Phenotypic diploidization in plant functional traits uncovered by synthetic neopolyploids in Dianthus broteri

Abstract Whole-genome duplication and post-polyploidization genome downsizing play key roles in the evolution of land plants; however, the impact of genomic diploidization on functional traits still remains poorly understood. Using Dianthus broteri as a model, we compared the ecophysiological behaviour of colchicine-induced neotetraploids (4xNeo) to diploids (2x) and naturally occurring tetraploids (4xNat). Leaf gas-exchange and chlorophyll fluorescence analyses were performed in order to asses to what extent post-polyploidization evolutionary processes have affected 4xNat. Genomic diploidization and phenotypic novelty were evident. Distinct patterns of variation revealed that post-polyploidization processes altered the phenotypic shifts directly mediated by genome doubling. The photosynthetic phenotype was affected in several ways but the main effect was phenotypic diploidization (i.e. 2x and 4xNat were closer to each other than to 4xNeo). Overall, our results show the potential benefits of considering experimentally synthetized versus naturally established polyploids when exploring the role of polyploidization in promoting functional divergence.


Introduction
Polyploidy, the genomic condition resulting from wholegenome duplication (WGD), is a widespread phenomenon that is considered to be a major driving force in evolution and diversification of flowering plants (Chen, 2007;Soltis et al., 2009;Tank et al., 2015;Landis et al., 2018). Although WGD is usually accompanied by hybridization (i.e. allopolyploidy), pure WGD (i.e. autopolyploidy) per se can induce immediate changes at the cellular and phenotypic levels as a result of the dramatic changing of the relationship between gene copy number and other cellular components (Doyle and Coate, 2019). Newly formed polyploids (hereafter referred to as neopolyploids) usually show an increase of cell size in concert with the DNA content ('nucleotypic effect'; Stebbins, 1971) and hence produce bigger organs (te Beest et al., 2012). In addition, changes in cell volume associated with WGD may promote phenotypic novelty (Soltis et al., 2014;Wendel, 2015). The most common phenotypic changes in neopolyploid plants include larger stomatal size, thicker leaf morphology, and more chloroplasts within each guard cell, which can modify physiological traits such as transpiration, and photosynthetic and growth rates (Masterson, 1994;Levin, 2002;Maherali et al., 2009;Liu et al., 2011).
Following WGD, polyploids start an adaptive process that can be divided into two evolutionary time-frames (Dodsworth et al., 2016). In the short term, neopolyploids need to survive and to establish by overcoming several intrinsic disadvantages (such as 'genomic shock' and aberrant meiosis; Comai, 2005), as well as extrinsic ones (i.e. minority cytotype exclusion, a frequency-dependent mating disadvantage; Levin, 1975). In this initial stage, polyploids need to outcompete their parents or ecologically differentiate from them. Hence, the adaptive shifts required for the establishment of neopolyploids (Soltis et al., 2014;Glick et al., 2016) usually bring them to inhabit new niches, especially under unstable environmental conditions (te Beest et al., 2012;López-Jurado et al., 2020). Polyploidymediated changes in physiological and ecological tolerances can enhance selective advantages and lead to the initial short-term survival and establishment of polyploids (Pandit et al., 2011;Soltis et al., 2014;Münzbergová and Haisel, 2019;Rice et al., 2019). In spite of the importance of ecophysiological aspects of niche divergence in autopolyploids, actual supporting evidence is scarce and is usually focused on drought tolerance (Maherali et al., 2009;Soltis et al., 2010;Liu et al., 2011;Spoelhof et al., 2017). Furthermore, the role of photosynthetic-related shifts in the establishment of polyploids has largely been overlooked (but see Vyas et al., 2007 andLópez-Jurado et al., 2020).
Once a polyploid has been effectively established, a longer-term process of evolution begins as it becomes a quasidiploid, through the diploidization process. The most common phenomenon during diploidization is genome downsizing caused by the loss of duplicate genes (fractionation) and small genomic fragments (Cheng et al., 2018), which might be noticeable when comparing the monoploid genome size (1Cx) among different ploidy levels (Balao et al., 2009). In addition, there are other evolutionary processes that act in parallel and shape the genomic, transcriptomic, and phenotypic structure of the mesopolyploids to allow their functional diploidization, such as gene subfunctionalization and neofunctionalization, gene-silencing, activation of transposable elements, or genome rearrangements (Dodsworth et al., 2016).
Notwithstanding all these well-known effects of WGD at the genomic, morphological, anatomical, physiological, and ecological scales (Ramsey and Schemske, 2002;Madlung et al., 2005;Soltis et al., 2014;Spoelhof et al., 2017), discerning the immediate impact of WGD from the short-term and long-term adaptive effects is challenging. Most previous studies have been conducted in naturally occurring polyploid complexes (Thompson and Merg, 2008;Balao et al., 2011;Manzaneda et al., 2012;Pegoraro et al., 2019) whose lineages have already experienced selection and diploidization across many generations. In contrast, synthetic neopolyploids provide an opportunity to study the immediate consequences of WGD on plant morphology and physiology (Maherali et al., 2009;Madlung, 2013;Spoelhof et al., 2017;Van Drunen and Husband, 2018). Artificial induction of autopolyploids using antimitotic chemicals (e.g. colchicine) has been successfully carried out in both plant breeding and research (Stupar et al., 2007;Vyas et al., 2007;Liu et al., 2011) and has provided valuable insights into the mechanisms by which polyploidization per se affects plant phenotypes. Discerning immediate WGD versus post-WGD phenotypic effects is possible by comparing the synthetic polyploids with parental diploids and natural polyploids (Martin and Husband, 2012;Husband et al., 2016;Pavlíková et al., 2017;Van Drunen and Husband, 2018). On the one hand, differences between diploids and synthetic neopolyploids in functional traits will reveal the instantaneous effects of WGD. On the other hand, comparisons between synthetic neopolyploids and naturally established polyploids might indicate which polyploidy-mediated shifts are adaptive (because they are retained in natural polyploids) and which ones are not advantageous in the long term.
Dianthus broteri (Caryophyllaceae) provides an excellent system for discerning the immediate effects of WGD from the subsequent evolutionary processes. This Iberian taxon encompasses the largest known autopolyploid complex within the genus (Balao et al., 2009), including diploid (2n = 2x = 30), tetraploid (2n = 4x = 60), hexaploid (2n = 6x = 90), and dodecaploid (2n = 12x = 180) monocytotypic populations. This complex seems to have evolved recently (0.9-2.1 Mya) following a rapid cytotypic adaptive divergence in concert with an extensive geographical disjunction and an ecological niche differentiation within an aridity gradient, with the higher cytotypes (6x and 12x) inhabiting more restricted and extreme habitats (Balao et al., 2010;López-Jurado et al., 2019). Diploids occupy the most benign Mediterranean subclimate, whilst tetraploids have enhanced ecological tolerances, encompassing the most diverse environmental conditions. This niche divergence is accompanied by numerous phenotypic differences (vegetative and reproductive) between the cytotypes (Balao et al., 2011), as well as different strategies regarding light-harvesting and photoprotection under temperature stress, therefore suggesting divergent adaptations (López-Jurado et al., 2020). All these phenotypic differences seem to be due to a mixture of both immediate and post-WGD effects. For example, stomatal sizes vary with ploidy according to the nucleotypic effect, and some floral parts show within-cytotype differences in phenotypic optima, such as in tetraploids due to their two unrelated lineages. In addition, genome-downsizing that is linage-dependent is evident from the 1Cx values (Balao et al., 2009) and higher cytotypes show increased levels of DNA methylation and epigenetic variability (Alonso et al., 2016).
Within this framework, the artificial induction of D. broteri neopolyploids provides a potential opportunity to distinguish between the influence of WGD per se and subsequent adaptive changes that vary across natural populations due to reproductive isolation and divergence by selective pressures (Maherali et al., 2009;Spoelhof et al., 2017;Van Drunen and Husband, 2018). In this study, we established a procedure for the induction of autotetrapolyploidy by treating diploid seeds with colchicine, and we then compared several ecophysiological traits of diploids and synthetic and natural tetraploids to address the following three questions. Is genome-downsizing a direct effect of polyploidization? Are there physiological shifts directly linked to polyploidization in D. broteri? And to what extent can post-polyploidization mechanisms explain the functional basis of the niche differentiation observed in naturally occurring populations of D. broteri?

Plant material and induction of synthetic neopolyploidy
Seeds were collected in 2018 from a diploid population (15 plants) of Dianthus broteri located in south-eastern Spain (37º00´N, 3º01´W). The ploidy level and DNA content of this population have previously been characterized by flow cytometry together with chromosome counts (Balao et al., 2009), and its morphology and ecophysiology are known to be representative of D. broteri diploids (Balao et al., 2011). Treatments with colchicine (Sigma-Aldrich) consisted of two different incubation times (24 h or 48 h) in combination with four different concentrations in water: 0% (w/v, control), 0.2%, 0.4%, and 0.6%. Seeds were incubated in 2-ml Eppendorf tubes in darkness on a shaker, after which they were rinsed with 25 ml of distilled water and sown on water-saturated filter paper in Petri dishes that were kept under a 12/12 h photoperiod at 25.5/17 ºC (130 µmol m -2 s -1 ). Germination was recorded every 2 d for 21 d, and the seedlings obtained were sown in individual 2.5-l pots filled with a mixture of a commercial organic substrate (Gramoflor) and perlite (3:1) and placed in a greenhouse with controlled temperature of 21-25 ºC, 40-60 % relative humidity (RH), and natural daylight ranging from 200-1200 μmol m −2 s −1 photosynthetic photon flux density (PPFD) incident on leaves during the day. The pots were adequately irrigated with tap water and seedling survival was checked weekly for 90 d after sowing (at which point all plants displayed sufficient vigour to be considered as definitely established). In addition, diploid and tetraploid plants from natural populations spanning the whole range of distribution of this species (Table 1) were also grown under the same greenhouse conditions. The aim of including such among-population variation was to make our results conservative (i.e. based exclusively on intrinsic characteristics of each group beyond the possible dissimilarities caused by local adaptations and genetic drift) and our conclusions robust.
For the ecophysiological characterization, 10 plants per group (i.e. 2x, 4x Neo , and 4x Nat ) were randomly selected 15 d prior to analysis and placed in a controlled-environment chamber (Aralab/Fitoclima 18.000EH, Lisbon, Portugal) with a 14/10 h photoperiod (300 μmol m −2 s −1 ) at 25/18 °C and 40-60 % RH. This period under controlled conditions was to avoid interference by possible border effects within the greenhouse.

Determination of DNA content and ploidy levels
The DNA content and ploidy level of the seedlings were estimated by flow cytometry (FC) following a specific protocol for D. broteri (Balao et al., 2009). The FC measurements were conducted using a Coulter CYTOMICS FC500-MPL (Beckman Coulter, Fullerton, CA, USA) equipped with a 20 mW argon-ion laser at 488 nm. Fresh leaf material (100 mg) from glasshouse-grown D. broteri seedlings was used for nuclear suspensions together with Pisum sativum L. cv. 'Ctirad' as the genome size control (2C nuclear DNA = 9.09 pg). All the FC measurements showed high quality (>5000 nuclei per peak and CV <5 %) and peak means were determined through manual gating using the Kaluza Analysis 2.1 software (Beckman Coulter). DNA content (2C) was estimated for the main FC peak (phase G 0 /G 1 ; Supplementary Fig. S1), which would correspond to ~1.80 pg for diploids and ~3.60 pg for tetraploids (Balao et al., 2009). Mixoploidy (mixture of diploid and polyploid cells within the same tissue) was found, with some FC samples showing a second large peak representing more than 40% of the sample events, plus a third peak corresponding to nuclei in mitosis (phase G 2 ) of the higher-ploidy-level cells. According to these results, each plant was classified as diploid (2n=2x), neotetraploid (2n=4x Neo ), diploid-tetraploid mixoploid (2n=2x+4x), or tetraploid-octoploid mixoploid (2n=4x+8x). The mixoploids were discarded for further analysis. Monoploid genome size (1Cx) was estimated as the amount of 2C nuclear DNA divided by the ploidy level. All 4x Neo (n=10) were induced from seeds collected in the diploid population of Laroles (marked with an asterisk).

Stomatal measurements
Stomatal measurements were performed 80-90 d after the colchicine treatment of seeds on 11 plants for each cytotype (all the confirmed 4x Neo plants, and randomly chosen plants for 2x and 4x Nat ). Epidermal impressions of the region near the leaf tip were taken on the adaxial surface of the youngest fully expanded leaf using Germolene New Skin Liquid Plaster (PharmaPac, UK) and examined under a Carl-Zeiss Axiophot photomicroscope equipped with a Sony DXC-390P Exwave HAD camera. The resulting images were analysed using ImageJ v.1.51j8, and the mean stomatal density together with mean lengths and widths were determined on the basis of 10 images per plant, at 10× magnification for density and 40× for the dimensions (Balao et al., 2011).

Measurement of leaf gas exchange
A LI-COR LI-6400 infrared gas analyser in an open system equipped with a Li-6400-02B LED light source was used for instantaneous gasexchange measurements and for constructing A/C i curves. Five fully developed leaves were randomly selected for measurement on each cytotype. Net photosynthetic rate (A N ), stomatal conductance (g s ), and intercellular CO 2 concentration (C i ) were recorded at a PPFD of 1000 µmol m −2 s −1 (with 15% blue light to maximize stomatal aperture), vapour pressure deficit of 2.0-3.0 kPa, ambient CO 2 concentration (C a ) of 400 µmol CO 2 mol −1 , leaf temperature of 25±2 °C, and 50±5% RH. In addition, leaves were dark-adapted for 10 min for measurement of dark respiration (R d ).
We estimated mesophyll conductance (g m ) and maximum carboxylation activity of Rubisco (V c,max ) by the curve-fitting method (Ethier and Livingston, 2004) and using the software package developed by Sharkey et al. (2007). A/C i responses were measured once the leaves had reached a steady-state under the conditions described above (~20 min), at which point C a was decreased in the following steps: 400, 350, 300, 250, 200, 150, 100, and 50 µmol mol −1 . The chamber conditions were then restored to their initial levels and C a was increased in the following steps to complete the curve: 500, 750, 1000, 1250, 1500, 1750, and 2000 µmol mol −1 . At each step, gas exchange was allowed to equilibrate to avoid significant variations in Rubisco activity (generally <180 s; Long and Bernacchi, 2003). CO 2 leakages into and out of the leaf chamber were determined using photosynthetically inactive leaves and the correction was applied to all the curves, as described previously by Flexas et al. (2013).

Analysis of leaf chlorophyll fluorescence
Modulated Chlorophyll a fluorescence was measured in dark-adapted leaves of each cytotype using a FluorPen FP100 PAM (Photon Systems Instruments, Czech Republic). The maximum quantum efficiency of PSII (F v /F m ) was determined by using a 0.8 s saturating light beam with an intensity of 8000 μmol m −1 s −1 (Schreiber et al., 1986), with 10 replicates per cytotype. In addition, Kautsky curves (which indicate the fast kinetics of Chlorophyll a) were obtained via OJIP tests implemented in the pre-programmed protocols of the FluorPen FP100 (n=10). F v /F m , absorbed energy flux (ABS/CS), trapped energy flux (TR/CS), electron transport energy flux (ET/CS), and dissipated energy flux (DI/CS) per leaf cross-section, together with performance index (PI) derived from OJIP were calculated according to Strasser et al. (2004Strasser et al. ( , 2010. These energy transduction fluxes on a leaf cross-section basis were employed due to differences in the density of reaction centres (RC/CS) among the cytotypes, according to López-Jurado et al. (2020). Maximum electron transport rate (ETR max ) was also obtained using the pre-programmed rapid light curve (RLC) protocol (n=6), which consisted of the exposure of dark-adapted leaves (30 min) to increasing light levels (100, 200, 300, 500, and 1000 µmol m −2 s −1 ) and recording the quantum yield at each step.

Statistical analyses
Data analyses were carried out using R v.3.6.3 (http://www.R-project. org/). The effects of the colchicine treatments (concentration and duration as main factors) on seed germination and seedling survival (dependent variables) were tested by fitting generalized linear models (GLMs) using a binomial error structure and the 'logit' link function for each dependent variable. The effectiveness of the colchicine treatments in producing viable polyploids was examined using the Fisher's exact test. One-way ANOVA and post hoc Tukey's tests were used for detecting differences between the cytotypes in the 1Cx DNA contents and the ecophysiological parameters derived from the OJIP tests and RLC protocol.

Effects of colchicine on germination, seedling survival, and neopolyploidy induction rate
Germination was significantly affected by colchicine concentration and incubation time, and there was no interaction between them (Table 2). Higher concentrations of colchicine resulted in lower germination success, from 61% in the control to 39% in seeds treated with 0.6 % colchicine. Seeds incubated for 48 h showed 29.3% higher germination overall than those incubated for 24 h (Fig. 1A). Seedling survival was also significantly affected by the colchicine concentration and by the incubation time, and again there was no interaction. An increase in colchicine concentration resulted in a decrease in survival rate, from 77.5% in the control to 13.74% in seeds treated with 0.6 % colchicine; however, in contrast to germination, survival was 20.8% lower overall in seedlings from the 48 h treatment compared to 24 h (Fig. 1B). A total of 94 plants were subjected to flow cytometry analysis (21 controls and 73 that survived the colchicine treatments) of which 74 were diploids (the 21 controls and 53 treated plants), 11 were tetraploids, and nine were mixoploids (eight 2x+4x and one 4x+8x). The frequency of polyploidy induction was significantly higher as the dose (i.e. concentration and incubation time) of colchicine increased (Fisher's exact test, P<0.05; Fig. 1C). The control treatment invariably failed to induce polyploidization. The highest induction of neopolyploidy (tetraploids 37.5% of the established plants) was obtained at 0.6 % colchicine concentration and Monoploid genome sizes of the 2x diploid, the 4x Nat natural tetraploid, and the 4x Neo neotetraploid. In the violin plots, white circles represent the median, thick black bars correspond to the interquartile range, and the thin lines show the 95% confidence interval. Different letters indicate significant differences among the cytotypes as determined using ANOVA followed by Tukey's test (P<0.05).
24 h incubation. The highest frequency of mixoploids (17.6 % of the established plants) was obtained after 48 h.
Photosynthetic performance was also affected by polyploidization, but 4x Nat did not always display the same trends than 4x Neo . Analysis of leaf gas exchange showed that the synthetic neopolyploids had a significant increase in net photosynthetic rate (A N ; F 2,13 =14.72, P<0.001), stomatal conductance (g s ; F 2,13 =17.71, P<0.001), and mesophyll conductance (g m ; F 2,11 =4.48, P<0.05) in comparison to the diploids and natural tetraploids, which showed similar values to each other (Fig. 3). A N and g s were almost doubled in 4x Neo compared to 2x and 4x Nat (P<0.05), which had similar values. The 4x Neo plants also showed ~2-fold greater g m than 2x (P<0.05), whilst 4x Nat showed intermediate values between those of 2x and 4x Neo . Although the results showed broadly similar patterns, no significant differences were found among the cytotypes for the other photosynthetic parameters that were measured, namely intercellular CO 2 concentration (C i ), maximum rate of Rubisco activity (V c,max ), and the rate of dark respiration (R d ) (Supplementary Fig. S2; ANOVA, P>0.05).
Analysis of leaf chlorophyll fluorescence also showed several distinct patterns of variation between the cytotypes (Fig. 4). Significant differences were found for maximum electron transport rate (ETR max ; F 2,20 =5.64, P<0.05) and maximum quantum efficiency of PSII (F v /F m ; F 2,48 =4.17, P<0.05). In both cases, higher values were found in 4x Neo compared to 2x, whilst vales in 4x Nat were intermediate (Fig.  4A, B), indicating a reduction in the polyploidy-mediated direct effect on these traits. Such an after-WGD reduction was not found in the performance index (PI), with 4x Neo and 4x Nat showing higher values (71.5% on average) than 2x (F 2,47 =8.70, P<0.001; Fig. 4C).
A different pattern was observed for the energy transduction fluxes. We did not observe a direct effect of WGD but a postpolyploidization divergence was found for the absorbed energy flux per cross-section (ABS/CS; F 2,47 =4.87, P<0.05; Fig.  4D). The values for 2x and 4x Neo were similar (P>0.05), whilst 4x Nat showed a 30.8% increase in ABS/CS compared to 2x (P<0.01). No significant difference was found between 4x Nat and 4x Neo (P>0.05). The same post-WGD divergence was also observed for the trapped energy flux per cross-section (TR/ CS; F 2,47 =5.31, P<0.01; Fig. 4E), with 4x Nat showing a 33.0% increase compared to the diploids. A similar trend was found in the electron transport energy flux per cross-section (ET/CS), but there were no significant differences among the cytotypes (F 2,47 =1.75, P>0.05), with mean values of 1717, 1960, and 2065 for 2x, 4x Neo , and 4x Nat , respectively ( Supplementary Fig.  S2). Finally, the dissipated energy flux per cross-section (DI/ CS) was only marginally affected by chromosome doubling (F 2,45 =2.57, P=0.09),with 4x Neo showing ~0.3-fold lower dissipation values than 2x and 4x Nat (Fig. 4F).

Discussion
Exposure to colchicine through seed imbibition was an efficient method for obtaining synthetic neopolyploids of Dianthus broteri, and our results could have implications for the optimization of autopolyploidy induction in general (Fig. 1). Fig. 3. Effects of polyploidization in Dianthus broteri on leaf gas exchange. Measurements were taken on diploids (2x), natural tetraploids (4x Nat ), and colchicine-induced neotetraploids (4x Neo ). (A) Net photosynthetic rate (A N ), (B) stomatal conductance (g s ), and (C) and mesophyll conductance (g m ). Data are means (±SD), n=5. Different letters indicate significant differences among cytotypes as determined using ANOVA followed by Tukey's test (P<0.05).
Despite polyploidization increasing in parallel with colchicine concentration, large reductions in germination and seedling survival at high dosages made treatments with intermediate doses more suitable for inducing neopolyploidy. This was consistent with results of previous studies (Thiebaut and Kasha, 1978;Chen et al., 2006;Lehrer et al., 2008;Omidbaigi et al., 2010;Xing et al., 2011;Pavlíková et al., 2017;Sadat Noori et al., 2017), although the different methodologies (different antimitotic agents and/or different target tissues such as shoot apical meristem or callus) and the species-specific responses to colchicine do not make it possible to establish a general protocol for neopolyploidy induction.
Quantification of nuclear DNA through flow cytometry provides a practical method for quickly determining ploidy levels (Balao et al., 2009) and it was also effective for discerning mixoploidy (Fig. 1C). Analysis of the monoploid genome size (1Cx) revealed remarkable genome size stability in nascent neopolyploids but a moderate genomic diploidization in the natural tetraploid populations (Fig. 1D), as previously shown by Balao et al. (2009). This is a common phenomenon in mesopolyploids (Dodsworth et al., 2016;Mandákova et al., 2010) and could be mainly due to gene loss and chromosomal rearrangements contributing to restore successful cyto-nuclear interactions (Sharbrough et al., 2017). Although genetic change seems to be subtle (but spread locally) in this D. broteri polyploid complex due to rapid diversification (Balao et al., 2010), genomic adjustments could have played a crucial role in adaptation after WGD, not only providing genomic stability but also accounting for transcriptomic distinctness, which frequently results in phenotypic novelty and ecophysiological differentiation Chelaifa et al., 2010;Soltis et al., 2014;Dodsworth et al., 2016). In this regard, genomic diploidization could have boosted local adaptations following the post-WGD niche expansion, and fostered the remarkable ecological features of D. broteri tetraploids, which show greater niche breadth compared with the rest of the cytotypes (López-Jurado et al., 2019) and also distinct light-harvesting and photoprotection strategies (López-Jurado et al., 2020).
The few previous studies that have examined the effects of autopolyploidy on plant physiology have focused on synthetic neopolyploids (Stupar et al., 2007;Liu et al., 2011;Dong et al., 2017;Wei et al., 2020) or on divergence in natural populations (Balao et al., 2011;Manzaneda et al., 2012;Thompson et al., 2014;Pacey et al., 2020). In contrast, our approach was able to distinguish not just the post-WGD effects on several physiological and development traits, but also the immediate effects that help to explain the divergent evolution of the ecological niche in the polyploids. For example, it is well established that stomatal characteristics have a great impact on the physiology of land plants, and climatic conditions usually operate as selective forces for these traits (Veselý et al., 2020). Our data indicated that autopolyploidization in D. broteri directly modified the size of guard cells (i.e. the 'gigas effect'), resulting in larger stomata with greater aperture area available for gas exchange (Fig. 2). Conversely, the density of stomata immediately after WGD decreased, but interestingly tetraploid populations in the wild did not show these changes in stomatal size and density. A reduction in stomatal size and an increase in density after WGD has also been shown in other polyploid systems (e.g. Maherali et al., 2009;Münzbergová, 2017). Thus, a 'phenotypic diploidization' could occur after WGD on these stomatal traits, which become diploid-like either due to directional selection after establishment (Vyas et al., 2007) or because the establishment of polyploid individuals with unfit stomatal characteristics is prevented (Comai, 2005;Soltis et al., 2014). In addition, stomatal downsizing in 4x Nat D. broteri, together with an increased number of stomata per unit area, would be related to the genome downsizing (i.e. genomic diploidization) in the natural tetraploids. Variation in cell size (related to the loss of DNA content) will play a crucial role in optimizing gas exchange, potential carbon gain, and water use after WGD (Simonin and Roddy, 2018;Lawson and Vialet-Chabrand, 2019;Roddy et al., 2020). Given the fact that higher stomatal densities have been associated with more arid climates (Carlson et al., 2016), the phenotypic and genomic diploidization in the higher-order neopolyploids (6x, 12x) of D. broteri could have also triggered adaptation to harsh environments (Balao et al., 2011). In addition, the remarkable within-cytotype variation of stomatal size in naturally occurring 2x and 4x populations (Fig. 2) suggests that local adaptations in D. broteri might reside in adjusting stomatal density rather than stomatal size .
Physiological traits also showed a mixture of immediate and post-WGD effects. However, we did not find any differences between cytotypes (either synthetic or natural polyploids) in maximum carboxylation rate of Rubisco (V c,max ), dark respiration (R d ), CO 2 concentration in the substomatal cavities (C i ), or the electron transport energy flux per leaf cross-section (ET/CS) ( Supplementary Fig. S2), so the few (if any) changes following WGD would be deleterious (e.g. under purifying selection). Interestingly, a previous study showed significant down-regulation in linear electron transport (i.e. ET/CS) only in the higher-ploidy levels (6x and 12x) of the D. broteri complex, suggesting a photoprotective adaptation to their stressful ecological niches (López-Jurado et al., 2020).
Phenotypic diploidization was evident in the stomatal and mesophyll conductances (g s and g m ; Fig. 3) and hence the increases gained through WGD were not maintained in 4x Nat . In agreement with our results (i.e. 4x Neo having higher conductances than 2x), neopolyploids frequently display increased g s and g m. (Vyas et al., 2007;Wang et al., 2019;Monda et al., 2016). This might be partially explained by structural differences between the ploidy levels, with greater stomatal area and larger intercellular spaces in the neopolyploids. Nevertheless, other non-structural factors directly affected by WGD could help to explain the differences in conductances, such as osmotic adjustment and metabolic regulation of stomatal opening (Mouhaya et al., 2010;Monda et al., 2016). It is worth noting that our physiological measurements were taken under optimal conditions and that a different picture might be seen under stressful ones (López-Jurado et al., 2020). The tendency of a return to diploid-like g s and g m values after WGD has not been previously reported to our knowledge, and a single study comparing diploids, neopolyploids, and established polyploids found that increases in g s and g m observed immediately after WGD tended to then increase even further (Vyas et al., 2007). However, this diploidization of conductances could accompany the structural diploidization of the stomata. Higher densities of smaller stomata are advantageous for gas exchange because they result in shorter diffusion path lengths (Franks and Beerling, 2009), and they can protect plants from xylem embolisms as well as facilitate the reduction of water loss (Li et al., 1996;te Beest et al., 2012).
The net photosynthetic rate (A N ), which correlates with g s and g m (among other parameters), showed higher values for 4x Neo whilst it did not differ between 2x and 4x Nat (Fig.  3). A greater photosynthetic capacity in combination with faster growth rate is common in neopolyploids (Ramsey and Schemske, 2002;Saleh et al., 2008;Münzbergová, 2017) and seem to be the rule in established mesopolyploids (Hull-Sanders et al., 2009;Ježilová et al., 2015;Liao et al., 2016). Accordingly, the noticeably lower net photosynthetic rate that we observed in natural tetraploids (similar to that in the diploids) would also be related to the genomic diploidization and the consequent stomatal adaptations to optimize the balance between CO 2 acquisition and water losses under a wide range of environmental conditions (Simonin and Roddy, 2018;Roddy et al., 2020).
The influence of phenotypic diploidization was also observed for the efficiency of the PSII apparatus, as indicated by the differences between the cytotypes in several chlorophyll fluorescence parameters (Fig. 4) (Maxwell and Johnson, 2000). The 4x Neo plants showed higher values for maximum electron transport rate (ETR max ) and maximum quantum efficiency of PSII (F v /F m ), and we also found that WGD per se decreased the dissipated energy flux per leaf cross-section (DI/CS). These results indicated that much of the energy absorbed by the 4x Neo plants would have been transformed in their photosystems and directed to the photochemical pathway (Flexas et al., 2013), which is consistent with their greater CO 2 assimilation capacity, as discussed above. Conversely, the higher DI/CS value in the natural populations could reflect a protective mechanism against the high irradiance typical of the Mediterranean dry season found at the localities inhabited by D. broteri (López-Jurado et al., 2019. Thus, environmental constraints may favour selection towards a return to the original diploid DI/CS values (i.e. phenotypic diploidization).
Notwithstanding this pattern of phenotypic diploidization being observed in leaf gas-exchange and in some leaf chlorophyll fluorescence parameters, other photosynthetically relevant variables showed an increase in post-WGD differentiation. While López-Jurado et al. (2020) did not detect significant differences between 2x and 4x cytotypes at non-stressful temperatures for absorbed and trapped energy fluxes per leaf cross-section (ABS/CS and TR/CS, respectively), our design allowed us to distinguish them (Fig. 4). Although there was not a clear immediate effect of WGD on these parameters, a post-WGD selection for elevated energy fluxes was revealed. A possible mechanistic explanation for the increases in ABS/CS and TR/CS in the natural tetraploids would be a preferential retention of dosagesensitive genes encoding components of both PSI and PSII, which would cause increased gene expression during the diploidization process (Coate et al., 2011). Synthetic tetraploids would display similar values to their progenitors because the WGD would maintain metabolic homeostasis (i.e. not alter molecular stoichiometry; He et al., 2013). An alternative explanation for the enhanced ABS/CS and TR/ CS in 4x Nat would be the neofunctionalization of duplicate genes encoding components associated with the two photosystems, with the alterations potentially enhancing their functionality (Coate and Doyle, 2013). It was notable that despite these differences in ABS/CS, TR/CS, and DI/ CS, the tetraploids conserved the photochemical superiority conferred by autopolyploidization per se (as mirrored by the values of performance index; Fig. 4C), indicating the necessity for natural polyploids of maintaining this feature.
Our study has highlighted the relative importance of WGD per se and post-WGD selection in determining the distinct phenotypes of the 2x and 4x cytotypes of D. broteri. The polyploidy-related shifts in ecophysiological traits that we have identified could provide selective advantages that lead to the initial short-term survival and establishment of polyploids, which usually occupy new niches out of the range of the diploid (te Beest et al., 2012). Although evidence for several alternative niche-evolution patterns has been found (Glennon et al., 2014;Pegoraro et al., 2019), divergence is one of the most commonly described in polyploid complexes (Thompson et al., 2014;López-Jurado et al., 2019). In D. broteri, the increased photosynthetic capacity observed in the synthetic neopolyploids (in comparison to diploids) would have the potential to facilitate their establishment through the exploitation of new niches, and would thus explain the wider ecological range of the tetraploids (López-Jurado et al., 2019).
In summary, the results of our study not only suggest that many important plant functional traits are directly affected by polyploidization, but they also reveal that the process of phenotypic diploidization can lead to the creative role of polyploidization being underestimated. Thus, phenotypic novelty in neopolyploids might be constrained by regulatory mechanisms related to physiology as well as by other evolutionary processes related to environmental pressures.

Supplementary data
The following supplementary data are available at JXB online. Fig. S1. Flow cytometry analyses of the plants arising from colchicine-treated seeds. Fig. S2. Intercellular CO 2 concentration, maximum carboxylation rate of Rubisco, dark respiration, and electron transport energy flux per leaf cross-section in the different cytotypes..