The linear mitochondrial genome of the quarantine chytrid Synchytrium endobioticum; insights into the evolution and recent history of an obligate biotrophic plant pathogen

Background Chytridiomycota species (chytrids) belong to a basal lineage in the fungal kingdom. Inhabiting terrestrial and aquatic environments, most are free-living saprophytes but several species cause important diseases: e.g. Batrachochytrium dendrobatidis, responsible for worldwide amphibian decline; and Synchytrium endobioticum, causing potato wart disease. S. endobioticum has an obligate biotrophic lifestyle and isolates can be further characterized as pathotypes based on their virulence on a differential set of potato cultivars. Quarantine measures have been implemented globally to control the disease and prevent its spread. We used a comparative approach using chytrid mitogenomes to determine taxonomical relationships and to gain insights into the evolution and recent history of introductions of this plant pathogen. Results We assembled and annotated the complete mitochondrial genome of 30 S. endobioticum isolates and generated mitochondrial genomes for five additional chytrid species. The mitochondrial genome of S. endobioticum is linear with terminal inverted repeats which was validated by tailing and PCR amplifying the telomeric ends. Surprisingly, no conservation in organisation and orientation of mitochondrial genes was observed among the Chytridiomycota except for S. endobioticum and its sister species Synchytrium microbalum. However, the mitochondrial genome of S. microbalum is circular and comprises only a third of the 72.9 Kbp found for S. endobioticum suggesting recent linearization and expansion. Four mitochondrial lineages were identified in the S. endobioticum mitochondrial genomes. Several pathotypes occur in different lineages, suggesting that these have emerged independently. In addition, variations for polymorphic sites in the mitochondrial genome of individual isolates were observed demonstrating that S. endobioticum isolates represent a community of different genotypes. Such communities were shown to be complex and stable over time, but we also demonstrate that the use of semi-resistant potato cultivars triggers a rapid shift in the mitochondrial haplotype associated with increased virulence. Conclusions Mitochondrial genomic variation shows that S. endobioticum has been introduced into Europe multiple times, that several pathotypes emerged multiple times, and that isolates represent communities of different genotypes. Our study represents the most comprehensive dataset of chytrid mitogenomes, which provides new insights into the extraordinary dynamics and evolution of mitochondrial genomes involving linearization, expansion and reshuffling. Electronic supplementary material The online version of this article (10.1186/s12862-018-1246-6) contains supplementary material, which is available to authorized users.


(Continued from previous page)
Results : We assembled and annotated the complete mitochondrial genome of 30 S. endobioticum isolates and generated mitochondrial genomes for five additional chytrid species. The mitochondrial genome of S. endobioticum is linear with terminal inverted repeats which was validated by tailing and PCR amplifying the telomeric ends. Surprisingly, no conservation in organisation and orientation of mitochondrial genes was observed among the Chytridiomycota except for S. endobioticum and its sister species Synchytrium microbalum. However, the mitochondrial genome of S. microbalum is circular and comprises only a third of the 72. 9 Kbp found for S. endobioticum suggesting recent linearization and expansion. Four mitochondrial lineages were identified in the S. endobioticum mitochondrial genomes. Several pathotypes occur in different lineages, suggesting that these have emerged independently. In addition, variations for polymorphic sites in the mitochondrial genome of individual isolates were observed demonstrating that S. endobioticum isolates represent a community of different genotypes. Such communities were shown to be complex and stable over time, but we also demonstrate that the use of semi-resistant potato cultivars triggers a rapid shift in the mitochondrial haplotype associated with increased virulence.
Conclusions: Mitochondrial genomic variation shows that S. endobioticum has been introduced into Europe multiple times, that several pathotypes emerged multiple times, and that isolates represent communities of different genotypes. Our study represents the most comprehensive dataset of chytrid mitogenomes, which provides new insights into the extraordinary dynamics and evolution of mitochondrial genomes involving linearization, expansion and reshuffling.
Keywords: Mitochondrial haplotypes, Pest introduction, Population dynamics, Fungal communities, Pathotype formation, Chytridiomycota Background Synchytrium endobioticum (Schilb.) Perc. is a chytrid fungus (Chytridiomycota) causing potato wart disease, which is one of the most important quarantine diseases on cultivated potato [1,2]. Host tissues infected with this fungus develop tumour-like galls (i.e. warts) rendering the crop unmarketable. The malformed tissue produces resting spores which are released into the surrounding soil when the host tissue decays, and once released, spores can remain viable and infectious for decades [3,4]. Individual isolates can be further characterized as pathotypes based on their virulence on a differential set of potato cultivars. Reported from many countries worldwide, S. endobioticum is believed to originate from potato-growing areas in the Andean region from where it was taken into the United Kingdom in the aftermath of the Irish potato famine (~1880s) [5]. Initially, the disease spread quickly and widely in Europe, but phytosanitary measures restricted its movement to other parts of the world [2,5]. Today, quarantine measures have been implemented through phytosanitary legislation world-wide. In addition, S. endobioticum is on the federal select agent program of the United States of America [6]. Restricting potato cultivation in infested fields and the use of resistant potato cultivars are the only known efficient control strategies. Characterization of potato wart resistance in potato cultivars is essential for efficient control.
Forty-five years after the first description of S. endobioticum [7], only a single pathotype was known, which is nowadays known as pathotype 1(D1). In 1941, wart development was discovered on formerly resistant potato cultivars (Edda, Edelragis, Parnassia, Primula, Sabina, and Sickingen) in Gießübel, Germany [8] and it was recognized that a new pathotype, pathotype 2(G1), of the fungus had emerged. Subsequently, new pathotypes were found in Germany, the Czech Republic, and Ukraine [9], and at present, 39 pathotypes are described with the most recent one being discovered in Piekielnik, Poland [10]. Although pathotype identity is essential for authorities to enforce suitable phytosanitary measures, these tests do not provide information on the migration and spread of the fungus, or the emergence of new pathotypes.
To infer intraspecific genetic variation, Gagnon et al. [11] developed polymorphic microsatellite loci for S. endobioticum. However, isolates of the same pathotype were found to display highly variable genotypes and none of the markers correlated with a specific pathotype. Apart from several Canadian isolates, no clustering based on geographical origin was observed. We have pursued a different approach, exploiting the mitochondrial genome to determine taxonomical relationships, and to gain insights into the introductions and population dynamics of this plant pathogen.
The mitochondrion is a cell organelle involved in ATP production, and is descended from an α-proteobacterium [12]. They have maintained the double membrane characteristic of their ancestors and retained a highly reduced bacterial-like genome [13,14]. The mitochondrial genome is typically assumed to be a single maternally inherited circular molecule that is present in high copy number, ranging from a few hundred to a few thousand copies per cell. Loci from the mitochondria are most frequently used in phylogenetic and population-level studies [15][16][17], because of their copy number, commonly available generic primers for amplification and sequencing, and phylogenetic signal on higher taxonomical levels [18,19]. With the introduction of next generation sequencing technologies, the number of complete mitochondrial genomes has rapidly increased [14]. However, complete mitochondrial genomes of the early diverging fungal lineages of the Chytridiomycota and Blastocladiomycota are scarce with only eight publically available complete and annotated mitochondrial genomes. This limits the possibilities for species level analyses using the mitochondrial genome in these divisions.
The aims for this study were to: (i) sequence, assemble and annotate the complete mitochondrial genome of S. endobioticum; (ii) perform phylogenetic analyses using the mitochondrial genes to determine its taxonomical relationships to 14 other species from the Chytridiomycota including Synchytrium taraxaci, which is the type species for the genus [20], and Synchytrium microbalum, the first culturable Synchytrium species to be described [21]; and (iii) perform network analyses using a panel of S. endobioticum isolates from different countries covering several pathotypes and identify potential linkages between genotypes, geographical origins and pathotypes.

Results
Assembly of the S. endobioticum mitochondrial genome We assembled and annotated the mitochondrial genome of S. endobioticum pathotype 1(D1) isolate MB42. Three independent assemblies using different datasets were performed and the results were combined to obtain a consensus mitochondrial genome; (i) from the MB42 reference genome, seven putative mitochondrial scaffolds were identified representing a total size of 68,068 base pairs (bp); (ii) using GRAbB, a single 63,627 bp putative mitochondrial scaffold was obtained; (iii) an assembly of PacBio data resulted in five putative mitochondrial scaffolds with a total size of 90,248 bp. Putative mitochondrial scaffolds were combined to create a single 72,865 bp consensus assembly. The assembly was verified by mapping the longread PacBio data to the consensus sequence which resulted in contiguous sequence with no reads that exceed the 5′ or 3′ terminal ends (Additional file 1: Figure S1). No circular mapping was obtained for the linear mtDNA, i.e. no reads were found that spanned the terminal ends of the consensus sequence, and terminal deoxynucleotidyl transferase (TdT) tailing followed by PCR amplification further verified the linear ends of the molecule. The ends of the linear molecule contain two identical 3529 bp repeats in inverted orientation, which we refer to as Terminal Inverted Repeats (TIR) (Fig. 1, Additional file 1: Figure S2).
The annotation of the S. endobioticum mitochondrial genome recovered the mitochondrial genes typically found in fungal species: atp6, atp8, atp9, cob, cox1, cox2, cox3, nad1, nad2, nad3, nad4, nad4L, nad5, nad6, rns and rnl. In addition, a reduced set of five tRNA recognising methionine, tryptophan, tyrosine, proline and leucine was found. Four homing endonuclease genes (HEGs) were identified encoded by introns of protein coding genes: three HEGs residing in cox1 introns carry the LAGLIDADG_1 Pfam domain (PF00961), whereas the HEG residing in the cob intron carries the LAGLIDADG_2 Pfam domain (PF03161). No presence/absence variations were found for the intron and HEGs content in other S. endobioticum isolates. Interestingly a gene coding for DNA polymerase type B (dpoB) containing the DNA_pol_B_2 Pfam domain (PF03175), which are also found in linear mitochondrial plasmids, was identified on the S. endobioticum mitogenome. The dpoB gene is partially repeated in the A + T rich region (Additional file 1: Figure S3).
Intergenic regions are found to have an unusual high G + C content (57.3%), with several intergenic spacers having G + C contents close to 70% (e.g. nad4-nad2: 68.5%, nad2-cob: 67.6%, and cox3-nad4: 65.5%). In contrast, the intergenic regions in the mitochondrial genome of a closely related species, S. microbalum, consist of 28.6% G + C, with the highest G + C percentage found in the nad6-nad4L intergenic spacer (34.2%). S. endobioticum has the highest intergenic G + C content in chytrid mitogenomes reported to date, which is only almost matched by the obligate biotrophic S. taraxaci (51.4%). For Chytridiomycota other than S. endobioticum and S. taraxaci, a mean intergenic G + C content of 34.3% (standard deviation 9.7%) was observed.
Of the five additional chytrid species sequenced in this study, three resulted in a single circular mtDNA scaffold containing all 14 mtDNA genes, i.e. C. confervae (225.6 kb), S. palustris (126.5 kb), and S. microbalum (23.8 kb). The assembly of P. hirtus resulted in three mtDNA scaffolds with a total size of 298.6 kb, in which all mtDNA genes were detected except for nad6. With three scaffolds we could not determine if the mitochondrial genome is circular. The mtDNA assembly of the obligate biotrophic species S. taraxaci extremely fragmented with 14 mtDNA scaffolds with a total size of 39.2 kb coding for all mtDNA genes. Annotation of the mtDNA scaffold of B. dendrobatidis JEL423 (DS022322; single scaffold, 175.3 Kb) resulted in the detection of all mtDNA genes except for nad2 (Additional file 2: Table S1).
Verification of the S. endobioticum linear mtDNA PCR amplification of the poly-C tailed terminal end to a primer site in the TIR (mtDNA_00787-rv), performed to verify the linearity of the mitogenome, resulted in an amplicon length of expected size (~850 bp). Sanger sequencing of the amplicon and alignment to the mitochondrial genome showed a perfect fit to the TIR region.
Long-range PCR amplification from the telomeric ends, over the TIR sequence, to a 5′ or 3′ specific primer site (mtDNA_03691-rv and mtDNA_69209-fw), performed to verify the presence of identical TIRs and 5′ and 3′ specific sequences, resulted in amplicons of the expected length (~3680 bp, and~3710 bp respectively), and Sanger sequence data of the amplicons verified the presence of the repeat on both telomeric ends of the linear molecule (Fig. 2). A linear concatemeric structure of the S. endobioticum mitochondrion can be ruled out as no sequence data could be mapped beyond the 5′ and 3′ ends, and no amplicons were obtained using a primer pointing outwards in the TIR (mtDNA_00787-rv). This primer anneals to the TIR and would act as both forward and reverse primer in a PCR reaction bridging the ends of both TIRs should a concatemeric structure exist.

Bayesian inference of phylogeny
A 16,329 nt alignment (including gaps) was constructed from concatenating individual mitochondrial genes from 44 isolates. The Bayesian phylogeny showed all nodes with a posterior probability of 1.00 (Fig. 3).

Mitochondrial haplotypes and intraspecies variation
Reference assembly of 29 additional S. endobioticum isolates to the MB42 mtDNA resulted in 72,941 bp to 72,775 bp mitogenomes with 50 to 163,200× mean coverage. Alignment of all 30 S. endobioticum mitogenomes resulted in a 73,030 bp alignment (including Fig. 1 Assembly and annotation of the linear S. endobioticum mtDNA genome. Tracks ❶ to ❸ represent putative mtDNA scaffolds from three different assemblies (respectively putative mtDNA scaffolds from the S. endobioticum MB42 genome assembly, reference guided assembly with GRAbB using S. endobioticum MB42 HiSeq data, and a reference guided assembly using S. endobioticum MB42 PacBio CCS) mapped to the linear mtDNA genome. Darker shades are used to indicate regions where scaffolds from the same assembly overlap. The narrow line in track 2 indicates a gap relative to the linear mtDNA genome. ❹ G + C content determined with a 100 bp sliding window in which orange ≤40% G + C, light green 40-50% G + C, green 50-60% G + C, and dark green > 60% G + C. ❺ mtDNA annotation track showing genes (green), rRNA (red), tRNA (purple) and the terminal inverted repeats (orange). All genes, rRNAs and tRNAs are orientated in the 5′ to 3'direction. ❻ the 72,865 bp linear mitochondrial genome of S. endobioticum gaps). Sequence homology between isolates was high at ≥99.21% and sequence variation was found mostly in the intergenic regions (96.8% of the variation). Two polish isolates, 03WS and 04WS, had identical mitochondrial genome sequences. Extraction of variable sites resulted in a 347 bp alignment (including gaps) that contained 122 parsimony-informative sites, which were used to determine mitochondrial haplotypes. Four main groups (i.e. mitochondrial lineages) are distinguished by > 20 mutations (Fig. 4). Clustering based on polymorphic sites in coding sequences of mtDNA genes showed a similar clustering with lower resolution (Additional file 1: Figure S5).

Composition of S. endobioticum populations
Read mapping to the mitochondrial genome showed unexpected variations of SNP frequency for parsimonyinformative sites for the majority of the sequenced S. endobioticum isolates. Next to the dominant type, low levels of the polymorphic base were observed. For instance, the polymorphism at position 5803 (reference: A, alternative: C) is fixed (> 98%) in nine isolates from group 1, but this SNP is also present in variable frequencies (12-60%) in seven isolates from groups 1, 2 and 3. Similar differences were found for all 122 parsimony-informative sites (Additional file 2: Table S2). In ten isolates from group 1 originating from Canada, Poland and the Netherlands, no signatures for multiple haplotypes were found (Fig. 4). An exception is found in a 40 bp hypervariable intergenic (atp9-cox1) region containing 13 polymorphic sites which was found to be present in all S. endobioticum isolates.

Pathotype formation
The dynamic process of pathotype formation is clearly illustrated by pathotype 6(O1) isolate E/II/2015. This isolate was obtained after two multiplications of pathotype 1(D1) isolate 01WS on the partial resistant potato Fig. 2 Verification of linearity of the S. endobioticum mtDNA. a Design of the verification experiments using a graphical representation of the 5′ and 3' TIRs (orange) and forward and reverse primer sites (green and light green). Three PCR reactions, performed after TdT tailing, are displayed: ❶ M13F_polyG/mtDNA_00787-fw to verify linear conformation of the mtDNA. This reaction takes place at both telomeric ends since the TIRs are inverse orientated. ❷ mtDNA_0007-fw/mtDNA_03691-rv to verify presence of the TIR with its specific flanking sequence on the 5′ end of the mtDNA. ❸ mtDNA_0007-fw/mtDNA_69209-fw to verify presence of the TIR with its specific flanking sequence on the 3′ end of the mtDNA. b Gel images corresponding with reactions described under A. The 1 kb-plus size marker (M) was used for amplicon size estimation. c Sanger sequence data mapped to the S. endobioticum mtDNA corresponding with the reactions described under A. The mtDNA is used as reference, and bases similar to the reference are shown in grey, and differences to the reference are highlighted (black). Phred scores for the individual peaks are shown as blue bars, and low quality data (Phred < 30) is annotated in pink. For reaction 1, the first 800 bases of the 5′ end TIR are shown, whereas for reactions 2 and 3 the full TIR including the specific flanking sequences are shown (~4 kb) Fig. 3 Bayesian tree (GTR model, G + I distributed sites) based on a concatenated alignment of atp6, atp8, atp9, cob, cox1, cox2, cox3, nad1, nad2, nad3, nad4, nad4L, nad5, and nad6 of 43 chytrid isolates with A. macrogynus (Blastocladiomycota) serving as outgroup. Bayesian posterior probabilities are displayed at branch nodes. Highlighted in bold is the S. endobioticum mtDNA reference isolate MB42. Species with linear mtDNA genomes are indicted with an asterisk "*" cultivar Erika in 2015. Before multiplication on this partial resistant cultivar a clear pathotype 1(D1) phenotype was obtained using the EPPO differential set, whereas after multiplication on the partial resistant potato cultivar, a clear pathotype 6(O1) phenotype was obtained ( Table 1). The resulting isolate E/II/2015 is virulent on cultivars Producent and Combi. Repeating the experiment, i.e. multiplying isolate 01WS on cultivar Erika, resulted in the same increased virulence and the same pathotype 6(O1) pathotype profile was obtained. After multiplication, the within isolate variation of E/II/2015 is higher compared to the 01WS isolate.

Discussion
We assembled and annotated the mitogenomes of 30 S. endobioticum isolates and five additional chytrid species, and compared these to publically available Chytridiomycota mitogenomes. Using the mitogenome sequences, we determined taxonomical relationships of S. endobioticum to 14 other chytrid species, and to performed network Median Joining haplotype network based on 141 polymorphic sites on the mitochondrial genome of S. endobioticum, of which 122 were parsimony-informative. Nodes in the network are coloured based on pathotype identity. Colours are used for pathotypes of major importance in Europe and Canada, whereas a greyscale is used for pathotypes of lesser importance [9]. Black nodes represent hypothetical ancestors. Marks on the branches indicate the number of mutations, and the numbers are shown on branches with > 5 mutations. No signatures for multiple mitochondrial haplotypes were detected for isolates with an asterisk "*" Table 1 Virulence profiles of S. endobioticum isolates 1(D1)_01WS and E/II/2015 using differential cultivars described in EPPO PM7/28 (1) Susceptible interactions (wart formation) are indicated with "S", intermediated reactions (sprouts partially proliferated) are indicated with "±", and resistant interactions are indicated with "-" analyses to identify potential linkages between S. endobioticum genotypes, geographical origins and pathotypes.
The linear mitogenome of S. endobioticum Mitochondrial genomes are typically assumed to have a circular monomeric conformation. However, after the description of a linear mitochondrial genome in Tetrahymena pyriformis [22], other linear mitochondrial genomes were reported in fungi, oomycetes, ciliates, Chlorophycean green algae and cnidarians, [23][24][25]. From our analyses, the S. endobioticum mitogenome was found to be linear, encoding a dpoB homolog and having two identical TIRs.
In contrast, the mitochondrial genome of the sister species S. microbalum is circular, lacks a dpoB homolog and comprises only a third of the size of the S. endobioticum mitogenome, suggesting recent linearization and expansion in the latter species. Integration of linear mitochondrial plasmids have been hypothesised to be cause of linearization of mitogenomes in medusozoan cnidarians [24] and yeasts [25], and the dpoB gene identified in the S. endobioticum genome is believed to be reminiscent of such event. Among chytrid fungi, a linear mitochondrial genome is described for H. curvatum [26] and B. dendrobatidis [27], of which the latter consists of three linear segments, each having inverted repeats at the termini and contains a dpoB ortholog.
As was found in other chytrid species [28,29], S. endobioticum contains a reduced set of tRNAs. The tRNA content of the mitochondrial genome determines the genetic code that applies for a given mitogenome. In particular, the presence of trnL(CTA), translating TAG to Leucine instead of introducing a stop codon, indicates that the Chlorophycean Mitochondrial Code (tbl16) applies to the mitogenomes of the species assembled and annotated in this study. The organisation and orientation of the mitochondrial genes, tRNAs and rRNAs are conserved between the linear mtDNA of S. endobioticum and the circular mitochondrial genome of S. microbalum (Additional file 1: Figure S4). Otherwise, conservation was observed neither in the organisation nor orientation of mitochondrial genes in the Chytridiomycota, underlining the long evolutionary distance among these species.

Bayesian inference of phylogeny
Clustering of species in their respective orders is identical compared to the nuclear rDNA based phylogeny (18S + 5.8S + 28S subunits) as described by James et al. [30], supporting the polyphyly in the order Chytridiales and monophyly in the genus Synchytrium as described by Smith et al. [31]. This further underlines the use and successful application of whole mitochondrial genomes for phylogeny providing the required resolution and robustness in such a wide and scarcely populated evolutionary branch.
The capacity of S. endobioticum for natural dispersion is limited, and the pest is mainly spread as a result of human crop exchange and agricultural activities [2,32]. Our observation that S. endobioticum isolates with identical multi-gene mitochondrial haplotypes can have a global distribution corroborates with distribution as a result of human activities. If regional or worldwide distribution would pre-date agricultural development, a much larger variation in mitochondrial haplotypes between isolates from distant locations would be expected. Based on the mitochondrial lineages identified, we conclude that S. endobioticum has been introduced at least three times in Europe, and that from those introductions the disease has spread in Europe and to other continents. Isolates from pathotype 18(T1), identified first in Germany in 1978 [33], cluster in mtDNA group 3, suggesting that the emergence of this pathotype was the result of a new introduction opposed to the gain of increased virulence of isolates already present in Europe. The forth group is represented by a single isolate originating from Peru and fits well with the hypothesis that more genetic diversity can be expected from the Andean region, the presumed centre of origin of this pathogen.
Pathotype formation in S. endobioticum is a dynamic and ongoing process. This is supported by the occurrence of the same pathotypes in different mitochondrial lineages, e.g. pathotypes 2(G1) and 6(O1) being found in both mitochondrial groups 1 and 2. Our data suggests that the emergence of those pathotypes have occurred independently in a different genetic background. Different SNP frequencies in read mappings demonstrate the presence of more than one mitochondrial haplotype in the majority of isolates. This was not observed in the other chytrid species sequenced in this study, with the exception of S. taraxaci; a closely related species with a similar obligate biotrophic pathogenic life style on its host dandelion. The number of SNPs was not high enough for the size of the fragments to assemble the individual haplotypes for isolates MB42 and LEV6574 using long read PacBio sequences and the recently published Canu assembler [34]. The composition of the S. endobioticum communities of different genotypes seem to be conserved. Isolates from group 1 typically portray low levels of diversity, whereas isolates from group 3 have high levels of diversity (Additional file 1: Figure S6). Ten isolates clustering together in group 1 are strikingly scarce in diversity, apart from the 40 bp atp9-cox1 intergenic region which found to be variable in all isolates. We suspect that this region could represent the origin of replication which may involve an RNA intermediate as was described for linear chromosomal fragments [35].
Inoculation experiments show that the use of semi-resistant potato cultivars triggers a rapid shift (increase) in virulence, which was associated with a shift in the mitochondrial haplotypes found in the isolate. In fact an increase in diversity was observed. Most likely this variation was already present, but below the detection limit of the sequencing depth. We postulate that this increased virulence was the result of a shift in the heterogeneous 1(D1) population of S. endobioticum due to the selection of virulent genotypes already present in the isolate. Functional variation could be present on the nuclear genome as we found evidence for genetic heterogeneity not only on the mitochondrial genome but also on some nuclear scaffolds in all sequenced isolates (data not shown). The alternative hypothesis that many new mutations could be induced during the two multiplications seems highly unlikely. The possible rapid increase in virulence due to selection of the population underlines that caution is needed when adopting the deployment of R genes in a control strategy, in particular when they provide partial or QTL based resistance.

Conclusions
The mitochondrial genome of S. endobioticum and other Chytridiomycota provides insight into the evolution of mitochondria. Of special interest are the linearization of what is often assumed to be a circular genome which happened multiple times during the evolution of Chytridiomycota, the extremely variable size range (19,5 to 298,5 Kb), and the remarkable high GC content for the intergenic regions (close to 70%). The mitochondrial genome also provides insight in the more recent history of this plant pathogen. From this study we conclude there were at least three independent introductions in Europe, most likely as the result of human activities. With one of those introductions, pathotype 18(T1) was introduced into Europe and so far this pathotype seems to be confined to Europe. A fourth mtDNA lineage obtained from a Peruvian isolate represents a genotype not found in other S. endobioticum isolates sequenced so far. Mitochondrial data shows that S. endobioticum isolates should not be considered as a single genotype, but rather populations of different genotypes. The observed population diversity seems to be consistent in the four different mitochondrial lineages identified in this study. Also, we conclude that pathotypes 2(G1) and 6(O1) have emerged at least twice independently, most likely by selection of virulent genotypes from diverse population as was observed for a pathotype 6(O1) derivative from a pathotype 1(D1) isolate.
Our data indicate that characterization by pathotyping alone is insufficient for characterising isolates, as underlying diversity may remain unnoticed. To keep track of genetic drift and selection of individuals from such populations we recommend sequencing of isolates before and after each multiplication, particularly when (partial) resistant potato cultivars are used. Obtaining genetically pure cultures may not be possible since S. endobioticum is an obligate biotroph. This situation may be similar for other plant pathogens in this genus as illustrated by the fact that a heterogeneous genotypic profile in the mtDNA of S. taraxaci was observed. To our knowledge S. endobioticum represents a unique case where although the number of (new) outbreaks of the disease is low, the genetic diversity is high.

Fungal material
A total of 30 S. endobioticum isolates were included in this study covering eleven different pathotypes and six isolates with unknown pathotype identity. Material was collected from countries in Europe, North America, South America and Asia (Table 2). Pathotype identification for the Canadian isolates was performed according to the Glynne-Lemmerzahl method [7]. The samples were provided either as wart material, resting spores or DNA extracts. When provided with wart material, resting spores were extracted following the procedure described by Bonants et al. [36]. For the Canadian isolates, resting spores were collected in a 37 μm sieve and treated with a 1/10 dilution of viscozyme (Sigma-Aldrich) in 100 mM phosphate buffered saline at pH 5.0 at 30°C for 16 h. Spores were layered onto a 50% solution of sucrose and centrifuging at 2000 x g for 20 min. Spores were collected in a sieve, washed, and further purified by pelleting through a 90% solution of Percoll (Sigma-Aldrich) at 1000 x g for 5 min. Cultures of Chytridium confervae (CBS 675.73), Powellomyces hirtus (CBS 809.83), Spizellomyces palustris (=Phlyctochytrium palustre) (CBS 455.65) and Synchytrium microbalum (JEL517) were maintained on ARCH medium and ARCH broth (Broth: peptone 2 g (Oxoid), malt extract 3 g (Oxoid), glucose 5 g ·L − 1 , pH 6. The medium contains 8 g·L − 1 agar (Oxoid)) at 21°C. Dandelion leaves containing Synchytrium taraxaci spores were   a. In this paper, pathotype identities are used based on the potato cultivar differential set as presented in EPPO standard PM7/28 (1). In some cases, additional cultivars were used resulting in a further differentiation of pathotype identity (n.a. when not applicable) collected from a meadow in Bennekom the Netherlands (GPS coordinates: 51.997314, 5.662961) in April 2015. S. taraxaci spores were extracted by gentle disruption of epidermal leaf tissue using a pipette tip. Extracted spores were washed with tap water and stored at − 20°C until DNA extraction.
Multiplication of pathotype 1(D1) isolate 01WS on potato cultivar Erika Eye pieces of cultivar Erika were inoculated with winter sporangia of pathotype 1(D1) isolate 01WS. After two months of incubation, maturated winter sporangia from small warts were removed and used for a second round of inoculation on cultivar Erika. After each multiplication the virulence was checked on a differential set of potato cultivars. When wart development on cultivars known to be resistant to pathotype 1(D1) was observed, a full virulence profile was determined on an extended set of potato cultivars. After two multiplications, isolate 01WS produced a pathotype 6(O1) virulence spectrum and the isolate was renamed to E/II/2015. The experiment was repeated at a different moment with similar results.

DNA extraction
A total of 200 μl suspensions containing ≥5000 S. endobioticum resting spores were homogenised in a Hybaid Ribolyser multiple bead beater (Thermo Electron Corporation, the Netherlands) at 5000 bpm for 100 s using three stainless steel beads (3.2 mm). For the WS and E/II/2015 isolates, resting spores were homogenised using a combination of vortexing and ultrasound sonication. Subsequently the Ultra Clean Soil DNA kit (MoBio) was used according to manufacturer's instructions to extract genomic DNA. For the Canadian isolates, DNA was extracted from resting spores as previously described [14]. Genomic DNA of the culturable chytrids was extracted from fungal cells grown in 8 mL ARCH broth, and 50 μL spore suspensions containing S. taraxaci spores using the Wizard Magnetic DNA Purification System for Food (Promega) following manufacturer's instructions. When sufficient starting material was available multiple DNA extracts were generated per isolate. The nuclear ITS region was amplified and sequenced to verify the species identity using primers ITS4 [37] and Chy18S-269A [31] as a check before Next Generation Sequencing (NGS).

Quantification of obligate biotrophs
S. endobioticum DNA was quantified using a S. endobioticum specific real-time PCR [38]. For the quantification of S. taraxaci, a species specific real-time PCR was designed based on publicly available ITS1-5.8S-ITS2 sequences covering 20 Synchytrium species (Additional file 1: Figure S7). DNA extracts with exponential amplification curves and Cq values below 30 were selected for NGS.
Assembly and annotation of the S. endobioticum mitochondrial genome A consensus mitochondrial sequence of S. endobioticum isolate MB42 pathotype 1(D1) was created using the results of three independent assemblies with two different datasets. . Contigs containing an Mfannot annotation and/or those with an e-value of <1e-20 when compared by blastn to other publically available chytrid mitochondrial genome sequences, were selected. In the second assembly, the GRAbB assembly tool [40] was used to reconstruct the S. endobioticum mitochondrial genome using the putative S. endobioticum scaffolds identified in the first assembly, as bait for the raw genomic library. The reads collected by the final iteration of the GRAbB run were used to perform de novo assemblies using three different assemblers using SPAdes v3.6 [41], Velvet v1.2.10 [42] and Edena v3.131028 [43]. Overlaps between contigs produced by the assemblers were manually identified and merged using the helper tools distributed with the GRAbB assembler until all regions were included in a single contig. Raw reads were mapped back to the final sequence and errors were corrected with Pilon v1.19 [44] iteratively until no further errors were identified. In the third assembly, the mitochondrial genome of S. endobioticum MB42 was assembled with Celera Assembler version 8.2 [45] using the PacBio corrected reads [35]. Using the largest scaffold as a backbone sequence, all putative mitochondrial scaffolds from the three independent assemblies were assembled in Geneious R10 [46] using default parameters. Conflicts were resolved using a majority rule. In addition, the Illumina HiSeq and PacBio reads from S. endobioticum MB42 were mapped to this consensus sequence to further identify and resolve conflicts using the quality based majority rule resulting in a final consensus S. endobioticum mitochondrial genome. This final consensus was verified using the PacBio reads mapping to the consensus, which were assembled using the Canu assembler [34]. The final consensus sequence was then annotated using Mfannot. Putative mitochondrial genes, tRNAs and rRNAs were manually curated. Amino acid sequences from predicted gene models were compared to those in Inter-ProScan [47] and the NCBI nr database by blastp and those with an e-value of <1e-20 with a known predicted function were kept as valid and those that did not meet these criteria were filtered out.

Molecular verification of mitochondrial genome linearity
A TdT tailing method based on [48,49] was used as a method to verify that the linear conformation of the S. endobioticum mitochondrial genome. Positive and negative controls (i.e. synthetic oligo resembling the mitochondrial genome sequence in a linear and circular form respectively, see (Additional file 1: Figure S8) were included to monitor the effectiveness of the TdT tailing. Sample DNA (30-45 ng/μL) was denatured at 95°C for 5 min, placed on ice and then immediately used in a 25 μl of 3′-TdT reaction consisting of 1× TdT buffer (TaKaRa), 0.01% BSA, 300 μmol of dCTP (TaKaRa), 7 U of TdT enzyme (TaKaRa), and 1 μL of sample DNA. TdT tailing was performed at 37°C for 30 min, followed by the inactivation of the TdT enzyme at 65°C for 10 min. Linearity of the mtDNA was tested by PCR amplification using the primers listed in Table 3. Amplification from the poly-C 3′ ends of the mtDNA into the terminal inverted repeat (TIR) was performed with primers M13F_polyG and mtDNA_00787-rv using a reaction mix containing 1× GoTaq flexi buffer (Promega), After electrophoresis, PCR products were excised from the gel, purified with the Wizard SV gel and PCR clean-up system (Promega) and submitted for sequencing using the generic M13F sequencing primer, and specific mtDNA amplification primers in separate reactions. The resulting sequence data were mapped to the assembled and annotated S. endobioticum consensus mitochondrial genome in Geneious.

Assembly and annotation mitochondrial genome of other species
Illumina NGS data of other chytrid species: C. confervae, P. hirtus, S. palustris, S. microbalum and S. taraxaci were used for de novo assembly using CLC genomics workbench (word size: automatic, bubble size: automatic, scaffolding: on) and scaffolds were updated using a read mapping approach (length fraction: 0.5, similarity fraction: 0.8) (van de Vossenberg et al., unpublished data). Scaffolds with relative high coverage were selected and annotated using Mfannot (tbl 16) as described above for S. endobioticum MB42.

Bayesian inference of phylogeny
NGS datasets from other isolates of S. endobioticum were mapped to the annotated S. endobioticum MB42 mitochondrial genome in CLC genomics workbench (length fraction: 0.8, similarity fraction: 0.9, non-specific match: map random, resolve conflicts: quality based) and regions with ≥5× coverage were extracted. Annotations from the S. endobioticum MB42 reference mitochondrial genome were transferred to the consensus sequence. Coding sequences of 14 mitochondrial genes were extracted from the consensus sequences and these were aligned with MAFFT v7.308 [50] to coding sequences of the same genes from publically available complete chytrid mitochondrial genomes (Additional file 2: Table S1). Allomyces macrogynus ATCC46923 (NC_001715), Harpochytrium sp. JEL94 (AY182005), Harpochytrium sp. JEL105 (AY182006), Hyaloraphidium curvatum SAG 235-1 (AF402142), Monoblepharella sp. JEL15 (NC_004624), Rhizophydium brooksianumb JEL136 (NC_003053), and Spizellomyces punctatus (NC_003052). In addition, a mitochondrial scaffold was identified from the Batrachochytrium dendrobatidis JEL423 genome (GCA_000149865.1) by blastn (e-value <1e-20) to other publically available chytrid mitochondrial genome sequences. B. dendrobatidis mitochondrial scaffold DS022322 was annotated using Mfannot (tbl 16). Alignments were concatenated and phylogenies were constructed using Bayesian inference (BI) [51]. Models for nucleotide substitution were obtained using JModelTest 2.1.10 [52] which was run on default settings and the best nucleotide substitution model, according to AIC, AICc, BIC and DT, was the GTR model with gamma distribution and estimation of invariable sites (GTR + G + I). BI was performed with the MrBayes v3.2.6 plugin incorporated in Geneious with a random starting tree and four Monte Carlo Markov Chains for 10 6 generations. Trees were sampled every 200 generations, and the first 10 5 generations were discarded as burn-in. Remaining trees were combined to generate a 50% majority rule consensus tree with posterior probabilities.

Mitochondrial haplotypes and intraspecies variation
The mitochondrial genomes of 30 S. endobioticum isolates were aligned with MAFFT v7.308 to estimate the relationship between S. endobioticum isolates. Low coverage terminal sequences (< 5× coverage) were trimmed from the alignment and variable sites were extracted with "Show variable sites only" from the online fasta sequence toolbox FaBox [53]. A Median Joining (MJ) network (ε = 0, uninformative sites = masked) was calculated with Popart v1.7 [54]. Variation within a isolate for a given polymorphic site was determined for informative sites in the MJ network with the basic variant calling tool, including only paired reads but allowing non-specific matches to include TIRs, in CLC genomics workbench (min coverage = 5×, min count = 2, min frequency = 10%, min base quality = 20, direction frequency filter = 5%).

Additional files
Additional file 1: Figure S1. PacBio read mapping to the S. endobioticum mtDNA. Figure S2. Dotplot internal repeat structure TIRs Figure S3. Dotplot repeat in AT-rich region containing dpoB gene Figure S4. Interspecies comparison of organisation and orientation of mitochondrial genes, tRNAs and rRNAs Figure S5. Haplotype network based on CDS of mtDNA genes S. endobioticum Figure S6. Within-strain diversity in haplotype network based on polymorphic sites of the S. endobioticum mtDNA Figure S7. Synchytrium taraxaci Taqman assay Figure S8. TdT-tailing controls. (DOCX 2313 kb) Additional file 2: Table S1. mitogenomes and mitochondrial genes included in the Bayesian inference (BI) of phylogeny Table S2. SNP percentages of polymorphic sites relative to the S. endobioticum 1(D1)_MB42 mitogenome Table S3. Mitochondrial haplotype classification of S. endobioticum isolates based on informative sites in coding sequences of seven mitochondrial genes Table S4.