Pervasive gene content variation and copy number variation in maize and its undomesticated progenitor

Abstract

Individuals of the same species are generally thought to have very similar genomes. However, there is growing evidence that structural variation in the form of copy number variation (CNV) and presence-absence variation (PAV) can lead to variation in the genome content of individuals within a species. Array comparative genomic hybridization (CGH) was used to compare gene content and copy number variation among 19 diverse maize inbreds and 14 genotypes of the wild ancestor of maize, teosinte. We identified 479 genes exhibiting higher copy number in some genotypes (UpCNV) and 3410 genes that have either fewer copies or are missing in the genome of at least one genotype relative to B73 (DownCNV/PAV). Many of these DownCNV/PAV are examples of genes present in B73, but missing from other genotypes. Over 70% of the CNV/PAV examples are identified in multiple genotypes, and the majority of events are observed in both maize and teosinte, suggesting that these variants predate domestication and that there is not strong selection acting against them. Many of the genes affected by CNV/PAV are either maize specific (thus possible annotation artifacts) or members of large gene families, suggesting that the gene loss can be tolerated through buffering by redundant functions encoded elsewhere in the genome. While this structural variation may not result in major qualitative variation due to genetic buffering, it may significantly contribute to quantitative variation.

Figure 1.

Structural variation affects many genes. The average log₂(other/B73) is plotted for all 2767 genes on chromosome 6 (A) or for 293 genes within a 20-Mb region of chromosome 1 (B) for eight genotypes. (Blue data points) UpCNV with more copies in the other inbred line relative to B73; (red data points) genes with significantly lower signal in the other line relative to B73 and are examples of DownCNV or PAV; (red arrows) several multigene structural variants that are observed in multiple genotypes; (black arrows) the position of several single gene structural variants that are observed in multiple genotypes.

Figure 2.

Distribution and frequency of structural variation throughout the maize genome. The physical locations of the 32,487 genes are plotted along the 10 maize chromosomes. The color of each gene indicates whether structural variation was observed and the type of variation and the y-axis indicates the number of genotypes that contain the structural variant.

Figure 3.

Enrichment for rare CNV/PAV in teosinte genotypes. (A) The number of genotypes containing each was determined and the percent of events was plotted. Only 10% of structural variants are detected in one or two genotypes, while over 60% of structural variant events are detected in at least six genotypes. (B) The plot shows the allele frequency distribution for structural variant events in teosinte (black) and maize (gray). The proportion of DownCNV/PAV that are observed in one to 16 genotypes is shown. Teosinte has an excess of DownCNV/PAV observed in a single genotype relative to maize genotypes. (C) A similar plot is shown illustrating the distribution of allele frequency for UpCNV in maize (gray) and teosinte (black) genotypes.

Figure 4.

Structural variation haplotype frequencies in subpopulations of maize. Each of the genotypes was assigned to a subpopulation based on pedigree information or structure analysis. The subpopulations are nonstiff stalk (NSS, n = 4), ex-plant varietal protection varieties (exPVP, n = 6), inbred teosinte (TeoI, n = 4), wild teosinte (TeoW, n = 10), or tropical (Trop, n = 5). The frequency of the structural variant within this subpopulation was used to perform hierarchical clustering of both the structural variants and the subpopulations. The color indicates the type and frequency of each structural variant, with blue indicating DownCNV/PAV and red indicating UpCNV. The brighter colors represent higher allele frequencies.

Figure 5.

Examples of CNV for previously characterized maize genes. The CGH data are summarized for three maize genes. For each genotype the average log₂ ratio for all probes from the gene is summarized as the height of the bar, and the standard deviation for the multiple probes is represented by the error bars.