Trait-associated SNPs provide insights

into heterosis in maize

Patrick S. Schnable

Iowa State University

China Agriculture University

Data2Bio, LLC

ICRISAT

19 February 2015

How to Translate Genomic Data into

Biological Understanding and Crop

Improvement?

2

B73 Reference Genome NGS data in NCBI SRA (Feb. 2014)

zmHapMap1

zmHapMap2

CAU resequencing

ISU Zeanome (RNA-seq)

Ames Diversity Panel

IBM RILs RNA-seq

CAAS resequencing

And many others

Schnable, Ware et al., Science, 2009

Te

ra (

10

12)

Ba

se

s

$32M

Associate Genes (or genetic markers)

with Traits

• Which of the ~50,000 maize genes control important traits?

• GWAS (Genome-wide association studies)

– Typically conducted on diversity panels

– By exploiting historical recombination events they yield higher resolution associations than QTL studies

– Identifies associations between genetic markers (e.g., SNPs) and traits

• Forward and reverse genetics

Y: phenotypic trait;

Pi: Fix effect of cross type population (N=4);

Sl: Fix effect of sub-population (N=25).

Approaches for GWAS

4

 

Y = u+ biPii=1

4

å + alSll=1

25

å + dSNP + e

• Single-marker GWAS approach

– SNP effects tested one at a time

– Using PLINK command line tool

• Stepwise regression approach

– SNPs fitted in a step-wise manner

– Using GenSel4 Stepwise (alpha=0.05,MaxMarkers=300)

• Bayesian-based approach

– SNPs fitted simultaneously into a model

– Using GenSel4 BayesC (chainLength=41,000, burnin=1,000)

GWAS for Yield-Related Traits

Kernel Count

Total Kernel Weight

Avg. Kernel Weight

Cob

Length

Cob Diameter

Cob Weight

Kernel Row Number

Jinliang Yang (杨金良)

Jeff Ross-Ibarra Lab, UC Davis

Yu, J. et al. Genetics 2008;178:539-551

Nested Association Mapping (NAM) Population

6

Four related populations (N=7,000

lines):

• NAM RILs (N=5,000 lines) + IBM

RILs (N=300 lines)

• Subset of MxRILs (N=300 lines;

IBM + NAM)

• Subset of BxRILs (N=800 lines;

IBM + NAM)

• NAM Partial Diallel (N=250 lines)

High Density Genotypic Data

• SNPs from three sources:

– Maize HapMap1* (1.6M)

– Maize HapMap2* (18.4M)

– Our RNA-seq SNPs from

5 tissues (4.9M)

7

# Concordance among overlapping variant sites

HapMap1

0.7 M

HapMap2

16.6 M

0.4 M

98.7%

1.2 M

96.6%

0.3 M

96.9%

0.2 M

RNA-seq

3.2 M

##

#

Imputation or

Projection

NAM RILs

BxNAM RILs

MxNAM RILs

NAM Diallels

*Gore, M.A., et. al.,

Science, 2009;

Chia, H-M, et. al., Nature

Genetics, 2012.

Merging and Filtering

Minor Allele Freq. (MAF) >= 0.1SNP Missing Rate < 0.6

Merged SNP set

13.0M

Phenotypic distributions

8

• CD=Cob Diameter, AKW=Avg. Kernel Weight, CL=Cob Length, CW=Cob

Weight, KC=Kernel Count, TKW=Total Kernel Weight

Based on ~100k observations/trait from 9 locations; ~20% our data and 80%

from: Brown, P. J., et. al., PLoS Genetics, 2011

Different GWAS Approaches are

Complementary

9

40/77 (52%) KAVs, representing 39 chromosomal bins

(bin size =100kb), have been cross-validated.

Genotyped TAS

Cross-validated TAS

Single-variant GWAS (-log10(P-Value))

Baye

sia

n-b

ased G

WA

S (

Model F

req)

Bayesian-based

and single-variant

N=16/21(76%)

Bayesian-based

N=9/26(35%)

Single-variant

N=10/15(67%)

Stepwise

regression

N=6/14(43%)

10

GWAS identified >1,000 associations

for seven yield-related traits

Heterosis

•Genotypes: B73, F1, Mo17

•Understand fundamental biology

•Predict hybrid performance

12

Schnable and Springer, Annu.

Rev. Plant Biol. 2013

Variation in percent heterosis across

traits

KRN=Kernel Row Number, CD=Cob Diameter, AKW=Avg. Kernel

Weight, CL=Cob Length, CW=Cob Weight, KC=Kernel Count,

TKW=Total Kernel Weight

Missing heritability

Inclusion of dominant gene action

improves predictions

13

Percentage of HPH

herita

bili

ty

Four GWAS populations Only Diallel population

Additive Dominance General

herita

bili

ty

Percentage of HPH

Missing heritability

Classical Models for Heterosis

Over-dominance

x

AA bb aa BB Aa Bb

Complementation

Zamir

Additive or dominant gene action Over-dominant gene action

Degree of Dominance for TASs

15

Degree of dominance (h), where d denotes dominant

effect and a denotes additive effect.

h =d

a

A A B BBA

a

d

positive

dominance

h > 0.5

negative

dominance

h < -0.5

additive

-0.5 <= h <= 0.5

Trait Associated SNP Effects

16 *Dominance includes true dominance, over-dominance and pseudo-overdominance

Phenotype (P) = Genotype

(G) + Environment (E) +

GxE

• Genotype: NGS revolution and GBS

• Environment: weather, soil type, water,

nutrients, disease pressure, agronomic

practices etc.

• GxE interactions complicate phenotypic

predictions, but offer fascinating avenues

of investigation

17

L

SL

L L

SL

S

S

L

S

SL S

SSL

L

L

SL

S

S

L

SL

L

L

SLS

SLL

S

L

L

S

SL

SS

SL LSL

S

The Drought Monitor focuses on broad-scale conditions. Local conditions may

vary. See accompanying text summary for forecast statements.S

SL

L

http://droughtmonitor.unl.edu/

U.S. Drought Monitor October 1, 2013

Valid 7 a.m. EDT

(Released Thursday, Oct. 3, 2013)

Intensity:D0 Abnormally Dry

D1 Moderate Drought

D2 Severe Drought

D3 Extreme Drought

D4 Exceptional Drought

Author: David Miskus

Drought Impact Types:

S = Short-Term, typically less than 6 months (e.g. agriculture, grasslands)

L = Long-Term, typically greater than

6 months (e.g. hydrology, ecology)

Delineates dominant impacts

NOAA/NWS/NCEP/CPC

E and GxE complicate

phenotypic predictions

• Strategies for dealing with “E” and “GxE”

– Study traits that are stable across E

– Conduct studies in controlled environments,

taking E and GxE out of the equation

– Control for and study the effects of E and GxE

statistically…embrace the opportunity to gain

a deeper understanding of the underlying

biology

18

Field-Based Phenotyping

• Sensors mounted on field-deployed

robots/UAV (expensive)

• Inexpensive, field-based sensors

• Unmanned Aerial Vehicles (UAVs)

RTK-GPS

John Deere Sub-Compact Utility Tractor Equipped

with Topcon Universal Auto-Steer System

20

Camera View Angle 2

Row 3

60-inch row spacing

Row 2Row 1 Row 4

GPS

Camera View Angle 1

GPS

Top-View

Back-View

4 m

Lead Screw Drive

3D TOF Cameras

“Next Generation Phenotyping”

Lie Tang Maria Salas

Fernandez

NIR Stereo Camera

21

Phenobot

22

Field-Based Phenotyping

• Sensors mounted on field-deployed

robots/UAV (expensive)

• Inexpensive, field-based sensors

• Unmanned Aerial Vehicles (UAVs)

Stop-Action Photography

for Phenomics

James Schnable Univ of NE

Yong Suk Chung Iowa State Univ

Dynamic Responses to Drought

Field-Based Phenotyping

• Sensors mounted on field-deployed

robots/UAV (expensive)

• Inexpensive, field-based sensors

• Unmanned Aerial Vehicles (UAVs)

Unmanned Aerial Vehicles (UAVs)

27

Phenotype (P) = Genotype (G) + Environment (E) + GxE

Predictive Models Will:

• Improve the accuracy of selection in plant breeding programs, thereby increasing the rate of genetic gain per year

• Enhance our ability to efficiently breed crops to withstand the increased weather variability associated with global climate change

• Improved ability to provide farmers with evidence-based recommendations for the appropriate varieties to plant in a given field, under a particular management practice in a given year, leading to greater farmer profits and enhanced yield stability

28

Summary

• DNA sequence variation (SNP) can explain 40-70% of genetic variation (considering only additive gene action) or 80-90% (including dominant gene action)

• Dominant effects explain much of the missing heritability

• Ratio of loci exhibiting positive dominant gene action to those exhibit negative dominant gene action is correlated with the degree of heterosis for that trait

• Determining which loci confer positive and negative heterosis for specific traits may increase our ability to predict hybrid performance

• Phenomics is a bottleneck in GWAS, GS and breeding

• Field-based sensors will allow us to study the genetics of dynamic traits rather than being limited to end-point traits

PSS has IP and equity interests in Data2Bio LLC

Data2Bio, LLC

31

•Founded in 2010, Data2Bio designs,

executes, analyzes and interprets

research projects involving next

generation sequencing

•Core strengths are experimental

design, genomics, bioinformatics, and

breeding support

•Academic and private-sector

customers on all continents except

Antarctica

•Proprietary genomic technologies

associated with DNA barcoding and

genotyping-by-sequencing (tGBS™),

as well as proprietary bioinformatic

pipelines