Evaluation of Lens culinaris germplasm against ascochyta blight by applying phenotyping and defense genes attributes Authors: Mohsen Modareskia, Hamid Hatami Maleki, Behrouz Shiran, Hamid Reza Pouralibaba Published: 2025 ABSTRACT AbstractLentil Ascochyta blight (caused by Ascochyta lentis) is one of the most important limiting factors of lentil cultivation and production in most regions of the world. Introducing resistance sources against the pathogen is a suitable strategy to conquer this biotic stress. In this study, A. lentis-lentil pathosystem was inspected through the evaluation of 79 lentil genotypes in greenhouse conditions. Afterward, germplasm population structure as well as genomic regions conferring in differentiation of disease response types were determined by defense-genes related SSRs. Also, comprehensive analysis of the lentil defense system in facing A. lentis was done by measuring transcripts of defense-related genes of PR-2 and RBP-hnRNPs at 24, 48, and 72 h post inoculation. Non-parametric analysis of variance showed significant differences among studied genotypes regarding disease severity and AUDPC, where 69.62%, 21.51%, and 8.86% of them showed susceptible, moderate resistant, and resistant reactions, respectively. Genomic fingerprinting of genotypes by using 9 SSR loci specific to defense genes, manifested 51 alleles. Among the studied SSR loci, RRM1 with PIC values of 0.74 possessed the highest efficacy in evaluation of the lentil germplasm. Discriminant analysis of principle component analysis, showed efficacy of defense-genes related SSRs in prediction of R, S, and MR reactions that were determined phenotypically in greenhouse conditions. Combination of genetic algorithm with quadratic SVM as supervised machine learning method revealed allele PP2C-7 of gene PP2C (has act in pathway of plant stress signaling) as important feature for prediction of lentil reaction types against A. lentis. Study of molecular response against A. lentis showed overexpression of genes RBP-hnRNPs (transcriptional factor) and PR-2 (anti-fungal compound), especially at 48 hpi and 72 hpi for resistant genotype. In total, screening lentil germplasm is continuous work regarding the co-evolution of pathogens and resistant genes, and in this way, artificial intelligence through defense genes relevant SSRs could reliably identify resistance type genotypes. AB-resistant lentil genotypes respond rapidly and rigidly against infection compared with susceptible genotypes.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-27651-0.Keywords: Ascochyta blight, Lentil, Defense-gene related SSR, Machine learning, Gene expressionSubject terms: Plant breeding, Plant stress responses Introduction Lentil (Lens culinaris L.), as one of the famous crops in the family Fabiaceae, possessed a high content of protein, and by having symbiosis with nitrogen-fixing bacteria, it could serve in soil fertility1. The breeding objective of lentil is to introduce genotypes with high yields and resistant reactions against biotic and abiotic stresses. In this regard, several biotic stress agents threatening lentil, among them Ascochyta blight (AB) of lentil caused by Ascochyta lentis, is a devastating necrotrophic fungal disease2. This foliar disease affects lentil`s seed quality as well as its seed yield, so that yield reductions of even up to 40% have been reported3. Albeit, either field practices or chemical control have been proposed, but none of them is efficient compared with introducing resistant genotypes4. Meanwhile, the last case is durable and eco-friendly as well. Several studies were carried out to evaluate the response of lentil germplasm against A. lentis5 with the aim of finding new resistance sources. According to Dadu et al.5, altering the aggressiveness of A. lentis and also the narrow genetic variability of the existing germplasm has resulted in pathogen overcoming even into resistant sources, and therefore, it is mandatory to continuously screen lentil germplasm to achieve stable lentil production. However, it is proved that expansion of lentil genotypes with multiple resistance genes could also mitigate the chance of simultaneous pathogen mutation for virulence factors for two or more effective genes. Accordingly, it is possible to integrate several resistance genes into interested genotypes6 and conquer this defect. Considering disadvantages of conventional methods in incorporation of resistance genes into lentil genotypes, a DNA marker-based approach has been addressed in an applicable manner6. Regarding mapping QTLs linked with AB using F27 and recombinant inbred lines8 populations, major QTLs with R2 values over 40% are employed in controlling resistance reactions and linked with resistance to AB in lentil. Hence, by the contribution of DNA markers linked with different resistance genes, it is possible to produce versatile lentil genotypes that possess multiple resistances through resistance gene pyramiding. In addition to identifying resistance resources against A. lentis, it is crucial to determine how the AB pathogen and host interact together in the way of defense or pathogenicity. In the facing of plant and pathogen, the activity of the pathogen is focused on conquering the host and consuming its resources, while the plant tries to recognize the pathogen and activate its defense mechanisms to deal with it and stop the microbial attack9,10. Hence, one of the important factors in plants after contamination is the speed of their response to the infection, and therefore, rapid and effective induction of defense responses distinguishes the resistant plant from the susceptible types11. Several studies related to the response of lentil against A. lentis showed that lentil recognize attacks from this fungal agent as soon as they contact it2,12. Accordingly, Mustafa et al.13, used defense-related cDNA probes from other leguminosae species to show how several resistant genes helped lentil accession “ILL7537” fight off A. lentis at 2, 6, and 24 h after infection (hpi). Several evidences2,14 have shown the existence of three distinct steps, including pathogen recognition (2 hpi), induced defense responses (6 hpi), and necrotic structural defense reactions (24 hpi) in lentil-A. lentis interaction. The aforementioned steps have also been verified by measuring physiological parameters such as reactive oxygen species (ROS) in lentil following infection by A. lentis2 which result in two main sections of morphological and physiological changes. Nowadays, new generation of sequencing technologies such as RNA sequencing facilitates comprehensive assessment of lentil-A. lentis interaction9,10,12,15 and in this way, several defense related genes were identified. For instance, Khorramdelazad et al.12 emphasized the key role of RBP-hnRNP (nuclear transcriptional factor) and PR-2 (anti-fungal compound) as candidate genes for controlling resistance against A. lentis in the genotype of “ILL7537” as resistant genotypes. Sari et al.9 reported that a signal transduction network arises after invasion of A. lentis and in this regard, preventing the ABA signaling pathway by A. lentis could suppress the salicylic acid, jasmonate and ethylene pathways as downstream defense responses. Sari et al.9 inferred that the resistant lentil genotype of “CDC Robin” possessed abundant frequencies of the candidate genes of receptor-like protein kinases, classified as PRR (the trans-membrane pattern recognition receptors), which recognize pathogen-associated molecular patterns (PAMP)16 and induce resistant responses. Findings of Garcia-Garcia et al.15 were also emphasized on the critical role of the jasmonate pathway in the resistance reaction of lentil against A. lentis. Moreover, Cao et al.10 using a comparative transcriptome study of susceptible and resistant lentil genotypes manifested that the suppression of gibberellin signaling and cell death associated with the hypersensitive response (HR) were linked with greater A. lentis resistance. This study aimed to (i) screen lentil germplasm, including exotic and indigenous genotypes, against A. lentis isolates identified from hotspots in Iran; (ii) identify SSR loci capable of predicting S, MR, and R lentil genotypes; and (iii) study the molecular response of identified susceptible and resistant genotypes by analyzing the expression of selected defense genes at several time points after inoculation. Materials and methods Plant material, fungal pathogen, and greenhouse experiment A total of 79 lentil genotypes obtained from the Iranian national lentil breeding program (Table 1) were tested for their resistance to A. lentis. In May 2018, lentil plants naturally infested with A. lentis were collected from fields located in Moghan (Ardebil province), Kamyaran (Kurdistan), Shirvan-Chardavol (Ilam province), and Gachsaran (Kohgiluyeh and Boyer-Ahmad province) regions of Iran. The stubbles were placed into large paper bags (20 cm in depth) and transferred into the lab. Infested leaves, stem and pod segments (1 × 1 cm) were produced, then surface sterilized in 0.5% NaClO for 10 min, dried on sterile paper, and cultured on lentil seed meal-dextrose agar (LDA: 2% lentil seed flour, 2% dextrose, and 2% agar in 1 L deionized water) amended with 1 ml/L chloramphenicol and incubated at 20 ± 2 °C, 12 h light/12 h dark with fluorescent lights (100 µmol m-2 s-1) supplemented with neagray radiation to stimulate sporulation. After one week, a dark gray colony appeared, and it was purified using the single spore method. The fungus was identified through colony morphology and spore dimension. For long-term storage, the purified culture was maintained on autoclaved stem pieces of lentils17. Conidia of seven-day-old colonies of A. lentis were prepared on LDA. Petri dishes were flooded with sterile distilled water, the surface of the culture was gently scratched with a sterile needle, and after macerating and squashing in a clay mortar, the harvested segments were transferred into an Erlenmeyer containing sterile water. The liquid was filtered through a four-layer cheesecloth, transferred into Falcone tubes, and centrifuged at 10,000 rpm for 10 min. The harvested conidial suspension was fixed at 1 × 106 spores per mL using a hemocytometer. Table 1Genotype`s genebank code and pedigree, disease reaction against AB, and inferred Q values obtained by population structure analysis (probability that an individual belongs to a subpopulation) of studied lentil genotypes. G37 (“Gachsaran”) is susceptible check in the experiment.CodeGenotype (pedigree)Disease reactionQ matrixCodeGenotype (pedigree)Disease reactionQ matrixCodeGenotype (pedigree)Disease reactionQ matrixQ1Q2Q1Q2Q1Q2G1FLIP 2011–5 L (ILL 6434 × ILL 6972)S0.9870.013G28SanaS0.9850.015G55×2009-5160K3R0.0310.969G2ILL 4605 × ADDA (2006-03-0GA-0GA-0GA-11)S0.9930.007G29KimiaS0.9740.026G56×2009-5160K1S0.0140.986G3ILL 6434 × ILL 8008 (2006-030G-0GA-0GA-11)S0.8430.157G30FLIP2011-43 L(ILL10947)-ILL 7537 X ILL 590S0.9920.008G57×2009-5212K2S0.0320.968G4ILL 10,707R0.9840.016G31FLIP2014-021 L(ILL11431)-ILL9977 x ILL 1005S0.9910.009G58×2009-51146K3S0.0170.983G5FLIP2003-2 LMR0.9160.084G32FLIP2014-032 L(ILL11442)-ILL5883 x ILL6458S0.9840.016G59×2009-5237K1S0.0170.983G608 S 40111-01S0.9790.021G33FLIP2012-244 L(ILL11219)-ILL7711XILL5480S0.9940.006G60×2009-5210K1S0.0120.988G709 S 96510-13-X2007S69-ILL8072 x ILL7162S0.0160.984G34FLIP2014-103 L(ILL11513)-ILL9892 x ILL7978 ICARDA 3 111 139 45 4.0 732 98 CS0.9950.005G61×2009-5223K4S0.0260.974G8010 S 96129-3-X2007S119-ILL10005 x ILL1005S0.0170.983G35ILL8006S0.9720.028G62ILL 7947R0.0180.982G909 S 82109-01-X2006S147-ILL1005 x ILL5883R0.0140.986G36FLIP2010-95 L(ILL10825)-ILL 7620 × 91,517S0.9830.017G6309 S-96S10-12R0.0070.993G10ILL 5988-MR0.010.99G37GachsarnS0.9590.041G6409 S-83184-01MR0.0060.994G1109 S 82109-05-X2006S147-ILL1005 x ILL5883MR0.0090.991G38SepehrMR0.9930.007G65ILL 6002MR0.0080.992G12ILL 6536-S0.010.99G39Yazd LocalS0.9960.004G66ILL 1323S0.0090.991G13ILL 6538-S0.8880.112G40BilesavarMR0.9940.006G67ILL 5883S0.0060.994G14ILL 4605-S0.9120.088G41FLIP97-10 LR0.990.01G68FLIP2012-8 LS0.0080.992G1509 S 96510-21-X2007S69-ILL8072 x ILL7162MR0.9760.024G420854-124-8MR0.9760.024G69FLIP2002-9 LS0.0080.992G16ILL7978-S0.9810.019G430854124-08S0.9930.007G70FLIP2002-15 LS0.0090.991G172009 S 96518-1-X2007S107-ILL7940 x ILL5883MR0.990.01G44FLIP2011-45 LS0.990.01G71FLIP2002-25 LS0.0120.988G182009 S 96501-1-X2007S61-ILL5883 x ILL6458S0.7520.248G45FLIP2011-42 LMR0.9920.008G72FLIP2002-29 LS0.0140.986G1909 S 83253-04-X2006S280-ILL8009xILL7711MR0.9870.013G46FLIP97-6 LMR0.990.01G73FLIP2014-021 LS0.010.99G2008 S 41226-02-X2005S206-X2004S99xX2004S27S0.9920.008G47FLIP2007-45 LS0.990.01G74FLIP2012-3 LS0.0090.991G212009 S 96501-5-X2007S61-ILL5883xILL6458MR0.9860.014G4808540124-08MR0.2970.703G75FLIP2012-196 LMR0.0080.992G22ILL 4605-S0.9830.017G49FLIP2003-2 LS0.4760.524G76FLIP2012-207 LS0.0090.991G2309 S 83194-04-X2006S149-ILL6434 x ILL5883S0.9850.015G50FLIP2011-45 LS0.0410.959G77FLIP2012-245 LR0.0080.992G242009 S 96101-4-S0.9820.018G51FLIP2007-55 LS0.2150.785G78FLIP2013-13 LS0.0210.979G25ILL 6002-S0.9760.024G52FLIP2007-95 LS0.0150.985G79FLIP96-59 LMR0.0180.982G26X2012S-154-X2012S-154-ILL6002xILWL118S0.9880.012G538068R0.0440.956G27X2012S-102-X2012S-102-(ILL7986 X ILWL 74)X(ILL4605 X ILL5677)S0.9780.022G54FLIP2007-56 LS0.0850.915 Genotype`s genebank code and pedigree, disease reaction against AB, and inferred Q values obtained by population structure analysis (probability that an individual belongs to a subpopulation) of studied lentil genotypes. G37 (“Gachsaran”) is susceptible check in the experiment. A pathogenicity test of recognized isolates was done on cultivar “Gachsaran”. The identified aggressive isolate was applied for phenotyping of lentil germplasm in greenhouse condition. In the process of phenotyping, the seeds of genotype were surface sterilized in 0.5% NaClO for 5 min, dried on paper and sown in small (5 × 8 × 6 cm) pots, two seeds per pot, filled with pasteurized clay loam soil (3:2:1 clay soil: river bed sand: compost). Three pots were considered for each genotype. Plastic trays measuring 97 × 55 × 27 cm were used to hold the pots, which were filled with pasteurized river-bed sand to a depth of 5 cm at the bottom. The plants were watered twice a week through this layer of sand. The trays were kept at 20 ± 2 °C, 12 h light/12 h dark photoperiod at 200 µmol m-2 s-1. The grown plants at the 4–6-leaf stage were inoculated with the spore suspension containing surfactant Tween 20 (0.01%) with a hand sprayer. After inoculation, the pods were kept under a plastic bag for 48 h to maintain leaf wetness under the above-mentioned temperature/light conditions. The disease symptoms were evaluated using the nine-digit scale18 after observing the first disease symptoms in the susceptible check (“FLIP2002-15L”), three times with one week interval. In the field condition, the aforementioned lentil germplasm was evaluated in Shirvan-Chardavol (33°36′N 46°40′E) in Ilam province of Iran. This region calculated as hotspot point in rainfed regions which AB disease occurs frequently. All genotypes conducted in randomized complete block design with two replication. In the following, artificial contamination using the same isolate which undertaken in greenhouse experiment was done and disease severity as well as AUDPC were measured. Designing SSR primers correspondence to A. lentis defense related genes In this study, the short read sequences correspondence to transcriptome study of lentil genotypes ILL7537 (resistant to A. lentis) and ILL6002 (susceptible to A. lentis)12 was downloaded and undertaken for further analysis. This data is available through National Center for Biotechnology Information (NCBI), Sequence Read Archive (SRA) with an accession number of SRP075524. After quality control and trimming of the reads through FastQC and Trimmomatic, de novo assembling was done using Trinity (Haas et al., 2013), to establish the full lentil defense-response transcriptome. Afterward assembling, 348,775 contigs were identified, which after differential gene expression analysis led to 3872 sequences (genes). The finally identified sequences were aligned across both non redundant and Uniprot databases to identify which sequence could achieve the 27 resistant genes predicted by Khorramdelazad et al. (2018). In this regard, 152 sequences were attributed to 25 out of 27 predicted defense related genes (Table S1). The alignment of 152 sequences with the SSR primer database of lentil19 through blastn revealed that 30 sequences out of 152 sequences belonged to 13 defense-related genes (Supplementary Table 1) and have been possessed SSR primers. Among these 30 sequences, 9 sequences from different defense-related genes (Table 2) were considered for further analysis, and their related SSR primers were implemented for fingerprinting of studied lentil germplasm using the method described in detained by Koul et al.20. Table 2(A) Pimer name, sequence, and molecular characters of primers designed regarding pre-defined defense-related genes and their related genomic information such as contig and locus ID. (B) sequences of defense gene and housekeeping gene used in qRT-PCR assay.PrimerDefense geneaContig IDLocus IDbSequenceAnnealing temp.Number of amplified allelesPICA RRM1.1RRM/RBD/RNP motif & hnRNPsTRINITY_R_lens33658_c0_g2_i1Locus_6418F: CGACCTTTGCAACAAACTCR: TTTCACCGATTGTTTCGTCA5870.74 Glyco81PR-2-O-glycosyl hydrolase family 17TRINITY_S_lens35815_c0_g1_i1Locus_22616F: CGCAGCAACAACATCAATTR: GAAAAGCATCCCCTTCACA5820.02 DEADP72 DEAD boxTRINITY_S_lens36802_c0_g1_i1Locus_8662F: GGTGATCCCAGCACGTAGTTR: ATCCAATGAGCGGAATCAAG5830.39 PP2CProtein phosphatase 2 ATRINITY_R_lens38450_c2_g2_i3Locus_11151F: CCACCTACTTCACCACCACR: GAAATCGGAACCAAAGCAAA58160.68 PRM1.2RRM/RBD/RNP motif & hnRNPsTRINITY_S_lens38932_c0_g1_i1Locus_10729F: AACACCAACACCAACACCAAR: TCTCCAGAAAGGACGCTTGT6050.67 zfRing-UBOXRING/U-box proteinTRINITY_S_lens40584_c2_g2_i9Locus_14141F: GCAAAGAGGGCACTGAAAAGR: GCACCAACCAAGCCATTATT5990.44 TMK1Receptor-like S/T kinase1TRINITY_S_lens42052_c5_g3_i1Locus_7737F: GATGGAGCTGATAACCGGAAR: ATCTGCCGACGTGAAAGAAT5930.16 ERF81Ethylene-responsive TFTRINITY_R_lens34326_c0_g2_i4Locus_12830F: ACATTGAAGGTGGTGGTGGTR: CAAGGACAGAATCTCCGAG5930.12 PGIP1Polygalacturonase inhibitorTRINITY_R_lens36289_c2_g1_i3Locus_4979F: GGGAAACTCCCATTTTCGTTR: ACTTGCAGCTTGGAAGAGG5830.16B PP2AReference geneF: GCCTCATTTGCAGCTGGTTTR: TACTCTCGTTCTAGGGTCCT RBP-hnRNPTranscriptional factorF: GAGAAAGATATTTGTTGGAGR: TGATCGTACATTACTACAACA PR2Anti fungal compoundF: GGCATGCTGGGAAACAATCTR: TGGCACACCTAACATGAGCT (A) Pimer name, sequence, and molecular characters of primers designed regarding pre-defined defense-related genes and their related genomic information such as contig and locus ID. (B) sequences of defense gene and housekeeping gene used in qRT-PCR assay. F: CGACCTTTGCAACAAACTC R: TTTCACCGATTGTTTCGTCA F: CGCAGCAACAACATCAATT R: GAAAAGCATCCCCTTCACA F: GGTGATCCCAGCACGTAGTT R: ATCCAATGAGCGGAATCAAG F: CCACCTACTTCACCACCAC R: GAAATCGGAACCAAAGCAAA F: AACACCAACACCAACACCAA R: TCTCCAGAAAGGACGCTTGT F: GCAAAGAGGGCACTGAAAAG R: GCACCAACCAAGCCATTATT F: GATGGAGCTGATAACCGGAA R: ATCTGCCGACGTGAAAGAAT F: ACATTGAAGGTGGTGGTGGT R: CAAGGACAGAATCTCCGAG F: GGGAAACTCCCATTTTCGTT R: ACTTGCAGCTTGGAAGAGG F: GCCTCATTTGCAGCTGGTTT R: TACTCTCGTTCTAGGGTCCT F: GAGAAAGATATTTGTTGGAG R: TGATCGTACATTACTACAACA F: GGCATGCTGGGAAACAATCT R: TGGCACACCTAACATGAGCT Gene expression assay Total RNA was extracted from the leaves of selected susceptible (“08S 40111-01”) and resistant (“ILL 7947”) genotypes at 24, 48, and 72 hpi (hours post inoculation) using the TMRNX plus kit (Cinagen, Iran) according to the manufacturer’s instructions. Extracted total RNA (34 µg for each reaction) was treated with the DNaseI enzyme to remove genomic DNA. The quantity and quality of extracted RNA were determined using a spectrophotometer and special RNA gel. Using a nanodrop, the concentration of RNA samples was equalized with each other, and then RNAs with a concentration of 250 ng/µl were used to make cDNA using a cDNA synthetase kit (Fermentas, USA) according to the manufacturer’s instructions. The synthesized cDNA was stored at −24 °C. The expression level of genes, including RBP-hnRNP (XM_051047539.1) as a transcriptional factor and PR-2 (JX475046.1) as anti-fungal compound12 which previously proved as differentiative genes among susceptible and resitant genotypes of lentil was measured using Real Time PCR method through Q-gene Rotor from Qiagene through 6 technical replications (Supplmentary file). The PCR reaction was performed with a final volume of 10 µl. The reaction program included: 95 degrees for 5 min and then during 04 cycles with 95 degrees for 34 s, 52 degrees for 24 s and 72 degrees for 24 s. Melting curve modeling was done from 14 to 95 degrees at the end of the reaction for each PCR to determine the specificity of the reaction. In this research, PP2A which proven having stable expression patterns at moderate levels across all time points and experimental groups12, was used as a reference gene. Finally, the expression level of studied genes accompanied with the reference gene was calculated through the delta CT method21. Statistical analysis Disease assessment was performed using procedure introduced by Nasir and Bretag18. Accordingly, score 1 = no visible disease symptoms; score 3 = leaf lesions only, chlorosis of affected leaves < 10% leaf drop; score 5 = leaf lesions, up to 25% leaf drop, stem flecks or lesions < 2 mm; score 7 = leaf lesions, up to 50% leaf drop, stem lesions > 2 mm; score 9 = leaf lesions, potential defoliation, stem girdling, potential plant death. Genotypes with a mean score of 3.0 or less for every isolate tested were considered resistant. Those with mean scores from 3.1 to 4.9 were considered to be moderately resistant, and those with a mean score of 5.0 and above were considered susceptible. The values for the areas under the disease progress curve were calculated in Microsoft Excel software, following this formula: Where yi is an assessment of a disease (percentage, proportion, ordinal score, etc.) at the ith observation, ti is time (in days, hours, etc.) at the ith observation, and n is the total number of observations. We performed the Kruskal-Wallis test as a nonparametric alternative to the randomized block design for the analysis of variance. In order to calculate molecular marker characteristics and discriminant analysis of principal component (DAPC) analysis, the nucleotide size of the amplified SSR allele was converted into 1 for presence and zero for absence of each allele in allelic format. Hence, the polymorphism information content (PIC) value was calculated for each primer in Microsoft excel software following this formula: PIC = 2ƒ (1 − ƒ). In which:ƒ= the frequency of present bands in the develop-ing gel; and 1-ƒ= frequency of absent bands. The R package of “adegenet”22 was used for DAPC analysis. In the following, different degrees of SVM (Support Vector Machine) and KNN (K-Nearest Neighbor) as supervised learning algorithms accompanied with genetic algorithm23 as feature selection algorithm were utilized to predict predefined groups of R, S, and MR. In this way, 51 SSR alleles produced by 9 SSR loci were considered features. In this process, 5-fold nested cross-validation was implemented to prevent data leakage and optimize generalization. Accordingly, dataset partitioned into 5 equal folds (outer loop) and then for each iteration, 4 folds used for feature selection + model training and one held-out fold for final validation. Final performance metrics represent the mean (± SD) across all 5 test folds. Finally, genetic algorithm applied for feature selection. This process was done through Matlab 2019b. Next, software STRUCTURE has been implemented to infer population structure among the studied genotypes, using a model-based Bayesian approach. In this regard, 10 independent runs were performed with the number of sub populations (K) set from 1 to 10, burn in time and MCMC (Markov Chain Monte Carlo) replication numbers both set to 500,000, and a model for admixture and correlated allele frequencies. Delta K(ΔK) based on the second order rate of change in the likelihood24 was used to represent the K value. Inferred ancestry estimates of individuals (Q matrix) were derived for the selected sub populations25. Results Lentil germplasm screening through A. lentis and defense-genes related SSRs In this study, four isolates of A. lentis were recognized in naturally infected fields of lentil. Pathogenicity test of A. lentis isolates on cultivar “Gachsaran” showed that the highest aggressiveness belonged to the Moghan isolate (data not shown). Hence, the aforementioned isolate was used for inoculation of studied lentil germplasm and further evaluation (Fig. 1). In this research, Kruskal-Wallis as a non-parametric method of analysis of variance, which is equal to parametric randomized complete block design, showed significant divergence among studied lentil genotypes considering DS and AUDPC (Table 3). For more information, the mean and standard error of DS and AUDPC for each genotype were addressed in Table S2. Simple distribution of DS and AUDPC values across lentil genotypes (Fig. 2) depicted genetic variability of studied lentil germplasm against A. lentis. Neither of the studied lentil genotypes showed an immune reaction against isolate Moghan (Table 1; Fig. 1). Among studied genotypes, genotype G6 (“08S 40111-01”) was detected as susceptible, with a maximum value of AUDPC (133) as well as DS (9). Likewise, the minimum values of DS (1.5) and AUDPC (22.75) were calculated for G62 (“ILL 7947”) (Fig. 2). Results showed that the majority of Iranian commercial cultivars, including G28 (“Sana”), G29 (“Kimia”), G37 (“Gachsaran”), and G39 (“Yazd”) were susceptible to A. lentis and just G38 (“Sepehr”) and G40 (“Bilesavar”) showed MR reactions. In general, 69.62%, 21.51%, and 8.86% of studied lentil germplasm possessed S, MR, and R reactions, respectively (Fig. 2). Interestigly, in the field condition (Table S3), the majority of genotypes showed resistant response albeit some susceptible and moderately resistant genotypes were also detected. Experimental execution in the greenhouse, (A) Lentil collection inoculated with A. lentis, Moghan isolate; (B) Phenotyping of disease reaction. For instance, b1- b5 get scores 1–5 regarding Nasir and Bertag (1998) scoring system as follows: b1 shows score 1 = no visible disease symptoms; b2 shows score 3 = leaf lesions only, chlorosis of affected leaves, < 10% leaf drop; b3 shows score 5 = leaf lesions, up to 25% leaf drop, stem flecks or lesions < 2 mm; b4 shows score 7 = leaf lesions, up to 50% leaf drop, stem lesions > 2 mm; b5 shows score 9 = leaf lesions, potential defoliation, stem girdling, potential plant death. Table 3Non-parametric analysis of variance for disease severity score and AUDPC by means of Kruskal-Wallis statistic.StatisticResistance criteriaDSAUDPCChi-square165.051183.03df7878P-value0.000.00 Non-parametric analysis of variance for disease severity score and AUDPC by means of Kruskal-Wallis statistic. AUDPC (blue bars) and DS (yellow bars) values of lentil genotypes inoculated with A. lentis in greenhouse condition. For each genotype, correspondent data is present within bar. Mean and standard error values related to each genotypes presented in Table S2. In the following, genomic fingerprinting of the studied lentil panel was inspected by using 9 functional SSR primer pairs, which specifically were correspondence with defense-related genes (Table 2). The number of alleles per locus varied between 2 (Glyco81) and 16 (PP2C) (Table 2). Polymorphism information criteria (PIC) as an index for clarifying DNA marker power in evaluation of genetic diversity of plant germplasm showed that loci RRM1 (correspondence to the defense gene of RRM/RBD/RNP motif & hnRNPs) with PIC values of 0.74 possessed the highest efficacy in evaluation of present lentil germplasm. In this study, SSR loci such as PP2C (PIC = 0.68) and PRM1.2 (0.67), by having PIC values more than 0.5, also depicted promising capability for this germplasm assessment. Regardless of resistance or susceptibility reaction of genotypes, a model-based Bayesian technique in the program STRUCTURE was utilized through SSR data to reveal the genetic structure of the analyzed germplasm. After using Evanno’s approach, the most likely value of K (number of subgroup) was found to be two (Fig. 3A). Table 1 shows the membership percentage of each accession to recognized subgroups (Q values). Accordingly, 79 studied lentil genotypes were partitioned into two main subgroups, including “green” and “red” (Fig. 3B). It is detected that a small number of genotypes, including G48 (“08540124-08”) and G49 (“FLIP2003-2L”), are located in the “Mix” subgroup (Table 1; Fig. 3B) in order to have Q values lower than %70 in the group (Spataro et al., 2011) and therefore, these genotypes have a more likely admixture genome. (A) Determination of optimal number of K (subpopulation) using Delta K; (B) Population structure with K = 2 for 79 lentil genotypes using 51 defense-gene related SSRs; (C) Discriminant analysis of principal components (DAPC) scatterplot drawn using 51 SSR markers across 79 lentil genotypes in the R package adegenet. Dots represent individuals with colors denoting cluster allocation; (D) No. Bands Freq. > = 5%, He = Expected Heterozygosity = 2×p×q, I = Shanon index. Supervised machine learning and differentiative resistance allele In the following, by considering resistance or susceptible reaction of lentil genotypes, discriminant analysis of principle component (DAPC) (Fig. 3C) was further carried out on SSR loci to assess the subgroups at K = 3, including Red (MR), Blue (S), and Yellow (R) colors. Based on the probability of cluster membership assignment, DAPC represented a good fit, and the majority of studied lentil genotypes were classified regarding their greenhouse response (Fig. 3C). Accordingly, implemented SSR loci could effectively distinguish pre-defined S, R, and MR subgroups from each other, although a little bit of overlapped genotypes was detected. Investigation of allelic pattern across identified subgroups (R, S, and MR) (Fig. 3D) revealed that He values were relatively higher for susceptible and moderately resistant subgroups, while they were the lowest for the resistant subgroup (Fig. 3D). Among the different genetic subgroups, the Shannon index as genetic diversity criteria were the highest for the S and R subgroups (Fig. 3C). Results showed that both R, and S subgroups have the maximum number of alleles with a frequency greater than 5%. A major deference between the results of DAPC and former admixture clustering was the tendency of DAPC analysis to assign all genotypes to a single cluster compared to admixture, which assigned admixed genotypes to multiple clusters based on K values. With prior knowledge about disease response class (R, MR, and S) of lentil genotypes, which was also proved by defense-genes related SSRs information, combination of support vector machine (SVM), and K-nearest neighbors (KNN) with genetic algorithm (GA) were used for feature selection (Table 4). Results showed maximum accuracy (78.48) in prediction of pre-defined classes (R, MR, and S) is devoted to GA + quadratic SVM method. As resulted (Table 4), 17, 11, 8, and 7 SSR alleles (features) were detected through each GA + linear SVM, GA + quadratic SVM, GA + cubic SVM, and GA + KNN methods, respectively. Notably, the “PP2C-7” feature was jointly found across all methods, which emphasized its functional role in discrimination of the three disease response classes of lentil. Table 4Application of genetic algorithm as feature selection method accompanying with supervised machine leaning methods (GA + linear SVM, GA + quadratic SVM, GA + cubic SVM, and GA + KNN) for identifying SSR allele which could predict R, S, MR types of lentil genotypes. In each case, accuracy of method and identified allele was shown.Genetic algorithm and supervised learning combinationGA + linear SVMGA + quadratic SVMGA + cubic SVMGA + KNN (k = 3)Accuracy75.9578.4875.9573.42Selected featurePRM1.1-1PRM1.1–4PRM1.1–6Glyco81-1Glyco81-2DEAD-2PP2C-1PP2C-7PP2C-15PRM1.2–3PRM1.2–4PRM1.2–5zfRing-1zfRing-2TMK1-2TMK1-3ERF81-2PRM1.1–2PRM1.1–4PRM1.1–7Glyco81-2DEAD-1PP2C-3PP2C-7PP2C-16PRM1.2–5zfRing-7PRM1.1–4PRM1.1–6PRM1.1–7DEAD-2PP2C-7PP2C-16zfRing-6ERF81-1PRM1.1–3PRM1.1–6DEAD-3PP2C-5PP2C-7PP2C-11PRM1.2–4 Application of genetic algorithm as feature selection method accompanying with supervised machine leaning methods (GA + linear SVM, GA + quadratic SVM, GA + cubic SVM, and GA + KNN) for identifying SSR allele which could predict R, S, MR types of lentil genotypes. In each case, accuracy of method and identified allele was shown. PRM1.1-1 PRM1.1–4 PRM1.1–6 Glyco81-1 Glyco81-2 DEAD-2 PP2C-1 PP2C-7 PP2C-15 PRM1.2–3 PRM1.2–4 PRM1.2–5 zfRing-1 zfRing-2 TMK1-2 TMK1-3 ERF81-2 PRM1.1–2 PRM1.1–4 PRM1.1–7 Glyco81-2 DEAD-1 PP2C-3 PP2C-7 PP2C-16 PRM1.2–5 zfRing-7 PRM1.1–4 PRM1.1–6 PRM1.1–7 DEAD-2 PP2C-7 PP2C-16 zfRing-6 ERF81-1 PRM1.1–3 PRM1.1–6 DEAD-3 PP2C-5 PP2C-7 PP2C-11 PRM1.2–4 Gene expression response of susceptible and resistant genotypes to A. lentis This study measured the expression patterns of two common defense-related genes (RBP-hnRNP and PR-2) to gain comprehensive and comparative insights, despite alleles like PP2C-7 being identified as differentiating agents between susceptible and resistant genotypes. The sequence alignment between two studied genes of RBP-hnRNP (XM_051047539.1) with 2362 bp length and PR-2 (JX475046.1) with 1332 bp length showed relatively low similarity between these sequences (average similarity of 28%) with abvious single nucleotide polymorphisms (Fig. 4). Gene expression assay showed significant difference between non-infected plants (control) and infected plants regardless of susceptible or resistant reaction (Fig. 5). Expression analysis of the RBP-hnRNP and PR-2 in both resistant (“ILL 7947”) and susceptible (“08S 40111-01”) genotypes showed clear up-regulated expression in the resistant genotype compared with the susceptible genotype, particularly at 48 hpi and 72 hpi for PR-2. At the first time stage after contamination (24 hpi), gene RBP-hnRNP (Fig. 5A) in contrast with gene PR-2 (Fig. 5B) had the highest expression in the resistant genotype (“ILL 7947”). Although the relative appearance of RBP-hnRNP in genotype “08S 40111-01” was significant at different hpi (Fig. 5A), the manifestation level was lower than that of the resistant genotype (“ILL 7947”). About PR-2 (Fig. 5B), its transcripts were not induced significantly at 24 hpi in both susceptible and resistant genotypes, while it had high expression at 48 hpi, higher in “ILL 7947” compared with the “08S 40111-01” genotype. (A) Sequence alignment between two genes of PR-2 (sequence 1) and RBP-hnRNP (sequence 2) (B) detailed information including sequence length, GC content, and similarity level. Gene expression patterns of defense genes of RBP-hnRNP (A) and PR-2 (B) in lentil susceptible ("08S 40111-01") and resistant (“ILL 7947”) genotypes over hours post inoculation (hpi). Letters above each column is resulted by mean comparsion among genotype×time interaction. Treatments with same letters shows non- significant difference Discussion Efforts for identifying resistance sources against important disease agents are calculated as basement during lentil germplasm management and its breeding schedule. Accordingly, the existence of richness in lentil germplasm as a treasure involving helpful genes is considered numerously26,27. Hence, in the present study, new inbred lines of lentil originated from several breeding programs of ICARDA accompanied with indigenous lentil varieties were evaluated under artificial infection of A. lentis isolate to assess new resistance sources for AB. In greenhouse conditions, the AB symptoms were obvious, which indicated the effectiveness of inoculation and the aggressiveness of separated isolates as prerequisites for screening germplasm reactions26. Results showed a large spectrum of resistant, moderately resistant, and susceptible reactions, with a major tendency toward susceptible and moderately resistant types. Meanwhile, the majority of genotypes showed resistant respond in field condition. In line with this finding, Tullu et al.26, evaluated 375 lentil accessions belonging to six wild species as well as 18 cultivated genotypes in both greenhouse and field conditions and addressed the dominant presence of AB resistance among wild species. Oppositly, in the other study28, through field examination of 148 Ethiopian lentil genotypes, a low frequency of resistant type versus other reaction types has been proved. Such discrepancy between greenhouse and field states is due to incomplete developing of AB disease in naturally infected nurseries26,28 which implying on imporance of both greenhouse and field phenotyping. In the present work, the existence of resistant reaction types just among external lentil genotypes supports the strategy of incorporating such germplasm into domesticated germplasm to achieve resistance against AB29. It is resulted that most of the released lentil varieties in Iran lack the resistance genes, and the identified resistant genotypes are promising genotypes that can be used as sources of parental materials for improving them. By applying this manner, resistant varieties will be achieved, which is the most practical and cost-efficient method for controlling and managing Ascochyta blight of lentil30. As a novel insight, in this study, SSRs relevant to defense-genes that were identified through lentil responses to infection by A. lentis were selected and implemented for evaluation of the genetic diversity of the presented lentil germplasm. Interestingly, a high level of genetic variability was found among lentil genotypes based on functional markers related to defense genes. In addition, multiple SSR alleles were recognized for each locus, which implies complicated control of the resistance reaction against AB in lentil. Among the studied SSRs, SSRs based on genes RRM1, PP2C, PRM1.2 and zfRing-UBOX had remarkable discrimination power in lentil germplasm regarding higher values of polymorphism information content. As expected, the population structure of studied lentil germplasm had not so much mixed genotype that was evident from the self-pollination nature of lentil as well as the issue that lentil breeding lines resulted from limited crosse31. Hence, broadening lentil gene pool with diverse genotypes even from wild species26 could speed up AB resistance breeding programs. Overall, the undertaken lentil germplasm in the present study, like other lentil germplasm introduced by Chowdhury et al.32 and Tullu et al.8, conferred genes for resistance to AB that could be mapped through DNA markers and furthermore applied in marker-assisted selection. In this regard, numerous studies were done through RAPD, SCAR, SSR, AFLP, and SSR markers to detect genomic regions controlling resistance to AB in lentil. In particular, Tar’an et al.6 could pyramid the genes for resistance to AB and anthracnose into lentil breeding lines. But, such aforementioned mapping studies and even other genetic analyses of resistance against AB in lentil need to recognize resistant and susceptible genotypes to construct a genetic population. Therefore, it is crucial to evaluate lentil germplasm not only through phenotyping but even by molecular technique to select genotypes concisely. In this regard, DAPC analysis by assuming three phenotypic classes of R, MR, and S was implemented by means of defense-genes relevant SSRs and showed the reliability of phenotypic classification in most cases. So, it is inferred that the applied SSRs could functionally be applied for screening lentil germplasm against A. lentis independent of growth stage. The robustness of DAPC analysis in prediction of subgroups through SSR data was proven in earlier studies33,34. Here in, to determine which SSR allele (s) as a feature (s) have a significant role in the construction of pre-defined subgroups, supervised machine learning was applied. Results show that allele PP2C-7 derived from PP2C defense-gene that act in pathways of plant stress signaling is an important feature that distinguished studied lentil genotypes in the presence of A. lentis infection. Therefore, without any phenotyping and by using the aforementioned defense-genes related SSR, it is possible to predict lentil genotype reaction to A. lentis and finally classify a germplasm. Although there are some studies about the usage of artificial intelligent approaches in the evaluation of plant germplasm such as olive35, potato36, grapevine37, and saffron38, but there is not any report about the evaluation of plant-pathogen interaction, especially with DNA marker data. Albeit, to somewhat the low number of applied samples may be proposed as confidence reduction agent for machine learning results but by incorporating expert knowledge into the model or feature selection process to guide which variables are most relevant, can increase the effective use of available data. From the point of view of molecular diagnostics, each resistance or susceptibility reaction could be verified molecularly and also clarify the mechanism underlying disease response in plants. Accordingly, Khorramdelazad et al.12 have introduced several genes that have their own roles in steps including fungal elicitor recognition and early signaling, structural response, biochemical response, transcription regulators, hypersensitive reactions, cell death, and systemic acquired resistance in A. lentis-lentil interaction. Regarding the significance of biochemical response and transcription regulators12, in this project two genes of PR-2 and RBP-hnRNPs as antifungal compounds and transcriptional regulators were measured in identified resistant (“ILL 7947”) and susceptible (“08S 40111-01”) genotypes infected by A. lentis. The reason for selection of these two genes was their high rate of amplification efficiency12 and so utilized for verifying machine learning based identified resistant (“ILL 7947”) and susceptible (“08S 40111-01”) genotypes. Moreover, the sequence alignment of PR-2 and RBP-hnRNPs genes confirmed their specificity and showed that cross-reaction is very unlikely due to the low sequence similarity. It prove that PR-2 primer do not accidentally bind to or detect an unrelated gene like RBP-hnRNP and vise versa. Transcripts recording showed, as a biochemical response to A. lentis, PR-2 gene as a type of pathogenesis-related (PR) proteins significantly increased in “ILL 7947” when compared to “08S 40111-01”. As clarified39, PR-2 play an important role in defense against fungi agents through catalyzing the hydrolytic cleavage of 1,3- β-D-glucosidic linkages in β−1,3-glucans existing in the fungal cell walls, and led to cell lysis and death in fungi. Moreover, PR-2 releases elicitors from the walls of fungi, which are recognized by plant receptor molecules and stimulate numerous downstream signaling and defense responses. Opposite to Khorramdelazad et al.12 which reported overexpression of PR-2 in resistant genotypes across all time points (2 hpi, 6 hpi, and 24 hpi), the present study showed a noticeable expression pattern at 48 hpi, and 72 hpi. But, in line with Sambasivam et al.2, oxidative burst phase as a signal for downstream gene expression occurs between 20 and 24 hpi after pathogen attack, and therefore, overexpression of PR-2 gene at time points after 24 hpi is not un-expectable. Here in, noticeable up-regulation was also detected for RBP-hnRNPs in the “ILL 7947” resistant genotype, particularly in all time points. This RNA-binding transcriptional factor plays a key role in regulating the transcription of genes in response to biotic and abiotic stresses in plants40. About this issue, Khorramdelazad et al.12 also revealed the expression of the majority of defence-related genes at 2 hpi and so, overexpression of this gene in the “ILL 7947” resistant genotype versus the “08S 40111-01” susceptible genotype in all time points is logical. Conclusion The current study showed the existence of genetic variability in the A. lentis-lentil pathosystem and the necessity of considering external lentil germplasm as a valuable resource having resistance genes. Population structure analysis of lentil germplasm concerned with defense-genes related SSRs showed distinguishing power of these functional SSRs in predicting R, S, and MR subgroups, which were defined phenotypically. By using a combination of genetic algorithms and supervised machine learning, the SSR allele of PP2C-7, designed based on the defense gene of PP2C (the defense gene that acts in the pathway of plant stress signaling), was detected as a key feature that could properly separate the aforementioned subgroups. Transcripts measurement for defense genes of PR-2 and RBP-hnRNPs in identified resistant (“ILL 7947”) and susceptible (“08S 40111-01”) genotypes showed up-regulation for both genes in the resistant genotype. Gene RBP-hnRNPs as transcriptional factor was overexpressed in a resistant genotype at 24 hpi, which depicts its activation upon invasion by A. lentis pathogen. In addition, gene PR-2 as a biochemical response, was overexpressed in resistant genotypes except for 24 hpi. This study also confirms the identity of genotype “ILL 7947” as a new resistant source against A. lentis, which has the potential to be utilized as a parental line for lentil AB breeding programs or even may be used as a commercial variety. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements The authors gratefully thank the University of Maragheh, Shahrekord University, and the Dryland Research Institute of Iran for supplying greenhouses and lab facilities. Author contributions HHM, HP and BS conceived and designed research. MM and HP conducted experiments. BS and HP supervised the process, and reviewed the manuscript, HHM analysed data and wrote the manuscript. All authors read and approved the manuscript. Data availability The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. Declarations Competing interests The authors declare no competing interests. Conflict of interest The authors assert that they have no confict of interest. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.