MECHANISMS OF GROUP II INTRON LOSS FROM THE MITOCHONDRIAL <i>COX2</i> GENE OF PLANTS

Eukaryotic evolution has been characterized by intron loss. In eukaryotes, the proposed mechanisms of intron loss are exonization, random genomic deletion, retroprocessing and most recently, horizontal gene transfer followed by gene conversion (HGT-GC). Here we investigate the mechanism of intron loss from the mitochondrial cox2 gene, which in previous studies has shown a variable intron distribution in plants. Through an extensive PCR survey that included 107 tracheophyte species, gene sequencing, and phylogenetic analyses, we determined that the variable intron distribution is better explained by intron loss rather than intron gain. We found no evidence supporting intron loss via exonization and random genomic deletion, while limited data was consistent with retroprocessing. Interestingly, the cox2 exon phylogeny did not fully recover organismal relationships, an indication of the effects of horizontal gene transfer. Four magnoliid intronless paralogs showed strong phylogenetic conflicts, and several members of other clades (Acorus, Ruscus, rosids, and asterids) were also found in unexpected positions, although with weaker support. Further research is needed to determine if these intronless paralogs arose via HGT-GC and to determine the role that HGT-GC has played in intron loss dynamics. 54 INTRODUCTION The organellar genomes of land plants are often interrupted by group I and II introns. Group II introns are particularly abundant in the mitochondrial genomes of angiosperms (Bonen 2008). In angiosperms, the evolutionary history of most mitochondrial group II introns is quite stable. However, the group II introns located in the cox2 gene are unusual in this regard since they have been repeatedly and independently lost across different lineages over time (Joly, Brouillet, Bruneau 2001; Kudla et al. 2002; Hepburn, Schmidt, Mower 2012). Thus, they serve as a great model for understanding the dynamics of intron loss. In plants, there are two described models for the biological mechanism of intron loss: random chromosomal deletion and retroprocessing (reviewed in detail in Chapters 1 and 2). Nevertheless, in 2012 we proposed a novel mechanism of intron loss, which involved horizontal gene transfer (HGT) and gene conversion (GC) (Hepburn, Schmidt, Mower 2012). This mechanism involves the horizontal transfer of an intronless gene that undergoes GC with the native intron containing copy, giving rise to a chimeric intronless gene. The latest model is plausible because HGT is a well documented and fairly common phenomenon in plant mitochondria (Nickrent et al. 1998; Bergthorsson et al. 2003; Won, Renner 2003; Bergthorsson et al. 2004; Davis, Wurdack 2004; Mower et al. 2004; Woloszynska et al. 2004; Davis, Anderson, Wurdack 2005; Richardson, Palmer 2007; Keeling, Palmer 2008; Hao et al. 2010; Mower et al. 2010). Furthermore, reports of horizontally acquired genes undergoing gene conversion with the native genes have emerged (Bergthorsson et al. 2003; Barkman et al. 2007; Richardson, Palmer 2007; Hao, Palmer 2009; Hao et al. 2010; Mower et al. 2010; Hepburn, Schmidt, Mower 2012). Perhaps the delay in reported cases is due, in part, because the combined effects of 55 HGT and GC can be difficult to detect, especially if the transferred or converted segment is small. In this chapter, I describe our efforts to identify which of the proposed mechanisms of intron loss is responsible for the widespread loss of introns from the cox2 gene. Through statistical and phylogenic work, it becomes apparent that there are several mechanisms responsible for the observed intron distribution in the cox2 gene of plants. We find some clear cases of retroprocessing, but we also find more evidence to support our own proposed HGT-GC mechanism. In contrast, we find no evidence for genomic deletion or for exonization. Materials and Methods Plant Materials Fresh leaves from a total of 107 samples representing three major vascular plant groups (monilophytes, gymnosperms and angiosperms) were obtained from the greenhouse at the Beadle Center and from the Earl G. Maxwell Arboretum at the University of Nebraska-Lincoln. Voucher information for most selected plants is provided in Table S1. Nucleic acid extraction, gene amplification & sequencing Total DNA and RNA samples were isolated from fresh leaf tissue as described by Hepburn et al. (2012). Several primer pair combinations were used to amplify the cox2 gene for these diverse plants. Most angiosperm species were amplified using degenerate cox2 primers labeled as UNI (Table S2). When these primer combinations failed, “EMBRYO” primers were used to amplify the gene. Gymnosperms and monilophytes were amplified using a combination of specific primers (GYMNO or PTERI respectively) in conjunction with degenerate primers (Table S2). To completely 56 sequence both introns for all species, primer walking was performed using primers from Table S3 and Table S4. All PCR reactions were initially carried out in a 60 ul reaction mixture using primer pairs from Table S2, GoTaq Flexi DNA Polymerase, and supplied reagents (Promega, WI, USA) following the manufacturer’s instructions. Each reaction was amplified on an Engine Dyad Peltier Thermal Cycler-0220 or in a C1000 Thermal Cycler (Bio-Rad, CA, USA). The PCR program parameters for all described assays included a predenaturation step of 94oC for 3 min, followed by 35 cycles of denaturation at 94oC for 30 s, annealing at 50oC for 45 s, and elongation at 72oC for 2 min, with a final step of 72oC for 5 min. All amplicons were purified and sequenced on both strands at the High-Throughput Genomics Unit (University of Washington, Seattle, USA). All newly generated sequences will be deposited in GenBank. cDNA synthesis, PCR amplification & RNA editing analysis cDNAs were synthesized and amplified as previously described (Hepburn, Schmidt, Mower 2012). Sites of RNA editing were determined experimentally for 46 species by comparing the DNA and cDNA sequences. For the rest of the species, edit sites were predicted using PREP-Aln (Mower 2009) with the detection cutoff value set to 0.2. Taxon sampling, alignments & phylogenetic analyses The mitochondrial cox2 gene was amplified from 107 species, while an additional 61 cox2 DNA sequences were retrieved from GenBank. The compiled dataset included 168 species representing 10 major groups within the tracheophytes (Table S1). 57 Sequences were assembled in CodonCode Aligner version 3.5 (CodonCode Corp., Dedham, MA) and sequence alignments were performed using MUSCLE inside the program MEGA version 5.0 (Edgar 2004; Tamura et al. 2011). Sampling was aimed to be as comprehensive as possible, although partial coding sequences (< 500 bp) were excluded from the alignments. Sequence alignments were manually refined using BioEdit version 5.0.6 (Hall 1999). Poor-quality alignment regions were removed by GBlocks version 0.91b (Talavera, Castresana 2007) and the parameters were set to be less stringent (by selecting the three options of the online version). All phylogenetic analyses were performed with phyML online version 3.0 (Guindon et al. 2010) using the Maximum Likelihood (ML) algorithm. The cox2 gene tree was built exclusively from cDNA sequences. In all the phylogenetic analyses the trees were rooted on Lycophytes. The general time-reversible (GTR) model was used as the substitution model. Base frequencies were set to empirical, while the proportion of invariable sites and the shape of the gamma distribution with four substitution rate categories (GTR+G+I) were estimated during the run. Tree improvement was performed by subtree pruning and regrafting with ten random starting trees. Tree support was evaluated by bootstrapping from 1000 replicates. Testing for character correlations & stochastic character mapping To determine if there were any character correlations between the number of introns and edit sites, we performed stochastic mapping simulations using SIMMAP Version 1.5 (Bollback 2006). SIMMAP is a Bayesian approach based on stochastic models that uses mutational mapping that is consistent with the distribution across the tips of the topology to estimate the posterior probability distribution from several ancestral states reconstructions, also known as stochastic realizations (Ronquist 2004; Bollback 2006).