Introduction
Different species in a shared habitat may be exposed to divergent selective agents, but one selective force common to natural populations across the globe is the diel cycling of environmental conditions at an unvarying pace of 24 h as a result of the Earth´s rotation. The prevalence of such selection pressure is evidenced by the evolution of the circadian clock, an endogenous timekeeper, across the tree of life, which allows for varied biological functions from the expression of individual genes to physiology and behavior to be appropriately timed relative to anticipated environmental oscillations (e.g., Yerushalmi and Green 2009). Experimental disruptions of the clock by loss-of-function mutations or by manipulations of the exogenous surroundings often reveal deleterious consequences on fitness components. For example,Drosophila melanogaster had lower fitness in unnatural day lengths than under 24-h cycles (Pittendrigh and Minis 1972); an arrhythmic Synechococcus bacterial strain was outcompeted by the wild type with a functional clock in cycling conditions (Woelfle et al. 2004), and in the common sunflower (Helianthus annuus ) an interruption of natural solar tracking reduced biomass and the number of pollinator visits (Atamian et al. 2016).
Circadian traits are measured in constant conditions where the clock can ”free-run” and where endogenous cycles often deviate from 24-h rhythms maintained under naturally cycling settings. The free-running length of one full cycle, the circadian period, is a key trait that is expected to be closely related to fitness: the circadian resonance theory postulates that peak fitness is achieved when the length of the endogenous circadian period matches that of the exogenous environmental cycle (Pittendrigh and Minis 1972). Consequently, 24-h circadian periods would be expected to be advantageous in natural environments, whereas deviations from 24 h in either direction would improve fitness in unnatural diel cycles shorter or longer than 24 h. These tradeoffs might arise from physiological costs associated with constantly re-entraining the clock to an environmental cycle that is out of sync with the endogenous circadian rhythm (Pittendrigh and Minis 1972). Evidence for this hypothesis has been reported in studies on experimental mutant genotypes of the plant model Arabidopsis thaliana and cyanobacteria that express marked variation in circadian period (Dodd et al. 2005, Ouyang et al. 1998, Woelfle et al. 2004; Rubin et al. 2017), but results from resonance experiments are not always fully compatible with the classical theory (e.g., Graf et al. 2011, Horn et al. 2019, Woodley of Menie et al. 2019).
While experimental genotypes with induced mutations are useful in dissecting the explicit molecular components of varied biological functions, they are not representative of patterns of genetic diversity that have arisen in natural environments. In the wild, various evolutionary processes determine the partitioning of genetic variation in quantitative traits among vs. within populations (Mackay et al. 2009, Mitchell-Olds et al. 2007), and populations may become genetically differentiated from each other as a consequence of stochastic events such as random genetic drift (Palumbi 2003, Slatkin 1987) or systematic spatially varying natural selection (Loveless and Hamrick 1984, Siol et al. 2010). Populations may also exhibit genetic isolation by distance, in which physically proximate populations are more similar than distant ones, if populations originate by chance from a limited number of genetically related founders or migration is restricted (McRae and Beier 2007, Slatkin 1987, Wright 1943).
Intraspecific genetic variation that tracks spatial heterogeneity in the environment characterizes many plant traits related to fitness, including timing of growth and reproduction (e.g., Méndez-Vigo et al. 2011) which have overlapping molecular genetic underpinnings with the circadian clock (Brachi et al. 2010). For example, flowering time is highly variable among natural genotypes of A . thalianasampled across the Iberian Peninsula, with lower-elevation origins generally flowering earlier in common garden (Vidigal et al. 2017). Circadian traits are no different from other quantitative traits when it comes to substantial naturally occurring genetic variation: Michael et al. (2003) documented a range of 6.5 h in circadian period in leaf movement among 150 A . thaliana genotypes sampled across the Northern Hemisphere and a latitudinal gradient spanning ~50°, while Rees et al. (2021) restricted their sampling to Sweden and described a range of 4.4 h in delayed fluorescence among 191 genotypes sampled between 55° and 63° N. Since diel cycles are uniformly 24 h in length across the globe, variation in circadian period has predominantly been examined in relation to latitude that correlates closely with photoperiod and its variability over the course of a year (e.g., Hut et al. 2013). Yet, latitude alone explains only a minor proportion (less than 8 %) of the observed genetic variation in circadian period in A . thaliana (Michael et al. 2003, Rees et al. 2021), and thus the environmental agents that help maintain variation in the clock remain to a large extent unidentified (Salmela and Weinig 2019). It is noteworthy that animals also exhibit significant natural variability in circadian period which in locomotor activity of insects varies by up to 8 h among mostly European populations ofPyrrhocoris apterus (Pivarciova et al. 2016) and inTribolium castaneum in Japan (Abe et al. 2021), with no evidence for marked linear latitudinal trends.
Beside studies on diverse pools of A . thaliana accessions, plant studies sampling multiple genotypes per geographic location (i.e., populations) have demonstrated that genetic variation in the clock concurrently segregates among and within populations (Greenham et al. 2017, Leinonen et al. 2020, Salmela et al. 2016) in a manner that is similar for instance to flowering time in natural populations ofA . thaliana on the Iberian Peninsula (Méndez-Vigo et al. 2013). For instance, in an annual Mimulus guttatus population from southern Oregon, maternal seed families exhibited a genotypic range of almost 4 h in circadian period (Greenham et al. 2017). Further, in a population of B . stricta from southeastern Wyoming, circadian period varied by 3.5 h among maternal families sampled within a few hundred meters, with evidence for a positive correlation between period length and first-year growth (Salmela et al. 2016). Together, these patterns point to the role of fine-grained regional environmental heterogeneity in shaping genetic variation in the circadian clock.
The contribution of environmental heterogeneity to the genetic diversity of circadian rhythms can be uncovered by sampling multiple populations across well-defined spatial gradients. Temperature conditions vary for instance with elevation such that growing seasons begin later and at longer day lengths at higher elevations, giving rise to genetic clines in quantitative traits even within relatively narrow spatial scales (e.g., Leinonen et al. 2020). Here, we sample plant populations intensively within a limited latitudinal but an 800-m elevational gradient in order to determine how quantitative genetic variation in circadian period is partitioned among and within Rocky Mountain populations of B . stricta , a short-lived and predominantly self-fertilizing perennial relative of A . thaliana with a wide North American distribution (Song et al. 2006). With the utilization of circadian assays of leaf movement across these populations and on over 3800 plants, we examine whether different evolutionary factors including stochastic and selective forces are associated with the spatial trends of variation in circadian period within a region that is small compared to previous studies on natural genotypes (e.g., Greenham et al. 2017, Michael et al. 2003, Rees et al. 2021) but that nonetheless comprises pronounced environmental heterogeneity. Based on a preceding study that found significant among- and within-population genetic diversity in circadian period in the same region but with only four populations and a smaller elevational gradient (Salmela et al. 2016), we hypothesize that (1) circadian period will exhibit genetic variation among and within the 30 populations,(2) that the range of variation among population will be larger than in Salmela et al. (2016) due to a greater range of environments sampled, and that (3) if variation in the clock contributes to adaptation to environmental heterogeneity, patterns of genetic variation among and within populations will correlate with elevation and associated local environmental variables.