2.2 Leaf Movement Assays
To determine circadian period in natural populations of B .stricta , we adapted the protocol from Tracking Rhythms in Plants (TRiP, Greenham et al. 2015). For each of the 30 populations, 10 to 20 family lines were selected for circadian analysis, and within each family, 18 replicates were included. We planted seeds into Redi-Earth potting mix (Sungro, Agawam, MA, USA) in 1-cm diameter pots and placed the pots into growth chambers (Percival Scientific, Perry, IN, USA) for germination using a 12-h/12-h light/dark cycle and 21 °C/18 °C day/night temperature cycles. After six days, seedlings were moved to 14-h/8-h light/dark cycle for entrainment; temperature conditions during entrainment simulated early May conditions in Laramie, WY with hourly cycling for temperature from 4 °C to 17.5 °C (Table S1). After seven days in the entrainment conditions, seedlings were transferred to imaging stands and placed in constant light and temperature (17.5 °C) conditions for 24 h prior to imaging. Images were taken every 15 min for five days to ascertain leaf positions, and stacked images of leaf movements were analyzed in MatLab (The MathWorks, Inc.) through the TRiP pipeline to determine circadian period (i.e., the duration of one endogenous cycle). These germination and entrainment windows were selected to allow sufficient time for plant growth and for cotyledons to be imaged, but before the development of the first true leaves, which can obscure cotyledon imaging.
We estimated period values through BioDare2 (Zielinski et al. 2014). Values for period were determined by two algorithmic methods: MFourFit (a curve-fitting method) and MESA (Maximum Entropy Spectral Analysis; an autoregressive model based on stochastic modelling). Settings for both analyses included removing estimates of period lower than 18 h and higher than 30 h. Traces from experimental replicates that displayed no rhythmicity of pattern were discarded. For each individual replicate measurement, we compared the difference between the two analyses and values that differed by more than 10 % (~2 h) were removed. Statistical outliers for each family line within the populations were removed. Mean values for each family line and subsequently each population were then determined, as well as the within-population range of values (the range from shortest to longest circadian period family within each population).