Descending Dysploidy and Bidirectional Changes in Genome Size Accompanied Crepis (Asteraceae) Evolution

The evolution of the karyotype and genome size was examined in species of Crepis sensu lato. The phylogenetic relationships, inferred from the plastid and nrITS DNA sequences, were used as a framework to infer the patterns of karyotype evolution. Five different base chromosome numbers (x = 3, 4, 5, 6, and 11) were observed. A phylogenetic analysis of the evolution of the chromosome numbers allowed the inference of x = 6 as the ancestral state and the descending dysploidy as the major direction of the chromosome base number evolution. The derived base chromosome numbers (x = 5, 4, and 3) were found to have originated independently and recurrently in the different lineages of the genus. A few independent events of increases in karyotype asymmetry were inferred to have accompanied the karyotype evolution in Crepis. The genome sizes of 33 Crepis species differed seven-fold and the ancestral genome size was reconstructed to be 1C = 3.44 pg. Both decreases and increases in the genome size were inferred to have occurred within and between the lineages. The data suggest that, in addition to dysploidy, the amplification/elimination of various repetitive DNAs was likely involved in the genome and taxa differentiation in the genus.


Introduction
Chromosomal changes, both numerical and structural, are acknowledged to be important mechanisms that accompany speciation and diversification in plants [1]. The large variation in chromosome numbers in the plant kingdom is the result of two major mechanisms-polyploidy and dysploidy [2]. Polyploidisation (whole genome duplication) and subsequent diploidisation seem to play a greater role in the evolution of angiosperms than in other eukaryotes [3]. There is evidence that most, if not all, angiosperms are ancient polyploids [4]. Dysploidy involves various types of structural rearrangements of chromosomes, which often result in decreases or increases in chromosome numbers [5,6]. The chromosome morphology can also be altered by the amplification/loss of DNA sequences [7]. Accumulation and/or loss of DNA sequences (mainly repetitive DNA) are mechanisms that lead to changes in the total karyotype length and genome size [8,9].
Crepis sensu lato (s.l.), which consists of approximately 200 annual and perennial species, is one of the largest genera in Asteraceae [10]. Most Crepis species are diploid with relatively low numbers of rather large and well-differentiated chromosomes. Several chromosome base numbers (x = 3, 4, 5, 6, and 11 [11]) and significant differences in genome size from 0.72 pg/1C to 32.75 pg/1C [12] have been reported in this genus previously. Thus, Crepis has been an excellent system for investigating the evolution of the karyotype [13,14]. The first cytogenetic and taxonomical study of numerous Crepis species defined several Table 1. Species names, collection numbers, places of seed collecting, voucher numbers of the analysed taxa, and the GenBank accessions numbers of the sequences obtained in this study.

DNA Amplification and Sequencing
Total genomic DNA was isolated from fresh leaf tissue using a modified CTAB method [24]. Genomic DNA from each sample was checked for quality and quantity using an ND-1000 (peqLab, Erlangen, Germany). The PCR amplification of the nuclear rDNA ITS (internal transcribed spacer; nrITS) region was carried out according to Venora et al. [25] using the primers ITS4 and ITS5 [26]. Based on the analyses of Shaw et al. [27], the four non-coding plastid DNA spacer regions (rpl32-trnL, rps16-trnK, ndhC-trnV and psbD-trnT), which have a high number of potentially informative characters in euasterids II, were selected for the analyses. The cpDNA markers were amplified using the primers from Shaw et al. ( [28]; Table S1). The PCR mixture for the amplification contained a 1x Color OptiTaq PCR Master Mix (EURx, Gdansk, Poland), 0.5 µM of each forward and reverse primer (Genomed, Warsaw, Poland), and 50 ng of the DNA template. The GeneAmpPCR System 9700 (Applied Biosystems, Waltham, MA, USA) and the conditions described in Blöch et al. [29] were used for the amplification. The PCR products were treated with E. coli Exonuclease I and FastAP Thermosensitive Alkaline Phosphatase (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions and the cycle sequencing was performed using a 3730xl DNA Analyzer (Applied Biosystems; Waltham, MA, USA) in a commercial facility (Genomed; Warsaw, Poland or Macrogen; Seoul, Korea). The template DNAs that were used for the nrITS and cpDNA amplifications were always derived from the same isolate. All of the sequences were deposited in GenBank (accession numbers in Table 1).

Sequence Alignment and Phylogenetic Analyses
Most of the phylogenetic analyses were performed using the DNA sequences of four concatenated cpDNA regions and the nrITS that was obtained in this study. Additionally, an extended nrITS dataset, which contained the sequences that were obtained in this study as well as previously published sequences were also analysed [12]. The sequences were assembled using DNA Baser version 3 (Heracle BioSoft S.R.L., Pitesti, Romania). Multiple sequence alignments were performed 20 times using webPRANK [30] and MergeAlign [31] was then used to obtain a consensus of the multiple sequence alignments. The phylogenetic relationships were inferred independently for the nrITS and for the four concatenated plastid regions using the maximum likelihood, as implemented in IQ-TREE version 0.9.5 [32]. The best model of sequence evolution for the ML analyses was determined using the Bayesian information criterion as implemented in IQ-TREE [33]. The best fit models were TIM3e + G4 for nrITS and TVM + F + G4 for the plastid data sets. The significance of the inferred relationships was assessed via bootstrapping with 1000 replicates. Picris hieracioides, Lactuca serriola, and Sonchus oleraceus were used as the outgroup taxa. The resulting phylogenetic trees were created using FigTree v.1.3.1 [34]. Bootstrap support values below 75 were excluded from the figures.

Chromosome Preparation and Karyotype Analyses
Young leaves were pretreated with 2 mM 8-hydroxyquinoline (Sigma, Burlington, MA, USA) for 2 h at room temperature and for 2 h at 4 • C, fixed in methanol:glacial acetic (3:1) and stored at −20 • C until use. The mitotic metaphase chromosome spreads were prepared according to Dydak et al. [35]. The chromosome preparations were stained with DAPI (4 ,6-diamidino-2-phenylindole) and analysed under a Zeiss AxioImager.Z.2 fluorescent microscope (ZEISS, Germany). The karyotype analyses were performed on at least ten high-quality mitotic metaphase spreads per accession. The chromosome length was measured using ImageJ software ver. 1.50 [36]. The chromosomes of each analysed cell were arranged into the karyotypes by decreasing length. The chromosome nomenclature of Levan et al. [37] was used. The degree of karyotype asymmetry was estimated according to Paszko [38]. The asymmetry index is an indicator of the levels of the heterogeneity of chromosome length and the positions of the centromeres in a given karyotype. The asymmetry index (A 1 ) was calculated using the following equation: where CV CL is the relative variation in chromosome length; CV CL = S CL X CL × 100 (S CL is a standard deviation and X CL is the mean chromosome length), whereas CV CI is the relative variation centromeric index CV CI = S CI X CI × 100 (S CI is the standard deviation and X CI is the mean centromeric index).

Genome Size Measurements
The genome sizes were measured using flow cytometry. Brachypodium hybridum Catalán, Joch.Müll., Hasterok & G. Jenkins ABR113 (2C DNA = 1.265 pg; [39]); Solanum lycopersicum L., Stupicke' (1.96 pg/2C DNA; [40]; Solanum pseudocapsicum L. (2.58 pg/2C DNA, [41]); Zea mays L. 'CE-777' (2C = 5.43 pg/2C DNA, [42]); Pisum sativum L. 'Ctirad' (9.09 pg/2C DNA, [43]); and Secale cereale subsp. cereale L. (16.01 pg/2C DNA, [43]) were used as the internal standards, depending on the genome sizes of the measured samples. The youngest fully developed leaves of the analysed Crepis species and the appropriate internal standard (the information about the standards were added to Table 2 in results section) were chopped together in a Petri dish in 500 µL of a nuclei extraction buffer using a razor blade (CyStain PI Precision P Sysmex 05-5022). The nuclei suspension was filtered through a 30 µm mesh (CellTrics, Sysmex, Kobe, Japan) and stained with a staining buffer containing propidium iodide, RNase (CyStain PI Sysmex Precision P 05-5022, Kobe, Japan) and 1% β-mercaptoethanol (Sigma, Burlington, MA, USA) according to the manufacturer's instructions. The samples were then analysed using a flow cytometer (CyFlow Space, Sysmex, Kobe, Japan) equipped with a 532 nm green laser. At least 10,000 nuclei were analysed for each sample. The sizes of the nuclear genomes were calculated as the linear relationship between the ratio of the 2C DNA peaks of a sample and the standard. Due to the high content of secondary metabolites, which prevented a flow cytometry analysis, the genome size could not be estimated for 12 species. The accepted CVs were less than 5%, except for C. kotschyana (5.78%) and diploid C. vesicaria (5.93%), for which measurements with CVs lower than 6% were accepted. The 1Cx-value and DNA content of one nonreplicated monoploid genome with chromosome number x were calculated according to Greilhuber et al. [44].

Ancestral State Reconstructions
The phylograms that resulted from the ML analysis (branch length information included) were used to infer the evolution of the chromosome numbers, the karyotype asymmetry, and genome size. The analyses were performed using the better supported cpDNA phylogram. Previously published genome size values for Crepis (Tables S2 and S3) [12] were added to the newly obtained results and together mapped on the phylogram resulting from the ML analysis of the extended nrITS dataset. The maximum likelihood analyses were performed under the CONST_RATE model (for the nrITS data set) or the CONST_RATE_DEMI model (for the cpDNA dataset), as implemented in ChromEvol 2.0. software [45]. For the ChromEvol analyses, the best-fit model was tested using an AIC test (Table S4). For the best-fitted model, the analyses were rerun with parameters that were fixed to those that were optimised in the first run using 10,000 simulations to compute the expected number of changes along each branch as well as the ancestral haploid chromosome numbers at the nodes. Picris hieracioides, Sonchus oleraceus, and Lactuca serriola were used to root the tree. Bootstrap support values less than 75 were excluded. The chromosome base numbers for Crepis pontana, C. acuminata, C. atribarba, C. modocensis, and C. intermedia were obtained from the literature [20,46] because of problems with obtaining good-quality meristematic tissue for the chromosomal analyses. The base chromosome number for the polyploid C. biennis (x = 5) was adopted from [47]. The evolution of the asymmetry index and genome sizes were analysed using maximum likelihood as implemented in the package phytools in the R software [48].

Phylogenetic Analysis
The phylogenetic relationships of the 45 Crepis species and Lapsana communis were inferred from analyses of their nrITS sequences and four plastid regions (rpl32-trnL, ndhC-trnV, rps16-trnK and trnT-psbD). A total of 61 nrITS and 244 cpDNA sequences (61 sequences per each of the four chloroplast markers) were analysed. Three species, Picris hieracioides, Lactuca serriola, and Sonchus oleraceus, were used as the outgroups. The length of the isolated nrITS regions among the analysed species ranged from 540 to 598 bp. The final alignment was 668 bp long (including gaps) with 241 Characters that were parsimony informative. Only one ribotype of the nrITS was amplified for all of the studied accessions. The total length of the analysed concatenated plastid DNA regions ranged from 2948 to 3577 bp (the data for the individual chloroplast markers are listed in Table S5). The concatenated alignment of the cpDNA was 5693 bp long (including gaps) with 720 characters that were parsimony informative.
Two major well-supported clades (nrITS BS100 and cpDNA BS100) corresponding to the evolutionary lineages of Lagoseris and Crepis s.s. (Figure 1) were consistently recovered in both the nrITS and cpDNA ML analyses. Crepis s.s., which contained most of the analysed species, was monophyletic in both of the nrITS and cpDNA analyses; however, the nrITS-and cpDNA-tree topologies differed in their branching patterns. The clades that were recovered in the cpDNA marker analyses were labelled with Roman numerals (clades I-IV) and the subclades with Roman numerals and lower-case letters (subclades IVa-IVc). The clades in the nrITS phylogenetic analyses were labelled with Arabic numerals (clades 1-4) and the subclades were labelled with Arabic numerals and lower-case letters (subclades 4a-4c). Four clades were distinguished in Crepis s.s. in all of the analyses. Clade 1 (BS81), which was recovered in the nrITS data analysis, included five perennial European species and the central Asian C. lyrata. This group of species was monophyletic in the nrITS phylogeny, but not in the cpDNA analysis in which the species were placed in two highly supported clades instead: clade I (BS99) and clade II (BS100; Figure 1). The second monophyletic clade 2 (BS96), which was recovered in the nrITS analysis, comprised nine, mostly annual Crepis species, with the exception of the perennial C. sibirica. These species were placed in subclade IVb (BS97) in the cpDNA tree together with C. setosa (Figure 1). Six European and/or Middle East perennial species, which were recovered in clade 3 (BS96) of the nrITS tree, were placed in subclade IVa (BS100) in the cpDNA phylogeny. Nearly half of the analysed Crepis s.s. species were placed in clade 4 in the nrITS phylogeny. Three highly supported subclades could be distinguished within this clade: (i) subclade 4a (BS90), which consisted of species that are native to North America; (ii) subclade 4b (BS86), which comprised C. biennis, C. setosa, and C. nicaeensis; and (iii) subclade 4c, which comprised both annual and perennial species from Europe and North Africa (from Algeria to Morocco). However, in the cpDNA phylogeny, the North American species together with C. tectorum, C. nigrescens, C. nicaeensis, and C. biennis formed the monophyletic clade III (BS100).
Three main evolutionary lineages were distinguished in the phylogenetic tree that resulted from the analysis of the newly obtained nrITS sequences and previously published data (species of Askellia. Lagoseris and Crepis s.s.; Figure S1; Table S3; [22]). Among the species of Crepis s.s., five main clades were identified, in congruence with earlier reports [22,23].

Chromosome Number
The chromosome numbers were analysed for 50 accessions (Table 2) representing 40 Crepis species and three accessions of L. communis. Most of the analysed Crepis species were diploids and only four polyploids were found, C. biennis (2n = 8x = 40), C. occidentalis (2n = 4x = 44), and C. vesicaria, which were represented by two polyploid accessions (both 2n = 4x = 16) and one diploid accession (2n = 2x = 8; Table 2; Figure S2). Five different chromosome base numbers were found among the diploid Crepis species: x = 3, 4, 5, 6, and 11 (Table 2; Figures 2,3 and S4). The most common chromosome base numbers were x = 4 found in 20 species and x = 5 found in ten species. Six species had x = 6 and only two species had a chromosome base number of x = 3 (C. capillaris and C. zacintha). The chromosome base number of x = 11 was found in C. occidentalis, a representative of the American Crepis agamic complex (2n = 4x = 44; Figure S5). L. communis from the Lagoseris clade was diploid with 2n = 14. Three main evolutionary lineages were distinguished in the phylogenetic tree that resulted from the analysis of the newly obtained nrITS sequences and previously published data (species of Askellia. Lagoseris and Crepis s.s.; Figure S1; Table S3; [22]). Among the species of Crepis s.s., five main clades were identified, in congruence with earlier reports [22,23]. chromosome base numbers were found among the diploid Crepis species: x = 3, 4, 5, 6, and 11 (Table 2; Figures 2, 3 and S4). The most common chromosome base numbers were x = 4 found in 20 species and x = 5 found in ten species. Six species had x = 6 and only two species had a chromosome base number of x = 3 (C. capillaris and C. zacintha). The chromosome base number of x = 11 was found in C. occidentalis, a representative of the American Crepis agamic complex (2n = 4x = 44; Figure S5). L. communis from the Lagoseris clade was diploid with 2n = 14.   The ML analysis (ChromEvol 2.0) was based on the cpDNA due to better support obtained for this marker. The ancestral chromosome base number for the common ancestor of Lagoseris and Crepis s.s. was inferred as x = 6 (pp = 0.97; Figure 4). The analysis of the cpDNA data sets suggested fourteen reductions (with expectation above 0.5) in the chromosome base numbers (Figure 4), three in the Lagoseris group and eleven in Crepis s.s. The ancestral base numbers for clades I and II was reconstructed as x = 6. The chromosome base number of x = 5 was inferred for the common ancestor of the species in clades III and IV (Figure 4). There was a further decrease in the chromosome base number from x = 5 to x = 4 in the evolutionary lineage of C. tectorum and C. nigrescens and in the evolutionary lineage of C. nicaeensis in clade III (Figure 4). A reduction in the chromosome base number from x = 5 to x = 4 accompanied the diversification of the common ancestor of the species in subclade IVa (Figure 4). Several events of descending dysploidy were also inferred for subclades IVb and IVc: (i) from x = 5 to x = 4 in the C. kotschyana lineage and C. setosa lineages; independently in the group that consisted of C. taraxacifolia, C. vesicaria, and C. The ML analysis (ChromEvol 2.0) was based on the cpDNA due to better support obtained for this marker. The ancestral chromosome base number for the common ancestor of Lagoseris and Crepis s.s. was inferred as x = 6 (pp = 0.97; Figure 4). The analysis of the cpDNA data sets suggested fourteen reductions (with expectation above 0.5) in the chromosome base numbers (Figure 4), three in the Lagoseris group and eleven in Crepis s.s. The ancestral base numbers for clades I and II was reconstructed as x = 6. The chromosome base number of x = 5 was inferred for the common ancestor of the species in clades III and IV (Figure 4). There was a further decrease in the chromosome base number from x = 5 to x = 4 in the evolutionary lineage of C. tectorum and C. nigrescens and in the evolutionary lineage of C. nicaeensis in clade III (Figure 4). A reduction in the chromosome base number from x = 5 to x = 4 accompanied the diversification of the common ancestor of the species in subclade IVa (Figure 4). Several events of descending dysploidy were also inferred for subclades IVb and IVc: (i) from x = 5 to x = 4 in the C. kotschyana lineage and C. setosa lineages; independently in the group that consisted of C. taraxacifolia, C. vesicaria, and C. polymorpha; in the lineages of C. aculeata and C. aspera; as well as for the common ancestor of the group, consisting of C. alpestris, C. oporinoides, C. pyrenaica, and C. capillaris; (ii) from x = 5 to x = 3 in C. zacintha; and (iii) from x = 4 to x = 3 in the C. capillaris lineage (Figure 4).  The chromosome base numbers have been mapped on the ML tree of concatenated cpDNA sequences using the maximum likelihood method implemented in ChromEvol 2.0 software [45]. The tree was rooted with Picris hieracioides, Lactuca serriola, and Sonchus oleraceus.

Karyotype Structure
A morphometric analysis of the chromosomes was performed for 47 accessions representing 38 Crepis species and three accessions of L. communis. The majority of the analysed species (38 accessions of 31 species) had symmetrical karyotypes ( Table 2). Only nine of the analysed Crepis species had karyotypes with a higher asymmetry index ( Table  2). The asymmetry index (AI) values were mapped onto the ML phylogenetic tree using the maximum likelihood ( Figures 5 and S6). The analysis was performed using better supported the cpDNA phylogenetic trees. The symmetrical karyotype was inferred as the ancestral state for the common ancestor of Lagoseris and Crepis s.s. (Figure 5). In the evolutionary lineage of Lagoseris, an increase in the karyotype asymmetry accompanied the speciation of C. praemorsa and L. communis. An increase in the karyotype asymmetry accompanied also the evolution of several Crepis s.s. species from subclades IVb in the cpDNA tree (C. alpina, C. sibirica, C. syriaca, C. rubra, and C. zacintha) and from subclade IVc (C. oporinoides, C. pyrenaica, and C. capillaris; Figure 5). The chromosome base numbers have been mapped on the ML tree of concatenated cpDNA sequences using the maximum likelihood method implemented in ChromEvol 2.0 software [45]. The tree was rooted with Picris hieracioides, Lactuca serriola, and Sonchus oleraceus.

Karyotype Structure
A morphometric analysis of the chromosomes was performed for 47 accessions representing 38 Crepis species and three accessions of L. communis. The majority of the analysed species (38 accessions of 31 species) had symmetrical karyotypes ( Table 2). Only nine of the analysed Crepis species had karyotypes with a higher asymmetry index ( Table 2). The asymmetry index (AI) values were mapped onto the ML phylogenetic tree using the maximum likelihood (Figures 5 and S6). The analysis was performed using better supported the cpDNA phylogenetic trees. The symmetrical karyotype was inferred as the ancestral state for the common ancestor of Lagoseris and Crepis s.s. (Figure 5). In the evolutionary lineage of Lagoseris, an increase in the karyotype asymmetry accompanied the speciation of C. praemorsa and L. communis. An increase in the karyotype asymmetry accompanied also the evolution of several Crepis s.s. species from subclades IVb in the cpDNA tree (C. alpina, C. sibirica, C. syriaca, C. rubra, and C. zacintha) and from subclade IVc (C. oporinoides, C. pyrenaica, and C. capillaris; Figure 5).

Figure 5.
Karyotype evolution in the analysed species of Crepis s.l. Genome sizes were mapped on the ML tree of concatenated cpDNA sequences using the maximum likelihood method implemented in Phytools. Karyotype formulas (symbols preceding the species names) and the increases in karyotype asymmetry (black arrows) are indicated for all species. The genome sizes, karyotype formulas, and asymmetry indexes of each species are presented in Table 2. The figure includes only species for which all three datatypes (chromosome number, karyotype formula, and asymmetry index) were provided. The tree was rooted with Picris hieracioides, Lactuca serriola, and Sonchus oleraceus.
Ten of the analysed species with a chromosome base number of x = 5 had five different karyotype formulas. The karyotypes of most of the species consisted of Figure 5. Karyotype evolution in the analysed species of Crepis s.l. Genome sizes were mapped on the ML tree of concatenated cpDNA sequences using the maximum likelihood method implemented in Phytools. Karyotype formulas (symbols preceding the species names) and the increases in karyotype asymmetry (black arrows) are indicated for all species. The genome sizes, karyotype formulas, and asymmetry indexes of each species are presented in Table 2. The figure includes only species for which all three datatypes (chromosome number, karyotype formula, and asymmetry index) were provided. The tree was rooted with Picris hieracioides, Lactuca serriola, and Sonchus oleraceus.

Genome Size
The genome sizes (1C values) of the 33 Crepis species and for L. communis were measured ( Table 2). The genome sizes varied nearly seven-fold among the diploid species, ranging from 1.03 pg/1C DNA (C. zacintha) to 7.46 pg/1C DNA (C. lacera). The genome sizes of the species of the Lagoseris evolutionary lineage were variable with nearly a six-fold difference (1.22 pg/1C DNA to 7.05 pg/1C DNA). In Crepis s.s., larger genome sizes were found in two species from clade I (mean 1C value = 4.78 pg), whereas clade II species had smaller genome sizes (mean 1C value = 2.44 pg). In clade III, two diploid species C. tectorum (3.06 pg/1C DNA) and C. nicaeensis (3.17 pg/1C DNA) had small genomes, whereas the octoploid C. biennis had a larger genome (10.45 pg/1C DNA). The Cx-value (the C-value of monoploid genome "x") of C. biennis was rather small (1C x = 2.61 pg/1C DNA; Figures 5 and S7) compared to the diploid species of this group. Three different groups of related species that differed in the patterns of their genome sizes distribution were identified in clade IV. The subclade IVa consisted of diploid species with larger genome sizes (mean 1C value = 6.35 pg) and, among them, C. lacera had the largest estimated genome size (1C value = 7.46 pg). The genome sizes of the species from clade IVc were on average 2.52 pg/1C DNA. The largest variation in genome sizes was observed among the species in subclade IVb in which the genome size of C. sibirica (1C value = 6.98 pg) was nearly seven times larger than that of C. zacintha (1C value = 1.03 pg; Table 2; Figures 5 and S7). C. vesicaria contained both diploid and tetraploid accessions (2n = 8 and 2n = 16). The genome sizes that were estimated for the diploid and tetraploid accessions (2.43 pg/1C and 2.78 pg/1C, respectively).
The DNA amounts were mapped onto the cpDNA ML phylogenetic tree using the maximum likelihood method (Figures 5 and S7). A genome size of 3.2 pg/1C DNA was inferred as the ancestral state for the Lagoseris lineage. Both decreases and increases in genome sizes were inferred to have accompanied speciation and diversification of this clade (Figures 5 and S7). The ancestral genome size for the entire Crepis s.s. lineage was reconstructed as 3.70 pg/1C. For clade I, 4.8 pg/1C was reconstructed as the ancestral state, whereas 2.50 pg/1C was recovered as the ancestral genome size for clade II. For the entire clade IV, 3.2 pg/1C was reconstructed as the ancestral genome size. For subclade IVa, the ancestral genome size was reconstructed to be 5.5 pg/1C, whereas for subclades IVb and IVc, 2.7 pg/1C and 2.9 pg/1C, respectively, were recovered. An increase in genome size was inferred for the species of clade I and the octoploid C. biennis (clade III). In contrast, a reduction in genome size accompanied the evolution of many species of clade II and subclades IVb and IVc (e.g., C. aurea and C. zacintha). The evolution of a common ancestor of subclade IVa was accompanied by an increase in the genome size, with further increases or decreases in the genome sizes inferred during the diversification and speciation of this clade.
The inferences of genome size evolution were additionally performed using nrITS tree based on an extended data set containing both newly obtained genome size data and previously reported genome size estimations ( Figure S8; Table S2). The extended data set included the species from the American agamic complex, which had relatively large genome sizes (Table S3). The observed values of the reconstructed ancestral states were slightly larger in comparison to the results of the analyses of the newly obtained data only. The inferred events of increases and reductions in genome size were very similar in both data sets.

Discussion
Resolving the phylogenetic relationships for Crepis provided the evolutionary framework for elucidating the evolution of the chromosome numbers and karyotype structure. Phylogeny based on the biparentally inherited nrITS of mostly diploid species of Crepis was found to be largely congruent with previously published results ( Figure S1; [22,23]). In the present study, a new clade comprising species of the North American Crepis agamic complex was identified. These species have not been included in previous analyses. The phylogenetic relationships inferred in this study were inconsistent with the sections defined by Babcock [20], in agreement with the earlier phylogenetic analyses [22]. Each clade and subclade contained species previously classified in different Babcock sections ( Figure S1). A phylogenetic analysis, based on the chloroplast markers, recovered mostly the same main groups of species as the nrITS-based analyses. However, the relationships among these clades were found to be incongruent between the nuclear and plastid markers. Such nucleo-cytoplasmic discordance has been well documented in previous studies of higher plants [49]. This incongruence could be explained by the convergent evolution of shared chloroplast sequences, incomplete lineage sorting of the ancestral polymorphisms, or introgressive hybridization [50,51]. The Crepis species analysed here were mostly diploids, but because Crepis species are primarily cross-pollinated and many of them hybridise extensively [20], introgression could have accompanied the speciation of individual species, as also suggested for other plant groups [52,53]. The incongruence could be also explained by technical reasons. Several nodes in nrITS phylogram in clade 4 were weakly supported and the analysed species represent only a quarter of the Crepis genome, which may influence the results. The incongruence between the nrITS and cpDNA trees concerning the position of C. biennis could suggest its allopolyploid origin. C. biennis is an octoploid species [47] whose origin and parental species are not yet known. The present analysis suggests a putative ancestral species for the two subgenomes of this polyploid taxon, species that are similar to the extant C. nicaeensis/C. setosa and another species related to C. tectorum (Figure 1). Additional molecular phylogenetic and cytogenetic studies are necessary to elucidate the origin of C. biennis.
This study reports the genome sizes for 33 species of Crepis s.l., eight of them for the first time. The genome sizes that were reported in the current study are largely in agreement with previous reports, except for two species: C. vesicaria and C. pannonica (Table 2; [12,54,55]). This discrepancy is probably the result of methodological factors as the previous data were obtained using Feulgen densitometry or flow cytometry but with different standards. In the case of the diploid C. vesicaria, the discrepancy (2.43 pg/1C in this study versus 4.18 reported by Enke et al. [12]) could result from the presence of the diploid and polyploid cytotypes within this species [20]. A strong correlation between genome size and life form has been suggested for Crepis in previous studies [12]. The current results, however, did not show an obvious correlation between these two characters. Both annual and perennial Crepis species groups showed high levels of variation of their genome sizes. Both one of the smallest (C. leontodontoides 1.06 pg/1C) and the largest genomes (C. lacera; 7.46 pg/1C) were found in the perennial diploid species. The correlation between genome size and plant life form has been extensively discussed in the literature. The annual life form, especially in weedy species, was inferred to be associated with a smaller genome size compared to its related perennials [56][57][58]. Such a correlation, however, must be interpreted with caution because other ecological and evolutionary factors might also influence genome size evolution [59]. The chromosome numbers of Crepis species analysed in the current study were in accordance with earlier reports [20,60]. Most of the analysed species were diploids, with only a few polyploids [20]. The tetraploids of C. vesicaria shared the same karyotype structure with its diploid accession, which suggests that they were of autopolyploid origin Analyses of chromosome number and genome size evolution in a phylogenetic framework revealed that there was no general trend in their evolution among angiosperms [61][62][63]. Both chromosome numbers and genome sizes could change very rapidly during the evolution and diversification of plant genomes [64][65][66]. A reduction in genome size was earlier suggested to accompany the evolution of the genus Crepis [12,67], with increases in genome sizes mainly correlated with polyploidisation [20]. However, the present data revealed a bidirectional evolution of genome size in both the Lagoseris and Crepis s.s. evolutionary lineages. The current study revealed that several diploid species have a genome size that is much larger than the ancestral reconstructed states of the lineages they belong to, e.g., C. pannonica or C. pulchra. This suggested a bidirectional evolution of genome size in at least some of the Crepis lineages. The proliferation of repetitive DNA sequences is considered to be the major cause of the increase in genome size next to polyploidy [7,24,68]. The mechanisms of the reduction of genome sizes and DNA removal are less understood, with an unequal homologous recombination or illegitimate recombination proposed to be most important [69,70]. The information about the genome structure in Crepis is still very limited and only further analyses, especially of repetitive DNA fraction, will allow better insight into the mechanisms of the evolution of genome size in this genus.
The ancestral states of the base chromosome number in a common ancestor of Lagoseris and Crepis s.s. was inferred to be x = 6, and descending disploidy was inferred as a main trend in the chromosome number evolution in Crepis (Figure 4). The obtained results are partly in agreement with the hypothesis of Babcock [20], who suggested that a decrease in the chromosome base number is the main trend in Crepis evolution. However, the derived base chromosome numbers (x = 5, 4, and 3) were found to have evolved independently several times in the evolutionary history of the genus. The chromosome base number of x = 5 has evolved at least three times during Crepis evolution. Dysploid change from x = 5 to x = 4 has occurred twice in Lagoseris and eight times in Crepis s.s. The base chromosome x = 3 has evolved twice in Crepis s.s. Nearly half of the descending dysploidy events were recovered for the same branches for which changes in genome sizes were inferred ( Figure 5). Some descending dysploidy events were accompanied by a reduction in genome sizes whereas others were accompanied by genome size increases and no common trend was observed. Similar patterns have been reported for Artemisia (Asteraceae; [64]) and other plant genera (e.g., Genlisea, Lentibulariaceae; [65]).
The chromosome base number was inferred to have increased only twice during the evolution of the analysed species. An allopolyploidisation or a duplication of the chromosome number followed by dysploidy is likely to have given rise to the base chromosome of x = 11 present in the American agamic complex. The evolution of this lineage was accompanied by an increase in genome size ( Figure S8), supporting the hypothesis of the polyploidy origin of this new base chromosome number of x = 11. In contrast, the increase in base chromosome number in the evolutionary lineage of Lapsana (from x = 6 to x = 7) was accompanied by the genome size downsizing. The chromEvol analyses inferred that the changes in chromosome number in the Lapsana lineage resulted from demiduplication. For demiduplication, a genome increase instead of decrease would be expected. Unfortunately, the direct ancestor of Lapsana and its genome size are unknown and thus this hypothesis cannot be tested. More detailed comparative analyses of the karyotype structure of Lapsana and related taxa are necessary to gain insight into the origin of base chromosome number of x = 7 in the Lagoseris lineage. Changes in the chromosome base numbers as well as in genome sizes in Crepis were most prevalent at the tips of the tree rather than early in the evolution of the genus. Similar trends have been reported for Trifolium, Hypochaeris, or Braschyscome [71][72][73].
Chromosomal rearrangements often accompany plant diversification and speciation [74][75][76]. Chromosome rearrangements generally have a low level of homoplasy. Therefore, it was hypothesised that species with the same chromosome number that evolved independently should differ in their chromosome structure [1]. Molecular cytogenetic methods and the availability of many chromosomal markers enables the comparative analysis of the karyotype in model or cultivated taxa (e.g., Brassicaceae and Brachypodium [77][78][79]. In wild species groups it is more challenging, but even comparative analyses of the morphology of chromosomes are indicative of occurrences of chromosomal rearrangements. However, the shape and size of the chromosomes can also be altered by a sequence amplification/loss [1,[7][8][9]. A comparative analysis of the Crepis karyotype formulas revealed that species with the same chromosome base number belonging to different evolutionary lineages differed in their karyotype morphologies ( Figure 5). Moreover, even the species in the same subclades (e.g., subclade IVc) possessing the same chromosome base number differed in karyotype structure ( Figure 5). These results supported the hypothesis that the derived base chromosome numbers (x = 5, 4, 3) might have evolved independently multiple times in Crepis. The changes in karyotype structure might lead to an increase in its asymmetry [38]. An increase in karyotype asymmetry was previously suggested as the main pattern of karyotype evolution in the genus Crepis [80]. The results of the present study did not support this hypothesis. An increase in karyotype asymmetry accompanied the speciation of only a few species in the different evolutionary lineages ( Figure 5). The increases in karyotype asymmetry accompanied both decreases and an increase in the chromosome base numbers ( Figure 5), as well as increases or decreases in the genome sizes. Some groups of closely related Crepis species had the same chromosome number and a similar karyotype structure (e.g., subclade IVa), which suggests that different mechanisms, e.g., an amplification of repetitive sequences, might have accompanied the diversification of these species.

Conclusions
Analyses of karyotype structure and genome sizes in a phylogenetic context enabled inferences about the ancestral base chromosome numbers and genome sizes in the Lagoseris and Crepis s.s. evolutionary lineages. A few events of base chromosome number changes, mostly descending dysploidy, explained the chromosome number distribution in the extant Crepis s.l. species best. Most of these changes accompanied the evolution of individual species or a small group of closely related, derived species. Analyses of the genome size evolution revealed both increases and decreases during the evolution of the genus. The changes in genome sizes occurred mainly at the tips rather than at the internal nodes of the tree. The present work contributes to a better understanding of the evolution of the genomes of Crepis and lays the foundation for more detailed comparative analyses of its karyotype structure and evolution, including repetitive sequences.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/genes12091436/s1, Figure S1: Maximum likelihood phylogenetic tree based on nrITS sequences incorporating of taxa analysed in the present study and available in GenBank accessions of other species from Crepis s.l. and related taxa. All the species analysed in this study are indicated in bold. Bootstrap support values are indicated at each node, Figure S2: Metaphase plates of Crepis species used to prepare the karyograms in Figure 2 Figure S9. Scale bar = 5 µm, Figure S5: Metaphase plate of Crepis occidentalis (A) and C. biennis (B), Figure S6: Ancestral character state reconstruction of the asymmetry indexes of karyotypes of analysed species of Crepis s.l. The asymmetry indexes have been mapped on the ML tree of concatenated cpDNA sequences using maximum likelihood method as implemented in phytools. The tree was rooted with Picris hieracioides, Lactuca serriola, and Sonchus oleraceus, Figure S7: Ancestral character state reconstruction of the genome size (1C) of species of Crepis s.l. The 1C DNA amounts have been mapped on the ML tree of cpDNA sequences using maximum likelihood reconstruction as implemented in phytools. Numbers next to the nodes indicate reconstructed ancestral 1C-values (in pg). The genome size of each species is presented in Table 2. The tree was rooted with Picris hieracioides, Lactuca serriola, and Sonchus oleraceus, Figure S8: Ancestral character state reconstruction of genome size (1C-value) of Crepis s.l. taxa analysed in the present study and available in literature. The analysis was conducted using maximum likelihood reconstruction as implemented in phytools, Figure S9: Metaphase plates of Crepis species used to prepare the karyograms in Figure S4 , Table S1: Sequences of primers used for PCR amplification and sequencing, Table S2: Species name and GenBank accessions numbers of the selected Crepis sequences from Enke et al. [22] used in this study, Table S3: Genome size of Crepis species analysed in present study and data retrieved from literature, Table S4: The ∆AIC scores and Akaike weights of each model tested in ChromEvol 2.0 software for nrITS and cpDNA data sets, Table S5: Length of analysed plastid DNA and nrITS (nuclear ITS: ITS1-5.8S rDNA-ITS2) data sets [81,82].  Data Availability Statement: All data generated or analysed during this study are available from the corresponding author on reasonable request.