Reconsidering the generation time hypothesis based on nuclear ribosomal ITS sequence comparisons in annual and perennial angiosperms

Abstract

Background: Differences in plant annual/perennial habit are hypothesized to cause a generation time effect on divergence rates. Previous studies that compared rates of divergence for internal transcribed spacer (ITS1 and ITS2) sequences of nuclear ribosomal DNA (nrDNA) in angiosperms have reached contradictory conclusions about whether differences in generation times (or other life history features) are associated with divergence rate heterogeneity. We compared annual/perennial ITS divergence rates using published sequence data, employing sampling criteria to control for possible artifacts that might obscure any actual rate variation caused by annual/perennial differences.

Results: Relative rate tests employing ITS sequences from 16 phylogenetically-independent annual/perennial species pairs rejected rate homogeneity in only a few comparisons, with annuals more frequently exhibiting faster substitution rates. Treating branch length differences categorically (annual faster or perennial faster regardless of magnitude) with a sign test often indicated an excess of annuals with faster substitution rates. Annuals showed an approximately 1.6-fold rate acceleration in nucleotide substitution models for ITS. Relative rates of three nuclear loci and two chloroplast regions for the annual Arabidopsis thaliana compared with two closely related Arabidopsis perennials indicated that divergence was faster for the annual. In contrast, A. thaliana ITS divergence rates were sometimes faster and sometimes slower than the perennial. In simulations, divergence rate differences of at least 3.5-fold were required to reject rate constancy in > 80 % of replicates using a nucleotide substitution model observed for the combination of ITS1 and ITS2. Simulations also showed that categorical treatment of branch length differences detected rate heterogeneity > 80% of the time with a 1.5-fold or greater rate difference.

Conclusion: Although rate homogeneity was not rejected in many comparisons, in cases of significant rate heterogeneity annuals frequently exhibited faster substitution rates. Our results suggest that annual taxa may exhibit a less than 2-fold rate acceleration at ITS. Since the rate difference is small and ITS lacks statistical power to reject rate homogeneity, further studies with greater power will be required to adequately test the hypothesis that annual and perennial plants have heterogeneous substitution rates. Arabidopsis sequence data suggest that relative rate tests based on multiple loci may be able to distinguish a weak acceleration in annual plants. The failure to detect rate heterogeneity with ITS in past studies may be largely a product of low statistical power.

Figure 1

Schematic of the tree topology used when simulating DNA sequence triplets for power analyses. The program used to simulate DNA sequences (Seq-Gen) continues to add nucleotide changes until threshold divergence values have been reached. These threshold divergence values were obtained by averaging the estimated sets of branch lengths from actual ITS sequences for 16 annual/perennial/outgroup comparisons (Table 1). These averaged values are given in Phylip format in Table 2. The threshold divergence values for the perennial-like taxon (D_IA-perennial) and the outgroup-like taxon (D_R-outgroup, D_R-IA) were kept constant for each outgroup (closer and further) and each set of nucleotide substitution parameters (ITS1-like, ITS2-like and Combined-ITS-like). To model rate heterogeneity, the threshold divergence value of the annual-like taxon (D_IA-annual) for each set of replicate simulations was determined by multiplying the perennial-like taxon divergence threshold (D_IA-perennial) by1.5 to 5 in steps of 0.5.

Figure 2

The frequency of significant rate heterogeneity by maximum likelihood relative rate tests along with the frequency of rate heterogeneity indicated by categorical rate comparisons in simulated data. Each data point represents the proportion of 1000 replicate simulated ITS-like sequence triplets with one of the ingroup taxon evolving with a substitution rate parameter between 1.5 and 5 times faster than the other ingroup taxon. In addition, the graphs show the proportion of 1000 replicates which correctly indicated that the taxon with the faster substitution rate parameter exhibited a higher substitution rate using qualitative substitution rate comparisons regardless of whether the maximum likelihood relative rate test rejected rate constancy. Simulated sequences were obtained with mean divergence and nucleotide substitution parameter sets estimated from actual ITS sequences (see Table 2).

Figure 3

Histograms of estimated substitution rate differences between annual-like and perennial-like species pairs in 1000 independent replicates that were simulated under the average nucleotide substitution parameter sets estimated from actual ITS sequences (see Table 2). The value axis gives the ratio of the faster substitution rate over the slower substitution rate. The rate difference ratio was assigned a negative value when the perennial-like taxon had a faster estimated substitution rate and was positive when the annual-like taxon had a faster estimated substitution rate.