Is GC bias in the nuclear genome of the carnivorous plant Utricularia driven by ROS-based mutation and biased gene conversion?

Abstract

At less than 90 Mbp, the tiny nuclear genome of the carnivorous bladderwort plant Utricularia is an attractive model system for studying molecular evolutionary processes leading to genome miniaturization. Recently, we reported that expression of genes encoding DNA repair and reactive oxygen species (ROS) detoxification enzymes is highest in Utricularia traps, and we argued that ROS mutagenic action correlates with the high nucleotide substitution rates observed in the Utricularia plastid, mitochondrial, and nuclear genomes. Here, we extend our analysis of 100 nuclear genes from Utricularia and related asterid eudicots to examine nucleotide substitution biases and their potential correlation with ROS-induced DNA lesions. We discovered an unusual bias toward GC nucleotides, most prominently in transition substitutions at the third position of codons, which are presumably silent with respect to adaptation. Given the general tendency of biased gene conversion to drive GC bias, and of ROS to induce double strand breaks requiring recombinational repair, we propose that some of the unusual features of the bladderwort and its genome may be more reflective of these nonadaptive processes than of natural selection.

Introduction

For centuries, carnivorous plants have held great scientific and public fascination. Darwin himself devoted an entire book to them in 1875.¹ The evolution of carnivorous plants has been modeled as a compromise between photosynthetic costs and the benefits of obtaining organic nitrogen (N) and phorphorus (P) in nutrient-poor habitats.^2,3 The carnivorous plant Utricularia, otherwise known as the bladderwort, possesses tiny traps of incredible complexity and belongs to the family Lentibulariaceae (order Lamiales), the most species-rich carnivorous plant group.⁴ Species of Utricularia and its sister genus Genlisea are characterized by unusually increased rates of nucleotide substitution as well as dynamic evolution of genome size, which varies from approximately 60–1,500 megabases.⁵ Bladderwort suction traps, which build up a strong, negative internal pressure via active ion transport, respond to tactile stimulus by equalizing that pressure in an extremely rapid influx that traps prey. This active transport process requires substantial energy, and correspondingly, in vivo experiments have demonstrated that respiratory rates are far greater in bladders than other plant parts.⁶ This remarkably greater respiration rate could increase intracellular concentrations of reactive oxygen species (ROS) through action of a specialized cytochrome c oxidase that may decouple proton pumping from electron transport in a reversible way.^7,8 In consequence, it has been postulated that ROS are a driving force behind the increased nucleotide substitution rates and dynamic genome evolution found in Utricularia.⁷

In recent work, we provided the first broad survey of nuclear gene transcripts in Utricularia.⁹ We particularly focused on the expression of genes involved in N and P uptake, hydrolase-related genes expressed during prey digestion, as well as genes involved in respiration and ROS production and scavenging. Our data suggested, for example, that whereas nitrogen absorption could in part take place in vegetative parts, phosphate uptake might occur mainly in traps. As additional support for their unusual respiratory physiology, global gene expression analysis showed that traps significantly overexpress genes involved in respiration. Furthermore, observed expression patterns of DNA repair and ROS detoxification enzymes were argued to correlate with response to increased respiration. Finally, direct measurement of ROS in situ and interspecific comparisons of organelle genomes and multiple nuclear genes provided further support for the hypothesis that increased nucleotide substitution rates in Utricularia may be due to amplified ROS-based mutagenesis. Here, we continue this analysis by providing information on DNA AT/GC content that is suggestive of mutational processes incurred by ROS, which include both nucleotide lesions and double strand breaks.

Nucleotide composition and its variation in Utricularia coding regions

In our previous work, molecular evolutionary rates in the Utricularia nuclear genome were assayed across a random set of 100 genes homologous to Conserved Orthologous Loci (COS II) available for several other asterid species (aligned sequences are available via FTP:www.langebio.cinvestav.mx/utricularia/). We used these data to estimate the GC content in the coding regions from these species. As expected, we found similar GC content between tomato (Solanum lycopersicum), potato (Solanum tuberosum) and pepper (Capsicum annuum,) three phylogenetically closely related species in the plant family Solanaceae. Although the GC content of coffee (Coffea canephora) is somewhat greater than Solanaceae, Utricularia gibba coding regions show considerably higher GC content, on average 4.5% more than other asterid plants (Fig. 1). To place these values in a greater context, we compared both coding and whole-genome GC contents in Utricularia, Solanaceae, and coffee with values from a number of angiosperms with completely sequenced genomes, including the asterid Mimulus, eight other eudicots, and four grasses (Fig. 2). Arranged in ascending order based on whole-genome GC content, two distinctive groups of species emerge: Utricularia plus the grasses, and all other eudicots. These groups are not at all correlated with genome size. Grass genomes are remarkable for having strong GC content heterogeneity that parallels the situation seen in mammals, i.e., richness of genes with short introns in GC-rich regions.¹³ The values reported for grasses in Figure 2 are merely averages throughout the genomes; indeed, strong bimodality is observed for coding region GC content in these species¹³ (results from CoGe not shown). The data for Utricularia are currently too fragmentary to assess such bimodality, but the GC content similarity to grasses may suggest common underlying molecular evolutionary processes distinct from those of “standard” eudicots.

Figure 1.

Average GC content estimated from a random set of 100 genes and their conserved orthologous Loci available for some asterid plant species [Coffea canephora (Coffee), Utricularia gibba (blanderwort), Capsicum annuum (pepper), Solanum tuberosum (potato) and Solanum lycopersicum (tomato)]. The maximum likelihood phylogenetic tree, left, was derived using the HyPhy program (http://www.datam0nk3y.org/hyphy/doku.php) and the GTR model with local variation.

Figure 2.

Genome size, coding and whole-genome GC content comparisons between Utricularia and other angiosperm species. All values are sorted by whole-genome GC. Two groups of angiosperms become clear by linear regression: Utricularia (in yellow) plus grasses, vs. all other eudicots shown. Whole-genome GC in Utricularia and grasses is more similar to coding GC in the other species. Coding GC contents from pepper, tomato, potato, coffee and bladderwort were assayed across the sequences used for reconstructing the tree in Figure 1.⁹ Genome size and whole-genome GC contents for these species, with the exception of Utricularia, were obtained from the literature.^10-12 GC content in the Utricularia nuclear genome was estimated from a preliminary draft assembly (~15x, Ibarra-Laclette et al., unpublished data). Additional data were obtained from CoGe through the OrganismView web tool (genomevolution.org/CoGe/OrganismView.pl).

In grasses, evidence suggests that the nonadaptive mechanism of biased gene conversion plays a major role in driving GC enrichment.¹³ Indeed, nonreciprocal exchanges during recombination appear to generally bias in favor of GC. For example, several studies have shown that ostensibly neutral A/T→G/C mutations can reach higher frequencies than those involving G/C→A/T, and correspondingly, others show that local recombination rates correlate with GC richness at silent sites (reviewed in ref. 14). In order to obtain a detailed profile of substitution biases for coding regions of bladderwort and its closest relative in the comparison, Coffea canephora, we used a concatenated super-matrix comprising all gene sequences for all species (the alignment contained five taxa and 87,713 nucleotide characters) to produce a maximum likelihood estimate with reference to their common ancestral node in the phylogeny. In particular, we searched for signatures related to recombinational GC enrichment. A general trend in organisms is that coding exons tend to be GC-enriched relative to introns or intergenic regions, and it has been argued that this may be directly due to particularly effective biased gene conversion in regions of adaptively constrained sequence similarity, where homologous recombination will be favored.^14,15 In addition to playing this direct role, recombination may indirectly reinforce exonic GC richness since G/C→A/T transitions due to cytosine deamination are less favored in GC-rich regions, which are more physically stable.^15,16 Although both coffee and bladderwort show a bias toward GC substitutions, that of Utricularia is much more pronounced, especially among A/T→G/C transitions (Table 1). Further analysis by codon position (Table 2) shows that in Utricularia, this transition bias is by far strongest at third positions of codons, which are silent for all amino acids except methionine and tryptophan. As such, a reasonable hypothesis is that recombinational forces have both directly and indirectly driven GC content increase in Utricularia, most particularly for A/T→G/C transitions at silent sites.

Table 1. Nucleotide substitutional biases detected in Utricularia gibba (ugb) and Coffea canephora (ccn) coding regions, as compared to a maximum-likelihood reconstructed ancestral sequence.

		ancestor*|Coffea	ancestor*|Utricularia
Number of nucleotides compared		67,193	64,756
Transitions	G:C→ A:T	2,282	2,788
	A:T→ G:C	2,561	4,603
Transversions	G:C→ T:A	720	1,417
	A:T→ C:G	1,116	1,936
	A:T→ T:A	1,309	1,908
	G:C→ C:G	836	1,398
GC bias rateδ		1.22	1.56
Total transitions		4,843	7,391
Total transversions		3,981	6,659
Transition/transversion bias		1.22	1.11
*The common ancestor for coffee and Utricularia was reconstructed using maximum likelihood phylogeny estimation as in Figure 1. δGC bias rate was obtained from the equation [(A:T→ G:C)+( A:T→ C:G)]/[( G:C→ A:T)+( G:C→ T:A)]. Table 2. Substitutions by codon position and their bias in Utricularia gibba coding regions.

		ancestor|Utricularia
		1st position Substitutions (%)	2nd position Substitutions (%)	3rd position Substitutions (%)
Transitions	G→ A	424 (3.02)	200 (1.42)	751 (5.35)
	C→ T	261 (1.86)	257 (1.83)	895 (6.37)
	A→ G	477 (3.40)	356 (2.53)	1,273 (9.06)
	T→ C	385 (2.74)	274 (1.95)	1,838 (13.08)
Transversions	G→ T	210 (1.49)	51 (0.36)	330 (2.35)
	C→ A	256 (1.82)	218 (1.55)	352 (2.51)
	A→ C	307 (2.19)	233 (1.66)	511 (3.64)
	T→ G	167 (1.19)	46 (0.33)	672 (4.78)
	A→ T	208 (1.48)	146 (1.04)	610 (4.34)
	T→ A	167 (1.19)	88 (0.63)	689 (4.90)
	G→ C	224 (1.59)	166 (1.18)	295 (2.10)
	C→ G	163 (1.16)	195 (1.39)	355 (2.53)
Totals		3,249 (23.12)	2,230 (15.87)	8,571 (61.00)

Biased gene conversion depends on heterozygosity for its operation. It has been demonstrated that selfing grass species, which tend to be more homozygous than outcrossers and therefore less efficient in recombination, have lower equilibrium GC contents.¹³ Some Utricularia species appear to be vegetatively reproducing, or at least have substantial clonal periods in their life cycles.^17-19 At first glance such life histories could suggest that heterozygosity and recombinational efficiency would be low in bladderworts. However, there are reasons to expect otherwise. First, the mutagenic action of ROS in somatic cells, and the propagation of these cells into new plants, could lead to substantial polymorphism within clonal populations, independent of sexual reproduction. Kameyama and Ohara¹⁸ observed such polymorphism among populations of an Utricularia species, but these authors suggested that sexual reproduction may be the dominant diversity generator; the data used were extremely limited AFLP markers, and the generality of the findings remains unclear. For example, genotypic diversity in another clonally reproducing aquatic plant, Butomus, appears to derive from somatic mutation.²⁰ Second, enhanced ROS production should increase the frequency of double strand breaks, which could then undergo recombinational repair and biased gene conversion when polymorphism exists. In considering the role that gene conversion might play in increasing GC bias, it is necessary to consider both the strength of the recombinational process and the permissivity of the population genetic environment in which such molecular evolution occurs.¹⁴ As just argued, recombination would be expected to be more frequent when ROS production is increased. As for population-level considerations, an otherwise low effective population size for a clonal species—best understood in terms of numbers of alleles in such a population—would be enormously increased by heritable somatic mutation. Just as natural selection has greater influence on the nucleotide composition landscape than random genetic drift in such large populations, so too would biased gene conversion.¹⁴ The fascinating difference is that biased gene conversion is a neutral process, whereby genome architectural change can occur without the action of natural selection. Again, it has recently been argued that GC biases in grasses may be largely explained by different recombinational profiles and the corresponding effects of biased gene conversion. Perhaps the Utricularia genome will also show architectural evidence for biased gene conversion, for example, shortening of introns through a neutral, selection-like process. Indeed, it is conceivable that most or all of the genome downsizing observed in Utricularia (Fig. 2) occurred under the influence of the neutral force of gene conversion as opposed to natural selection.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.