This project aimed to develop drought tolerant maize lines for Asia using Marker Assisted Recurrent Selection (MARS) and genome-wide selection (GWS). Yellow drought tolerant inbred lines were produced from bi-parental crosses between CIMMYT-Asia lines and African drought tolerant donors. Multiple cycles of selection resulted in 50-100% higher genetic gains for grain yield under drought compared to phenotype-only selection. The project provided training in maize breeding, phenotyping, and molecular breeding techniques. Challenges included delays in genotypic data and germplasm exchange. Future work could evaluate populations based on observed heterosis rather than extrapolated data.

1. Asian Maize Drought Tolerance
(AMDROUT) Project
SP3 PROJECT G4008.56
Principal Investigator: B. S. Vivek

2. Project Details
Grant Period:
 (Start: Nov 08) (End: Oct 2013)
 No cost extension to July 2014

3. What?
Drought tolerant lines
 using Marker Assisted Recurrent Selection (MARS) based on
genome-wide selection (GWS)
Yellow drought tolerant inbred lines
Proof of concept for MARS-GWS
Scientists trained in molecular breeding
Principle Outputs

4. Why?
80% of maize in Asia is grown under rain-fed conditions and
prone to drought stress

5. Drought in Asia
 …wheat and maize harvests will
be strongly affected by droughts
…unless states ..can quickly
adapt their agricultural practices
 China, Indonesia and Pakistan
were relatively well-placed to
adapt to climate change. India
was found to have one of the
lowest capacities to adapt ….its
maize production (in C and N
India)

6. 2012: The worst drought US
witnessed in recent decades
Source: Bob Nielsen (Purdue University)
• Sixth worst
departure from yield
trend since 1866!
• Yield estimate would
be 23% lower than
predicted in 2012…
• Drought coupled
with heat stress
• Possible global
implications in terms
of maize prices
Wherever you are, …you need stress resilience
to ensure stable performance

7. Collaboration  Dr. B. S. Vivek, CIMMYT-India
 Girish Kumar Krishna, CIMMYT-India
 V. Vengadessan , CIMMYT-India
 P. H. Zaidi, CIMMYT-India
 Le Quy Kha, NMRI, Vietnam
 Pichet Grudloyma, NSFCRC, Tak Fa,
Thailand
 I.S. Singh, Krishidhan Seeds, India
 R. Babu, CIMMYT-India
 Eureka Ocampo, Institute of Plant
Breeding, UPLB, Philippines
 Fan Xingming, YAAS, Kunming, China
 M. Azrai, ICERI, Maros, Indonesia
 R.P. Singh, Syngenta, India
 J. Burgeño, CIMMYT-Mexico
 J. Crossa, CIMMYT-Mexico

8. Final irrigation before imposing drought
Drought Phenotyping
Optimal
Management
Managed
Drought
Drought expression at flowering
Typical genotypic variability
under drought

9. Form S2 x tester (CML474)
Evaluate test crosses (At least
3 drought + 3 optimal sites)
Calculate marker effects
(MARS-GWS)
Form C1 with best S2 lines
based on phenotype data
Genotype S2 families: 350
polymorphic whole genome SNPs:
Kbioscience KASPar assay
Form c1F2
(recombine best
looking c1 plants)
C2
S6
Genotype-only
selection (24
plants with
highest GEBVs
recombined)
•CML470
•VL1012767
•VL1012764
•CML472
S2
F2 (S1)
F1
P1 P2x4 elite yellow Asian
lines (lack drought
tolerance)
2 African white drought
tolerant donors
•CML444
•CML440
Populations
•AMDROUT1: CML470/CML444
•AMDROUT2: VL1012767/CML444
•AMDROUT5/6: VL1012764/CML444//CML472xCML440
Genotype C1 plants
Calculate Genomic
Estimated Breeding
Values (GEBVs) using
marker effects
How? Genome Wide Selection (GWS)

10. Mean
Lines developed by pedigree
selection
Lines selected for recombination
from C0 phenotyping
Cycle 3 MARS lines
Population of random lines
extracted from a cross
MARS increases the
frequency of favorable
alleles
Moves the mean of the
selected population beyond
the original distribution
MARS / GWS

11. Sl.
No.
AMDROUT Population
Number
Population
Type
Data from K BioScience
Marker effects
available
GEBVs
available
No. of individuals in
the population
No. of SNPs
assayed
1 AMDROUT1 F2:F3 294 340 Yes No
2 AMDROUT1 Cycle-1 242 318 Yes
3 AMDROUT1 Cycle-2 347 278 Yes
4 AMDROUT2 F2:F3 188 376 Yes No
5 AMDROUT2 Cycle-1 258 351 Yes
6 AMDROUT2 Cycle-2 346 271 Yes
7 AMDROUT5 F2:F3 197 421 Yes
8 AMDROUT6 F2:F3 183 340 Yes
9 AMDROUT5/6 Cycle-1 352 425 Yes
10 AMDROUT5/6 Cycle-2 252 399 Yes

12. @60% gain
@40% gain
@80% gain
0.0
1.0
2.0
3.0
4.0
5.0
6.0
0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
days
t/ha
Cycles of selection
GY
ASI
Genetic Gains for Grain Yield under Drought:
AMDROUT1 1 location
3 reps
2 row plots
H 0.74
c1 c2: Marker-only selection
@ per cent
gain over F2

13. @50% @40%
@60%
0.00
0.50
1.00
1.50
2.00
2.50
3.00
t/ha
Cycles of selection
GY
GY
Genetic Gains for Grain Yield under Drought:
AMDROUT2
1 location
3 reps
2 row plots
H 0.74
@ per cent
gain over F2
c1 c2: Marker-only
selection

14. AMDROUT1 cycles
AMDROUT1c2 (Phenotypic
selection followed by genotype-
only selection)
AMDROUT1c1F2 (Two cycles
of phenotype-only selection)

15. Therefore:
For grain yield under drought: one cycle of
phenotypic selection followed by one cycle of
genotype-only selection gives 50-100% higher
genetic gain compared to two cycles of phenotype-
only selection
GWS is a useful methodology to improve source
(breeding) populations

16. Moving on…..
Nov 2013: Test crosses with cycles of selection
May 2014: Drought and well watered trial
information on per se and test cross performance
of cycles of selection
August 2014: Draft publication
Meanwhile:
Wet season data is being collected
Second season of drought screening will be
completed by May 2014

17. Germplasm Outputs
Cycles of selection from bi-parental crosses between
CIMMYT-Asia lines and African DT donors
 AMDROUT1c3, AMDROUT2c3, AMDROUT(5x6)c3
 AMDROUT1c2, AMDROUT2c2, AMDROUT(5x6)c2
 AMDROUT1c1F2, AMDROUT1c2F2, AMDROUT2c2F2,
AMDROUT(5x6)c1F2, AMDROUT(5x6)c2F2
Early generation lines from above cycles of selection
DH lines from the above cycles?

18. Genotyping Outputs
1215 SNP data of 175 advanced lines
Genetic linkage maps of two populations viz.
AMDROUT1 and 2
Marker effects of SNPs for 4 bi-parental populations

19. Other Outputs
Proof of Concept publication on MARS-GWS method
Trained scientists
 Cambridge Molecular Breeding
 Maize breeding
 Drought phenotyping
 Data management
 Visiting scientists
 On site visits
 Statistics, molecular breeding course planned

20. Looking further ahead: Movement of
AMDROUT germplasm to farmers’ field

21. SWOT: What worked?
 An opportunity to learn and implement new technology
which promises to give higher genetic gain, especially under
drought.
 Main outputs from this project: yellow drought tolerant
selection cycles starting from white African germplasm as
donor. Yellow maize covers most of the Asian region, hence
having such breeding material on hand again will ensure
faster genetic gain. This output will be used in future
breeding either as drought tolerant donor or for deriving
inbred lines.
 Excellent training opportunity (for field evaluation and data
analysis) for collaborators.
 CIMMYT locations continue to give excellent data, along
with couple of other partner sites.

22. SWOT: What was difficult?
There were issues with timely delivery of genotypic
data from KBioscience, UK; but this issue has been
taken up and resolved.
Germplasm export to collaborators was slow and
timely evaluation of trials was difficult.
Getting good heritability of grain yield was difficult in
a lot of partner locations.

23. SWOT: What was difficult?
 An unforeseen insect attack by pollen eating blister beetles
(Cylidrothorax tenuicollis), hitherto unseen, almost derailed the
recombination of AMDROUT1 and 2
 Similar to a locust swarm

24. SWOT: What was difficult?
Cross selection was a challenge because tropical
maize (unlike temperate maize)
is very diverse (while temperate maize has a
narrower based and a better defined heterotic
behavior)
has a shorter history of genetic improvement
has inconsistent pedigree information
has poorly maintained historical data

25. SWOT: What I would do different in the
future?
Selection of breeding populations on observed
heterosis (evaluation of 3-ways between the target
populations and a tester panel) rather than
extrapolated heterosis (which was done here
through a Design II study);
 Heterotic partner known before inbreds are developed
 Direct use of observed heterosis
 Less reliance on extrapolation and historic information
which is often poorly managed

26. SWOT: What else could have been
different?
Perhaps ……..
Using multi-parent synthetics as the breeding
material (as opposed to the bi-parental populations
used here). For tropical maize, the onus on
selection of the best bi-parental crosses and the
right testers is too huge a risk. This risk is magnified
by working with few populations (4 in this case).
Using multi-parent synthetics would have
potentially reduced this risk.