Introduction
As the bearers of grain, the grass panicle (or inflorescence) has been the target of selection for thousands of years (Doust, 2007). There is enormous diversity in panicle architecture within and among grass species (Coen & Nugent, 1994). Panicle architectures are critical determinant of interspecies differences in plants morphology and life history, and are often measured as variation in panicle length, branching (number, length, and pattern), and flower number and size borne on each branch type. Simple panicles may have only primary branches, while complex panicles can possess many secondary and tertiary branches (Bommert & Whipple, 2018; Glemin & Bataillon, 2009). In wild grasses, branching pattern plays an important role in wind pollination and affects the number and size of seeds, which ultimately influences seed yield and plant fitness (Friedman & Harder, 2004). In domesticated species, there is a direct association between panicle architecture and seed productivity (Brown et al., 2006; Crowell et al., 2016; Wang & Li, 2005). Analysis of the phylogenetic distribution of panicle variation in the grasses suggests that different panicle architecture have arisen independently many times, and homoplasy across the grass phylogeny has obscured the mechanisms of panicle diversity (Doust & Kellogg, 2002; Kellogg, 2000). These traits likely evolve in response to natural selection mediated by pollinator (Friedman & Harder, 2004) and environmental variation such as light (Vogler et al., 1999), drought (Mal & Doust, 2005), nutrient availability (Dorken & Barrett, 2004), soil water availability (Caruso, 2006), and intraspecific competition (Wolfe & Mazer, 2005). Given the importance of inflorescence architecture to the fitness and productivity of both wild and domesticated species, it is of great interest to understand genetic variation in panicle architecture.
Organisms often respond to changing resource availability and environmental signals through phenotypically plastic changes (Sultan, 2000). Panicle traits often display plasticity in response to different environmental cues, as panicle development involves complex regulatory mechanisms and these mechanisms interact with environmental signals (Adriani et al., 2016; Bai et al., 2016; Tu et al., 2019). For example, Adriani et al. (2016) found that secondary branch number was the most variable (or plastic) trait of panicle architecture in response to light resources in rice. Secondary branching was also found to be influenced by water deficit more than primary branching in a rice recombinant inbred family (Liu et al., 2010). Abiotic components such as heat, drought, and light affect panicle development and ultimately panicle architecture (Adriani et al., 2016; Wu et al., 2017; Wu et al., 2016; Mitchell et al, 1997). In addition to environmental effects on variation in panicle traits, standing genetic variation within species in panicle traits is also common (Jamal et al., 2009; Brown et al., 2006; Hong & Yan, 2004; Ungerer et al., 2002).
Genetic variation in phenotypic plasticity in response to the environment is better known as genotype-by-environment interactions (G x E) (Des Marias et al., 2013). Quantitative studies of G x E in many crops (e.g., maize, rice) have identified important quantitative trait loci (QTL) impacting many panicle traits (Kovi et al., 2011; Leng et al., 2017; Zhao et al., 2017). For example, Leng et al. (2017) identified 17 QTL for five panicle related traits in a double haploid population in rice. Among these QTL, six QTL showed QTL-by-environment interactions, indicating that panicle related traits are susceptible to environmental influence. Zhao et al. (2017) found that 11 out of 19 QTL were involved in QTL-by-environment interactions for tassel primary branching number in maize under different watering environments. G x E is common in QTL studies and identifying G x E and the pattern of interactions is of great interest to understand the genetic architecture underlying phenotypic traits.
Pleiotropy is the phenomenon of a single gene affecting multiple distinct traits (Williams, 1957) and is an important driver of trait integration and modularity (Armbruster et al., 2014; Klingenberg, 2008). Pleiotropy contributes to the genetic correlation among traits and therefore has broad implications in genetics, development and adaptive evolution (Armbruster et al., 2014; Auge et al. 2019; Pigliucci and Preston 2004;). In plants, the formation of all aboveground organs, such as leaves, tiller, internodes, and inflorescences is mainly dictated by the activity and determinacy of the shoot apical and axillary meristems. During the vegetative-to-reproductive transition, many developmentally related traits originate from the same meristem and complex environmental signals may affect the different type of meristems by similar regulatory networks (Wang & Li, 2008; Xue et al., 2020). Therefore, the loci for vegetative and reproductive development-related traits frequently show pleiotropy. For example, most flowering time pathway genes show pleiotropic effects on tiller number and yield potential in crops (Auge et al. , 2019). Genes involved in hormone pathways frequently affect both vegetative growth and reproductive development (Lee et al., 2019), in part through their developmental impacts on meristems (Azizi et al., 2015). QTL mapping is one of the many approaches that have been used to estimate genome-wide pleiotropy, and fits squarely in the context of developmental pleiotropy (Paaby & Rockman, 2013). Pleiotropic effects identified through overlapping QTL locations have been observed in panicle development in sorghum and rice (Brown et al., 2006; Endo-Higashi & Izawa, 2011; Komatsu et al., 2001; Miura et al., 2010; Yu et al., 2017), but these patterns have not been investigated in native perennial grasses.
Switchgrass (Panicum virgatum L.) has been championed as a potential biofuel crop since it was selected by the US Department of Energy (US DOE) as a model grass species for bioenergy in the early 1990s (Hohenstein & Wright, 1994; McLaughlin, 1993). Its potential for high biomass production on marginal land, adaptation to a wide range of environments, and ecosystem service such as carbon sequestration, water flow management and erosion control, makes switchgrass an excellent candidate for filling bioenergy needs (Mitchell, Vogel, & Uden, 2012; Robertson et al., 2017). Switchgrass is a warm-season C4 perennial grass native to the North America, with a range that extends from the eastern seaboard west to the Rocky Mountains and from southern Canada south to the Texas Coastal Plain and Northern Mexico (Casler, 2007; Hopkins, 1995). Two major distinctive populations have been classified in the past based on morphology and habit preference, northern upland and southern lowland ecotypes (Porter Jr, 1966). A very recent study based on a resequenced switchgrass diversity panel was able to define a third coastal ecotype, which is broadly sympatric with the lowland ecotype but possesses upland leaf characters and lowland plant morphotype (Lovell et al., 2020, in review).
Information on panicle morphology is limited in switchgrass, although panicle length differences have been reported between switchgrass ecotypes and cultivars (Porter Jr, 1966; Price, 2014; Van Esbroeck, 2003). We hypothesize that panicle evolution in switchgrass may be related to selection on aspects of mating system and degree of investment in vegetative versus sexual reproduction, especially in the context of seedling establishment in differing habitats. For example, lowland switchgrass has a restricted bunch grass growth form and occurs primarily in patchy distributions along riparian areas. In contrast, upland switchgrass has a rhizomatous spreading growth form that occurs in many prairie habitats. Pattern of pollen dispersal across patches, or aspects of seed establishment (e.g. seed size/number tradeoffs or disturbance regimes) likely differ in these habitats and may have driven divergence in panicle form. Panicle morphology and its relationship to seed quality may be important targets of selection and breeding, as consistent seed production will be critical to meet the demands for large-scale biofuel production (Das & Taliaferro, 2009; Vogel, 2000).
In this study, we evaluated the genetic architecture of panicle traits across ten field sites in the central US to investigate the importance and nature of G x E for panicle traits. To accomplish this goal, we planted clonal divisions of progeny from a four-way outbred mapping population derived from upland and lowland germplasm, along with the four grandparents and F1 hybrids, at ten field sites spanning a large latitudinal gradient. Three panicle traits including panicle length (PL), primary branching number (PBN) per panicle, and secondary branching number (SBN) on the panicle, were assessed at each site to investigate: (1) the genetic architecture underlying these three traits, (2) the sensitivity of QTL and their effects across different environments (i.e., QTL x E), (3) the extent of pleiotropy between panicle and other traits, and (4) the environmental factors contributing to QTL x E interactions. Finally, we identified candidate genes potentially involved in regulating panicle architecture in switchgrass.