Plant materials
Two populations of doubled haploid (DH) lines were provided by InterGrain Pty Ltd (Perth, Western Australia, Australia). The populations had been developed using wheat × maize wide-crossing system at the Department of Primary Industries and Regional Development (DPIRD). One population (designated 05Y002) consisted of a random sample of 247 lines derived from the F1 generation of a cross between the cultivars Calingiri and Wyalkatchem, both of which carry the tsn1 gene for insensitivity to ToxA. The other population (designated 03Y260) consisted of a random sample of 97 lines from a cross between a breeding line, IGW2574 and Annuello, neither of which carries tsn1. Wheat cultivars ranging in their tan spot resistance were used as controls. These included H45 [moderately resistant (MR)], Magenta (MR), Cunderdin [moderately resistant to moderately susceptible (MRMS)], EGA Bonnie Rock (MRMS), Mace (MRMS), Arrino [moderately susceptible to susceptible (MSS)], Reeves (MSS), Machete [susceptible (S)], Correll [susceptible to very susceptible (SVS)], Datatine (SVS), Yitpi (SVS) and Gutha [very susceptible (VS)].
Genetic map development
DNA was isolated from leaf tissue sampled from one seedling per line using a mini-prep ball bearing extraction method (Rogowsky et al. 1991) with some modifications (Pallotta et al. 2000). DNA from each line was assayed on a DArT marker array (Akbari et al. 2006) by Diversity Arrays Technology Pty Ltd (Canberra, ACT, Australia), with each DArT marker scored based on the presence or absence of hybridisation. Simple-sequence repeat (SSR) markers were assayed using Multiplex-Ready technology (Hayden et al. 2008). Several single-nucleotide polymorphisms (SNPs) were assayed using KASP technology (LGC Limited, Teddington, Middlesex, UK).
Within each population, a Chi-squared test was used to compare the observed segregation ratio of each marker to the expected ratio of 1:1. Markers showing significant segregation distortion (p < 0.01) or markers for which data were available for fewer than 80% of the lines were excluded from map construction. The remaining markers were grouped into linkage groups and initially ordered within linkage groups using the program Multipoint (http://www.multiqtl.com/). Marker order was obtained using the MSTmap (Wu et al. 2008) function in the ASMap package (Taylor and Butler 2017) available in the R statistical computing environment (R Core Team 2016). Genetic distances between marker loci were determined using the hidden Markov algorithm derived by Lander and Green (1987). Linkage groups were assigned to chromosomes and oriented within chromosomes via comparisons with published maps. Final linkage maps are presented in Supplementary Material 1. Missing allele scores were imputed using the flanking marker algorithm of Martínez and Curnow (1992).
Based on the results of QTL mapping, one chromosome from each cross (1A for IGW2574/Annuello and 2A for Calingiri/Wyalkatchem) was selected for more detailed analysis. Additional SNPs were assayed on these chromosomes. For some SNPs, KASP primer sequences (Supplementary Material 2) were obtained from CerealsDB (http://cerealsdb.uk.net; Wilkinson et al. 2012). For others, KASP primers were designed from new sequences. Some of these sequences were kindly provided by Dr. Timothy March of the University of Adelaide. Others were obtained via application of DArTseq technology (Diversity Arrays Technology Pty Ltd) to the mapping parents and selected lines. KASP markers were designed and assayed using software (Kraken), instruments (a SNPLine system) and reagents from LGC Limited. For each of the two chromosomes, a new linkage map was constructed using only the KASP marker data.
Phenotypic evaluation
The two DH populations, their parents and the control lines were phenotyped at South Perth, Western Australia (S31°59.20
′
, E115°53.13
′
) during 2009, 2010, and 2011; at Horsham, Victoria (S36°74.56
′
, E142°11.13
′
) during 2010 and at Toowoomba, Queensland (S27° 32.00
′
, E151° 56.15
′)
during 2011, 2012, and 2013.
South Perth, Western Australia
Inoculum was prepared using a modified method of Evans et al. (1993). Five 4 mm2 plugs of isolate WAC11137 (obtained from WAC, Plant Pathology Culture Collection, Perth, Western Australia) were excised from the periphery of a freshly growing culture on potato dextrose agar and transferred to 200 ml of potato dextrose broth (PDB) in a 500-ml Erlenmeyer flask. Flasks were incubated at room temperature and light conditions on an orbital shaker at 250 rpm for 5 days. The mycelial mat was poured onto a sterile nylon screen and squeezed with a sterile spatula to remove the excess PDB. A sample of 10 g of the mycelial mass was homogenised in 20 ml of sterile distilled water for 30 s at 13,500 rpm using a TH-220 tissue homogenizer (Omni International, Marietta, GA, USA) with a G10-195 saw tooth probe. A sample of 20 ml of the homogenised mycelial suspension was added to 1 l of liquid clarified V8 juice agar medium maintained at 48 °C. Flasks were swirled to distribute the homogenised mycelium in the liquid agar medium and poured into Petri plates. Plates were incubated for 48 h at 24 °C in darkness after which aerial mycelium was appressed to the agar surface using three drops of sterile distilled water and a sterile bent rod. Plates were then incubated for 24 h at 24 °C under a combination of fluorescent and near-ultraviolet light (50 μmol/m2/s) to induce conidiophore production and then for 24 h at 16 °C under darkness to induce conidial production. Plates were examined under a dissection microscope to verify conidial production and then stored in air tight plastic bags at −20 °C, with spores remaining viable for up to 1 year. For inoculation, plates were thawed at room temperature for 1 h and spores harvested by spinning agar discs in 0.5% gelatine solution (e.g., 50 discs in 500 ml of gelatine solution) at 175 rpm for 5 min in a 4-l glass beaker. Approximately 80,000–130,000 spores per Petri plate were produced using this method. The spore concentration was adjusted to 3000 spores/ml in 0.5% gelatine solution for all inoculations.
During 2009, both populations, their parents and the control lines were evaluated for resistance in an irrigated field nursery in a randomised block design with three replications. The three replicate blocks of the experiment were sown at one-week intervals, in the second, third and fourth weeks of May. Each experimental unit consisted of a 30-cm row in which 15 seeds were sown. Rows were 20 cm apart. Plots were fertilised with a mixture of superphosphate, urea and potash (6:4:1) at a rate of 100 kg/ha at planting and at 8 weeks after sowing. Plots were protected from powdery mildew infection with 250 g/ha of Quinoxyfen and 125 g/ha Bupirimate applied at 4-week intervals for 12 weeks. Blanket spray inoculations using conidial suspensions were done for each block when most lines had started to tiller (Zadoks growth stage (Z) 21; Zadoks et al. 1974) and again at half head emergence (when most lines had reached Z55). To promote humidity for fungal infection, plots were watered prior to inoculation. After inoculation, blocks of plots were covered by plastic tents and shade cloth (84–90% cover factor) for 48 h. Seven days after the first (tillering-stage) inoculation, disease severity was assessed for each plot using a 0–5 scale where 0 is no disease and 5 is high disease. Two weeks after the second (heading-stage) inoculation, percentage leaf area diseased (PLAD) was assessed on the flag leaf and the leaf below (flag-1) of five plants in each plot. Mean PLAD values were calculated for each plot and used for analysis. The Zadoks growth stage of each plot was recorded 12 weeks after sowing.
In 2010, a similar experiment was conducted, but with modifications to the trial design and inoculation protocol in order to overcome possible effects of plant maturity and height on disease expression at the adult plant stage (Shankar et al. 2008). Lines were sown in plots consisting of two 10-cm rows 10 cm apart, with up to ten seeds sown per row and with 30 cm between adjacent plots. Plots were fertilised and protected from powdery mildew infection as described above. Individual plots were inoculated at different times as they reached heading (Z55), by spraying flag leaves with the conidial suspension to run-off. High humidity was ensured by watering the site just before inoculation and by using plastic bags secured over PVC rings (15 cm high, 30 cm diameter) to cover individual plots for 48 h after inoculation (Shankar et al. 2008). Before being used to cover the plots, the plastic bags were misted internally with water. To shade the inoculated plants from direct sunlight, the plastic bags themselves were covered with shade-cloth bags (84–90% cover factor). At 390 °C thermal days (sum of mean daily temperatures) after inoculation, PLAD was scored on the flag leaves of five individual plants in each plot and a mean was calculated and used for analysis.
For phenotypic evaluation of seedlings, glasshouse experiments were conducted during both 2009 and 2010 with both populations, their parents and the control lines. Lines were sown in 120-mm-diameter pots containing a sand-loam mix with 1 g of Osmocote (slow release fertiliser). The experiments were conducted in three-replicate randomised block designs, with four lines sown within each pot. Temperatures were maintained at 24 °C during the day and 22 °C at night. At Z12.5, seedlings were spray inoculated to run-off with the conidial suspension as described above. Inoculated plants were incubated for 24 h in a humidifier. Nine days after inoculation, a 0–5 scale was used to rate disease symptoms on the leaves that had been fully emerged at inoculation. In 2010, an additional experiment was conducted in which the two populations were inoculated at tillering (Z22) and rated 9 days later.
In 2011, the IGW2574/Annuello population, its parents and the control lines were grown in a controlled environment with 24/22 °C day/night temperatures. The experiment was conducted in a randomised block design with three replicates. Four seeds per line were sown within each pot. Growing medium, fertiliser, seedling inoculation and the initial disease rating were the same as described above for the glasshouse experiments. Immediately after the initial disease rating, plants were provided with a 20-h photoperiod consisting of 12 h of natural day light and 8 h of high pressure sodium light with an active radiation of 400–500 μmol/m2/s. Plants were fertilised with soluble all purpose Thrive N:P:K 25:5:8.8 (Yates Australia, Padstow NSW) at a concentration of 0.8 g/l and a rate of 60 ml/pot on a weekly basis and with a trace element solution of Liberal BMX (BASF) at a concentration of 0.5 g/l and a rate of 30 ml/pot on a fortnightly basis. At heading (Z55), flag leaves of individual pots were inoculated as described above. Fourteen days after this inoculation, disease severity symptoms were rated on a percentage scale.
Horsham, Victoria
In the field, during 2010 three replicates of each line of the two populations, parents and controls were sown in randomised block design in an irrigation bay on 18 May 2010. Each experimental unit consisted of a 50-cm row in which approximately 20 seeds were sown. Rows were 30 cm apart. One week prior to sowing the field site was pre-drilled with mono-ammonium phosphate fertiliser (70 kg/ha) treated with flutriafol fungicide (75 g/ha) to suppress stripe rust (Puccinia striiformis) development. Infection was established by spreading approximately 1 t/ha of wheat stubble naturally infected with locally occurring P. tritici-repentis from the previous year in June 2010. The site was flood irrigated twice during season to stimulate disease development. Disease severity was rated on a 1–9 scale where 1 is low disease severity and 9 is high disease severity at tillering (Z22) and booting (Z45).
For glasshouse experiments, inoculum was prepared using methods described by Mclean et al. (2010) for Pyrenophora teres f. maculata. Three virulent isolates 07-0003, 03-0025, 03-0152 (obtained from the culture collection of Agriculture Victoria, Horsham) were grown on potato dextrose agar under white fluorescent and growlux lights at 20 °C for 5–7 days. Two 4 mm2 plugs of each cultured isolate were then sub-cultured onto V8 juice agar plates and incubated in the conditions described above for 6 days. Inoculum was prepared by scraping conidia and mycelium from the surface of the plates using an electric toothbrush. Two seeds of each line were sown into 5-cm pots containing potting mixture, fertiliser and trace elements. Lines were arranged in a four-replicate randomised complete block design under natural light at 20 ± 2 ºC. Seedlings were inoculated at the two-leaf stage (Z12) with a spore and mycelium suspension with concentration of ~80,000 parts per ml of spores and mycelial fragments. Inoculated plants were incubated at 95–100% humidity at a temperature of 20 ± 1 ºC for 72 h with the first 24 h in darkness and the following 48 h under a 12-h photoperiod. Inoculated seedlings were then returned to a glasshouse for 8 days to allow symptom development, after which the seedlings were assessed for symptom severity using a 1–9 scale.
Toowoomba, Queensland
During 2011, both populations, their parents and the control lines were sown into a field nursery in a two-replicate randomised block design under shade cloth. To enhance pathogen sporulation and infection, humidity was increased two or three times weekly, as necessary, at sunset using rainwater supplied micro misters for 30–60 min. Each plot was a 20-cm-long single row with 20 cm between plots. To avoid drought stress, plants were irrigated approximately fortnightly using drip irrigation. Infection was established by spreading between rows, infected stubble from the previous year. At early tillering (Z22), tan spot symptom severity was rated on a 1–9 scale.
During 2012–2013, both populations, their parents and the control lines were assessed in a controlled environment room at 23 ± 1 °C with each lighting fixture containing both Sodium Vapour and Metal Halide bulbs emitting PAR at 400–500 μmol/m2/s. Prior to seedling rating plants were grown with 14-h day and 10-h night and post rating were switched to 20-h day and 4-h night to reduce the time to head emergence. Plants were grown at 20–22 °C in a 14-h day in 55-mm Square Native Tube pots, 160 mm high (Garden City Plastics, Monbulk, VIC, Australia), containing Searles Native Mix potting soil (Searles Pty Ltd, Kilcoy, QLD, Australia). Seeds were pre-germinated at 5 °C and four seeds per entry were planted in each pot. The statistical layout was a partially replicated design. For all of the experiments conducted in Queensland, an isolate BRIP28204 originally isolated from wheat in the Darling Downs QLD in 1999 was used (Queensland Plant Pathology Herbarium, BRIP, Brisbane, Australia). The isolate, sub-cultured a maximum of three times after isolation from infected leaves, was grown on thick plates of clarified V8 juice agar medium (approx. 15 ml per 9 cm diameter Petri plate) and incubated for 5 days in the dark at 25 °C after which aerial mycelium was appressed to the agar surface using three drops of sterile distilled water and a sterile bent rod. Plates were then incubated at 22 °C for 8 h under a combination of fluorescent and near UV light to induce conidiophore production and then for 16 h under darkness to induce conidial production. Sporulating plates were stored at −70 °C for up to 1 year. For inoculation, sporulating plates were thawed at room temperature and spores harvested by applying a strong spray of deionized H2O to the culture surface to dislodge spores. One drop of tween was used per 100 ml of spore suspension and the suspension was kept on ice and with constant agitation until inoculation. For both adult and seedling screening spore concentration was adjusted to 120 ± 30 spores per ml and at growth stage Z12.5 the top 2–3 seedling leaves were inoculated with 1 ml of spore suspension. Inoculated plants were incubated in the dark for about 36 h with a humidifier operating 30 min on and 30 min off. When supplementary susceptible control plants exhibited expected disease reactions (at around 9 days after inoculation), disease was assessed on a 1–9 scale on the leaves that had been fully emerged at inoculation. Immediately after rating, plants were returned to the growth room and foliar and soil fertilised with half strength soluble Thrive fertiliser (Yates Australia, Padstow NSW), on a weekly basis. When individual plants had reached growth stage Z55, flag leaves of individual pots were inoculated at around Z55 as above with approximately 3 ml of spore suspension and rated 14 days later using a 1–9 scale.
Statistical methods
Single experiment linear mixed model
Tan spot severity measurements recorded as percentages were scaled to proportions and logit transformed to ensure model assumptions were adequately satisfied. The remaining tan spot severity traits (recorded as 0–5 and 1–9 rating scales) and data for developmental traits (Zadoks growth stage, days to heading and plant height) were analysed on their original scales. Each trait was then initially analysed using a linear mixed model that partitions and accounts for genetic and non-genetic information arising from the field, glasshouse or controlled-environment. The linear mixed model had the form:
$$\varvec{y} = \varvec{X\beta } + \varvec{Zu} + \varvec{Z}_{\varvec{g}} \varvec{g} + \varvec{e,}$$
(1)
where y is a vector of trait observations and Xβ is the fixed component of the model containing a term that captures the mean effects of the DH population, parents and control lines. This component may also contain covariate information such as terms to adequately capture linear trends that may exist in the environment of the experiment. The Zu is a random component consisting of terms used to model non-genetic variation due to the experimental design such as blocks or benches. The component Z
g
g was used to model the underlying genetic variation of the trait among the DH lines. Where appropriate, the residual error, e, was used to capture extraneous variation or correlation that may arise from known dependencies between observations in an experiment. For example, in the field trials, the error term consisted of a residual variance as well as a correlation structure parameterized as a separable AR1 × AR1 (AR1 = auto-regressive process of order 1) that models the dependency of the observations due to the proximity of neighbouring plots in the experiment. The set of effects (u, g, e) were considered to be mutually independent. For each of the fitted trait models, the best linear unbiased predictors (BLUPs) of the DH lines were extracted and broad-sense heritabilities were calculated using the formula derived by Cullis et al. (2006).
Multiple experiment linear mixed model
For each of the populations, the complete set of tan spot severity traits collected across the experiments were analysed in a multi-experiment (ME) linear mixed model of the form described in (1). To ensure a uniform trait measurement, tan spot ratings that were taken on a 0–5 or 1–9 scale in any experiment were scaled proportions and then logit transformed. The ME linear mixed model consisted of the complete set of non-genetic terms associated with each of the single experiment models. It also contained a genotype-by-experiment term (Smith et al. 2005) to capture the genetic variation of the DH lines in each experiment as well as the genetic correlation of the DH lines between each pair of experiments. From the fitted ME linear mixed model, the estimated genetic correlation matrix was extracted and summarised.
QTL analyses
For all traits in each experiment, the whole genome average interval mapping (WGAIM) approach was used for QTL analysis (Verbyla et al. 2007, 2012). This used an extension of the linear mixed model defined in (1) by incorporating a whole genome approach for detection and selection of QTL. Preceding the analysis, pseudo-intervals were calculated at the mid-points between adjacent markers using Verbyla et al. (2007). A working linear mixed model was then proposed that includes all pseudo-intervals as random covariates with a single variance parameter. This parameter was then tested for significance using a likelihood ratio test of the working model against the initial linear mixed model. If it was significant, an outlier statistic is calculated for every interval and the interval with the largest outlier statistic is selected as a QTL. The interval was then removed from the contiguous block of intervals and placed in the model as a separate random covariate. This forward selection process was repeated until the variance parameter associated with the remaining random pseudo-interval effects was not significant. The additive set of selected QTL was then summarised with their location on the genome, the sizes of the QTL effects, their LOD scores of significance and their percentage contribution to the total genetic variance.
For each of the populations, an ME linear mixed model QTL analysis was conducted using the ordered set of KASP markers spanning the chromosome containing the main QTL of interest. This was achieved through the incorporation of individual markers as a marker-by-interaction fixed component in the ME linear mixed model. Each marker was then considered in turn and a separate ME QTL linear mixed model was fitted for all markers spanning the chromosome (Bonneau et al. 2013). To detect the marker most likely linked to the putative QTL, the marker-by-experiment term for each ME QTL linear mixed model was tested under a null hypothesis that all single marker effects across experiments were simultaneously zero. A Wald statistic (Kenward and Roger 1997) was calculated for each marker and used to form an ME QTL profile spanning the chromosome. The peak of this Wald statistic profile was considered to be the marker most likely linked to the QTL and the KASP markers at this location were then summarised across individual experiments showing their LOD (logarithm base 10 of odds) scores. Linkage maps of chromosomal regions were drawn using MapChart (Voorrips 2002). Additionally, to understand the reduction in disease symptoms achievable when the favourable allele is used with these markers, the marker effect sizes were back-transformed from the logit scale and the average reduction in the severity of tan spot was calculated from a fixed level of 50% disease severity.
Computations
All models were analysed using the flexible linear mixed-modelling software ASReml-R (Butler et al. 2009) available in the open-source statistical software platform R (R Core Team 2016). For all single experiment QTL analyses, the R package wgaim (Taylor et al. 2011; available at the Comprehensive R Archive Network, http://CRAN.R-project.org/package=wgaim) was used. The package uses the linear mixed model functions in ASReml-R for fitting QTL models and also contains tools for post model diagnostics and summary of the QTL. For the multi-experiment QTL analysis, unpackaged R code was used to implement the marker scanning algorithm.
Development and evaluation of lines with resistance alleles at three loci
Marker-assisted single-seed descent was used to develop lines that carry alleles at two QTL with tsn1. One line was selected from each population based on good resistance and the presence of resistance-associated marker alleles in the QTL region (on 1A in IGW2547/Annuello and on 2A in Calingiri/Wyalkatchem). These lines were crossed with each other. All F1s were ‘grass clumps’. These were grown at a constant 26 °C temperature and treated with 0.5 g/l of gibberellic acid at 2-day intervals (Yao and Canvin 1969) to induce flowering. One hundred and ninety-four F2 seeds were obtained. Progeny were progressed through subsequent generations by single seed descent with marker-assisted selection applied for the resistance allele of each QTL, for the tsn1 allele and for homozygosity at the Vrn-A1 phenology locus. Sixteen F5 lines homozygous for the resistance alleles at the three tan spot resistance loci and at Vrn-A1 were selected. These lines, along with the four original parents and a larger set of lines (not discussed in this paper) were assessed at the seedling and adult stages in a controlled environment and at the adult stage in the field at South Perth. At the seedling stage, a rating scale of 0–5 was used. At the adult stage, the percentage of diseased flag leaf was estimated.
Tan spot severity measurements recorded as percentages were logit transformed and all traits were analysed using the single experiment linear mixed model modelling method described above. In this validation model, the grandparents of the two populations, the parents selected for this crossing scheme and the 16 derived F5 lines were fitted as fixed effects to ensure accurate comparative hypothesis testing of their means could be achieved. The remaining lines were fitted as random effects to adequately capture their genetic variation. For each trait, the best linear unbiased estimator (BLUE) for each line was extracted from the fitted model. A least significant difference (LSD) was then calculated at an alpha level of 0.05 and used to broadly compare BLUEs. For percentage traits, all inference was performed on the scale of the transformation. For these traits, the extracted BLUEs were back-transformed to simplify interpretation.