Transcript Recurrent Selection - Crop and Soil Science

PBG 650 Advanced Plant Breeding
Module 10: Recurrent Selection
Recurrent selection
•
Cyclical selection of populations
– form families
– evaluate in trials
– recombine selections
•
Pedigree selection and improvement of elite lines
are also cyclical processes, but the population
structure is not so clearly defined
•
Selfing and introgression of new germplasm are
common features of both selection systems
•
Recurrent selection and development of lines can be
integrated into a comprehensive system
Bernardo, Chapt. 10
Rationale for recurrent selection
•
Selfing systems:
– Fixation of alleles is so rapid that the impact of selection is limited
– Probability of obtaining segregants with all of the favorable alleles
controlling a quantitative trait is small
Example: with 5 loci, all alleles have p=0.5
1/32 chance to get all of the good alleles
•
Recurrent selection:
– systematically increases the frequency of favorable alleles
Example: with 5 loci, all alleles have p=0.6
1/13 chance to get all of the good alleles
– maintains the genetic variation within a population to permit continual
progress from selection
Recurrent selection in practice
Why is it not used more often? (Bernardo)
•
Easy to apply in cross-pollinated crops; difficult in selfpollinating crops
– male sterility systems can be used
•
Objectives are long-term
– several generations needed to complete a cycle
• Immediate output is an improved open-pollinated variety, not
a line or hybrid
•
Need to choose one or a few populations for selection
– not as much opportunity for speculation in use of germplasm
•
Nonetheless, there are many examples of widescale use of
varieties developed from recurrent selection schemes
Expected selection response
Progeny
Phenotype
•
b
•
•
OP
•
••
•
•
•
•
•
•
• ••
• • • • •••
Expected
•
• • • ••
• •
•
•
•
•
•
•
• ••
•
•
• •
•
•
•
•
selection
•• ••• • •• ••• • •••• •••
• • •
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
• •• • •• • ••• ••• • • • • •••• •• •• • ••
response
•
•
•
•
•
•
•
•
•
•• •
•
• •• ••••• • •••• ••• •• • •• • • •• •••••• • •••• ••• •• • •••• •• ••• • ••• • • •
•
R = h2S
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
• • ••• •••• • •••• •• •••• •••• • •• • •• • ••• ••••• •• • • • ••
•
•
•
•
• • •• •• • ••• • ••• ••• ••• • •••• •••••••••••••• •••••••••• •••• •••••••• •••• •• •
•
• • •• • ••• •• • •• •• ••• •• ••••• •••••• •••••• ••• ••• ••• ••• ••• • • ••
•••
• •• ••• • •• •••••• •••••• ••• •••• •••• •••• •• ••••• • •• •• ••••• ••••••• •• ••• • • ••••
• • • •• •• • • •• ••••• ••••••••• •••••••••• •••• ••• ••• •• ••• ••• • • • ••• ••• •••• •• •
•
•
••
• ••
•
•
• • • •• • •• •• •• • •••• •• •••• ••• • ••• • • ••••• ••• • •• •
••
• • Mean of
•
•
•
•
•
•
•
•
• • ••• • ••••• • •• • ••
•
•
•••• • •••
••
•••
•
•
•
selected
•
•
• •• • • • • • ••• • •••••• • • •
•
•
• • • •
parents
•
••
••
•
•
Selection
differential
S
Selected parents
Truncation point
Mean of all possible parents
(selection candidates)
Source: Lecture by Jean-Luc Jannink at Iowa State, 2004
 h2
Mid-parent
Phenotype
Response to selection
R=h2S
X0
S
Selection differential
S  XS  X0
Response to selection
Xs
60
70
80
90
100
110
Select best
10% of C0
120
130
140
150
120
130
140
150
Recombine to form C1
R  X1  X0
X1
Realized heritability
R 113  100
2
h  
 0.75
S 117  100
R
60
70
Falconer and Mackay, Chapt. 11
80
90
100
110
Predicting response to selection
R=h2S
• Need estimates of h2 and the selection
differential
• In theory, h2 is only applicable for a single
generation, because heritability depends on
gene frequencies. In practice, predictions
seem to work for 5-10 generations.
Selection differential
• S can be predicted if we can assume:
– normal distribution of phenotypes
– truncation selection
• Standardized selection differential (i)
S = iσP
i = S/σP = z/p
p = proportion selected
z = height of curve at
truncation point
i = standard deviations
from the mean
Values of standardized selection differential
p
i
p
i
0.90
0.20
0.09
1.80
0.80
0.35
0.08
1.86
0.70
0.50
0.07
1.92
0.60
0.64
0.06
1.99
0.50
0.80
0.05
2.06
0.40
0.97
0.04
2.15
0.30
1.16
0.03
2.27
0.25
1.27
0.02
2.42
0.20
1.40
0.01
2.67
0.15
1.55
0.005
2.89
0.10
1.76
0.001
3.37
Becker, 1984 – Appendix Tables 2 and 3 (infinite population size)
Response to selection
R=h2S
S = iσP
R=ih2σP
•

h 

2
2
A
2
P
R
i
2
A
P
Applies to individual plants in a population
– Selections made before flowering + controlled matings among
selected individuals
– Mass selection + selfing of selected plants
Family selection
(O)
Parental plant
in reference
population
(R)
Recombination
unit
(X)
Selection Unit
(progeny mean)
(W)
Individual in
improved
population
Cov(X,W) determines expected gain from selection
Hallauer, Carena and Miranda (2010) Chapt. 6
Intrapopulation Improvement
Method
Progenies tested
Recombination
unit
Mass selection (both parents)
Individual plants
Individual plants
Mass selection (one parent)
Individual plants
outcrossed seed
Half-sib (progeny selected)
Half-sib families
Half-sib families
Half-sib (parent is selfed)
Half-sib families
S1 family
Modified ear-to-row
Half-sib families
outcrossed seed
Full-sib
Full-sib families
Full-sib families
S1 family
S1 family
S1 family
S2 family
S2 family
S2 family
Intrapopulation Improvement
Method
Mass selection (both parents)
Mass selection (one parent)
Half-sib (progeny selected)
Half-sib (parent is selfed)
S1/Testcross
Modified ear-to-row
Full-sib
S1 family*
S2 family*
Expected Gain
i A  P
2
i(1/2) A
i(1/4) A2
i(1/2) A2
i(1/2) A2
i (1/8) A2
i(1/2) A2
2
i A  P
2
i (3/2) A
Generations/Cycle
2
1
P
 Phs
 Phs
 Phs
 Phs
 Pfs
1
2
3
4
1
2
3
S1
P
4
S2
σP is the square root of variance; pertains to selection units
*additive variance for inbred progeny includes an additional
component that is a function of the degree of dominance
Phenotypic variance of families
Half-sibs
P 
HS
1
re
Full-sibs
P 
FS
1
re
      
S1 families
P 
S1
1
re
     
S2 families
P 
1
re
      
S2
    
2

2

2

2

1
e
1
e
1
e
1
e
2
GE
2
GE
2
A
1
2
2
GE
2
GE
2
A
1
4
2
A
3
2
2
D
1
4
1
4
2
A
2
D
3
16
2
D
 
Error variance
2GE 
Variance due to genotype x environment interactions
2

r = # replications
e = # environments
Interpopulation Improvement
Method
Progenies tested
Recombination unit
Reciprocal recurrent
Half-sib families
S1 families
Reciprocal full-sib
Full-sib families
S1 families
Testcross
Testcrosses
S1 families
Reciprocal recurrent selection
•
Half-sibs evaluated (Design I matings)
A0
S1 recombined
A0 females
HS
yield trials
HS
yield trials
B0 females
B0
•
A1
A1 x B1 (improved cross)
S1 recombined
B1
Full-sibs evaluated
A0
S1 recombined
A1
FS
B0
yield trials
S1 recombined
Full-sib RRS
• plants must have two “ears”
• twice the number of plants
can be evaluated
• continue to inbreed and
evaluate specific crosses
A1 x B1 (improved cross)
B1
Interpopulation Improvement
Method
Expected Gain
Generations
per cycle
Reciprocal recurrent
(i (1/2)
2
A ( P1)
 P(P1hs ) )  (i (1/2)
2
A (P 2 )
 P(P 2hs ) )
3
Reciprocal full-sib
(i (1/2)(
2
A ( P1)

2
A (P 2 )
)  P( P1xP 2 fs)
3
Testcross
Depends on choice of tester, but typically
• Cross P1 plants to inbred line from P2
• Cross P2 plants to inbred line from P1
3
Phenotypic variance of families for RRS
r = # replications
e = # environments
P(P1hs) 
P(P1xP2fs) 
1
re
1
re

2
(P1)
 
1
e
2
GE(P1)
 
1
2
2
A(P1)
2
2
2  1e GE
 12 (2A(P1)  2A(P2) )  14 D(P1P2)
Comprehensive breeding program
•
Development of breeding populations from diverse
sources such that the performance of the
population cross is maximized while maintaining
high levels of genetic variance within each
population
•
Application of an effective recurrent selection
procedure
•
Development of inbreds from each population with
good combining ability and recycling of superior
inbreds back into the base populations
Eberhart et al., 1967
Increasing selection response
• Increase the selection differential (reduce
proportion selected)
•
•
•
•
Increase the coefficient of A2
Increase A2
Reduce nongenetic effects
Reduce generations/cycle or increase
generations/year
Choice of selection method
I. Breeding Objectives
•
Open-pollinated varieties, synthetics or hybrids
– Status of commercial seed sector
– Strategy for distribution of seed
•
•
Elite variety or genetic resource
Target production environments
– Low or high inputs?
– Narrow or broad adaptation?
•
Number of traits, relative importance of traits
Choice of selection method
II. Genetic, Environmental, External Factors
• Heritability of the trait(s)
• Extent of GXE
• Type of gene action
• Effects of inbreeding on the trait
• Expected gain per cycle
• Number of seasons per cycle
• Growing seasons per year and availability of offseason nurseries
• Seed quantities required for screening
• Costs and resources available
Maize families – seed quantity issues
Family
Crosses
Seed quantity
Comments
Half-sibs
1. Collect pollen in bulk
and cross to a female
plant
2. Take pollen from one
male and cross to
several females
1. One ear
1. Controlled pollinations or
by detasseling
2. ~4 ears
2. Full-sib families within
half-sibs
Full-sibs
Cross two plants
One or two ears  With or without reciprocals
S1, S2, etc.
Self pollination
One ear
Testcrosses
1. Cross one male plant
to a female tester and
self
2. Cross an S1 line to a
tester (population or
inbred line)
1. ~4 ears
2. many ears
 Seed quantities decrease
with inbreeding
 Can increase a line by
selfing or by sib-mating
 Controlled pollinations or
by detasseling (if S1 line is
female)
one ear  at least four single-row plots
Maintenance of Maize Streak Virus Resistance
Modified full-sib family selection
Year 1
Main season,
savanna zone
Evaluate full-sib families in target environments
Year 2
First season,
forest zone
Recombine selected full-sib families by making
plant to plant crosses between families
Year 2
Second season
(high disease
pressure)
1) Plant F1 families ear-to-row under MSV
infestation
2) Remove susceptible plants and offtypes
before flowering
3) Make reciprocal crosses between best plants
in good rows to generate new full-sib families
Off-season: data entry and analysis
Reciprocal S1 Testcross Selection (modified)
Year 1
First season
 Self
Year 1
 Evaluate ~500 S1 families (2 reps, 2 loc)
Second season  Select for disease resistance and other
(high disease
highly heritable traits
pressure)
 Testcross to the reciprocal population
Year 2
Main season
 Evaluate ~ 200 testcrosses (3 reps, 4 loc) in
the target environments
 Select for yield and other agronomic traits
Year 2
Off-season
 Recombine selected S1 families
• Can stagger populations so that one is at the S1 stage and the other is
at the testcross stage each year
Meadowfoam - use of blue bottle flies as pollinators
S1 testcross selection in meadowfoam
Year 1 (spring)
– Self ~300 plants in the greenhouse with blue bottle flies
Year 1 (fall) + Year 2 (spring)
– Plant rows of ~5 seeds per family in isolation with bees, 2 blocks
– Reject S1 families with poor agronomic characteristics (disease,
insect damage, small seeds, etc)
– Harvest ~6000 seeds per family in bulk
Year 2 (fall) + Year 3 (spring)
– Evaluate ~150 testcross families in yield trials, select ~30
Year 3 (fall)
– Recombine S1 seed of selected families in greenhouse
•
•
Further selfing of selected S1 lines
Evaluation of experimental varieties in yield trials
Balancing resources for recurrent selection
• *Daylength can be controlled in the greenhouse to complete
a generation in four months
• Makes efficient use of greenhouse space
• New experimental varieties can be evaluated every year