Initiation of winter dormancy and its release, vegetative growth, and development of reproductive organs occur sequentially following plant perception of environmental stimuli like changes in day length and temperature. These stimuli trigger the expressions of genes that regulate the growth of different plant structures such as shoots, stems, flowers, rhizomes, and roots. The identification of the genomic regions associated with seasonal growth changes offers the possibility of tapping the genetic potential of switchgrass to produce higher biomass yield through the extension of the growth period.
Through our observation of most switchgrass growth in the field, it peaks during the long days and warm temperature of the late spring (May-June), starts flowering when sensing the gradual decrease in daylengths in mid-summer (July to early August), and finally undergoes senescence (September -October) and becomes dormant (November to January) when days are short and cold in the winter. Identifying the QTLs associated with these growth changes would enable using them as markers for selecting progenies with delayed dormancy and flowering, and with early spring emergence. In our study, we used FRH and NDVI as indicators of dormancy status, SE for dormancy exit, and FD as an indication for the plant reaching the end of its growth cycle.
It is important to understand the mechanism underlying stand persistence of switchgrass so that the changes made to the growth cycle will not negatively affect its survival. Switchgrass persistence is particularly impacted by C and N translocation from the shoots to crowns, rhizomes, and roots where lower mobilization can result in plant death [29]. This can happen in the case of low tillering and disruption of senescence [29]. Sarath et al. [29] suggested a longer growth period of the southern germplasms compared to the northern germplasms when planted in a northern environment. However the plants do not have a defined period of senescence before winter dormancy, and hence do not fully cycle C and N. As a consequence, the non-adapted germplasm suffers from the loss of structures crucial to perenniality [29]. In southern locations where winter usually starts at a later date and generally less cold than northern locations, planting non-dormant or semi-dormant cultivars can potentially increase biomass yield, as long as the plant can properly translocate nutrients to the belowground storage organs before winter starts. In addition to death caused by poor C and N translocation, switchgrass deaths can also be caused by severe frost, suggesting a non-adaptation of cellular mechanisms to freezing and lack of cold acclimation [29].
We did not phenotype cold tolerance in this study but recognize the importance of this trait for warm-season plant survival under low temperatures. We observed some mortality of the B6 parent that might be a result of intolerance to freezing temperature as it originated from a Florida collection. The B6 parent has been observed to persist and grow very well through winter at the more southern location Tifton, where winter is milder than the location where we carried out this study (Figure 2b). It has also been observed to grow during winter in the greenhouse where other parents (AP13 and VS16) were gone completely dormant, typical of a nondormant genotype (Figure 2a). AP13 is a lowland genotype that is known to have a high biomass yield while VS16 is an upland genotype with lower yield but better tolerance to cold. Both populations exhibited a continuous bell-shaped distribution for all four phenotypes, suggesting they are quantitative traits controlled by many genes. This explains the large number of QTL that were successfully mapped for all traits.
Heritability was calculated to see if traits can be passed on to progenies without being affected by the environmental changes. Although we did not test the progenies in different locations, the different years bring enough environmental differences because of the variability in rainfall, temperature, nutrient status, etc. We also clipped the plants at a later date in 2018, thus the FRH and NDVI values were smaller in 2018 compared to 2017 for both populations. The difference in the year impact was shown by the significant year and genotype by year interaction for the majority of the traits. In our study we found FD to have the highest heritability, followed by FRH, NDVI, and SE.
Correlation of traits with dry biomass yield was calculated to test the hypothesis that an extended growth period leads to more biomass accumulation in these two F1 populations. We have previously found a positive correlation in a switchgrass diversity panel consisting of 17 lowland, 5 intermediate, and 14 upland accessions [32]. In the current study biomass weight was confirmed again to be positively correlated with FRH and NDVI, while negatively correlated with SE, for both populations. The positive correlation between biomass weight and FD was weak for the AB population and non-significant for the BV population. This means biomass weight increases with the increase in growth and plant greenness in the fall, and earlier plant emergence. Later flowering date increases the biomass yield only to a smaller magnitude in AB, which indicates the disadvantage of using the trait for indirect selection of high biomass yield, particularly in the southern region with longer growing season and switchgrass is observed to flower in early summer. We found a high correlation between FRH and NDVI (0.78 r in AB and 0.63 r in BV), intermediate correlations between FRH and SE (-0.38 r in AB and -0.44 r in BV) and between NDVI and SE (-0.40 r in AB and -0.37 r in BV), and low or non-significant correlations between FD and other traits. For highly correlated traits, we think a similar gene pathway or gene action is involved in controlling the expression of the traits.
We have successfully mapped 18 QTLs for FRH, 18 QTLs for NDVI, 21 QTLs for SE, and 30 QTLs for FD. We found a higher number of QTL in the BV population, specifically in the VS16 map. This can be explained by the higher genetic divergence of the parents used for the cross [43]. We found that the BV population contained more polymorphic markers and that the parents have the largest genetic distance. Since the parents were originally adapted to different latitudes they should have more variants contributing to traits segregation in the progenies. In the case of linkage mapping in a pseudo-testcross population (with separate parental genetic maps), the variation within the parental genome is captured through the trait-marker association. On the other hand for the AB population, since both parents originally adapted to southern latitudes, they theoretically have less genetic diversity, thus fewer variants contributing to trait segregation. Another reason might be the high rates of segregation distortion of alleles in the progenies of the AB population [43], which may have resulted in a lower percentage of mappable markers and possible dropout of alleles associated with the traits understudied [44, 45].
We found some QTLs reoccurring in the second year of evaluation (using LS means); the reoccurrence of these QTLs suggests the high heritability of the genes linked to those QTLs. There are 3 common FD QTLs and 1 common FRH QTLs mapped for both years. As both FD (0.62 – 0.88 H2) and FRH (0.55 – 0.64 H2) have high trait heritability, this explains why common QTLs were found for these two traits. The markers located near the QTL positions can potentially be used for progeny screening as they give more confidence in the trait expression.
QTLs colocalization indicates either pleiotropic gene action or different genes that are closely linked. For the latter, the association between different traits can be broken after a few cycles of recombination. For pleiotropic gene action, we can utilize the QTL to simultaneously select for multiple favorable traits such as lower dormancy level, early emergence, and later flowering/maturity. Colocalization of QTLs could also explain why certain traits are highly correlated with each other. We found 8 colocalized regions with FRH and NDVI QTLs, and these two traits were highly correlated with each other (0.78 r in AB and 0.63 r in BV). FD has the lowest number of colocalized regions with other traits, as also shown by its low correlation with other traits.
There are several QTL mapping studies done in switchgrass, these include mapping of QTL for spring green-up, flowering time, developmental traits, and biomass weight [46-53]. Dong et al. [52] conducted QTL mapping for reproductive maturity and found the QTL in LG 1a, 2b, 3a, 3b, 7a, 8b, and 9a. In our study, we also mapped flowering QTL in LG 1N (1a), 3K (3a), 3N (3b), and 7K (7a). Milano et al. [54] found QTL for flowering date in LG 2K, 4K, 5K, 5N, 9K, and 9N. In our study, we also found flowering QTL in LG 4K, 5N, and 9K. Tornqvist et al. [47] did QTL mapping of heading and anthesis dates and mapped the QTL in LG 2K, 2N, 3K, 4K, 4N, 7N, 8N, and 9K. We also mapped the flowering QTL in LG 2N, 3K, 4K, and 9K. Ali et al. [46] performed QTL mapping of spring green-up and days to flower. They identified spring green-up QTL in LG 1K, 1N, 2N, 3K, 3N, 4N, 5K, 6K, 6N, 8K, 8N, 9K, and 9N. Days to flower QTL were mapped in 1K, 1N, 2K, 2N, 3K, 4K, 5K, 5N, 6K, 6N, 7K, 7N, 8N, and 9N. In our study, we also found the spring emergence QTL in LG 1K, 1N, 2N, 5K, 9K, and 9N, and flowering date QTL in LG 1K, 1N, 2N, 3K, 4K, 5N, 6N, and 7K. Poudel et al. [55] found QTL associated with tiller numbers phenotyped during spring emergence after undergoing staged freezing treatment in LG 1K, 5K, 5N, and 9K. We mapped the spring emergence QTL in the same LGs.
Flowering time was the focus of many QTL mapping studies due to its potential in extending the growth period and increase biomass yield. However, our study showed that biomass yield was correlated more with winter dormancy than flowering date, suggesting that winter dormancy QTL can be used to screen for plants with potential high biomass yield to be grown in the southern region. To date, there are no published QTL mapping studies on winter dormancy besides the few transcriptomic studies aiming to identify gene pathways and processes involved during senescence and dormancy [56-58]. Poudel et al. [59] developed genomic selection models to predict southern germplasms’ winter survival in northern regions. They found higher prediction accuracy with better genetic relatedness between the training and validation populations. The study did not phenotype winter dormancy per se but the survival rate of plants after freezing winter seasons which is a function of adequate senescence, cold acclimation, and cold tolerance. Our study is the first report on QTL (and their genomic regions) associated with winter dormancy and the trait implication in the accumulation of more biomass through the extension of the growth period. The markers found within the QTL interval are potential genomic resources that can be used in marker-assisted breeding programs. These markers merit more investigations in future work to validate the status and level of association with the traits in other populations with different genetic backgrounds.