Premise: Apomictic plants (reproducing asexually through seed) often have larger ranges and occur at higher latitudes than closely related sexuals, a pattern known as geographical parthenogenesis (GP). Explanations for GP include differences in colonizing ability due to reproductive assurance and direct/indirect effects of polyploidy (most apomicts are polyploid) on ecological tolerances. While life history traits associated with dispersal and establishment also contribute to the potential for range expansion, few studies compare these traits in related apomicts and sexuals. 

Methods: We investigated differences in early life history traits between diploid-sexual and polyploid-apomictic Townsendia hookeri (Asteraceae), which displays a classic pattern of GP. Using lab and greenhouse experiments, we measured seed dispersal traits, germination success, and seedling size and survival in sexual and apomictic populations from across the range. 

Key Results: While theory predicts that trade-offs between dispersal and establishment traits should be common, this was largely not the case in T. hookeri. Apomictic seeds had both lower terminal velocity (staying aloft longer when dropped) and higher germination success than sexual seeds. While there were no differences in seedling size between reproductive types, apomicts did, however, have slightly lower seedling survival than sexuals. 

Conclusions: These differences in early life history traits, combined with reproductive assurance conferred by apomixis, suggest that apomicts achieve a greater range through advantages in their ability to both spread and establish. 

Study system

Townsendia hookeri is a diminutive perennial member of the sunflower family (Asteraceae) with two forms: diploids that reproduce sexually and autopolyploids (mainly triploids) that reproduce via gametophytic apomixis (Beaman1957; Thompson and Whitton 2006). Sexual individuals are self-incompatible, while apomicts set seed autonomously, without the need for pollen to fertilize the endosperm. Townsendia hookeri produces single-seeded diaspores with persistent feathery pappus bristles (Strother 1993) usually associated with wind dispersal (Andersen 1992). Despite differing in ploidy and reproductive mode, the two forms are macro-morphologically indistinguishable in the field but can be distinguished by pollen traits. Polyploids produce larger viable pollen grains than diploids and have much lower pollen viability. To date, our lab group has characterized more than 90 populations throughout the species’ range and found a tight association between low pollen viability and larger genome size (Thompson & Whitton 2006; Garani 2014; Lee 2015), supporting Beaman’s (1957) contention that diploids are sexual and polyploids are apomictic (Thompson & Whitton 2006; Thompson et al. 2008).

Sexual populations have a much smaller range than the apomicts and primarily occur between Boulder, CO and Laramie, WY (Lee 2015). Apomictic populations extend from southern WY along the eastern side of the Rocky Mountains as far north as BC, Canada. In addition, a small number of diploid-sexual and polyploid-apomict populations occurs in a disjunct distribution in the Yukon territory (Thompson & Whitton 2006; Garani 2014).

Seed dispersal traits and terminal velocity measurements and analysis

We compared diaspore traits of apomicts and sexuals using field-collected diaspores (technically cypselae, but from now on referred to as “seeds”) obtained between 2008 and 2013 from five diploid sexual and seven polyploidapomictic populations from across T. hookeri’s range, including northern populations in British Columbia and the Yukon territory (Appendix S1, Table S1; see the Supplementary Data with this article). 

We selected two seeds from each of up to ten maternal plants per population, using only filled and darkly coloured seeds (traits that indicate viability; Garani 2014) with an intact pappus. In order to restore the pappus to a comparably open state (relative to the somewhat flattened state that resulted from storage in the collection envelopes), we placed seeds on wet filter paper in a sealed petri dish for ∼24 hours, then removed them and allowed them to air-dry for another 24 hours. At this point, the pappus had achieved a more regular form consistent with what is seen at the time of field collection. For each seed, we estimated the mass using an analytical balance, recorded the number of pappus bristles, and measured the length of two bristles from the center-apex using an ocular micrometer. We measured the angle of attack, the maximum angle across the open pappus bristles centered on their point of attachment to the seed proper, using a protractor (see Figure 2 in O’Connell and Eckert (2001) for a diagram depicting diaspore traits).

We estimated terminal velocity by dropping individual seeds down a clear 120 cm long acrylic glass tube. Seeds were dropped by holding the pappus with tweezers and releasing. After conducting preliminary trials, we determined that the seeds would reach terminal velocity over the first 70 cm of the drop. We then recorded the bottom 50 cm of the fall with a video camera (shooting at 30 frames per second). We calculated the terminal velocity of each seed by dividing the drop time by drop distance (50 cm), taking great care throughout not to damage or disturb the structure of the pappus.

We analyzed differences in seed traits (terminal velocity, angle of attack, bristle length, number of bristles, and seed mass) between reproductive modes using mixed effects models. Terminal velocity, angle of attack, and bristle length were analyzed using linear mixed effect models (LMMs), and number of bristles was analyzed used generalized linear mixed effects models (GLMMs) with a log link function for Poisson data. Reproductive mode was a fixed factor; population and mom (nested within population) were included as random factors. In order to test for differences in variance in terminal velocity between reproductive modes, we used Fligner-Killeen’s Test of Homogeneity of Variances (fligner.test() in R) which is robust to departures from normality. 

We used linear models to test the effects of each diaspore trait on terminal velocity separately. Traits were centered and scaled using the scale() function in R with default options, which subtracts the mean from each value and divides by the standard deviation.  We explored the relationship between diaspore traits and terminal velocity using simple models that included each trait as a fixed effect (without the random effects of population or mom). 

Germination in the lab

For the germination and greenhouse experiments, we collected seeds from six diploid sexual populations and six polyploid apomict populations in the spring of 2013 and 2014 (Appendix S1, Table S1). Seeds were stored at room temperature in paper coin envelopes. The ploidy of each population used in this study was assessed previously by flow cytometry (Lee 2015). 

We chose 11 maternal plants (hereafter known as “moms”) from each population, and 32 dark, filled seeds from each mom. We assessed germination in agar-filled petri dishes (hereafter referred to as “plates”). Each plate was divided into four quadrants, with four seeds from a single mom randomly assigned to each quadrant. Within quadrants, seeds were spaced to avoid contact. Due to the large number of seeds, we filled the plates over the course of three days (beginning on June 18, 2014). We stacked the plates on the lab bench at room temperature (21°C) away from direct sunlight and re-ordered the stacks every 3–4 days. We checked the plates every 1–2 days until no new seeds had germinated for 10 days, and scored germination as successful based on the emergence of both a radicle and a pair of cotyledons.

Seedling traits

We transplanted seeds from agar into racks of cone-tainer pots (Proptek - Watsonville, California) 2–3 days after germination over a period of two weeks. We filled the pots with a well-draining soil mixture (4 potting soil: 2 sand: 1 perlite) and placed the seedling racks in growth chambers (Conviron, various models - Winnipeg, Canada). Growth chambers were set for 12h light (20°C) / 12h dark (10°C) cycles. We hand-watered seedlings daily in the week following transplanting and every two days after that. We rotated the racks between chambers once per week to account for potential differences among growth chamber models. After most seedlings reached a height of at least 2 cm (early July 2013), we hardened the seedlings in a greenhouse at the University of British Columbia Farm (Vancouver, Canada). The greenhouse environment was regulated by Argus Controls (Surrey, Canada) (set to 12h light at 20°C / 12h dark at 12°C), and the seedlings were bottom-watered on flood tables. Because the seedlings grown here were subsequently transplanted in a common garden experiment (the results of which are in Hersh 2020), we moved the plants outside to harden them to UV light (3–5 hours per day). After two weeks of hardening, we took our final measurements, censusing all plants for survival and estimating plant size using total number of leaves, and the length of one fully expanded leaf per seedling. Thus, our measurement of seedling size reflects the size achieved from the day of germination until the census approximately two months later. While not every seed germinated on the same day (and therefore did not have equal time to grow until the census), 75% of seeds that germinated did so within a five-day period, so we expect the variation in total growth time had minimal effects on our growth estimates. 

Analysis of germination and seedling traits

We used GLMMs to analyze differences in germination success (logit link - binomial) and germination speed (log link - poisson, due to speed being rep- resented by day counts) between reproductive modes. Reproductive mode was a fixedfactor; plate, population, and mom (nested within population) were included as random factors. Similar to terminal velocity above, we tested for differences in variance in germination success and germination speed using Fligner-Killeen’s Test. We used GLMMs to analyze differences between reproductive modes for seedling survival (logit link - binomial) and leaf number (log link - poisson), and LMMs for leaf length. Reproductive mode was a fixed factor; rack, population, and mom (nested within population) were included as random factors. Including rack as a random effect resulted in model singularities for the leaf length and leaf number models, so we removed it from those analyses. Because populations were the main level of replication, we used subsequent mixed models to determine whether there were differences among populations, with population as a fixed effect and the same random factors as above. 

We performed the mixed-effect analyses using the lmer and glmer functions implemented in the lme4 package (Bates et al. 2015) and fixed effect analyses using the lm function in R version 3.6.1 (R Core team 2019). We evaluated the strength of effects by using a combination of likelihood ratio tests (LRTs; comparing fully fitted model to a model with the tested term removed) and model predictions using the ggeffects package (Lüdecke 2018). We considered that we had strong evidence for differences between groups when both the LRT P-values were below 0.05 and confidence intervals from model predictions were largely non-overlapping. In cases where only one of these criteria was met (i.e., LRT P-value below 0.05 but substantial overlap in predicted confidence intervals), we considered that the evidence for differences was moderate.