Dynamic virulence‐related regions of the plant pathogenic fungus Verticillium dahliae display enhanced sequence conservation

Abstract Plant pathogens continuously evolve to evade host immune responses. During host colonization, many fungal pathogens secrete effectors to perturb such responses, but these in turn may become recognized by host immune receptors. To facilitate the evolution of effector repertoires, such as the elimination of recognized effectors, effector genes often reside in genomic regions that display increased plasticity, a phenomenon that is captured in the two‐speed genome hypothesis. The genome of the vascular wilt fungus Verticillium dahliae displays regions with extensive presence/absence polymorphisms, so‐called lineage‐specific regions, that are enriched in in planta‐induced putative effector genes. As expected, comparative genomics reveals differential degrees of sequence divergence between lineage‐specific regions and the core genome. Unanticipated, lineage‐specific regions display markedly higher sequence conservation in coding as well as noncoding regions than the core genome. We provide evidence that disqualifies horizontal transfer to explain the observed sequence conservation and conclude that sequence divergence occurs at a slower pace in lineage‐specific regions of the V. dahliae genome. We hypothesize that differences in chromatin organisation may explain lower nucleotide substitution rates in the plastic, lineage‐specific regions of V. dahliae.


| INTRODUC TI ON
Microbes colonize near all habitats on earth, even those that are characterized by extreme conditions with respect to salinity, temperature, pH, radiation and pressure (Pikuta, Hoover, & Tang, 2007).
How microbes can adapt to their environment, especially when it concerns heterogeneous environments, is a fundamental question in evolutionary biology. It remains a major challenge to determine what the molecular consequences are of adaptive mutations in response to dynamic environments, and how these translate into conditionally adaptive phenotypes. During their life cycle, many microbes colonize other organisms that act as their hosts. These symbiotic interactions between microbes and their hosts can range from commensalistic to either beneficial or parasitic. Plant-pathogen interactions are often exquisite models for the study of the molecular processes of (microbial) adaptation to fluctuating (host) environments, due to the short generation time of the microbe, the typically well-characterized interaction in the wild and in the laboratory, and more and more frequently the genetic tractability of the pathogen and the host.
To establish their parasitic relationships, pathogenic microbes evolve repertoires of secreted proteins, so-called effectors, that mediate host colonization often by deregulating host immunity (Cook, Mesarich, & Thomma, 2015;Dodds & Rathjen, 2010). As plants have evolved immune receptors that recognize various molecular patterns that betray microbial invasion as so-called invasion patterns, receptors evolved that can detect effectors or their activities (Cook et al., 2015;Dodds & Rathjen, 2010). Consequently, pathogens and their hosts are typically engaged in co-evolutionary arms races in which plant pathogen effector repertoires are subject to selective forces that often result in rapid diversification. Interestingly, effector genes are often not randomly organized in genomes of filamentous plant pathogens (Dong, Raffaele, & Kamoun, 2015). For instance, effector genes of the oomycete potato late blight pathogen Phytophthora infestans reside in repeat-rich regions that display increased structural polymorphisms and enhanced levels of positive selection (Haas et al., 2009;Raffaele et al., 2010). Consequently, it has been proposed that many filamentous pathogens have a bipartite genome architecture with housekeeping genes residing in a conserved core genome and effector genes in dynamic and repeat-rich compartments; a "two-speed" genome (Croll & McDonald, 2012;Raffaele & Kamoun, 2012). It is hypothesized that this compartmentalization facilitates the rapid evolution of effector repertoires to mediate continued symbioses between pathogen and the plant host. Often, repeat-rich genome regions display signs of such accelerated evolution with structural variations such as presence/absence polymorphisms (Raffaele et al., 2010) or chromosomal rearrangements (Faino et al., 2016;de Jonge et al., 2013). Furthermore, such regions can also display increased substitution rates (Cuomo et al., 2007;van de Wouw et al., 2010), including increased levels of nonsynonymous substitutions (Raffaele et al., 2010;Sperschneider et al., 2015;Stukenbrock et al., 2010).
Verticillium is a genus of soil-borne Ascomycete fungi containing notorious plant pathogens of numerous crops ) that infect their hosts via the roots and then colonize xylem vessels, resulting in vascular occlusion and wilt disease (Fradin & Thomma, 2006). Currently, 10 Verticillium species are described, which are divided in two phylogenetic clusters, i.e., clade Flavexudans and clade Flavnonexudans (Inderbitzin et al., 2011).
Verticillium species are thought to have a predominant, if not exclusive, asexual reproduction as a sexual cycle has never been described for any of the species (Short, Gurung, Hu, Inderbitzin, & Subbarao, 2014). Nevertheless, Verticillium dahliae still appears to have the machinery for sexual recombination, as mating types and meiosis-specific genes occur (Short et al., 2014). However, a severely skewed mating type ratio in the global V. dahliae population (99% vs. 1%; Short et al., 2014) combined with the low mobility of V. dahliae due to its soil-borne nature makes sexual recombination only a marginal phenomenon at most. In V. dahliae, the most notorious plant pathogen within the genus that infects hundreds of plant species , mechanisms different from meiotic recombination were shown to contribute to the genomic diversity, including large-scale genomic rearrangements, horizontal gene transfer, and transposable element (TE) activity (Faino et al., 2016;de Jonge et al., 2012de Jonge et al., , 2013Seidl & Thomma, 2014;Shi-Kunne, van Kooten, Depotter, Thomma, & Seidl, 2019). These mechanisms often converge on lineage-specific (LS) regions that are enriched in TEs and in in planta-induced effector genes (Faino et al., 2016;de Jonge et al., 2013;Klosterman et al., 2011). We previously reported that sequence conservation in coding as well as noncoding regions than the core genome.
We provide evidence that disqualifies horizontal transfer to explain the observed sequence conservation and conclude that sequence divergence occurs at a slower pace in lineage-specific regions of the V. dahliae genome. We hypothesize that differences in chromatin organisation may explain lower nucleotide substitution rates in the plastic, lineage-specific regions of V. dahliae.

K E Y W O R D S
comparative genomics, effector, genome evolution, mutagenesis, two-speed genome, Verticillium wilt LS regions of V. dahliae are largely derived from segmental duplications (Faino et al., 2016) that are known as important sources for functional diversification (Magadum, Banerjee, Murugan, Gangapur, & Ravikesavan, 2013). To study the evolution of the LS regions, in the present study we exploited comparative genomics across the Verticillium genus to identify differential rates of sequence diversification to further characterize the two-speed genome of V. dahliae.
For PD670, PD660, PD659 and PD736, two libraries (500 bp and 5 kb insert size) were prepared and sequenced using an Illumina High-throughput sequencing platform. In total, ~18 million pairedend reads (150 bp read length; 500 bp insert size library) and ~16 million mate-paired read (150 bp read length; 5 kb insert size library) were produced per strain. We assembled the genomes using the A5 pipeline (default settings; Tritt, Eisen, Facciotti, & Darling, 2012), and we subsequently filled the remaining sequence gaps using SOAPdenovo2 (default settings; Luo et al., 2012). After obtaining final assemblies, we used quast (Gurevich, Saveliev, Vyahhi, & Tesler, 2013) to calculate genome statistics. Gene annotations for V. dahliae strain JR2 and other Verticillium species were obtained from previous studies (Faino et al., 2016;Shi-Kunne et al., 2018), except for V. isaacii strain PD660 that was annotated with the Maker2 pipeline in this study according to Shi-Kunne et al. (2018) and Holt and Yandell (2011).

| Comparative genome analysis
Repetitive elements were identified using repeatmodeler (version 1.0.8) based on known repetitive elements and on de novo repeat identification, and genomes were subsequently masked using repeatmasker (version 4.0.6; sensitive mode; Smit, Hubley, & Green, 2015). To prevent assigning high sequence identities to repetitive elements, sequence alignments were performed to a repeat-masked reference genome using nucmer (option --maxmatch), which is part of the mummer package (version 3.1; Kurtz et al., 2004).
Linear plots showing alignments within and closely adjacent JR2 LS regions were plotted with the r package genoPlotR (version 0.8.7; Lineage-specific sequences were defined by alignment of different strains to a reference using nucmer (version 3.1, option --maxmatch; Kurtz et al., 2004) and regions were determined using bedtools version 2.25.0 (options sort, merge and genomecov; Quinlan & Hall, 2010).
Lineage-specific regions of V. dahliae and V. tricorpus were delimited based on interspecific nucleotide alignments with nucmer, which is part of the mummer package (version 3.1, option --maxmatch; Kurtz et al., 2004). Genome regions were high sequence conservations started or ended in combination with absence/ presence polymorphisms were determined as start or end of the LS region, respectively (Table S6). The pairwise identity of the genome-wide and LS regions between V. dahliae/V. tricorpus and other haploid Verticillium species was calculated using nucmer (option -maxmatch) by dividing the respective V. dahliae/V. tricorpus query sequences into nonoverlapping windows of 500 bp and align them to the repeat-masked genomes of Verticillium species (Table 1).
Only sequences with a 1-to-1 alignments and with a minimum alignment length of 500 were considered (delta-filter −1 -l 500). In this fashion, genome-wide sequence identity between Verticillium species was also calculated to determine the sequence divergence across the Verticillium phylogenetic tree (Figure 4). Instead of using V. tricorpus and V. dahliae as reference genomes, V. klebahnii and V. alfalfae were used as reference for clade Flavexudans and clade Flavnonexudans respectively, to calculate the genome-wide nucleotide identity differences with the remaining species within the same clade. Nucleotide identity differences with remaining species were calculated in increasing order of phylogenetic distance with the reference species. Identity differences were evenly assigned to the phylogenetic branches in increasing order of phylogenetic distance.
A pan-LS-genome was constructed based on following  (Table S6). Repeat masked regions were removed from the pan-LS-genome using bedtools version 2.25.0 (Quinlan & Hall, 2010). Additionally, duplicated regions (≥90% identity, ≥100 bp) in the pan-LS-genome were determined using nucmer (version 3.1, option --maxmatch; Kurtz et al., 2004) and subsequently removed with using bedtools version 2.25.0 (options merge, subtract and getfasta; Quinlan & Hall, 2010). The fraction of pan-LS-genome that is present in every individual Verticillium strain was determined using nucmer (version 3.1, option --maxmatch; Kurtz et al., 2004). The clade pan-LS-genomes were constructed by combining all the pan-LS-genome regions that are present in the Verticillium clade isolates, which was then also removed from duplicate regions.  To compare the rate of synonymous and nonsynonymous substitutions between the core and LS regions, Ka and Ks of orthologs of JR2 and TAB2 were determined using the Nei and Gojobori method (Nei & Gojobori, 1986) in paml (version 4.8; Yang, 2007). Significance of positive selection was tested using a Z test (Stukenbrock & Dutheil, 2012). Z-values > 1.65 were considered significant with p < .05. Secreted proteins were predicted by SignalP4 (Petersen, Brunak, Von Heijne, & Nielsen, 2011). To compare coding regions of genes and intergenic regions, sequence identities were retrieved by blast (version 2.2.31+) searches between strains V. dahliae JR2 and V. nonalfalfae TAB2 (Altschul, Gish, Miller, Myers, & Lipman, 1990).
The coding regions of genes were aligned to each other and the best hits (sequence identity) with a minimal coverage of 80% with each other were selected. Intergenic regions of V. dahliae strain JR2 were fractioned in 5 kb windows with bedtools version 2.25.0 (options makewindows and getfasta) and similarly queried to the genome of V. nonalfalfae strain TAB2 (Quinlan & Hall, 2010). Hits with a maximal bit-score and minimal alignment of 500 bp to a window were selected.

| Tree building and ortholog analysis
The phylogenetic trees of the Verticillium genus were previously generated using 5,228 single-copy orthologs that are conserved among all of the genomes (Shi-Kunne et al., 2018 (Stamatakis, 2014). The robustness of the inferred phylogeny was assessed by 100 rapid bootstrap approximations.

| LS sequences reside in four regions of the genome of V. dahliae strain JR2
Previously, four LS regions were characterized for V. dahliae strain JR2; one on chromosome 2 and 4, and two on chromosome 5 (Faino F I G U R E 3 Regions with particularly high interspecific sequence identity. All Verticillium strains mentioned in Table S2 were

| LS regions share increased sequence identity to other Verticillium species
Next, we extended our analysis to other Verticillium species. While most of the V. dahliae strain JR2 genome aligns with V. nonalfalfae strain TAB2 with an average sequence identity of ~92%, particular regions display increased sequence identity, even up to 100% ( Figure S1). Intriguingly, the regions with increased sequence identity co-localize with LS regions (Faino et al., , 2016

| High interspecific sequence identity of LS regions is not unique to V. dahliae
To investigate whether other Verticillium species similarly carry LS regions that display high interspecific sequence identity, we performed alignments using V. tricorpus strain PD593 as a reference because of its high degree of completeness with seven of the nine scaffolds probably representing complete chromosomes ( in only a single genomic region of 41 kb on scaffold 1 (Figure 3b).
Like for V. dahliae strain JR2, sequences of other Verticillium species aligned with high identity to V. tricorpus PD593: V. isaacii, V. klebahnii and V. zaregamsianum display a median genome identity of ~95%, while other haploid Verticillium species display ~88%-89% median genome identity (Table 1). Notably, regions that display significantly higher sequence identity localized at the LS region on scaffold 1, but also to an additional region of 23 kb on scaffold 6 (Figures 2b and   3b). It is likely this concerns an LS region that could not be identified based on the two V. tricorpus strains used in our analysis, and thus will also be referred to as LS region. For Verticillium strains with total alignments of at least 100 kb of high-identity sequences, the fraction of high-identity sequences that aligned to LS genome regions ranged from 49% for V. nubilum (PD621) up to 84% for V. albo-atrum (PD747; Table S4). As expected, the sequence identity to six of the eight other haploid Verticillium species was significantly higher in LS genome regions compared to the core genome (Table 1). No increase in sequence identity was found in alignments with V. alfalfae strain PD683 and V. zaregamsianum strain PD739 as only few sequences of LS genome regions could be aligned (Table 1).

| LS regions are unlikely to originate from horizontal DNA transfers among Verticillium species
High interspecific sequence identity of particular genomic regions, such as observed here for LS regions, could occur through two different mechanisms: (a) horizontal transfer of sequences between species, in this case horizontal transfer of LS sequences between Verticillium species, or (b) differences in nucleotide substitution rates between genomic regions, in this case between core and LS genome regions. To explain the high sequence identity through the occurrence of interspecific horizontal DNA transfers, a minimum of five such transfers must have occurred that involve V. dahliae to explain the differences in median sequence identities of LS and core genome regions with the other Verticillium species (Figure 4, Table 1).
Similarly, at least five such transfers are required to explain the composition of the V. tricorpus genome relative to its sister species If LS regions were transferred between species through a limited number of events, a depletion of interspecific sequence identity variation in LS regions may be observed. Thus, we calculated the variance in interspecific sequence identity for V. dahliae LS regions and genome-wide (Table 1). Sequence identities in LS regions varied less than in the core genome in alignments to V. alfalfae and to V. nonalfalfae, two species that diverged only recently from V. dahliae (Table 1). In contrast, sequence identities with V. albo-atrum, V. nubilum and V. isaacii displayed more variation in LS genome regions than within the core genome regions (Table 1). No significant differences between LS and core genome regions were found in alignments with V. tricorpus, V. klebahnii and V. zaregamsianum. A similar pattern is observed for V. tricorpus as LS regions display lower sequence identity variation than the core genome in alignments to V. albo-atrum and V. dahliae, whereas the opposite is true for alignments to V. nonalfalfae and V. klebahnii (Table 1). Thus, although it needs to be noted that there is a significant difference in size between core and LS genomic regions, there seems to be no genus-wide trend towards lower interspecific sequence identity variation in LS regions compared to the core genome, which may be interpreted as an argument against horizontal transfer of these regions.

F I G U R E 5
The evolution of lineage-specific (LS) region genes. The left tree shows the phylogenetic relationships between the haploid Verticillium species and their division into clade Flavexudans (FE) and clade Flavnonexudans (FNE). The middle and right phylogenetic trees are of a particular LS region gene from V. dahliae and from V. tricorpus, respectively. Genes indicated in red have particularly high sequence identity. The robustness of the phylogeny was assessed using 100 bootstrap replicates. For all V. dahliae and V. tricorpus LS region genes with at least one homolog in clade Flavexudans and one homolog in clade Flavnonexudans, phylogenetic trees were constructed including all their Verticillium homologs. The number of trees in congruence and incongruent with the Verticillium clade dichotomy are indicated in the green and red box, respectively. In these assessments, complete trees were analysed in case the tree did not contain paralogs for any of the Verticillium species. In case of paralogs that indicate gene duplication events, only a subclade of the phylogenetic tree was used for assessment. The number in the white box is the number of trees that was inconclusive [Colour figure can be viewed at wileyonlinelibrary.com]   Verticillium species, with five highly conserved orthologs in species that generally differ from 5.2% to 11.2% in genome-wide nucleotide identity with V. dahliae, whereas sequence identities between the highly conserved Chr5g02240 orthologs differs only from 0% to 0.8% (2,460 total nucleotides; Figure 5).
Moreover, in accordance with the overall Verticillium phylogeny, the V. nonalfalfae homolog clusters with that of V. dahliae. In total, 80% of the trees displayed a phylogeny where the Flavexudans and Flavnonexudans species clustered, whereas only 5% of the trees did not yield this typical dichotomy and 15% of the trees did not contain sufficient homologous sequences to be conclusive ( Figure 5). Similarly, the majority (56%) of phylogenetic trees for V. tricorpus LS genes obey to the Flavexudans and Flavnonexudans dichotomy ( Figure 5). Like for V. dahliae gene Chr5g02240, the phylogenetic tree for homologs of V. tricorpus gene Chr006g10380 shows the expected segregation F I G U R E 6 Diversity of pan-LS-genome contents across the Verticillium genus. A pan-LS-genome was constructed based on sequences from V. dahliae JR2, V. alfalfae PD683, V. tricorpus PD593 and V. klebahnii PD401 (black bars). The bar size next to the species names in the Verticillium phylogenetic tree is representative for the amount of the pan-LS-genome that is present in the individual isolates.

| Pan-LS-genome distribution across the Verticillium genus
Considering that horizontal transfer is unlikely, the high sequence identity between Verticillium LS sequences indicates that their origin is ancestral and predates speciation, and that lower num- The proportion of the pan-LS-genome differed markedly between Verticillium strains and ranged from 12% for V. nubilum strain PD621 up to 58% for V. dahliae strain JR2 ( Figure 6, Table   S5). Notably, by using a limited number of isolates in the consensus reconstruction, retentions are probably biased towards strains that are phylogenetically closer related to the species that were used to compose the pan-genome. However, V. alboatrum strains contained considerably more of the pan-LS-genome compared to V. zaregamsianum and V. isaacii strains, despite its phylogenetically more distant relation to V. klebahnii and V. tricorpus ( Figure 6, Table S5). Moreover, LS contents do not only differ considerably between species but also within species.

| Increased sequence conservation is probably not driven by negative selection
To study the impact of the increased sequence conservation on gene evolution in more detail, substitution rates of LS region genes were compared with those of core region genes. We determined the rates of nonsynonymous (Ka) and synonymous (Ks) substitutions for LS versus core genes. In total, 48% (68 out of 142) of the LS genes could not be used for Ka and Ks determination, as we did not observe any substitutions when compared to their corresponding V. nonalfalfae orthologs. In contrast, almost all core genes (8,583 out of 8,584) display nucleotide substitutions when compared with their V. nonalfalfae orthologs. Whereas the Ka is not different (two-sided Wilcoxon rank-sum test, p < .05) between genes in LS regions (median = 0.015, n = 74) and the core genome The resulting increase in Ka/Ks ratio's for genes in LS regions suggests that negative selection is probably not responsible for the slower divergence of gene sequences. Rather, LS regions seem to encounter a lower rate of synonymous nucleotide substitutions that lead to lower numbers of neutral/synonymous substitutions in genes in LS regions when compared with core genes. In congruence with this hypothesis, sequence comparisons of coding and intergenic regions between V. dahliae and V. nonalfalfae revealed that increased sequence conservation similarly occurs in coding regions as well as intergenic sequences (Figure 7b), suggesting that the increased sequence conservation is driven by a mechanism that affects whole LS regions, rather than by selection.

| D ISCUSS I ON
Genomes of many filamentous plant pathogens are thought to obey to a two-speed evolution model (Croll & McDonald, 2012;Dong et al., 2015;Möller & Stukenbrock, 2017). Verticillium dahliae is similarly thought to evolve under a two-speed regime, as LS regions that are of significance for host interactions display increased structural variation and TE activity (Faino et al., , 2016de Jonge et al., 2013).
Additionally, LS regions are extremely plastic with abundant presence/absence polymorphisms (Figure 1; Faino et al., 2016;de Jonge et al., 2013). Intriguingly, although LS regions are enriched in segmental duplications (Faino et al., 2016), which can be an incentive for evolutionary diversification (Magadum et al., 2013), LS sequences display a remarkably high degree of sequence identity to other Verticillium species (Figures 2,3; Table 1). Principally, sequences with increased identities between distinct taxa can originate from horizontal transfer, a phenomenon that has been implicated in the pathogenicity of various filamentous plant pathogens (Soanes & Richards, 2014). For instance, Pyrenophora tritici-repentis, the causal agent of wheat tan spot, acquired a gene from the fungal wheat pathogen Phaeosphaeria nodorum enabling the production of the host-specific toxin ToxA that mediates pathogenicity on wheat (Friesen et al., 2006). However, horizontal transfer is not likely to explain our observations, as the increased sequence identity is observed genus-wide and concerns every species within the genus (Table 1) Considering that horizontal transfer is extremely unlikely, we argue that V. dahliae LS regions are subject to processes that mediate increased sequence conservation. This increased sequence conservation of LS regions would not be a consequence of negative selection on coding regions, as intergenic regions display similarly increased conservation levels ( Figure 7b). Moreover, LS genes display a similar fraction of nonsynonymous substitutions (Ka) as core genes, but carry significantly less synonymous substitutions (Ks; Figure 7a).
Consequently, high interspecific sequence identity of LS genome regions is probably due to generally lower synonymous nucleotide substitution rates in these regions when compared with the core genome.
Repressed levels of synonymous substitutions were previously found for repeat-rich dispensable chromosomes of the fungal wheat pathogen Zymoseptoria tritici (Stukenbrock et al., 2010). However, this observation was not attributed to lower substitution rates, but rather the consequence of a lower effective population size of these dispensable chromosomes (Stukenbrock et al., 2010). As sex is probably rare to nonexistent in Verticillium, selective sweeps are an unlikely explanation for increased sequence conservation of LS region as, in case of Verticillium species with low to nonexistent meiotic recombination, entire genomes would sweep to fixation and not specific genome regions (Shapiro, Leducq, & Mallet, 2016). Moreover, sequence conservation is observed between species and fixation of the same genome region across species boundaries is unlikely, as different Verticillium species conceivably encounter distinct selection pressures due to differences in their biological features, such as pathogenicity and host ranges . Thus, the increased sequence conservation as observed here is unprecedented and perhaps counter-intuitive. Previously, increased substitution rates have been associated with two-speed genome evolution (Cuomo et al., 2007;Dong et al., 2015). For example, repeat-induced point (RIP) mutagenesis increases sequence divergence of particular effector genes of the oilseed rape pathogen Leptosphaeria maculans that are localized near TEs (van de Wouw et al., 2010). However, accelerated evolution through increased SNP frequencies is not consistently observed for two-speed genomes, as no significant difference in SNP frequencies between core and repeat-rich genomic regions was found in P. infestans (Raffaele et al., 2010). Nevertheless, accelerated evolution of LS regions can also be established through other means, such as presence/absence polymorphisms. The wellcharacterized V. dahliae LS effector Ave1 is highly conserved, as an identical copy occurs V. alfalfae strain VaMs102 that displays a genome-wide average sequence identity of 92% (de Jonge et al., 2012).
Moreover, no Ave1 allelic variation is hitherto found in the V. dahliae population as well as in V. alfalfae and V. nonalfalfae populations (de Jonge et al., 2012;Song et al., 2017). Since Ave1 is recognized by the tomato immune receptor Ve1 (Fradin et al.., 2009), evasion of recognition occurs through various Ave1 deletion events from the population (Faino et al., 2016;de Jonge et al., 2012).
Mechanisms that can explain the observed increased sequence conservation in repeat-rich LS regions remain unknown. Mutations mostly originate from DNA polymerase errors and there is no immediate reason why these error rates would diverge in LS regions.
Possibly, the depletion of nucleotide substitutions can be associated with a differential chromatin organisation of LS regions. Intriguingly, a study into chromatin structure in the human genome noted that regions of open chromatin displayed lower mutation rates which was hypothesized to be a result of these regions being more accessible to repair mechanisms (Prendergast et al., 2007). However, repeat-rich regions such as the LS regions in V. dahliae are thought to be associated with densely organised chromatin, referred to as heterochromatin (Galazka & Freitag, 2014). In Z. tritici, repeat-rich conditionally dispensable chromosomes are enriched for histone modifications associated with heterochromatin, in contrast to core chromosomes that are largely euchromatic and transcriptionally active (Schotanus et al., 2015). Generally, heterochromatin is associated with suppression of genomic structural alterations such as recombination.
Nevertheless, heterochromatic regions of Z. tritici are enriched for structural variations as they are enriched for duplications and deletions (Seidl, Cook, & Thomma, 2016). Thus, further research is needed to investigate whether differences in chromatin organisation can explain lower rates of SNP frequencies that occur in the plastic LS regions of V. dahliae.

| CON CLUS ION
The two-speed genome is an intuitive evolutionary model for filamentous pathogens, as genes important for pathogenicity benefit from frequent alternations to mediate continued symbiosis with the host. However, filamentous pathogens comprise a heterogeneous group of organisms with diverse lifestyles (Dean et al., 2012;Kamoun et al., 2015). Consequently, it is not surprising that accelerated evolution is driven by different mechanisms between species.
In V. dahliae, acceleration evolution is merely achieved through presence/absence polymorphisms, as nucleotide sequences are highly conserved in LS regions. Perhaps, deletion of recognized effectors leads to a more rapid immunity evasion than sequence alterations through nucleotide substitutions (Daverdin et al., 2012). Thus, the quick fashion of host immunity evasion through the deletion of effector genes can be evolutionary advantageous over allelic diversification, especially for soil-borne pathogens with a small effective population size that have little means of mobility.