Finding and Comparing Syntenic Regions among Arabidopsis and the Outgroups Papaya, Poplar, and Grape: CoGe with Rosids^1,[W]

Rosid Genome Facts and Crucial Definitions

Comparing Arabidopsis, poplar, papaya, and grape genomes is particularly useful because of their phylogenetic relationships and placement of paleotetraploidies relative to speciation events. The phylogenetic tree of Figure 1 is from the Missouri Botanical Garden trees Web site (http://www.mobot.org/MOBOT/research/APweb/welcome.html in March 2008; Rodman et al., 1993; Soltis et al., 1999, 2003; Hall et al., 2002; Davis et al., 2005; Jansen et al., 2007; Zhu et al., 2007) and has been modified to highlight whole genome duplication (WGD) events and their subsequent fractionation back toward the gene content of the diploid ancestor. The WGD events identified in Figure 1 are referenced individually as follows. The Arabidopsis lineage has undergone two sequential WGD events (Bowers et al., 2003) since its divergence from the papaya lineage deep in the order Brassicales. Each Arabidopsis genome duplication event is followed by extensive fractionation (see definitions) that reduced the genome's gene content back toward that of the pretetraploid ancestor. Meanwhile, papaya's lineage has had no such WGD events (Ming et al., 2008).

Whereas both Arabidopsis and papaya are in the fabids (eurosid I) clade, poplar represents the sister rosid lineage (malvids/eurosid II) whose lineage underwent one minimally fractionated WGD event (Tuskan et al., 2006). Grape (Jaillon et al., 2007; Valesco et al., 2007) is just basal to the rosids or a basal rosid (Soltis et al., 1999, 2003; Jansen et al., 2007; Zhu et al., 2007), and, like papaya, has not had any WGD events in its own lineage since the paleohexaploidy (Jaillon et al., 2007) or gamma polyploidy (Bowers et al., 2003) shared by grape and all rosids. Arabidopsis now has an excellent outgroup in papaya, and all Brassicales have a useful outgroup in grape. Poplar provides an additional outgroup for Brassicales research provided that its own lineage-specific WGD event is reverse fractionated to infer the genomic structure of its preduplicated ancestor. Comparison across multiple outgroups is a powerful tool that, for example, permits unambiguous discrimination between gene gain in Arabidopsis versus gene loss in any one outgroup.

Many eukaryotic gene lineages are characterized by gene family expansions over evolutionary time (Lespinet et al., 2002; Koonin, 2005). Several gene families of Arabidopsis have been shown to have expanded by (1) tandem duplication (Rizzon et al., 2006); (2) retention and/or the avoidance of fractionation following WGD events (see citations below); and (3) transposition (Freeling et al., 2008). Each of these modes of gene family expansion creates paralogs that potentially duplicate the function of the original gene. If retained, this functional duplication sets the stage for biased gene expansion (Blanc and Wolfe, 2004; Seoighe and Gehring, 2004; Maere et al., 2005; Thomas et al., 2006) and subsequent subfunctionalization (Force et al., 1999). Selection against dosage effects (haploinsufficiencies) that might occur due to particular fractionations of a tetraploid (Birchler et al., 2005) is probably a major principle in explaining gene content trends in gene expansion (Freeling, 2008), and probably in increasing morphological complexity as well (Freeling and Thomas, 2006).

Some words have complex or multiple meanings. We use the following definitions and provide examples in the CoGe with Rosids online tutorial and later in this article.

Collinearity

Collinearity is a fact (induction) indicating that a chromosome or segment will tend to have genes in the same order and orientation as they were in an ancestor. As divergence time between any two genomic regions increases, collinearity decreases by virtue of chromosomal or single-gene aberrations, such as inversions, translocations, and deletions.

Synteny

Synteny is an inference (deduction) based on collinearity data that two or more chromosomes or segments are derived from a common ancestor. Genomic regions with collinear features are syntenous, but syntenous segments can have zero shared features (e.g. fully fractionated gene content). If noncollinear gene orders can be explained by chromosomal aberrations, then synteny may often be inferred. Single-gene transpositions obfuscate synteny, as do rearrangement-prone regions (Fortna et al., 2004; Gordon et al., 2007), and can obscure attempts to reconstruct evolutionary history. Genes derived from syntenic regions can be paralogs (e.g. homologs retained from tetraploidy) or orthologs (e.g. from speciation). The term syntelog might be used to identify homologous genes in reference to ancestral chromosomal position only.

Single-Gene Transposition

Single-gene transposition refers to the insertion of one gene into a new location (Freeling et al., 2008). This chromosomal event uniquely destroys collinearity and synteny. Some synonyms: ectopic duplication (Leister, 2004) and ectopic translocation (Ameline-Torregrosa et al., 2007).

Fractionation

Fractionation is the loss mechanism by which a duplicated gene, chromosomal segment, or genome tends to return to its preduplication gene content, but not necessarily gene order (Lockton and Gaut, 2005). Fractionation of a tetraploid results in the loss of one of the initial homologs, but not both. A species is polymorphic for duplications arising by local duplication and by single-gene transposition and may be polymorphic for exact patterns of post-tetraploidy fractionation as well. There is no one diploid of a species, so the term diploidizaiton is not always an accurate synonym for fractionation.

Subfunctionalization

Subfunctionalization is the selectively neutral tendency of a duplicated cis-acting unit of function (gene) to lose dispensable sequences (functions) on one, but not both, duplicates, such that the ancestral function is spread over both duplicates (Force et al., 1999). This idea was originally used to explain the over-retention of gene pairs following duplication (Lynch and Force, 2000), but gene content change data following tandem duplication versus WGD events supports dosage effect rather than subfunctionalization as the duplicate retention mechanism (Freeling and Thomas, 2006; Freeling, 2008). In any case, postretention duplicate genes often subfunctionalize (He and Zhang, 2005; Rastogi and Liberles, 2005; Duarte et al., 2006; Ganko et al., 2007; Woolfe and Elgar, 2007).

Plant CNS

Plant CNS is a significantly conserved sequence of nonprotein-coding plant DNA associated with a pair of syntenous (orthologous or homologous/paralogous) genes or segments as evidenced by unfiltered BLASTn hits below the e-value of a 15/15 exact nucleotide match (or equivalent). True CNS identification requires divergence times long enough to ensure that identified regions of sequence similarity are caused by selection for function rather than neutral carryover. Maize (Zea mays)-rice (Oryza sativa; Guo and Moose, 2003; Inada et al., 2003) and Arabidopsis-Arabidopsis α-homologs (Haberer et al., 2004; Thomas et al., 2007) exhibit useful divergence times.

No introduction to comparing homologous DNA sequences is complete without emphasizing the importance of the relative level of sequence divergence. For example, as two homologous noncoding sequences diverge from a common ancestor, at some point in their evolution, nonfunctional regions will be removed or randomized with respect to one another. If two homologous sequences are not sufficiently diverged, then CNSs might be unselected carryover from the ancestor. Such conserved regions cannot be assumed to have function and represent false-positive noise. Detecting functional noncoding sequences based on sequence similarity can yield false positives that may be filtered by employing e-value (or other metrics) cutoff values. However, exact cutoff metrics depend on the alignment algorithm used. Previous work has identified that a BLASTn e-value cutoff for a 15/15 exact nucleotide match or a Lagan window of 21 bp at 70% identity are sufficient criteria when comparisons are under approximately 20 kb (Lyons and Freeling, 2008). When divergence is optimal, CNSs are often more conserved than protein-encoding exons because of wobble in the third codon position. As two sequences diverge beyond the optimum, CNSs become less conserved than exons, and ultimately become undetectable. The degradation of conservation in functional noncoding sequences has been explained by a process known as binding site turnover (Ludwig et al., 1998; Frith et al., 2006; Moses et al., 2006).

Sometimes divergent genomes fractionate to such a large extent that synteny is difficult to detect. For an extreme example, if a post-tetraploid lineage has complete fractionation whereby every duplicated gene is lost from one homologous genomic region or the other, then there is no ability to detect synteny by finding a collinear series of putatively homologous genes. An example of extreme fractionation exists within the genomes of both grape and papaya as evidenced by the three genomes of the ancient pre-rosid hexaploidy located in Figure 1. Two of these three genomes are fractionated so heavily that synteny is barely apparent; the third genome helps to evaluate these data (Lyons et al., 2008). The genomes of any dicot and any monocot are not collinear in any overall way (M. Freeling, unpublished data). If this is due to one or more independent heavily fractionated paleotetraploidies in their lineages, synteny could conceivably be tested if we had a genome of one of the outgroup Anthophyta taxa that predates the monocot-dicot split: Magnoliales, Winteraceae, Laurales, or Illicales. In theory, even fully fractionated tetraploidies can be reconstructed using outgroups.

CoGe contains tools for assessing synteny, divergence levels, and CNS discovery by supporting efficient acquisition, evaluation, and refinement of syntenic alignment analyses. Our platform should prove to be particularly useful for de novo discovery by allowing researchers to quickly specify genomic regions for comparative analyses, adjust alignment algorithms and settings, add or remove genomic regions, extend or contract regions being analyzed, and rerun analyses. Results are generated on the fly and displayed in an interactive system for quick interpretation and subsequent iterative refinement of the analysis.

Introducing the CoGe Online Tutorial Titled CoGe with Rosids

The CoGe with Rosids online tutorial (http://tinyurl.com/4a23pk) is a step-by-step walk-through for finding and evaluating syntenic regions among the sequenced rosid genomes and encourages researchers to follow along with their own analyses. At the top of each tutorial page are the links that comprise the titles of the various lessons: Genome Duplications, Gene Lists, GEvo, Finding Synteny, Putting Syntelogs Together, Overview and Analysis, Fractionation, Annotation Errors, Local Duplications and Insertions, CNSs/Subfunctionalization, and Inversions. Each lesson is an entrance to CoGe.

A typical question that CoGe aids in answering is starting with one Arabidopsis gene of interest and ending with an analysis of multiple syntenic chromosomes within rosid species. CoGe's workflow for answering this follows.

Method for Generating Arabidopsis-Papaya-Grape Two Poplar Gene List with GEvo Links

Arabidopsis (Arabidopsis thaliana; The Arabidopsis Information Resource [TAIR], version 7) was blasted against papaya (Carica papaya; EVM, 1.0 release) and grape (Vitis vinifera; genoscope version 1) using BLASTn using default settings and an e-value filter of 0.001. The top hit to each was taken as a potential syntelog. To generate syntelogs to poplar (Populus trichocarpa), which has had a relatively recent genome duplication event, DAGchainer (Haas et al., 2004) was used. If a single matching syntenic region was found, a single poplar gene was included (if one matched the Arabidopsis query sequence). If two matching syntenic regions were found, a gene from each region was used (if there were matches to the Arabidopsis query sequence). These sets of genes were then used to generate links to GEvo with appropriate anchors for each matching genome.

Method for Generating Four Arabidopsis-Papaya Gene List with GEvo Links

We analyzed the collinearity between the longest 200 papaya scaffolds and the five Arabidopsis chromosomes. We used BLASTP (E-value < 1e-5, top five hits) to search for similarities among complete sets of annotated papaya (EVM, 1.0 release) and Arabidopsis genes (TAIR, version 7). For simplicity, the papaya genes were renamed according to their positions on the scaffolds. We applied a multiple genome and subgenome alignment algorithm MCscan (Tang et al., 2008) to identify and display collinearity. Briefly, the program first enumerates pairwise collinearity between all pairs of papaya scaffolds and Arabidopsis chromosomes through a dynamic programming procedure very similar to DAGchainer. The default scoring scheme (configurable) is min (log₁₀ e-value, 50) match score for one gene pair, and −1 gap penalty for each 10-kb distance between any two consecutive gene pairs. All tandem matches are collapsed using one representative gene pair that has the lowest BLASTP e-value. This is followed by a statistical test used in Wang et al. (2006), and an e-value is calculated for each block. The syntenic blocks that exceed score 300 and e-value <1e-10 are retained. These pairwise collinear segments are then grouped into blocks and progressively realigned, using papaya scaffolds as reference. This creates multiple alignment views of several genomic regions that show similar gene orders. Finally, for each syntenic block, we manually removed regions that are derived from a more ancient duplication event (γ). This pipeline generates a gene list that contains many multialigned blocks that show one papaya region usually matching to up four Arabidopsis regions.

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Information S1. At-Cp-Vv-2Pt.xls.

• Supplemental Information S2. At4-Cp.xls.