1. China Agriculture University
Presence/Absence Variation (PAVs) in Maize: Phenotypic Variation, Heterosis, and Domestication (with some final comments about cassava)
GCP21-II
Kampala, Uganda
22 June 2012
Patrick S. Schnable
Iowa State University
China Agriculture University
Data2Bio, LLC

2. The $30M B73 Maize Genome Sequencing Project
Schnable, Ware et al., 2009
The Maize Genome Sequencing Project, Rick Wilson, PI

3. Genome Projects are Analogous to the Lewis & Clark Expedition
Expensive and require extensive planning/coordination
Generates lots of information that requires subsequent analysis
Exploration of the unknown; expect surprises

4. Outline CNV and Presence-Absence Variation (PAVs)
The origin of “recurrent de novo CNV”
Revisit the domestication bottleneck in light of SV
Relevance to cassava?

5. Outline CNV and Presence-Absence Variation (PAVs)
The origin of “recurrent de novo CNV”
Revisit the domestication bottleneck in light of SV
Relevance to cassava?
Kai Ying Yan Fu
应开 傅延

6. Structural Variation (CNV & PAV)
What is overall level of (genic) SV?
Does SV contribute to phenotypic diversity?

7. Array-based Comparative Genome Hybridization (CGH)
Nimblegen’s HD2 Array (~2.1M probes)
Probes designed using a “frequency masked” 200 bp tile-path through the draft B73 genome sequence
Genotypes: B73, Mo17 (different heterotic groups)
Springer et al., PloS Genetics, 2009 7 7 7

8. Several hundred intact, expressed, phylogenetically conserved genes exhibit CNVs and PAVs
Segmentation Results
Springer et al., PloS Genetics, 2009
Beló A et al. Theor Appl Genet. (2010) (Rafalski Lab)

9. 2 Mb deletion on Ch 6*
*Includes ~2 dozen genes, incl. resistance gene (Xu Mingliang, 徐明良); this large deletion also identified by Antoni Rafalski (CGH) and Ed Buckler (Re-Seq)

10. CNV and PAV Loss (blue) & CNV Gain (red) Intervals relative to B73
Outer to Inner rings:
Teosinte vs. B73
Tx303 vs. B73
Hp301 vs. B73
Mo17 vs. B73
~10,000 YBP

11. Re-sequencing Six Inbreds
Identified PAVs
5.4X coverage/inbred
~150 Genes Present among Six Inbreds are Missing from B73
Lai et al., 2010

12. Classical Models for Heterosis
Complementation
Over-dominance x
AA bb
aa BB
Aa Bb
Zamir
Complementation of PAVs in pairs of inbreds could contribute to heterosis; PAVs could also play a role in over-dominance 12 12

13. Deletions can be favorable
Removal of traits lost during domestication
Ion uptake machinery (Heavy metal resistance in wheat)
Cyanide release (chemical defense) in white clover (Olsen et al., 2007; 2008)
Favorable rice QTL are in some cases PAVs:
qPE9-1, panicle erectness; Zhou et al. (2009) Genetics
semi-dwarf1, height; Ashikari et al. (2005) Science
GW5, grain width; Weng et al. (2008) Cell Res

14. Outline CNV and Presence-Absence Variation (PAVs)
The origin of “recurrent de novo CNV”
Revisit the domestication bottleneck in light of SV
Relevance to cassava?
Sanzhen Liu
刘三震

15. Detection of De Novo CNV in Human Trios
Mom
(no CNV)
Dad
“Kid”
(de novo CNV) X
Detection of Recurrent De Novo CNV in Human Trios
Mom
(no CNV)
Dad
“Kid”
(de novo CNV) X
Mom
(no CNV)
Dad
“Kid”
(de novo CNV) X

16. Novel CGH Patterns
Liu et al., Plant J, 2012

17. Reciprocal Gene Loss Model to explain speciation
Lynch and Force, 2000

18. Segregation of Non-Allelic Homologs (SNH) Generates “Recurrent De Novo CNV”
Model Predicts:
Changes in “gene complement” among RILs (gains and losses)
Should affect multiple RILs
Affected genes should have non-allelic positions in B73 and Mo17
Figure 3.

19. Segregation of Non-Allelic Homologs Generates “Recurrent De Novo CNV” and Novel Phenotypes
Consistent with model:
Losses and gains in gene content validated by Seq-Capture experiments (~200 segments in 2 RILs)
Specific losses (as detected by PCR) observed in multiple RILs (12/14 genes lost in 25% or 12.5% of RILs)
Affected genes are in non-allelic positions in the B73 and Mo17 genomes
Inbreds can have different gene complements than parents
Strong statistical support for association between gene loss and yield component traits in IBM RILs
Figure 3.

20. Association of Gene Loss with Traits
Losses of 2/14 (14.3%) tested segments are significantly associated with phenotypic variation:
Reduced yield component traits (adjusted p-values=0.03 and 0.01).
Increased tiller number (adjusted p-value=0.01).  
This rate (14.3%) is substantially higher than the 0.1% (N=670) of 515,620 control pairs of unlinked SNP markers that similarly exhibit associations with the same set of traits, identified via a two-dimension genome-wide scanning using a set of 1,016 SNP markers  

21. Outline CNV and Presence-Absence Variation (PAVs)
The origin of “recurrent de novo CNV”
Revisit the domestication bottleneck in light of SV
Relevance to cassava?
Kai Ying 应开
Camile
Rustenholz

22. Teosinte, the wild ancestor of maize
Zea mays sp parviglumis
Maize
Zea mays sp mays
Domestication ~ 10,000 years ago

23. Genes Selected During Domestication Have a Molecular Signature
Yamasaki, M., et al. Plant Cell 2005;17:
Copyright ©2005 American Society of Plant Biologists

24. CNV and PAV Loss (blue) & CNV Gain (red) Intervals relative to B73
Outer to Inner rings:
Teosinte vs. B73
Tx303 vs. B73
Hp301 vs. B73
Mo17 vs. B73
~10,000 YBP

25. Hypothesis: maize lacks some teosinte genes
B73
Mo17
Tx303
CML277
Oh7B
Maize inbreds
Domestication and crop improvement
Teo. 1
Teo. 2
Teo. 3
Teo. 4
Teo. 5
Teosinte
Genes conserved in maize and teosinte
Non-B73 genes
Non-maize genes

26. 3,774 contigs NOT aligned with B73 reference genome
1,000s of expressed genes in teosinte are missing from the B73 reference genome
B73 × Teosinte Ac3660
B73 × Teosinte Ac3662
F1 teosinte Ac3660
F1 teosinte Ac3662
RNA-Seq
190M reads,
14Gb
ABySS assembly
≥ 300bp
63,464 contigs
Alignment against B73 reference genome v2
(≥90% identity, ≥50% coverage)
59,690 contigs aligned with B73 reference genome
3,774 contigs NOT aligned with B73 reference genome

27. Number of maize lines where the gene is present
Extensive validation (sequence capture & CGH) identified 72 expressed teosinte genes that are absent from all tested (N=92) maize genomes
92 diverse maize lines
Hybridization
Synthesis
2,836 potential PAVs
Number of genes
Logratio ≥1
Genotyping CGH array
Number of maize lines where the gene is present

28. Map locations of T+ M- PAVs
26 isolated genes
11 clusters of 2 genes or more

29. Presence rate of 72 T+ M- PAVs in 91 teosinte accessions
Genotyping teosinte diversity panel via PCR
% of teosinte accessions
Questionable
Low presence rate (≤50%)
High presence rate (≥75%)

30. Random model – low frequency PAVs
16/72 (22%) of T+M- PAVs are present in <50% of tested teosintes. Low frequency T+M- PAVs could be lost via random processes, such as drift.
Genetic bottleneck
Drift or other random mechanisms
Maize
Teosinte

31. Direct Selection Against T+M- Genes
Selection AGAINST a PAV; multiple haplotypes in maize possible (depending on LD in teosinte)
Maize
Teosinte
Genetic bottleneck
Direct
49/72 (68%) T+M- PAVs are present in more than 75% of the tested teosintes
Selection (direct or indirect) can explain the loss of high frequency T+M- genes

32. Indirect Selection Against T+M- Genes
Selective sweep
Maize
Selection FOR a domestication allele in LD and coupling with the absence of a PAV indirectly selects against the PAV
Teosinte
Indirect
27/49 high-frequency PAVs (55%) co-localize with low diversity regions
49 T+M- PAVs (68%) are present in more than 75% of the tested teosintes
Selection (direct or indirect) can explain the loss of high frequency T+M- genes

33. Summary (Part I)
Maize haplotypes exhibit extensive SV (CNV and PAVs) that affects several hundred genes (supported by CGH, PCR, and re-sequencing results: both WGS and exome capture)
SV provides a testable hypothesis for heterosis (potentially making heterosis more predictive)
SV may help explain extraordinary level of phenotypic diversity in maize. CNVs and PAVs that are not in LD with SNPs could contribute to some of “missing heritability” in GWAS experiments.
“Recurrent de novo CNVs” can arise via meiotic segregation (SNH Model), yielding non-parental gene complements that have phenotypic consequences (transgressive segregation?)
Genetic variation arising from SNH model would NOT be detected in typical genome scans

34. Summary (Part II)
It is widely accepted that allelic diversity is reduced by domestication. We now know that not only alleles but entire genes can be lost during domestication
~2,000 expressed genes present in teosinte are missing from the B73 genome. 72 of these genes are missing from all other tested maize lines.
Teosinte genes failed to pass through the domestication bottleneck for a variety of reasons (selection for or against haplotypes and random processes).
Teosinte genes that were lost inadvertently during domestication, may have potential in crop improvement (e.g., biotic and abiotic stress resistance).

35. Outline CNV and Presence-Absence Variation (PAVs)
The origin of “recurrent de novo CNV”
Revisit the domestication bottleneck in light of SV
Relevance to cassava?

36. What About Cassava? Interesting questions:
How common are PAVs among cassava CVs? (Steve Rounsley)
Do wild relatives of cassava contain genes that are absent from breeding germplasm? (Wenquan Wang)
Do these missing genes confer agronomically relevant traits? (e.g., resistance to biotic or abiotic stresses) (implications for positional cloning experiments)
Does complementation of key PAVs contribute to heterosis in cassava? (Ismail Rabbi)

37. The Maize Genome Sequencing Project
Collaborators
Srinivas Aluru
Dan Nettleton
Nathan Springer
Brad Barbazuk
The Maize Genome Sequencing Project
Jeffrey Jeddeloh
Mike Scanlon (PI, Cornell)
Jianming Yu, M. Timmermans,
G. Muehlbauer,
D. Jannick-Buckner
Jinsheng Lai 38

38. CHINA AGRICULTURAL UNIVERSITY
Sanzhen Liu Camile
刘三震 Rustenholz
Kai Ying Yan Fu Wei Wu
应开 傅延 吳薇
CHINA AGRICULTURAL UNIVERSITY 39 39