Stowaway miniature inverted repeat transposable elements are important agents driving recent genomic diversity in wild and cultivated carrot
Miniature inverted repeat transposable elements (MITEs) are small non-autonomous DNA transposons that are ubiquitous in plant genomes, and are mobilised by their autonomous relatives. Stowaway MITEs are derived from and mobilised by elements from the mariner superfamily. Those elements constitute a significant portion of the carrot genome; however the variation caused by Daucus carota Stowaway MITEs (DcStos), their association with genes and their putative impact on genome evolution has not been comprehensively analysed.
Fourteen families of Stowaway elements DcStos occupy about 0.5% of the carrot genome. We systematically analysed 31 genomes of wild and cultivated Daucus carota, yielding 18.5 thousand copies of these elements, showing remarkable insertion site polymorphism. DcSto element demography differed based on the origin of the host populations, and corresponded with the four major groups of D. carota, wild European, wild Asian, eastern cultivated and western cultivated. The DcStos elements were associated with genes, and most frequently occurred in 5′ and 3′ untranslated regions (UTRs). Individual families differed in their propensity to reside in particular segments of genes. Most importantly, DcSto copies in the 2 kb regions up- and downstream of genes were more frequently associated with open reading frames encoding transcription factors, suggesting their possible functional impact. More than 1.5% of all DcSto insertion sites in different host genomes contained different copies in exactly the same position, indicating the existence of insertional hotspots. The DcSto7b family was much more polymorphic than the other families in cultivated carrot. A line of evidence pointed at its activity in the course of carrot domestication, and identified Dcmar1 as an active carrot mariner element and a possible source of the transposition machinery for DcSto7b.
Stowaway MITEs have made a substantial contribution to the structural and functional variability of the carrot genome.
KeywordsTransposable elements Insertional polymorphism TE-gene association Mariner DcSto Daucus carota
DcSto-intron length polymorphism
Daucus carota Stowaway
Carrot double haploid plant used for reference genome assembly
Miniature inverted repeat transposable elements
Parallel insertion sites
Singular enrichment analysis
TE-associated structural variation
Target site duplication
Transcription start site
Unique insertion sites
Transposable elements (TEs) are discrete segments of DNA capable of changing their genomic location in a process called transposition . Based on the mechanism of transposition, TEs are divided into two classes, class I (retrotransposons), mobilised via an RNA intermediate, use a ‘copy and paste’ mechanism, while class II (DNA transposons) are mobilised by ‘cut and paste’ or ‘copy and paste’ mechanisms of DNA that do not require a reverse transcription step. In both classes, there are autonomous elements that possess enzyme-encoding genes required for mobilisation, non-autonomous elements which can still be mobilised by their autonomous counterparts, and inactive defective copies .
Miniature inverted-repeat transposable elements (MITEs) are small in size (< 800 base pairs, bp), usually AT-rich sequences with no coding capacity. They are mobilised by related autonomous class II trans-acting elements. Despite their small size, they may account for a significant portion of plant genomes, representing up to 10 and 13.8% for rice and mulberry, respectively [3, 4]. This extremely efficient proliferation of MITEs, as compared to their ancestral autonomous elements, might be caused by their higher affinity for transposase, resulting from lower cis-requirements for enzyme recognition and by the presence of subterminal and/or internal enhancers of nucleoprotein complex formation [5, 6]. Some MITEs families, such as mPing elements in rice, may preferentially target single-copy gene-rich regions , and escape the epigenetic control system because MITE-derived trans-acting siRNAs do not share sequence similarity with the coding region of the source of the transposase . All these features, coupled with their small size, make them abundant in plant genomes, and frequently present in the vicinity of genes.
Currently, the pivotal role of TEs in the evolution of plant genomes is becoming more widely recognised. Mobilisation of TEs leads to structural variations that contribute to the genomic diversity of their host, some of which can be adaptive. Among other effects, the role of TE insertional hotspots in the formation of biosynthetic gene clusters was proposed, based on the analysis of genes of the terpene biosynthesis pathway in eudicots . TE insertions can impact gene expression in many ways. By insertion upstream, within, or downstream of coding regions, TEs may provide new regulatory features that can affect gene expression . In addition, RNA-directed methylation (RdDM), which has a role in repetitive DNA control and defense against viruses, can lead to epigenetic changes upon TE insertion that may produce epialleles for adjacent genes . In crops, such modifications can affect agronomically important traits, such as observed with flowering time variation due to MITE insertions into the quantitative trait locus Vegetative to generative transition 1 (Vgt1) .
Therefore, one of the main challenges of crop genomics is to critically evaluate the extent of species-wide TE-associated structural variation (TEASV), in order to better understand the dynamics of genome evolution. Several bioinformatics tools have been developed that allow for the identification of TEASV from resequencing data generated by next-generation sequencing (NGS) (reviewed in ). However, only a few genome-wide comparative analyses of TEASV have been published, almost exclusively for autogamous species. Moreover, most of these were focused on the global TE landscape, and thus they were biased towards the most numerous TE families.
TE-derived variants are also sources of functional variation, as shown by TE variants linked to changes in DNA methylation in Arabidopsis thaliana  and flowering traits in maize [15, 16]. A study associating TE-derived variants with phenotypic traits related to maize adaptation to a temperate climate revealed more candidate genes than analysis using single nucleotide polymorphisms . This suggests that TEs may be able to quickly enhance host adaptability under adverse environmental conditions. An adaptive role for TEs has also been suggested for A. thaliana  and Capsella rubella .
To date, only a few reports have focused on the global analysis of MITEs at the species level. Mining of MITEs in 19 A. thaliana ecotypes yielded a total of 2406 copies grouped into 212 families . In another report, 20 MITE families were annotated in B. rapa, B. oleracea, and A. thaliana. Of these, only four were present in A. thaliana, indicating that amplification and diversification of most Brassica MITE families took place after divergence of the Arabidopsis and Brassica lineages. Moreover, some MITE families were significantly enriched in B. rapa and B. oleracea; therefore, they were likely activated after divergence of those species . Rice mJing elements were frequently inserted into introns (16.67%) and into the 2 kb regions flanking genes (45.83%) . Rice accessions differ dramatically in terms of mJing copy number, ranging from 18 to 150 in japonica and African cultivated rice, respectively. This suggested multiple amplification bursts, which most likely occurred before the amplification burst of mPing, another well-characterised active MITE in rice . Some rice lines showed a sharp increase of mPing copies, from less than 10 copies in indica to 1000 in a temperate japonica cultivar Gimbozu EG4 . Association of MITE insertions with coding regions was also shown for 18 wheat Stowaway element families, with 5.1% of more than 19,000 MITEs being transcribed, and 52–63% insertion sites being located within 100 bp of genes [23, 24]. A comparative analysis revealed specific proliferation of two MITE families in the A genome and one in the B genome, suggesting their possible impact on genome diversification during speciation .
Carrot (Daucus carota) is a diploid species with 2n = 2x = 18, and a relatively small genome of 473 Mb . It is an allogamous species, suffering from inbreeding depression. Cultivated carrot is a biennial root vegetable and the most economically important species of the Apiaceae family, and is grown around the world in temperate and subtropical regions . D. carota has been domesticated relatively recently, about 1100 years ago. Wild carrot is widespread in temperate regions of the world. While the Mediterranean basin is considered the centre of biodiversity for Daucus spp. , Central Asia has been identified as the place of origin of domesticated carrots [25, 28]. The species has four major structural groups: European wild D. carota, which show remarkable morphological diversity and are grouped into several subspecies, referred to as D. carota complex; Asian wild D. carota subsp. carota; eastern cultivated carrots, which are mostly primitive landraces, often producing yellow or purple storage roots; and western cultivated carrots, which include advanced orange cultivars. Cultivated and wild carrots can easily hybridise, and a considerable amount of genetic variation is exhibited both between and within the groups, with no apparent signature of a domestication bottleneck .
The carrot reference genome assembly of a double haploid plant (DH1) has been published recently . The repetitive fraction constituted 46% of this carrot genome. DNA transposons comprised 13.6% of the genome and 30% of the total repetitive DNA. Approximately 2.3% of the assembled portion of the genome was attributed to MITEs, of which Stowaway-like elements constituted around 0.5% . Carrot Stowaway-like MITEs (DcStos) had previously been reported to be abundant and highly polymorphic [25, 29]. In this current study, we used 14 DcSto families for a systematic genome-wide analysis of TEASVs in 31 resequenced genomes from cultivated and wild carrot accessions. The accessions were representative of the four structural groups of D. carota, as described above. DcSto insertions were comprehensively annotated and their chromosomal distribution was analysed. In addition, we identified a DcSto family likely active in cultivated carrot.
Distribution of DcSto elements in D. carota
Abundance and distribution of the 14 DcSto families in D. carota
Number of DcSto insertion sites
2 kb upstream
2 kb downstream
Verification of in silico results for 28 DcSto insertion sites in the carrot genome by DcS-ILP
Comparison of in silico / DcS-ILP results
Accession / insertion site combinations
empty / homozygous empty
occupied / homozygous occupied
occupied / heterozygous (empty + occupied)
empty / homozygous variant of different size (‘empty’ for DcSto)
empty / heterozygous (empty + variant of different size)
occupied / heterozygous (occupied + variant of different size)
Total correct calls
empty / heterozygous (empty + occupied)
empty / homozygous occupied
occupied / homozygous empty
Total incorrect calls
All DcSto families had similar densities in all the nine carrot chromosomes (Additional file 2: Figure S2); however, they differed in terms of their copy number, from 155 copies for DcSto11 to 3633 copies for DcSto6 (Table 1). The differences likely reflected their ability to proliferate once integrated in the genomes of D. carota. The proliferation coefficient (PrC), i.e., the proportion of the total number of insertions per family divided by the average number of insertions per genome, ranged from 6.33 for DcSto11 to 17.50 for DcSto7b (Table 1). PrC values correlated with the intra-family similarity (Additional file 2: Figure S3), indicating that PrC was a good measure of the recent expansion of particular DcSto families.
Insertional polymorphism of DcStos
Among the 18,518 insertion sites, only two of them harboured the same element insertion in all 31 genomes of D. carota (Additional file 2: Table S2), while 22 insertions (0.12%) were present in all genomes of cultivated carrots (Additional file 2: Table S3). We observed a high proportion of insertions in only one of the 31 D. carota genomes (66.2%; Additional file 2: Table S4), which we subsequently referred to as unique insertion sites (UIS). It was important to note that most UIS were likely not ‘unique’ in absolute terms, but only in relation to the collection of 31 plants of different origin investigated in this current study. Thus, the majority of them represented DcSto insertions occurring less frequently, but likely still shared among populations of D. carota. In general, the number of UIS in the cultivated carrot accessions was relatively more uniform than in those representing the wild carrots. In addition, the wild carrots, especially those of European origin, had a higher proportion of UIS per genome, as compared to the cultivated carrots.
DcStos and the structure of genetic diversity for D. carota
DcSto families differed in terms of their contribution to the total copy number in individual genomes (p-value = 2.2e-16). The differences in DcSto distribution reflected the classification of the investigated accessions into four major groups, European wild (Ssp1-Ssp5 and W1-W3), Asian wild (W4-W8), eastern cultivated (C1-C6), and western cultivated (C7-C13 and I1-I4; Fig. 1b), in agreement with the previously reported population structure of D. carota [25, 28]. This was further confirmed by the genetic diversity structure inferred from global DcSto insertion polymorphisms. The four major groups were clearly distinguishable as non-overlapping clusters (Fig. 1c and d).
The eastern cultivated carrots were characterised by fewer DcSto1 and DcSto8 copies, while they carried more DcSto7b copies. Different proportions were observed in western cultivated carrots, which had slightly less DcSto6 copies. Within the Asian wild carrots, as in the case of the eastern cultivated carrots, DcSto1 and DcSto8 families were less numerous. By contrast, the eastern cultivated and the Asian wild accessions largely differed in the number of Dcsto7b copies, which were overrepresented in the former and underrepresented in the latter. Generally, European wild carrots were the most diverse in terms of DcSto distribution. Within this group, accessions Ssp1, Ssp2, Ssp3, and W2 had more DcSto6, DcSto8, and DcSto1 copies, while Ssp5 (D. carota subsp. capillifolius) was characterised by more DcSto2 and DcSto7a copies (Fig. 1b).
Localisation of DcSto copies in relation to genes was non-random and family-specific
Gene ontology (GO) enrichment analysis revealed that DcSto copies inserted in upstream or downstream regions of genes were significantly associated with those involved in the regulation of transcription (biological process, p-value = 4.06e-13; Additional file 2: Figure S4) and transcription factor activity (molecular function, p-value = 1.85e-08; Fig. 2e). By contrast, DcStos elements inserted in introns did not show an association with any particular GO term, except for marginally significant family-specific signals not related to transcription regulation. For UTRs, the number of DcSto insertions were too low to find reliable associations, except for DcSto7b insertions in 5’UTR regions, which were significantly associated with genes encoding transcription factors (Additional file 2: Table S5).
DcSto insertion hotspots
Within all identified insertion sites, 292 (1.6%) were parallel insertion sites (PIS), i.e., insertion sites of different DcStos into precisely the same genomic position. Within PIS, 95% harboured insertions of DcSto copies from two different families, while the remaining 5% carried alternative insertions of three or more different copies. More than 63% of PIS were localised in the vicinity of genes or within the body of genes (Additional file 3).
The co-occurrence of copies from two different DcSto families in PIS was positively correlated (p = 7.03e-12) with the cumulative number of all insertion sites of those families, and was negatively correlated (p = 8.80e-03) with the genetic distance between terminal inverted repeats (TIRs) of those families (Additional file 2: Figure S13, Additional file 2: Table S6). Thus, families with more copies were more frequent in PIS; however, DcStos elements carrying more similar TIRs were also relatively more frequently inserted into the same site.
DcSto7b elements have been active in the course of carrot domestication
Dcmar1 might provide the transposition machinery for DcSto7b
The evidence for recent mobilisation of DcStos elements described above prompted us to search for autonomous elements that could have been involved in the process. Eleven copies of mariner-like elements were found in the carrot DH1 genome (Additional file 2: Table S7), ranging from 1922 bp (Dcmar9) to 4940 bp (Dcmar6) and carrying 24- to 32-nt-long TIRs. One mismatch between the 5′ and 3′ TIRs was present in Dcmar1, Dcmar5, and Dcmar10, while TIRs of the remaining elements carried more mismatches (Additional file 2: Table S7). The C-terminal part of the predicted transposases of eight Dcmars had a complete DD39D motif, characteristic of mariner elements. Three elements lacking the conserved region of the MLE domain (Dcmar9, Dcmar10 and Dcmar11) were classified as internally truncated and were not further considered.
The first two aspartic acids of the DD39D motif were predicted to be Mg2+ binding sites for all eight Dcmars elements, while the helix-turn-helix (HTH) DNA binding motif was predicted for six of them, with at least a 90% probability (Additional file 2: Table S7). However, all features required for mariner transposition, as defined by Claeys Bouuaert and Chalmers , were only found with Dcmar1, a 4353 bp-long element inserted in chromosome 8 (position 25,189,375–25,193,731 in the reference genome DH1).
We investigated the transcriptional status of Dcmars elements, using RNAseq reads of DH1 . The Dcmar1 transposase was expressed in four of 20 tissues, callus, whole opened flowers (2 cm umbels at anthesis), bracts (2 cm umbels), and flower buds, while no transcripts attributed to other Dcmars elements were found.
At least one copy of Dcmar1 was present in 14 of 19 genomes of cultivated carrots (74%), but only 4 of 12 wild carrot genomes (33%; Additional file 2: Figure S14b). However, PCR amplification of the region spanning the Dcmar1 insertion site in the reference genome DH1 revealed an absence of the element at that genomic location in all other D. carota genomes investigated in this study (Additional file 2: Figure S14c). In silico identification of the position of Dcmar1 elements was largely consistent with results from PCR assays. For all plants but one, we identified at least one putative insertion site, at 17 different genomic locations. Of these, eight Dcmar1 insertion sites were associated with genes. In the case of the eastern cultivated accession C4 from Afghanistan, probably only one insertion site was present in the genome; however, due to the insertion of Dcmar1 into an intron of one version of paralogues with high similar sequences, we were not able to determine its precise position (Additional file 2: Figure S14d). Only in case of wild carrot accession W2 from Portugal was the PCR assay positive; however, the presence of a Dcmar1 copy was not confirmed by in silico analysis. The data suggested that Dcmar1 elements had been actively transposing.
Transposable elements have been recognised as major drivers of the evolution of eukaryotic genomes. They have been involved in the creation of structural and functional novelty, making them an essential and long-standing part of genomes . In particular, MITEs and LTR retrotransposons, being ubiquitous in plants, have dynamically shaped genome structure and altered gene function in a variety of ways . A number of recent reports have provided clues of functional interactions of MITEs with host genes [4, 34, 35, 36, 37, 38, 39, 40, 41]. The rice genome has been used as a model for investigating MITE-host interactions, as despite its moderate size it harbours an exceptionally rich and diverse collection of almost 180,000 MITE copies divided into 338 families , including mPing, the most thoroughly studied currently active MITE family . However, less than 15% of all full-length MITE insertions were polymorphic between subspp. Indica and japonica . Being a dicot species, carrot provides an alternative plant host model to rice. Our results point at a much higher level of insertional polymorphism of carrot Stowaway MITEs, perhaps more similar to that of active mPing elements in rice . This might be attributed to the recent or ongoing activity of some DcSto families, but also to the fact that contrary to rice, carrot is an allogamous species. De la Chaux et al.  reported a major reduction of TE abundance in autogamous A. thaliana, as compared to its close allogamous relative A. lyrata, consistent with such theoretical expectations .
DcSto mining strategy
Fourteen DcSto families were selected as they were the most abundant in the carrot reference genome . A global annotation of MITEs in carrot also revealed the presence of several other Stowaway-like MITE families that were usually less numerous (data not shown). It was possible that some of them acquired much higher copy numbers in the resequenced genomes, as suggested by up to four-fold differences in the global number of analysed DcStos elements, as well as within-family differences.
The genome-wide comparative analysis yielded a catalogue of 18,518 structural variants caused by MITE copies belonging to 14 DcSto families across 31 D. carota accessions. Of these, less than 2000 copies were attributed to the reference genome DH1. Previously, Iorizzo et al.  identified around 4000 DcSto copies in the carrot genome. However, in this current study a more stringent approach was used for DcSto mining, which resulted in a generally lower number of catalogued insertion sites. Iorizzo et al.  used the web tool TIRfinder , which identifies all sequences that share common structural features, i.e., specified target site duplications (TSD) and TIRs in assembled sequences. All elements meeting these criteria, regardless of their genomic location, were reported. By contrast, RelocaTE analysis  was used in the current study. It retrieved insertion sites from raw sequencing reads based on a similarity search using stringent cut-off parameters, filtering out reads mapping to genomic regions comprising repetitive sequences. Therefore, elements too divergent from the TE family consensus and those inserted into repetitive regions or in the vicinity of structural variants were not reported. This was why in this current study only complete MITEs residing in unique genomic regions were mined, and consequently, the total number of DcSto elements obtained for the reference genome was lower than previously reported. Nevertheless, the same approach was systematically applied to all carrot accessions, and the results were comparable. In addition, the in silico mining results were extensively validated by PCR, with the reliability of the mapping tool RelocaTE being largely confirmed.
DcSto distribution and the genetic diversity of D. carota
DcSto insertions were extremely polymorphic between cultivated and wild carrots, and within both groups. Global DcSto insertion polymorphism revealed a genetic diversity that mirrored previous reports using single nucleotide polymorphisms [25, 47]. It showed that informative DcSto insertional polymorphisms, i.e., those present in at least two genomes, allowed grouping of the accessions. The high rate of unique insertion sites observed might have resulted from insufficient sampling, especially for the wild D. carota group; however, it might also indicate current transpositional activity. This latter option was supported by the observation that copy numbers of particular DcSto families differed among accessions belonging to the same group. For example, the Portuguese accessions (members of the wild European group) were enriched in DcSto6 and Dcsto8 elements, while subsp. capillifolius carried more copies of the DcSto2 family. This might suggest amplification bursts of different MITE families in geographically separated populations of wild carrots. In addition, while both eastern cultivated carrots and eastern wild carrots had similarly low frequencies of DcSto1 and DcSto8, they differed with respect to the numbers of DcSto7b elements. The sharp increase in the number of DcSto7b copies in eastern cultivated carrots, as compared to the sister clade of Asian wild carrots, suggested the activity of these elements might have been significant during the early stages of domestication.
DcSto insertions in the context of the carrot genome
DcSto copies showed similar distribution patterns across all carrot chromosomes typical for MITEs, i.e. depletion around centromeres and enrichment in genic regions. Previously, fluorescence in situ hybridisation had revealed DcSto signals along chromosome arms, and their absence at centromeres, telomeres, and nucleolar organiser regions . Iorizzo et al. reported that DcStos elements did not show any deviation from a random distribution across the carrot reference genome . Even so, the current study indicated that individual DcSto families were characterised by contrasting distribution patterns in terms of their association with genes. Enrichment of DcSto7b copies was observed within the 2 kb region upstream of transcription start sites (TSSs) and in 5′ UTRs, and depletion in introns, while DcSto6 showed the opposite tendency. This might reflect a genuine preference for insertion of different families into specific sections of genic regions, or random insertions followed by selection acting on non-neutral insertions. The latter scenario would change frequencies of insertions observed in different genic segments, depending on the age of the insertions. DcSto7b copies are very similar and the family has been shown to have had a single very recent peak of activity . By contrast, DcSto6 elements are an older family, which has likely experienced several peaks of activity . If the selection hypothesis is true, it would imply selection has acted against DcSto6 insertions in sequences upstream of genes and in 5′ UTRs and/or retention of insertions in introns.
To date, comparative analyses of TE distribution has usually been generalised for larger groups. A global distribution of MITEs in Citrus resembled that reported in this current study for DcStos elements in carrot . However, analyses focusing on particular TE families have revealed specific patterns [4, 50]. Notably, in contrast to carrot Stowaways, mulberry Tc1/mariner superfamily MITEs were the only group that was not preferentially inserted near genes. Nevertheless, they had the highest ratio of the total number of transcribed MITEs to the total number of genes . This supported our hypothesis that detailed analysis of individual families was essential for better understanding of the impact of TEs on the host genome.
Prevalence of low frequency DcSto insertions
We observed a high level of low frequency DcSto insertions, with most of them referred to as UIS if present in only one of the 31 investigated genomes, a phenomenon also reported for other species [17, 51, 52]. As proposed by Uzunović et al., localisation of TEs in genic region may be limited due to negative selection . On the other hand, the prevalence of UIS may result from an ongoing TE activity. For rice retrotransposon families, Carpentier et al. suggested that the presence of both low- and high-frequency insertion sites indicated continuous transposition, while high numbers of low-frequency insertions indicated their recent mobilisation . In general, most carrot DcSto families produced higher proportions of UIS in wild carrots, as compared to the cultivated carrots. With the absence of any significant domestication bottleneck in carrots [25, 28], such a difference was not expected, unless the transpositional activity of DcStos had been elevated in the wild genepool. Alternatively, one might speculate that TE insertional polymorphisms are a more sensitive indicator of a domestication bottleneck than SNPs, due to non-neutrality of some gene-associated insertions.
DcSto insertion hotspots
This current study showed that more than 1.5% of all DcSto insertion sites were occupied by more than one DcSto element in exactly the same position in different genomes, which we named parallel insertion sites (PIS). The occurrence of different Stowaway MITE insertions in orthologous positions has been reported previously, e.g., it was studied for the β-amylase gene in Poaceae [54, 55]. However, it has never been addressed in the context of whole genomes. In this current study, we showed that it was a relatively frequent phenomenon and new insertions appeared fast enough to produce a series of insertion variants within the species. Notably, different copies at the same insertion site usually came from families sharing more similar terminal sequences. This might suggest that they utilised the same source of transposase, which resulted in parallel targeting to the same chromosomal positions. In carrots, 63% of all PIS were located within 2 kb of the nearest gene. As such, it will be important to reveal if these variants show functional variability in these genes. Recently, the importance of variation sources resulting in the occurrence of parallel mutations has been highlighted .
DcStos as a source of variation in genic regions
Carrot DcStos elements, like other MITEs, were frequently associated with genes. The current study showed that 73% of all DcSto copies were inserted in the vicinity of genes, and particular DcSto families differed in their distribution within genic regions. A similar distribution of MITEs, enriched upstream of TSSs and depleted within the body of genes, was observed for Stowaways elements in potato  and mPing elements in rice . Some 9738 carrot genes, including 61 tRNA genes, were associated with at least one DcSto element. On average, 3% of all annotated genes were associated with DcStos elements in an individual carrot genome, ranging from 337 genes for accession W1 to 1490 genes for accession C9 (Additional file 2: Table S8). It was likely that these insertions were important for the fine-tuning of the expression of these associated genes . Indeed, the non-random association of DcSto insertions with particular groups of genes, most notably transcription factors, indicated functional importance of these associations. However, none of the gene-associated DcSto insertions was fixed in D. carota. Nevertheless, they might provide a rich source of variability in the fine-tuning of certain regulatory networks and constitute a basis for selection. A more extensive sampling across carrot germplasm will be required to verify if some of these insertions show signatures of selection during domestication.
A recent genome-wide analysis of TEs showed that they were very important for rapid genome modifications, providing phenotypic variability important for adaptation. In A. thaliana ecotypes, genes carrying polymorphic TE insertions were enriched for defense and immune response functions important for adaptation to new ecological niches . At least two genes with TE insertions were likely positively selected, contributing to the adaptation of that species. Similarly, TEs were involved in the rapid adaptation and the “genetic paradox of invasion” of C. rubella. In comparison to its outcrossing relative, C. grandiflora, C. rubella promoter regions were enriched in TE sequences . Variability resulting from polymorphic insertion sites of Stowaway MITEs and an altered methylation status of surrounding sequences may impact adaptation to local environmental conditions, as reported for wild emmer wheat [59, 60]. It was likely that DcStos, especially DcSto7b, had contributed to the phenotypic variation of cultivated carrots. Carrot has been domesticated relatively recently ; however, it shows a remarkable diversity of cultivar types and storage root traits . It was tempting to speculate that at least some of the observed variability among carrot cultivars could have resulted from selection on variants resulting from the insertion of DcSto elements.
DcSto7b elements were activated upon domestication
To date, only a few active MITEs have been described for which an accompanying autonomous class II element was proposed. These include Stowaway family dTStu1 element in potato , and Tourist and hAT-related MITE families in rice [7, 21, 64, 65, 66, 67]. The current study indicated that the DcSto7b family had been mobilised in the course of carrot domestication and might still be active in cultivated carrots. The high proportion of UIS of DcSto7b elements in cultivated carrots was notable in relation to the opposite trend for the remaining DcSto families. Several lines of evidence suggested very recent activity by DcSto7b elements, namely the highest PrC value (Table 1), more copies in the cultivated carrot accessions (Fig. 1b), and more UIS as compared to other DcSto families (Fig. 4). This was further supported by the highest intra-family similarity of individual copies of DcSto7b in the DH1 reference genome, as reported previously .
We hypothesised that Dcmar1, a related autonomous Mariner-like element, provided the transposition machinery for the mobilisation of DcSto7b elements. Dcmar1 was present only in a subset of the studied carrot accessions, which showed higher DcSto7b copy numbers, being more frequently present in genomes of cultivated carrots. The insertion site of Dcmar1 in the DH1 reference genome was unique, with the same position being empty in all the remaining 30 plants. Therefore, Dcmar1 itself, was likely a currently active element.
This current study described the landscape of carrot Stowaway MITEs, providing insight into their importance in shaping the structural and functional variability of the carrot genome. Extreme insertional polymorphism of carrot Stowaways was identified, likely resulting from their recent mobilisation, as well as diversification from amplification bursts among carrot accessions. In particular, the DcSto7b family had likely been active in the course of domestication. Moreover, DcSto insertions were commonly present within genic regions, and were non-randomly associated with specific groups of genes, including those encoding transcription factors, with independent insertions of MITEs in the same genomic positions being relatively common events (comprising 1.6% of all insertion sites). Further analyses of carrot MITEs will be needed to understand the mechanisms responsible for their successful amplification and the extent of their functional impact on genes and on the phenotype of carrots.
To identify DcSto insertions, we used sequencing data from 31 resequenced genomes of D. carota (NCBI Sequence Read Archive, accession SRP062070, under umbrella project PRJNA285926; Additional file 2: Table S1), comprising 13 wild and 18 cultivated carrot accessions, along with the assembled carrot reference genome and its raw reads . DNA from the 31 resequenced plants (excluding Ssp3 and including C14) was amplified using a REPLI-g Mini Kit (Qiagen), following the manufacturer’s protocol.
In silico mining of DcSto insertions
Raw reads were pre-processed by removing low quality reads and trimming adapters using Trimmomatic version 0.35 , with parameters minqual = 28, minlen = 50, LEADING:28, TRAILING:28, SLIDINGWINDOW:10:28, and MINLEN:50, and quality was controlled using fastqc .
To identify insertion sites of the 14 DcSto families we used RelocaTE  with consensus sequences representing DcSto families . RelocaTE allowed identification of TE insertions from unassembled short reads. In brief, short reads were aligned to a reference/consensus TE sequence, matching reads were trimmed to remove the TE sequence, and the remaining read fragments were aligned to the reference genome to identify the regions flanking the TE insertions . The following RelocaTE parameters were use: -bm 12, −bt 11, −m 0.2 and -r 1. As the method included a mapping step, we first examined whether there were differences in the percentage of reads aligning to the reference genome. The mapping quality was evaluated with bwa-mem , using previously described parameters . Next, files containing information about insertion sites for each DcSto family/genome combination were merged and converted into a binary matrix using a custom script, with absence and presence of a TE insertion being scored as 0 and 1, respectively.
Due to differences in genome coverage, we calculated correlation between the depth of coverage and the number of identified insertion sites. The Shapiro-Wilk’s normality test was performed, and the non-parametric correlation was tested using Spearman’s rank-based correlation, with the results were plotted in R using the ‘ggpubr’ package v.0.2 .
A binary matrix for the 31 accessions was used to calculate the number of DcSto insertion sites, UIS, i.e., those present in only a single accession, and the number of PIS, i.e., those with different copies of DcSto elements inserted in different genomes at exactly the same position. Genomic distribution of DcSto insertion sites and genes was plotted using the ‘ggplot2’ R package .
The presence of DcSto insertions in the context of genic regions, divided into five categories of 2 kb upstream sequences, 5’UTRs, coding sequence (cds), introns, 3’UTRs, and 2 kb downstream sequences, were determined based on the National Center for Biotechnology Information (NCBI) carrot genome annotation file GCF_001625215.1_ASM162521v1_genomic.gff, using BEDTools v.2.26.0 . The same resource was used to calculate the total length of each of the five genic categories. Singular enrichment analysis (SEA) of the DcSto-associated genes was carried out using the Phytozome annotation file (Dcarota_388_v2.0.annotation_info.txt) and AgriGO v.2.0 , to define biological processes (BP), cellular components (CC) and molecular functions (MF).
All correlation tests were calculated and plotted using the ‘Corrplot’ R package . Family distribution of DcSto insertion sites within the five genic categories was calculated based on a contingency table of data representing the number of occurrences of each DcSto family in defined segments, using the Pearson chi-squared test. Due to a low number of DcSto11 insertions, a simulated p-value based on 2000 replicates was used. Pearson residuals were calculated using a contingency table containing data representing the total number of insertion sites of each DcSto family in individual genomes. The matrix of Pearson’s correlation coefficients was calculated to test interconnection between the sum of copy numbers for families that were inserted into the same position (PIS), the number of their common occurrences in PIS, and the genetic distance between each pair of DcSto consensus sequences. Intra-family genetic distance was calculated for all copies representing each family identified in the DH1 genome, as reported by Iorrizo et al. .
The binary matrix for the 31 genomes was used to calculate the genetic distance based on the Jaccard coefficient, with the ‘vegan’ R package . This was a conservative approach, where only the presence of a common insertion was considered informative. The values were used for principal coordinate analysis (PCoA) using the ‘ape’ package in R .
Finally, a gff3 file was prepared, where for each insertion the ‘start position’ referred to the second nucleotide (A) of the target site (TA), while the ‘end position’ referred to the first nucleotide of the DcSto element, in the case of insertions present in the reference genome DH1, or to the first nucleotide following the target site, in the case of insertions not mapped to DH1. For each insertion, a note containing information about its genomic position was given, as well as the LOC number of the adjacent gene, when the DcSto copy was inserted in a genic region (less than 2 kb from the gene). The ID field contained information about the DcSto family to which the copy was attributed, and comma separated codes of accessions carrying the insertion.
Identification of autonomous elements
Autonomous mariner-like elements were mined from the DH1 reference genome assembly using TIRfinder , with tirMask: CTCCCTYYSKYMC, tsdMask: TA, tirSeqMismatches: 1, tsdSeqMismatches: 0, tirMaskMismatches: 0 and tsdMaskMismatches: 0. Coordinates and sequences of identified elements were manually inspected to remove redundant sequences. FGENESH , GENEID  and Augustus  gene prediction tools were used to identify coding regions in all mined TE sequences.
Predicted proteins in TEs were aligned with transposase sequences of known plant mariner-like elements, from Ppmar1 (NCBI accession no. HM581665), Soymar1 (NCBI accession no. AF078934) and OSMAR1 (Repbase accession no. AC135425), using ClustalW . The presence of a highly conserved fragment of the mariner transposase starting from the first two aspartic acids of the DDD motif, previously used for phylogenetic analysis of plant mariner-like transposases , was manually inspected. Elements lacking the DDD motif were removed from further analysis. For the remaining proteins of putative autonomous elements, HTH motifs  and iron binding sites  were identified. The basic local alignment search tool (BLAST) was used to compare the corresponding mRNAs with carrot DH1 RNAseq short reads from 20 tissues (Sequence Read Archive SRP062159) .
Phylogenetic analysis of putative autonomous mariner-like and DcStos elements was conducted with Mega v.6.06 software . Evolutionary distances were computed based on 50 nt-long sequences of both TIRs using the p-distance method , and were used to calculate a neighbour joining tree . Bootstrap values were obtained based on 1000 replicates.
To identify genomic positions of Dcmar1 elements in the resequenced genomes, cleaned Illumina reads were analysed by the TRACKPOSON method . One cultivated carrot accession, I4, was not included in the analysis, as only forward reads were available. To avoid false positives from DcSto MITEs, TIRs were removed from the Dcmar1 query sequence prior to analysis, leaving only the internal portion of the sequence specific to the Dcmar1 element. In order to precisely determine genomic positions, sequences flanking Dcmar1 elements were reconstructed from unmapped paired reads, manually verified, and aligned with the DH1 carrot reference genome using BLAST analysis. The presence of Dcmar1 TIRs in the reconstructed sequences provided a confirmation of the results of in silico mining.
Experimental verification of DcSto insertion sites identified by RelocaTE analysis
Thirty-nine DcSto insertion sites located in introns, and six sites characterised by parallel insertions, were selected for validation. For PCR, site-specific primers were as described by Stelmach et al. , or were designed de novo using Primer3  (Additional file 2: Table S9). Reaction mixes contained about 20 ng REPLI-g-amplified genomic DNA, 1 mM forward and reverse primers, 0.25 mM dNTPs (Thermo Fisher Scientific), 0.5 U Taq DNA polymerase (Thermo Fisher Scientific), and 1x Taq buffer with MgCl2 (Thermo Fisher Scientific). Amplification took place at 94 °C for 1 min, followed by 30 cycles of 94 °C for 30 s, 56 °C/58 °C for 30 s, and 68 °C for 2 min, and finally 68 °C for 6 min. Products were separated by 1% agarose gel electrophoresis, and were purified with a GeneJET Gel extraction kit (Thermo Fisher Scientific), and cloned into pGEM-T (Promega). Cloned DNAs were extracted using the Wizard Plus SV Miniprep DNA Purification System (Promega) and sequenced by the Sanger method (Genomed SA, Poland). Nucleotide sequences were manually aligned using BioEdit .
The presence of Dcmar1 elements in D. carota accessions was verified using a pair of primers, DcMar1_499_F: 5′ GCC GAC ATA CGA ATC CTG TCA 3′ and DcMar1_499_R: 5′ TTG TGG CTT CCT TCT GCT GTA 3′, anchored in the DDD domain of the Dcmar1 element. The presence of Dcmar1 in the DH1 insertion site was screened across D. carota accessions with one of the above DDD-anchored primers in combination with a corresponding forward or reverse primer flanking the insertion (DcMar1_499_flank_F: 5′ TGT TCT TAG CAG CGG TAG CAC and DcMar1_499_flank_R: 5′ GTT GGT GTT TAC ACT GGA GGT TG 3′). As a positive control for the PCRs, a single-copy carrot genomic fragment was amplified with primers CULT-q-orf6-F 5′ CTT CTC GTA CAA CTG AGC C 3′ and CULT-q-orf6-R 5′ GCT TAG CAA GTA CAA GGG AA 3′ . Fragments were amplified in 10 μl reactions containing 20 ng REPLI-g-amplified genomic DNA, 1 mM forward and reverse primer, 1 mM forward and reverse control primer, 0.25 mM dNTPs (Thermo Fisher Scientific), 0.5 U Taq DNA polymerase (Thermo Fisher Scientific), and 1x Taq buffer with MgCl2 (Thermo Fisher Scientific). Amplification took place at 94 °C for 1 min, followed by 30 cycles of 94 °C for 30 s, 56 °C for 30 s, and 68 °C for 2 min for the DDD test and 10 min for the DH1 site, and then 68 °C for the final elongation for 6 min for the DDD test and 20 min for the DH1 site.
AM-P and DG designed the study; AM-P performed in silico analyses; AM-P, KS and KK performed laboratory verification; AM-P and DG edited the manuscript. All authors read, reviewed, and approved the final manuscript.
The research was financed from (1) funds for basic research on crop improvement granted by the Polish Ministry of Agriculture and Rural Development, (2) MINIATURA1 (grant number 2017/01/X/NZ9/00930) granted by the Polish National Science Center, and (3) funds for the statutory activity of the Faculty of Biotechnology and Horticulture, University of Agriculture in Krakow, granted by the Polish Ministry of Science and Higher Education.
Ethics approval and consent to participate
Consent for publication
The authors declare that they have no competing interests.
- 5.Loot C, Santiago N, Sanz A, Casacuberta JM. The proteins encoded by the pogo-like Lemi 1 element bind the TIRs and subterminal repeated motifs of the Arabidopsis Emigrant MITE: consequences for the transposition mechanism of MITEs. Nucleic Acids Res. 2006;34(18):5238–46.PubMedPubMedCentralCrossRefGoogle Scholar
- 14.Stuart T, Eichten SR, Cahn J, Karpievitch YV, Borevitz JO, Lister R. Population scale mapping of transposable element diversity reveals links to gene regulation and epigenomic variation. eLife. 2016;5:e20777.Google Scholar
- 26.Simon P.W. Economic and Academic Importance. In: Simon P, Iorizzo M, Grzebelus D, Baranski R, editors. The Carrot Genome. C. Cole, series editor. Compendium of Plant Genomes. Springer Nature Switzerland AG; 2019:1–8.Google Scholar
- 27.Spooner D.M. Daucus: Taxonomy, Phylogeny, Distribution. In: Simon P, Iorizzo M, Grzebelus D, Baranski R, editors. The Carrot Genome. C. Cole, series editor. Compendium of Plant Genomes. Springer Nature Switzerland AG; 2019:9–26.Google Scholar
- 35.Kuang H, Padmanabhan C, Li F, Kamei A, Bhaskar PB, Ouyang S, et al. Identification of miniature inverted-repeat transposable elements (MITEs) and biogenesis of their siRNAs in the Solanaceae: new functional implications for MITEs. Genome Res. 2009;19(1):42–56.PubMedPubMedCentralCrossRefGoogle Scholar
- 46.Robb SM, Lu L, Valencia E, Burnette JM, Okumoto Y, Wessler SR, et al. The use of RelocaTE and unassembled short reads to produce high-resolution snapshots of transposable element generated diversity in rice. G3-Genes Genome Genet. 2013;3(6):949–57.Google Scholar
- 50.Quadrana L, Silveira AB, Mayhew GF, LeBlanc C, Martienssen RA, Jeddeloh JA, et al. The Arabidopsis thaliana mobilome and its impact at the species level. eLife. 2016;5:e15716.Google Scholar
- 61.Ellison S. Carrot domestication. In: Simon P, Iorizzo M, Grzebelus D, Baranski R, editors. The Carrot Genome. C. Cole, series editor. Compendium of Plant Genomes. Springer Nature Switzerland AG; 2019:77–92.Google Scholar
- 62.Simon PW, Freeman RE, Vieira JV, Boiteux LS, Briard M, Nothnagel T, et al. Carrot. In: Prohens J, Carena MJ, Nuez F, editors. Handbook of crop breeding, vol 1. Vegetable breeding. Heidelberg: Springer; 2008. p. 327–57.Google Scholar
- 69.Andrews S. FastQC: a quality control tool for high throughput sequence data. 2010. http://www.bioinformatics.babraham.ac.uk/projects/fastqc of subordinate document. Accessed 20 Aug 2019.Google Scholar
- 80.Blanco E, Parra G, Guigó R. Using geneid to identify genes. Curr Protoc Bioinformatics. 2007;18(1):4–3.Google Scholar
- 87.Nei M, Kumar S. Molecular evolution and phylogenetics. Oxford: Oxford University Press; 2000.Google Scholar
- 89.Rozen S, Skaletsky H. Primer3 on the WWW for general users and for biologist programmers. In: Bioinformatics methods and protocols. Totowa: Humana Press; 2000. p. 365–86.Google Scholar
- 90.Hall TA. BioEdit: a user-friendly biological sequence alignment editor and analysis program for windows 95/98/NT. Nucleic Acids Symp Ser. 1999;41(41):95–8.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.