Mapping of Novel Leaf Rust and Stem Rust Resistance Genes in the Portuguese Durum Wheat Landrace PI 192051

Leaf rust caused by Puccinia triticina Erikss. (Pt) and stem rust caused by Puccinia graminis f. sp. tritici Erikss. & E. Henn (Pgt) are serious constraints to production of durum wheat (Triticum turgidum L). The objective of this study was to identify leaf rust resistance (Lr) and stem rust resistance (Sr) genes/QTL in Portuguese durum landrace PI 192051. Four Pt-isolates, representing three virulence phenotypes (BBBQJ, BBBSJ & EEEEE) and six Pgt-races TTKSK, JRCQC, TKTTF, QFCFC, TPMKC and TMLKC were used to evaluate 180 recombinant inbred lines (RILs) derived from the cross Rusty (rust susceptible) × PI 192051-1 (rust resistant) at the seedling stage. The RILs were also phenotyped at the adult-plant stage in a stem rust nursery in Ethiopia in 2017. The RILs were genotyped using the Illumina iSelect 9K wheat SNP array. PI 192051-1 carries a previously unidentified major Sr gene designated as QSr.ace-7A on chromosome arm 7AS and Lr gene Lr.ace-4A in the pericentromeric region of chromosome 4A. In addition, three minor Sr QTL QSr.ace-1A, QSr.ace-2B and QSr.ace-4A were mapped in PI 192051-1 on chromosomes 1AL, 2BL, and 4A, respectively Lr.ace-4A could be co-located or tightly linked to QSr.ace-4A. Markers linked to the identified QTL/genes can be used for marker assisted selection. These findings enrich the genetic basis of rust resistance in both durum and common wheat.

A few cataloged Lr genes (Lr3a, Lr14a, Lr27+Lr31, Lr61, Lr72, and LrCamayo) have been reported in durum wheat cultivars (Herrera-Foessel et al. 2007, 2008a, 2008bHuerta-Espino et al. 2009a). In addition, the majority of the durum cultivars grown globally appear to carry single racespecific resistance genes. Consequently, this has favored selection for Pt races with virulence to all these genes in just a few years after their deployment (Ordoñez and Kolmer 2007a;Huerta-Espino et al. 2009a, 2009bGoyeau et al. 2012;Herrera-Foessel et al. 2014). The large majority of durum cultivars are susceptible to the durum type Pt races. Therefore, there is an urgent need to identify new Lr genes in durum wheat.
Stem rust is a historically devastating disease of common and durum wheat. Pgt race TTKSK (Ug99) and its evolving variants have virulence to several widely used wheat stem rust resistance (Sr) genes Singh et al. 2011). Currently, more than 70 Sr genes are characterized in wheat. Approximately 31 genes continue to be effective against at least one race of the Ug99 lineage Rouse et al. 2011Rouse et al. , 2014a. Around half of these genes were introgressed from wild wheat relatives (Rouse et al. 2014a;Singh et al. 2015) and only a few genes have been mapped in durum wheat. Resistance to the Ug99 lineage in North American durum cultivars is mainly due to Sr13 alleles, of which Sr13a was first identified in Khapstein, a bread wheat derivative of emmer (T. turgidum, L. ssp. dicoccum) cv. Khapli (Knott 1962, Klindworth et al. 2007, Zhang et al. 2017. A possible Sr8 allele designated as Sr8155B1 is another resistance source in durum wheat found to confer resistance to a variant of the Ug99 race group TTKST, but not to race TTKSK . Besides the limited number of characterized Sr genes/alleles in durum wheat cultivars, the continuous emergence of virulent Pgt-races including TRTTF, TTRTF, JRCQC, and TKTTF that do not belong to the Ug99 lineage have been reported (Olivera et al. 2012(Olivera et al. , 2015rusttracker.cimmyt.org 2017). The recent stem rust outbreak in Sicily, Italy in 2016 caused by race TTRTF, which is unrelated to the Ug99 group, represents a serious threat to the common wheat and durum production in the Mediterranean basin (rusttracker.cimmyt.org 2017).
Landraces may carry new sources of resistance that can be exploited to enrich the narrow resistance spectrum currently found in adapted cultivars. The use of landraces is limited in most breeding programs (Bonman et al. 2007;Newton et al. 2010;Bux et al. 2012;Gurung et al. 2014), which has created a diversity bottleneck resulting in lack of disease resistance (Tanksley and McCouch 1997). Use of landraces like 'PI 192051' from the USDA-National Small Grains Collection (NSGC) could contribute to enhanced rust resistance in durum wheat. PI 192051 was found to be highly resistant to several Pt races (Aoun et al. 2016(Aoun et al. , 2017 and Pgt-races (Olivera et al. 2012;Chao et al. 2017). Aoun et al. (2017) reported that F 1 plants of the cross Rusty · PI 192051-1 were resistant to Pt-race BBBQJ collected from durum wheat in California. This Pt race has a similar virulence pattern to the Mexican race BBG/BN that caused the leaf rust outbreak in Northwestern Mexico in 2001 (Singh et al. 2004;Kolmer 2015). Segregation of F 2:3 families indicated the presence of a single gene conferring resistance to race BBBQJ. The objective of the present study was to map leaf rust and stem rust resistance genes/QTL in PI 192051 and to develop SNP markers tightly linked to the identified loci.

Population development
The population used in this study was developed by crossing 'Rusty' with a single plant selection of PI 192051, PI 192051-1 (Aoun et al. 2017). PI 639869), a durum accession that is a near universally susceptible to wheat stem rust, was selected and released in 2004 by the USDA-ARS Northern Crops Science Laboratory, Fargo, ND, and North Dakota State University (Klindworth et al. 2006). Durum wheat landrace PI 192051 from Lisboa, Portugal where it is known as 'Amarelo de Barba Branca' was deposited in the NSGC in 1950. In this study, Rusty was the susceptible genotype and PI 192051-1 was resistant to both leaf rust and stem rust (Table 1). The hybrid population was advanced by single seed descent at North Dakota Agricultural Experiment Station Greenhouse Complex, Fargo, ND, USA and 180 F 6 RILs were produced.

Leaf rust evaluation
The F 6 -RILs were screened at the seedling stage with four Pt-isolates virulent on durum wheat in the biosafety level-2 facility at the Fargo Agricultural Experiment Station Greenhouse Complex. Pt isolates collected from Ethiopia, Morocco, Tunisia, and USA (California) were designated as Eth50-4, Mor38-2, Tun20-4, and CA1.2, respectively. The virulence/avirulence phenotypes of these isolates were based on seedling stage infection types using 20 Thatcher near-isogenic lines (NILs) as described by Long and Kolmer (1989). Both CA1.2 and Mor38-2 were race BBBQJ (virulent to genes LrB, Lr10, Lr14b, and Lr20), and Tun20-4 was race BBBSJ (virulent to LrB, Lr10, Lr14a, Lr14b, Lr20). Isolate Eth50-4 was avirulent on seedlings of the cultivar Thatcher and was given the race designation EEEEE. The RILs were evaluated for leaf rust response in twice replicated randomized complete blocks (RCBD) with exception of the experiment for isolate CA1.2 that was replicated thrice. In each replicate 8-10 plants/RIL were tested. The parents of the cross, the common wheat cultivar Thatcher and the leaf rust susceptible durum line 'RL6089' were included in each 50-cell tray as checks. Two replicates of Thatcher NIL differentials were included in each experiment to confirm the virulence phenotype of Ptisolates. The seedlings were grown under the same greenhouse conditions as described by Kertho et al. (2015). Inoculum increase, inoculation process, and greenhouse conditions under which the inoculated plants were grown until disease screening were as described by Aoun et al. (2016).
Leaf rust infection types (ITs) were evaluated 12-14 days after inoculation on the second leaf using a 0-4 scale (Long and Kolmer 1989; et al. 1995), where IT 0 = no visible symptom, ; = hypersensitive flecks, 1 = small uredinia with necrosis, 2 = small-to medium-size uredinia surrounded by chlorosis, 3 = medium-size uredinia with no chlorosis or necrosis, and 4 = large uredinia with no necrosis or chlorosis. Larger and smaller uredinia than expected for each IT were designated with + and 2, respectively. Seedlings showing ITs of 0 -2 + and 'X' (a mixture of low and high ITs evenly distributed on the leaf surface) were considered resistant, whereas seedlings showing ITs of 3-4 were considered susceptible (Long and Kolmer 1989;McIntosh et al. 1995). The RILs that showed only resistant plants across replicates were considered homozygous resistant (HR) and the RILs that showed only susceptible plants across replicates were considered homozygous susceptible (HS). RILs that showed both resistant and susceptible plants were classified as segregating (Seg).

Stem rust evaluation
The RILs were also tested at seedling stage with African Pgt-races  (Rowell 1984). The spores were then suspended in mineral oil (Sotrol 170, Phillips Petroleum, Borger, TX, USA), then sprayed onto the primary leaves of the seedlings. The inoculated seedlings were placed in a humidity chamber in darkness for 14-16 h at 18°, then retained for 3-4 h under florescent light to enhance spore germination. The plants were then placed in the greenhouse at 18 6 2°and 16 h photoperiod for 10-12 days when ITs were scored using the Stakman 0 -4 scale (Stakman et al. 1962). Plants showing ITs 0 -2 + 3 were considered resistant and those with IT of 3 -4 were considered susceptible. The classification of RILs into HR, HS, and Seg was conducted as described for leaf rust.
For both leaf rust and stem rust data, the x2 test for goodness-of-fit at 95% level of confidence was used to assess the deviation of observed segregation from 1HR: 1HS theoretically expected of RILs at F 6 and F 7 generations.
Field evaluation of the RIL population at the adult-plant stage was carried out at the international stem rust nursery at the Ethiopian Institute for Agricultural Research Center in Debre Zeit (EIAR-DZ) in 2017. This nursery is an international durum wheat screening site for stem rust as part of the Borlaug Global Rust Initiative's Durable Rust Resistance in Wheat Project. The station is located at 1,900 m above sea level at 8°44' N latitude and 38°85' E longitude. This center is a hotspot for wheat stem rust and can be used for two cropping seasons (July-November and January-May) (Letta et al. 2013).
A total of 160 F 7 -RILs were phenotyped for stem rust response in Ethiopia in 2017. Rusty, PI 192051-1 and the local check 'Arendeto' were included four times (every after 40 entries). The RILs, parents and Arendeto were planted in hill plots with 20-30 seeds/RIL. Stem rust spreaders of susceptible wheat cultivars were artificially inoculated 2-3 times with races TTKSK (Ug99) and JRCQC starting from stem elongation stage. Other races such as TKTTF, TRTTF, RRTTF that are known to be present in this region (Olivera et al. 2012(Olivera et al. , 2015 may have been present in the nursery. Stem rust severity and response were assessed twice at the soft-dough stage of plant development on June 8 th and June 23 rd , 2017. Disease severities were scored following a modified Cobb scale (Peterson et al. 1948;Roelfs et al. 1992). The lines were classified, based on the host response into resistant (R), moderately resistant (MR), intermediate (M), moderately susceptible (MS), and susceptible (S) as described by Roelfs et al. (1992). A combination of two categories of host response on the same line was possible.

SNP genotyping and linkage mapping
Leaf tissue from Rusty, PI 192051-1, and 180 F 6 -RILs were collected, lyophilized for 24 h, and ground as described by . The DNA was extracted using the CTAB protocol (Riede and Anderson 1996) modified by Liu et al. (2006). The DNA was diluted to 50 ng/ml and genotyped at the USDA-ARS Small Grain Genotyping Lab at Fargo using the Illumina iSelect 9K wheat SNP array (Cavanagh et al. 2013) following the manufacturer's protocol (Illumina Inc., SanDiego, CA, U.S.A.). The SNP genotypes were manually scored using version 1.0 of the polyploid clustering module for Illumina GenomeStudio version 1.9.4 (Illumina Inc.).
Markers were sorted according to the parents and all monomorphic markers were removed. Heterozygous genotypes were converted into missing data and polymorphic markers with ,20% missing data were used for mapping. The maps were generated using MapDisto version 1.7.7.0.1.1 (Lorieux 2012) with a LOD score of 4 and maximum recombination frequency of 0.3. The Kosambi mapping function (Kosambi 1943) was used to convert recombination frequencies between SNPs into map distances in centiMorgans (cM). The robustness of the map was evaluated using "ripple order" and "check inversions" commands. Problematic loci that increased the map length by more than 4 cM were identified using the "drop locus" command. After removing each problematic marker, the ripple order and check inversion commands were applied to revalidate the robustness of the maps. The developed linkage maps covered all 14 tetraploid wheat chromosomes and the linkage groups were assigned to chromosomes by comparison with the tetraploid wheat consensus map (Maccaferri et al. 2015).

QTL analysis
The seedling rust responses of the RILs were converted from the 0-4 Stakman scale to a 0-9 linear scale as described by Zhang et al. (2014). In the case of segregating RILs, the weighted means were calculated, where the weighted mean of the RIL= (the most predominant linearized IT Ã 2 + the least predominant linearized IT)/3. The means of the replicates were used for quantitative trait locus (QTL) mapping. The stem rust response data at adult-plant stage were converted into coefficient of infection (CI) and then used for QTL analysis. The CI was obtained by multiplying the severity and a constant for host response, where immune = 0.0, R = 0.2, MR = 0.4, MS = 0.8, S = 1.0, RMR = 0.3, MRMS= 0.6 and MSS = 0.9 (Yu et al. 2011). Pairwise correlations between traits were calculated and plotted on R 3.4.1 (R Core Team 2016) using the 'corrplot' package (Wei 2013). Correlation values were considered significantly different from zero at P value # 0.05. QTL analysis was conducted using QGene 4.0 (Joehanes and Nelson 2008) using three QTL analysis methods: composite interval mapping (CIM), multiple interval mapping (MIM) and MIM-based on a general linearized framework (MIM-GLZ). For CIM and MIM, the QTL analysis was based on transformed phenotypic data (Ln (1+ y) where y= phenotypic response. For MIM-GLZ, the QTL analysis was performed on the non-transformed data. The significant LOD threshold values were determined by performing permutation tests (1000 iterations) at an experiment-wide error of a= 0.05 and the coefficient of determinations (R 2 ) were calculated and used to determine the amount of phenotypic variation explained by the QTL. The 95% confidence intervals of QTL were estimated using the 2-LOD drop method as described by Lander and Botstein (1989). QTL were considered co-localized if their confidence intervals overlapped. After identifying the positions of the QTL on specific linkage groups, co-segregating SNPs were excluded to reduce marker redundancy of the maps. SNP marker flanking sequences were BLAST against the Chinese Spring wheat genome sequence (IWGSC_RefSeq_v1.0) to identify the physical locations of the SNP markers linked to the identified QTL.

Data Availability
The RILs of the population Rusty · PI 192051-1 are available upon request. Table S1 contains phenotypic data used in this study and Table  S2

Leaf rust response
Evaluation of 155-172 F 6 -RILs for response to the four Pt isolates showed bimodal distributions of ITs ranging from 0; to 3 + . The lowest ('0;') and highest ('3 + ') ITs corresponded to those of the resistant and susceptible parents, respectively ( Figure 1 and Table 1). Grouping the RILs into HR and HS showed a segregation ratio of 1HR: 1HS based on x2 test (P $ 0.05) for all four Pt isolates (Table 2 and Table S1). This suggests that the leaf rust resistance to each race in PI 192051-1 was conferred by a single gene. Small differences in number of HR and HS RILs in the four tests were due to differences in numbers of RILs tested with each Pt isolate (Table 2 and Table S1). There were significant correlations (Pearson correlation coefficients (r) were 0.9 to 1, P value ,0.05) between ITs of the RILs for the four studied Pt isolates, suggesting that the same gene conferred resistance to these Pt isolates ( Figure S1).

Stem rust phenotypic evaluation
For stem rust seedling tests, 137-172 RILs were evaluated with three US races and three African races (Table 1). The median ITs of the resistant parent PI 192051-1 to African Pgt races TTKSK, JRCQC and TKTTF were '2 -', '2', '22 + ', respectively, whereas the median IT of PI 192051-1 was '2' to all the three US races. For all the Pgt races, the lowest and highest IT observed in the RILs were similar to those of the resistant parent PI 192051-1 and susceptible parent Rusty, respectively ( Figure 2 and Table 1). The pattern of segregation of RILs screened did not fit a ratio expected for a single gene (based on x2 test at 95% level of confidence) except for Pgt-race QFCFC (Table 3 and Table S1). We observed 35 and 23 segregating families to Pgt races JRCQC and TKTTF, respectively, which is higher than the expected number of segregating RILs at the F 6 generation. This deviation is likely caused by the intermediate nature of the resistant reactions (ITs 22+ to 2+3) to races JRCQC and TKTTF. The complex intermediate reactions make classification of homozygous vs. segregating families difficult.
The RILs were evaluated in two consecutive scoring dates (Eth2017_ 1 and Eth2017_2) in the field nursery in Ethiopia (Eth2017), The median infection responses of Rusty were 30S and 50SMS whereas those of PI 1902051-1 were 5MS and 20MSMR at Eth2017_1 and Eth2017_2, respectively. Arendeto showed median disease responses of 30MSS and 40S at Eth2017_1 and Eth2017_2, respectively. Transgressive Figure 1 Distributions of the seedling response data for Puccinia triticina isolates BBBQJ_CA1.2, BBBQJ_ Mor38-2, BBBSJ_Tun20-4, and EEEE_Eth50-4 for recombinant inbred lines (RILs) from the cross Rusty · PI 192051-1. Distribution is expressed as the percent individuals within a linearized Stakman scale (0-9). Median phenotypes for PI 192051-1 and Rusty are indicated on the graph. segregation led to responses higher than that of Rusty (Figure 2, Table1, and Table S1). There was high correlation between the RIL data obtained at the two scoring dates (r = 0.7, P , 0.05). Responses to Pgt races at the seedling stage were significantly correlated (0.4 # r # 0.6, P , 0.05) with the Eth2017 adult-plant stage data. RIL seedling responses to Pgt races QFCFC, TMLKC, and TKTTF had relatively low but significant correlations (r = 0.2, P , 0.05) with RIL responses to Pt isolates ( Figure S1).

Linkage mapping
A total of 1,138 polymorphic SNPs with ,20% missing data were identified using Illumina's iSelect 9K SNP array. These markers were used to generate a genetic map with a total length of 1,436.24 cM distributed across 20 linkage groups covering the 14 chromosomes of the tetraploid wheat. The number of markers per linkage group varied from 16 SNPs (seven unique loci) in linkage group 2A-1 to 107 SNPs (56 unique loci) in linkage group 5B. The lengths of the linkage groups ranged from 2.8 cM in linkage group 7A-1 to 150.4 cM in linkage group 5B. A total of 176 SNPs showed evidence of segregation distortion from 1:1 ratio based on x2 goodness of fit test at 99% level of confidence. Majority of the distorted markers were located on the linkage groups 2B (34 markers), 3B-2 (49 markers) and 6B (69 markers). For the remaining linkage groups, there number of distorted markers ranged from zero to six SNPs ( Figure S2, Table S3).

Lr gene mapping
Leaf rust phenotypic data for all the four Pt isolates showed bimodal distributions and the segregation ratio were not significantly different from 1HR: 1HS based on the x2 test for goodness-of-fit at 95% level of confidence ( Figure 1 and Table 2). The phenotypic data were converted into binary data based on classification as HR or HS. The binary phenotypic data were then merged with the genotypic data for linkage mapping using MapDisto. QTL analysis was also performed on the linearized leaf rust scores (0-9 scale).
The Lr gene in PI 192051-1 conferring resistance to the four Pt isolates was mapped to chromosome 4A. The Lr gene region was delimited by IWA232 and IWA1793 (11.9-15.9 cM) covering a 4.0 cM region (Figure 3 and Table 4). Lr.ace-4A-CA (to race BBBQJ_ CA1.2) and Lr.ace-4A-Mor (to race BBBQJ_ Mor38-2) were mapped to the same genomic position and at 0.7 cM from Lr.ace-4A-Eth (to race EEEEE_Eth50-4) and 1.9 cM from Lr.ace-4A-Tun (to race BBBSJ_ Tun20-4) (Figure 3). These small deviations in the mapping positions of the identified Lr gene region could be explained by the different numbers of RILs evaluated per isolate. Thus, it is most likely that resistance to these four Pt isolates was conferred by the same Lr gene designated as Lr.ace-4A. The tests for races BBBQJ_ Mor38-2 and BBBQJ_CA1.2 had the highest numbers of RILs evaluated (Table 2), therefore the mapping resolution should be higher for Lr.ace-4A-CA and Lr.ace-4A-Mor. Lr.ace-4A had LOD scores of 64.14, 74.64, 44.28, and 52.91 and accounted for 78, 81, 72, and 81% of phenotypic variations (R 2 ) to the Pt-isolates BBBQJ_CA1.2, BBBQJ_ Mor38-2, EEEEE_Eth50-4, and BBBSJ_Tun20-4, respectively.
Eighty-seven SNPs representing 17 unique loci were mapped on the linkage group corresponding to chromosome 4A with a total length of 107.2 cM ( Figure S2). The mapping region of Lr.ace-4A spanned the centromere with flanking markers IWA232 (159,507,011 bp) and IWA1793 (570,267,558 bp) on 4AS and 4AL, respectively based on BLAST of the SNP marker flanking sequences against the genome sequence of Chinese Spring wheat (RefSeq_v1.0) on the International Wheat Genome Sequencing Consortium (IWGS) website. Each marker flanking Lr.ace-4A had sets of co-segregating markers on either side of the centromere, where IWA232 co-segregated with markers IWA126, IWA8416, IWA4359, IWA6377, IWA4253, and IWA4254 and the SNP IWA1793 co-segregated with IWA8341 ( Figure S2). The marker IWA232 (11.9 cM) on 4AS was the closest to Lr.ace-4A-CA, Lr.ace-4A-Mor, and Lr.ace-4A-Eth therefore, Lr.ace-4A is probably located on chromosome 4AS very close to the centromere. The SNP IWA232 allele "T' is associated with leaf rust resistance while the marker allele 'C' is associated with susceptibility.
RILs with the susceptibility 'T' allele of QSr.ace-7A closest marker IWA8390 had average stem rust IT scores of 7.7, 7.6, 7.9, 8.8, 8.9, and 8.6 to Pgt-races QFCFC, TPMKC, TMLKC, JRCQC, TKTTF and TTKSK, respectively and an average stem rust severity of 50% in the field test Eth2017_2. The RILs with the resistance 'C' allele of the SNP IWA1805 had average stem rust IT scores of 5.3, 5.3, 5.4, 5.3, 6.1, and 4.8 to Pgt-races QFCFC, TPMKC, TMLKC, JRCQC, TKTTF and TTKSK, respectively, and an average stem rust severity of 31.3% in field test Eth2017_2. The distribution of RILs based on alleles at IWA8390 had little overlap. For instance only four RILs with the IWA8390 susceptibility 'T' allele from Rusty had stem rust IT scores , 6 and none of the RILs with the resistance allele from PI 192051-1 had a stem rust IT score . 6. In the field test Eth2017_2 the distribution of RILs based on n alleles at IWA8390 also had little overlap where only three RILs with the resistance allele had severities .50% and only five RILs with the IWA8390 susceptibility allele had severities less than 30%. Very similar results were recorded for the other QSr.ace-7A flanking marker IWA1805 ( Figure S4). A minor effect stem rust QTL, designated QSr.ace-4A, was identified using CIM on the chromosome 4A (Figure 4 and Table 4). QSr.ace-4A had LOD values of 3.11, 3.56, and 3.06 and explained 10, 11, and 8% of the variation in stem rust response caused by races QFCFC, TMLKC, and TKTTF respectively. It was flanked by markers IWA603 and IWA4657 (0 -15.6 cM), with IWA7521 (5.9 cM) being the closest marker to the QTL peak for races QFCFC and TMLKC, and IWA4657 on chromosome 4AL being the closest marker to the QTL peak for race TKTTF. Markers IWA568, IWA4431, IWA482, IWA4432, IWA483, and IWA569 co-segregated with IWA4657 ( Figure S2). QSr.ace-4A spanned the centromere with flanking markers IWA603 and IWA4657 located on chromosomes 4AS and 4AL, respectively. Based on the physical positions of the QTL flanking markers, QSr.ace-4A interval was within 37,813,793 -581,470,783 bp ( Figure 4 and Table 4).
Epistatic interactions were investigated by grouping RILs based on their QSr.ace.7Aand QSr.ace-4A genotypes and statistically comparing the disease means of each RIL group to QFCFC, TMLKC, and TKTTF using Tukey's honest significant difference test (HSD) (Tukey 1949) at 95% level of confidence (Table 5). The RILs with both QSr.ace-7A and QSr.ace-4A had lower linearized stem rust ITs compared to that of the RIL with only QSr.ace-7A, however this difference was statistically significant only for race TKTTF based on Tukey's test (P , 0.05). This suggests that QSr.ace-4A enhances QSr.ace.7A resistance to Pgt race TKTTF. The infection types of the RILs carrying only QSr.ace-4A and those with neither QTL were not significantly different (Table 5).
Minor QTL QSr.ace-1A conferring resistance to Pgt race TPMKC on the distal end of the chromosome 1AL was identified using CIM. Ninety-four SNPs, representing 34 unique loci spanned 144.3 cM. QSr.ace-1A peaking between markers IWA3409 and IWA5734 (131.7 -144.3 cM) with a LOD score of 3.03 explained 10% of the variation in stem rust response ( Figure S3 and Table 4). Based on the physical positions of the flanking markers, the QSr.ace-1A physical confidence interval was within the region 571,580,052 -593,399,125 bp. IWA8523 (141.2 cM) was the closest marker to QSr.ace-1A. IWA3409 co-segregated with IW2450 and IWA5734 co-segregated with IWA3799, IWA3492, and IWA5806 ( Figure S2). RILs with both QSr.ace-1A and QSr.ace-7A had significantly lower stem rust ITs compared to RILs with only QSr.ace-7A based on Tukey's test (P , 0.05). This suggests that QSr.ace-1A enhances QSr.ace.7A resistance to Pgt race TPMKC. The presence of only QSr.ace-1A did not confer detectable resistance to race TPMKC (Table 5).
Even though the segregation ratio of the RILs to Pgt races TTKSK and JRCQC did not fit a single gene model based on x2 test at 95% level of confidence, no additional QTL to QSr.ace.7A was identified using CIM. Therefore, other QTL mapping methods MIM and MIM-GLZ were used to investigate the possibility of other loci being involved in resistance to these races. The MIM-GLZ algorithm is known to have greater power to overcome model assumption violations in phenotypic data compared to traditional algorithms such as CIM and MIM that rely on transformed phenotypic data (Joehanes 2009;Dahleen et al. 2012;Zurn et al. 2018). The use of MIM-GLZ helped to detect QSr.ace-2B-a to race JRCQC flanked by IWA2261 and IWA6399 (90.9 -92.7) at the distal end of chromosome 2BL. BLAST of the flanking SNP sequences against the Chinese Spring genome sequence showed that QSr.ace-2B-a was within 650,285,204 -665,668,837 bp. QSr.ace-2B-a had a LOD score of 5.30 (LOD threshold = 4.79) and explained 13% of the phenotypic variation. MIM revealed identified QSr.ace-2B-b to race TTKSK flanked by IWA6399 and IWA7955 (92.7 -105.3) on chromosome 2BL. The physical positions of the flanking markers showed that n

Figure 3
Mapping of leaf rust resistance gene Lr.ace-4A in PI 192051-1 to P. triticina isolates BBBQJ_CA, BBBQJ_Mor, BBBSJ_Tun, and EEEEE_Eth at seedling stage. QTL analysis was performed using composite interval mapping. The QTL analysis LOD threshold is indicated with the blue line. Co-segregating markers were excluded from this map. SNPs and their cosegregating markers are presented in Figure S2.
QSr.ace-2B-b was within 665,668,638 -720,480,388 bp. This QTL with a LOD score of 5.98 (LOD threshold = 3.72) explained 14% of the phenotypic variation. Markers flanking both QTL also had co-segregating markers ( Figure S2). Markers IWA469/IWA5414 (92.2 cM) and IWA2318/IWA6399 (92.7 cM) were important in detecting QSr.ace-2B-a, whereas IWA5414 (92.2 cM) and IWA6399 (92.7 cM) were necessary in mapping QSr.ace-2B-b. The genomic and physical regions of both QSr.ace-2B-a and QSr.ace-2B-b overlapped suggesting they were likely same QTL conferring resistance to races TTKSK and JRCQC. These QTL are henceforth designated QSr.ace-2B ( Figure S3 and Table  4). RILs with both QSr.ace-2B and QSr.ace-7A had lower linearized stem rust infection types (4.35 to TTKSK and 5.21 to JRCQC) compared to those of other RIL genotypes, however, this difference was not significantly lower than that of RILs genotypes carrying QSr.ace-7A alone based on Tukey's test at 95% level of confidence (Table 5).

DISCUSSION
New sources of resistance are essential to effectively protect durum production against continuously and rapidly evolving rust pathogens. In the current study PI 192051-1 showed resistance to durum type-Pt isolates collected in Ethiopia, Morocco, USA, and Tunisia. This genotype also carries stem rust resistance that is effective against both African and North American Pgt races. A Lr gene Lr.ace-4A, and four Sr QTL, QSr.ace-7A, QSr.ace-1A, QSr.ace-2B and QSr.ace-4A, were identified in PI 192051-1.Linked SNPs to these stem rust resistance QTL and leaf rust resistance gene and the SNP alleles associated with resistance were identified. Primer sequences for development of Kompetitive allele specific PCR (KASP) markers for all wheat Illumina SNPs, including those identified as associated with resistance in this study, are publicly available at http://polymarker.tgac.ac.uk/ (Ramirez-Gonzalez et al. 2014, 2015. Lr.ace-4A conferring resistance at the seedling stage was mapped to the centromere region of chromosome 4AS. The only previously cataloged Lr gene on chromosome 4AL and close to the centromere is Lr30, identified in common wheat cultivar Terenzio (Dyck and Kerber 1981). To our knowledge, Lr30 has not been reported in durum and also appears to be quite rare in common wheat germplasm. For instance, a Lr gene postulation study on a worldwide common wheat collection of 275 accessions showed that only two accessions from North America possibly carried Lr30 (Dakouri et al. 2013). An evaluation of PI 192051-1 with North American common wheat Pt-race TNRJJ that is virulent on Thatcher NIL carrying Lr30 showed that PI 192051-1 was highly resistant (IT ;) (M. Aoun, unpublished). However, PI 192051-1 could have Lr72 or another durum gene that gives resistance to common type wheat races. Based on the Lr30 map on the GrainGenes website (https://wheat.pw.usda.gov/GG3/), this gene is flanked by IWA4359 (physical position: 531,679,637 bp) and IWA2585 (physical position: 544,177,878 bp). IWA4359 is one of the flanking makers of both Lr.ace-4A and Lr30. There was an overlap of the physical regions of the durum gene Lr.ace-4A and the common wheat gene Lr30. Lr30 was reported as a recessive resistance gene (Dyck and Kerber 1981) which suggests it may be different from the dominant resistant gene Lr.ace-4A in PI 192051-1. However, it is possible that expression differs in a durum background compared to a common wheat background. There are no reports of cataloged Lr genes in durum wheat on chromosome 4A thus, Lr.ace-4A is considered a novel Lr gene in durum.
Markers associated with leaf rust resistance on 4A were observed in an association mapping study (AM) using the USDA-NSGC from which PI 192051 was selected (Aoun et al. 2016). The AM revealed five SNPs associated with leaf rust response on 4A. One of these was IWA1570 (59.9 cM) within the mapped region of Lr.ace-4A (51.3-64.0 cM) based n All QTL were identified using composite interval mapping except for the QTL on 2B that were mapped using multiple interval mapping (MIM) and MIM-based on a general linearized framework (MIM-GLZ) for races TTKSK and JRCQC respectively. b The underlined nucleotide is the SNP allele associated with stem rust resistance.
c 95% confidence intervals of the QTL were estimated using the 2-LOD drop method as described by Lander and Botstein (1989). f Pgt races QFCFC, TPMKC, TMLKC, JRCQC, TKTTF, and TTKSK were tested at seedling stage in the greenhouse. Eth2017_1, Eth2017_2 are adult plant stem rust phenotypes from two dates in Ethiopia 2017.
g IWA7521 is the closest marker to the QTL peak to races QFCFC and TMLKC and IWA4657 is the closest marker to the QTL peak TKTTF.
on the tetraploid wheat consensus map (Maccaferri et al. 2015) and marker physical positions on the Chinese Spring genome sequence. PI 192051-1 is resistant to several durum type Pt isolates virulent to Lr3a, Lr27+31, Lr61, and Lr72 (Aoun et al. 2016). This Portuguese landrace showed a high level of adult-plant stage resistance in field trials conducted in USA, Mexico (Aoun et al. 2016), Morocco, and Ethiopia (Aoun and Acevedo, unpublished). Therefore, future evaluation of the Rusty · PI 192051-1 population to other Pt isolates at seedling and at adult plant stages in several geographical locations could reveal other genes in PI 192051-1. One of the Pt-isolates used in this study, BBBSJ_Tun20-4 is virulent to the widely used Lr14a, therefore the deployment Lr.ace-4A in breeding programs could provide an additional source of resistance particularly in the regions where virulence to Lr14a is prevalent (Ordoñez and Kolmer. 2007a;Goyeau et al. 2006;Soleiman et al. 2016). PI 192051-1 also carries QSr.ace-7A on chromosome arm 7AS conferring seedling resistance to both North American and African Pgtraces as well as to Pgt-races at the adult-plant stage in Eth2017. Chao et al. (2017) reported that PI 192051 was resistant at seedling stage to many USA Pgt races, including BCCBC, MCCFC, QFCSC, QTHJC, RCRSC, RKQQC, and TTTTF, but was susceptible to the Ethiopian race TRTTF. The same study showed a high resistance level in PI 1902051 at the adult plant stage in stem rust nurseries at St. Paul, USA and Debre Zeit, in Ethiopia. Since cataloged Sr genes on chromosome 7AS have not been characterized in tetraploid or hexaploid wheat, QSr.ace-7A is a novel Sr gene in wheat. A number of QTL on 7A, associated with stem rust resistance at both the seedling and adult-plant stages have been reported in AM and linkage mapping studies in durum wheat (Haile et al. 2012;Letta et al. 2013;Chao et al. 2017). Chao et al. (2017) showed that IWA7200 was associated with stem rust response in the durum lines in the USDA-NSGC. The SNP IWA7200 (68.7 cM) on 7AS is close to the mapped position of QSr.ace-7A (IWA8390 at 62.4 cM), based on the tetraploid wheat consensus map (Maccaferri et al. 2015) and also based on the physical positions of the markers on the Chinese Spring genome sequence. Letta et al. (2013Letta et al. ( , 2014 reported two DArt markers on 7AS (wPt-2799 and wPt-7885) associated with response to Pgt-races in field trials in Ethiopia in a durum panel of 183 cultivars and breeding lines. In another study in Ethiopia Haile et al. (2012) identified QTL for stem rust resistance on chromosome 7A in the durum cultivar Sebatel. One of the QTL on 7AS (QSr.1PK-7A.1) was flanked by SSR markers gwm974 and gwm631, while the second (QSr.1PK-7A.2) was thought to be Sr22 which is located on 7AL. However, Sr22 is a diploid wheat-derived gene and its presence in durum would be unlikely. Since the markers flanking QSr.1PK-7A.1 are absent in the tetraploid consensus map and the SSR flanking sequence information is not available on the GrainGenes database, comparison between the genomic locations of QSr.ace-7A and QSr.1PK-7A.1 was not possible.
Three minor stem rust resistance QTL were inherited from PI 192051-1 (QSr.ace-1A, QSr.ace-2B and QSr.ace-4A). QSr.ace-2B is close to the position of Sr9 (Rouse et al. 2014a). Since JRCQC and TTKSK are virulent to Sr9e, QSr.ace-2B could be Sr9h that is effective against race TTKSK and originally from durum cv. Gaza (Nirmala et al. 2016). In an AM study by Letta et al. (2014) gwm1300 was associated with seedling resistance to races TTTTF and TTKSK. This SSR marker is very close to the position of QSr.ace-2B based on marker positions in the tetraploid consensus map.
In this study we observed an overlap between the genomic regions of QSr.ace-4A (0 -15.9 cM) and Lr.ace-4A (11.9 -15.9 cM) and a significant correlation between stem rust responses and leaf rust responses of the RILs (r= 0.2, P # 0.05). This suggests Lr.ace-4A is co-located or linked to a stem rust resistance gene with minor effect (QSr.ace-4A) to races QFCFC, TMLKC and TKTTF. The Sr genes Sr7 (Knott 1959, Knott and Anderson 1956, McIntosh et al. 1995 and SrND643 thought to be an allele of Sr7 (Basnet et al. 2015) are present on chromosome 4AL. Saini et al. (2018) reported QSr.rwg-4A (likely to be Sr7a) in the Figure 4 Mapping of stem rust resistance QTL (QSr.ace-4A and QSr.ace-7A) in the population Rusty · PI 192051-1 at seedling stage to races QFCFC, TPMKC, TMLKC, JRCQC, TKTTF, and TTKSK and at adult plant stage in stem rust nursery in Ethiopia in 2017. All QTL were identified using composite interval mapping. The QTL analysis LOD threshold is shown with the blue horizontal line. Co-segregating markers were excluded from these maps. SNPs and their cosegregating markers are presented in Figure S2.
durum cultivar Lebsock at the distal end of chromosome 4AL whereas QSr.ace-4A was located in the pericentromeric region of chromosome 4A, thus QSr.ace-4A is unlikely to be Sr7.
QSr.ace-1A on chromosome 1AL was associated with resistance to Pgt-race TPMKC at the seedling stage. A cataloged Sr gene mapped to chromosome 1AL was not previously detected; however, QSr.ipk-1 conferring stem rust resistance at the adult plant stage in the durum cultivar Sebatel was located on 1AL (Haile et al. 2012). The SSR markers Xbarc148 and Xbarc119 flanking QSr.ipk-1 are distant from the position of QSr.ace-1A based on marker locations in the tetraploid consensus map and the marker physical positions on the Chinese Spring genome sequence. Rouse et al. (2014b) mapped QSr.cdl-1AL, an adult plant resistance QTL, at the distal end of chromosome1AL in Thatcher, a derivative of Iumillo durum. QSr.cdl-1AL was tightly linked to the DArt marker XwPt6869 which was associated with epistatic interactions for leaf rust resistance in durum wheat (Singh et al. 2013).
In this study, we identified significant positive (enhanced resistance) QTL interactions. The two minor effect QTL, QSr.ace-4A and QSr.ace-1A, enhanced significantly the seedling resistance of QSr.ace-7A to Pgt races TKTTF and TPMKC, respectively. However, these QTL does not provide resistance in the absence of QSr.ace-7A. This implies that breeders will not grain much by introgressing only the minor QTL, thus QSr.ace-4A and QSr.ace-1A will be useful only if accompanied by QSr.ace-7A.
This study revealed a number of stem rust and leaf rust resistance genes/QTL with major (Lr.ace-4A and QSr.ace-7A) and minor effects (QSr.ace-1A, QSr.ace-2B, and QSr.ace-4A) in durum landrace PI 192051-1. This is the first study to identify a Lr gene on chromosome 4A (Lr.ace-4A) of durum wheat. Stem rust resistance QTL QSr.ace-2B on chromosome 2BS is possibly Sr9h, and the remainder QTL (QSr.ace-1A, QSr.ace-4A, and QSr.ace-7A) appear to be uncharacterized. QSr.ace-7A was mapped on chromosome arm 7AS where no previously cataloged Sr gene has been identified in either durum or common wheat. Lr.ace4A could be the same or closely linked to a minor QTL for stem rust resistance named QSr.ace-4A. Genes/QTL that confer resistance to multiple diseases are suggestive of new molecular structures, as well as allowing simultaneous breeding. Three of the resistance genes/QTL in PI 192051-1 (Lr.ace-4A, QSr.ace-4A, and QSr.ace-7A) were effective against diverse Pt or Pgt races from different countries. Markers closely linked to these QTL will facilitate their introgression into adapted durum wheat cultivars.

ACKNOWLEDGMENTS
We thank Amanda Swank and the wheat personnel at the EIAR-Debre Zeit Research Center for their technical support. We also thank Mohamed Salah Gharbi and Filippo Bassi for kindly providing durum wheat leaf samples infected with P. triticina from Tunisia and Mororcco respectively. This work was supported by the North Dakota Wheat Commission and the Delivering Genetic Gain inWheat Project (BMGF Grant Number OPP1133199).  a All QTL were identified using composite interval mapping except for the QTL on 2B that were mapped using multiple interval mapping (MIM) and MIM-based on a general linearized framework (MIM-GLZ) for race TTKSK and JRCQC respectively.

LITERATURE CITED
b Disease scores were based on a linearized Stakman scale (0-9) for all races. c Numbers in the same column followed by the same letter are not significantly different at P = 0.05.