Two populations of F₈ recombinant inbred lines (RIL) were used in this study. They were developed using a single seed descent from their F2 generations. The population (109 RILs) derived from a cross between the Australian domesticated breeding line 83A:476 (resistant to GLS) and Morocco wild-type P27255 (susceptible to GLS) was used for genetic map construction and QTL mapping for GLS confirmation. Another 93 RIL₈ population derived from the cross between two Australian cultivars Tanjil (resistant to GLS) and Unicrop (susceptible to GLS) was used to validate the QTL and the molecular markers in this study. Unicrop was one of the three susceptible varieties released prior to the introduction of GLS resistance, while Tanjil was released in 1998 as part of the lineage with the GLS resistance. A total of 18 historical and current commercial cultivars and breeding or wild ecotypes (Tanjil, Yorrel, Coromup, Mandelup, Merrit, 83A:476, Moonah, Unicrop, Kalya, Tallerack, P26167, P27255, Quillinock, 75A:258, P26603, Bo7212, P26668, and P27221) were used to analyze GLS alleles and the gene evolution. These germplasms are from Australian varieties and European wild ecotypes.

Two isolates (S. beticola: WAC12986 and S. vesicarium: WAC13136) from the Western Australian Culture Collection (Thomas et al., 2011; Vaghefi et al., 2020) were used for phenotyping in this study. Isolates were grown on 20% V8 agar medium at 22°C with a 12:12 h light/dark cycle for 14 days (Everts and Armentrout, 2001). Then spores were harvested and suspended in sterile distilled water with 0.1% Tween-20. The final concentration of spores was adjusted to 1.5 × 10⁵ ml^–1. Three replicate pots, each containing seven 14-day-old seedlings of each line were spray inoculated with the spore suspension. The pots were placed randomly. Inoculated plants were placed in an intermittent misting growth chamber with 90% of shade of natural daylight. Two days later, the plants were transferred to the greenhouse at 20°C. Disease scoring was conducted 14 days after inoculation. Disease resistance was scored on the first two fully expanded leaves on a 0–5 scale (0: no symptoms; 1: one to two lesions < 1 mm on each leaflet; 2: less than five small lesions on each leaflet; 3: 1- to 5-mm lesions, often coalescing with associated chlorosis; 4: some leaflets completely necrotic or fallen; 5: all leaflets completely necrotic or complete defoliation). Plants scoring 0 and 1 were considered resistant, while those scoring greater than 3 were considered susceptible (Thomas et al., 2011; Ahmad et al., 2016). Genotype effects on disease severity were analyzed by ANOVA with differences compared by LSD. The disease assessments of the two RIL populations were carried out in 2008, 2009, and 2013.

Identification of MFLP markers linked to GLS resistance followed the same method as described by previous studies (Yang et al., 2002; Lin et al., 2009). MFLP method combines the concept of microsatellite-anchor primer and AFLP technique. Genomic DNA was digested by restriction enzyme MseI. One AFLP MseI adaptor was ligated onto the restriction fragments. PCR was performed using one MseI primer in combination with one microsatellite-anchor primer (Yang et al., 2001). Seventeen plants were used in the MFLP analysis. Eleven of these plants, including parental line 83A:476 and 10 RIL₈, were resistant to GLS. The other six plants, including parental line P27255 and five RIL₈, were susceptible to GLS. These plants were genotyped with MFLP markers. The genotypes of the markers close to GLS should be associated with the disease resistance.

PCR bands from the MFLP markers linked to GLS were excised from MFLP gels and boiled for 1 min in TE buffer. The products were re-amplified by PCR, and electrophoresed in 1% agarose gel. DNA fragments were isolated with agarose gel and purified with a gel extraction kit (Qiagen), and then sequenced using the BigDye Terminator system (Applied Biosystems). Blastn searches were performed on the sequences in the lupin genome database (Wang et al., 2020). The chromosome and physical positions of MFLP markers were anchored.

The lupin parental lines (Unicrop, 83A:476, and P27255) were shotgun sequenced with 10 times coverage using the second-generation sequencing method, and the sequences were obtained from the China National GeneBank Database¹ and were available under the project accession CNP0001034. The RIL₈ population from the cross of 83A:476 and P27255 were genotyped using the RAD method (Yang et al., 2013b; Zhou et al., 2018). The protocol was described by Chutimanitsakun et al. (2011) with minor modifications that EcoRI was used. Sequencing libraries (100 bp) were constructed by using a unique multiplex identifier (MID) barcode (Baird et al., 2008). The RAD product sequencing was carried out on Solexa HiSeq2000. Sequencing raw data were divided into different lupin lines according to the MID barcodes (Baird et al., 2008), and then the MID barcodes were removed from the raw sequencing data. The 92-bp RAD reads with the same DNA sequences were clustered into one read tag. Tags containing over 100 reads were considered to be from repetitive regions, so they were removed to avoid the detection of SNP/indel variants (Catchen et al., 2011). Genome assembly of the two parental lines was conducted before alignment.

MapQTL5.0 was used to detect QTL for gray leaf spot disease. The genetic map of the population between 83A:476 and P27255 was obtained from our previous studies (Zhou et al., 2018). Formatted files including the genetic map, genotypic data, and phenotypes were imported into MapQTL5.0. The threshold of LOD value for QTL mapping was 3.0. First, interval mapping was performed, and then the marker with the highest LOD value was selected for MQM mapping analysis.

High-density markers across the QTL region were genotyped in the RIL population. The recombinant lines were selected to narrow down the QTL for gray leaf spot resistance. The left border and right border of the gene were determined by comparing the genotypes and phenotypes.

The software CLC Genomic Workbench 21.0.3 was used to map the sequence data to the reference narrow-leafed lupin “Tanjil” genome. The 17 accessions excluding the reference cv. Tanjil were shotgun sequenced with 10 times coverage. All the sequencing read data were obtained from the China National GeneBank Database (see text footnote 1) and were available under the project accession CNP0001034. The parameters were as follows: match score 1, mismatch cost 2, linear gap cost, length fraction 0.5, similarity fraction 0.8, and non-specific match handling with map randomly. Tracks were created, and all the genome sequences used in this study were combined. The SNP and indel variants were detected and generated. For variant detection, fixed ploidy variant parameters were as follows: ploidy 2 and required variant probability 90%, ignore broken pairs, minimum coverage 10, minimum count 2, and minimum frequency 20%.

Gray leaf spot (GLS) alleles and gene evolution were investigated among 18 narrow-leafed lupin accessions. SNPs were identified by mapping their reads to the Tanjil reference genome. SNP variants were obtained from the variants detection. The SNP genotype data were imported into Geneious 8.0 for phylogenetic tree construction. Geneious Tree Builder function was used, and the parameters were as follows: genetic distance model of Jukes–Cantor with the build method of UPGMA.

For the gray leaf spot disease phenotypes, all the plants of parental line 83A:476 were resistant to GLS (scoring 0), while all the plants of P27255 were susceptible (scoring 4 or 5, mean 4.6). The parental phenotypes were significantly different (p < 0.05). Among the 109 tested F₈ RILs, 51 RILs were susceptible to GLS, and the other 58 RILs were resistant (Figure 2). The segregation of susceptible:resistant in the F₈ population fit the expected 1:1 ratio (χ² = 0.33, χ²₀.₀₅,₁ = 3.84, (χ² < χ²₀.₀₅,₁), indicating that a single gene controls GLS resistance in 83A:476.

In the beginning, we wanted to develop molecular markers linked to GLS and use immediately in breeding programs, so an efficient method was adopted to identify the locus. A total of 1,657 polymorphic MFLP markers were identified between parental lines 83A:476 and P27255. These markers were genotyped in 10 resistant RILs and 5 susceptible RILs as well. Among these markers, three markers (D145, D280, and D300) co-segregated with GLS resistance in these RILs (Table 1). All the resistant RIL₈ exhibited the same genotype to 83A:476, and the susceptible RILS showed the same genotypes to P27255.

The PCR products from PAGE gel were amplified and sequenced. The amplicon sequences of the three markers are listed in Supplementary Table 1. The markers did not perform very well (data not shown) among the lupin germplasms maybe because these markers were located outside of the GLS gene, so we continued to identify the gene and develop diagnostic markers. Recently, the release of the lupin genome sequence enables us to anchor the MFLP markers to their physical positions and to mine the genes based on the lupin genome (Wang et al., 2020). The sequences excluding forward and reverse primers were used to conduct a blast in the lupin cv. Tanjil genome. All the three markers (D145, D280, and D300) were anchored to the lupin linkage group LG-19 with a physical position of 2.156–2.218 Mb region. D280 and D300 were anchored to the 2.156 Mb region, while D145 was mapped to the 2.218 Mb region.

QTL mapping and fine mapping were used to validate the candidate gene. The genetic map from 83A:476 and P27255 population was constructed in our previous studies (Zhou et al., 2018). The genetic map contains 27,892 markers (Supplementary Data Sheet 1). A simplified map containing one marker for each locus was used for QTL mapping. QTL analysis was conducted in the RIL population. A single major QTL for gray leaf spot resistance was identified on chromosome 19. The closest marker DAFWAsnp_16977 (29.33 cM) can explain 97.9% phenotypic variations and with the LOD value of 74.83 (Figure 3 and Supplementary Figure 1).

The recombinant lines around the QTL region (23.58 to 32.78 cM) were investigated in the RIL₈ population. A total of 16 recombinant lines were obtained based on the genotypes (Figure 4). Figure 4 showed that the genotypes of QTL marker DAFWAsnp_16977 were consistent with their gray leaf spot resistance in this population. From these recombinant lines, the gene was narrowed down by the lines W/D039 and W/D056. The recombination point of W/D039 was between the markers DAFWAsnp_20112 (R genotype) and DAFWAsnp_16977 (S genotype), while the GLS resistance of this line was susceptible, so the gene must be on the right side of the marker DAFWAsnp_20112. For the line W/D056, the recombination point was between the markers DAFWAsnp_19170 (S genotype) and DAFWAsnp_20338 (R genotype), and the GLS resistance was susceptible, so the gene should be on the left side of the marker DAFWAsnp_20338. In a word, the gene was narrowed down between the two markers DAFWAsnp_20112 and DAFWAsnp_20338. These markers were anchored to narrow-leafed lupin cv. Tanjil genome, and the physical positions of the markers were 2,157,779 and 2,399,114 bp, respectively. The QTL was finally fine-mapped to a 241-kb region. The three MFLP markers (D145, D280, and D300) overlapped with this fine mapping region.

A total of 21 genes were annotated in the QTL mapping region (Table 2). Two genes, LOC109334326 and LOC109334327, were associated with disease resistance. LOC109334326 is a TMV resistance protein N-like, and LOC109334327 is a putative disease resistance protein RGA3 (Wang et al., 2020).

Genomic sequence analysis revealed that there were six SNPs in the LOC109334326 region between 83A:476 and P27255 (Figure 5). Two SNPs were in the 5′-UTR regions, while the other four SNPs were in the coding region. Only one SNP (T/C) at 2,190,169 in the coding region induces amino acid substitution from leucine in 83A:476 (R) to proline in P27255 (S) (Figure 5).

To further investigate these two genes, we used two additional Australian cultivars Tanjil and Unicrop. Tanjil is resistant to gray leaf spot disease, while Unicrop is susceptible. Unicrop was re-sequenced and mapped to the Tanjil reference genome. The segregation of susceptible:resistant in the F₈ population fit the expected 1:1 ratio (χ² = 3.76, χ²₀.₀₅,₁ = 3.84), (χ² < χ²₀.₀₅,₁). It indicated that a single gene controlled GLS resistance in the Tanjil and Unicrop population. Interestingly, the result indicated that there was only one SNP variation from T (Tanjil) to C (Unicrop) at 2,190,196 bp in the 241-kb fine-mapping region on chromosome 19, and this SNP variation from the gene LOC109334326 results in the same amino acid substitution from leucine in Tanjil (R) to proline in Unicrop (S). It indicates that this is the candidate gene for gray leaf spot resistance.

For another disease resistance RGA3 gene LOC109334327, 17 SNPs were detected between the parental lines 83A:476 and P27255 (Table 3), but no variations exist between Tanjil and Unicrop. Furthermore, no SNPs were identified between the tolerant line 83A:476 and the susceptible cultivar Unicrop. It indicates that this gene is not the candidate gene.

To confirm the Tanjil carrying the same GLS QTL/locus, which was mapped from 83A:476 and P27255 population, a dominant SNP resistant marker was developed for the GLS candidate gene (forward: TCTGATGGCGTACATGTGTAA, reverse: TACTCTGCCCTCACTGACCT). This marker was tested in 93 RIL₈ derived from a cross between Tanjil and Unicrop. The genotypes co-segregated with their GLS resistance (Supplementary Table 2).

The gene alleles were investigated among Australian and European lines. SNP variants of the gene (LOC109334326) were analyzed among 18 narrow-leafed lupin accessions consisting of Australian varieties, Australian breeding lines, and European wild lupin (Tanjil, Yorrel, Coromup, Mandelup, Merrit, 83A:476, Moonah, Unicrop, Kalya, Tallerack, P26167, P27255, Quillinock, 75A:258, P26603, Bo7212, P26668, and P27221). The genomic sequences of these accessions were obtained from the China National GeneBank Database (see text footnote 1) and were available under the project accession CNP0001034. The sequence reads were mapped to the reference genome Tanjil, and SNP variants were obtained (see Supplementary Data). The phylogenetic tree (Figure 6) showed that they were classified into two major groups. All the European wild lupin accessions, together with the Australian variety/breeding lines Quillinock and 75A:258, were in one group, while all the other Australian varieties/breeding lines were clustered into another group. Furthermore, all the resistant (Tanjil, Yorrel, Coromup, Mandelup, Merrit, and 83A:476) and susceptible varieties (Moonah, Unicrop, Kalya, and Tallerack) from the major group were divided into different sub-groups. For the resistant gene, two alleles were identified. Australian resistant cultivars Tanjil, Yorrel, Coromup, Mandelup, and Merrit carried the same allele, while the breeding line 83A:476 carried a different GLS allele. It indicated that new GLS-resistant allele from lupin germplasm was incorporated into the breeding line 83A:476 during the breeding program.