Genetic Variance for Fusarium Crown Rot Tolerance in Durum Wheat

Tolerance to the cereal disease Fusarium crown rot (FCR) was investigated in a set of 34 durum wheat genotypes, with Suntop, (bread wheat) and EGA Bellaroi (durum) as tolerant and intolerant checks, in a series of replicated field trials over four years with inoculated (FCR-i) and non-inoculated (FCR-n) plots of the genotypes. The genotypes included conventional durum lines and lines derived from crossing durum with 2-49, a bread wheat line with the highest level of partial resistance to FCR. A split plot trial design was chosen to optimize the efficiency for the prediction of FCR tolerance for each genotype. A multi-environment trial (MET) analysis was undertaken which indicated that there was good repeatability of FCR tolerance across years. Based on an FCR tolerance index, Suntop was the most tolerant genotype and EGA Bellaroi was very intolerant, but many durum wheats had FCR tolerance indices which were comparable to Suntop. These included some conventional durum lines, V101030, TD1702, V11TD013*3X-63 and DBA Bindaroi, as well as genotypes from crosses with 2-49 (V114916 and V114942). The correlation between FCR tolerance and FCR-n yield predictions was moderately negative indicating it could be somewhat difficult to develop high yielding FCR-tolerant genotypes. However, FCR tolerance showed a positive correlation with FCR-i yield predictions in seasons of high disease expression indicating it could be possible to screen for FCR tolerance using only FCR-i treatments. These results are the first demonstration of genetic diversity in durum germplasm for FCR tolerance and they provide a basis for breeding for this trait.

100 based on different procedures for inoculation (Li et al. 2008;Mitter et al. 2006; Yang et al. 2010). An 101 outdoor pot assay known as the "terrace" system involving growing plants on terraces in open ended 102 tubes containing 0.24 g of FCR inoculum has been routinely used for screening bread wheat lines in 103 South Australia (Wallwork et al. 2004), but this method suffers from high variability (Liu and Ogbonnaya 104 2015). FCR screening in field-based disease nurseries is also commonly practised (for example, Martin  115 of sources, using the "terrace" system, and found partial resistance in four T. dicoccum lines. Ma et al. 116 (2012) reported absence of variation for resistance in 400 unspecified durum genotypes using a 117 glasshouse test. In many studies the assessment of resistance has been based on seedling studies 118 which might have been able to identify only the best resistance but not the intermediate or partial 119 resistance which might be expressed in adult plants (Wallwork et al. 2004). Also, all the screening and 120 pre-breeding efforts in Australia (e.g. Martin et al. 2013), to date, have focussed on symptom-based 121 assessment of resistance to FCR but the correlation of this resistance with yield or yield loss is 122 uncertain. The ability of durum lines developed from pre-breeding research with improved resistance to 123 FCR to maintain or improve production levels in the presence of this disease has yet to be determined. 140 previous years to determine the most reliable trial protocols, including the most appropriate trial design 141 and sowing date because FCR data tends to be highly variable (data not included). Each field trial had 142 four replications. Plots were 2 m wide and 10 m long. Within a trial, genotypes were grown as both 143 inoculated with FCR (FCR-i), and as non-inoculated bare seed (FCR-n) side-by-side using a split plot 144 design with genotypes as main plots and FCR treatments allocated randomly to the subplots. Rainfall  150 seed at a rate of 2 g Fp inoculum/m row, as described by Dodman and Wildermuth (1987

315
The , = 1, 2, 3, were assumed to have a factor analytic structure of order 1 (denoted as FA (1) where is a 4 × 4 diagonal matrix whose non-zero elements are the specific variances for each 329 environment. One of the variance parameters in 2 is set to one to ensure identifiability of the  396 Higher levels of disease developed in the FCR inoculated plots and produced visually observable 397 differences between FCR inoculated plots and the non-inoculated controls next to them. FCR 398 inoculated plots were generally less vigorous with lower biomass and they took longer to reach ear 399 emergence relative to the FCR treated plot (data not presented). These differences resulted in lower 400 yield in the treated plots (Fig 3).  Table 3.
409 In 2016, the variance components for CRI from FCR-i and FCR-n were small and were not statistically 411 significantly different from zero. This was consistent with the low FCR disease pressure in that year 412 due to the high rainfall (Fig 1). In contrast, in 2017 the variance components for CRI from FCR-i and 413 FCR-n were large, with the FCR-i being significantly different from zero. There was a strong 414 correlation between CRI from FCR-n and FCR-i in 2017.

415
416  In the next step, five variance models which incorporated correlation between environments were 459 fitted to the variance structure of the VFE effects and these are summarised in Table 5. These 460 variance models included two separable and three non-separable models. The AIC values (Table    Table 7 presents the REML estimates of the rotated loadings for FCR tolerance and yield for FCR-n 486 respectively. Table 8 presents the REML estimate of the genetic correlation between environments for 487 FCR tolerance, and between FCR tolerance and yield predictions for both FCR-n and FCR-I, for each 488 pair of environments. There were moderate positive correlations between trials for FCR tolerance 489 (Table 8 and Fig 4) . The complexity of the genotype by environment interaction for FCR tolerance is 490 less than that for FCR-n yield, which is reflected in the loadings for each of these two traits. Since the 491 loadings for the first factor are all the same sign it is therefore possible to derive the overall 492 performance for FCR tolerance for these environments.    (Fig 1). The REML estimates 514 of the correlations between the FCR-n and FCR-i effects for each genotype (Table 6)     557 than expected yield loss and thus a moderate to high correlation between FCR tolerance values 558 between years (Fig 4). The two check varieties performed as expected with EGA Bellaroi showing the 559 third lowest tolerance and Suntop showing the highest tolerance.

560
561 In this study, there was a consistent and reasonably strong genotype effect showing low yield loss in 562 certain genotypes. It was therefore considered better to use a simpler measure of tolerance that directly 563 relates to the ability of a genotype to tolerate the disease and to produce a relatively higher yield rather 564 than percentage yield loss or the regression method. This method of using the simple difference 565 between yield of FCR-n and FCR-i yields as the measure of FCR tolerance of genotypes using EBLUPs 566 from a robust MET analysis has provided an objective method of determining the tolerance status of 567 durum genotypes. This approach has also produced a very useful ranking of genotypes which agrees 568 with results from other datasets (unpublished DBA data). However, some authors have used 575 trials and a key condition for the use of a covariate is that it cannot be affected by the treatment applied 576 (Elashoff 1969). There was little difference between the genotypes within FCR treatments for CRI in 577 2016, possibly due to better growing conditions which limits disease expression. Also, there was little 578 correlation between CRI from FCR-i and FCR-n plots in 2016. The range of CRI increased in the 2017 579 season (Fig 2), most likely due to dry conditions, and there was a strong correlation between CRI from 580 FCR-i and FCR-n plots.