A genome size and phylogenetic survey of Mediterranean Tripleurospermum and Matricaria (Anthemideae, Asteraceae)

The study of genome size variation can contribute valuable information on species relationships as well as correlate to several morphological or ecological features, among others. Here we provide an extensive report on genome sizes on genus Tripleurospermum and its closely related genus Matricaria, which are two typically Mediterranean genera particularly widespread and diverse in Turkey, the origin of most of the populations here studied. We analyse and discuss genome size variation in the first relatively complete molecular phylogenetic framework of Tripleurospermum (based on ITS and ETS ribosomal DNA–rDNA–regions). We find cases of intraspecific genome size variation, which could be taxonomically significant. Genome downsizing is also detected as the typical response to polyploidisation in Tripleurospermum taxa, being most conspicuous at the tetraploid level. Several positive correlations with genome size, including those with pollen and stomatal size or cypsela length, among others, are also found. Remarkably, taxa presenting rhizomes tend to present higher genome sizes, confirming a trend to accumulate nuclear DNA in such species, which could be explained by the nutrient reserves availability in their storage organs, allowing genome expansion, or by the lower rates of sexual reproduction in rhizomatous taxa.


Introduction
Tripleurospermum Sch.Bip. is a genus of family Asteraceae, tribe Anthemideae, comprising between 30 to 40 species, depending on the authors and systematic treatments. Distributed in the Northern hemisphere, mainly in the Mediterranean basin including Europe, temperate Asia and North Africa [1,2,3], some taxa are also present in North America. It is particularly abundant and diverse in Turkey, its main center of diversity, where around 30 Tripleurospermum taxa can be found half of which endemic [4]. One of its species, T. inodorum (L.) Sch. Bip., is widespread as a weed [1]. Another Anthemideae genus with an important presence in the Mediterranean basin and closely related to Tripleurosperum is Matricaria L. The genus a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 comprises six species [2,5], three of them present in Turkey. Its species M. chamomilla L., a part from being widely known for its medicinal properties, is also a widespread weed [6,7].
As in many Anthemideae and mostly due to close morphological affinity Tripleurospermum (subtribe Anthemidinae), Matricaria (subtribe Matricariinae) and, to a lesser extent, Anthemis L. (subtribe Anthemidinae), among others, have been confused both taxonomically and nomenclaturally [4,8,9,10]. However, although some Tripleurospermum species were formerly assigned to Matricaria, later it was recognized as a different genus based on fruit features. On the one hand, the morphology of Tripleurospermum cypselas differ from that of those of Matricaria in its shape and ornamentation [2]. Besides, Tripleurospermum species have a tetrasporic embryo sac, a character shared with the genus Anthemis, while Matricaria and other Anthemideae present monosporic embryo sac [11,12]. Molecular phylogenetic researches based on tribe Anthemideae are also consistent with this difference and both genera appear well separated in different and supported clades [13,14].
Although karyology can also contribute to clarify relationships between closely related taxa [15] such as Matricaria and Tripleurospermum, little differences have been found until present regarding their chromosomes. Karyological knowledge on both genera is mostly based on chromosome counts and in some (few) karyotypes. Species of Tripleurospermum and Matricaria have the same and exclusive base chromosome number, x = 9, which is also the most common in tribe Anthemideae and in family Asteraceae as a whole, where it is considered the ancestral condition [16]. In certain Tripleurospermum species, triploid, tetraploid and pentaploid populations have been detected [17,18], whereas in Matricaria only the diploid level has been recorded [19].
The study of genome size variation has shown its utility in systematic and evolutionary research in many plant groups. A first flow cytometric approach in Tripleurospermum was performed by our research group [17], where we measured genome sizes by flow cytometry for seven of its species and related them with features such as ploidy level, life cycle and environmental factors. Until then, only Nagl and Ehrendorfer [20] had estimated, through Feulgen cytodensitometry, the genome size for Tripleurospermum maritimum, the most widespread species of the genus. Very recently, Certner et al. [21] studied genome size variation by flow cytometry in mixed ploidy populations of T. inodorum. As for Matricaria, previous studies have contributed genome size information, using either Feulgen cytodensitometry [20,22] or flow cytometry [23,24].
The main purpose of this study is to address genome size variation in the two closely related genera Tripleurospermum and Matricaria, increasing the sample to embrace most of the species of both genera, for many of which we will provide first estimates. We will test relationships between genome size and morphological, karyological and ecological features of the species in a phylogenetic framework constructed for this purpose. This may also help clarifying systematic relationships within and between these genera and other closely related Anthemideae, as genus Anthemis.

Plant materials
Cypselas of 64 populations, corresponding to 30 Tripleurospermum species and subspecies (42 populations), four Matricaria species and subspecies (21 populations) and one Anthemis species were collected from the wild for genome size assessments and molecular phylogenetics. We have included the species Anthemis macrotis (Rech.f.) Oberpr. & Vogt in the analysis since (i) it was relevant as a genus closely related to both genera and had been previously considered a member of Matricaria (M. macrotis Rech.f.), although a more recent taxonomic study based on molecular markers best placed it as a member of Anthemis [9], and (ii) no previous genome size estimate was available for the species. None of the studied species is listed as endangered or protected in any national or international legal regulation. The accessions are listed in Table 1 following the taxonomic names in the Flora of Turkey the East Aegean Islands [25], Inceer and Hayırlıoğlu-Ayaz [4] and Inceer [26]. The map in Fig 1 indicates

Molecular techniques: DNA extraction, amplification and sequencing
Total genomic DNA was extracted using either the CTAB method [27] as modified by [28] [29]. The PCR profile used for amplification was 94˚C 3min; 30 × (94˚C 20 s; 55˚C 1 min; 72˚C 1 min); 72˚C 10 min. Only the ITS4 primer was used for sequencing in most cases. ETS region-Double-stranded DNA of the ETS region was amplified with the ETS1f and 18SETSr primers [30]. The PCR profile used for amplification was 94˚C 3min; 30 × (94˚C 20 s; 50˚C 1 min s; 72˚C 1 min); 72˚C 10 min. Both ETS1f and 18SETS were used as sequencing primers, and also the internal primers AST1F and AST1R [31] were used occasionally.

Flow cytometric measurements
For flow cytometric measurements of leaf tissue were obtained from seeds grown in pots in the greenhouse of the Faculty of Pharmacy and Food Sciences, Universitat de Barcelona. Five individuals per population of each species were studied, and two samples of each were individually processed. Petunia hybrida Vilm. 'PxPc6' (2C = 2.85 pg) and Pisum sativum L. 'Express Long' (2C = 8.37 pg) were used as the internal standards [32]. Fresh leaf tissue for the standard and the target species were chopped together in 600 μl of LB01 buffer (8% Triton X-100) [33] supplemented with 100 μg/ml ribonuclease A (RNase A, Boehringer, Meylan, France) and stained with 36 μl of 1 mg/ml propidium iodide (Sigma-Aldrich, Alcobendas, Madrid) to a final concentration of 60 μg/ml, and kept on ice for 20 min. The fluorescence measurements were performed using an Epics XL flow cytometer (Coulter Corporation, Miami, FL, USA) at the Centres Científics i Tecnològics, University of Barcelona. More details about the method are in [34]. The data were submitted to the Genome Size in Asteraceae Database (GSAD) [35].

Phylogenetic analyses and reconstruction of character evolution
The nuclear ribosomal DNA dataset (1075 concatenated bp) includes ITS (excluding the 5.8S) and 5' ETS sequences (636 and 439 bp, respectively) for 45 taxa out of which 34 belong to Tripleurospermum, four to Matricaria and the remaining seven to different species of genera Achillea (GenBank accession number-hereafter GB-: AY603251), Anacyclus L. (GB:    (2) Ploidy level.
(4) Holoploid genome size and standard deviation in brackets. The two sequence matrices obtained with the nuclear molecular markers were manually edited and concatenated (S1 File) with BioEdit v. 7.1.3.0 [37] and MAFFT [38]. 54 gaps or indel characters were coded with FastGap [39] as 0/1 (absence/presence) and added to the data matrix as a separate partition.
The phylogenetic analyses were performed in the CIPRES Science Gateway [40]. Bayesian Inference phylogenetic analysis was performed in MrBayes v.3.2.6 [41] using the GTR+I+G model previously determined from jModeltest v.2.1.6 [42] under the Akaike information criterion [43]. The Markov chain Monte Carlo (MCMC) sampling approach was used to calculate posterior probabilities (PP). Four consecutive MCMC computations were run for 100,000,000 generations, with tree sampling every 10,000 generations. The first 25% of tree samples were discarded as the burn-in period. PP were estimated through the construction of a 50% majority rule consensus tree.

Statistical analyses
All data manipulations and statistical analyses were performed with RStudio, v.0.98.1078, a user interface for R [44]. The phylogenetic generalised least squares (PGLS) algorithm as implemented in the nlme package for R (Version 3.1-118) was used to analyse variation of genome size with respect to karyological (chromosome number, ploidy level) and morphological traits (pollen polar axis and equatorial diameter, size of the stomata, length and width of the cypsela, size of the plant, capitulum type (homogamous, heterogamous), presence of mucilage in cypsela, presence of rhizome), as well as other features such as life cycle, altitude, habitat and invasive behavior, following Olanj et al. [45]. The information used for the analyses is presented in S1 Table, together with genome sizes of the accompanying species of other genera, which have been obtained from the GSAD database [35]. Additionally, regular statistical  analyses of regression, one-way analysis of variance (ANOVA) and Shapiro-Wilk test for normality were performed, without considering phylogenetic relationships between taxa. The packages ape and geiger were also required for the phylogenetic-statistical analyses, as well as the package agricolae for LSD tests. Since in most cases datasets were not normally distributed, we also performed non-parametric tests such as Spearman rank correlation, the Kruskal-Wallis test by ranks and multiple comparison tests after Kruskal-Wallis (using the pgirmess package).

Phylogenetic relationships between Tripleurospermum and Matricaria
Here we contribute a preliminary molecular phylogenetic framework for Tripleurospermum and related genera, in which to analyse genome size variation together with other traits of the species. Genus Tripleurospermum appears as monophyletic and strongly supported (PP = 1.00) in our phylogenetic reconstruction (Fig 2). All Tripleurospermum taxa possess a 106 bp insertion in the 5' ETS which clearly defines the group. Relationships within the genus point to two large groups highly supported (PP = 0.99 and 0.94), and several other clades at lower hierarchical levels also show strong statistical support (PP = 1.00). Although the support is moderate (PP = 0.83) the sister taxa for Tripleurospermum could be species from genus Anacyclus. Species from genera Anthemis, Tanacetum and Achillea occupy intermediate positions between Tripleurospermum and Matricaria. The latter genus also appears as monophyletic and strongly supported (PP = 1.00). Both genera, Tripleurospermum and Matricaria, are clearly independent in this phylogenetic framework, although both grouped with the remaining genera in a highly supported clade whose outgroup is Artemisia.

Genome size variation
The present study expands genome size knowledge in Tripleurospermum up to 83% of the accepted taxa (25 out of 30 taxa according to the Euro+Med PlantBase [46]). There are also some species, such as T. melanolepis (Boiss. & Buhse) Pobed. and T. transcaucasicum (Manden.) Pobed., that do not appear in this database but are included as accepted names in The Plant List ( [47], accepting 38 species in the genus), for which we also contribute new genome sizes. Our contribution to Matricaria expands to 50% the genome size knowledge on the genus, while for Anthemis it is much smaller, since we only add one species to the about 150 accepted species. The basic chromosome number of all taxa (Tripleurospermum, Matricaria and Anthemis) is x = 9 and ploidy level of the 64 populations here studied ranges from diploid (46 populations, 71.9%), to pentaploid (albeit only one population of T. subnivale), being the remaining 17 populations (26.6%) tetraploid. Table 1 provides holoploid genome sizes (2C), ranging from 4.65 to 13.11 pg in Tripleurospermum and from 4.6 to 5.39 pg in Matricaria. The average half peak coefficient of variation (HPCV) corresponding to 10 samples of five individuals was 4.5% for the problem plant and 3.2% for the standards, indicating the good quality of the flow cytometric assessments. We found low intraspecific genome size differences in several cases in which two or more populations had been assessed in this study: 0.4% between three populations of T. hygrophilum (Bornm.) Bornm., 2.34% between two populations of T. parviflorum (Willd.) Pobed., 2.55% between two populations of T. heterolepis (Freyn. & Sint.) Bornm. and 2.64% between two populations of Matricaria aurea. Moderate intraspecific differences were detected between between 11 populations of Matricaria chamomilla (8.66%) and seven populations of M. chamomilla var. recutita (L.) Fiori (5.18%), four populations of T. baytopianum E. Hossain (7.95%) and two populations of T. inodorum (8.25%). The highest percentage difference was found between two populations of T. conoclinum (15.44%) and also between three populations and two varieties of T. oreades (15.07%). Within Tripleurospermum, two supported clades can be clearly distinguished (A and B in Fig 1). Ploidy levels are significantly different between these clades (Pearson's Chi-squared test X 2 = 18.514, p = 1.656Áe -05 ), in which species in clade A are all tetraploid except six (out of 21), while species in clade B are all tetraploid with the exception of one (out of 12). This is also reflected as significantly different holoploid genome sizes between both clades (p = 6.47Áe -05 ). However, holoploid genome sizes (2C) of diploid species in both clades are not significantly different (p = 0.221). Table 2 summarizes the main results of the statistical analyses performed. Using the phylogenetically generalized least squares method (PGLS) we found that holoploid genome size  (2C) is significantly and positively correlated with ploidy level (p = 0.0000) and chromosome number (p = 0.0000). Monoploid genome size (1Cx) decreases significantly with both (p = 0.0000), which points to certain degree of genome downsizing. Other morphological characters that have produced significant relationships with genome size, considering phylogenetic relationship between taxa, are: size of the plant (positive correlation, p = 0.0216), presence of mucilage in cypselas (negative correlation, p = 0.0436) and presence of rhizome (positive correlation, p = 0.0034). Regarding the habitat, species living in alpine environments present significantly higher genome sizes than those living in the remaining habitats (2C = 9.23, p = 0.0007) while species inhabiting lower, ruderal areas, show significantly lower C-values (2C = 5.51, p = 0.043). As for the geographical distribution, differences are not significant (though nearly) between species with a large (2C = 6.26, p = 0.0568), medium (2C = 6.45, p = 0.063) or restricted (2C = 6.50, p = 0.0523) area. The other features tested against genome size (pollen polar axis and equatorial diameter, dimensions of cypsela, size of stomata, capitulum type, altitude and life cycle) did not show any phylogenetically significant correlation. However, when the correlations were performed without considering the phylogenetic relationships between taxa, pollen polar axis and equatorial diameter were significantly and positively correlated with genome size (p = 0.002 and p = 0.003, respectively), as it was the length of the cypsela (p < 0.0001), the size of the stomata (p < 0.0001) and the altitude (p = 0.0071).

Phylogenetic placement of Tripleurospermum and Matricaria, and a mention to Anthemis macrotis
Understanding phylogenetic relationships within tribe Anthemideae has always been problematic. As Oberprieler et al. [2] stated, while the circumscription of the tribe is clear [1], the subtribal classification has caused considerable difficulties through its taxonomic history. Both Tripleurospermum and Matricaria appear as monophyletic and highly supported, and independent from each other, confirming previous works on morphological [48] and molecular bases [2]. Our results allow, in this respect, confirming the placement of Tripleurospermum and Matricaria in different subtribes, as proposed previously (Anthemidinae and Matricariinae, respectively) [5]. Phylogenetic relationships above Tripleurospermum are overall consistent with the recent work of Vitales et al. [49], placing species from genus Anacyclus as the likely sister group of the former genus. Besides, the species Anthemis macrotis occupies an intermediate position between Tripleurospermum and Matricaria, confirming [9], now with an additional molecular marker (ETS), which best placed the former Matricaria macrotis as a member of Anthemis on molecular bases. The former inclusion of this "rare and enigmatic species of the East Aegean" in Matricaria was based on certain morphological features (absence of receptacular scales) yet others (e.g. indumentum, achene traits) pointed to its best treatment as Anthemis, which was later confirmed [9]. As for intraspecific structuring, despite extensive research in the genus, in particular from the morphological [50,51], anatomical [12,52], palynological [53], chemical [54,55] and karyological [18,56] points of view, little is known about phylogenetic relationships between Tripleurospermum species. The reconstruction here presented has been performed with the purpose of providing a phylogenetic framework in which to analyse genome size variation, although it gives, in addition, a quite resolved picture of the interspecific relationships within the genus. As indicated previously, two large and significantly supported clades are apparent (Fig 1A and 1B) and these appear to differ on ploidy levels of their species (see the different sizes of the bars representing genome sizes in both clades in Fig 1). It is also observed that clade A is the largest and most diversified, containing 19 taxa and 18 species as compared to clade B with only 13 taxa and 11 species. Polyploidy has been linked to higher diversification rates [57] and it is likely that this has been the case in Tripleurospermum, in which, additionally, the polyploid taxa are more widely distributed geographically than the diploid.
Several clades observed in this phylogenetic reconstruction reflect what previous work (mostly floras) had already stated for certain species. In particular, species like T. repens, T. heterolepis, T. sevanense and T. callosum, inhabiting similar areas in North East Anatolia, had been considered closely related [25,51]. They appear in the phylogeny in a strongly supported clade (Fig 2A); similarly, T. melanolepis, T. subnivale, T. caucasicum and both varieties of T. oreades are allied in the Flora of USSR, in the Flora of Turkey and in other taxonomic treatments [8,25]; besides, T. caucasicum and T. oreades are usually considered as synonyms [46,47], whereas these species are not considered as synonyms by some authors [8,25,26] [8].
These taxa are also members of the same highly supported clade in the molecular phylogeny ( Fig 2B). Species like T. maritimum and T. inodorum (Fig 2C) also form a moderately supported clade, which is consistent with the fact that they have also been considered as synonyms of each other in several occasions (indeed one of the populations was recorded as T. perforatum, considered synonym of T. inodorum; similarly, T. maritimum has sometimes been quoted as T. inodorum ssp. maritimum). Other consistent groupings, such as the one formed by both populations of T. conoclinum and T. parviflorum, are probably explained by the sympatric habitat of both species (the three populations inhabit in close proximity in the area of Izmir Province). Other groupings may also respond to sympatry, such as the one uniting T. corymbosum, T. elongatum, T. transcaucasicum and T. microcephalum (Fig 2D): the studied populations are found in the close areas of Agrı, Ardahan and Kars Provinces, respectively, from East Anatolia; in the Flora of Turkey [25] also a close relationship between T. corymbosum and T. transcaucasicum had previously been suggested on a morphological basis.

Genome size variation and genome downsizing in polyploids
The present study contributes new genome size data for 22 taxa of Tripleurospermum, for the species Matricaria aurea and one variety of M. chamomilla (M. chamomilla var. recutita) and for the species Anthemis macrotis, for whose genus only five genome size estimates were previously available. Genome sizes of seven Tripleurospermum species and one variety (T. callosum, T. elongatum, T. maritimum, T. melanolepis, T. oreades, T. oreades var. tchihatchewii, T. repens and T. sevanense) have been reassessed with respect to previous results [17], which are consistent through both research works. As for Matricaria, the values reported for M. chamomilla var. chamomilla are also consistent with previous genome size estimates of diploid populations of the species [58]. Regarding M. discoidea (synonym of M. matricarioides) there is a remarkable difference (23%) with one of the populations measured previously [22] with Feulgen microdensitometry (4.6 vs 5.66 pg). Average genome sizes of Tripleurospermum (at diploid level) and Matricaria are not significantly different (5.19 and 4.98, respectively), however the average genome size of Anthemis species (6.47) is significantly different from both genera (p = 0.0057), which adds evidence to the separation of Anthemis macrotis (2C = 5.55 pg) from Matricaria.
Although the C of the term "C-value" stands for constancy of genome size within a species [59], examples accumulate which show that, many times, this is not the case [29,60,61]. Although generally attributed to karyotype variations (e.g., polyploidy, aneuploidy or presence of B-chromosomes) many research works have observed intraspecific genome size variation beyond chromosomal features. However, well-known methodological variation or presence of staining inhibitors in plants must always be taken into account when discussing intraspecific genome size variation. Intraspecific genome size changes were explained by Ceccarelli et al. [62] as a result of quantitative modulations of DNA repeats and transposable elements. Genuine intraspecific variation, even within a population, has been documented for several species [63]. As discussed earlier [45], it is difficult to set a cut-off point from which a given percentage should be considered significant. In the previous genome size study on some Tripleurospermum taxa we found low genome size differences between most populations [17]. Here, with a much larger sampling in terms of both taxa and populations, we detect a moderate to high degree of variation between the two varieties of T. oreades, two populations of T. inodorum and two populations of T. conoclinum. Murray [64] related intraspecific genome size variations to microevolutionary differentiations, which could be taxonomically significant. This could be the case of the two varieties of T. oreades, whose distinct phylogenetic placement (Fig 2) would also support their separation as distinct species beyond genome size, as proposed previously on the basis of morphological and karyological characters [8,56]. The case of T. inodorum could be explained by its widespread distribution and possible introgression or hybridization with its congeners [65]. Interestingly, however, one of the populations analysed of T. inodorum was formerly given the name of T. perforatum, considered as a synonym of T. inodorum, which is indeed the valid name for the species [47]. The fact that both populations of T. inodorum appear closely related in the phylogenetic tree would support this consideration; yet, genome size of the population named as T. inodorum is 8.25% lower than that formerly named as T. perforatum. So this difference might also be interpreted in terms of microevolutionary differentiations indicating certain speciation processes, still not enough manifested morphologically. The same could have happened between the two populations analysed of T. conoclinum: remarkably, the genome size of population 264 is closer to T. parviflorum than to that of the other T. conoclinum population 262, forming a highly supported clade in the phylogeny. As for Matricaria, the interpopulational differences detected in this study between M. chamomilla var. chamomilla and M. chamomilla var. recutita are moderate (below 10%), despite having assessed a relatively large number of populations (11 and seven, respectively). Our results are almost 50% lower than previously reported (2C = 7.7 pg) [20] for a population of the same ploidy level. Similarly, our genome size assessment of M. matricarioides is about 23% lower than previously reported by the same authors (2C = 5.66 pg). The different techniques used in our study (flow cytometry) and in the previous (Feulgen densitometry) could partly explain the discordance between data. However, results obtained for T. maritimum in the present study (2C = 5.28 pg) are almost equal to the previous data provided by Nagl and Ehrendorfer [20] (2C = 5.25 pg) so we cannot discard genuine interpopulational differences.
Genome downsizing is a well-known and common finding in polyploid systems [66]. It is a widespread biological response to polyploidisation, which may lead to the further diploidisation of the polyploid genome. In the case of Tripleurospermum, there is significant genome downsizing between diploids and tetraploids, which have around 20% less than expected monoploid genome size on average. However, the only pentaploid species of our dataset apparently does not undergo downsizing, as its monoploid genome size (1Cx = 2.62) is equivalent to that of the diploids (1Cx = 2.60). Genome downsizing has been detected in genera closely related genera to Tripleurospermum such as Artemisia [67], although the highest levels of reduction in the monoploid genome were usually found at higher ploidy levels contrary to our findings in the pentaploid T. subnivale. Maybe, the likely allopolyploidisation process responsible for the appearance of the pentaploid cytotype could imply the expansion of certain genomic repeats, which would increase its size.

Genome size and phenotypic traits
Correlations between genome size at the phenotype scale are supported by many research works pointing to a positive relationship with traits such as seed and leaf mass [68,69,70], pollen size [45], and stomatal size [71], among others. In this regard, we have also detected a positive correlation between pollen dimensions, stomatal size and length of the cypsela, which support the direct effect of genome size on cell size (nucleotype theory, i.e. the indirect influence of DNA in development by the physical-mechanical effects of its mass) [72]. At a higher phenotype scale, we found comparable results with the height of the plant. Knight and Beaulieu [73] showed that the strength of the correlations between genome size decreased in predictive power with increasing phenotypic scale, yet our findings argue for such correlations also at higher scales in Tripleurospermum. Closely related to this argument, rhizomatous Tripleurospermum species show significantly higher genome sizes than those without rhizomes, as pointed previously from a more limited sampling on the genus [17]. Also in the Asteraceae genus Artemisia, the largest genome sizes are found in plants showing vegetative multiplication [74]. Much earlier, Rees and Jones [75] had also observed larger genome sizes in species with vegetative reproduction than in those presenting sexual reproduction from genus Lolium. We had previously argued that the presence of rhizomes could be related with higher rates of asexual reproduction and therefore lower incidence of meiosis as a controlling mechanism of genome size expansion, which may explain higher genome sizes in species showing such vegetative organs. Besides, Veselý et al. [76] explained that the universal tendency of geophytes (to which rhizomatous species belong) to possess higher genome sizes than their non-geophytic relatives could be related with the nutrient reserves availability in their storage organs. Polyploidy has also been found strongly associated with vegetative reproduction than diploids [77] and this is consistent with our findings since many rhizomatous species of our sample are polyploid (though not all).
The negative correlation between the presence of slime in cypselas (a mucilaginous layer that protects the seed and promotes its germination) and genome size/ploidy level in Tripleurospermum is consistent with the observations of [78,79], which did not detect any slime production in certain polyploid Artemisia, while most diploids showed it. It was argued that the ability to form slime may depend on ploidy level [79]. In high Artemisia polyploids developmental abnormalities have been observed, such as disturbances in cypsela production or germination. Differences in gene expression have also been found in potato autopolyploid series at higher ploidies [80], so it might be possible that there is some dysfunction within the expression of genes responsible for mucilage production in cypselas of polyploid taxa, possibly explaining the usual absence of slime in polyploid Tripleurospermum taxa.

Genome size, environment and ecology
The genome sizes of the diploid plants studied fall within the category of small genomes (2C values between 2.8 and 7 pg), whereas those of polyploids belong to the category of intermediate genomes (2C values between 7 and 28 pg) [81], and no large and very large genomes are present in the current set. This is consistent with the absence of extremophilous taxa among those considered, in agreement with the large genome constraint hypothesis [82], postulating that plants with large genomes are rare in extreme environments.
It has been suggested [83,84] that an increased DNA amount was an adaptation to altitude. In the same line [85,86,87] found a high rate of polyploid taxa in different alpine systems. The selective advantages that polyploidy may confer could explain a better colonizing ability of alpine habitats by polyploids. Although polyploidy is probably not essential in determining species adaptation to alpine environments [34], we found that alpine Tripleurospermum taxa are usually polyploid and present higher genome sizes. Consistently, species inhabiting ruderal areas present lower C-values and genome size is positively correlated with altitude. However, this correlation may not be true in all systems and the relationship between genome size, polyploidy and altitude is probably multifactorial. For example, Mas de Xaxars et al. [34] found mostly diploid species in the clade comprising high mountain Artemisia. Indeed, our previous work on Tripleurospermum's genome size variation [17] detected a negative correlation with altitude, although the sampling was much more reduced.
Some of the taxa here studied have a wide distribution, and are even considered as invasive (e.g. the weed T. maritimum), while some others have a more restricted geographical range, being endemic to very narrow areas (e.g. T. baytopianum, T. corymbosum, T. heterolepis). Although we failed to find significant differences, species with a wider distribution tend to present smaller genome sizes than those from restricted areas. Similar findings were reported previously in other groups [67]. There is evidence that a small genome size could be related to invasiveness [69,88] and hence a wider distribution, although the reasons of this relationship are badly understood. Perhaps, a small(er) genome size would favor faster generation times that may contribute to wider geographical distributions.

Concluding remarks
Molecular phylogenetic data, not always in agreement with classical classifications, have proved themselves here useful to confirm the monophyly of both genera addressed, Tripleurospermum and Matricaria. This was already suggested by morphological traits, irrespective of the fact that several taxa have been the object of nomenclatural combinations under both genera. We have also shown here that the study of genome size variation within a genus, or closely related genera, can provide interesting hypotheses to understand certain morphological or ecological traits observed in plants, such as the results obtained on e.g. higher genome sizes in rhizomatous taxa or absence of slime in polyploids. Our interpretations may be backed, and may be applied in a more general scope, if similar studies in other genera raise similar conclusions, but such works are still scanty in the literature.
Supporting information S1 File. DNA sequence matrix of ITS+ETS rDNA regions for the studied species. (FAS) S1 Table. Data used for the statistical analyses together with source references. (DOCX)