A pair of effectors encoded on a conditionally dispensable chromosome of Fusarium oxysporum suppress host-specific immunity

Many plant pathogenic fungi contain conditionally dispensable (CD) chromosomes that are associated with virulence, but not growth in vitro. Virulence-associated CD chromosomes carry genes encoding effectors and/or host-specific toxin biosynthesis enzymes that may contribute to determining host specificity. Fusarium oxysporum causes devastating diseases of more than 100 plant species. Among a large number of host-specific forms, F. oxysporum f. sp. conglutinans (Focn) can infect Brassicaceae plants including Arabidopsis (Arabidopsis thaliana) and cabbage. Here we show that Focn has multiple CD chromosomes. We identified specific CD chromosomes that are required for virulence on Arabidopsis, cabbage, or both, and describe a pair of effectors encoded on one of the CD chromosomes that is required for suppression of Arabidopsis-specific phytoalexin-based immunity. The effector pair is highly conserved in F. oxysporum isolates capable of infecting Arabidopsis, but not of other plants. This study provides insight into how host specificity of F. oxysporum may be determined by a pair of effector genes on a transmissible CD chromosome.

P athogenic fungi often carry chromosomes that are not necessary for growth in the non-pathogenic state 1,2 . Analogous to the well-characterized virulence plasmids in bacteria, the number of these dispensable chromosomes in individual isolates can vary. In plant pathogenic fungi, dispensable chromosomes that are associated with virulence are generally referred to as supernumerary, 'B', or conditionally dispensable (CD) chromosomes 1 . When pathogenic fungi lack CD chromosomes, they can grow in vitro, but often exhibit attenuated or no virulence 1,2 . The functions of CD chromosomes in some plant pathogenic fungi are associated with suppression or deactivation of host-specific factors. In Fusarium solani, for example, a CD chromosome carries phytoalexin detoxifying genes 3 . In contrast, the CD chromosomes of Alternaria alternata and Cochliobolus carbonum harbor host-specific toxin genes 2,4 . Therefore, CD chromosomes can be crucial determinants of host specificity that are defined by phytotoxin activity or by defense against chemicals such as phytoalexins.
Fusarium oxysporum causes devastating diseases of more than 100 plant species, including economically important crops such as tomato, banana, and melon 5 . Individual isolates of F. oxysporum have different host ranges and are classified into formae speciales (ff. spp.) based on the susceptibility of plant species to infection. Although much is known about the genetics and pathology of F. oxysporum, the precise molecular mechanisms of host specificity remain unclear. So far, CD chromosomes have been identified in the tomato-infecting pathogen F. oxysporum f. sp. lycopersici (Fol) and in F. oxysporum f. sp. radicis-cucumerinum (Forc), a cucurbit-infecting pathogen [6][7][8] . The Fol isolate 4287 and the Forc isolate 016 each contain a single virulence-associated CD chromosome that is transferable to other isolates [7][8][9] . Horizontal transfer of the CD chromosomes from Fol4287 or Forc016 converts non-pathogenic F. oxysporum isolates into pathogens of their respective hosts 6,7,9 . Part of this phytopathogenic conversion is often due to the expression of CD-encoded effectors that modulate host immunity against infection, such as Secreted In Xylem (SIX) effectors that are, as their name indicates, secreted into xylem elements during infection 10,11 . A total of fourteen SIX genes (SIX1 to 14) have been identified from Fol 10 . The CD chromosome of Fol4287 contains all of the SIX genes except SIX4, which is not present in Fol4287 6,12 , but is present in certain other Fol isolates. The CD chromosome of Forc016 contains SIX6, SIX9, SIX11, and SIX13 homologs 7 . SIX1, SIX3, SIX5, and SIX6 from Fol are involved in overcoming resistance in tomato and the SIX6 homolog from Forc016 is crucial for virulence in cucumber 7,10 . However, their molecular mechanisms as virulence factors are as yet unknown.
Arabidopsis-infecting isolates of F. oxysporum are useful as a model pathosystem. There are at least three ff. spp. that cause disease on Arabidopsis: f. sp. conglutinans, f. sp. matthiolae, and f. sp. raphani 13 . F. oxysporum f. sp. conglutinans (Focn) can also infect other Brassicaceae plants such as cabbage (Brassica oleracea var. capitata). The SIX1 gene is required for full virulence on cabbage in Focn 14 , but the Focn factor(s) that are required for virulence on Arabidopsis have not been identified. We have previously shown that the Focn isolate Cong:1-1 (FocnCong:1-1) harbors SIX1, SIX4, SIX8, and SIX9 homologs on multiple chromosomes of different sizes 15 . Although these chromosomes are presumed to be conditionally dispensable in Focn, their status as CD chromosomes has not been established.
Here we report, through analyses of chromosome-deficient FocnCong:1-1 strains and through horizontal chromosome transfer, that FocnCong:1-1 has multiple CD chromosomes. Importantly, we identified individual CD chromosomes that are required for virulence on Arabidopsis, cabbage, or both. Furthermore, we identified a pair of effector genes on a CD chromosome that are required for suppression of Arabidopsis-specific phytoalexin-based immunity.

Results
Chromosome-level genome assembly of FocnCong:1-1. We assembled the FocnCong:1-1 genome sequence into 198 contigs with an N50 of 1.271 Mb. To improve contiguity, we further performed optical mapping using two restriction enzymes. The final assembly consisted of 22 scaffolds (SCs) with an N50 SC length of 4865 kb and a 99.1% complete BUSCO score (Table 1). For gene prediction, we generated transcriptome data from axenic culture and plant infections, resulting in a total of 21,781 genes, among which are eight presumptive effector genes (SIX1, SIX4, SIX8, SIX9, and FOA1-FOA4) that were previously known from Arabidopsis-infecting F. oxysporum 16 , as well as the homologous genes of FOA1 and FOA4, which were named FOA1b and FOA4b, respectively. We did not detect homologs of any other SIX genes. To find unknown effectors, 1467 putative secreted proteins were screened for proteins with an effector-like structure using the EffectorP v1 and/or v2 algorithm 17,18 . A total of 263 secreted proteins were predicted as effectors by both EffectorP v1 and v2. This prediction did not include FOA1, which is involved in the suppression of pattern-triggered immunity 16 , nor its homolog FOA1b. Therefore, a total of 265 proteins, including FOA1 and FOA1b, were defined as high-confidence effector candidates (Table 1 and Supplementary Data 1).
The F. oxysporum genome is composed of core genomic regions that are conserved among Fusarium species, and additional accessory genomic regions that are conserved in certain isolates 19 . Comparative analysis with the Fol4287 genome as a reference indicated that (i) the FocnCong:1-1 SCs have no homology with known accessory genomic regions in Fol4287 (chr01B; chr02B; chr03; chr06; chr14; chr15) 6 , (ii) similarly, there are genomic regions of FocnCong:1-1 that have no homology with Fol4287, and (iii) the non-homologous genomic regions are enriched in transposable elements (TEs) (Fig. 1a). All known effector genes except FOA4 are located in the TE-rich genomic region in FocnCong:1-1 as follows: SIX1 (in SC8), SIX4 (SC9), SIX8 (SC10), SIX9 (SC3), FOA1 (SC5), FOA1b (SC10), FOA2 (SC9), FOA3 (SC3), and FOA4b (SC10) (Supplementary Fig. 1). FOA4 (SC12) may be a pseudogene since it is not expressed either in vitro or in planta (Supplementary Data 1 and 2). TEs are suspected to be involved in the generation of genomic variations leading to environmental adaptation and, in the case of pathogens, they may have been involved in the acquisition of the ability to infect particular hosts 20 . Therefore, chromosomes containing TE-enriched genomic regions have a high potential to be CD chromosomes. Recently, a chromosome-level genome assembly of the Arabidopsis-infecting F. oxysporum isolate Fo5176 was reported 21 . The genomes of FocnCong:1-1 and Fo5176 are very similar, sharing from 93.2% to 94.3% of their total scaffold/contig lengths (>95% identity, 10 kb). Synteny analysis revealed that (i) SC16 and SC18 of FocnCong:1-1 correspond to chromosome 14 (chr14) of Fo5176, and (ii) SC10 and SC20 to chr16 (Fig. 1b), indicating that these SCs constitute, or contribute to the respective chromosomes. Due to the observations (i) and (ii) above, we refer to the chromosomes carrying these sequences as chr SC16/SC18 and chr SC10/SC20 in FocnCong:1-1, respectively.
Chr SC10/SC20 is involved in suppression of CYP79B2/CYP79B3mediated immunity. A CD chromosome from the Arabidopsisinfecting anthracnose fungus Colletotrichum higginsianum has been reported to be involved in suppression of plant immunity that is dependent on tryptophan (Trp)-derived secondary metabolites 24 . We investigated whether CD chromosomes of FocnCong:1-1 encode products that also suppress specific immunity. For this experiment, we used the Arabidopsis double mutant cyp79b2/cyp79b3 that lacks the ability to synthesize Trpderived secondary metabolites 25 . Among the chromosomedeficient FocnCong:1-1 mutants (HS2 to HS6) with attenuated virulence to Arabidopsis Col-0 WT, only FocnCong:1-1 HS5 (ΔSC9/chr SC10/SC20 ) showed the same level of virulence on cyp79b2/cyp79b3 plants as was observed for its parent strain ΔSIX4 ( Fig. 4a and Supplementary Fig. 10a, b). These results suggest that chr SC10/SC20 plays a key role in suppressing Trpderived secondary metabolite-dependent immunity. FocnCong:1-1 HS6 (ΔSC5/SC8/SC9/chr SC10/SC20 ) was substantially less virulent on cyp79b2/cyp79b3 plants. This is likely because SC5 or SC8 are involved in virulence other than through suppression of Trpbased immunity. In addition, we found that the cyp79b2/cyp79b3 double mutant was resistant to all tested SC3-deficient Focn-Cong:1-1 mutants (HS2 to HS4; Fig. 4a and Supplementary  Fig. 10a, b), possibly due to some deficiency of conidial formation in these mutants (Fig. 2c).
To investigate which step or steps of infection the CD chromosomes contribute to, a histological analysis was performed using GFP-labeled FocnCong:1-1 strains in Arabidopsis. Focn-Cong:1-1 ΔSIX4-GFP always colonized xylem vessels of roots, often reaching stem elements in Arabidopsis WT, whereas FocnCong:1-1 HS2-GFP lacking SC3 germinated on root surfaces but showed almost no colonization in xylem vessels of roots or stems (Fig. 4b), confirming its deficiency in growth in planta. FocnCong:1-1 HS5-GFP (ΔSC9/chr SC10/SC20 ) colonized root xylem vessels, but the frequency of stem colonization was low in Arabidopsis WT. In cyp79b2/cyp79b3 double mutants, however, FocnCong:1-1 HS5-GFP frequently colonized the stems as was observed for FocnCong:1-1 ΔSIX4 in WT (Fig. 4b). These results suggest that chr SC10/SC20 is implicated in the ability to colonize beyond root xylem vessels into stems, and conversely that CYP79B2/CYP79B3 participate in inhibition of FocnCong:1-1 colonization of stems.

A pair of effectors are involved in virulence on Arabidopsis.
Because chr SC10/SC20 is likely to encode effectors that contribute to suppression of Arabidopsis-specific immunity, we searched for genes encoding potential effectors, and found a total of twelve effector candidate genes located on chr SC10/SC20 (Supplementary Data 1). Expression profiling revealed that FocnCong_v001893 (SIX8) and FocnCong_v001894 were highly expressed during infection ( Fig. 5a and Supplementary Data 1). Interestingly, SIX8 is adjacent to FocnCong_v001894, with an intergenic distance of 1057 bp on SC10 (Fig. 5b). The intergenic region contains a miniature impala inverted repeat (mimp-IR) sequence, which is related to TE sequences ( Fig. 5b and Supplementary Fig. 11). A mimp-IR is also often located in the upstream regions of SIX and other effector candidate genes in Fol, Forc, and the meloninfecting pathogen F. oxysporum f. sp. melonis 7,12,29 . To determine whether SIX8 and FocnCong_v001894 are involved in virulence on Arabidopsis, a genome fragment containing the SIX8-FocnCong_v001894 locus was introduced into FocnCong:1-1  Fig. 14a). In contrast, Arabidopsis WT was resistant to the other FocnCong:1-1 HS5 transformants that contained only SIX8 or FocnCong_v001894 (Fig. 5c and Supplementary Fig. 14a). It should be noted that virulence of knockout mutants that lack the SIX8-FocnCong_v001894 locus in FocnCong:1-1 was significantly lower than for WT ( Fig. 5d and Supplementary Figs. 14-16), suggesting that both SIX8 and FocnCong_v001894 are necessary for virulence on Arabidopsis. We therefore designated FocnCong_v001894 as Pair with SIX Eight1 (PSE1).
Genetic and functional conservation of the SIX8 and PSE1 loci. Next, we investigated whether the SIX8-PSE1 pair is conserved in Arabidopsis-infecting F. oxysporum isolates. Comparative analysis of highly contiguous and available genome assemblies of F. oxysporum isolates (Supplementary Table 1) showed that the SIX8-PSE1 locus is completely conserved in Fo5176 and in the stock-infecting pathogen F. oxysporum f. sp. matthiolae (Fomt) PHW726, which can infect Arabidopsis 13,30 , but not in isolates that cannot infect Arabidopsis (Fig. 6a, b). For example, the banana-infecting pathogen F. oxysporum f. sp. cubense (Focb) tropical race 4 (TR4), which threatens banana production worldwide, has SIX8 but not PSE1. In the other non-Arabidopsisinfecting isolates, except Fol4287, neither SIX8 nor PSE1 is present. Fol4287 has multiple copies of SIX8 and its homolog SIX8b 12,31 but PSE1 is not present in the published Fol4287 gene annotation 6 . However, we found three loci similar to the SIX8-PSE1 locus in chromosomes 2, 3, and 14 of Fol4287. At these loci, adjacent to SIX8, there is a PSE1-like gene (PSL1) differing in the C-terminal 10 amino acids ( Fig. 6b and Supplementary Fig. 17). Furthermore, multiple SIX8b loci contain TEs inserted into adjacent PSE1 sequences. For example, a transposase gene was found in the first intron of the PSE1 homologs in two loci of chromosome 3 and another locus in chromosome 6 ( Fig. 6b and Supplementary Fig. 18). Similarly, a presumptive transposase was found immediately upstream of the potential-but-unannotated PSE1 homolog in another locus in chromosome 6 (Fig. 6b). Thus, TE insertion seems to have disrupted the PSE1 adjacent to SIX8b in Fol4287.

Discussion
Here we report the identification of a CD chromosome in F. oxysporum that is required for virulence on Arabidopsis. This CD chromosome encodes a pair of effectors (SIX8 and PSE1) that are involved in suppressing Arabidopsis-specific immunity, and are conserved in the other F. oxysporum isolates capable of infecting Arabidopsis. The mode of action potentially involves defense against, or suppression of, the phytoalexin camalexin. We also report that another CD chromosome is required for pathogenicity on cabbage. In addition, certain CD chromosomes are involved in conidial formation.
In plant pathogenic fungi, CD chromosomes associated with virulence are usually not involved in vegetative growth 1,2 . In this sense, SC3 and chr SC10/SC20 in FocnCong:1-1 are atypical CD chromosomes that affect conidial formation (Fig. 2c). Although the reduced virulence of SC3-deficient FocnCong:1-1 mutants  Supplementary Fig. 14a. d Disease index of Arabidopsis Col-0 WT challenged with FocnCong:1-1 WT, ΔSIX8-PSE1, an ectopic transformant (ect) or water (mock) at 28 dpi was scored as described in Methods. Results of three independent experiments were combined. n denotes the number of plants investigated. Asterisks represent significant difference from WT (*p < 0.05, Mann-Whitney U-test). Representative images of Arabidopsis at 28 dpi are shown in Supplementary Fig. 14b. (HS2, HS3, and HS4) on Arabidopsis and cabbage (Fig. 2c, d) may be due to deficiency in the ability to form conidia, or to regulatory step(s) that has multiple unexplored phenotypic effects, we cannot exclude the possibility that yet-unknown effectors located on SC3 are implicated in virulence. Interestingly, SC3 contains a region partly syntenic to chromosome 11, which is a core chromosome of Fol4287 (Fig. 1a). This syntenic region may contain dose-effective genes involved in conidial formation. In contrast to SC3, chr SC10/SC20 negatively regulates conidial formation but positively contributes to virulence on Arabidopsis but not on cabbage (Fig. 2c, d), possibly representing a trade-off between vegetative growth and virulence to a particular host.
FocnCong:1-1 carries multiple CD chromosomes that have distinct virulence functions against specific hosts. For example, the CD chromosome chr SC10/SC20 -deficient FocnCong:1-1 HS5 is less virulent on Arabidopsis, but is able to develop severe disease on cabbage (Fig. 2d and Supplementary Fig. 5). This result may be explained by the fact that FocnCong:1-1 HS5 maintains the CD chromosome SC8, which harbors a gene, SIX1, required for full virulence on cabbage 14 . Consistently, FocnCong:1-1 HS6, which lacks both SC8 and chr SC10/SC20 , lost pathogenicity on both cabbage and Arabidopsis ( Fig. 2d and Supplementary Fig. 5), and introduction of SC8 into HS6 restored virulence on cabbage ( Fig. 3c and Supplementary Fig. 9). Thus, we conclude that chr SC10/SC20 and SC8 are responsible for host-specific virulence on Arabidopsis and cabbage, respectively.
The target of the CD chromosome chr SC10/SC20 effector is likely to be CYP79B2/CYP79B3-mediated immunity in Arabidopsis, because the loss of chr SC10/SC20 attenuated virulence of FocnCong:1-1 HS5 to WT, but not to the cyp79b2/cyp79b3 double mutant ( Fig. 4a and Supplementary Fig. 10). CYP79B2/CYP79B3 had not previously been implicated in resistance to F. oxysporum.
For instance, Kidd et al. 32 reported that susceptibility of cyp79b2/ cyp79b3 to F. oxysporum Fo5176 was not different from WT. Consistent with this report, our study shows that virulence of FocnCong:1-1 on cyp79b2/cyp79b3 is comparable to WT (Fig. 4a and Supplementary Fig. 10). Thus, only the use of CD chromosome-deficient mutants allowed us to uncover the involvement of CYP79B2/CYP79B3 in resistance to F. oxysporum. Furthermore, histological analysis suggests that CYP79B2/ CYP79B3-mediated immunity may be associated with inhibition of root-stem translocation of FocnCong:1-1 (Fig. 4b). CYP79B2/ CYP79B3 is responsible for synthesis of Trp-derived secondary metabolites, including sulfur-containing compounds that are characteristic of the Brassicaceae 26 . These sulfur-containing antimicrobial compounds differ among Brassicaceae species; for example, camalexin is produced in Arabidopsis, but not in cabbage 28 . Our results suggest that FocnCong:1-1 can overcome the Arabidopsis-specific immunity conferred by PAD3, a camalexin synthetic gene ( Fig. 4d and Supplementary Fig. 10c), when the CD chromosome chr SC10/SC20 that encodes the paired effectors SIX8 and PSE1 is present. This pair of effectors is highly conserved in Arabidopsis-infecting F. oxysporum isolates, but not in other isolates (Fig. 6), thus the presence of a particular CD chromosome that harbors these effector genes would contribute to the determination of host specificity.
In this study, FocnCong:1-1 HSs were generated by treatment with the mitosis inhibitor benomyl. In the generation process, a genome rearrangement, but not just a chromosome loss, has occurred at least in HS1, HS2, and HS5 (Fig. 2a). We also investigated phenotypes in an additional HS mutant with the same karyotype as HS5 (HS5L: HS5-like mutant; Supplementary  Figs. 22 and 23). Like HS5, HS5L showed virulence on cyp79b2/ cyp79b3 and pad3 plants, but not on Col-0 WT plants. We cannot rule out the possibility that these genome rearrangements affect phenotypes. In addition to the results of HS5L, however, the return of HS5 virulence on Arabidopsis in two independent HS5 transformants containing FocnCong1-1 SIX8-PSE1 (Fig. 5c) supports the conclusion that the SIX8-PSE1 pair is required for virulence on Arabidopsis.
We identified SIX8 and PSE1 as a gene pair adjacent but encoded on opposite DNA strands (head-to-head orientation) ( Fig. 5b and Supplementary Fig. 11). Head-to-head orientation of effector genes has been documented for other SIX genes in F. oxysporum. For instance, in Fol, a pair of effector genes SIX3 (also known as AVR2) and SIX5 are also adjacently located in a headto-head transcriptional orientation 12,33,34 . Both SIX3 and SIX5 are required for not only full virulence in a susceptible host, but also disease resistance in tomato lines containing the resistance gene I-2 [33][34][35] , and the gene products are thus likely to function as a pair. The close head-to-head orientation may ensure coordinated transcription to produce both proteins at similar levels. Such system would be suitable for maintaining tight stoichiometry of two proteins in a complex. Indeed, SIX5 interacts with SIX3 at plasmodesmata in plant cells, facilitating cell-to-cell movement of SIX3 33,34 . Unlike the SIX3-SIX5 pair, however, we failed to detect direct interaction between SIX8 and PSE1 in a yeast two-hybrid assay ( Supplementary Fig. 24). We cannot exclude the possibility that SIX8 indirectly interacts with PSE1, e.g., via host target(s), or the yeast system may not be suitable for detecting interactions of these proteins. Alternatively, SIX8 and PSE1 may act independently. As bioinformatic analysis of SIX8 and PSE1 protein sequences gives no known domain annotations, identification of host targets of SIX8 and PSE1 will be required to clarify functions of the paired effectors. It is also notable that disruption or loss occurs in only PSE1, but not in SIX8, in certain non-Arabidopsis infecting F. oxysporum isolates. Perhaps PSE1, but not SIX8, is recognizable in plants that carry corresponding resistance proteins, leading to its disruption or loss to avoid detection.
In this work we demonstrate that the host range of F. oxysporum depends on CD chromosomes. In this respect, it is interesting that certain isolates, such as Fol4287 and Forc016, have only a single virulence-associated CD chromosome, whereas FocnCong:1-1 has multiple CD chromosomes, each of which encodes host-specific effectors. Because the FocnCong:1-1 genome is very large (68.8 Mb) compared to most known F. oxysporum genomes, such as Fol4287 (59.9 Mb) 6 and Forc016 (52.9 Mb) 7 , FocnCong:1-1 is likely to have expanded its host range by acquiring and maintaining additional CD chromosomes. Indeed, Masunaka et al. 36 have shown that a field isolate of A. alternata carrying two putative CD chromosomes has a wide host range. In that case, host-specific toxin genes on different chromosomes determine host range 36 . In the case of F. oxysporum, host specificity can be determined, at least in part, by effectors, as seen in this study. Further functional analyses of the SIX8-PSE1 paired effectors and their derivatives will be needed to dissect out the molecular mechanisms underlying effector-based host specificity in F. oxysporum.

Methods
Fungal strains and plants. Fungal strains used in this study are listed in Supplementary Table 2. For pre-incubation, all strains were incubated on potato dextrose agar (PDA; Nissui Pharmaceutical Co.) at 28°C in the dark. For bud cell production, all strains were grown in NO 3 medium (0.17% yeast nitrogen base without amino acids, 3% sucrose and 1% KNO 3 ) at 120 strokes per minute (spm) for 4 days at 28°C in the dark. For gene expression profiling ( Supplementary  Data 1 and 2), mycelia were harvested after 10 days of incubation on PDA at 28°C. Bud cells were collected from NO 3 medium by filtration with a nylon mesh and centrifugation. Hyphae trapped with the nylon mesh were collected. Mycelia from PDA, bud cells, and hyphae were stored at -80°C until RNA isolation.
Bioassays. For evaluation of disease severity, 14-day-old Arabidopsis and cabbage cv. Shikidori roots were injured with a forceps or a plastic peg and then irrigated with 1 ml of FocnCong:1-1 bud cell suspension (1 × 10 7 cells/ml). Inoculated Arabidopsis plants were grown at 28°C for 10 h under light and 14 h dark in a growth chamber. An Arabidopsis disease index was scored at 28 or 29 days postinoculation (dpi) as: 0, no symptoms; 1, dwarf; 2 yellowing, vein clearing or wilting of one to a few leaves; 3, wilting of a whole plant; 4, dead. A cabbage disease index was also scored at 28 or 29 dpi as: 0, no symptoms; 1, yellowing lower leaves; 2, yellowing lower and upper leaves; 3, whole plant wilting; 4, dead.
For gene expression profiling ( Supplementary Data 1 and 2), 20-or 21-day-old Arabidopsis and 17-day-old cabbage cv. Shosyu roots were irrigated with 1 ml of bud cell suspension (1 × 10 7 cells/ml). At 3 dpi and 10 dpi, infected roots were washed with water to remove soil. The roots were stored at -80°C until RNA isolation.
For observation of colonization of Arabidopsis by FocnCong:1-1, roots of 14day-old Arabidopsis were cut to approximately 1 cm lengths from the border between roots and stems and soaked in bud cell suspension (1 × 10 7 cells/ml) for 1 min, then transferred to square plates containing soil. At 12 dpi, roots approximately 5 mm below soil surface were observed by an Olympus BX51 epifluorescence microscopy (Olympus) with excitation of 488 nm for GFP. Images were obtained with an Olympus DP74 digital camera (Olympus) and edited with cellSens (Olympus).
Fungal growth assays. FocnCong:1-1 strains were grown on PDA for 8 days at 28°C in the dark from a freezer stock. For measurement of colony diameter, mycelium agar disks were collected from the growing edge of a colony using sterile plastic straws and placed in the center of fresh PDA plates. After 8 days, colony diameter was measured. For quantification of conidial formation, 17-day-old colonies were soaked in 10 ml of water and scraped with a colony spreader. Conidial suspensions were filtrated through a nylon mesh to remove mycelia and conidia were quantified at OD 600 with a WPA CO 8000 Cell Density Meter (WPA) or by counting using haemocytometer.
To generate the FocnCong:1-1 SIX8-PSE1 locus disruption vector, the flanking regions of SIX8 and PSE1 were amplified from the FocnCong:1-1 SIX8-PSE1 locus complementation vector as a template and assembled with an hph cassette using NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs). The assembled fragment was cloned into pCR™8/GW/TOPO ® .
Constructs for yeast two-hybrid assays were generated from cDNAs of SIX8 and PSE1 without signal peptide sequences or a stop codon by amplification from cDNA generated from mRNA isolated from FocnCong:1-1-infected Arabidopsis. Amplicons were inserted into pENTR™/D-TOPO ® (Invitrogen), and then into yeast expression vectors pDEST-DB and pDEST-AD 38 using Gateway™ LR Clonase™ II Enzyme Mix (Invitrogen) as described by the manufacturer.
Genome sequencing and assembly. For PacBio sequencing, genomic DNA of FocnCong:1-1 was isolated using CTAB and 100/G genomic tips (QIAGEN) as described in the 1000 Fungal genomes project (http://1000.fungalgenomes.org). The genome was sequenced on five PacBio RSII cells and assembled by the Hierarchical Genome Assembly Process (HGAP) v4 within SMRT Link (v5.1.0). Default values were kept and the expected genome size was set to 70 Mb.
For optical mapping, genomic DNA was isolated using a Blood and Cell Culture DNA Isolation Kit (Bionano Genomics) as described by the manufacturer. Genomic DNA was labeled with an NLRS Labeling Kit (Bionano Genomics) with BspQI and BbvCI as described by the manufacturer. The labeled DNA was scanned using a Bionano Irys platform. Bionano maps from two enzymes (BspQI and BbvCI) (Bionano Solve v3.2) were merged with PacBio sequence assemblies to produce long hybrid scaffolds. Completeness of gene space within the assembly was assessed through the presence of conserved single-copy genes using BUSCO version 3.0.2 40,41 . Analysis with the Sordariomyceta data set (3725 genes) indicated the presence of 3690 genes (99.1%) in the assembly (Table 1). Whole-genome alignments were performed with nucmer (with -maxmatch) in MUMmer 3.23 42 .
For genome sequencing of FocnCong:1-1 ΔSIX4 and HSs, genomic DNA was isolated using DNeasy Plant Mini Kits (QIAGEN). Illumina NovaSeq 6000 or HiSeq 2500 paired-end sequencing was used for FocnCong:1-1 ΔSIX4 and HSs, except for HS3, using a library with a mean insert size of 550 bp. Illumina NextSeq 500 single-end sequencing was used for FocnCong:1-1 HS3, from library preparation with a mean insert size of 350 bp. The Illumina sequence library was quality-filtered using the FASTX Toolkit 0.0.13.2 (Hannonlab) with parameters -q20 and -p50. Reads containing "N" were discarded. Quality-filtered libraries were aligned with the FocnCong:1-1 genome using CLC Genomic Workbench 20 using default settings. , followed by a temperature gradient from 55 to 95°C. Standard curves were generated using serial dilutions of cDNAs from Arabidopsis infected with FocnCong:1-1 at 10 dpi for SIX8 and PSE1 and cDNAs from bud cells for FocnCong:1-1 TUB2. FocnCong:1-1 TUB2 was used as a reference gene. Primers used for qPCR are listed in Supplementary Table 3. RNA sequencing. Using the extracted RNA, strand-specific shotgun type of RNA library was prepared using the Breath Adapter Directional sequencing protocol 43 . Briefly, mRNA was extracted and fragmented using magnesium ions at elevated temperature. The polyA tails of mRNA was primed by an adapter-containing oligonucleotide for cDNA synthesis. 5′ adapter addition was performed by breath capture technology to generate strand-specific libraries. The final PCR enrichment was performed using oligonucleotides containing the full adapter sequence with different indexes. After cleanup and size selection, concentration of library was measured by microplate photometer Infinite ® 200 PRO (TECAN) to pool libraries for Illumina sequencing systems. The libraries were sequenced on an Illumina NextSeq 500 platform. The Illumina sequence library was quality-filtered and aligned as above. Transcription levels for each transcript were calculated as TPM (transcripts per million).
Analysis of repeat elements. Repeat element prediction was performed using the genome sequences of eight F. oxysporum strains in the NCBI database that had contig N50 values greater than 1 Mb (last accessed on November 24, 2019) as described in Gan et al. 49 . Code used for this analysis is available at: https://github. com/pamgan/colletotrichum_genome. The details of genome sequences used for this analysis are shown in Supplementary  50 , and LTRPred 51 (https://github.com/HajkD/LTRpred). Sequences that were longer than 400 bp from TransposonPSI, LTR_retriever, and LTRPred were combined and used as queries for BLASTx against RepBase 52 peptide sequences bundled in RepeatMasker open-4.0.9-p2 (http://www.repeatmasker.org). Lastly, these sequences were used as queries for BLASTn against each fungal genome. Only sequences with more than five hits (BLASTn E-value cutoff = 1E-15) and/or with a hit to a RepBase peptide (BLASTx E-value cutoff = 1E-5) were retained for further analysis. Sequences from all sources were combined using VSEARCH v2.14.0 53 , using 80% identity as the cutoff threshold. Consensus sequences were classified using RepeatClassifier (from RepeatModeler open-1.0.11). Known Fusarium repeat sequences registered in Dfam_Consensus-20181026 and RepBase-20181026 were extracted, except for those that were annotated as artefacts, simple repeats, or low complexity sequences. The custom repeat library was created by combining the consensus sequences and known Fusarium repeat sequences, and used as input for RepeatMasker open-4.0.9-p2. The "one code to find them all" 54 was used to reconstruct repeat elements.
Chromosome loss and transfer. A chromosome loss experiment was performed according to VanEtten et al. 22  . Hyphae were removed with a nylon mesh, and bud cells were collected by centrifugation at 1630 g for 10 min. Supernatant was discarded and the remnant with bud cells was spread on M100 plates containing 2% agar and 0.04% Triton X-100 (Wako), and the inoculated plate was overlaid with an autoclaved filter paper. Plates were incubated at 28°C for 1 to 3 days, then the filter paper was transferred onto M100 medium containing hygromycin B (100 μg/ml) and incubated at 25°C overnight. Hygromycin Bsensitive isolates were selected by comparing the plates, and then chromosome loss patterns were verified by PCR ( Supplementary Fig. 2) using primers listed in Supplementary Table 3.
Contour-clamped homogeneous electric field (CHEF) gel electrophoresis. CHEF gel plugs were made by resuspending protoplasts in STE (1 M sorbitol, 25 mM Tris-HCl pH 7.5, 50 mM EDTA). Protoplast concentration was adjusted to 4 × 10 8 cells/ml and added to the same amount of 1.2% low melting agarose gel (Bio-Rad) solution. Protoplast suspensions (2 × 10 8 cells/ml) containing 0.6% low melting agarose gel were added to 50-well dispensable mold plates (Bio-Rad). Plugs were immersed in 10 ml of NDS (1% N-lauroyl sarcosinate sodium salt solution, 0.01 M Tris-HCl, 0.5 M EDTA) and incubated at 65 spm for 14 to 20 h at 37°C. NDS was replaced with 0.05 M EDTA three times every 30 min. Plugs in 0.05 M EDTA were stored at 4°C until use.
CHEF gel electrophoresis was done according to Inami et al. 58 . Briefly, chromosomes were separated on 1% SeaKem ® Gold Agarose (Lonza) in 0.5×TBE buffer at 4 to 7°C for 260 h using a CHEF Mapper System (Bio-Rad). Switching time was 1200 to 4800 s at 1.5 V/cm with an included angle of 120°. The running buffer was exchanged every two or three days. Chromosomes of Hansenula wingei (Bio-Rad) were used as a DNA size marker. Gels were stained with 3×GelGreen (Biotium) to visualize chromosomes.
Statistics and reproducibility. All statistical analyses were performed in EZR 60 . Welch's t-test was used to analyze the statistical significance for continuous variables (e.g., OD 600 value of conidial suspensions), whereas Mann-Whitney U-test was used for evaluation of disease severity. The reproducibility was determined by using independent biological replicates as indicated in the figure legends. Individual values for data plots are included in Supplementary Data 3.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The Whole Genome Shotgun project of FocnCong:1-1 has been deposited at DDBJ/ENA/ GenBank under the accession RSAI00000000 (BioProject number PRJNA506492 and BioSample number SAMN10461798). The version described in this paper is version RSAI01000000. RNA sequencing data from culture medium and plant infections have been deposited in NCBI's Gene Expression Omnibus (GEO) and are accessible through GEO Series accession number GSE157823. The source data underlying Fig. 2c, d, 3b, c, 4a, b, d, 5a, c, d and 6c are provided as Supplementary Data 3. Other data are available by reasonable request.