Identification and evaluation of resistance to powdery mildew and yellow rust in a wheat mapping population

Deployment of cultivars with genetic resistance is an effective approach to control the diseases of powdery mildew (PM) and yellow rust (YR). Chinese wheat cultivar XK0106 exhibits high levels of resistance to both diseases, while cultivar E07901 has partial, adult plant resistance (APR). The aim of this study was to map resistance loci derived from the two cultivars and analyze their effects against PM and YR in a range of environments. A doubled haploid population (388 lines) was used to develop a framework map consisting of 117 SSR markers, while a much higher density map using the 90K Illumina iSelect SNP array was produced with a subset of 80 randomly selected lines. Seedling resistance was characterized against a range of PM and YR isolates, while field scores in multiple environments were used to characterize APR. Composite interval mapping (CIM) of seedling PM scores identified two QTLs (QPm.haas-6A and QPm.haas-2A), the former being located at the Pm21 locus. These QTLs were also significant in field scores, as were Qpm.haas-3A and QPm.haas-5A. QYr.haas-1B-1 and QYr.haas-2A were identified in field scores of YR and were located at the Yr24/26 and Yr17 chromosomal regions respectively. A second 1B QTL, QYr.haas-1B-2 was also identified. QPm.haas-2A and QYr.haas-1B-2 are likely to be new QTLs that have not been previously identified. Effects of the QTLs were further investigated in multiple environments through the testing of selected lines predicted to contain various QTL combinations. Significant additive interactions between the PM QTLs highlighted the ability to pyramid these loci to provide higher level of resistance. Interactions between the YR QTLs gave insights into the pathogen populations in the different locations as well as showing genetic interactions between these loci.


Introduction
Powdery mildew (PM) and yellow rust (YR), caused by Blumeria graminis f. sp. tritici (Bgt) and Puccinia striiformis f. sp. tritici (Pst) respectively, are the most devastating diseases of wheat (Triticum aestivum L.) in cool climate regions [1,2]. Approximately 6 million ha of wheat in China is grown in areas prone to PM or YR epidemics. Major epidemics occurred in these regions in 1990 with grain yield losses due to PM epidemics estimated at 1.4 million tonnes [3] and to YR at 2.65 million tonnes [2,4]. Growing resistant cultivars is an effective, economical and environmentally safe approach to control these diseases [5].
Plant disease resistance can be classified as either qualitative or quantitative [6]. Qualitative resistance is generally conferred by single genes with large effects against the pathogen and can be observed in both seedling and adult plant stages. To date, 50 loci containing 78 resistant genes/alleles to PM [7,8] and 74 genes (Yr1-Yr74) to YR [9,10] have been identified in bread wheat and its relatives. This type of resistance has a strong tendency to be overcome by new races, particularly when a single gene is deployed over large areas. However while the gene remains effective, it strongly affects the presence and frequency of specific pathotypes in the field [11][12][13]. The majority of this type of resistance loci have been overcome by the pathogen with only a few, including Pm2, Pm4, Pm21 and Pm30, still effective against prevailing Bgt isolates [13,14]. In YR, only Yr5, Yr10 and Yr15 are still effective against the prevalent Chinese Pst races of Pst-CRY32, Pst-CRY33 and Pst-V26 [15].
In contrast, quantitative resistance is mediated by multiple genes or quantitative trait loci (QTLs) [6] and is most commonly observed in adult plants grown under field epidemic conditions. This type of locus often show partial but additive effect against the majority of isolates, and is considered to be broad-spectrum and durable [6], making it highly valuable to breeding programs. Although individual adult plant resistance (APR) genes or QTLs often confer partial and inadequate resistance, combinations of such genes can result in "near-immunity" [16,17]. So far, 119 resistance QTLs for PM [18] and more than 140 QTLs for YR [19] have been identified, with nearly every chromosome harboring at least one resistance locus.
Wheat cultivars containing combinations of effective resistance genes are likely to provide long-lasting control of PM and YR diseases. Cultivar XK0106 showed high level resistance to PM [20] and YR [21] in both seedling and adult stages, while E07901 exhibited APR to both PM and YR [21]. The two cultivars are promising breeding sources with favorable agronomic traits, but little is known about the genetic basis of their resistance to both diseases. The objectives of this study were to 1) map QTLs responsible for resistance to PM and YR in a population derived from E07901 and XK0106 with SSR and SNP marker genetic linkage maps and 2) assess effectiveness of the detected QTLs alone or in combination in different environments.
YR were selected based on their QTL complements and used for the evaluation of QTL efficacy against PM and YR separately in four different environments. Chancellor and Mingxian 169 were used as susceptible checks in PM and YR field trials respectively. Thirty lines with known Pm (S1 Table) and four near-isogenic lines containing key Yr genes (S2 Table) were used in seedling tests.
Isolate Bgt6-11, being incompatible with XK0106 and compatible with E07901, was used for the PM seedling assays, while five Chinese Pst races including Pst-CYR29, Pst-CYR32, Pst-CYR33, Pst-Su11-4 and Pst-V26 were used for the YR seedling assays. The first four of these races have been predominant in China since the 1980s [4], and the race Pst-V26 was first isolated from Chuanmai42 in 2008 and is virulent to Yr24/26 [23]. Sixteen differential Bgt isolates were listed in S1 Table. Seedling assays for PM and YR Seedling resistance assays for PM were evaluated using a detached leaf segment method. Seeds of XK0106, E07901, F 1 , Chancellor and the DH lines were germinated and planted in square pots of 12×12×12 cm and grown to the two-leaf stage (10 days after planting). Leaf segments, 3 cm in length, were cut from the middle part of the primary leaf and placed on 0.5% water agar (w/v) supplemented with 50 mg L -1 Benzimidazole in clear plastic boxes with the abaxial epidermis facing upwards. Three independent replicates were used for each DH line. Inoculation was performed by blowing the spores into a plastic tower at a density of 4×10 3 conidia cm -2 . The leaf segments were then incubated in a growth cabinet with 80% relative humidity and a 12 h light 12 h dark photoperiod at 18±1˚C. Infection type (IT) was scored on a 0-4 scale [24] at 12 days post inoculation (dpi), when the susceptible control Chancellor showed fully developed disease symptoms (IT 4). All lines were classed into two groups according to IT with resistant lines scoring between zero and two, and susceptible lines scoring three or four. XK0106, E07901 and a set of lines with known Pm gene to 16 differential Bgt isolates were also evaluated using the above method (S1 Table).
Seedling resistance assays for YR were evaluated under controlled greenhouse conditions. Thirty seeds of parents, Mingxian169 and single gene lines were sown separately in square pots of 12×12×12 cm. Seedlings at the two-leaf stage (14 days after planting) were inoculated with urediniospores of the five respective Pst races. The inoculated plants were incubated at 10 ± 1˚C in a dew chamber in the dark for 24 h, and then transferred to a greenhouse at 17 ± 2˚C. IT was scored 20 days after inoculation using a 0-9 scale described by Line and Qayoum [25]. Plants with an IT of 0 to 3 were considered resistant, 4 to 6 considered intermediate and 7 to 9 susceptible.

Adult-plant assessment of PM and YR
Field trials for the QTL analysis were conducted at the Hubei Academy of Agricultural Sciences Nanhu farm (30.2856 N, 114.1839 E, 27 m) in Wuhan, Hubei Province. The PM trials were conducted during the wheat cropping seasons of 2011, 2012 and 2013 while the YR trials were conducted in 2010 and 2013. Trials had two replicates and were designed as randomized complete blocks with repeating checks. Each plot consisted of two 1.5 m rows with row spacing of 25 cm. Approximately 100 seeds were sown in each row and the susceptible check, Chancellor or Mingxian169, was planted every 20 plots and around the test lines to ensure ample PM or YR inoculum. The spreader rows of the susceptible checks were exposed to artificial inoculation of PM at stem elongation (Growth stage 30 according to [26]) with mixed conidia from the 16 Bgt isolates that were used in the seedling tests. A similar approach was adopted for YR with mixtures of Pst-CYR32 and Pst-CRY33 urediniospores suspended in the light weight mineral oil Soltrol 170 (Chempoint.com) applied to the spreader plots at the tillering stage (Growth stage 25, [26]). Powdery mildew severity (PMS) or yellow rust severity (YRS) was scored at the mid-grain filling stage (Growth stage 75, [26]). The three upper leaves of 15 randomly selected plants were assessed using a 0-9 scale [27] for PMS and a modified Cobb scale [28] for YRS. Disease severity of 15 plants was averaged to obtain the mean PMS or YRS for each plot.
Further field evaluations were conducted on selected lines from the DH populations that contained different combinations of QTLs. PM evaluations were conducted during 2014 at four sites including Nanhu farm in Wuhan, Wolong farm (32.

Statistical analysis
Chi-square analysis was performed to predict the minimum number of loci contributing to resistance against PM in XK0106 according to the segregation ratio of IT to isolate Bgt6-11.

Genomic DNA extraction, SSR and SNP genotyping
Young leaves of the parents and DH lines were collected and frozen in liquid nitrogen. Genomic DNA was extracted using the CTAB protocol [29]. Three hundred and ninety-five SSR markers were randomly selected from the 21 Somers consensus chromosome maps [30] to test for polymorphisms between the parents. All of the polymorphic SSR markers were used to analyze genotypes of the 388 DH lines. Two STS markers, CINAU15 [31] and CINAU17 [32], and one EST-SSR marker Xedm129 [33] associated with Pm21 were also evaluated in the population.
The PCR assays for the SSR, EST-SSR and STS markers were conducted in an EDC-810 PCR Thermocycler (Dongsheng, Beijing, China) in a reaction mixture (10 μL) containing 10 mM Tris-HCl, pH 8.3, 50 mM KCl, 1.5 mM MgCl 2 , 0.2 mM dNTPs, 25 ng of each primer, 50 ng genomic DNA and 0.75U Taq DNA polymerase. Amplifications were performed at 94˚C for 5 min, followed by 40 cycles at 94˚C for 45 s, 50-60˚C (depending on specific primers) for 45 s, and 72˚C for 1 min, with a final extension at 72˚C for 10 min. The PCR products (2 μL) were mixed with an equal amount of loading buffer and separated on 8% nondenaturing polyacrylamide gels (39 acrylamide: 1 bisacrylamide). Gels were silver stained and photographed.
The genotyping of the sub-population was conducted at the Genome Center of the University of California, Davis. The DNA of 80 DH lines and the parents were extracted and then genotyped through the 90K Illumina iSelect SNP array [34] following the manufacturer's protocol. SNP allele clustering and genotype calling was performed with Genome Studio software v2010.3. Each of the SNP clusters were manually examined to correct imperfect calling of automated clustering. SNP markers with ambiguous SNP calling between parents and/or with a negative hybridization response in most lines were removed from the data set.

Genetic map construction and QTL analysis
Initial linkage group (LG) construction using the polymorphic SSR markers was completed with Joinmap 4.0 [35]. Linkage analysis and marker ordering were carried out using the regression mapping algorithm with a threshold log-likelihood (LOD) ratio !3.0 with the recombination values being converted to genetic distances using the Kosambi mapping function.
LGs were assigned to chromosomes by reference to Somers consensus maps [30].
Initially, mean PMS and YRS of all the lines from each year were used to identify QTLs with the SSR LG map. Composite Interval Mapping (CIM) was performed with WinQTL Cartographer version 2.5 [36] using Model 6 with five markers as controls and employing a window size of 10 cM. Significant thresholds for QTL detection were calculated for each dataset using 1,000 permutations with a genome-wide error rate of 0.05. Phenotypic variance (R 2 ) explained by a QTL was obtained by the square of the partial correlation coefficient. Genetic maps were drawn using MapChart2.2 (http://www.wageningenur.nl/en/show/Mapchart.htm).
The SNP marker-LGs were constructed using MultiPoint software (http://www.multiqtl. com). Prior to map construction, all non-polymorphic SNP markers between parents as well as those markers with greater than 20% missing data were omitted. Eleven lines with poor quality data were also omitted. Segregation of the remaining SNP markers were subjected to Chi-square tests and severely distorted markers deviating from the expected segregation ratio (1:1) at the probability level p = 0.001 were excluded from further analyses. A maximum threshold rfs value of 0.05 to 0.15 with a 0.01 step was used to initially group the markers into different LGs. Multipoint linkage analysis of loci within each LG was then performed with the maximum likelihood (ML) mapping algorithm and the marker order was further verified through re-sampling for quality control via jack-knifing [37]. Markers with known chromosomal locations on the 90K_consensus_map ( [34]; http://wheat.pw.usda.gov/cgi-bin/grain genes/report.cgi?class=mapdata;name=Wheat_2014_90KSNP) were used to assign LGs to chromosomes. Redundant SNP linked markers were removed with the remaining SNP markers being outlined in S3 Table. The complete marker dataset is supplied in S4 Table. These were also used to draw chromosome maps using MapChart 2.2 (http://www.wageningenur.nl/ en/show/Mapchart.htm). This approach was repeated with the combined sets of SSR and SNP markers.
QTLs were mapped on the combined marker LGs using the phenotypic data of 80 remained DH lines of PM and YR. CIM was performed with WinQTL Cartographer version 2.5 [36] with the same parameters as described above. Significant thresholds for QTL detection were calculated for each dataset using 1,000 permutations with a genome-wide error rate of 0.05 (significant) and 0.1 (suggestive). Phenotypic variance (R 2 ) explained by a QTL was obtained by the square of the partial correlation coefficient.

Phenotypic evaluations of PM and YR in seedling tests
XK0106 and F 1 lines of the cross between E07901 and XK0106 were highly resistant to Bgt6-11 (IT 0), whereas cultivar E07901 was highly susceptible (IT 4). The seedling assay of the DH population segregated in a 1:1 ratio (Table 1), indicating that a single gene was involved in XK0106 resistance. This was supported by the IT data for this population as it fitted a Ushaped frequency distribution. However there were a number lines with ITs of 1 to 3, suggesting the possibility that other minor QTLs may have been present in altering seedling IT.
The PM reaction patterns of XK0106 and E07901 to 16 differential Bgt isolates were compared with those of lines possessing known genes. The responses of XK0106 showed an identical pattern to that of Yangmai5/sub.6v (Pm21) with immunity (IT 0) to all the test isolates, while E07901 was susceptible (IT3 or 4) to all isolates except BgtE01 (S1 Table).
The YR seedling ITs showed that XK0106 was resistant (IT 0-3) to four out of five Pst races but susceptible to Pst-V26 (IT 8). This compatible reaction was similar to that observed in the Avocet Yr26 NIL. E07901 was susceptible or moderately susceptible to all of the races tested, indicating that it doesn't have Yr10, Yr15 or Yr24/26 (S2 Table).

Phenotypic evaluation of PM and YR at the adult-plant stage
The mean severity of PM varied from 7.9% to 34.7% over the three years of testing, with 2012 being the highest ( Table 2). XK0106 maintained its immunity in all of these seasons while E07901 ranged between 11.8% and 61.5%. The PMS of the 388 DH lines showed an L-type distribution with approximately half of the lines having a severity of zero while the rest of the lines were continuously distributed (Fig 1A). A number of lines consistently had a higher PMS than that of the susceptible parent E07901 (Table 2).
XK0106 was immune in both of the environments tested while E07901 had YRS scores of 10.0 and 26.5%. Mean YRS of the 388 lines in the DH population ranged between 6.0 and 17.0% and the severity of single DH lines varied between 0 to 100% in each of the two years ( Table 2). The frequency distribution of DHs for YRS was continuous with a pronounced skewness towards resistance ( Fig 1B).

Genetic linkage mapping and QTL analysis
Overall, 134 (33.9%) of the 395 SSR markers showed polymorphisms between the parents. This was further reduced to 117 useful markers when run on the entire population and led to the construction of a genetic map with 26 linkage groups. Apart from chromosomes 1A and 1B, all other wheat chromosomes had between one and three LGs. Chromosome 3B had the greatest coverage with ten SSR markers, while chromosomes 3A, 4B and 6D had the least with only 2 markers each. Chromosome 2B had the longest genetic distance (130.5 cM), while chromosome 5A had the shortest distance (3.9 cM) (S3 Table). The SNP marker set dramatically increased the marker density and chromosome coverage. In all, 11,746 (14.4%) out of 81,587 SNPs showed polymorphisms between parents and 11,330 could be incorporated into the map (including the 117 SSR markers). This resulted in 33 LGs and each was assigned to the different chromosomes of wheat according to 90K_consensus_ map information [34]. Each chromosome contained at least 1 LG and the group D chromosomes had the least representation of markers (225). The genetic map spanned 3,351 cM with an average density of one marker every 2.5 cM. Chromosomes 5B and 4D had the largest (119) and the fewest (14) number of markers, respectively. Chromosome 5A had the longest genetic distance (266.9 cM) and chromosome 4D had the shortest genetic distance (48.7 cM) (S3 Table).
Composite interval mapping was conducted on both the SSR map containing 117 markers from 388 lines as well as the SSR+SNP map containing 11,330 marker from 80 lines. All QTLs identified in the SSR analysis were also identified in the combined marker analysis, although the former analysis always had much higher LOD scores ( Table 3).
The most significant QTL detected for PMS was derived from XK0106 and was located on chromosome 6A. It was designated QPm.haas-6A and was identified in the seedling test (LOD 27.8) and in the field in 2011, 2012 and 2013 (respective LODs of 4.1, 26.3 and 5.9)( Table 3, Fig 2). QPm.haas-2A was derived from E07901 and was located on chromosome 2A. It was significant in the seedling test (LOD 7.9) and in the field in 2011 (LOD 6.7), and was a suggestive QTL (LOD 3.0) in 2013. QPm.haas-3A and QPm.haas-5A were significant in the field assays in 2011 and 2012 respectively. The former QTL, located on chromosome 3A, was derived from   Fig 2).

Resistance evaluation of QTLs to PM and YR
The effects of the different PM QTLs were further tested in four environments in 2014 by selecting three DH lines each that contained various combinations of the different QTLs   A similar process was completed for YR where three DH lines were selected for each of the QTL combinations and tested for YRS in four environments in 2015. There were no lines that contained the combination of all three QTLs. Lines containing QYr.haas-1B-1 showed immune responses in three environments but had the same score as the null lines in Gangu. QYr.haas-1B-2 had a partial effect in reducing YRS, but this was again ineffective in Gangu while QYr.haas-2A had a partial effect that was significant in all environments. The combination of QYr.haas-1B-1 with either QTL was not significantly different from the scores of QYr. haas-1B-1 in Wuhan, Xiangyang and Kunming. However, QYr.haas-1B-1 and QYr.haas-2A did reduce disease severity by more than QYr.haas-2A alone in Gangu. Finally, QYr.haas-1B-2 and QYr.haas-2A acted additively to reduce disease severity in all environments by more than that observed when these QTLs were present alone (Table 5).

Discussion
This study identified resistance loci to PM and YR in a DH population derived from a cross between E07901 and XK0106 and evaluated the effectiveness of these loci in different combinations and environments. One QTL (QPm.haas-6A) for PMS was detected at seedling and adult plant stages in all environments and its seedling reactions and genomic location identified it as Pm21. A further three QTLs were found for PMS, QPm.haas-2A, QPm.haas-3A and QPm.haas-5A. QPm.haas-2A was significant in the seedling test and two field environments and is likely a previously unidentified gene. The YRS study identified three QTLs with QYr. haas-1B-1 having a strong effect in both QTL field environments, while QYr.haas-1B-2 and QYr.haas-2A were only effective in 2013 and 2010 respectively. Seedling pathotype testing and the genomic location also suggest QYr.haas-1B-2 had not previously been identified, although QYr.haas-1B-1 is likely Yr24/26 and QYr.haas-2A is Yr17. A novel strategy was developed to

Comparison of QTLs to known resistance genes
The major seedling resistance of Pm21 was located on chromosome 6A and associated with well characterized markers, in particular, 6V-CINAU15, which is deemed a functional marker for this gene [31]. Indeed this marker, along with the other Pm21 associated markers of 6V-CINAU17 and Xedm129 [32,33], all occurred within the QTL region of QPm.haas-6A. Furthermore, SNP markers also associated with this QTL, RAC875_rep_c69836_475, RAC875_ c48891_87 and the peak marker RAC875_c68978_220, have also been placed in this region through consensus maps [34]. This designation is also supported by seedling tests to 16 B. graminis isolates where XK0106 was immune to all isolates and matched the pattern produced by Yangmai5/sub.6v, a line with the Pm21 containing translocation. Pm21 was introduced into common wheat through the translocation T6VS-6AL derived from 6VS of Haynaldia villosa (2n = 2x = 14, VV) [38]. As it gives high levels of resistance to PM, the T6VS 6AL translocation has been widely used in breeding programs since 2002, particularly in powdery mildew prevalent provinces including Sichuan, Guizhou, Gansu and Jiangsu. Cultivars released such as Neimai8, Neimai836, Neimai10, Neimai11 and Mianmai39 have all been widely planted in Sichuan Province and contain Pm21 [31]. XK0106 originated from Sichuan Province and as its resistance has now been confirmed to contain Pm21, it must also be derived from a T6VS-6AL translocation line. There have been two other introgressions containing PM resistance genes on chromosome 6A, MIRE, introgressed from T. dicoccum [39] and MIG from T. dicoccoides [40]. Map positions clearly differentiate these loci from Pm21 [40]. These 6A introgressions show the value of wild species in contributing useful resistances to the common wheat gene pool.
QPm.haas-2A was flanked by the markers BS00065434_51 and RAC875_c5082_841 with the peak marker being wsnp_JD_c289_450995 (position 164.2 cM Fig 2A). A consensus map of Wang et al. [34] placed this QTL towards the telomere of 2AL. Several QTLs including QPm. inra-2A [41], QPm.vt-2A [42,43], Qpm.ttu-2A [43] and Qpm.crag-2A [44] have been identified on chromosome 2A. Li et al. [18] reviewed all PM QTLs and located QPm.inra-2A on the short Powdery mildew and yellow rust QTLs in a wheat mapping population arm of 2A. QPm.vt-2A was near the centromere on 2AL and was associated with Xgwm312 [42]. On our map this marker is 90 cM (Position 93.8 cM Fig 2A) from the QTL peak marker. The map positions therefore clearly differentiate QPm.haas-2A from the two aforementioned QTLs. Qpm.ttu-2A and Qpm.crag-2A were located near the telomere of 2AL with the former being tightly linked to Xwmc658 [43]. Again in our map this marker is over 28 cM from the QTL peak (Position 192.5 cM Fig 2A). Mingeot et al. [44] mapped Qpm.crag-2A to the same locus as the seedling resistance gene Pm4b, and described the QTL as a residual effect of the defeated gene. S1 Table shows numerous differences between the seedling reactions of Armada, a Pm4b carrying line, and E07901, the QPm.haas-2A donor. This also indicates that Qpm.haas-2A is different from Qpm.crag-2A (Pm4b) and is therefore likely a new QTL for PM. The Qpm.haas-3A and Qpm.haas-5A loci had minor effects in 2011 and 2012, respectively. It was difficult to judge the relationship of these QTL with other known QTLs on chromosome 3A and 5A as there was an absence of shared markers between our maps and other reported maps. However these QTLs and their associated markers could be useful to pyramid minor genes for durable resistance.
Excalibur_c43567_282 and Xgwm413 were associated with QYr.haas-1B-2 and both have been located to chromosome 1BS on consensus genetic maps [34]. Apart from Yr24/26, there are several other genes that have been identified on this chromosome including Yr10 [43], Yr15 [44], YrCH42, YrH52 [48], Yr29/Lr46 [49], Yr64andYr65 [9]. More recently, Yr24/26 and YrCH42 have been shown to be identical due to their similar genomic position and reaction patterns against 26 Pst isolates [45]. The closest known YR genes to QYr.haas-2B-2 are Yr10 [48] and Yr15 [49] and both of these were clearly differentiated from QYr.haas-1B-2 through seedling pathotype tests. Despite detailed mapping, YrH52 could not be separated from Yr15 and as both are derived from T. dicoccoides, they have yet to be clearly identified as different loci [50,51]. E07901 (QYr.haas-1B-2 donor) was MS to S against five pathogens tested, while the Yr10 NIL had a resistant reaction to CYR29 and Sul1-4 and the Yr15 NIL had a resistant reaction to all pathotypes. Furthermore, these genes are rarely used in current breeding programs in China [15,25] with Yr15 being derived from T. dicoccoides [49,50] and presents with significant linkage drag. QYrco.wpg-1B.1 has also been reported in this region as a QTL that had both race specific seedling reactions and robust APR [52]. The marker Xpsp3000 mapped 2-4.4 cM proximal to the seedling reaction QTL QYrco.wpg-1B.1 and 1.2cM from Yr10 [53], suggesting a very similar location for these loci, although pedigree data suggested that they were different genes. QYr.haas-1B-2 could be differentiated from QYrco.wpg-1B.1 with the marker Xgwm413. This marker was 3.2 cM proximal to QYr.haas-1B-2, yet 44 cM proximal to QYrco.wpg-1B.1 [52]. Furthermore, Xpsp3000 and Xgwm413 are 59 cM apart on the Somers Consensus map [30].
QYr.haas-1B-2 is unlikely to be any of the other gene identified on 1B. YrH52 is from T. dicoccoides and has only been introgressed into T. durum with the gene containing segment suffering from negative crossover interference [50]. Yr29/Lr46 is a single locus that is located towards the telomere of chromosome 1BL [54] while Yr64 and Yr65 have only recently been introduced from T. durum and are yet to be deployed in hexaploid wheat [9]. All of these data suggest that QYr.haas-1B-2 is likely a new QTL for YR.
Another QTL for YR was mapped to the telomeric end of chromosome 2AS (QYr.haas-2A) with the SSR markers Xbarc124 and Xwmc177. Several YR resistance genes or QTLs have been reported on chromosome 2AS including the race-specific gene Yr17 [41], as well as QYr.ufs-2A [55], QYr.uga-2AS [56] and QYr.ucw-2AS [57]. Yr17 was also located towards the teleomere of 2AS, and was associated with Xgwm636 [41]. This marker, along with the two QYr.haas-2A associated markers, are within 7 cM of each other on the Somers consensus map [30]. As QYr. haas-2A could not be differentiated from Yr17 by neither map position nor seedling assays, we made the conservative assumption that QYr.haas-2A was likely to be Yr17, although testing with an avirulent pathotype would be required to confirm this.

Resistance evaluation of QTLs to PM and YR
Durable resistance to rust diseases under severe epidemics has been achieved through the combining of three minor loci to control leaf rust [58] and up to five to control yellow rust [59]. Although QPm.haas-6A (Pm21) is still an effective major gene, given its wide-spread deployment over a large area in numerous Chinese cultivars, there is a strong possibility that it will breakdown in the coming years. This study investigates the role the minor QTLs could play in the absence of Pm21. The additive effect of various combinations of these minor QTLs were investigated in several environments by selecting three lines that contained each of the various QTL combinations. This novel approach allowed detailed investigations of the major QTLs without having to grow out the entire mapping population. This has the advantage of being able to grow more replicates of each line, increasing the plot size and being able to take more detailed notes on each plot. These factors result in more accurate scores for each line tested. There is a disadvantage in not being able to identify new QTLs that may be environment specific, however such QTLs are often relatively minor in effect.
Disease pressure had a significant impact upon whether single loci could reduce disease severity. Huanggang and Jingzhou had the lowest disease severity of powdery mildew as evidenced by scores of lines with the null QTL combination. In these low severity sites, both QPm.haas-2A and QPm.haas-3A significantly reduced disease severity, yet had no effect at the high disease sites of Wuhan and Xiangyang. QPm.haas-5A had little effect when present by itself (Table 4). It is doubtful however that any of these loci in isolation would provide much protection of yield even under moderate disease pressures.
Combinations of QTLs identified interesting additive effects. QPm.haas-2A and QPm.haas-5A combined well and reduced disease severity in all environments despite the lack of effect of QPm.haas-5A alone. This is indicative of an epistatic interaction between these two loci. Contrastingly, the QPm.haas-3A/QPm.haas-5A combination was no better than the 3A QTL alone. Such as situation has been previously observed in another YR QTL study where both QTLs on chromosomes 3D and 5D gave moderate protection in isolation and that was no different than the protection provided when they were combined. However these QTLs were clearly additive with all other loci [17]. Such an interaction suggests they may share part of a similar pathway in their respective defense responses and the two PM QTLs identified herein follow a similar pattern.
Lines with all three PM QTLs fared little better than lines with the QPm.haas-2A/QPm. haas-5A combination, again highlighting the non-additive effect of QPm.haas-3A. This is important information for breeders who wish to pursue additive resistances to achieve durability. The two minor genes with additive effects that have been combined in this study reduced disease severity by two thirds in the stronger epidemic environments and further reduced it to negligible levels in less severe sites. It is hoped that, as is the case for the rust diseases, by recombining two to three more loci, near-immunity may be reached. Indeed such loci are readily available. A number of pleiotropic, durable APR loci that give intermediate levels of resistance to leaf rust, yellow rust and stem rust have been identified, and recent work has shown they also have effects against PM. These loci have been termed Lr34/Yr18/Sr57/Pm38 [60], Lr46/Yr29/Sr58/Pm39 [61] and Lr67/Yr46/Sr55/Pm46 [62]. Sound molecular markers are available and it would be a straight-forward breeding exercise to recombine these with the QTLs described herein, in an attempt to generate near-immune lines.
A similar approach of selecting lines with various QTLs was adopted with YR. This gave insights not only into additive effects of different loci, but also of pathogen composition in the various environments. This was most clearly demonstrated with the prevalence of Pst-V26 in Gangu where QYr.haas-1B-1 containing lines scored high, but were immune at other sites. Virulence to Yr24/26 was first detected in China on wheat cultivar Chuanmai42 in the Sichuan Basin in 2008 [63]. Virulence was subsequently shown in Gansu on lines 92R137 and Gui-nong22 [64], and steadily increased throughout the region [65]. This race is not yet dominant in populations in Wuhan, Xiangyang and Kunming, and QYr.haas-1B-1 (Yr24/26) showed excellent resistance in these three sites in 2015. Furthermore, the very low scores of lines combining QYr.haas-1B-1 with other QTLs reflected the overriding response strong seedling resistances have when combined with intermediate levels of resistance. It seems likely that Pst-V26 also has virulence to QYr.haas-1B-2 as this QTL was not only ineffective in isolation in Gangu, but so were the lines that combined it with the Yr24/26 locus.
QYr.haas-1B-2 and QYr.haas-2A had significant effects in reducing YR severity but did not create the immune response as observed with Yr24/26. The QYr.haas-2A effect is consistent with Yr17 where resistance is often incomplete, which can be influenced by genetic background and growing conditions [7]. This locus was still effective in Gangu and partially reduced disease severity. Furthermore, when Yr17 was combined with either Yr24/26 or QYr. haas-1B-2, disease scores in Gangu were further lowered. This suggests that there were mixed isolates in the field, some with virulence to Yr17 and other with virulence to the other QTLs.

Genetic map and the number of QTL detected
In this study, two marker sets were used to empirically investigate the effectiveness of population size and marker density in identifying QTLs. A QTL analysis was initially undertaken with a sparse genetic map (117 SSR markers) but with a large population size (388 lines). A subsequent analysis used a high density genetic map (11,330 markers) with a small population size (80 lines). The QTLs detected in the SSR map spanned much longer chromosome segments and this is not surprising given the low marker density of 10.6 cM per marker compared to the much higher density in the SNP map (2.5 cM per marker). Furthermore, the LOD scores in the SSR map were generally two to four times higher than in the SNP map and is a reflection of the vastly larger population sizes giving much greater confidence in the QTLs observed. However, most telling was the total number of QTLs identified, with seven different loci proving significant in the SNP map, while only three were apparent in the SSR map. This is again due to the greater genome coverage afforded by the SNP map as additionally identified QTLs were mostly in regions without any SSR coverage. The only exception was QYr.haas-1B-2 that was derived from E07901. However this was close to the XK0106 derived QYr.haas-1B-1 which provided immunity that would mask the more minor effects of the former QTL in the sparse SSR map.
In conclusion, major seedling resistance genes were found for both pathogens and these corresponded to Pm21, Yr24/26 and Yr17. Two new QTLs were likely identified in QPm.haas-2A and QYr.haas-1B-2, along with two other minor PM QTLs. QTL combination studies showed the ability to pyramid some of the PM QTLs is a starting point for developing nearimmune lines based on QTLs, and gave insights into the pathogen populations in the YR sites.
Finally, empirical testing of overlapping mapping populations with different marker densities and population sizes highlighted the usefulness of SNP platforms, even in relatively small populations.
Supporting information S1