Evaluation of wheat genotypes resistance to Fusarium head blight in Paraguay

ABSTRACT Fusarium head blight (FHB), or scab, caused by a Fusarium spp. complex, is an important wheat disease in Paraguay. Among the strategies used to control it, the genetic resistance is considered highly efficient and cost effective. This study aimed to evaluate and compare the effects of F. graminearum on six wheat genotypes, including two comparison varieties, in two seeding dates. The genotypes were artificially inoculated in the spike, at the flowering stage, by injecting a pool of four pathogenic F. graminearum isolates. The FHB development was evaluated by scoring the disease incidence and severity, percentage of diseased spikelets and damaged kernels, as well as using the area under the disease progress curve. Besides the kernel infection, its impact on the development of mycotoxins (deoxynivalenol) and interactions with the genotypes were also evaluated. The results identified an advanced breeding line (Lin 84) with a resistance level to FHB comparable to that of the universally known resistance sources (Sumai 3 and Frontana). The other three genotypes (Caninde 11, Caninde 12 and Caninde 21), in spite of presenting a higher grain yield potential, were evaluated as moderately susceptible to susceptible. These results suggest that, although it is possible to transfer the FHB resistance to a higher agronomic type, combining such resistance with a higher grain yield potential remains an ongoing challenge.


INTRODUCTION
The wheat production generates an immense economic impact worldwide, being one of the most Fusarium head blight (FHB), or scab, caused by a Fusarium spp. complex, is an important wheat disease in Paraguay. Among the strategies used to control it, the genetic resistance is considered highly efficient and cost effective. This study aimed to evaluate and compare the effects of F. graminearum on six wheat genotypes, including two comparison varieties, in two seeding dates. The genotypes were artificially inoculated in the spike, at the flowering stage, by injecting a pool of four pathogenic F. graminearum isolates. The FHB development was evaluated by scoring the disease incidence and severity, percentage of diseased spikelets and damaged kernels, as well as using the area under the disease progress curve. Besides the kernel infection, its impact on the development of mycotoxins (deoxynivalenol) and interactions with the genotypes were also evaluated. The results identified an advanced breeding line (Lin 84) with a resistance level to FHB comparable to that of the universally known resistance sources (Sumai 3 and Frontana). The other three genotypes (Caninde 11, Caninde 12 and Caninde 21), in spite of presenting a higher grain yield potential, were evaluated as moderately susceptible to susceptible. These results suggest that, although it is possible to transfer the FHB resistance to a higher agronomic type, combining such resistance with a higher grain yield potential remains an ongoing challenge.

PALAVRAS
Besides its severe impact on grain yield and quality, the fungus produces secondary metabolites (mycotoxins), including deoxynivalenol (FDA 2010, . The disease control must include multi-faceted strategies addressing genetic resistance, proper chemical application and efficient growing practices. Cultural practices primarily involve crop rotation and incorporating crop residues, in order to reduce the field inoculum load. While methods using preventive chemical control are prevalent, their efficacy has not been fully documented, and the use of resistant cultivars constitutes the most effective strategy to reduce the impact of the disease (Mazzilli et al. 2007, Díaz de Ackemann & Kohli 2013, Reis & Carmona 2013. Schroeder & Christensen (1963) defined three types of genetic resistance against FHB. Type I is resistant to initial infection and evaluated based on disease incidence in the presence of natural inoculum or by forced-artifice spray infection (Miedaner et al. 2003). Type II refers to resistance against the spread of spike infection (e.g., test by point injections). The third type is related to the accumulation of mycotoxins, as proposed by Mesterházy (1995). It is well documented that all types of FHB resistance are complex, being based on multiple genes, with small effects.
From China, the Sumai 3 cultivar has been reported as a source of Type I resistance to FHB. Catbird, another advanced breeding line, developed by the Centro Internacional de Mejoramiento de Maíz y Trigo (CIMMYT), and derived from a Chuan Mai#18/Bagula cross, is a source for Type II resistance, located on chromosome 7D (Cativelli et al. 2013). Catbird and Sumai 3 are considered to be stable sources of resistance to FHB, and have been widely used in breeding programs in South America and throughout the world . Another variation frequently used in South America is the Frontana cultivar, from Brazil. It presents Type I resistance and is useful for comparisons with candidates for this type of resistance (Alves et al. 2013a).
Crop breeding programs must identify genotypes that produce higher yields and grain quality, while simultaneously reducing damages caused by diseases such as FHB (Pereyra & Lori 2013, Steiner et al. 2017. The incidence and severity of a disease may help to characterize its phenotypes (Engle et al. 2003, Sharan et al. 2004). The evaluation of relative severity over time allows calculating the area under the disease progress curve (AUDPC), and provides a better phenotyping (Siou et al. 2014). Considering its importance to the wheat production in Paraguay, this study aimed to evaluate the resistance to FHB in certain advanced wheat genotypes.

MATERIAL AND METHODS
Six wheat genotypes, presenting different reactions to Fusarium head blight (FHB), were obtained from the Instituto Paraguayo de Tecnología Agraria (IPTA) ( Table 1). These were seeded on two seeding dates (May and June 2016), at the Centro Multidisciplinario de Investigaciones Tecnológicas -Universidad Nacional de Asunción (CEMIT-UNA), San Lorenzo, Paraguay. Seeding on two dates was done to expand the period of testing and observations. Seeding was carried out using a randomized blocks design (plots of 1 x 0.5 m 2 each), with six replications per group (inoculated and a noninoculated control), using the following varieties: Sumai 3, Frontana (resistant), Caninde 12 and Caninde 11 (susceptible). A substrate composed of a mixture of peat and gravel, at a ratio of 3:1, was used to plant the seeds, and urea (46 %) was used in two phases (80 kg ha -1 at tilling and 40 kg ha -1 at stem elongation). A standard chemical protection was applied in accordance with general recommendations for wheat crops before the anthesis (Kohli et al. 2012).
The presence of trichothecene synthesis pathway genes, related to the production of deoxynivalenol, was determined by Tri5 marker (Doohan et al. 1999, Watson & Wang 2012. Upon verification of their aggressiveness, four mono-spore isolates were selected to evaluate the wheat genotypes (Table 2).
Fungal isolates were cultivated in a modified-CLA medium and incubated for 10 days, at a temperature of 22 ± 5 ºC, under continuous illumination (Cazal et al. 2014). Afterwards, the plates were scraped with 10 mL of sterile distilled water with 0.01 % Tween 20, and briefly shaken. Conidial suspension for each isolate was adjusted to 6 x 10 4 conidia mL -1 , and then mixed to obtain the isolate pool used for artificial inoculation (Mazzoni & Peixoto 2016).
Artificial spike inoculation with pathogenic isolates (Table 2) was performed by injecting a macro-conidial suspension (Engle et al. 2003) at flowering (61 to 65 anthesis) (Lancashire et al. 1991). The inoculation method involved micro-pipetting 1 mL of suspension onto the central spikelets, which were kept in a polythene bag for 24 h. The relative humidity of the chamber was maintained at around 80 %, with spray irrigation. Each spike was taken as an experimental unit and all spikes (inoculated and non-inoculated controls) were evaluated.
The FHB incidence was categorized by the presence or absence of the disease, and severity was calculated for three weeks (22 days) (Stack & McMullen 2011). The percentage of diseased spikelets was calculated as the number of Fusarium-diseased spikelets over the total number of spikelets, in each spike considered. The percentage of Fusariumdamaged kernels was calculated as the number of damaged kernels over the total number of kernels on each spike. The area under the disease progress curve (AUDPC) was used to combine multiple observations from four data points (8, 15, 22 and 30 days after infection) into a single value (Malbrán et al. 2012, Simko & Piepho 2012. The empty grain percentage and 1,000-grain weight were considered yield components (Velazquez & Formento 2012). The mycotoxin content (deoxynivalenol) was evaluated using immunofluorescence, on a Vertu lateral flow reader (Vicam 2011). For the detection of deoxynivalenol, a sub-sample of 5 g was used, derived from the total sample of blocks per treatment, which were previously homogenized. The experiment was performed in triplicate.
Using multivariate analysis, the infection responses (FHB effects) of the Lin 84 and Caninde 21 genotypes were compared to both resistant and susceptible reference genotypes. The data set was generated from 78 Lin 84 observations and 131 Caninde 21 observations. For quantitative analysis, the FHB severity and AUDPC variables were used. Comparisons were made using non-parametric multivariate analysis of variance NP-Manova (Anderson 2001), with 9,999 permutations and Euclidean distance differences.
The normality tests used the Shapiro-Wilk analysis, and the variance homogeneity was verified  Frontana and Sumai 3, known worldwide for their resistance, demonstrated the lowest FHB incidence rates [respectively 7.2 % and 2.9 % for the first seeding date (May), and respectively 39.8 % and 30.9 % for the second seeding date (June)]. The findings are consistent with those reported by other authors who observed that Frontana is considered a source of resistance to the initial infection progress (Bainotti et al. 2013), and thus slower the infection development (Alves et al. 2013b). In both cases, the FHB infection and its development were highly dependent on the local environmental conditions, which, in this case, were ideal for the second seeding date in June. However, Caninde 11 (moderately susceptible) and Caninde 12 (susceptible) demonstrated higher disease incidences for both seeding dates (Figure 1). In other words, while less than ideal environmental conditions are enough to promote higher FHB infection rates in susceptible germplasm, they are not sufficient to afford observational differences between moderately resistant and resistant genotypes.
Lin 84 and Caninde 21 were the new genotypes included in the study. Lin 84 presented disease incidences of 12.7 % and 33.3 %, respectively for the first and second seeding dates, and was classified as a moderately resistant genotype; while Caninde 21 presented 30.4 % and 71.5 %, respectively for the first and second seeding dates, and was considered a susceptible genotype.
The differences in the FHB incidence in the different genotypes and over the different seeding dates were supported by Bayes Factor tests. Generalized linear modeling, employing logistic regression, was applied to explain and predict the best model for incidence determinations. Taking into account the Akaike information criterion, the "genotype * climate" model explains the incidence for the genotypes in this study, where the incidence varies together with the tested genotype and climatic conditions (Table 3).   Our results also confirmed the lower FHB severity observed in the resistant cultivars Sumai 3 and Frontana (Table 4). The mycotoxin content analysis revealed cultivars below 1 ppm (Table 4) for deoxynivalenol. This low concentration of mycotoxin for these genotypes may be explained by the presence of Qfhs.ndsu-3BS, encoding a glucosyl transferase which is effective only against Fusarium strains that produce deoxynivalenol or structurally similar trichothecenes (Lemmens et al. 2005).
While Sumai 3 has been reported by some researchers as carrying a Type II resistance (Ittu et al. 2005, Niwa et al. 2014, Lahlali et al. 2016, its variable deoxynivalenol levels are attributed to its quantitative trait locus QTL aforementioned, which tends to accumulate different deoxynivalenol levels (Zhou et al. 2002, Dweba et al. 2017, thus confirming the general complexity of the FHB response. Our results revealed that Lin 84 also presents a Type II resistance, as described by Mesterházy (1995), and, in spite of the fact that it presented marginally infection levels higher than the resistant comparatives (Figure 1), it accumulated a lower deoxynivalenol concentration (Table 3).
On the other hand, the national varieties Caninde 11, Caninde 12 and Caninde 21 were all susceptible, revealing FHB severity values between 7.5 % and 12.7 %, and deoxynivalenol concentrations between 2.83 ppm and 4.14 ppm. In spite of the relatively low FHB infection rates under field conditions over the years, in this study, the new cultivar Caninde 21 was found to be moderately susceptible, as verified by the higher deoxynivalenol concentration (Table 3). Caninde 12 was identified as the most susceptible cultivar, presenting the highest severity for FHB and deoxynivalenol concentration.
It is also interesting to observe the progress of the disease over the 30-day period after infection on the seeding dates ( Figure 2). The data shows that, for both the seeding dates, Frontana, Sumai 3 and Lin 84 presented similar median values for AUDPC, being well below those for Caninde 11, Caninde 12 and Caninde 21.
In both the seeding periods, Frontana presented low incidences, thus showing a difficulty for initial infection or Type I resistance (Figure 1). However, once the fungus overcomes the mechanical barrier of the spike, the disease progresses very quickly, as shown by the higher AUDPC values, in comparison to Sumai 3 and Lin 84 (Figure 2).
The disease severity, as evaluated by the incidence and progress of the disease, was low for Sumai 3 (Figures 1 and 2). This is attributed to its Type II resistance (ability to slow the infection progress) reported in other studies (Kubo et al. 2013, Niwa et al. 2014). Our study partially supports the results obtained by Mendes et al. (2018), who reported the lowest AUDPC in the resistant genotypes (BRS, Parrudo and Frontana), as compared to susceptible genotypes ( Figure 2). In this study, the AUDPC for Frontana was lower than the susceptible genotypes for the first seeding date; but, for the second seeding date (ideal conditions for the disease development), its susceptibility was comparable to the already susceptible genotypes.
A significant finding of this study is the identification of Lin84, an advanced breeding line, which did not present a level of FHB severity lower than that for the susceptible comparisons (Table 3), but did present a significantly lower disease development over the period of 30 days post infection ( Figure 2). Significantly, besides being comparable to the resistant genotypes Sumai 3 and Frontana, it also accumulates comparatively lower deoxynivalenol concentrations. Such a relationship was mentioned by Lemmens et al. (2005), who reported an association between disease progress (Type II resistance) and lower deoxynivalenol concentration. Our results confirmed a higher value of AUDPC in all the susceptible varieties (Caninde 11, Caninde 12 and Caninde 21); although the Caninde 11 values were markedly lower than the other two for both seeding dates.
The correlation analysis for the various infection parameters and deoxynivalenol concentration revealed some interesting results (Table 5). The susceptible genotypes (Caninde 11, Caninde 12 and Caninde 21) presented a high positive correlation with the deoxynivalenol concentration for all phenotypic variables (disease severity, AUDPC and Fusariuminfected kernels). The same was not true for the resistant genotypes (Frontana, Sumai 3 and Lin 84), with a positive correlation observed between the deoxynivalenol content and the FHB severity (r = 0.759; p < 0.05), but not for AUDPC or for the percentage of diseased kernels. In the susceptible genotypes, a high positive correlation between the deoxynivalenol concentration and the AUDPC (r = 0.941; p < 0.01) was observed for FHB severity and Fusarium-damaged kernels (r = 0.880; p < 0.05, for both cases) (Table 5).   These results support Spanic et al. (2019), who observed positive correlations between the AUDPC and deoxynivalenol concentration in inoculated wheat grains (r = 0.91; p < 0.01); and also Mesterházy et al. (2005), who observed highly significant correlations among FHB, Fusarium-damaged kernels, yield loss and deoxynivalenol contamination. It should be mentioned that all genotypes under study presented positive and significant correlations between FHB severity and Fusariumdamaged kernels, for resistant (r = 0.935; p < 0.01) and susceptible (r = 1; p < 0.01) genotypes. This was also reported by Hernandez (2010), who observed similar correlations for winter wheat.
The FHB impact on grain yield (by genotype) was assessed by comparing the average number of spikes infected with Fusarium with each respective control group, for the percentage of empty grains and 1,000-grain weight (Figure 3).
In both cases, significant statistical differences were observed, confirming variable degrees of yield loss in the artificially inoculated genotypes. Sumai 3 presented a significant difference for 1,000-grain weight values, as compared to its control plot, while no differences were observed for Frontana and Lin 84 (Figure 3a). On the other hand, Caninde 12 and Caninde 21 presented significant differences between their inoculated spikes and control plots, for empty grain percentages. Similar effects were also observed for the resistant genotype Sumai 3 (Figure 3b). Although these results confirm the impact of FHB on grain yield and its components, the extent of variability present among the genotypes is an indicative of the difficulty in using a single parameter to identify resistant genotypes under development.
An additional consideration is that, under field conditions, genotypes behave differently. Caninde 11 and Caninde 12 demonstrated a better agronomic performance and higher grain yield potential, corroborated by systematic surveys conducted by the Instituto Paraguayo de Tecnologia Agraria (IPTA) (data not shown here). Similar results were reported in   (Bainotti & Donaire 2015). The Bohemia variety, in spite of its lower expression of FHB and relatively higher grain yield, accumulated higher concentrations of deoxynivalenol (Chrpová et al. 2010). Finally, to classify two new wheat lines (Lin 84 and Caninde 21) for their reactions to FHB, NP-Manova was used to compare their infections with the previously known resistant and susceptible genotypes in the study ( Table 6). The analyses revealed that the FHB reaction of Lin 84 was not significantly different from the resistant genotype, while the performance of Caninde 21 was similar to the known susceptible genotype. Thus, they were respectively classified as resistant and susceptible to FHB. CONCLUSIONS 1. It was possible to identify a new breeding line (Lin 84) as a source of Type II resistance against the Fusarium head blight (FHB), in a superiorly adapted agronomic plant type. The high yield potential Caninde 21 was classified as susceptible; 2. The controlled artificial inoculation of newly developed wheat genotypes is an effective tool for classifying the resistance or susceptibility to FHB. Yet, no single parameter, such as disease incidence or severity, may adequately guarantee the results. A combination of various characters, such as disease severity, rate of disease progress over time (area under the disease progress curve), infected or empty grain percentages, and, most importantly, mycotoxin (deoxynivalenol) buildup concentration are essential to categorize genotypes with unknown potential for FHB resistance.

ACKNOWLEDGMENTS
This study was funded by a project of the Consejo Nacional de Ciencia y Tecnología (Conacyt) of Paraguay, through the PROCIENCIA program, with financial resources from the Fondo para la Excelencia de la Educación y la Investigación (FEEI), in collaboration with the CAPECO and CEMIT-UNA. We thank the excellent assistance of interns and students who contributed to this study (Cinthia Rojas and Carolina Toledo) and are especially grateful to the staff of the different institutions involved in the project, as well as the PhD Rosa Oviedo, for their critical comments on the manuscript.