Climate change is intensifying droughts across the globe, challenging species to adapt to novel conditions. While plant physiological and phenological responses to drought are well-documented, less is known about how water scarcity affects the evolution of selfing across species ranges. According to the selfing syndrome hypothesis, in environments where selfing confers a fitness advantage, selection should favour floral traits associated with increased selfing relative to outcrossing. We used a field experiment near the northern range edge of the scarlet monkeyflower (Mimulus cardinalis) to test this hypothesis both spatially (among leading-edge, central, and trailing-edge populations) and temporally (between cohorts separated by a period of historic drought). Although populations from different range positions showed genetic differentiation in some floral traits, these differences did not consistently support predictions of the selfing syndrome hypothesis. Contrary to the predictions of reduced investment in floral rewards and increased selfing ability at range edges, the sugar content of nectar was greater and autogamous seed set was smaller in leading-edge than central populations, herkogamy tended to be greater in trailing-edge populations relative to leading-edge and central ones, and nectar volume did not vary predictably among regions. There was no support for the evolution of selfing syndrome from the predrought ancestors to the postdrought descendants. Instead, in leading-edge populations, descendants evolved greater sugar content relative to ancestors, and there were no other differences between ancestors and descendants in any other trait or region. Overall, these findings suggest that mating system evolution in M. cardinalis likely reflects a complex interplay of regional factors, including range position, historical adaptation, and local environmental variability, rather than simple stress-induced shifts towards selfing.

This study takes advantage of a common garden that is part of a larger resurrection experiment focused on quantifying quantitative genetic parameters and evolutionary responses to a historic period of drought across the range of M. cardinalis (Figure 1; Diffenbaugh et al. 2015). For this larger study, seeds were collected from six populations of M. cardinalis (two each from its northern, central, and southern range; Figure 1) in 2010 (“pre-drought” ancestors) and 2017 (“post-drought” descendants), before and after a period of historic drought, respectively (Sheth et al. 2025; Sheth & Angert 2016; Vtipil & Sheth 2020; Wooliver et al. 2020). Locality information for each population is reported in Sheth & Angert (2016). To control for maternal and seed storage effects, plants were crossed for one generation following a nested paternal half-sibling design to allow for the estimation of quantitative genetic parameters (Sheth et al. 2025; Wooliver et al. 2020). After five weeks of greenhouse growth, seedlings were transplanted into three common gardens across the range. Due to logistical reasons that prevented the collection of floral trait data in the central and southern gardens, this study focuses on data from the northernmost garden in Eugene, Oregon (Friends of Buford Park and Mount Pisgah Native Plant Nursery). The garden had 10 blocks, with all six populations and cohorts represented in each block, for a total of 5,468 individuals. Large sample sizes and pedigreed crosses were needed for estimating additive genetic variances and covariances in a suite of traits associated with drought adaptation in the larger experiment, but were not required for the current study evaluating the evolution of floral traits across space and over time. Thus, we randomly selected 180 plants (30 from each region and cohort) for our study of floral traits. 

To quantify mating system evolution across space (among populations) and time (between cohorts), we compared floral rewards, morphology, and seed set between populations using a pollination exclusion experiment. Two buds per plant were enclosed in mesh bags to prevent pollinator access before they opened. For one bagged flower per plant, we measured three traits associated with the mating system.  First, nectar volume indicates the amount of reward that the flower provides to pollinators, such that flowers producing less nectar may exhibit a higher degree of selfing than flowers producing more nectar (CITATION). We measured nectar volume in micro-liters using a 30 μL microcapillary tube inserted into the nectary. Volume was calculated from the height of the nectar column, measured with digital calipers (Carol Ann Kearns & David William Inouye 1993). Second, nectar sugar content provides information about the quality of the floral reward, with greater nectar sugar content generally associated with outcrossing. We used a refractometer (model SR0017-ATC from manufacturer Xindacheng) to measure nectar sugar content. Samples were diluted with 50 microliters of deionized water, and the undiluted sugar content (measured in degrees Brix, Brix°, where 1°Brix equals 1 gram of sucrose in 100 grams of solution) was calculated using the formula:

𝑑𝑖𝑙𝑢𝑡𝑖𝑜𝑛 𝑓𝑎𝑐𝑡𝑜𝑟 = 𝑡𝑜𝑡𝑎𝑙 𝑠𝑎𝑚𝑝𝑙𝑒 𝑣𝑜𝑙𝑢𝑚𝑒 (𝑛𝑒𝑐𝑡𝑎𝑟 + 𝑤𝑎𝑡𝑒𝑟)/𝑛𝑒𝑐𝑡𝑎r volume

𝑢𝑛𝑑𝑖𝑙𝑢𝑡𝑒𝑑 𝐵𝑟𝑖𝑥° = 𝑑𝑖𝑙𝑢𝑡𝑖𝑜𝑛 𝑓𝑎𝑐𝑡𝑜𝑟 × 𝑑𝑖𝑙𝑢𝑡𝑒𝑑 𝐵𝑟𝑖𝑥°

Third, herkogamy, the spatial separation of stigma and anthers in a flower, should influence the probability of selfing, with shorter absolute distance between anthers and stigma associated with higher selfing rates (Opedal 2018). To estimate herkogamy, we used digital calipers to measure the distance between the uppermost anther pair and the stigma. To evaluate the efficacy of selfing, we collected fruits resulting from the second bagged flower on each plant to assess autogamous seed set as plants senesced. Due to the high seed count per fruit and multiple fruits per plant, we estimated the total seed set per fruit based on mass (Angert 2006). On a subset of fruits, we counted the number of seeds using photographs (Nikon D750 camera) of seeds on a white background. Images were analyzed with ImageJ to isolate the seeds, and the seeds were then weighed on an analytical balance. From this subset, seed number and seed mass were used to build a relationship to predict seed number for the remaining fruits based on their seed mass using the predict() function.