Transcript Plant Genetics 2003 - Biology Department | UNC Chapel Hill

The genetic architecture of crop
domestication: a meta-analysis
María Chacón, Todd Vision, Zongli Xu
Department of Biology
University of North Carolina at Chapel Hill
October 23, 2003
Domestication
• “Domestication involves genetic changes in
populations tending to infer increased fitness
for human-made habitats and away from
fitness for wild habitats.” (Harlan 1995)
• Domestication syndrome: The stereotypical set
of adaptations to human habitat seen in crops
Quantitative trait locus (QTL) mapping in
wild x domesticated crosses
• Genetic architecture of domestication
– Number of QTL
– Effect sizes
– Mode of action
– Chromosomal locations
• Limitations
– Underestimate QTL #
– Overestimation of effect size in small samples
– QTL are located to large chromosomal segments
– Difficult to distinguish linked vs. pleiotropic QTL
• Mapping populations differ in
– Statistical power
– Ability to measure dominance
QTL mapping
Parents
X
F1
QTL
F2 genotype
QTL
F2 Phenotype
QTL map in rice
(Cai and Morishima, 2002)
Received wisdom regarding
domestication QTL (DQTL)
• Few loci of major effect
• Domestication alleles tend to be recessive
• DQTL tend to be clustered among and
within linkage groups
• DQTL tend to be homologous among
related crops (e.g. fruit weight QTL in the
Solanaceae)
Crop systems
Pulses
Common bean
Cowpea
Fruit crops
Eggplant
Pepper
Tomato
Watermelon
Vegetable crops
Lettuce
Grain crops
Maize
Pearl millet
Rice
Sorghum
Wheat
Wild rice
Industrial crops
Cotton
Sunflower
Sugarcane
Questions
• What is the effect of study power on
– The # DQTL per trait?
– The effect sizes of the DQTL?
• Do DQTL tend to be recessive even for polygenic
traits?
– What is the effect of breeding system?
– What does the pattern suggest about the origin of the
domestication alleles?
• Clustering of DQTL among and within linkage groups
– Is it an artifact of pleiotropy?
– Is the pattern of clustering consistent with the major hypothesis
concerning its origin
Questions
• What is the effect of study power on
– The # DQTL per trait?
– The effect sizes of the DQTL?
• Do DQTL tend to be recessive even for polygenic
traits?
– What is the effect of breeding system?
– What does the pattern suggest about the origin of the
domesticated alleles?
• Clustering of DQTL among and within linkage groups
– Is it an artifact of pleiotropy?
– Is the pattern of clustering consistent with the major
hypothesis concerning its origin
Gene action
Genotype
A2A2
Genotypic
value
-a
0
Additive
Recessive
-1.25
-1.00
-0.75
-0.25
0
A1A2
A1A1
d
+a
Dominant
0.25
d/a=gene action of the A1 allele
0.75
1.00
1.25
Expectations for gene action
of domestication alleles
• Domestication alleles are recessive (Lester,
1989, Ladizinsky, 1998)
• If adaptation uses new mutations
autogamous are expected to fix more
recessive alleles than allogamous (Orr and
Betancourt, 2001)
• If adaptation uses standing variation, the
probability of fixation of alleles is independent
of dominance (Orr and Betancourt, 2001)
Gene action of domestication alleles
Crop
Dominant
Rice
22
Tomato
3
Sorghum
2
Eggplant
16
Total
54
Autogamous
(37%)
Maize
11
Sunflower
32
Pearl millet
22
Wild rice
13
Total
78
Allogamous
(34%)
Total
132
Codominant Recessive d/a
Mating system
7
15
0.03
Autogamous
11
16
0.10
Autogamous
1
6
1.40
Autogamous
7
24
0.78
Autogamous
26
67
(18%)
(46%)
15
17
0.11
Allogamous
17
29
0.13
Allogamous
14
31
0.25
Allogamous
13
16
-0.43
Allogamous
59
93
(26%)
(40%)
85
160
Average d/a = 0.570 (autogamous), 0.015 (allogamous)
Two-tailed paired t-test: p<0.31
Findings
• Domestication alleles are not always
recessive
• Autogamous and allogamous crops have
equal proportions of recessive and
dominant domestication alleles
• Results are more compatible with the
predictions of the ‘standing variation’
model than the ‘new mutation’ model
Questions
• What is the effect of study power on
– The # DQTL per trait?
– The effect sizes of the DQTL?
• Do DQTL tend to be recessive even for polygenic
traits?
– What is the effect of breeding system?
– What does the pattern suggest about the origin of the
domesticated alleles?
• Clustering of DQTL among and within linkage groups
– Is it an artifact of pleiotropy?
– Is the pattern of clustering consistent with the major
hypothesis concerning its origin
Why might DQTL be clustered?
• Predicted from some population genetic models (Le
Thierry D’Ennequin et al. 1999)
– Assuming
• DQTL could arise throughout the genome
• Introgression from wild relatives
– Selection will prefer linked QTL in disequilibrium
– Clustering should be more apparent in allogamous than
autogamous crops
• Potential for methodological artifact
– One pleiotropic QTL would be detected multiple times
– This would give the false appearance of clustering
– Conservative set of QTLs chosen to reduce problems of
pleotropic QTL (one per trait category per locus)
Classification of domestication traits
Earliness
Growth habit
Increase in yield
Gigantism
Seed dispersal
Days to flower
Heading date
Fruiting date
Ripening date
Plant height
No. tillers/plant
No. of branches
Average length of
nodes
No. of nodes
Kernel No./spikelet
Kernel No./plant
Kernel No./panicle
Grain yield
Spikelet No./ spike
Spikelet No./panicle
Spikelet density
Spike No./panicle
Cupules/rank
No. of rows of cupules
Fruit number
Fruit yield
Seed size/weight
Panicle length/weight
Spike length/weight
Fruit diameter/weight
/length
Shattering rate
Brittle rachis
Awn length
Full data
set
Reduced
data set
How to test for QTL clustering
• Clustering among linkage groups
– Measured by a X2 goodness of fit test
• Clustering within linkage groups
– Measured by simulation (randomly
assigning same number of QTL and
measuring distance between neighboring
QTL)
Clustering of DQTL among and within LGs
Crop
Outcrossing
rate
Clustering
among
Clustering
within
Rice
<1%
yes
no
Rice
<1%
yes
yes
Rice
<1%
yes
yes
Common bean
1-5%
yes
yes
Tomato
1-5%
yes
no
Tomato (r)
1-5%
yes
no
Wheat
1-5%
yes
yes
Wheat (r)
1-5%
yes
no
Pepper
12-15%
yes
yes
Pepper (r)
12-15%
no
yes
Cowpea
12-15%
yes
yes
Eggplant
12-15%
yes
yes
Eggplant (r)
12-15%
no
no
Sunflower
25-40%
yes
no
Sunflower (r)
25-40%
yes
no
Pearl millet
25-40%
yes
yes
Pearl millet
25-40%
yes
yes
Maize
>40%
no
no
Wild rice
>90%
no
yes
Clustering of DQTL in common bean
Non-clustering of DQTL in maize
Clustering of non-domestication QTL
Crop
Cross type
Rice
DxD
Sunflower D x D
Outcrossing
rate
Clustering
among
Clustering
within
<1%
no
yes
25-40%
no
no
Maize
DxD
>40%
no
yes
Rice
WxD
<1%
yes
yes
Cowpea
WxD
12-15%
yes
yes
Alternative explanations
• Are QTL clustered because they map to gene dense
regions?
– Suggested for wheat (Peng et al. 2003)
• Preliminary test in rice using high density transcript
map (6591 ESTs, Wu et al. 2002)
– Counted number of QTLs and markers in 5cM windows
– Average # of markers/windows = 4.41
– Weighted avg. # of markers/window for QTL = 3.49
QTL homology
• Observed for QTL in several systems
– Cereals (grain size, flowering time, shattering)
– Solanaceae (fruit size, shape)
– Beans (seed size)
• Not necessarily domestication trait specific
• If clusters reflect chromosomal regions that
are particularly liable to contain QTL
– Some correspondence in the location of QTL
among related species is to be expected
– So do homologous QTL really correspond to the
same genes?
Summary
• DQTL number and effect size
– Trend toward less DQTL and larger effect sizes in low power
studies
– Some major DQTL detected in powerful studies (e.g.
sugarcane)
• Mode of gene action and origin of DQTL alleles
– d/a is not significantly different between allogamous and
autogamous crops
– Results consistent with ‘standing variation’ model
• Clustering of DQTL
– Does not appear to be an artifact of pleiotropy
– Not consistent with introgression hypothesis
– Appears to reflect inherent differences among regions of the
genome
Acknowledgements
• All those who helped provide
supplemental data from their QTL
studies:
– John Burke (sunflower)
– Lizhong Xiong (rice)
– Valerie Poncet (pearl millet)
– Raymie Porter and Ron Phillips (wildrice)
Statistical power
• Power
– Probability of rejecting the null hypothesis (absence of QTL) when
it is false = probability of detecting a QTL when it is present
– Calculated by simulation
• Assumptions
– Single codominant QTL
– Constant small additive effect
– Constant environmental variance
Power of study and # DQTL detected
# DQTL detected per trait
20
18
16
14
12
10
8
6
4
2
0
0
0.2
0.4
0.6
Power
0.8
1
1.2
Power and effect size of DQTL
100
90
80
70
60
50
40
30
20
10
0
0
0.1
0.2
0.3
0.4
0.5
Power
0.6
0.7
0.8
0.9
1