The influence of experimentally induced polyploidy on the relationships between endopolyploidy and plant function in Arabidopsis thaliana

Abstract Whole genome duplication, leading to polyploidy and endopolyploidy, occurs in all domains and kingdoms and is especially prevalent in vascular plants. Both polyploidy and endopolyploidy increase cell size, but it is unclear whether both processes have similar effects on plant morphology and function, or whether polyploidy influences the magnitude of endopolyploidy. To address these gaps in knowledge, fifty‐five geographically separated diploid accessions of Arabidopsis thaliana that span a gradient of endopolyploidy were experimentally manipulated to induce polyploidy. Both the diploids and artificially induced tetraploids were grown in a common greenhouse environment and evaluated with respect to nine reproductive and vegetative characteristics. Induced polyploidy decreased leaf endopolyploidy and stem endopolyploidy along with specific leaf area and stem height, but increased days to bolting, leaf size, leaf dry mass, and leaf water content. Phenotypic responses to induced polyploidy varied significantly among accessions but this did not affect the relationship between phenotypic traits and endopolyploidy. Our results provide experimental support for a trade‐off between induced polyploidy and endopolyploidy, which caused induced polyploids to have lower endopolyploidy than diploids. Though polyploidy did not influence the relationship between endopolyploidy and plant traits, phenotypic responses to experimental genome duplication could not be easily predicted because of strong cytotype by accession interactions.

investigating the functional significance of WGD and its effects on phenotype and fitness.
One of the most common effects of WGD is that it often increases cell size (del Pozo & Ramirez-Parra, 2015;Melaragno, Mehrotra, & Coleman, 1993;Otto & Whitton, 2000;te Beest et al., 2012). This effect causes the surface area:volume ratio of a cell to decrease which can have consequences at the tissue and whole-organism level (Bennett, 1971(Bennett, , 1972Levin, 1983;Otto & Whitton, 2000). For example, tissue composed of larger cells with low surface area:volume ratios should have higher net volume of intracellular storage space and lower net volume of cell wall than an equal volume of tissue comprised of smaller cells. This trade-off between intracellular storage and cellular wall could affect a tissue or organism in many ways, altering an organism's function and response to environment. For example, WGD might be favoured in environments characterized by drought, where enhanced water storage capacity relative to biomass investment could facilitate survival through periods of water limitation (De Rocher, Harkins, Galbraith, & Bohnert, 1990;Schwinning & Ehleringer, 2001). However, such an adaptation would come at the expense of greater diffusion barriers to gas exchange, which could limit photosynthesis and offset increased water storage with reduced water use efficiency while also causing weakened structural support (Corneillie et al., 2019;Niklas, 1992Niklas, , 1994. Whole genome duplication that affects every cell in an organism (including gametes) is called polyploidy. Polyploidy can result from duplication of genomes in an interspecific hybrid (allopolyploidy) or of a single species (autopolyploidy) (Comai, 2005;del Pozo & Ramirez-Parra, 2015;Levin, 1983;Otto & Whitton, 2000;Soltis, Buggs, Doyle, & Soltis, 2010;te Beest et al., 2012). Autopolyploidy is a useful system for studying the phenotypic effects of polyploidy because autopolyploidy primarily reflects changes in genome size, whereas allopolyploids exhibit impacts of hybridization as well as polyploidy (Comai, 2005;del Pozo & Ramirez-Parra, 2015;Levin, 1983;Otto & Whitton, 2000;Soltis et al., 2010;te Beest et al., 2012).
The effects of autopolyploidy on phenotype reflect the downstream impacts of increased cell size (a nucleotypic effect that is independent of the informational content of the genome; Bennett, 1971Bennett, , 1972 plus any additional genetic effects of increased gene copy number, gene expression, and postduplication evolutionary change (Comai, 2005;del Pozo & Ramirez-Parra, 2015;Levin, 1983;Otto & Whitton, 2000;Soltis et al., 2010;te Beest et al., 2012). Indeed, many of the phenotypic changes seen in polyploid plants may be caused by increased cell size relative to cell number (although a compensation effect where cell number decreases to offset increased cell size can also occur; see Hisanaga, Kawade, & Tsukaya, 2015) which could also explain why polyploid plants often exhibit larger body, organ, and leaf size than their progenitors (Comai, 2005; Pozo & Ramirez-Parra, 2015; Müntzing, 1936;Levin, 1983;Otto & Whitton, 2000;Soltis et al., 2010;te Beest et al., 2012).
In contrast to polyploidy, endopolyploidy refers to WGD events that occur within some somatic cells and tissues of a given individual. Similar to polyploidy, endopolyploidy often increases cell size (although this can depend on tissue type; see Katagiri et al., 2016), which may affect phenotypes in the same way as polyploidy.
However, unlike polyploidy, the degree of endopolyploidy in a particular accession can vary with environmental stimuli (Jovtchev, Barow, Meister, & Schubert, 2007;Scholes & Paige, 2015). This plasticity allows endopolyploid cells and tissues to produce a range of growth-related phenotypes in response to prevailing environmental conditions. Endopolyploidy can also be genetically determined and vary among accessions from different environments (for an excellent review of endopolyploidy in seed plants see Barow, 2006;Gegas et al., 2014;Pacey, 2018).
Comparative analyses among species suggest that a trade-off may occur between polyploidy and endopolyploidy. For example, species with small genomes or low polyploidy are more likely to exhibit high levels of endopolyploidy, whereas species with large genomes or high polyploidy often have little or no endopolyploidy (Bainard, Bainard, Henry, Fazekas, & Newmaster, 2012;Barow & Meister, 2003;De Rocher et al., 1990;Nagl, 1976). These patterns suggest there are developmental or structural constraints on maximum cell size for species with large genomes or high polyploidy (Barow, 2006;De Rocher et al., 1990). However, these comparisons among species cannot be used to infer whether polyploidy actually causes species with large genomes to have reduced endopolyploidy.
Experimental studies that manipulate polyploidy are required to determine if increased polyploidy causes a reduction in endopolyploidy.
Empirical studies of the effects of induced autopolyploidy on the expression of endopolyploidy have variable results. For example, in Datura stramonium, Hyoscyamus niger and Portulaca grandiflora, synthetic autopolyploids have decreased endopolyploidy compared to diploids (Mishiba & Mii, 2000;Weber, Georgiev, Pavlov, & Bley, 2008). By contrast, the degree of endopolyploidy in A. thaliana appears to be insensitive to induced autopolyploidy (del Pozo & Ramirez-Parra, 2014). Jovtchev et al. (2007) observed lower endopolyploidy in natural A. thaliana tetraploids compared to diploids; however, the tetraploids and diploids were from different accessions, and thus polyploidy could not be determined as the cause of reduced endopolyploidy. The apparent insensitivity of A. thaliana endopolyploidy to experimentally induced autopolyploidy could be due to its relatively small genome size (Schmuths, Meister, Horres, & Bachmann, 2004). Alternatively, past studies have been based on only two accessions (Col-0 and Ler); hence the results to date may not reflect the general effects of autopolyploidy on endopolyploidy in this species (del Pozo & Ramirez-Parra, 2014;Rédei, 1992). It is not known if endopolyploidy in naturally occurring accessions would respond differently to induced autopolyploidy. Because A. thaliana is known to show extensive natural variation for the degree of endopolyploidy, it could serve as an important model system to make general inferences about how inducing autopolyploidy influences the expression of endopolyploidy (Gegas et al., 2014;Pacey, 2018).
To explore the relationship between induced polyploidy and endopolyploidy, we examined the degree of endopolyploidy in natural diploid and synthesized autotetraploid accessions of A. thaliana ( Figure 1). Though A. thaliana has experienced at least two ancient WGD events, polyploidy has been lost over evolutionary time in this species, and nearly all wild populations function as diploids (del Pozo & Ramirez-Parra, 2015). We used a common garden experiment to address the following two questions. Does induced polyploidy decrease endopolyploidy in natural accessions of A. thaliana? Does the degree of endopolyploidy affect growth and reproductive traits differently in diploids versus tetraploids? We hypothesize that induced polyploidy decreases endopolyploidy in natural accessions of A.
thaliana because there are developmental or structural constraints on maximum cell size. We also hypothesize that the degree of endopolyploidy will not affect growth and reproductive traits differently in diploids versus tetraploids because endopolyploidy should influence cell size independently of polyploidy.

| Source populations and polyploid synthesis
To explore how induced polyploidy affects endopolyploidy and growth/reproductive traits in A. thaliana, seed families from 55 randomly selected geo-referenced diploid accessions of A. thaliana that span its native geographic range (collected as part of the 1,001 genomes project and made available through the Arabidopsis Biological Resource Center) and are facultative in their vernalization requirement were used (Pigliucci, 2002;Weigel & Mott, 2009; Appendix Table A1). Seeds from each accession were germinated on an agarose medium, bathed in colchicine (0.5% for 1-2 hr) to induce polyploidy and then transplanted to soil (Yu, Haage, Streit, Gierl, & Torres Ruiz, 2009;A. Green & B. C. Husband, unpublished). Inflorescences on plants from colchicine-treated seeds were checked for cytotype using flow cytometry and seeds were collected from each inflorescence separately (to prevent mixing of seeds from other inflorescences and parents) and then stored in their own individual vial (A. Green & B. C. Husband, unpublished). These seeds, which were one generation removed from the colchicine treatment, were unlikely to show unintended effects of colchicine (Münzbergová, 2017). To confirm this, a previous study compared vegetative and reproductive traits between colchicine-treated and -untreated diploids and found no difference (A. Green & B. C. Husband, unpublished). To minimize this concern further, in this study, we used diploid seed that had received a colchicine treatment so any observed differences between ploidy levels were not confounded with colchicine treatment.

| Growth conditions and experimental design
To overcome seed dormancy and synchronize germination, diploid and tetraploid seeds of each accession were stratified at 3°C in the dark for 72 hr in Parafilm ® enclosed petri dishes containing moist filter paper. Three germinating seeds (that came from the same inflorescence seed vial) for each accession and cytotype replicate were placed on the soil of a single 7.3 cm deep, 377 cm 3 dark green pot (KORD Products) containing Sunshine Mix #4 (Sun Gro Horticulture).
Plants were thinned to one per pot when true leaves developed (the healthiest seedling by appearance was kept). All pots were randomly placed in eight-liter white trays, which each hold up to 18 pots (ITML Horticulture Products Inc.), watered weekly with 18-9-18 fertilizer mixed at a rate of 200 ppm until trays overflowed and then allowed to fully drain (which required 10 min). Trays were also watered (without fertilizer) every 2 days after day 14 postpotting until they overflowed and then allowed to fully drain (which required 10 min).
Plants were grown in a controlled environment (23°C day, 20°C night, 16 hr daylight) at the University of Guelph Phytotron greenhouse. Supplemental lighting using 600 W high-pressure sodium light fixtures (P.L. Light Systems Inc.) with SON-T bulbs (Philips) that delivered ~300 µmol/m 2 /s at greenhouse bench level were used when natural sunlight was inadequate (≤400 µmol/m 2 /s) during the assigned 16 hr photoperiod. Plants were grown in two groups of three randomized temporal blocks (the first group of three blocks was grown 1 week apart in November 2015 while the second group of three blocks was grown 1 week apart in June 2016 using the same methods) with each block containing one individual from each accession and cytotype. Each plant from the November 2015 group of three blocks was measured for day of bolting (when the primary stem first emerged), F I G U R E 1 Organism photo of an experimentally induced tetraploid (4x; right) beside its natural diploid (2x; left) progenitor Arabidopsis thaliana accession (Stw-0, CS1538) leaf endopolyploidy index (EI), stem EI, and stem height (single block N = 110, three combined blocks from November 2015 N = 330 with N = 3 for each accession and cytotype). Due to the destructive nature of flow cytometry, the second group of three temporal blocks (grown in June 2016) was needed and used to measure leaf size, leaf dry mass, specific leaf area (SLA), leaf water content, and chlorophyll concentration (single block N = 110, three combined blocks from June 2016 N = 330 with N = 3 for each accession and cytotype).

| Flow cytometry and endopolyploidy index
To estimate the degree of endopolyploidy in each individual, the largest rosette leaf was harvested on the first day of bolting (when the primary stem first emerged). The primary stem from these same individuals was harvested on the first day of anthesis (or fruiting if anthesis was not clearly visible) and its length was measured. To prepare leaf tissue for flow cytometry, the right half of the leaf blade (when the abaxial epidermis was facing upwards and the petiole was pointed toward the researcher) was cut off with a razorblade without including its midrib (to reduce tissue heterogeneity). To prepare stem tissue, all petioles, cauline leaves and flowers (to reduce tissue heterogeneity) were removed. Nuclei from leaf and stem tissue were then isolated separately by finely chopping with a fresh razor blade in Galbraith's buffer (Galbraith et al., 1983). Propidium iodide (100 µg/ml) and RNAse (0.5 µg/ml) were added to the buffer to stain DNA and degrade interfering RNA, respectively. The mixture of tissue and buffer was filtered (30 µm filters; Partec GmbH) to remove tissue fragments and allowed to stain for 20-60 min. samples and tissues. The 2C peak was prominent and readily detected in stem tissue, and thus was used as a marker to confirm the location of the 2C peak in leaf tissue for each accession. At least 5,000 nuclei were counted for leaf and stem samples from each individual plant and the ploidy identified from their fluorescence peaks (representing 2C, 4C, 8C etc. values and ranging from 2C to 2,048C as indicated by their respective flow cytometry histograms; Figure 3). The extent of endopolyploidy in each tissue was then calculated using the cycle value or Endopolyploidy Index (EI) equation (Barow & Meister, 2003): where n #c refers to the number of nuclei of the # C value and where # ranges from 2, 4, 8, 16 to the maximum detected. Note that EI is a single value measure of the number of rounds of endopolyploidy in a sample over its base genome size (2C). The 2C peak of a diploid is 2x (C = amount of DNA within an unreplicated gametic genome, x = number of sets of chromosomes) while the 2C peak of a tetraploid is 4x ( Figure 3). Aside from this difference (which is not included in the calculation), the EI calculation is the same whether it is a diploid or tetraploid plant. Therefore, the lack of a diploid measure in tetraploids does not affect the EI calculation and makes their EI comparable to the EI of a diploid.
The EI was used over other measures of endopolyploidy due to its simplicity. Reporting nuclei number for each ploidy level (2C, 4C, 8C etc.) provides more detailed information on the distribution of ploidy; however, it is more cumbersome, statistically, to evaluate variation in endopolyploidy (Gegas et al., 2014). Using the mean C-value of an individual sample provides a single continuous value; however, it places an overemphasis on higher ploidy levels because of the exponential nature of increasing ploidy (Barow & Meister, 2003). Note that if two samples with different combinations of endopolyploidy gave the same EI, they were considered to be equivalent in their degree of endopolyploidy regardless of whether they were functionally equivalent.

| Plant functional traits
To compare the relationships between endopolyploidy and plant functional traits between diploids and induced tetraploids, we measured days to bolting, leaf EI, stem EI, and stem height in the first set of temporal blocks. The second set of temporal blocks was used to measure leaf size, leaf dry mass, leaf water content, leaf chlorophyll concentration, and SLA on the largest rosette leaf on the first day of bolting. Specific leaf area was included because it is considered a key plant functional trait and has been hypothesized to be higher in leaves with larger cells than equally sized leaves with smaller cells because larger-celled leaves should have relatively less cellular wall mass due to their change in cellular surface area:volume ratio (Shipley, Lechowicz, Wright, & Reich, 2006;Wright et al., 2004).
Apparent chlorophyll content was taken as the average of SPAD meter (SPAD-502, Minolta Camera Co. Ltd.) measurements at the base, middle, and tip of the leaf (Ling, Huang, & Jarvis, 2011). Leaf size (cm 2 ) was measured using a leaf area meter (LI-3100, LI-COR Inc.). Leaves were then weighed, dried in a drying oven at 60°C for 48 hr, and then reweighed to obtain dry mass (g). Leaf water content (g) was calculated as the wet mass minus the dry mass divided by the dry mass. Specific leaf area (cm 2 /g) was calculated as the leaf area divided by dry mass.

| Statistical analyses
To test the hypothesis that induced polyploidy decreases endopolyploidy in natural accessions of A. thaliana, we used a two-way ANOVA of individual plant values with accession, cytotype, accession * cytotype, and block as fixed factors. The coefficient of variation (CV) for traits was calculated by dividing their standard deviation by their mean. To test the hypothesis that the regression slope between endopolyploidy and each growth/reproductive trait will be equal for diploids and tetraploids, we used an ANCOVA on accession means with cytotype as a fixed factor and EI as a covariate.
We tested for homogeneity of slopes by looking at the interaction between cytotype and EI, where a nonsignificant effect means that the slopes are not statistically different. Leaf dry mass, leaf size, chlorophyll concentration, and days to bolting violated the ANOVA and ANCOVA assumptions of normality and equality of variances and hence were log 10 transformed. All statistical analyses were F I G U R E 2 Flow cytometry manual gating examples for a single diploid (2x) Arabidopsis thaliana plant. (a) Leaf debris gating, (b) leaf nuclei cluster gating, (c) stem debris gating, (d) stem nuclei cluster gating. The analysis software used only allowed for nine nuclei cluster gates so 1024C and 2048C nuclei were counted manually and added to total gated events F I G U R E 3 Flow cytometry histograms of low, medium and high endopolyploidy index (EI) Arabidopsis thaliana leaves that span diploid (2x) and tetraploid (4x) cytotypes. X-axes represent log FL3-Height (670 nm) fluorescence while Y-axes represent the number of individual nuclei detected for each ploidy level (2C, 4C, 8C etc). Note that the 2C peak of a diploid is 2x (4C = 4x, 8C = 8x etc.) while the 2C peak of a tetraploid is 4x (4C = 8x, 8C = 16x etc.) completed with SPSS 24 (IBM) and graphed with Sigma Plot 12.5 (Systat software Inc.).  Table 1).
The effect of induced polyploidy on leaf and stem EI was uniform among accessions (i.e., no cytotype by accession interaction), decreasing mean values in most cases (Figure 4a,b, Table 1, Appendix   Table A1). However, accessions responded differently to induced polyploidy in five other traits measured (Figures 4 and 5, Table 1).
Induced polyploidy increased days to bolting by up to 71% in one accession and decreased it by as much as 4% in another ( Figure 4c, Table 1). Stem height increased by up to 39% and decreased by as much as 78% ( Figure 4d, Table 1), and leaf dry mass increased by up to 134% and decreased by as much as 26%, depending on accession ( Figure 5a, Table 1). Induced polyploidy increased leaf size up to 125% and decreased it by 35% ( Figure 5d, Table 1), and SLA increased by 30% and decreased by 29%, depending on accession ( Figure 5e, Table 1). The

| D ISCUSS I ON
We found evidence for a trade-off between induced polyploidy and endopolyploidy in natural accessions of A. thaliana. Our hypothesis that induced polyploidy would decrease the degree of endopolyploidy was supported; specifically, induced polyploidy lowered endopolyploidy in leaf and stem tissue by 15% and 18%, respectively (Figure 4a,b,  Figure 5d).
Induced polyploidy did not influence the relationship between endopolyploidy and growth/reproductive traits (Figures 6 and 7,  Galbraith, Harkins, & Knapp, 1991;Yang & Loh, 2004). Systemic endopolyploidy in A. thaliana appears to be associated primarily with growth, as increasing endopolyploidy results in increased cell size and consequently organ size (Cookson, Radziejwoski, & Granier, 2006;Galbraith et al., 1991;Melaragno et al., 1993).   Beest et al., 2012). That days to bolting was the functional trait that was most affected by induced polyploidy (increasing by 19%) is con-

sistent with this explanation and may indicate that generation time in
A. thaliana is particularly susceptible to changes in base genome size ( Figure 4c, Table 1).
Our study provided novel insights about how induced polyploidy and endopolyploidy interact. It provides experimental evidence to support the hypothesis that inducing polyploidy decreases the degree of endopolyploidy (Figure 4a,b, Table 1, Appendix Table A1).
Moreover, it shows that this trade-off may be influenced by the

This work was supported by a Natural Science and Engineering
Research Council (NSERC) Discovery Grant to H. Maherali and B.
C. Husband. We thank A. Green for creating the induced tetraploid lines, and P. Kron and M. Alie for help with flow cytometry analysis.
We also thank M. Mucci, T. Slimmon and S. Couling for help in the Phytotron.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest to declare. Note: The 2C peak of a diploid is 2x (4C = 4x, 8C = 8x etc.) while the 2C peak of a tetraploid is 4x (4C = 8x, 8C = 16x etc.). N = 2-3 for each specific accession and cytotype, N = 1 for 4x Sij-1.