Comparative alignments of regions from the pathogenicity chromosomes of Fom005 and Forc016 were performed and visualized using PyGenomeViz v1.0.0 with default MUMmer alignment settings (Shimoyama, 2024). For this alignment, 200 kb segments were selected from each strain: 1,000,000 – 1,200,000 bp of the ChrRC of Forc016 (GenBank: CM008298.1) and 318,000 – 518,000 bp of contig 129 from Fom005 (GenBank: GCA_001703205.2). Gene annotation from Forc016 sequences were transferred to the corresponding Fom005 region by importing features in SnapGene v8.0.3 (https://www.snapgene.com), based on local sequence similarity. For the alignment of ECC1a (positions 38,884-49,385 bp) and ECC1b (positions 161,609-172,113 bp) loci from Fom005, the ECC1b region was reversed to account for orientation prior to visualization using the same PyGenomeViz settings. Other genes in these regions were further investigated by retrieving annotated gene names and functional description from the UniProt database (Bateman et al., 2025).

Protein sequence alignment was conducted using ClustalW through the Jalview platform v2.11.4.1 (Waterhouse et al., 2009). The resulting multiple sequence alignments were visualized using ESPript 3.0 (ENDscript – https://endscript.ibcp.fr) (Robert and Gouet, 2014).

The protein structures of ECC1 homologs were predicted using AlphaFold3 (Abramson et al., 2024). Structures were visualized and converted to PDB format using ChimeraX v1.10 (Meng et al., 2023). The resulting PDB file from ECC1a^Fom was used as input for structural similarity analysis with the DALI server (Holm, 2020). A pairwise structure comparison was performed against ToxA from Pyrenophora tritici-repentis (Sarma et al., 2005) (PDB entry: 1ZLD) and four known ToxA-like Fo effector structures: FOXGR_015533, SIX7, SIX8 and Avr2 (SIX3), for which PDB files were previously published (Yu et al., 2024).

FOSC assemblies downloaded from GenBank are listed in Supplementary Table S1 . Assemblies were selected based on having a high N50 where reference-guided assemblies were excluded, or based on being described as f. sp. melonis, cucumerinum, niveum, or radicis-cucumerinum. To infer a species phylogeny, we first identified BUSCO genes (busco version 5.7.1, with –metaeuk and hypocreales_odb10) and used a custom Python script to select BUSCO genes that were present as single-copy in all assemblies. These were aligned with Muscle (v 5.3.linux64) with default settings and trimmed with TrimAl (v1.5.rev0, -automated1: optimized for maximum likelihood inference). Subsequently, these trimmed alignments were concatenated with second custom Python script. We then used ModelFinder in IQTree to identify an optimal substitution model (GTR+F+I+R9) and IQtree with UFBoot (-m MFP -B 1000 -bnni -alrt 1000) to infer a consensus maximum likelihood phylogeny from this concatenated alignment with 1000 bootstrap replicates.

To infer the gene tree of ECC1, we used megablast with default settings to identify homologs of the gene sequences of ECC1a^Fom , ECC1a^{For c} and ECC1b (including introns) in the genome assemblies listed in Supplementary Table S1 . We selected all hits with an E -value < 0.001 for which more than 50% of the query sequence was represented in the alignment. We used a one-line awk script to convert coordinates returned by BLAST to bed format, used bedtools slop to add 50 bp of flanking sequence where possible, merged overlapping regio with bedtools merge and used bedtools getfasta to obtain sequences in fasta format. We inferred a multiple sequence alignment with mafft (with –adjust direction accurately because sequences may be in different orientations). We inspected the resulting alignment and manually removed the 50 bp flanks in AliView (Larsson, 2014). We then inferred a gene tree with IQ-tree (Minh et al., 2020), implementing ModelFinder (Kalyaanamoorthy et al., 2017) to identify the best substitution model (K2P+G4), and UFBoot (Thi Hoang et al., 2017) for bootstrapping (1000 replicates).

Musk melon (Cucumis melo cv. ‘Cha-T’) and cucumber (Cucumis sativus cv. ‘Paraiso’) seeds were germinated and inoculated in a climate-controlled greenhouse at 25°C, with a relative humidity of 65% and 16/8h light/dark cycles.

Fusarium oxysporum (Fo) strains Fo f. sp. radicis-cucumerinum 016 (Forc016; ‘33’) (Lievens et al., 2007) and Fo f. sp. melonis 005 (Fom005; ‘Fom0123’) (Alvarez et al., 2005) were grown on Czapex Dox Agar (Difco) plates at 25°C in the dark.

For ORF deletion mutants of the ECC1 homologs, the ORF was replaced with a hygromycin (HPH)-GFP cassette. The HPH-GFP cassette was excised from pPK2-HPH-GFP (Michielse et al., 2008) using HindIII and EcoRV (ThermoFisher Fermentas). Homologous flanking regions of 1kb upstream (primers FP10575 and FP10576) and downstream (primers FP10577 and FP10578) of the ORF were inserted adjacent to the cassette using Gibson cloning (NEBuilder^® HiFi DNA Assembly Cloning Kit, New England Biolabs (UK) Ltd.). The resulting constructs are referred to as pΔECC1aFom-hphGFP, pΔECC1aForc-hphGFP, and pΔECC1b-hphGFP. Primers are listed in Supplementary Table S2 .

For CRISPR/Cas9-mediated ORF disruption, the same HPH-GFP cassette was used, flanked by homologous regions of 400bp around the targeted double-strand break site within the ORF. These regions were amplified using primers FP10856 and FP10857 and primers FP10858 and FP10859. The constructs were named pCRISPRΔECC1aFom, pCRISPRΔECC1aForc, and pCRISPRΔECC1b.

For ectopic complementation of ECC1 homologs, constructs were generated by Gibson cloning the promoter region (~1kb upstream of ORF), the ORF and the terminator (~400 bp downstream of ORF) upstream of a phleomycin cassette in pRW1p (Houterman et al., 2008), digested by EcoRI and HindIII (ThermoFisher Fermentas). An additional ~600bp homologous flanking region downstream of the ORF was inserted downstream of the cassette to facilitate homologous recombination. The terminator was duplicated since it could be a shared terminator for both the ECC1 gene and the neighboring predicted gene encoding a NPP1 domain-containing protein (NPP1d) ( Figure 1 ). The resulting constructs were named pECT-ECC1aFom-phleo, pECT-ECC1aForc-phleo, and pECT-ECC1b-phleo. Assembly was done using the following primer sets ( Supplementary Table S2 ): primers used for introducing the upstream and downstream region are FP11990+FP11991 and FP11992+FP11993 for ECC1a^Fom , FP11990+FP11991 and FP11992+FP11994 for ECC1a^Forc and FP11995+FP11996 and FP11992+FP11993 for ECC1b.

To enable in locus complementation, pBV1n was generated by replacing the phleomycin cassette in pRW1p with a nourseothricin resistance cassette amplified from pZPnat1 (GenBank AY631958.1), using FP12743+FP12744 and FP12756+FP12757 and assembled by Gibson cloning. This vector served as a backbone for generating pLOC-ECC1aFom-nat, pLOC-ECC1aForc-nat, and pLOC-ECC1b-nat. Inserts containing the promoter (~1kb upstream of ORF), the ORF and the terminator (~ 400 bp downstream of ORF) were cloned upstream of the cassette, while a 600bp downstream flanking region was inserted downstream, using FP12739+FP12742 and FP12745+FP12746 for ECC1a^Fom , FP12739+FP12742 and FP12745+FP12747 for ECC1a^Forc and FP12741+FP12742 and FP12745+FP12746 for ECC1b.

The pHis-parallel1-_NLSCas9 (Addgene Catalog #112065) plasmid was used to express Cas9 in BL21 GOLD (DE3) cells as described (Pokhrel et al., 2022). Protein expression was induced with 0.3 mM IPTG at 25°C for 18 hours. For purification, BL21 GOLD cells were resuspended in Cas9 Lysis Buffer (20 mM Tris–HCl, 300 mM NaCl, 10 mM imidazole, 1 mM PMSF, pH 8.0) containing ~0.1 mg/mL Lysozyme (Sigma-Aldrich) and incubated for 1h at 4°C. Cells were lysed by disrupting them four times with a French Press (ThermoFisher Scientific). The lysate was clarified by centrifugation at 50,000 x g for 1 hour at 4°C using a Beckman Coulter Avanti J-E centrifuge equipped with JA 25.50 rotor. Recombinant Cas9 was purified using ÄKTA (Cytiva) with His Trap FF (5 mL) column (Cytiva). The columns were washed with Cas9 Wash Buffer (20 mM Tris–HCl, 500 mM NaCl, 25 mM imidazole, pH 8.0) and proteins were eluted using Cas9 Elution Buffer (20 mM Tris–HCl, 500 mM NaCl, 500 mM imidazole, pH 8.0). Proteins were concentrated using Amicon Centriplus Centrifugal Filter Devices YM-100 (Millipore, 100 kDa cut-off). Amicon Ultra – 4 Centrifugal Filters Ultracel – 50K (Millipore, 50kDa cut-off) were used for the final concentrating steps. For buffer exchange, PD10 desalting columns (Cytiva) were used. Recombinant Cas9 was stored in Cas9 storage buffer (20 mM HEPES, 300 mM NaCl, 10% glycerol, pH 7.5).

sgRNA were screened for potential off-target sites using the CRISPR gRNA Design Software from Geneious v2023.2 (https://www.geneious.com) and blastn (megablast, default options, NCBI). sgRNAs were generated using the New England Biolabs EnGen^® sgRNA Synthesis Kit (S. pyogenes) as described (Pokhrel et al., 2022). Oligos generated for the in vitro RNA synthesis are given in Supplementary Table S3 .

sgRNA cleavage efficiency was checked by in vitro cleavage assay as described (Pokhrel et al., 2022), with some minor alterations in the mastermix. The mastermix consisted of 1x Cas9 nuclease buffer (20 mM HEPES, 100 mM NaCl, 5 mM MgCl₂, and 0.1 mM EDTA, pH 6.5), 0.5 μg sgRNA, 0.5 μg Cas9, 100 ng of DNA template and DEPC water to a final volume of 20 μL. Reactions were incubated for 1 hour at 37°C. After incubation, samples were treated with Proteinase K (ThermoFisher Scientific) for 10 minutes at 37°C. The cleavage activity was visualized by gel electrophoresis on a 0.8-1% agarose gel.

Protoplast isolation was performed using a protocol based on methods previously described (Brückner et al., 1992; Tudzynski et al., 1999; Janevska et al., 2018). Briefly, the Fo pre-cultures were prepared using 100 mL of Darken medium (87 mM sucrose, 7.6 mM (NH₄)₂SO₄, 0.7 g/100 mL CaCO₃, 15 g/L corn steep solids) and used to inoculate 100 mL ICI main culture medium (403 mM D-glucose•H₂O, 5.9 mM MgSO₄ • 7 H₂O, 3.6 mM KH₂PO₄, trace elements (1:500; 36 mM FeSO₄•7 H₂O, 0.6 mM CuSO₄•5 H₂O, 5.6 mM ZnSO₄•7 H₂O, 0.6 mM MnSO₄• H₂O, 0.08 mM (NH₄)₆Mo₇O₂₄•4 H₂O), 5.6 µM L-glutamine) with 0.5% (v/v) pre-culture grown for 3 days at 150 rpm and 25°C (Janevska et al., 2018). Young mycelium was harvested using Miracloth, washed with sterilized MQ and KCl/CaCl₂ Buffer (1.2 M KCl, 50 mM CaCl₂•2 H₂O), and treated with enzyme solution (4 g/L Lysing enzymes (Sigma-Aldrich), 0.2 g/L Lyticase (Sigma-Aldrich), 0.2 g/L Yatalase (Takara), 0.2 g/L Albumin Fraktion V (Merck) dissolved in KCl/CaCl₂ buffer). After filtration using glass filters (VitraPOR Por. 1/2, ROBU) and centrifugation, protoplasts were washed and resuspended in 1x STC buffer (1.2 M sorbitol, 10 mM Tris, 50 mM CaCl₂, pH 7.5) (Brückner et al., 1992). The protoplasts were then diluted to a final concentration of 2 × 10⁷ protoplasts/mL.

Fo protoplast transformation was performed using a protocol based on methods previously described (Brückner et al., 1992; Tudzynski et al., 1999; Janevska et al., 2018; Pokhrel et al., 2022). Per transformation, RNPs were assembled in a 50 µL reaction containing 1x Cas9 nuclease buffer, 20µg recombinant Cas9 and 20µg sgRNA. The mix was incubated at 37°C for 20 min. The 1:1 Cas9:sgRNA ratio was previously determined by the in vitro nuclease assay as optimal. Per transformation, 200 µl protoplasts were mixed with 50 µL of the RNPs and 300–400 ng amplified donor DNA, and incubated for 20 min at RT. The RNP mixture was transferred to 1.6 mL PEG solution (50% w/v PEG 4000, 50mM CaCl₂, 10 mM Tris, pH 7.5) (Brückner et al., 1992). After an incubation of 10 min at RT, the reaction was terminated by adding 3.2 mL 1xSTC. The transformation mixture was mixed with Regeneration Medium (RM) (700 mM sucrose, 0.5 g/L yeast extract, 20 g/L agar) and incubated O/N at RT before adding the selective layer containing antibiotics to a final concentration of 100-150 µg/mL hygromycin (Duchefa Biochemie) for the knockout mutants or 50 µg/mL nourseothricin (Jena Bioscience) or 100 µg/mL zeocin (InvivoGen) for the complementation and gene replacement strains. Edits at ECC1 loci were verified by PCR analysis ( Supplementary Figures S2 , S3 ) and Sanger sequencing.

Agrobacterium tumefaciens-mediated Fo transformation was used to obtain ectopic complementation and gene replacement strains in Fom005 and Forc016 as described previously (Takken et al., 2004; Michielse et al., 2008). Monosporic isolates of transformants were obtained on Potato Dextrose Agar (Difco) as described (Li et al., 2020b).

To test the virulence of the Fo transformants, melon seedlings (nine days old) and cucumber seedlings (seven days old) were inoculated with water (mock), Fom005 (WT), Forc016 (WT) or the ECC1 mutants at 25°C. The plants were inoculated with 10⁷ spores/mL via the root dip method described previously (Rep et al., 2004). Spores were collected of five-day-old Fusarium cultures grown in liquid NO₃ medium (0.17% yeast nitrogen base, 3% sucrose, 100mM KNO₃). The number of plants per treatment varied per plant species per replicate and is specified in the corresponding figure legends. Disease progression was assessed 14 days post inoculation (dpi) by measuring plant fresh weight and scoring disease severity using a disease index ranging from 0-4, where 0 indicated no symptoms; 1, slight discoloration (browning)/root rot symptoms, only at tip of main root; 2, discoloration or root rot symptoms and stem lesions visible aboveground, growth distortion; 3, very clear root rot symptoms of the entire root system, often with a large lesion extending above the cotyledons, severe growth distortion and wilting; 4, plant either dead or very small and wilted ( Supplementary Figure S4 ).

Statistical analyses were performed in R version 4.4.2 (R Core Team, 2024). Normality was assessed using the Shapiro-Wilk test, which indicated that data did not follow a normal distribution. Fresh weight data were analyzed using Kruskal-Wallis tests, followed by Dunn’s post hoc test with Benjamini-Hochberg correction for multiple testing. Disease severity scores were assessed using Mann-Whitney U tests with Benjamini-Hochberg correction.

Ten-day-old melon seedlings and seven-day-old cucumber seedlings were inoculated using the root-dip method as described above, with a modification: to allow sufficient tissue collection at early timepoints, the roots were trimmed to approximately 2 cm (instead of 1 cm), prior to inoculation with wild-type Fom005, Forc016 or Milli-Q (mock treatment). After inoculation, the seedlings were potted in vermiculite supplemented with nutrients. The roots of three seedlings were harvested per replicate at 2-, 4-, 7- and 10-days post-inoculation (dpi), flash-frozen in liquid nitrogen and subsequently freeze-dried.

Freeze-dried infected root material was disrupted using 4 mm metal beads in a tissue lyser (Qiagen) at 30 Hz for 2 min. The entire root system was used as input for RNA extraction. RNA was extracted using the RNeasy Plant Mini Kit (Qiagen) including Appendix D: Optional On-Column DNase Digestion with the RNase-Free DNase Set (Qiagen). An additional DNase treatment was performed with RNase-free DNase I (ThermoFisher Scientific) according to the manufacturer’s instructions. Then, 1 µg of total RNA was used for cDNA synthesis by RevertAid H Minus Reverse Transcriptase (ThermoFisher Scientific) following the manufacturer’s protocol.

Probes and primers were designed using IDT PrimerQuest™ Tool (Owczarzy et al., 2008) and Primer3Plus version: 3.3.0 (Untergasser et al., 2012). TaqMan assays were performed on a QuantStudio 3 Real-Time PCR System (ThermoFisher Scientific). The 10 µL reactions contained (final concentration): 0.25 U of DreamTaq DNA Polymerase (ThermoFisher Scientific), 1x DreamTaq buffer, 2 pmol of each probe, 5 pmol of each primer, 0.2 mM dNTPs (each), 1 µL of cDNA. Multiplex reactions were performed for targets ECC1a and ECC1b, whereas EF1α reactions were run separately (simplex). The TaqMan RT PCR program was set as follows: 2 min at 95°C; 45 cycles of [15 s at 95°C, 48 s at 68°C, 12 s at 68°C (data collection)]. Each sample was run in three technical replicates. A no-template control (NTC), where Milli-Q replaced the template, was included, as well as mock-inoculated plant samples as a second negative control. Primers and probes are listed in Supplementary Table S5 .

Previous analyses have shown that ECC1a is located 150 kb upstream of a homolog, ECC1b (Li et al., 2021). ECC1b is highly similar to ECC1a^Fom , and an identical homolog is present in Forc016. Synteny in this entire region is highly conserved between Fom and Forc ( Figure 1A ). ECC1a and ECC1b are part of a ~3.3 kb segmental duplication, which also includes a gene that encodes a secreted protein with a necrosis inducing protein (NPP1)-like domain, which we here call NPP1d. To determine whether ECC1a and ECC1b are part of a larger family and whether they are also present in other cucurbit-infecting isolates, a dataset of 149 Fo genome assemblies was compiled. Of these, 99 are from strains that are known to, or predicted to, infect a member of the cucurbits ( Supplementary Table S1 ) while the other 50 are from strains that are pathogenic on other plant species or non-pathogenic isolates isolated from soil or asymptomatic hosts. We inferred a phylogeny for the strains in this dataset, and observed, consistent with previous analyses (Sabahi et al., 2021), that f. sp. melonis, f. sp. cucumerinum and f. sp. niveum are polyphyletic, i.e. members of the same forma specialis cluster in different clade in the phylogeny ( Supplementary Figure S1 ).

We then searched for ECC1 homologs in our dataset with BLAST and found that ECC1a is part of a family of effectors that is present in many, but not all, strains that infect melon, watermelon and/or cucumber. By inferring a gene tree of the ECC1 gene family, it was found that ECC1 homologs can be grouped into four subfamilies, where ECC1a^Fom and ECC1b belong to the same subfamily, but ECC1a^Forc does not ( Figure 2 ). Not counting sequences that are disrupted by assembly errors (i.e. located at start or end of contigs, or interrupted by an assembly gap), these four subfamilies could be further subdivided into 13 different sequence types: clades in which the phylogenetic distance between members is zero (indicated with number 1-4c in Figure 2 ). Some sequence types are specific to a single forma specialis. Subfamily 4, which includes ECC1a^Fom and ECC1b, is present only in f. sp. melonis, except for the ECC1b gene that is present in Forc016. In contrast, subfamily 2, that includes ECC1a^Forc , is present in f. sp. melonis, niveum and cucumerinum, and both f. sp. melonis and f. sp. cucumerinum strains carry the exact ECC1a^Forc genotype. Surprisingly, no copy of ECC1a^Forc or ECC1b was found in the assemblies of Forc031 and Forc024, while these are very closely related to Forc016 and have the same host range. Closer inspection revealed that these assemblies carry partial copies of ECC1a^Forc /ECC1b that correspond to the parts that are identical between these genes, interrupted by a gap in the assembly. This suggests that detection of ECC1 failed due to an assembly error: collapse of the 3.3 kb segmental duplication that ECC1a and ECC1b are located on. Notably, strains that share an ECC1 sequence type are not necessarily phylogenetically closely related, suggesting that these genes have transferred horizontally between cucurbit-infecting strains ( Figure 2 ; Supplementary Figure S1 ). Based on these phylogenetic analyses, we predict that ECC1a^Fom and ECC1b are important for infection of melon, given the fact that they are present in most melon-infecting strains in our dataset. Moreover, we predict that ECC1a^Forc may contribute to virulence towards melon and cucumber, since it is present in multiple distinct lineages that group into these formae speciales, and it is highly expressed in Forc016 during infection of cucumber (Van Dam et al., 2017; Li et al., 2021).

To obtain more information on the potential function of ECC1 homologs, protein structure predictions were generated for all three ECC1 homologs using AlphaFold3 ( Figure 3 ). The resulting models revealed a β-sandwich fold, which is a characteristic structural feature of ToxA-like effectors. Structural similarity between ECC1a^Fom and ToxA (from Pyrenophora tritici-repentis), as well as known ToxA-like Fo effectors (FOXGR_015533, SIX7, SIX8 and Avr2 (SIX3)) was assessed using the DALI server resulting in Z-scores of 7.4 (ToxA), 7.5, 6.8, 6.6 and 5.5, respectively. Since Z-scores between 2 and 8 are generally indicative of structural homology (Holm, 2020), these results suggest that ECC1^Fom is structurally related to the ToxA and the ToxA-like effector family. Interestingly, when reviewing the protein sequence alignment ( Figure 1C ), potential Kex2 processing sites (LxxR motif) were found in the sequence of the three homologs. Kex2 sites have been found before in fungal ToxA-like effectors and in other effectors from Fo (Outram et al., 2021). These results suggest that ECC1 may be part of the ToxA-like effector family and could be Kex2 pro-domain-processed (K2PP).

To study the function of ECC1a^Fom, ECC1a^Forc and ECC1b in different genetic backgrounds, single- and double knockout and complementation strains were generated, and strains in which a knockout of ECC1a was complemented with the ECC1a gene from the other forma specialis. Initially, using PEG-mediated protoplast transformation, ECC1 open reading frame (ORF) deletion mutants were generated in Fom005 and Forc016. From this transformation event only three out of the 57 transformants screened contained the desired ORF deletion ( Supplementary Table S4 ). To increase efficiency, we adapted an RNP-based CRISPR-Cas9 approach to generate ORF disruption mutants in Fom005 and Forc016. In addition, this strategy enabled efficient generation of double knockout mutants. An overview of the mutants is given in Supplementary Table S4 .

Next, to verify that any observed phenotypes in the knockout mutants are caused by the respective gene deletions, we also generated complementation strains of the single knockout mutants by reintroducing the native gene either at the original locus or ectopically. To generate in locus complementation strains (complemented with the endogenous native gene) and gene replacement strains (complemented with the homolog from the other forma specialis), we again employed the CRISPR/Cas9 system. Ectopic complementation and gene replacement strains were generated via Agrobacterium tumefaciens-mediated transformation (ATMT). Together, for each type of single knockout, two to four complementation or gene replacement strains were generated. Finally, we tested the virulence of all these mutants towards melon and cucumber in disease assays.

To investigate whether ECC1a or ECC1b impact virulence of Fom on cucumber by Fom, we compared the virulence towards cucumber of knockout mutants in Fom005 with that of the wild- type strains. No obvious differences in growth or colony morphology were observed for the ECC1 deletion mutants under standard in vitro culture conditions. Virulence was quantified by scoring disease severity ( Supplementary Figure S4 ) and measuring fresh weight, and representative pictures of the plants were taken ( Supplementary Figures S5 – S8 ). Overall, there were no significant differences in disease severity between cucumber plants inoculated with ORF deletion mutants and ORF disruption mutants ( Supplementary Table S4 ). As expected, wild-type Fom005 was non-pathogenic on this host ( Figures 4A, C ). As ECC1a^Fom acted as a non-host avirulence gene in Forc016 (Li et al., 2021), we reasoned that deletion of ECC1a^Fom in Fom may result in an increase in virulence towards cucumber. However, neither the single nor the double knockout mutants of Fom005 showed a significant reduction in plant fresh weight or disease symptoms on this plant species ( Figures 4A, C ). Together, these results indicate that ECC1 homologs alone are not sufficient to restrict the host range of Fom.

To determine whether ECC1a^Forc and ECC1b play a role in virulence of Forc016 towards cucumber, the same setup was used to test the impact of single and double knockout mutants on disease. As anticipated, wild-type Forc016 caused severe disease symptoms and significantly reduced fresh weight in cucumber plants compared to mock ( Figures 4B, D ). In contrast to results observed in Fom005, fresh weight of plants inoculated with ECC1a^Forc or ECC1b single and double knockout mutants significantly differed from those observed for the Forc016 wild type. Moreover, disease symptoms were consistently significantly less severe in all knockout mutants relative to those of the wild-type strain. The ECC1 double knockout mutants did not consistently show a larger reduction in virulence on cucumber as compared to the single knockout mutants. Together, these results suggest that both ECC1a^Forc and ECC1b contribute to cucumber infection by Forc016.

To verify that the observed phenotype is specifically due to inactivation of the targeted gene, knockout strains were complemented with the original gene. Complementation of Forc016 ECC1a knockout strains with ECC1a^Forc restored virulence on cucumber to varying degrees ( Figures 5E–H ). In locus complementation fully restored wild-type levels of disease symptoms and fresh weight consistently throughout several repetitions, whereas ectopic transformation resulted in either full or partial restoration of virulence ( Supplementary Figures S9 , S11 ). In contrast, despite consistent reduction in virulence for several independent Forc016ΔECC1b mutants ( Figures 5B, D ), complementation with ECC1b, either ectopically or in locus, failed to restore the wild-type phenotype to Forc016ΔECC1b-1. We conclude that, while results are mixed for ECC1b, ECC1a^Forc contributes to virulence towards cucumber in Forc016.

Having established that ECC1a^Forc contributes to virulence on cucumber of Forc016, we then asked whether replacing ECC1a ^Fom with ECC1a ^Forc in a Fom strain would result in gain of pathogenicity to cucumber. Complementation of Fom005ΔECC1a-1 with ECC1a^Forc did not lead to disease development ( Figures 5A–D ). This indicates that, while ECC1a^Forc contributes to virulence in Forc016, its presence in the Fom005 background is not sufficient to gain pathogenicity towards cucumber. In contrast, replacing ECC1a^Forc with ECC1a^Fom in locus in Forc016 significantly reduced virulence towards cucumber ( Figures 5E, G ), which corresponds with results from a previous study in which ECC1a^Fom was ectopically introduced into a Forc016 background (Li et al., 2021). Together, these data show that ECC1a^Forc contributes to Forc virulence towards cucumber and confirm that ECC1a^Fom acts as an avirulence factor for cucumber.

Next, we investigated whether ECC1a and ECC1b contribute to disease on melon by testing severity of root rot and wilt symptoms in disease assays using single and double knockout mutants in Fom005 and Forc016. As in the cucumber infection assays, there were no significant differences in fresh weight and disease severity between melon plants inoculated with ORF deletion mutants and ORF disruption mutants. Fom005 mutants lacking ECC1a, or ECC1b, or both homologs, were significantly less virulent than the wild-type strain ( Figures 6C, D ). Most Forc016 ECC1a knockout strains remained virulent on melon, but Forc016 ECC1b knockout mutants showed reduced symptom severity compared to Forc016 wild type ( Figure 6D ). As expected, based on these results, double knockout mutants in Forc016 showed a similar phenotype as single ECC1b knockout mutants. Taken together, these results indicate that, except for ECC1a^Forc, ECC1 homologs in both Forc and Fom contribute to virulence on melon.

To confirm that the reduced virulence of the ECC1a and ECC1b knockout mutants was due to gene deletion or disruption, we assessed whether complementation of the single knockout mutants would restore virulence ( Figure 7 ). Complementation of Fom005ΔECC1a-1 with ECC1a^Fom only partially restored virulence: plant fresh weight was comparable to that of the knockout strain ( Figure 7A ), and only one strain with an ectopic insertion of ECC1a (Fom005ΔECC1a-1+ECC1aFom-2) induced more severe disease symptoms than the Fom005ΔECC1a-1 background ( Figure 7C ). Complementation of Fom005ΔECC1b-1 with ECC1b^Fom also partially restored virulence: only the Fom005ΔECC1b-1+ECC1b-1 strain showed full restoration of virulence ( Figures 7B, D ), and other complementation strains showed intermediate phenotypes: more severe than the knockout, but less severe than wild-type. On melon, Forc016ΔECC1a-1 retained its virulence as observed previously ( Figure 6 ), and reintroducing ECC1a^Forc had no significant effect on this phenotype ( Figures 7E, G ). Although independent Forc016ΔECC1b mutants showed consistent loss of virulence ( Figure 6 ), introducing ECC1b failed to restore the loss of virulence to wild-type phenotype to Forc016ΔECC1b. Therefore, the exact role of ECC1b in Forc016 infection of melon remains unclear. In contrast, both ECC1a and ECC1b appear to contribute to Fom005 virulence on melon.

ECC1a^Fom and ECC1a^Forc are identified as each other’s orthologs based on synteny conservation and therefore could be predicted a priori to have a conserved function. On the other hand, their high sequence divergence indicates functional diversification of these orthologs, e.g. with respect to a role in virulence in a specific host. To test this, ECC1a was replaced with ECC1a from the other forma specialis, introduced either in locus and/or ectopically, and virulence of these mutants on melon was assessed. Replacement of ECC1a^Forc with ECC1a^Fom in the Forc016ΔECC1a-1 background yielded inconsistent results: the in locus replacement strain remained virulent, whereas the ectopic strain showed reduced symptom severity ( Figures 7E,G ). Conversely, replacing ECC1a^Fom with ECC1a^Forc in Fom did not restore virulence on melon ( Figures 7A, C ). These results indicate that ECC1a^Fom is not functionally interchangeable with ECC1a^Forc in melon infection and that ECC1a^Fom has a forma specialis-specific virulence function.

Having found that ECC1 homologs contribute to virulence of Fom and Forc on their respective hosts, we next investigated whether they are expressed during different stages in infection. To assess this, transcript levels of ECC1a and ECC1b relative to the Fo housekeeping gene Translation Elongation Factor 1 alpha (EF1α) were quantified using qPCR at 2-, 4-, 7- and 10-days post inoculation (dpi) of melon and cucumber. While the knockout and complementation assays indicated that ECC1b contributes to virulence of Fom005 to melon, ECC1b expression was not detected during melon infection. In contrast, ECC1a^Fom was expressed and expression peaked at 4 dpi ( Figure 8 ). As Fom005 can colonize cucumber plants, albeit without causing disease, expression of ECC1 homologs during cucumber infection was also assessed. It was found to be comparable to that of melon infection, with similar relative expression levels at its peak at 4 dpi. This indicates that the lack of impact of ECC1a^Fom on virulence on cucumber is probably not due to a lack of expression.

In contrast, both ECC1a^Forc and ECC1b were expressed during Forc016 infection of melon and cucumber. ECC1a^Forc and ECC1b are expressed during early stages of infection: transcript levels peaked at 4 dpi and returned to very low basal levels at 10 dpi. This suggests ECC1a^Forc and ECC1b play a role in initial host colonization by Forc016.