Tree Genetics & Genomes (2010) 6:915929

DOI 10.1007/s11295-010-0301-2

ORIGINAL PAPER

Genomic mapping and testing for quantitative trait loci

in tea (Camellia sinensis (L.) O. Kuntze)
S. M. Kamunya & F. N. Wachira & R. S. Pathak &
R. Korir & V. Sharma & R. Kumar & P. Bhardwaj &
R. Chalo & P. S. Ahuja & R. K. Sharma

Received: 7 November 2008 / Revised: 4 May 2010 / Accepted: 6 May 2010 / Published online: 8 June 2010
# Springer-Verlag 2010

Abstract The tea industry is significant in the economies

of tea-growing countries. Prospects of improving yield of
made tea genomic information were explored using
clones from a cross between clones TRFCA SFS150 and
AHP S15/10. The 42 clones were tested in two distinct
tea-growing regions in Kenya. Bulk segregant analysis
was performed followed by complete genotyping. Out of
260 informative markers, 100 markers that showed 1:1
segregation were used to construct a linkage map. The
map contained 30 (19 maternal and 11 paternal) linkage
groups that spanned 1,411.5 cM with mean interval of
14.1 cM between loci. Based on the map, quantitative
trait loci (QTL) analysis was done on yield data over

20032007 across the two sites, Timbilil and Kangaita.

Twenty-three putative QTLs were detected, 16 in five
different linkage groups for Timbilil, two in two groups
for Kangaita, and the rest were associated with unassigned markers. No QTL was detected at both sites, which
showed strong genotype site interaction (G E) but
highly effective within-site heritability (h^2 generally>
0.7). Problems of overestimated and spurious QTL effects
arising from the smallness of the population should be
mitigated by generally high within-site heritability. At
least two unassigned markers associated with yield at
Kangaita over the whole study period, suggesting potential as candidate markers for site-specific marker-assisted
selections. Implications of the results with respect to
mapping population, G E, and marker-assisted selection
are discussed.

Communicated by R. Burdon

Keywords Camellia sinensis . Genotypes environment

interaction . QTL mapping . Tea . Yield . Kenya

S. M. Kamunya : V. Sharma : R. Kumar : P. Bhardwaj :

P. S. Ahuja : R. K. Sharma (*)
Division of Biotechnology, Institute of Himalayan Bioresource
Technology, IHBT, (CSIR),
Post Box 6, Palampur, Himachal Pradesh 176061, India
e-mail: mrk_sharma@yahoo.com
e-mail: ramsharma@ihbt.res.in
S. M. Kamunya : F. N. Wachira : R. Korir : R. Chalo
Tea Research Foundation of Kenya,
P.O. Box 820, Kericho 20200, Kenya
F. N. Wachira
Biochemistry and Molecular Biology Department,
Egerton University,
P.O. Box 536-20115, Njoro, Kenya
R. S. Pathak
Department of Crops, Horticulture and Soil Sciences,
Egerton University,
P.O. Box 536-20115, Njoro, Kenya

Introduction
Tea is one of the most widely consumed soft beverages
in the world and plays a significant role to the economy of
all the tea-producing countries, including Kenya where tea
is the leading foreign exchange earner and export commodity. However, the future of tea industry depends upon
the availability of high-yielding, black tea quality and
drought-tolerant tea clones. The tremendous improvement
in Kenyan tea production over the years is largely attributed
to development and release of high-yielding and betterquality clones, which have gradually replaced most of the
pioneer seedling plantations (Wachira 2002). Kenyan tea
has a world-class standing for high black tea quality and

916

yields. However, it has not yet been possible to produce

clones with optimum leaf yield and tolerance to important
biotic and abiotic stress factors such as the ever-increasing
incidence of drought (Kamunya and Wachira 2004;
Kamunya et al. 2004). Although the newly developed
and released cultivars are products of rigorously and
highly protracted breeding and clonal selection (Gazi
1978), tea improvement process would have been much
enhanced with the application of reliable and stable
selection tools that are not prone to the continually
changing environment.
Most of the agronomic traits of tea are quantitative in
nature and therefore not amenable to easy manipulations in
breeding programs without elaborate and long-term field
testing in at least more than one environment in order to
determine their inheritance, adaptability, and stability. It is
envisaged that identification of molecular markers linked to
major quantitative trait loci (QTLs) for the most desirable
traits would be utilized for early selection of elite clones,
thereby saving on time, cost of maintenance, and land
resources. Though not necessarily genes themselves,
QTLs are stretches of DNA that are closely linked to the
genes that underlie the trait in question. QTLs can be
molecularly identified [for example, with random amplified polymorphic DNA (RAPD), simple sequence repeat
(SSR), or amplified fragment length polymorphism
(AFLP)] to help map regions of the genome that contain
genes involved in specifying a quantitative trait. This can
be an early step in identifying and sequencing these genes.
Based on detection of QTLs and selection of advantageous
traits, markers linked to complex traits can be used to
select against negative characteristics, in a negative
selection program, while others could be used to select
parents that would give rise to progeny with desired
genotypes. Molecular markers are better utilized in QTL
analysis if they are already placed in linkage maps, which
enable precise positioning of QTLs. While such a linkage
map for tea exists (Hackett et al. 2000), there is no report
on QTL analysis in tea to date. However, similar studies
on other tree crops have demonstrated the feasibility of
marker-assisted selection (Brown et al. 2003; Grattapaglia
and Sederoff 1994; Missiaggia et al. 2005; Yang et al.
1997; Zhang et al. 2004), even though genotype by
environmental (GE) interactions and varying biotic and
abiotic factors on a seasonal or yearly basis markedly
affected detection of QTLs.
From the foregoing, the current study set out to
identify and dissect QTLs controlling yield in tea with
the aim of initiating marker-assisted selection and
breeding in tea improvement program. To our knowledge, this is the first report on performing actual QTL
mapping in tea. Findings are preliminary, and the QTL
are necessarily putative.

Tree Genetics & Genomes (2010) 6:915929

Materials and methods

Mapping population and experimental sites
Genetic materials for construction of a linkage map and
mapping of QTLs governing attributes of agronomic
importance included the pseudo-test progeny of two
heterozygous parental clones TRFCA SFS150 (female)
and AHP S15/10 (male). Clone TRFCA SFS150 is a
Malawian Assam type that is a moderate yielder, has
tolerance to drought, cold, and pest, has sparse pubescence
on shoots, and has moderate levels of caffeine (2.9%),
while clone AHP S15/10 is an Assam-type Kenyan local
selection that is high-yielding, with dense pubescent shoots,
but susceptible to water stress and contains moderate levels
of caffeine (3.0%). Although the two clones belong to the
Assam taxon, the cross was chosen on the basis of the
differing parental attributes and moderate genetic similarity
of 67% (Wachira 2002). The cross comprising 42 clonal
progeny was established in 2000 in two sites, one each at
the two research stations of Tea Research Foundation of
Kenya, in Timbilil (Kericho District) and Kangaita
(Kirinyaga District) (Table 1). Cuttings were collected from
individual seedling bushes (ortets), rooted, and raised in the
nursery for 1 year prior to field transplanting. The trial was set
up in a completely randomized block design with three
replications in clonal plots of 30 ramets spaced at 0.61 m
within rows and 1.22 m between rows (i.e., 13,448 plants per
hectare) in each site. Each replicate was surrounded by a
guard row of clone TRFK 303/1199. The trial was brought
into bearing following the recommended management
practices and applied with fertilizer at rate of 150 kg N per
hectare per year in the form of NPKS 25:5:5:5 compound
fertilizer (Anon 2002). The trial was subjected to maintenance
pruning in 2004 at Kangaita and 2005 at Timbilil.
Crop yield evaluation
Yield data collection of the plucked two leaves and a bud
commenced in February 2001 and continued up to
December 2007. Harvesting was carried out at average
intervals of 7 to 10 days depending on availability of crop.

Phenotypic data analysis

The cumulative yield data were converted from green leaf
weight to annual mean yield by dividing it by the number of
years since first plucking. The green leaf yield was converted
to made tea per hectare by a conversion factor of 0.225 prior
to analysis. Only yield data for 2003, 2006, 2007, and longterm annual yield means for 2006 (ANYLD06) and 2007
(ANYLD07) were subjected to QTL analysis.

917

Tree Genetics & Genomes (2010) 6:915929

Table 1 Climatic data between 2000 and 2007 and site characteristics for Timbilil and Kangaita
Year/site characteristics

Timbilil

Kangaita

Rainfall (mm)

Annual mean air temp (C)

Rainfall (mm)

Annual mean air temp (C)

2000
2001
2002
2003
2004

1,684
2,287
2,039
2,062
1,916

16.6
16.5
16.8
16.8
16.4

1,685
1,903
2,222
2,277
1,686

15.1
15.1
15.2
15.3
15.0

2005
2006
2007
Mean
Rainfall distribution
Location
Altitude
Soil classification

1,961
2,558
2,586
2,137
Even
0 22 S, 35 21 E
2,180 m
Clay

16.7
16.9
16.6
16.7

1,626
2,812
2,196
2,051
Weakly bimodal
0.5 S, 37.3 E
2,100 m
Sandy loams

15.6
15.5
15.6
15.3

The analysis of variance for yield across sites and years

was carried out using plot means on MSTAT-C software
based on factorial design combined over locations and years
with genotypes established in a randomized complete block
design within locations. However, owing to marked clone by
site interactions, variance components for heritability
estimation followed a simplified within-site randomized
block design model as shown in Table 2. The following
mathematical model was employed to facilitate derivation of
components of variance, statistical analysis, and data
interpretation:
Yij m gi bj eij ;
where Yij represents the phenotypic mean, gi the effect of ith
genotype, bj the effect of the j which is the randomized
complete block, and eij the residual error. All the effects were
considered random. Variances for the various effects were
estimated from the mean squares and their expected
composition shown in Table 2. Thus,


b
b2g =b
b2g = s
b2g s
b2w =b ;
h 2g s
s 2p s
where s 2g , s 2P , and s 2W are the components of genotypic
(clones), phenotypic, and residual (error) variances, respectively; and h2g is the effective heritability. Analysis of
variance for the data was performed using MSTAT-C
statistical software. The standard errors of heritability
estimates were calculated according to Wright (1976).
Genetic correlations among yield traits as measured in space
and time were also computed using the MSTAT-C software.
Spearmans rank correlation analysis was further applied
on clonal means at respective sites to corroborate the effects
of genotypeenvironment interaction on the traits assessed

at the two test sites. Rank correlations were obtained by

StatistiXL1.8 software.
DNA extraction, purification, and quantification
Young fresh leaf (two leaves and a terminal bud) was
harvested from the experiments and freeze-dried overnight
at 20C prior to DNA extraction using a modified protocol
of Gawel and Jarret (1991). All the samples were checked
for quality and quantity and stored at 20C.

Bulk segregant analysis

Bulk segregant analysis (BSA; Michelmore et al. 1991) was
used to target the genomic regions associated with yield. In
this case, two bulked DNA samples for yield obtained at
Timbilil were constructed using equal amounts of DNA
from ten top- and bottom-performing progeny. Two
hundred and fifty-two random 10-mer primers, 96 AFLP
primer combinations, and 15 SSR primers pairs (Freeman et
al. 2004) were then screened on the parents and all the
bulked DNA samples. Primers distinguishing both the
parents and corresponding bulks were subjected for further
BSA with individual progeny and fingerprinting of the
entire population.
RAPD analysis
For RAPD analysis, modified polymerase chain reaction
(PCR) conditions first described by Williams et al. (1990)
were used. The reaction was conducted in a 25-l volume
containing 20 ng genomic DNA template, 20 ng of a single

Tree Genetics & Genomes (2010) 6:915929

918
Table 2 Expectations of mean
squares for yield of F1 clonal
progeny

g and b are numbers of

genotypes and replications,
respectively

Source of variation

DF

MS

Components of variance

Genotypes
Blocks
Error
Total

g1
b 1
(g1) (b1)
gb1

MS3
MS2
MS1

s 2W bs 2g
s 2W gs 2b
s 2W

decamer primer (Operon Technologies, Alameda, CA),

200 M each of dNTPs 1X Taq polymerase buffer with
1.5 mM MgCl2, and 0.5 U of Taq polymerase (Bangalore
Genei Pvt Ltd., Bangalore, India). All the PCR reactions
were performed on I-Cycler PCR system (Bio-Rad,
Australia) as follows: 94C for 5 min, 45 cycles of 94C
for 1 min, 37C for 1 min, 72C for 2 min, and one
extension cycle of 72C for 7 min. All the amplified
products were resolved on 1.5% agarose gels and stained
with ethidium bromide and sized with 50- and 100-bp Plus
DNA ladders as size standards (MBI Fermentas, Lithuania).
DNA fragments were visualized under UV light and
documented with the Gel Doc XR system (Bio-Rad,
Australia).
AFLP analysis
AFLP analysis was performed following the protocol of
Key-gene N.V. (Zabeau and Vos 1993) with slight modifications. Genomic DNA (250 ng) was restricted with MseI
and EcoRI (1.5 U of each) in restriction buffer (50 mM Tris
HCl, pH 7.5, 50 mM magnesium acetate, 250 mM potassium
acetate) in a total volume of 25 l. EcoRI and MseI adapters
were subsequently ligated to the digested DNA fragments.
The adapter-ligated DNA was diluted 1:10 and preamplified
using the following PCR cycling parameters: 20 cycles at
94C for 30 s, 56C for 60 s, and 72C for 60 s. The
preamplified DNA was then diluted 1:50 and used as a
template for selective amplification using+3 primers (EcoRI
and MseI). The EcoRI primer was labeled with 33P-ATP and
T4 polynucleotide kinase. The PCR parameters were: one
cycle at 94C for 30 s, 65C for 30 s, and 72C for 60 s. The
annealing temperature was lowered by 0.7C per cycle
during the next 12 cycles, followed by another 23 cycles at
94C for 30 s, 56C for 30 s ,and 72C for 60 s. PCR
reactions were carried out in I-Cycler PCR system (Bio-Rad,
Australia). The AFLP reaction products were mixed with an
equal volume of formamide dye [98% (v/v) formamide,
10 mM ethylenediaminetetraacetic acid (EDTA), 0.025%
bromophenol blue, and 0.025% (v/v) xylene cyanol] and
denatured at 94C for 4 min, then quickly cooled on ice.
Three microliters of each sample was loaded onto a 6% (w/v)
denaturing polyacrylamide gel and run in 1X Trisborate
EDTA electrophoresis buffer (0.089 M Tris, 0.089 M boric
acid, 0.002 M EDTA). The gel was dried for 2 h in a gel

drier before subjecting it to autoradiography for 3 to 14 days

at 70C depending on the signal intensity. The size of each
fragment was estimated using a 20-bp DNA ladder as a size
standard (M/s Cambrex Molecular appl.).
SSR analysis
Fifteen tea SSR primers developed by Freeman et al. (2004)
were utilized in the current study. Amplification was carried
out in a 10-l reaction mixture consisting of 10 PCR
assay buffer (Bangalore Genei Pvt. Ltd., India) with similar
component as mentioned for RAPD, except 12 ng each of
forward and reverse primers (Life Technologies, USA)
were used to amplify the microsatellites using the following
cycling parameters: initial denaturation at 94C for 5 min,
followed by 35 cycles of 94C for 1 min, 55C for 1 min,
72C for 2 min, and finally a primer extension cycle of
7 min at 72C. Two methods were used to obtain microsatellite markers to resolve the amplified products. Preferably, amplified products were resolved in 3% metaphor
agarose stained with ethidium bromide. However, allele
difference of5 bp was resolved by 6% polyacrylamide gel
and markers detected by silver staining.
Marker scoring
Alleles arising from codominant markers such as SSRs
were scored as a and b to denote markers from female
and male parents, respectively, while heterozygosity was
denoted by h. However, in dominant markers like RAPD
and AFLP, heterozygotes could not be distinguished from
one of the homozygous genotypes, and therefore, c was
used to denote genotype h-or-b or not a and d to
denote h-or-a or not b. Missing genotype observations,
where they arose, were denoted by . Only data from
intensely stained unambiguous bands were used for
statistical analysis. The names of individual marker loci
were described using the PCR primer code followed by the
actual size in base pairs.
Linkage analysis
As molecular data for outcrossing pedigrees, such as the
current population, produces different mating types within a
single parental cross (as opposed, for example, to mating of

919

Tree Genetics & Genomes (2010) 6:915929

inbred crosses where all loci refer to a particular mating

type (Williams 1998)), a mixture of backcross mating and
intercross (F2) types at some marker loci were obtained.
Informative mating types were the backcross with a
heterozygous maternal parent, i.e., Aaaa (or AA) or
heterozygous paternal parent aa (or AA)Aa, leading to a
segregation ratio 1:1 for both dominant and codominant
markers. In the intercross mating type, both paternal and
maternal are in heterozygous state AaAa leading to 3:1
segregation ratio for dominant markers and 1:2:1 for
codominant markers. Thus, all informative markers were
individually examined and categorized according to mating
type using Map Manager QTXb2.0 [21]. Then, single-locus
segregation analysis was carried out to identify distortions
in Mendelian ratios that are larger than that likely to occur
by random fluctuation alone.
A linkage map was constructed using Map Manager
QTX2.0 (Manly et al. 2001) using the pseudo-testcross
analysis as previously proposed by Grattapaglia and Sederoff
(1994). The Kosambi (1944) mapping function was used
with the threshold value of P=0.001 to convert the
recombination fraction into map distance. For this purpose,
the make linkage groups function was used. Additionally,
the ripple function was used to improve the order of loci in
a linkage group by testing alternative orders created by local
permutations of the locus order.
QTL analysis
Each of the measured traits was assessed for normal
distribution using a chi-square (2) goodness-of-fit test
using a computer software StatistiXL1.8 (www.statistixl.
com). QTL mapping was performed using genome-wide
single-marker regression analysis, interval, and where
necessary, composite interval mapping in Map Manager
QTX2.0 (2001). Thus, mean phenotypic trait data for each
site were computed and entered along with the mapping
data. The marker regression function (P=0.01 to 0.0001)
was used to declare the presence of a putative QTL
associated with each set of trait data. With this threshold,
an overall false-positive rate of 1% was expected, given
the average marker distance per linkage group. The locus
with the highest likelihood ratio statistic (LRS) for each set
of trait data was added to the background. Composite
interval mapping was then applied using the interval
mapping function. The locus in the background for each
trait was used to control for other QTLs. For each linkage
group carrying significant QTL, confidence intervals were
estimated by bootstrap resampling, and interval map figures
and histograms representing the confidence intervals of
peak LRS values were generated. Suggestive, significant,
and highly significant effects were generated by determining LRS thresholds for each trait and linkage group by at

least 1,000 permutations using the permutation test

function at P<0.01 (Churchill and Doerge 1994). Composite interval mapping was used to test for multiple QTLs and
ascertain the position of QTLs in the linkage groups (Zeng
1993, 1994) via Map Manager QTX2.0. Additionally, the
total phenotypic variance explained for each trait by all
putative QTLs was calculated with multiple regression
analysis using the trait as the dependent variable and the
previously identified markers, linked to the QTLs, as
treatments. Phenotypic variance explained by the QTL at
a particular locus was presented as percent and by
coefficient of determination (R2) in the multiple regressions.
The interaction or epistatic effects between two loci were
searched by interaction function set at stringent value of
P=105.

Results
Variation of yield across sites
Variations of annual mean yield (2001 to 2006 denoted as
ANYLD06) as assessed at both Timbilil (T) and Kangaita
(K) experimental sites are shown in Fig. 1a, b. The
ANYLD06-T among the progeny ranged from 1,670 to
2,412 kg made tea per hectare (mt/ha), while ANYLD06-K
ranged from 672 to 2,428 kg mt/ha. The F1 means for the
two sites were 2,180 and 1,504 kg mt/ha for ANYLD06-T
and ANYLD06-K, respectively. The midparent values
(MPV) were 2,163 and 2,013 kg mt/ha for ANYLD06-T
and ANYLD06-K, respectively. The midparent heterosis
for ANYLD06-T was marginal at 0.8%, while the one for
ANYLD06-K was high at 25.3%. While the best progeny
at Timbilil gave 11.5% more yields over MPV, the best
progeny at Kangaita outperformed MPV by 20.6%. The
normality of distribution for the trait as measured in both
sites when subjected to chi-square goodness-of-fit test
showed continuous distribution, possible site differences
notwithstanding. It is apparent from the Fig. 1a, b that yield
as assessed in the two sites did not behave the same way, as
the F1 means were higher and lower than the parental
means for Timbilil and Kangaita, respectively. In this case,
the data did not exhibit consistency as the average
heterozygotic values did not match either homozygotic
values nor were they midway between the two parental
values for both sites.
Genetic parameters
The analysis of variance (ANOVA) for the yield data for the
mapping family as recorded in the two sites and years as
well as interactions between them presented in Table 3
reveals highly significant site, clone, year, clonesite, and

Tree Genetics & Genomes (2010) 6:915929

920

Ranking on ANYLD06 performance across the two

sites was applied to further study the effects of genotype
environment interaction on the yield as assessed in the two
test sites. Spearmans rank correlations obtained by
StatistiXL1.8 software (www.statistixl.com) revealed considerable GE interaction (rs =0.08, P=0.59), further indicating
why none of the QTLs had congruent effects on yield in the
two sites (see Table 4).
Bulk segregant analysis for yield

Fig. 1 Frequency distribution of annual mean yield (ANYLD06-T) in

St 463 progeny at Timbilil site (a); frequency distribution of annual
mean yield (ANYLD06-K) in St 463 progeny at Kangaita site (b). P1,
P2, and F1 refers to means for parent 1 (female), parent 2 (male), and
first filial generation, respectively

siteyear interactions, while replication (site), cloneyear,

and clonesiteyear interactions were not significant (P>
0.05). Heritability estimates for yield within and across
years varied from 0.2 to 0.88 at Timbilil site, while it
remained more or less the same at Kangaita (0.68 to 0.73;
Table 4). Genetic correlation estimates exhibited in Table 5
were highly significant between various years records
within site but not between sites save for T06 and K03,
K06, K07, ANMK06, and ANMK07, which were weakly
significant (P<0.05).

Table 3 ANOVA for yield

across sites and years

Three interactions, yearclone,

siteyearclone, and replication(site)clone, which were
nonsignificant (F<1), have been
pooled with the residual to make
up error B

This experiment set out to rapidly identify QTLs associated

with annual mean yield up to 2006 since first plucking
(ANYLD06-T) at Timbilil site and then confirm them in the
entire population as well as in the second experimental site
at Kangaita. Two hundred and fifty-two RAPD decamer
primers, 15 SSR, and 96 AFLP primer pairs were screened
using BSA method with DNA bulks constructed using ten
each high- and low-yielding progenies. Of these, 21 RAPD
primers, 20 AFLP, and six SSR primer pairs produced
polymorphic bands between parents and corresponding
bulks. However, following exploratory genotyping the
individual progeny with informative primers, it was
discovered that, although some marker loci could discriminate the bulks and their respective parents during the
screening process, in most cases the pattern was not
reproducible upon genotyping. The patterns were mostly
confounded by appearance of recombinants in either of the
two classes. Thus, all polymorphic primers across the three
marker systems were utilized in complete genotyping of the
entire population.
Linkage analysis
Forty-seven primers used in complete genotyping generated
260 (50 RAPD, 11 SSR, and 199 AFLP) informative
markers. Of these markers, 149 dominant markers showed
1:1 segregation ratio (backcross=BC) and 118 displayed
1:3/3:1 segregation ratio (F2). Six codominant markers

Source of variation

Degrees of freedom

Mean square

F value

Prob

Site
Year
Replication (site)
Siteyear
Error A
Clone
Clonesite
Error B
Total

1
4
4
4
16
41
41
1,148
1,259

477,666,217
159,886,252
12,093,697
48,661,730
4,495,637
2,746,054
2,073,311
138,953

106
35.6
2.69
10.8

0.0000
0.0000
0.069
0.0002

16.3
12.3

0.0000
0.0000

921

Tree Genetics & Genomes (2010) 6:915929

Table 4 Genotypic heritability estimates for yield across sites and
years of recording
Experimental site

Trait

h2se

Kangaita

2003 yield
2006 yield
2007 yield
Annual mean
Annual mean
2003 yield
2006 yield
2007 yield
Annual mean
Annual mean

0.680.03
0.710.03
0.720.03
0.720.03
0.730.03
0.730.03
0.490.02
0.200.01
0.880.03
0.830.03

Timbilil

yield up to 2006
yield up to 2007

yield up to 2006
yield up to 2007

segregated according to 1:1 ratio while another six markers

exhibited distorted segregation in the form of 1:2:1.
Similarly, 52 markers of F2-type population also displayed
segregation distortion. Of 149 dominant markers, 93 were
maternally inherited, while the rest were paternally
inherited. The 155 (149 dominant+6 codominant) markers
that showed 1:1 segregation ratio were used to construct a
linkage map using Map Manager QTX2.0 (2001). The
Kosambi mapping function (1944) that assumes crossover
interference was employed with the threshold value of P=
0.0001 (with LOD score varying from 2.4 to 7.8) to convert
recombination frequency to map distance. Thus, the make
linkage groups function was used, followed automatically
by ripple function to improve the order of the loci in a
linkage group by testing alternative orders created by local
permutations of the locus order. Under these criteria, 100
markers generated 30 linkage groups, while 55 remained
unassigned to any linkage group. The linkage groups so
constructed were ordered sequentially from longest to the

shortest (Fig. 2). However, three markers each in linkage

groups 2 and 4 and one marker in linkage group 22 were
found to be redundant and had to be temporary removed
with hide redundant loci function to allow for QTL
mapping of various traits to proceed. The 30 linkage groups
comprised 19 maternal linkage groups and 11 paternal
linkage groups arising from 69 female-inherited and 31
male-inherited markers, respectively. The maternal linkage
groups spanned 1,012 cM, while the paternal map covered
a total length of 399.5 cM, with mean distance between
markers being 14.7 and 12.9 cM, respectively. The 30
linkage groups formed a consensus map that spanned
1,411.5 cM with mean interval between markers of
14.1 cM. The two parental genomes were treated as a
single entity for QTL detection and interval mapping for
yield as recorded over the period in the two different sites.
Mapping putative quantitative trait loci
Single-marker genome-wide regression analysis
Although yield data recording ensued in 2001, QTL
analysis utilized yield data for 2003, 2006, 2007, and
annual mean yields for 2006 (ANYLD06) and 2007
(ANYLD07) as the most representative data for the two
sites. Yield data collected for the first two years may have
been biased against some clones that were slow to establish
and therefore might not have been suitable for QTL
analysis. Moreover, the Kangaita site was subjected to
maintenance pruning in 2004, while Timbilil got the same
treatment in 2005. QTL analysis was done separately for
each site and period.
Single-marker regression analysis of trait values for
statistical association with genotypes of marker loci in the
progeny detected a total of 23 significant (P<0.01) putative

Table 5 Genetic correlation between yield means for the various years yield records
K03

K06

K07

ANMK06

ANMK07

T03

K03
K06
K07
ANMK06
ANMK07
T03
T06

1
0.88***
0.86***
0.95***
0.95***
0.08 (ns)
0.34*

1
0.96***
0.96***
0.97***
0.02 (ns)
0.42**

1
0.93***
0.95***
0.04 (ns)
0.48**

1
0.99****
0.02 (ns)
0.42**

1
0.01
0.44**

1
0.57***

T07
ANMT06
ANMT07

0.23 (ns)
0.11 (ns)
0.15 (ns)

0.23 (ns)
0.24 (ns)
0.27 (ns)

0.22 (ns)
0.29 (ns)
0.30 (ns)

0.19 (ns)
0.18 (ns)
0.17 (ns)

0.42**
0.89***
0.89***

0.27
0.77***
0.76***

0.24 (ns)
0.20 (ns)
0.23 (ns)

T06

T07

1
0.35*
0.50**

ANMT06

1
0.99***

ANMT07

Key: K03, K06, K07, ANMK06, and ANMK07 are yield records for Kangaita site in 2003, 2006, 2007, annual mean yield up to 2006, annual
mean yield up to 2007, respectively. The same applies for the Timbilil site denoted by T
*P<0.05 (significance of genetic correlations); **P<0.01 (significance of genetic correlations), ***P<0.001 (significance of genetic correlations)

922

Tree Genetics & Genomes (2010) 6:915929

Fig. 2 Linkage map of tea:

paternal linkage groups are
parenthesized with M (male).
Scale of linkage groups is indicated on the left in centimorgans
(cM). LOD thresholds varying
between 2.4 and 7.8 with an
average of 3.2 were used to
construct the map

QTLs controlling various yield traits across the two sites

for the 5years used in the analysis (Tables 6, 7, 8, 9, and
10). Sixteen of the markers were associated with Timbilil
yield traits, while seven were detected in Kangaita site.
Besides, 15 QTLs were inherited from the female parent,
while eight descended from the male parent. Further, five

loci in each of the two sites were not assigned to any

linkage group. Surprisingly, none of the markers was
mutually detected in the two sites for the entire period
under consideration. Another interesting observation is that
at least two loci (EAGC/MCAC02) and OPO-9 were
consistently identified having significant associations (P<

923

Tree Genetics & Genomes (2010) 6:915929

Table 6 QTL analysis for yield recorded in 2003 (YLD-2003)
Site

Linkage group

Locus

LRS

Percent

QTL pos. (cM)

Add

Dom. allele source

Timbilil

Unlinked
Unlinked
Unlinked
Unlinked
Unlinked
Unlinked
Unlinked
Unlinked

OPW-1
EAGC/MCAG01
OPT-2
EACC/MCAG13
EAGC/MCAC02
EACC/MCAC16
EAGC/MCAG13
OPO-9

8.2
7.1
8
7.4
9.7
7
7.4
9.8

18**
16**
17**
16**
21***
15**
16**
21***

236.14
165.19
175.27
168.93
446.86
476.36
483.71
418.81

P
P
M
M
P
M
M
P

Kangaita

Typically, the sign of additivity (Add) for each QTL implies which QTL had increasing or decreasing effect on the trait depending on the parent
contributing the dominant allele; QTL localization was done by interval mapping; The LRS indicates significance of potential association; Percent
is the proportion of the total variance attributable to a QTL at a particular locus; P and M designate paternal and maternal alleles, respectively
*P<0.01; **P<0.001; ***P<0.0001

0.001) with yield from 2003 to 2007 at Kangaita site

(Tables 6, 7, 8, 9, and 10). The amount of variance
explained by each locus also remained more or less the
same throughout the experimental period for the two loci.
The two loci, however, had opposing main effects but of
dissimilar magnitude, although the dominant alleles were
paternally inherited. Multiple regression of the two
unlinked loci with ANYLD07 showed that QTLs linked
to them explained 38% (P<0.05) of the total phenotypic
variance.
The pattern was, however, different for Timbilil site.
Significant associations (P < 0.001) were revealed for
different loci in 2003, 2006, and 2007. However, when
annual mean yields were taken into account for 2006 and
2007, two unlinked markers (OPW-1 and OPT-2) were
detected for both 2003 and 2006 but not for 2007 (see
Table 9). Still, four of the ten loci identified in 2006
(ANYLD06) were also significantly associated with
ANYLD07 in 2007. At Timbilil, marker OPW-1 was highly
significantly associated with yield in 2003 (YLD-2003) as
well as overall annual yield means up to 2006 (ANYLD06),
accounting for 18% (P<0.01) and 29% (P<0.0001) of the
phenotypic variance, respectively, yet it did not fall in any

of the linkage groups (Tables 6, 7, 8, and 9). It could also

be noted that most of the unlinked markers had more effect
on the trait than the linked ones.
Seven QTLs associated with long-term yield
(ANYLD06) whose explained influence ranged from 15%
to 23% of phenotypic variance were detected in different
linkage groups at Timbilil. A cluster of four of these QTLs
mapped on linkage group 1 (Table 9). Interestingly, when
marker OP26-1 with highest LRS was used a cofactor in
composite interval mapping, one of the new unlinked
markers (EAGC/MCAC02) that emerged with significant
association with ANYLD06 (14%: P<0.01) had also been
detected at Kangaita, accounting even for higher amount of
phenotypic variance (25%: P<0.0001) than at Timbilil
(Table 9). The locus though differing in level of magnitude
in affecting the trait in the two sites was consistent in its
increasing effect, in spite of site differences. Multiple
regression involving ANYLD06 and the three loci OPG-2,
OPO-2, and OP26-7 mapped on linkage group 1 that were
significantly associated with ANYLD06 and ANYLD07 at
Timbilil resulted in significant regression coefficient (P<
0.05) with R2 of 44%. Since the three loci have positive
main effects (Tables 9 and 10), their combined effect may

Table 7 QTL analysis for yield recorded in 2006 (YLD-2006)

Site

Linkage group

Locus

LRS

Percent

QTL pos. (cM)

Add

Dom. allele source

Timbilil

LG 12
LG 16
Unlinked
Unlinked

OPO-4
OPO-7
EAGC/MCAC02
OPO-9

10.4
7.6
16.2
10.6

22***
17**
32****
23***

2
1

195.44
172.2
492.69
386.6

M
P
P
P

Kangaita

Column titles are as described in Table 6. **** denotes <0.00001

Tree Genetics & Genomes (2010) 6:915929

924
Table 8 QTL analysis for yield recorded in 2007 (YLD-2007)
Site

Linkage group

Locus

LRS

Percent

QTL pos. (cM)

Add

Dom. allele source

Timbilil

LG 1
LG 11
LG 8
Unlinked
Unlinked

EAGC/MCAA03 (400)
EACC/MCAC02 (865)
EAGC/MCAG02 (910)
EAGC/MCAC02 (790)
OPO-9 (OPO-11-400)

8.1
8.7
9.2
17.3
11.9

18**
19**
20***
34****
25***

3
3
5

197.15
204.72
1,616.73
1,654.44
1,333.11

M
M
M
P
P

Kangaita

Column titles are as described in Table 6. **** denotes <0.00001

have been complementary, which might have inflated the

R2. Addition of OPO-7 loci on linkage 16, which also had a
positive effect, in the multiple regression analysis did not
change the significance level of regression coefficient nor
the magnitude of R2.

locus OPG-2 alone. When the locus with highest LRS in

linkage group 1 was placed in the background, epistatic
effects were not detected. Interaction effects for
ANYLD07-T were more or less the same as for
ANYLD06-T, while no epistatic effects were detected for
Kangaita yield data.

Epistatic effects
Significant epistatic effects (interactions) between loci
EAGC/MCAA01 and EAGC/MCAA03 as well as between
EAGC/MCAA01 and OPG-2 were detected for
ANYLD06-T (Table 11). The first two loci were not
detected during single-point regression analysis although
their combined interaction effects were highly significant.
Similarly, the interaction effect between loci EAGC/
MCAA01 and OPG-2 was higher than that detected for

Localization of QTLs by simple and composite interval

mapping
Some variables associated with highly significant QTLs
merited further attention and were therefore placed in
various linkage groups using interval and composite
interval mapping. In general, interval mapping was
employed to localize the QTLs of all the assigned markers
that had significant association with yield recorded in

Table 9 QTL analysis for annual mean yield from 2001 to 2006 (ANYLD06)
Site

Linkage group

Locus

LRS

Timbilil

Unlinked
Unlinked
Unlinked
LG 1
LG 1
LG 1

OPW-1 (OPW-04)
OPT-2 (OPT-18-1300)
OP26-1 (OP-26-08-380)
OPG-2 (OPG-07-2800)
OPO-2 (OPO-02-900)
OPT-1 (OPT-18-2500)

14.4
8.4
7.6
7.4
9.4
7.1

29****
18**
17**
16**
20***
15**

LG 1
LG 3
LG 3
LG 16
Unlinked
Unlinked
Unlinked
LG 2
LG 8

OP26-7 (OP-26-15-1031)
EACT/MCTA08 (70)
EACT/MCTC01 (355)
OPO-7 (OPO-07-300)
EAGC/MCAC02 (790)
EACC/MCAC16 (120)
OPO-9 (OPO-11-400)
EAGC/MCAG05 (725)
EAGC/MCAG02 (910)

10
6.8
7
10.6
12
8.2
12.7
8.4
6.8

21***
15**
15**
23***
25***
18**
26***
18**
15**

Kangaita

Column titles are as described in Table 6. **** denotes <0.00001

QTL Pos. (cM)

Add

Dom. Allele source

2
1
3

215.26
127.14
121.98
125.11
139.01
117.62

P
M
P
M
M
M

3
13
3
9

3
9

137.27
119.94
116.85
143.71
387.77
405.13
740.08
411.44
386.8

M
M
M
P
P
M
P
P
M

925

Tree Genetics & Genomes (2010) 6:915929

Table 10 QTL analysis for annual mean yield from 2001 to 2007 (ANYLD07)
Site

Linkage group

Locus

LRS

Percent

QTL pos. (cM)

Add

Dom. allele source

Timbilil

LG 1
LG 1
LG 1
LG 16
LG 2
LG 8
Unlinked
Unlinked

OPG-2 (OPG-07-2800)
OPO-2 (OPO-02-900)
OP26-7 (OP-26-15-1031)
OPO-7 (OPO-07-300)
EAGC/MCAG05 (725)
EAGC/MCAG02 (910)
EAGC/MCAC02 (790)
EACC/MCAC16 (120)

7.6
9.4
9.2
9.1
8.1
7.4
13.1
7.7

17**
20***
20***
20***
18**
16**
27****
17**

2
1
5
9
4
9

164.36
181.05
172
175.02
1,240.26
1,229.37
1,237.72
1,207.31

M
M
M
P
P
M
P
M

Unlinked

OPO-9 (OPO-11-400)

12.6

27****

1,143.59

Kangaita

Column titles are as described in Table 6. **** denotes <0.00001

different years. Thus, in Timbilil, the nearest QTL to a

marker was positioned at 1 cM for loci OPO-7 (Table 7)
and OPO-2 (Tables 7, 9, and 10) in linkage groups 1 and
16, respectively, while the furthest was placed at 13 cM
from locus EACT/MCTA08 in linkage group 3 (Table 9).
At Kangaita, a QTL was localized 3 cM from locus EAGC/
MCAG05 in linkage group 2 (Table 9) and 9 cM from locus
AGC/MCAG02 in linkage group 8 (Tables 9 and 10). It is
worth noting that the analysis revealed that the trait was
controlled by more than one QTLs, which could have been
influenced by the prevailing environmental factors, clonal
maturity, and differing site characteristics (see Table 1)
typical of the sites that straddle the Great Rift Valley.
Some QTLs, however, had small increasing or decreasing
effects, while others had major negative or positive effects
on the trait. For example, while a paternal unlinked locus
EAGC/MCAG01 explaining 16% of phenotypic variance
had a net decreasing effect (Table 6) in Timbilil, a major
increasing QTL accounting for 34% of phenotypic
variance was identified in 2007 yield records at Kangaita
(Table 8).
By and large, yield appears to be under the influence of
multiple putative QTLs. For example, at Timbilil, two
moderate QTLs were detected and localized by composite
interval mapping at 2.0 cM away from markers OPG-072800 and OPO-07-900, respectively, in linkage group 1
(Fig. 3). The two loci respectively explained 16% and 15%

of the phenotypic variance. With 18.1-cM interval between

the two loci, it is possible that the two detected QTLs could
be overlapping. Multilocus model by composite interval
mapping indicated that the joint action of mapped QTLs
could only account for 21% of phenotypic variance.
Additionally, a major QTL influencing ANYLD06-T which
accounted for 23% of phenotypic variance was detected and
positioned 9 cM from marker OPO-07-300 (Fig. 4). Dearth
of closely linked markers in chromosome 16 might have
affected precise localization of the QTL. On the other hand,
in Kangaita site, one QTL was detected and localized at
2.7 cM from marker EAGC/MCAG-725 in linkage group 2
(Fig. 5). This QTL accounted for 18% of phenotypic
variance.

Discussion
In studies of quantitative traits, the observed variation is
attributed to the segregation of several to many naturally
occurring genes, for each of which the effects of the allelic
differences on the phenotype are generally small compared
with the effects of the environment (Kearsey and Pooni
1996). Yield is one the traits that have been found to be
typically under the influence of many genes (QTLs) with
the environment largely masking the effects of individualgene action (Crouzillat et al. 2000; Faleiro et al. 2006).

Table 11 Interaction effects for yield (ANYLD06-T) detected at P=1.0e5

Trait

1st LG

Locus1

2nd LG

Locus2

LRS

IX

Main1

Main2

ANYLD06-T

LG 1
LG 1

EAGC/MCAA01
EAGC/MCAA01

LG 1
LG 1

EAGC/MCAA03
OPG-2

26.7
27.9

9.1
10.4

2.2
2.2

3.7
9.4

LRS LRS for association, IX interaction LRS, Main1 LRS for locus 1 main effect, Main2 LRS for locus 2 main effect

926

Tree Genetics & Genomes (2010) 6:915929

specific linkage groups, 19 being maternal and 11 paternal

groups, each of which had 69 and 31 markers, respectively.
The maternal map spanned 1,012 cM, while the paternal
one had a total length of 399.5 cM, with mean distance
between markers of 14.7 and 12.9 cM, respectively. The
combined map accumulated a total length of 1,411.5 cM,
with a mean interval between loci of 14.1 cM. Few markers
notwithstanding, the maternal map is comparable to the one
constructed by Hackett et al. (2000), which covered
1,349.7 cM, with an average distance of 11.7 cM between
loci. The map was constructed with 124 markers. In their
work, Hackett et al. (2000), however, did not construct a
paternal map as in the current study. The availability of 30
linkage groups in the current study, however, made it
possible to detect a total of 23 QTLs that significantly
associated with the yield for five yield records across the
two trial sites. With 30 linkage groups comprising 19
maternal and 11 paternal groups, it was possible to detect
QTL significantly associated with either the female or male

Fig. 3 Linkage group 1 showing position of the multiple YLD-T

QTLs at 2 cM each from markers OPG-07-2800 (upper arrow) and
OPO-07-900 (lower arrow). The LRS line (L1) was obtained by
composite interval mapping YLD-T at regular intervals between
marker loci. Bars represent the estimated confidence interval by
bootstrap resampling in Map Manager QTX. Fine vertical lines from
left to right represent suggestive (2.8), significant (8.4), and highly
significant (15.0) threshold LRS values obtained by 1,000-permutation test at P=0.01

When considering perennial crops such as tea, QTL

analysis is likely to be confounded not only by the highly
varying weather pattern but also the heterogeneous edaphic
and cultural practices leading to significant and unpredictable GE interactions (Bradshaw 1998). In the current
study, tea yield, the most important attribute in tea given
that farmers are paid on the basis of amount of leaf they
deliver to factory, was studied for 7 years in two main
experimental sites of the tea-growing region. Consequently,
breeders interest in this character as well as other attributes
such as tea quality that in combination would enhance
Kenyan tea competitiveness in world market has recently
been markedly intensified. Thus, the present study set out
to identify molecular markers that could be used in markerassisted selection, thereby shortening the development and
release for commercialization of elite varieties that generally takes at least 20 years (Gazi 1978) to develop under
conventional breeding and clonal selection.
Although a total of 260 markers were generated in the
current study, only 100 markers could be assigned to

Fig. 4 Postulated linkage map of chromosome 16. Arrow shows

position of the YLD-T QTL at 9.0 cM from marker OPO-07-300. The
LRS line (L1) was obtained by composite interval mapping YLD-T at
regular intervals between marker loci. Bars represent the estimated
confidence interval by bootstrap resampling in Map Manager QTX.
Fine vertical lines from left to right represent suggestive (5.8),
significant (11.2), and highly significant (18.4) threshold LRS values
obtained by 1,000-permutation test at P=0.01. Line (L2) is additive
regression coefficient

Tree Genetics & Genomes (2010) 6:915929

Fig. 5 Postulated linkage map of chromosome 2. Arrow shows

position of the YLD-K QTL at <3.0 cM from marker EAGC/MCAG725. The LRS line (L1) was obtained by interval mapping YLD-K at
regular intervals between marker loci. Bars represent the estimated
confidence interval by bootstrap resampling in Map Manager QTX.
Fine vertical lines from left to right represent suggestive (5.9),
significant (11.9), and highly significant (19.0) threshold LRS values
obtained by 1,000-permutation test at P=0.01. Line (L2) is additive
regression coefficient

loci, thereby increasing the power of QTL detection and

dissection.
The current study used both BSA and complete
genotyping experiments to identify QTLs associated with
yield using pseudo-testcross strategy in combination with
RAPD, AFLP, and SSR marker. As pointed out by
Grattapaglia and Sederoff (1994), the term pseudotestcross which is used as the testcross mating configuration
is unknown a priori as in a conventional testcross where the
tester is homozygous recessive for the locus of interest.
Rather, the configuration is inferred a posteriori after
analyzing the parental origin and genetic segregation of
the marker in the progeny of a cross between highly
heterozygous parents with no prior genetic information.
Based on this approach, the two experiments were able to
detect QTLs for the two traits although at different levels of
significance. The lowest probability level for the search of
QTL was P=0.01, which translates into 1% false positives.
This less stringent criterion was applied where no QTL
could be detected at lower P value.
Although the availability of a saturated linkage map
would have allowed precise localization of putative QTLs
for the yield variables, the complexity arising from mixed
mating types in what is considered an F1 population might
have negated the attainment of this ideal objective. While

927

the 118 markers that showed 1:3 or 3:1 segregation ratio,

characteristic of an F2 population, were discarded, the small
size of our mapping population comprising 42 clonal
progeny should not preclude QTL detection (Ortiz 1996).
Moreover, the current study, being the first of its kind,
opens opportunities of exploring further QTL analysis of
tea using populations arising from either clonal F2s or
doubled haploids (Liu 1998). The 23 putative yield QTLs
as assessed across space and time provide a promising
starting point. Our small population would have been very
conducive to overestimated and spurious OTLs, as has also
been pointed out by different workers (Beavis 1998; Sewell
et al. 2000; Utz et al. 2000; Lerceteau et al. 2001; Lee
2005), so our findings are very tentative, especially with
regard to marker-assisted selection. The high heritability
estimates for the trait, irrespective of its measurement in time
and space, may nonetheless mitigate the problem (Lee 2005).
Even so, using more recent software, coupled with Bayesian
analysis, may substantially shorten the list of detected QTLs.
Moreover, availability of a large mapping population that is
genotyped with highly informative markers such as identified in the present case would result in dense linkage map
that would not only lead to identification of tightly linked
QTLs but also would be of great importance in works aimed
at map-based cloning of genes.
The magnitude of genotype by environment interaction
is corroborated by phenotypic variation as displayed in
histograms, ANOVA of the mapping population across
sites, and lack of any congruent locus affecting the traits in
any of the years under consideration and across environments. This is further exhibited by highly variable
heritability estimates within Timbilil (ranging from 0.2 to
0.88), but fairly stable and high estimates from year to year
at Kangaita. Yield recording was done more or less
simultaneously in the two sites, yet none of the yield traits
seems to have been influenced by the same QTL. Such a
high heritability particularly for yield has been documented
in a previous study (Kamunya and Wachira 2004). The
complex environmental factors prevailing in the two sites
might have predominant effects on QTLs governing the
traits. This implies that multiple QTLs with small to
moderate effects exist for each yield trait and it only
requires significant environmental shift to trigger a QTL to
act or to shut down. Kearsey and Pooni (1996), while
commenting on nature and causes of GE, have demonstrated that the genetical variance among a collection of
genotypes may alter with the environment, meaning that the
effects of given allele substitutions may be quite different in
one environment than in another. This is also illustrated by
the low estimated genetic correlations between sites for
yield (Table 4). However, the few loci that could be
consistently associated with yield over the period within a
site, especially at Kangaita, will need to be validated for

Tree Genetics & Genomes (2010) 6:915929

928

any marker-assisted selections. For example, markers

EAGC/MCAC-790 and OPO-11-400 that consistently
associated with yield at Kangaita for the entire period
under investigation indicated their potential as candidate
markers for site-specific marker-assisted selection. This
seems to be in line with Kearsey and Pooni (1996), where
they have emphasized that for optimal performance of a
strain or variety in a particular environment, selection
should be carried out in that environment. Nevertheless, it
is possible that with a bigger population, such useful but
unassigned markers would be placed in some linkage
groups making localization of QTLs associated with them
more precise and meaningful. Various workers have come
up with recommendations on ways of enhancing phenotypic evaluations (Collard et al. 2005). Among those
recommended include use of a large population size,
recombinant inbred lines, double haploids, or clones in
replicated experiments across sites and over time (George et
al. 2003; Hackett 2002; Collard et al. 2005). Besides
cloning the progeny, to mitigate the shortcoming of
population size, establishing them in replicated trials within
and between sites including long-term observation is
another factor that had been carefully considered in the
current study.
The indication of significant epistatic effects between
loci EAGC/MCAA01 and EAGC/MCAA03 as well as
between EAGC/MCAA01 and OPG-2 that were detected
for ANYLD06-T suggests that yield is a product not only of
individual sparsely localized QTLs but of interactions from
other unlinked or linked QTLs. The first two loci were not
detected during single-point genome-wide regression analysis although their combined interaction effects were highly
significant. Similarly, the interaction effect between loci
EAGC/MCAA01 and OPG-2 was higher than that detected
for locus OPG-2 alone. The interaction effects for yield
seems also to be influenced by GE interactions as it was
only detected for ANYLD07-T at more or less the same
level as for ANYLD06-T, while no such effects were
detected for Kangaita yield data at any of the period under
consideration.
Considering the long time span involved in developing
elite varieties (20 years) with majority of desirable
attributes, adoption of molecular markers linked to important traits can be of marked help in this venture. Where
putative QTLs have been tightly linked to markers, there is
need to verify and confirm their effects using independent
populations constructed from the same parental genotypes
or closely related genotypes used in the primary QTL
mapping studies. Integration of molecular markers in
breeding and clonal selection would not only reduce the
number of clones/seedlings for field testing but also lead to
increased genetic gain per unit of time by decreasing
evaluation periods.

Acknowledgements We would like to thank the Institute of

Himalayan Bioresource Technology, IHBT (CSIR), India for providing support and necessary facilities for conducting the molecular
work. The first author is thankful to the Council of Scientific and
Industrial Research (CSIR), India and the Academy of Sciences for
the Developing World (TWAS), Italy, for providing financial
assistance in form of Postgraduate CSIR-TWAS Fellowship. The Tea
Research Foundation of Kenya also supported the study.

References
Anon (2002) Tea growers handbook, 5th edn. Tea Research
Foundation of Kenya, Kericho
Beavis WD (1998) QTL analyses: power, precision and accuracy. In:
Paterson AH (ed) Molecular dissection of complex traits. CRC,
New York, pp 145162
Bradshaw HD Jr (1998) Case history in genetics of long-lived plants:
molecular approaches to domestication of a fast-growing forest
tree: Populus. In: Paterson AH (ed) Molecular dissection of
complex traits. CRC, New York, pp 219228
Brown GR, Bassoni DL, Gill GP, Fontana JR, Wheeler NC, Megraw
RA, Davis MF, Sewell MM, Tuskan GA, Neale DB (2003)
Identification of quantitative trait loci influencing wood property
traits in loblolly pines (Pinus taeda L.). III. QTL verification and
candidate gene mapping. Genetics 164:15371546
Churchill GA, Doerge RW (1994) Empirical threshold values for
quantitative trait mapping. Genetics 138:963971
Collard BCY, Jahufer MZZ, Brouwer JB, Pang ECK (2005) An
introduction to markers, quantitative trait loci (QTL) mapping
and marker-assisted selection for crop improvement: the concepts. Euphytica 142:169196
Crouzillat D, Menard B, Mora A, Phillips W, Petiard V (2000)
Quantitative trait loci analysis in Theobroma cacao using
molecular markers: yield QTL detection and stability over
15 years. Euphytica 114:1323
Faleiro FG, Queiroz VT, Lopes UV, Guimaraes CT, Pires JL, Yamada
MM, Araujo IS, Pereira MG, Schnell R, Filho GAS, Ferreira CF,
Barros EG, Moreira MA (2006) Mapping QTLs for witches
broom (Crinipellis perniciosa) resistance in cacao (Theobroma
cacao L.). Euphytica 149:227235
Freeman S, West J, James C, Mayes S (2004) Isolation and
characterization of highly polymorphic microsatellites in tea
(Camellia sinensis). Mol Ecol Notes 4:324326
Gawel NJ, Jarret RL (1991) A modified CTAB DNA extraction
procedure for Musa and Ipomoea. Plant Mol Biol Rep 9:262266
Gazi MS (1978) Distributional pattern of yield and vacancy of tea in
Bangladesh. Tea J 14(2):1922
George MLC, Prasanna BM, Rathore RS, Setty TAS, Kasim F, Azrai
M, Vasal S, Balla O, Hautea D, Canama A, Regalado E, Vargas
M, Khairallah M, Jeffers D, Hoisington D (2003) Identification
of QTLs conferring resistance to downy mildews of maize in
Asia. Theor Appl Genet 107:544551
Grattapaglia D, Sederoff RR (1994) Genetic linkage maps of
Eucalyptus grandis and Eucalyptus urophylla using pseudotestcross: mapping strategy and RAPD markers. Genetics
137:11211137
Hackett C (2002) Statistical methods for QTL mapping in cereals.
Plant Mol Biol 80:524535
Hackett CA, Wachira FN, Paul S, Powell W, Waugh R (2000)
Construction of a genetic linkage map for Camellia sinensis (tea).
Heredity 85:346355
Kamunya SM, Wachira FN (2004) Preliminary observations on
inheritance of yield and some physiological traits in tea. Tea 25
(2):2431

Tree Genetics & Genomes (2010) 6:915929

Kamunya SM, Wachira FN, Chalo RM (2004) Evaluation of newly
developed clones of tea (Camellia sinensis) for yields, drought
tolerance and quality: Preliminary indications. Tea 25(1):1219
Kearsey MJ, Pooni HS (1996) The genetical analysis of quantitative
traits. Stanley Thorne, Cheltenham, p 381
Kosambi DD (1944) The estimation of map distances from recombination values. Ann Eugen 12:172175
Lee C (2005) Selection bias in quantitative trait loci mapping. J Hered
96(4):363367
Lerceteau E, Szmidt AE, Anderson B (2001) Detection of quantitative trait
loci in Prunus sylvestris L. across years. Euphytica 121:117122
Liu BH (1998) Computational tools for study of complex traits. In:
Paterson AH (ed) Molecular detection of complex traits. CRC,
Boca Raton, pp 4379
Manly KF, Cudmore RH Jr, Meer JM (2001) Map Manager QTX,
Cross-platform software for genetic mapping. Mamm Genome
12:930932
Michelmore RW, Paran I, Kesseli RV (1991) Identification of markers
linked to disease-resistance genes by bulked segregant analysis: a
rapid method to detect markers in specific genomic regions by
using segregating populations. Proc Natl Acad Sci U S A
88:98289832
Missiaggia AA, Piacezzi AL, Grattapaglia D (2005) Genetic mapping
of Eef1, a major effect QTL for early flowering in Eucalyptus
grandis. Tree Genet Gen 1:7984
Ortiz R (1996) Statistical basis of marker identification. In: Crouch
JH, Tenkouano A (eds) DNA marker-assisted improvement of
the staple crops of Sub-Saharan Africa. Proceedings of the
Workshop on DNA markers at IITA held by the Crop
Improvement Division, IITA, Ibadan, Nigeria, 2122 August
1996, pp 2734
Sewell MM, Bassoni DL, Megraw RA, Wheeler NC, Neale DB
(2000) Identification of QTLs influencing wood property traits in

929
loblolly pine (Pinus taeda L.). I. Physical wood properties. Theor
Appl Genet 101:12731281
Utz HF, Melchinger AE, Shon CC (2000) Bias and sampling error of
the estimated proportion of genotypic variance explained by
quantitative trait loci determined from experimental data in maize
using cross validation and validation with independent samples.
Genetics 154:18391849
Wachira FN (2002) Genetic diversity and characterisation of Kenyan
tea germplasm. A tea Genetic Diversity (TGD) project. TGD
final project document, Kericho, Kenya
Williams CG (1998) QTL mapping in outbred pedigrees. In: Paterson
AH (ed) Molecular dissection of complex traits. CRC, New York,
pp 8194
Williams JGK, Kubelik AR, Livak KJ, Rafalski JA, Tingey SV (1990)
DNA polymorphisms amplified by arbitrary primers are useful as
genetic markers. Nucleic Acids Res 18:62316235
Wright JW (1976) Introduction to forest genetics, New York
Academic Press, Inc.
Yang HY, Korban SS, Kruger J, Schmidt H (1997) The use of
modified bulk segregant analysis to identify a molecular marker
linked to a scab resistance gene in apple. Euphytica 94:175182
Zabeau M, Vos P (1993) Selective restriction fragment amplification:
a general method for DNA fingerprinting. European Patent
Application. 92402629 (Publ Number 0 534858 A1)
Zeng ZB (1993) Theoretical basis for separation of multiple linked
gene effects in mapping quantitative trait loci. Proc Nat Acad Sci
U S A 90:1097210976
Zeng ZB (1994) Precision mapping of quantitative trait loci. Genetics
136:14571468
Zhang WK, Wang YJ, Luo GZ, Zhang JS, He CY, Wu XL, Gai JY,
Chen SY (2004) QTL mapping of ten agronomic traits on the
soybean (Glycine max L. Merr.) genetic map and their association
with EST markers. Theor Appl Genet 108:11311139