Seed size and capitulum position drive germination and dormancy responses to projected warming for the threatened dune endemic Cirsium pitcheri (Asteraceae)

Abstract Among coastal plant species at risk from rapid environmental changes is the North American Great Lakes dune endemic Cirsium pitcheri. Despite being listed as federally threatened, little is known about how C. pitcheri seed attributes influence germination and dormancy‐break patterns in the context of climate change. Following a previous work where we found differences in the number and weight of C. pitcheri seeds among capitulum positions and study sites, here we examine the effects of seed attributes (capitulum position, seed weight, and site of origin) on the proportion and timing of C. pitcheri seed germination under temperature treatments that simulate projected warming in the Great Lakes (20/10, 25/10, and 30/10°C day/night). Our results demonstrate that C. pitcheri produces diverse cohorts of seeds with seed attributes that significantly influence the timing and probability of germination over a 3‐year soil seed bank. Cirsium pitcheri seed germination proportions were highest at 20°C and decreased successively at 25 and 30°C. Seeds from terminal capitula also had higher germination proportions and took longer to germinate than those from secondary capitula. Lastly, the effect of seed weight on germination probability depended on site of origin and capitulum position, with all effects varying in size and significance over time. Ultimately, our results highlight the considerable differences in germination patterns exhibited by seeds from different capitulum positions and sites of origin and provide insight into the dormancy‐break patterns that C. pitcheri might experience under predicted temperature rise in the Great Lakes region of North America.

specificities are especially vulnerable to these disturbances as they affect the establishment and survival of individuals in changing dune landscapes (Martínez & Psuty, 2004).
Habitat and other environmental changes induced by climate change are expected to further exacerbate dune plant population declines (Frosini et al., 2012;Seabloom et al., 2013;Staehlin & Fant, 2015). While plants growing in dune ecosystems are adapted to withstand environmental stresses like resource limitation, low moisture, high temperatures, and periodic burial (Brown & McLachlan, 2002;Hamzé & Jolls, 2000;Martínez & Psuty, 2004), the increased intensity and frequency of natural and anthropogenic ecosystem perturbances predicted by climate models may significantly reduce the appropriate climate envelopes for dune plant species (Frosini et al., 2012;Seabloom et al., 2013). Changes in abiotic factors like precipitation and temperature will also likely affect plant developmental responses and impact life stage performances of subsequent plant generations (Gray & Brady, 2016;Walter et al., 2016).
Among at-risk dune species with narrow habitat specificities is the Great Lakes, North American dune endemic Cirsium pitcheri (Torr. ex Eaton) Torr. & A. Gray (Asteraceae, tribe Cynareae). As a monocarpic perennial species, C. pitcheri flowers and sets seeds only once in its lifetime after growing as a rosette for 4-8 years (Havens et al., 2012;Jolls et al., 2019;Loveless, 1984). Flowering in this species is determinate, beginning with terminal capitula at the apex of main stems and progresses basipetally to axillary, secondary, and tertiary capitula that are produced at the tips of first-and second-order branches, respectively (Gijsman et al., 2020;USFWS, 2002). At the time of dispersal in late July, C. pitcheri seeds are dormant and are attached to a long pappus that helps parachute them across the dune (Chen & Maun, 1999;Hamzé & Jolls, 2000;Keddy & Keddy, 1984;USFWS, 2002). Once dispersed, C. pitcheri seeds are subjected to multiple stressors in the infertile dune sand substrate including low moisture, high sand temperatures, and burial (Hamzé & Jolls, 2000). With an ephemeral between-year soil seed bank, a combination of sand burial and multiple episodes of low temperature (cold) stratification is required to break seed dormancy (Chen & Maun, 1999;Hamzé & Jolls, 2000). Studies investigating C. pitcheri seed germination have demonstrated that seed attributes like mass and source influence seed dormancy (Hamzé & Jolls, 2000), as well as seedling emergence (Staehlin & Fant, 2015) and survival in dune environments (Chen & Maun, 1999).
The successful production of viable seeds for C. pitcheri has even greater import given its threatened status (COSEWIC, 2000).
Population viability models for C. pitcheri predict its extinction time to be <20 years (Havens et al., 2012;Jolls et al., 2015). Factors like shoreline development (USFWS, 2002), lake level fluctuations (Staehlin & Fant, 2015), and seed predation by the non-native weevil, Larinus planus (Fabricius, 1792) (Havens et al., 2012;Louda & O'Brien, 2002) are expected to contribute to C. pitcheri population declines across its range. Further population declines are also expected due to climate change as temperatures in the Great Lakes region rise (Hayhoe et al., 2010) and early life stages, like seed germination and seedling emergence, are impacted (Staehlin & Fant, 2015). However, despite the bleak predictions for C. pitcheri population persistence, little is known about the impact of climate change on C. pitcheri seed dormancy-break patterns and how seed attributes affect the timing and probability of seed germination in the context of climate change.
Following previous work where we found differences in the number and weight of C. pitcheri seeds produced in terminal and secondary capitula between two study sites (Gijsman et al., 2020), we conducted an experiment that examined the effect of seed attributes (capitulum position, site of origin and seed weight) on C. pitcheri seed germination in the context of climate change. We first broadly assessed the effect of seed attributes on cumulative C. pitcheri seed germination at three temperature treatments that simulate projected temperature rise in the Great Lakes and then constructed a more fine-grained distribution of germination times and probabilities by tracking the timing and germination of individual seeds. Altogether, this study aims to improve current knowledge on C. pitcheri germination ecology and better predict its population dynamics and responses to climate change. at SC, 1,277 at WFD) (see Gijsman et al., 2020 for collection methods). For each site and year, we haphazardly selected a subsample of the total number of viable seeds produced by flowering plants in terminal and secondary capitula. However, because flowering plants across sites and years produced different numbers of capitula, sample sizes for seeds from each site, capitulum position, and year vary slightly (see Table S1 for specific sample sizes and experiment replicate numbers).

| Seed preparation and germination trials
Prior to our germination trials, we surface-sterilized seeds through submersion in a 0.25% sodium hypochlorite (bleach) solution for 1 min and rinsed seeds with distilled water to ensure the removal of contaminating microbes. Any excess moisture from the surface of the seeds was removed using lint-free Kim wipes. For germination trials tracking cumulative seed germination, we plated a randomized subsample of 20 seeds from a given capitulum position and site of origin onto Petri dishes lined with dampened, sterilized germination paper, and randomly assigned Petri dishes to one of three temperature treatments (20/10, 25/10, and 30/10°C day/night, hereafter "20, 25, and 30°C"). These temperatures fall within ranges previously tested for C. pitcheri (Chen & Maun, 1998, 1999Staehlin & Fant, 2015) and simulate predicted temperature scenarios that are expected to influence its range (Hayhoe et al., 2010;Vitt et al., 2010).
To investigate the timing and probability of germination for individual C. pitcheri seeds, we weighed seeds to the nearest 10 −4 g using a Sartorius Mettler Toledo balance prior to surface sterilization and tracked seed attributes (site of origin, capitulum position, seed weight) of individual seeds using 96-well plates. We randomly assigned each seed to a temperature treatment and plated seeds onto Petri dishes lined with two layers of differently colored germination paper, to facilitate seed tracking and visualization of individual seed germination. We hole-punched the top layer of germination paper, labeled holes with a seed-tracking letter and plated up to 15 seeds for a given capitulum position and site of origin.
Once prepared, we placed Petri dishes into incubators set to 3°C for a cold stratification period of 8 weeks, to break seed dormancy by simulating natural winter conditions C. pitcheri seeds must overcome to germinate (Chen & Maun, 1998). However, despite surface sterilization with bleach and the use of sterile equipment, fungal growth was a significant problem in our experiment, particularly for seeds from WFD. To control the spread of fungus, we treated Petri dishes with a dilute solution of captan fungicide (3.5 g/L), a tissue culture and seed sterilization treatment (Falloon, 1987;Payamnour et al., 2014).
After the 8-week cold stratification period, we placed Petri dishes into light-controlled incubators, alternating 12/12 day/night hours at 20/10, 25/10, and 30/10°C. We tracked seed germination, defined as radicle emergence >1 mm, every 2 days for up to 10 weeks. Once germinated, we transferred seeds to labeled starter trays with a soil mix composed of two-parts potting mix and one-part sand and grew seedlings in their respective germination temperatures until large enough to be transplanted to sandboxes at the Chicago Botanic Garden (CBG) for use in future dune restoration projects.
As C. pitcheri seeds are viable for up to 3 years in the soil seed bank and require multiple consecutive periods of cold stratification to fully break dormancy (COSEWIC, 2000), we placed any remaining ungerminated seeds back into cold stratification at 3°C for an additional 8 weeks, followed by a subsequent warm-temperature germination period of 10 weeks at their previously assigned temperature.
We subjected all ungerminated seeds to up to three cold stratification-germination rounds to simulate an approximate 3-year seed bank. Any seeds that germinated within a 3°C cold stratification period were included in germination proportions within each round of cold stratification-germination. For germination trials tracking individual seed germination, we recorded the following additional time-to-event data over the course of each cold stratification-germination round: binomial germination success (y/n), germination within cold stratification round (y/n), number of germination and cold stratification rounds experienced prior to germination, and the number of days taken to germinate from the start of the experiment (day 0 = first day of first cold stratification round).

| Statistical analyses
We used generalized linear models (GLMs) with a binomial distribution to test for the effects of site of origin, temperature, capitulum Values are fitted cumulative germination proportions and standard errors for each cold stratification-germination round.
TA B L E 1 Cirsium pitcheri cumulative germination proportions for each cold stratification-germination round at temperature treatments 20/10, 25/10, and 30/10°C (day/night) position, and their interactions on cumulative C. pitcheri germination in each cold stratification-germination round. For each cold stratification-germination round, we selected minimal adequate models that best described our data using a stepwise, backward elimination approach that sequentially tested the removal of a seed attribute predictor term or interaction with a threshold of 5% (p ≤ .05) for rejecting simpler models (Crawley, 2015).
For individual seed germination trials, we used GLMs with a binomial distribution to test for the effects of site of origin, capitulum position, temperature, seed weight, and their interactions on total C. pitcheri seed germination proportions at the end of all cold stratification-germination rounds. We also tested for differences in the mean time to seed germination among seed attribute predictors using a GLM with a negative binomial distribution due to overdispersion with Poisson models.
To investigate the rate of C. pitcheri seed germination over time, A value greater than one represents an increase in the probability of an event taking place within a treatment group compared to a reference, less than one a reduction and equal to one represents no difference. We conducted our comparisons of hazard ratio estimates within levels of seed attribute predictors, using the site of origin level "SC," capitulum level "terminal capitula" and temperature level "20°C" as references. We conducted all analyses in R (R Core Team, 2016) and used the survival (Therneau, 2015) and survminer (Kassambara et al., 2019) packages for time-to-event analyses and ggplot2 package to graph results (Wickham, 2016).

| Cumulative seed germination
Cirsium pitcheri cumulative seed germination proportions were consistently higher for seeds from terminal capitula and at 20°C across all cold stratification-germination rounds (Table 1, Figure 1). By the end of the third cold stratification-germination round, seeds from secondary capitula at WFD had higher germination proportions than at SC across all temperatures (Table 1, Figure 1). However, cumulative germination proportions of all seeds at 25 and 30°C were more than 50% and 70% lower than those at 20°C, respectively, for both sites of origin and capitulum positions ( Figure 1).
Models that best fit the germination data differed across cold stratification-germination rounds. In the first round, models included all three individual predictor effects (site of origin, capitulum position, temperature), as well as the site-capitulum position (ANOVA, Deviance = 5.83, p = .02) and temperature-capitulum position two-way interactions (ANOVA, Deviance = 8.45, p = .01).
Models for the second and third rounds, however, included the individual predictor effects, their three-way interaction and all twoway interactions (ANOVA, round two: Deviance = 9.66, round three: Deviance = 10.1, both p < .01).

| Individual seed germination
When tracking the germination of individual seeds over time, seeds from SC had higher total germination proportions than WFD across   Figure 2a). Overall, seeds from terminal capitula at SC had the highest total germination proportions which consecutively decreased at increasing temperature treatments (Figure 2a). For seeds from WFD, total germination proportions did not differ among capitulum positions, but also decreased at increasing temperature treatments (Figure 2a).
At both sites of origin, seeds from terminal capitula took longer to germinate than those from secondary capitula across all temperature treatments except 30°C (Table 2, Figure 2b). For SC, seeds from terminal and secondary capitula had comparable mean times to germination at 20 and 25°C, but took longer to germinate at 30°C. The opposite was the case for seeds from WFD, for which seeds from terminal and secondary capitula took longer to germinate at 20°C, followed by 25°C, and then 30°C (Figure 2b).

| Effect of seed weight on germination probability
Individual seed weight and its interactions with seed attribute predictors was a significant factor influencing the germination probability of individual seeds at the three temperature treatments (ANOVA, Deviance = 8.73, p = .012). At SC, the germination probability of individual seeds increased concurrently with seed weight across all temperature treatments regardless of capitulum position (Figure 3).

| Hazard ratio estimates for predictors
Compared to seeds from SC, seeds from WFD were more than 70% less likely to germinate in the first and second cold stratificationgermination rounds, but as likely to germinate as those from SC in the third round ( Figure 5). Seeds from secondary capitula were as likely to germinate as those from terminal capitula in the first round, but more than 45% less likely to germinate in the second and third rounds. Compared to 20°C, seeds at 25 and 30°C, respectively, were 22% and 58% less likely to germinate in the first round. In the second and third rounds, seeds were more than 50% and 70% less likely to germinate at 25 and 30°C, respectively ( Figure 5).

| D ISCUSS I ON
Our study demonstrates that C. pitcheri is capable of producing phenotypically diverse cohorts of seeds, with seed attributes that influence the probability, timing and rate of germination of seeds at temperature treatments that simulate projected warming in the Great Lakes. Seed germination events were also broadly distributed over a 3-year seedbank, which we simulated by subjecting seeds to three consecutive periods of cold stratification and warm-temperature germination (cold stratification-germination round).

| Effect of capitulum position and site of origin on germination
In general, seeds from terminal capitula had higher germination proportions and took longer to germinate than those from secondary capitula at both sites of origin. Seeds from WFD also took longer to germinate than those from SC across all temperature treatments The effects of seed weight on dormancy and germination, more specifically, have also been shown to depend on a seed's developmental position in a fruit or on a parental plant (Baskin & Baskin, 2014;Diggle, 1995;Susko & Lovett-Doust, 2011). We had previously found that C. pitcheri seeds from terminal capitula are significantly heavier than those from secondary capitula and that flowering plants at both sites in 2017 produced fewer viable seeds in terminal and secondary capitula in 2016 (Gijsman et al., 2020).
Following up on that work, here, we used seeds from senescing flowering plants in 2016 and 2017 to investigate the effects of C. pitcheri seed attributes on cumulative and individual seed germination, respectively. Hence, while we cannot make a direct comparison of our results for seeds collected in different years, it is likely that interannual variation in C. pitcheri seed production influenced seed attribute effects in our study. In addition, the position-dependent effects on seed germination observed in this study likely reflect resource allocation strategies employed by maternal plants, resulting in the seed weight differences among capitulum positions seen in C. pitcheri plants at both sites.

| Effect of seed weight on germination
Seed weight significantly influenced the germination probability of individual seeds at both sites, although to different extents. At SC, heavier seeds from both terminal and secondary capitula had higher germination probabilities than lighter seeds across all temperature treatments. In contrast at WFD, these effects were not as clear-cut.

TA B L E 3
Log-rank test pairwise temperature comparisons of Kaplan-Meier survival curves for Cirsium pitcheri seeds from terminal and secondary capitula at SC and WFD in 2017 For seeds from terminal capitula, germination probabilities increased with seed weight at 20 and 25°C and decreased at 30°C, but these effects were not significant. Given that terminal capitula at WFD produced few viable seeds in 2017 (Gijsman et al., 2020), our sample sizes for this capitulum position were small, which may have precluded our ability to distinguish any effects. For seeds from secondary capitula at WFD, seed weight significantly predicted germination probabilities at 25 and 30°C. At these high temperature treatments, heavier seeds were less likely to germinate than lighter seeds. These results are in accordance with previously published findings for C. pitcheri which indicate that seed weight influences seed dormancy and germination (Chen & Maun, 1999;Hamzé & Jolls, 2000).
Larger C. pitcheri seeds have also been reported to have higher chances of infection from seed-and soil-borne pathogens (Chen & Maun, 1999). These results are especially noteworthy given that L. planus oviposits eggs at higher proportions in secondary capitula at WFD (Gijsman et al., 2020). When combined with the lower germination probability of heavier seeds produced in secondary capitula at WFD, our results highlight the potential impacts of L. planus on C. pitcheri that go beyond seed predation. In addition, fungal infection for seeds from WFD was a substantial issue in our study.
While scarification of the seed coat by predating insects can aid with water imbibition and break seed dormancy (Han et al., 2018;Karban & Lowenberg, 1992), it can also impact seed quality, increase mold damage by disseminating fungal spores (Caneppele et al., 2003;Sinha, 1984), and impede seed germination (Dalgleish et al., 2012;Koptur, 2009;Tomaz et al., 2007). Although we did not assess the level of damage that seeds from WFD endured from predation by L. planus, our results suggest that it may impact the quality and dormancy of C. pitcheri seeds that survived predation. To better ascertain these potential effects, future studies should directly investigate the impact of L. planus on C. pitcheri seed germination and early life stages.

| Effect of temperature on germination
Germination proportions for C. pitcheri seeds were highest at 20°C and decreased successively at the 25 and 30°C temperature treatments. The probability of seed germination at the higher temperature treatments also decreased over time. At 25°C, seed germination in the second cold stratification-germination round was more than 50% less likely to occur than at 20°C, with that number increasing to 70% at 30°C. Such reductions in seed germination at 25 and 30°C suggest a critical temperature tolerance limit for C. pitcheri germination.
Survival curves describing the germination of seeds over time also indicate that germination rates differed at the different temperature treatments and across cold stratification-germination rounds. For seeds from WFD, survival curves at the different While it is difficult to predict the extent to which the seed attribute effects in this study may impact the establishment and survival of C. pitcheri populations, our results suggest that climate change is a noteworthy threat that could increase the species' risk of extinction (Havens et al., 2012). Climate projections for the Great Lakes region predict annual temperatures to increase by 1.4 ± 0.6°C in the near-term (2010-2039) (Hayhoe et al., 2010). Winter and spring precipitation is also projected to increase by 20% under a low emission scenario (Hayhoe et al., 2010). In conjunction with our results, predicted shifts in growing seasons and frequent extreme climate events suggest that climate change will significantly impact the timing and outcomes of seedling recruitment processes that are essential for the regeneration of C. pitcheri plant populations in dune ecosystems.

| Implications for C. pitcheri populations
For C. pitcheri, variation in seed production, germination rate, and timing is likely an important bet-hedging strategy that ensures diversity in the timing of population establishment events in a heterogeneous environment (Giesel, 2003;Sales et al., 2013) and under changing environmental conditions (Lu et al., 2017). In addition, because flowering anthesis is asynchronous in C. pitcheri, differences in the floral phenology between terminal and secondary capitula further contribute to the temporal variation in population establishment events. Terminal capitula in C. pitcheri develop, mature, and set seeds that are also heavier, earlier than secondary capitula (Gijsman et al., 2020;USFWS, 2002). These differences in floral phenology and seed production between capitulum positions therefore may result in contrasting seed performances for C. pitcheri populations, with weight-dependent trade-offs in dispersal, burial, dormancy, germination, and seedling emergence and establishment (de Ruiz, 2002;Lu et al., 2017;Susko & Lovett-Doust, 2000;Turnbull et al., 2004;Venable & Brown, 1988). For instance, C. pitcheri seed germination and seedling emergence have been found to be negatively correlated with seed burial depth (Chen & Maun, 1999;Hamzé & Jolls, 2000). Yet, seedlings F I G U R E 5 Hazard ratio estimates and 95% confidence intervals for predictor effects (site of origin, capitulum position, temperature) in Accelerated Failure Time (AFT) models at each cold stratification-germination round. Stars denote predictor significance with respect to predictor reference level (*p < .05, ***p < .001) emerging from larger seeds produce longer roots that enhance their establishment in the dunes (Chen & Maun, 1999 Similar strategies and germination patterns have been reported for plant species with two flowering-fruiting events. For example, seed production in Centaurea eriophora (Asteraceae) depends on the size and flowering period of capitula (de Ruiz, 2002).
Seeds from small capitula, which flower later than large capitula, are lighter and initially produce seedlings that are less robust (de Ruiz, 2002). Likewise, Gurvich et al. (2004) found that seeds from early-flowering Bidens pilosa (Asteraceae) plants (referred to as early type) are heavier, have faster germination rates and lower germination probabilities than those from normal-type plants.
In both cases, the prolongation of flowering and fruiting periods resulted in the production of seeds with variable traits that are essential for the initial establishment of populations in the face of environmental change (de Ruiz, 2002;Gurvich et al., 2004).
Nonetheless, it is important to note that while such reproductive strategies can produce a wide array of seeds, their success is also highly dependent on site dynamics and climatic conditions such as temperature, the latter which our study indicates has strong effects on seed germination.

| CON CLUS ION
In this study, we found that C. pitcheri produces seeds with germination events that are widely distributed over a 3-year soil seed bank and with seed attributes that significantly influence the timing and probability of germination under projected warming. Our results also indicate that terminal capitula produce seeds with the highest germination probabilities and that high temperatures drastically impact C. pitcheri germination. With climate change posing a significant threat to C. pitcheri populations, actions that increase the likelihood of germination and seedling survival, such as selectively using larger seeds or seeds sourced from terminal capitula, may help ensure population persistence and success in conservation and restoration efforts.

ACK N OWLED G M ENTS
Funding for this research was provided to FG by the Office of Undergraduate Research and Weinberg College of Arts and Sciences at Northwestern University. This work was conducted under U.S.
Fish and Wildlife and Wisconsin DNR permits that authorize C. pitcheri research and seed collection. We thank Kayri Havens, Robert Hevey and Chicago Botanic Garden student interns for help with data collection. We also thank Stuart Wagenius for help with statistical analyses.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflicts of interest.