Nội dung

McCarthy et al. BMC Plant Biology (2019) 19:162 https://doi.org/10.1186/s12870-019-1771-5 RESEARCH ARTICLE Open Access Early consequences of allopolyploidy alter floral evolution in Nicotiana (Solanaceae) Elizabeth W. McCarthy1,8, Jacob B. Landis1,2,3, Amelda Kurti1, Amber J. Lawhorn1, Mark W. Chase4,5, Sandra Knapp6, Steven C. Le Comber7, Andrew R. Leitch7 and Amy Litt1* Abstract Background: Polyploidy has played a major role in angiosperm evolution. Previous studies have examined polyploid phenotypes in comparison to their extant progenitors, but not in context of predicted progenitor phenotypes at allopolyploid origin. In addition, differences in the trends of polyploid versus diploid evolution have not been investigated. We use ancestral character-state reconstructions to estimate progenitor phenotype at allopolyploid origin to determine patterns of polyploid evolution leading to morphology of the extant species. We also compare trends in diploid versus allopolyploid evolution to determine if polyploidy modifies floral evolutionary patterns. Results: Predicting the ancestral phenotype of a nascent allopolyploid from reconstructions of diploid phenotypes at the time of polyploid formation generates different phenotype predictions than when extant diploid phenotypes are used, the outcome of which can alter conclusions about polyploid evolution; however, most analyses yield the same results. Using ancestral reconstructions of diploid floral phenotypes indicate that young polyploids evolve shorter, wider corolla tubes, but older polyploids and diploids do not show any detectable evolutionary trends. Lability of the traits examined (floral shape, corolla tube length, and corolla tube width) differs across young and older polyploids and diploids. Corolla length is more evolutionarily labile in older polyploids and diploids. Polyploids do not display unique suites of floral characters based on both morphological and color traits, but some suites of characters may be evolving together and seem to have arisen multiple times within Nicotiana, perhaps due to the influence of pollinators. Conclusions: Young polyploids display different trends in floral evolution (shorter, wider corolla tubes, which may result in more generalist pollination) than older polyploids and diploids, suggesting that patterns of divergence are impacted by the early consequences of allopolyploidy, perhaps arising from genomic shock and/or subsequent genome stabilization associated with diploidization. Convergent evolution in floral morphology and color in Nicotiana can be consistent with pollinator preferences, suggesting that pollinators may have shaped floral evolution in Nicotiana. Keywords: Ancestral character state reconstruction, Evolution, Flower color, Flower morphology, Geometric morphometrics, Hybridization, Nicotiana, Polyploidy * Correspondence: amy.litt@ucr.edu 1 Department of Botany and Plant Sciences, University of California, Riverside, Riverside, CA 92521, USA Full list of author information is available at the end of the article © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. McCarthy et al. BMC Plant Biology (2019) 19:162 Background Polyploidy, or whole genome duplication, is a widespread phenomenon in angiosperms. All angiosperms have had at least one whole genome duplication in their evolutionary history [1], ~ 15% of speciation events in angiosperms and ~ 31% in ferns involve polyploidy [2], and 24% of extant vascular plants are neopolyploid [3]. Polyploidy may increase adaptability to new environments [4], but newly established polyploids are rare and therefore are at a disadvantage because they are much more likely to receive pollen from diploids, which may be incompatible due to the difference in ploidy [5], or may self-fertilize, leading to inbreeding depression. Many crop species, such as wheat, oilseed rape, coffee, and cotton, are allopolyploid, involving both whole genome duplication and interspecific hybridization [6]. The merger of two distinct genomes in one allopolyploid nucleus may result in ‘genomic shock’ [7], which yields changes in gene expression [8–12], chromosomal rearrangements [13, 14], increase of transposable element activity [15, 16], alterations of physiological processes [17, 18], changes in morphology [19, 20], and niche shifts [21]. These processes and their results can isolate newly formed allopolyploids from their diploid progenitors and may facilitate their establishment as a new species. The new combinations of traits that can result from genomic shock associated with allopolyploidy may allow them to respond differently than diploids to evolutionary pressures. In long-term evolution experiments with yeast grown on poor carbon-source media, tetraploid yeast adapted to the medium more rapidly than either haploid or diploid yeasts [22, 23]. Tetraploid yeast also accumulated a greater diversity of adaptive mutations, suggesting that tetraploids may have evolutionary potential that diploid and haploid yeasts lack [23]. However, diploid yeast consistently displays higher growth fitness than haploid, triploid, and tetraploid yeasts in multiple environmental contexts in short-term growth experiments [24]. Although growth in yeast and the evolution of complex traits in angiosperms may be on a different scale, they are both controlled by regulatory networks and biochemical pathways. Therefore, these yeast results suggest that tetraploids may be at a fitness disadvantage in the short-term, but may be more adaptable in the long-term, especially in harsh and stressful conditions [22–25]. The ability of polyploids to adapt to harsh environments has been proposed as one hypothesis for the persistence and increased diversification of polyploids after major ecological events such as the mass extinction event at the Cretaceous-Paleogene boundary [25–27]. Previous studies in angiosperms have investigated allopolyploid phenotypic evolution with respect to plant biomass [19, 28], photosynthetic capacity [29], non-photochemical quenching [18], defense response Page 2 of 19 to herbivory [17], and flower morphology and color [20, 30–32]. These studies have compared allopolyploid phenotypes to those of their diploid progenitors to evaluate whether allopolyploids display novel traits or combinations of traits, but these studies have not addressed whether allopolyploids follow different evolutionary trends than diploids. In addition, these studies do not take into account the fact that the diploid progenitor species have also been evolving since allopolyploid origin. Therefore, the phenotypes of the progenitors at the time of allopolyploid origin may have been different from those of the extant species that are used to evaluate allopolyploid phenotypes and evolution. This divergence of phenotypes is particularly likely to be true in older allopolyploid species. Using extant progenitors may, thus, introduce error into our interpretation of how allopolyploids have evolved. Previously, studies have reported differences in the short- and long-term consequences of allopolyploidy for genome structure [13, 15, 16, 33, 34], but less is known about short- and long-term consequences of allopolyploidy on phenotype, and none, as far as we are aware, has used ancestral character-state reconstruction to predict the phenotype of the progenitor diploids at the time of polyploid origin. In this study, we use ancestral character state reconstructions to compare the evolutionary responses of allopolyploids and diploids and to determine whether using reconstructed progenitor phenotypes modifies our conclusions about polyploid evolution. Nicotiana consists of 76 species, about half of which arose from six independent allotetraploid events at different time points (ca. 0.4, 0.6, 0.7, 1.4, 4.3, and 6 million years ago; Table 1; [35]. In addition, synthetic allopolyploids that were created in the lab ( [19]; K.Y. Lim, Queen Mary, University of London) from the same progenitor species as natural allopolyploids are available. Nicotiana has been well studied phylogenetically [36–40], and putative parentage of all allopolyploid species/groups has been determined. Nicotiana displays considerable diversity in floral morphology and color [20, 31, 41], facilitating study of the effects of allopolyploidy on floral evolution. Nicotiana allopolyploids can display transgressive morphologies that fall outside the range of their diploid progenitors and are thought to have evolved shorter, wider corolla tubes than expected, assuming the nascent polyploid has a morphology predicted by the morphologies of the flowers of extant diploids [20]. In this study, we compare floral evolution in allopolyploids to that observed in diploids to address the following questions. 1) Do allopolyploids have novel suites of floral characters not found in diploids? 2) Do allopolyploids display different evolutionary trends in floral morphology than diploids? 3) Are there differences between phenotypic McCarthy et al. BMC Plant Biology (2019) 19:162 Page 3 of 19 Table 1 Nicotiana allotetraploids (except section Suaveolentes) and their diploid progenitors and ages [35] Polyploid Section Maternal Progenitor Paternal Progenitor Approximate Age (myo) N. tabacum Nicotiana N. sylvestris N. tomentosiformis 0.6 N. rustica Rusticae N. paniculata N. undulata 0.7 N. arentsii Undulatae N. undulata N. wigandioides 0.4 N. clevelandii Polydicliae N. obtusifolia N. attenuata 1.4 N. quadrivalvis Polydicliae N. obtusifolia N. attenuata 1.4 N. nesophila Repandae N. sylvestris N. obtusifolia 4.3 N. nudicaulis Repandae N. sylvestris N. obtusifolia 4.3 N. repanda Repandae N. sylvestris N. obtusifolia 4.3 N. stocktonii Repandae N. sylvestris N. obtusifolia 4.3 evolution immediately following allopolyploidy versus that observed over longer time scales? 4) Do reconstructed progenitor phenotypes alter interpretation of polyploid evolution compared with predictions using values from extant diploids directly? Results Concatenated dataset yielded well-supported tree In order to determine whether allopolyploids and diploids display different evolutionary trends, we reconstructed ancestral character states, which requires a well-supported phylogenetic tree representing species relationships and detailed character states for extant species. Previous phylogenetic studies in Nicotiana [36–39] elucidated species relationships and hybrid origins with strong support, but often lacked support for backbone nodes because they were based on single DNA marker sequences. Our concatenated dataset, which uses sequences obtained from these previous studies with additional sequences generated in this study (Additional file 1: Table S1), produced a wellsupported tree with > 70% bootstrap support from maximum likelihood (ML) analyses for all except five nodes and > 0.95 posterior probabilities from Bayesian analyses for all except four nodes. In these analyses the backbone is also well supported (Fig. 1). In our tree, the positions of allopolyploid sections, and therefore the inferred diploid progenitors, are congruent with those found in previous studies [36–38]. In addition, our results suggest that diploid Nicotiana can be separated into two large clades, consisting of 1) sections Undulatae, Paniculatae, Tomentosae, and Trigonophyllae and 2) sections Alatae, Sylvestres, Petunioides, and Noctiflorae. Sister relationships were observed for sections Alatae and Sylvestres, Petunioides and Noctiflorae, and Undulatae and Paniculatae, whereas section Tomentosae was sister to sections Undulatae and Paniculatae, and section Trigonophyllae was sister to sections Undulatae, Paniculatae, and Tomentosae. Floral variation in Nicotiana Geometric morphometric analysis of floral shape in Nicotiana allopolyploids and diploids yielded a similar morphospace based on principal components 1 and 2 to that obtained previously [20]. The morphospace consists of two diagonal axes: round to stellate floral limb outline, and relatively small to relatively large floral tube opening (‘relative’ because all shapes are scaled to the same size in this analysis; Additional file 2: Figure S1). Principal component 1 (PC1) accounts for 58.84% of the variation in the dataset and PC2 for 19.41% of the variation. Across the corolla size dataset, corolla tube length ranged from 0.84 to 9.36 cm and tube width ranged from 0.14 to 1.65 cm (Additional file 2: Figure S1) based on floral averages calculated from measurements of five replicate photographs from each flower. Floral evolution in diploids versus allopolyploids To determine evolutionary trends of diploid morphology, we used ancestral character state reconstruction to predict substantial shifts in floral morphology across the diploid-only tree. A few examples of substantial shifts include: a shift to a more stellate floral outline on the branch leading to the most recent common ancestor of N. plumbaginifolia and N. longiflora (Fig. 2), a shift to a longer corolla tube on the branch leading to N. sylvestris (Fig. 3a), and a shift to a smaller corolla tube width on the branch leading to N. miersii (Fig. 3b). The number of shifts in the evolution of floral shape (22 shifts; Fig. 2) is similar to that seen in the evolution of tube length (23 shifts; Fig. 3a), whereas shifts in the evolution of tube width are less common (13 shifts; Fig. 3b). In addition, 71% of branches that have shifts have them in more than one trait (Fig. 3c), demonstrating that shifts in multiple traits tend to co-occur. We tested whether each floral trait displayed phylogenetic signal, that is, whether closely related species tend to have similar morphology, using both Blomberg’s McCarthy et al. BMC Plant Biology (2019) 19:162 Page 4 of 19 Fig. 1 Phylogenetic tree of diploid and allopolyploid Nicotiana species. Tree reconstructed from the concatenated dataset based on maximum likelihood (ML) and Bayesian analyses is well-supported at almost all nodes. Plain nodes: ML bootstrap > 70%, posterior probability > 0.95; node with orange dot: ML bootstrap > 70%, posterior probability < 0.95; nodes with blue dot: ML bootstrap < 70%, posterior probability < 0.95; nodes with red circles: nodes of polyploid origin for estimating reconstructed progenitor phenotypes. Side flower photographs to scale, bar = 5 cm K and Pagel’s λ. All floral traits showed less phylogenetic signal than predicted by a Brownian motion model of trait evolution (K < 1, Table 2), but only the results for tube length failed to reject the null hypothesis of no phylogenetic signal. We obtained similar results for Pagel’s λ; tube width and both PC1 and PC2 were significantly different from λ = 0 (no phylogenetic signal), but tube length was not (Table 2). These results suggest that the evolution of tube length is less constrained by phylogeny than that of the other floral traits. To estimate the progression of floral morphological evolution in diploids, we quantified the direction and magnitude of changes between successive internal nodes and between extant taxa and their reconstructed most recent ancestor on the diploid tree. The changes seen between reconstructed internal nodes are represented by the arrows in Fig. 4. We compared these with the direction and magnitude of the morphological change in allopolyploids as measured by the distance between the progenitor midpoint (the average of the means of each progenitor species) and the mean of each allopolyploid species/accession, following the methods of McCarthy et al. [20]. We then compared trends in floral morphological evolution between diploids and allopolyploids (Fig. 5). Based on the graphs in Fig. 5, diploids do not display any clear trends in evolution because the estimated progression of evolution is not significantly different from a uniform circular distribution around the origin in either McCarthy et al. BMC Plant Biology (2019) 19:162 Page 5 of 19 Nicotiana wigandioides Nicotiana undulata Nicotiana solanifolia Nicotiana cordifolia Nicotiana knightiana Nicotiana paniculata Nicotiana benavidesii Nicotiana raimondii Nicotiana tomentosiformis Nicotiana kawakamii Nicotiana setchellii Nicotiana otophora Nicotiana obtusifolia var obtusifolia Nicotiana obtusifolia var palmeri Nicotiana sylvestris Nicotiana alata Nicotiana mutabilis Nicotiana rastroensis Nicotiana bonariensis Nicotiana forgetiana Nicotiana langsdorffii Nicotiana plumbaginifolia Nicotiana longiflora Nicotiana noctiflora Nicotiana petunioides Nicotiana corymbosa Nicotiana pauciflora Nicotiana acuminata Nicotiana miersii Nicotiana attenuata Fig. 2 Ancestral character state reconstructions of floral limb shape on a diploid tree. Reconstructed values of floral limb shape represented by thin plate splines from the geometric morphometric morphospace (obtained using reconstructed (PC1, PC2) coordinates) at each internal node. Substantial shifts (greater than 10% of the range of shape variation) in floral limb shape marked on branches with blue lines. Front flower photos scaled to the same size to show only changes in shape floral limb shape (Moore-Rayleigh, R* = 0.0237, N = 58, p < 0.999, significance threshold of α = 0.05 is 0.0036 after Bonferroni correction) or in tube length and width (Moore-Rayleigh, R* = 0.365, N = 58, p < 0.90; Fig. 5e, f; Table 3). Similarly, overall trends in the direction of evolution in floral limb shape in allopolyploids are not significantly different from a uniform circular distribution when the complete polyploid dataset is analyzed (Moore-Rayleigh, R* = 0.433, N = 20, p < 0.90) or when young (0–0.7 million years old (myo); Moore-Rayleigh, R* = 1.27, N = 13, p < 0.01, significance threshold of α = 0.05 is 0.0036 after Bonferroni correction) and older (1.4–4.3 myo; Moore-Rayleigh, R* = 1.202, N = 7, p < 0.025, significance threshold of α = 0.05 is 0.0036 after Bonferroni correction) allopolyploids are analyzed separately (Fig. 5a, c; Table 3). For the complete polyploid dataset, patterns of evolution in corolla length and width are significantly different from a uniform circular distribution (Moore-Rayleigh, R* = 1.57, N = 20, p < 0.001; Fig. 5b, d; Table 3), suggesting that polyploids tend to evolve shorter, wider corolla tubes as also concluded in McCarthy et al. [20]. For the young polyploids, patterns of evolution are again skewed toward shorter and wider corolla tubes (Moore-Rayleigh, R* = 1.65, N = 13, p < 0.001); however, older polyploids are not significantly different from a uniform circular distribution (Moore-Rayleigh, R* = 0.437, N = 7, p < 0.90; Fig. 5b, d; Table 3). These results suggest that polyploids have diverged along a more similar path than diploids, especially early in allopolyploid evolution. Reconstructed progenitor phenotypes do not alter interpretation of allopolyploid evolution in Nicotiana We hypothesized that the difference between reconstructed and extant progenitor morphology would increase with polyploid age (polyploid parentage and age are found in Table 1). To test this, we measured the distance between extant and reconstructed progenitor phenotypes for Nicotiana polyploids of different ages in our floral trait morphospaces. About half (8 of 14) of the reconstructed phenotypes showed differences from their extant counterparts (Fig. 6a, b). As predicted, the distance between McCarthy et al. BMC Plant Biology (2019) 19:162 Page 6 of 19 a 1.061 8.088 Nicotiana wigandioides Nicotiana undulata Nicotiana solanifolia Nicotiana cordifolia Nicotiana knightiana Nicotiana paniculata Nicotiana benavidesii Nicotiana raimondii Nicotiana tomentosiformis Nicotiana kawakamii Nicotiana setchellii Nicotiana otophora Nicotiana obtusifolia var obtusifolia Nicotiana obtusifolia var palmeri Nicotiana sylvestris Nicotiana alata Nicotiana mutabilis Nicotiana rastroensis Nicotiana bonariensis Nicotiana forgetiana Nicotiana langsdorffii Nicotiana plumbaginifolia Nicotiana longiflora Nicotiana noctiflora Nicotiana petunioides Nicotiana corymbosa Nicotiana pauciflora Nicotiana acuminata Nicotiana miersii Nicotiana attenuata 1.334 Nicotiana wigandioides Nicotiana undulata Nicotiana solanifolia Nicotiana cordifolia Nicotiana knightiana Nicotiana paniculata Nicotiana benavidesii Nicotiana raimondii Nicotiana tomentosiformis Nicotiana kawakamii Nicotiana setchellii Nicotiana otophora Nicotiana obtusifolia var obtusifolia Nicotiana obtusifolia var palmeri Nicotiana sylvestris Nicotiana alata Nicotiana mutabilis Nicotiana rastroensis Nicotiana bonariensis Nicotiana forgetiana Nicotiana langsdorffii Nicotiana plumbaginifolia Nicotiana longiflora Nicotiana noctiflora Nicotiana petunioides Nicotiana corymbosa Nicotiana pauciflora Nicotiana acuminata Nicotiana miersii Nicotiana attenuata b 0.186 length=0.017 c Nicotiana wigandioides Nicotiana undulata Nicotiana solanifolia Nicotiana cordifolia Nicotiana knightiana Nicotiana paniculata Nicotiana benavidesii Nicotiana raimondii Nicotiana tomentosiformis Nicotiana kawakamii Nicotiana setchellii Nicotiana otophora Nicotiana obtusifolia var obtusifolia Nicotiana obtusifolia var palmeri Nicotiana sylvestris Nicotiana alata Nicotiana mutabilis Nicotiana rastroensis Nicotiana bonariensis Nicotiana forgetiana Nicotiana langsdorffii Nicotiana plumbaginifolia Nicotiana longiflora Nicotiana noctiflora Nicotiana petunioides Nicotiana corymbosa Nicotiana pauciflora Nicotiana acuminata Nicotiana miersii Nicotiana attenuata Fig. 3 Ancestral character state reconstructions of corolla tube length and width on a diploid tree. Reconstructed values of corolla tube length (a) and width (b) represented as a heat map across the tree; red = short/narrow, blue = long/wide. Substantial shifts (greater than 10% of the range of tube length or width variation) marked on branches with black (length) or pink (width) lines. c Tree with all shifts in floral limb shape (blue), length (black), or width (pink) to determine on which branches shifts in multiple traits occur extant and reconstructed diploid progenitor phenotypes increased with allopolyploid age (Fig. 6c). To determine whether the use of reconstructed phenotypes alters the interpretation of allopolyploid divergence, we compared the results of our analyses of allopolyploid evolution using reconstructed versus extant progenitor phenotypes. Extant and reconstructed progenitor midpoints differed in 60% of cases (6 of 10) based on floral limb shape and corolla tube length and width data (Fig. 4a, b, c; Additional file 3: Figure S2; Additional file 4: Figure S3), and McCarthy et al. BMC Plant Biology (2019) 19:162 Page 7 of 19 Table 2 Phylogenetic signal tests Floral trait Pagel’s λ Blomberg’s K Randomization test (H0: no signal) H0: no signal PC1 0.549 p = 0.002 0.857 p = 0.00013 PC2 0.786 p = 0.001 0.99995 p = 3.4 × 10−5 −5 Tube 0.275 length p = 0.213 6.61 × 10 p=1 Tube width p = 0.002 0.918 p = 1.68 × 10−5 0.790 these changes in the progenitor midpoint resulted in differences in the direction of allopolyploid divergence in 7 and 27% of allopolyploids in floral limb shape and tube length and width, respectively (Fig. 5a-d). For example, N. nesophila, N. repanda, and N. stocktonii, which have the same two progenitors, have corollas that are longer and wider, shorter and narrower, and shorter and wider, respectively, compared to their extant progenitor midpoint (Fig. 4a, c). When compared to their reconstructed progenitor midpoint, however, they are all longer and narrower (Fig. 4b, c). Comparison of the results of Moore-Rayleigh tests based on either extant or reconstructed progenitor phenotypes show that reconstructed progenitor phenotypes do not change the response of allopolyploids in floral evolution in most of these analyses. For floral limb shape, trends in the direction of evolution using reconstructed progenitor phenotypes are not significantly different from a uniform circular distribution when the complete polyploid dataset is analyzed (Moore-Rayleigh, R* = 0.314, N = 20, p < 0.90) or when young (Moore-Rayleigh, R* = 1.19, N = 13, p < 0.025, significance threshold of α = 0.05 is 0.0036 after Bonferroni correction) and older (Moore-Rayleigh, R* = 1.28, N = 7, p < 0.01, significance threshold of α = 0.05 is 0.0036 after Bonferroni correction) polyploids are analyzed separately (Fig. 5a, c; Table 3). For corolla tube length and width, patterns of evolution are skewed toward shorter and wider corolla tubes for young polyploids (Moore-Rayleigh, R* = 1.64, N = 13, p < 0.001), but are not significantly different from a uniform circular distribution for older polyploids (Moore-Rayleigh, R* = 0.234, N = 7, p < 0.90) as observed for extant phenotypes (Fig. 5b, d; Table 3). In contrast, corolla tube length and width evolution is not significantly different from a uniform circular distribution when the complete polyploid dataset is analyzed based on reconstructed progenitor phenotypes (Moore-Rayleigh, R* = 1.01, N = 20, p < 0.10), whereas it is skewed towards shorter and wider tubes when extant progenitor phenotypes are used (Fig. 5b, d; Table 3). Allopolyploids sometimes display suites of floral characters not observed in diploids To determine whether allopolyploids display suites of floral characters that are not found in diploids, we identified evolutionary shifts in floral characters and determined whether any of these shifts represent convergent evolution. When only morphological characters were used, the two N. quadrivalvis accessions and the four N. tabacum accessions were placed in the same convergent regime with the following characters: a stellate floral shape with a relatively large tube opening compared to floral limb breadth, medium tube length (average from 2.90– 4.67 cm), and large tube width (average from 0.67–0.94 cm; Additional file 5: Figure S4). Both N. quadrivalvis copies were included in this convergent regime, but it included only the maternal copy of the N. tabacum accessions; the paternal copy of N. tabacum grouped with its paternal progenitor and related diploids. Because each allopolyploid copy has a different phylogenetic context due to the different evolutionary histories of the progenitors that contributed each copy, the surface program can place the two copies in different regimes even though the same morphology is entered for both copies. Nevertheless, these results suggest that N. quadrivalvis and N. tabacum allopolyploids possess a suite of floral characters distinct from those found in Nicotiana diploids. In addition, both copies of the related allopolyploids N. nesophila, N. repanda, and N. stocktonii were grouped in a convergent regime with the following characters: a stellate floral shape with a relatively small tube opening compared to floral limb breadth and a long (average from 4.28–5.10 cm) and narrow (average from 0.31–0.42 cm) corolla tube (Additional file 5: Figure S4). Although this is unsurprising since the same morphology was input for both copies for each species, it suggests that these allopolyploid species display a suite of floral characters that is not shared with any diploid species, based on morphological data. Other allopolyploids were grouped with either their maternal progenitor, paternal progenitor, or both (Additional file 5: Figure S4). In the analyses with only color characters and with both morphological and color characters, all allopolyploids are grouped with either their maternal, paternal, or both progenitors (Fig. 7; Additional file 6: Figure S5). However, these analyses indicate that several convergent regimes are present within Nicotiana. It should be noted that the floral color PCA was performed with spectra that were normalized to the same area under the curve in order to group spectra with the same shape, and thus most likely similar pigments, instead of focusing on the brightness or concentration of pigment. Therefore, convergent regimes may include species with varying floral color saturation, i.e. some light flowers and some dark flowers, but should reflect differences in floral hue. In the floral color only analysis, the four convergent regimes identified correspond to green-, magenta/purple-, pink-, and UV-reflecting white-flowered species (Additional file 6: Figure S5), suggesting that these floral colors arose multiple times in Nicotiana. McCarthy et al. BMC Plant Biology a b c d (2019) 19:162 Page 8 of 19 Fig. 4 Reconstructed phenotypes in the context of polyploid and diploid evolution. Convex polygons enclose the space taken up by all flower averages and the colored point represents the species mean for each species/accession. a-c Allopolyploid section Repandae in tube length and width in the context of extant progenitor phenotypes (a), reconstructed progenitor phenotypes (b), and both (c). Allopolypoids have filled polygons; diploid progenitors have outlined polygons and are labelled with ♀ for maternal and ♂ for paternal. The progenitor midpoint is denoted with a black square. Reconstructed progenitors are marked with dotted circles and the reconstructed progenitor midpoint is a square with a dotted black outline. Colored lines connect progenitor midpoints with allopolyploid means (solid = extant; dashed = reconstructed). d Diploid section Alatae in tube length and width. Black dots represent reconstructed phenotypes at internal nodes on the phylogenetic tree. Arrows denote direction of evolution based on phylogenetic relationships In the morphology and color analysis, convergent regimes are similar to those obtained with the color only analysis, but with a few differences, suggesting that flowers with the same colors tend to have similar morphology. The ‘green’ regime also includes N. knightiana and groups species with flowers that are green, have a round shape with a relatively large opening compared to floral limb breadth, medium width (average from 0.38–0.65 cm), and some variation in length (average from 1.64–4.24 cm; Fig. 7). The ‘UV-reflecting white’ regime no longer includes N. nudicaulis, but does include N. cordifolia and is marked by species with flowers that have high UV to visual spectral ratio (N. cordifolia is purple, not UV-reflecting white), mostly round floral shape, medium length (average from 1.89–2.20 cm, except for N. pauciflora: 5.51 cm), and medium width (average from 0.35–0.60 cm; Fig. 7). The species composition of the ‘pink’ regime is identical to that recovered using only floral color data and comprises of flowers that are pink, magenta, or red (to humans) with stellate shape and relatively large tube opening compared to floral limb breadth, wide tubes (average from 0.81–1.33 cm), and some variation in tube length (average from 2.36–4.56 cm; Fig. 7). The fourth convergent regime, the ‘white, stellate’ regime, was not found in the color-only analyses and contains species with white stellate flowers, relatively small tube opening compared to floral limb breadth, and long corolla tubes (average 3.21–8.09 cm; Fig. 7). The presence of several convergent regimes across Nicotiana suggests that suites of floral characters may be evolving together, perhaps due to the influence of pollinators. In addition to convergent regimes, the surface analyses detected shifts in floral traits that were not associated with convergent evolution. Both these unique shifts and the convergent regimes tend to correspond to shifts in at least two traits identified using the ancestral reconstruction analyses (Figs. 3c and 7). McCarthy et al. BMC Plant Biology (2019) 19:162 Page 9 of 19 a b c d e f Fig. 5 Trends in allopolyploid versus diploid floral divergence. Trends in evolution for extant allopolyploid floral limb shape (a), extant allopolyploid tube length and width (b), reconstructed allopolyploid floral limb shape (c), reconstructed allopolyploid tube length and width (d), diploid floral limb shape (e), and diploid tube length and width (f). Lines represent vector from the progenitor midpoint to the allopolyploid mean in (a-d) and the origin represents the progenitor midpoint. Young allopolyploids (0–0.7 myo) shown in red; older allopolyploids (1.4–4.3 myo) shown in light blue. In (e-f), lines represent the difference in reconstructed values between successive nodes on the tree and the origin represents the older node. Labels in the quadrants denote the phenotype toward which the vectors in that quadrant are evolving Discussion Corolla length evolution is less labile in allopolyploids than diploids and may play a role in speciation Evolutionary differences among young polyploids, older polyploids, and diploids are apparent in the relative lability of different floral traits. In polyploids, we measure lability as being less likely to overlap with the expected phenotype, represented by the mean of the progenitors (our null hypothesis). Polyploid groups with a lower percentage of accessions overlapping with the expected phenotype were McCarthy et al. BMC Plant Biology (2019) 19:162 Page 10 of 19 Table 3 Moore-Rayleigh test results (** denotes significance: ɑ = 0.05 after Bonferroni correction is 0.0036) Trait Group Extant/Reconstructed N R* P-value Shape All diploids – 58 0.0237 p < 0.999 Shape All polyploids Extant 20 0.433 p < 0.90 Shape Young polyploids Extant 13 1.27 p < 0.01 Shape Old polyploids Extant 7 1.20 p < 0.025 Shape All polyploids Reconstructed 20 0.314 p < 0.90 Shape Young polyploids Reconstructed 13 1.19 p < 0.025 Shape Old polyploids Reconstructed 7 1.28 p < 0.01 Length/Width All diploids – 58 0.365 p < 0.90 Length/Width All polyploids Extant 20 1.57 p < 0.001** Length/Width Young polyploids Extant 13 1.65 p < 0.001** Length/Width Old polyploids Extant 7 0.437 p < 0.90 Length/Width All polyploids Reconstructed 20 1.01 p < 0.10 Length/Width Young polyploids Reconstructed 13 1.64 p < 0.001** Length/Width Old polyploids Reconstructed 7 0.234 p < 0.90 considered more labile. In diploids, we measure lability by testing whether a trait displays phylogenetic signal; traits with no phylogenetic signal are more labile than traits that show phylogenetic signal. Because testing phylogenetic signal requires the input of a phylogenetic tree, we cannot use this metric to analyze polyploids. Although we cannot use the same metric to measure lability in both polyploids and diploids, we can rank the lability of the three floral traits (most labile to least labile) and compare this ranking order across young polyploids, older polyploids, and diploids to determine whether the same or different traits are relatively more labile in the three groups. Our previous analyses suggest that corolla tube length may be less labile than floral limb shape or tube width in Nicotiana polyploids [20], which show less overlap with the expected phenotype. However, when we separate young and older polyploids and re-examine these data, we see a different pattern. In young polyploids, we observe greater overlap with the expected phenotype for tube length than floral limb shape, and least overlap for tube width [20]. Thus, tube width is the most labile trait in young polyploids, followed by floral limb shape, with tube length as least labile trait. In contrast, older polyploids are most labile in floral limb shape, followed by tube length, with tube width the least labile [20]. In diploids, corolla tube length is the most labile (no phylogenetic signal), whereas corolla tube width and floral limb shape are less labile (significant phylogenetic signal; Table 2). Thus, floral traits differ in lability across groups, suggesting that they may be under different evolutionary pressures. Differences in the relative lability of corolla tube length may be linked to pollination. Length of corolla tubes is an important factor in the fit between flower and pollinator [42–44], and as a result, shifts in corolla tube or nectar spur length may facilitate reproductive isolation between species [45–47]. Consistent with this hypothesis, six of the 11 sister species pairs in our diploid tree show opposing shifts in corolla tube length: one towards a longer tube and the other towards a shorter tube (Fig. 3a). These 12 shifts comprise over half of the shifts in corolla tube length observed. These results, along with evidence that tube length evolution is not constrained by phylogeny, suggest that shifts in corolla tube length may play a role in species divergence, perhaps via pollinator-mediated selection. Pollinator relationships have only been elucidated for a subset of Nicotiana species; however, several of the sister species pairs that display shifts in tube length belong to section Alatae, the best studied section of Nicotiana in terms of pollination. These three pairs of sister species not only show shifts in tube length, but also shifts in their primary pollinators. Between sister species N. alata and N. mutabilis, longer-tubed N. alata is pollinated by hawkmoths [48], and shorter-tubed N. mutabilis is pollinated by hummingbirds [49]; hawkmoth-mediated selection has been shown to drive evolution of longer corolla tubes [45]. Shorter-tubed N. bonariensis is visited by small moths with tongues of similar length to its corolla tube, whereas its sister species, longer-tubed N. forgetiana, is primarily pollinated by hummingbirds [50]. Longer-tubed N. longiflora is pollinated by hawkmoths, but its shorter-tubed sister species, N. plumbaginifolia, is self-pollinating [50], consistent with the idea that outcrossing species are more likely to be subject to constant selective pressure from pollinators [50, 51] and therefore to maintain a longer corolla tube. These results demonstrate that shifts in corolla tube length can be correlated with specialization toward different pollinator