Gene flow can have rapid effects on adaptation and is an important evolutionary tool available when undertaking biological conservation and restoration. This tool is underused partly because of the perceived risk of outbreeding depression and loss of mean fitness when different populations are crossed. In this article we briefly review some theory and empirical findings on how genetic variation is distributed across species ranges, describe known patterns of gene flow in nature with respect to environmental gradients, and highlight the effects of gene flow on adaptation in small or stressed populations in challenging environments (e.g., at species range limits). We then present a case study involving crosses at varying spatial scales among mountain populations of a trigger plant (Stylidium armeria: Stylidiaceae) in the Australian Alps to highlight how some issues around gene flow effects can be evaluated. We found evidence of outbreeding depression in seed production at greater geographic distances. Nevertheless, we found no evidence of maladaptive gene flow effects in likelihood of germination, plant performance (size), and performance variance, suggesting that gene flow at all spatial scales produces many offspring with high adaptive potential. This case study demonstrates a path to evaluating how increasing sources of gene flow in managed wild and restored populations could identify some offspring with high fitness that could bolster the ability of populations to adapt to future environmental changes. We suggest further ways in which managers and researchers can act to understand and consider adaptive gene flow in natural and conservation contexts under rapidly changing conditions.

Study system

We examined F1 hybrid performance of the thrift-leaved trigger plant, Stylidium armeria, a species common throughout the montane and high elevation woodland areas of southeastern Australia (with the current focus on the Australian Alps) (Figure 3). The alpine areas in Australia form a rare ecosystem, with treeless alpine vegetation covering ~0.15% of the continent, and like other alpine environments around the world, they are highly vulnerable to the effects of climate change (Hughes, 2003). Cuttings from wild plants were harvested from various sites throughout the Victorian and New South Wales high country. Outcrossing of the different populations was performed and the F1 progeny of these outcrossings were germinated under controlled nursery conditions. Stylidium armeria is morphologically variable throughout its distribution, with differences thought to be related to surrounding vegetation, soil type and climatic factors (Raulings & Ladiges, 2001). The pollination unit is a zygomorphic flower, which is characterized by the fusion of staminate and pistillate tissues into a motile, protandrous column, which is “triggered” when pollinators, usually native bees, land on the corolla (Armbruster & Muchhala, 2009).

Twelve populations were sourced, 4 from each of 3 mountain regions within the Australian Alps (Bogong High Plains, Victoria; Mount Buller region, Victoria; and Kosciuszko region, New South Wales) to include in the experimental crosses (Table 1, Figure 4). The 3 regions vary in distance from each other between ca. 50-200 km. Stylidium armeria is genetically highly differentiated between Victorian alpine regions based on pooled SNP data, comparable to differentiation seen for alpine herbaceous plants, which tend to be more differentiated than alpine shrubs (Bell et al., 2018). Within each region, the four populations differ in elevation by 174-227 m and distance by 3-15 km from each other. Within each population, at least 30 plants were collected as cuttings from rhizomes, each including a basal rosette of leaves. Each collection spanned at least a 100 m² area. Plants were first transferred in Fall 2011 after collection into planting tubes with a “native mix” soil medium used at Burnley Campus (University of Melbourne) glasshouses, which consists of pine bark, peat and sand. After winter, plants were transferred individually into 7-inch pots. After pollinations were initiated some plants died due to lack of soil drainage, which may have contributed to some failed pollinations.

Experimental crosses 

Experimental pollinations between all populations, including within population crosses, were conducted to examine outcrossing effects of three broad spatial categories of gene flow: within site/population (WP), the site of collection; within range (WR), between sites within a mountain range; and between mountain ranges (BR). Crosses were completed during summer and fall of 2012 (443 total crosses). Within each population, sire plants were chosen randomly during flowering and were mated with up to 2 dams from each population, including their originating population. Dams were chosen randomly within each population, with replacement. Plants within each population were used only once as pollen donors (sires) to individuals in their population and to all other populations, but could also serve as pollen receivers (dams) to other sires in the study. Two replicate flower pollinations were made for each cross. Crosses were completed with equal directionality between populations; that is to say, each population served as both pollen donor and pollen receiver to each other population. This design produced more possible crossing combinations in increasing order of geographic scale: WP = 12 possible types; WR = 18 possible types; BR = 48 possible types. We used this design to increase sampling size and variety in anticipation of outbreeding depression from longer-distance gene flow crosses, and we considered WP crosses to serve as a natural control against which to compare WR and BR crosses. To test for background or contamination pollination, 69 individual flowers, randomly chosen, were marked to test for unintended seed set rate. Thirty flowers were randomly chosen and self-pollinated to test for self-incompatibility.

Of the total pollinations observed (including crosses, self-pollinations, and tests of inadvertent pollination), 262 yielded seeds (59.1% of 443). Some crosses may have been unsuccessful due to plant stress that occurred from lack of soil drainage, but this effect was independent of population of origin. Capsules were harvested when mature and dried, and the resulting seeds were later counted and the seed lots weighed to produce an average seed weight (number of seeds divided by total weight). F1 seeds were sowed into 7 x 8-cell seedling trays during autumn of 2014. The cells were partially filled with the native mix and covered with a layer of “seed raising mix,” which consisted of 5 parts medium-grade pine bark, 5 parts fine pine bark, 1 part coarse sand, and 1 part sieved peat, including the additives Saturaid (1500 g/m²) and dolomite (750 g/m²). Each replicate included three seeds sown per cell. Depending on seed availability, 1-15 replicates were sown per seed family and all were completely randomized across 50 trays included in the study, for a total of 2,800 replicates sowed. Following the completion of sowing, trays were treated with Regen Smokemaster 2000, a smoke water solution applied at a rate of 100 mL per 1L of water, sprayed per square meter in order to trigger earlier germination. Trays were randomly rotated once weekly, during which trays were monitored for evidence of germination and survival. Seedlings (520 total) were transferred into larger pots with new potting material after several months’ growth. Plants were grown for 235 days between the first germinant and the end of the growth trial. Plant sizes were estimated by multiplying the longest aboveground height and width measurements of surviving plants (493 in total). Plant size was also standardized by dividing the final plant size by the number of days since germination. Percent germinated plants, seed number, average seed weight, final plant size, and the variance in final plant size were all compared to assess the outcomes of outcrossing among the three general spatial categories, WP, WR, and BR.

Statistics

Our main question was, what is the effect of crossing type among the three spatial categories? We used means comparison methods (ANOVA–and non-parametric Kruskal-Wallis tests by ranks) to test for differences among crossing types. We also used generalized linear models (GLM) to model additional effects, including the variation of specific crosses within WP, WR, and BR categories. Models incorporating all cross types simultaneously could not run due to too many missing rows of data (e.g., a row with a WP cross type cannot have a BR cross type, and vice versa). However, we did test these effects in reduced models, separating WP and WR effects from BR effects (see Supplemental Information). Plant size data were square root-transformed to meet parametric assumptions and analysis of variance (ANOVA) was used to detect significant differences among crossing categories. Because transformations of seed number and average seed weight data still failed to meet parametric assumptions, we instead used Kruskal-Wallis tests to test for differences among crossing types for these two variables. Chi-squared analysis was used to test for differences in germination. Levene’s test of equality of variances was used to test for significant variance differences among crossing types in plant size. Finally, we tested for the effect of spatial distance of plant crosses using the estimated Haversine distance between populations on the above seed and plant size variables. Crosses within populations were assigned a distance of 0 km. Because crossing distance data failed to meet assumptions of parametric analyses, we used Spearman’s rho (ρ) rank correlations to test the effect of crossing distance on seed and plant size traits. All analyses were carried out in the JMP statistical package (version 16.0.0, SAS Institute, Inc., 2021).