Highly flexible infection programs in a specialized wheat pathogen

Abstract Many filamentous plant pathogens exhibit high levels of genomic variability, yet the impact of this variation on host–pathogen interactions is largely unknown. We have addressed host specialization in the wheat pathogen Zymoseptoria tritici. Our study builds on comparative analyses of infection and gene expression phenotypes of three isolates and reveals the extent to which genomic variation translates into phenotypic variation. The isolates exhibit genetic and genomic variation but are similarly virulent. By combining confocal microscopy, disease monitoring, staining of ROS, and comparative transcriptome analyses, we conducted a detailed comparison of the infection processes of these isolates in a susceptible wheat cultivar. We characterized four core infection stages: establishment, biotrophic growth, lifestyle transition, and necrotrophic growth and asexual reproduction that are shared by the three isolates. However, we demonstrate differentiated temporal and spatial infection development and significant differences in the expression profiles of the three isolates during the infection stages. More than 20% of the genes were differentially expressed and these genes were located significantly closer to transposable elements, suggesting an impact of epigenetic regulation. Further, differentially expressed genes were enriched in effector candidates suggesting that isolate‐specific strategies for manipulating host defenses are present in Z. tritici. We demonstrate that individuals of a host‐specialized pathogen have highly differentiated infection programs characterized by flexible infection development and functional redundancy. This illustrates how high genetic diversity in pathogen populations results in highly differentiated infection phenotypes, which fact needs to be acknowledged to understand host–pathogen interactions and pathogen evolution.


| INTRODUC TI ON
Plants are colonized by a broad diversity of microbial species. Many plant-associated microbes are commensals that utilize the plant as a scaffold for their proliferation by feeding on readily available nutrients. Other microbial species engage in more intimate interactions with plants, either as mutualistic symbionts where both partners benefit from the interaction or as antagonistic symbionts where the microbial species exploits the plant for growth and reproduction (Zeilinger et al., 2015). Mutualistic and antagonistic interactions between plants and co-existing microbial species rely on molecular interactions between plant receptors and signaling pathways and microbial molecules including microbe-associated molecular patterns (MAMPs) and effectors. The successful establishment of such intimate plant-microbe interactions, in one or the other form, depends on the ability of the microbial partner to suppress and interfere with the immune responses of the plant (Dodds & Rathjen, 2010;Jones & Dangl, 2006;Lo Presti et al., 2015) and to colonize and reproduce associated to plant tissues (Haueisen & Stukenbrock, 2016;van der Does & Rep, 2017). These symbioses are therefore considered to be highly specialized and the result of co-evolution.
In a small number of model systems, key molecules involved in the antagonistic interaction of plants with their specialized pathogens have been characterized, for example, the Ustilago maydis effector Tin2 that interferes with the anthocyanin biosynthesis pathway in its host maize (Tanaka et al., 2014) or the tyrosine phosphatase HopAO1 of Pseudomonas syringae which inhibits activation of a plant pattern recognition receptor (Macho et al., 2014). While the relevance of these molecular interactions for the studied hostpathogen genotype combinations is undisputed, the consequences at the population level have so far been poorly addressed. While many plant pathogens are known to be highly specialized to their host, substantial genetic variation is found in many species. Genetic variation can translate into phenotypic variation that can be important for pathogen populations to persist in changing environments (Fisher, Hawkins, Sanglard, & Gurr, 2018;Möller & Stukenbrock, 2017). However, host specialization may require a highly specialized and definite infection program. An intriguing question is which consequences genetic and phenotypic variation entails for the diversity of host interactions in pathogens that are specialized and host-specific?
In this study, we addressed the extent of phenotypic variation in a fungal plant pathogen characterized by a high level of genomic variability. Our aim was to study phenotypic diversity beyond virulence phenotypes that are limited to the description of quantitative virulence which is based on the quantification of disease symptoms at a particular time of infection. We therefore considered morphological and temporal infection development as well as stage-specific gene expression to characterize the infection phenotypes of pathogen individuals. We used the wheat pathogen Zymoseptoria tritici (syn. Mycosphaerella graminicola) as a model to investigate how the development of disease symptoms and the transcriptional program induced during infection vary among three field isolates from geographically distinct locations. Zymoseptoria tritici has served as a prominent model in population genetic studies of crop pathogens, and genetic variation has been assessed from local scales in individual lesions on single leaves up to continental scales. The amount of genetic variation in a Z. tritici field population is comparable to the variation found on a continental scale suggesting very high levels of genetic variation in local populations and little subdivision between populations (Linde, Zhan, McDonald, & a., 2002;McDonald et al., 2016;Zhan, Pettway, & McDonald, 2003). Zymoseptoria tritici has been described as hemibiotroph (Ponomarenko, Goodwin, & Kema, 2011) or latent necrotroph (Sánchez-Vallet, McDonald, Solomon, & McDonald, 2015) with a prolonged epiphytic stage (Fones, Eyles, Kay, Cowper, & Gurr, 2017). Infections are biphasic and characterized by a long asymptomatic, biotrophic phase, followed by necrotrophic growth where the fungus degrades and takes up nutrients from dead host cells (Ponomarenko et al., 2011;Rudd et al., 2015). In spite of the general importance of this pathogen in wheat fields, the life cycle and infection biology of Z. tritici are so far poorly understood.
The haploid genome of Z. tritici comprises a high number of accessory chromosomes ranging from 400 kb to 1 Mb in size in the reference isolate IPO323 Wittenberg et al., 2009). Recent studies provide evidence for the presence of virulence determinants on the accessory chromosomes; however, the genes responsible for these effects have so far not been identified (Habig, Quade, & Stukenbrock, 2017). Furthermore, several genome-wide association and quantitative trait loci mapping studies have linked a variety of phenotypic traits to genetic variants and candidate genes (Hartmann, Sánchez-Vallet, McDonald, & Croll, 2017;Lendenmann, Croll, & McDonald, 2015;Lendenmann, Croll, Stewart, & McDonald, 2014;Mirzadi Gohari et al., 2015;Stewart et al., 2017;Zhong et al., 2017).
Here, we investigated how infection of a susceptible host by genetically and morphologically distinct isolates results in similar quantitative virulence. By combining confocal microscopy, disease monitoring, reactive oxygen species (ROS) localization, and transcriptome analyses, we compiled a detailed characterization of infection phenotypes of three Z. tritici isolates. We hypothesized that high genetic diversity not only increases the evolutionary potential of the pathogen but also results in a variety of host-pathogen interactions that cause a range of different infection phenotypes. Our combined comparative analyses enabled us to characterize infection morphology and gene expression of the three Z. tritici isolates, including a core infection program and isolate-specific infection phenotypes. We conclude that host specialization of Z. tritici entails a substantial amount of variation in terms of the temporal, spatial, and molecular host-pathogen interactions. We speculate that this phenotypic variation is important for the pathogen to rapidly respond to changing environments, and we underline the need of considering variation at this level in the study of pathogen evolution and the development of disease control strategies.  ; 85 genes located on chromosome 18 were not considered. d 10,745 genes of IPO323 (90.76%) found by nucleotide blast for Zt10.

| Transcriptome analyses of Z. tritici isolates during wheat infection
Per isolate and infection stage, the two most representative samples were selected as biological replicates for transcriptome sequencing (  Information Table S3).
To analyze genomic distances between differentially expressed genes and transposable elements (TEs), we annotated TEs as described in Grandaubert et al. (2015) for Zt05 (Supporting Information Table S6) and Zt10 (Supporting Information Table   S7) and used the published TE annotation of IPO323 for Zt09 . Distances between genes of interest and the closest annotated TEs were calculated with bedtools v2.26 (Quinlan & Hall, 2010). Likewise, we used ChIP-seq peak data (Schotanus et al., 2015) to calculate distances between genes and the closest H3K9me3 and H3K27me3 peaks. Statistical analyses were performed in R. For a detailed overview, see Supporting Information Text S1.

| Pulsed-field gel electrophoresis
A non-protoplast protocol (Supporting Information Text S1) was used to produce DNA plugs for separation of small chromosomes (~0.2 to 1.6 Mb) by pulsed-field gel electrophoresis (PFGE) (Stukenbrock et al., 2010). Chromosomal DNA of Saccharomyces cerevisiae (Bio-Rad, Munich, Germany) was used as standard size marker. Gels were stained in 1 µg/ml ethidium bromide solution, and chromosome bands were detected with Thyphoon Trio™ (GE Healthcare, Munich, Germany). Alignments were analyzed using a custom python script to extract unique DNA blocks with a minimum length of 1 bp.

| The three Z. tritici isolates Zt05, Zt09, and Zt10 exhibit high levels of genomic variation
We selected three Z. tritici isolates Zt05, Zt09, and Zt10 that belong to genetically different clades within the species (Grandaubert, Dutheil, & Stukenbrock, 2017) but produce the same amount of disease symptom on the susceptible wheat cultivar Obelisk (see below). These isolates were previously collected in Denmark, the Netherlands, and Iran, respectively (Supporting Information Table   S1), and vary in their tolerance to abiotic stressors (Supporting Information Text S1, Figure S3, Table S8). A previous population genomic study using the 39.7 Mb genome of the isolate IPO323  as reference addressed the extent of genetic variation in these isolates and identified 500,177 single nucleotide polymorphisms (SNPs) in Zt05 and 617,431 SNPs in Zt10, indicating a considerable genetic distance between the three isolates (Grandaubert et al., 2017).
Previous comparative genome analyses have shown that Z. tritici is highly polymorphic at the level of chromosome structure and gene content, including chromosomal rearrangements, genomic orphan regions, and isolate-specific gene presence/absence patterns Plissonneau et al., 2016;. To further assess variation in genome structure and content, we performed a karyotype analysis by pulsedfield gel electrophoresis (PFGE) and generated high-quality genome assemblies based on long-read SMRT sequencing of Zt05 and Zt10 (Supporting Information Table S9). The PFGE analyses revealed extensive variation in the karyotypes of the three isolates with no small chromosomes of the same size (Supporting Information Figure S4).
The PFGE results suggest that Zt05 and Zt10 possess at least seven and four putative accessory chromosomes, respectively, and show length polymorphisms of the smallest core chromosomes 12 and 13 consistent with variation reported in a previous study (Mehrabi, Taga, & Kema, 2007). In spite of the pronounced differences in genome structure, alignment of the three high-quality genome assemblies showed a high extent of synteny. By whole-chromosome synteny analyses using SyMAP, we identified large syntenic DNA blocks for all 21 chromosomes of IPO323 in the Zt05 assembly, while Zt10 lacked homologs of chromosomes 18, 20, and 21 (Supporting Information Figure S5 and Table S9).
To identify the genes that are shared between Zt05, Zt09, and Zt10, we performed nucleotide BLAST analyses using the coding sequences of the 11,839 annotated genes of the reference IPO323 as input . We identified 11,138 IPO323 genes (94.08%) in Zt05 and 10,745 (90.76%) in Zt10. The gene presence/absence patterns correlate with the absence of large syntenic DNA blocks of chromosomes 18, 20, and 21 in Zt10 (Supporting Information Figure S5B and Table S9). Consequently, 91% of genes on core chromosomes are shared, while only 49% (313 of 643) of the genes located on accessory chromosomes are present in all three isolates (Supporting Information Table S9). Similarly, only 85% (370 of 434) of the previously identified genes encoding candidate secreted effector proteins (CSEPs) (Stukenbrock & Dutheil, 2018) were found in all isolates, indicating that the effector repertoire of Z. tritici is characterized by presence/absence polymorphisms. In total, 10,426 genes were present in the three isolates and considered to be Z. tritici core genes that we included in our further analyses (Supporting Information Table S10). In summary, the genome comparison of Zt05, Zt09, and Zt10 shows a high extent of variation at single nucleotide positions as well as structural variation including differences in the total gene content that is typical in natural isolates of this highly polymorphic pathogen.

| Virulence of the three Z. tritici isolates is very similar but disease develops at different speeds
We compared the virulence phenotypes of Zt05, Zt09, and Zt10 on the highly susceptible wheat cultivar Obelisk and evaluated infections 28 days post inoculation (dpi) by categorizing the percentage of leaf area affected by necrosis (Figure 1a and Supporting Information Figure S6) and covered with pycnidia, the asexual fruiting bodies (Figure 1b and Supporting Information Figure S6).

Although we observed different levels of necrosis (two-sided
Mann-Whitney U tests, p ≤ 0.0048), we found no significant differences in the pycnidia levels of the three isolates (two-sided Mann-Whitney U tests, p ≥ 0.034) (Supporting Information Figure   S7). As previously observed , quantitative necrosis and pycnidia levels were only weakly linked and coverage of necrotic lesions with pycnidia was highly variable (Supporting Information Figure S1). Hence, for quantitative F I G U R E 1 In planta phenotypic assay demonstrates similar pycnidia levels of Zymoseptoria tritici isolates on the susceptible wheat cultivar Obelisk. Quantitative differences in (a) necrosis and (b) pycnidia coverage of inoculated leaf areas were manually assessed at 28 days post inoculation based on six symptom levels: 0 (without visible symptoms), 1 (1% to 20%), 2 (21% to 40%), 3 (41% to 60%), 4 (61% to 80%), and 5 (81% to 100%). The three isolates caused different levels of necrosis (two-sided Mann-Whitney U tests, p ≤ 0.0048), but pycnidia levels were not different (two-sided Mann-Whitney U tests, p ≥ 0.034) virulence of Z. tritici, the production of pycnidia is considered the primary measure as it directly reflects reproductive fitness rather than differences in host sensitivity, for example, sensitivity toward pathogen toxins that contribute to necrosis   (1) isolate Zt05 at 7 dpi, (2) Zt09 at 11 dpi, and (3) Zt10 at 9 dpi. Scale bars = 25 µm and quantified accumulation by image analysis of the stained leaves.
H 2 O 2 is involved in restricting colonization of resistant wheat cultivars but also plays an important role during necrotrophic fungal growth in compatible infections (Shetty et al., 2003(Shetty et al., , 2007. We observed ROS accumulation coinciding with the onset of necrosis ( Figure 2c and Supporting Information Figure S8). However, 11 to 14 days after inoculation, we observed no ROS (Figure 2b,

| Zymoseptoria tritici infection is characterized by four core developmental stages
Next, we aimed to morphologically characterize host colonization of the three Z. tritici isolates. We conducted detailed confocal microscopy analyses in which we scanned 101 leaves harvested between 3 and 28 days after inoculation (Supporting Information Table S1).
Analyses of large z-stacks of longitudinal optical sections allowed us to infer the spatial and temporal fungal colonization on and in infected leaves. First, we focused on the commonalities in host colonization shared by the isolates. We identified a sequence of four stages that we define as the core infection program of Z. tritici

| Highly differentiated infection phenotypes of the three Z. tritici isolates on Obelisk wheat
While we clearly recognized the four core infection stages for the isolates Zt05, Zt09, and Zt10, we also observed differences (Supporting Information Text S1) that mainly relate to the timing of transitions between the stages and the extent of fungal proliferation. immunity (Jones & Dangl, 2006) and acquires energy initially mainly from stored lipids and only later from host-derived nutrients (Rudd et al., 2015). Delayed stomatal penetration and reduced internal leaf colonization increase the role of the epiphytic phase where the pathogen is exposed to environmental influences and control measures but could still be transferred to a more favorable host environment (Fones et al., 2017). We speculate that the different extent of fungal growth reflects colonization of different niches associated with the host tissue, deviating strategies to avoid host recognition as well as distinct capabilities to store lipids and exploit the limited nutrient resources during biotrophic colonization.
Taken together, the infection development of the studied Z. tritici isolates is highly divergent, although the final infection outcome is the same (Figure 1). Thereby, instead of one strictly defined infection program, Z. tritici exhibits a variety of host-pathogen interactions that represent equally successful strategies for reproduction in a susceptible wheat cultivar. In particular, we find that infection development of the wheat pathogen can be highly flexible with respect to the timing of the lifestyle transition and the spatial distribution of infecting hyphae inside host tissue.

| Generation of isolate-and stagespecific transcriptomes based on confocal microscopy analyses
Given the morphological and temporal differences in infection devel- To account for the temporal variation between Zt05, Zt09, and Zt10, we combined confocal laser scanning microscopy and RNAseq and conducted microscopy analyses of tissue from the wheat leaves used for RNA extraction and transcriptome sequencing.
Infected leaves were collected at up to nine time points per isolate, and samples for transcriptome sequencing were selected based on their morphological infection stage (Supporting Information Figure   S2; Table 1, Supporting Information Table S2). Thereby, we generated stage-specific RNA-seq datasets corresponding to the four core infection stages, allowing us to compare the isolate-specific expression profiles at the same stage of infection development.
We obtained 89.2 to 147.5 million single-end, strand-specific reads per replicate (total >2.7 billion reads) that were quality trimmed and filtered. Between 4.54% (early infection) and 76.4% (late infection) of the reads could be mapped to the genome of the respective isolate, reflecting the infection stage-specific amount of fungal biomass (Table 1, Supporting Information Table S3; Text S1).
Across all isolates, transcriptomes of stages A and B, representing biotrophic growth, cluster together and are clearly different from transcriptomes of stages C and D that likewise cluster and represent necrotrophic growth of Z. tritici (Supporting Information Figures S9 and S10). Exploring the transcriptome datasets based on gene read counts shows the greatest variation of biological replicates for Zt10 at stage C (Supporting Information Figure S11), possibly reflecting variability in the infection development of the two biological replicates.

| Core Z. tritici transcriptional program during wheat infection
The mean expression of genes located on accessory chromosomes was 6-to 20-fold lower than the expression levels of genes located on core chromosomes (Supporting Information Table S11). We performed differential gene expression analyses to compare transcription of the 10,426 Z. tritici core genes. We hypothesized that genes involved in stage-specific infection development and transitions between the four core stages have dynamic, stage-specific expression profiles that are shared among the three Z. tritici isolates. In total, we identified only 597 genes (5.6%) that fulfilled these two criteria: differential expression between infection stages (DESeq2, p adj ≤0.01, |log 2 fold change| ≥2) and shared expression kinetics-and hence represent putative determinants of the Z. tritici core infection stages ( Figure 4a). Interestingly, 79 of these genes were differentially expressed between several infection stages, suggesting dynamic, wave-like expression kinetics (Supporting Information Figure S12).
A total of 246 genes were differentially expressed (Supporting Information Table S12)  HAUEISEN Et Al. (Kubicek, Starr, & Glass, 2014); polyketide synthases; and cytochrome P450s. These transcriptional changes reflect the previously described metabolic reprogramming of Z. tritici during the transition from biotrophic to necrotrophic growth (Rudd et al., 2015). Instead of feeding from intracellular lipids, the fungus switches to utilize plant-derived nutrients (Rudd et al., 2015) and rapidly develops large hyphal networks and the primal structures of the asexual pycnidia in the infected wheat mesophyll tissue.
Between stages C and D, only 74 genes were differentially expressed (Supporting Information Table S14)  These genes may be involved in determining the stage-specific infection development and include candidates for core virulence determinants (see below).

| Core biotrophic and necrotrophic effector candidates with shared expression profiles in Z. tritici isolates
Given their importance in plant-pathogen interactions, we particularly focused our analyses on genes encoding candidate secreted effector proteins (CSEPs) (Lo Presti et al., 2015). Zymoseptoria tritici F I G U R E 4 Zymoseptoria tritici core transcriptional program during wheat infection and isolate-specific expression during the four infection stages. Numbers of significantly differentially expressed genes across all isolates (a) between the four core Z. tritici infection stages and (b) between the isolates within the infection stages (between Zt05 and Zt09: orange arrows, between Zt05 and Zt10: purple arrows, between Zt09 and Zt10: green arrows). Small arrows (↑) with stage or isolate names indicate the number of genes specifically up-regulated during that stage or in that isolate for the respective comparison. Differential gene expression analyses performed with DESeq2. Genes were considered to be significantly differentially expressed if p adj ≤ 0.01 and |log 2 fold change| ≥ 2. *Indicates significant enrichment of effector candidates among differentially expressed genes (Fischer's exact tests, p < 0.001). Effector candidates encode secreted proteins putatively involved in modulating molecular host-pathogen interactions (Lo Presti et al., 2015) CSEP-encoding genes were previously predicted (Stukenbrock & Dutheil, 2018) using the machine learning approach EffectorP (Sperschneider et al., 2015) and are significantly enriched among the core differentially expressed genes (p ≤ 2.7 × 10 −13 , Fischer's exact tests) (Figure 4a), indicating highly dynamic transcription of core effectors during all stages of wheat infection. We filtered the differentially expressed CSEP genes according to their expression profiles (Supporting Information Figures S13 and S14) to identify putative key genes facilitating biotrophic and necrotrophic growth in wheat (Tables 2 and 3).  Figure S13) that mostly encode hypothetical proteins. In comparison with Zt05 and Zt09, expression of the biotrophic effectors in Zt10 is in general lower, possibly reflecting the strongly limited biotrophic colonization of this isolate (Figure 3b).
Thirty five CSEP genes are specifically up-regulated at stage C (Supporting Information Figure S14) and represent candidates for necrotrophic core effectors (Table 3, Supporting Information Table   S16). These genes may be involved in the transition from biotrophic to necrotrophic growth and the induction of necrosis. Nine CSEP genes encode putative plant cell wall-degrading enzymes and cutinase-like proteins, demonstrating that the lifestyle switch to necrotrophy involves intensified degradation of plant tissue and cell wall components. The gene Zt09_chr_9_00038 encodes a putative hydrophobin; hydrophobins are small fungal-specific proteins with various functions (Aimanianda et al., 2009;Wösten, 2001), i.a. as toxins in plant-pathogen interactions (Takai, 1974). Likewise strongly induced during the transition to necrotrophy is the gene Zt09_chr_7_00263 that encodes a putative secreted metalloprotease, which are known fungal virulence factors in animal and plant pathogens (Karimi Jashni et al., 2015;Naumann, Wicklow, & Price, 2011;O'Connell et al., 2012;Vu et al., 2014).
In summary, our comparative approach allowed us to identify a set of core Z. tritici effector candidates, which are consistently expressed during infection of the cultivar Obelisk in the different genetic background of Zt05, Zt09, and Zt10.

| Isolate-specific transcriptional changes during wheat infection
The 597 genes that we identified as differentially expressed between the stages show the same expression profile in each of the three isolates and we consider them as part of the core Z. tritici transcriptional infection program. However, we observed that transcript levels of many genes strongly deviate between the isolates during the specific infection stages.
To further study how the infection phenotypes of the Z. tritici isolates relate to differences in gene expression, we compared expression profiles during the infection stages (Figure 4b). In total, 2,377 (~22.8%) of the 10,426 shared genes are differentially Note. Summary of core Z. tritici biotrophic effector candidate genes that were identified based on their specific expression profiles within the Z. tritici core transcriptional program during wheat infection. Functional annotation, PFAM, and GO term information from Grandaubert et al. (2015).
expressed between the Z. tritici isolates during wheat infection (Table 4, Supporting Information Tables S17, S19) reflecting the extent of spatial, temporal, and quantitative differences in development that we observe by confocal microscopy (Figure 3a).  Table S19). Figure 5 exemplifies the isolate-specific expression kinetics of five CSEP genes. These five genes encode one hypothetical effector (Zt09_chr_12_00427) and secreted proteins with various functions: a hydrophobin (Zt09_ chr_9_00020) also shown previously to be differentially expressed Note. PCWDE: putative plant cell wall-degrading enzyme. Summary of core Z. tritici necrotrophic effector candidate genes that were identified based on their specific expression profiles within the Z. tritici core transcriptional program during wheat infection. Functional annotation, PFAM, and GO term information from Grandaubert et al. (2015).
between Swiss field isolates (Palma-Guerrero et al., 2017), a DNase (Zt09_chr_2_01162), and the ribonuclease Zt6 (Zt09_chr_3_00610), which possesses ribotoxin-like activity and is cytotoxic against plants and various microbes (Kettles, Bayon, Sparks, et al., 2017). The gene Zt09_chr_4_00039 encodes a protein with homology to the phytotoxin cerato-platanin of Ceratocystis fimbriata which was shown to induce necrosis and defense responses in plane trees (Pazzagli et al., 1999). During all infection stages, Zt09_chr_4_00039 is significantly higher expressed in Zt09 and might contribute to the higher necrosis levels caused by Zt09 (Figure 1a).
In addition to the differences in the expression of CSEP genes, we also noted isolate-specific expression patterns for genes located on accessory chromosomes. For example, three neighboring genes located on chromosome 19 in Zt09 (Zt09_chr_19_00071 to Zt09_ chr_19_00073) are significantly higher expressed in Zt10 during all four infection stages (Supporting Information Figure S15;  (Schotanus et al., 2015). In Fusarium graminearum, the histone modification H3K27me3 is associated with gene clusters encoding secondary metabolites and pathogenicity-related traits (Connolly, Smith, & Freitag, 2013). It is possible that variation in the To test this, we assessed the distances of all genes to the closest annotated transposable element. In the genomes of all three isolates, we found that isolate-specific differentially expressed genes are located significantly closer to transposable elements than genes that were not differentially expressed (Mann-Whitney U tests, p < 2.2 × 10 −16 ). Although more than 50% of these differentially expressed genes are located downstream of the closest transposable element, we did not observe an overall enrichment of TEs in their putative promotor regions 2 kb upstream (Supporting Information  (Schotanus et al., 2015).
Our findings indicate that during host infection, chromatin state of repeat-rich genome compartments is highly dynamic and changes between "active" euchromatin and "repressive" heterochromatin, as suggested in Leptosphaeria maculans (Soyer et al., 2014 Figure S18) and demonstrated conservation of previously identified patterns, such as transcriptional silencing of the right arm of chromosome 7 (Kellner et al., 2014;Rudd et al., 2015) in Zt05, Zt09, and Zt10 (Supporting Information Figure S17). This chromosomal segment has characteristics of an accessory chromosome, as it is significantly enriched with H3K27me3 that mediates transcriptional silencing (Schotanus et al., 2015). While syntenic chromosomal regions generally have a similar composition of transcribed and silenced loci, the fine-scale distribution of transcriptional cold and hot spots is clearly different between the genomes of the three isolates studied here. showing that "host specialization" in Z. tritici involves a very flexible strategy to exploit wheat tissue for growth and reproduction. This flexibility may be facilitated by the fact that pathogenesis does not involve the formation of complex penetration or feeding structures that require fine-tuned developmental programs. In powdery mildew which need to form appressoria as well as haustoria to establish host infections, comparative transcriptomics during early pathogenesis identified a comparatively low number of isolate-specifically expressed genes in two barley powdery mildew strains (Hacquard et al., 2013) and showed that transcriptional programs are similar even during incompatible infections and interactions with nonhost species (Hacquard et al., 2013;Hu et al., 2018).

| D ISCUSS I ON
As necrotic lesions are usually composed of several distinct Z. tritici genotypes (Linde et al., 2002), it would be highly relevant to investigate whether strains in one lesion have similar or different infection phenotypes. Isolates with different infection strategies colonizing the same leaf could complement each other or have antagonistic effects. Future multi-isolate studies must however consider possible developmental asynchrony during infection and should focus on isolates showing similar temporal infection development.
Our findings likewise underline the importance to consider not only quantitative infection outcomes but also qualitative aspects, like the timing of disease progress over an appropriate period of time, to be able to compare the infections of different pathogen isolates and detect meaningful differences in their infection phenotypes (Habig et al., 2017;Meile et al., 2018).
An intriguing question that emerges from our analyses is which factors cause deviation in gene expression phenotypes in Z. tritici.
Genetic variants associated with transcriptional regulation likely contribute to differences in gene regulation. However, we hypoth-

ACK N OWLED G M ENTS
We thank Ronny Kellner for insightful comments to a previous version of this manuscript, Julien Y. Dutheil for support with comparative transcriptome analyses, and Petra Happel for help with the in vitro stress assays. This work was supported by intramural funding of the Max Planck Society and a personal grant from the State of Schleswig-Holstein to Eva H. Stukenbrock. The funders had no role in study design, data collection and analyses, decision to publish, or preparation of the manuscript.

CO N FLI C T O F I NTE R E S T
None declared.

DATA ACCE SS I B I LIT Y
All generated RNA-seq datasets have been deposited at the NCBI