De novo sequence assembly and characterization of the floral transcriptome in cross- and self-fertilizing plants

Abstract

Background: The shift from cross-fertilization to predominant self-fertilization is among the most common evolutionary transitions in the reproductive biology of flowering plants. Increased inbreeding has important consequences for floral morphology, population genetic structure and genome evolution. The transition to selfing is usually characterized by a marked reduction in flower size and the loss of traits involved in pollinator attraction and the avoidance of self-fertilization. Here, we use short-read sequencing to assemble, de novo, the floral transcriptomes of three genotypes of Eichhornia paniculata, including an outcrosser and two genotypes from independently derived selfers, and a single genotype of the sister species E. paradoxa. By sequencing mRNA from tissues sampled at various stages of flower development, our goal was to sequence and assemble the floral transcriptome and identify differential patterns of gene expression.

Results: Our 24 Mbp assembly resulted in ~27,000 contigs that averaged ~900 bp in length. All four genotypes had highly correlated gene expression, but the three E. paniculata genotypes were more correlated with one another than each was to E. paradoxa. Our analysis identified 269 genes associated with floral development, 22 of which were differentially expressed in selfing lineages relative to the outcrosser. Many of the differentially expressed genes affect floral traits commonly altered in selfing plants and these represent a set of potential candidate genes for investigating the evolution of the selfing syndrome.

Conclusions: Our study is among the first to demonstrate the use of Illumina short read sequencing for de novo transcriptome assembly in non-model species, and the first to implement this technology for comparing floral transcriptomes in outcrossing and selfing plants.

Figure 1

Histogram of the frequency of different contigs sizes in transcriptome assemblies of Eichhornia samples. Blue bars represent the distribution of contig sizes for the initial de novo Oases assemblies and the red bars represent the final consensus transcriptome after being processed in our pipeline. The distribution has shifted to the right indicating more long contigs.

Figure 2

Proportion of each of the assembled sequences that is aligned to its top BLASTx protein hit versus the proportion of the top hit which is covered by the assembled query sequence. This plot demonstrates the size of our assembled transcripts relative to their homologs in other plant species. Dense clusters of points along the top of the figure represent loci entirely aligned to their respective protein hits and points along the right are genes fully covering their best BLAST protein hit.

Figure 3

Distribution of genes in the transcriptome assembly assigned to broad GO categories. Cellular components (blue) and biological processes (red). Percentages indicate the proportion of sequences assigned within each subcategory using the program BLAST2GO, which assigns putative function to each transcript based on similarity to proteins with GO annotations from other organisms.

Figure 4

Pairwise correlations of gene expression between the four genotypes: Eichhornia paniculata - Brazil, Jamaica and Nicaragua and Eichhornia paradoxa. Above the diagonal are the correlation coefficients for data plotted below the diagonal. Each pair below the diagonal is expression level plotted on a log scale, measured in FPKM, which estimates gene expression from the number of reads that are derived from each transcript in our samples. All correlations are significant at P < 0.00001.