Discussion
The Teff lines evaluated in this study exhibited a wide phenotypic variation (Supp. Fig.7a), which is comparable to previous literature reports (Assefa et al., 2002, Girma et al., 2019, Nigu el al., 2016). For example, the GY range in the current field experiment was equivalent to 0.24-0.6 t/ha, where the national average farmer’s yield is around 1 t/ha, and 2.5 t/ha under experimental conditions in Ethiopia (Berhe et al., 2011, Girma et al., 2019, Nigu et al., 2016, Zhu, 2018). Teff has a potential for yielding 4.6-5 t/ha if lodging can to be resolved (Hailu et. al., 2000). Harvest Index values previously reported (Assfa et al., 2002) were of a similar range to our field experiment (Table 3). In addition, phenological values (Table. 6) were in accordance with previously reported ranges for Teff cultivation (Assfa 2001 b & 2002).
Within the genetic material tested, no correlation was found between TDM and GY, whereas in other reported positive (Braha et al , 2017) or negative correlation (Chanyalew et al., 2010, Lule & Mengistu, 2014). However, when analyzing the white and brown separately (and excluding the white cultivar which appears to be much different) there appears to be some degree of correlation (r=0.6 and 0.7 for the white and black respectively), yet not statistically significant. We found correlation between GY and HI (0.8***) that was in agreement with previous reports (Lule & Mengistu, 2014).
Plant height was previously reported to range between 74 and 116 cm (Assfa, 2002). Under our field conditions, within the growing season MPH was 58-78 cm (Table. 8), and in the greenhouse off-season MPH was 180-200 cm (Table 1), which was more than double field growth but with narrower range across the lines. Differences in day length and other environmental factors may account for these differnces. Some of the lines were ranked similarly in both experiments (Table 1 and 6) in terms of MPH.
Since there are two duplicate genes for grain color in Teff, which are known to be dominant (Berhe et al., 2001), the small fraction of the brown seeds found within the seeds propagated from the initial collected panicle (0.064, 0.105 and 0.015 for 53-1, 53-2, 53-3 respectively, Table 1, Fig. 6) can only be explained by an external foreign pollination. The increase in brown seed ratio in 53-1 from the first collected generation to the greenhouse next-generation, from 0.021 to 0.064 - as would be expected from the segregation of o heterozygosity of the grain color loci (A/a). The small fraction of A/a in the collected panicle would be expected to triple (a total of: 1 A/A and 2 A/a) over the course of a single generation. The data also support the hypothesis that the brown lines are half-siblings (hybrids) to their white counterparts; these half-siblings share the maternal side but differ in the paternal one. It is very likely that these hybridizations were most probably wind-driven. As oppose to 53-1 No dark brown seeds could be detected within the grain of the collected panicles of these two lines, so there must have been undetectable light brown seeds that were in a heterozygous state A/a and were later segregated.
The brown lines in this study (pure and segregating) exhibited an overall advantage over the white lines in terms of directing their biomass towards grain production (Table 3, PCA Fig. 5, excluding the case of the ‘white commercial cultivar’). The white commercial cultivar, the pure brown 44A-163-B and some of the segregating brown lines were the highest ranked for GY (Table 3, PCA Fig. 5).
A clear pattern emerges that all brown hybrids being earlier to flower (along with the white commercial cultivar, Table. 6) as compare to their white half-siblings. Therefore, it is possible that the pollen donor/s was/were a relatively early flowering type . Also the significant differences in plant height, observed between 53-2-W and 53-2-B (76 cm vs. 58 cm, Table 7) may indicate that the pollen donor in this case has a shorter stature than the maternal line.
53-3-B, which originated from the segregation of the mixed color line 53-3 into brown and white seeds (Table 1), was especially interesting. This line exhibited relatively low TDM and high HI (Table 3), low EGC (Table 4), and high Chl levels (Table 5). Contrastingly, this line was also relatively tall as indicated by its high MPH (Table 7) and thin stems (Table 4). In terms of sensory evaluation, 53-3-B was the most promising line with its preferable taste, smell and appearance (Tables 9 and 10).
Our hypothesis is that each of the three half-siblings originated from a different pollen donor was strengthened by the large variations in Chl levels between the three half-siblings. 53-2-B had the lowest Chl levels, and 53-1-B and 53-3-B exhibited the highest levels among the collection (Table 6 and Fig. 3). Interestingly, 53-2 which presented the lowest Chl levels had the highest GY among the three. Another result was the grouping of the traits: Chl Tot, Chlb, flavor and AGW that was negatively correlated with TDM (Fig. 5). This is in agreement with the literature that suggests that a smaller plant may contain denser leaves and more chloroplast and Chl per gr FW of leaf tissue (Fritschi & Ray., 2007). The positive correlation obtained between flavor and Chl may be indirect however, these correlations may have importance for future breeding programs as initial phenotypes for selection.
Following the genetic model of lemma color (Berhe et al., 2001), it appears that both the maternal line of 53-2 and 53-3 had the basic p1 and/or p2 genes in the background of the dominant C gene thus resulting in red lemma color, and that the hybridization with an unknown brown donor introduced a P1 or P2 thus resulting in purple lemma color. There seems to be differences that may have come from maternal differences between 53-2 and 53-3 in lemma color, because 53-3-W was not red as 53-2-W, but rather gray (Fig. 3).
In this work we characterize in detail the lodging phenomena in the studied Teff lines. The simplified scoring system for lodging (Caldicott & Nuttall, 1979), which is commonly used, does not take into account at which growth stage lodging starts nor the uniformity of lodging within a plot. Therefore we choose to document plot-height over the course of the reproductive period (Fig. 2), as well as to calculate the days to 75% lodging in a plot. This detailed inspection and documentation explains mechanisms related to Teff lodging (Fig. 2 & Supp. Fig. 7-8), which are very much context-dependent in terms of environmental conditions. In that respect, the lodging resistance which was the reason for collecting these lines to begin with, (Supp. Fig. 1). It appears that since the collected plants growing randomly as single plants within a homogeneous lodging inclined genetic population as well as occurring at very low-density planted areas, showed a lodging-resistant phenotype. In well-irrigated, well fertilized and low-stress conditions of our experiments, completely lodging-resistant line was not found, and lodging seems to be a flash-mob phenomenon were several plants start a lodging movement that sweeps the reset of the field. However, we show that our in-depth documentation and interpretation of lodging here may be useful for future breeding of -lodging resistance characteristics in Teff.
High yielding lines tend to lodge at harvest time (Yu et al., 2007, Davison & Laca, 2010). The lack of variation in lodging resistance may be a result of unfavorable associations of lodging resistance with productivity promoting traits such as plant height, panicle length, grain and shoot biomass (Kebebew et al., 2011, Nigus Eet al., 2016, Yu et al., 2007). Improvement of lodging related traits, such as culm length, overall-height and diameter of the culm internodes, through breeding is expected to be a demanding task due to their relatively low heritability and lack of reliable genetic advance-estimates (Assefa et al., 2001a). Therefore, increasing our understanding of the lodging phenomena and its phenotyping can improve our ability to breed for high yielding lodging-resistant cultivars. This study shows that the nature of lodging is variable in terms of timing and strength (Fig. S7a). In terms of timing, we observed an early type of lodging which was most likely triggered by the fast inflorescence weight increase exhibited by the white commercial cultivar (Fig. 3 and Fig. S7b) and 53-1-B (Fig. 3). Other lines were ‘strong’ enough to carry the inflorescence most of the grain-filling period, such as 44A-163-B and 44B-163-W (Fig S7b). Therefore, the rate in which panicle increases in weight, prior and throughout grain filling appears to vary and may be important from a breeding perspective.
Surprisingly, the white commercial cultivar which was the first to lodge, was mostly stable in plot-height once lodged (around 35 cm above ground surface) during the entire grain filling and was the best yielding line. In the white commercial cultivar, despite stem weakness and the plant being bent towards the ground, the panicles are mostly above ground level. This pattern creates a medium level of lodging that appears to be different from the strong lodging where the plant is heavier and is totally bent to the ground (Fig S7b: 44B-163-W and 53-2-2-W).
Despite the large number of studies screening Teff lines (Assefa et al., 2003, Zeid et al., 2012, Girma et al., 2019, Nigus et al., 2016) there have been only few which include in-depth characterization of different lines. In addition, there are hardly any studies that present agronomic as well as sensory traits side by side to analyze possible links between them . The processing, eating, and nutritional quality of food products may be greatly influenced by Teff variety. Therefore, it is necessary to assess the genetic diversity of this crop for potential improvements of agronomic as well as edible traits. Newly breed varieties must be subjected to sensory analysis for consumer acceptance to make the research efforts commercially meaningful (Zhu, 2018). We report a significant genotypic effect on most of the sensory traits evaluated (Tables 9-10). The effectiveness of flour grinding was found to be crucial for texture across of all market samples. The white cultivars were ranked higher than the rest of the lines. The size of the grain may also effect Injera acidity and odor, as AGW was found to be positively correlated with odor intensity and negatively with acidity.
The current study growing conditions (pots and field experiments) were not favorable to observe the lodging resistance which was observed during original seed collection. The experiment didn’t replicate the specific environmental context of a single seed developing in low density planted area and/or surrounded by a homogenous population of the plots which it was collected from. A wide genotypic variance was found in the current study for stem width and plant height. However, under an abundance of water and nutrient and, at high plant density, a thick stem does not ensure lodging resistance. It appears however that under low density and/or some sort of environmental stress may lead to increased stem lignification and hardening allowing the plant to carry the grain load at an erect posture. Future experimentation to test this hypothesis may include of combinations of agro-technics implementations such as: seed-coverage to enlarge seeds, using a mixture of Teff lines, reducing sowing density to reduce plant density, and the introduction of controlled stresses such as water deficiency and salinity. It is clear that late maturing, thick stem and tall Teff varieties possess deeper root systems than early maturing lines of shorter height (Ayele et al., 2001). Therefore, a combination that include lines that are characterized by thick stems with stress can improve lodging resistance. Integration between a wide range of parameters and the correlations obtained between agronomic and sensory traits may improve our ability to breed towards a “real world” better end-product.