Historical Genomics of US Maize: Domestication and Modern Breeding

Genome-Wide View of Breeding History and Selection in North American Maize

Dr. Jeffrey Ross-Ibarra University of California at Davis

Patterns of diversity across the genome are reflective of the historical processes of selection, drift, and mutation that have acted on a species during its evolution. Here, we apply population genetic analysis of two genomic datasets to assess the spatial and temporal signatures of evolution in the maize genome. We first describe genome-wide patterns of evolution during two epochs of selection, domestication and modern breeding, using full-genome resequencing of a number of maize and teosinte lines. Then, using a chronologically-stratified panel of corn belt lines, we dissect temporal patterns of ancestry and selection in greater detail across the 80+ years of North American corn breeding. We find distinct impacts of selection during domestication and modern breeding on diversity and gene expression, identify candidate regions targeted by selection, and describe changes in genetic structure and ancestry during the evolution of modern corn belt lines.

HISTORICAL GENOMICS OF MAIZE

Jeffrey Ross-Ibarrawww.rilab.org

Acknowledgements

Ed Buckler (Cornell, USDA)

Jeff Glaubitz (Cornell)

Major Goodman (NC State)

Mike McMullen (Missouri, USDA)

Chi Song (BGI)

Nathan Springer (Minnesota)

Doreen Ware (CSHL, USDA)

Xun Xu (BGI)

Matthew Hufford Joost van Heerwaarden Tanja Pyhäjärvi Jer-Ming Chia


��

��

��

��

��

��

�

��

��

��

��

��

��

��

��

��

�

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

�

Spatial resolution -- genome resequencing

Temporal resolution -- SNP genotyping


��

�

��

��

��

��

��

�

��

��

��

��

��

��

��

��

��

�

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

�



ρ

0.00

20.

006

0.01

00.

010.

020.

030.

04

ρρππ

lower 1% cent10cM/Mb

ρ

Chr 10

Maize Hapmap I: low-copy fraction of genome

Gore et al. 2009 Science

• Re-sequenced low-copy portion of genome of 27 inbreds

• High diversity genome, but large regions of low recombination

• Identified low diversity putative targets of selection

• No comparison to older material

• Cannot distinguish domestication from modern improvement

Chia et al. Submitted

Maize Hapmap II: full genome resequencing

75 resequenced genomes & >21M SNPs75 resequenced genomes & >21M SNPs

Sequenced Genomes

Sequence Depth

Locality

Improved 35 5.04 X 13 Temperate, 13 Tropical,3 Mixed, 6 Chinese

Landraces 23 5.34 X 9 South American, 14 North American

parviglumis 14 3.81 X4 Jalisco/Nayarit,

9 Balsas, 1 Oaxaca1 1Oaxaca

mexicana 2 6.83 X 1 Jalisco, 1 Morelos

Tripsacum 1 12.62 X Colombia

parviglumis

Landraces

Improved Lines

Domestication

Improvement

Hufford et al. Submitted

Maize

Parviglumis

Maize

Parviglumis

Genic Tajima's D

Tajima's D

De

nsity

-4 -2 0 2 4

0.0

0.1

0.2

0.3

0.4

0.5


Maize

Parviglumis

Maize

Parviglumis

Genic Tajima's D

Tajima's D

De

nsity

-4 -2 0 2 4

0.0

0.1

0.2

0.3

0.4

0.5

De

nsity

0.0

0.1

0.2

0.3

0.4

0.5

Genome-wide Tajima's D

Tajima's D

De

nsity

-4 -2 0 2 4

0.0

0.1

0.2

0.3

0.4


Chen et al. 2010 Genome Research

COG3-like FMP25-like frq PAC10-like SEC14-like NSL1-like NCU02261B

Carib

LA

Out

Ellison et al. 2011 PNAS

• Cross-Population Composite Likelihood Ratio

• Identifies extended regions of differentiation

• Compares to simulated model

• Define selected regions, estimate s

Artificial selection in the maize genome

• Stronger selection during domestication (s~0.015 vs. s~0.003)

• Selection strength consistent with archaeology & natural selection

• Continued selection on domestication genes (18% overlap)

• ~3,000 genes in selected regions. Identified ~1100 candidates

• 5-10% of selected regions had no genes


he maize genome

log s

Fre

qu

en

cy

-6 -5 -4 -3 -2

05

00

01

50

00

25

00

0

domestication

improvement

Gore et al. Science 2009

Mb chr 8

proportion

FST

Mb chr 8

proportion

0.0

0.2

0.4

0.6

0.8

1.0

fixed shared unique temperate unique tropicalemfixed shared unique temperate unique tropical

0 20 40 60 80 100 120 140 160 180

00

.05

0.1

50

.25

0.3

5

Population-specific selection

Population-specific selection


• Many population-specific loci

• Stronger overall estimate of selection

• Still weaker than domestication

• Lower power to rule out drift

Candidate genes for maize improvement


Selection on the transcriptome

• Expression at >18K genes of 27 maize and teosinte using Nimblegen array

• Domestication directly selected on candidate gene expression

• Breeding has worked with highly expressed genes

• Modern breeding selected for heterosis in expression

Candidates: DOM IMP

Change in expression * --

Decreased variance * *

Tissue specificity# --

overall higher expression

Dominance in crosses# -- inter > intra

# Data from Stupar et al. 2008 BMC Plant Bio; Sekhon et al. 2011 Plant J.


Regions with no genes likely regulatory

teosintemaize

Swanson-Wagner et al. Submitted

Expr

essi

on

Genetic load in modern maize

• ~1% of SNPs disrupts predicted protein

• 10,117 premature stop-codons

• 2557 (or 12%) high-confidence genes has a stop-codon in ≥1line

• 870 genes have a segregating stop-codon in ≥ 2 of improved lines

Chia et al. Submitted

Artificial selection across the maize genome


��

��

��

��

��

��

�

��

��

��

��

��

��

��

��

��

�

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

��

�



Detailed historical look at selection during modern breeding

• Larger sample size

• Increased chronological resolution

• Take into account heterotic group structure

• Focus on corn belt varieties

Resolve caveats of resequencing approach

Genotyping and Sampling

• 400 N. American corn belt lines

• Genotype on public Illumina 55K

• Tracked population structure, ancestry, and selection over time

99

94

70

137

land races

<1950 inbreds

1960-1970 inbreds

ex-PVP

0

1

2

3

van Heerwaarden et al. Submitted

(Oh43, W22, B14)

(B73, 207, Mo17)

Population structure through time




C



C

C



C

C



C

C



IodentNSSSS

Number and diversity of ancestors decreases through time

ance

stry


c.

IodentNSSSS

effe

ctiv

e #

anc

esto

rsdi

vers

ity o

f anc

esto

rs

SS

0.0

0.1

0.2

0.3

0.4

unde

fined

O

H43

C

I31A

H

Y

B8

B

10

B14

B

96

B37

I2

05

OH

07

CI3

A

CI1

87-2

C

I540

I2

24

WD

456

O

h316

7B

Tr9-

1-1-

6

FE2

LE

773

N

D20

3

B73

A

634

A

635

N

28

H10

0

B64

B

14A

B

68

CM

105

B

84

A63

2

N19

6

LH19

5

LH20

5

LH19

6

LH19

4

991

LH

74

PH

G39

LH

132

G

80

7800

4

7801

0

4676

A

7800

2A

794

LP

5

PH

T55

H

8431

1 2 3

Ancestral contributions of individual lines

Ancestral contributions of individual lines

IDT1 2 3

0.0

0.1

0.2

0.3

0.4

unde

fined

O

H43

W

F9

A37

5

A50

9

A55

6

I198

B

2

H5

H

Y

B16

4

G

ND

203

B

10

B14

L3

17

I205

C

I187

-2

C49

A

I224

A

H83

Tr

9-1-

1-6

C

11

IDT

A

638

M

14

207

P

HN

34

PH

P76

P

HW

86

PH

G50

P

HG

35

PH

G71

IB

014

P

HG

83

LH15

0

PH

G29

P

HG

72

PH

G84

P

HZ5

1

PH

V78

P

HK

42

PH

N11

P

HH

93

PH

J33

P

HN

73

PH

R62

Weak landrace structure still evident in modern inbreds

ance

stry

pro

port

ion

IodentNSSSS

Selection and frequency over time

Duvick et al. 2004 Plant Breeding Reviews

advantageous alleles show continual increase in frequency over time

•Identify genetic structure across all time periods (PCA, Tracy-Widom)

•Treat time period as a phenotype for association mapping

•For each SNP, compare two models:

•M0: Frequency determined by drift and population structure (covariance matrix of all SNPs)

•M1: Frequency determined by drift, population structure AND environment (time)

•Bayes Factor for each SNP

Time as a phenotype to identify selection

0

1

2

3

Coop et al. 2010 Genetics

Temporal association identifies candidates of selection

• 236 Candidate regions (1021 genes) with consistent BF across runs

• Include stress response, lignin biosynthesis, auxin response

• Two of the top candidates ZmErecta (US7847158) and ZmArgos (US7834240) patented for plant growth, yield

ssociation identifies candidates

te regions (1021 genes) with consistent BF

0 1 2 3

0.0

0.2

0.4

0.6

0.8

1.0

era

p

Temporal association identifies candidates of selection

• 236 Candidate regions (1021 genes) with consistent BF across runs

• Include stress response, lignin biosynthesis, auxin response

• Two of the top candidates ZmErecta (US7847158) and ZmArgos (US7834240) patented for plant growth, yield

Div

ersi

ty

Genome Sequence

Selective Sweep

Selective sweeps

No abnormal haplotypes, ancestry at selected sites

0 1 2 3

01

23

4p p p p p p p p

era

ratio

obs

erve

d/ra

ndom

ized

No abnormal haplotypes, ancestry at selected sites

Genome-wide ancestry not determined by selection

Weak correlation between good alleles and contribution to modern lines

<1950’s inbreds

<1980’s inbreds

ex-PVP lines

selected regions

% a

nces

try

Weak correlation between good alleles and contribution to modern lines

Structure and selection in corn belt maize

Some final thoughts on adaptation...

modern corn

phenotypic space

FISHER-ORR MODEL OF ADAPTATION

modern corn teosinte

phenotypic space


modern corn teosinte

phenotypic space


landraces

phenotypic space


landraces

phenotypic space


Brown et al. 2011 PLoS Genetics

• Effect sizes of ear vs. other consistent with Fisher-Orr model

• Larger effects → larger fitness difference → stronger selection

Thank You!

Education

Historical Genomics of US Maize: Domestication and Modern Breeding