1. Introduction
The evolution of gender polymorphisms in plant populations is characterized by complexity on many levels (Charlesworth, Reference Charlesworth2006). For example, gender expression is evolutionarily linked to many aspects of floral evolution, including variation in flower size and overall plant life-history. In species that exhibit gender dimorphism, such as dioecy, sex-specific drivers of floral evolution are relatively easy to study in isolation. Indeed, quantitative genetic models have been applied to good effect to understand sex-specific floral evolution leading to sex differences in flower size and other secondary sex characteristics (Meagher, Reference Meagher, Geber, Dawson and Delph1999). Given that dioecy is based on very specific genomic effects, such as sex-limited gene expression and in the extreme the evolution of sex chromosome heteromorphism, it is possible that floral evolution in dioecious species is driven by genomic processes that are specific to dioecy. On the other hand, genomic effects underlying floral evolution, and in particular flower size evolution, might be independent of gender expression. In the present paper, we explore whether genomic effects that influence flower size in dioecious species are also manifested in closely related hermaphroditic species.
One of the long-standing mysteries in genomic evolution is the observation that much of the genome is composed of repetitive DNA in the form of transposable elements and other forms of repetitive sequences (Flavell et al., Reference Flavell, Jones, Lonsdale and O'Dell1983; Heslop-Harrison, Reference Heslop-Harrison2000; Kidwell & Lisch, Reference Kidwell and Lisch2001; Meagher & Vassiliadis, Reference Meagher and Vassiliadis2005). Such repetitive DNA has typically been regarded as superfluous to the function of the genome in generating phenotypes, and the introduction of repetitive sequences has often been interpreted as a deleterious effect (Charlesworth et al., Reference Charlesworth, Sniegowski and Stephan1994). More recently, our continuing investigation of the relationship between variation in repetitive DNA, manifested as variation in nuclear DNA content and flower size, suggests that repetitive DNA may play an effective role in adaptive evolution (Meagher et al., Reference Meagher, Gillies and Costich2005).
The discovery of a negative correlation between nuclear DNA content and flower size in Silene latifolia (Meagher & Costich, Reference Meagher and Costich1994) has been supported by investigation of changes in DNA content as a correlated response to selection on flower size (Meagher & Costich, Reference Meagher and Costich1996), and the effect has been observed across a range of related species in Silene sect. Elisanthe (Meagher & Costich, Reference Meagher and Costich2004). This work has intriguing implications for the relationship between DNA content variation and phenotypic evolution. We attribute the observed relationship to indirect impacts of repetitive DNA on patterns of gene expression. For example, tandem repetitive DNA is known to influence local protein-DNA interactions; and dispersed repetitive DNA, such as transposable elements, is known to affect gene expression through insertion into regulatory regions of genes. On the basis of our observations, we have proposed that repetitive DNA underlying DNA content variation is a major driver in quantitative phenotypic variation (Meagher & Costich, Reference Meagher and Costich1996; Meagher & Vassiliadis, Reference Meagher and Vassiliadis2005).
A consistent feature of species of Silene sect. Elisanthe is that they are all dioecious, and there is a well-developed model of sex chromosome evolution within this group (Ainsworth, Reference Ainsworth1999; Charlesworth, Reference Charlesworth2002; Westergaard, Reference Westergaard1958). Indeed, S. latifolia was one of the first known examples of the XX (female) and XY (male) mode of sex determination (Blackburn, 1923, and Winge, 1923, cited in Lengerova et al., Reference Lengerova, Moore, Grant and Vyskot2003). This species has also been an important object of study in the development of evolutionary models of Y chromosome evolution, in which the Y chromosome is predicted to accumulate non-coding repetitive DNA sequences (Nicolas et al., Reference Nicolas, Marais, Hykelova, Janousek, Laporte, Vyskot, Mouchiroud, Negrutiu, Charlesworth and Moneger2005).
Given the presence of sex chromosomes in dioecious Silene species, and the tendency of sex chromosomes to accumulate repetitive DNA, it seems plausible that dioecious species undergo genomic evolution in ways that differ from what one might expect in hermaphroditic species. For example, one might speculate that accumulation of transposable elements in the non-combining region of Y chromosomes could provide a reservoir that could then elevate probabilities of transposable element establishment to other parts of the genome, such as the autosomes, through lateral transmission. Thus, we query whether the observed relationship between nuclear DNA content and flower size observed in dioecious Silene is a peculiarity of sex chromosome evolution. More specifically, would a negative correlation between flower size and nuclear DNA content appear in the absence of dioecy?
The present study investigated nuclear DNA content and flower size variation in Silene sect. Siphonomorpha, a group comprised of hermaphrodite species. We addressed the following questions: What is the extent of nuclear DNA content and flower size variation among sect. Siphonomorpha species? Does the negative correlation between nuclear DNA content and flower size previously observed in dioecious species of Silene hold up for hermaphroditic species? What insights into genome evolution are suggested by the pattern of distribution of nuclear DNA content across species of sect. Siphonomorpha (hermaphroditic) in contrast to sect. Elisanthe (dioecious)?
2. Materials and methods
This study focuses on five species in Silene sect. Siphonomorpha: S. andryalifolia, S. coutinhoi, S. fernandezii, S. mellifera and S. nutans. These species are native to the Iberian peninsula, though S. nutans is more widespread (Talavera, Reference Talavera, Castroviejo, Laínz, González, Montserrat, Garmendia, Paiva and Villar1990). In May–June of 1998, populations of each species were located in Spain (Fig. 1, Table 1); floral measurements (calyx diameter, petal claw length [=calyx length] and petal limb length [=1/2 of corolla diameter]) were taken from up to three flowers from 25 plants per population or as many plants as were present (N ranged from 12 to 25; Table 1). Seeds were collected during the last week of July 1998 and stored at room temperature in plastic bags until they were planted in late July 2000 and grown under glass at the University of St Andrews (UK). It was our intention to raise plants in the greenhouse to flowering so as to obtain floral measurements under uniform conditions, but these species are long-lived perennials and flowering in the greenhouse even after 7 years of cultivation remains too sporadic to yield a useful sample size. Consequently, our analyses of flower size variation in this paper are based on the original field measurements.
Fig. 1. Map of Spain showing locations of the sampled populations of Silene sect. Siphonomorpha (circles, S. andryalifolia; diamonds, S. coutinhoi; squares, S. fernandezii; triangles, S. mellifera; inverted triangles, S. nutans). Precise coordinates for each locality are given in Table 1.
Table 1. Collection localities for Silene sect. Siphonomorpha populations sampled
Map reference numbers for each population as well as latitude and longitude (in degrees) were those used in Fig. 1. Sample sizes for field-based floral measurements (N field) and greenhouse-based DNA content assays (N gh) are also shown.
Flow cytometric assays of nuclear DNA content (Costich et al., Reference Costich, Meagher and Yurkow1991) were performed on leaf material from greenhouse-raised plants (overall number of plants per population accession in our greenhouse population are shown in Table 1). A concern with flow cytometric assays is that apparent DNA content differences might be due to cytosolic interference from secondary compounds (Price et al., Reference Price, Hodnett and Johnston2000). We have tested for such effects in Silene by including leaf samples from plants that differ in DNA content in a single preparation, thus exposing nuclei from both plants to the same suite of secondary compounds (Meagher & Costich, Reference Meagher and Costich2004). This control procedure provided no evidence of cytosolic impacts on DNA fluorescence. We have recently conducted further tests for cytosolic impacts in Silene by investigating the relationship between variation in nuclear DNA fluorescence and variation in the fluorescence of the internal standard chicken red blood cell (CRBC) and again found no evidence of cytosolic effects (Looseley & Meagher, in prep.).
Two assays of nuclear DNA content emphasize different features of the genome (Costich et al., Reference Costich, Meagher and Yurkow1991). Red fluorescence of propidium iodide (PI-DNA) measures overall DNA content, whereas red fluorescence of PI in the presence of chromomycin (PI+CA3-DNA) provides an AT-biased measure that targets repetitive DNA since repetitive motifs in plants are typically AT-biased (Wang et al., Reference Wang, Weber, Zhong and Tanksley1994). For both assays, estimates of DNA content were obtained by dividing the mean fluorescence of Silene latifolia nuclei by the mean for an internal CRBC standard and multiplying by 2·33, the DNA content (in picograms) of CRBC (Arumuganathan & Earle, Reference Arumuganathan and Earle1991). To estimate DNA content in the PI+CA3-DNA assay, the multiplier of 2·33 was adjusted by multiplying by the mean fluorescence of CRBC among samples with PI+CA3 and dividing by the mean fluorescence of CRBC among samples with only PI. In comparing the relationship between PI-DNA and PI+CA3-DNA in the present study with previously published results for species of sect. Elisanthe (Meagher & Costich, Reference Meagher and Costich2004), earlier estimates of PI+CA3-DNA were adjusted in the same manner. Because the DNA content estimates for sect. Siphonomorpha and sect. Elisanthe were obtained at different times and using different flow cytometers, we did not do a quantitative comparison of the two sets of results, but rather limit our consideration to a qualitative comparison.
All statistical analyses were conducted using SAS version 9.1.3 (SAS Institute, 2004). Population effects were considered as nested within species in analyses of variance. Correlations between nuclear DNA content and floral dimensions were based on population-level means across the 15 species by population combinations (listed in Table 1). We used population means because we could not directly measure DNA content on plants in the field and, as noted above, we were not able to measure floral traits on plants in the greenhouse. Our estimate of the correlation between these two sets of traits confounds species- and population-level effects. In principle, it would be desirable to take into account phylogenetic relationships among species in determining the correlation between nuclear DNA content and flower size using a method such as independent contrasts (Felsenstein, Reference Felsenstein2004). That was not possible in our study because the phylogenetic relationships among these species have not been included to date in modern phylogenetic work on the genus Silene (B. Oxelman, pers. comm.) Species-level correlations among floral characters within individual plants were calculated by pooling across populations. Numbers of plants included in measures of floral dimensions in field populations and estimation of DNA content in greenhouse populations are indicated in Table 1.
3. Results
There was evidence of DNA content variation both within and among species of sect. Siphonomorpha (Table 2a, Fig. 2). PI-DNA showed variation at all levels, suggesting that there is substantial potential for the evolution of DNA content variation overall. PI+CA3-DNA differed among species, but did not show strong variation among populations within species. The difference between these two measures of DNA content variation reflects variation in the repetitive DNA component of the genome. The estimated difference between these two measures showed significant variation among species but not among populations within species.
Fig. 2. Relationships between phenotype and PI-DNA or PI+CA3-DNA for calyx diameter (a, b), petal claw length [=calyx length] (c, d) and petal limb length (e, f) for Silene sect. Siphonomorpha (circles, S. andryalifolia; diamonds, S. coutinhoi; stars, S. fernandezii; triangles, S. mellifera; squares, S. nutans). DNA content is expressed in picograms, and floral dimensions are expressed in millimetres. Correlation estimates are shown in Table 3.
Table 2. Nested analysis of variance (ANOVA) results testing for differences among species populations
a MS(Population[Species]) used as the error MS.
There was significant variation in flower size both within and among species of sect. Siphonomorpha (Table 2b, Fig. 2). Interestingly, calyx diameter, which shows sexual dimorphism in dioecious Silene species, especially in S. latifolia, and which shows extensive variation within and among species of sect. Elisanthe, did not exhibit significant variation among species of sect. Siphonomorpha. On the other hand, petal claw and petal limb length varied significantly among as well as within species.
There was a strongly significant correlation between the two measures of DNA content variation (Table 3, Fig. 3a). In terms of relationship to flower size (Table 3, Fig. 2), there was a significant negative correlation between each of the two DNA content measures and petal limb, which is similar to what has been observed previously in dioecious S. latifolia. However, there was no evidence of correlation between DNA content and calyx diameter or petal claw length. There was also no significant level of correlation among the three floral dimensions as measured at the among-population level (Table 3), but there was a significant correlation among floral traits within individual plants for each of the five species (Table 4).
Fig. 3. Impact of CA3 on DNA content estimates for Silene sect. Siphonomorpha (PI+CA3-DNA <5·5 pg: circles, S. andryalifolia; diamonds, S. coutinhoi; stars, S. fernandezii; triangles, S. mellifera; squares, S. nutans) and sect. Elisanthe (PI+CA3-DNA >5·5 pg: circles, S. latifolia; diamonds, S. marizii; triangles, S. diclinis; squares, S. dioica; filled symbols, males; open symbols, females). Data for sect. Elisanthe are derived from Meagher & Costich (Reference Meagher and Costich2004).
Table 3. Correlations by population and species between nuclear DNA content and flower measurements for Silene sect. Siphonomorpha
The sample size (populations) is N=15 throughout.
Table 4. Phenotypic correlations among floral dimensions for species of Silene sect. Siphonomorpha
There is evidence that the genomes of sect. Siphonomorpha contain repetitive DNA to a similar extent to levels found in sect. Elisanthe, even though the genomes of the former are smaller (Fig. 3). This is reflected in the fact that PI+CA3-DNA estimates are lower than PI-DNA estimates across species in both of these sections.
4. Discussion
There was considerable intra- as well as interspecific nuclear DNA content variation found in sect. Siphonomorpha. This was true for both the PI-DNA and PI+CA3-DNA assays, which suggests that the AT-rich repetitive DNA fraction of the genome follows a similar distribution to that of overall DNA content. Thus, hermaphroditic species of sect. Siphonomorpha show variation that is similar in pattern to that found in dioecious species of sect. Elisanthe.
There are several different types of repetitive DNA contributing to overall variation (Meagher & Vassiliadis, Reference Meagher and Vassiliadis2005). The alternative assay used in this study, PI+CA3-DNA, focuses on AT-rich repetitive sequences, which in plants are generally characteristic of tandem repetitive sequences (Wang et al., Reference Wang, Weber, Zhong and Tanksley1994), such as satellite and microsatellite DNA. This category of DNA falls squarely into the historical definition of non-coding ‘junk’ DNA. However, our previously observed correlation between a flower size measure and PI+CA3-DNA in the dioecious S. latifolia suggests that, on the contrary, this class of repetitive DNA has phenotypic impacts. This finding is supported by the negative correlation observed in five hermaphrodite species of sect. Siphomorpha in the present study between floral limb length and overall nuclear DNA content (PI-DNA) as well as the AT-biased measure (PI+CA3-DNA). Although tandem repeats are generally not transcribed, and are therefore non-coding in that sense, the proximity of such tandem repeats to expressed genes can moderate their expression (Zuckerkandl, Reference Zuckerkandl1997). Thus, accumulation of repetitive sequences may lower expression levels in general, while reduction of repetitive sequences may have the opposite effect. In general, the process of natural selection is not tied to conventional models of allelic substitution, but rather will incorporate whatever features of the genome are capable of being transmitted to the next generation, akin to the phenomenon of ‘bricolage’ as coined by Jacob (Reference Jacob1982; Meagher & Vassiliadis, Reference Meagher and Vassiliadis2005). Indeed, the role of ancillary non-coding DNA in phenotypic evolution is an area of active investigation (Meagher & Vassiliadis, Reference Meagher and Vassiliadis2005; Pennisi, Reference Pennisi2007; Zuckerkandl, Reference Zuckerkandl2002).
Another category of repeated sequences that accumulate in the genome is transposable elements that occur as dispersed repeats (McClintock, Reference McClintock1984). These typically have sequence-specific insertion sites such that one can model their evolutionary dynamics in the genome using a population dynamics approach (Charlesworth & Charlesworth, Reference Charlesworth and Charlesworth1983). Transposable elements are considered to be a ubiquitous (or nearly so) feature of eukaryotic genomes (Kidwell & Lisch, Reference Kidwell and Lisch2001). One widespread class of transposons, the LTR-retrotransposons, has been thoroughly characterized (Kumar & Bennetzen, Reference Kumar and Bennetzen1999). LTR-retrotransposons consist of a series of genes that code for enzymes involved in their own replication and transmission within the genome, so they are capable of rapid evolutionary change within the genome of a particular lineage. This class of repetitive DNA is also likely to contribute to variation in genome size (Kidwell, Reference Kidwell2002) and may play a role in phenotypic evolution as well (Pennisi, Reference Pennisi2007). We are presently investigating methods for directly assaying LTR-retrotransposon copy number in Silene spp. (Meagher and Yahr, in prep.; Meagher and Loosely, in prep.) in order to determine their potential for contributing to floral evolution in this group.
Silene sect. Siphonomorpha and sect. Elisanthe both showed striking differences between the two DNA assays, suggesting that DNA content variation in the two is strongly driven by AT-rich repetitive DNA. Of course, this is only one of the categories of repetitive DNA likely to contribute to DNA content variation. However, we would expect on theoretical grounds to find more accumulation of certain types of repetitive sequences in dioecious sect. Elisanthe species because of the presence of a Y chromosome, which should accumulate repetitive sequences in the non-recombining region (Charlesworth & Charlesworth, Reference Charlesworth and Charlesworth2000; Charlesworth et al., Reference Charlesworth, Charlesworth and Marais2005). In the case of LTR-retrotransposons, sequences that accumulate on the Y chromosome could also spread through lateral transfer to other chromosomes and become a more widespread genomic feature. Although the overall DNA content of the hermaphroditic species was lower, there was still evidence of considerable intraspecific as well as interspecific variation, with impacts on flower size (petal limb). Even though these species do not have the Y chromosome effects, there are other aspects of their reproductive biology that could contribute to accumulation of repetitive DNA. For example, given the extent of the evolution of morphologically based outbreeding mechanisms in Silene, such as gynodioecy and dioecy (Desfeux et al., Reference Desfeux, Maurice, Henry, Lejeune and Gouyon1996), it is reasonable to assume that hermaphroditic species of Silene are self-compatible, and thus potentially selfing. The only species of sect. Siphonomorpha that has been studied with regard to its mating system is S. nutans, and this species does show evidence of being locally inbred (Van Rossum & Prentice, Reference Van Rossum and Prentice2004). Inbreeding is another process that potentially leads to accumulation of repetitive DNA (Charlesworth & Charlesworth, Reference Charlesworth and Charlesworth1995). Further investigation of the mating system in natural populations of sect. Siphonomorpha, in conjunction with investigation of the distribution of LTR-retrotransposons and other specific repetitive DNA fractions, would shed further light on the dynamics of repetitive DNA and its role in floral evolution.
In conclusion, intra- and interspecific variation in DNA content, with impacts on phenotypic evolution, as reported in Silene latifolia, are not a peculiarity of dioecy, but rather appear to be more widespread. Meagher & Costich (Reference Meagher and Costich1996) suggested that such effects of repetitive DNA could be an underlying contributor to measures of quantitative genetic variation, as opposed to the conventional model which assumes infinitesimal effects of allelic substitution across multiple coding loci (Bulmer, Reference Bulmer1980). The present results suggest that impacts of nuclear DNA content on quantitative variation in flower size is more widespread. Further investigation into the role of repetitive DNA in phenotypic evolution is warranted.
We thank Salvador Talavera for his help in locating field sites and identifying species of sect. Siphonomorpha, Laura Meagher for her help in collecting seed, Harry Hodge for his help cultivating plants in the greenhouse and conducting flow cytometry analysis, Bengt Oxelman for his advice on Silene systematics, and Daniel Barker for his advice on phylogenetically based comparative methods. We thank P. E. Gibbs, M. E. Looseley and L. R. Meagher for their comments on an earlier draft of this paper. This work was supported by US NSF grant 9726580/0096215 and UK NERC grant NER/T/S/2001/00297.
We dedicate this paper to Deborah Charlesworth, whose contributions to our understanding of the reproductive biology of plants played a critical role in the development of our thinking in this area.
