Wheat (Triticum aestivum
L.), one of the most important calorie sources available to humankind, provides approximately 19% of daily caloric needs worldwide [1
]. Furthermore, according to the Foreign Agricultural Service (FAS) a division of the United States Department of Agriculture (USDA), global consumption of wheat was over 730,000,000 metric tons in the 2018/19 marketing year [2
]. The demand for wheat and other staple foods is expected to increase as the global population continues to rise. For example, wheat demand is projected to increase by 60% by the year 2050 in developing countries [1
]. In addition to increasing production in a sustainable way to meet the growing demand for wheat, preventing contamination and other grain quality and food safety issues caused by diseases are among the most daunting challenges facing agricultural researchers today.
Head scab or Fusarium head blight (FHB) of wheat, caused by Fusarium graminearum
, is an example of such a disease. It results in damaged discolored grain contaminated with deoxynivalenol (DON), often referred to as vomitoxin, due to the acute and chronic disease symptoms that develop after consumption of diseased grain. DON toxicity symptoms include nausea, vomiting, diarrhea, abdominal pain, headache, dizziness, fever, and with enough exposure, death [3
]. As a result, health organizations around the world have placed stringent regulations on DON levels in wheat products for both direct human consumption and feedstock sources for animals [4
]. In the US, according to Aakre et al., 2005, the current FDA advisory levels for DON are: 1 ppm in finished wheat products for direct human consumption such as flour, 10–30 ppm in feedstock for consumption by ruminating animals, 10 ppm for chicken feedstock with an additional recommendation that it makes up no more than 50% of the diet, 5 ppm in swine feed and no more than 20% of the diet, and 5 ppm in all other animal feed produced from wheat grain or grain by-products and no more than 40% of the animal’s total diet [5
]. In addition to DON contamination, yield reduction and quality diminution are other consequences of severe FHB infection [6
]. Diseased kernels are less dense than healthy kernels and directly reduce yield. Furthermore, severely damaged kernels are expelled along with chaff during combine harvest resulting in additional yield loss [6
]. Damaged kernels also decrease market value due to reduced test weight and flour yield [9
]. Therefore, limiting DON accumulation and reducing the proportion of fusarium damaged kernels (FDK) via genetic resistance are major goals of wheat and other small grain breeders.
DON concentration can be quantified and used as a basis for selection, but testing each genotype in several genetically variable populations is expensive and time-consuming. Thus, plant breeders focused on enhancing FHB resistance in early generation material have traditionally relied on their eyes (extensive in-season field phenotyping-based decisions). Small grain breeders often utilize visual selection schemes such as the pedigree method, the bulk method, and mass selection, to generate improved plant cultivars. Mass selection, an antediluvian, but still extensively used breeding scheme, has been employed in breeding for enhanced disease resistance for many years. Rex Bernardo formally defines mass selection, in Essentials of Plant Breeding
, as the “selection of several to a large number of plants on the basis of individual-plant performance” [10
]. Furthermore, mass selection is a simple method that exposes a segregating population to different environmental conditions and utilizes intense natural, and/or artificial selection pressure to alter the genotypic frequencies of a population [11
]. Deliberately infecting a field with F. graminearum
and only harvesting plants that show no signs or symptoms of FHB would be an extreme example of mass selection. Therefore, in theory, mass selection relies on the assumption that only progeny from superior individuals will be present in the next generation. In addition, it is important to note that mass selection is most effective when dealing with a highly heritable trait, improving more complicated traits like yield are not typically goals for the mass selection method [10
Unfortunately, FHB resistance in wheat is a complex quantitative trait, strongly influenced by the environment, with low to moderate heritability [12
]. Consequently, phenotypic mass selection for enhanced head scab resistance done visually is subjective, and effectiveness can vary widely depending on the experience level of the person recording the ratings, disease pressure, natural lighting, maturity level of the plant, and other environmental factors. This is not to say that wheat breeders have been unsuccessful at using phenotypic selection to develop cultivars more resistant to head scab. Wheat breeders have identified and incorporated single QTL, like Fhb1
a major effect gene discovered in the Chinese spring wheat Sumai 3 [8
], via traditional methods into breeding material. However, resistance to FHB employs both-well studied major genes and other minor genes. Although traditional breeding methods that center around known individual QTL have proven useful in breeding for increased FHB resistance in wheat, these methods can be improved to provide more locally adapted wheat cultivars. Therefore, a scheme that eliminates the inherent subjectivity of the visual selection method and increases the frequency of minor effect QTL is necessary.
Not only is FHB resistance a difficult trait to visually phenotype and select for, but a changing climate will also further complicate things by expediting the need for new cultivars more resistant and adapted to a wider array of environments than those currently available [14
]. For example, a European study aimed at predicting wheat phenology and DON indicated that climate change will cause plants to flower and reach maturity 1 to 2 weeks earlier, and DON concentrations will increase up to 3x
what is currently considered typical in the regions where the study was conducted [15
]. In addition, Backhouse et al. using a modeling approach, found a positive correlation between climate and pathogenic Fusarium species and predicted that future conditions will be conducive for FHB epidemics in regions such as Mexico, North Africa, and Western Siberia where high DON and FDK are not currently a problem [16
]. Furthermore, the genetic composition of current varieties provides partial resistance to FHB. However, environmental conditions largely influence whether or not resistance genes will be expressed [17
]. Therefore, a higher-throughput less subjective selection method is needed to gradually accumulate new regional smaller effect resistance QTL, and more rapidly enhance head scab resistance in wheat and other small grains [18
Optically separating diseased from non-diseased grain has potential as an en masse selection method to identify and enhance resistance in genetically variable populations. In addition, small breeding programs can use the proportion of damaged kernels obtained by optically sorting grain (FDKos) as an additional consideration during germplasm development (selecting parent breeding material with enhanced head scab resistance). The objective of this study was to determine if optically sorting seed from breeding material segregating visually for scab resistance over several generations could be used to generate lines with enhanced FHB resistance (lower DON and FDK values). Three-hundred F4 derived breeding lines from five unique 3-way crosses were grown in an inoculated disease nursery, harvested by hand, threshed, sorted, and the accepted (non-scabby) seed used to plant the subsequent filial generation over several years in Lexington, KY.
One of the most important challenges of this century is to increase crop yields, in order to maintain access to essential food resources like wheat, under a changing climate [14
]. Devastating plant diseases like FHB are major hurdles in the way of solving issues concerning food security, safety, and quality. Yield reduction, low test weight, reduced percentage of high and low molecular weight glutenins and mycotoxin contamination associated with FHB infection threaten food security, diminish food safety and reduce quality [6
]. Furthermore, the consequences of FHB infection will likely be exacerbated by ongoing climate change [15
]. This coupled with an ever-increasing global population has expedited the need for new germplasm more resistant to head scab. On the basis of the results of this study, it is our opinion that an optical seed sorter has potential to reduce DON accumulation and kernel damage. In this study, we tested the potential of an optical seed sorter as an en masse selection method to identify and enhance disease resistance, by exposing genetically variable breeding material to a head scab epidemic, optically separating diseased from non-diseased grain, and planting only non-diseased grain the subsequent filial generation.
We observed that within line selection with the optical sorter only (no among line selection) was not effective at lowering average DON accumulation in ppm, our primary target trait (Table 2
). These results are not unexpected, considering the selection material was in the F4:5
generation prior to any within line selection with the optical sorter. Genetic variation, necessary for progress in plant breeding, is less within an F4
derived breeding line than, for example, an F2
]. Had these breeding lines been F2
populations, or even F2
derived breeding lines more genetic, variation would have been available and greater within line selection progress (reductions in average DON accumulation each cycle) may have been observed. Indirect within line selection for lower DON with the optical sorter is not supported by the broad sense heritability values of DON (0.91 Table 6
, 0.74 Table 4
) and FDKos (0.94 Table 6
, 0.68 Table 4
). Falconer showed that for indirect selection to be superior to direct selection, the genetic correlation must be high and heritability of the trait to be selected must exceed that of the other trait [33
]. Whenever phenotypic correlation is substituted for genotypic correlation, Q, the ratio of indirect to direct selection expressed as the product of the phenotypic correlation coefficient and the ratio of the square roots of the heritabilities can be calculated (e.g., Q = rp
]). The phenotypic correlations (DON to FDKos) were 0.55 (2017–2018) and 0.31 (2019). Using data gathered during within line selection Q = 0.55 × (0.82/0.86) = 0.52, and, Q = 0.31 × (0.97/0.95) = 0.32 using data gathered during the final evaluation. Therefore, Q does not support indirect selection for lower DON with the optical sorter (FDKos).
Interestingly, whenever within line selection was coupled with among line selection on FDKos, consistent reductions in DON and kernel damage traits (FDKos and FDKvs) were observed (Table 2
). Therefore, in practice, indirect among line selection using FDKos values obtained with the optical sorter successfully reduced DON accumulation each cycle of selection. Indirect selection (C3
DON = 0.7, C3
DON as % KY02C-3005-25 = 92) did not result in final DON concentrations below that of direct selection (C3
DON = 0.4, C3
DON as % KY02C-3005-25 = 53) using DON measurement (Table 2
and Table 3
). This result agrees with our estimates of Q, which indicated that direct selection would be more successful than indirect selection on FDKos for lower DON. However, indirect selection using FDKos did consistently reduce DON each cycle of selection (C1
= 1.1, C2
= 0.9, C3
= 0.7), whereas direct selection did not consistently reduce DON each cycle of selection (C1
= 0.6, C2
= 0.7, C3
= 0.4). In addition, indirect selection for lower DON using FDKos as the basis of selection resulted in average DON levels below that of the resistant check and consistent reductions in kernel damage traits (FDKos and FDKvs). Direct selection using DON concentration actually resulted in increases in final average FDKvs and FDKos. Although indirect among line selection with the optical sorter did not result in final DON levels lower than direct selection, final DON levels obtained via indirect among line selection using FDKos were better than that of the resistant check. In addition, kernel damage traits were significantly lower than in previous generations. These findings indicate that the optical sorter is useful for identifying lines with lower DON values while at the same time selecting for lower kernel damage, which are both important traits for ensuring secure access to safe food. The reductions achieved in all three traits using the indirect selection on FDKos were not achieved with direct selection for a lower DON. Furthermore, although indirect within line selection is not supported by estimates of Q obtained in this experiment, performing indirect selection in early generation populations (e.g., F2
) for lower DON with the optical sorter is supported by the results of among line selection on FDKos.
FDKvs another measure of kernel damage due to FHB infection in wheat was also used as the basis of indirect among line selection for lower DON. Indirect selection for lower DON using FDKvs determined with the vacuum sorter was not supported by the broad sense heritability values of DON (0.91 Table 6
, 0.74 Table 4
) and FDKvs (0.80 Table 6
, 0.83 Table 4
). The phenotypic correlations (DON to FDKvs) were 0.41 (2016–2018) and 0.41 (2019). Therefore, Q = 0.41 × (0.91/0.86) = 0.45 and Q = 0.41 × (0.89/0.95), both estimates are < 1 and do not support indirect selection (using FDKvs as the basis of selection) over direct selection for lower DON accumulation. Our data support this notion (Table 2
). Neither indirect selection on FDKvs (C1
= 0.6, C2
= 1.0, C3
= 0.7) or direct selection (C1
= 0.6, C2
= 0.7, C3
= 0.4) consistently reduced DON each cycle of selection. Both selection schemes resulted in average C3
DON levels < 100% of KY02C-3005-25, but direct selection for lower DON (53% of KY02C-3005-25) outperformed indirect selection on FDKvs (92% of KY02C-3005-25). Interestingly, similar to what was observed with direct selection for lower DON, indirect selection on FDKvs resulted in increases in kernel damage traits. Indirect selection for lower DON on FDKvs actually increased FDKvs each cycle of selection, indicating that direct selection for lower FDKvs values does not work. FDKos increased during indirect selection on FDKvs for lower DON from C1
but then decreased from C2
. These results indicate that single trait direct and indirect selection for lower DON is achievable using DON concentrations, FDKvs and FDKos estimates as the basis of selection, however, single trait indirect selection on FDKos lowers DON, FDKvs, and FDKos simultaneously. The other single trait selection procedures (direct on DON and indirect on FDKvs) do not achieve consistent reductions in both kernel damage traits.
In addition to direct and indirect selection on single traits, an index was created and used to base selection. Both direct selection and indirect selection on FDKos for lower DON achieved promising reductions in DON concentrations with each additional cycle of selection. However, each of the two single trait selection procedures (direct on DON and indirect on FDKos) were better than the other at lowering specific traits. Much lower average C3
DON levels were achieved with direct (0.4 ppm, 53% of KY02C-3005-25) than with indirect selection on FDKos (0.7 ppm, 92% of KY02C-3005-25). Kernel damage traits (FDKvs and FDKos) were lowered when using indirect among line selection on FDKos, whereas kernel damage increased with direct selection on DON. Therefore, we thought it logical to combine the two traits (FDKos and DON) in an index to attempt to lower DON at a level comparable to direct selection on DON concentrations while simultaneously lowering kernel damage traits to the degree observed with indirect selection on FDKos. Index selection was incredibly successful at lowering all three traits (DON, FDKvs, and FDKos) simultaneously (Table 2
and Table 3
). Index selection achieved final FDKvs (3.3%, 132% of KY02C-3005-25) values lower than all other selection scenarios we explored. Final average C3
FDKos values (13.5%, 142% of KY02C-3005-25) achieved with the index were slightly lower than values obtained with direct selection for low FDKos (13.7%, 144% of KY02C-3005-25). In addition, index selection resulted in final average DON concentrations of 0.4 ppm, the same as what was observed for direct DON measurement.
These findings suggest that optically sorting grain is an effective breeding strategy for lowering final DON accumulation and limiting kernel damage associated with head scab infection. Average HT increased with additional cycles of sorter selection. This indicates that optically sorting grain to enhance head scab resistance may be most appropriate during pre-breeding (germplasm improvement) for FHB resistance, not improved agronomic characteristics. At a minimum, FDKos values obtained by optically sorting grain can be used as an additional criterion during germplasm development and when selecting parental breeding material with enhanced head scab resistance. In addition, FDKos estimates can be utilized in conjunction with direct DON measurement to form a very successful selection index that effectively lowers DON accumulation and associated kernel damage traits over generations. The optical sorter selection method described herein should provide small breeding programs focused on delivering FHB resistant germplasm with a useful tool to identify and select head scab resistance in wheat and other small grains.