 specific locus
amplified fragment sequencing (SLAF-seq) SNP markers
(Zhang et al., 2016) had been previously constructed, with
QTLs for fiber quality traits ( Jamshed et al., 2016) and
for boll weight (Zhang et al., 2016) identified. The stabil-
ity and reliability of the aforementioned QTLs still awaits
verification. The aims of this study were to construct a
genomic-scale, high-density genetic map using chip SNP
markers and to identify and/or verify stable QTLs for FS
on the basis of these previous studies (Sun et al., 2012;
Jamshed et al., 2016; Zhang et al., 2016).
MATERIALS AND METHODS
Plant Materials
An intraspecific F6:8 RIL population with 196 lines was used
in this study. The development procedure of the population
was detailed in the report of Sun et al. (2012). Briefly, the RIL
population was developed from an upland cotton cross between
strain 0153, an upland cotton cultivar with strong FS, and
sGK9708, which is derived from CRI41 and has wide adapt-
ability and high yield potential. The cross was made in 2001 and
196 F2:3 families were randomly selected in 2004 in Anyang. A
random plant was selected in each of the F2:6 families in Anyang
in 2006 to form F6:7 families. The F6:7 families were planted in
Hainan Province in the winter of 2006. F6:8 families and later
generations were regarded as RILs. For convenience in descrip-
tion and research, all the RILs were regarded as F6:8 (Sun et
al., 2012). The phenotypic evaluations of this population were
conducted across 5 yr and six locations, including in 2007 in
Anyang; in 2008 in Anyang, Quzhou, and Linqing (Sun et al.,
2012; Zhang et al., 2015b; for further information), in 2009 in
Anyang, Quzhou, and Akesu; in 2010 in Anyang, Gaoyi, and
Zhengzhou; and in 2013 in Anyang (Zhang et al. 2015b, 2016;
Jamshed et al., 2016; for further information).
Phenotypic Data Analysis
The sample collection and phenotypic data analysis of FS were
described in Zhang et al. (2015b, 2016). Briefly, 30 fully opened
bolls in each of the 196 lines of the RIL population were sam-
pled during harvesting season in September each testing year
and the fibers were sampled. High-volume instrumentation
(HVI) testing was done using HFT9000 (Premier Evolvics
Pvt. Ltd, India) instruments with HVICC calibration to test
the FS in the Cotton Quality Supervision, Inspection and Test-
ing Center, Ministry of Agriculture, Anyang, China. One-way
ANOVA and Microsoft Excel were used to test the significance
of the difference in FS between two parents and provide uni-
variate phenotypic data on the population, including mean
value, standard deviation, skewness, and kurtosis.
Analysis of SNP Markers
The IntlCottonSNPConsortium_70k chip was developed by
an international consortium and manufactured by Illumina Inc.
(San Diego, CA). It contains approximately 70,000 SNP probes,
in which all of the then-published SNP development works, as
well as some sets of in-press SNP data at the time of chip man-
ufacture, were included (Hulse-Kemp et al., 2015). The chip
primarily targets G. hirsutum-specific SNPs, but also contains
SNPs from Gossypium barbadense L., Gossypium tomentosum Nutt.
ex Seem, Gossypium mustelinum Miers ex G. Watt., Gossypium
armourianum Kearney, and Gossypium longicalyx Hutch. & Lee
(Hulse-Kemp et al., 2015).
The IntlCottonSNPConsortium_70k Chip
Hybridization and Result Analysis
The genomic DNA of both parents and 196 F6:8 RILs were
extracted with the TaKaRa MiniBEST Plant Genomic
DNA Extraction Kit (TaKaRa, Dalian, China). Two
parental and 196 F6:8 lines were genotyped using the
IntlCottonSNPConsortium_70k chip. The procedure of
chip hybridization was performed according to the workflow
described in the Illinium HD Assay Ultra manual provided by
chip manufacturer Illumina Inc. and also described by Hulse-
Kemp et al. (2015) and Zhang et al. (2014). Single-nucleotide
polymorphism genotyping usually produced three clear clus-
ters: AA homozygote, BB homozygote, and AB heterozygote.
The SNP markers were screened and discarded through sev-
eral steps based on the following criteria: first, the SNPs that
did not show clearly three defined AA, BB, and AB clusters;
second, the genotypes of SNPs of one or both of the parents
that showed heterozygosis; third, the SNPs that had no poly-
morphism between parents; fourth, the SNPs that had missing
data of more than 40%; and finally, the SNPs with an extremely
significant segregation distortion (P < 0.001). The remaining
SNP loci were used for further genotyping analysis and genetic
map construction. A region on the map with more than three
consecutive adjacent loci that showed significant (0.001 < P <
0.05) segregation distortion was defined as a segregation distor-
tion region (SDR) (Zhang et al., 2014).
SSR Markers Used in the Map Construction
All the SSR markers used in this study were identified by
Sun et al. (2012). The SSR markers also underwent screen-
ing procedures based on the criteria set for the SNP marker, as
described in the section about IntlCottonSNPConsortium_70k
chip hybridization and the results analysis.
Linkage Map Construction
The SNP and SSR markers (Sun et al., 2012) were partitioned
initially into linkage groups and then located to the chromo-
somes based on genome database sequence (Zhang et al., 2015a).
Considering that the data may have genotyping errors and dele-
tions when scanning and calling SNP genotype, which may
greatly reduce the quality of the linkage map, a high map strategy
was used to order the SNP alleles and correct genotyping errors
within all chromosomes (Liu et al., 2014). We first determined
primary marker orders by their locations on chromosomes, and
according to the relationship between ordered markers (Jansen
776 www.crops.org crop science, vol. 57, march april 2017
et al., 2001; van Ooijen, 2011), genotyping errors or deletions
were corrected using the SMOOTH algorithm (van Os et al.,
2005). After that, we ordered the map again and then we reran
SMOOTH to correct the new ordered genotypes. After four or
more cycles, a high-quality linkage map with 26 chromosomes
was obtained. Map distances were estimated using the Kosambi
mapping function (Kosambi, 1943).
QTL of Fiber Strength Identification
Quantitative trait loci were identified using both a single-envi-
ronment model (Model I) and a combined multiple-environment
model (Model II), respectively. Model I analysis was performed to
identify additive QTLs in a single environment. When the marker
intervals of QTLs in different environments overlapped com-
pletely or partially, the QTLs were considered as the same ones.
The QTLs identified in at least three environments were regarded
as stable ones. Model II analysis was performed to identify additive
QTLs and epistatic QTLs in combined multiple environments.
Model I analysis was performed with the composite inter-
val mapping (CIM) method (Zeng, 1994) using the WinQTL
Cartographer 2.5 software (Wang et al., 2001) with a map-
ping step of 1.0 cM, five control markers. The threshold score
of logarithm of odds (LOD) was determined by performing
a 1000-permutation test at a significant level of P < 0.05.
Nomenclature of QTLs identified with Model I followed Suns
description (Sun et al., 2012).
Model II analysis was performed with the inclusive com-
posite interval mapping (ICIM) method using the MET
function of the software QTL IciMapping 4.1 (Li et al., 2007,
2015b; Meng et al., 2015). Basically, two categories of QTLs,
additive and epistatic, were identified based on IciMapping 4.1.
For the additive analysis, a mapping step of 1.0 cM with a prob-
ability in stepwise regression (PIN) value of 0.001 was used.
For the epistasis analysis, a mapping step of 1.5cM with a PIN
value of 0.0001 was used. The threshold LOD score of both
additive and epistasis analyses was determined by performing a
1000-permutation test at a significance level of P < 0.05. The
additive QTLs identified by Model II were named using meaq
(mutienvironment additive QTL), followed by the abbreviation
of the trait, and then suffixed with the number of the chromo-
some and QTL number. The epistatic QTL pairs identified by
Model II were named using meeq (mutienviroment epistatic
QTL), followed by the abbreviation of the trait, and then with
the number 1 or 2 to identify the member of the pair and,
finally, the identification number of the QTL pair.
Congruence Analysis of the QTLs
Identified by Models I and II
When the confidence interval of the QTL indentified by Model I
was the same as or overlapped with that of the QTL identified by
Model II or vice versa, the two QTL were regarded as the same.
Congruence Analysis of the QTLs
with Previous Reports
Meta-analysis was performed using Biomercator 4.2 (Arcade et
al., 2004), as described by Said et al. (2013), to construct a con-
sensus linkage map based on the QTL database of cottonqtlgb.
org. The previously reported QTLs were compared based on
this consensus map and common markers. The QTLs resulting mapped, including 1028 SNP markers and 44 SSR ones,
from meta-analysis (meta-QTLs) were named using the term
which spanned a linkage distance of 1380.9 cM with an
metaqFS-chromosome number-QTL number in the same chro-
average marker interval of 1.29 cM. In the Dt subgenome,
mosome. The QTLs in the current study were compared with
previous meta-QTLs using the physical positions of markers in there were 1321 markers mapped, including 1288 SNP
the QTL confidence intervals based on the sequence database of markers and 33 SSR ones, which spanned a linkage distance
Zhang et al. (2015a). If the positions of the markers associated with of 1484.83 cM with an average marker interval of 1.15 cM.
two QTLs spanned an overlapped region in the physical map, par- The largest chromosome was c16, consisting of 178 markers
tially or completely, the two QTLs will be considered as the same. and covering a genetic length of 194.96 cM, with an average
marker interval of 1.10 cM. The smallest chromosome in the
map was c6, consisting of only 46 markers and covering a
RESULTS genetic length of 52.85 cM, with an average marker interval
Phenotypic Data Analysis of 1.17 cM. (Table 2, Supplemental Table S4).
The P-value in the one-way ANOVA analysis was <0.001,
suggesting a significant difference in FS between the parents. Quality of the Genetic Map
The descriptive statistical analysis results of the parents and The quality of the genetic linkage map was mainly assessed
RIL population throughout 11 environments are listed in by the criteria of gaps in the map, segregation distortion of
Table 1. An approximately normal distribution was suggested mapped markers, and of collinearity analysis between the
based on the absolute skewness value of the mean value of FS genetic linkage map and the physical map.
in the RIL population, indicating that the population was The map contained 80 gaps, which were larger than
suitable for map construction and QTL identification. 5 cM in the genetic map. The largest gap in the map
The heritability of FS was analyzed using fiber was 28.17 cM and was on the c8. The chromosome c12
measurement data from 10 of the 11 environments of the contained 10 gaps >5 cM, while c2, c3, c6, c21, and c25
RILs. The results are shown in Supplemental Table S1. contained no such gaps (Table 2, Supplemental Table S4).
The heritability of FS was high, ranging from 0.7919 to Among the total 2393 markers in the map, there were
0.8492 for single lines and from 0.8833 to 0.9169 for an 834 markers that showed significant segregation distortion
average per environment (P < 0.01). These results were in (0.05 < P < 0.001). These segregation distortion markers
agreement with previous reports (Jamshed et al., 2016). (SDMs) were dispersed in all the linkage groups. Among
Genotypes, environments, and the interaction between the 834 SDMs, 313 of them were distributed in the At
them significantly contributed to the phenotypes of FS subgenome of upland cotton, whereas 521 of them
(Supplemental Table S2). Correlation analysis revealed that were distributed in the Dt subgenome of G. hirsutum.
FS of the RILs were highly significantly correlated (P < Chromosome c14 contained 217 SDMs, accounting for
0.001) across all the environments (Supplemental Table S3). 69.80% of all the markers mapped on it. Chromosome c11
contained the smallest number of SDM (three). The lowest
Genetic Map Construction percentage of SDMs was on c21 (5.5%). In total, the SDMs
The final genetic map consisted of 2393 markers, including aggregated into 30 SDRs in the whole map, with 15 of
2316 SNP and 77 SSR markers. The map spanned a total them located in the At subgenome and the remaining 15
distance of 2865.73 cM, with an average marker interval of located in Dt subgenome of G. hirsutum (Table 2).
1.20 cM. In the At subgenome, there were 1072 markers
Table 1. The results of the statistical analysis of the parents and the whole population.
Parents Recombinant inbred line population
Environment P1 P2 P1P2 P-value Minimum Maximum Range Average SD Variance Skewness Kurtosis
07ay 33.75 25.65 8.10 <0.001*** 23.65 36.75 13.10 29.02 2.24 5.01 0.40 0.66
08ay 32.60 25.60 7.00 <0.001*** 24.65 35.30 10.65 29.62 1.98 3.93 0.22 0.18
08lq 33.40 24.20 9.20 <0.001*** 23.55 38.70 15.15 29.76 2.37 5.60 0.36 0.54
08qz 34.50 26.70 7.80 <0.001*** 25.35 37.75 12.40 31.06 2.32 5.39 0.30 0.19
09ay 33.20 25.60 7.60 <0.001*** 22.90 35.90 13.00 29.11 2.26 5.11 0.28 0.17
09qz 37.30 26.95 10.35 <0.001*** 28.35 39.75 11.40 33.38 2.43 5.93 0.45 0.07
09aks 33.90 25.60 8.30 <0.001*** 24.20 36.30 12.10 29.46 1.98 3.92 0.27 0.15
10gy 32.25 23.95 8.30 <0.001*** 22.85 34.90 12.05 28.33 2.28 5.22 0.22 0.34
10ay 29.75 25.40 4.35 <0.001*** 23.95 33.85 9.90 28.60 1.91 3.66 0.27 0.29
10zz 33.30 27.95 5.35 <0.001*** 25.75 36.40 10.65 30.38 2.02 4.07 0.19 0.07
13ay 32.00 25.70 6.30 <0.001*** 25.87 32.81 6.94 31.81 2.52 6.29 0.20 0.03
07ay, Anyang in 2007; 08ay, Anyang in 2008; 08lq, Linqing in 2008; 08qz, Quzhou in 2008; 09ay, Anyang in 2009; 09qz, Quzhou in 2009; 09aks, Akesu in 2009; 10gy, Gaoyi
in 2010; 10ay, Anyang in 2010; 10zz, Zhengzhou in 2010; 13ay, Anyang in 2013.

P1: 0153; P2: sGK9708.

crop science, vol. 57, march april 2017 www.crops.org 777

Table 2. Detailed information of the genetic map
No. No. Total No. Genetic Average distance Max Gaps
Chromosome SSR SNP Markers SDM SDM distance between markers gap > 5cM X2 P SDR
% cM
c1 5 107 112 35 31.25 125.42 1.13 8.27 5 5.48 0.20 3
c2 4 54 58 9 15.52 74.52 1.31 4.34 0 2.26 0.48 1
c3 7 100 107 27 25.23 98.42 0.93 4.46 0 4.21 0.35 1
c4 6 60 66 8 12.12 87.29 1.34 7.48 1 4.20 0.25 0
c5 5 119 124 45 36.29 102.49 0.83 5.56 1 6.89 0.19 2
c6 3 43 46 20 43.48 52.85 1.17 4.54 0 14.43 0.19 1
c7 2 97 99 9 9.09 123.93 1.27 10.32 3 2.83 0.39 0
c8 0 21 21 8 38.10 125.75 6.29 28.17 6 15.80 0.27 1
c9 3 91 94 24 25.53 104.60 1.13 7.92 7 6.90 0.29 2
c10 0 68 68 38 55.88 95.37 1.42 6.93 4 21.46 0.16 2
c11 2 28 30 3 10.00 121.62 4.19 11.44 10 2.71 0.41 0
c12 2 89 91 11 12.09 106.61 1.19 5.51 3 3.41 0.24 0
c13 5 151 156 76 48.72 162.05 1.05 5.18 1 5.24 0.09 2
c14 6 305 311 217 69.78 169.93 0.55 5.96 1 10.64 0.12 2
c15 4 27 31 7 22.58 95.96 3.20 10.16 6 4.38 0.21 0
c16 3 175 178 72 40.45 194.96 1.10 13.59 5 11.55 0.22 3
c17 1 60 61 7 11.48 96.45 1.61 13.37 3 2.52 0.43 0
c18 1 96 97 8 8.25 126.18 1.31 13.89 4 2.64 0.27 0
c19 1 56 57 11 19.30 100.59 1.80 9.71 6 3.81 0.39 1
c20 1 62 63 19 30.16 92.47 1.49 10.75 5 10.49 0.24 2
c21 2 89 91 5 5.50 103.81 1.15 4.79 0 1.78 0.46 0
c22 2 64 66 8 12.12 98.40 1.51 9.89 3 2.29 0.43 1
c23 0 131 131 62 47.33 114.01 0.88 8.30 2 16.12 0.14 2
c24 1 46 47 7 14.90 89.03 1.94 5.46 1 6.95 0.29 1
c25 8 121 129 93 72.09 93.43 0.73 3.93 0 18.49 0.06 3
c26 3 56 59 5 8.47 109.62 1.89 7.68 3 3.19 0.38 0
Total 77 2316 2393 834 34.85 2865.73 1.20 28.17 80 30

Collinearity analysis of SNP and SSR marker loci
between the linkage map and the physical map (Zhang et
al., 2015a) is shown in Fig. 1. The results showed that the
SNP and SSR markers used in constructing the genetic
linkage map had a reasonable consistency with their
corresponding locations on the physical map over the
whole genome. Chromosomes c1, c2, c3, c7, and c13 in
the At subgenome and c14, c18, c20, c22, c24, and c26 in
the Dt subgenome showed some deviations in collinearity
analysis (Fig. 1, Supplemental Table S5)
Additive QTLs for Fiber Strength
Identified by Model I
Based on Model I analysis, a total of 63 additive QTLs for
FS were identified across 11 environments. Chromosomes
c17, c23, and c26 had no additive QTL, and the remaining
chromosomes were identified to harbor at least one
additive QTL each, among which 16 additive QTLs were
identified as stable ones. Briefly, qFS-c71, detected in
all the 11 environments, was identified to be within the
marker interval of i46669Gh-i21754Gh, explaining 4.72
to 13.55% of the observed PVs (Fang et al., 2014). qFS-
c251, detected in nine environments, was identified to
be within the marker interval of i57381Gb-i54356Gb,
explaining 4.04 to 12.49% of the observed PVs. qFS-c41,
qFS-c72, and qFS-c73 were detected in six environments,
were identified to be within the marker intervals
of i47003Gh-BNL2572, i21754Gh-i24033Gh, and
i40250Gh-i01461Gh, and explained 5.45 to 10.70, 6.05 to
8.73, and 4.24 to 9.19% of the observed PVs, respectively.
The remaining stable additive QTLs, namely qFS-c112
and qFS-c252 (detected in five environments), qFS-c61,
qFS-c111, qFS-c131, and qFS-c212 (detected in four
environments), and qFS-c11, qFS-c191, qFS-c201, qFS-
c211, and qFS-c221 (detected in three environments) are
detailed as in Fig. 2, Table 3, and Supplemental Table S6.
Additive QTLs for Fiber Strength Identified
by Model II
Based on Model II analysis, a total of 16 additive QTLs
were identified, among which four were mapped on c7,
three on c4, two each on c3 and c16, and one each on
c9, c13, c18, c23, and c25. A major QTL, meaqFS-c71,
was identified to be within the marker interval of
i30531Gh-i43170Gh, with 13.92% of the observed total
PVs explained. Three additive QTLs, meaqFS-c73,
meaqFS-c74, and meaqFS-c131, were identified to
have significant effects, and were located in the marker

778 www.crops.org crop science, vol. 57, march april 2017

Fig. 1. Collinearity between the genetic map and the physical map: (a) collinearity of the At subgenome between the genetic map and the
physical map and (b) collinearity of the Dt subgenome between the genetic map and the physical map.

Fig. 2. The logarithm of odds (LOD) and the observed phenotypic variation values of the stable quantitative trait loci identified by Model I.

crop science, vol. 57, march april 2017 www.crops.org 779

Table 3. The detail information of stable additive quantitative trait loci (QTL) of fiber strength identified by the Model I. LOD,
logarithm of odds; PV, phenotypic variation.
LOD LOD_L LOD_R
Chromosome QTL Environment Position Value Additive PV Marker interval (P < 0.01) (P < 0.01)
cM %
c1 qFS-c11 10zz 69.31 5.00 1.09 10.87 MUSS422 68.80 70.80
10ay 70.31 3.01 0.74 7.14 TMB1931i39578Gh 66.90 73.20
08qz 71.31 2.27 1.01 6.05 i39578Gh 70.00 72.20
c4 qFS-c41 10ay 81.21 2.93 0.54 6.60 i47003GhBNL2572 76.70 85.90
08qz 82.21 2.91 0.93 6.67 i47003GhBNL2572 77.80 85.90
10gy 82.21 2.23 0.76 5.45 i47003GhBNL2572 77.50 85.90
09ay 82.91 3.89 0.95 8.19 BNL2572 80.40 85.90
09qz 82.91 4.91 0.88 10.70 BNL2572 81.60 85.20
09aks 82.91 2.41 0.75 4.99 BNL2572 80.20 85.70
c6 qFS-c61 07ay 15.41 2.17 0.60 4.83 i19782Ghi17220Gh 13.90 17.00
08lq 15.41 2.67 0.69 5.84 i19782Ghi17220Gh 13.90 16.40
08qz 15.41 1.93 0.58 4.22 i19782Ghi17220Gh 13.90 17.70
09ay 15.41 2.11 0.62 4.58 i19782Ghi17220Gh 13.90 17.00
c7 qFS-c71 08ay 1.01 4.45 1.10 13.55 i46669Gh 0.00 2.00
09ay 1.01 3.17 0.74 6.97 i46669Ghi39902Gh 0.00 9.20
09qz 1.01 2.21 0.91 4.72 i46669Ghi39902Gh 0.00 9.10
10gy 1.01 4.08 0.73 8.89 i46669Ghi39902Gh 0.00 9.70
13ay 1.01 3.24 1.11 7.44 i46669Ghi39902Gh 0.00 9.20
07ay 2.01 5.27 1.17 10.60 i46669Ghi39902Gh 0.00 9.30
08lq 2.01 5.61 1.26 11.08 i46669Ghi39902Gh 0.00 9.40
08qz 2.01 2.91 0.95 5.13 i46669Ghi39902Gh 0.00 8.50
09aks 2.01 2.31 0.52 4.92 i46669Ghi21754Gh 0.00 11.60
10ay 2.01 4.53 0.61 9.21 i46669Ghi39902Gh 0.00 10.20
10zz 2.41 2.53 0.53 5.17 i46669Ghi39902Gh 0.00 10.60
c7 qFS-c72 07ay 13.11 4.10 1.12 8.14 i36368Gh 12.20 13.60
08ay 13.11 3.19 0.82 6.66 i21754Ghi36368Gh 10.60 13.80
08lq 13.11 4.42 1.23 8.73 i43900Ghi31430Gh 11.80 14.50
09ay 13.11 2.86 0.71 6.05 i21754Ghi24033Gh 10.20 23.80
10gy 14.21 2.93 0.62 6.15 i43900Ghi28452Gh 11.60 20.10
10ay 16.01 3.64 0.56 7.39 i43900Ghi28452Gh 11.60 20.10
c7 qFS-c73 08lq 92.51 3.05 0.85 7.91 i42029Gh 90.50 99.40
08qz 95.51 2.69 0.71 7.48 i42029Gh 90.50 100.80
13ay 96.51 2.35 0.80 7.46 i42029Ghi40250Gh 90.50 102.90
09ay 97.51 3.12 0.73 9.19 i40250Gh 92.00 100.80
10ay 98.51 2.42 0.55 6.18 i40250Ghi01461Gh 90.80 105.40
08qz 102.81 2.45 0.51 4.24 i40250Gh 100.80 103.50
c11 qFS-c111 07ay 15.51 2.41 0.69 5.31 i61083Gt 14.50 21.80
10ay 15.51 2.97 0.64 6.28 i61083Gt 11.40 21.20
10zz 15.51 2.86 0.72 6.10 i61083Gt 11.30 21.00
09ay 19.51 2.65 0.78 6.87 i61083Gti15922Gh 14.80 22.60
c11 qFS-c112 07ay 36.91 1.87 0.62 4.17 i45937Ghi50525Gb 34.90 43.50
09ay 36.91 3.23 0.79 6.76 i45937Gh 35.90 39.60
09aks 36.91 1.88 0.61 4.16 i06946Ghi50525Gb 32.10 40.30
10ay 36.91 2.36 0.58 5.11 i45937Gh 35.70 39.70
10zz 36.91 1.98 0.57 4.40 i45937Gh 35.90 39.40
c13 qFS-c131 10ay 121.11 3.49 0.78 7.80 i24050Ghi44161Gh 120.40 122.60
10zz 121.11 3.35 0.94 7.17 i49771Ghi66252Ga 120.20 123.90
09aks 121.51 2.02 0.70 4.52 i49771Ghi66252Ga 120.20 122.90
13ay 124.21 3.84 0.80 8.40 i24050Ghi23021Gh 121.30 127.20
c19 qFS-c191 08ay 20.91 2.28 0.53 5.22 i09003Ghi08957Gh 14.50 24.60
10zz 20.91 1.85 0.45 4.42 i09003Ghi08957Gh 15.30 24.30
08lq 22.71 2.09 0.52 4.36 i09003Ghi08945Gh 16.30 25.70
10zz, Zhengzhou in 2010; 10ay, Anyang in 2010; 08qz, Quzhou in 2008; 10gy, Gaoyi in 2010; 09ay, Anyang in 2009; 09qz, Quzhou in 2009; 09aks, Akesu in 2009; 07ay,
Anyang in 2007; 08lq, Linqing in 2008; 08ay, Anyang in 2008; 13ay, Anyang in 2013.

LOD_L indicates the genetic position where the QTL starts, while LOD_R indicates the genetic position where the QTL ends in the linkage map at the significance level of
P < 0.01.

780 www.crops.org crop science, vol. 57, march april 2017

Table 3. Continued.
LOD LOD_L LOD_R
Chromosome QTL Environment Position Value Additive PV Marker interval (P < 0.01) (P < 0.01)
cM %
c20 qFS-c201 10ay 32.91 2.66 0.75 5.79 i00501Gh 32.20 34.00
10zz 32.91 2.43 0.64 5.35 i00501Gh 32.20 34.00
13ay 32.91 2.12 0.74 4.67 i00501Ghi12249Gh 32.20 34.50
c21 qFS-c211 07ay 9.61 2.00 0.67 4.33 i06825Ghi14092Gh 7.80 10.80
09ay 9.61 2.20 0.70 4.69 i06825Ghi14092Gh 7.80 10.80
10ay 9.61 2.00 0.50 4.25 i14092Ghi33389Gh 8.80 11.70
qFS-c212 09ay 24.31 2.49 0.92 6.03 i37987Ghi06951Gh 22.40 27.20
10gy 24.31 1.93 0.86 4.90 i26476Ghi06953Gh 20.00 28.50
10ay 25.31 3.14 0.84 7.72 i37987Ghi06951Gh 21.00 26.40
09aks 26.41 2.25 0.61 4.90 i06951Ghi06953Gh 26.30 28.50
c22 qFS-c221 09ay 53.81 4.00 1.35 8.45 i23482Ghi44365Gh 53.40 54.50
10gy 53.81 2.60 0.87 5.57 i23482Ghi30443Gh 53.40 56.00
09aks 54.81 1.79 0.40 4.09 i35236Ghi17784Gh 53.30 60.10
c25 qFS-c251 07ay 76.61 5.92 1.30 11.37 i57381GbNAU1192 75.80 77.60
08ay 76.61 3.59 1.49 5.95 i57381GbNAU1192 75.80 77.60
08lq 76.61 4.00 1.49 7.53 i57381GbNAU1192 75.80 77.60
08qz 76.61 1.96 0.91 4.04 i57381GbNAU1192 75.80 77.90
10ay 76.61 4.29 1.37 8.51 i57381Gbi10989Gh 75.80 77.70
10gy 76.61 2.08 1.18 4.42 i57381Gbi10989Gh 75.80 78.00
10zz 76.61 5.08 1.59 10.26 i57381GbNAU1192 75.80 77.60
13ay 76.61 2.54 1.35 5.15 i57381Gbi54356Gb 75.80 82.60
07ay 81.71 6.12 0.92 12.49 i51495Gbi54356Gb 81.10 82.50
c25 qFS-c252 13ay 86.01 2.07 1.10 4.46 i48814Ghi33995Gh 82.60 86.90
07ay 86.71 6.21 0.92 12.68 i33995Gh 85.10 88.60
08lq 86.71 2.89 1.03 5.56 i33995Gh 85.80 88.00
10ay 86.71 1.99 0.77 4.06 i47444Gh-i33995Gh 84.20 88.30
10zz 86.71 2.03 0.83 4.26 i33995Gh 85.00 88.20

10zz, Zhengzhou in 2010; 10ay, Anyang in 2010; 08qz, Quzhou in 2008; 10gy, Gaoyi in 2010; 09ay, Anyang in 2009; 09qz, Quzhou in 2009; 09aks, Akesu in 2009; 07ay,
Anyang in 2007; 08lq, Linqing in 2008; 08ay, Anyang in 2008; 13ay, Anyang in 2013.

LOD_L indicates the genetic position where the QTL starts, while LOD_R indicates the genetic position where the QTL ends in the linkage map at the significance level of
P < 0.01.

intervals of i01461Gh-i40901Gh, i00986Gh-i00200Gh,
and i13102Gh-i26701Gh, with 4.97, 5.21, and 5.21%
of the observed total PVs explained, respectively.
The locations and effects on the PVs of the remained
additive QTLs, namely meaqFS-c31, meaqFS-c32,
meaqFS-c41, meaqFS-c42, meaqFS-c43, meaqFS-c72,
meaqFS-c91, meaqFS-c161, meaqFS-c162, meaqFS-c181,
meaqFS-c231, and meaqFS-c251, are detailed and listed
in Fig. 3 and Supplemental Table S7.
Epistatic QTLs for Fiber Strength
Identified by Model II
Epistatic QTLs were identified in pairs with Model II, and
a total 156 pairs of epistatic QTLs were identified (Fig. 4,
Supplemental Table S8). All chromosomes were identified
to harbor epistatic QTLs with no obvious preferences
(Fig. 4b and 4c). The chromosomes that harbored the
largest number of epistatic QTL numbers were c5 in the
At subgenome and c18 and c23 in the Dt subgenome,
harboring 13, 15, and 13 epistatic QTLs, respectively. The
chromosomes that harbored the fewest epistatic QTLs were
c2 and c8, both in the At subgenome, harboring six and
seven epistatic QTLs respectively (Fig. 4a). Among them,
only one pair of epistatic QTLs, meeqFS-1155 and meeqFS-
2155, could explain >4% (4.29%) of the observed total PV
of epistasis. Eight pairs of epistatic QTLs could explain from
3 to 4% of the observed total PV of epistasis. One hundred
and forty-four pairs of epistatic QTLs could explain from 1
to 3% of the observed total PV of epistasis, and three pairs
of epistatic QTLs could explain <1% of the observed total
PV of epistasis (Fig. 4, Supplemental Table S8).
Comparison and Congruence Analysis
of the QTLs Identified by Models I and II
Comparing the additive QTLs identified, four stable QTLs
identified by Model I, qFS-c41, qFS-c71, qFS-c131, and
qFS-c252, had the same or overlapped confidence intervals
with four of the 16 QTLs identified by Model II, meaqFS-c43,
meaqFS-c71, meaqFS-c131, and meaqFS-c251, respectively.
The confidence interval of the QTL qFS-c73, identified
by Model I, harbored two QTLs identified by Model II,
meaqFS-c72 and meaqFS-c73 (Table 3, Supplemental Table

crop science, vol. 57, march april 2017 www.crops.org 781

 Fig. 3. The total logarithm of odds (LOD) and
phenotype variation (PV) values, the main effect
LOD and PV values, and the environment effect
LOD and PV values of the additive quantitative
trait loci identified by Model II. Figure (a) shows
LOD values and (b) shows PV values.

Fig. 4. The epistatic quantitative trait

locus (QTL) pairs and their logarithm
of odds (LOD) values: (a) the
number of epistatic QTLs in each
chromosome, (b) the interactions of
the epistatic QTL pairs, and (c) the
LOD value of the epistatic QTL pairs
in the whole genome.

782 www.crops.org crop science, vol. 57, march april 2017

S7). The QTL meaqFS-c32, identified by Model II, also had with the genetic map constructed with SSR markers, the
overlapping confidence intervals with the QTL qFS-c32, linkage map constructed with SNP makers had significantly
which could only be detected in one environment by Model increased marker density and much better genome
I. The remaining QTLs were exclusively identified by either coverage (Guo et al., 2015; Ma et al., 2015; Tan et al., 2015;
Model I or II (Table 3, Fig. 3, Supplemental Tables S6 and S7). Wang et al., 2015). In the current study, the SNP markers
The results of congruence analysis between the additive identified using the IntlCottonSNPConsortium_70k chip
QTLs identified by Model I and the epistatic QTLs identified greatly increased the density of the map over the one
by Model II are shown in Table 4. Among the 156 pairs constructed with SSR markers by Sun et al. (2012), i.e.,
of epistatic QTLs, one pair was identified to be congruent 30 SNP markers were added between the SSR markers
with additive QTLs qFS-c201 and meaqFS-c91, identified TMB1688 and NAU1085 on chromosome 7, 128 SNP
by Models I and II, respectively. Twenty epistatic QTL markers were added between the SSR markers TMK12 and
pairs were identified so that one of each pair was an additive BNL 1421 on chromosome 13, seven SNP markers were
QTL identified by Model I and/or Model II. The results added between SSR markers BNL1047b and TMB2388,
showed meaqFS-c43 and qFS-c41 and meaqFS-c71 and and 25 SNP markers were added between SSR markers
qFS-c71 to be the same QTLs, respectively. Only qFS-c41 NAU3243 and BNL3806b on chromosome 25 (Fig. 5).
and qFS-c71 are referred to in the following analyses and The addition of SNP markers significantly increased the
discussions. qFS-c71 had epistatic interaction with three resolution of the constructed linkage map (Wang et al.,
epistatic QTLs, meeqFS-111, meeqFS-215, and meeqFS- 2015; Ning et al., 2014).
2149, on chromosomes 2, 9, and 26, respectively. qFS-c11 As to the current linkage map constructed by the
and qFS-c201 identified by Model I and meaqFS-c231 IntlCottonSNPConsortium_70k chip, gap, collinearity,
identified by Model II had epistatic interaction with and segregation distortion analysis of the mapped markers
two epistatic QTLs, namely meeqFS-256 and meeqFS- confirmed its reliability (Fig. 1, Table 2) for further study
290, meaqFS-c91 and meeqFS-295, and meeqFS-1114 applications. An overall 34.85% of the mapped markers
and meeqFS-1115, respectively. Each of the remaining showed distorted segregation. These SDMs were unevenly
11 additive QTLs identified by Model I or Model II had distributed not only on chromosomes, but also between
epistatic interaction with one epistatic QTL locus. the At and Dt subgenomes, with c14 containing the largest
number of SDMs and c14 and c25 having the highest
DISCUSSION percentage of SDMs (Table 2). This severe segregation
Genetic Map Construction distortion in our population was first observed by Sun et
Using SNP Markers al. (2012), then subsequently by Jamshed et al. (2016) and
Due to their abundance and stability, SNP markers are Zhang et al. (2016). Segregation distortion has also observed
increasingly being applied to genetic map construction in other populations (Shen et al., 2005, 2007; Zhang et al.,
(Wang et al., 2014, 2015; Tian et al., 2015). Compared 2009; Tang et al., 2015). Various factors including genetic
Table 4. The epistasis effects between additive and/or epistatic quantitative trait loci (QTLs).
Phenotypic variation explained by epistasis
QTL 1 Chromosome QTL 2 Chromosome Total Main Environmental
meeqFS-111 2 qFS-c71 7 1.99 1.78 0.22
meaqFS-c71 7 meeqFS-215 9 2.17 2.04 0.13
qFS-c71 7 meeqFS-215 9 2.17 2.04 0.13
meaqFS-c41 4 meeqFS-231 12 2.04 1.95 0.09
qFS-c11 1 meeqFS-256 17 2.01 1.93 0.08
meaqFS-c43 4 meeqFS-269 18 1.79 1.63 0.16
qFS-c41 4 meeqFS-269 18 1.79 1.63 0.16
meaqFS-c91 9 qFS-c201 20 2.52 2.35 0.17
qFS-c11 1 meeqFS-290 21 1.74 1.55 0.20
meeqFS-191 6 qFS-c212 21 2.05 1.75 0.30
qFS-c201 20 meeqFS-295 21 1.87 1.71 0.16
qFS-c73 7 meeqFS-299 22 2.17 2.07 0.10
meaqFS-c131 13 meeqFS-2100 21 1.83 1.61 0.22
meeqFS-1103 3 qFS-c221 22 1.63 1.45 0.18
meeqFS-1114 1 meaqFS-c231 23 1.80 1.71 0.10
meeqFS-1115 6 meaqFS-c231 23 2.44 2.19 0.25
qFS-c111 11 meeqFS-2122 24 1.71 1.64 0.08
qFS-c211 21 meeqFS-2124 24 1.87 1.69 0.18
qFS-c61 6 meeqFS-2126 24 1.68 1.40 0.28
meaqFS-c161 16 meeqFS-2137 25 1.68 1.43 0.25
qFS-c71 7 meeqFS-2149 26 1.84 1.78 0.06

crop science, vol. 57, march april 2017 www.crops.org 783

Fig. 5. Comparison of the maps constructed by simple sequence repeat (SSR) markers and SSR + single-nucleotide polymorphism
(SNP) markers: (a) comparison of the maps of chromosome seven constructed by SSR markers and SSR + SNP markers, (b) comparison
of the maps of chromosome 13 constructed by SSR markers and SSR + SNP markers, and (c) comparison of the maps of chromosome
25 constructed by SSR markers and SSR + SNP markers.

784 www.crops.org crop science, vol. 57, march april 2017

drift (Zhang et al., 2009) and environment factors (Jamshed
et al., 2016) may have led to the segregation distortion.
Within in the current population, the composition
characteristics may also account for some of the segregation
distortion (Jamshed et al., 2016).
The collinearity analysis indicated that the linkage
map described herein did not have full coverage of the
physical map, which is illustrated by the total map length
of 2865.73 cM (Table 2). The current estimate is a 4500-
cM physical map in tetraploid cotton, which includes G.
hirsutum and G. barbadense (Rong et al., 2004). Few of the
cotton linkage maps constructed ever reach this estimation
(Sun et al., 2012; Ning et al., 2014; Hulse-Kemp et al.,
2015; Wang et al., 2015).
The genome sequencing database of G. hirsutum
reveals an estimation of At and Dt subgenomes of 1.7 Gb
and 800 Mb, respectively, in physical size (Zhang et al.,
2015a; Li et al., 2015a). Although the physical size of the
At subgenome was almost twice that of the Dt subgenome,
the total linkage map length of the At subgenome was
slightly shorter than that of the Dt subgenome (Table
2, Supplemental Table S4). The At and Dt subgenomes
showed different crossover or recombination rates, with
1 cM of linkage map length representing an average of
1231 kb in the At subgenome chromosomes and 539 kb in
the Dt subgenome chromosomes. Both values are a little
bit higher than the estimations based on the genome sizes
of Gossypium arboreum L. and Gossypium raimondii Ulbr.
(Hulse-Kemp et al., 2015).
Another intraspecific linkage map was also constructed
with the CottonSNP63K array (Hulse-Kemp et al., 2015).
It covered a total genetic distance of 3490 cM, with several
sizable gaps, the largest of which was 56 cM in c21. The
Hulse-Kemps map (Hulse-Kemp et al., 2015) consisted of
more than 7000 SNPs, with only 1176 cosegregation bins.
In our linkage map, 2393 markers were mapped with a
total genetic distance of 2865.73 cM, with a few moderate
gaps and small cosegregation bins. Common SNP markers
between the current study and the Hulse-Kemps map
were searched for, but due to the lack of full marker
information of cosegregation bins in the Hulse-Kemps
map (Hulse-Kemp et al., 2015), only a few common SNP
markers were found. The map also featured a consensus
with SSR markers identified by Sun et al. (2012). (Table
2, Supplemental Table S4). This consensus provides direct
evidence of the consistency of the current linkage map
with the previous one (Sun et al., 2012) constructed with
the same RIL population of different generations.
Reliability of the Epistatic QTLs
Identified by Model II
In this study, permutation tests of 1000 times were
used to control the strictness of the epistatic QTL pair
identification. As the epistasis effect can have a high false
crop science, vol. 57, march april 2017 www.crops.org 785
positive rate, after permutation tests of 1000 times, the
software still identified 156 pairs QTL pairs endowed with
a high enough LOD value to indicate their reliability. The
percentage of the observed total PV of epistasis that the
epistatic QTL pairs could explain might also be taken
into consideration to assess the reliability of the QTL pair.
Therefore, the QTLs pairs that had a higher LOD value
and could also explain a higher percent of the observed
total PV of epistasis should be considered first have true
epistasis effect (Fig. 4c, Supplemental Table S8). Still,
whether all the 156 QTLs pairs have epistasis effect or not,
is still needed to be confirmed in further studies.
Epistasis Effect Analysis of the Additive QTLs
Identified by Models I or II
Obviously, the additive QTLs identified, whether by Model
I or II, have biased distributions on the chromosomes
throughout the whole genome. This phenomenon was also
observed in various studies (Shen et al. 2005, 2007; Sun et
al., 2012; Wei et al., 2014; Tan et al., 2015; Shang et al.,
2015; Zhang et al., 2015b; Jamshed et al., 2016). The epistatic
QTLs identified by Model II do not show significant bias
distributions throughout the genome except that some
chromosomes harbor a few more or less epistatic QTLs.
We reported a total of nine additive QTLs identified
by Model I and seven additive QTLs identified by Model
II that have an epistatic interaction with an epistatic QTL
locus. The additive QTL qFS-c71 had interaction with
three epistatic QTL loci. It was also confirmed by both
models to have a major contribution to the formation of
FS. The QTL qFS-c11 had interaction with two epistatic
QTL loci. It was one of the few stable QTLs that had
a favorable alleles derived from the low FS value parent
sGK9708 and was also one of the QTLs that could explain
the largest observed PVs. qFS-c201 is a newly identified
stable QTL that had an epistatic interaction with two
QTLs, the additive QTL meaqFS-c91 and the epistatic
QTL meeqFS-295 identified by Model II. meaqFS-c231
was the only additive QTL identified by Model II to have
interactions with two epistatic QTL loci. These additive
QTLs with epistasis effects need special attention when
they are applied to MAS in future breeding programs,
especially the additive qFS-c201 and meaqFS-c91, as
their mutual epistatic interaction indicates a synergistic
effect when both are selected.
Congruence Analysis
with Previously Reported QTLs
Composite interval mapping (Model I) is a commonly used
method to detect QTLs. When CIM is used to identify QTLs
across multiple environments, the related trait is regarded as
multiple ones; therefore, CIM is unable to identify the QTL
by environment interactions based on multiple environments.
The algorithm used in CIM also cannot completely ensure
that the effect of QTL at the current testing interval is not
absorbed by the background marker variables and may thus
result in biased estimation of the QTL effect (Zeng, 1994; Li
et al., 2007). Inclusive composite interval mapping (Model II)
was designed to improve CIM with the modified algorithm
of the latter being able to render CIM more inclusive of
all marker data (Li et al., 2007, 2015b; Meng et al., 2015).
Especially when phenotypic data are evaluated and collected
through multiple locations and/or years, the built-in
environment interactions provided by multiple environment
data would be interpreted and investigated effectively with
the ICIM method. Such built-in environment interactions
could contribute useful information to the efficient use of
MAS in cotton breeding and better understanding of genetic
architecture of important quantitative traits. Therefore,
the two QTL identification models in the current study
emphasize different considerations. Model I considered the
QTLs identified in a single environment, and thus the stability
of a QTL was detected across multiple environments. Model
II emphasized an integration of various environments to
identify QTLs. The advantage of Model II was its detection
of the epistatic QTLs besides the major effects of the additive
QTLs. Therefore, it is reasonable that some of the QTLs
were solely identified by one of the two models. The QTLs
identified in both models should be emphasized first to do
the next step of research activities.
As one of the most important component traits of
fiber quality, FS, has been the focus in plenty of QTL
identification studies (Said et al., 2013). Several major
QTLs identified from different studies have been uploaded
into the database cottonqtlgb.org (Zhang et al., 2003;
Shen et al., 2005, 2007; Said et al., 2013; Fang et al., 2014;
Ning et al., 2014; Tan et al., 2015). We meta-analyzed
the previously reported QTLs and compared them with
QTLs identified in the current study (Supplemental Table
S9). Compared with the metaQTL, of the 16 stable QTLs
identified by Model I and 16 additive QTLs identified by
Model II, seven QTLs identified by Model I (qFS-c73,
qFS-c111, qFS-c112, qFS-c201, qFS-c211, qFS-c212,
and qFS-c221) and seven QTLs identified by Model II
(meaqFS-c41, meaqFS-c42, meaqFS-c72, meaqFS-c73, Arcade, A., A. Labourdette, M. Falque, B. Mangin, F. Chardon,
meaqFS-c74, meaqFS-c161, and meaqFS-c162) were
newly identified additive FS QTLs. All the remained
QTLs were reported in prior research.
Among these previously reported QTLs, the most
interesting finding is that the QTLs reported in the current
study are more precisely mapped. The QTL identified on c7
provides a very good explanation example for such finding.
The QTL metaqFS-c71 was previously report by several
researchers as a major QTL controlling FS (Supplemental
Table S9) (Zhang et al., 2009; Sun et al., 2012; Zhang et
al., 2012; Fang et al., 2014; Ning et al., 2014; Jamshed et al.,
2016). In the confidence interval of this QTL, we identified
two FS-controlling QTLs, qFS-c71 and qFS-c72, each
786 www.crops.org crop science, vol. 57, march april 2017
of which has a much more concise confidence interval,
explaining 4.72 to 13.55% and 6.05 to 8.73% of the
observed PVs, respectively, in the current study (Table
3). qFS-c41 could also be an important QTL identified
in this study, as it was not only identified as stable across
multiple environments and confirmed by both Models I
and II simultaneously but also was identified by previous
studies (Supplemental Table S9) (Said et al., 2013; Jamshed
et al., 2016). The physical positions of the markers in the
confidence interval (Supplemental Table S10) indicate that
qFS-c41 could possibly be the same as metaqFS-c41, as
the physical positions of the markers in their confidence
intervals are located at less than 2 Mb (Supplemental Tables
S9 and S10). Chromosome c23 was estimated to harbor
five QTLs conferring FS from meta-analysis (Supplemental
Table S9). The only QTL that was identified by Model
II to have two epistatic QTL loci, meaqFS-c231, could
be metaqFS-c235, as their physical positions were quite
adjacent (Supplemental Tables S9 and S11). The QTLs for
which the physical positions of their markers are in adjacent
confidence intervals may need further confirmation as to
whether they were the same or not.
Supplemental Material Available
Supplemental material for this article is available online.
Acknowledgments
This work was funded by the Natural Science Foundation of
China (31371668, 31471538), the National High Technology
Research and Development Program of China (2012AA101108),
the National Agricultural Science and Technology Innovation
Project for CAAS, and the Henan Province Foundation with
cutting-edge technology research projects (142300413202).
Thanks to the Biomarker Technologies Corporation (Beijing,
China) for providing the software Highmap and help in the
genetic map construction with Highmap. Thanks to the
Quantitative Genetics Group of CAAS (Beijing, China)
providing the software ICIMAPPING and help in QTL
identification with Model II.
References
A. Charcosset et al. 2004. BioMercator: Integrating genetic
maps and QTL towards discovery of candidate genes.
Bioinformatics 20:23242326. doi:10.1093/bioinformatics/
bth230
Burstin, J., P. Marget, M. Huart, A. Moessner, B. Mangin, C.
Duchene et al. 2007. Developmental genes have pleiotropic
effects on plant morphology and source capacity, eventually
impacting on seed protein content and productivity in pea.
Plant Physiol. 144:768781. doi:10.1104/pp.107.096966
Fang, D.D., J.N. Jenkins, D.D. Deng, J.C. McCarty, P. Li, and
J. Wu. 2014. Quantitative trait loci analysis of fiber quality
traits using a random-mated recombinant inbred population
in Upland cotton (Gossypium hirsutum L.). BMC Genomics
15:397. doi:10.1186/1471-2164-15-397
Guo, Y.S., G.L. Shi, Z.D. Liu, Y.H. Zhao, X.X. Yang, J.C. Zhu et
al. 2015. Using specific length amplified fragment sequencing
to construct the high-density genetic map for Vitis (Vitis
vinifera L. Vitis amurensis Rupr.). Front. Plant Sci. 6:393.
Hulse-Kemp, A.M., J. Lemm, J. Plieske, H. Ashrafi, R. Buyyarapu,
D.D. Fang et al. 2015. Development of a 63K SNP array
for cotton and high-density mapping of intraspecific and
interspecific populations of Gossypium spp. G3: Genes,
Genomes, Genet. 5:1187209 doi:10.1534/g3.115.018416
Jamshed, M., F. Jia, J.W. Gong, K.K. Palanga, Y.Z. Shi, J.W.
Li et al. 2016. Identification of stable quantitative trait loci
(QTLs) for fiber quality traits across multiple environments in
Gossypium hirsutum recombinant inbred line population. BMC
Genomics 17:197. doi:10.1186/s12864-016-2560-2
Jansen, J., A.G. de Jong, and J.W. van Ooijen. 2001. Constructing
dense genetic linkage maps. Theor. Appl. Genet. 102:1113
1122. doi:10.1007/s001220000489
Kosambi, D.D. 1943. The estimation of map distance
from recombination values. Ann. Eugen. 12:172175.
doi:10.1111/j.1469-1809.1943.tb02321.x
Lacape, J.M., J. Jacobs, T. Arioli, R. Derijcker, N. Forestier-
Chiron, D. Llewellyn et al. 2009. A new interspecific,
Gossypium hirsutum G. barbadense, RIL population: Towards
a unified consensus linkage map of tetraploid cotton. Theor.
Appl. Genet. 119:281292. doi:10.1007/s00122-009-1037-y
Lejeune-henaut, I., E. Hanocq, L. Bethencourt, V. Fontaine,
B. Delbreil, J. Morin et al. 2008. The flowering locus Hr
colocalizes with a major QTL affecting winter frost tolerance
in Pisum sativum L. Theor. Appl. Genet. 116:11051116.
doi:10.1007/s00122-008-0739-x
Li, F.G., G.Y. Fan, C.R. Lu, G.H. Zhao, C.S. Zou, R.J. Kohel
et al. 2015a. Genome sequence of cultivated Upland cotton
(Gossypium hirsutum TM-1) provides insights into genome
evolution. Nat. Biotechnol. 33:524530. doi:10.1038/
nbt.3208
Li, S.S., J.K. Wang, and L.Y. Zhang. 2015b. Inclusive composite
interval mapping of QTL by environment interactions
in biparental populations. PLoS ONE 10(7):e132414.
doi:10.1371/journal.pone.0132414
Li, H.H., G.Y. Ye, and J.K. Wang. 2007. A modified algorithm
for the improvement of composite interval mapping. Genetics
175:361374. doi:10.1534/genetics.106.066811
Liu, D.Y., C.X. Ma, W.G. Hong, L. Huang, M. Liu, H. Liu et
al. 2014. Construction and analysis of high-density linkage
map using high-throughput sequencing data. PLoS ONE
9(6):e98855. doi:10.1371/journal.pone.0098855
Ma, J.Q., L. Huang, C.L. Ma, J.Q. Jin, C.F. Li, R.K. Wang et
al. 2015. Large-scale SNP discovery and genotyping for
constructing a high-density genetic map of tea plant using
specific-locus amplified fragment sequencing (SLAF-
seq). PLoS ONE. 10(6):e0128798. doi:10.1371/journal.
pone.0128798
Maccaferri, M., A. Ricci, S.G. Milner, E. Noli, P.L. Martelli, R.
Casadio et al. 2015. A high-density, SNP-based consensus map
of tetraploid wheat as a bridge to integrate durum and bread
wheat genomics and breeding. Plant Biotechnol. J. 5:648663.
Meng, L., H.H. Li, L.Y. Zhang, J.K. Wang. 2015. QTL IciMapping:
Integrated software for genetic linkage map construction and
quantitative trait locus mapping in biparental populations.
The Crop J. 3:26283. doi:10.1016/j.cj.2015.01.001
Ning, Z.Y., H. Chen, H.X. Mei, and T.Z. Zhang. 2014. Molecular
crop science, vol. 57, march april 2017 www.crops.org 787
tagging of QTLs for fiber quality and yield in upland cotton
cultivar Acala-Prema. Euphytica 195:143156. doi:10.1007/
s10681-013-0990-3
Park, A.H., U.S. Alabady, M. Ulloa, B. Sickler, T.A. Wilkins,
J.Z. Yu et al. 2005. Genetic mapping of new cotton fiber
loci using EST-derived microsatellites in an interspecific
recombinant inbred line cotton population. Mol. Genet.
Genomics 274:428441. doi:10.1007/s00438-005-0037-0
Rong, J.K., C. Abbey, J.E. Bowers, C.L. Brubaker, C. Chang, P.W.
Chee et al. 2004. A 3347-locus genetic recombination map of
sequence-tagged sites reveals features of genome organization,
transmission and evolution of cotton (Gossypium). Genetics
166:389417. doi:10.1534/genetics.166.1.389
Said, J.I., Z.X. Lin, X.L. Zhang, M.Z. Song, and J.F. Zhang. 2013.
A comprehensive meta QTL analysis for fiber quality, yield,
yield related and morphological traits, drought tolerance,
and disease resistance in tetraploid cotton. BMC Genomics
14:776. doi:10.1186/1471-2164-14-776
Shang, L.G., Q.Z. Liang, Y.M. Wang, X.C. Wang, K.B. Wang,
A. Abduweli et al. 2015. Identification of stable QTLs
controlling fiber traits properties in multi-environment
using recombinant inbred lines in upland cotton (Gossypium
hirsutum L.). Euphytica 205:877888. doi:10.1007/s10681-
015-1434-z
Shen, X.L., W.Z. Guo, Q.X. Liu, X.F. Zhu, Y.L. Yuan, and T.Z.
Zhang. 2007. Genetic mapping of quantitative trait loci for
fiber quality and yield trait by RIL approach in upland cotton.
Euphytica 155:371380. doi:10.1007/s10681-006-9338-6
Shen X.L., W.Z. Guo, X.F. Zhu, Y.L. Yuan, J.Z. Yu, R.J. Kohel
et al. 2005. Molecular mapping of QTLs for fiber qualities in
three diverse lines in Upland cotton using SSR markers. Mol.
Breed. 15:169181. doi:10.1007/s11032-004-4731-0
Sun, F.D., J.H. Zhang, S.F. Wang, W.K. Gong, Y.Z. Shi, A.Y.
Liu et al. 2012. QTL mapping for fiber quality traits across
multiple generations and environments in upland cotton.
Mol. Breed. 30:569582. doi:10.1007/s11032-011-9645-z
Tan, Z.Y., X.M. Fang, S.Y. Tang, J. Zhang, D.J. Liu, Z.H. Teng et
al. 2015. Genetic map and QTL controlling fiber quality traits
in upland cotton (Gossypium hirsutum L.). Euphytica 203:615
628. doi:10.1007/s10681-014-1288-9
Tang, S.Y., Z.H. Teng, T.F. Zhai, X.M. Fang, F. Liu, D.J. Liu et
al. 2015. Construction of genetic map and QTL analysis of
fiber quality traits for upland cotton (Gossypium hirsutum L.).
Euphytica 201:195213. doi:10.1007/s10681-014-1189-y
Tian, H.L., F.G. Wang, Jr., H.M. Zhao, L. Yi, L. Wang, R.
Wang, et al. 2015. Development of maizeSNP3072, a high-
throughput compatible SNP array, for DNA fingerprinting
identification of Chinese maize varieties. Mol. Breed. 35(6):1
11. doi:10.1007/s11032-015-0335-0
Timmerman-Vaughan, G.M., T.J. Frew, R. Butler, S. Murray,
M. Gilpin, and K. Falloon. 2004. Validation of quantitative
trait loci for Ascochyta blight resistance in pea (Pisum sativum
L.), using populations from two crosses. Theor. Appl. Genet.
109:16201631. doi:10.1007/s00122-004-1779-5
van Ooijen, J.W. 2011. Multipoint maximum likelihood mapping
in a full-sib family of an outbreeding species. Genet. Res.
93:343349. doi:10.1017/S0016672311000279
van Os, H., P. Stam, G.F. Visser, G.F. Rechard, and H.J. van Eck.
2005. SMOOTH: A statistical method for successful removal of
genotyping errors from high-density genetic linkage data. Theor.
Appl. Genet. 112:187194. doi:10.1007/s00122-005-0124-y
Wang, S.C., C.J. Basten, and Z.B. Zeng. 2001. Windows QTL
Cartographer 2.5. North Carolina State Univ., Raleigh, NC.
http://statgen.ncsu.edu/qtlcart/WQTLCart.htm (accessed 1
Feb. 2013).
Wang, S.C., D. Wong, K. Forrest, A. Allen, S.M. Chao, B.E.
Huang et al. 2014. Characterization of polyploid wheat
genomic diversity using a high-density 90000 single
nucleotide polymorphism array. Plant Biotechnol. J. 12:787
796. doi:10.1111/pbi.12183
Wang, Y.K., Z.Y. Ning, Y. Hu, J.D. Chen, R. Zhao, H. Chen
et al. 2015. Molecular mapping of restriction-site associated
DNA markers in allotetraploid upland cotton. PLoS ONE
10(4):e0124781. doi:10.1371/journal.pone.0124781
Wei, Q.Z., Y.Z. Wang, X.D. Qin, Y.X. Zhang, Z.T. Zhang, J.
Wang et al. 2014. An SNP-based saturated genetic map
and QTL analysis of fruit-related traits in cucumber using
specific-length amplified fragment (SLAF) sequencing. BMC
Genomics 15:1158. doi:10.1186/1471-2164-15-1158
Zeng, Z.B. 1994. Precision mapping of quantitative trait loci.
Genetics 136:14571468.
Zhang, D.D., Y.P. Hua, X.H. Wang, H. Zhao, L. Shi, and F.S.
Xu. 2014. A high-density genetic map identifies a novel major
QTL for boron efficiency in oilseed rape (Brassica napus L.).
PLoS ONE 9(11):e112089. doi:10.1371/journal.pone.0112089
Zhang, K., J. Zhang, J. Ma, S.Y. Tang, D.J. Liu, Z.H. Teng et al.
2012. Genetic mapping and quantitative trait locus analysis of
fiber quality traits using a three-parent composite population
in upland cotton (Gossypium hirsutum L.). Mol. Breed. 29:335
348. doi:10.1007/s11032-011-9549-y
788 www.crops.org crop science, vol. 57, march april 2017
Zhang, T.Z., Y. Hu, W.K. Jiang, L. Fang, X.Y. Guan, J.D. Chen
et al. 2015a. Sequencing of allotetraploid cotton (Gossypium
hirsutum L. acc. TM-1) provides a resource for fiber
improvement. Nat. Biotechnol. 33(5):531537. doi:10.1038/
nbt.3207
Zhang, T.Z., Y.L. Yuan, J. Yu, W.Z. Guo, and R.J. Kohel. 2003.
Molecular tagging of a major QTL for fiber strength in upland
cotton and its marker-assisted selection. Theor. Appl. Genet.
106:262268. doi:10.1007/s00122-002-1101-3
Zhang, Z., J.W. Li, J. Muhammad, J. Cai, F. Jia, Y.Z. Shi et al.
2015b. High resolution consensus mapping of quantitative trait
loci for fiber strength, length and micronaire on chromosome
25 of the upland cotton (Gossypium hirsutum L.). PLoS ONE
10(8):e0135430. doi:10.1371/journal.pone.0135430
Zhang, Z., H.H. Shang, Y.Z. Shi, L. Huang, J.W. Li, Q. Ge et al.
2016. Construction of a high-density genetic map by specific
locus amplified fragment sequencing (SLAF-seq) and its
application to quantitative trait loci (QTL) analysis for boll
weight in upland cotton (Gossypium hirsutum.). BMC Plant
Biol. 16:79. doi:10.1186/s12870-016-0741-4
Zhang, Z.S., M.C. Hu, J. Zhang, D.J. Liu, J. Zheng, K. Zhang, et
al. 2009. Construction of a comprehensive PCR based marker
linkage map and QTL mapping for fiber quality traits in
upland cotton (Gossypium hirsutum L.). Mol. Breed. 24:4961.
doi:10.1007/s11032-009-9271-1