Dissecting the genetic basis of resistance to Soil-borne cereal mosaic virus (SBCMV) in durum wheat by bi-parental mapping and GWAS

Soil-borne cereal mosaic virus (SBCMV), the causative agent of wheat mosaic, is a Furovirus challenging wheat production all over Europe. Differently from bread wheat, durum wheat shows greater susceptibility and stronger yield penalties, so identification and genetic characterization of resistance sources are major targets for durum genetics and breeding. The Sbm1 locus providing high level of resistance to SBCMV was mapped in bread wheat to the 5DL chromosome arm (Bass in Genome 49:1140–1148, 2006). This excluded the direct use of Sbm1 for durum wheat improvement. Only one major QTL has been mapped in durum wheat, namely QSbm.ubo-2B, on the 2BS chromosome region coincident with Sbm2, already known in bread wheat as reported (Bayles in HGCA Project Report, 2007). Therefore, QSbm.ubo-2B = Sbm2 is considered a pillar for growing durum in SBCMV-affected areas. Herein, we report the fine mapping of Sbm2 based on bi-parental mapping and GWAS, using the Infinium 90 K SNP array and high-throughput KASP®. Fine mapping pointed out a critical haploblock of 3.2 Mb defined by concatenated SNPs successfully converted to high-throughput KASP® markers coded as KUBO. The combination of KUBO-27, wPt-2106-ASO/HRM, KUBO-29, and KUBO-1 allows unequivocal tracing of the Sbm2-resistant haplotype. The interval harbors 52 high- and 41 low-confidence genes, encoding 17 cytochrome p450, three receptor kinases, two defensins, and three NBS-LRR genes. These results pave the way for Sbm2 positional cloning. Importantly, the development of Sbm2 haplotype tagging KASP® provides a valuable case study for improving efficacy of the European variety testing system and, ultimately, the decision-making process related to varietal characterization and choice. Supplementary Information The online version contains supplementary material available at 10.1007/s00122-024-04709-7.


Introduction
Soil-borne cereal mosaic virus (SBCMV) is a furovirus affecting the production of both tetraploid durum (Triticum turgidum var.durum Desf.) and hexaploid bread wheat (Triticum aestivum L.) across Europe, especially in France, Italy, Germany, Poland, and Denmark (Budge et al. 2008;Kanyuka et al. 2004;Maccaferri et al. 2011a;Rubies-Autonell et al. 2003).SBCMV was initially considered the European strain of soil-borne wheat mosaic virus (SBWMV), a member of the genus Furovirus which affects mainly wheat crops in U.S., and later recognized by the International Committee on Taxonomy of Viruses as a separate species (Kanyuka et al. 2004).
Fields heavily infected with SBCMV have been shown to suffer a yield reduction of ca.40-50% in susceptible winter wheat cultivars in the UK and up to 70% in highly susceptible durum wheat in Italy (Clover et al. 1999;Kühne 2009;Vallega and Rubies Autonell 1985;Vallega et al. 2003).
The name "soil-borne" is due to the infection mechanism which takes advantage of the plasmodiophorid vector Polymyxa graminis, a soil inhabitant fungal-like endoparasitic protist slime mold, endoparasite of wheat.As a vector, P. graminis produces resting spores that are easily spread mainly by water, wind, and machinery.SBCMV nests inside the resting spores of the vector during periods of adverse environmental conditions and long rotations of non-host crops, persisting and accumulating in soil for decades.When conditions reverse to favorable for disease development (e.g., presence of susceptible hosts and high soil moisture, particularly in the fall, when extended periods of high soil moisture favor the movement of Polymyxa graminis), the virus is released from the resting spores and, once vectored through zoospores, enters root hair cells (Kühne 2009).From there, the virus can reach the aerial part of the plant and infect the leaf tissue, causing symptoms ranging from mild mottling to severe leaf mosaic.Moreover, vigor and plant height are also affected, with susceptible cultivars showing stunting (Kim et al. 2017;Kühne 2009;Perovic et al. 2009;Roberts 2014).
Due to the nature and persistence of the natural inoculum, agronomic practices like crop rotation or delayed sowing are mostly ineffective once the soil is infected.Chemical defense is neither effective nor environmentally friendly.As a consequence, the only possible option is the development and deployment of resistant cultivars (Kanyuka et al. 2003;Ordon et al. 2009).
The first and most effective SBCMV resistance locus identified in the bread wheat germplasm was Sbm1, located on chromosome 5DL (Bass et al. 2006).Resistance was detected in bread wheat cv.Cadenza by assessing a doubled haploid population obtained from the cross Avalon (susceptible cultivar) × Cadenza (resistant cultivar).Resistance, mainly of "translocation type" preventing the infection to spread from the roots to the stem and leaves (Kanyuka et al. 2004;Ordon et al. 2009), was found to be controlled by a major locus named Sbm1 mapped to the long arm of chromosome 5D tagged by microsatellite Xbarc110 and Xwmc765 in a region of approximately 17 cM (Bass et al. 2006).Due to the relevance of Sbm1 for hexaploid wheat breeding, it has been recently fine mapped (Liu et al. 2020) and confined in the 546,651,779-547,273,461-bp interval of Chinese Spring RefSeq v1.0 pseudomolecule (International Wheat Genome Sequencing Consortium (IWGSC), 2018), harboring 17 candidate genes, with pto-interacting protein 1 (PTI1, TraesCS5D01G531200) identified as the most promising one.After Bass et al. (2006), another major QTL named Sbm2 was mapped in the same Avalon × Cadenza population by Bayles et al. (2007) in chromosome arm 2BS.Sbm2 acts in a distinct and complementary way to Sbm1, suggesting the possibility to pyramid these two sources of resistance in bread wheat.Taken together, these two major Sbm loci enable breeding for bread wheat varieties with valuable SBCMV resistance.
Conversely, durum wheat lacks the D genome; thus, Sbm1 resistance locus is not directly exploitable for durum wheat breeding.The potentially higher impact of SBCMV on grain yield loss and the lack of known sources of resistance in durum wheat as compared to hexaploid wheat prompted the search for valuable sources of resistance in the durum wheat germplasm.Using a RIL population from Meridiano (resistant) × Claudio (moderately susceptible), Maccaferri et al. (2011a) identified QSbm.ubo-2B, a major QTL for SBCMV resistance in durum wheat located in the distal telomeric region of chromosome 2BS, close to the wPt-2106 DArT® marker and three SSR markers (Xwmc661-Xgwm210-Xbarc35). The large proportion of variance explained (PVE) by this locus (55.8-68.0%for ELISA and up to 91.6% for symptom severity) suggested QSbm.ubo-2B as a potentially major target for durum wheat breeding, marker-assisted selection, and QTL cloning.QSbm.ubo-2B was confirmed in a durum wheat panel (Maccaferri et al. 2011a) and in a second mapping population, Simeto (susceptible) × Levante (resistant) with a PVE of 60-70% for symptom severity.Meta-QTL analysis mapped QSbm.ubo-2BS on a two cM-wide CI centered to wPt-2106, proximal to Xwmc661-Xgwm210-Xbarc35, which allowed for the first molecular assay to trace the QTL resistant allele.Based on the tetraploid wheat consensus map (Maccaferri et al. 2015) and meta-QTL analysis, QSbm.ubo-2BS (Maccaferri et al. 2012) was mapped in a region coincident with the Sbm2 locus in bread wheat (QSbm.ubo-2BS= Sbm2).This makes QSbm.ubo-2BS = Sbm2 a valuable target for basic and applied research aimed at its fine mapping and positional cloning.
Currently, the use of single-nucleotide polymorphisms (SNPs, Yuan et al. 2014) has become more efficient, cheaper, and robust than other types of markers and related sequencing techniques in both bread and durum wheat, thanks to the development of species-specific SNP arrays like the Illumina Infinium iSelect 90 K wheat SNP array (Cavanagh et al. 2013;Wang et al. 2014).Maccaferri et al. (2015) developed a consensus map for tetraploid wheat harboring 30,144 markers, with high-density gene-derived SNPs as useful anchor points to isolate target loci.SNPs abundance in the wheat genome, together with the development of high-throughput fluorescence-based marker assays, like KASP® (Kompetitive Allele Specific PCR, Semang et al. 2014), makes them ideal for fine mapping and MAS breeding purposes.
The publication of the durum wheat reference cultivar Svevo (Maccaferri et al. 2019) allowed us to easily switch from the genetic to the physical map, accelerating both fine mapping and candidate gene identification.Genomewide association studies (GWAS) coupled with dense SNPs arrays enable efficient exploitation of large phenotypic datasets to search for and fine map QTLs (Korte and Farlow 2013).
The present work aimed at fine mapping QSbm.ubo-2BS using a combination of mapping populations and GWAS.The development of KASP® markers was pursued with the double objective of providing genic markers anchored to the durum wheat Svevo genome assembly suitable for fine mapping of Sbm2 and, at the same time, assays ready to implement MAS.These results were integrated with candidate genes, identified within the QTL region projected upon the durum wheat cv.Svevo, hexaploid wheat cv.Chinese Spring, and the wild emmer wheat Zavitan (Triticum turgidum ssp.dicoccoides, Körn.Ex Asch.& Graebn.)reference genomes.Moreover, the results of an extensive survey of the QSbm.ubo-2BS = Sbm2-resistant/susceptible haplotype distribution in durum wheat panels grown in Europe and worldwide are reported.

Mapping populations and cultivated germplasm panels
A recombinant inbred line (RIL) population developed by the University of Bari (Gadaleta et al. 2009) composed of 3,075 lines obtained from cv.Svevo (hereafter referred to as Sv, resistant) and Ciccio (hereafter referred to as Cc, susceptible) was used for Sbm2 fine mapping.Svevo harbors the Xwmc661-Xgwm210-Xbarc35-wPt-2106 resistance tagging haplotype at Sbm2 while Ciccio carries the susceptible haplotype, similar to Simeto (Maccaferri et al. 2011a).Cultivar Svevo was sequenced, assembled, and annotated at gold-standard level (Maccaferri et al. 2019).Ciccio is adapted to Southern Mediterranean conditions and is a derivative of Valnova founder, highly susceptible to SBCMV.A second population includes 181 RILs developed by Produttori Sementi Bologna SpA from a cross between cv.Meridiano (hereafter referred to as Mr, resistant) and Claudio (hereafter referred to as Cl, susceptible).The Mr × Cl RIL population was described in Maccaferri et al. (2011aMaccaferri et al. ( , 2012)).
The cultivated durum wheat panels herein considered provide a representation of the cultivated germplasm worldwide spanning a breeding period from 1914/1920 to 2020.A total of 549 varieties/breeding lines included in this study.A first panel, referred as the "UNIBO Durum Panel" (Maccaferri et al. 2011b), was used to perform GWAS and assess linkage disequilibrium extent in the Sbm2 region and the repeatability and information content of diagnostic Illumina SNP markers converted to fluorescent-based high-throughput KASP® markers.The UNIBO Durum Panel included 258 breeding lines and cultivars released from 1970 to early 2000 from Italy, Spain, CIMMYT, ICARDA, North America, Southwestern USA, France, Austria, and Australia (Maccaferri et al. 2011b).This panel was field phenotyped for SBCMV response over three years.A wider panel including 291 more recent and diverse European varieties was assembled and analyzed in the framework of the Horizon 2020 project "Inno-Var, next-generation variety testing for improved cropping on European farmland" and other sources.The complete panel was used to investigate the distribution of SBCMVresistant and SBCMV-susceptible haplotypes worldwide and across breeding decades.

Enrichment of the Sbm2 region with KASP® markers
Based on the consensus map of Maccaferri et al. (2015) and Svevo genome sequence, KASP® assays were developed from 16 Illumina SNPs previously mapped in the Sbm2 region.Features of these SNPs are reported in Table 1.Primers were designed using PolyMarker (http:// Polym arker.tgac.ac.uk) followed by manual curation.Probe sequences tagged with fluorescent dyes FAM and HEX were added to the 5′-end of A and B primers, respectively.KASP® marker development protocol is described in Method S1.
Table 1 Summary information with regard to the Illumina Infinium iSelect 90 K wheat SNPs chosen to develop polymorphic KASP® markers in the Sbm2 confidence interval.Illumina 90 K wheat SNPs ID code, SNP marker name, newly developed KASP® marker name, genetic position in the durum wheat consensus map, physical location in durum wheat cv.Svevo genome, and polymorphism in durum / tetraploid recombinant inbred lines.Markers are listed from the distal -5'-to the proximal -3'-2BS chromosomal region cM, (= centiMorgans) are reported as cumulative genetic distances on chromosome 2B according to the consensus map from Maccaferri et al. (2015) a Genetic map distances estimated based on the tetraploid wheat consensus map published in Maccaferri et al. (2015) Genome-wide association study on the UNIBO Durum Panel The UNIBO Durum Panel was phenotyped in a highly infected field nursery located in Cadriano, Bologna, Italy (44°35′N 11°27′E).The propagule soil content in the field nursery was progressively increased, maintained high, and evenly spread in the nursery soil by repeatedly adopting a wheat-to-wheat rotation with the highly susceptible durum cv.Grazia.This ensured spreading of virulent P. graminis spores as clusters of resting spores in root tissue debris or infective propagules directly released in soil and evenly distributed throughout the field.
Phenotypic data were obtained from unreplicated, augmented design field trials over three years (2005,2007,2010).Cvs.Svevo, Ciccio, Meridiano, and Claudio were used as common checks across blocks (five blocks per experimental field).Broad-sense heritability was calculated with the R package heritability (Kruijer et al. 2015).
Single-plot data were spatially adjusted according to a moving average based on the double adjacent plots.One representative score per year was chosen for further analysis based on correlation among dates, average infection level, and cultivar discrimination power.Dates considered were April 22, 2005, March 6, 2007, and April 7, 2010, for SEV, March 14, 2007, and April 7, 2010, for ELISA.Phenotyping was carried out only for years with nurseries showing an infection level of SEV ≥ 3.00 in susceptible parents and checks throughout the field nursery.Overall infection levels were also concomitantly monitored by ELISA (ELISA ≥ 0.70).Consequently, 2006Consequently, , 2008Consequently, , and 2009 nurseries were not considered for further analysis.
Adjusted visual scores and ELISA values were used to calculate best linear unbiased estimates (BLUEs), based on a mixed linear model including genotypes as fixed and all other factors as random (Bates et al. 2014).
GWAS was performed in Tassel 5 (Bradbury et al. 2007) based on genotypic data from Illumina 90 K SNPs, newly developed KASP® markers, wPt-2106 and visual score and ELISA adjusted BLUES.Genotypic data were converted to hapmap format, keeping molecular markers with MAF ≥ 5%, missing data ≤ 50% and by removing genotypes with missing data ≥ 50%.Missing SNPs were subsequently imputed with Beagle v5.4 (Browning and Browning 2022).
A kinship matrix was calculated with all non-redundant SNP markers identified in Haploview 4.2 (Gabriel et al. 2002) with the tagger function set to r 2 = 1.0.Kinship based on identity-by-state (IBS) similarities among accessions was calculated in TASSEL.Each marker was evaluated for association with phenotypic data using a mixed linear model (MLM;Bradbury et al. 2007) which included the kinship matrix (MLM + K).A second round of analysis was carried out using the marker most associated with Sbm2 region as covariate to highlight additional QTL peaks other than the major one on chromosome 2B.IWB6584 was chosen as the Sbm2 tagging marker to be used as covariate.
An additional set of 30 RILs from the Meridiano × Claudio RIL population (Mr × Cl, Maccaferri et al. 2011a) with recombinations between KUBO-3 and KUBO-9 were selected for detailed KUBO genotyping.For each population, genotypic data were used to construct a genetic map was assembled using JoinMap v. 4 (Van Ooijen 2006).Marker order and mapping distances between markers were calculated using the maximum likelihood algorithm and Haldane's mapping function.The two maps were compared to check for the order of KASP® markers in the region flanked by KUBO-13 and KUBO-9.Once marker order was determined, genotypic and phenotypic data were integrated to define the restricted interval where QSbm.ubo-2BS was most probably located.Mapping and association analysis were carried out in Windows QTL Cartographer v. 2.5 by single-marker analysis (Wang et al. 2012).Interval mapping (IM) analysis was also carried out to calculate the LOD score in the interval between KUBO-13 and KUBO-9.The threshold value was set at LOD 3.0 (rounded) via permutations.Walk speed was set at 0.5 cM.
Marker physical positions were determined by BLAST analysis while co-linearity among intervals was compared using MUMmer4 (Marçais et al. 2018) and manual curation as described in Method S3.

Projection of the Sbm2 support interval into the Svevo physical map for candidate gene identification
The KUBO-27 to KUBO-40 interval (Figs. 1, 2, S5, S6, and S7) was considered to retrieve annotated genes from the Svevo genome reference assembly (Maccaferri et al. 2019).Gene functions were recovered using the functional annotation program AHRD, considering the longer transcript as the representative one.GO terms (GO Slim) in R Bioconductor (version 3.4) package GOstats (Falcon and Gentleman 2007; R Core Team 2017) were used to categorize genes according to their molecular functions and biological processes.Gene intervals were explored with gProfiler software (Raudvere et al. 2019).Functional profiling was computed extracting GO molecular functions and biological processes from all the genes of the interval.

Sbm2 haplotypes present in the UNIBO Durum Panel
Based on GWAS results obtained from the Durum Panel UNIBO, the region most associated with Sbm2 was identified as coincident with the haploblock containing markers KUBO-27, KUBO-29, KUBO-1, KUBO-3, KUBO-41, and KUBO-40 (listed from distal to proximal region of chromosome 2BS, Figs. 1, 2, S5, S6, and S7).UNIBO Durum Panel SNP genotype calls in the Sbm2 region in the KUBO-13 to KUBO-9 interval.This interval was used to inspect the local Sbm2 specific SNP haplotype distribution patterns in cultivars/breeding lines based graphically on a two-color mode (Fig. 3).In the Sbm2 region, the SNP alleles belonging to the resistant parents/founders Levante, Meridiano, and Neodur (showing identity by descent, IBD, in the region) were dark blue colored while the alternative alleles were pale red colored.Additionally, the local Sbm2 haplotypes depicted  2002) method in Fig. 3 were clustered based on local genetic similarities (simple matching genetic similarity algorithm, GS, also corresponding to identity by state, IBS) and Ward's clustering algorithm.Simple matching GS were computed in Tassel 5 and Ward clustering in pheatmap R package as in Kolde and Kolde (2015).Linkage disequilibrium among local SNPs and linkage blocks were determined using Haploview (Barrett et al. 2005), defining haploblocks by the algorithm developed by Gabriel et al. (2002).Box plots of SEV and ELISA phenotypes were produced for each haplotype detected at this block.

KASP® marker design and validation
Fine mapping of loci of breeding interest and downstream diagnostic (haplotype-based) marker-assisted selection require the development of locally dense molecular marker maps to saturate the confidence interval regions.Therefore, in the case of Sbm2, sequences of 16 Illumina Infinium 90 K wheat SNPs mapped in the Sbm2 region by multiple tetraploid mapping populations (Maccaferri et al. 2015) were deployed to develop high-throughput, fluorescence-based KASP® assays, subsequently used to genotype both RILs and breeding germplasm panels for fine mapping.
Nineteen out of the 35 tested assays (KUBO-1, KUBO-3, KUBO-5, , reported based on map order, from the distal 5′-end to the proximal 3′-end region of the QTL on chromosome arm 2BS) proved to be functional, highly repeatable, and codominant.These are ideal for genotyping, resulting in a 54.2% SNP-to-KASP® conversion rate (Illumina SNP and KASP® sequences in Table 1 and Table S1a, respectively).
KUBO-13 and KUBO-9, positioned at the two borders of Sbm2 region 2.5 cM upstream and 8.1 cM downstream of wPt-2106, respectively, were selected to identify recombinants in the Sv × Cc RILs (Table 1 and Fig. S5).
The DArT® marker wPt-2106, close to the Sbm2 peak LOD region, was sequenced and converted to two PCRbased assays, one based on allele-specific (ASO) primers and detectable through horizontal gel electrophoresis assay, and the other based on the high-throughput high-resolution melting technique (Table S1b).Detailed information is reported in Method S1 while Fig. S1 shows PCR assay results for wPt-2106 DArT marker converted to wPt-2106 Allele-Specific Oligonucleotide (ASO), wPt-2106 High-Resolution Melting (HRM, Wittwer et al. 2003).
The newly developed KASP® markers were used to genotype the UNIBO Durum Panel and on RIL populations and validated by comparison with the former Illumina 90 K SNP genotypic calls.KASP® diagnostic marker accuracy results are reported in Table S2 while their amplification plots are shown in Fig. S2.

GWAS based on the UNIBO Durum Panel
The UNIBO Durum Panel was used as an independent resource to fine map Sbm2and to validate the newly developed KASP® markers (genotyping information reported in Method S2).The UNIBO Durum Panel was field evaluated across three years in a dedicated nursery in Cadriano, Bologna, Italy for both visual symptom severity (SEV) and virus content in the leaves (ELISA), as reported in Table S3.SEV and ELISA traits were highly related (Figure S3), with an r value of 0.69.
The UNIBO Durum Panel showed a population genetic structure including eight well-defined groups of varieties/ breeding lines corresponding to the main breeding programs/pedigree worldwide, originated from well-known "founders" (Fig. 4a, Data_S1).
SBCMV response (both SEV and ELISA) appeared to be highly differentiated according to population structure, as shown in the box-plot phenotypic distributions based on population structure reported in Fig. 4b.
Germplasm of worldwide success as CIMMYT'60 (Jori C69 founder), CIMMYT'70 (Yavaros C79 founder), CIM-MYT'80 (Altar C84 founder), the Italian (Valnova founder), and the North American and French germplasm (cvs.Kyle, Langdon, and Neodur, Vic, founders) are represented in the panel.CIMMYT'60, CIMMYT'70, and the North American/French groups showed a higher frequency of resistance as compared to other groups like the ICARDA temperate, while Italian varieties originated from Valnova and the CIM-MYT'80 varieties originated from Altar C84 were all highly susceptible (Fig. 3).Additionally, resistance was observed in high frequency in old varieties derived from West Asian landraces (Haurani/Omrabi group, Fig. 3).membership values together with phenotypic and genotypic data are reported in supplemental Data_S1 and Data_S2.
GWAS showed a major peak locus on chromosome 2B, corresponding to QSbm.ubo-2BS = Sbm2 (Fig. 5).The LOD of this peak was 21.6 for visual score (SEV) and 18.2 for ELISA; R 2 (= PVE) was 36.9% for visual score (SEV) and 34.1% for ELISA, respectively (Fig. 1).The three peak markers (IWB72375, IWB67715, and IWB6584) from the two SEV and ELISA phenotypes were coincident.Detailed marker-trait association results for the Sbm2 region are reported in supplementary material section (Data_S2).Notably, the association results for SEV and ELISA were highly consistent.
The region corresponding to the peak of significance on Svevo genome extended from IWB11421 (KUBO-27) at 13,011,332 Mb to IWB6584 at 16,464,631 Mb, very close to the previously mapped wPt-1601 at 16,497,383 Mb (Fig. 1 and Table 2).In this interval of ca.3.5 Mb, 14 out of 35 markers showed very high LOD values, clearly identifying the Sbm2 hotspot region.The entire region is tagged by the series KUBO-27, wPt-2106, KUBO-29, KUBO-3, KUBO-41, and KUBO-40 (reported from distal -5′-to proximal -3′-2BS chromosomal region, details of association results reported in Table 3), thus identifying the corresponding haplotypes highlighted in Fig. 3.Among KASP® markers (Table 3), the most closely associated is KUBO-27.Overall, − log 10 (P) ranged from 19.36 for KUBO-27 from    2 and 3), while the lowest was obtained for KUBO-13 (5.7%, ELISA; Tables 2 and 3) Sbm2 allelic distribution appeared to be highly related to population structure, as all the founders of groups associated with resistance including ICARDA dryland, CIMMYT'60, CIMMYT'70, and North American / French varieties carried the Sbm2resistant allele, while founders of ICARDA temperate, Italian, and CIMMYT'80 groups carried the Sbm2-susceptible allele (Data_S2).
Several minor putative QTLs across the genome resulted from the GWAS (Table S4).Up to 20 putative minor QTLs at − log 10 (P) > 3 in GWAS were mapped, eight of which were common to both SEV and ELISA (Tables S4a and S4b, for SEV and ELISA, respectively).
After completing the GWAS, the QSbm.ubo-2BS= Sbm2, the additional minor QTLs, and population structure Q values were modeled in GLM (Table S5).Major and minor QTL and population structure effects on SEV and ELISA were tested based on a multi-marker general linear model (MM-GLM + Q).The p-values and adjusted R 2 of the full Sbm2 and minor QTL models are reported in Table S5.The population structure Q memberships accounted for 21.5 and 13.3% of SEV and ELISA blues values, respectively, a finding mainly explained by the allele distribution at Sbm2, whose resistant and susceptible alleles were differently distributed according to main population structure partitioning.Therefore, we modeled the effect of Sbm2 together with the additional minor QTLs suggested by GWAS in a GLM excluding population structure as covariate.Notably, Sbm2 accounted for 54.6 and 46.9% of the total phenotypic variation for SEV and ELISA, respectively.Minor QTLs were mostly confirmed by GLM for either SEV or ELISA, or both, except for IWB48240_2B and IWB26062_4A, which were not validated by the GLM.Sbm2 × other QTL interactions were mostly non-significant.Importantly, the addition of minor QTL effects and their interactions raised the total variation explained by the global genetic effects up to 67.6 and 54.7% for SEV and ELISA, respectively (Table S5).
The distribution of resistant and susceptible alleles at Sbm2 and minor QTLs in the UNIBO durum panel is detailed in Data_S3.Resistant alleles at minor QTLs were distributed across multiple varieties, from a minimum of four to a maximum of 18 alleles per variety.The linear model plot reports the cumulative number of resistant alleles across different genotypes in the X-axis, based on the tag-QTL marker scores, whereas the Y-axis reports the BLUES of SEV and ELISA data.Based on this data, linear regression was computed to estimate the overall effect of minor QTL combinations on the phenotypic data (Data_S3).
The global effect of the susceptible-to-resistant ideal progressive substitution of alleles in germplasm is exemplified by the histograms reported in Data_S3, reported separately for varieties carrying either the resistant or susceptible allele at Sbm2.
In the 148 Sbm2 resistant varieties, the cumulated substitution effect of minor QTLs was estimated (based on the regression reported in Data_S3) to improve SEV only marginally from 1.63 (four minor resistant alleles) to 0.92.However, the effect was more marked on the ELISA value that decreased from 0.92 (4 minor alleles) to 0.40 (18 alleles).
Minor QTL allelic substitution effect impacted more evidently on the group of 110 Sbm2 susceptible varieties, as expected.In this group, the SEV response of varieties decreased from 3.81 (varieties cumulating four minor resistant alleles only) to 1.82 (varieties cumulating 16 cumulated minor resistant alleles) and, for ELISA, the average value decreased from 1.45 (four resistant alleles) to 0.94 (16 resistant alleles), thus providing a substantial improvement in resistance.

Fine mapping of Sbm2 based on the Svevo × Ciccio and Meridiano × Claudio RIL populations
Genetic maps constructed for each population are presented in Fig. S4.KASP® order was coincident in the two maps.In the Mr × Cl map, KUBO-27 was in a recombination bin including KUBO-29 and wPt-2106, while in the Sv × Cc map it was separated by a recombination event.KASP® and wPt-2106 genotypes of the Sv × Cc RILs are reported in Table S5a.Crossing overs were not evenly distributed in the target interval.Only seven RILs showed recombination inside the target interval and were therefore informative for fine mapping (Table S5b).In total, 433 recombinant Sv × Cc RILs in the KUBO-13 to KUBO-9 interval were selected and scored for SBCMV resistance in 2016 and 2017.BLUEs frequency distribution obtained from corrected SEV data is reported in Figure S6.Data presented a mean value of 2.58 with extreme values of 0.29 and 4.27.The distribution histogram showed two peaks, one centered on a SEV score of 1.53 (medium resistant) and one on a SEV score of 3.16 (medium susceptible), with heritability value of 0.83.
The joint genotypic and phenotypic analysis of the Sv × Cc recombinant lines allowed exclusion of the two distal regions marked by KUBO-13 and KUBO-9 flanking markers.Sixty-five resistant lines showed that the QTL was most probably downstream of KUBO-13 and 140 susceptible lines showed that it was upstream of KUBO-9 (Table S6a).Three recombinant susceptible lines (#652, #2077, #3089) together with three recombinant resistant lines (#1450, #126, #1166) showed that the QTL is most probably upstream of KUBO-1 located at 13,011,332 bp in Svevo (Table S6a).However, only one additional recombinant line (#139) provided evidence for the QTL to be downstream of KUBO-27 (Table S6a).Therefore, Sv × Cc RILs positioned the supporting interval of Sbm2 in 0.2 cM between KUBO-27 and KUBO-1, corresponding to 2.661 Mb.
As to the Mr × Cl RIL population, beside the confirmation of results from the Sv × Cc populations, 18 RILs were informative for fine mapping (Table S6b) and confirmed the interval downstream of KUBO-13 (supported by five RILs) and upstream of KUBO-1 (supported by two recombinant lines) as the most probable for harboring Sbm2.Singlemarker analysis (Fig. 6) showed that wPt-2106 was the marker most associated with the phenotype.Consistently with GWAS, KUBO-27 was the most associated with the phenotype among all newly developed KASP® markers.The result of the Interval Mapping analysis is reported in Fig. 6.As expected, the LOD value peaked in the region identified by KUBO-27, wPt-2106, KUBO-29, and KUBO-1, with a peak LOD value of 86.

Comparison among genomes for marker and gene order in Sbm2 region
A comparison of marker order in Sbm2 region among the three wheat genomes (cvs.Zavitan, Svevo, and Chinese Spring) is shown in Figure S6.Order inconsistency among physical maps of the three genomes was observed for 26 markers, mostly located in a few regions that showed rearrangements among assemblies, as in the case of a group of eight consecutive markers (IWB71947, IWB35524, IWB72156, IWB62546, IWB41644, IWB73884, IWB29097, and IWB72157) that identified an inversion of ca. 2 kb in Svevo.The three wheat genomes were therefore compared for gene collinearity in the interval IWB73347 (KUBO-13, position: 9,986,862 bp) and IWB10512 (KUBO-9, position: 24,313,794 bp), including Sbm2.This analysis further confirmed the local inversion spanning the region tagged by IWB45885, IWB4596, IWA7545, IWB26871, IWB26232, IWA3589, and IWB26233 (Fig. S7).No inversions were detected between Zavitan and Chinese Spring assemblies, indicating that the local inversions in Svevo genome could be an artifact of assembly.Additional analysis using longread sequencing data is required to confirm this observation and its potential effects on differential gene expression.

Sbm2 molecular haplotypes in cultivated durum germplasm
With the aim to define the linkage disequilibrium and haplotype patterns in the Sbm2 region, SNP data from the UNIBO durum panel were subjected to local disequilibrium analysis using the linkage block method implemented in the Haploview software.Considering the extended region between KUBO-13 and KUBO-9, six main linkage blocks of various Mb size were identified (Fig. 2).The target interval on chromosome 2B is comprised between 9,986,912 and 24,313,794 bp.Based on the block partitioning, both fine mapping and GWAS results indicate that hapblock-3 harbors the QTL hotspot.
SNPs were partitioned into six haplotype blocks (hapblocks) as in Fig. 2 and their − log 10 (P) of association with SBCMV response are reported as horizontal heat bars in the lower side of the figure (F-G panes).
Detailed local SNP-based haplotypes present in the UNIBO Durum Panel varieties for the KUBO-13 to KUBO-9 region are reported in Fig. 3 and in Data_S2.In Fig. 3, varieties were clustered based on the identity-by-state (IBS) genetic distance matrix and Ward's clustering obtained using SNPs from the Sbm2 region only (Fig. 3, A pane), thus reflecting the genetic relationships among varieties based on Sbm2 haplotypes only.In Fig. 3, clustered varietal haplotypes are shown side by side with their respective whole-genome population structure Q values (eight Q subpopulations, reported as heat bar), SEV-and ELISA-BLUES values, all reported as vertical heat bars aligned from left to right (Fig. 3, B and C panes).Haplotypes of varieties/breeding lines at Sbm2 are shown as horizontal bars adopting a two-color scheme where the SNP allele from the resistant Meridiano, Levante, and Neodur parents was considered as the main reference allele and hence is dark blue colored, while the alternative SNP allele, carried by susceptible parents, is pale red colored (Fig. 3, D, E, and F panes).
Considering the region as a whole, two main haplotype groups (H1 and H2) were identified and were clearly associated with resistance (H1-R) and susceptibility (H2-S and H3-S), respectively.Within resistant and susceptible haplotype groups, sub-haplotypes were defined based on haplotype differentiation, termination, and/or recombination events among haplotypes (indicated as black arrows, numbered from 1 to 6).Additional haplotypes were observed in a group of a few landraces only with a much more complex haplotype structure, mostly associated with a susceptible response (bottom side of the figure).R-and S-haplotype groups showed population structure effects, with H1-R being more frequent in North American, French, CIMMYT'70, Desert Durum, and some ICARDA temperate varieties.Conversely, H2-S was more represented in ICARDA dryland, Italian, French, and Spanish varieties derivatives of Valnova/ Simeto while H3-S was specifically found in highly susceptible varieties derived from the CIMMYT'80 Altar84 founder.
SNPs differentiating the H1-R group from the two H2-S and H3-S groups can already be identified, from left to right, in hapblock-1 (IWB73347 = KUBO13, A/G, indicated by black arrow 1) and hapblock-2 (including all SNPs of the block, reported in Data_S2).However, these SNPs are highly recombined with the phenotype and have low discrimination power, indicating that hapblock-1 and -2 are distant from the causal sequence.For instance, at IWB73347 = KUBO-13, the G variant (in pale red) is present in relatively high frequency in both R and S varieties.Similarly, all SNPs included in hapblock-2 showed variants common to both R and S varieties and were therefore unable to discriminate between resistant and susceptible varieties, while the rare variants discriminated between modern and ancient germplasm only.
Within hapblock-3, a main H1-R haplotype associated with resistance and two haplotypes associated with susceptibility, H2-S and H3-S, can be clearly distinguished.Interestingly, they showed a core haplotype region common to both S haplotypes and highly differentiated from H1-R, comprised between IWB11421 = KUBO-27 (black arrow 2 in Fig. 3) and IWB28973 = KUBO-1 (black arrow 3).From KUBO-1 downwards, still in hapblock-3, the two S haplotypes start to differentiate as to each other and differentiate from H1-R for a wide segment.Therefore, based on GWAS results, the KUBO-27 to KUBO-1 interval is the main candidate to harbor the Sbm2 causal sequence.
Overall, these observations are consistent with the results obtained from the fine mapping experiment with the biparental populations.

Haplotype distribution in a panel of worldwide durum wheat varieties across decades
To broaden the survey regarding Sbm2 haplotypes in cultivated durum wheat, a larger panel of 549 varieties was investigated and genotyped with diagnostic markers KUBO-27, KUBO-29, and KUBO-1.Based on the genotyping results, haplotypes were classified in categories: resistant haplotype, susceptible haplotype, recombinant haplotype of type-1, type-2, type-3, double recombinant haplotype, and admixed.The proportion of accessions represented in each haplotype category is reported in Figure S8.The resistant haplotype was the most frequent, being detected in 53% of the accessions, followed by the susceptible haplotype detected in 31% of accessions.The three categories of recombinant haplotype type-1, type-2, and type-3 represented in total 11%, while accessions with a double recombinant haplotype were less than 1%.
The distribution of resistant, susceptible, and recombinant haplotypes based on varietal origin is reported in Fig. 8.For all origins, varieties with the resistant haplotype represented more than 45% of the total, with Portugal-Spain having the minor proportion of resistant accessions (46%).The highest proportion of resistant varieties was observed for the Canada-USA (81% resistant varieties) and the lowest for CIMMYT-Mexico (49% resistant varieties).In Europe, the resistant haplotype was mostly represented in France (51%), even if the proportion was very similar to that observed for the other European countries, ranging from 46 to 51%.ICARDA materials showed the presence of the resistant haplotype in relatively high frequency (72%).The susceptible haplotype was poorly represented in Canada-USA (14%) and ICARDA (25%), while ranged from 35 to 44% for all the other locations.
Time course haplotype distribution trends were inspected from 1955 to 2020 (Figure S9).Notably, the registration of resistant varieties in North America underwent an outburst in the 1995-2000 interval.Additionally, in Italy and France the release of varieties with the resistant haplotype increased starting from 2000.As to Portugal-Spain, the release of varieties is more evenly distributed throughout the years, with a major release of resistant varieties in the interval 1995-2000.Release of varieties from ICARDA was concentrated in the first half of the considered period, from 1985 to 2000, and most of the novel varieties presented the resistant haplotype (Fig. 8).

Candidate gene analysis at Sbm2
The hapblock-3 hotspot interval between KUBO-27 and KUBO-1 up to KUBO-40/IWB6584/wPt-1601 corresponded  S8), of which 52 were high-confidence (HC) and 41 low-confidence (LC) genes.Genes present in this region appeared to be involved mainly in metabolic, oxidation-reduction and protein phosphorylation processes (Fig. 9).Part of those genes are involved in plant defense response, hence representing candidate genes putatively involved in SBCMV resistance.Such genes belong to defensins, disease resistance protein (NBS-LRR class) family and NBS-LRR-like resistance proteins   S7).In the fine-mapped interval, we observed: 17 cytochrome p450, three receptor kinases, two defensins, one high-confidence and one low-confidence, and three NBS-LRRs in addition to further genes involved in secondary metabolic pathways and transporter activity.Defensins class included an HC gene and a LC gene.As regards to the NBS-LRR genes (HC), two of them encoded for NBS-LRR resistance proteins, while one was considered an NBS-LRR like resistance protein.
Based on the gene ontology enrichment (GO), the molecular functions ascribed to mono-oxygenase activity, terpene biosynthesis, and ion binding are highly represented in the Sbm2 confidence interval between markers KUBO-27 and KUBO-40, based on the logarithm of adjusted p-value (Fig. 9).With regard to the biological processes, general metabolic pathways are highly frequent such as terpenoid and isoprenoid biosynthetic processes and protein phosphorylation (Figure S10), classes known to be involved in the plant-pathogen interaction.
The genes included in the interval between KUBO-27 and KUBO-40 were also explored for KEGG terms using Blast-Koala bioinformatic tool (Kanehisa et al. 2016).As to the organism system attribution, the main represented pathway is "plant-pathogen interaction," which reports the metabolic pathways involved in the recognition between a host organism and a virulent pathogen.As shown in Fig. S11, the gene NBS-LRR RPM1 (TRITD2Bv1G007240) appears to be involved in this pathway as responsible for the hypersensitive response (defense response) as a consequence of the recognition between the NBS-LRR RPM1 product and the pathogen avirulent protein (avr).
Additionally, the gene content in the Sbm2 confidence interval was compared to the candidate genes identified in the fine-mapped Sbm1 region on chromosome 5DL of bread wheat by Liu et al. (2020).The Sbm1 interval was positioned between 546,651,779 bp and 547,273,461 bp, harboring 17 candidate genes from CS RefSeq v1.0 pseudomolecule (IWGSC, 2018).The candidate gene functions belong to Ser/Thr kinase proteins, phosphorylation proteins for signal transduction, membrane transporter binding proteins, transcription factor proteins, and secondary metabolism proteins.Functions such as secondary metabolism, signal transduction by protein phosphorylation, and in particular Ser/Thr proteins and transcription factors are shared with gene functions reported for Sbm2 (Table S8).However, Liu and colleagues (2020) identified as the most promising candidate a pto-interacting protein 1 (PTI1) (TraesC-S5D01G531200) based on RNA-Seq experimental data.This specific gene function was not identified among the Sbm2 candidate genes which, on the other hand, represented other promising candidates.With the aim to find ortholog and/ or paralog genes shared between Sbm1 and Sbm2 durum and bread wheat intervals, the ENSEMBL plant database where orthologs and paralogs of each annotated gene are listed based on alignment identity, common domains, and molecular function was surveyed.The comparison between Sbm2 and bread wheat CS RefSeq v1.0 genome was also performed, retrieving orthologs in CS reference genome and paralogs in Svevo durum wheat genome.The results of the ENSEMBLE search are reported in Data_S4 and Data_S5.As to Sbm1 5D genes, no orthologs/paralogs were identified in the Sbm2 region of chromosome 2B.Most of the Sbm1 5D orthologs mapped to 4A and 5B chromosomes, as expected based on the wheat 4AL/5AL translocation (Devos et al. 1995), and not on chromosome 2 group.On the other hand, mapping of Sbm2 orthologs and paralogs, showed absence of the corresponding orthologs/paralogs in Sbm1 5D in Chinese Spring and Svevo genomes.

Fine mapping of Sbm2
This work was based on the previous knowledge gathered on the genetic basis of SBCMV resistance in durum wheat and was aimed at fine mapping the unique major QTL responsible for this mechanism known so far in durum (Maccaferri et al. 2012).The fine mapping strategy took advantage of the combination of both production of novel recombinants obtained by artificial crossing and use of GWAS in a panel of varieties cultivated worldwide.
Based on these results, the most informative Illumina SNPs mapped in the tetraploid wheat consensus map and in the Svevo genome assembly (Maccaferri et al. 2015 and2019) were successfully converted to KASP® assays paving the way for haplotype tracing, germplasm characterization, and MAS (Koebner and Summers 2003;Maccaferri et al. 2022).SNP markers proved very effective for MAS when coupled with a high-throughput technology which allows greater flexibility as compared to the original SNP array technology (Shavrukov 2016).The 69% SNP-to-KASP conversion rate allowed for an effective exploitation of SNP markers singled out with the Illumina SNP array.The probability of success in converting an SNP to a KASP® marker, however, varied greatly according to each SNP.For some SNPs, the design of suitable codominant KASP® assays was particularly difficult or impossible, as already reported by Makhoul et al. (2020).
To fine map Sbm2, a combination of GWAS and linkage mapping in progenies derived from two different populations, namely the Sv × Cc and the Mr × Cl RILs, was used (Salvi and Tuberosa 2005).Recombinants were identified with two flanking KASP® in a 2.0 cM interval.RIL observations confirmed what was reported by Maccaferri et al. (2011a;2012) on Sbm2 as a major locus for SBCMV resistance in durum wheat.The divergence of the phenotypic response from a normal distribution toward a bimodal distribution supports this hypothesis.However, the absence of a clear bimodality suggests the influence of additional QTLs with minor effects, an expected outcome of QTLome dissection (Salvi and Tuberosa 2015).
The spatially adjusted model applied to phenotypic data analysis effectively allowed for accounting micro-environmental variation and uneven incidence of SBCMV infection across the field, scoring/sampling date, and seasons.It is well known that SBCMV infection effects are strongly influenced by environmental factors such as temperatures and soil humidity that impact both vector and virus spreading, hence causing spatial patchiness of infection incidence (Bayles et al. 2007;Liu et al. 2014).Moreover, effects of cold weather on plant growth can worsen virus symptoms.For these reasons, implementing growth chamber experiments under controlled conditions is highly desirable to standardize infection effects and to increase heritability.
The frequency of informative recombination events between KUBO-13 and KUBO-9 obtained from the two mapping populations was highly heterogeneous and considerably lower in Sv × Cc as compared to Mr × Cl.This was possibly due to local cross-specific suppression of recombination (Akhunov et al. 2003;Conley et al. 2004), supporting the need for diverse genetic materials.Maccaferri et al. (2019) reported that the chr.2BS telomere of the Svevo genome is highly recombinogenic, with an average physical-to-genetic ratio of 1.8 Mb/cM.For the interval between KUBO-13 and KUBO-9, our results demonstrated a physical-to-genetic ratio of 1.3 and 2.5 Mb/cM for Sv × Cc and Mr × Cl RILs, respectively.However, in correspondence of the fine-mapped interval of QSbm.ubo-2BS interval between KUBO-27 and KUBO-1, the ratio was 3.1 Mb/cM for Mr × Cl and 13.5 Mb/cM for Sv × Cc.
The complementary GWAS analysis conducted on modern cultivars from around the world allowed mapping Sbm2 to the same short interval obtained from the RIL fine mapping experiments, hence providing an independent validation of results, while enabling the evaluation of allelic distribution and diagnostic power of significant SNPs in cultivated germplasm.All markers were validated and the assessment of their diagnostic power highlighted wPt-2106_ASO, wPt-2106_HRM,  as the best markers for MAS.
The potential to predict the phenotype at the causal locus among genotypes belonging to different breeding origins/ lineages is important for wide application of MAS in diverse genetic backgrounds and contributes to define the breeding value of a marker (Terracciano et al. 2013).A drawback of SNP markers, as compared to SSR markers, is their biallelism, as SNPs fit with the infinite site and allele mutation model (Kimura and Crow 1964;Kimura 1969).This strongly limits the chance of a single SNP marker to consistently match with the causative mutations and haplotypes at the QTL.Notably, the availability of multiple SNP markers tightly associated with the target locus allows for more accurate haplotype-based screening, a more effective approach for increasing the diagnostic power of MAS (Meuwissen et al. 2001;Mucha and Wierzbicki 2012) and a more effective breeding by design approach (Peleman and Voort 2003).
In this study, several concatenated SNPs highly diagnostic of the allelic state at Sbm2 have been identified and some of them were successfully transformed into high-throughput KASP® or HRM markers, in sufficient number to perform haplotype-based characterization.This enables to trace identity by descent of Sbm2 alleles at high level of confidence.The combination of KUBO-27, wPt-2106, KUBO-29, and KUBO-41 provides the best multi-marker haplotype assay available to ascertain the presence of the resistant-Sbm2 allele for germplasm characterization, including diverse breeder's materials, segregating individuals, and lines from artificial crosses, while characterizing selected breeding lines as possible candidate for varietal registration.
Until now, Sbm2 has been considered the only major QTL to provide SBCMV resistance in durum wheat (Bayles et al. 2007;Maccaferri et al. 2012).If additional major sources of resistance are found in Triticum turgidum and/or other tetraploid wheat germplasm, or in chromosomes A and/or B of bread wheat, it would be useful to pyramid them into new varieties to increase both virus resistance and durability; hence, the KASP® markers developed in this study are instrumental.
Recently, the Jmv1 locus for resistance against the Japanese soil-borne wheat mosaic virus, a Furovirus also transmitted by Polymyxa graminis, has been identified in chromosome 2H of barley (Okada et al. 2020 and2023).A joint comparative analysis of the two loci in Triticeae would increase the chance to better understand the causal locus and mechanism of resistance.

Candidate genes in Sbm2 interval
The Svevo genome enabled the identification of the gene(s) responsible for Cdu-B1, the major locus underlying the cadmium concentration QTLome in grain, mainly based on candidate gene analysis and translational genomics, previously hindered by the low recombination rate in the target region (Maccaferri et al. 2019).A similar approach could be used to dissect the Sbm2 region.Other solutions are represented by alternative cloning approaches which do not rely on positional fine mapping, like the MutRenSeq method proposed by Steuernagel et al. (2016).
A glance at the list of candidate genes proposed in this study identified three gene categories as most promising for underlying SBCMV resistance: defensins, disease resistance protein (NBS-LRR class) family, and NBS-LRR-like resistance proteins.Plant defensins are a family of folded antimicrobial peptides thought to be effective against fungal pathogen (Thomma et al. 2002), while NBS-LRR resistance proteins represent the largest class of plant resistance genes (R genes) and R genes are involved in resistance against diverse pathogens, including viruses (Calil and Fontes 2017).
Specifically to SBCMV, Vallega et al. (2006) proposed a hypothesis about a generalized resistance to virus accumulation in durum wheat, i.e., a mechanism controlling virion accumulation in plants by limiting the spread of the virus particles from the roots to the shoot.Kanyuka et al. (2004), Ordon et al. (2009), andPerovic et al. (2009) reported a potential Sbm1 mechanism defined as "translocation resistance," which prevents the virus to spread from the roots to the stem and leaves in SBCMV-resistant hexaploid cultivars.However, a more complete explanation about the mechanism of SBCMV defense in wheat has not yet been sufficiently defined.Thus, it would be inappropriate to narrow down the search of candidate genes to the R gene category only.Singh and Sharma (2015) reported that several terpenoids have a role in plant defense, acting as phytoalexins, low molecular weight compounds, known to exhibit antimicrobial properties in rice (Prisic et al. 2004) and deterrent for insects and herbivores in corn, lima bean, poplar, and cotton (Rodriguez-Saona and Crafts-Brandner 2003; Arimura et al. 2004;Mithofer 2005;Schnee et al. 2006).Ten genes involved in terpene synthase activity were detected in the Sbm2 region, representing the second most frequent gene category.Another uncertainty lies in disentangling the direct resistance against the virus from the indirect resistance against the Polymyxa vector, particularly because P. graminis is not well known as Polymyxa betae.Additionally, with the objective of providing an overall view of the entire list of genes present in the confidence interval, we did not exclude low-confidence genes.High-/low-confidence classification based on a single reference genome is not a valid criterion to discriminate putative candidate genes.Jupe et al. (2012), in a survey of NBS-LRR genes in the potato genome, showed that many of these genes previously annotated as partial/low confidence by the Consortium (2011) were actually functional.
Herein, an overall view on the list of genes presents in the confidence interval of the Svevo durum wheat reference genome and resistance donor was outlined, providing a starting point for a deeper and detailed scan of functions and biological processes.

Conclusions
In this study, SNP markers were used to dissect the genetic basis of SBCMV resistance in durum wheat.High-density SNP arrays allowed for marker enrichment of QSbm.
ubo-2BS = Sbm2 region, while simple, cheap, and robust SNP-specific assays are required for targeted high-throughput screening such as fast evaluation of large segregant populations and efficient marker-assisted selection.Among the available technical solutions, KASP® technology is currently the most used worldwide.Fine mapping of Sbm2 was achieved, with Sbm2 now delimited to a 0.2-cM interval including wPt-2106 and flanked by overlapping the target QTL.This study provides a series of markers organized into haplotypes closely associated and overlapping to the target QTL, hence highly diagnostic for resistance.Such markers are instrumental for breeding programs to select SBCMVresistant genotypes in durum wheat, with the aim of increasing the frequency of Sbm2-resistant haplotype.This work paves the way for widening the investigation to bread wheat germplasm, where Sbm2 has also been described.
Additionally, the effective development of Sbm2 haplotype tagging KASP® presented herein can be considered as a case study to innovate plant variety testing in order to improve the efficacy and accuracy of European variety tests as well as the assessment of the value for cultivation and use and the decision-making process.
The recent availability of the genome assembly of the parental line and germplasm founder Svevo durum wheat allowed us to compare the genetic and physical maps at the target region and identify putative candidate genes.Therefore, this study provides the basis for further refining the Sbm2 interval and paves the way for the positional cloning of this locus.AGROALIMENTARE (SO.FI.A) project, the support of the Interdepartmental Centre for Industrial Agrofood Research-CIRI Agrofood, and the European Union's Horizon 2020 research and innovation program project "INNOVAR-Next generation variety testing for improved cropping on European farmland," under grant agreement No 818144.

Fig. 1
Fig. 1 SBCMV response GWAS association results for markers in the Sbm2 region on chromosome 2BS.Minus log 10 (P panel A) and R 2 (%, panel B) GWAS analysis results based on the UNIBO Durum Panel performed with visual score (SEV, shown in dark blue) and ELISA (shown in pale red) BLUEs of field data collected in 2005

Fig. 2
Fig. 2 Local linkage disequilibrium (LD) matrix based on pairwise r 2 values in the KUBO13-KUBO9 chromosome 2BS interval harboring Sbm2.The r 2 matrix has been generated in Haploview (Barrett

Fig. 3
Fig. 3 UNIBO Durum Panel local haplotypes in the KUBO13-KUBO9 chromosome 2BS interval harboring Sbm2.Haplotype-based UPGMA identity-by-state IBS genetic similarity was used to group durum panel genotypes and landraces (panel A).Each genotype was associated with population structure Q file at Q = 8 (B) and with SEV and ELISA phenotypic data (C).Critical SNP and KUBO KASP® markers are shown in (D).Haploblocks detected in Haploview are shown in (E) according to Fig. 5. Nucleotide-based haplotypes were

Fig. 4
Fig. 4 Genetic population structure and SBCMV response phenotypic distributions of varieties and breeding lines included in the UNIBO Durum Panel.A neighbor joining tree of the main eight subpopulation groups detected in the UNIBO Durum Panel based on Illumina 90 K SNP.B Box-plot phenotypic distributions of UNIBO

Fig. 5
Fig. 5 Manhattan plots of GWAS association analysis results for SBCMV response performed on the UNIBO Durum Panel.A GWAS for symptom severity scores (SEV; years 2005, 2007 and 2010).B

a
Genetic map positions (cM) are reported as in the consensus map ofMaccaferri et al. (2015) b Physical map positions (bp) are reported as the start positions obtained by blasting the marker sequence on the durum wheat genome reference Svevo(Maccaferri et  al. 2019) c Strand defines based on the direction of the marker sequence d Best linear unbiased estimates (BLUEs) were calculated based on LUEs of A) symptom severity scores (SEV) data from 2005, 2007, and 2010 (SEV_ BLUEs) and of ELISA data collected in 2007 and 2010 (ELISA_BLUES) in Cadriano (Bologna, Italy), spatially corrected Marker Position (cM) a Position (bp) b Strand ( ±) c −

Fig. 6
Fig. 6 Single-marker QTL analysis results for KUBO KASP® markers and wPt-2106 mapped in Sv x Cc population.Interval mapping was based on SEV trait data collected in 2016 (3 dates) and 2017 (5 dates) in Cadriano (Bologna, Italy) for Sv x Cc RILs

Fig. 7
Fig. 7 Distribution of phenotypic values (y-axis) for UNIBO Durum Panel varieties and breeding lines, based on the main three haplotypes detected in hapblock3 (x-axis) for SEV (panel A) and ELISA (panel B) phenotypic values (BLUEs).Phenotypic data were collected in

Fig. 8
Fig. 8 Proportion of SBCMV-resistant, SBCMV-susceptible, and recombinant haplotypes in durum wheat varieties based on their origin.The number in brackets refers to the total number of accessions.

Fig. 9
Fig. 9 Molecular functions (A) and biological processes (B) of genes present in the support interval of Sbm2 based on Gene Ontology

Table 2
Peak markers in Sbm2 chromosome region resulted from the GWAS analysis on the UNIBO Durum Panel

Table 3
Maccaferri et al. (2015)KUBO KASP® markers and wPt-2106 in GWAS performed on the UNIBO Durum Panel a Genetic positions (cM) are reported as in the consensus map ofMaccaferri et al. (2015) b Physical positions (bp) are reported as the start positions obtained by blasting the marker sequence on the durum wheat genome reference Svevo (Maccaferri et al. 2019) c Strand defines the direction of the marker sequence d Best linear unbiased estimates (BLUEs) were calculated based on symptom severity scores (SEV) data from 2005, 2007 and 2010 (SEV_ BLUEs) and ELISA data collected in 2007 and 2010 (ELISA_BLUES) in Cadriano (Bologna, Italy), spatially corrected