Maize Genetics Cooperation Newsletter
vol 87 2013
RESPONSES TO AGGREGATE
TRAIT SELECTION FOR CHILO PARTELLUS
(SWINHOE) RESISTANCE IN MAIZE (ZEA MAYS L.) POPULATION
MUTINDA,
C.J.M; S.O. AJALA AND P.O. AYIECHO
Selection
indices are efficient ways of simultaneously improving a number of quantatively inherited
traits in maize (Zea Mays L.). Different
selection indices have been improved (Smith, *1936; Hazel,1943;
Williams,1962) while Elston(1963) further improved on
these by proposing weight free indices. More recently, Mulamba
and Mock (1978) developed a parameter free index, the rank summation indices (RSI)
to improve density tolerance in maize (Zea
Mays L.). However, further
studies comparing relative efficiencies of different indices suggest that
simpler indices, that are parameter and weight free, are favoured
(Subandi et al 1973; crosbie et al 1980).
Relative
efficiency of a breeding procedure is dependent on the rate of improving and
ease of handling. Predicting responses to selection helps in comparing
different methods. Studies predicting progress from a single trait selection
are common in literature but those predicting responses to index selection are
very few. Crosbie et al (1980) observed that linear
index proposed by Baker (1974), the Elston (1963)
weight free index(EWF) and rank summation(RSI) of Mulamba and Mock(1978), combined simplicity of use, freedom
from use of estimate genetic parameters, good selection differential and
predicted gains in each trait and in the aggregate genotype. Pesek and Baker (1969a) proposed the formula for predicted
gain for each trait due to index selection while Mock and Eberhart
(1972) further suggested a formula for calculating predicted gain in the
aggregate trait and concluded that index selection for cold tolerance was about
as efficient as single trait selection. Opeke (1983)
noted that responses to index selection for seedling vigour
though inferior to single trait selection would improve grain yield. Subandi et al (1973) observed that results from earlier
studies on selection indices could be summarized as follows: in general,the index was superior to
other selection procedures in both predicted and actual genetic advances ,an
index aimed at improving a trait
gives greater gains than selection based only on that trait and for an index to be effective, the genetic
correlations between the trait included in the index and the traits to be
changed must be high.
ICZ3(IC-90-W1),
a white grained early to medium population was used in this study. A total of
144 S2 lines and 140 S2 test cross progenies derived from
ICZ3 were evaluated for damage parameters (leaf feeding, dead heart and stem
tunneling) caused by Chilo partellus and agronomic traits including gain yield in
two locations. The experimental locations were (MPFS) and Ungoye
which are ICIPE testing sites. Mbita Point field
Station has bimodal rainfall distribution with two distinct peaks. The early
season (long rains) starts from late march and ends in late September or early
October, the late season (short rains) starts from late September or early
October to December.
It
is situated on the shores of lake Victoria in Western Kenya (latitude 00
25�-00 30� South, Longitude 340 15 East and altitude
1240M. Ungoye is 35 KM from Mbita
Point field Station with similar rainfall distribution pattern and also
situated along the lake region.
In
each site, the 144 S2 lines and 140 S2 test cross
progenies were planted in two replicate experiments. The genotype
were grown in a randomized complete block design with single row
plots. Each row was 5.0m long but
separated into two 2.25m halves with a space of 0.5m in the middle. Spacing was
0.75m between rows and 0.25m between hills. Each hill was planted with two
plants but later thinned to one three weeks after germination to give a maximum
of
10 plants/2.25m row and a density of approximately 53,333 plants/ha. All plants
in one half of the row were artificially infested with 30 first instar C.partellus larvae reared on artificial diet (Ochieng etal1985) three weeks after emergence. Appropriate
culture practices, such as fertilizer application, weeding, bird or monkey scaring were carried out as
deemed necessary during the season.
Data
on foliar lesions and dead heart were taken at four weeks after infestation. Foliar lesions was score on a 1-9 scale (1=resistant and
9=susceptible) while dead heart was assessed as the proportion of plants in a
plot showing the symptom. Extent of stem tunneling by the larvae was estimated
at harvest as the percentage of the plant height. Other agronomic data recorded
were plant height, stand at harvest, number of ears harvested, mean length of
five ears per plot, moisture content at harvest and grain yield. Grain yield
was obtained as grain weight adjusted to 13% moisture content. Yield reduction
was calculated as the difference between the yield of the un
infested control and the infested.
Dead
heart and stem tunneling data for each location were transformed into arc-sine
values before subjecting to analysis of variance (ANOVA). On this transformed
scale, error variances were highly homogenous according to Barlett�s
test (Barlett, 1939). Combined ANOVA was therefore
carried out. Two Rank Summation Indices (RSIs) were constructed to determine
the ranking of each line within the population for suitable response. The first
index(RSI-1) was
obtained by ranking the means of each leaf feeding (LF), dead heart(DH)
and stem tunneling(ST) for each line, summing the ranking of the line to obtain
its aggregate performance compared with other lines within the same population.
A second (RSI-2) was obtained using the three traits and grain yield. Rank
Summation Index (Mulamba and Mock, 1978) was
summarized as follows;
RSI=ΣRi�s
Where
Ri is the rank of the mean of each of the desired
traits.
RS1-1=Aggregate performance of a genotype
using the ranking of leaf feeding, dead heart and stem tunneling.
RSI-2=
Aggregate performance of a genotype based on ranked means of leaf feeding, dead
heart, stem tunneling and grain yield. Thus the lowest possible values for the
two indices would be three and four respectively, characterizing a line in a
line particular progeny type that ranked first for all traits. An entry with
the least damage for foliar feeding, dead heart and stem tunneling and highest
grain yield will rank first for the four traits.
Expectations
of mean squares (EMS) from analysis of variance were used to estimate genotypic
(σ2g), genotype x environment (σ2ge)
interaction, error (σ2) and phenotypic (σ2ph)
components of variance, while expectations of mean cross products (EMCP) from
analysis of covariance were used to estimate genotype correlations. Standard
errors (S.E) for each of the variances (σ2i) except phenotypic
variance were calculated as (Hallauer, 1971):
S.E. σ2i= [2/C 2 {msi 2/ (dfi=2}]
½
While
that for phenotypic variance was computed as
S.E
σ2 ph= [(1/re2){msg/(dfi+2)}] ½
where
msi, dfi and C2 are
mean squares, degree of freedom and coefficient of the component in the EMS for
trait I respectively, and msg is the mean square for
genotype, r=number of replicates and e=number of environments or locations.
Habitability (σh2) estimates were calculated as proportions of
total variance due to genetic causes with S.E also calculated as proportions of
S.E of σ2g to σ2 ph. Entry means across
locations and replicates were used to calculate simple correlations and step
wise multiple regressions. Predicted responses (ΔG) for single trait
selection were calculated as:
(ΔG)=k.
ph. h2
Where
k (k=1.76 for selection intensity of 10%) is the standard selection
differential, ph is the phenotypic standardized
deviation and h2 represents heritability for the trait under
consideration.
RSI
values were subjected to both analysis of variance and covariance and the
information obtained from EMS and EMCP were used to estimate variance components
and heritability. Predicted response to selection for RSI was then calculated
using the above formula. This was
then compared with the formula of Mock and Eberhart (1972)
for calculating gains from aggregate selection as follows:
ΔH=aiΔgi
Where
is the economic weight for the ith and Δgi, which was calculated, using the formula of Pesek and Baker (1969a), is the predicted response for
trait due to index selection.
Economic
weights were –1,-1,-1 and 1 for foliar, dead
heart, stem tunneling damages and grain yield, respectively. Coefficient (b
values) used in the estimation were obtained by solving the equation bi= (Xij)-1(gij)(ai) where Xij and gij are variance covariance matrices of phenotypic and
genotypic values respectively for the four traits in each of the progeny types.
Correlated
responses due to single and aggregate trait selection created by RSI were
calculated as:
CRy(x) = ix.hx.hy.rgx.y
phy (Falconer, 1960)
Where
ix= selection intensity applied to trait x, .hx
and .hy are square roots
of heritability estimates for traits x and y, respectively, rgx.y
is the genetic correlation between the two traits, and phy is the square root of phenotypic variance
for trait y.
Estimates
of perimeter components of variance obtained for each of the two progeny types
presented in Table 1. for most traits, genetic(σ2e)
and environmental(σ2e), and phenotypic (σ2ph)
variances exceeded twice their standard errors. Generally, the genotypic
variances for most traits were large enough for selection purposes. Except for
a few cases the estimates of genotype by environment variances (σ2ge)
were either negative or smaller than their respective standard errors (se). Also,
most of the genetic variances were larger for S2 progeny types than
for the test cross hybrids corresponding to high heritability estimates in the
former than the later. Heritability estimates for parameters of resistance,
grain yield and selection indices in most cases were moderate for the S2 families
thus suggesting that simultaneous improvement of these traits in the desired
direction should be possible, and especially so with the use of selection
indices to effectively combine the traits. However, for the test cross hybrids,
the estimates were low for the majority of the traits.
� TABLE 1:
Correlations
of parameters of resistance to C. partellus (leaf
feeding and stem tunneling) with mature plant characteristics, including grain
yield (Table 2), were generally negative. Dead heart showed highly
significant correlations with stand count, ear length, ear number and moisture
% at harvest for the two progeny types. Rank summation index(RSI-1)
involving the three parameters of resistance namely, leaf feeding, dead heart
and stem tunneling showed highly significant (P<0.01) correlations with the
four agronomic traits as opposed to those involving RS-2, which were generally negative
apart from a few cases.
TABLE 2: Simple linear correlations of Chilo partellus resistance parameters including rank summation index (RSI) on mature plant traits and gain yield from test cross hybrids and S2 progenies combined for Mbita Point Field
Station (MPFS) Ungoye locations of western Kenya
|
|
Progeny type |
Leaf feeding |
Dead heart |
Stem tunneling |
RSI-1 |
RSI-2 |
|
Plant height(cm) |
(i) (ii) |
0.11 0.05 |
-0.07 -0.03 |
0.29** 0.22** |
-0.13 -0.13 |
0.00 -0.02 |
|
Stand count |
(i) (ii) |
-0.09 -0.08 |
0.19* 0.30** |
-0.11 -0.08 |
0.31** 0.48** |
-0.04 0.08 |
|
Ear length |
(i) (ii) |
-0.07 -0.14 |
0.40** 0.32** |
-0.09 -0.15 |
0.95** 0.99** |
-0.13 0.21** |
|
Ear number |
(i) (ii) |
0.01 -0.01 |
0.26** 0.25** |
0.09 -0.03 |
0.34** 0.45** |
-0.14 0.12 |
|
Moisture (%) |
(i) (ii) |
-0.09 0.04 |
0.27** 0.24** |
-0.14 -0.14 |
0.49** 0.61** |
-0.14 -0.02 |
|
Grain yield(t/ha) |
(i) (ii) |
-0.03 -0.01 |
-0.04 -0.03 |
-0.05 -0.04 |
-0.01 -0.02 |
-0.08 0.04 |
*,** significant at P<0.05 and 0.01, respectively. (i)=testcrosses (ii)= progenies
The possible contribution of each of the damage parameters to grain yield reduction was examined using step-wise multiple regressions. Results obtained (Table 3) indicated that in the testcrosses, stem tunneling accounted for at least 45% of the total variation in grain yield reduction (R2=0.45). In the two progeny types, stem tunneling had the greatest contribution towards grain yield reduction (R2 being 0.36 for S2 lines and 0.45 for the testcross)
TABLE 3: Unstandardized partial regression coefficients (b-values), coefficients of determination (R2) and R2 change ΔR2 from step-wise multiple regression of grain yields on parameters of resistance in each of the progeny types.
|
|
trait |
b-value |
R2 |
R2 |
|
Test crosses |
Leaf feeding Dead heart% Stem tunneling |
-0.01 1.40 -0.14 |
-0.01 0.01 0.45 |
0.01 0.00 0.44 |
|
S2 lines |
Leaf feeding Dead heart% Stem tunneling |
0.002 -0.09 0.02 |
0.20 0.27 0.36 |
0.20 0.07 0.09 |
Predicted
direct responses to selection for gain yield, parameters of resistance i.e. leaf
feeding, dead heart and stem tunneling due to index selection were much lower
than when single trait selection was carried out for each of the traits (Table
4). Opeke (1983) noted that relative to single trait
selection, index selection usually gave lower progress for selection because
superiority of a trait is negated by mediocrity in other traits in the index. Response
due to index selection (ΔH) was higher for test cross hybrids than that of
the S2 progenies while Rank Summation Index (RSI), more progress was
achieved in S2
progenies than in the test cross hybrids. In effect, although either of the
methods would result in aggregate improvement , actual
gains in each progeny would depend on the
selection method used. Rank Summation Index (RSI) gave more than double
the progress of the aggregate trait selection in S2 progenies (Table
4).
TABLE
4: Predicted direct response( G/CYCLE) to single trait
selection for parameters of resistance, the aggregate trait
Created By Rank Summation Index (RSI) and to index selection in each of the two progeny types.
|
|
|||||
|
Single trait selection |
|||||
|
Test crosses S2 lines |
Grain yield(t/ha) 0.32 0.40 |
Leaf feeding -0.08 -0.16 |
Dead heart(%) -0.12 -0.05 |
Stem tunneling(%) -0.22 -0.79 |
RSI 4.48 10.75 |
|
Test crosses S2 0.12 -0.09 -0.06 -0.08 4.20 Lines 0.10 -0.08 -0.04 -0.63 3.55 |
|||||
Predicted
correlated responses in grain yield when selection was done for parameters of
resistance and the rank summation indices are presented in
Table 5. When these gains were expressed as percentage of
the means of their respective families in the two progeny types, they were
lower than those expected from direct selection for grain per se, in all
cases, except, for RSI-2 in S2 families.
TABLE 5: Predicted correlated responses (per cycle) in grain yield (t/ha) when selection was done for parameters of
Resistance including summation index (RSI) in ICZ3 population.
Selection criteria test cross hybrids S2 families
Leaf feeding -0.02 -0.05
(-0.84) (-2.04)
Dead heart (%) (-3.45) -0.23
Stem tunneling (%) -0.25 (-9.39)
(-4.32) -0.32
RSI-1 -0.002 (-13.06)
(-0.03) -0.001
RSI-2
-0.64
0.20
(
(-11.50)
(8.16)
()= Correlated responses expressed as % of the overall mean yield of the respective progeny type
Studies
suggesting approaches aimed at reducing limitations associated with selection
index construction have been reported (Williams, 1962; Elston,1963;
Pesek and Baker, 1969b,1970) but problems in
assigning appropriate economic importance (weight) to each trait and those
associated with extensive computation still exist. RSI therefore, has the advantage
of not only giving appreciable progress for aggregate gain but also the ease
with which they can be handled.
Aggregate
trait selection in the progeny types would result in the improvement of other
traits including those not include in the formation of the index e.g grain yield and plant height.
However,
increase in height is an undesirable character commonly associated with yield
in tropical maize germplasm (Miranda Filho, 1985). Excessively tall plants can lead to stalk
lodging especially in windy weather (Ajala, 1990).
The association or correlated response to selection of a trait or other
unselected traits occur either due to linkage response to selection of a trait
or other unselected traits occur either due to linkage or pleiotropy
(Fakorede and Mock,1982).
Correlation
between the resistance parameters and the selection indices with mature plants
traits including grain yield were generally very low, except for a few
agronomic traits, implying that damage levels could not be used as a measure of
expected grain yield for the materials studied. Such findings are in agreement
with that of Ajala et al (1993).
Since
grain yield is of paramount importance to the breeder, possible contribution of
each of the damage parameters examined using step-wise multiple regressions
indicated that in both the test crosses and S2 progenies leaf
feeding seemed to contribute less towards yield reduction than stem tunneling
and dead heart. Mohyuddin and Attique
(1978) and Pathak and Othieno(1990) attributed
yield reduction in maize to be caused more by dead heart. Results obtained in
this study do not seem to occur with the observation of these researchers.
However, Ajala and Saxena(1994) using
correlations, step wise multiple regressions and path coefficient analyses to
study the interrelationship among the three damage parameters (foliar lesions,
dead heart and stem tunneling) and their contribution to grain yield reduction
showed that yield loss caused by Chilo partellus is primarily due to stem tunneling of the
plants.
The
primary objective of the study was to improve maize population (ICZ3) for
resistance to the spotted stem borer, Chilo partellus. Data presented herein showed that use of RSI
is feasible and will improve grain yield. Use of index coefficients requires that
appropriate economic weights be placed on each progeny type and determination
of the economic importance (weight) for each trait is arbitrary. This therefore
strengthens the argument in support of RSI as a better index.
