Comparative genome analysis identifies necrosis-inducing effectors in the lentil pathogen Ascochyta lentis in isolates with distinct virulence profiles

Abstract

Background

Ascochyta blight of lentil is a devastating fungal disease caused by the hemibiotrophic fungal pathogen Ascochyta lentis. This disease leads to significant yield loss and reduced seed quality when not managed properly. Fungal pathogens employ a suite of effectors to manipulate their host. Here, we present telomere-to-telomere assemblies, genome annotations, and comparative genome analysis of two A. lentis isolates, AlKewell and P94-24, with opposing virulence profiles to identify and characterise effector-like elements.

Results

The two genomes are highly syntenic and only contain a small number of structural variations, mostly limited to repeat-rich regions, more specifically long terminal repeat retrotransposons. Protein coding genes were ranked for their effector likelihood using Predector. The top 12 candidate effectors were characterised by agroinfiltration into Nicotiana benthamiana, of which 11 candidates induced necrotic lesions including the two candidates present only in AlKewell. Another major difference is the presence of Penelope-like elements of type Coprina at several chromosomal ends in the genome of P94-24 and their absence in AlKewell. AlKewell on the other hand contains two biosynthetic gene clusters which are heavily truncated in P94-24 and are predicted to encode genes involved in the synthesis of known fungal toxins, one of which is an effector candidate that induced necrosis on N. benthamiana.

Conclusions

These genomic resources present new interesting effector candidates and other putative pathogenicity elements that should advance our understanding of the complex lentil-pathogen interaction.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-026-12790-z.

Background

Ascochyta blight (AB) is a fungal disease affecting many cool-season legumes including chickpea (Cicer arietinum), field pea (Pisum sativum), faba bean (Vicia faba), and lentil (Lens culinaris). In lentil, AB is caused by the hemibiotrophic host-specific fungal pathogen Ascochyta lentis (teleomorph: Didymella lentis) resulting in necrotic lesions on all above ground parts of the plant [1, 2]. A. lentis is a host specific fungal pathogen that has been found to only infect lentil, and is the sole pathogen responsible for lentil AB [1]. It spreads via infected seed and plant debris, rain splash dispersal of asexual conidia, and wind dispersal of ascospores after mating [3]. Infected leaves are often shed, and severely infected plants die due to lesions girdling the stem. Infected pods can lead to seeds that are unable to germinate or act as inoculum for seedling infection. Altogether AB leads to severe yield and grain quality loss [4, 5]. In 2022, Australia was the second largest exporter of lentil with 1 Mt and the third largest producer, making AB an economically important disease [6, 7]. In Australia, A. lentis is responsible for annual control costs and losses of AU$15.3 M and can lead to up to 40% crop loss, although the economic loss due to seed staining can be up to 70% of the value [3, 4, 8, 9].

Plant pathogenic fungi follow a variety of lifestyles to colonise their host, and they use different host colonisation strategies to acquire nutrients for fungal growth. Biotrophic fungal pathogens like Cladosporium fulvum, and Ustilago maydis establish close relationships with their respective hosts and divert nutrients from living plant cells [10]. Biotrophs are known to secrete avirulence protein effectors to suppress their host’s immune response. In resistant hosts with the corresponding R (resistance) protein this leads to effector-triggered immunity (ETI) and activation of the hypersensitive response (HR), localised cell-death triggered by effector recognition [11]. Necrotrophic fungal pathogens like Pyrenophora tritici-repentis and Parastagonospora nodorum however, kill host cells by secreting necrotrophic effectors into the apoplast and live off the released nutrients and degraded plant cell components. Secreted molecules include carbohydrate- and cell wall-degrading enzymes, broad host proteinaceous effectors such as necrosis and ethylene-inducing peptide-like protein (NLP), and host-specific proteinaceous effectors such as ToxA [12–14]. Hemibiotrophs such as Colletotrichum higginsianum and Leptosphaeria maculans have an early biotrophic phase to establish plant infection and then switch to a necrotrophic lifestyle [15, 16]. Due to the recent discovery of the avirulence gene AlAvr1 in A. lentis and the delayed onset of necrotic lesions in lentil, this pathogen is described as a hemibiotroph [2].

Recent advances in sequencing technologies have revolutionised biology research resulting in high quality genome assemblies of non-model organisms. Sequencing technologies such as Oxford Nanopore routinely produce long reads in the 10 s to 100 s of kilobase range that bridge repetitive and AT-rich regions typical for many fungal genomes and generate telomere-to-telomere (T2T) assemblies. The proportion of the genome consisting of repeats varies significantly between species. Genome analysis of 18 plant pathogenic fungi reported a range of repeat content from 0.02% to 29.8% [17] and it can be as high as 80.04% in the case of the cucurbit powdery mildew pathogen Podosphaera xanthii [18] or 93% as in the Asian soybean rust fungus Phakospora pachyrhizi [19]. Several fungal effectors have been found within or in close proximity to repetitive regions, specifically regions that contain transposable elements [20]. The wheat virulence gene ToxA for example has been horizontally transferred between three wheat pathogens, P. nodorum, P. tritici-repentis, and Bipolaris sorokiniana, and was found to be part of a transposon named ToxhAT [21], which itself has recently been found to be part of a much larger and recently discovered class of giant transposable elements called Starships [22]. Repeat-rich genomic regions are problematic to assemble from short reads alone and result in fragmented assemblies which then make it difficult to study potential effectors in their proper genomic context.

We have been taking advantage of two A. lentis isolates that represent two distinct pathotypes to identify pathogenicity factors responsible for host manipulation. AlKewell is pathogenic on lentil cultivar PBA Hurricane XT and PBA Bolt, but avirulent on Nipper, while P94-24 is virulent on cultivar Nipper but avirulent on PBA Hurricane XT and PBA Bolt (Fig. 1) [2]. We have previously discovered the avirulence gene AlAvr1 which comes in two isoforms, AlAvr1-1 in P94-24 and AlAvr1-2 in AlKewell. We showed that AlAvr1-1 acts as an avirulence effector for lentil cultivars PBA Hurricane XT and PBA Bolt, and that P94-24 knockout strains for AlAvr1-1 become virulent towards PBA Hurricane XT. Knockouts of AlKewell AlAvr1-2, however, do not lead to virulence of the isolate on Nipper [2, 23].

Fig. 1.

Lentil cultivars Nipper, PBA Hurricane XT, PBA Bolt infected with Ascochyta lentis isolates AlKewell (top panel) and P94-24 (bottom panel). Disease symptoms on the susceptible cultivar appear as brown lesions on leaves, pods and stem often with black spots or dark fruiting bodies in the centre

Here we set out to generate further genomic resources for the two A. lentis isolates that represent both pathotypes. We generated T2T assemblies for both isolates using the Oxford Nanopore platform and polished the resultant assemblies with Illumina short reads. Gene prediction was performed using various tools together with in vitro and in planta RNA-seq data. Final gene models were manually curated and used for effector ranking, and effector activity validated by agroinfiltration into Nicotiana benthamiana. The complete genome assemblies led to the prediction of several potential pathogenicity factors and other elements that could help our understanding of the mechanisms of fungal pathogenesis for A. lentis.

Methods

Fungal culture and DNA extraction

Ascochyta lentis isolates AlKewell (collected near Kewell, Victoria, Australia in 2001) and P94-24 (collected near Turretfield, South Australia, Australia in 1994) were grown in Yeast Extract Glucose (2-YEG) (yeast extract 2 g/L and glucose 10 g/L) medium for three days at 180 rpm at room temperature. Mycelium was harvested and ground to a powder under liquid nitrogen. Genomic DNA for Illumina sequencing was extracted according to Xin and Chen [24]. High molecular weight DNA for Nanopore sequencing was extracted according to Debler et al. [25] using a modified CTAB protocol.

Genome sequencing

P94-24 was sequenced on an Illumina NextSeq 500 instrument. Genomic DNA (50 ng) was used as input to the Nextera DNA library preparation kit. AlKewell was sequenced by Novogene (Beijing, Chaoyang District, China) on an Illumina HiSeq 3000 in PE150 mode. Nanopore libraries were generated with the SQK-LSK109 ligation sequencing kit (Oxford Nanopore Technologies, Oxford, United Kingdom) together with the native barcoding kit EXP-NBD104 (Oxford Nanopore Technologies, Oxford, United Kingdom) and sequenced on a MinION Mk1B on an R9.4.1 flowcell following the manufacturer’s instructions.

RNA sequencing

AlKewell and P94-24 were grown in 2-YEG as described earlier to harvest mycelium for the in-vitro samples. Both isolates were also grown on ½ potato dextrose agar (PDA) plates at room temperature (18 °C-22°C) under 12-h dark 12-h light under a long-wave UV lamp. Mature pycnidia were harvested after 14 days by flushing the plates with water and gently rubbing with the plunger of a 10 mL syringe. Lentil cultivars PBA Bolt, PBA Highland XT, and Nipper were inoculated with AlKewell and P94-24 pycnidiospore suspensions containing 1 × 10⁶ spores/mL and 0.025% Tween 20. Infected leaves were harvested at 1, 3, 5 and 7 dpi and flash frozen in liquid nitrogen. Mycelium and plant material were ground to a fine powder under liquid nitrogen and total RNA was extracted with the Direct-zol RNA Miniprep kit (Zymo Research, Irvine, CA, USA). RNA was assessed for DNA contamination by PCR of the actin gene using primers Actin-F (CTGCTTTGAGATCCACAT) and Actin-R (GTCACCACTTTCAACTCC). Samples showing DNA contamination were further processed with the TURBO DNA-free kit (Thermo Fisher, Waltham, MA, USA) until no more actin PCR products were detectable. RNA integrity was checked by gel electrophoresis of 500 ng RNA on a 1% agarose gel. RNA for timepoints 1 and 3 dpi were pooled and referred to as ‘early’, timepoints 5 and 7 dpi were pooled and referred to as ‘mid’. RNA (1500 ng) was prepared in 20 µL nuclease free water and shipped to the Australian Genome Research Facility (AGRF, Melbourne, Australia) for sequencing. RNA libraries were prepared using TruSeq mRNA Library Prep and 100 bp paired-end sequencing was performed on the Illumina NovaSeq 6000. Data was downloaded from AGRF and paired reads were trimmed with FastP 0.23.2 [26] (–trim_poly_g –cut_right –disable_quality_filtering). Trimmed read pairs were mapped against the genome assemblies with hisat2 2.1.0 and bam files generated with samtools 1.13 [27, 28].

For processing the RNA-seq data for differential expression analysis we used the nextflow pipeline ‘nf-core rnaseq’ (version 3.18.0) with options –trimmer fastp and read filtering against the Lens culinaris assembly CDC Redberry V2.0 retrieved from https://knowpulse.usask.ca/genome-assembly/Lcu.2RBY/ [29, 30]. After removing plant reads the mean percentage of fungal reads for the early timepoint was 0.13%, for the mid timepoint 0.42% and for the in vitro samples 99.84%. Differential expression analysis was performed with the nextflow pipeline ‘nf-core differential abundance’ (version 1.5.0) which uses the R package DEseq2 [31, 32].

Genome assembly

Illumina data was trimmed using Trimmomatic 0.38 [33]. Raw Nanopore data was basecalled with Guppy 5.0.7 using the ‘sup’ model (guppy_basecaller -c dna_r9.4.1_450bps_sup.cfg –barcode_kits EXP-NBD104 –trim_barcodes –detect_mid_strand_barcodes –min_score_mid_barcodes 60). Genomes were assembled with Canu 2.1.1 [34] (genomeSize = 45 m -fast), polished once with racon 1.4.21 [35] (-m 8 -x −6 -g −8 -w 500 -t 14 –no-trimming) and once with medaka 1.4.3 (-m r941_min_sup_g507). Contigs were reordered and reoriented relative to the A. lentis reference genome Al4 [36] with ragtag 2.1.0 (scaffold) [37], but the generated scaffolded assemblies containing Ns were only used as a reference to reorient the draft contigs against Contigs were manually inspected in Geneious Prime 2026.0.2 (Biomatters, Auckland, NZ) by mapping the long reads back onto the reordered contigs and overhanging ends were extended until split contigs could be merged. New joins were confirmed by mapping the Canu-corrected long reads back with Minimap 2.2.0 [38]. This was also used to extend contigs without telomeres as far as possible. Final assemblies were polished with Illumina data using pilon 1.24 [39] after mapping the reads against the assemblies with hisat2 2.1.0 (–max-intronlen 2000) [40].

Genome annotation and analysis

De novo genome annotation was performed with GeneMark-ES 4.68 [41] (–fungus). Augustus 3.3 [42] was first trained with the RNA-seq data and then run using the new model. RNA-seq reads were assembled with StringTie 2.1.6 [43], the output GTF file converted to GFF3 with AGAT [44] (agat_convert_sp_gxf2gxf.pl -g stringtie.gtf -o stringtie.agat.gff3) and genes predicted with CodingQuarry 2.0 [45] in pathogen mode. BRAKER1 [46] was run with RNA-seq data and BRAKER2 [47] was run with the odb10 fungi protein database downloaded from orthoDB. Results of BRAKER1 and BRAKER2 were combined with Tsebra 1.0.3 [48]. All annotations and RNA-seq data were loaded onto WebApollo [49] provided by Bioplatforms Australia and manually curated. The final geneset was exported and translated protein sequences run through Interproscan 5.51–85 [50] and antiSMASH 7.0 [51]. Funannotate 1.8.9 was used to combine Interproscan and antiSMASH results and a Python script [52] was used to clean up EC number tags in the final GFF file to make it NCBI compliant. Proteomes were used for effector candidate ranking by Predector 1.2.7 [53]. Nucmer 4.0.0.rc1 [54] with the –mum option was used to align the genomes and output files used to generate circos plots using Circos 0.69–8 [55]. Linear genome comparison diagrams were generated with PlotSR 1.0.0 [56] and SyRI 1.6.3 [57].

Repeat annotation

Repetitive sequences were predicted and annotated using the EarlGrey 7.0.2 pipeline [58], utilising RepeatMasker 4.2.2 [59], RepeatModeler 2.0.7 [60], RepeatCraft [61], LTR_FINDER [62], RepBase 2,018,026 [63] and Dfam 3.9 (Partition 16 – fungi). Repetitive elements were not further manually curated with exception of the Penelope-like element described herein.

Phylogenetic analysis of telomere-specific retrotransposon family Coprina-1_Alen

Sequences of the 13 Coprina elements in P94-24 were extracted between their flanking telomeric repeats and aligned using the MAFFT v7.490 [64] plugin in Geneious Prime 2026.0.2. The consensus sequence was extracted and open reading frames (ORFs) predicted, translated and annotated using the InterProScan [50] plugin within Geneious Prime.

Penelope-like elements of type Coprina SLL, Pt1, Sm1, Sm2, Cc1, Pc1 and Pc2, Neptune and Poseidon were extracted from RepBase-20181026. Athena Av1 (ACY69870), AvM (EU643486, EU643487) and AvO (EU643486), Penelope Dv (XP_070067300) and Coprina-1_LeEd (PV524095) were downloaded from GenBank and Chlamys, Hydra, Naiads, Physarum and PLE-X elements were downloaded from the supplementary data by Craig et al. [65]. This resulted in a total of 135 sequences. ORFs were predicted and translated in the DNA sequences. All translated ORFs were annotated by InterProScan. When multiple ORFs were present (Athena elements, Coprina-1_Alen), the protein with an annotated reverse transcriptase (RT) domain was chosen for further analysis. RT domain containing proteins were aligned using MAFFT in Geneious Prime and the alignments trimmed to keep the sequence between the DKN and Thumb domains, including the conserved RT 1 to 7 domains [66]. A tree was generated from the trimmed alignment using the Geneious Tree Builder function with the Jukes-Cantor genetic distance model and the Neighbour-Joining method. A consensus tree was generated from 1,000 bootstrap replications.

Agroinfiltration

The top 12 candidate effector genes (Table S1) were cloned and constructs infiltrated into N. benthamiana as described by Debler et al. [67]. Details of primers used in this study are listed in Table S2. Briefly, mature fungal effector candidates (without their native signal sequence) had attB tails for Gateway cloning added via PCR using the Platinum SuperFi high fidelity DNA Polymerase Mastermix (Thermo Fisher, Waltham, MA, USA) according to the manufacturer’s instructions. These constructs were then inserted into the pDONR/zeo entry vector via Gateway BP Clonase II (Thermo Fisher, Waltham, MA, USA). Entry clones were further recombined with destination plasmids pEAQ-HT-DEST1-His and pEAQ-HT-DEST1-LcPR1-His using Gateway LR Clonase II (Thermo Fisher, Waltham, MA, USA). Both Gateway steps used only 25% of the volumes of all reagents as suggested by the manufacturer. Plasmids were transformed via heat shock at 42 °C for 30 s into competent OmniMAX 2 E. coli cells (Thermo Fisher, Waltham, MA, USA). Constructs were confirmed via Sanger sequencing (Macrogen, Seoul, South Korea) and electroporated into A. tumefaciens strain GV3101::pMP90 by combining 0.2 µL purified plasmid at 100 ng/µL and 80 µL electrocompetent A. tumefaciens in a Bio-Rad Gene Pulser Xcell electroporator using the predefined program ‘Agr’. A. tumefaciens transformants were selected on LB Agar containing 30 µg/mL kanamycin, 50 µg/mL rifampicin and 25 µg/mL gentamicin grown at 25 °C for two to three days. Colonies carrying the correct expression clones were grown in 10 mL LB containing 30 µg/mL kanamycin, 50 µg/mL rifampicin and 25 µg/mL gentamicin on a shaking incubator at 27 °C at 250 rpm for 24 h. Turbid cultures were harvested by centrifugation at 4,000 × g at 4 °C for 15 min. The supernatant was discarded, and the pellet resuspended to OD₆₀₀ of 0.5 in 10 mM MES-MgCl₂. Acetosyringone (Sigma-Aldrich, St Louis, MO, USA) was added to a final concentration of 400 µM and the cultures were incubated at 27 °C at 100 rpm for at least 3 h. Candidates together with the positive necrosis control construct A. lentis NLP2 and the negative control construct GFP were infiltrated into the abaxial surface of six-week-old N. benthamiana leaves using a 1-mL needleless syringe. Cell death was observed seven days after infiltration. Leaves were harvested and fluorescence images taken on a ChemiDoc MP (BioRad, Cal, USA) imaging system. True colour images were acquired using an Epson flatbed scanner.

Results

High quality draft genome assembly and annotation

The two A. lentis isolates AlKewell and P94-24 were selected for whole genome sequencing, assembly and annotation, as these isolates represent two distinct pathotypes with contrasting responses on Australian lentil cultivars Nipper, PBA Hurricane XT and PBA Bolt (Fig. 1) [2, 23]. Figure 1 shows typical AB symptoms on resistant and susceptible lentil cultivars. The symptoms on compatible disease interactions are observed as necrotic lesions on leaves, stems and pods, general chlorosis, and leaf curling. This is particularly evident for Nipper with P94-24 as well as with AlKewell on PBA Hurricane XT and PBA Bolt. In contrast, incompatible interactions show healthy green leaves with no signs of infection symptoms. The genomes were assembled from Oxford Nanopore long reads (30 × and 61 × coverage for AlKewell and P94-24, respectively) and polished with Illumina short reads (128 × and 81 × coverage for AlKewell and P94-24, respectively). Sequencing statistics are summarised in Table 1. Draft assemblies were mapped against the A. lentis isolate Al4 reference genome [36] (Genbank reference GCA_004011705.1) and the contigs reordered to match the Al4 contig order and orientation. Split contigs in the drafts were manually merged and the breakpoints inspected by mapping the long reads back against the merged contigs, making sure the joints were supported by sequencing data. The respective mitochondrial genomes were assembled into individual single contigs and trimmed to remove sequence duplication due to their circular nature. Both draft assemblies consist of 22 complete uninterrupted chromosomes with telomeres on both ends, a complete circular mitochondrial genome, and one chromosome with only one telomere. This unresolved chromosome in both assemblies contains rDNA repeats that could not be resolved to the telomere (Fig. 2a and b, Chr 18). The AlKewell and P94-24 draft assemblies are 41.71 Mb and 42.11 Mb in size with respective GC content of 48.4% and 48.3%. Assembly size and GC content are close to the A. lentis reference Al4 (41.92 Mb and 48.4%) [36] with the only major difference being the Al4 mitochondrial genome, which is a concatenation of 1.5 copies.

Table 1.

Summary of assembly and statistics for the sequencing of Ascochyta lentis isolates, AlKewell and P94-24

Gene annotation statistics	AlKewell	P94-24
Genome size (bp)	41,712,406	42,114,786
GC content	48.4%	48.3%
Coverage Nanopore	29.6 x (1,235,943,995 bp)	60.6 x (2,553,630,806 bp)
Read length average (read N50)	5,042 bp (15,675 bp)	5,816 bp (12,135 bp)
Coverage Illumina	127.7 x (5,325,898,100 bp)	81.2 x (3,420,064,701 bp)
Largest contig	3,960,074 bp	4,140,525 bp
Assembly N50	1,824,960 bp	1,804,940 bp
Complete nuclear chromosomes (T2T)	22	22
Incomplete nuclear chromosomes	1	1
Complete circular mitochondrial genome (size)	Yes (52,500 bp)	Yes (52,190 bp)
Annotated genes (proteins)	13,938 (14,282)	13,927 (14,273)
Unique proteins	29	35
Predicted secreted proteins	2,154 (1,244–2,136)	2,152 (1,237–2,140)
Predicted secondary metabolite clusters	50	47
Predicted number of CAZymes (predicted by at least 2 of 3 tools)	567	567
Predicted tRNAs	117	117
BUSCOs (5.4.3) vs ascomycota odb10 (1706 BUSCOs)	97.6%	97.6%
Complete and single-copy	1,664	1,664
Complete and duplicated	2	2
Fragmented	6	7
Missing	34	33
GO terms assigned	35,966	35,459
Biological Process	11,749 (82.3%)	11,607 (81.3%)
Molecular Function	13,042 (91.3%)	12,764 (89.4%)
Cellular Component	11,174 (78.2%)	11,088 (77.7%)
Percentage of genome repeats	22.92%	23.75%
Percentage of genome RIP	22.18%	22.67%
Percentage of repeats RIPed	84.09%	87.47%

BUSCO Benchmarking Universal Single-Copy Orthologs, CAZymes Carbohydrate-active enzymes, GO Gene Ontology, RIP Repeat induced point mutations

Fig. 2.

Overview of the AlKewell (a) and P94-24 (b) genomes. Nuclear contigs labelled as Chr 01 to Chr 23, the complete mitochondrial genome is labelled as mito. Telomeres are represented by black dots at contig ends. The red end without a telomere symbol on both Chr 18 represents rDNA repeats that could not be resolved to the telomere. Location of Coprina-1_Alen PLEs in P94-24 is represented by a red diamond. Blue track above contigs represents locations of annotated genes, orange track underneath represents annotated repetitive sequences. Purple boxes indicate predicted biosynthetic gene clusters and red dots on the chromosomes indicate the location of the 12 effector candidates described here. c Linear comparison of parts of chromosome 1. The top track in black represents annotated repeats, the blue track below represents annotated genes. The red dot is one of the Predector top 50 effector candidates. Syntenic regions are represented in grey between the two genomes and the numbers below represent physical distance in Mb

Genes were predicted using multiple de novo and RNA-seq-supported gene prediction tools, after which gene models were manually curated by integrating all available predictions and mapped RNA-seq reads in a WebApollo [49] genome browser [68]. This resulted in 13,938 and 13,927 protein coding genes for AlKewell and P94-24, respectively. Including splice variants these generate 14,282 and 14,273 proteins respectively of which 2,154 and 2,152 contain predicted secretion signal peptides. A reciprocal BLAST analysis identified 29 unique proteins in AlKewell and 35 unique proteins in P94-24 (Table S3). The percentage of predicted secreted proteins in AlKewell and P94-24 is similar with 15.45% for both, yet slightly lower than that of Al4 with 18%. The Predector pipeline [53] uses SignalP, DeepSig, and Phobius [69–71] to predict signal peptides and the presented percentage represents a prediction by any one of the tools used. The actual spread between the tools, however, ranges from 1,244 (8.9%) – 2,136 (15.3%) for AlKewell, 1,237 (8.8%) – 2,140 (15.3%) for P94-24 and 1,129 (9.7%) – 1,808 (15.5%) for Al4. The difference between AlKewell and P94-24 compared to Al4 is due to the higher number of protein coding genes added during manual annotation and the incorporation of RNA-seq data described here, while the annotation of Al4 used de-novo prediction by AUGUSTUS [72]. Genome completeness was assessed by BUSCO [73] (version 5.4.3 using ascomycota odb10). Scores for both assemblies reached 97.6% indicating that the genomes have captured the gene-rich space. P94-24 has a slightly higher number of annotated repeats with 23.73% compared to AlKewell’s 22.92%. Carbohydrate-active enzymes (CAZymes) were predicted using the dbCAN 2.0 webserver, implementing prediction tools HMMER against the dbCAN and dbCAN-sub databases and DIAMOND against the CAZy database [74–77]. Hits were counted when at least two of the three tools used agreed on the same CAZyme class. Both isolates contain 567 predicted CAZymes. This is proportionally similar to the CAZyme content of A. lentis Al4 (491) with 11,638 annotated genes. The CAZyme class break down was the same for both isolates, the majority (274) classified as glycoside hydrolases followed by 114 with auxiliary activities, 94 glycosyl transferases, 45 carbohydrate esterases, 35 polysaccharide lyases and five carbohydrate-binding modules. CAZymes are produced by many fungi to break down plant cell walls. Functions were assigned using Gene Ontology (GO) terms with Blast2GO 6.0.3 [78]. Of AlKewell’s proteins 11,749 (82.3%), 13,042 (91.3%), and 11,174 (78.2%) were assigned to biological processes, molecular function, and cellular component, respectively. For P94-24 the respective results were 11,607 (81.3%), 12,764 (89.4%), and 11,088 (77.7%).

AlKewell and P94-24 chromosomes are highly syntenic

To evaluate the synteny between the two genomes, the chromosomes of AlKewell and P94-24 were reordered according to the published Al4 reference using RagTag 2.1.0 [37] and compared against each other with Nucmer 4.0 [54]. Both genomes are highly conserved and small differences are limited to repetitive regions (Fig. S1). Figure 2c shows a linear comparison of parts of chromosome 1, exemplifying that while the overall repeat content remains similar (22.05% in AlKewell vs 22.99% in P94-24), major structural variations and synteny breaks are co-located with repeat expansions and contractions. Repeat analysis and classification show that these are almost exclusively long terminal repeat (LTR) retrotransposons of type Ty3/Gypsy (Table 2). The A. lentis genomes described here contain several small inversions, most of which are also limited to repetitive regions. For example, an inversion on chromosome 4 spans 169 kb in P94-24 and 128 kb in AlKewell, the region in P94-24 contains several additional Ty3/Gypsy LTRs as well as one gene of unknown function as part of one of the additional Ty3/Gypsy LTRs. Another inversion on chromosome 5 (26 kb) contains the same seven genes in both isolates, but the whole region is flanked by Ty3/Gypsy LTRs. Most presence/absence differences between the two isolates are limited to repeat-rich subtelomeric regions.

Table 2.

Comparison of annotated repeats found by EarlGrey 7.0.2 using RepBase-20181026 and Dfam 3.9 (Partition 16 – fungi) in the genomes of AlKewell, P94-24 and Al4

	AlKewell	P94-24	Al4
DNA	626 (646,037 bp)	516 (544,205 bp)	633 (677,001 bp)
Rolling Circle	95 (150,011 bp)	110 (80,015 bp)	93 (106,959 bp)
Penelope	35 (18,398 bp)	13 (59,149 bp)	21 (5,369 bp)
LINE	67 (43,137 bp)	110 (111,927 bp)	60 (39,727 bp)
SINE	1 (106 bp)	1 (106 bp)	1 (106 bp)
LTR	3,099 (6,935,814 bp)	3,152 (7,435,302 bp)	2,956 (6,971,127 bp)
Ty3/Gypsy	1,920 (5,218,958 bp)	1,921 (5,583,724 bp)	1,947 (5,362,533 bp)
Ty1/Copia	1,178 (1,712,514 bp)	1,220 (1,846,188 bp)	1,016 (1,610,800 bp)
Other (Simple Repeat, Microsatellite, RNA)	9,124 (548,359 bp)	9,049 (481,266 bp)	8,900 (646,216 bp)
Unclassified	2,484 (1,405,978 bp)	2,595 (1,455,175 bp)	16 (2,284 bp)
Total Interspersed Repeats	6,407 (9,199,481 bp) 22.05%	6,501 (9,685,879 bp) 22.99%	6,301 (9,328,109 bp) 22.22%

LINE Long interspersed nuclear element, SINE Short interspersed nuclear element, LTR Long terminal repeat retrotransposons

Repeat content

Repeat sequences were predicted using the EarlGrey pipeline [58]. The genomes of AlKewell and P94-24 contain 22.05% and 22.99% repetitive sequence, respectively. These values are slightly higher than the 18% reported by Lee et al. [36] in the analysis of the A. lentis reference isolate Al4 that used the PiRATE Galaxy server [79] for repeat annotation, which uses a slightly different suite of tools and is therefore not directly comparable. We reannotated the Al4 assembly using the EarlGrey pipeline resulting in 22.22% of the sequence classified as repeats. Comparing the predicted repeat types of AlKewell, P94-24, and Al4 suggests that Al4 is more closely related to AlKewell than P94-24, especially regarding the difference of Penelope-like elements (Table 2, Fig. S2). The most prevalent retroelements in all three isolates are long terminal repeat retrotransposons of the Ty3/Gypsy superfamily, covering 12.5%, 12.8% and 13.3% of the genomes of AlKewell, Al4 and P94-24, respectively.

Characterisation of Penelope-like element Coprina-1_Alen in P94-24

One major difference in the repeat content between AlKewell and P94-24 is the presence of Penelope-like elements (PLEs) near telomeres in P94-24 and their scattered presence in AlKewell and Al4. All 13 PLEs present in P94-24 are located at the very end of their respective chromosomes (Fig. 2b, red diamonds). Closer inspection reveals additional telomeric repeat sequences between the annotated PLE and the rest of the chromosome for all but a severely degraded copy on chromosome 23. Of the 13 PLEs, ten contain two ORFs and two elements contain three (Fig. 3a). Functional annotation with InterProScan predicts a coil domain and several disordered regions for ORF1 and a reverse transcriptase (RT) domain followed by a Thumb domain for full length ORF2. The elements with an additional ORF3 (chromosome 11 and 15) contain a 5 bp deletion and a 1 bp insertion, respectively, leading to a premature stop codon after the RT domain in ORF2 thereby separating the Thumb domain into ORF3. The element on chromosome 23 is severely truncated, possibly due to insertion of the neighbouring Ty3/Gypsy retrotransposon, and contains several mutations not present in the other elements. All the PLEs are oriented in the same direction with their ORFs directed towards the centromere (away from the telomere). To ensure that these elements are real and not assembly artefacts one of the sequences was used as bait and all matching uncorrected Nanopore reads from P94-24 were extracted and assembled into contigs. These were then used in a BLASTN search against the draft assembly. This led to the addition of two PLEs to chromosome ends that were missed by the assembler and the collapsing of one tandem PLE into a single copy. This was further verified by mapping all the Nanopore reads back against the final assembly and manual inspection of their locations. No reads containing PLE sequences could be found in the AlKewell raw data, further indicating this to be a P94-24-specific property. The average length of PLEs in P94-24 (4,500 bp), AlKewell (525 bp) and Al4 (255 bp) and their scattered presence in AlKewell and Al4 suggest the latter to be annotation artifacts. We aligned all P94-24 PLEs and designed two forward and two reverse primers (Table S2, black arrows Fig. 3a) against conserved regions between all PLEs generating a 1014 bp (PLE_F1 + PLE_R1), 981 bp (PLE_F1 + PLE_R2), 892 bp (PLE_F2 + PLE_R1) or 859 bp (PLE_F2 + PLE_R2) product. Testing different primer combinations, we settled on PLE_F2 + PLE_R1 (Fig. S3) for future PLE screenings. This PCR further confirmed the absence of PLEs in AlKewell. To investigate if these elements are transcriptionally active, we examined P94-24 in vitro RNA-seq data for reads mapping to the identified PLE loci (Fig. 3b). We detected transcriptional activity across 12 of the 13 elements with coverage peaks roughly matching the annotated ORFs. Although due to the very high homology of 12 of the elements some peak borders may represent mapping artifacts.

Fig. 3.

Characterisation of Coprina-1_Alen. a Nucleotide alignment of PLE sequences extracted from 13 chromosome ends in P94-24. Extracted sequences contain telomeric repeats on both ends (except for chromosome 23). Black arrows on top indicate locations of diagnostic primers PLE_F1, PLE_F2, PLE_R1 and PLE_R2. b Mapped Illumina RNA-seq reads from P94-24 in vitro cultures overlayed onto their respective Coprina-1_Alen element. Black arrows represent telomeric repeats. c Model of the Coprina-1_Alen consensus element. T = telomeric repeats, R1 and R2 = repeated sequence (LTR), C = coil domain, NLS = nuclear localisation signal, DKN = conserved DKN motif, RT = reverse transcriptase domain containing conserved domains 1–7, Thumb = thumb domain. Elements have been reoriented so that the telomeric repeat on the left side in panels (a), (b) and (c) represents the end of the respective chromosome and the telomeric repeat on the right the border between the PLE and the rest of the chromosome

We extracted the consensus nucleotide sequence from the multiple sequence alignment of the 12 full length elements and annotated the proposed model (Fig. 3c). The two open reading frames are flanked by two direct repeats (Fig. 3c R1 & R2) of 63 bp and 255 bp respectively. The repeats at the 3’ end are separated by 104 bp. The element on chromosome 13 contains the same 104 bp between the two annotated repeats on its 5’ end leading to two direct repeats of 422 bp each. ORF2 contains a nuclear localisation signal (NLS) as well as the conserved DKN, RT and Thumb domains typical for PLEs. To determine the class of PLE we extracted sequences for Coprina SLL, Pt1, Sm1, Sm2, Cc1, Pc1 and Pc2, Neptune and Poseidon from RepBase-20181026. Athena Av1 (ACY69870), AvM (EU643486, EU643487) and AvO (EU643486), Penelope Dv (XP_070067300) and Coprina-1_LeEd (PV524095) were downloaded from GenBank and Chlamys, Hydra, Naiads, Physarum and PLE-X elements were downloaded from the supplementary data by Craig et al. [65]. Sequence alignment of the RT domain containing protein places our element into the Coprina clade of PLEs (Fig. 4, complete tree Fig. S4). The consensus sequence of Coprina-1_Alen has been submitted to Genbank under Accession PX883740.

Fig. 4.

Phylogenetic relationships between different Penelope-like element families. Neighbour-joining phylogram based on the alignment of the translated ORF containing the reverse transcriptase domain trimmed to include the conserved DKN, RT 1–7 and Thumb domains. Branch support values from 1,000 bootstraps. For readability the tree branches containing Naiad, Chlamys and PLE-X families were collapsed. The full tree is available in Fig. S4

Biosynthetic gene clusters

We used fungiSMASH 7.0 to investigate the difference in biosynthetic gene clusters (BGCs). A summary of the predicted BGCs and types is listed in Table 3, and their respective locations are indicated in purple in Fig. 2a and b. Several BGCs are in the subtelomeric regions of both isolates, of which two stand out. Chromosome 3 of AlKewell contains a 50 kb predicted fungal-ribosomally synthesised and post translationally modified peptides (RiPP)-like cluster in one of its subtelomeric regions containing 10 genes. The Minimum Information about Biosynthetic Gene cluster (MIBiG) repository of known BGCs identified two of them,, M3J07_002490 and M3J07_002492, as UstYa domain containing proteins involved in the synthesis of phomopsin A, a mycotoxin produced by Phomopsis leptostromiformis, a fungal pathogen of lupin causing lupinosis in livestock [80]. Between those two genes resides one of the effector candidates identified by Predector (AlenEC002, ranked 45). This gene is absent from P94-24 and has no BLAST hits against the nr database. Most of this region (37 kb) is missing from P94-24, which instead carries a Penelope-like element (PLE) at the end of its chromosome. The second cluster is also located at the very end of chromosome 6 in AlKewell. At this location fungiSMASH predicts two overlapping BGCs, that contain a fungal RiPP-like protein and a non-ribosomal peptide synthase (NRPS). The gene encoding the NRPS in AlKewell is 9.6 kb long and predicted to be a cyclic-depsipeptide synthetase. This gene is truncated to 2 kb in P94-24, where the open reading frame ends in the telomere.

Table 3.

Biosynthetic gene clusters (BGC) predicted by fungiSMASH 7.0

	AlKewell	P94-24
Fungal-RiPP-like	12	11
Indole	2	2
Non-ribosomal peptide (NRP)-metallophore	1	1
NRPS	10	9
NRPS-like	10	9
Type 1 PKS (T1PKS)	11	11
Type 3 PKS (T3PKS)	1	1
Terpene	3	3

NRPS Non-ribosomal peptide synthase, PKS Polyketide synthase, RiPP Ribosomally synthesised and post translationally modified proteins

Effector ranking and shortlisting using Predector

All protein coding genes were ranked for their effector likelihood by Predector [53]. We focused on the top 50 effector candidates. Table S4 lists the top 50 candidates for each isolate side by side with their respective ranks, nucleotide and amino acid differences, and all the properties Predector uses for ranking (complete Predector results for both isolates in Tables S5a & S5b). Nucleotide and amino acid differences between AlKewell and P94-24 were determined using SnpEff 5.2 [81]. Single nucleotide polymorphism (SNP) analysis revealed that of the total 78,148 SNPs between the two isolates, only 3,904 SNPs in coding sequences led to amino acid changes resulting to 1,936 genes affected by this change (Table S6). A total of six genes in the top 50 effector candidates led to amino acid changes across 10 SNPs, two of which were characterised in this study (Table 4).

Table 4.

Shortlisted effector candidates and their properties

Candidate ID	AlKewell ID	P94-24 ID	Protein identity	MW (kDa)	Cys	BLASTp vs NCBI nr/nt database best hit outside A. lentis	Predector hit	Foldseek hit using AlphaFold3 model
AlenEC001$*	XPS75409.1	NA	NA	10.67	8	None		No hits (AlphaFold3 pTM: 0.18)
AlenEC002$*#	XPS70260.1	NA	NA	19.28	10	None		No hits (AlphaFold3 pTM: 0.19)
AlenEC003$*	XPS77111.1	XPS99982.1	100%	7	8	Hypothetical protein Ascochyta rabiei (E-value 7.43e-21, coverage 100%, identity 45.2%)		No hits (AlphaFold3 pTM: 0.27)
AlenEC004$*	XPS72150.1	XPS95023.1	100%	5.7	8	Hypothetical protein Akantomyces lecanii (E-value 3.74e-25, coverage 100%, identity 62.3%)		Penicillium brevicompactum – bubble protein (AlphaFold3 pTM: 0.72, Foldseek E-value: 5.28e-3)
AlenEC005$*	XPS69095.1	XPS91967.1	100%	24.4	4	Hypothetical protein Botrytis paeoniae (E-value 5.13e-58, coverage 93.02%, identity 47.3%)		Borytis porri—uncharacterised protein (AlphaFold3 pTM: 0.84, Foldseek E-value: 8.16e-20)
AlenEC006$*	XPS67726.1	XPS90588.1	100%	10.7	6	Hypothetical protein Colletotrichum fioriniae (E-value 3.40e-53, coverage 100%, identity 68.9%)		Colletotrichum simmondsii—uncharacterised protein (AlphaFold3 pTM: 0.73, Foldseek E-value: 6.06e-5)
AlenEC007$*	XPS71014.1	XPS93874.1	100%	4.9	8	None		No hits (AlphaFold3 pTM: 0.43)
AlenEC008$*	XPS76720.1	XPS99594.1	100%	8.4	12	Cytochrome p450 Lasiodiplodia theobromae (E-value 4.55e-33, coverage 89.9%, identity 57.4%)		Diplodia corticola—Cytochrome p450 (AlphaFold3 pTM: 0.75, Foldseek E-value: 2.51e-5)
AlenEC009$*	XPS70623.1	XPS93482.1	100%	10.18	8	Hypothetical protein Ascochyta rabiei (E-value 5.30e-42, coverage 99.1%, identity 61.8%)		No hits (AlphaFold3 pTM: 0.23)
AlenEC010$*	XPS81080.1	XPT03937.1	100%	13.38	6	Ecp7 Fusarium agapanthi (E-value 7.57e-49, coverage 100%, identity 55.9%)	RiSLM, Mg1LysM, MgxLysM, MoSlp1, CfEcp6	Fusarium oxysporum—LysM domain-containing protein (AlphaFold pTM: 0.79, Foldseek E Value: 5.38e-17)
AlenEC011$*	XPS75390.1	XPS98274.1	99.35% (1AA)	15.14	6	Hypothetical protein Ascochyta rabiei (E-value 5.26e-78, coverage 100%, identity 75.2%)		Didymella exigua—uncharacterised protein (AlphaFold pTM: 0.67, Foldseek E-value: 2.47e-14)
AlenEC012$*^	XPS72974.1	XPS95848.1	99.4% (1AA)	15.25	10	AvrLm3 Leptosphaeria maculans (E-value 5.14e-26, coverage 94.58%, identity 36.1%) Ecp11-1 Cladosporium fulvum (E-value 2.42e-22, coverage 87.35%, identity 36.5%)	LmAvrLm3	Leptosphaeria maculans—AvrLm3 (AlphaFold pTM: 0.86, Foldseek E-value: 3.68e-14)

Cys number of Cysteines

^*Deeploc extracellular prediction

^{^}Localizer nucleus prediction

^#Localizer chloroplast prediction

^$contains signal peptide

The top 50 list contains 14 matches to published effectors, including a homologue of the necrosis and ethylene-inducing peptide-like (Nep1-like) protein (NLP) at rank 15 and 13 for AlKewell and P94-24, respectively. NLPs are present in many plant pathogens and lead to necrotic lesions and ethylene production in dicots [82]. We have previously used NLP homologues of the pea pathogen Peyronellaea pinodes and the chickpea pathogen Ascochyta rabiei as positive necrosis controls for Agroinfiltration experiments [67]. Predector detected a total of five potential homologues to known NLPs, only two of which contain the NPP1 domain (Pfam ID PF05630) including the hallmark hexapeptide “GHRHDWE” which forms part of an acidic cation-binding pocket important for their activity [83]. The other three contain a Vps62 domain (Pfam ID PF06101) required for cytoplasm to vacuole targeting of proteins [84]. Part of the Vps62 domain in these three proteins is a short “HVGDWEH” peptide similar to the “GHRHDWE” heptapeptide motif of the NPP1 domain, the underlined amino acids represent the core of the hexapeptide in NLPs, and mutation of each individual one to an alanine has been shown to abolish the protein’s activity [85]. This may be the reason Predector assigned these to homologues of NLPs. The two NPP1-containing proteins ranked at 15 and 147 (AlKewell) and at 13 and 148 (P94-24) while the Vps62-containing ones ranked at 448, 2,142 and 9,342 (AlKewell) and 447, 2,155 and 9,333 (P94-24). The NLP homologue ranked at 15 in AlKewell and 13 in P94-24 is homologous to the NLP2 protein of P. pinodes and A. rabiei we previously used in agroinfiltration experiments [67] and will be referred to as NLP2 going forward. To validate the A. lentis NLP2 homologue as an effector, we cloned the NLP2 homologue with the addition of an N-terminal signal peptide for apoplast targeting and agroinfiltrated into N. benthamiana alongside the previously validated P. pinodes and A. rabiei homologues [67]. All three generated necrotic lesions (Fig. 5) and we continued to use the A. lentis NLP2 homologue for further experiments as the positive necrosis control. We used RNA-seq data derived from A. lentis AlKewell and P94-24 infections of PBA Bolt and Nipper to inform effector candidate selection. The availability of RNA-seq data from PBA Bolt serves as a (near identical or close) replacement for PBA Hurricane XT due to their closely shared genetic background derived from the parent line 96-047L*99R060 [2]. Both cultivars exhibit comparable resistance responses to A. lentis infection (Fig. 2), particularly in their recognition of the avirulence gene AlAvr1-1 [2, 23]. For the purpose of candidate selection, we consider PBA Bolt's RNA-seq data a reliable proxy in this study.

Fig. 5.

Homologues of the necrosis and ethylene-inducing peptide 1-like proteins-2 (NLP2) from the pea pathogen Peyronellaea pinodes, the chickpea pathogen Ascochyta rabiei and the lentil pathogen Ascochyta lentis expressed via transient Agrobacterium tumefaciens-mediated transformation (agroinfiltration) of Nicotiana benthamiana. GFP acts as the negative necrosis and positive protein expression control. The same leaves were scanned in true colour and imaged on a ChemiDoc to visualise the GFP negative necrosis control and highlight the necrotic areas

Effector candidate characterisation

To validate the effector-likelihood of the proteins ranked by Predector, we characterised the top 10 AlKewell proteins and their corresponding homologues in P94-24, together with two proteins in the top 50 present only in AlKewell (summarised in Table S4). We generated agroinfiltration constructs for these candidates and their properties are listed in Table 4. For each candidate we cloned the mature part without its native signal peptide into two destination plasmids, pEAQ-HT-DEST1-His and pEAQ-HT-DEST1-LcPR1-His [67]. After infiltration into Nicotiana benthamiana the protein expressed from pEAQ-HT-DEST1-His is expected to remain inside the transformed cell’s cytoplasm, while the protein expressed from pEAQ-HT-DEST1-LcPR1-His is expected to be exported to the apoplast because the L. culinaris PR1 protein signal peptide is fused to the candidate effector’s N-terminus.

Representative effector candidates infiltrated into N. benthamiana leaves are shown in Fig. 6 (Replicates are in Fig. S5) and their respective expression profiles in lentil cultivars PBA Bolt and Nipper in Fig. 7. Of the 12 candidates, two are only present in AlKewell (AlenEC001 and AlenEC002), three have one nucleotide difference between the AlKewell and P94-24 version of which two lead to an amino acid change (AlenEC011P/K and AlenEC012P/K). Of these 12 candidates 11 induced necrotic lesions after agroinfiltration into N. benthamiana.

Fig. 6.

Agroinfiltration of effector candidates into Nicotiana benthamiana. ‘ + ’: A. lentis NLP2: positive necrosis control; ‘G’: GFP, negative necrosis control; ‘SP’: effector candidate expressed with N-terminal LcPR1 signal peptide; ‘noSP’: effector candidate expressed without signal peptide. The same leaves were scanned in true colour and imaged on a ChemiDoc to visualise the GFP negative necrosis control and highlight the necrotic areas

Fig. 7.

RNA-seq differential expression results of AlKewell and P94-24 on lentil cultivars PBA Bolt and Nipper. Normalised gene expression counts were obtained using DESeq2’s median-of-ratios method. Displayed are expression results for the 12 effector candidates described here as well as AlActin as a control for abundant expression in vitro and during infection, AlAvr1 as a control of the known avirulence gene with high expression on its compatible host Nipper early during infection, and AlNlp2 as positive necrosis control used in agroinfiltration experiments

AlenEC001 and AlenEC002 are only present in AlKewell, and both showed necrotic activity on N. benthamiana. These two as well as the other ten candidates have all the typical hallmarks of an effector (small, secreted, high percentage of cysteines, Table 4), however, AlenEC002 is predicted to be localised in the nucleus while AlenEC001 is predicted to be secreted into the apoplast. In agroinfiltration transient expression, AlenEC001 with the LcPR1 signal peptide induced necrosis while the construct without did not. This suggests that AlenEC001 interacts with a plant cell surface receptor which in turn triggers a necrotic response. On the other hand, AlenEC002 without the LcPR1 signal showed positive necrosis while the construct with the signal peptide did not, matching the prediction that it is targeted to the nucleus. It also suggests AlenEC002 may be a host selective effector which relies on a lentil-specific receptor to gain entry into the plant cell. This gene resides in a 30 kb region at the 5’ end of chromosome 3 missing from P94-24 and is part of a small cluster of four genes flanked by Ty3/Gypsy LTRs on both sides. Two of the genes in this cluster are predicted as UstYa mycotoxin biosynthesis proteins by InterProScan and the whole region is predicted to be part of a fungal-RiPP-like cluster by fungiSMASH. Structural prediction for both candidates returned models with very low confidence scores (pTM 0.18 and 0.19) and no hits in any of the databases used by Foldseek. RNA-seq analysis shows no expression for AlenEC001 but several reads covering the gene can be seen in the genome browser [68] for the in vitro sample of an AlKewell AlAvr1-2 knockout strain. AlenEC002 shows similar amounts of expressed transcripts for both PBA Bolt and Nipper, with significantly more during the later stage of infection.

Candidates AlenEC003, AlenEC004, AlenEC007, AlenEC008, and AlenEC009 are identical in both isolates, and all show necrotic lesions for the apoplast-targeted version only, while AlenEC005 and AlenEC006, also identical between isolates, show necrotic lesions for both the intracellular and the apoplast-targeted versions. Of these seven candidates only AlenEC008 produces BLAST hits to named proteins, specifically 50% homology to cytochrome p450 of the oak pathogen Diplodia corticola and the broad host fruit tree pathogen Lasiodiplodia theobromae. The predicted folded protein (pTM 0.75) also matches the D. corticola cytochrome p450 model. Of these the only other candidate with structural similarity to a characterised protein is AlenEC004. It is similar to the Penicillium brevicompactum bubble protein, a small cysteine rich protein suspected to function as a killer toxin [86]. None have annotated protein domains but all do contain between one and four RxLR-like motifs as defined by the pattern [R⁄K⁄H]x[L⁄M⁄I⁄F⁄Y⁄W]x [87]. It has been shown for other plant pathogenic fungi that effectors with RxLR-like motifs can be imported by RxLR-mediated binding to phosphatidylinositol-3-phosphate (PI3P), which is abundant on the outer surface of plant cell membranes [88]. Even though all these candidates have the same sequence, AlenEC003 shows a differential expression response on the respective cultivars. During the mid to late stage of infection AlKewell produces significantly more transcript on cultivar PBA Bolt, while P94-24 produces significantly more on cultivar Nipper. AlenEC004, AlenEC006 and AlenEC008 only show very little expression limited to the in vitro sample. AlenEC005 is expressed early in PBA Bolt for both isolates while P94-24 produces more transcript on Nipper than AlKewell. Expression levels drop during the later stage to the same level for both pathogens on both cultivars. Expression of AlenEC007 in AlKewell is only observed during infection but remains at a comparable level across timepoints and cultivars. Expression in P94-24 at the early timepoint is higher in Nipper but at a similar level in both cultivars later in the infection. AlenEC009 exhibits a similar pattern to AlenEC007 with consistent levels in AlKewell but higher levels in P94-24 early on Nipper.

On the other hand, AlenEC011 and AlenEC012 each have one single nucleotide polymorphism (SNP) between AlKewell and P94-24, respectively, which leads to an amino acid change. Both candidates induce necrotic lesions for both the intra- and the extracellular versions in N. benthamiana. AlenEC011 only shows expression in vitro with a few reads mapping from the Nipper infection. The expression pattern for AlenEC12 however is very similar to the avirulence gene AlAvr1. Higher levels in P94-24 on Nipper but low levels on PBA Bolt early and comparably low levels at the later timepoint, while levels for AlKewell remain similarly low throughout. AlenEC012 has some amino acid sequence homology to the avirulence genes AvrLm3 from Leptosphaeria maculans (36.1% identity over 94.58% of its sequence) and Ecp11-1 from Cladosporium fulvum (36.5% identity over 87.35% of its sequence). The predicted structures of both AlenEC012P and AlenEC012K (pTM 0.86) do however show very high structural homology to both AvrLm3 (RMSD 0.782 over 630 atoms) and Ecp11-1 (RMSD 0.600 over 582 atoms) (Fig. S6).

AlenEC010 is the only candidate among the 12 infiltrated that does not induce necrotic lesions for both agroinifiltration constructs. It is also the only one with an annotated protein domain, specifically a LysM domain. It has amino acid sequence homology to the Fusarium agapanthi effector candidate Ecp7 (55.9% identity over 100% of its sequence) and structural homology to the Fusarium oxysporum LysM domain-containing protein (Foldseek E Value 5.38e-17). This candidate might therefore be involved in hiding the pathogen during the early stage of infection by sequestering fungal chitin instead of eliciting a response from the host plant.

Discussion

The complex interaction between host and pathogen requires a suite of virulence-associated elements to promote successful infection. Our analysis of the two A. lentis isolates AlKewell and P94-24 which exhibit opposing pathogenicity patterns towards specific lentil cultivars provides new insights into the genomic basis of pathogenicity in this important lentil pathogen. Building upon our previous identification of the first A. lentis avirulence effector AlAvr1, identified through map-based cloning using a bi-parental A. lentis population [2], we have now generated near-complete, high-quality genome assemblies and annotations of both isolates representing the two A. lentis pathotypes.

The genome architectures of AlKewell and P94-24 are highly conserved and closely resemble the published Al4 reference genome [36] in size, structure and repeat content. By integrating multiple gene prediction algorithms with RNA sequencing data and manual curation of the gene models, we have created the most comprehensive A. lentis gene catalogue to date. This effort has identified an additional 2,300 and 2,289 protein-coding genes in AlKewell and P94-24, respectively, compared to Al4. When considering splice variants detectable only through RNA-seq data, these numbers rise to 2,644 and 2,635, further expanding the known gene repertoire of A. lentis.

The primary genomic differences between isolates are found in expanding or contracting repeat-rich regions, predominantly containing Ty3/Gypsy-type long terminal repeat (LTR) retrotransposons and shorter unclassified repeats. Ty3/Gypsy elements are widespread across plants, animals, and fungi, often representing the largest group of retrotransposons in fungal genomes [89]. Similar variation in Ty3/Gypsy content has been observed in other fungi, such as in Pleurotus ostreatus, where these elements comprised 2.2% of one isolate's genome but only 0.3% of another [17]. These transposon-rich regions are particularly interesting as they have been shown to harbour effector genes that can move with the transposon between isolates or to other related species [21]. Transposable elements can make up significant amounts of fungal genomes, with studies reporting TE content ranging from 0.02% to 29.8% in various Ascomycetes and Basidiomycetes, reaching as high as 58% in Tuber melanosporum and 62.7% in Erysiphe necator [17, 90, 91].

A notable discovery was the presence of Penelope-like elements (PLEs) of type Coprina exclusively in P94-24. PLEs, first observed causing hybrid dysgenesis in Drosophila virilis, are now known to occur in most eukaryotic genomes [66]. They are classified into two subclasses: those containing an open reading frame for a polyprotein with reverse transcriptase (RT) and GIY-YIG family endonuclease (EN) domains (Neptune, Poseidon, Penelope, EN(+)), and those with only the RT domain (Athena and Coprina, EN(-)) [92–94]. Coprina PLEs, first identified in fungi, have also been identified in the Kuruma shrimp Marsupenaeus japonicus, the clubmoss Selaginella moellendorffii and the diatom Phaeodactylum tricornutum [66, 95].They typically locate near telomeres and have proliferated extensively in some species, such as T. melanosporum and Agaricus bisporus, where they occupy almost every chromosome end [90, 96]. The two classes of EN(-) PLEs, Athena and Coprina, are typically distinguished by the presence of one ORF in Coprina and two ORFs in Athena [97]. The element present in A. lentis however contains two and in some cases due to indels three ORFs, yet a phylogenetic analysis of the RT domain-containing protein places it in the Coprina clade. Phylogenetically, PLE reverse transcriptases cluster closest to telomerase reverse transcriptases (TERTs), suggesting a common ancestor and mode of action [93, 97]. The proposed mechanism for EN-independent terminal retrotransposition involves sequence homology between the PLE RNA intermediate and temporarily unprotected telomeres, followed by reverse transcription by the PLE-encoded RT and telomere addition by the endogenous telomerase machinery [97]. The discovery of hammerhead ribozyme (HHR) motifs in PLE flanking sequences supports this model through self-cleavage of the RNA intermediate [98].

The restricted proliferation capacity of Coprina elements due to their lack of an EN domain and dependence on unprotected telomeres, explains their relatively low numbers compared to other retrotransposons. Furthermore, this lack of a GIY-YIG endonuclease suggests these elements are less likely to cause widespread genomic instability via random cleavage and may instead contribute to telomere maintenance dynamics. We detected transcriptional activity mapping to the PLEs confirming these elements are not merely genomic fossils but remain biologically active. This activity suggests these elements may play a role in telomere maintenance in P94-24, distinct from the telomerase-only maintenance likely utilised by AlKewell. This parallels the system in Drosophila, where the organism lacks standard telomerase and instead relies on specific retrotransposons to maintain telomere function, creating arrays functionally analogous to telomeric repeats [99]. However, this colonisation of chromosome ends appears to influence effector evolution. For instance, the AlKewell subtelomeric region of chromosome 3 carries a predicted biosynthetic gene cluster with two genes showing homology to genes involved in the synthesis of the mycotoxin phomopsin A. The effector candidate AlenEC002 resides within this BGC between the two predicted mycotoxin synthesis genes. Most of this cluster, including AlenEC002, is absent in P94-24 and is replaced by a Coprina PLE, indicating that PLE proliferation may drive the deletion or competitive exclusion of virulence-associated gene clusters. Their phylogeny suggests vertical transmission, with instances of partial or complete losses as observed in related genera like Drosophila and Caenorhabditis [66, 100]. The presence/absence variation between the described A. lentis isolates could serve as a useful marker for tracking isolate ancestry as more whole genome sequences become available.

Secondary metabolites (SMs) are compounds produced by many microorganisms that are not directly involved in growth or reproduction and include antibiotics like penicillin. The genes responsible for SM production typically cluster in biosynthetic gene clusters (BGCs) containing all necessary components for precursor production, modification, export, and regulation [101]. Genes involved in BGCs include non-ribosomal peptide synthases (NRPS), type 1 polyketide synthases (PKS), terpene and indole synthases, and ribosomally synthesised and post translationally modified peptides (RiPP). We identified two BGCs unique to AlKewell: a fungal RiPP-like cluster spanning approximately 50 kb with homology to known toxin synthesis genes, and an NRPS coding for a cyclic-depsipeptide synthetase. Similar cyclic-depsipeptide synthetases produce host-selective toxins in other plant pathogens, such as AM toxin in Alternaria alternata and HC toxin in Cochliobolus carbonum, which cause tissue damage in susceptible apple leaves and Northern corn leaf spot disease, respectively [13, 102]. Current knowledge of A. lentis BGCs and SMs is limited, with only disrupted ascochitine genes identified in Al4 [103] and several anthraquinone derivatives (lentisones and lentiquinones A, B & C) characterised from culture extracts [104, 105]. These compounds exhibit phytotoxic activity against lentil and antimicrobial properties against Bacillus subtilis, and notably, are unique to A. lentis despite anthraquinone's widespread occurrence in plants and fungi [106].

Of particular interest is the effector candidate AlenEC002 (Predector rank 45), located within one of the AlKewell-specific BGCs. This nuclear-localised protein, containing an NLS 'KKKR' motif, induces necrosis in N. benthamiana when expressed without its signal peptide. The absence of lesions in the apoplast-targeted version suggests that nuclear localisation is essential for its activity. Its AlKewell-specific presence and high expression levels on Nipper and PBA Bolt makes it a candidate host-selective effector. This gene resides in a 30 kb region at the 5’ end of chromosome 3 which is absent in P94-24 and is instead occupied by a Coprina PLE. This absence highlights the potential trade-off between repetitive element expansion and effector retention in subtelomeric regions. However, functional validation through knockout studies faces technical challenges due to its subtelomeric location and the limited availability of selective markers for A. lentis transformation. Current transformation success rates using Agrobacterium tumefaciens-mediated transformation (ATMT) are around 1%, and Hygromycin B remains the only viable selective agent, precluding double knockouts or complementation experiments [107].

Our comprehensive effector prediction pipeline identified multiple promising candidates, with 11 out of 12 tested candidates inducing necrotic lesions in N. benthamiana. These candidates display typical effector characteristics: small size, secretion signals, and high cysteine content (Table 4). The sole non-necrosis-inducing candidate, AlenEC010, contains a LysM domain for chitin binding. LysM-domain-containing effectors are known to suppress chitin-triggered immunity by sequestering chitin oligomers away from host receptors, effectively concealing fungal presence [108, 109]. AlenEC010 has amino acid sequence homology to characterised LysM domain-containing effectors from various fungi, including RiSLM (Rhizophagus irregularis), Mg1LysM and MgxLysM (Zymoseptoria tritici), MoSlp1 (Magnaporthe oryzae), and CfEcp6 (C. fulvum) as well as structural homology to a F. oxysporum LysM domain-containing protein. LysM domain-containing proteins are known to hide their respective pathogen from host chitinases and chitin binding receptors by sequestering chitin oligomers [110–113]. High expression levels of AlenEC010 throughout the infection process and the lack of a necrotic response align with A. lentis' hemibiotrophic lifestyle, suggesting a role in immune evasion rather than immune system activation.

Two notable candidate genes, AlenEC011 and AlenEC012, contain SNPs differentiating the AlKewell and P94-24 variants, and both induce necrosis in N. benthamiana regardless of cellular targeting. While AlenEC011 shows sequence homology to uncharacterised proteins in other Ascochyta species (A. rabiei 72%, A. clinopodiicola 75%) yet barely any detectable transcript in PBA Bolt and Nipper (but for a few mapped reads), AlenEC012 has a Predector match to the L. maculans AvrLm3 and a PHIbase hit to the C. fulvum Ecp11-1 avirulence proteins. It also has significantly increased expression in P94-24 early on Nipper compared to AlKewell. AvrLm3 and Ecp11-1 share little amino acid sequence homology (36.8% pairwise identity) but are structurally similar and therefore considered to be functional homologues [114]. They are part of the LARS-fold (Leptosphaeria AviRulence and Suppressing) family of effectors which also includes AvrLm4-7, AvrLm5-9, and AvrLmS-Lep2 [114–116]. AlenEC012 has a similarly low pairwise amino acid identity to both AvrLm3 (36.7%) and Ecp11-1 (36.5%). However, a structural alignment using the AlphaFold 3 generated structures of all four proteins (without their respective signal peptides) shows very high similarities (Fig. S6) suggesting that AlenEC012 is a functional homologue of both AvrLm3 and Ecp11-1. As recently demonstrated, this family exhibits extreme sequence plasticity while retaining a conserved core fold, a feature that allows specific resistance proteins (like Rlm3) to provide broad-spectrum recognition of fungal pathogens [116]. AlenEC012 may therefore target similar host cellular machinery to AvrLm3, with its sequence divergence reflecting adaptation to the specific immune receptor repertoire of lentil. Its expression pattern is similar to the AlAvr1 avirulence gene with high levels of expression early in P94-24 on cultivar Nipper. Like AvrLm3, AlenEC012 is located within a RIP-degenerated region dominated by Ty3/Gypsy retrotransposons, yet unlike AvrLm3, which shows presence/absence variation in L. maculans [117], AlenEC012 appears conserved across the currently available A. lentis genomes, though broader sampling is needed to confirm this pattern.

The very low transcript levels for candidates AlenEC001, AlenEC004, AlenEC006, AlenEC008 and AlenEC011, contrasted by their ability to cause necrosis in N. benthamiana and their high ranking by Predector are curious. AlenEC001 is located in the subtelomeric region, 30 kb from the telomere in between two TcMariner transposons, which have both been subject to RIP. AlenEC006 is also located next to a TcMariner transposon which is part of a larger 60 kb repeat cluster of large stretches of Ty3/Gypsy LTRs separated by TcMariners. This whole repeat cluster is also RIP-degenerated and parts of AlenEC006 appear to also have been subjected to RIP. The genome defence mechanism RIP, triggered by duplicated sequences during the sexual reproductive cycle of the pathogen, converts cytosine to thymine, mainly to stop the spread of transposable elements [118, 119]. This strategy however is known to sometimes spread to neighbouring genes, inadvertently inactivating them or their regulatory elements in the process [120]. AlenEC004, AlenEC008 and AlenEC011 are located in gene rich regions and have enough in vitro RNA-seq reads to support their respective gene models. However, apart from a few partial reads of AlenEC011 during Nipper infection, they show no expression in planta. This could point to these candidates being reliant on a host specific signal to induce expression, which wasn’t present in the cultivars used in this study, or they might be generic effectors needed for the pathogen to survive on a non-host species. The only known isolation of A. lentis from another host, grasspea (Lathyrus sativus), led to the description of the new variant A. lentis var. lathyri. While still able to mate with A. lentis this variant shows significant morphological differences and is not able to infect other legume hosts [121]. The related pathogen of chickpea, Ascochyta rabiei, has however been isolated from different non-symptomatic plants grown in the vicinity of chickpea, both legumes and cereals [122–124].

Conclusions

This study significantly enhances the genomic resources available for A. lentis research by providing two near-complete genome assemblies from isolates with contrasting virulence profiles and manually curated gene models. These resources lay the foundation for developing an A. lentis pangenome, which would facilitate deeper understanding of pathogen evolution. Furthermore, our characterised effector candidates provide valuable insights into host–pathogen interactions in the lentil Ascochyta blight pathosystem, potentially informing future disease management strategies. The combination of in silico effector prediction and ranking together with in planta expression data and candidate pre-screening in the model plant N. benthamiana further fine tunes the pipeline of candidates to take forward for testing in lentil. This is important due to the low efficiency of targeted manipulation of A. lentis when creating knockouts and the difficulty of agroinfiltration into lentil.

Abbreviations

AB

Ascochyta blight

ATMT

Agrobacterium tumefaciens-Mediated transformation

BGC

Biosynthetic gene cluster

BUSCO

Benchmarking Universal Single-Copy Orthologs

CAZymes

Carbohydrate-degrading enzymes

CRE/SLACS

Crithidia-related element/spliced leader-associated conserved sequence

DNA

Deoxyribonucleic acid

dpi

Days post-inoculation

EN

Endonuclease

ETI

Effector-triggered immunity

GFP

Green fluorescent protein

GO

Gene Ontology

HHR

Hammerhead ribozyme

HR

Hypersensitive response

LINE

Long interspersed nuclear element

LTR

Long terminal repeat

MIBiG

Minimum Information about Biosynthetic Gene cluster repository

MITE

Miniature inverted-repeat transposable element

NLPs

Necrosis and Ethylene-inducing Peptide-like proteins

NRPS

Non-ribosomal peptide synthase

PCR

Polymerase chain reaction

PI3P

Phosphatidylinositol-3-phosphate

PKS

Polyketide synthase

PLE

Penelope-like element

RiPP

Ribosomally synthesised and post translationally modified peptides

RNA

Ribonucleic acid

RT

Reverse transcriptase

SM

Secondary metabolites

SNP

Single nucleotide polymorphism

SP

Signal peptide

T2T

Telomere to telomere

TE

Transposable element

TERT

Telomerase reverse transcriptase

Authors’ contributions

**Johannes W. Debler:** Conceptualisation; investigation; methodology; data curation; formal analysis; writing – original draft; writing – review and editing. **Robert C. Lee:** Conceptualisation; supervision; writing – review and editing. **Lina M. Farfan-Caceres:** Methodology – Illumina sequencing; writing—review and editing. **Karam B. Singh:** Conceptualisation; supervision; writing – review and editing. **Lars G. Kamphuis:** Conceptualisation; supervision; funding acquisition; writing – review and editing. **Bernadette Henares:** Conceptualization; supervision; project administration; methodology; writing – review and editing.

Funding

This research was supported by the Australian Grains Research and Development Corporation (GRDC) and Curtin University co-investment in the Centre for Crop and Disease Management (CCDM) under grant CUR00023.

Data availability

All data generated or analysed during this study are included in this published article and its supplementary information files.—All sequencing data described in this paper has been deposited with the NCBI Genbank (BioProject PRJNA926806; AlKewell GCA_016097375.2 and P94-24 GCA_022607105.2) For details see Table S7. Genomes for AlKewell and P94-24 can be interrogated on the Apollo web browser on https://pulsepathogens.genome.edu.au/apollo/394538/jbrowse/index.html and https://pulsepathogens.genome.edu.au/apollo/394707/jbrowse/index.html respectively.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.