Aphanomyces euteiches causes disease in Lathyrus sativus with a globally diverse, polygenic resistance landscape Sara Rodriguez-Mena, Mario González, Ehsan Valiollahi, Carmen Santos, Diego Rubiales, Maria Carlota Vaz Patto Abstract AbstractKey messageAphanomyces euteiches inoculation of a global Lathyrus sativus collection allowed the identification of sources of resistance and associated candidate genes.AbstractAphanomyces euteiches is a major concern in legume crops with limited efficient methods for its management. Although Aphanomyces root rot has been widely studied in pea and lentil, the effect of this pathogen on grass pea (Lathyrus sativus) has been poorly explored. This study aimed to characterise the response of a globally diverse grass pea collection to A. euteiches and to identify genomic regions associated with any potential resistance. A diverse panel of 169 grass pea accessions was inoculated with the A. euteiches RB84 isolate. Twenty days after inoculation, foliar and root symptoms were evaluated. A wide range of responses was observed within the panel, from susceptibility to partial or even complete resistance. Along with these phenotypic data, an improved acquired 12,974 single-nucleotide polymorphism (SNP) data set was used to perform a genome-wide association study. The association study identified 20 resistance-associated SNPs, across five of the seven grass pea chromosomes, suggesting a polygenic nature. In silico analysis identified seven putative candidate genes associated with resistance involved in vesicle trafficking, signal transduction, nucleotide metabolism, fatty-acid metabolism, cell wall synthesis and ethylene signalling. This study confirms the compatible grass pea × A. euteiches interaction, identifying for the first time, sources of resistance and proposing resistance-associated loci with co-located candidate genes for future precision breeding purposes.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1007/s00122-026-05372-w. Introduction Aphanomyces root rot, caused by the oomycete Aphanomyces euteiches, is one of the major limitations in pea (Pisum sativum) and lentil (Lens culinaris) production worldwide (Wu et al. ) and has also been reported on other legumes like some Lathyrus species (Lathyrus cicera, L. hirsutus, L. latifolius, L. odoratus and L. tingitanus), alfalfa (Medicago sativa), snap bean (Phaseolus vulgaris), vetch (Vicia sativa), faba bean (V. faba), clovers (Trifolium pratense and T. repens), and several weed species (Papavizas and Ayers ; Gaulin et al. ). This pathogen primarily affects the root system, causing necrosis and discolouration of roots and stems, which leads to yellowing of the leaves and, in severe cases, plant death (Hughes and Grau ). It is a genetically diverse pathogen, with different isolates, pathotypes, and field populations identified (Kälin et al. ). Traditionally A. euteiches pea isolates have been classified according to their virulence patterns on differential host sets, distinguishing 11 pathotypes (I to XI) in pea and two races (race 1 and race 2) in alfalfa. (Giles et al. ; Moussart et al. ). Although A. euteiches exhibited host specificity, some isolates, such as the pea isolate RB84, could infect several legume species (Leprévost et al. ). In cross-infection scenarios, non-specific hosts generally develop root symptoms, whereas aerial ones are absent or delayed (Becking et al. ). Due to their capacity to fix atmospheric nitrogen, legumes play an essential role in sustainable agriculture. An expansion in legume-cultivated areas could also lead to a spread of A. euteiches, and consequently its impact (Ditzler et al. ). Grass pea (Lathyrus sativus) is a valuable legume crop due to its resistance to drought, flood, salinity, and several pests and diseases, and is considered a model crop for sustainable agriculture (Gonçalves et al. ). Although a major constraint on most legume crops, A. euteiches has not yet been reported as a problem on grass pea, probably due to the decline of this crop cultivation over the past decades in regions where this pathogen is endemic. However, the renewed interest in grass pea cultivation (Gonçalves et al. ; Solovieva et al. ) highlights the need to investigate the potential effect of Aphanomyces root rot in this crop. The ability of A. euteiches to produce long-lasting resting structures (oospores) that can survive in soils over 10 years without a host, combined with its capacity to infect multiple legume species, makes it particularly difficult to control. Some commercial seed treatments, such as Rancona Trio (ipoconazole, carbathiin, and metalaxyl; UPL Agrosolutions, Ontario, Canada) and Zeltera Pulse (ethaboxam, inpyrfluxam, mandestrobin, and metalaxyl; Nufarm, Alberta, Canada), inhibit A. euteiches infection at the early seedling stage, but they are ineffective when infection occurs at later plant growth stages (Willsey et al. ). The incorporation of green manures from Brassicaceae and other soil amendments (e.g., spent lime) has shown efficacy in suppressing Aphanomyces root rot, although their large-scale application is limited (Hossain et al. ). Therefore, cultural practices remain the primary management strategy for this pathogen, including long-term crop rotation of at least six to eight years with non-host species, cultivation in well-drained fields, and avoidance of soil compaction (Wikström et al. ). Additionally, determining the inoculum potential of the soil prior to sowing is highly recommended in regions where A. euteiches is endemic, avoiding planting in high infested soils (Chatterton et al. ). Overall, the development of resistant legume varieties represents a key strategy for sustainably managing this disease. Significant research efforts have focused on understanding the genetic basis of plant responses to this pathogen, leading to the identification of sources of resistance and associated genomic regions with the objective of increasing the efficiency of resistance breeding programmes (Bonhomme et al. ; Jacquet and Bonhomme ; Wu et al. ; Leprévost et al. ; Rodriguez-Mena et al. ). Genetically close legumes susceptible to Aphanomyces root rot (P. sativum, V. faba, L. culinaris and M. truncatula) have been shown to harbour syntenic QTLs (Leprévost et al. ). However, grass pea was not included in previous studies, leaving the presence of syntenic QTLs in this species unexplored. A more comprehensive understanding of the genomic conservation and diversity of resistance loci among co-cultivated legumes could support more rational crop rotation strategies to manage this pathogen with higher efficiency (Mundt ). Furthermore, intensifying the cultivation of crops that share similar genetic factors may exert selective pressure on pathogen populations, potentially accelerating adaptation and reducing the durability of host genetic resistance (Papaïx et al. ). As observed in other closely related legumes, grass pea is also expected to exhibit quantitative resistance to Aphanomyces root rot. Although monogenic complete resistance can be highly effective, the mixed reproductive system combining sexual and asexual stages of this pathogen and its high population genetic diversity make quantitative resistance a more durable long-term strategy (Mundt ; Kälin et al. ). Moreover, pyramiding multiple resistance QTLs that vary in effect across pathogen populations or act at different stages of the pathogen life cycle represents a promising approach to enhance both the resistance and its durability (Pilet-Nayel et al. ). The aim of this study was to evaluate the response of grass pea to A. euteiches inoculation, clarify the genetic basis of any depicted resistance, and identify associated genetic markers and candidate genes through a genome-wide association study (GWAS) approach. For this purpose, a worldwide diverse collection of 169 grass pea accessions was inoculated with the A. euteiches RB84 isolate under controlled conditions. Foliar and root symptomatology were assessed 20 days after inoculation. These phenotypic data were then combined with an improved acquisition of a previously generated high-density single-nucleotide polymorphism (SNP) data set to identify SNPs significantly associated with grass pea resistance to A. euteiches and to propose potential candidate genes. Material and methods Plant material The 169 grass pea accessions evaluated belong to the collection described in Sampaio et al. (). Among them, 67 accessions had Mediterranean origin, 49 South Asian, 18 East European, 15 Sub-Saharan, 12 North Asian, 2 South American, 2 North American and 4 with unknown origin. Seeds also had differences in seed colour, with 96 of them dark and 73 light (based on visual score), and in seed size, with 45 of them large (100 seed weight ≥ 18 g) and 124 small (100 seed weight < 18 g). For the experiments, seeds were scarified and surface-sterilised by immersion in a 10% of sodium hypochlorite solution for 10 min followed by rinse with water for 10 min. After sterilisation, seeds were placed in Petri dishes wrapped with moist paper. A cold shock was applied by incubating the plates at 4 oC in darkness for 24 h. Plates were then transferred to a growth chamber and kept in darkness at 23 ± 2 oC for 3 days. Germinated seeds were sown in 200 mL plastic pots (6 × 6 × 10 cm) filled with autoclaved perlite. Pots were placed in plastic trays (8 × 6 pots per tray) and maintained under controlled conditions (chamber at the Institute of Sustainable Agriculture 24 ± 2 oC, 65 RH, 12:12 h D:N with 150 μmol·m−2·s2). Plants were irrigated every 3−4 days with Hoagland nutrient solution (Hoagland and Arnon ). The accessions were evaluated using a completely randomised block design across three temporally independent rounds. Each round included four biological replicates per accession, randomly assigned to individual trays. The susceptible pea (P. sativum) cultivar “Messire” was included as a positive control in each tray, following the protocol described by Rodriguez-Mena et al. (). Inoculation and disease assessment Zoospores of A. euteiches were produced as described by Rodriguez-Mena et al. (). Zoospore concentration was adjusted to 1000 zoospores/mL using a Fuchs-Rosenthal hemocytometer. Ten-day-old plants were inoculated by applying 5 mL of the suspension at the base of each stem. Similar as it is performed in pea, foliar and root symptoms were assessed 20 days post-inoculation (Rodriguez-Mena et al. ). Foliar symptoms (FS) were rated on a scale from 0 (healthy plant) to 5 (dead plant) (Fig. a), while root symptoms (RS) were evaluated using a scale from 0 (healthy plant) to 9 (dead plant) (Fig. b). Phenotypic data analysis Statistical analyses were performed using R software version 4.4.2 (R Core Team ) with “ggplot2” package for data visualisation (Wickham ). First, outliers were detected by the interquartile range method and removed in the required cases (Estaghvirou et al. ; Tripodi ). The effect of the experimental components (round, repetition and genotype) on the evaluated symptoms was assessed with a Kruskal–Wallis test due to their ordinal nature. Then, the correlation between FS and RS was calculated using the stat_poly_eq function from “ggmisc” package (Aphalo et al. ). Effects of seed characteristics, including geographical origin, colour and size, on both symptoms were evaluated with Kruskal–Wallis test followed by a Dunn test in the required cases (p-value < 0.05) (functions Kruskal.test and DunnTest, Bonferroni method from the “stats” package (R Core Team )). Foliar and root rot symptoms were analysed as ordinal variables using a cumulative link mixed model with logit as link function (clmm function from the “ordinal” package (Christensen )) (Leprévost et al. ). Phenotypic data were adjusted to the completely random model: FS/RS = Genotype + Round + Round:Genotype + Round:Repetition where Genotype represents the 169 grass pea evaluated accessions, Round the 3 temporally separate experiments, Genotype:Round the interaction between round and genotype components and Round:Repetition the interaction between round and the 4 biological replicates contained in each round according to the performed block design. The variance components of this model were used to calculate broad-sense heritability (H2) of both symptoms using the formula: H2=σG2/[σG2+(σE2/r)] where σ2G is the genetic variance, σ2E the residual variance and r the number of temporal rounds performed (Nakagawa and Schielzeth ; Leprévost et al. ). The same model but, with genotype as fix term, was used to obtain best linear unbiased estimates (BLUEs) (emmeas function of the “emmeas” package (Lenth )). SNP data acquisition improvement DNA from each grass pea accession was previously extracted from young leaves and genotyped using two high-throughput genotyping‐by‐sequencing platforms, DArTseqTM genotyping (Kilian et al. ) and genotype‐by‐sequencing (GBS) performed by BGI (Beijing Genomic Institute, Beijing, China) (Sampaio et al. ). In the present study, to improve SNP data acquisition, we took advantage of the recently published grass pea reference genome (accession GCA_963859935.3) (Vigouroux et al. ) by calling additional SNPs from the GBS data and by mapping the DArTseq-derived markers onto the genome assembly. For GBS data, variants were identified using GBS-SNP-CROP v.4.1 (Melo et al. ). Clean, demultiplexed reads were mapped to the seven primary chromosomes of the L. sativus reference genome using the GBS-SNP-CROP-5.pl script, which implements BWA-MEM (Li ) and SAMtools (Li et al. ) to generate standard alignment files. Variant and genotype calling were then performed using the GBS-SNP-CROP-7.pl script. The resulting SNP genotype matrix was formatted for compatibility with TasselGUI (Glaubitz et al. ) for subsequent analysis. For DArTseq markers, the genomic flanking sequences were aligned to the L. sativus reference genome using the BLASTn tool (e-value < 1 × 10–5) implemented in OmicsBox v2.0, and markers with no significant hit or with multiple hits showing identical alignment probability (bit score, percentage similarity and e-value) to different genomic regions were considered unmapped and excluded from further analyses. Marker quality filtering was performed by removing those with a minor allele frequency (MAF) below 0.05 or with missing data/heterozygosity greater than 0.25, and by removing redundant markers shared between DArTseq and GBS data sets. As described in Sampaio et al. (), a subset of markers was selected by considering 0.1 Mpb interval between markers within each chromosome to be used in the principal components calculation to determine population structure. Linkage disequilibrium (LD) was estimated for all quality-filtered markers as the squared coefficient of correlation between marker pairs (r2), after correction for the population structure using the principal components scores from Eigenanalysis. These analyses were performed in Genstat® software, 24th edition (VSN International, ). Association mapping analysis BLUEs from FS and RS were used along with the high-quality-filtered SNP markers for association mapping. A single-trait genome-wide association study was performed using GAPIT 3.0 (Wang and Zhang ), testing four models: single locus mixed linear model (MLM), multiple locus mixed linear model (MMLM), multi-locus Bayesian information and linkage disequilibrium iteratively nested keyway (BLINK) and fixed and random model circulating probability unification (FarmCPU). The optimal number of principal components (PCs) was determined using the Bayesian Information criterion (BIC)-based model selection approach. Kinship matrix (VanRaden method) was also included to account for relatedness among individuals (Astle and Balding ; Zhang ). To evaluate the genomic inflation or deflation, lambda (λ) of each model was calculated (estlambda2 function of the “QQperm” package (Wang )) and Quantile–Quantile plots performed for each model and trait. Models with λ below 0.9 or higher to 1.2 were not considered. Significantly associated markers for each model were identified using false discovery rate (FDR) correction with an adjusted p value threshold of 0.05 (Chen et al. ). Unlike the more conservative Bonferroni correction, FDR provides a less restrictive alternative that increases the likelihood of detecting meaningful associations that might otherwise be overlooked (Storey and Tibshirani ; Osuna‐Caballero et al. ). P-value adjustment was performed using the p.adjust function from the “stat” package (R Core Team ). Phenotypic variance explained (PVE) was calculated from the significant associated markers using the formula: PVE=VQTL/Vphenotype in which VQTL represent the variance of the marker calculated as 2·MAF·(1 − MAF)·effect2 and Vphenotype the genotypic variance of the adjustment model (Resende et al. ). To visualise the results, Manhattan plots were generated displaying the -log10 (p-value) for each marker along the seven L. sativus chromosomes highlighting the markers with significant association. Additionally, MG2V version 2.1 was used to represent significantly associated markers distribution across the seven grass pea chromosomes (Chao et al. ). Partially resistant major allele frequency Based on their foliar symptom and root rot symptoms average values, the 169 grass pea evaluated accessions were classified into three resistance groups. For FS, the groups were defined as resistant (FS ≤ 1), moderately susceptible (1 < FS ≤ 3) and susceptible (3 < FS ≤ 5). For RS, the thresholds were resistant (RS ≤ 5), moderately susceptible (5 < RS ≤ 7) and susceptible (7 < RS ≤ 9). For each significantly associated marker, the number of reference, alternative and heterozygotes or missing values was counted within each group. The most frequent allele in resistant group was designated as “partially resistant major allele”. Its frequency was then calculated for each group by dividing the number of partially resistant major allele by the total number of genotypes in that group. Candidate genes A tentative list of putative candidate genes involved in the response was established by exploring the genomic regions surrounding significantly associated SNPs. Based on previously calculated recombination, nearby markers were checked and considered linked when r2 > 0.2. If linked markers were present, the search window encompassed the genomic interval between them. In the absence of linked markers, the candidate region was limited to the position of the significant SNP itself (Leitão et al. ). Gene annotation within these regions was manually performed using NCBI genome browser (https://www.ncbi.nlm.nih.gov/datasets/genome/GCA_963859935.3/) (Vigouroux et al. ). When possible, corresponding protein FASTA sequences of putative candidate genes were retrieved and subsequently used for functional annotation via Mercator4 v2.0 (https://mapman.gabipd.org/app/mercator) (Schwacke et al. ). Results Grass pea responses to Aphanomyces euteiches The evaluated grass pea collection presented a wide range in foliar and root symptomatology. FS average values ranged from 0.08 to 4.50 and RS from 1.64 to 8.73 (Table ). Most of the evaluated accessions presented average FS and RS superior to 1 and 5, respectively, to the tested isolate. Still, 12 resistant accessions with average FS < 1 and RS < 5 were identified, two of them being highly resistant (FS < 0.5 and RS < 2) (Fig. ). Average FS and RS values of 4.59 and 8.48 observed in the susceptible pea control “Messire” confirmed the adequacy of the inoculation. Genotype differences were mainly responsible for foliar and root symptomatology variance, with a higher broad-sense heritability associated with RS (Table ). A moderate correlation was observed between FS and RS (R2 = 0.71, p-value < 0.05) (Fig. and Fig. ). Significant differences in both symptoms were observed based on seed characteristics (Kruskal–Wallis and Dunn test pvalue < 0.05). Regarding geographical origin, Mediterranean accessions presented significantly lower RS than South Asian ones, with non-significant differences among the rest of the origin groups. With respect to seed size and colour, accessions with large and light seeds displayed lower levels of disease symptoms (Fig. and Table ). Genotyping data improvement and marker positions The DarTseqTM and GBS genotyping combined resulted in 26,876 SNPs, of which 5651 were high-quality (Sampaio et al. ). After the improvement of the original SNPs scoring by BGI and DArTseq using the reference grass pea genome (Vigouroux et al. ), a total of 44,383 SNPs were detected of which 12,974 SNPs (29.23%) passed the quality control criteria. The quality-filtered SNP markers provided a comprehensive coverage of the seven grass pea chromosomes and were used for association mapping analysis (Fig. ). SNP marker list and the marker scores per accession are available in Table . The subset of markers used to calculate the principal component is listed in Table . Genetic regions associated to grass pea resistance against Aphanomyces euteiches To detect significant markers associated with grass pea resistance against A. euteiches, a GWAS was conducted testing 4 different models: MLM, MLMM, BLINK, and FarmCPU. Although all the genomic inflation factor (λ) and Q-Q plots revealed that the population structure was efficiently controlled, only BLINK and FarmCPU detected significantly associated markers (Fig. ). A total of 20 significant markers were identified, two of them common between BLINK and FarmCPU. FS had 13 significantly associated SNPs and RS eight (Fig. ). FarmCPU detected a higher number of significantly associated markers than BLINK (Table ). Sixteen significant associated SNPs were identified by FarmCPU, 9 of them linked with FS, 6 with RS and a common one between both phenotypic traits. On the other hand, BLINK detected 6 significant associated markers, 3 associated with FS and 3 with RS. The 20 significantly associated markers were distributed along 5 of the seven grass pea chromosomes, with none of them located in chromosomes 2 or 3 (Fig. ). Markers in chromosomes 1 and 7 were only detected with FarmCPU, while chromosomes 4, 5, and 6 contained markers detected with both models. The markers SNP032149 and SNP007517, in chromosomes 5 and 6, respectively, were significantly associated with RS in both models, and SNP007517 was also associated with FS using FarmCPU (Fig. and Table ). The PVE of the significant markers detected with BLINK ranged between 4.22 to 11.62% and with FarmCPU markers from 0.13 to 9.58%. (Table ). Marker allele frequency based on symptomatology Based on the average values of foliar and root symptoms, the 169 grass pea evaluated accessions were classified into 3 categories: resistant, moderately susceptible, and susceptible. For FS, 14 accessions were classified as resistant (FS ≤ 1), 92 as moderately susceptible (1 < FS ≤ 3), and 63 as susceptible (3 < FS ≤ 5). For RS, 18 accessions were classified as resistant (RS ≤ 5), 95 as moderately susceptible (5 < RS ≤ 7), and 56 as susceptible (7 < RS ≤ 9). Differences in the partially resistant major allele frequency (most frequent in resistant group) were observed between resistant groups across the significantly associated markers, with the clearest differences occurring in FS-associated markers. For FS-associated markers, the frequency of the partially resistant major allele was significantly higher in the resistant group compared to the susceptible group in all cases except SNP035419. In contrast, for RS associated markers, this exception occurred with SNP011771, SNP032149, SNP035033 and SNP040001. In these cases, the lower frequency in the resistant groups was justified by a high number of heterozygotes or missing values among the resistant accessions (Fig. and Tables and ). In silico identification of candidate genes Using an LD decay threshold of r2 = 0.2, SNP035033, SNP035419, and SNP038082 were found to be in LD with SNP035035, SNP035415, and SNP038086, respectively. For these SNP pairs, the candidate gene search window was extended to span the region between the linked markers. In the rest of the cases, the search window was restricted to the position of the associated SNP (Table ). Seven of the twenty significantly associated markers were located within annotated L. sativus genes. Of them, three were associated with FS, three with RS and one was common to both phenotypic traits. Most of the candidate genes were located on chromosome 6 (three), while the remaining four were distributed across chromosomes 1, 4, 5 and 7, with one candidate gene identified on each of these chromosomes. Diverse biological functions were associated with the candidate genes. In chromosome 1, g3062.t1 encoded a wax ester synthase. Chromosome 4 harboured g14004.t1 which coded a component of the exocyst complex. Chromosome 5 contained g21044.t1 gene that encoded a protein implicated in fatty-acid metabolism. Chromosome 6 contained two protein kinase genes (g24166.t1 and g24512.t1) and an ethylene receptor gene (g24922.t1). Finally, g29459.t1, a callose synthase coding gene, was located on chromosome 7 (Table ). Discussion This study characterised the response of a diverse global grass pea collection to A. euteiches, confirming grass pea susceptibility to this pathogen and identifying two resistant and 10 partially resistant accessions in this crop. The broad range of FS and RS observed in the collection presented a suitable base, together with an improved acquired SNP data set, to study the genetic architecture behind this response. The genome-wide association study identified 20 significantly associated markers linked to resistance across five of the seven grass pea chromosomes, confirming its polygenic nature and allowing the proposal of potential candidate genes involved in the response. To our knowledge, this is the first application of association analysis to explore grass pea resistance to A. euteiches. A wide range in FS and RS was observed among the 169 evaluated grass pea accessions, confirming the susceptibility of this crop to A. euteiches. Nevertheless, when comparing RS severity with results from similar pea infection studies, grass pea appears to be, on average, less susceptible than pea, with most of the evaluated accessions classified as partially susceptible in both of the evaluated symptoms. Also, as previously described in pea, a significant moderate correlation was observed between FS and RS (Rodriguez-Mena et al. ). In contrast to pea, this study identified two accessions with very high levels of resistance, a situation that has never been described in pea, where only partial resistance has been described (Desgroux et al. ; McGee et al. ; Rodriguez-Mena et al. ). Notably, grass pea accession PI283570_A and accession PI283580 exhibited complete resistance to the tested isolate, supporting the potential existence of intra-specific genetic variability in response to A. euteiches infection, as previously reported in faba bean (Leprévost et al. ), or the existence of race-resistant accessions in alfalfa (Giles et al. ). Otherwise, the pea origin of the tested isolate caused a non-host-specific and less aggressive infection (Moussart et al. ); the identification of grass pea-specific isolates could further validate these results. Additionally, ten accessions showed partial resistance. One of these partially resistant accessions is STUDENICA, a commercial variety with desirable agronomic traits that presents a promising candidate for cultivation in soils infested with A. euteiches (Gonçalves et al. ). Among the tested resistant accessions, PI283580, BGE17185 and PTLS1006 also demonstrated low symptom levels in response to another soil-borne pathogen, Fusarium oxysporum f. sp. pisi (Sampaio et al. ). Given the high frequency of co-infection of A. euteiches and Fusarium spp. in pea, accessions with resistance to both pathogens represent valuable resources for breeding broadly resistant cultivars (Sharma et al. ; Trenk et al. ). Regarding seed characteristics, light-coloured, large seed accessions showed lower FS and RS compared to dark, small seed accessions. This aligns with the lower RS observed in Mediterranean seeds, which are typically larger and lighter, compared to the predominantly small, dark seeds of South Asian origin. The same characteristics were associated with more resistant accessions to Fusarium oxysporum f. sp. pisi in the evaluated collection that could indicate similar mechanisms behind grass pea resistance to these two soil-borne pathogens (Sampaio et al. ). The broad phenotypic variation among the evaluated grass pea after A. euteiches inoculation, together with the high heritability associated with both foliar and root symptoms, supports the suitability of this collection to perform genetic studies. The performed GWAS identified 20 significant molecular markers, 12 associated with FS, 7 with RS, and one common marker between both symptoms. Among these, two markers, SNP032149 and the common marker between symptoms SNP007517, were consistently detected by two models (BLINK and FarmCPU), reinforcing their potential relevance to Aphanomyces root rot resistance in grass pea. Clear allelic differences between resistant and susceptible groups were observed for the associated markers. These differences were more pronounced for foliar-related SNPs than for root-related ones, likely due to a higher incidence of missing/heterozygotes in accessions classified as resistant based on RS. The significant markers identified in this study were distributed across five of the seven grass pea chromosomes. This pattern, together with the moderate proportion of PVE per marker, supports the notion that complex polygenic resistance to A. euteiches is present in grass pea, as previously reported in other legumes (Leprévost et al. ). However, the resistance nature of the two described resistant accessions requires further investigation to determine whether they exhibit monogenic or polygenic resistance. Given the high genetic diversity in A. euteiches populations and its hemibiotrophic lifestyle, polygenic quantitative resistance may provide a more durable long-term management strategy than monogenic based resistance (Mundt ; Kälin et al. ). The GWAS highlighted six regions, two on chromosome 5 and four on chromosome 6, that harbour closely located significantly associated markers, suggesting the presence of potential quantitative trait loci associated with Aphanomyces root rot resistance in grass pea. The chromosome 5 regions, spanned from 193,375,403 (SNP005290) to 241,052,867 (SNP005347) and from 508,589,669 (SNP031472) to 561,173,187 (SNP032149). The four regions of closely located markers in chromosome 6 were from 395,296,031 bp (SNP035346) to 408,765,593 bp (SNP035419), a second from 478,139,587 bp (SNP035948) to 478,197,292 bp (SNP035962), a third from 525,489,889 bp (SNP007517) to 525,490,364 bp (SNP007520) and the last from 616,109,525 bp (SNP038080) to 616,109,553 bp (SNP038082). The chromosomal distribution of associated markers differed from those in other susceptible species (pea, lentil, faba bean or barrel medic), lacking the identification of a conserved genomic pattern with previous studies (Leprévost et al. ). Further investigation is required to confirm the roles of these candidate loci. Examining the genomic location of the significantly associated markers led to the identification of 7 candidate genes potentially involved in A. euteiches resistance in grass pea. These genes are predicted to participate in various biological processes, including vesicle trafficking (g14004.t1), signal transduction (g24166.t1), nucleotide metabolism (g24512.t1), fatty-acid metabolism (g3062.t1 and g21044.t1), cell wall synthesis (g29459.t1) and ethylene signalling (g24922.t1). Notably, serine/threonine protein kinases (g24166.t1) and ethylene-related proteins (g24922.t1) have previously been associated with A. euteiches response in other legumes (Nyamsuren et al. ; Afzal et al. ; Badis et al. ; Desgroux et al. ; Jacquet and Bonhomme ; Rodriguez-Mena et al. ; Pandit et al. ). Serine/threonine protein kinases are involved in signalling and plant defences, while the ethylene pathway leads to lignin accumulation (Djébali et al. ; Desgroux et al. ). Lignin accumulation enables the formation of a protective ring around the central cylinder in infected roots, restricting pathogen entry and preserving root function (Djébali et al. ). Regarding the other candidate genes, evidence from non-Aphanomyces root rot studies suggests that they may also play defence-related roles against A. euteiches. For instance, infection increases the presence of several fatty acids (g3062.t1 and g21044.t1) in pea which has been proposed to mediate reactive oxygen species (ROS) production, contributing to R gene–mediated resistance (Hossain et al. ). Similarly, a gene involved in callose synthesis (g29459.t1) may play a role in reinforcing cell walls at infection sites, as callose and phenolic compound deposition have been shown to limit fungal hyphal growth at the infection point (German et al. ). Although components of the exocyst complex (g14004.t1) have been associated with plant immunity, their specific role in A. euteiches resistance remains poorly understood (De la Concepcion ). Likewise, the contribution of the in silico detected genes required further investigation. In planta validation of these genes could clarify their role in the resistance and contribute to characterise the mechanism behind grass pea response to A. euteiches. Phenotypic evaluation of grass pea under field conditions is required to assess the potential impact of A. euteiches in natural environments and to validate the identified markers and associated candidate genes. Evaluating additional A. euteiches isolates would also clarify whether the observed susceptibility in grass pea is specific to RB84 or applicable to other isolates. Such analysis would also reveal whether the genomic regions identified in this study have a major or minor effect depending on the tested pathotype, as occurs in pea (Leprévost et al. ). The RB84 isolate was selected in this study due to its documented capacity to infect several legume species, confirming also in this work its ability to infect grass pea (Leprévost et al. ). However, it is a pea isolate, which implies that the characterised response is non-host specific. Should grass pea isolates exist, they might elicit a stronger response and potentially overcome the complete resistance identified in this study. Altogether, this study confirmed a compatible interaction between grass pea and A. euteiches RB84 isolate, but with variable levels of polygenic resistance. The tested collection exhibited a wide range of foliar and root symptomatology after inoculation, allowing the identification of resistant and partially resistant accessions, which can be exploited in future breeding programmes. The performed GWAS allowed the identification of significant associated SNPs on five of the seven grass pea chromosomes, explaining only a percentage of the observed variability, consistent with the quantitative nature of resistance to A. euteiches described in other legumes. A notable concentration of significant markers on chromosomes 5 and 6, suggests the presence of resistance-related loci in these regions. Additionally, the significant associated markers enabled the identification of seven candidate genes potentially involved in the grass pea response to A. euteiches. Among them, genes encoding serine/threonine protein kinases and ethylene receptors are noteworthy due to their previously reported roles in A. euteiches defence in other legume species. Nevertheless, functional validation is required to clarify their role in the response and to guide their use in precision breeding approaches. These might also be potentially valuable for pea breeding if crossability barriers are overcome. Therefore, this study opens new avenues to study grass pea response against A. euteiches as well as proposed resistance-associated loci and candidate genes involved in the response. Author contribution statement SRM contributed to experiment design, phenotypic data collection, phenotypic and genotypic data curation, data analysis and visualisation and wrote the original draft; MG contributed to conceptualization, experiment design, phenotypic data acquisition, supervision and writing review and edition. EV contributed to genomic data curation and analysis; CS contributed to genomic data curation and analysis, writing review and edition; DR contributed to conceptualization, writing review and edition, funding acquisition and project administration; MCVP contributed to conceptualization, supervision, funding acquisition and writing review and edition. Footnotes Contributor Information Sara Rodriguez-Mena, Email: srodriguez@ias.csic.es. Mario González, Email: mgonzalezromero@ias.csic.es.