Genome size variation at constant chromosome number is not correlated with repetitive DNA dynamism in Anacyclus (Asteraceae)

Abstract

Background and Aims

Changes in the amount of repetitive DNA (dispersed and tandem repeats) are considered the main contributors to genome size variation across plant species in the absence of polyploidy. However, the study of repeatome dynamism in groups showing contrasting genomic features and complex evolutionary histories is needed to determine whether other processes underlying genome size variation may have been overlooked. The main aim here was to elucidate which mechanism best explains genome size evolution in Anacyclus (Asteraceae).

Methods

Using data from Illumina sequencing, we analysed the repetitive DNA in all species of Anacyclus, a genus with a reticulate evolutionary history, which displays significant genome size and karyotype diversity albeit presenting a stable chromosome number.

Key Results

By reconstructing ancestral genome size values, we inferred independent episodes of genome size expansions and contractions during the evolution of the genus. However, analysis of the repeatome revealed a similar DNA repeat composition across species, both qualitative and quantitative. Using comparative methods to study repeatome dynamics in the genus, we found no evidence for repeat activity causing genome size variation among species.

Conclusions

Our results, combined with previous cytogenetic data, suggest that genome size differences in Anacyclus are probably related to chromosome rearrangements involving losses or gains of chromosome fragments, possibly associated with homoploid hybridization. These could represent balanced rearrangements that do not disrupt gene dosage in merged genomes, for example via chromosome segment exchanges.

INTRODUCTION

Genome size (GS) is a basic trait of organisms, and correlates with cytogenetic, ecophysiological, systematic and other biological or non-biological (e.g. geography) factors (Pellicer et al., 2018). It is also important for genomic studies as it provides baseline, but crucial, information to evaluate the possibilities of performing extensive or complete DNA sequencing of a given taxon (e.g. Arabidopsis Genome Initiative, 2000; Nystedt et al., 2013). Today GS data are available for more than 15 000 species of eukaryotic organisms (Pellicer et al., 2018). Only in land plants, GS spans up to an astonishing 2500-fold variation, from 0.122 Gbp/2C in Genlisea tuberosa (Fleischmann et al., 2014) to ~300 Gbp/2C in the monocot Paris japonica (Pellicer et al., 2010) and the fern Tmesipteris obliqua (Hidalgo et al., 2017). At lower taxonomic levels, however, the range of variation is not so wide, although there is contrasting dynamism across different families and genera. For instance, in the broadly studied family Asteraceae, GS ranges from 0.44 Gbp/2C (Conyza canadensis) to 64.02 Gbp/2C (Crepis barbigera) (Vitales et al., 2019a). Such a large diversity among plant species has been studied using phylogenetic, cytogenetic and genomic approaches (e.g. Lysak et al., 2008; Garcia et al., 2009; Dvorak et al., 2018). However, there is still much discussion about which are the most determinant evolutionary processes and mechanisms influencing plant GS (Bennetzen et al., 2005; Fedoroff, 2012).

Certainly, GS diversity in plants is partially related to changes in chromosome numbers (Bennett and Leitch, 2005), and polyploidy is recognized as one of the main drivers of GS diversity (Soltis et al., 2015). In the absence of the latter, repetitive DNA activity is the most relevant factor because differential expansion and contraction of DNA repeats accounts for most GS differences (Pellicer et al., 2018). Indeed, the bulk of the genome is made up of repetitive DNA, mostly transposable elements (TEs), which usually reach high copy numbers, thus contributing substantially to GS (Kejnovsky et al., 2012). Additionally, TE dynamism may affect gene positioning by duplication and shuffling of gene sequences or induce structural changes such as chromosomal rearrangements (Kidwell, 2002; Feschotte and Pritham, 2007; Ren et al., 2018). Given the abundance of TEs in genomes, including gene regions, it has been proposed that the activity of TEs can also contribute to plant adaptation and evolution (Lisch, 2013). Activation and subsequent proliferation of TEs may be also triggered by stressful events such as hybridization, in some cases contributing substantially to GS variation (e.g. Ungerer et al., 2006). The proliferation of TEs in hybrids has been proposed to be the result of various mechanisms: for example, a breakdown in the nuclear immune system due to genomic shock (Lisch, 2009), or a regulated response to stress that can be switched on and off (Grandbastien et al., 2005) as hypothesized by McClintock (1984). Nonetheless, the molecular basis and the factors behind this, as well as the tempo of these processes, remain poorly known.

Recently, approaches using high-throughput sequencing (HTS) techniques have been crucial to gain critical insights into our understanding of GS evolution, complementing previous cytogenetic, genomic and genetic tools (Weiss-Schneeweiss et al., 2015). A suitable genomic approach is similarity-based clustering of low-coverage genome sequencing reads (Novák et al., 2010), as it is remarkably efficient and cost-effective to study the evolution of repetitive DNA in non-model plant groups lacking reference genome information. Using this approach, most case studies have found that differential accumulation or deletion of a small number of highly abundant TE families can explain substantial GS differences among related taxa (e.g. Macas et al., 2015 in Fabeae; Harkess et al., 2016 in Asparagus; Mascagni et al., 2017 in Helianthus). Other studies using this approach reported that changes on GS were associated with joint proliferation or contraction of many TE lineages, all of them contributing similarly to GS variability among species (e.g. Novák et al., 2014 in Musaceae; Kelly et al., 2015 in Fritillaria). In addition, a few studies in different plant groups have shown considerable GS variation that is not attributable to TE activity, but to amplification of simple sequence repeats (e.g. Myburg et al., 2014 in Eucalyptus; Ågren et al., 2015 in Oenothera). In summary, most published studies using this methodology confirm the critical role of repeatome dynamics in driving GS evolution in the absence of polyploidy.

The genus Anacyclus (Asteraceae) comprises eight species from the Western Mediterranean region (Vitales et al., 2018), and represents an excellent case to study GS variation in an evolutionarily convoluted group. The interest of evolutionary biologists regarding Anacyclus dates back to the 1970s and 1980s, when a series of cytogenetic studies highlighted substantial GS variation from 9.37 Gbp/2C in Anacyclus homogamos to 15.69 Gbp/2C in A. radiatus (Nagl and Ehrendorfer, 1974; Schweizer and Ehrendorfer, 1976; Humphries, 1981), despite having a constant chromosome number of 2n = 18. Schweizer and Ehrendorfer (1976) also reported a large intra-individual diversity of Giemsa C-banded karyotypes, suggesting a high incidence of structural chromosome heterozygosity in the genus. Based on further cytogenetic and crossing experiments, Humphries (1981) suggested that speciation in Anacyclus had been accompanied by chromosome re-patterning triggered by a complex reticulate evolution. Agudo (2017) provided additional genetic, morphological and reproductive evidence for a substantial role of hybridization along the evolutionary history of this genus, and Agudo et al. (2019) reported a gradual GS variation in contact zones where hybridization currently occurs. Rosato et al. (2017, 2018) further explored the cytogenetic variability of the genus in all Anacyclus species with a thorough sampling of populations and individuals. Using fluorescence in situ hybridization (FISH), these authors reported outstanding diversity in the number and location of ribosomal DNA sites as well as interstitial telomeric-like repeats (ITRs) at both intra- and inter-specific levels, which could have been partly caused by the reticulate history of Anacyclus.

In this context, and based on previous phylogenetic and cytogenetic results, two main non-exclusive hypotheses can be invoked to explain GS variation in the genus. On the one hand, as shown above, hybridization events might have resulted in genomic shocks (McClintock, 1984) that activated the repeat machinery (Lisch, 2009) and led to changes in GS. On the other hand, hybridization may also lead to chromosomal reorganizations (e.g. Robertsonian translocations: Fuchs et al., 1995; Rieseberg, 2001; Lysák and Schubert, 2012), which could potentially result in certain chromosome segments from the parental genome donors being inserted or deleted in the resulting hybrid genome, and hence affecting their overall nuclear DNA content (Schubert, 2007). If the first hypothesis is true, one would expect to find that GS variation among species is related to the differential amplification and/or elimination of one or many repeats. In contrast, if chromosomal reorganizations are responsible for GS diversity, repeatome composition would probably remain stable, with no signature of amplification or removal processes associated with changes in DNA content. Thus, to elucidate which mechanism better explains GS evolution in the genus, here we (1) characterize the repeatome of Anacyclus species, (2) analyse the evolutionary dynamism of the most abundant repeats, and (3) integrate the results of our repeatome analyses with previous cytogenetic data and phylogenetic reconstructions to gain insight into the evolution of GS in the genus.

MATERIALS AND METHODS

Plant material, GS estimation and genomic DNA sequencing

Voucher and seed samples from eight Anacyclus species – representing the whole taxonomic diversity of the genus at the species level – as well as from Heliocauta atlantica, sister to the genus Anacyclus, were collected in the field (see Table 1 for collection details). Specimen vouchers are deposited at the Real Jardín Botánico, Madrid, herbarium (MA) and further accession information is available in Supplementary Data Table S1. Twenty seeds per species were sown in the glasshouse of the Botanical Institute of Barcelona and plants were cultivated until flowering. After pooling equal proportions of silica-gel dried leaves from five individuals per species, genomic DNA was extracted at the Beijing Genomics Institute (BGI; Shenzhen, China) using the phenol–chloroform method. The quality and concentration of the DNA products were assessed by agarose gel electrophoresis using a Qubit 4 fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). Paired-end libraries of the whole genome with an average insert size of 500 bp were prepared and sequenced on the Illumina HiSeq 4000 platform (Illumina, San Diego, CA, USA) at BGI. For each species, ~1.5–1.9 Gb of raw data (equivalent to ~0.15× of their respective GS) were generated with pair-end 125-nt read length. Complete replicates, including independent library preparations and sequencing, were performed for all samples. The raw sequencing data reported in this study were deposited in the NCBI short-read archive (SRA) under BioProject accession number PRJNA556380. The quality of the raw sequencing reads was assessed using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/).

Table 1.

Estimation of genome size and clustering details of Anacyclus species

Species	Code	Genome size 2C (Gbp)	Genome size SD	Individual clustering		Comparative clustering
				No. of reads	Coverage (×1C)	No. of reads (0.0065×)
Anacyclus clavatus (Desf.) Pers.	CLA	10.42	0.02	1 426 554	0.034×	271 062
Anacyclus homogamos (Maire) Humphries	HOM	10.16	0.02	838 817	0.021×	264 197
Anacyclus linearilobus Boiss. & Reuter	LIN	13.14	0.03	1 089 707	0.021×	341 498
Anacyclus maroccanus (Ball) Ball	MAR	9.62	0.03	928 410	0.024×	250 466
Anacyclus monanthos (L.) Thell.	MON	11.54	0.01	1 172 602	0.025×	299 796
Anacyclus pyrethrum (L.) Link	PYR	13.10	0.04	1 472 815	0.028×	340 481
Anacyclus radiatus Loisel.	RAD	12.34	0.02	1 376 587	0.028×	320 139
Anacyclus valentinus L.	VAL	8.22	0.01	9 96 776	0.030×	213 849
Heliocauta atlantica (Litard. & Maire) Humphries	HEL	9.92	0.03	1 858 037	0.047×	258 094

Nuclear DNA contents from the same sequenced specimens were estimated at the Jodrell Laboratory of the Royal Botanic Gardens, Kew (RBGK) with a CyFlowSL Partec flow cytometer (Partec GmbH, Göttingen, Germany) following the one-step protocol of Doležel et al. (2007) with minor modifications as described by Clark et al. (2016). Measurements were made using the internal standard Petroselinum crispum ‘Champion Moss Curled’ (4.401 Gbp/2C; Obermayer et al., 2002) and the ‘general purpose buffer’ (GPB; Loureiro et al., 2007) supplemented with 3 % polyvinylpyrrolidone, with two replicates for each individual.

Clustering analysis using RepeatExplorer

Repeat identification by similarity-based clustering of Illumina paired-end reads was performed following the RepeatExplorer (RE) pipeline (Novák et al., 2013) implemented on the GALAXY server (https://galaxy-elixir.cerit-sc.cz). First, Illumina FASTQ files were filtered to avoid the presence of adapters, reads with indeterminate bases (N) and a minimum quality score of 10 using CLC Genomic Workbench 10.0.1 (CLC-BIO, Aahrus, Denmark). After converting filtered FASTQ reads to interlaced FASTA format, clustering analysis was performed on these data employing default RE settings (minimum overlap = 55 and cluster size threshold = 0.01 %). The cluster size threshold for detailed analysis was set to 0.01 % as recommended by Novák et al. (2013). Automated repeat classification – based on cluster connections via paired-end reads, BLASTn and BLASTx similarity searches to RE custom databases of repetitive elements and information on conserved protein domains – was used as a draft for final manual annotation and quantification of clusters. Plastid DNA clusters were removed prior to downstream analyses. For individual clustering analyses of each species, we used the maximum number of randomly sampled reads that could be processed by RE implemented on the GALAXY server using the ‘large and long’ option (see Table 1). Reads from two independently sequenced libraries of each taxon were analysed in parallel to detect biases due to library preparation and/or sequencing performance.

In addition to the individual analyses, a simultaneous comparative repeat analysis of all Anacyclus genomes – as well as of Anacyclus plus Heliocauta genomes – was performed by clustering a combined dataset made by pooling reads representing 0.65 % of the genome of each species (Table 1). Reads of 125 nt from each taxon were labelled with a unique three-character prefix and combined in a single dataset. Comparative clustering analysis was performed following the same RE settings used for individual analyses. Repeat annotation was carried out as for the individual analyses.

Analyses of repeat dynamism on individual genomes

The abundance of repeat families in the genome of Anacyclus and Heliocauta species was summarized by parsing the annotation file obtained from each RE analysis. Basic statistics such as the relative contribution of a given repetitive element to the genome composition of the species or the correlation between GS and abundance of the most important repeat families was calculated using R (www.r-project.org). To visually compare the association between repeat abundances, GS and the evolutionary relationships among the species, ancestral nuclear DNA contents (2C) were estimated under maximum likelihood using the fastAnc function and the phylogenetic tree of Oberprieler (2004), and plotted on the phylogeny with the contmap function (Phytools; Revell, 2012).

To estimate the relative timing of amplification for the main retrotransposon lineages in all species, the sequence conservation of the reverse transcriptase (RT) domain of these elements was analysed. The protein domain tool of RE was used to identify and extract conserved regions of RT protein domains for the most abundant Ty3-gypsy and Ty1-copia retrotransposon lineages. Then, a number of Illumina reads proportional to the GS of the different species was mapped onto RT domain encoding sequences and counted using CLC Genomics Workbench 10.1 (CLC-BIO). Following Mascagni et al. (2017), we kept fixed mismatch cost, deletion cost and insertion cost at 1, changing similarity and length fraction at 0.9, 0.7 or 0.5, respectively. The ratio between the number of reads mapping onto a given retrotransposon lineage at low, medium or high stringency reflects the sequence conservation level of the elements on that species. Assuming similar evolutionary rates in each species, the higher the sequence conservation the more recent the amplification of that repetitive element in a genome.

The signatures of repeat removal in HTS data are more difficult to detect than repeat amplifications (Macas et al., 2015). However, we investigated the role played by ectopic recombination between the long terminal repeats (LTRs) of the same element as a mechanism of retrotransposon elimination. Proportions of solitary LTRs (solo-LTRs), resulting from ectopic recombination, were estimated for the most abundant transposon lineages in each species following Macas et al. (2015). Basically, this method results in a value (Rsf) representing the estimated ratio of solo-LTRs to full-length elements in the genome. Larger values of Rsf suggest a higher contribution of ectopic recombination to genome shrinkage in the analysed species.

Comparative analyses of repeat dynamics

To compare the contribution of particular RE clusters to the genome of individual species, a graph-based annotation of the simultaneous comparative repeat analysis was constructed using custom R scripts (supplied by Petr Novák, Institute of Plant Molecular Biology, Biology Centre ASCR). The genomic abundance (i.e. copy number) of the largest shared clusters among each Anacyclus species was then represented together with their phylogenetic relationships and the differences in GS.

In addition, pairwise scatterplots based on the number of reads for the clusters characterized among the most important repeats were constructed following McCann et al. (2018). They were constructed for all sister species recovered from the phylogeny of Oberprieler (2004), as a measure of repeat contribution to differences in GS among pairs of taxa. Lines on those graphs represent the ratio of GS between two species, so that deviation from this line could suggest contribution of a given cluster to differences in GS among those pairs of species.

Phylogenetic analyses based on the repeatome of Anacyclus

Phylogenetic analysis based on differential genomic repeat abundance among the species was performed with the program TNT (Goloboff et al., 2003), following the approach of Dodsworth et al. (2014). First, a matrix with repeat abundances obtained from the most-abundant clusters (i.e. those representing more than 0.01 % of input reads) in the combined RE analysis was converted to Hennig86 format. Cluster abundances were cube-root transformed so that all data would fall within the range 0–65 (as required by the TNT software). Tree inference was performed using maximum parsimony (MP), calculating 100 000 bootstrap replicates with symmetrical resampling. Finally, to explore potential reticulation in the genus, a filtered supernetwork was constructed in SPLITSTREE4 (Huson and Bryant, 2006) from 1000 random bootstrap trees obtained on the MP analysis.

Additionally, we performed a phylogenetic reconstruction based on the pairwise similarities between reads obtained from the comparative repeat analysis of RE. This novel approach makes use of the observed/expected inter-specific read similarity matrices generated for each of the top most abundant clusters obtained from the output of the comparative analyses of RE. These similarity matrices were transformed into distance matrices by calculating the inverse of the values and neighbor-joining trees were constructed using the function NJ from the R package ape. All resulting trees were exported to SPLITSTREE4 in Newick format to construct a supernetwork summarizing the genetic distances among the repeatomes of Anacyclus species. A more detailed explanation of this new phylogenetic approach can be found in Vitales et al. (2019b).

Finally, to explore potential signatures of reticulation and compare them with phylogenetic inferences based on repetitive elements, the nuclear ribosomal DNA (rDNA) sequences of Anacyclus species were used as a phylogenetic marker. The 35S rDNA unit was assembled following a combined approach involving de novo and reference-guided assembly methodology. The reference sequence for each species was obtained from the longest contig of the cluster identified as ribosomal DNA by RE. The different regions constituting the rDNA units on these contigs were determined by Blast searches to the 35S rDNA sequences of Arabidopsis thaliana (GenBank accession number: X52322) using Geneious Prime 2019.1.3 (Biomatters, Auckland, New Zealand) via default settings. Then, quality trimmed (Q < 20, 0.01 probability error) Illumina reads from each species were mapped to their corresponding 35S rDNA reference sequences with the following mapping settings in CLC: mismatch cost value 2, insertion cost value 3, and deletion cost value 3, with both the length fraction value and the similarity fraction value set at 0.8. A consensus sequence was obtained from the mapping, allowing the ‘ambiguous’ mode to take into account possible intra-genomic variability (>25 % of the reads) among rDNA copies (i.e. coded in IUPAC degenerate bases). The structure and position of rRNA genes, internal transcribed spacers (ITS) and external transcribed spacer (ETS) were determined by comparison with reported Arabidopsis thaliana sequence in Geneious and Blast searches. The assembled 5S rDNA region was short (<500 bp) and did not contain many informative characters, so it was not included in further analysis. For phylogenetic reconstruction, the consensus 35S sequences of all Anacyclus species were first aligned using MAFFT 7 (Katoh and Standley, 2013). Then, SPLITSTREE4 (Huson and Bryant, 2006) was used to construct a Neighbor-Net network (Bryant and Moulton, 2004) transforming sequence divergences to uncorrected P-distances and handling ambiguous characters as average states.

RESULTS

Genome size evolution in Anacyclus

Nuclear DNA contents for the accessions of Anacyclus and H. atlantica analysed in this study are listed in Table 1. They represent the first estimates for most of these species based on flow cytometry, while estimates for Anacyclus clavatus, A. homogamos and A. valentinus are within the range given by Agudo et al. (2019). Previous nuclear DNA contents based on Feulgen cytophotometry (Nagl and Ehrendorfer, 1974) are quite similar to those obtained here. Nuclear DNA content among the species ranged between 8.22 Gbp/2C in A. valentinus and 13.14 Gbp/2C in A. linearilobus. The GS of H. atlantica, sister to Anacyclus, showed an intermediate value of 9.92 Gbp/2C. Figure 1 illustrates the distribution of GS values along the phylogenetic tree of Anacyclus based on the reconstruction by Oberprieler (2004). This representation shows contrasting patterns of GS evolution along different lineages, without any general trend towards genome expansion or reduction. Instead, some pairs or groups of closely related species present the largest differences in GS values (e.g. A. linearilobus and A. valentinus, 4.92 Gbp/2C; A. pyrethrum and A. maroccanus, 3.48 Gbp/2C).

Fig. 1.

Genome size evolution and comparative repeat composition of Anacyclus species. (A) Phylogenetic tree of Anacyclus species (from Oberprieler, 2004) with colour gradients along the tree branches indicating inferred genome size changes. (B) Species-specific genomic abundance of the ten largest clusters – sorted by their size in Anacyclus. Rectangles are coloured according to the type of repetitive element and their size is proportional to the total length of individual repeats on each species (acronyms as in Table 1).

Repeat composition of Anacyclus genomes

The number of reads per species and the corresponding genome coverage analysed with RE are reported in Table 1. Clusters making up at least 0.01 % of the genome (representing moderately to highly abundant repetitive elements) ranged from 72.9 % of nuclear DNA content in A. valentinus to 82.7 % in H. atlantica. Most of these high-copy repeats were successfully characterized and assigned to particular repeat families (78–85 % of the repeatome, depending on the species), whereas the rest were annotated as unclassified (Supplementary Data Table S2). The global repeat composition estimated from the clustering analyses of individual genomes is summarized in Fig. 2. Repeats annotated as retrotransposons were the most abundant, constituting 72–83 % of the repeatome depending on the species. The relative abundance of these repeats was similar in all Anacyclus species, whereas H. atlantica showed a considerably different profile for some repeat types (Fig. 2). The most abundant repeats corresponded to the Maximus/SIRE lineage of the Ty1-copia type (33–39 % of the repeatomes in Anacyclus species, and 52 % in H. atlantica), as well as Athila (18–24 % in Anacyclus, and 12 % in Heliocauta) and Tekay (12–14 %) lineages of Ty3-gypsy type. Other lineages of Ty1-copia and Ty3-gypsy retrotransposons (e.g. Angela or Ogre lineages) were less abundant (<5 %), while DNA transposons comprised a smaller but noticeable proportion of the repeatome (0.2–2.0 %). Regarding tandem repeats, there was considerable variation in the abundance of DNA satellites between species (1.79–4.53 % in Anacyclus, and 0.41 % in Heliocauta), with some particular clusters accounting for >1 % of the genome in certain species. Ribosomal DNA was characterized in one or two clusters, representing between 0.34 % (A. pyrethrum) and 0.78 % (A. homogamos) of the repeatome of these species.

Fig. 2.

Repeat composition of Anacyclus species and Heliocauta atlantica (acronyms as in Table 1). The abundances (percentage of repeatome) are detailed by type of repeat.

The clustering results described above, as well as the downstream analyses of repeat dynamics given below, are based on reads derived from one of the sequencing replicates per species. Comparing these results with those obtained from the other experimental replicate dataset (including independent library preparation and sequencing processes) revealed only minor differences in repeat composition of the genomes. The highest variability between replicates was found for the estimated abundance of the Tekay element in A. homogamos datasets, showing a difference of 335.6 Mbp/2C between replicates (i.e. 3.3 % variation in genome proportion). Average differences in genome proportion for whole groups of repeats ranged between 1.2 % (i.e. Tekay elements) and 0.02 % (i.e. LINE repeats), and the global repeat composition estimated from the two replicate datasets was essentially equal (Supplementary Data Fig. S1). No significant differences between replicates were observed in the results from downstream repeat dynamism and phylogenetic analyses (results not shown).

Repeatome and GS of Anacyclus species

Maximum and minimum abundances (in Mbp/2C) of different repeat types in Anacyclus species and correlations between repeat amounts and GS variation in the genus are shown in Table 2. We observed a significant strong correlation (R² = 0.992, P = 1.40E-07) between GS and the abundance of all the repeats taken together. Analysing the repeat types separately, the strongest correlation with GS corresponds to the most abundant elements in the genome. Indeed, the most prolific retrotransposons [i.e. Tekay (R² = 0.880, P = 0.000564), Athila (R² = 0.786, P = 0.00334), Maximus (R² = 0.861, P = 0.000888) and Angela (R² = 0.899, P = 0.000331)] showed significant positive correlations between their abundance in the genome and the total nuclear DNA amount of the species.

Table 2.

Maximum and minimum abundances (Mbp/2C) and correlation of repeat amounts with genome size variation among Anacyclus species

	Min.	Max.	R 2	P-value
Ty1-copia Maximus/SIRE	1958	4098	0.861	0.0008879
Ty1-copia Angela	178	470	0.8994	0.0003313
Ty1-copia other & unknown	56	138	0.09588	0.4555
Ty3-gypsy Tekay	830	1462	0.8802	0.0005636
Ty3-gypsy Athila	1416	2388	0.7862	0.003337
Ty3-gypsy Ogre	68	208	0.1608	0.3249
Ty3-gypsy other & unknown	0	70	0.2856	0.1724
Satellites	186	356	0.002245	0.9113
LINE	0	10	0.0001918	0.974
DNA transposons	14	186	0.3309	0.1357
rDNA	32	60	0.004529	0.8742
All repeats	5994	10410	0.9923	1.40E-07

The comparative clustering approach allowed simultaneous analysis of approximately 2.5 million reads, which accounted for 0.65 % for each Anacyclus and H. atlantica genomes (Table 1). Almost 2 million reads were grouped in 260 clusters representing moderate or highly abundant repeat families shared among the different taxa. The proportion of homologous clusters (or repeat types) found in the genomes of Anacyclus and H. atlantica is shown in Supplementary Data Fig. S2. Most repeats are shared by all Anacyclus species whereas many of them are absent from the H. atlantica genome. In contrast, some particular clusters are either exclusive or much more abundant in Heliocauta than in Anacyclus species. The differences in the genomic abundance of homologous clusters within Anacyclus species against the phylogenetic tree can be visually seen in Fig. 1 and Supplementary Data Fig. S3, obtained from comparative clustering analysis of Anacyclus data alone. It is apparent from these representations that all high-copy-number repeats are present in each species, with varying abundances of individual repeats usually proportional to their GS. Likewise, copy number of most Ty1-copia and Ty3-gypsy repeats was proportional to the variation in GS between sister species (Fig. 3). Deviation from the expected ratios was only observed for some Ty3-gypsy clusters when comparing A. clavatus and A. monanthos. More heterogeneous repeat abundances (in relation to GS) were observed between non-sister species (Supplementary Data Fig. S4), particularly for Ty1-copia Maximus and Ty3-gypsy Chromovirus clusters (e.g. A. pyrethrum vs. A. maroccanus or A. pyrethrum vs. A. homogamos).

Fig. 3.

Pairwise scatterplots of the number of reads from each sister Anacyclus species in repeat clusters from the comparative analysis (acronyms as in Table 1). The slope of the dotted line is equal to the ratio of the genome sizes between the two species (i.e. repeats along the abline are found in the same genomic proportions in species compared).

Repeat dynamism in Anacyclus

Sequence conservation of the most abundant retrotransposons (i.e. Maximus-SIRE, Athila and Tekay) was inferred by mapping Illumina reads to their RT DNA sequences at different stringency levels (Fig. 4). The more conserved a sequence is, the more recent the proliferation of the element in question should be. Our results showed that RT domains of Tekay elements were more divergent for high-stringency mapping conditions, whereas Maximus and Athila RT domains were more conserved in most of the species. Overall, the level of sequence conservation varied between species and LTR–retrotransposon families. For instance, the RT sequences of A. pyrethrum showed the highest divergence level for Maximus/SIRE elements, whereas they appeared to be amongst the most conserved for Athila elements. Sequence conservation of the RT domains of these highly abundant Ty1-copia and Ty3-gypsy elements did not show any apparent association with the GS of the species. For example, A. linearilobus and A. valentinus (i.e. presenting the largest and the smallest GS values within the genus, respectively) showed a similar moderate conservation level for the RT sequences of Maximus/SIRE elements.

Fig. 4.

Relative number of mapped Illumina reads on sets of reverse-transcriptase species-specific domains belonging to the most abundant transposon lineages of Anacyclus at three different stringency parameters (acronyms as in Table 1).

Finally, to investigate the effect of retrotransposon elimination by ectopic recombination affecting the LTR regions of the same element, we compared the presence of solo-LTRs among Anacyclus genomes (Table 3; Supplementary Data Fig. S5). We were only able to identify LTR 3′ end and 5′ untranslated region from the most abundant retrotransposon repeats (i.e. Maximus/SIRE, Tekay and Athila elements), while the approach could not clearly recognize these LTR junctions within less abundant repeats. Rsf values ranged from 0.04 to 0.88 for all elements within Anacyclus species, while most of the Rsf values for LTR families, and the mean values per species, were below 0.5 (Table 3), indicating that highly abundant retrotransposons are predominantly full-length. The average estimated ratio of solo-LTRs to complete elements was slightly higher on Tekay elements (Rsf = 0.46) than on Athila (Rsf = 0.37) or Maximus (Rsf = 0.17) lineages. The species with the highest Rsf values were A. homogamos, A. monanthos and A. clavatus, all of them showing medium-sized genomes (Table 3; Supplementary Data Fig. S5). The species with the largest GS (i.e. A. linearilobus and A. pyrethrum) showed the lowest average Rsf values. However, no significant correlation was found between GS and Rsf values estimated for any of the elements (Spearman’s rank correlation; P > 0.05 in all cases).

Table 3.

Estimated ratios of solo-LTR to complete elements (Rsf) for the three most abundant repetitive elements in Anacyclus.

Species	2C (Gbp)	Ty1-copia Maximus/SIRE	Ty3-gypsy Athila	Ty3-gypsy Tekay	Mean Rsf value
LIN	13.14	0.04	0.20	0.29	0.18
VAL	8.22	0.38	0.49	0.17	0.35
CLA	10.42	0.04	0.50	0.53	0.36
MON	11.54	0.07	0.46	0.88	0.47
RAD	12.34	0.18	0.44	0.33	0.32
HOM	10.16	0.41	0.31	0.77	0.50
MAR	9.62	0.12	0.22	0.61	0.32
PYR	13.10	0.12	0.36	0.08	0.19

Phylogenetic reconstructions based on Anacyclus repeats

The phylogenetic reconstructions based on rDNA sequences, repeat abundances and repeat sequence similarities are presented in Fig. 5. The alignment of rDNA sequences consisted of 6615 nucleotide positions, including 5′ETS, 18S gene, ITS1, 5.8S gene, ITS2 and 26S gene. The 3′ETS region was difficult to align and therefore excluded from further analyses. The final aligned matrix comprised 287 variable positions of which 104 included intragenomic polymorphisms. The Neighbor-Net network showed evidence of reticulation, as indicated by the presence of several alternative splits. Overall, the phylogenetic relationships among the species are consistent with previous systematic studies of Anacyclus (e.g. Oberprieler, 2004), except for A. clavatus, which appears in a reticulation between A. pyrethrum and the group comprising A. linearilobus, A. valentinus, A. monanthos and A. radiatus.

Fig. 5.

Evolutionary relationships among Anacyclus species based on repeatome information. (A) Neighbor-Net network inferred from rDNA sequences. (B) Filtered supernetwork summarizing a random selection of 10 000 bootstrap trees from the maximum parsimony analysis based on cluster abundances of the most abundant repeats in Anacyclus. (C) Filtered supernetwork summarizing neighbor-joining trees obtained from the interspecific sequence similarity matrices of the most abundant repeats in Anacyclus.

The networks resulting from repeat abundances and repeat similarities data showed well-defined topologies, in both cases with species relationships supported by the majority of the trees used to infer each reconstruction. The structure of the networks revealed substantial conflict between these phylogenetic inferences. Only the evolutionary relationship between A. clavatus and A. pyrethrum was consistent on both approaches. The remaining species appeared to be grouped according to GS resemblance in the network based on repeat abundances. In contrast, the phylogenetic relationships inferred using a supernetwork constructed upon repeat sequence similarity data showed higher agreement with the network based on rDNA sequences, as well as with previous evolutionary hypothesis for the genus and the geographical distribution of species. All these phylogenetic analyses were confirmed using the second sequencing replicate of our dataset from which we obtained the same results.

DISCUSSION

Despite having the same chromosome number, the eight Anacyclus species show considerably different GS values, the smallest values representing around 60 % of the largest. Some species showed similar GS values to those reconstructed for their most recent common ancestor, whereas in other cases, both genome up- and downsizing trends were retrieved. Overall, these values point to a bi-directional GS evolution, where genome expansions and contractions take place in parallel along separate lineages of Anacyclus (Fig. 2). By examining and characterizing the repetitive DNA content in Anacyclus, we have shown that the size of these genomes is not determined by the differential proliferation/removal of one or a few high-copy-number TE families, proposed to be the most common pattern in plants with relatively small genomes (El Baidouri and Panaud, 2013). In contrast, our results indicate that the relative abundance of all kinds of repeats have remained relatively constant along the evolution of this group, irrespective of species GS (Figs 1 and 3).

This pattern of similar repeatome composition despite GS diversity could be explained by different mechanisms, depending on the putative role played by TE activity during genome evolution. As hypothesized based on previous cytogenetic and phylogenetic data (Humphries, 1981; Agudo, 2017; Rosato et al., 2018), hybridization events could have triggered transposon dynamism during the evolutionary history of Anacyclus. By relaxing epigenetic silencing mechanisms (Lisch, 2009), genomic shock (McClintock, 1984) could affect different TE lineages at the same time. This is consistent with the revealed picture showing that overall repeat abundance is correlated with GS. However, if larger genomes were explained by higher DNA amplification rates, we would expect the majority of large-sized genomes to show evidence of recent amplification of repetitive DNA (Mascagni et al., 2017). Instead, our sequence divergence analyses suggest that the conservation levels of highly abundant TEs are decoupled from the actual GS of the species (Fig. 5). Therefore, recent amplification of these repeats would not be the most likely cause of GS differences.

Another potential cause for contrasting GS values between Anacyclus species could be the differential ability to remove DNA efficiently from repeat copies as they are amplified (e.g. Kelly et al., 2015). The two species with the largest GS values (i.e. A. linearilobus and A. pyrethrum) showed overall low proportions of solo-LTRs, which might suggest that ectopic recombination between the LTRs of the same element was less active in those two cases. This and/or other non-tested genome shrinkage mechanisms (e.g. illegitimate recombination; see Hawkins et al., 2009; Ren et al., 2018) might occur without bias with respect to repeat type, such that the proportions and types of repeats would remain constant even when GS does not. However, compared to the values obtained by Macas et al. (2015) in the tribe Fabeae, the estimated ratios of solo-LTRs to complete elements are generally lower in all Anacyclus species, lacking significant correlation with their GS values. Moreover, the contrastingly large and small GS values found among very closely related Anacyclus species (e.g. the sister species A. linearilobus and A. valentinus) are difficult to explain by the sole action of deletion processes (i.e. without the occurrence of genome expansion too). In summary, contrary to the predominant view that plant homoploid genomes basically expand or contract as a consequence of TE dynamics (Hawkins et al., 2009; El Baidouri and Panaud, 2013; Kelly et al., 2015), our results suggest that other mechanisms have played an important role in GS differences across Anacyclus species.

The alternative hypothesis we propose to explain GS evolution in the genus involves chromosomal rearrangements probably associated with hybridization events during the evolutionary history of Anacyclus. Based on extensive cytogenetic, morphological and systematic data, Humphries (1981) stated that speciation in Anacyclus was associated with chromosome re-patterning derived from hybridization among the taxa. Specifically, he suggested that losses and gains of DNA content in the genus are linked to structural changes in the chromosomes, such as paracentric inversions and reciprocal translocations, occurring as a result of the reticulate evolution of Anacyclus (Humphries, 1981). More recently, FISH experiments by Rosato et al. (2017, 2018) revealed an unusual high abundance and diversity of interstitial telomeric repeats (ITRs) and 45S rDNA sites at intra- and interspecific levels in the genus. Interestingly, all cases of interstitial 45S rDNA co-occurred with ITRs in close proximity in the same chromosome arms, a pattern that can be attributed to chromosomal rearrangements such as translocations affecting both rDNA and telomeres (Tsujimoto et al., 1999; Raskina et al., 2008). Homoploid hybridization between two genomes that differ in size can lead to chromosome arm exchanges (or other balanced rearrangements that do not disrupt gene dosage), which create transgressive GS values and heterozygous karyotypes (Fig. 6; Schubert and Lysak, 2011; Danilova et al., 2017). If the repeatome composition in hybridizing species is mainly homogeneous, as in Anacyclus, these rearrangements are likely to impact the overall GS without altering significantly the relative proportion of dispersed repeats (e.g. transposons).

Fig. 6.

Scheme representing hypothetical recombination events between homologous chromosomes derived from distinct genomes (i.e. from homoploid hybridization) leading to chromosome arm exchanges, which would result in different genome sizes. The colours represent the distinct genomic origin of chromosomes and chromosome arms. Arrows indicate the direction of the recombination events and dash lines show the hypothetical chromosome segments that are exchanged.

This hypothesis is also supported by the results of our phylogenetic reconstructions based on repeat abundances and repeat sequence similarities among Anacyclus species (Fig. 5). By comparing these two phylogenetic approaches, we attempted to test whether the repeatome (as well as the GS) evolution could be better explained by repeat dynamics or by chromosomal rearrangements. Dodsworth et al. (2014) showed that, assuming repeats evolve in accordance with random genetic drift, repeat abundances contain consistent phylogenetic signal. Therefore, if repeatomes in Anacyclus were mainly the result of gradual expansion (e.g. retrotransposition) and contraction (e.g. recombination-based deletion) events, we would expect repeat abundances and repeat sequence similarities to evolve together (i.e. by random drift processes) across species. In this case, the phylogenetic hypothesis based on repeat abundances should be equivalent to those based on pairwise read similarities. In contrast, if repeatome evolution was mainly determined by dissimilar chromosome segment exchanges between parental subgenomes (Fig. 6), the copy number of repetitive elements would change abruptly after reorganization events, while sequence similarity would be tracing chromosome rearrangement (and hybridization) history. In this second scenario (i.e. chromosomal rearrangements driving GS evolution), we would expect that repeat abundances are disconnected from repeat sequence similarities among species, and thus phylogenetic inference based on the two data sources would differ. The contrasting topologies of phylogenetic networks obtained from both approaches indicate that repeat abundances did not evolve together with repeat sequence similarity. This would suggest that, apart from more or less gradual amplification and deletion processes, repeatome evolution within this genus probably experienced other kinds of large-scale genomic restructuring mechanisms impacting GS. Altogether, our results fit with a scenario in which hybridization accompanied by deep chromosome rearrangements shaped GS evolution and homoploid hybrid speciation in Anacyclus.

The role of homoploid hybridization in evolution (Abbott et al., 2013) remains unclear, both when focusing on adaptive introgression (Suarez Gonzalez et al., 2018) and when focusing on homoploid hybrid speciation (Schumer et al., 2014; Nieto Feliner et al., 2017). Furthermore, the influence of reticulation events on the karyotype of homoploid hybrids has been insufficiently studied (Danilova et al., 2017), and prevents us from drawing strong conclusions without falling into excessive speculation. Nonetheless, our results suggest that reticulate evolution could have contributed to shape GS diversity through enhanced chromosomal rearrangements. This hypothesis is now experimentally testable using comparative chromosomics (a term encompassing genome sequencing, cytogenetics and cell biology; for a review see Deakin et al., 2019) to trace the history of chromosome segments/arms. Finally, the study of additional groups showing similar patterns to Anacyclus (i.e. homoploid hybridization and/or karyological polymorphism) will be necessary to clarify whether this mechanism of GS evolution is more common than previously thought across land plants.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following.

Figure S1: Repeat composition of Anacyclus species and Heliocauta atlantica comparing replicated datasets.

Figure S2: Comparative repeat composition of Anacyclus species and Heliocauta atlantica.

Figure S3: Genome size evolution and comparative repeat composition of Anacyclus species.

Figure S4: Pairwise scatterplots of the number of reads from non-sister Anacyclus species in repeat clusters from the comparative analysis.

Figure S5: Graphical representation of the genomic abundances of major types of LTR retrotransposons and the estimated Rsf values in Anacyclus species.

Table S1: Additional sampling information of the Anacyclus specimens analysed in the study.

Table S2: Global repeat composition estimated from the clustering analyses of both replicates from individual Anacyclus genomes.

FUNDING

This work was supported by the Catalan government (grant number 2017SGR1116) and by the Spanish government [grant numbers CGL2010-18039, CGL2013-49097-C2-1-P, CGL2013-49097-C2-2-P, CGL2016-75694-P (AEI/FEDER, UE), RYC-2014–16608 to S.G. and RYC-2016–21176 to O.H.].