Seed traits are pleiotropically regulated by the flowering time gene PERPETUAL FLOWERING 1 (PEP1) in the perennial Arabis alpina

Abstract The life cycles of plants are characterized by two major life history transitions—germination and the initiation of flowering—the timing of which are important determinants of fitness. Unlike annuals, which make the transition from the vegetative to reproductive phase only once, perennials iterate reproduction in successive years. The floral repressor PERPETUAL FLOWERING 1 (PEP1), an ortholog of FLOWERING LOCUS C, in the alpine perennial Arabis alpina ensures the continuation of vegetative growth after flowering and thereby restricts the duration of the flowering episode. We performed greenhouse and garden experiments to compare flowering phenology, fecundity and seed traits between A. alpina accessions that have a functional PEP1 allele and flower seasonally and pep1 mutants and accessions that carry lesions in PEP1 and flower perpetually. In the garden, perpetual genotypes flower asynchronously and show higher winter mortality than seasonal ones. PEP1 also pleiotropically regulates seed dormancy and longevity in a way that is functionally divergent from FLC. Seeds from perpetual genotypes have shallow dormancy and reduced longevity regardless of whether they after‐ripened in plants grown in the greenhouse or in the experimental garden. These results suggest that perpetual genotypes have higher mortality during winter but compensate by showing higher seedling establishment. Differences in seed traits between seasonal and perpetual genotypes are also coupled with differences in hormone sensitivity and expression of genes involved in hormonal pathways. Our study highlights the existence of pleiotropic regulation of seed traits by hub developmental regulators such as PEP1, suggesting that seed and flowering traits in perennial plants might be optimized in a coordinated fashion.


| INTRODUC TI ON
Plant life cycles are characterized by discrete phase transitions, such as the initiation of reproduction and the germination of seeds. The timing of age-dependent transitions and their coordination with environmental cues strongly determines fitness (Bradford, 2005;Pianka & Parker, 1975;Takada, 1995). Both the timing of seed germination and flowering have evolved in response to predictable cues, such as seasonal thresholds of day length and temperature (Fenner, 1998;Imaizumi & Kay, 2006;Roff, 1996). The identification of genes regulating phenology is an important ongoing task in plant biology.
However, the roles that genetic elements play in regulating phenological traits in annuals, which reproduce only once, may differ in perennials, which reproduce many times.
The regulation of flowering in response to a prolonged period of cold (vernalization) is an important phenological trait that differs among annual and perennial Brassicaceae species. In Arabidopsis thaliana L., the MADS-box transcription factor FLOWERING LOCUS C (FLC) regulates flowering in response to vernalization Sheldon, Rouse, Finnegan, Peacock, & Dennis, 2000). It does so by repressing flowering activators, such as SUPPRESSOR OF OVEREXPRESSION OF CO1 (SOC1) and FLOWERING LOCUS T (FT) (Deng et al., 2011;Helliwell, Wood, Robertson, Peacock, & Dennis, 2006;Mateos et al., 2017;Searle et al., 2006).
Prolonged exposure to cold causes epigenetic changes resulting in the stable and irreversible repression of FLC mRNA, which in turn permits flowering to occur after vernalization (Sheldon et al., 2000).
Among A. thaliana accessions, there is variation in the depth and duration of vernalization required for the stable silencing of FLC, and intraspecific variation in FLC expression between accessions is associated with adaptive changes in flowering time, including local adaptation (Ågren, Oakley, Lundemo, & Schemske, 2017;Mendez-Vigo, Pico, Ramiro, Martinez-Zapater, & Alonso-Blanco, 2011;Shindo et al., 2005). In Arabis alpina L.-a perennial herb in the Brassicaceaethe FLC ortholog PERPETUAL FLOWERING 1 (PEP1) also represses flowering prior to vernalization, but the spatiotemporal variation of this repression facilitates A. alpina's perennial life history. After vernalization, PEP1 expression is high in axillary branches to suppress flowering and ensure the return to vegetative development (Lazaro, Obeng-Hinneh, & Albani, 2018;Wang et al., 2009). Different PEP1 alleles have been observed in A. alpina accessions possessing independent mutations that render PEP1 inactive. These mutations result in lack of a vernalization requirement to flower and perpetual flowering (Albani et al., 2012).
Optimal control theory models have been developed to predict the optimal time to initiate flowering, and in the absence of extrinsic factors, an optimal strategy for an annual plant is to allocate all resources to reproduction as soon as current reproductive value is higher than the residual reproductive value (Hirschfield & Tinkle, 1975;Williams, 1966). In perennials, the phases of vegetative growth, flowering and seed production are indistinct, and therefore, maximizing fitness requires understanding the vegetative growth-reproduction trade-off, as well as how early vegetative growth can increase future reproductive opportunities (Iwasa & Cohen, 1989). The expected fitness of future reproduction is therefore determined by intrinsic factors such as the length of flowering duration, the possibility of vegetative reproduction and seed quality.
Seed germination is a second important developmental transition undertaken by plants. The timing of germination is strongly regulated by seed dormancy mechanisms. Seed dormancy is the ability of viable seeds to abstain from germination despite favourable conditions and is an innate quality that is characterized by unresponsiveness to signals that promote germination (Bewley, 1997). Seed dormancy determines fitness, since seeds with shallow dormancy may germinate too early in the growing season, whereas seeds with high dormancy may germinate too late (Finch-Savage & Leubner-Metzger, 2006). Dormancy also allows plants to establish multiyear seed banks and thereby to spread reproductive risk among many seasons (Venable & Brown, 1988). Plant phytohormones such as abscisic acid (ABA) and gibberellic acid (GA) are the primary regulators of seed dormancy. ABA inhibits germination, whereas GA promotes germination, and together, these hormones antagonistically exert control over seed dormancy (Footitt, Douterelo-Soler, Clay, & Finch-Savage, 2011).
Seed longevity denotes the ability to retain viability despite ageing-related stresses such as oxidation and genetic degradation. Longevity is also a critical life history trait, since seed survival is a major factor in the eco-evolutionary dynamics of seed banks.
Predictable decreases in seed viability occur through time, and longevity may also be subject to significant intraspecific variation Sletvold & Agren, 2015). Many factors affect seed longevity, but plants have two main strategies to avoid the loss of seed viability: protection and repair (Sano et al., 2016). ABA may also play an important role in regulating seed longevity; in A. thaliana, abi1-5 mutants, which have compromised endogenous ABA biosynthesis, have lower viability after four years of natural ageing .
Here, we report that PEP1 affects plant and seed mortality as well as seed dormancy in a way that is functionally divergent from the role FLC plays in regulating seed traits in A. thaliana. We demonstrate that pep1 mutants and perpetual flowering accessions show lower seed dormancy and longevity, reduced sensitivity to exogenous ABA and altered expression of genes conferring ABA sensitivity, including ABSCISIC ACID INSENSITIVE (ABI) 3 and 5. To our knowledge, this is the first study demonstrating a pleiotropic link between flowering time genes and seed dormancy or longevity has been found in a perennial species.

| Plant material
Arabis alpina L. (Brassicaceae) is an herbaceous mat-forming perennial present in alpine and subalpine habitats across Europe, North and East Africa, Western Asia and North America (Assefa, Erich, Taberlet, Nemonissa, & Brochmann, 2007;Ehrich et al., 2007). In Europe, A. alpina shows clines of variation in mating system (outcrossing vs. selfing) across much of this range, and seed germination and seedling establishment occur throughout the growing season (Tedder et al. 2015;Torang et al 2015;Laenen et al. 2018). In addition to reproduction through seeds, plants can also propagate clonally by stoloniferous growth (Buehler, Graf, Holderegger, & Gugerli, 2012;Torang et al. 2015). Its perennial life history is characterized by a vegetative phase (during which plants are not able to respond to flowering inductive stimuli) and the ability to keep vegetative growth after flowering. Plants initiate flower buds in response to (and during) prolonged cold and flower very rapidly after snow melt (Lazaro et al., 2018;Torang et al. 2015;Wang et al., 2009) (Albani et al., 2012).
In this study, we used plant material that was characterized by Albani et al. (2012). This material included six natural accessions of A. alpina that were sourced from different mountain habitats throughout Europe by the collectors listed in Supporting information Table   S1 of Albani et al. (2012). These accessions included three seasonal flowering accessions (Paj, Ara and Sty) and three perpetual flowering accessions (Dor, Tot and Wca). As per Albani et al. (2012), the Ara and Paj accessions were sourced from the Valle de Arán (Catalan: Val D'Aran) in the Spanish Pyrenees and the Cantabrian mountains (at an altitude of 1,400 m), respectively. The Dorfertal (Dor) accession originated from the High Tauern National Park (1,650 m) in the Austrian East Tyrol. The Totes Gebirge (Tot), West Carpathian (Wca) and South Tyrol (Sty) accessions were originally sourced from the Totes Gebirge (Austria), West Carpathian (Austria) and South Tyrol (Italy) mountain ranges (1,600 m), respectively. To better examine the effect of PEP1 on flowering and seed traits, we also included the pep1-1 and pep1-2 mutant alleles. These genotypes were derived from a mutagenesis screen of the accession Paj and have also been described previously (Nordström et al., 2013;Wang et al., 2009).

| Plant growth and phenotyping in the greenhouse and garden
Sterilized seeds were stratified (at 4°C for 48 hr), then germinated in filter paper lined Petri plates for four days in a growth chamber providing a 16 hr photoperiod of fluorescent light at 23°C ± 0.1°C.
Following germination, seeds were sown in 9 × 9cm pots. Plants were grown in long days (LD; 16 hr:8 hr day:night) in a climate-controlled greenhouse until ready for vernalization. To avoid precocious flowering of the accessions that do not require vernalization to flower, seasonal accessions were grown for eight weeks, and perpetual genotypes for five weeks prior to vernalization. Subsequently, the plants were split into different groups that will either experience cold in a controlled environment chamber or outside in a garden. gov/gmd/grad/solcalc/) using the latitude and longitude described above. Temperature data were analysed using the WeatherData package in r.
For both greenhouse and garden plants, seeds within the same cohort were harvested at the same time by selecting siliques that were formed within a five-day window. For Greenhouse Experiment 1, we harvested siliques at 73 days after the end of vernalization and for Greenhouse Experiment 1 at 75 days after the end of vernalization. Plants flowered less rapidly in the garden, and therefore, seed sampling was performed 125 days after the last temperature drop below 0°C (i.e., 125 days after 15 March, so seeds were collected on 18-20 July). All seeds from marked siliques were harvested within five days of silique maturation. Because after-ripening is associated with the loss of sensitivity to gibberellic acid (GA), dormancy germination trials and the GA hormone sensitivity trial were begun within four days of seed harvest. Controlled deterioration tests and the paclobutrazol (PBZ) and abscisic acid (ABA) sensitivity trials were begun after after-ripening had permitted dormancy release.

| Assessments of seed traits
Seed dormancy was assessed by conducting germination assays at different time points as seeds after-ripened. We made 14 assessments, once per week from 0 to 13 weeks after seed harvest. All assays were performed using four to six independent biological replicates for each accession; six independent biological replicates were used for Paj and the pep1-1 mutant, and four for all other accessions. Each of these biological replicates was assayed by four technical replicates for each time point tested. For each assay, a sample of 30-80 sterilized seeds was placed on moistened filter paper in 5cm Petri plates and germinated without stratification in a climate chamber programmed for a 25°C:20°C day:night temperature regime with a 12 hr:12 hr day:night photoperiod cycle. Photographs of the plate were taken seven days after the start of germination, and seed counts were performed in triplicate from these photos. The number of days of dry storage (i.e., after-ripening) required for 50% germination (DSDS50) was computed by modelling time to germination using a GLM with a logit link (see: Hurtado et al., 2012), and the DSDS50 was the interpolated point at which the germination rate of the seed sample was 50% ( To assess endogenous differences in seed longevity under controlled conditions, we subjected seeds to artificial ageing using a controlled deterioration test (CDT). To artificially stress seeds, open PCR tubes containing sterilized seeds were placed in an airtight box containing a saturated salt (KCl) solution that fixed the ambient humidity at 83%. Three replicate seed samples were then incubated at 37°C for 0, 1, 2, 3, 4, 6, 8, 10, 12 and 14 days. After incubation, seeds were stratified for 48 hr at 4°C then germinated in a climate chamber programmed for a 25°C:20°C day:night temperature regimen with 12 hr:12 hr day:night photoperiod cycle. Photographs of seeds were produced 10d after germination, and seed counts were performed in triplicate from these photographs. We computed the half-viability period (P50) for all accessions tested, using a method previously described (Nagel et al., 2016). All assays were performed using four to six independent biological replicates for each accession; six independent biological replicates were used for Paj and the pep1-1 mutant, and four for all other accessions. Each biological replicate was assayed by four technical replicates at each time point.
To assess differences in seed sensitivity to exogenous plant hormones, we conducted a series of seed hormone sensitivity trials for assessing seeds from all genotypes and experimental conditions. We conducted a hormone sensitivity trial on seeds collected just after harvest for gibberellic acid (GA4+7), and trials for seeds collected 15 weeks after harvest for abscisic acid (ABA) and paclobutrazol (PBZ). For each trial, samples of 30-80 sterilized seeds from four biological replicates per accession were placed on dry filter paper in 5 cm Petri plates and 1.5 ml of purified water (produced by a Millipore Milli-Q water purification system) containing a given concentration of GA, ABA or PBZ. Seeds were then germinated without stratification in a climate chamber programmed for a 25°C:20°C day:night temperature regime with a 12 hr:12 hr photoperiod cycle for 10 days. All assays were performed using four to six independent biological replicates for each accession; six independent biological replicates were used for Paj and the pep1-1 mutant, and four for pep1-2. Each biological replicate was assayed by four technical replicates at each concentration level for all hormonal treatments.

| Analyses of gene expression
Relative expression of target genes was determined using RT-qPCR.
Total RNA was extracted from dry and 24-hr imbibed seeds using a QIAGEN RNeasy RNA purification kit, and cDNA synthesis was carried out using a QIAGEN Quantitect Reverse Transcription kit. qPCR trials were performed in 96-well plates on a Bio-Rad CFX96 Touch Real-Time PCR system. For each gene of interest, we performed PCR using 150 ng of cDNA and 0.5 μM F + R qPCR primers in 10 μl reaction volumes of master mix from a Bio-Rad IQ SYBR Green Supermix kit. Observed Ct values were standardized using two housekeeping genes known to be stably expressed in seeds, AaKU70 (an ortholog of At1g16970 in A. alpina) and AaWU40 (an ortholog of At2g43770; Dekkers et al., 2012). These genes were selected for their stability and efficiency based on a preliminary assessment of a range of housekeeping genes. Relative expression was determined using the comparative C T method. Mean values were determined using three technical replicates each of four biological replicates, and asymmetrical relative error was computed using the method of Livak and Schmittgen (2001).
We tested the expression levels of a variety of genes that may affect seed traits. Specifically, we tested the expression of PEP1, the A. alpina ortholog of the seed dormancy regulator DELAY OF GERMINATION 1 (AaDOG1) and the A. alpina orthologs of genes involved in hormone biosynthesis or signalling that have been shown to play a role in seed dormancy in A. thaliana. These included genes linked to ABA such as the 9-cis-epoxycarotenoid dioxygenases (AaNCED6 and AaNCED9), the cytochrome P450 CYP707A genes (AaCYP707A1 and AaCYP707A2) and the ABA-INSENSITIVE genes (AaABI2, AaABI3, AaABI4 and AaABI5) but also genes associated with GA such as the gibberellin 2-oxidases (AaGA2ox1, AaGA2ox2, AaGA2ox3, AaGA2ox4, AaGA2ox6 and AaGA2ox8), the gibberellin 3-oxidases (AaGA3ox1 and AaGA3ox2), the gibberellin 20-oxidases (AaGA20ox2 and AaGA20ox4), the protein phosphatase SHEWANELLA-LIKE PROTEIN 2 (AaSLP2) and DELLA RGA-like 2 (AaRGL2). A complete list of the primers used for qPCR is shown in Supporting information Table S1, and the genes referred to in this article can be found in the GenBank/EMBL databases under the accession numbers shown in Supporting information Table S2.

| Statistical analyses
To test whether PEP1 was associated with significant differences in the allocation of reproductive effort, we used ANOVAs to test whether accession groups-that is seasonal accessions (Ara, Paj and Sty), perpetual accessions (Dor, Tot and Wca) and pep1 mutants (pep1-1 and pep1-2)-differed in the proportion of total reproductive effort (i.e., the total number of siliques produced) allocated to the main stalk. Models analysing data from the greenhouse population (which included data from both Replicate Experiments 1 and 2) included one response variable (i.e., the proportion of reproductive effort allocated to the main stalk) and three fixed effects (accession group, replicate-i.e., Replicate Experiment 1 or 2-and their interaction), while models analysing data from the garden population included only accession group as a fixed effect. Where accession group was a significant predictor of the response, we used Tukey HSD tests to conduct pairwise comparisons between accession groups; these HSD tests used false discovery rate (FDR) corrections to ensure that the global alpha remained 0.05. Given the particular usefulness of examining the phenotypic differences between the Paj accession and the pep1-1 and pep1-2 mutants, we conducted similar pairwise comparisons between these particular genotypes for all response variables. We ensured that residual variation was homoscedastic and that all models fit the data by examining residual plots.
The same factorial/one-way ANOVA approach combined with post hoc FDR-corrected pairwise comparisons between means of accession groups and genotypes (i.e., Paj/pep1-1/pep1-2) was used to determine whether accession group was a significant predictor of other flowering and seed traits we measured in both the greenhouse and garden populations, as well as whether significant differences existed between accession group and genotype means. For instance, this approach was used to determine whether accession group was a significant predictor of flowering traits in the experimental garden population, including (a) the mean date of the onset of flowering (measured in days elapsed since 1 January 2016); (b) the mean date of the end of flowering; (c) the duration of flowering; and/or (d) the date of mean flowering (measured as the date on which the 50th percentile flower was produced). We also used this approach to assess differences in survivorship among accession groups in the experimental garden. To analyse seed trait data from the greenhouse and the experimental garden, we used the factorial and one-way ANOVA approaches, respectively, to predict whether accession group was a significant predictor of seed dormancy (i.e., DSDS50) and longevity (i.e., P50). Post hoc tests were used to evaluate the significant differences between accession group and genotype means for these traits. We used slightly different models to analyse germination tests after natural ageing or at different temperature treatments. To determine whether genotype was a significant predictor of differences in germination after natural ageing, we used an ANCOVA model that included genotype (Paj or pep1-1) as a fixed effect and the duration of natural ageing as a covariate (as well as their interaction), to predict differences in seed germination rate. To determine whether genotype was a significant predictor of seed germination rate at different germination temperatures, we used a factorial ANOVA model that included genotype, germination temperature and their interaction as fixed effects. Finally, we used one-way ANOVAs to test whether genotype (i.e., Paj, pep1-1 or pep1-2) was a significant predictor of seed sensitivity to plant hormones, including ABA, GA and paclobutrazol. Separate seed trait tests were conducted on greenhouse and experimental garden data. All statistical analyses were carried out in r version 3.5.0.

| PEP1 influences flowering traits and plant mortality
To assess the differences in reproductive output between plants with and without functional PEP1 alleles, we compared flowering traits among seasonal and perpetual accessions and mutants. Accessions  Table S3). In contrast, pep1-1 and pep1-2 plants produced flowers continuously until the end of the experiment (98 days after EV) and showed a lower proportion of their reproductive effort allocated to the main stalk (pep1-1 mean = 0.29, SE = 0.02; pep1-2 mean = 0.34, SE = 0.02; Supporting information Table S3). Similarly, other accessions in the seasonal accession group (Ara and Sty) showed an early peak in flowering and a higher proportion of reproductive effort on the main inflorescence, while accessions in the perpetual accession group (Dor, Tot and Wca) continued flowering throughout the experiment and showed a lower proportion of reproductive effort on the main inflorescence (Figure 1c, d).
Statistical analysis revealed that accession group was a strong predictor of the total number of flowers produced (Table 1) Table S4). However, mutant genotypes and perpetual accessions also significantly differed in the mean number of flowers produced (Supporting information Table S4). A separate analysis showed that there was no difference in the mean number of seeds produced per silique in Paj (mean = 22.6 seeds, SE = 3.94) and pep1-1 (mean = 22.9, SE = 4.52) plants (Supporting information Figure S1). Next, we examined the proportion of reproductive effort invested in the inflorescence on the main shoot axis compared to axillary branches. An ANOVA revealed that seasonal accessions (mean = 0.77, SE = 0.02) invested a significantly higher proportion of their reproductive effort on the main inflorescence than perpetual accessions (mean = 0.29, SE = 0.01) or mutant genotypes (mean = 0.31, SE = 0.01; Figure 1f, Table 1; Supporting information Table S4). Moreover, we found no significant difference in the proportion of reproductive effort on the main stalk invested by perpetual accessions and mutant genotypes (Supporting information Table S4). However, despite the consistent differences in flowering behaviour between accession groups, there was considerable variation between the accessions and mutant genotypes within accession groups as well (Figure 1g).
In the experimental garden, we also found differences in flowering schedules among accession groups (Figure 2a (Table 2). However, we did not find a significant difference between seasonal and perpetual accession groups (Supporting information Table S4). As in the greenhouse, the seasonal accession group (mean = 0.81, SE = 0.02) invested a higher proportion of reproductive effort on the main inflorescence than the perpetual accession group (mean = 0.45, SE = 0.03) or the mutant group (mean = 0.43, SE = 0.02; Supporting information Table S4). No significant difference in the proportion of reproductive effort on the main inflorescence was detected between the perpetual and mutant groups (Supporting information Table S4) Table S4). However, we found significant differences in the date of mean floweringthat is the date by which 50% of all flowers had been produced  Table S4); however, we found no significant differences in flowering duration between the mutant and perpetual accession groups (Supporting information Table S4). Taken together, these results show that the presence of a functional PEP1 allele is associated with profound differences in flowering behaviour, including a later onset of flowering, a later date of mean flowering and a shorter total duration of flowering.
We also found differences in survivorship between accessions.  Figure   S2).

| Perpetual flowering accessions and pep1 mutants have reduced seed dormancy
We conducted germination trials to examine whether seeds produced by different accessions showed variation in the amount of after-ripening required to release seed dormancy. The number of days of dry storage to reach 50% germination (DSDS50) was determined for each accession, with a higher DSDS50 value indicating higher seed dormancy.
In both the greenhouse and the garden, plants from the seasonal accession group showed higher seed dormancy than plants from the perpetual accession or mutant groups. In seeds matured in greenhouse plants, germination rates immediately after harvest were low for all genotypes. Seasonal accessions required 11-12 weeks of dry storage to reach 100% germination, while perpetual accessions and the pep1 mutants reached 100% germination after only 6-9 weeks of dry storage (Figure 4a Table S4). Pairwise comparisons between Paj and the pep1-1 and pep1-2 mutants also showed that Paj seeds (mean = 61.8; SE = 0.9) had higher DSDS50 values than either mutant (pep1-1 mean = 37.8; SE = 1.1; pep1-2 mean = 25.1; SE = 1.5; Supporting information Table S3). Differences in DSDS50 were not influenced by temperature, since seeds from Paj and pep1-1 showed no difference when germinated at 10°C  Figure S3). Moreover, we found that the mean DSDS50 of the mutant genotype accession group was slightly higher than that of the perpetual accession group (Supporting information Table S4).
In the garden, germination rates upon harvest were considerably higher than in the greenhouse, although the trend among accession groups was similar (Figure 4a, b). Seasonal accessions reached near 100% germination after 6-8 weeks of after-ripening, while perpetual  Hours O c t 1 5 N o v 1 5 D e c 1 5 J a n 1 6 F e b 1 6 M a r 1 6 A p r 1 6 M a y 1 6 J u n 1 6 J u l 1 6 A u g 1 6 S e p 1 6 O c t 1 6 N o v 1 6 D e c 1 6 J a n 1 7 F e b 1 7 M a r 1 7

(d)
O c t 1 5 N o v 1 5 D e c 1 5 J a n 1 6 F e b 1 6 M a r 1 6 A p r 1 6 M a y 1 6 J u n 1 6 J u l 1 6 A u g 1 6 S e p 1 6 O c t 1 6 N o v 1 6 D e c 1 6 J a n 1 7 F e b 1 7 M a r 1 7 O c t 1 5 N o v 1 5 D e c 1 5 J a n 1 6 F e b 1 6 M a r 1 6 A p r 1 6 M a y 1 6 J u n 1 6 J u l 1 6 A u g 1 6 S e p 1 6 O c t 1 6 N o v 1 6 D e c 1 6 J a n 1 7 F e b 1 7 M a r 1 7 O c t 1 5 N o v 1 5 D e c 1 5 J a n 1 6 F e b 1 6 M a r 1 6 A p r 1 6 M a y 1 6 J u n 1 6 J u l 1 6 A u g 1 6 S e p 1 6 O c t 1 6 N o v 1 6 D e c 1 6 J a n 1 7 F e b 1 7 M a r 1 7 0 20 40 60 accessions required 3-5 weeks of after-ripening (Figure 4b, d, f). In the garden, nondormant seeds produced by perpetual accessions were often found to germinate shortly after being shed. Due to the relatively high mortality of pep1-1 and perpetual accession plants in the garden, it was not uncommon to find a cloud of small germinated seeds and small seedlings overgrowing a mother plant (Figure 4g).

| Perpetual flowering accessions and pep1 mutants have low seed longevity
To determine whether seeds produced by different accessions and mutants had altered seed longevity, we used both artificial and natural seed ageing tests. Controlled deterioration tests (CDTs) were used to assess seed survival in response to exposure to high   Table S4). We again found that the P50 values of seeds from Paj plants (mean = 8.8; SE = 0.3) were significantly higher than the P50 values of seeds from pep1-1 (mean = 4.8; SE = 0.2) or pep1-2 plants (mean = 6.8; SE = 0.2).
To corroborate the results of the controlled deterioration tests, we performed germination tests on Paj and pep1-1 seeds that had been naturally aged for one to four years. This test showed that Paj seeds showed significantly higher germinability than pep1-1 seeds after one, two and four years of natural ageing (Figure 5e; Supporting information Table S6). Thus, accession group was a strong predictor of P50 in all environments, with the presence of a functional PEP1 being associated with higher P50 values and thus higher seed longevity.

| pep1 mutants have reduced sensitivity to ABA and reduced expression of genes implicated in ABA signalling
To examine whether PEP1 influences seed sensitivity to phytohormones, we conducted seed germination trials of Paj and pep1 mutants in the presence of varying concentrations of ABA, GA and paclobutrazol (PBZ), an inhibitor that prevents endogenous GA biosynthesis. We found that Paj seeds were more sensitive to exogenous ABA than were pep1-1 and pep1-2 seeds, and the greatest difference in ABA sensitivity between accessions was between Paj  Table S3).

| Expression of genes implicated in ABA signalling in seeds is reduced in pep1-1
We also compared the expression of genes that correlate with seed dormancy in A. thaliana in dry and imbibed pep1-1 and Paj seeds.
Among genes tested, the ABA signalling genes AaABI2, AaABI3 and AaABI5 showed the largest differences between the two genotypes; each of these was more highly expressed in Paj compared to pep1-1 dry and 24 hr-imbibed seeds (Figure 7a-d). In addition, the expression of AaNCED6 and AaNCED9 was higher in pep1-1 seeds than in Paj seeds. In A. thaliana, NCED6 and NCED9 are required for seed development-nced6 and nced9 knockout mutant plants produce less ABA and show reduced seed dormancy (Lefebvre et al., 2006). In addition, impaired expression of ABI2, ABI3, ABI4 and ABI5 in A. thaliana has been linked to ABA insensitivity and reduced seed dormancy (Finkelstein & Somerville, 1990;Kermode, 2005;Lopez-Molina, Mongrand, McLachlin, Chait, & Chua, 2002).
Our results also show that the expression of AaGA2ox6 was upregulated (Supporting information Figure S4) and the expression of AaGA2ox8 was downregulated in pep1-1 relative to Paj (Figure 7).
Although in our data the detected GA2ox enzymes show the opposite expression pattern between the two genotypes, the higher expression of GA2ox enzymes might be associated with increased seed dormancy.

| PEP1 regulates multiple life history traits in the perennial life history of A. alpina
In temperate climates, many species-both annual and perennialflower seasonally in the spring or early summer, concentrating flowering during favourable conditions. Perennials overwinter both as seeds and as mature plants, and to avoid wasting reproductive effort by flowering too late, the flowering episode in perennials is restricted. In A. alpina, PEP1 contributes to perenniality in two waysby preventing axillary branches from undergoing the floral transition and limiting flowering duration (Wang et al., 2009). In this study, we observed differences among genotypes with and without lesions in PEP1 for each of these traits, both in plants reared in the greenhouse and in the garden. The intrinsic and environmental factors governing the decision to invest resources in a present reproductive episode instead of a future one (or vice versa) are complex, but previous work has shown that the primary resource trade-off concerns vegetative development and reproduction (Lloyd, 1980).
We found that the perpetual accession and pep1 mutant accession groups realized higher reproductive effort and emphasized perennial traits to a lesser degree than did genotypes in the seasonal accession group (which included Paj). Seasonal accessions in both greenhouse and garden environments invested a higher proportion of their reproductive effort on the main stalk compared to axillary branches. In contrast, most axillary branches in perpetual accessions flowered and produced seeds. In the garden, we also observed strong differences in plant mortality; perpetual genotypes (and the pep1-1 mutant) were less likely to survive periods of winter cold than seasonal accessions. We can therefore hypothesize that winter cold influences perpetual accessions to a greater degree than seasonal ones. This difference may be due to direct effects of PEP1, which has been shown to regulate cold response by binding to cold-regulated genes (COR; Mateos et al., 2017) or by differences in intrinsic resource allocation schedules between seasonal and perpetual genotypes. Under greenhouse conditions, the pep1-1 mutant produced more flowers than Paj (Figure 1e; Supporting information Table S3) and did not complete flowering by the end of the experiment. Since we found no differences in the mean number of seeds per silique (Supporting information Figure S2), genotypes in the perpetual accession group are likely capable of producing many more seeds than genotypes in the seasonal accession group, at least in the first year.
A strategy involving high reproductive effort early in life at the cost of survival or future reproduction is characteristic of annuals.
However, in contrast to the behaviour of perpetual accessions of A.
alpina, annuals in predictable, seasonal environments die after reproduction and therefore experience strong selection to complete reproduction quickly. In addition, annuals experience strong selection to establish a multiyear seed bank by producing seeds that can remain dormant. In both the greenhouse and the experimental garden, the reproductive behaviour of perpetual accessions resembles a compromised perenniality rather than true annuality. However, our data were collected in an experimental garden, and the adaptive sig-

| The role of PEP1 on seed traits has diverged from FLC
In A. thaliana, there is strong evidence that the floral repressor FLC also affects seed traits. A study by Chiang et al. (2009) demonstrated that higher expression of FLC was linked to low dormancy and found that the effect of FLC appeared to be temperature-dependent. That is, differences in germination rates were apparent only when seeds were germinated in cool (10°C) conditions. In addition, data from 52 accessions also revealed a positive relationship between FLC expression and seed germination at 10°C but not at 22°C. Consistent with this, mutants in autonomouspathway genes that ordinarily repress FLC (i.e., FCA, FY and FPA) show reduced germination (Auge, Blair, Karediya, & Donohue, 2017;. However, the role of FLC in germination and dormancy can vary, possibly depending on experimental or environmental conditions. For instance, seeds from the flc-101 mutant and those produced by Col-0 wild-type plants showed similar dormancy levels (Liu et al., 2011). Furthermore, Chen et al. (2014) found that seeds from the flc-21 mutant showed lower dormancy than did those from the Ler wild-type background. This was observed when seeds were germinated both at 22 and 10°C. Blair et al. (2017) argued that since the temperature range facilitating seed germination itself depends on dormancy, the temperature at which FLC regulates germination may also vary with dormancy. In this study, our results indicate that the role of  Tables S3 and S4). These differences were consistent regardless of whether seed maturation occurred in the greenhouse or the garden (Figures 4, 5). In addition, while we found differences in germination fraction between pep1-1 (mean germination fraction = 0.20; SE = 0.06) and Paj seeds (mean = 0.17; SE = 0.04) germinated at 10°C, the magnitude of these differences was no greater than those at 23°C (Supporting information Figure   S3, Table S7).
Seasonal flowering accessions also consistently showed higher seed longevity. These results differ from previous studies of A.
thaliana, where seeds with high seed dormancy were found to have low seed longevity .
Furthermore, while in general seeds produced by garden plants consistently showed reduced seed dormancy relative to seeds produced by greenhouse plants, the differences in seed dormancy and longevity between Paj and pep1 mutants-and between seasonal and perpetual accessions overall-were clear in all cases, suggesting a genetic contribution to these differences. In A. thaliana, differences in seed longevity between FLC mutants have not  (Michaels, He, Scortecci, & Amasino, 2003). Both tests revealed that Col had significantly higher seed longevity than Ler. This suggests that FLC expression is associated with higher seed longevity, which agrees with the results we report here for A. alpina. However, Ler and Col-0 plants also have other genetic differences which may contribute to differences in seed longevity. Natural variation studies have identified QTLs that map to FLC that are also associated with traits that have been linked to variation in seed longevity, such as circadian period length and water use efficiency (Clerkx, El-Lithy et al., 2004;McKay et al., 2003;Swarup et al., 1999). In general, the mechanism by which FLC may affect seed longevity in A. thaliana is unknown and requires further investigation.
The differences in gene regulation between Paj and pep1-1 differ from the ones reported in studies of FLC and seed dormancy. In A. thaliana, FLC expression is positively correlated with expression of the ABA catabolism gene CYP707A2, as well as the GA biosynthesis genes GA20ox1 and GA3ox1 (Chiang et al., 2009). In general, the actions of these genes increase the concentration of bioavailable GA and reduce the concentration of ABA, and Chiang et al.
(2009) speculated that FLC was associated with lower seed dormancy due to altered regulation of GA biosynthesis or signalling.
In A. alpina, we did not find differences in sensitivity to exogenous paclobutrazol or GA between pep1-1 and Paj seeds, whereas we did find differences in sensitivity to exogenous ABA. We also found no differences in CYP707A2, GA20ox1 or GA3ox1 expression between Paj and the pep1-1 mutant. Instead, GA2ox6 and GA3ox2 were upregulated and GA2ox8 was downregulated in pep1-1 relative to Paj ( Figure 7; Supporting information Figure S4). In a recent ChIP-seq comparison of the targets of FLC and PEP1 in leaves and apices in A.
alpina, Mateos et al. (2017) reported that AaGA3ox2 and AaGA2ox8 were direct targets of PEP1, but the corresponding orthologs were not direct targets of FLC. In addition, this study also reported that AaABI5 was a direct target of PEP1, but that ABI5 was not a direct target of FLC. Here, we found that genes involved in ABA sensitivity-including NCED6, NCED9, ABI2, ABI3, ABI4 and ABI5-were differentially expressed between Paj and pep1 seeds. Two MADS-box transcription factors similar to PEP1-AGAMOUS-LIKE-21 (AGL21) and AGL67-affect seed traits in A. thaliana, and AGL67 in particular is seed-specific and has also been linked to altered ABA signalling (Bassel et al., 2011). Because ABA and GA biosynthesis and signalling are regulated by complex networks of feedback regulation, we can hypothesize that differential expression of GA-related genes is due to knock-on effects from differential expression of ABA-related genes or vice versa. thaliana have demonstrated that higher seed dormancy is associated with lower seed longevity (Nguyen and Bentsink (2015), Rajjou & Debeaujon, 2008, Shen et al., 2018. However, other studies have shown that there is no relationship between these traits (Debeaujon, Leon-Kloosterziel, & Koornneef, 2000;Thompson, Ceriani, Bakker, & Bekker, 2003 perpetual accessions in the field should be conducted to test these hypotheses in a natural environment. In addition, differences in seed longevity may permit seasonal accessions to distribute reproductive risk over a longer period of time-that is among years by mother plants versus by dormant seeds.
In A. thaliana, there is evidence that intraspecific variation at even a single locus may result in substantial life history changes. For example, the bHLH transcription factor SPATULA (SPT) controls the germination response to cold and light by integrating light and temperature signalling in the seed (Penfield et al., 2005). Interestingly, there is evidence that SPT alleles vary among A. thaliana accessions, thereby giving rise to a range of possible seed responses to chilling; this in turn can facilitate the production of a variety of life history strategies, including extreme summer annual and winter annual strategies (Penfield & Springthorpe, 2012;Springthorpe & Penfield, 2015

| CON CLUS I ON AND OUTLOOK
Identifying the conditions in which an annual life history can obtain a demographic advantage over a perennial one has been a major project in life history theory (Hughes, 2017;Hughes & Simons, 2014).
However, few models explicitly consider the role played by genetic elements in the regulation of life history traits, or the degree to which trade-offs can be constrained by pleiotropy. In A. alpina, previous studies have shown that genotypes with lesions in PEP1 show compromised perenniality and continuous flowering.
However, in this work we report that these genotypes-as well as the pep1 mutants-also show reduced seed dormancy and longevity.
Thus, PEP1 regulates multiple stages of the A. alpina life cycle and facilitates perenniality both by limiting reproductive allocation and by preventing precocious seed germination. In the controlled environment in which we compared the relative fecundity and mortality of different natural accessions of A. alpina, perpetual accessions invested more in seed production but produced seeds that germinated quickly. Moreover, these accessions appeared to compensate for high winter mortality by producing more seedlings. We speculate that since fitness optima vary among years and environments, the pleiotropic effects of flowering time regulators such as PEP1 may prevent the simple, independent optimization of life history strategies. However, the adaptive significance of PEP1's pleiotropic effect on flowering and seed traits in natural environments should be assessed by a field experiment where measures of lifetime fitness can be directly compared.
Pleiotropic regulation of different major developmental transitions affecting different life history stages is not predicted by theoretical life history models, which generally consider the optimization of the timing of such transitions to be distinct questions, and therefore assume that they can be optimized independently (Young, 1981). For instance, annual-semelparous and perennialiteroparous life history strategies are thought to be adaptations to divergent adult and juvenile mortality rates. However, where pleiotropic regulation prevents the independent optimization of traits contributing to adult and juvenile fitness, the assumption that this trade-off exists may be too simplistic. In annual species, differences in the timing of the initiation of flowering have been noted to have important knock-on effects on the time in which seeds are produced, and hence on the optimal degree of dormancy that seeds should have. Thus, life history models of Arabidopsis have suggested that flowering time optimization may constrain the range of available seed environments and sustain life history strategies (Springthorpe & Penfield, 2015). Perennial species face a complex set of trade-offs that annuals do not face because they must invest in growth and reproduction in successive years.
However, models of optimal perennial reproduction emphasize the sensitivity of evolutionarily stable strategies for perennials to the values of age-specific traits including juvenile and adult survivorship, reproductive rates and growth rates (Iwasa & Cohen, 1989;Kozlowski & Wiegert, 1987;Pugliese & Kozlowski, 1990;Wang, Li, & Wang, 2016). Consequently, genetic or developmental constraints-such as pleiotropy-that prevent the independent optimization of vegetative and reproductive traits may have even more profound evolutionary implications for perennials than for annuals. Thus, understanding the nuanced role of major hub regulators such as PEP1 on fitness traits at multiple life stages may be specifically important for understanding natural variation in life history strategies in wild populations of perennial species. Maria C. Albani https://orcid.org/0000-0002-8215-020X