Epitranscriptomic RNA editing resolves Mus81 DNA repair tradeoffs in heat tolerance and meiosis

Abstract

Maintaining genome integrity is essential, yet how DNA repair is balanced across life stages remains poorly understood. Here we uncover an epitranscriptomic mechanism in the fungal pathogen Fusarium graminearum that alleviates a trade-off between heat-stress adaptation and sexual reproduction. We show that FgMus81 acts independently of its nuclease activity and canonical partner FgMms4, and has dosage-dependent, stage-specific functions: restrained levels support meiosis, whereas elevated levels promote heat-stress survival. We identify a sexual stage-specific A-to-I RNA editing event that recodes FgMus81 (N420D) and tunes its abundance to meet meiotic demands without compromising stress resilience. Notably, both pre-editing and post-editing isoforms support meiotic interhomolog crossovers, but the post-editing isoform impairs mitotic recombination. Conservation of this editing across Sordariomycetes suggests evolutionary selection for stage-specific control. Together, these findings reveal an epitranscriptomic switch that partitions Mus81 functions across life stages and identify adaptive RNA editing as a regulator of homologous recombination in fungal pathogens.

Maintaining genome integrity during different life stages is critical. Here, the authors reveal an RNA editing mechanism in the fungal pathogen Fusarium graminearum that adjusts Mus81 protein function to balance heat-stress survival and sexual reproduction. This stage-specific editing fine-tunes DNA repair for adaptation and reproduction.

Introduction

Maintaining genome integrity is a universal requirement, yet the optimal deployment of DNA repair activities is not constant across life stages and environments. Double-strand breaks (DSBs) and replication-associated lesions arise from endogenous metabolism as well as external stresses, and their misrepair threatens viability and genome stability^1,2. In eukaryotes, homologous recombination (HR) provides a high-fidelity route to repair such lesions using homologous DNA templates, but the extent to which HR components are required—and how their activities are tuned—can differ between vegetative growth, stress conditions, and sexual reproduction^3–5.

Programmed DSBs generated by Spo11 initiate meiotic recombination and are indispensable for faithful chromosome segregation and gamete formation^6–8. Mitotic HR primarily repairs DSBs and replication-associated lesions using the sister chromatid via the noncrossover pathway, preserving genomic stability and preventing chromosomal rearrangements⁹. In contrast, meiotic HR promotes genetic diversity through crossovers between homologous chromosomes during meiosis I^8,9. In both mitotic and meiotic contexts, a central step in HR is catalyzed by the Rad51 recombinase, which promotes homology search and strand invasion to form joint-molecule intermediates¹⁰. These intermediates must then be processed through competing outcomes, including noncrossovers and crossovers. In many eukaryotes, a major class of meiotic crossovers is produced through a pathway involving ZMM proteins and the MutLγ complex Mlh1–Mlh3, which promotes resolution of recombination intermediates toward crossovers with interference^7,11,12. In parallel, additional structure-selective endonucleases can process persistent or atypical joint molecules, providing alternative routes to resolve recombination and replication-associated DNA structures. These include Mus81–Mms4, Slx1–Slx4, and Yen1, which can cleave branched intermediates (e.g., nicked junctions or related structures) to yield crossover or noncrossover products depending on substrate configuration and cellular context^13,14. Importantly, the relative contributions of Mlh1–Mlh3 versus Mus81–Mms4/Slx1–Slx4/Yen1 are known to vary across organisms and conditions, implying that pathway choice and the division of labor among these factors are evolutionarily and physiologically plastic^7,11,12.

Mus81 is best characterized as functioning together with a regulatory partner—Mms4 (Eme1 in Schizosaccharomyce pombe; Eme1/Eme2 in mammals)—to form a structure-selective endonuclease complex whose activity depends on this partnership^15,16. The activity of the Mus81–Mms4 complex is critical for processing replication and recombination intermediates, especially under replication stress^17,18. It resolves toxic DNA structures such as stalled replication forks, nicked Holliday junctions, and under-replicated DNA at common fragile sites^18–20, which, if left unresolved, would disrupt chromosome segregation and cause genetic instability. Because inappropriate cleavage of HR intermediates can be harmful, Mus81–Mms4 activity is typically subject to tight regulation, including cell-cycle-dependent control and regulated access to DNA substrates, ensuring that cleavage occurs at the appropriate time and place^21–24. Nevertheless, how the Mus81–Mms4 complex–centered processing is coordinated with other joint-molecule resolution pathways, and how this coordination is modulated by environmental stress versus sexual development, remains insufficiently understood in filamentous fungi. This knowledge gap is particularly relevant for species in which growth-stage stress adaptation and meiosis impose potentially conflicting constraints on genome maintenance activities.

Fusarium graminearum is the main causal agent of Fusarium head blight (FHB), a devastating disease that severely threatens global cereal crop yields^25,26. Sexual reproduction is central to its life cycle, as forcibly discharged ascospores from sexual fruiting bodies (perithecia) are the primary inoculum for FHB epidemics²⁷. Meiotic recombination during sexual development greatly increases genetic diversity, enabling the rapid emergence of diverse genotypes that can adapt to environmental changes and host defenses^28,29. This high-frequency genetic reshuffling accelerates the evolution of fungicide resistance and enhanced virulence, sustaining aggressive field populations^30–32.

As a homothallic fungus, F. graminearum can self-fertilize and readily forms perithecia under laboratory conditions, making it a suitable model for studying the molecular basis of sexual development²⁷. Recent research has revealed that adenosine-to-inosine (A-to-I) mRNA editing acts as a key post-transcriptional regulatory mechanism during the sexual stage in F. graminearum and related filamentous ascomycetes^33–35. This epitranscriptomic modification is catalyzed by the Tad2-Tad3-Ame1 ternary complex, which is conserved within Sordariomycetes^36,37. While Tad2-Tad3 is responsible for canonical tRNA editing at the wobble position (A34), the fungal-specific factor Ame1—expressed only during sexual development—enables the complex to target mRNAs via its interaction with Tad3³⁶. Although the editing machinery differs, A-to-I mRNA editing is also a widespread phenomenon in animals and bacteria^37–40.

Functionally, inosine is interpreted as guanosine by the cellular translation machinery, so A-to-I editing can alter codons and recode proteins, similar to A-to-G substitutions at the RNA level. Genome-wide studies have identified tens of thousands of A-to-I editing sites in the perithecia of F. graminearum and Neurospora crassa, mainly within coding regions^41,42. While the crucial roles of Ame1 and some specific A-to-I editing events in fungal sexual development have been established^36,40,43,44, the functions of most editing sites remain unclear. This raises key questions about why RNA editing is maintained as a regulatory mechanism instead of fixing adaptive changes in DNA, and what advantages temporary RNA modifications offer over permanent genomic mutations^45,46.

In this work, we dissect key HR components in F. graminearum and uncover an epitranscriptomic mechanism that resolves a trade-off between heat-stress growth adaptation and sexual reproduction. We show that FgRad51-dependent HR is dispensable for asexual development under standard conditions but essential for sexual reproduction. Among downstream factors, FgMus81 emerges as a pivotal node linking development and environment: loss of FgMUS81 compromises ascospore formation and causes pronounced heat-stress defects in growth and pigmentation. Strikingly, these roles are uncoupled from the canonical Mus81 partner FgMms4 and from nuclease catalytic activity, indicating a noncanonical mode of action. We further identify a sexually induced, site-specific A-to-I editing event in FgMUS81 (N420D) that is required for normal sexual development. Mechanistically, the post-editing isoform exhibits reduced protein stability, attenuates a FgMus81-dependent mitotic HR readout, and supports normal meiotic progression. In contrast, forcing expression of the pre-editing isoform elevates FgMus81 output and disrupts meiotic nuclear division in a dosage-sensitive manner. A fixed genomic A-to-G substitution that mimics the edited codon phenocopies the meiotic benefits but incurs a cost to heat-stress survival, highlighting an antagonism in optimal FgMus81 dosage across different contexts and illustrating the adaptive advantage of flexible RNA editing over irreversible mutation. Comparative analyses reveal both conservation and dynamic retention of the N420D editing site across Sordariomycetes, consistent with evolutionary selection for stage-specific attenuation of Mus81 function. Together, our results establish N420D RNA editing as a single-nucleotide rheostat that fine-tunes Mus81 dosage to reconcile the competing demands of meiosis and thermal adaptation. This highlights RNA editing as an adaptive mechanism for life-cycle stage–specific optimization of genome maintenance and identifies it as a potential regulatory vulnerability for disease control.

Results

FgRad51 HR is required for sexual but not asexual development

To define the contribution of HR to development in F. graminearum, we identified homologs of core and downstream HR factors, including the recombinase gene FgRAD51 (FG1G17440) and representative meiotic crossover-associated genes FgMLH1 (FG4G25380), FgMLH3 (FG1G24760), FgMUS81 (FG1G17120), FgMMS4 (FG1G25820), FgSLX1 (FG3G17370), and FgYEN1 (FG2G38190) (Fig. 1a and Supplementary Table 1). Targeted deletion mutants were generated using the split-marker strategy⁴⁷. For each gene, at least two independent transformants with comparable phenotypes were obtained (Supplementary Table 2), and a representative strain is presented.

Fig. 1. Biological roles of homologous recombination pathway genes in F. graminearum.

a Schematic of the homologous recombination (HR) pathway showing the core recombinase gene and representative downstream meiotic crossover-associated genes. Genes highlighted in red were selected for knockout in this study. Abbreviations: DSB double-strand break; dHJ double Holliday junction; CO, crossover; NCO, non-crossover. b Colony morphology, perithecia production, and ascospore formation of wild-type PH-1 and indicated HR gene deletion mutants on PDA after 3 days and carrot agar plates at 8 days post-fertilization (dpf). Scale bars: 0.5 mm (white), 20 μm (black). c Growth inhibition rate (%) after treatment with methyl methanesulfonate (MMS, 0.05%), camptothecin (CPT, 50 nM), hydrogen peroxide (H₂O₂, 0.02%), or hydroxyurea (HU, 1 mM) for 3 days. d Percentage of asci with abnormal ascospore numbers produced by indicated strains at 8 dpf (30 perithecia per replicate). e Colony growth of indicated strains on PDA plates at 25 °C or under heat stress (30 °C) after 2.5 days, with corresponding growth inhibition rates shown. f Colony growth of indicated strains on PDA plates at 25 °C or under heat stress (30 °C) after 2.5 days, and ascospore production on carrot agar plates (CA) at 8 dpf. Scale bar: 20 μm. Bar graphs show relative expression of FgMUS81 in 7-dpf perithecia, growth inhibition rate (%), and percentage of asci with abnormal ascospore numbers at 8 dpf (30 perithecia per replicate) in the indicated strains. For all bar graphs, data are presented as mean ± SD from three biological replicates (n = 3). Different letters denote statistically significant differences (one-way ANOVA with Tukey’s multiple comparisons test, P < 0.05). Source data are provided as a Source Data file.

Across all mutants, vegetative colony morphology and radial growth on potato dextrose agar (PDA), as well as conidiation in carboxymethyl cellulose (CMC) medium, were indistinguishable from the wild type (Fig. 1b and Supplementary Fig. 1a), indicating that these HR factors are not required for asexual development under the conditions tested. By contrast, at 8 days post-fertilization (dpf), Fgrad51 exhibited a severe defect in perithecium formation. Moreover, the few perithecia that developed contained only a small number of asci, most of which were empty and lacked ascospores (Fig. 1b). Together, these data indicate that FgRad51-dependent HR is largely dispensable for vegetative growth and asexual reproduction but is essential for sexual reproduction in F. graminearum.

FgRad51, not crossover factors, confers DNA stress resistance

To test whether HR contributes to genome maintenance during vegetative growth, we challenged the indicated mutants with multiple genotoxic agents, including methyl methanesulfonate (MMS), camptothecin (CPT), hydrogen peroxide (H₂O₂), and hydroxyurea (HU). MMS induces DNA alkylation, CPT stabilizes topoisomerase I–DNA cleavage complexes and generates replication-dependent DSBs, H₂O₂ triggers oxidative damage and strand breaks, and HU causes replication stress by depleting dNTP pools⁴⁸.

Fgrad51 was hypersensitive to all four agents (Fig. 1c), consistent with a central role of FgRAD51 in damage tolerance. In contrast, mutants lacking crossover-associated downstream factors (Fgmus81, Fgmms4, Fgslx1, Fgyen1, Fgmlh1, and Fgmlh3) showed no detectable changes in sensitivity under any tested condition (Fig. 1c). These results indicate that FgRad51-dependent HR is a major determinant of resistance to diverse DNA-damaging and replication-stress conditions during vegetative growth in F. graminearum, whereas the tested crossover-associated factors are dispensable in these assays. Alternatively, the absence of an observable sensitivity phenotype could reflect functional redundancy among structure-selective endonucleases and/or crossover-associated pathways, whereby loss of any single component is compensated by others.

FgMus81 is required for sexual development and heat tolerance

Among the crossover-associated factors examined, only Fgmus81 displayed an overt defect in sexual development (Fig. 1b and Supplementary Fig. 1a). Specifically, Fgmus81 showed reduced ascospore production, with 15.6% of asci containing fewer ascospores than the wild type (Fig. 1d). Notably, although the MutLγ pathway (Mlh1–Mlh3) is a principal meiotic crossover mechanism in many eukaryotes^9,12, Fgmlh1 and Fgmlh3 did not exhibit discernible sexual defects in F. graminearum, supporting a species-specific requirement for FgMUS81 during sexual development.

Despite wild-type sensitivity to MMS, CPT, H₂O₂, and HU (Fig. 1c), Fgmus81 was markedly hypersensitive to elevated temperature. At 30°C, Fgmus81 showed reduced vegetative growth and diminished pigment accumulation relative to the wild type (Fig. 1e), a phenotype not observed in other crossover-associated mutants (Supplementary Fig. 1b). Fgrad51 also exhibited heat-associated growth defects, but did not show the pigmentation defect (Fig. 1e). Collectively, these data identify FgMUS81 as an important determinant of cellular fitness under heat stress. The pigmentation defect specific to Fgmus81 is consistent with the possibility that FgMus81 may prevent the build-up of downstream toxic recombination/repair intermediates under heat stress, thereby indirectly supporting pigment production.

FgMus81 acts independently of FgMms4 and nuclease activity

Unexpectedly, deletion of FgMMS4, which encodes the conserved regulatory partner of Mus81, did not recapitulate the sexual-development or heat-stress phenotypes of Fgmus81 (Fig. 1b, e and Supplementary Fig. 1), despite a physical interaction between FgMms4 and FgMus81 (Supplementary Fig. 2). These results suggest that the contributions of FgMus81 to sexual development and heat-stress tolerance are largely independent of FgMms4 in F. graminearum.

To determine whether FgMus81 nuclease activity is required for these functions, we generated two putative catalytic-dead alleles: FgMUS81^2A (D362A, D363A), targeting residues conserved in Saccharomyces cerevisiae⁴⁹, and FgMUS81^5A (D294A, E297A, D327A, D362A, D363A), additionally targeting residues implicated in human Mus81 catalysis¹⁶. Quantitative reverse-transcription (RT)-PCR indicated that transcript levels of both alleles were comparable to the wild type (Fig. 1f), arguing against altered mRNA stability. Notably, both alleles displayed wild-type growth and pigmentation at 30°C and normal ascospore formation (Fig. 1f), indicating that FgMus81-dependent heat-stress tolerance and sexual development do not require canonical nuclease activity. This nuclease-independent function provides a parsimonious explanation for the absence of heat-stress and sexual development phenotypes in Fgmms4 and uncouples the roles of FgMus81 in heat-stress adaptation and sexual reproduction from its DNA cleavage activity.

Site-specific N420D editing enables FgMus81 sexual roles

Notably, with the exception of FgRAD51, all crossover-associated factor genes examined in F. graminearum harbored nonsynonymous A-to-I mRNA editing sites that resulted in amino acid substitutions (Supplementary Fig. 3). Among these genes, only FgMUS81 contained nonsynonymous sites with editing levels exceeding 50%. Specifically, FgMUS81 carried three highly edited sites (each > 50%) (Fig. 2a).

Fig. 2. N420D RNA editing essential for FgMUS81 function in sexual development.

a Schematic of FgMUS81 protein domains and three highly edited sites (> 50% editing level). For each site, the pre-editing residue, editing position, and post-editing residue are indicated. The editing codon and neighboring codons of wild-type (WT), uneditable (UE), and post-editing (PE) alleles are shown. Edited and mutated bases are highlighted in red and green, respectively. b Representative sequencing traces of cDNA from 7-dpf perithecia of PH-1 (WT) and uneditable (UE) mutants, showing regions flanking the three RNA editing sites in FgMUS81. Red dashed lines and black triangles indicate editing and mutated sites, respectively. c Ascospore production in PH-1 and indicated mutants on carrot agar plates at 8 dpf. Superscripts UE and PE denote multi-site uneditable and post-editing mutants, respectively. Letters preceding numbers indicate single-site uneditable mutants; numbers preceding letters indicate single-site post-editing mutants. Red triangles mark asci with fewer than eight ascospores. Scale bar: 20 μm. Bar graph shows the percentage of asci with abnormal ascospore numbers at 8 dpf (mean ± SD, n = 3; 30 perithecia per replicate). Different letters denote statistically significant differences (one-way ANOVA with Tukey’s multiple comparisons test, P < 0.05). Source data are provided as a Source Data file.

To examine the functional importance of these sites, we generated a multi-site uneditable mutant (FgMUS81^UE) by simultaneously mutating the −1 thymine (T) base required for editing specificity^41,44 to guanine (G) at all three highly edited sites (Fig. 2a). Because these T-to-G substitutions were synonymous, they were expected to abolish editing without altering the encoded amino acid sequence. Complete loss of A-to-I editing in FgMUS81^UE was confirmed by sequencing RT–PCR products (Fig. 2b). Notably, FgMUS81^UE exhibited ascospore development defects comparable to those of Fgmus81 (Fig. 2c), indicating that RNA editing is required for FgMUS81 function during sexual development.

To further test whether the post-editing isoform is sufficient for normal sexual reproduction, we generated a post-editing mutant (FgMUS81^PE) in which the editing adenosines were replaced with guanosines in the corresponding genomic positions, thereby constitutively encoding the edited amino acids (Fig. 2a). FgMUS81^PE displayed normal sexual development (Fig. 2c), supporting the conclusion that the post-editing isoform—rather than the pre-editing isoform—is required for normal sexual reproduction.

To pinpoint the functionally critical editing event(s), we next constructed single-site uneditable mutants for each of the three highly edited sites. While the other two single-site uneditable mutants exhibited wild-type phenotypes, unediting at position 420 (FgMUS81^N420)—corresponding to the Asn-to-Asp substitution (N420D)—caused ascospore formation defects similar to those observed in FgMUS81^UE (Fig. 2c). However, the single-site post-editing mutant FgMUS81^420D displayed normal sexual development. These results indicate that site-specific N420D RNA editing is essential for FgMUS81 function in sexual development.

Edited FgMus81 supports meiotic but not mitotic recombination

To delineate the respective contributions of the pre-editing and post-editing FgMus81 isoforms to recombination, we first quantified mitotic HR using transformation-based gene-targeting assays, in which plasmids carrying homology to a chromosomal locus integrate via HR. We transformed the indicated strains with pKNTG carrying the FG1G36140 fragment and scored HR by PCR-confirmed targeted integration. The uneditable mutant (FgMUS81^N420) exhibited an HR rate comparable to that of the wild type, whereas the post-editing mutant (FgMUS81^420D) showed a significantly reduced HR rate, similar to Fgmus81 (Fig. 3a). Consistently, an independent assay at the FgPKS12 locus—using the frequency of white colonies as a readout of HR-mediated disruption—revealed a significant reduction in HR in FgMUS81^420D and Fgmus81 but not in FgMUS81^N420, relative to the wild type (Fig. 3a). Together, these results indicate that the post-editing FgMus81 isoform is less proficient in promoting mitotic HR during vegetative growth than the pre-editing isoform.

Fig. 3. Function of pre-editing and post-editing FgMus81 isoforms in mitotic and meiotic recombination.

a Function of FgMus81 isoforms in mitotic recombination. Schematic illustrates the transformation-based assay workflow. Blue circle: pKNTG plasmid carrying FG1G36140 gene (purple) and G418 resistance cassette (magenta). Orange circle: pUC19 plasmid containing the 5’ region of the FgPKS12 coding sequence (light green). Dark green rectangle: endogenous FgPKS12 locus. Correct integrations were identified by PCR primers or colony morphology. A check mark indicates gene function or successful PCR amplification; a cross mark indicates loss of function or PCR failure. Bar graph shows integration rates (% of transformants with expected integration) in indicated strains (mean ± SD, n = 3). Different letters indicate significant differences (one-way ANOVA, Tukey’s multiple comparisons test, P < 0.05). b Function of FgMus81 isoforms in meiotic interhomolog recombination. The schematic shows assay design for measuring meiotic recombination in PH-1 × Gz3639 outcrosses. Orange and blue rectangles represent chromosomes from PH-1 and Gz3639, respectively. Positions of FgMUS81, zeocin resistance (zeo), and hygromycin resistance (hph) cassettes are indicated. Bar graph shows frequency of dual-resistant single-ascospore colonies (interhomolog recombination products) in mutants (mean ± SD, n = 6). Statistical significance assessed by two-sided t test (ns, not significant, P > 0.05). c Meiotic crossover patterns revealed by whole-genome sequencing of 10 single-ascospore isolates from outcrosses (chromosome 1 shown as representative). Blue and orange bars indicate chromosomal segments inherited from Gz3639 and PH-1, respectively. Bar graph shows the number of interhomolog crossover events per group (mean ± SD, n = 10). Statistical significance assessed by two-sided t test (ns, not significant, P > 0.05). Source data are provided as a Source Data file.

We next tested whether the two isoforms differentially affect meiotic interhomolog recombination. Post-editing and uneditable FgMUS81 mutants were generated in the F. graminearum Gz3639 background. To quantify interhomolog crossovers, selectable markers (hph and zeo) were inserted near opposite ends of chromosome 2 in the PH-1 and GZ3639 backgrounds, respectively (Fig. 3b). Among double-resistant single-ascospore progeny derived from crosses between these backgrounds, the apparent interhomolog crossover rate in FgMUS81^420D was indistinguishable from that in FgMUS81^N420 (Fig. 3b). Moreover, whole-genome sequencing of 10 randomly selected single-ascospore isolates (intraspecific F1 hybrids) enabled reconstruction of crossover maps using SNP markers, revealing no significant differences in either the number or genomic distribution of interhomolog crossovers among FgMUS81^420D, FgMUS81^N420, and the wild type (Fig. 3c and Supplementary Fig. 4). Collectively, these data suggest that although the post-editing FgMus81 isoform is compromised in mitotic HR, both isoforms are competent to support meiotic interhomolog crossover recombination in F. graminearum.

Finally, to assess whether this mitotic HR function requires the canonical Mus81 partner, we measured HR-mediated transformation efficiency at the FgPKS12 locus in Fgmms4. In contrast to Fgmus81 and FgMUS81^420D, Fgmms4 showed no detectable reduction in HR relative to the wild type (Supplementary Fig. 5). This genetic dissociation suggests that FgMus81 can promote mitotic HR events in F. graminearum independently of FgMms4.

N420D editing attenuates FgMus81 to ensure normal meiosis

In the wild type, most asci at 6 dpf completed meiosis followed by post-meiotic mitosis, yielding eight nuclei that were each encapsulated by a forespore membrane and ultimately producing eight morphologically well-defined ascospores per ascus. By contrast, Fgmus81 displayed pronounced nuclear division defects: the proportion of asci containing eight nuclei was reduced by 59%, accompanied by a marked increase in asci containing only one to four nuclei (Fig. 4a). These observations indicate that FgMUS81 is required for proper meiotic nuclear division and subsequent ascospore formation.

Fig. 4. Regulatory role of N420D editing of FgMus81 in heat-stress tolerance and meiosis.

a DAPI-stained asci at 6 dpf visualized by epifluorescence microscopy. DIC, Differential interference contrast microscopy; DAPI, 4’,6-diamidino-2-phenylindole. Scale bar: 10 μm. Bar graph shows the percentage of asci with different nuclear numbers at 6 dpf (mean ± SD, n = 3, 20 perithecia per replicate). Different letters indicate significant differences within each nuclear-number group (one-way ANOVA, Tukey’s multiple comparisons test, P < 0.05). b Colony morphology on PDA after 2.5 days, with or without heat stress (30 °C), and ascospore production on carrot agar plates (CA, 8 dpf) in PH-1 (WT) and indicated mutants. Red triangles indicate asci with fewer than eight ascospores. Scale bar: 20 μm. Bar graphs show growth inhibition rate (%) under heat stress (mean ± SD, n = 3) and the percentage of asci with abnormal ascospore numbers at 8 dpf (mean ± SD, n = 3, 30 perithecia per replicate). Different letters indicate significant differences (one-way ANOVA, Tukey’s multiple comparisons test, P < 0.05). c Relative expression of FgMUS81 in 7-dpf perithecia of the indicated strains. Data represent mean ± SD from three biological replicates (n = 3). d Western blots of total proteins extracted immediately from hyphae of each strain after 0, 5, and 10 hours of cycloheximide (CHX) treatment (in vivo), and of extracted total proteins incubated at 30 °C for 0, 4, and 8 hours (in vitro). Proteins were probed with anti-Flag and anti-GAPDH antibodies, with GAPDH as a loading control. The bar graph shows relative FgMus81 protein levels (anti-FLAG/anti-GAPDH). Source data are provided as a Source Data file.

Consistent with this requirement, the uneditable mutant (FgMUS81^N420) exhibited nuclei division defects in asci resembling those of Fgmus81, albeit less severe, whereas the post-editing mutant (FgMUS81^420D) was indistinguishable from the wild type (Fig. 4a). Given that FgMUS81^N420 promoted higher mitotic HR than FgMUS81^420D during vegetative growth (Fig. 4a), we hypothesized that although FgMUS81 is essential for sexual development, excessive activity/dosage of the pre-editing isoform can be detrimental to meiotic progression, thereby contributing to the defects observed in FgMUS81^N420.

To test this idea, we introduced an additional copy of FgMUS81^N420 into the FG1G36140 locus in the FgMUS81^N420 background, generating a two-copy strain (d-FgMUS81^N420). d-FgMUS81^N420 exhibited more severe ascospore formation defects than the original FgMUS81^N420 mutant (Fig. 4b, c). Similarly, introducing an extra copy of FgMUS81^420D into the FgMUS81^420D background (d-FgMUS81^420D) also impaired ascospore formation, but the defect was weaker than that observed in d-FgMUS81^N420 (Fig. 4b, c). Together, these data support a model in which excessive FgMus81 dosage compromises ascospore development, and N420D editing helps restrain FgMus81 function during meiosis to promote proper nuclear division and ascospore formation.

N420D editing reduces FgMus81 protein stability

Quantitative RT-PCR analysis showed that FgMUS81 transcript levels were comparable between FgMUS81^N420 and FgMUS81^420D (Fig. 4c), suggesting that N420D editing does not measurably alter FgMUS81 mRNA abundance.

To evaluate protein stability, we generated strains expressing Flag-tagged pre-editing or post-editing proteins (FgMus81^N420-Flag and FgMus81^420D-Flag). Although the corresponding mRNA levels were similar (Supplementary Fig. 6a), cycloheximide (CHX) chase assays during vegetative growth revealed a weaker FgMus81^420D-Flag signal than FgMus81^N420-Flag after 5 h of CHX treatment (Fig. 4d), consistent with faster turnover of the post-editing isoform in vivo. After multiple rigorous attempts, we were unable to extract sufficient FgMus81 protein from sexual tissues for reliable Western blot analysis (Supplementary Fig. 6b). However, purified FgMus81^420D-Flag protein showed more extensive degradation than purified FgMus81^N420-Flag in an in vitro stability assay (Fig. 4d). These results were reproducible across independent experiments (Supplementary Fig. 6c). Since this instability is observed in both vegetative cells and cell-free systems, it suggests that the increased degradation is an intrinsic property of the edited protein rather than a stage-specific regulatory event. Collectively, these data indicate that N420D editing decreases FgMus81 protein stability and thereby reduces FgMus81 dosage.

N420D editing resolves FgMus81 trade-offs in heat tolerance and meiosis

These results raise an important question: why is the edited residue (D) not constitutively encoded in the genome, but instead generated via stage-specific RNA editing? Given that FgMUS81 contributes to heat-stress tolerance, we compared heat-stress phenotypes among FgMUS81^420D, the wild type, and FgMUS81^N420. At 30 °C, FgMUS81^420D exhibited reduced radial growth and diminished pigmentation, closely resembling Fgmus81, whereas FgMUS81^N420 displayed wild-type-like growth (Fig. 4b). These data suggest that the post-editing isoform alone is insufficient to support normal vegetative performance under heat stress, consistent with its reduced protein stability and its impaired proficiency in mitotic HR assays (Figs. 3a, 4d).

Interestingly, both two-copy strains (d-FgMUS81^N420 and d-FgMUS81^420D) displayed wild-type-like growth under heat stress (Fig. 4b). This observation suggests that increasing FgMUS81 dosage can compensate for the reduced stability of FgMus81^420D with respect to heat-stress adaptation, and that elevated FgMus81 dosage per se does not compromise heat tolerance. Together with the sexual-development phenotypes, these results support a trade-off model in which lower FgMus81 dosage/function favors normal sexual development, whereas higher dosage/function is beneficial for growth adaptation under heat stress.

Consistent with our earlier observations that Fgmus81 did not alter sensitivity to MMS, CPT, H₂O₂, or HU during vegetative growth (Fig. 1c), neither FgMUS81^N420 nor FgMUS81^420D displayed detectable differences from the wild type in these genotoxin assays (Supplementary Fig. 7). Thus, the mitotic function of FgMus81 revealed by HR-based transformation appears to be particularly relevant to heat-stress adaptation rather than to the tested DNA damage responses. Notably, the reduced mitotic HR observed in Fgmus81 and FgMUS81^420D, but not in Fgmms4 (Fig. 3a and Supplementary Fig. 5), further supports a specific contribution of FgMus81 to this vegetative recombination-associated phenotype that is at least partially separable from FgMms4.

Evolutionary dynamics of N420D editing in Sordariomycetes

To examine whether N420D editing extends beyond F. graminearum, we analyzed publicly available RNA-seq datasets from the sexual stages of 10 Sordariomycetes species that retain an editable adenosine at the corresponding site. A-to-I editing at this position, with variable editing levels, was detected in three additional Fusarium species and in two more distantly related taxa (Chaetomium globosum and Colletotrichum cliviicola) (Fig. 5a and Supplementary Fig. 8). In contrast, no evidence of editing at this site was detected in Neurospora, Pyricularia oryzae, or Sordaria macrospora, consistent with the absence of the preferred local sequence context and/or predicted hairpin-like RNA structures surrounding this site in these species (Fig. 5a). Notably, the N420D site was not predicted to reside in a hairpin-like structure in Fusarium neocosmosporiellum or C. cliviicola, suggesting that such structures may be transient, condition-dependent, or not reliably captured by in silico prediction. By comparison, editing at two other sites without known functional importance (K90E and K221E) is less conserved and detected only within Fusarium (Supplementary Fig. 9). Together, these analyses suggest that N420D editing is conserved in a subset of Sordariomycetes but remains evolutionarily flexible.

Fig. 5. Evolutionary dynamics of the N420D editing site in fungal MUS81.

a A-to-I editing events at the N420D site in MUS81 orthologs of 10 Sordariomycetes species. Sequence context (-2 to +4 positions), editing levels, and predicted RNA secondary structures (-30 to +30 nt) are shown. Red triangles mark the editing site. b–d Amino acid states, sequence logos, and predicted RNA secondary structure percentages at the N420D editing site in MUS81 orthologs across Ascomycota (b), Sordariomycetes (c), and Fusarium (d) taxa. Pie charts show proportions of species with no identified MUS81 orthologs (NA), pre-editing residue (N), post-editing residue (D), or other amino acid states (others) at the N420D editing site. The number of species examined per group is indicated in brackets. Red triangles mark the editing site.

To further investigate the evolutionary history of this site, we identified MUS81 orthologs from 581 representative ascomycete species, including 343 Sordariomycetes and 154 Fusarium species (Supplementary Data 1). MUS81 orthologs were broadly conserved across most ascomycete lineages but were not detected in Dothideomycetes, consistent with a lineage-specific loss of MUS81 in this class (Fig. 5b). Parsimony analysis suggests that the pre-editing residue N represents the ancestral state across ascomycetes, as it predominates at this position in most examined lineages except Saccharomycotina (Fig. 5b). The genomically encoded D state was also observed in multiple lineages, with higher frequencies in Leotiomycetes, Lecanoromycetes, and Pezizomycetes. In contrast, both the N residue and the corresponding editable adenosine are relatively conserved in Sordariomycetes (Fig. 5b), a clade in which A-to-I RNA editing is prevalent.

Although core sequence preferences for A-to-I editing are broadly conserved among filamentous ascomycetes, predicted hairpin-like RNA structures encompassing the N420D site were enriched in Sordariomycetes (Fig. 5b), suggesting that acquisition of local RNA secondary structure may represent a derived feature that facilitates editing in this clade. Among Sordariomycetes lineages, the fraction of species with predicted hairpin-like structures varied substantially (Fig. 5c), consistent with either multiple independent gains or an ancestral origin followed by repeated loss. Within Fusarium, most species retained predicted hairpin-like structures despite apparent losses in some lineages (Fig. 5d), suggesting overall conservation of the editing potential during Fusarium evolution. Moreover, although cytosine (C) is inferred to be the ancestral −1 base in Fusarium, independent substitutions to uracil (U)—the preferred − 1 base for efficient editing—occurred in several lineages, including the Sambucinum species complex (which includes F. graminearum). Conversely, in the Incarnatum–equiseti species complex, this position was replaced by G, a strongly disfavored − 1 base for A-to-I editing^41,44. These observations suggest that N420D editing is likely ancestral within Fusarium but has been independently optimized or attenuated across lineages.

Collectively, the recurrent gain and loss of local RNA structural features—and thus the capacity for editing—highlights an evolutionary landscape in which some lineages favor flexible, editing-mediated regulation, whereas others adopt fixed genomic encoding, potentially reflecting differences in ecology and life history.

Discussion

Our genetic dissection delineates an HR framework in F. graminearum and reveals an unexpected epitranscriptomic mechanism that tunes a structure-selective nuclease to reconcile developmental and environmental demands. We show that FgRAD51 is largely dispensable for vegetative growth and asexual development under standard laboratory conditions, yet it is essential for sexual reproduction. By contrast, several downstream crossover-associated factors tested here—Fgmlh1, Fgmlh3, Fgmms4, Fgslx1, and Fgyen1—do not individually produce overt vegetative genotoxin sensitivity phenotypes and are also individually dispensable for sexual development, whereas Fgmus81 displays clear defects in ascospore formation. Together, these results establish a functional hierarchy in which FgRad51-dependent HR is critical for sexual reproduction and for tolerating diverse genotoxic and replication-stress conditions, while the contribution of canonical crossover-associated branch factors is more context dependent under the conditions assayed here.

Loss of FgMUS81 disrupts meiotic nuclear division and ascospore formation, as evidenced by the marked reduction in asci containing eight nuclei and the concomitant increase in asci with fewer ascospores (Fig. 6a). Although this phenotype is compatible with an important role for FgMus81 in processing recombination intermediates, it does not by itself establish that the majority of meiotic crossovers are generated through a Mus81-dependent Class II pathway. Indeed, our data argue against a simple reassignment of crossover control to either the MutLγ pathway (FgMlh1–FgMlh3) or Mus81 alone, because Fgmlh1 and Fgmlh3 are individually dispensable for sexual development, and, importantly, editing status does not measurably alter the interhomolog crossover landscape in our mapping assays: the pre-editing allele FgMUS81^N420 supports apparently normal interhomolog crossover number and distribution. A parsimonious interpretation is that FgMus81 contributes to safeguarding meiotic chromosome segregation by preserving the proper timing and integrity of joint molecule resolution, thereby preventing the accumulation of toxic or aberrant joint molecules that impede nuclear division, rather than functioning solely as a dedicated crossover-generating nuclease. Similar damage-limiting roles of Mus81 family nucleases in resolving atypical recombination or replication intermediates have been described in other systems⁵⁰.

Fig. 6. Models for FgMus81 function and adaptive N420D RNA editing.

a Deployment of FgMus81 function in F. graminearum. FgMus81 supports sexual development and heat-stress tolerance independently of FgMms4 and canonical nuclease activity, yet is dispensable for vegetative growth under the generic genotoxic stresses tested. FgMus81 output is antagonistically optimized across life stages: higher FgMus81 dosage/function promotes vegetative growth adaptation at elevated temperature but is detrimental during meiosis, whereas restrained FgMus81 dosage/function is beneficial or required for normal meiotic progression and ascospore formation but compromises heat-stress adaptation. This antagonism reflects a dosage/function-dependent life-cycle trade-off centered on FgMus81. Defective asci with abnormal nuclear divisions and impaired ascospore formation are shown in blue. b Resolution of the dosage/function trade-off by sexual stage–restricted A-to-I (N420D) editing of FgMUS81 mRNA. During meiosis, when recombination intermediates require the timely and accurate regulation, both deficiency and excess of FgMus81 dosage can disrupt their processing, resulting in defective meiotic nuclear division. Under heat-associated replication stress, increased replication-associated DNA intermediates instead create a higher demand for FgMus81 output. The post-editing isoform FgMus81^420D is less stable than the pre-editing isoform FgMus81^N420, thereby lowering protein dosage. An uneditable A allele produces only FgMus81^N420, supporting heat-stress tolerance but causing meiotic defects, whereas a genomically encoded G allele produces only FgMus81^420D, supporting normal meiosis and ascospore development but compromising heat-stress tolerance. Thus, sexual stage–restricted N420D editing acts as a post-transcriptional brake that reduces FgMus81 dosage during meiosis while preserving the high-stability isoform for heat-stressed vegetative growth.

Beyond meiosis, our data further indicate that FgMus81 becomes particularly important under thermal stress, revealing a second context in which its genome-maintenance function is limiting. A striking feature of FgMus81 in F. graminearum is its selective contribution to cellular fitness under thermal stress. Fgmus81 confers pronounced heat-associated defects in radial growth and pigment accumulation, yet it does not broadly sensitize cells to MMS, CPT, H₂O₂, or HU in our assays (Fig. 6a). This separation suggests that the relevant lesions under elevated temperature differ from those induced by classical genotoxins. The specificity of meiotic crossover modulation by temperature, reported in several plants and animals^51–53, supports the idea that thermal stress generates a recombination landscape distinct from that produced by generic DNA damage. The pigmentation defect in Fgmus81 at 30°C may therefore be an indirect consequence of genome instability or perturbed developmental/metabolic regulation triggered by unresolved DNA intermediates, although the mechanistic linkage between recombination stress and pigment production remains to be determined. By analogy to the role of Mus81 in completing DNA synthesis at under-replicated loci in mammalian cells^17,18,54,55, it is conceivable that an FgMus81-dependent, replication-associated repair branch becomes particularly important under heat stress in F. graminearum. Direct tests of this idea will require assays that monitor replication dynamics and DNA intermediate accumulation under temperature challenge, as well as genetic epistasis with replication-stress response factors.

Perhaps the most unexpected result is that FgMus81’s roles in meiosis and heat-stress tolerance are genetically separable from those of its canonical partner FgMms4. In contrast to the established Mus81–Mms4/Eme1 heterodimer paradigm in yeast and mammals^15,16, deletion of FgMMS4 did not phenocopy loss of FgMUS81. We confirmed that the targeted FgMUS81 and FgMMS4 loci are bona fide orthologs of genes encoding the Mus81–Mms4/Eme1 heterodimer (Supplementary Figs. 10, 11). We further verified that the F. graminearum genome encodes only a single MMS4 homolog by exhaustive BLAST searches, and that the Fgmms4 strains used in this study are bona fide knockouts by PCR and whole-genome sequencing (Supplementary Fig. 12). Together, these controls make misannotation, redundancy, or incomplete gene disruption unlikely explanations for the phenotypic separation. Moreover, putative catalytic-dead FgMUS81 alleles retained wild-type-like sexual development and heat-stress phenotypes, arguing that the key functions of FgMus81 in these contexts are not mediated by canonical nuclease catalysis. Notably, transformation-based mitotic HR assays further dissociate FgMus81 from FgMms4: reduced mitotic HR was observed in Fgmus81 and FgMUS81^420D, but not in Fgmms4. Together, these findings suggest that FgMus81 can function through a noncanonical mechanism—potentially acting as a scaffold, a structural tether, or a competitive regulator independent of its nuclease activity—to regulate the timely and accurate processing of joint molecules, thereby supporting meiotic progression and adaptation to heat stress. Consistent with this possibility, recent proteomic evidence in yeast suggests the presence of Mms4-independent Mus81 associations during meiotic prophase⁵⁶, raising the broader possibility that noncanonical Mus81 configurations may extend beyond F. graminearum. Elucidating the precise noncanonical mechanism(s) underlying these Mms4- and nuclease-independent functions will be an important next step.

The most conceptually significant discovery is that site-specific A‑to‑I mRNA editing acts as an essential post-transcriptional rheostat for HR nucleases during meiosis. Our data support a model in which the primary role of the N420D editing event is to modulate FgMus81 dosage. The post-editing FgMus81^420D isoform is less stable than the pre-editing FgMus81^N420 form. The uneditable mutant, which exclusively produces the pre-editing FgMus81^N420 isoform, displayed normal mitotic HR rates but showed pronounced meiotic defects. Overexpression of the pre-editing isoform exacerbated these defects, supporting a model in which excessive FgMus81 dosage during meiosis is harmful. Together, these findings argue that excessive FgMus81 dosage/function is detrimental during meiosis, and that N420D editing provides a post-transcriptional brake that enables faithful nuclear division and ascospore development. This mechanism parallels the well-established need for tight regulation of nucleases by protein phosphorylation^23,24. This establishes A-to-I editing as a mechanistic layer in genome integrity control, extending the scope of the epitranscriptome to include dynamic tuning of protein dosage. Notably, the same dosage principle can explain the antagonistic requirements for FgMus81 across different life stages: during meiosis, where recombination intermediates are precisely regulated, excessive FgMus81 may promote aberrant processing of these intermediates, resulting in accumulation of toxic DNA species and disruption of meiotic nuclear division; whereas under heat-induced replication stress, the increased burden of replication-associated DNA intermediates renders FgMus81 dosage a limiting factor for their effective resolution. This framework naturally predicts stage- and environment-specific optima for FgMus81 function, which we explore below.

Our results also inform a long-standing evolutionary question: why maintain RNA editing rather than encode the seemingly advantageous residue in the genome? We uncover a stage- and environment-dependent antagonism. The high-stability, pre-editing FgMus81^N420 isoform benefits mitotic growth and performance under elevated temperature but is detrimental during meiosis. Conversely, constitutive genomic encoding of the edited residue (420D) supports normal sexual development but compromises heat-stress adaptation. This antagonism reflects a life-cycle trade-off: restrained FgMus81 dosage/function is beneficial (or required) for normal meiosis, whereas higher FgMus81 dosage/function supports growth adaptation under elevated temperature. RNA editing provides a flexible solution by enabling stage- and context-dependent tuning of FgMus81 dosage (Fig. 6b). Such conditional tuning exemplifies adaptive plasticity and may be particularly advantageous for a pathogen whose life cycle requires both environmental persistence and efficient ascospore production²⁷. Given that forcibly discharged ascospores constitute the primary inoculum for FHB epidemics, the marked reduction in ascospore production caused by editing deficiency is expected to reduce inoculum output. Although the epidemiological consequences remain to be quantified under field conditions, our findings nominate the editing event itself—and the cis-elements and trans-acting factors that enable it—as potential leverage points for strategies aimed at disrupting stage- or environment-specific fitness programs.

Consistent with the proposed constraint at this site, our comparative analyses suggest that the N420D editing event is conserved yet evolutionarily flexible across Sordariomycetes. Editing is detectable in several Fusarium species and in more distantly related taxa, whereas it is absent in others, consistent with variation in local sequence context and predicted RNA secondary structure. Broader phylogenetic surveys indicate that the ancestral state at this position is likely Asn (N), while genomically encoded Asp (D) appears repeatedly in multiple lineages, implying recurrent resolution of a shared functional constraint through distinct molecular routes (editing versus hard-coded substitution). The gain and loss of predicted hairpin-like structures in Sordariomycetes further supports dynamic evolution of the editing-permissive architecture, potentially reflecting differences in ecological pressures and life-history strategies.

In sum, our study identifies FgRad51-dependent HR as a key determinant of sexual reproduction and genotoxin tolerance in F. graminearum and reveals that site-specific A-to-I editing of FgMUS81 tunes FgMus81 protein stability to reconcile competing demands of meiosis and thermal adaptation. These findings expand the conceptual scope of RNA editing in fungal biology from transcript diversification to quantitative control of genome maintenance factors. Future work should focus on elucidating the precise noncanonical mechanism(s) by which FgMus81 acts independently of FgMms4 and canonical nuclease activity, defining how editing is selectively installed on meiotic transcripts, and determining how heat stress shapes a distinct recombination/replication landscape in which FgMus81 dosage becomes limiting.

Methods

Strain and culture conditions

We routinely cultured wild-type Fusarium graminearum strains PH-1 and Gz3639²⁵, along with their derived mutants and transformants, on potato dextrose agar (PDA) plates (20% potato, 2% dextrose, 1.5% agar) at 25 °C. We assessed colony growth rate and morphology on PDA plates after 3 days of incubation. To test strain sensitivity to various stresses, we supplemented PDA plates with one of the following agents: 0.05% methyl methanesulfonate (MMS), 50 nM camptothecin (CPT), 0.02% hydrogen peroxide (H₂O₂), or 1 mM hydroxyurea (HU). We then incubated the strains for 3 days. We harvested conidia from 5-day-old liquid cultures grown in carboxymethyl cellulose (CMC) medium (1.5% carboxymethylcellulose, 0.1% NH₄NO₃, 0.1% KH₂PO₄, 0.05% MgSO₄·7H₂O, and 0.1% yeast extract).

For sexual reproduction assays, we gently pressed down aerial mycelia from 7-day-old cultures grown on carrot agar plates (1.5% agar, 250 g/L carrot) using 1 mL of sterile 0.1% Tween-20. Subsequently, we incubated the plates under black light at 25 °C. We examined perithecium formation using an Olympus SZX16 stereoscope. The morphology of asci and ascospores was observed with an Olympus BX-51 microscope.

Gene deletion and site-directed mutagenesis

We performed gene deletions of DSB repair pathway genes using the split-marker approach⁴⁷. We amplified approximately 1.0 kb fragments upstream and downstream of each target gene. We then fused these fragments, via overlapping PCR, to the 5’- and 3’-termini of a recyclable marker module that conferred hygromycin resistance and sensitivity to the nucleoside analog 5-fluoro-2’-deoxyuridine⁴¹. Following transformation of PH-1 protoplasts, we selected transformants using hygromycin B (H005, MDbio, China) at a final concentration of 300 μg/ml.

For site-directed mutagenesis to generate specific uneditable and post-editing mutants, we constructed allelic fragments containing the desired mutations by overlapping PCR. Each allelic fragment included approximately 1.0 kb upstream and downstream homologous arms. We then transformed these fragments into the corresponding gene deletion mutant protoplasts. We screened transformants resistant to 25 μg/mL floxuridine (HY-B0097, MCE, USA) by PCR, and the desired mutations were confirmed by DNA sequencing. We used quantitative PCR to verify the absence of unintended integration of allelic fragments in the transformants. For each mutation, we identified at least two independent mutants exhibiting similar phenotypes. All generated transformants and primers used are listed in Supplementary Table 2 and Supplementary Data 2, respectively.

Reverse-transcription (RT)-PCR

We extracted total RNA from 12-hour hyphae or 7-day post-fertilization (dpf) perithecia of each strain. We performed cDNA synthesis using the RevertAid Master Mix (Thermo Scientific) according to the manufacturer’s instructions. We gel-purified RT-PCR products and directly sequenced them. We visualized Sanger sequencing traces using SnapGene Viewer 4.3 (https://www.snapgene.com/snapgene-viewer/)⁴⁴. We quantified relative expression levels by real-time RT-PCR using the 2^⁻ΔΔCT method⁵⁷, with the GzUBH gene serving as an internal control. The primers used are listed in Supplementary Data 2.

Overexpression of FgMUS81^N420 and FgMUS81^420D alleles

To achieve ectopic overexpression of the FgMUS81^N420 and FgMUS81^420D alleles at a defined genomic locus, we utilized the FG1G36140 locus, as its deletion does not alter the wild-type phenotype³⁶. First, in the FgMUS81^N420 and FgMUS81^420D mutant backgrounds, we replaced the FG1G36140 gene with a recyclable marker module. This generated recipient strains such as FgMUS81^N420ΔFG1G36140.

Next, approximately 1.0 kb fragments upstream and downstream of the FG1G36140 gene were amplified. These fragments were then fused, via overlapping PCR, to the FgMUS81^N420 or FgMUS81^420D allele (including its native promoter), creating the overexpression constructs. The resulting constructs were transformed into the protoplasts of the FgMUS81^N420ΔFG1G36140 or FgMUS81^420DΔFG1G36140 mutants, respectively. Transformants resistant to floxuridine were screened by PCR. RT-PCR was performed to confirm that the transformants overexpressed the FgMUS81 alleles in 7 dpf perithecia.

Single ascospore isolation

To facilitate outcrosses, we replaced the key MAT gene MAT1-2-1 in the Gz3639 strain with a GFP-G418 resistance marker, thereby generating a Gz3639 mat1-2-1 strain that exhibited GFP fluorescence. We also constructed uneditable and post-editing mutants of FgMUS81 in the Gz3639 mat1-2-1 background.

For outcrosses, we inoculated 7-day-old cultures of Gz3639 mat1-2-1, Gz3639 mat1-2-1 FgMUS81^N420, and Gz3639 mat1-2-1 FgMUS81^420D strains (grown on carrot agar plates) with 1 mL of conidial suspension (1 × 10⁶ conidia/mL). These conidial suspensions were derived from PH-1, FgMUS81^N420, and FgMUS81^420D strains in the PH-1 background, respectively.

Fifty perithecia were randomly collected from each outcross near a flame under a dissecting microscope. These were then crushed with fine forceps and transferred into 100 μL of sterile water in a centrifuge tube. The tube was immediately sealed and vortexed to obtain an ascospore suspension. We transferred 10 μL of this suspension onto a 2% agar plate. By adjusting the focal plane of the dissecting microscope to the medium surface, we located individual ascospores and transferred a single ascospore to the plate edge using a capillary tube. The ascospore’s position was marked on the bottom of the plate. Plates were sealed with Parafilm and incubated at 25 °C for 8 h. Subsequently, we excised a small piece of medium containing the germinating ascospore using a sterile needle and transferred it to a PDA plate, where single ascospore isolates were obtained after further incubation at 25 °C.

DNA-seq and whole-genome meiotic crossover profiling

For each outcross (wild-type, uneditable mutants, and post-editing mutants), we randomly selected ten single ascospore isolates for whole-genome DNA sequencing. The Gz3639 mat1-2-1 strain was also sequenced as a reference. DNA was extracted using the CTAB method. Sequencing libraries were prepared and subsequently sequenced on an Illumina NovaSeq 6000 system (2 × 150 bp paired-end) at Novogene Bioinformatics Institute (Tianjin, China). Low-quality reads and adapter sequences were removed using Trimmomatic⁵⁸ with default settings. The raw DNA-seq data have been deposited in the Genome Sequence Archive at the National Genomics Data Center, China National Center for Bioinformation, under the accession numbers provided in Supplementary Table 3.

We downloaded the latest genome assembly (YL1) of F. graminearum PH-1⁵⁹ from FgBase (http://fgbase.wheatscab.com/) to serve as the reference genome. DNA-seq reads from each strain were mapped to the PH-1 reference genome using Bowtie 2 v2.5.1⁶⁰ with default parameters. SAM files were then converted to BAM format and indexed using SAMtools v1.17⁶¹. To avoid biases in SNP calling caused by amplification artifacts, duplicated reads were removed using the MarkDuplicates tool from the Picard package v3.1.0 (https://broadinstitute.github.io/picard/). SNPs were identified using BCFtools v1.17⁶¹, and variant calls were filtered based on sequencing depth and quality scores. We utilized 76,117 high-quality SNP sites, identified between Gz3639 mat1-2-1 and PH-1, as genetic markers to detect crossover events in each single ascospore isolate using custom scripts (https://github.com/qinhuwang/FG.crossover). Crossover points were defined as the midpoint between two adjacent markers flanking a crossover event.

Genetic assays of meiotic recombination rates

To evaluate meiotic recombination rates, we inserted two selectable markers into intergenic regions at opposite ends of chromosome 2 in both PH-1 and Gz3639 background strains. Specifically, in the PH-1 background, a hygromycin resistance cassette (hph) was integrated into both FgMUS81^N420 and FgMUS81^420D strains at the 5’-terminus of chromosome 2 (69040-70803 bp). Concurrently, in the Gz3639 background, a zeocin resistance cassette (zeo) was inserted into both Gz3639 mat1-2-1 FgMUS81^N420 and Gz3639 mat1-2-1 FgMUS81^420D strains at the 3’-terminus of chromosome 2 (8938050-8939537 bp).

For outcrosses, Gz3639 mat1-2-1 FgMUS81^N420 zeo or Gz3639 mat1-2-1 FgMUS81^420D zeo strains were used as the female parent, and these were outcrossed with the male parent strains FgMUS81^N420 hph or FgMUS81^420D hph, respectively. We randomly collected ascospores from each outcross and plated 100 ascospores per plate on PDA medium supplemented with 300 μg/mL hygromycin B and 700 μg/mL zeocin (SL4140, Coolaber, China). Crossover recombination efficiency was quantified by counting the number of dual-resistant colonies.

Genetic assays of mitotic recombination rates

To evaluate mitotic recombination rates, we employed transformation-based assays. For the first assay, the FG1G36140 gene was cloned into the pKNTG plasmid⁶², and the resulting pKNTG-FG1G36140 construct was transformed into protoplasts of PH-1, Fgmus81, FgMUS81^N420, and FgMUS81^420D mutants. Transformants were selected on TOP plates (0.3% yeast extract, 0.3% casein acid hydrolysate, 1.5% agar, and 20% dextrose) supplemented with 400 μg/mL G418 (HY-17561, MCE). Plasmid integration was confirmed by PCR using primers targeting the G418 resistance cassette (G418-F and G418-R). To detect recombination events, a forward primer (up-F, approximately 1.0 kb upstream of FG1G36140) and a reverse primer (down-R, approximately 1.0 kb downstream) were designed. If the plasmid integrated into the FG1G36140 locus via homologous recombination, a PCR product was amplified using the up-F and G418-R primer pair. Conversely, if no integration occurred, no product was detected with this primer pair, but another product was detected using the up-F and down-R primer pair. As a control, wild-type PH-1 genomic DNA yielded no product with the up-F/G418-R pair but produced an expected band with the up-F/down-R pair, confirming the integrity of the locus in the absence of plasmid integration.

A similar strategy was applied to assess recombination at the FgPKS12 locus. This locus was chosen because its disruption leads to a readily observable phenotype: white colony⁶³. The 5’-part of the FgPKS12 coding region was cloned into the pUC19 vector to generate pUC19-FgPKS12, and this construct was transformed into protoplasts of PH-1, Fgmus81, Fgmms4, FgMUS81^N420, and FgMUS81^420D mutants. Integration of pUC19-FgPKS12 into the endogenous FgPKS12 locus via homologous recombination disrupted gene function. This disruption resulted in the formation of two tandemly repeated FgPKS12 copies, with one copy lacking the 3’-part of the coding region and the other lacking its promoter, leading to white colony transformants.

Yeast two-hybrid assays

To investigate protein-protein interactions between FgMus81 and FgMms4, yeast two-hybrid assays were performed. The open reading frames (ORFs) of FgMUS81 and FgMMS4 were amplified and subsequently cloned into the pGBKT7 (bait) and pGADT7 (prey) vectors (TaKaRa Bio, Tokyo, Japan), respectively. The resulting constructs were then co-transformed pairwise into the yeast strain AH109. Interactions were assessed by evaluating growth on SD-His-Leu-Trp plates, as previously described³⁶. Additionally, the activity of Mel1 α-galactosidase was evaluated on SD-His-Leu-Trp plates supplemented with X-α-gal.

Western blotting

To facilitate protein detection, FgMUS81^N420-3×FLAG and FgMUS81^420D-3×FLAG constructs were generated and transformed into PH-1 protoplasts. Transformants were selected on TOP plates supplemented with G418, and their integration was confirmed by PCR.

To assess protein stability in vivo, conidia from the generated transformants were harvested and inoculated into YEPD medium (0.3% yeast extract, 1% peptone, 2% dextrose), followed by 12 hours of shaking at 25 °C. Subsequently, 25 μg/mL cycloheximide (CHX; Merck, USA) was added. Hyphae were collected after 0, 5, and 10 h of incubation at 30 °C. Sexual tissues (perithecia) were collected at 7 days post-fertilization (7 dpf) from carrot agar plates. For in vitro protein stability assessment, conidia were inoculated in YEPD medium and shaken for 12 h. The resulting hyphae were harvested, total proteins extracted, and then incubated at 30 °C for 0, 4, and 8 h prior to Western blotting.

Total proteins were isolated and separated on 10% SDS-PAGE gels, then transferred to nitrocellulose membranes. To prevent non-specific antibody binding, the membranes were incubated in 5% skim milk for 1 h. Subsequently, proteins were incubated overnight at 4 °C with either anti-FLAG (dilution:1:5000, Merck, F3165, Germany) or anti-GAPDH (dilution:1:5000, Sangon Biotech, D110016-0200, China) antibodies. Following three washes with 1 × TBST (100 mM Tris-HCl, pH 7.5; 136 mM NaCl; 0.1% Tween-20), the membranes were incubated with corresponding secondary antibodies at room temperature for 1 hour. After additional washes with 1 × TBST, a chemiluminescent substrate was applied for 5 minutes, and signals were detected using a chemiluminescence imaging system.

Bioinformatic analysis

To identify genes encoding core and downstream homologous recombination factors in F. graminearum, we employed a robust reciprocal BLASTp search strategy using protein sequences from S. cerevisiae and S. pombe as reference queries. The A-to-I RNA editing sites in F. graminearum used in this study were obtained from our previous work³⁴ (see online Supplemental Material). To identify MUS81 orthologs and analyze the N420D editing site, genome sequences of 581 representative ascomycete species were downloaded from the NCBI Genome database. MUS81 orthologs, along with their corresponding amino acid states at the N420D editing site and flanking DNA sequence context, were identified using tBLASTn. Only hits with e-values < 1 × 10⁻⁵ were considered as homologous sequences.

For RNA editing analysis, RNA-seq data from the sexual stage of 10 Sordariomycetes species were obtained from the NCBI SRA database (details regarding these datasets are documented in Supplementary Table 3). RNA-seq reads were mapped to their respective genomes using HISAT2 v2.2.1⁶⁴, as previously described³⁶. Subsequently, N420D editing events were manually inspected using IGV.

The phylogenetic tree for the genus Fusarium was constructed based on our prior research⁴³. Evolutionary relationships among major Sordariomycetes and Ascomycota groups were inferred from the MycoCosm tree (https://mycocosm.jgi.doe.gov/mycocosm/home).

Sequence logos flanking the editing sites were visualized using ggseqlogo v0.1⁶⁵. RNA secondary structures, encompassing 30 nucleotides upstream and downstream of each editing site, were predicted using RNAFold⁶⁶. The distributions of secondary structure types were analyzed with forgi v2.0.3⁶⁷, and the predicted structures were visualized using RNAstructure v6.5⁶⁸.

Statistics & reproducibility

The investigators were not blinded to group allocation during data collection and analysis, as blinding was not feasible due to the nature of the experiments. However, objective and standardized methods were employed to minimize bias. No formal statistical method was used to predetermine sample sizes. Instead, sample sizes were selected based on standard practices in the field and prior experience to ensure sufficient biological replicates for reliable and reproducible results. All statistical analyses were performed in R. P-values were calculated using two-sided t tests with adjustments for multiple comparisons or one-way ANOVA followed by Tukey’s multiple comparisons test. To ensure reproducibility, each experiment was independently repeated at least twice, yielding consistent results. All blot images presented represent consistent findings from independent experiments. For phenotype observations, at least two independent mutants or transformants with three biological replicates each were used per experiment, although only representative images are shown.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

H.L. and M.W. conceived and designed the experiments; M.W., J.L., P.C., and M.L. performed the experiments; H.L., C.F., Q.W., C.J., and J.R.X. contributed advice and reagents; M.W., J.L., Q.W., and H.L. analyzed data; H.L. and M.W. wrote the paper; and H.L. provided supervision. M.W. and J.L. contributed equally to this paper.

Peer review

Peer review information

Nature Communications thanks Yuange Duan, Hokyoung Son and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Data supporting the findings of this work are available within the paper and its Supplementary Information file. DNA-sequencing data generated for this study and public RNA-sequencing data used in this study can be accessed from the Genome Sequence Archive using the accession numbers listed in Supplementary Table 3. Source data are provided in this paper.

Code availability

The custom codes used in this study for crossover event identification have been deposited in GitHub (https://github.com/qinhuwang/FG.crossover).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-71219-z.