Study species
Arabidopsis lyrata subsp. lyrata (Brassicaceae) is a small, insect-pollinated, short-lived perennial native to the Great Lakes region of North America. It grows in relatively dry habitats with porous soils, such as sand dunes and rocky outcrops (Mable et al. 2005). Like many other Brassicaceae, this plant is usually characterized by sporophytic SI (Mable et al. 2003) and thus obligately outcrossing, although hand-pollinations have indicated that SC individuals occur at low frequencies in otherwise SI populations (Mable et al. 2005). A few populations consist of only SC plants and are characterized by a mating system with high selfing rates (Foxe et al. 2010) and shorter-life spans (Gorman et al. 2020b). Evidence suggests that there have been at least two relatively recent (<10,000 years ago) independent transitions to selfing (Hoebe et al. 2009; Foxe et al. 2010; Mable et al. 2017). In line with this recent origin, the SC populations have not evolved a clear selfing syndrome (Carleial et al. 2017). There is no evidence that pollinators discriminate between mating types (Gorman et al. 2020a).
Flowers on A. lyrata inflorescences have overlapping periods of stigma receptivity and pollen dehiscence, but SC plants do not produce seeds in the absence of pollinators (e.g., when grown in growth chambers). Although this means that selfing is not autonomous for SC plants, the close physical proximity of the anthers and stigma (Carleial et al 2017) can easily result in autogamous self-pollination (e.g., during a pollinator visit). Indeed, when conducting hand crosses, bud-emasculations of recipient SC plants are needed to prevent contamination by self-pollen (Li et al. 2019). In addition to (facilitated) autogamous self-pollination, pollinators can cause geitonogamous selfing when moving among flowers within A. lyrata plants, which often display multiple flowers (in this study, plants displayed between 1 and 46 flowers at a time) and pollinators frequently move among flowers within plants (Gorman et al. 2020a).
Common garden experiment
Our study is based on a common garden experiment (described in more detail in Gorman et al. 2020a, b) designed to test a series of hypotheses related to the occurrence of reproductive isolation between SC and SI lineages of A. lyrata. The experimental plants were the product of crosses made between field-collected seeds from North American A. lyrata populations with known breeding and mating systems. These populations represented six SC populations (Foxe et al 2010), and six SI populations selected for their close geographic proximity or comparable latitude to the SC populations. The whole experiment included additional cross-types, but in this study we exclusively used plants from SC cross-types (SC × SCwithin and SC × SCbetween, i.e., progeny of within- and between-population crosses among plants from the five SC populations) and SI cross-types (SI × SIwithin and SI × SIbetween, i.e., progeny of within- and between-population crosses among plants from the five SI populations). Testing our main hypothesis required comparisons between breeding systems, not between different cross-types within each breeding system. Nevertheless, as explained in more detail below, we initially accounted for within- versus between-population cross-types in our statistical models. Additional information on these crosses is provided in the Supplementary Materials (Supplementary Tables S1 and S2).
Plants for the experiment were propagated by first sowing seeds from each of the within- and between-population cross-types listed above in peat-based substrate in individual 3.5-inch pots between March 20 and 22, 2018. Seeds were germinated in microclimate chambers (Conviron TCR, Manitoba, Canada) until seedlings produced two true leaves. Microclimate conditions followed 11-h days with a 21/18 °C day/night cycle at 95% humidity. Up to three haphazardly chosen seedlings from each seed family were transplanted into individual Ray Leach “Cone-Tainers”™ (Tangent, Oregon, USA) between April 18 and May 1, 2018. The plants were moved to the common outdoor garden located at Trent University (Peterborough Ontario) on May 10 (prior to flowering) and placed into positions following a randomized block-design (Gorman et al. 2020a). A total of 1509 individuals were located in three replicate 1 × 9 m blocks (1, 2, and 3), and each block was subdivided into three quadrats (A, B, and C), yielding a total of nine quadrats (1A, 1B, 1C; 2A, 2B, 2C; and 3A, 3B, 3C). Each block contained 9 trays of cone-tainers, each with approximately 20 cone-tainers, yielding up to 180 equidistantly spaced plants per 3 m2 quadrat (60 plants/m2). Each seed family and cross-type was randomly distributed within and between quadrats (Gorman et al. 2020a).
The main flowering period was between June 1 and July 14, 2018. The date of peak flowering was similar among cross-types, which also had broadly overlapping flowering phenologies (Gorman et al. 2020a). On the date of peak flowering, plants produced an average of 12 flowers per day, and there were no strong differences in the maximum number of flowers produced among cross-types (Gorman et al. 2020a). The predominant floral visitors were solitary bees and hoverflies, and both types of pollinators visited SC and SI plants at similar frequencies (Gorman et al. 2020a).
Seed collection
To estimate outcrossing rates for plants from SC and SI cross-types, we sampled 78 plants (38 SC plants from 21 SC × SCwithin and 17 SC × SCbetween, and 40 SI plants from 14 SI × SIwithin and 26 SI × SIbetween) from the common garden experiment. Seed families were sampled by collecting one ripe fruit from a randomized subset of plants from the first block of the experimental garden (block 1). Because mature fruits representing the cross-types in our sampling scheme were not always available during the period over which seeds were collected (July 10–31, 2018), cross-types were unevenly represented in our final sample. Collected fruits were placed individually in 1.5 ml microtubes, which were left open to dry at room temperature for 5–7 days before they were sealed and stored in a refrigerator at 4 °C until use.
DNA extraction and genotyping
Beginning on August 21, 2018, seeds were removed from fruits and soaked in distilled water for up to 24 h. The seed coat was then removed from each seed prior to being ground into a semi-fine powder using micro pestles. DNA was extracted using QuickExtract™ Plant DNA Extraction Solution (Lucigen, Wisconsin, USA) following the manufacturer’s instructions, and eluted to a final volume of 50 µl. DNA was then purified using E.Z.N.A plant DNA kits (Omega Bio-tek, Georgia, USA) following the manufacturer’s instructions, which were modified so that we began with the filtration and DNA precipitation step. Final eluted volumes were 60 µl.
To estimate outcrossing rates, we genotyped seeds at four simple sequence repeat loci (athzfpg, atts0392, lyr417, adh-1; Clauss et al. 2002; Mable and Adam 2007) previously used for mating-system estimation for A. lyrata (Foxe et al. 2010). The forward primers were fluorescently labeled with either hex (atts0392, lyr417, adh-1) or 6-fam (athzfpg). DNA amplifications were conducted as single reactions with a final volume of 10 µl. Three of the four amplification reactions (atts0392, lyr417, adh-1) involved the use of 3.55 µl dH2O, 2 µl 5 × colorless GoTaq™ reaction buffer (Promega, Madison Wisconsin), 1 µl BSA, 0.8 µl MgCl2, 0.2 µl dNTPs, 0.1 µl forward primer, 0.1 µl reverse primer, 0.25 µl GoTaq™ DNA polymerase (Promega), and 2 µl DNA template. Amplification for these three loci involved the following run conditions: denaturation at 95 °C for 2 min, followed by 35 cycles of denaturation at 95 °C for 45 s, annealing at 57 °C for 30 s, extension at 72 °C for 1 min, and a final extension of 72 °C for 5 min. The fourth reaction (athzfpg) involved the use of 2.1 µl dH2O, 2 µl 5 × colorless GoTaq™ reaction buffer (Promega), 2 µl BSA, 1.2 µl MgCl2, 0.2 µl dNTPs, 0.1 µl forward primer, 0.1 µl reverse primer, 0.3 µl GoTaq™ DNA polymerase (Promega), and 2 µl DNA template. This fourth locus was amplified using the following run conditions: denaturation at 95 °C for 2 min, followed by 40 cycles of denaturation at 95 °C for 45 s, annealing at 55.8 °C for 30 s, extension at 72 °C for 1 min, and a final extension of 72 °C for 5 min. All amplification reactions were conducted using MasterCycler epGradient thermocyclers (Eppendorf, Hamburg, Germany).
Amplification products were diluted (1:10 dilutions) and genotyped using an automated sequencer (Applied Biosystems 3730 DNA analyzer, Applied Biosystems, Foster City, California) with ROX500 (Applied Biosystems) size standard for reference. Genotypes were analyzed using GeneMapperID-X software (v. 4.0 Applied Biosystems). Genotype scores were corrected manually after visual inspection of electropherograms. In some cases, amplification failed or provided equivocal allele scores. These seeds were excluded from analysis, yielding a total sample of 800 seeds from 77 seed families (38 SC and 39 SI).
Outcrossing rate estimation
Rates of outcrossing (t) were estimated using the MLTR software program (v. 3.4; Ritland 2002). This program uses a maximum likelihood procedure to estimate t from estimated allele frequencies in the population and inferred maternal genotypes (i.e., inferred from the segregation of alleles among progeny from the same seed parent). The calculations assume mixed-mating, with selfing occurring at a rate s, where s = 1 – tm, and where the subscript m refers to estimation of t across multiple loci. Our estimates of tm were made at the individual level using 10–12 seeds per plant for a total sample of 800 seeds from 77 seed families (38 SC and 39 SI). As a low frequency of null alleles had been reported for our loci (Foxe et al. 2010), we ran the MLTR program assuming that null alleles were present.
Patterns of insect visitation
Opportunities for self-pollination via the transfer of pollen among flowers within plants (i.e., via geitonogamous pollen transfer) were investigated by recording videos of the movement patterns of floral visitors (hereafter referred to as pollinators) in small arrays of plants. In each array, four to six plants (depending on their size) were taken from the randomized blocks that formed part of the larger common garden experiment (so this sample of plants therefore included plants not sampled for outcrossing rate estimation), such that the plants that formed an array were chosen randomly regardless of the cross-type and populations they represented (for more details on these methods, see Gorman et al. 2020a). Before starting each video segment, the number of simultaneously open flowers per plant in the array was recorded. To avoid potential effects of disturbance associated with starting and stopping the recordings, videos were trimmed to a 10-min segment that excluded the beginning and end of the recording. A total of 140 videos were analyzed, representing 23.3 h of video footage from 379 plants, 123 of which were included in more than one video recording. For each video, we recorded the type of pollinator (distinguishing between the categories solitary bee and hoverfly), and the associated duration of the visits. In addition, for each visitor, we recorded the path taken through the array from the first flower they visited; from that flower we tracked whether pollinators moved to another flower on the same plant, or to a flower on a different plant. We tracked this pattern of within- versus between-plant pollinator movements until the pollinator left the video frame. On average, each flower received 2.07 ( ± 0.13 SE) visits per hour and pollinators displayed no preference for SC versus SI cross-types (Gorman et al. 2020a). Movements outside of the video frame were tallied as between-plant movements. Based on these movement patterns, we calculated the proportion of within- versus between-plant pollinator movements for each plant and used the average values as empirical estimates of pg and px for the calculations described below.
Floral display and opportunities for geitonogamy
We used our observations of pollinator movements within versus between inflorescences to estimate the total proportion of geitonogamously visited flowers per plant. To do this, we assumed that geitonogamous pollination depended on the frequency of within- versus between-plant pollinator movements, that each pollinator movement was independent of inflorescence size, flower position, and of previous movements (i.e., pollinators moved xenogamously to another plant with fixed probability px, or visited another flower on the same plant with probability pg = 1 – px), that each flower had an equal probability of receiving an initial pollinator visit, that pollinators did not re-visit flowers on an inflorescence, and therefore that the number of pollinator visits per plant per day was finite and limited by the total number of open flowers per day. In practice, probabilities pg and px are likely to vary among plants that, for example, were visited by different types of pollinators (Brunet and Sweet 2006) or had different numbers of flowers (Harder and Barrett 1995); however, our observations were not sufficiently detailed to include these aspects of geitonogamous pollination. Based on the observations described in more detail in Gorman et al. (2020a), we further assumed that pollinators did not preferentially visit particular cross-types or otherwise alter their behavior when visiting different cross-types. Using these assumptions, the empirical estimates of pg and px based on patterns of insect visitation, and the daily record of open flowers per plant (Gorman et al. 2021a), we calculated the daily per-plant probability of geitonogamy π as follows:
$$pi = p_g cdot p_x + p_g^2 cdot p_x + p_g^3 cdot p_x + cdots + ,p_g^{n – 1} = 1 – p_x,$$
where n is the number of open flowers per plant per day (also see Supplementary Methods). Daily values of π were therefore equal to 0 (for plants with fewer than two open flowers) or pg (for plants with more than two flowers). We used these daily estimates of π to calculate the average proportion of geitonogamously visited flowers per plant, G as follows:
$$G = frac{{mathop {sum}nolimits_{i = 1}^d {f_ipi _i} }}{{mathop {sum}nolimits_{i = 1}^d {f_i} }},$$
where d is the total number of days over which each plant produced multiple flowers and f the number of open flowers per plant on day i, such that G is the weighted average of π. Put another way, values for G for a given plant represent the average per-visit probability of geitonogamy (π) scaled to reflect variation in the number of flowers displayed per day. Under this approach values of G increased with average display size and reached a maximum of G = pg for plants that always displayed multiple flowers.
Statistical analyses
To evaluate differences in outcrossing rates between plants from SC and SI populations, we first used linear mixed-effects model with tm as the dependent variable, cross-type (SI × SIwithin, SI × SIbetween, SC × SCwithin, SC × SCbetween) as the independent variable, and maternal ID (seed family) as a random grouping variable. We used a contrast-matrix approach to evaluate the overall difference in outcrossing rates between plants from SC versus SI populations using the “contrasts” option in the lmer function. This analysis indicated that there were no differences in outcrossing rates between cross-types representing the same breeding system (i.e., no differences between SI × SIwithin and SI × SIbetween, or between SC × SCwithin and SC × SCbetween; Supplementary Materials). Accordingly, and because we had no a priori reason to expect differences in outcrossing rates between cross-types with the same breeding system, we simplified this analysis so that breeding system (SI versus SC) was the independent variable. Mixed-model parameters were calculated using the lmer function from the lme4 package (v. 1.1–27; Bates et al. 2015) in R (v. 4.1.0; R Core Team 2021) and significance of fixed effects (the SC versus SI contrast) was assessed using type-2 tests obtained from the Anova function from the car package (v. 3.0–10; Fox and Weisberg 2019).
We used permutation tests to evaluate the extent to which any observed differences in outcrossing rates between plants from SC and SI populations may have been driven by the particular sample of seed families in our subsample of plants from the larger experiment. In each of 5000 permutations, we randomly sampled (with replacement) outcrossing rates from 800 seeds (400 from each of the SC and SI cross-types). While resampling data points, all other information related to the sampled data was retained, including the breeding system of the maternal parent and the maternal parent’s source population. We used these resampled datasets to re-run the mixed-model calculations using the same procedure as described in the previous paragraph. From these bootstrapped analyses, we report the upper and lower 95% confidence intervals for the difference in mean tm between SC and SI and test-statistic values of mixed-model analyses of permuted data, and present the distribution of P values in the Supplementary Materials (Supplementary Fig. S1).
We evaluated whether plants from SI versus SC cross-types differed in terms of their expected probability of geitonogamous pollination using a linear mixed-effects model with G as the dependent variable, breeding system (SI versus SC) as the independent variable, and maternal ID (seed family) as a random effect using the lmer function. Significance of the fixed effect was evaluated using the Anova function, as in the analysis of outcrossing rates outlined above.
Source: Ecology - nature.com
