Symptomatic nuts (Tonda Gentile Romana cv) were collected from a mature hazelnut orchard located in the Viterbo area (VT) (Latium, Italy; latitude 42°16′00.0″, longitude 12°17′00.0″, altitude 275 m). Small fragments of approximately 2 mm² were taken from the nut surface and placed onto potato dextrose agar (PDA) plates supplemented with 0.2 g⋅L^–1 streptomycin sulfate. The Petri plates were then incubated at 25°C until the fungal colonies were grown enough to be singularly further transferred onto new PDA plates to finally obtain monosporal isolates.

For morphological characterization, the cultures were grown on Spezieller Nahrstoffarmer agar (SNA) (Leslie and Summerell, 2006). After 10 days, the cultures were examined using a Nikon SMZ128 stereomicroscope (Tokyo, Japan), and the images were captured using Alexasoft TPS5000H CMOS camera (Florence, Italy). For microscopic analysis, the samples were prepared by mixing with lactophenol blue dye using Leica DM6 B optical microscope (Wetzlar, Germany), the images were captured using Leica DFC 7000 T camera (Wetzlar, Germany), elaborated using Leica Application Suite X program (version 4.12) (Wetzlar, Germany), and all the morphological characteristics were evaluated according to The Fusarium Laboratory Manual (Leslie and Summerell, 2006).

For each isolate, about 100 mg of mycelium was put into a sterile 2 ml microtube with 1 ml of lysis buffer (1 mM EDTA pH 8, 10 mM Tris–HCl pH 8, 100 mM NaCl, SDS 1%) for DNA extraction. After 5 min of centrifugation at 12,000 rpm, the supernatant was transferred into a new microtube, and the DNA was precipitated with isopropanol and 70% ethanol. To characterize the isolates at the genus/species level, two genes were amplified using two different pairs of primers, one targeting the internal transcribed spacer (ITS) (White et al., 1990) and the second one targeting the Fusarium specific translation elongation factor 1-alpha (Edel-Hermann et al., 2015), and the amplicons were Sanger sequenced at Eurofins genomics (Eurofins Genomics GmbH, Konstanz, Germany).

To confirm that the isolate in subject was responsible for the observed symptoms on fruits, the following pathogenicity tests were carried out. For inoculum preparation, the PT isolate of the Fusarium sp. was grown on SNA at 25°C for 10 days (Leslie and Summerell, 2006). Then, conidia were scraped from the fungal cultures and filtered through layers of cheesecloth to remove mycelial fragments. The resulting conidial suspension was quantified using a hemocytometer and diluted to a final concentration of 10⁶ spore/ml. Inoculations were performed using two techniques: (i) inoculation of young to fully formed fruits with 20 μl of conidial suspension (10⁶ conidia/ml) dropped between nut and bracts; and (ii) inoculation of the hazelnut mesocarp with 20 μl of conidial suspension (10⁶ conidia/ml). Control fruits were inoculated with water. Inoculated and control fruits were placed in a moist chamber at room temperature for 12 days. The appearance of any symptoms was monitored daily.

The fungal culture was initiated using ∼1 × 10⁶ conidia inoculated in 100 ml of Czapek Dox Yeast broth (Leslie and Summerell, 2006) and grown for 3 days at 25°C. Next, the genomic DNA was extracted following a modified Cetrimioum bromide (CTAB) method (Graham and Henry, 1997) and performed as described below.

The total mycelium grown in liquid media was lyophilized using a vacuum pump. An aliquot of 250 mg of lyophilized mycelium was fine grounded using a pestle and mortar in liquid nitrogen and transferred into a 2 ml tube. Then, 500 μl of CTAB buffer (10% CTAB, 25 mM EDTA pH 8.0, 200 mM Tris–HCl pH 8.0, 2.50 M NaCl) and 5 μl proteinase K (20 mg/ml) were added to the sample and the solution was gently mixed and incubated overnight at 56°C. After 5 min of agitated incubation with 500 μl of phenol:chloroform:isoamyl alcohol (25:24:1) and 10 min of 12,000 rpm centrifugation, the aqueous clear phase was transferred to a new tube and incubated with 500 μl of chloroform:isoamyl alcohol (24:1). After 10 min of centrifugation at 12,000 rpm, the aqueous phase was collected into a new tube and 0.6 VOL of cold isopropanol was added, followed by 30 min of centrifugation at 12,000 rpm for the DNA precipitation. Two cleaning/precipitation steps using 1 ml of 70% cold ethanol were performed by 10 min of centrifugation at 12,000 rpm. The pellet was dried, resuspended into a 100 μl of ultrapure sterile water, and treated with RNase at 37°C for 30 min. The DNA integrity was evaluated on a 1% agarose gel electrophoresis run, whereas the DNA purity was checked using Nanodrop™ spectrophotometer (Thermo Fisher Scientific). DNA sequencing was performed on a MinION Mk1b device (Oxford Nanopore Technologies, ONT, United Kingdom) using a R9.4.1 Flow Cell (ONT), after library preparation using the SQK-RBK004 Rapid Barcoding Kit (ONT). An aliquot of the same DNA sample was sequenced at Eurofins Genomics (Eurofins Genomics GmbH, Konstanz, Germany) with the genome sequencer Illumina NovaSeq 6000 S2 using the paired-end sequencing.

Illumina reads quality was evaluated using FastQC (Andrews, 2010), NovaSeq 6000 adapters were trimmed using Trimmomatic version 0.39 (Bolger et al., 2014), and low-quality reads were removed using Sickle (Joshi and Fass, 2011). To get the draft genome sequence, four different assemblers were used separately: SPAdes version 3.11.1 (Nurk et al., 2013), Minimap2 version 2.12-r849-dirty (Li, 2018) in combination with Miniasm version 0.3-r179 (Li, 2016), MaSuRCA version 3.4.2 (Zimin et al., 2013), and Canu version 2.1.1 (Koren et al., 2017). SPAdes was used in a first trial using only Illumina reads and in a second run for a hybrid assembly using both Illumina and Nanopore reads. A hybrid assembly was also performed using MaSuRCA, whereas Miniasm-Minimap2 and Canu used only Nanopore reads. For all the assemblers, default parameters were used, except for the expected genome size that was set at 40 Mb in Canu.

The obtained assemblies were further polished using Nanopolish version 0.11.1 (Loman et al., 2015), Racon version 1.3.3 (Vaser et al., 2017), or Pilon version 1.23 (Walker et al., 2014). The assembly quality statistics, before and after polishing, were evaluated using QUAST version 5.0.2 (Gurevich et al., 2013), while BUSCO version 5.beta.1 (Simão et al., 2015) was used to assess the assembly completeness, using hypocreales_db10 as ortholog lineage dataset, which consists of a set of 4,494 conserved profiles.

The draft genome assembled using Canu, chosen for downstream analysis, was structurally annotated following the de novo MAKER pipeline version 3.01.03 (Holt and Yandell, 2011), using the built-in RepeatModeler to mask repetitive elements, SNAP and AUGUSTUS for an ab initio gene prediction, and Est2Genome and Protein2Genome to further refine introns and exons boundaries using Exonerate and tRNAscan-SE to identify the genes related to the tRNA biosynthesis. Transcripts and proteins concatenated from four closely related Fusarium species (F. fujikuroi, F. graminearum, F. oxysporum f. sp. lycopersici, and F. verticillioides) were given as gene models, and the maximum intron size was set as 2,500. Functional annotation was performed using BLASTp and SwissProt as a database.

The same annotation pipeline was applied to annotate the genomes of F. culmorum, F. circinatum, F. oxysporum f. sp. koae 44, F. pseudograminearum CS3270, F. solani IlSc-1, F. tricinctum INRA104, F. tricinctum NRRL25481, F. tricinctum T6, F. verticilloides BRIP 53263, BRIP 53590, and all the F. avenaceum isolates, except for the already annotated Fa05001 used for the following comparative genomics analysis.

The genome sequence of 27 others Fusarium species, together with their annotated proteins when available, were downloaded from the NCBI genome databases and used for phylogenetic and comparative genomics analysis (Table 1). Notably, the only F. lateritium genome sequence available in the NCBI database under the accession number GCA_014898835.1 refers to a strain isolated from a symptomatic elm tree in Pineville, Louisiana, United States (Kim et al., 2020).

For a phylogenetic analysis, the gene sequences of six housekeeping genes (EF-1α, RPB1, RPB2, beta tubulin, ITS, and LSU) were extracted from each genome and concatenated and aligned using MUSCLE (Edgar, 2004). The resulting alignment was used to build a maximum likelihood (ML) tree using raxmlHPC (Stamatakis, 2014) and visualized in a dendrogram using FigTree version 1.4.4³. A second ML phylogenetic tree was built using raxmlHPC on the core genome SNPs identified during a pan genome analysis performed using Panseq, with the run mode set to pan, the fragment size at 500 nucleotides, the percentage of identity at 90%, and the core genome threshold to 28 genomes, in order to find out the sequences in common among all the strains (Laing et al., 2010). To better discriminate between isolates belonging to the F. tricinctum species complex (FTSC), a phylogenetic ML tree was built on the alignment of the concatenated sequence of the DNA-directed RNA polymerase II largest (RPB1) and second largest subunit (RPB2) nucleotide sequence, as previously applied by O’Donnell et al. (2013); Ponts et al. (2020), and Crous et al. (2021). MUSCLE and RAxML were used as previously described, and the 63 Fusarium isolates used in this analysis are reported in Supplementary Table 1.

Orthologous proteins were identified using OrthoFinder, and the results were used to build a species tree, visualized using FigTree as well (Emms and Kelly, 2019). The average nucleotide identity (ANI) analysis was performed using the pyani script and ANIb as algorithm for the alignment (Pritchard et al., 2016).

The putative secreted proteins involved in pathogenesis were identified using SignalP version 5.0b with the cutoff set ≥ 0.5 (Almagro Armenteros et al., 2019a). TargetP version 2.0 was used to identify signal peptide (SP), mitochondrial and chloroplast transit peptide (mTP and cTP, respectively), and potential cleavage sites (CS) (Almagro Armenteros et al., 2019b). Prediction of transmembrane proteins was performed using TMHMM version 2.0 (Krogh et al., 2001). Pathogen Host Interactions database (Urban et al., 2019) was used to find the similarity with pathogenicity and virulence-related genes, experimentally tested for roles in pathogenicity. Instead, Carbohydrate-Active Enzyme (CAZy) database (Lombard et al., 2014) was used to identify families of enzymes related to degradation, modification, and creation of glycosidic bonds, focusing on the Cell Wall Digestion Enzymes (CWDE).

The biosynthetic gene clusters (BGCs) were automatically searched and analyzed using AntiSMASH6 (Blin et al., 2021). The presence of the emerging mycotoxin enniatin was verified by blasting the available gene sequences from several Fusarium species (EF029060.1, NW_022158785.1, NW_022158526.1, KP000028.1, NC_030995.1, ENA| Z18755, NW_023502434.1, NW_023501408.1, and NW_023501343.1) (Fraeyman et al., 2017; Ponts et al., 2018).

A Fusarium sp. strain was isolated from symptomatic nuts collected from a hazelnut field located in the VT (Italy), which was being monitored in the framework of the PANTHEON project and named as Fusarium sp. isolate PT (Figures 1A,B). The morphological analysis performed to look at both macroscopic and microscopic characteristics revealed a mycelia color variable from white in the first 3–4 days to pale orange. Abundant medium-long, thin macroconidia with walls parallel for most of the spore length were observed, in the absence of microconidia and rare monophialides (Figures 1C,D). The characteristics found on the sample resembled the ones from the F. lateritium species when compared with The Fusarium Laboratory Manual (Leslie and Summerell, 2006). The sequenced ITS and the EF-1α region showed the highest identities (98.22 and 99.69% respectively) with F. lateritium sequences when blasted in the NCBI database (Supplementary Table 2).

All fruits except the negative control developed symptoms similar to the one observed in the field (Figure 2). The Fusarium sp. isolate PT was consistently re-isolated from the inoculated nuts but not from the control fruits. The fungus re-isolated from diseased fruits showed the same morphology of the original isolate as well as the ITS, LSU, and β-tub, and EF-1α sequences were identical to those previously obtained, thus confirming Koch’s postulates.

A clean and high molecular weight DNA was obtained, with both 260/230 nm and 260/280 nm ratios falling between 1.8 and 2 and was further sequenced using both ONT and Illumina technologies. ONT MinION run produced ∼701 k reads (1.92 Gbp; ∼73 × coverage), with a mean read length of 2,740 bp and an N50 of 4,588. Illumina NovaSeq 6000 S2 sequencing produced 2 × ∼2.5 M reads (2 Mbp × 385 Mbp; ∼20 × coverage).

After low-quality reads and adapters removal done by sickle and trimmomatic, respectively, the total number of the paired-end Illumina reads resulted to be 4.8 M in total, with an average length of 151 bp and a GC content of 47%. Instead, the average length of the Nanopore reads, affected by the starting DNA fragments material, resulted in an average length of 2,740 bp, with a GC content of 47%.

The reads were assembled following several pipelines and, thus, different algorithms whose results are described by QUAST statistics (Table 2). SPAdes draft genome derived either by only Illumina reads or by a hybrid assembly between Illumina and Nanopore reads, resulted to be the one with the lowest N50, the richest in undetermined bases N (171 and 48, respectively) and assembled in the highest number of contigs (70), and thus, it was discarded for further downstream analysis. The genome assembly derived from Miniasm-Minimap2 without any polishing yielded 70 contigs and 40.28 Mb of genome length with an N50 of 1.028 Mb that slightly increased to 1.029 Mb after five iterative polishing steps. MaSuRCA assembly resulted in 53 contigs arranged in a total length of 40.44 Mb and an N50 of 1.49 Mb. One step of Pilon polishing seemed to be unnecessary since the quality did not change much, proving that MaSuRCA itself with its POLCA polishing step is already enough to reach a good quality genome. Overall, Canu assembly resulted to be the one with the best statistics, with a total genome length of 40.51 Mb arranged in only 27 contigs.

Then, the completeness of the three best quality assemblies was further evaluated by looking for the presence of 4,494 conserved ORF among the Hypocreales order (BUSCO). In fact, the polishing steps were needed to increase the quality and completeness of each of the three assemblies as well as to reduce the fragmented or missing BUSCO, leading to 99.6, 99.8, and 99.5% of completeness, respectively (Supplementary Dataset 1).

Based on the assembly statistics results, the draft genome assembled using Canu was chosen for downstream analysis. Its sequence was further verified by mapping the Illumina reads back on the 27 contigs used as reference to finally reach a high-quality consensus sequence (BWA version 0.7.17-r1188, samtools version 1.2, Burrows and Wheeler, 1994; Li et al., 2009). The genome was then annotated following MAKER pipeline and deposited on the NCBI genome database under the accession number JAHMRZ000000000.

The aligned concatenated sequences of EF-1α, RPB1, RPB2, beta tubulin, ITS, and LSU (around 7,100 nucleotides) of the Fusarium isolates under comparison showed clear differences perfectly represented by the ML tree in Figure 3A. Interestingly, our Fusarium sp. isolate PT clustered together with the FTSC and not directly with the other F. lateritium strain, as one could expect from the symptoms associated to the NGN. This clustering is due to the low sequence similarity, ranging from 82 to 86% of similarity, respectively, between the EF-1α, RPB1, RPB2, and beta-tubulin genes of the sequenced Fusarium sp. isolate PT and the F. lateritium, even though the ITS and LSU sequences were 100 and 98% identical (Supplementary Dataset 2). On the contrary, the sequence similarity of these genes with the ones from the F. tricinctum species complex were, indeed, higher but never reaching the 100% similarity (Supplementary Dataset 2).

Accordingly, to further disentangle the phylogenetic relationship among the Fusarium strains, a ML tree was built on 4,319 SNPs found in the core genome alignment derived from Panseq. As shown in Figure 3B, the clustering of our Fusarium sp. isolate PT within the FTSC was confirmed. The same clusterization was obtained when the orthologous proteins were identified and a species tree was built using OrthoFinder (Figure 3C). The average nucleotide identity (ANI) performed using blastn gave an overall picture of the sequence identity between the Fusarium strains under comparison, as shown by the heatmap in Figure 3D, with the Fusarium sp. isolate PT again among the FTSC isolates.

Finally, a selection of 63 Fusarium isolates, belonging to the FTSC for which both RPB1 and RPB2 nucleotide sequences were available, was used to build an additional ML phylogenetic tree. As in the previous trees, the Fusarium sp. PT isolate constitutes an independent branch within the dendrogram (Supplementary Figure 1).

It is well-known that proteins are secreted in many Fusarium species during the colonizing stages (Ma et al., 2010), and for this reason, we went looking for the transcripts involved in pathogenesis.

In fact, 564 analogous genes involved in pathogenicity were identified from the PHI database. Among these, 24 genes were detected to be putative secreted proteins using SignalP. The 408 carbohydrate-active enzymes were detected, approximately 220 of which encoding glycoside hydrolases (GHs) and 79 genes encoding glycosyltransferases (Gts) (Table 3). Among the 1,288 proteins with a signal peptide, TargetP identified 525 proteins with a mitochondrial transit peptide, whereas 1,813 proteins resulted to have a possible cleavage site (Supplementary Figure 2). The predicted transmembrane helices found using TMHMM are reported in Supplementary Dataset 3.

Seventeen biosynthetic gene clusters belonging to T1PKSs, NRPs, and terpene synthases (Tss) were found using antiSMASH 6. Among the T1PKSs gene cluster, the fujikurin A, B, C, and D were found with 83% of similarity, bikaverin with 42 or 57% of similarity, and fusarielin H and oxyjavanicin with 50% of similarity. Besides gibepyrone A with 40% of similarity, the ACT-Toxin II shared 100% identity with the sequence deposited in the database. The NRP gene cluster is represented mainly by ilicicolin H (50% of similarity) and chrysogine (83%), whereas BGCs belonging to bassianolide, beauvericin, and fusariodione A showed a similarity below 20%. Finally, the koraiol and α-acorenol terpenes showed 100% of similarity, whereas squalestatin S1 and gibberellin were of 40% similarity.

The complete sequence of the enniatin gene was also identified, sharing 85 and 86% of similarity with the gene deriving from F. scirpi and F. tricinctum strain INRA104, respectively (Supplementary Table 3).