• Chromosome number and the drivers of its variation were examined in 6 main angiosperm clades . In particular, as chromosome number can be used as a proxy for effective recombination rate, we investigated whether low chromosome numbers are favoured in individuals living in unstable environments.
• We tested our hypothesis by relating chromosome numbers with other plant traits and environmental variables. To this end, we used a phylogenetic comparative approach that estimates phylogenetic inertia and adaptation in chromosome number based on an Ornstein-Uhlenbeck process.
• The phylogenetic effects in chromosome number varied among the examined clades, but were generally high. Chromosome number resulted poorly related to large scale climatic conditions, while a stronger relation with environmental categorical variables was found. Specifically, open, disturbed, drought-prone habitats selected for low chromosome numbers, while shaded, stable environments with good availability of water and nutrients selected for high chromosome numbers. In addition, low chromosome numbers were associated with small flowers clustered in flower-like inflorescence and high chromosome numbers with perennial herbs, and especially with woody plants.
• Altogether, our findings confirm our expectations and, by considering chromosome number as a proxy for effective recombination rate, we argue that environmental stability favour higher recombination rates with respect to unstable environments. In addition, by comparing the results of the different models, testing for evolvability of 2n and of x, we were able to provide insight into the ecological significance of polyploidy.
Key words : adaptation, angiosperms, chromosome number, endemics, phylogenetic inertia, phylogenetic comparative analysis, plant-climate interactions.
Introduction
An integral component of plant speciation is chromosome evolution (Stebbins, 1966; Grant, 1981). That the genetic changes associated with polyploidy and other forms of chromosome variation (e.g. dysploidy) stimulate and provide the genetic support for ecological differentiation and adaptation is largely accepted (Levin, 2002). The crucial role played by polyploidy was widely demonstrated (see, e.g. Soltis & Soltis, 2012, Cuypers & Hogeweg, 2014, Shimizu-Inashugi et al., 2016; McIntyre & Strauss, 2017), while more recently Escudero et al. (2014) showed that dysploidy can have a higher evolutionary impact than polyploidy, in the long run. Altogether, these findings suggest that chromosome number variation, as the result of the genomic mutations cited above, should be influenced by selection to habitat and life-histories of plants. Particularly, granted that the genetic recombination acts as trade-off between the opposite needs for immediate fitness and evolutionary adaptability (Grant, 1958), chromosome number should clearly play a central role in this balance (Stebbins, 1966).
In the last decades, large datasets of chromosome counts have been set-up focusing on several countries and taxonomic groups (see Peruzzi & Bedini, 2014 for a review), allowing to describe the variation of mean chromosome number at distinct geographical levels (Bedini & Peruzzi, 2015 and literature cited therein). Further, mean chromosome number was analysed at various taxonomic levels in order to detect significant differences among ranks (Bedini et al., 2012), and methods were proposed to quantify the heterogeneity of chromosome number in taxonomic and/or geographic groups of organisms (Peruzzi et al., 2014).
Despite the interest in chromosome number evolution over the last decades, in the absence of studies linking changes of chromosome number to natural selection, it is impossible to identify the adaptive function of such variation. At present, an adaptive role has been demonstrated for other genomic phenotypic traits, such as genome size (e.g. Vidic et al., 20009, Kang et al., 2014, Carta & Peruzzi, 2016). On the contrary, concerning chromosome number variation, evolutionary hypotheses have been raised by a limited set of preliminary works (see Escudero et al., 2012 and literature cited therein). It is generally admitted that chromosome number and genome size are not necessarily correlated in angiosperms and gymnosperms, while there is a highly significant positive correlation in ferns and lycophytes (Nakazato et al., 2008). Further, Grant (1981) argued that in angiosperms, species with a basic chromosome number higher than 14 should be considered palaeo-polyploid, thus linking high chromosome numbers with putative polyploidy. Unbalanced (i.e. odd) chromosome numbers often cause significant phenotypic changes and severely impact plant growth and sexual fitness (Blakeslee, 1921), with the exception of apomictic plants and those with holocentric chromosomes, for which no deleterious effect (i.e. counterselection) is expected (Stebbins, 1971, Hipp et al., 2009). It has been postulated that descendant dysploid chromosome number changes, coupled with transition to annual life form, is the main trend in angiosperm (Stebbins, 1971); however, this soon resulted in an improper generalization, like many other karyological assumptions concerning, for instance, the direction of variation in karyotype asymmetry (Siljak-Yakovlev & Peruzzi, 2012). Stebbins (1966) also hypothesized that a chromosome number reduction by dysploidy should be expected in plants occupying pioneer habitats, in order to avoid excessive segregation and recombination of genes. Similar concepts were expressed by Darlington (1958), who considered the chromosome number as one of the variables (together with chiasmata frequency) playing a role in his recombination index. Overall, both authors, together with Grant (1958), agreed with the view that low recombination rates should be favoured in individuals living in unstable environments, in order to quickly develop populations; on the contrary, environmental stability is expected to select for increased recombination rates because loss of alleles is overbalanced by those rare allelic combinations with high fitness.
The possible relation among chromosome number, plant traits, and environmental parameters was never quantitatively tested so far. Accordingly, the main aim of our paper is to use the dataset of Italian endemic vascular plants to test these relationships for 6 main angiosperms clades (Monocots, Fabids, Malvids, Caryophyllales, Lamiids, Campanulids). To this end, we used a phylogenetic comparative approach, that estimates phylogenetic inertia (i.e. resistance to adaptation) and adaptation in chromosome number based on an Ornstein-Uhlenbeck process (Hansen et al., 2008). Hence, we were able to estimate relationships between chromosome numbers, plant traits and environmental factors in an evolutionary framework. Finally, by comparing the results of the different models testing for evolvability of 2n and of x, we were able to assess the relative contribution of polyploidy to 2n and provide insight into its ecological significance.
Materials and Methods
Chromosome data
Chromosome counts for 1364 accessions of 801 vascular plants endemic to Italy have been extracted from Chrobase.it (http://bot.biologia.unipi.it/chrobase/ last accessed 12 May 2016). Chrobase.it is an online dataset of chromosome counts for the Italian vascular flora (Peruzzi & Bedini, 2014), hosting cytogenetic data for endemic and non-endemic species. For this study we only selected counts of endemic plants because they are one of the most sensitive components of a flora, being often restricted to ecologically selective habitats (Thompson, 2005), for which we are confident that the environmental variables calculated in the present study can be a good proxy for the total ecological requirements of the species. Most counts have been associated to an exact geographic locality in the database. For those chromosome counts lacking precise information (< 10%), we identified an approximate locality based on the restricted distribution range of the species (Peruzzi et al., 2015).
Mean chromosome numbers were estimated for each species, while within-species variation to be incorporated into the phylogenetic analysis was not estimated separately for each species, because within-species samples are limited to few counts (Garamsegi, 2014). Hence, we estimated the pooled variance across the species and used it, weighted by sample size, to estimate the observation variance of the individual species, as recommended by Hansen & Bartoszek (2012). Trial analyses revealed that log transformation improved model fit by over 500 log-likelihood units relative to untransformed data, thus mean chromosome numbers were log transformed prior to analysis (Hipp, 2007).
Chromosome number evolution was analysed in 6 major angiosperm clades (Monocots, Fabids, Malvids, Caryophyllales, Lamiids, Campanulids; APG IV, 2016) including a minimum of 67 and a maximum of 200 taxa per clade. The whole dataset is listed in the Supporting Information Table S2.
Phylogeny
The topology of the phylogentic tree is based on the backbone phylogeny of European flora (Durka & Michalski 2012). Subsequently, lower level nodes and species were resolved and assembled onto the backbone tree, using more than 90 phylogenetic and systematic studies (see Supporting information Table S1 and Figure S1).
Mean diploid (2n) and basic (x, see Peruzzi, 2013) chromosome numbers for each taxon were visualized on the phylogenetic tree (Fig. 1) with the plotsimmap function in the package phytools (Revell, 2012) of R (R Core Team, 2015).
Climatic, ecological and morphological data
Climatic data associated with the sampling sites were downloaded from the Worldclim database at 2.5 min scale (http://www.worldclim.org; Hijmans et al., 2005). The considered climate data were: mean annual temperature (°C), temperature seasonality (SD × 100), temperature continentality (°C), mean annual precipitation (mm) and precipitation seasonality (coefficient of variation). Means and standard deviations were estimated for each species within a buffer of 10 min over georeferenced sites using Qgis v. 2.18 (Quantum Gis, http://www.qgis.org).
Data about morphological traits and habitat characteristics, for each of the considered species, were retrieved from literature (Pignatti, 1982 and Table S2). To characterise the relevant features of vegetative morphology and reproductive strategies, we considered the growth form (annual herb, geophytes, perennial herb, and woody), the flower size (large, small, inconspicuous) and whether the flowers are clustered in flower-like inflorescence or not. Habitats were classified using categorical variables ecologically meaningful in terms of stability vs. instability of the habitats and reliable vs. unreliable resource availability. For this purpose, we classified the species into three categories according to soil moisture (dry, moist, wet), habitat light (open, semi-shaded, forest) and soil nutrient contents (oligo-, meso-, eu-trophic).
Phylogenetic comparative analysis
We investigated whether the evolution towards optima of diploid (2n) and basic (x) chromosome numbers is influenced by climatic variables (continuous predictors), habitat characteristics or plant traits (categorical predictors) within different angiosperm clades.
To this end, we used the phylogenetic comparative method implemented in the R program SLOUCH, conceived to study adaptive evolution of a trait along a phylogenetic tree (Hansen et al., 2008). With this approach, the response trait is modelled as an Ornstein–Uhlenbeck model, assuming that the trait has a tendency to evolve towards a ‘primary’ optimum θ, defined as the average optimal state that species will reach in a given environment when ancestral constraints have disappeared (Hansen, 1997).
To facilitate the interpretation of parameter estimates, the tree is scaled to 1.0 total length (from the root to the tip in the ultrametric tree). The two main parameters returned by the model are the phylogenetic half-life (t1/2) and the stationary variance (vy). Phylogenetic half-life represents the average time to evolve half the distance from the ancestral phenotype towards the predicted optimal phenotype, and thus is a quantification of the phylogenetic inertia (Hansen, 1997). A half-life above zero indicates that adaptation is not instantaneous, while when t1/2 = 0 means that there is no evolutionary lag. The stationary variance is the stochastic component of the model, which can be interpreted as evolutionary changes in the response trait to unmeasured selective forces and genetic drift.
Phylogenetic half-life in a model that only includes the intercept is a measure of the phylogenetic effect in the response trait. A half-life = zero in such a model means that the response variable is not phylogenetically structured, while a half-life > 0 indicates that there is an influence of phylogeny on the data; when the half-life shows high values, this can attributed to an underlying continuous Brownian motion process. Indeed, a phylogenetic effect can be due to lag of adaptation, adaptation towards phylogenetically structured optima, or a combination of both. Hence, when a trait exhibits phylogenetic signal, it is not immediately clear whether this is due to the trait itself having strong inertia, or if the trait is evolving in response to a phylogenetically structured variable. The intercept-only model is contrasted with a model that also includes a predictor variable. This type of model is regarded as adaptation model because it tests whether the response traits evolve towards optima influenced by a predictor. By comparing a model with and without the inclusion of predictor variables, it is possible to determine how much of the phylogenetic signal can be attributed to phylogenetic inertia. No reduction in the t1/2 suggests that phylogenetic signal can be entirely attributed to phylogenetic inertia; on the contrary, when a trait evolves in response to a variable, a reduction in the half-life for the response trait should be observed.
The adaptation models which include continuous predictors (i.e. the climatic variables in our study) are fitted using maximum likelihood for estimation of t1/2 and vy and generalized least squares for estimation of the regression parameters in an iterative procedure (Hansen et al., 2008). In addition, the SLOUCH model assumes that the predictors have a longer phylogenetic half-life than the model residuals, and this is well supported by the variables involved in our study. The model returns parameters of an optimal regression and of a phylogenetic regression. The former is the relationship between the response and predictor variable that is predicted to evolve free of ancestral influence (absence of inertia). Therefore, the slope of this regression must be steeper than that of the phylogenetic regression.
To evaluate the effect of the categorical predictors on the evolution of chromosome number, the ANOVA and ANCOVA extensions implemented in SLOUCH were used. Categorical character-states were mapped onto the phylogeny using parsimony reconstruction.
Intercept-only models are compared to the adaptation models using the Akaike information criterion corrected for small sample-size (AICc); models with AICc scoring zero-two units lower are considered substantially better while if more than two units lower than they are considered significantly better (Burnham & Anderson, 2004). Finally, model interpretations were based on comparisons of t1/2 and vy of the adaptation models with the intercept-only model, together with the amount of variation in chromosome number that the models explain. All statistical analyses have been carried out in R v3.2.3.
Results
Phylogenetic effects in chromosome number
The phylogenetic effects in chromosome number varied among the examined clades, but were generally moderate to large (Table 1 ). Diploid chromosome numbers (2n) exhibited significant phylogenetic effects (t1/2 > 0) and the supported values (values with a log-likelihood until two units lower than the maximum log-likelihood) ranged from a moderate to strong phylogenetic effect, not exceeding 1.0 total length (t1/2 < 1), with the exceptions of Malvids and Caryophyllales. Indeed, the best estimate for Malvids was t1/2 = 0.01, with a support region from 0 to 0.05, suggesting that the trait is evolving rapidly, nearly instantly on the timescale of this phylogeny. On the other side, the half-life value for Caryophyllales, instead, was many times the total tree length (infinity), indicating that 2n evolves as if by Brownian motion in this clade.
Compared to diploid numbers, the basic chromosome number (x) showed stronger phylogenetic effects, with half-life that include t1/2 > 1 and a supporting region that include t1/2 = ∞. However, the wider support regions suggest that estimates of t1/2 for x are more uncertain than those for 2n.
It should be noted that the stronger phylogenetic effects exhibited by x are associated with coefficients of variation (CV) lower than 0.5, while the CV calculated for 2n are larger and this was followed by weaker phylogenetic effects. Mean chromosome numbers (± sd) and CV for each clade are as follows. Monocots (2n): 33.7 ± 25.1, CV = 0.74; Monocots (x): 10.8 ± 5.2, CV = 0.48; Fabids (2n): 32.1 ± 17.9, CV = 0.56; Fabids (x): 10.5 ± 2.8, CV = 0.27; Malvids (2n): 26 ± 20.3, CV = 0.78; Malvids (x): 8.8±2.4, CV = 0.28; Caryophyllales (2n): 27 ± 11.3, CV = 0.42; Caryophyllales (x): 10.2 ± 2.2, CV = 0.22; Lamiids (2n): 29.7 ± 18.4, CV = 0.62; Lamids (x): 9.89 ± 3.6, CV = 0.37; Campanulids (2n): 25.6 ± 14.1, CV = 0.55; Campanulids (x): 10.6 ± 4.4, CV = 0.42.
Adaptation and inertia in chromosome number
We found clear evidence for adaptation of chromosome number to environmental and morphological predictors, but in several cases, and especially for basic chromosome number (x), a Brownian motion model best explained the evolution of chromosome number on the phylogeny (Tables 1 and 2).
The effects of the continuous (climatic) predictors were generally weak. Indeed, only eight of the 60 models that used climatic predictors had lower AICc and a half-life reduction (indicating that not all phylogenetic effect is due to phylogenetic inertia), compared to the model without predictor (Table 1). Nevertheless, even in these cases (all these models refer to 2n), the half-life reduction was small and the AICc decrease never exceeds 2 units which means there is a weak tendency for chromosome numbers to evolve towards the optimum. Hence, these results should be regarded with caution as they represent only a tentative indication of climatic effects on chromosome numbers.
In the eight above mentioned models, the climatic variables that explained, at least marginally, variation in chromosome number were mean temperature, temperature continentality and precipitation seasonality. Overall, the association of mean temperature with chromosome number was negative with the slope of the phylogenetic regression nearly flat in Caryophyllales and Lamiids, while it was steeper in other clades, despite explaining less than 4% of the variance. The relation, albeit not significant, is, instead, positive in Fabids. A negative relation with chromosome number was also found for precipitation seasonality, while temperature continentality was generally positively associated with chromosome number.
For categorical predictors, the relation with both 2n and x is overall robust with 32 out of the 84 models outperforming the model without predictor (Table 2). Chromosome number is mostly affected by habitat categories (light, moisture and nutrient), but in some clades also by morphological categories (growth form and flower size). The gain in AICc values for the outperforming models achieved more than 2 units with percentage of the variance explained largely exceeding 5%. These models also exhibited a significant reduction in half-life, which indicated that chromosome number evolved in response to categorical variables while some, but not all, of the phylogenetic effect in chromosome number is due to phylogenetic inertia. Specifically, open habitat selected for low chromosome numbers while shaded, stable environment (e.g. forest) selected for higher chromosome numbers in Monocots, Fabids, Campanulids (for both 2n and x) and in Mavids and Lamiids (for 2n only) (Fig. 2). Habitat moisture and nutrient availability were positively associated with chromosome number, although some optimal values (θ) had little biological meaning. This particularly applies to eutrophic and wet categories due to the low number of branches attributed to these categories. Hence, we regard these results with caution; nevertheless, the overall patterns of the models are robust and understandable. Chromosome number is also positively related with plant morphological traits, namely growth form (life forms with longer life cycles had higher chromosome numbers) and flower size, but in a few cases also negatively associated with flowers clustered in flower-like inflorescence.
Discussion
In this study, we found phylogenetic effects in chromosome number ranging from moderate to large, rejecting the hypothesis of species independence. Our study also demonstrates that phylogenetic inertia is a significant component in all models (t1/2 > 0). Nevertheless, we found evidences for adaptation of chromosome number to environmental and morphological predictors, especially for 2n, while, not surprisingly, x exhibited a lower degree of variation and less significant adaptive evidences, providing insight into the ecological significance of polyploidy. Hence, although the majority of variation remains unexplained, in the light of strong phylogenetic inertia, even accounting for a small amount of the total variation, indicates that a Brownian motion process inadequately explains the evolution of chromosome number due to the action of natural selection (Felsenstein, 1985; Hansen, 1997).
From our results it seems that chromosome number is poorly related to large scale climatic conditions. Only three climate variables showed modest significant relation indicating that species with higher chromosome numbers tend to occur in sites with lower annual temperature, lower precipitation seasonality and higher continentality. On the contrary, a stronger relation with categorical variables was found. Specifically, open, disturbed, drought-prone habitats selected for low chromosome numbers, while shaded, stable environments with good availability of water and nutrients selected for high chromosome numbers and consequently for increased recombination rates (Grant, 1958). Bell (1982) claimed that morphological features associated with a rapid rate of production of many propagules are correlated with low chromosome number, while Stebbins (1966) argued that chromosome number reduction is related with shorter life cycles and higher chromosome numbers (especially polyploids) with polycarpic perennials (Gustafsson, 1948). As we found that low chromosome numbers are associated with small flowers clustered in flower-like inflorescence and higher chromosome numbers with perennial herbs, especially geophytes, the above mentioned speculations are confirmed by the present study. Furthermore, as higher chromosome numbers were exhibited by woody plants, where polyploidy is rare, we argue that considering their long cycles, this life form show clear evidence for increased recombination rates counterbalanced by smaller chromosome sizes (Stebbins, 1971) caused by the structural and physiological constrains on cell size (Beaulieu et al., 2008). In the present study several associations among chromosome number, plant traits and environmental factors were found. However, as mentioned by other authors (Hipp, 2007; Escudero et al., 2012), it should be reminded that there is no reason to expect chromosome number per se to affect plant fitness. Instead, chromosome number variation can be driven by cellular processes that affect meiosis and mitosis. Hence, we also interpret the evolution towards an optimal state as a karyotypic equilibrium determined by the rates of mutations (Pardo-Manuel de Villena & Sapienza, 2001), rather than an adaptive optimum.
While categorical predictors and, to a less extent, some climatic variable (mean temperature and precipitation seasonality) supported the hypothesis that environmental instability and seasonality select for low chromosome numbers (i.e. decreased recombination rates), temperature continentality suggests that chromosome number selection is also governed by other, still not fully interpretable, reasons. It should be noted, however, that models showing modest significant effects of climatic predictors are those modelling 2n evolution, pointing out the relative contribution of polyploidy and its ecological significance on diploid chromosome number. The increase of polyploid taxa with latitude and in colder sites with continental climates has been already shown by early authors (Löve & Löve, 1957; Hair, 1966; Hanelt, 1966; Stebbins, 1971). The recent availability of large chromosome number databases (Peruzzi & Bedini, 2014; Rice et al., 2015) has spurred further research on this subject, spanning a substantially wider taxonomic space and confirming this trend across the whole Arctic flora (Brochmann et al., 2004) and at other geographical scales (Bedini & Peruzzi, 2015 and literature cited therein). Nevertheless, even among categorical predictors, models fitted on 2n are generally exhibiting stronger association with environmental and morphological predictors than the models fitted on x, again suggesting a putative ecological significance of polyploidy. Particularly, we speculate whether genome duplication, through polyploidy, might ensure the required recombination rates and its relative fitness when the karyotypic equilibrium cannot immediately approach to a selective state depending on the environment.
Results are generally similar among clades, and, especially for Monocots and Campanulids are quite robust; nevertheless, some additional notes may be highlighted. Monocots exhibit a high variation in chromosome number and include genera (e.g. Poales) in which holocentric chromosomes occur, which is a peculiar condition allowing for rapid diversification and range of chromosome number (Escudero et al., 2016). However, the large presence of geophytes in this clade, also explains the results obtained, especially for 2n. Indeed, in geophytes, the presence of large cells is assumed to be an advantage during the rapid development of the plant body and cell expansion (Grime & Mowforth, 1982); as a consequence, they show higher tolerance to genome duplication, with a high frequency of polyploids. The performance of the OU models on Malvids might be biased by low sampling (Cooper et al., 2016), while in Fabids the results are partly obscured by the co-occurrence of dysploidy and polyploidy phenomena, especially in Mediterranean taxa (e.g. Genista). In Caryophyllales, results are biased by the predominance of taxa belonging to the genus Limonium, notwithstanding, model fit on 2n and x are largely comparable, pointing out for other grounds, like habitat patchiness/stochasticity and high frequency of hybridization, rather than effect of polyploidy on speciation and chromosome number changes.
Conclusions
The main goal of the present study was to disentangle phylogenetic constrains from potential ecological determinants in chromosome number and the SLOUCH programme, through the phylogenetic half-life (t1/2), indeed allowed to quantify phylogenetic inertia (Hansen, 1997). The dataset includes more than half of Italian endemics covering 27 orders and 6 main angiosperms clades, which makes this study a comprehensive interspecific assessment of the relation among chromosome number, plant traits and environmental factors in angiosperms. Yet, having demonstrated that phylogenetic inertia is a significant component in chromosome number evolution, future studies can be extended using a different approach where not only the primary optima but also the α-parameter (strength of selection) and σ-parameter (strength of the drift) differ among niches, allowing to test for clade-dependent rates of adaptation (Beaulieu et al., 2012). At this stage, our study affords non-stochastic demonstrations for chromosome number variation. In addition, whilst phylogeny is a strong predictor of trait value, especially for x, a simple phylogenetic explanation is inadequate.