Segmental Duplications Drive the Evolution of Accessory Regions in a Major Crop Pathogen

Many pathogens evolved compartmentalized genomes with conserved core and variable accessory regions which carry effector genes mediating virulence. The fungal plant pathogen Fusarium oxysporum has such accessory regions often spanning entire chromosomes. The presence of specific accessory regions influences the host range, and horizontal transfer of some accessory regions can modify the pathogenicity of the receiving strain. However, understanding how these accessory regions evolve in strains that infect the same host remains limited. Here, we define the pan-genome of 69 diverse Fusarium strains that cause Fusarium wilt of banana, a significant constraint to global banana production. In this diverse panel of Fusarium strains infecting banana, we analyzed the diversity and evolution of the accessory regions. Accessory regions in Fusarium strains infecting the same banana cultivar are highly diverse, and we could not identify any shared genomic regions and in planta induced effectors. We demonstrate that segmental duplications drive the evolution of accessory regions. Furthermore, we show that recent segmental duplications and aneuploidy occur specifically in accessory chromosomes and cause the expansion of accessory regions in F. oxysporum. Taken together we conclude that extensive recent duplications drive the evolution of accessory regions in Fusarium, which contribute to the evolution of virulence.


Introduction
The interaction between fungal plant pathogens and their hosts drives rapid evolution (M€ oller & Stukenbrock, 2017).Plant hosts have evolved sophisticated immune systems to detect pathogens, while adapted pathogens, in turn, secrete effectors to deregulate immune responses and to support host colonization (Rovenich et al., 2014;Cook et al., 2015).To facilitate this coevolutionary cycle, many filamentous plant pathogens evolved compartmentalized genomes where effector genes are localized in distinct genomic regions (Dong et al., 2015;Torres et al., 2020).
These effector-rich compartments show extensive genetic variation and are enriched for repetitive sequences such as transposable elements (Raffaele & Kamoun, 2012;Dong et al., 2015;M€ oller & Stukenbrock, 2017;Seidl & Thomma, 2017;Torres et al., 2020).This spatial separation is often referred to as the 'two-speed' genome (Dong et al., 2015) and is thought to facilitate the rapid diversification of effector gene repertoires to enable pathogens to evade host resistances or host jumps (S anchez -Vallet et al., 2018).
Fusarium oxysporum is a genetically diverse fungal species complex that consists of three clades that can be further separated into different lineages, some of which are recently reclassified as separate species (Maryani et al., 2019).Members of this species complex can collectively infect > 100 economically important crops.By contrast, individual strains typically only infect a single host plant (Edel-Hermann & Lecomte, 2019).Fusarium oxysporum evolved a two-speed genome organization where specific genomic regions are characterized by extensive presence-absence polymorphisms between strains (Ma et al., 2010;van Dam et al., 2017;Fokkens et al., 2018;Zhang et al., 2020).These variable accessory regions (ARs) can be embedded in core chromosomes or can encompass entire accessory chromosomes (Ma et al., 2010;Fokkens et al., 2018).Interestingly, shared ARs in otherwise genetically diverse strains have been linked to the adaptation toward the same host (Fokkens et al., 2018;Henry et al., 2021).Moreover, ARs can transfer the capacity to infect a specific host between isolates (Ma et al., 2010;Ayukawa et al., 2021).For example, the transfer of accessory chromosome 14 from a tomato-infecting F. oxysporum strain (Fol4287) to a nonpathogenic strain converts the nonpathogenic isolate into a tomato-infecting pathogen (Ma et al., 2010).This transfer of pathogenicity is likely linked to effector genes that are localized in ARs such as some of the 14 well-studied Secreted In Xylem (SIX ) effectors (Rep et al., 2002).Consequently, the presence and absence of effectors can group genetically diverse F. oxysporum strains based on their host range (van Dam et al., 2016).Whole-genome sequencing of diverse F. oxysporum can provide novel insights into the origin of host specificity (Fokkens et al., 2018;Henry et al., 2021).For example, a recent population study of F. oxysporum isolates infecting chickpea suggests that, contrary to expectations (Gordon & Martyn, 1997;Henry et al., 2021), F. oxysporum might undergo sexual and clonal reproduction (Fayyaz et al., 2023), which could impact the constitution and the dynamics of ARs and, ultimately, influence pathogenicity.However, despite the importance of ARs for host adaption, little is known about the constitution of ARs in F. oxysporum strains that cause disease in the same host, and the origin and processes that shape the evolution of ARs.
Fusarium wilt of banana (FWB), which is caused by a suite of F. oxysporum species (Maryani et al., 2019), is a major constraint for banana production and threatens food security in tropical and subtropical countries where > 400 million people depend on banana cultivation (FAO, 2020).In F. oxysporum causing FWB, three different races can be distinguished based on their pathogenicity toward subsets of banana varieties; race 1 (R1) strains cause FWB in Gros Michel, race 2 (R2) strains infect Bluggoe, and tropical race 4 (TR4) strains infect Gros Michel, Bluggoe, and Cavendish (Ploetz, 2015).In addition to TR4, subtropical race 4 (STR4) strains can cause disease in Cavendish banana under environmental stress conditions, for example caused by lower temperatures in subtropical regions (Ploetz, 2006).All races infect additional locally important banana varieties.TR4 has developed into a pandemic over the last 60 yr (Su et al., 1986;Ordonez et al., 2015;van Westerhoven et al., 2022b) and is of particular concern as it causes FWB in Cavendish, the banana variety that dominates global productions (> 50%) and export trade (> 95%; FAO, 2020).Despite the impact of FWB on banana production, we know little about the ARs in F. oxysporum strains that infecting banana and their relation to pathogenicity.
Here, we analyze the dynamics of ARs in a suite of F. oxysporum strains causing FWB.To gain insights into the genomic structure, we generate seven chromosome-level reference genome assemblies for R1, R2, and TR4 strains, and use a pan-genomic framework together with a global panel of 69 strains to gain insights into the evolution of ARs in genetically diverse F. oxysporum strains.

Gene expression analysis
Gene expression patterns were compared between in vitro growth and 8 d post infection.Cavendish 'Grand Naine' banana plants were grown in a glasshouse compartment (28 AE 2°C, 16 h light, and c. 85% relativity humidity) and acclimatized under plastic for 2 wk to maintain high humidity.The roots of c. 2.5-monthold plants were dip inoculated with 10 6 spores ml À1 .RNA was isolated from the roots of inoculated plants at 8 d post inoculation and from mycelium grown on PDA medium.Samples were ground, and RNA extraction was performed using the Maxwell Plant RNA kit (Promega, Madison, WI, USA) following the manufacturer's instructions.The quality of the RNA samples was tested by agarose electrophoresis and quantified by Nanodrop (Thermo Fisher, Waltham, MA, USA), and subsequently sent to KeyGene (Wageningen, the Netherlands) for sequencing.Quantifications of gene transcripts from sequenced RNA-Seq reads were performed using Kallisto (Bray et al., 2016).
To identify orthologous groups (OGs), we used BROCCOLI (v.1.1;Derelle et al., 2020) on the predicted protein-coding genes from the 69 F. oxysporum strains infecting banana; per gene, only the longest transcript was included.Orthologous groups present in all isolates were considered core, OGs found in > 55 genomes (80%) were considered softcore, and genes found in < 55 of the genomes were considered accessory.
We determined the number of nonsynonymous and synonymous substitutions (dN/dS) per orthologous group.All genes in an OG were aligned using MAFFT (v.7.490;Katoh et al., 2002) and nonsynonymous and synonymous substitutions per gene pair in the OG were inferred based on a codon-guided nucleotide alignment, created by PAL2NAL (v.14;Suyama et al., 2006), with CODEML (from PAML, v.4.9;Yang, 2007).Differences between groups of genes were further analyzed using custom python scripts and compared using a two-sided Mann-Whitney U-test in SCIPY (v.1.10.1;Virtanen et al., 2020).(v.2.5.4;Emms & Kelly, 2019) was used to detect the presence of homologs in 308 phylogenetically distinct fungal species from the joint genome institute; per fungal family, one fungal genome assembly was randomly selected (Table S2).

ORTHOFINDER
Homologs that occur at least once outside the phylum

Detection of duplication events
We detected gene duplication events by self-comparisons of the translated protein sequences (only the longest isoform) of the chromosome-level genome assemblies of F. oxysporum infecting banana, as well as Fusarium oxysporum f. sp.lycopersici (strain Fol4287; Ma et al., 2010;GCF_000149955.1),Fusarium graminearum (Cuomo et al., 2007;GCF_000240135.3),Fusarium verticillioides (Cuomo et al., 2007;GCF_000149555.1),and Fusarium solani (Mesny et al., 2021;GCF_020744495.1).Per Fusarium genome assembly, the predicted protein sequences were compared with BLAST and the best five hits were retained after filtering for an evalue 1e À10 and query coverage larger than 50%.Then, BLAST alignments were used to detect collinear duplicated blocks with MCscanX (Wang et al., 2012); blocks of five genes were considered collinear, and matches were filtered with the parameters matchscore 50 and e-value 10 À5 .Subsequently, we classified duplicated genes into dispersed, proximal, tandem, or segmental duplications with MCscanX's 'duplicate_gene_classifier' script.

Fusarium oxysporum strains infecting banana are genetically diverse
To study the occurrence and evolution of ARs in Fusarium oxysporum strains infecting banana, we sequenced and assembled 69 strains (Fig. 1a; Table S1) that were isolated from all major banana growing regions and classified into R1 (39 strains), R2 (two strains), and TR4 (28 strains; Fig. 1a,b; Table S1).De novo assembly of seven strains infecting different banana varieties (three TR4, two R1, and two R2) sequenced with long-read sequencing technology achieved chromosome-level contiguity containing 12-15 contigs, most of which represent complete chromosomes flanked by telomeric repeats (Tables S3, S4).While genome assemblies obtained solely from short-read data yielded more fragmented assemblies, all assemblies contained at least 96.9% of the single-copy BUSCO genes.The assemblies range from 43 to 51 Mb in size (Fig. 1b) and encode between 15 664 and 17 865 predicted proteins, of which 551-671 are effector candidates, and consist of 2.6-8.9% of repeats (Fig. 1e S1), we conducted a phylogenetic analysis based on 3811 single-copy orthologous genes (2226 902 amino acid positions).As expected, F. oxysporum strains infecting banana are polyphyletic and associated with F. oxysporum clades 1, 2, and 3 (Fig. 1a; O'Donnell et al., 1998;Maryani et al., 2019).We observed low nucleotide diversity between TR4 strains (pi-nucleotide diversity 0.0018) compared with a 1009 higher diversity between R1 strains (pi-nucleotide diversity 0.185), corroborating that most TR4 strains evolved from a single recent clonal origin (Ordonez et al., 2015;Maryani et al., 2019).Based on their genetic diversity, F. oxysporum strains infecting banana have recently been separated into different species, and the lineage that encompasses TR4 strains (Fig. 1d) is now referred to as Fusarium odoratissimum (Ordonez et al., 2015;Maryani et al., 2019).

Fusarium oxysporum strains causing Fusarium wilt of banana have a compartmentalized genome
To identify ARs in F. oxysporum strains infecting banana, we performed whole-genome alignments of the seven chromosome-level genome assemblies to the reference genome assembly of the tomato-infecting F. oxysporum strain 4287 (Fol4287; Ma et al., 2010).We observed 11 homologous chromosomes between the F. oxysporum strain II5 (TR4; F. odoratissimum) and Fol4287, which are considered the core genomes (chromosomes 1-11; Fig. 2a).In addition to the core chromosomes, we observed two large ARs specific for strain II5.The first spans 1.8 Mb, occupying 27% of chromosome 1 in proximity to one telomere and the second region is 1.1 Mb in size and constitutes the entire contig 12 (Fig. 2a).Contig 12 only contains one telomere (Fig. 2b), and therefore, it remains currently unknown whether this AR represents an extra chromosome or is attached to one of the core chromosomes.Importantly, however, we assembled a similar contig in TR4 strain M1, and a similar contig is present in the independently assembled TR4 strain UK0001/Eden (Warmington et al., 2019), which strongly suggests that this contig represents an independent AR.
To further analyze the diversity of ARs, we identified ARs in our global collection using a pan-genomic approach based on allvs-all whole genome alignments.ARs were defined as regions longer than 5 kb that are absent in > 55 of 69 strains (80%).We identified a varying amount of ARs in F. oxysporum strains infecting banana (Fig. 1c), ranging from 6.7 Mb (15% of the genome size) in strain Indo110 (R1) to 15.5 Mb (30%) in strain C135 (R2), in line with the 19 Mb (29%) of ARs in the reference strain Fol4287 (Ma et al., 2010).Next to the two ARs in chromosome 1 and contig 12, our pan-genomic approach identified an additional 8 Mb of ARs in strain II5, typically localized at subtelomeric regions (defined as the first and last 10% of each chromosome; Fig. 2b).These sub-telomeric regions contain 1287 genes, of which 80 are predicted effector genes (6.6%).We also identified that chromosomes 9 (42.6%ARs, 87 single nucleotide polymorphisms (SNPs) kb À1 ), 10 (27% ARs, 80 SNPs kb À1 ), and 11 (36% ARs, 92 SNPs kb À1 ) are less conserved than the other core chromosomes (on average 20% ARs, 59 SNPs kb À1 ;   S1), which resembles the situation of the previously identified 'fast-core' chromosomes in Fol4287 (Fokkens et al., 2018).
We show that ARs can be defined based on sequence conservation and differ from the core genome in gene, repeat, and GC content.Consequently, principal component analyses distinguished ARs from the core genome based on these four features (Fig. 2c).Unanticipatedly, this analysis also separated the large ARs on chromosome 1 and contig 12 from the smaller ARs localized at the sub-telomeres (Fig. 2c); this separation is driven by the slightly lower GC content in sub-telomeric regions (44% vs 47%).A reduced GC content can be caused by repeat-induced polymorphisms (RIP) and fungal defense mechanisms that introduces C-to-T mutations in repetitive regions (Galagan & Selker, 2004).In the sub-telomeric region, 25.4% of the repetitive elements are likely subjected to RIP (i.e.composite RIP index > 0), in contrast to only 7.7% in the other ARs, suggesting that repetitive elements at the sub-telomeres are older, or alternatively, that RIP has been more active.S2).In TR4, the ARs from strain II5 on chromosome 1 and contig 12 show extensive similarity to the corresponding chromosomes in strain M1 (Figs 3a, S2).Instead, isolate 36102 (F.odoratissimum), causing milder symptoms in Cavendish than TR4 strain II5, carries only the AR on chromosome 1 and does not encode a region similar to contig 12 (Fig. S2).None of the ARs from strain II5 are present in the strains representing race 1 and race 2. Similar to the TR4 strains, the accessory contigs of two race 2 strains correspond to each other (Fig. S2), but no shared ARs are found among R1 strains.Moreover, we observed that no ARs are conserved among different races; for example, only limited genetic material is shared between ARs of TR4 and R2 (Fig. S2).The diversity of ARs was further corroborated by pairwise analyses of all 69-strains, which revealed that, on average, TR4 shared 91% of the ARs, whereas R1 shared only 24% (Fig. 3b).In line with the high number of shared ARs in TR4 (F.odoratissimum; Fig. 3b), the amount of shared ARs was the highest among R1 strains belonging to the same Fusarium species/lineage (60% shared ARs; FigS 3b, S3); however, R1 strains from different species/lineages share very few ARs (Fig. 3b).
To quantify to what extent F. oxysporum strains infecting banana share ARs with F. oxysporum strains infecting different plant hosts, we identified and compared ARs in a set of 55 additional F. oxysporum strains that are pathogenic to different plant hosts (van Dam et al., 2016).We observed that genetically related strains share most ARs (Fig. S4); however, we also noticed that F. oxysporum strains infecting banana encode more diverse ARs than for instance the tomato-infecting strains; on average, Fol strains share 62% ARs, with a minimal 30% of the ARs shared between any two Fol strains (Fig. S4).By contrast, banana-infecting strains on average share 44% of the ARs and a pair of strains can share as little as 12% (Fig. S4).Thus, we demonstrate that ARs are highly variable among F. oxysporum strains infecting banana, and importantly, strains do not share one host-associated chromosome in contrast to previously reports for F. oxysporum infecting tomato (Fokkens et al., 2018).

Evolutionary dynamics of the accessory genome
Fusarium oxysporum strains infecting banana do not share large ARs (Fig. 3b), but we reasoned that ARs nevertheless might share genes that contribute to the pathogenicity of banana.We therefore grouped 1.1 million protein-coding genes predicted in 69 strains into 22 612 OGs and subsequently analyzed the gene-content diversity.The pan-genome consisted of 12 101 core groups (53%) that are present in all 69 strains, 2395 softcore groups present in at least 80% of the strains, 5595 (25%) accessory groups present in < 80% of the strains, and 2521 unique genes (Fig. 4a).Importantly, the pan-genome based on our collection captures the diversity of protein-coding genes in F. oxysporum strains infecting banana as we did not observe an increase in the pan-genome size after including > 40 strains (Fig. 4a).The conserved core genes are, as expected, enriched with housekeeping genes, while accessory genes are enriched with genes encoding secondary metabolites (Fig. S5).No Lines (red and blue) indicate aligned regions between strains based on nucleotide similarity.Small alignments are present between R1 and TR4 (blue lines); however, these are rearranged and often involve repetitive sequences (yellow blocks).(b) R1 strains (39 strains, highlighted in blue) have a diverse accessory content sharing on average only 20% of the ARs, whereas TR4 strains (28, highlighted in purple) share most of their ARs (91%).Phylogenetically related F. oxysporum strains, here indicated by a recently proposed species name (Maryani et al., 2019), share more ARs than distantly related strains.
New Phytologist (2024) 242: 610-625 www.newphytologist.comÓ 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation enrichment of effector genes is found in any of the three categories (core, accessory, or softcore); the pan-genome consists of 739 gene families encoding predicted effectors that are evenly distributed over the different gene categories (3.7% of accessory genes and 3.5% of the core genes).The role of accessory genes in host infection is suggested by gene expression, as accessory genes are upregulated 8 d after banana infection, with a median log 2 fold change of 1.59 (compared with the in vitro control; Fig. 4b), comparable to the upregulation of effector genes (median log 2 fold change of 2.4).The core genes, however, show significantly less increase in gene expression upon plant infection (median log 2 fold change = 0.25, P < 0.05, twosided Mann-Whitney U-test; Fig. 4b).Although this suggests a role of accessory genes in host-pathogenicity, we observed a  highly diverse gene content in R1 strains with only 246 out of 7832 accessory genes (3%) being shared among all R1 strains (Fig. 4f).TR4 strains, by contrast, have a more similar gene profile and share 2493 of the 4406 accessory genes (57%).In TR4 strain II5, ARs have a different gene composition than the core regions, with ARs consisting mostly of accessory genes (405 out of 463; 87%); however, four core and 38 soft-core genes are also found in ARs (Fig. 4e), indicating that some shared genes can be found between otherwise nonconserved regions.ARs seem to contain recently evolved genes, as most AR genes (71%) have homologs in closely related fungi (Fig. 4g,h) and only 29% are considered 'old', that is with homologs in all fungi, which contrasts to the 53% old genes in the core genome (7764, out of 14 774 genes; Fig. 4g,h).Additionally, accessory genes are significantly shorter (315 aa) and evolve under relaxed selective pressure (dN/dS = 0.20), compared with the core genes (469 aa, dN/dS = 0.089; P < 0.05, two-sided Mann-Whitney U-test; Fig. 4b,c), both hallmarks of more recent genes (Wolf et al., 2009).This suggests that most genes in the ARs evolved recently and that these regions serve as cradles for the evolution of novel genes.In addition, many genes on the ARs (177 out of 463) belong to expanded gene families based on the defined orthogroups (Fig. 4e), suggesting that gene duplications play a role in the evolution of ARs in F. oxysporum strains infecting banana.

Variable effector repertoire in Fusarium oxysporum genomes
Effector gene repertoires define host range (van Dam et al., 2016;Brenes Guallar et al., 2022), and we anticipated that effector presence-absence profiles would distinguish F. oxysporum strains infecting banana from F. oxysporum strains infecting other hosts.We predicted candidate effector genes in all 124 F. oxysporum isolates based on their proximity to MIMP elements that are known to co-localize with some effector genes in F. oxysporum (van Dam et al., 2016;Brenes Guallar et al., 2022), which yielded 398 MIMP-associated effector gene families (160 of these were predicted in the complete set of 739 candidate effectors).The predicted MIMP-associated effector repertoires clustered F. oxysporum isolates in host range (van Dam et al., 2016;Brenes Guallar et al., 2022;Fig. 5).While banana-infecting strains form a clear cluster, only four MIMP-associated effectors were identified that occur in most (i.e.> 80%) of banana-infecting strains but are consistently absent (i.e.present in < 20%) in other F. oxysporum strains.Moreover, races within F. oxysporum strains infecting banana are not clearly separated; TR4 strains have very similar effector profiles, yet R1 strains differ considerably and do not encode shared effectors.Interestingly, two STR4 strains carry an effector profile remarkably similar to the effector profile of TR4 strains (Fig. 5).This suggests that this effector profile

Research
New Phytologist contributes to the pathogenicity of these STR4 strains to Cavendish banana.However, this effector profile is not STR4 specific as the other two STR4 strains encode effector profiles similar to R1 strains (Fig. 5).This difference can be the result of misclassification of STR4 strains, because STR4 infects Cavendish only under environmental stress conditions and this environmental component makes it difficult to distinguish true STR4 isolates from race 1 isolates (Ploetz, 2006).To further investigate effectors shared between TR4 and STR4, an accurate screening strategy to distinguish true STR4 needs to be conducted.We here reveal a diverse effector repertoire in R1 strains, together with the variable ARs and gene content, this suggests that the species encompassing R1 utilize varying molecular mechanisms to support banana infection.

Recent segmental duplications drive accessory genome expansion in Fusarium oxysporum
We observed diverse ARs in F. oxysporum strains infecting banana with a high abundance of evolutionary young genes and genes that evolved via gene duplications (Fig. 4e).Duplications have been previously reported in F. oxysporum, including F. oxysporum strains infecting banana (Kistler et al., 1995;Ma et al., 2010;Vlaardingerbroek et al., 2016;Armitage et al., 2018;Li et al., 2020).To better understand the role of gene duplications in the origin and diversification of ARs, we characterized gene duplications in F. oxysporum, using MCscanX (Wang et al., 2012) and classified them into four categories (dispersed, proximal, tandem, and segmental; Fig. 6a).Remarkably, most genes in TR4 strain II5 (8831 genes) have been affected by gene duplication during their evolution (Figs 6b,S6).This high number of duplications include old gene duplications, recognized as distant homologs that acquired considerable number of mutations over time.When we apply a more stringent percent identity filtering, the number of duplicated genes is reduced, yet the number of segmental duplicated genes is least affected (Table S5), indicating that the segmental duplicated genes acquired less mutations and thus evolved more recently.We observed similar results for other F. oxysporum strains as well as for other Fusarium species; F. verticillioides, F. graminearum, and F. solani (syn.Neocosmospora solani), suggesting that duplications played a major role in Fusarium evolution (Fig. 6b).Interestingly, we observed that segmental duplicated genes occur specifically in ARs (72 out of 124 in II5) and sub-telomeres (48 out of 124 in strain II5) in all chromosome-level F. oxysporum assemblies (Figs 6b, S6).In F. solani, which contains accessory chromosomes analogous to those observed in F. oxysporum (Coleman et al., 2009), 139 segmental duplicated genes were identified.By contrast, in F. verticillioides and F. graminearum, for which no accessory chromosomes have been described, no segmental duplicated genes are found (Fig. 6c-e), underscoring the link between accessory chromosomes and segmental duplications in the Fusarium genus.Moreover, segmental duplicated genes in strain II5 are upregulated during infection (8 d post inoculation; median log 2 -fold change = 2.69; Fig. S7; Table S6), suggesting a role of these genes in the infection process.These duplications are possibly driven by TE activity (Wicker et al., 2010;Faino et al., 2016), and we observe that segmental duplications in strain II5 occur closer to transposable elements (median distance of 1734 kb, P < 0.05, two-sided Mann-Whitney U-test) than other duplication types (dispersed 7721 bp, proximal 7503 bp, and tandem 5941 bp), suggesting that TE activity might influence segmental duplications.
The highest number of segmental duplicated genes ( 2887) is observed in Fol4287, carrying the largest accessory genome (19 Mb, 30% of the total genome size; Figs 6d, S6) and 2366 (82%) of the segmental duplicated genes are in ARs.Interestingly, many segmental duplicated genes are shared between accessory chromosomes 3 and 6 (Fig. S8), as observed previously (Ma et al., 2010), indicating that accessory chromosomes evolved by inter-chromosomal segmental duplication that drives the expansion of ARs in Fusarium.
Duplications of AR genes significantly affect effector repertoires as 336 out of 669 predicted effectors in II5 evolved via duplications, and nine effectors are part of segmental duplications.Among the segmental duplicated effectors is SIX1, an effector that is essential for full virulence to banana (Widinugraheni et al., 2018).Three copies of SIX1 (SIX1a,b,c) are present in the AR on chromosome 1 of the TR4 strains M1, 36 102, and II5.SIX1a and SIX1b are part of a segmentally duplicated block (Fig. 6g), whereas SIX1c is a proximal duplication that shares 82% amino acid sequence identity with SIX1a and 73% sequence identity with SIX1b.By contrast, the R1 strain CR1.1 contains only one copy of SIX1, resembling SIX1c.The duplicated blocks with SIX1 in strain II5 are interspersed with nonduplicated genes (Fig. 6g), indicating that over time these blocks further diverged by gene gains and losses.Interestingly, the segmentally duplicated SIX1 blocks also share similarity to a region on contig 12 (Fig. 6g), yet no copy of SIX1 is present, and thus, SIX1 is either lost or has been gained on chromosome 1 after the segmental duplication between chromosome 1 and contig 12 occurred.Segmental duplications therefore contribute to the evolution of virulence factors such as SIX1, which is essential for successful banana infection (Widinugraheni et al., 2018).
To estimate the relative timing of the duplication events, we used the synonymous substitution rate (dS) between duplicated gene pairs as a proxy of time.Dispersed duplicated genes had the highest average dS value (3.08), suggesting that these are the most ancient duplicates, while gene pairs that arose via segmental duplication evolved more recently, with a significantly lower average dS value of 0.61 (Fig. 6f).To assess whether even more recent large-scale duplications are present in our global panel, we determined the read depth of all isolates sequenced with short reads against II5.Based on the genome-wide average read depth (c.409 coverage), we determined that contig 12 has been entirely duplicated (read depth of c. 809 coverage), while a section (position c. 0.6-1.1 Mb; c. 1609 coverage) occurs four times (Fig. 6h), suggesting additional copy number changes in strain II5 next to the previously observed segmental duplications (Fig. 6h).Similar copy number changes were observed when aligning the reads to a repeat masked genome assembly (Fig. S9), indicating that the coverage increase is not due to the expansion of repetitive sequences in the re-sequenced genome or to the collapse of repeats in our assembly.Additionally, the alignment of the long-read sequences also supports the observed copy number increased (Fig. S9).Interestingly, no copy number variation of the AR of chromosome 1 is observed.The partial copies of contig 12 possibly originate from additional duplications, yet it currently remains unclear whether these are attached to (core) chromosome(s), or whether they might give rise to an additional smaller accessory chromosome.To resolve the origin of this duplicated region, we mapped the long reads to the assembly.We observed many clipped reads in this region; however, clipped regions did not align to another chromosome, suggesting that contig 12 or part of it are not attached to another chromosomal arm.Interestingly, the copy number increase in contig 12 can also be observed in three additional TR4 strains from the Middle East, but not in the other 26 TR4 strains (Fig. 6h).Collectively,

Discussion
Many fungal plant pathogens evolved compartmentalized genomes with conserved core regions and variable ARs (Dong et al., 2015;Torres et al., 2020).Some members of the F. oxysporum species complex are important plant pathogens and are well-known to have extensive ARs that can encompass entire chromosomes (S anchez -Vallet et al., 2018), which have been linked to pathogenicity toward specific hosts.Here, we use a pangenomic framework to study the occurrence, composition, and evolution of ARs in a global collection of banana-infecting F. oxysporum strains, belonging to the several distinct lineages that have recently recognized as distinct species (Maryani et al., 2019).Fusarium oxysporum strains carry up to 15 Mb of ARs that contain in planta induced effector genes as well as most SIX effectors.ARs are highly divergent among strains and races, and consequently, we cannot identify shared ARs or accessory chromosomes that can be linked to pathogenicity toward banana.We furthermore demonstrate that members of the Fusarium genus evolved by extensive duplications and that ARs specifically are shaped by recent segmental duplications in F. oxysporum and F. solani.Lastly, we uncover that an accessory contig in strain II5 underwent a recent copy number change, suggesting that these processes drive the emergence and evolution of ARs in F. oxysporum.
Fusarium wilt of banana is a major threat to food security as banana is a major staple crop in tropical and subtropical regions (Drenth & Kema, 2021;van Westerhoven et al., 2022b).Compared with R1 strains, which caused epidemics in Gros Michel bananas in the 1900s, TR4 strains emerged more recently (1967) and caused the ongoing FWB pandemic in Cavendish plantations (van Westerhoven et al., 2022b).TR4 strains are genetically highly similar, corroborating that a single clone that circumvents Cavendish resistance drives the ongoing pandemic (Ordonez et al., 2015;van Westerhoven et al., 2022b).By contrast, R1 strains are genetically diverse, which is possibly due to prolonged co-evolution of R1 strains with a plethora of genetically diverse banana varieties in the center of origin in Southeast Asia (Perrier et al., 2011).Extensive genetic diversity among F. oxysporum strains that infect the same host is not uncommon (Baayen et al., 2000;van Dam et al., 2016;Henry et al., 2021;Fayyaz et al., 2023), yet their ARs and effector profiles are typically remarkably similar (van Dam et al., 2016;Fokkens et al., 2018;Brenes Guallar et al., 2022).We, however, observe that the ARs and effector profile of R1 strains show extensive variation, underscoring their diversity and possibly suggesting that different mechanisms contribute to disease in banana (Henry et al., 2021;Brenes Guallar et al., 2022).F. oxysporum strains that infect strawberry have two distinct ARs, one that contributes to yellowing and the other to wilting (Henry et al., 2021).Quantitative differences in banana corm-discoloration have been noted between R1 strains (Garcia-Bastidas, 2019), and our results suggest that this might relate to differences in effector profiles, yet further functional assays are needed to elucidate the link between specific effectors and phenotypic variation.
We identify abundant and recent segmental duplications in F. oxysporum ARs, as well as a copy number change in contig 12 in TR4 strain II5.Previous studies have observed large-scale duplications and deletions in the ARs of F. oxysporum during in vitro chromosome transfer (Vlaardingerbroek et al., 2016;Li et al., 2020), yet our results show that duplication of ARs also occurs in F. oxysporum strains sampled directly from infected banana plants.Although variations in chromosome number or large-scale duplications typically come with a fitness cost (Todd et al., 2017), it can also provide genetic variation necessary for adaptation, for example increasing virulence or fungicide resistance (Sionov et al., 2010;Ropars et al., 2018).We speculate that the observed large-scale duplications play a major role in the origin and evolution of ARs in Fusarium.In the wheat pathogen Zymoseptoria tritici, chromosomal duplication, together with breakage and fusion, shapes the evolutionary dynamics of ARs (Croll et al., 2013;M€ oller et al., 2018).Similar dynamics of chromosome duplication, fission, and fusion may underly the recent segmental duplications observed in F. oxysporum.For example, the fusion of a (partially) duplicated contig 12 with the arm of chromosome 1 in strain II5 could explain the observed homology between the ARs, or alternatively, the fission of chromosome 1 could have led to the emergence of contig 12.The process underlying these chromosome dynamics remains unclear.Fusarium oxysporum has long been considered to evolve strictly asexually (Gordon & Martyn, 1997); however, the presence of an active sexual cycle in F. oxysporum populations has recently been proposed (McTaggart et al., 2022;Fayyaz et al., 2023) and we similarly observe reticulation between strains that caused FWB (Fig. S10), which supports an evolutionary history in which ancient or infrequent sexual recombination is followed by clonal expansion of selected lineages.In this scenario, incomplete chromosome pairing and nondisjunction during meiosis might drive copy number variation of chromosomes, segmental duplications, and ARs diversification.We show that ARs are similar within individual F. oxysporum lineages but are genetically distinct when comparing strains from genetically distant lineages.Individual lineages accumulate genomic changes over time, and this process might be sufficient to explain ARs diversity.If new F. oxysporum lineages indeed arise from a meiotic cycle followed by clonal expansion, recombination, gain, or loss of ARs during meiosis might further explain the diversity of ARs between lineages.Similarly, the absence of meiotic recombination during subsequent clonal expansion would explain the similarity of ARs within clonal lineages.However, we observe copy number variation within individual clonal lineages, for instance the duplication of contig 12 in some F. oxysporum strains, which demonstrates that this variation can occur independent of a meiotic cycle.Thus, chromosome dynamics in F. oxysporum require further elucidation, and so far, the mechanisms leading to chromosomal duplications are unclear; they might arise from nondisjunction during mitosis or meiosis (Fayyaz et al., 2023), from incomplete chromosome loss following heterokaryon formation (Harrison et al., 2014;Shahi et al., 2016) chromosome transfers when strains acquire the identical chromosome or parts of it several times.In contrast to the structural variation in the ARs, the core chromosomes in F. oxysporum are remarkably stable.Although ARs are likely more tolerant to structural variation as they encode fewer essential genes, additional processes can influence genome stability.First of all, the accessory genome is enriched in transposable elements and the activity of these elements can contribute to structural variation including gene duplications (Faino et al., 2016;Torres et al., 2021;Stalder et al., 2022;Wang et al., 2022).Possibly the activity of transposable elements in the accessory genomic regions gave rise to the observed segmental duplications, for example, in plant genomes, large-scale duplications can arise through double-stranded break repair upon TE-induced double-stranded breaks (Wicker et al., 2010;Wang et al., 2022).Moreover, the presence of histone modifications such as trimethylation of lysine 27 on histone 3 (H3K27me3), a histone modification often enriched in ARs (Fokkens et al., 2018;M€ oller et al., 2019;Cook et al., 2020;Torres et al., 2020) in fungal plant pathogens, has been implied in genome instability (Seidl et al., 2016;M€ oller et al., 2019).H3K27me3 also occurs in the ARs of F. oxysporum infecting tomato (Fol4287), but can also be found in smaller core chromosomes ('fast-core' chromosomes, 9-11;Fokkens et al., 2018) that do not seem to undergo segmental duplications or copy number changes, and thus, additional factors likely influence chromosome stability in F. oxysporum.
Extensive duplications similarly occur in other plant pathogens (Dutheil et al., 2016;M€ uller et al., 2019;Wyka et al., 2021;Wacker et al., 2023) and are thought to be important drivers in co-evolutionary arms races with their hosts.Understanding the evolution of ARs in Fusarium oxysporum can facilitate the discovery of new effector genes and provide insights into effector diversification.Knowledge of effector profiles is crucial for designing effective disease control strategies and supports the identification of durable resistances in crops (Vleeshouwers et al., 2008).Table S1 Genomes of Fusarium oxysporum strains infecting banana analyzed in this study.
Table S2 Fungal genomes used in our analyses to estimate the gene age.
Table S3 Statistics of the long-read nanopore sequencing data of seven Fusarium oxysporum strains.
Table S4 Genome assembly statistics of seven Fusarium oxysporum strains sequenced with nanopore long-read sequencing technology.
Table S5 Number of duplicated genes as reported compared with the number of duplicated genes using a stricter filtering approach.
Table S6 Gene expression statistics per gene duplication category in Fusarium odoratissimum strain II5, both in vitro and 8 d post infection on Cavendish banana.
Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors.Any queries (other than missing material) should be directed to the New Phytologist Central Office.
, f), similar to previous F. oxysporum genome assemblies (van Dam et al., 2016; Warmington et al., 2019; Zhang et al., 2020).To assess the genetic diversity of F. oxysporum strains infecting banana in relation to 55 publicly available F. oxysporum strains infecting eight different host plants (van Dam et al., 2016; Table

Fig.
Fig.S1), which resembles the situation of the previously identified 'fast-core' chromosomes in Fol4287(Fokkens et al., 2018).We show that ARs can be defined based on sequence conservation and differ from the core genome in gene, repeat, and GC content.Consequently, principal component analyses distinguished ARs from the core genome based on these four features (Fig.2c).Unanticipatedly, this analysis also separated the large ARs on chromosome 1 and contig 12 from the smaller ARs localized at the sub-telomeres (Fig.2c); this separation is driven by

Fig. 2
Fig.2Fusarium oxysporum strains infecting banana carry accessory regions (ARs).(a) The tropical race 4 (TR4) reference strain II5 (Fusarium odoratissimum, TR4) contains 11 core chromosomes (chromosomes 1-11) as well as two ARs (1.8 Mb on chromosome 1 and entire contig 12).Core chromosomes align to Fol4287 in contrast to the ARs that are highlighted in orange.(b) The ARs (orange) in II5 carry most Secreted In Xylem (SIX ) genes and show an increased repeat and lower gene content.A similar pattern is found in the sub-telomeric regions (green, first and last 10% of the chromosome).(c) Principal component analysis on gene-, repeat-, GC content, and coverage of 69 F. oxysporum strains separates the genomic regions of II5 into three different clusters.ARs (orange and green) are separated from the core region (grey) and further separated into those ARs localized in subtelomeric regions (green) and those on chromosome 1 and contig 12 (orange).

Fig. 3
Fig.3Accessory regions (ARs) in Fusarium oxysporum strains infecting banana differ significantly.(a) Accessory contig 12 of the F. oxysporum tropical race 4 (TR4) reference strain II5 (F.odoratissimum) is shared with TR4 strain M1; however, this contig shows little similarity to race 1 (R1) strain CR1.1.Lines (red and blue) indicate aligned regions between strains based on nucleotide similarity.Small alignments are present between R1 and TR4 (blue lines); however, these are rearranged and often involve repetitive sequences (yellow blocks).(b) R1 strains (39 strains, highlighted in blue) have a diverse accessory content sharing on average only 20% of the ARs, whereas TR4 strains (28, highlighted in purple) share most of their ARs (91%).Phylogenetically related F. oxysporum strains, here indicated by a recently proposed species name(Maryani et al., 2019), share more ARs than distantly related strains.

Fig. 4
Fig.4Evolutionary dynamics of genes in the different genomic compartments in Fusarium oxysporum strains infecting banana.(a) Global collection of F. oxysporum strains infecting banana has a pan-genome size of 20 091 orthologous groups (OGs, 22612 when unique genes are included), of which 12 101 are present in all strains (grey).The pan-genome is saturated after adding 40 genomes.The order of genomes was randomly sampled 10 times, dots show the average pan-genome size, and the error bars indicate the SD.(b) Accessory genes are upregulated in planta at 8 d post inoculation (dpi) of banana plants.Average log 2 fold change in accessory genes (2416 in II5) and effector genes (629 in II5) shows these genes are typically upregulated upon plant infection, which contrasts the core (11 740 genes, log 2 -fold change = 0.252) and BUSCO genes (4364 genes, log 2 -fold change = À0.167).The median log 2 fold change between all gene categories is significantly different (P < 0.05, two-sided Mann-Whitney U-test).Dotted lines indicate the median value.(c) Sequence conservation estimated by the number of nonsynonymous and synonymous substitutions (dN/dS) values.BUSCO gene families (4443 OGs) and core gene families are most conserved, whereas the accessory genes evolve under more relaxed selective pressure.All gene categories show significant variation in dN/dS value (P < 0.05, Mann-Whitney U-test).Dotted lines indicate the median value.(d) Distribution of protein length separated by different gene categories shows that accessory gene families (5595 OGs, pink) are shorter (median length of 315 aa per OG) than core gene families (median length 469 amino acids per OG, 12101 OGs, grey).All gene categories differed significantly from each other (P < 0.05, Mann-Whitney U-test).Boxes indicate the interquartile range of the values, the median values are indicated by horizontal lines, and the whiskers depict the minimum and maximum values.(e) Accessory regions (ARs) in II5 (orange, contig 12, and chromosome 1A) consist mostly of accessory genes (pink, 405 out of 463, 87%), and 28% of these genes are part of expanded orthogroups (129 out of 463).Accessory genes can also be found in core chromosomes.(f) Gene content is variable between race 1 strains, and only 20% of the genes are present in 80-90% of the race 1 strains.Most genes (60%) in tropical race 4 (TR4) are present in all TR4 strains (90-100%).(g) Gene families in strain II5 (Fusarium odoratissimum, TR4) differ to which extend they have homologs in 308 fungal phyla; most genes are present in related F. oxysporum strains, but only a few gene families (853) are conserved among all considered fungi.(h) ARs contain most recent genes, present in Fusarium (51) and Hypocreales (160).Genes in core regions are evolutionarily older with more genes that have homologs in 'all fungi' (7764 out of 11 740 genes).

Fig. 5
Fig.5Predicted effector profile clusters Fusarium oxysporum strains infecting the same host.Fusarium oxysporum strains infecting banana (yellow) are separated from F. oxysporum strains infecting other hosts.In the set of 124 Fusarium isolates, 398 Miniature IMPala (MIMP) associated effector families are predicted.The race 1 strains (blue) have diverse effector profiles and do not cluster together, whereas the effector profiles of tropical race 4 strains (purple) are highly similar.Nine Secreted In Xylem (SIX) effectors (blue) are identified in the 398 MIMP-associated effector families and are present in F. oxysporum strains infecting tomato, from which they are originally identified, as well as in other F. oxysporum genomes.

Fig. 6
Fig.6Segmental duplications are involved in the expansion of accessory regions (ARs) in Fusarium oxysporum.(a) Four different types of duplications are distinguished.(b) Segmental duplications (pink) occur mainly in the ARs in strain II5 (tropical race 4 (TR4); Fusarium odoratissimum).Sub-telomeric regions have more dispersed duplicated genes (blue) than the chromosomal arms.On average, the read coverage of II5 mapped onto the II5 reference genome is 409, whereas contig 12 has a coverage between 809 and 1609, suggesting copy number variations.(c) The schematic tree depicts the relationship of Fusarium species with (d) their genome sizes.(e) Extensive segmental duplications are only found in F. oxysporum and Fusarium solani and are virtually absent in Fusarium verticillioides and Fusarium graminearum.(f) The age of duplications was estimated from the number of synonymous substitutions (kS) between duplicated gene pairs.Segmental duplications occurred most recently (kS = 0.119).kS distributions differ significantly between all duplication types (P < 0.05, two-sided Mann-Whitney U-test).Dotted lines indicate the median values.(g) Genes encoding the virulence factor Secreted In Xylem (SIX) 1 (red) are part of a segmental duplication within chromosome 1.SIX1a and SIX1b are segmentally duplicated, and SIX1c is a proximal duplication sharing 82% amino acid sequence similarity to SIX1a.Light red blocks indicate segmental duplications; brown lines between the genes (blue arrows) highlight the genes involved in the segmental duplication.Not all genes in a block are part of the segmental duplication, for example a region on contig 12 is similar to the SIX1 regions on chromosome 1, yet SIX1 is absent on contig 12. (h) Sequencing read coverage of F. oxysporum strain II5 (TR4; F. odoratissimum) and three additional II5 strains from the Middle East (JV11, ISR5, and JV14) show a twofold increase relative to the genome-wide coverage.The other 26 strains do not show an increase in coverage, although smaller duplications are found throughout contig 12.

Fig. S3
Fig.S3Fraction of shared accessory regions in different subgroups of Fusarium oxysporum strains infecting banana.

Fig. S4
Fig. S4 Accessory regions are shared between genetically related Fusarium oxysporum strains.

Fig. S6
Fig.S6Duplication types found in the core, softcore, and accessory regions in the seven chromosome-level genome assemblies of Fusarium oxysporum strains infecting banana and the Fusarium oxysporum strain infecting tomato (Fol4287).

Fig. S7
Fig. S7 Distribution of log 2 -fold gene expression changes per gene in strain II5 (TR4; Fusarium odoratissimum), comparing gene expression between in vitro growth with 8 d post inoculation of Cavendish banana.

Fig. S10
Fig. S10 Splitstree network shows indications of reticulation that might suggest recombination between some Fusarium oxysporum strains infecting banana.Methods S1 Supporting materials and methods.

Ó
2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation New Phytologist (2024) 242: 610-625 www.newphytologist.com Dotted lines indicate the median value.(d) Distribution of protein length separated by different gene categories shows that accessory gene families (5595 OGs, pink) are shorter (median length of 315 aa per OG) than core gene families (median length 469 amino acids per OG, 12101 OGs, grey).All gene categories differed significantly from each other (P < 0.05, Mann-Whitney U-test).Boxes indicate the interquartile range of the values, the median values are indicated by horizontal lines, and the whiskers depict the minimum and maximum values.(e) ).The median log 2 fold change between all gene categories is significantly different (P < 0.05, two-sided Mann-Whitney U-test).Dotted lines indicate the median value.(c)Sequence conservation estimated by the number of nonsynonymous and synonymous substitutions (dN/dS) values.BUSCO gene families (4443 OGs) and core gene families are most conserved, whereas the accessory genes evolve under more relaxed selective pressure.All gene categories show significant variation in dN/dS value (P < 0.05, Mann-Whitney U-test).Ó 2024 The Authors New Phytologist Ó 2024 New Phytologist Foundation New Phytologist (2024) 242: 610-625 www.newphytologist.com