Recurrent Chromosome Destabilization Through Repeat-Mediated Rearrangements in a Fungal Pathogen

TL;DR

This study identifies specific genetic elements that cause recurring chromosomal instability in a fungal pathogen, leading to degenerative cycles and reduced virulence.

Contribution

The paper identifies transposable elements and repeat-induced mutations as key drivers of chromosomal instability in Zymoseptoria tritici.

Findings

Fragile sites in Zymoseptoria tritici trigger reproducible chromosomal rearrangements through nonallelic recombination.

A transposable element family is linked to chromosomal instability and reduced pathogen virulence.

Repeat-induced point mutation contributes to hypermutation in duplicated sequences, accelerating chromosomal degeneration.

Abstract

Genomic instability caused by chromosomal rearrangements has severe consequences for organismal fitness and progression of cancerous cell lines. The triggers of destabilized chromosomes remain poorly understood but likely co-locate with fragile sites. Here, we retrace a runaway chromosomal degeneration process observed in the fungal pathogen Zymoseptoria tritici using telomere-to-telomere assemblies across an experimental progeny. We show that the same fragile sites triggered reproducible, large-scale rearrangements through nonallelic recombination. Across our four-generation progeny, chromosomal rearrangements were accompanied by nondisjunction events leading to aneuploid progeny with up to four chromosomal copies. We identify a specific transposable element family co-locating with fragile sites, likely triggering ongoing repeated chromosomal degeneration. The element has recently been associated with lower virulence of the pathogen and has undergone an expansion of copy numbers across the genome. Chromosome sequences are also targeted by repeat-induced point mutation, a genome defense mechanism actively leading to hypermutation on duplicated sequences. Our work identifies the exact sequence triggers that initiate chromosome instability and perpetuate degenerative cycles. Dissecting proximate causes leading to runaway chromosomal degeneration could expand our understanding of chromosomal evolution beyond fungal pathogens.

Figures

Pedigree analyses revealing repeated occurrence of enlarged chromosome 17. a) Pedigree of four rounds of meiosis. The colors and symbols indicate whether the isolate carries a parental, unrelated (not carried by a parent), rearranged, or no chromosome 17 according to pulsed-field gel electrophoresis (PFGE) analysis. b) PFGE of chromosomes from the progeny (A66.2 and A2.2) and the parents (1E4 and 1A5), adapted from Croll et al. (2013). Orange arrows indicate the enlarged chromosome 17, and the blue arrow indicates the parental chromosome 17. c) Schematic and read coverage of the region in the progeny A66.2. The colors indicate the different blocks involved in chromosomal rearrangements. The coverage and breakpoints of progeny A66.2 reads mapped to the parent 1A5. Black and gray horizontal dashed lines indicate mean coverage and 2× mean coverage of the core chromosomes, respectively. Red dots indicate the mean coverage in 1 kb windows (regions with excessive, >300× coverage were removed). Dashed vertical lines indicate the chromosomal breakpoints identified from mapped reads (at positions B and E, turquoise and continuing in gray). Solid vertical lines indicate the positions of loci amplified by PCR (gray, amplification; red, no amplification). d) Dotplot of the assembled chromosome for progeny A66.2 compared to the parental chromosome. Inverted regions are indicated in blue. e) Mapping location and span of split reads: (i) deletion between B and C and (ii) translocation. At the bottom is a schematic of the resulting enlarged chromosome in the progeny. f) Schematic representation of the breakpoints and rearrangements between the two chromatids of 1A5 that generated the enlarged chromosome 17 (0 to A]BC[DEBA to 0) recovered in progeny A66.2. Arrows indicate the region of the two Styx copies, where nonallelic recombination might have occurred. The hypothetical smaller rearrangement product (FBCDEF) was not observed in PFGEs but is expected to have been produced based on the rearrangement pattern of the larger variant.

Sequence of chromosome 17 rearrangements tracked through four rounds of meiosis. Chromosome rearrangements are shown according to breakpoint positions A to F identified through split read mapping. Arrows linking different letters A to F indicate translocations and deletions for the successive rounds of meiosis. Panels on the right show schematic representations of each chromosome 17 variant present in each progeny. V0 labels the variant found after the first round of meiosis in both A2.2 and A66.2. An in-depth visualization of these first two meioses is available in Fig. S2. V1 to V5 label the different variants found over all progenies.

Reconstructed chromosome 17 variants based on sequence rearrangement breakpoints. For each progeny the mapped read coverage is shown relative to the 1A5 reference genome sequence. Black and gray horizontal dashed lines indicate mean coverage and 2×/3×/4× mean coverage of the core chromosomes, respectively. Red dots indicate the mean coverage in 1 kb windows. Vertical dashed lines indicate the chromosomal breakpoints A to F (see Figs. 1 and 2). Solid vertical lines indicate the positions of loci amplified by PCR (gray, amplification; red, no amplification). Below each coverage plot, arrows show variants reconstructed for each progeny. If multiple variants were detected, variants are labeled with V1 to V4. Variant labels are independent between progenies. Arrow color is based on the level of duplication (absent, single-copy, or multiple-copy regions). The progeny is represented by a) the progeny of the first meiotic round, b) second, c) third, and d to e) fourth round.

Characterization of repeat-induced point mutation (RIP) acting on duplicated sequences. a) Read coverage along chromosome 17 and evidence for a duplicated region 0 to B (0 to A in Ztprog11). The upper schematic shows how RIP mutations are introduced into duplication regions, causing an excess of adenine and thymine with RIP acting on guanine or cytosine. The lower schematic shows how mapped PacBio reads in duplicated regions were used for SNP calling, revealing potentially RIP-like mutations. b) Overview of progeny over generations with barplots summarizing the number of RIP-like and other mutations detected per progeny generation in region 0 to B (0 to A in Ztprog11). Dotplots show the percentage of reads carrying the alternative allele (parent 1A5). The identified copy numbers of the region 0 to B (0 to A in Ztprog11) are indicated. Blue dashed lines show the expected percentage of reads, confirming a new mutation carried by at least one copy of the duplicated regions. c) Breakdown of all mutations detected in duplicated regions of each progeny. Mutations at particular dinucleotides are summarized individually per progeny.

Recent expansion of the Styx transposable element underpinning the chromosomal rearrangements. a) The length and location of coding regions of the long and short copies of Styx. Schematic and read coverage of the region in the progeny A66.2. The colors indicate the different blocks involved in chromosomal rearrangements. Dotplot of the consensus sequence showing duplicated regions indicating the terminal inverted repeats (red circle). b) Copy number, GC content, and length distribution of Styx copies recovered from a global panel of Z. tritici reference genomes. A dark gray background indicates genomes of sister species Z. ardabiliae (Za17), Z. passerini (Zpa63), Z. brevis (Zb87), and Z. pseudotritici (Zp13). c) Phylogeny of Styx copies in Z. tritici, with Z. pseudotritici as an outgroup. The color scale indicates the GC content. The two full-length copies found on chromosome 17 of the parent 1A5 are indicated by a star. d) Density plot of branch length against GC content of individual Styx copies. The most recent burst of Styx copies shows small branch lengths and high GC content.

Full text

Introduction

Meiosis is a highly conserved process in eukaryotes, whereby homologous chromosomes pair, undergo recombination, and separate into daughter cells. Aberrations during the faithful transmission of chromosomes through meiosis can have serious consequences for an organism. Nondisjunction events resulting in additional or fewer chromosomal copies occur frequently in humans and are the leading causes of miscarriage (Hassold and Hunt 2001). Similarly, nondisjunction events and chromosome rearrangements occur in somatic lines and may lead to genomic instability, which is a hallmark of cancers (Negrini et al. 2010). However, factors that cause fragility of precancerous genomes or that trigger meiotic errors are largely unknown. Chromosomal breakage is a major factor contributing to instability and occurs often at specific loci referred to as fragile sites (Durkin and Glover 2007). Although the locations of many mitotic fragile sites have been identified in the human genome (Kumar et al. 2019), how they contribute mechanistically to breakage remains poorly understood. Fragile sites for nonallelic homologous or ectopic recombination during meiosis remain largely unknown, possibly because major rearrangements tend to be lethal and are quickly purged from the population by purifying selection. Investigations of nonlethal rearrangements offer a promising approach to unravel sequence determinants of fragile sites.

Accessory chromosomes in plant and fungal genomes are powerful models to study nonlethal chromosome rearrangements. These chromosomes (also called supernumerary or B) are present in a karyotype in addition to the regular chromosomes, contain no essential genes, and show broad presence/absence within species (Camacho et al. 2000). It is unclear what mechanisms generate these chromosome rearrangements and whether fragile sites are involved. For example, the accessory chromosomes of Magnaporthe oryzae contain many rearrangements and frequent interchromosome translocations between core and accessory chromosomes (Langner et al. 2021). Accessory chromosomes are hypothesized to originate from core chromosomes through rearrangements involving the terminal regions of core chromosomes. Zymoseptoria tritici, a haploid fungal pathogen of wheat, exhibits some of the most extreme degrees of structural variation observed within fungal species, including large differences in terms of chromosome length, transposable element (TE) and gene content, and recombination rate, as well as telomere and centromere structures (Croll et al. 2013; Schotanus et al. 2015; Plissonneau et al. 2016, 2018; Sánchez-Vallet et al. 2018; Fouché et al. 2018a; Badet et al. 2020). Chromosome rearrangements occur frequently during meiosis, with higher rates of rearrangements observed in the accessory chromosomes (Croll et al. 2013; Fouché et al. 2018b).

We analyzed a rearranged accessory chromosome of Z. tritici, which was generated repeatedly during meiosis through an unknown mechanism (Croll et al. 2013). Here, we show how the rearranged chromosome was generated and how further rearrangements occurred through subsequent rounds of meiosis. We also find evidence that a transposable element acts as the specific sequence trigger of the chromosome rearrangement and degenerative cycles.

Results

Origin of a Complex Chromosome Rearrangement

The rearrangement of the Z. tritici accessory chromosome 17 was first discovered in two F1 progeny A66.2 and A2.2 (Fig. 1a) (Croll et al. 2013). The enlarged nature of chromosome 17 in these progeny was confirmed by pulsed-field gel electrophoresis (PFGE) as well as southern hybridization (Fig. 1b) (Croll et al. 2013). To assess how frequently rearrangements of chromosome 17 occur through meiosis, we screened an additional 48 progeny from the same cross with a segment-specific PCR assay for ∼500 bp regions of coding sequences at regular intervals along chromosome 17 and found that anomalies of chromosome 17 were most likely restricted to the two initially discovered progeny A66.2 and A2.2 (Fig. 1a, Fig. S1). A previous screen based on restriction site-associated DNA sequencing of a further 228 progeny from this cross revealed four progeny with likely rearranged chromosome 17 (Fouché et al. 2018b). Finally, we assessed the frequency of rearranged chromosome 17 variants using whole-genome sequencing of 150 field-collected isolates matching the sampling location of the parental isolates. Based on read coverage variation, the population carries likely three disomic or partially duplicated variants of chromosome 17 (Singh et al. 2021).

Pedigree analyses revealing repeated occurrence of enlarged chromosome 17. a) Pedigree of four rounds of meiosis. The colors and symbols indicate whether the isolate carries a parental, unrelated (not carried by a parent), rearranged, or no chromosome 17 according to pulsed-field gel electrophoresis (PFGE) analysis. b) PFGE of chromosomes from the progeny (A66.2 and A2.2) and the parents (1E4 and 1A5), adapted from Croll et al. (2013). Orange arrows indicate the enlarged chromosome 17, and the blue arrow indicates the parental chromosome 17. c) Schematic and read coverage of the region in the progeny A66.2. The colors indicate the different blocks involved in chromosomal rearrangements. The coverage and breakpoints of progeny A66.2 reads mapped to the parent 1A5. Black and gray horizontal dashed lines indicate mean coverage and 2× mean coverage of the core chromosomes, respectively. Red dots indicate the mean coverage in 1 kb windows (regions with excessive, >300× coverage were removed). Dashed vertical lines indicate the chromosomal breakpoints identified from mapped reads (at positions B and E, turquoise and continuing in gray). Solid vertical lines indicate the positions of loci amplified by PCR (gray, amplification; red, no amplification). d) Dotplot of the assembled chromosome for progeny A66.2 compared to the parental chromosome. Inverted regions are indicated in blue. e) Mapping location and span of split reads: (i) deletion between B and C and (ii) translocation. At the bottom is a schematic of the resulting enlarged chromosome in the progeny. f) Schematic representation of the breakpoints and rearrangements between the two chromatids of 1A5 that generated the enlarged chromosome 17 (0 to A]BC[DEBA to 0) recovered in progeny A66.2. Arrows indicate the region of the two Styx copies, where nonallelic recombination might have occurred. The hypothetical smaller rearrangement product (FBCDEF) was not observed in PFGEs but is expected to have been produced based on the rearrangement pattern of the larger variant.

To investigate at base-pair resolution how the enlarged chromosome 17 was generated in progeny A66.2 and A2.2, we sequenced the isolates with long-read PacBio technology. The recency of the duplication creates a significant challenge for the chromosomal assembly as duplicated regions are not expected to show sequence divergence and collapse during assembly. We used a combination of coverage and split long-read mapping approaches to overcome this challenge. First, we mapped reads from the progeny to the 1A5 genome and identified regions of chromosome 17 with higher than average coverage compared to the core chromosomes (1 to 13). Mean read coverage transitioned sharply along chromosome 17 (Fig. 1c). We found that region 0 to B in progeny A66.2 and A2.2 each had approximately double the mean coverage of the core chromosomes, implying that this region was duplicated (Fig. 1c, Figs. S2 and S3, Table S1). We found that the region 0 to B is duplicated in the progeny genome and that the duplicated sequences are connected to two distinct locations of the chromosome at C and E (Fig. 1d, Figs. S2 and S3). Further supporting evidence included a breakpoint at the end of segment 0 to B, with reads showing split mapping and linking two distinct regions of the chromosome (Fig. 1e, Fig. S2, Table S2). Reads mapping to this position were evenly split between canonically mapping across positions B to C and split reads mapping from A to B and continuing in an inverted orientation from position E toward D (Fig. 1e, Fig. S2).

The rearranged chromosome 17 carries a single-copy region between C and E. A lack of coverage after position E toward F suggests that this region is absent in the rearranged chromosome. We confirmed the absence of region E to F by PCR (Figs. S2 and S3). A lack of coverage between B and C indicates a deletion of this region. Using the information on coverage and reads spanning distinct chromosomal regions, the chromosome 17 of progeny A66.2 and A2.2 are composed of region 0 to B, followed by region C to E, and a second copy of region 0 to B in an inverted orientation (Fig. 1c; Fig. S2, Table S2). We identified the putative centromere to be located between positions 405,779 and 415,898 based on sequence homology to the canonical chromosome 17 (Schotanus et al. 2015). Therefore, the resulting enlarged chromosome 17 is expected to carry a single copy of the centromeric region (Fig. 1e). Both progeny most likely inherited an identical, enlarged chromosome 17 generated during the first round of meiosis through the same nonallelic recombination between sister chromatids at locations B and E (Fig. 1f). The rearrangement should also have produced a second, shorter chromosomal variant. However, the shorter variant was not recovered in progeny, despite being predicted to carry a centromere. We reconstructed the sequence of the rearranged chromosome 17 to be 819 kb in length, matching the approximate length identified by PFGE (Fig. 1b).

Sustained Chromosome Degeneration in Subsequent Rounds of Meiosis

We investigated the fate of the enlarged chromosome 17 in further rounds of meiosis by performing several backcrosses (Fig. 2a). Progeny strains were selected based on abnormalities discovered through chromosomal size pattern screens following separation by pulsed-field gel electrophoresis (PFGE). New variants of chromosome 17 were already generated in the second generation (Figs. 2b and 3b, Fig. S3). Chromosome 17 of Ztprog11 carries a duplicated region, 0 to A, and the second copy of 0 to A is joined to position E in an inverted orientation (Figs. 2b and 3b, Figs. S3 and S4). Ztprog01 has two copies of chromosome 17, including a small and a large variant consistent with nondisjunction. One copy lacks the segment between B and D (Figs. 2b and 3b, Figs. S3 and S5). During the third round of meiosis, ZtProg19 stably inherited a chromosome 17 variant (Figs. 2d and 3c, Figs. S3 and S7). Based on progeny and read mapping evidence, the second progeny ZtProg45 carries multiple variants. Variant V2 may have undergone nondisjunction (2× V1 is inherited with 1× V4) or one copy of V1 is present with one copy of V2 and V3 each, indicating rearrangements at breakpoints B and D as well as nondisjunction (Figs. 2c and 3c, Figs. S3 and S6). Therefore, two small variants are present together with one large variant.

Sequence of chromosome 17 rearrangements tracked through four rounds of meiosis. Chromosome rearrangements are shown according to breakpoint positions A to F identified through split read mapping. Arrows linking different letters A to F indicate translocations and deletions for the successive rounds of meiosis. Panels on the right show schematic representations of each chromosome 17 variant present in each progeny. V0 labels the variant found after the first round of meiosis in both A2.2 and A66.2. An in-depth visualization of these first two meioses is available in Fig. S2. V1 to V5 label the different variants found over all progenies.

Reconstructed chromosome 17 variants based on sequence rearrangement breakpoints. For each progeny the mapped read coverage is shown relative to the 1A5 reference genome sequence. Black and gray horizontal dashed lines indicate mean coverage and 2×/3×/4× mean coverage of the core chromosomes, respectively. Red dots indicate the mean coverage in 1 kb windows. Vertical dashed lines indicate the chromosomal breakpoints A to F (see Figs. 1 and 2). Solid vertical lines indicate the positions of loci amplified by PCR (gray, amplification; red, no amplification). Below each coverage plot, arrows show variants reconstructed for each progeny. If multiple variants were detected, variants are labeled with V1 to V4. Variant labels are independent between progenies. Arrow color is based on the level of duplication (absent, single-copy, or multiple-copy regions). The progeny is represented by a) the progeny of the first meiotic round, b) second, c) third, and d to e) fourth round.

In the last round of meiosis, the region B to D was deleted in Ztprog8 and Ztprog64 (Figs. 2d and 3d and ii, Figs. S3, S8, and S9). Ztprog64 carries four chromosome 17 variants (three small and one large variant) that are missing the region B to D (Figs. 2d and 3i and ii, Figs. S3 and S9). Ztprog9 most likely inherited the enlarged chromosome 17 without further rearrangements from the previous generation (Figs. 2d and 3d, Figs. S3 and S10). Ztprog30 has the same chromosome complement as Ztprog01 (Figs. 2d and 3e, Figs. S3 and S11). The enlarged chromosome 17 was highly unstable through further rounds of meiosis and degenerative cycles via nonallelic recombination as well as nondisjunction occurred. Rearrangements at all breakpoints except A were observed several times in the experiment.

Degeneration of the Duplicated Sequence Through Repeat-Induced Point Mutations

Fungal genomes encode repeat-induced point mutations (RIP), a defense mechanism targeting duplicated sequences that dramatically reduces TE activity. We predicted that the presence of a large, duplicated region in chromosome 17 would trigger RIP (Fig. 4a). To detect RIP-like mutations, we mapped reads from progeny to the parental chromosome 17 and identified regions with transitions from G/C to A/T (Fig. 4b; Fig. S4). We found an overrepresentation of RIP-like mutations in region 0 to B. As a control, we analyzed the rearrangement-free progeny Ztprog08 for RIP-like mutations and found indeed no overrepresentation of RIP-like mutations, indicating RIP not being active. Hence, the genomic defenses are specifically active against the new duplications appearing in degenerate chromosome 17 variants. RIP-like mutations predominantly targeted CpA dinucleotides (or TpG) (Fig. 4c), matching evidence from other model fungi (Gladyshev et al. 2017). Additionally, CpG dinucleotides were also frequently targeted. New RIP-like mutations were detected after a single round of meiosis, suggesting that the RIP genome defense did not yet reach saturation. RIP-like mutations never targeted the putative centromere (405,779 to 415,898 bp; Fig. S4). Our results show that massive rearrangements and nondisjunction trigger the genome defense RIP, contributing to an increased mutation rate in the duplicated region.

Characterization of repeat-induced point mutation (RIP) acting on duplicated sequences. a) Read coverage along chromosome 17 and evidence for a duplicated region 0 to B (0 to A in Ztprog11). The upper schematic shows how RIP mutations are introduced into duplication regions, causing an excess of adenine and thymine with RIP acting on guanine or cytosine. The lower schematic shows how mapped PacBio reads in duplicated regions were used for SNP calling, revealing potentially RIP-like mutations. b) Overview of progeny over generations with barplots summarizing the number of RIP-like and other mutations detected per progeny generation in region 0 to B (0 to A in Ztprog11). Dotplots show the percentage of reads carrying the alternative allele (parent 1A5). The identified copy numbers of the region 0 to B (0 to A in Ztprog11) are indicated. Blue dashed lines show the expected percentage of reads, confirming a new mutation carried by at least one copy of the duplicated regions. c) Breakdown of all mutations detected in duplicated regions of each progeny. Mutations at particular dinucleotides are summarized individually per progeny.

Chromosomal Rearrangements Triggered by a Recent TE Burst

To understand the mechanism triggering the initial chromosome 17 rearrangement and sustaining the degeneration, we examined the breakpoint sequences. We identified a full copy of the DNA transposon Styx (also known as REP9) at position B and a partial copy of the same family at position C in the parental chromosome of 1A5 (Fig. 5a). Styx was previously described as a negative regulator of virulence (Wang et al. 2021). Furthermore, Styx was shown to proliferate in the progeny and pathogen populations across continents (Feurtey et al. 2023; Badet et al. 2024). At position B, the element is at 582 bp from the breakpoint. A second complete copy of Styx was found at position D (∼8 kb in length) (Fig. 5a). Styx has 21 copies in the 1A5 genome, including three copies on chromosome 17 (Fig. 5b). Using NCBI BLAST, we predicted four putative coding sequences in full-length copies of Styx (Fig. 5a), of which one shows weak homology to RNase H and an integrase (Badet et al. 2024). We analyzed Styx copy numbers in 19 completely assembled genomes of Z. tritici (Badet et al. 2020) and in the genomes of the sister species Z. passerinii, Z. brevis, and Z. pseudotritici (Feurtey et al. 2020). Styx has been reported at higher copy numbers in the sister species but nearly absent in genomes sampled from the center of origin of Z. tritici (Badet et al. 2024). Our findings support a scenario of Styx being present in the common ancestor at low copy numbers and undergoing independent bursts in the sister species, as well as in Z. tritici populations in North and South America, Europe, and Australia, resulting in increased copy numbers in the parental and other isolates (Fig. 5b) (Badet et al. 2024). The TE copy on chromosome 17 was introduced following a recent burst (Fig. 5c). TE copies created during the burst are characterized by short terminal branch lengths and high GC content (Fig. 5d). The creation of the enlarged chromosome 17 was likely mediated by nonallelic homologous recombination between Styx copies at position B and a different sequence with microhomology at position E. We examined the regions 1 kb up- and downstream from positions A, B, and E for similar repetitive sequences. Positions B and E carry a similar 6 bp repeat that may have ultimately triggered the rearrangement (Tables S5 to S6). Finally, near position E, the repeat is 34 bp away from the breakpoint. Similarly, positions A and E had two similar 3 bp repeats (Tables S6 to S7).

Recent expansion of the Styx transposable element underpinning the chromosomal rearrangements. a) The length and location of coding regions of the long and short copies of Styx. Schematic and read coverage of the region in the progeny A66.2. The colors indicate the different blocks involved in chromosomal rearrangements. Dotplot of the consensus sequence showing duplicated regions indicating the terminal inverted repeats (red circle). b) Copy number, GC content, and length distribution of Styx copies recovered from a global panel of Z. tritici reference genomes. A dark gray background indicates genomes of sister species Z. ardabiliae (Za17), Z. passerini (Zpa63), Z. brevis (Zb87), and Z. pseudotritici (Zp13). c) Phylogeny of Styx copies in Z. tritici, with Z. pseudotritici as an outgroup. The color scale indicates the GC content. The two full-length copies found on chromosome 17 of the parent 1A5 are indicated by a star. d) Density plot of branch length against GC content of individual Styx copies. The most recent burst of Styx copies shows small branch lengths and high GC content.

Discussion

We reveal the dynamics of an unusual chromosomal rearrangement in a fungal plant pathogen. Using split long-reads, we identified the exact breakpoints and retraced the degeneration through four rounds of meiosis. We propose that the primary degenerative rearrangement was caused by nonallelic recombination between a TE and a region with microhomology to the TE. The degenerated chromosome was composed of a large, duplicated, and inverted region connected to a single-copy chromosomal segment near the centromere. The sequences serving as triggers for the rearrangement in the first round of meiosis were likely the transposable element Styx, being present in multiple copies on chromosome 17 and sharing microhomology. The functional and fitness consequences of the chromosomal rearrangements remain unknown, because of limitations in our experimental pedigree. This is because for each chromosomal variant, progeny screening would have to be conducted in a further backcross to account for mutations co-segregating with a rearranged chromosome. Alternatively, experimental transfer protocols of rearranged chromosomes may be investigated. Chromosome 17 is dispensable and repetitive and encodes only a few genes without apparent homology (Badet et al. 2020; Singh et al. 2021; Goodwin et al. 2007). Field population surveys identified likely partial deletions and duplications, indicating that chromosomal rearrangements are also produced in the environment and at noticeable rates (Singh et al. 2021). Chromosome 17 might thus be under largely relaxed selection, and chromosomal rearrangements are tolerated.

However, chromosomal rearrangements co-locating with Styx copies were also detected in core chromosomes of the same cross (Badet et al. 2024). Subsequent rounds of meiosis increased the spectrum of rearrangement breakpoints consistent with runaway chromosomal degeneration. Chromosomal segments affected by the rearrangement were repeatedly affected multiple times, and three of these segments co-localize with copies of Styx. Concurrent with the degeneration, nondisjunction events of chromosome 17 increased along the progeny, characterized by disomy and trisomy. The frequency of nondisjunction events increased with the presence of chromosome 17 variants in both parents, suggesting that opportunities for mispairing of chromatids increase the likelihood of aberrant segregation. Not all of the observed rearrangements could be resolved at the sequence level. In particular, it remains unclear how the B to C deletion was triggered. It is unlikely that this deletion was already present in the cell pool of the parental strain, as all experiments including the original crossing and sequencing were handled from freezer stocks. Furthermore, the B to C deletion indicates that the chromosome 17 rearrangement might be even more complex than described here. The degenerating chromosome was furthermore affected by the genomic defense mechanism RIP, introducing mutations into recently duplicated sequences. RIP on very large duplicated sequences might have a titration effect, leading to less dense RIP mutations in smaller duplicated sequences (Singh and Kasbekar 2008). Overall, the identity of the paired parental genotypes had a major influence on the likelihood of rearrangements observed in the progeny.

The copy-number amplification in progeny Ztprog64 and Ztprog30 shares hallmarks of breakage-fusion-bridge (BFB) cycles observed in many cancer lines (Ciullo et al. 2002; Hellman et al. 2002; Marotta et al. 2012, 2013, 2017; Bianchi et al. 2019). Following the initial degeneration, chromosome 17 was stabilized in pairings of a single rearranged chromosome variant with a parental chromosomal variant (ie Ztprog8 and Ztprog9). This stabilization suggests that the progression of the degenerative cycle was interrupted through proper segregation. The chromosome 17 variant in Ztprog8 was generated through nonallelic homologous recombination between two Styx copies. The repair of chromosomes through recombination between repeats is known to produce intermediary chromosomes that can ultimately produce stable karyotypes (Hoang et al. 2010). The reconstructed chromosome 17 variants pinpointed the most likely sequence triggers for the observed rearrangements. The observation that specific locations on the chromosome were repeatedly involved in creating new rearrangements during chromosome degeneration indicates that these locations are fragile sites (ie sites frequently co-locating with chromosome rearrangements). Three out of four of these fragile sites co-locate with copies of Styx. The tendency of Styx to trigger chromosomal degeneration is likely widespread within the species, as the same TE is also responsible for an inter-chromosomal rearrangement (Badet et al. 2024). Isolates collected near the geographic center of origin of the pathogen in the Middle East carry no or only a few Styx copies in the genome, and the expansion to higher copy numbers appears restricted to European genotypes and their descendants (Badet et al. 2020). A separate and likely independent burst of amplification of Styx appears to have occurred in the sister species of Z. tritici, where no chromosomal rearrangements on chromosome 17 have been described yet (Oggenfuss and Croll 2023; Badet et al. 2024). The significant consequences for chromosomal integrity could mean that the activity of Styx and the presence of fragile sites are under strong selection.

Taken together, our results indicate that specific TE sequences can trigger runaway chromosome degeneration. Nonallelic homologous recombination drives the deleterious rearrangements at the onset of the process with nondisjunction events in subsequent rounds of meiosis. Specific regions on chromosomal segments are preferentially amplified, consistent with patterns observed during degenerative (mitotic) BFB cycles in cancer cell lines (Marotta et al. 2012; Umbreit et al. 2020). BFB was first discovered in maize by McClintock in dicentric chromosomes going through cycles of degeneration (McClintock 1938, 1941) and is also known to occur in animals (Toledo et al. 1993; Bi et al. 2004) and fungi (Rank et al. 1988; Hackett et al. 2001). BFB cycles are initiated via telomere-telomere fusions of chromosomes with degraded or missing telomeres (Maciejowski and de Lange 2017). The centromeres of the dicentric chromosome are pulled in opposite directions during anaphase, generating a bridge that breaks apart and results in daughter cells with different lengths of the chromosome that lack telomeres (Smith et al. 1990; Ma et al. 1993; Toledo et al. 1993; Coquelle et al. 1997; Marotta et al. 2013). At the karyotype level, chromosome 17 undergoes cycles consistent with BFB cycles. However, the likely absence of centromere duplications is inconsistent with classic BFB cycles. We pinpoint an alternative mechanism driving amplification of chromosomal regions in a pattern resembling BFB and involving ectopic recombination. Recombination serves both as the initial trigger to create unstable chromosomes and as the mechanism to maintain the degenerative process. Nondisjunction amplifies the process following the initial meiosis, producing the aberration. We show that large-scale chromosome rearrangements can spontaneously occur in natural pairings of fungal individuals, bearing hallmarks of degenerative processes in somatic cell lines such as those observed in cancers.

Methods

Establishment of a Four-Generation Progeny

To detect how chromosomal rearrangements evolved on chromosome 17, we analyzed four generations of crosses, including two backcrosses of the haploid progeny A66.2 (Fig. 1a). The initial cross between parental strains 1E4 and 1A5 and subsequent crosses were described earlier (Croll et al. 2013; Badet et al. 2021). We used previously generated telomere-to-telomere PacBio assembly of parental strains 1E4, 1A5, and 3D7 (Plissonneau et al. 2016, 2018). Parental genome assemblies were validated using high-density genetic maps (Croll et al. 2013; Lendenmann et al. 2014). Crosses were performed by coinfecting wheat leaves with asexual conidia from the parental strains of opposite mating types, according to an established crossing protocol (Kema et al. 1996): Conidia of both parents were sprayed onto wheat plants in equal concentration and incubated outdoors for 40 to 60 days. Ascospores were isolated over several days by incubating infected wheat leaves on wet filter paper inside Petri dishes. Wheat leaves were covered with upside-down Petri dish lids filled with water agar, enabling the capture of vertically ejected ascospores. Ascospores captured on the water agar were left to germinate and inspected for contaminants. Only progeny isolates from single ascospores were selected. Each germinating ascospore was transferred to an individual culture plate for clonal propagation. The mycelium produced by each ascospore was used for DNA extraction. Progeny mycelium was grown in yeast sucrose broth (YSB) liquid medium for 6 to 7 d at 20 °C prior to DNA extraction.

Chromosome Segment PCR Assay

In order to assess the presence–absence polymorphism of chromosome 17 segments, we used previously designed PCR assays to amplify ∼500 bp regions of coding sequences at regular intervals along the chromosome 17 of the reference strain IPO323 (Croll et al. 2013). Detailed information on primer binding sites in the genome of 1A5 is available in Table S3. PCR reactions were performed in 20 µl volumes with 5 to 10 ng genomic DNA, 0.5 mM of each primer, 0.25 mM dNTP, 0.6 U Taq polymerase (DreamTaq, Thermo Fisher Scientific, Inc.), and the corresponding PCR buffer. In order to avoid false negatives, we included a primer pair for a conserved microsatellite locus in each PCR mix (Goodwin et al. 2007). The amplification protocol was described in more detail previously (Croll and McDonald 2013). Successful PCRs produced an additional band that was clearly distinguishable from the PCR product associated with the amplified chromosome region. PCR products were analyzed on agarose gels. Data was visualized using the R package gplots (https://github.com/talgalili/gplots).

DNA Extraction for PacBio Sequencing

Progeny DNA from each cross was extracted using a modified version of the cetyltrimethylammonium bromide (CTAB) DNA extraction protocol developed for plant DNA extractions (Allen et al. 2006). Fungal cultures were grown for 5 to 7 d in YSB broth and lyophilized overnight. Approximately 60 to 100 mg of dried material was crushed with a mortar and pestle. The phenol-chloroform-isoamyl alcohol extraction step was performed twice and the washing step three times. In the last step, the DNA pellet was resuspended in 100 µl of sterile water.

Preparation of Fungal Material for Molecular Karyotyping

DNA from intact chromosomes was extracted from conidia from fungal cultures, embedded in agarose gels by the in situ digestion of cell walls, using a modified nonprotoplasting method (McCluskey et al. 1990). We included seven Z. tritici isolates confirmed with PCR testing to have inherited chromosome 17 from each of the crosses. Isolates were transferred from stocks maintained in glycerol at −80 °C to yeast malt agar plates and incubated for 3 to 4 d in the dark at 18 °C. Conidia were then isolated by washing the plates with sterile water and transferring 600 to 800 μl of suspended conidia to new YMA plates. The plates were again incubated for 2 to 3 d as described above. Conidia were harvested by washing the plates with sterile distilled water and filtering through sterile Miracloth (Calbiochem, La Jolla, California, United States) into 50 ml Falcon tubes. The volume was adjusted to 50 ml by adding more distilled water, and the suspension was centrifuged at 3,750 rpm at room temperature for 15 min with a clinical centrifuge (Allegra X-12R, Beckman Coulter, Brea, California, United States). The pellets were resuspended in 1 to 3 ml TE buffer (10 mM Tris-HCl, pH 7.5; 1 mM EDTA, pH 8.0) and vortexed gently. The spore concentration of the solution was calculated using a Thomas hemocytometer cell counter. The 1.5 ml spore suspensions with a concentration between 8 × 10^7^ and 2 × 10^8^ spores/ml were transferred to 50 ml Falcon tubes and incubated at 55 °C in a water bath for a few minutes. We added 1.5 ml prewarmed (55 °C) low-melting-point agarose prepared in TE buffer (2% w/v; molecular biology grade, Biofinex, Switzerland). The solution was mixed by gentle pipetting. An aliquot of 500 μl was solidified on ice for approximately 10 min in a precooled plug casting mold (Bio-Rad Laboratories, Switzerland). Agarose plugs were incubated in 15 ml Falcon tubes containing 5 ml of a lysing solution [0.25 M EDTA, pH 8.0, 1.5 mg/ml protease XIV (Sigma, St. Louis, Missouri, United States), 1.0% sodium dodecyl sulfate (Fluka, Switzerland)]. Plugs were incubated for ∼24 h at 55 °C. The lysing solution was changed once after ∼18 h and gently mixed every few hours. Plugs with whole chromosomal DNA were washed three times for 15 to 20 min in ∼5 ml of a 0.1 M EDTA (pH 9.0) solution and then stored in the same solution at 4 °C until use.

Pulsed-Field Gel Electrophoresis

To detect length differences in chromosome 17, PFGE was performed using a Bio-Rad CHEF II apparatus (Bio-Rad Laboratories, Hercules, California, United States). Chromosomal plugs were placed in the wells of a 1.2% (w/v) agarose gel (Invitrogen, Switzerland) to separate small chromosomes up to 1 Mb. Chromosomes were separated at 13 °C in 0.56× Tris-borate-EDTA buffer (Sambrook and Russell 2001) at 200 V with a 60 to 120 s pulse time gradient for 24 to 26 h. Gels were stained in ethidium bromide (0.5 mg/ml) for 30 min. Destaining was performed in water for 5 to 10 min. Photographs were taken under ultraviolet light with a Molecular Imager (Gel Doc XR+, Bio-Rad Laboratories, Switzerland). As size standards, we used chromosome preparations of Saccharomyces cerevisiae (Bio-Rad Laboratories, Switzerland).

Southern Hybridization for A66.2 and A2.2

Southern blotting and hybridization were performed following standard protocols (Sambrook and Russell 2001). First, hydrolysis was performed in 0.25 M HCl for 30 min, and then DNA was transferred onto Amersham Hybond-N+ membranes (GE HealthCare, Switzerland) overnight under alkaline conditions. DNA was heat-fixed onto the membranes at 80 °C for 2 h. Membranes were prehybridized overnight with 25 ml of a buffer containing 20% (w/v) SDS, 10% BSA, 0.5 M EDTA (pH 8.0), 1 M sodium phosphate (pH 7.2), and 0.5 ml of sonicated fish sperm solution (Roche Diagnostics, Switzerland). Probes were labeled with ^32^P by nick translation (New England Biolabs, Inc.) following the manufacturer's instructions. Hybridization was performed overnight at 65 °C. Blots were subjected to stringent wash conditions with a first wash in 16× SSC and 0.1% SDS and a second wash with 0.26× SSC and 0.1% SDS. Both washes were performed at 60 °C. Membranes were exposed to X-ray film (Kodak BioMax MS) for 2 to 3 d at −80 °C. We used the same probe as described earlier (Table S4) (Croll et al. 2013).

PacBio Library Preparation

PacBio SMRTbell libraries were prepared using 15 to 31 µg of high-molecular-weight DNA. The libraries were size-selected with an 8 kb cutoff on a BluePippin system (Sage Science, Inc.). After selection, the average fragment length was 15 kb. PacBio sequencing was run on a PacBio RS II instrument or Sequel at the Functional Genomics Center, Zurich, Switzerland, using P4/C2 and P6/C4 chemistry, respectively.

Assembly of Chromosome 17 and Breakpoint-Junction Analyses

Each chromosome 17 for all four generations was then assembled (Fig. 1a). We mapped the reads to the reference genome 1A5 using minimap2 version 2.17 (Li 2018) with the parameters --secondary=no -ax map-pb. We compared the coverage of regions of chromosome 17 to the mean coverage of the core chromosomes (1 to 13) to estimate copy numbers of distinct regions. We then identified breakpoints by analyzing regions with >15 reads that either end or start at a specific position using BEDTools bamtobed version 2.29.2 (Quinlan and Hall 2010) and extracted split reads in this region. We then assembled draft chromosomes by using the structural information from reads showing split alignments, and by joining individual breakpoints (Table S2). Reads were mapped to the assembled chromosomes with minimap2, and we counted the number of reads spanning each established junction point (Table S2). Established chromosome 17 assemblies were error-corrected with Quiver version 2.1.2 (Chin et al. 2013).

Characterization of Sequence Features of Breakpoints

Transposable elements consensus sequences were annotated on the parent 1A5 chromosome 17 sequence with RepeatMasker version 2.10.0 based on previously described TE family consensus sequences (Badet et al. 2020). The cutoff value was set to 250, and both simple repeats and low-complexity regions were discarded. We used nucmer version 4.0.0rc1 and dotPlotly (https://github.com/tpoorten/dotPlotly/) to visualize the Styx TE consensus sequence (Marçais et al. 2018). To further classify Styx, we analyzed conserved domains in the consensus sequence with online BLASTx on the nonredundant NCBI protein database. We detected four putative coding regions consistent with previous analyses of the TE (Badet et al. 2024). As a proxy for repeat-induced point mutations, we calculated the GC content for each copy with geecee from EMBOSS version 6.6.0 (Rice et al. 2000). To search for microhomology at rearrangement breakpoints, we analyzed tandem repeats within 1,000 bp of the breakpoint locations A to E using mreps with the parameters -exp 3 and -res 5 to allow for fuzzy detection of degenerate repeats (Kolpakov et al. 2003).

Genome-wide Description of the Styx TE Family

We extracted all annotated copies of Styx in the 19 reference-quality Z. tritici genomes and the sister species (Badet et al. 2020; Feurtey et al. 2020) with SAMtools faidx version 1.9 (Li et al. 2009). Multiple sequence alignments of all copies were made with MAFFT version 7.453 using the following parameters: --reorder --localpair --maxiterate 1000 --nomemsave --leavegappyregion (Katoh and Standley 2013). We located the putative coding region 4 in the multiple sequence alignment and extracted the sequence with extractalign from EMBOSS. We excluded empty hits (ie not containing coding regions) with trimAl version 1.4rev15 and sequences with more than 20% of gap sites with SeqKit version 0.11.0 (Capella-Gutiérrez et al. 2009; Shen et al. 2016). To remove large blocks of gaps and regions that represent rare insertions, we used Gblocks version 0.91b with the parameters: -t=d -b4=5 -b5=h (Castresana 2000). We used RAxML version 8.2.12 to create phylogenetic trees in three rounds (Croll et al. 2013): First, we generated 20 ML trees each with a different starting tree and extracted the tree with the best likelihood with the following parameters: raxmlHPC-PTHREADS-SSE3 -T 4 -m GTRGAMMA -p 12345 -# 10 --print-identical-sequences. Second, we made a bootstrap search for support values with the following parameters: raxmlHPC-PTHREADS-SSE3 -T 4 -m GTRGAMMA -p 12345 -b 12345 -# 50 --print-identical-sequences. Finally, we drew bipartitions on the best ML tree with bootstrapping: raxmlHPC-PTHREADS-SSE3 -T 4 -m GTRGAMMA -p 12345 -f b --print-identical-sequences. We imported the tree into R with read.tree from the package treeio version 1.10.0 (https://github.com/YuLab-SMU/treeio), converted it to a tibble object with as.tibble from package tibble version 3.0.1 in tidyverse version 1.3.0, and added the GC content per sequence as a variable with left_join from the package dplyr version 0.8.5 in tidyverse. The tree was then recreated with as.phylo and as.treedata in treeio (Wickham et al. 2019; Wang et al. 2020). We chose a sequence from Z. pseudotritici as an outgroup to root the Z. tritici tree. We visualized the tree with GGTree version 2.0.1 (Yu et al. 2017, 2018).

Repeat-Induced Point Mutation Analysis

To detect the impact of RIP in the progeny, we checked for biases of mutations for specific dinucleotides. We mapped PacBio reads to the reference genome 1A5 using minimap2 (Li 2018) as described above. We performed SNP calling with the software Longshot version 0.4.0 and included a minimum mapping quality cutoff of -q 30 (Edge and Bansal 2019). All SNPs with a “dn” tag, indicating mapping issues, were removed. We allowed for 20% errors in the uncorrected PacBio reads. We expected the alternate allele at polymorphic sites to be shared by ≥80% of the mapped reads for single-copy regions, shared by ≥40% of the reads in regions with evidence for duplications (based on read coverage), and shared by ≥27%, ≥20%, and ≥16% of the reads in regions with three, four, and five copies (based on read coverage), respectively. We also excluded regions with ≤50% or ≥150% of the expected read coverage for further analyses to reduce erroneous variant calls due to inconsistent read mapping.

Supplementary Material

evag037_Supplementary_Data