An Overview of Arthropod Genomics, Mitogenomics, and the Evolutionary Origins of the Arthropod Proteome

Abstract

Arthropods represent the largest majority of animal biodiversity and include organisms of economic interest and key model species. It is thus unsurprising that the genome of an arthropod, the fruit fly Drosophila melanogaster, was among the very first to be sequenced (Adams et al. 2000) and that to date, about 21 Drosophila genomes as well as a variety of other arthropod genomes have been sequenced. Despite this promising start, current sampling is biased towards economically relevant species, and a suitable close outgroup to the arthropods, which is necessary to polarise genomic studies, is still missing. Among the suitable outgroups to the Arthropoda, the Nematoda represent one of the largest components of the extant animal biomass, and their economic importance is comparable to that of the more biodiverse arthropods. As with the Arthropoda, the importance of the nematodes is reflected in the fact that the very first animal genome to be sequenced was that of the nematode Caenorhabditis elegans (The C. elegans genome consortium 1998). Despite the nematodes being phylogenetically close to the arthropods (Aguinaldo et al. 1997; Copley et al. 2004; Dopazo and Dopazo 2005; Philippe et al. 2005; Irimia et al. 2007; Roy and Irimia 2008; Dunn et al. 2008; Belinky et al. 2010; Hejnol et al. 2009; Holton and Pisani 2010), this group is composed of highly derived species, both genetically and morphologically. Accordingly, their genomes are unlikely to be of great utility in understanding arthropod genome evolution. Some genomic data (mostly in the form of transcriptomes) are now available for other smaller ecdysozoan phyla, and some genomes (Priapulida and Tardigrada) are on the horizon. Nonetheless, enough genomic information is now available for the Arthropoda (Table 3.1) to justify an investigation into the evolution of their genome. Such an analysis, however, is intimately dependent on the availability of a robust phylogenetic background, and to a lesser extent, robust divergence times for the nodes in the background phylogeny.

Table 3.1 The most important of the available Arthropod genomes

Appendix: Methods for the Analyses Presented in this Chapter

3.1.1 A. Generation of the Onychophoran Transcriptome

Total RNA was extracted from three individuals of “Peripatoides novaezealandiae complex” (Trewick 1998), which were commercially purchased, using TriZol©. A transcriptome-wide cDNA library was generated and sequenced using two IlluminaHiseqII lanes at TrinSeq (Trinity College Dublin, Institute of Molecular Medicine, Genome Sequencing Laboratory) to an estimated coverage of <100, using 100-bp paired end reads. Row data were inspected for its quality and assembled using Abyss (Simpson et al. 2009) with k-mer of 45. This resulted in ~27,000 assembled transcripts (with lengths variable between ~70 and 1,750 base pairs). Approximately 17,000 of these transcripts had a significant blast hit against an annotated gene, while ~5,000 hit a known gene of unknown function. This set of ~22,000 genes was used to investigate the origin of orphan genes in Arthropoda. However, the 5,000 non-annotated genes were not considered for the Blast2go analysis (see below).

3.1.2 B. Mitogenomic Compositional Analyses

We downloaded a set of mitochondrial genomes of 90 arthropods in order to represent the whole phylum as homogenously as possible. Coding genes were extracted and processed with DAMBE (Xia and Xie 2001) to obtain composition for each codon position.

3.1.3 C. Phylogenetic Analyses

We investigated whether the low posterior probabilities observed for some nodes by Campbell et al. (2011) were caused by the presence of unstable taxa. We estimated leaf stability indices (Thorley and Wilkinson 1999) using P4 (Foster 2004) and performed Bayesian bootstrap analysis under CAT+G—the same model used by Campbell et al. (2011)—using the entire data set of Campbell et al. (2011). To perform the Bayesian bootstrap analyses, 100 bootstrapped data sets were generated starting from the alignment of Campbell et al. (2011). For each bootstrapped data set, a Bayesian analysis (2 independent runs) was performed under CAT+G (using Phylobayes; Lartillot et al. 2009). Results from each Bayesian analysis were summarised to generate a Bayesian majority rule consensus tree, and the resulting 100 trees were then summarised to generate a bootstrap majority rule consensus (results in Fig. 3.3).

3.1.3.1 Identification of Novel Gene Families

We downloaded the entire proteomes for the taxa in Fig. 3.4 and used MCL (Enright et al. 2002) to define protein families. A Perl script written by LC was used to partition these gene families with reference to their taxon coverage. This allowed the identification of protein families that are exclusive and universally distributed within each one of the clades in Fig. 3.4. These protein families must have been present in the clade’s last common ancestor (LCA) and must have been gained along the stem lineage of the considered clade. Because different genomes have different numbers of protein coding genes, the absolute numbers of newly acquired protein coding families for each internode can be misleading. We thus normalised numbers of orphan families by dividing these numbers by the total number of protein coding genes in the set of considered genomes (sum of the values in bold at the tips of Fig. 3.4). The normalised orphan counts (N-orph) can be interpreted as the fraction of some, abstract, pan-metazoan genome that was acquired at each internode of Fig. 3.4. Finally, we calculated rates of new orphan acquisition per million of years, dividing the N-orph values by the length of the internode along which the N-orph was acquired. As above, this allows the amount of orphan families gained each million year, along each internode in Fig. 3.4, to be expressed as proportions of a reference (abstract) “pan-metazoan” genome. The estimates of divergence times of Erwin et al. (2011) were used to calculate branch durations in million of years. For each internal node in our phylogeny, we also estimated (using squared parsimony—as implemented in Mesquite—http://mesquiteproject.org) the expected size of the genome of the corresponding LCA. This was done to allow evaluation of what proportion of each LCA genome was gained via new orphan family acquisition, along the corresponding stem lineage. Because squared parsimony is unlikely to be a particularly robust estimator of ancestral size, we suggest these numbers should be considered with caution, and only to represent a rough approximation of the true LCA-genomes dimensions.

Once the orphan gene families were identified for every internode of Fig. 3.4, BLAST2Go (www.blast2go.com) was used to obtain functional information for each of these families. For each protein family, the BLAST2Go analysis was performed for one protein family member only, and we assumed, by homology implication that all the other proteins in the same orphan family had the same (or similar) function.