Evolutionary stasis in pollen morphogenesis due to natural selection
Summary
- The contribution of developmental constraints and selective forces to the determination of evolutionary patterns is an important and unsolved question. We test whether the long-term evolutionary stasis observed for pollen morphogenesis (microsporogenesis) in eudicots is due to developmental constraints or to selection on a morphological trait shaped by microsporogenesis: the equatorial aperture pattern. Most eudicots have three equatorial apertures but several taxa have independently lost the equatorial pattern and have microsporogenesis decoupled from aperture pattern determination. If selection on the equatorial pattern limits variation, we expect to see increased variation in microsporogenesis in the nonequatorial clades.
- Variation of microsporogenesis was studied using phylogenetic comparative analyses in 83 species dispersed throughout eudicots including species with and without equatorial apertures.
- The species that have lost the equatorial pattern have highly variable microsporogenesis at the intra-individual and inter-specific levels regardless of their pollen morphology, whereas microsporogenesis remains stable in species with the equatorial pattern.
- The observed burst of variation upon loss of equatorial apertures shows that there are no strong developmental constraints precluding variation in microsporogenesis, and that the stasis is likely to be due principally to selective pressure acting on pollen morphogenesis because of its implication in the determination of the equatorial aperture pattern.
Introduction
Evolutionary stasis is common, but its causes remain unknown (Hansen & Houle, 2004; Eldredge et al., 2005; Futuyma, 2010). Stasis in a character of a lineage can result from a failure of any variation (caused by developmental constraints) or from the failure of the variation to be fixed (caused by stabilizing selection) (Charlesworth et al., 1982; Maynard Smith, 1983; Maynard Smith et al., 1985; Williams, 1992; Gould, 2002), but the relative importance of these two causes is unclear. Here, we investigate the evolutionary forces responsible for stasis in pollen development in flowering plants.
Alberch (1982) proposed a simple test for whether the evolutionary stasis of a morphological character is caused by stabilizing selection or by developmental constraints: study the variation of the character in the absence of the selective pressure. If the character remains unchanged, then developmental constraints are likely to be acting; if the character changes, then it is likely that its stability was due to stabilizing selection. Such a test is usually not possible as we can neither remove selection, if it is present, nor can we wait long enough to be confident that continued stasis is not due to that selection. The evolutionary history of eudicot pollen, however, provides a natural experiment of the type Alberch proposed. The natural experiment we exploit here is that a minority of eudicots produce special pollen types, in which a portion of the selective pressure acting on the pollen morphogenetic system (microsporogenesis) must be absent. The role of this selective pressure in maintaining pollen stasis can thus be tested by investigating whether there is an increase of developmental variation in clades of species producing these special pollen types, as compared with clades of species producing normal pollen grains. This test is greatly facilitated by the fact that the clades with special pollen types are phylogenetically nested within clades in which other members have the normal pollen types, thus providing clear evolutionary transitions from one condition to the other.
The wall of pollen grains exhibits impressive morphological variation (Fig. 1). This diversity is caused by variation in pollen morphogenesis, which takes place during microsporogenesis. Microsporogenesis is the process by which, during male meiosis, the four daughter cells that will become the pollen grains are separated from each other by intersporal walls made of callose. Microsporogenesis is variable, as its three principal determinants (the cytokinesis type, the mode of callose deposition between meiotic daughter cells and the form of the tetrad of daughter cells) present several possible states (Fig. 2). By determining all possible combinations of variants of these three features, a morphospace of the theoretically possible microsporogenesis patterns can be defined (Fig. 3). This morphospace contains 20 possibilities but they do not appear to be observed equally in nature: some have never been observed, and others dominate large clades. In the eudicot clade (an old and speciose clade with c. 165 000 species) the huge majority of species have the same microsporogenesis type (no. 17 on Fig. 3) (Wodehouse, 1935; Blackmore & Crane, 1998; Nadot et al., 2008; Matamoro-Vidal et al., 2012). By contrast, in the monocot clade and in early-divergent angiosperms, microsporogenesis is quite variable (Furness & Rudall, 1999b; Furness et al., 2002; Penet et al., 2005; Sannier et al., 2006). Evolution of microsporogenesis seems thus to represent a case of evolutionary stasis in eudicots relative to its evolution in other angiosperm groups. Whether this stasis is the result of stabilizing selection or developmental constraints restricting the production of variation of microsporogenesis is an open question (Walker, 1974; Furness & Rudall, 2004).
Developmental constraints acting on microsporogenesis in eudicots could, for example, be expressed as an absence of genetic variance, or as robustness relative to mutations, as described in other biological systems (Bradshaw, 1991; Blows & Hoffmann, 2005; Stoltzfus & Yampolsky, 2009; Wagner, 2011). Alternatively, from a selective point of view, variation in microsporogenesis could be limited by the fitness effects due to the resulting variation in pollen morphology. In particular, a selective pressure could be acting on microsporogenesis through its implication in the determination of the aperture pattern (i.e. the number and the position of the apertures on the pollen wall), which is a major component of pollen morphology. Apertures are areas where the external pollen wall is thin or absent (Fig. 1), providing sites at which pollen tube growth is initiated (Edlund et al., 2004). Changes in aperture number affect the longevity and fertilization efficiency of pollen grains (Dajoz et al., 1991, 1993; Till-Bottraud et al., 1999), and also the accommodation of changes in pollen volume due to water gains and losses (harmomegathy) (Wodehouse, 1935; Halbritter & Hesse, 2004; Katifori et al., 2010; Firon et al., 2012).
Microsporogenesis variation has been shown to affect the aperture pattern in eudicots with equatorial apertures (e.g. Hulskamp et al., 1997; Albert et al., 2010, 2011). This is achieved by modifying the modalities of callose deposition, thus altering pollen exine wall synthesis at the level of the cell membrane (callose is an essential component for pollen exine wall synthesis; Dong et al., 2005). Figure 3 shows the equatorial aperture patterns associated with each microsporogenesis pattern, according to a model of aperture pattern ontogeny (Ressayre et al., 2002a,b). The microsporogenesis pattern more common in eudicots (pattern no. 17) results in the production of pollen with three (but sometimes more) equatorial apertures. This is the pollen type more common in the eudicot clade, which is often referred as the ‘tricolpate’ clade (Judd & Olmstead, 2004), tricolpate being a term for pollen with three furrow-like apertures. The other microsporogenesis patterns allow less variation in pollen morphology, and result in pollen with lower aperture numbers on average (Fig. 3).
It could be that the microsporogenesis pattern no. 17 has been under stabilizing selection in eudicots because the resulting pollen morphologies have better performance than the morphologies produced by other microsporogenesis patterns (Furness & Rudall, 2004). In a previous study focused on the Euphorbiaceae family, we developed a conceptual approach allowing Alberch's logic to be applied to our system and thus to test whether the stasis on microsporogenesis is due to selection related to performance of pollen morphology (Matamoro-Vidal et al., 2012). Here, we extend the approach to eudicots as a whole. The approach consists of studying the variation of microsporogenesis in a context where the putative selective pressure acting on it must be absent. For this, looking at microsporogenesis in species with nonequatorial pollen aperture patterns (e.g. lacking apertures (inaperturate), or with irregular or global aperture patterns) (Fig. 1i–l) is appropriate because in these, a partial or total disconnection of aperture position from microsporogenesis occurs (Wodehouse, 1935; Blackmore & Crane, 1998; Ressayre et al., 2002a; Furness, 2007; Albert et al., 2009; Matamoro-Vidal et al., 2012). Because of this disconnection, the effects of the selective pressures, if any, acting on microsporogenesis due to aperture positioning should be released or even absent in these species.
We thus propose the following rationale in order to discriminate between two hypotheses. The first hypothesis is that the stasis in microsporogenesis is maintained by selection related to aperture pattern function. Under this hypothesis, microsporogenesis in species producing pollen with typical equatorial eudicot aperture patterns (those shown in Fig. 3) is the target of a selective pressure based on aperture pattern determination (Fig. 4a1). However, in species presenting atypical non-equatorial aperture patterns (i.e. inaperturate, global or irregular aperture patterns), the aperture pattern is not determined by microsporogenesis. Thus, the aperture pattern no longer imposes any selection on microsporogenesis, irrespective of any other selection occurring on pollen morphology in these species (Fig. 4b1). Under this hypothesis, one would expect to observe variation in the characteristics of microsporogenesis in clades with atypical aperture patterns that are nested within clades in which other members have equatorial apertures. In our second hypothesis, stasis in microsporogenesis is maintained by developmental constraints. In this case, in species producing pollen with typical eudicot apertures patterns, microsporogenesis cannot vary because the characteristics of microsporogenesis are constrained (Fig. 4a2). Therefore, in species with atypical aperture patterns, even if the phenotype of the pollen does not depend on the microsporogenesis, the constraints should remain because disconnecting microsporogenesis from aperture pattern determination is not predicted to have any consequence on the constraints (i.e. on the robustness of microsporogenesis variation relative to mutations). Consequently, microsporogenesis should remain stable, and we should observe continued stasis in morphogenesis in species with atypical aperture pattern (Fig. 4b2).
Inaperturate, global and irregular aperture patterns have evolved independently numerous times throughout the eudicot lineage (Furness, 2007). Each of these occurrences provides a phylogenetically independent event suitable for testing the effect on microsporogenesis of the selective pressure related to aperture pattern function. Previous studies (Albert et al., 2009; Matamoro-Vidal et al., 2012) have shown that microsporogenesis is highly diverse in eudicot species producing inaperturate pollen; however, these works were based on the study of a single clade of inaperturate species. Here, we considerably extend these works by studying the variation of microsporogenesis in 18 phylogenetically independent samples of species which produce pollen with atypical pollen aperture patterns dispersed throughout the eudicot clade.
Materials and Methods
Species sampling and assessment of microsporogenesis
Data on microsporogenesis and on the pollen aperture pattern were obtained by collecting information from the literature for 35 species and by making our own observations for 46 additional species. The sampling includes species from 29 eudicot families representing 18 orders (Table 1, Fig. 5). Microsporogenesis was assessed by dissecting flower buds at different developmental stages. For most species, the flower buds were obtained from fresh material collected from the living collections of botanic gardens. Buds were obtained by growing seeds obtained from botanic gardens in the laboratory glasshouse (Orsay, France) for the following species: Vernonia cinerea, Matthiola incana and Linum strictum. For each species, the pollen morphology, tetrad form, cytokinesis type and mode of callose deposition during intersporal wall formation were recorded according to a described protocol (Ressayre, 2001). When more than one form of tetrad was observed for a given species, a survey of the proportion of each tetrad form was conducted by counting, within one or several buds of an individual, a number of tetrads ranging from c. 200 to 300 (Table 1).
Taxa | Aperture pattern | Tetrad form (% of each state and n) | Callose deposition | Cytokinesis | Species origin or reference |
---|---|---|---|---|---|
Adoxaceae | |||||
Viburnum tinus L. | Triaperturate | Td | CpP | Sim | PBL |
Apocynaceae s.l. | |||||
Acokanthera oppositifolia (Lam.) Codd | Triaperturate | Td | CpP | Sim | MNHN (no. 6061) |
Amsonia tabernaemontana Walter | Triaperturate | Td | CpP | Sim | MNHN (no. 22808) |
Carissa bispinosa (L.) Desf. Ex Brenan | Triaperturate | Td (98.5%) - Rh (1.5%) (300) | CpP | Sim | MNHN (no. 4355) |
Apocynum cannabinum L. | Global | Td (26.6%) - Rh (29.7%) - Tg (43.2%) - Lin (0.5%) (222) | CfP | Succ | MNHN (no. 228011) |
Periploca graeca L. | Global | Td (15%) - Rh (85%) (318) | CpT | Sim | MNHN (no. 22829) |
Trachelospermum jasminoides (Lem.) | Global | Lin | CfP | Succ | MNHN (no.15708) |
Apocynum androsaemifolium L. | Inaperturate | Td - Rh - Tg - Lin - T | CfP | Sim & Succ | Frye & Blodgett (1905) |
Cynanchum callialata Buch.-Ham. ex Wight | Inaperturate | Lin | CfP | Succ | Devi (1964) |
Hemidesmus indicus (L.) R. Br. | Inaperturate | T (majoritary) - Rh - Lin - Td | CfP | Sim | Devi (1964) |
Hoya carnosa (L.) R. Br. | Inaperturate | Lin | CfP | Succ | Schill & Dannenbaum (1984) |
Pergularia daemia (Forsk.) Chiov. | Inaperturate | Lin - T | CfP | Succ | Devi (1964) |
Ochrosia elliptica Labill. | Irregular | Td (14.6%) - Rh (82.7%) - Tg (2.7%) (295) | CpT | Sim | PBL |
Aquifoliaceae | |||||
Ilex aquifolium L. | Triaperturate | Td (99%) - Rh (1%) (101) | CpP | Sim | PBL |
Asteraceae | |||||
Achillea millefolium L. | Triaperturate | Td | CpP | Sim | PBL |
Catananche caerulea L. | Triaperturate | Td | CpP | Sim | Blackmore et al. (2007) |
Felicia bergeriana O.Hoffm. ex Zahlbr. | Triaperturate | Td | ? | Sim | Sharma & Murty (1978) |
Galinsoga parviflora Cav. | Triaperturate | Td (96%) - Rh (4%) (102) | CpP | Sim | PBL |
Helianthus annuus L. | Triaperturate | Td | ? | Sim | Gotelli et al. (2008) |
Solidago canadensis L. | Triaperturate | Td - Rh (rare) | ? | Sim | Pullaiah (1978) |
Vernonia cinerea (L.) Less. | Inaperturate | Td (29%) - Rh (71%) (110) | CpT | Sim | Collected seeds |
Berberidaceae | |||||
Epimedium rubrum E.Morren | Triaperturate | Td (97.5%) - Rh (2%) - Tg (0.5%) (294) | CpP | Sim | PBL |
Nandina domestica Thunb. | Triaperturate | Td | ? | Sim | Furness (2008) |
Podophyllum peltatum L. | Triaperturate | Td (99.5%) - Rh (0.5%) (400) | CpP | Sim | RBG (no. 1969 - 19646) |
Mahonia aquifolium Nutt. | Irregular | Tg | CpT | Succ | PBL |
Mahonia japonica (Thunb.) | Irregular | Tg - Irregular | CpT | Succ | Furness (2008) |
Brassicaceae | |||||
Arabidopsis thaliana (L.) Heynh. | Triaperturate | Td | CpP | Sim | Nadot et al. (2008) |
Capsella bursa-pastoris (L.) Medik. | Triaperturate | Td | CpP | Sim | PBL |
Matthiola incana (L.) R. Br. subsp. glabra | Inaperturate | Td (97.3%) - Rh (2.7%) (300) | CpP & CpT | Sim | RBG (no. 310) |
Caryophyllaceae | |||||
Cerastium glomeratum Thuill. | Global | Td | CpP | Sim | MNHN (no. 23327) |
Silene latifolia Poir. | Global | Td (81.5%) - Rh (17.9%) - Tg (0.5%) (189) | CpP | Sim | PBL |
Silene pendula L. | Global | Td (81%) - Rh (19%) | ? | Sim | MNHN |
Stellaria media (L.) Vill. | Global | Td (98%) - Rh (2%) (222) | CpP | Sim | PBL |
Convolvulaceae | |||||
Calystegia sepium (L.) R. Br. | Global | Td (99%) - Rh (1%) (100) | CpP | Sim | PBL |
Stictocardia beraviensis Hallier f. | Global | Td (93.5%) -Rh (6.5%) (200) | ? | Sim | MNHN |
Cucurbitaceae | |||||
Bryonia dioica L. | Triaperturate | Td | CpP | Sim | PBL |
Euphorbiaceae s.s. | |||||
Dalechampia spathulata (Scheidw.) Baill. | Triaperturate | Td | CpP | Sim | Matamoro-Vidal et al. (2012) |
Hevea brasiliensis Muell. | Triaperturate | Td | ? | Sim | Rao (1964) |
Hura crepitans L. | Triaperturate | Td | CpP | Sim | Matamoro-Vidal et al. (2012) |
Macaranga tanarius (L.) Müll. Arg. | Triaperturate | Td | CpP | Sim | Matamoro-Vidal et al. (2012) |
Mercurialis annua L. | Triaperturate | Td | CpP | Sim | Matamoro-Vidal et al. (2012) |
Mercurialis perennis L. | Triaperturate | Td | CpP | Sim | Matamoro-Vidal et al. (2012) |
Pedilanthus tithymaloides (L.) Poit. subsp. Tithymaloides | Triaperturate | Td | CpP | Sim | Matamoro-Vidal et al. (2012) |
Ricinus communis (L.) | Triaperturate | Td | CpP | Sim | Matamoro-Vidal et al. (2012) |
Manihot esculenta Crantz. | Global | Td (80%) - Rh (20%) (289) | CpP & CpT | Sim | Matamoro-Vidal et al. (2012) |
Baloghia inophylla (G.Forst) P. Green | Inaperturate | Td (70%) - Rh (25%) - Tg (5%) (215) | CpP & CpT | Sim | Matamoro-Vidal et al. (2012) |
Codiaeum variegatum var. pictum (Lodd.) Müll. Arg. | Inaperturate | Td (72%) - Rh (20%) - Tg (8%) (291) | CpP/CpT/CfT | Sim & Succ | Albert et al. (2009) |
Eremocarpus setigerus (Hook.) Benth. | Inaperturate | Td (51%) - Rh (29%) - Tg (19%) (228) | CpP & CpT | Sim | Matamoro-Vidal et al. (2012) |
Jatropha curcas L. | Inaperturate | Td - Rh | ? | Sim | Liu et al. (2007) |
Jatropha gossypiifolia L. | Inaperturate | Td (89%) - Rh (10%) - Tg (1%) (293) | CpP & CpT | Sim | Matamoro-Vidal et al. (2012) |
Jatropha integerrima Jacq. | Inaperturate | Td (61%) - Rh (32%) - Tg (7%) (188) | CpP & CpT | Sim | (Matamoro-Vidal et al. (2012) |
Jatropha podagrica L. | Inaperturate | Td (83%) - Rh (15%) - Tg (2%) (218) | CpP & CpT | Sim | (Matamoro-Vidal et al. (2012) |
Fabaceae | |||||
Uraria crinita (L.) Desv. ex DC. | Triaperturate | Td | CpP | Sim | Liu & Huang (1999) |
Vicia sepium L. | Triaperturate | Td | CpP | Sim | PBL |
Juglandaceae | |||||
Juglans regia L. | Global | Td | CpP | Sim | PBL |
Lamiaceae | |||||
Lamium purpureum L. | Triaperturate | Td | CpP | Sim | PBL |
Linaceae | |||||
Linum strictum L. | Triaperturate | Td (95.5%) - Rh (4.5%) (354) | CpP | Sim | MNHN (no. 07-43) |
Reinwardtia cicanoba (Buch.-Ham.ex D.Don) Hara | Global | Td (20.8%) - Rh (78%) - Lin (0.3%) - T (0.9%) (1277) | CpP | Sim | RBG(no. 1973-12448) |
Linderniaceae | |||||
Torenia fournieri | Triaperturate | Td | CpP | Sim | Vagi et al. (2004) |
Malvaceae | |||||
Abutilon sp. ‘Red’ | Triaperturate | Td | CpP | Sim | MNHN |
Hibiscus waimeae A. Heller | Global | Td (19%) - Rh (70.5%) - Tg (10.5%) (278) | CpP & CpT | Sim | RBG (no. 1988-4056) |
Lavatera maritima Gouan | Global | Td (94%) - Rh (6%) (300) | CpP | Sim | MNHN |
Nepenthaceae | |||||
Nepenthes maxima Nees | Inaperturate | Td | CpP | Sim | JBVP |
Onagraceae | |||||
Chamerion angustifolium L. | Triaperturate | Td (98%) - Rh (1%) - Tg (1%) (100) | CpP | Sim | Nadot et al. (2008) |
Orobanchaceae | |||||
Orobanche hederae Duby | Inaperturate | Tg (91%) - Irreg (9%) (101) | CfP | Succ | PBL |
Papaveraceae | |||||
Chelidonium majus L. | Triaperturate | Td | CpP | Sim | Nadot et al. (2008) |
Meconopsis cambrica (L.) Vig. | Triaperturate | Td (98%) - Rh (2%) (600) | CpP | Sim | MNHN (no. 56808) |
Meconopsis horridula Hook. f. & Thomson | Inaperturate | Td - Rh - Tg - Irreg | CpP | Sim & Succ | MNHN (no. 47941) |
Phyllanthaceae | |||||
Phyllanthus juglandifolius Willd. | Global | Td (90.4%) - Rh (9.6%) (239) | CpP | Sim | PBL |
Plantaginaceae | |||||
Plantago lanceolata L. | Global | Td (69%) - Rh (18%) - Tg (13%) (100) | CpP | Sim & Succ | PBL |
Veronica chamaedrys L. | Triaperturate | Td | CpP | Sim | PBL |
Primulaceae | |||||
Anagallis arvensis L. | Triaperturate | Td | CpT | Sim | PBL |
Ranunculaceae | |||||
Helleborus foetidus L. | Triaperturate | Td | CpP | Sim | Ressayre et al. (2005) |
Ficaria verna Huds. | Triaperturate | Td (99%) - Rh (1%) (101) | CpP | Sim | PBL |
Rosaceae | |||||
Chaenomeles japonica (Thunb.) Lindl. | Triaperturate | Td | CpP | Sim | PBL |
Salicaceae | |||||
Salix dasyclados Wimm. | Triaperturate | Td | CpP & CpT | Sim | MNHN |
Populus italica Du Roi | Inaperturate | Td (85%) - Rh (11%) - Tg (4%) (295) | CpT | Sim | PBL |
Sapindaceae | |||||
Acer platanoides L. | Triaperturate | Td | CpP | Sim | PBL |
Solanaceae | |||||
Nicotiana tabacum L. | Triaperturate | Td | CpP | Sim | Ressayre et al. (2003) |
Solanum nigrum L. | Triaperturate | Td | CpP | Sim | PBL |
Tropaelaceae | |||||
Tropaeolum majus L. | Triaperturate | Td | CpP | Sim | Bolenbaugh (1928) |
- The last column indicates the origin of the species (with the botanical garden accession number when available) for which new data are provided or the reference for data obtained from the literature. For the tetrad form, the percentage of each form and the number of counted tetrads (n) are provided (when available). Botanical gardens: JBL, Jardin Botanique de Launay, Orsay, France; JBVP, Jardin Botanique de la Ville de Paris, France; MNHN, Muséum national d'Histoire naturelle, Paris, France; RBG: Royal Botanic Gardens, Kew, UK. Tetrad form: Td, tetrahedral; Tg, tetragonal; Rh, rhomboidal; Lin, linear; T, T-shaped tetrad. Callose deposition: CpP, centripetal according to the cleavage plane; CpT, centripetal according to the tetrad; CfT, centrifugal according to the tetrad. n, number of tetrads counted to establish the percentages. Microsporogenesis: Sim, simultaneous; Succ, successive. ‘?’ denotes missing data.
Phylogeny, character optimization and test of correlated evolution
Phylogenetic relationships of the species for which we acquired data on pollen morphology and microsporogenesis were obtained by using available DNA sequences of rbcL and matK genes (Table S1). Nucleotide sequences were aligned separately for each gene using MAFFT (Katoh et al., 2002) and combined in a single matrix. The tree was rooted with Ranunculales, which are sister to all other eudicots (Soltis et al., 2011). The phylogenetic relationships of species were inferred under the assumption of maximum likelihood (ML) using PhyML 3.0 (Guindon & Gascuel, 2003; Guindon et al., 2010). A preliminary analysis resulted in a topology presenting some slight discrepancies with previous works (Soltis et al., 2011). Thus, we reran an analysis in which the topology was constrained by the results of a large-scale angiosperm phylogeny (Soltis et al., 2011) to define relations among families, and of phylogenetic studies focusing at lower taxonomic levels to define relations within the families for which more than two species were sampled: Apocynaceae s.l (Potgieter & Albert, 2001; Ionta & Judd, 2007); Asteraceae (Liu et al., 2002; Goertzen et al., 2003); Berberidaceae (Wang et al., 2009); Brassicaceae (Beilstein et al., 2008; Jaen-Molina et al., 2009); Caryophyllaceae (Greenberg & Donoghue, 2011); Euphorbiaceae (Wurdack et al., 2005); Malvaceae (Baum et al., 2004; Small, 2004) and Papaveraceae (Kadereit et al., 1997). The program was asked to optimize branch lengths only. The model of nucleotide substitution was the general time reversible model (GTR). Parameter values were estimated during running. Nucleotide frequencies were: (A) = 0.28995; (C) = 0.18194; (G) = 0.20172; (T) = 0.32640. The estimated proportion of invariable sites was 0.256 and a gamma distribution was assumed for rates at invariable sites.
Pollen and microsporogenesis characters were optimized onto the phylogeny using ML, as implemented in Mesquite software (Maddison & Maddison, 2008). Character coding for the optimization analysis is shown in Table 2. Heteromorphism (the coexistence of several states of the character within a single stamen) was observed for microsporogenesis features. Callose deposition and cytokinesis type were coded as heteromorphic when two or more character states were observed within a single stamen. For the mode of callose deposition, heteromorphism resulted most times in the co-occurrence of the states CpP and CpT. The other case of heteromorphism for the mode callose deposition was the co-occurrence of all the states of this character. These two cases were thus coded as states as such (Table 2). The stages of microsporogenesis for which the mode of callose deposition and the cytokinesis type are visible are very ephemeral, whereas the tetrad form is relatively easy to observe. This allowed us to adopt a quantitative rule for heteromorphism in the case of the tetrad form: the tetrad form was coded as heteromorphic when several states were observed within a single stamen and none of them was above the frequency of 0.95.
Character | Character states | ||||
---|---|---|---|---|---|
State 1 | State 2 | State 3 | State 4 | State 5 | |
Aperture pattern | Triaperturate | Inaperturate | Global | Irregular | – |
Aperture pattern (binary) | Triaperturate | Inaperturate, Global or Irregular | – | – | |
Tetrad form | Tetrahedral | Linear | Tetragonal | Heteromorphic | – |
Tetrad form (binary) | Tetrahedral | Other (nontetrahedral or heteromorphic) | – | – | – |
Mode of callose deposition | CpP | CpT | CfP | Heteromorphic 1 (CpP & CpT) | Heteromorphic 2 (CpP, CpT, CfP & CfT) |
Mode of callose deposition (binary) | CpP | Other (non-CpP or heteromorphic) | – | – | – |
Cytokinesis type | Simultaneous | Successive | Heteromorphic | – | – |
Cytokinesis type (binary) | Simultaneous | Other (successive or heteromorphic) | – | – | – |
- Alternative binary codings were defined for the purpose of the tests of correlated evolution, which can only be performed with binary characters. CpT, centripetal according to the tetrad; CpP, centripetal according to the cleavage plane; CfT, centrifugal according to the tetrad; CfP, centrifugal according to the cleavage plane.
We tested for a correlation between changes in pollen morphology and variation in microsporogenesis features using the BayesDiscrete method of the BayesTraits computer package (Pagel, 1994). This method requires a binary coding of the characters. The two character states defined for the aperture pattern were triaperturate (all the species with equatorial apertures studied were triaperturate) or other (patterns with irregular, global or absent apertures). Features of microsporogenesis were coded as binary characters as follows: (1) cytokinesis, simultaneous or other (successive or heteromorphic); (2) mode of callose deposition, CpP or other (heteromorphic or CpT or CfP or CfT) and (3) tetrad form, tetrahedral or other (heteromorphic or nontetrahedral tetrads). This coding is suitable for our aim of testing whether there is variation in microsporogenesis once aperture pattern is not triaperturate (i.e. in any configuration where a portion of the selective pressures acting on the microsporogenesis is presumably relaxed).
BayesDiscrete tests for correlated evolution among pairs of binary traits by taking into account the phylogenetic distribution of the different states of these traits were performed. However, the result of the test does not rely on any reconstruction of the traits' ancestral character states. The method consists of estimating the likelihood of two models of evolution: one in which the two traits evolve independently on the tree (independent model) and one that assumes correlated evolution between traits (dependent model). The independent model has four parameters (the two transition rates between states 0 and 1 for each character). The dependent model has height parameters which are the allowed transition rates (qij) between each of the four possible states of the pair of traits: (0,0); (0,1); (1,0); and (1,1). Evidence of correlated evolution is found when the dependent model fits better than the independent model. We used the reversible-jump Markov chain Monte Carlo (RJ MCMC) method (Pagel & Meade, 2006), which estimates the Bayesian posterior distributions of the likelihoods of the data given the model (dependent or independent). The log-harmonic mean of the likelihoods is compared between two runs (one assuming correlated evolution and the other independent evolution). The relative fit of the two models is tested using the Bayes factor [BF = 2(log [harmonic mean(dependent model)])/log[harmonic mean (independent model)])]. Any positive value of the Bayes factor favours the model assuming correlated evolution, but conventionally a ratio of > 2 is taken as ‘positive’ evidence, > 5 is ‘strong’ and > 10 is ‘very strong’ evidence. Analyses were run for 10 million generations and sampled every 1000th generation. The first million generations were discarded as burn-in. We used a gamma prior on the rate coefficients. Since no a priori information was available about the mean and variance of the rate coefficients, we did not specify the mean and the variance of the gamma distribution: these two parameters were seeded from a uniform (0–10) hyperprior distribution (priors 1). This allows one to remain relatively uncommitted about the details of the prior distribution (Pagel & Meade, 2006). Nevertheless, we measured the sensitivity of our results to the choice of priors by running analyses in which we specified exponential priors seeded from a hyperprior with a uniform (0–30) distribution (priors 2).
When evidence was found that a pair of traits evolves in a correlated fashion, the transition rates (qij) of the dependent model could be used to show the evolutionary sequence between the ancestral and derived states of the two traits in a flow diagram. The significance of each transition rate was examined by quantifying the percentage of times a transition rate was assigned a value of zero (Z) in the dependent model RJ MCMC chain. Transition rates that were rarely assigned to zero (Z < 10%) were considered probable events (as in Higginson et al., 2012) whereas transition rates often assigned to zero (Z > 90%) were considered improbable. Rates with intermediate Z values (10% ≤ Z ≤ 90%) were considered as uncertain.
BayesTraits allows one to take into account phylogenetic uncertainty and branch length heterogeneity in the analyses. For this, Bayes factors were estimated from a sample of 200 tree topologies obtained from a Bayesian analysis. This sample was obtained by analysing our alignment matrix with Bayesian inference using MRBAYES version 3.1.2 (Huelsenbeck & Ronquist, 2001; Ronquist & Huelsenbeck, 2003). The evolutionary model was set to the GTR model with a gamma distributed rate variation across sites and an estimated proportion of invariable sites. Analyses were performed as two independent runs of one million generations each, with sampling every 100th generation. Likelihood values reached a plateau within 100 000 generations (checked with Tracer v.1.5; (Rambaut & Drummond, 2007)). These were deleted as burn-in. The tree files of the two runs were combined in a single file from which 200 trees were randomly sampled.
Results
Data on microsporogenesis were obtained for 44 species producing triaperturate pollen and 37 producing pollen with atypical aperture patterns (inaperturate, irregular or global). Reconstruction of the ancestral character state of the aperture pattern indicates that the ancestral state of our sample is triaperturate and that the species with atypical aperture patterns from 18 samples that are phylogenetically independent (Figs 5, 6). For the tetrad form, the mode of callose deposition and cytokinesis type, the ancestral states were tetrahedral, CpP and simultaneous, respectively (Fig. 6). For the tetrad form, 14–15 transitions from the tetrahedral form to other states were observed (Fig. 6a). For mode of callose deposition, 10 transitions from the CpP mode to other states were found (Fig. 6b); and for cytokinesis type, there were only six transitions from the ancestral simultaneous type to other states (Fig. 6c).
The 44 triaperturate species investigated here carry out microsporogenesis via tetrahedral tetrads, CpP mode of callose deposition and simultaneous cytokinesis (Figs 5, 6; Table 1), which is the typical eudicot microsporogenesis pathway. Additional modes of callose deposition were observed in two triaperturate species only. In Anagallis arvensis, the mode of callose deposition was CpT, and in Salix dasyclados, both CpP and CpT modes of callose deposition were observed. Some slight variation was observed in the tetrad form for eight triaperturate species. Rhomboidal tetrads were observed in addition to the typical tetrahedral state, but at low frequency (< 5% of the counted tetrads; Table 1).
With regard to the species producing atypical aperture patterns, pollen morphology could not be associated with any specific microsporogenesis pattern (Table 1, Figs 5, 6). Very diverse microsporogenesis character states were found, such as successive cytokinesis, rhomboidal, linear, tetragonal or irregular tetrads; and centrifugal callose deposition (Fig. 7). For inaperturate species, the tetrad form could be exclusively tetrahedral (e.g. Nepenthes), linear (e.g. Hoya) or, in most of the cases (16 out of 18 species), heteromorphic (i.e. several forms of tetrad were observed within the same stamen). Callose deposition in inaperturate species was also very variable, as it could be exclusively CpP (e.g. Meconopsis horridula), CpT (e.g. Vernonia), CfP (e.g. Apocynum) or heteromorphic (8 species). The cytokinesis type in inaperturate species was either simultaneous (e.g. Matthiola), successive (e.g. Orobanche) or heteromorphic (e.g. Codiaeum). In the same way, there was a complete lack of association between pollen morphology and microsporogenesis in species producing pollen with global aperture patterns. Regarding species with irregular aperture patterns, the tetrad form was either tetragonal (Mahonia) or heteromorphic (Ochrosia); the mode of callose deposition was CpT, and the cytokinesis type was either successive (Mahonia) or simultaneous (Ochrosia). Overall, for the species with atypical aperture patterns, microsporogenesis was altered (i.e. different from the typical eudicot microsporogenesis pathway) for at least one of the three microsporogenesis features in 32 species, whereas microsporogenesis remained identical to the typical eudicot microsporogenesis pathway in five species only (Fig. 5). Intra-specific variation of microsporogenesis was observed in 29 out of 37 species with atypical aperture patterns.
We detected correlated changes in the three pairs of traits tested (pollen morphology/tetrad form: Bayes factor (BF) c. 49; pollen morphology/callose deposition: BF c. 12; pollen morphology/cytokinesis type: BF c. 18) (Table 3), indicating strong correlations between the transition from triaperturate to atypical aperture patterns and the appearance of variation for each of the microsporogenesis traits. This result was not sensitive to the choice of priors (Table 3). Figure 8 shows the most probable evolutionary route from the ancestral state of triaperturate pollen and typical microsporogenesis (simultaneous cytokinesis – Fig. 8b; tetrahedral tetrad – Fig. 8c; and CpP mode of callose deposition – Fig. 8d) to the derived state of an atypical pollen morphology and an atypical or variable microsporogenesis. Our results suggest that the loss of the triaperturate morphology always preceded alterations of any of the microsporogenesis features (q13 > q12), and that this loss created conditions favourable to the appearance of variation of the microsporogenesis (q34 > 0). The alternative route in which the variation of the microsporogenesis appears first, in a context where pollen morphology is triaperturate, is poorly supported (q12 c. 0 for the cytokinesis type and for the tetrad form using priors 1 or 2; for the mode of callose deposition with the priors 1 only).
Trait X | Trait Y | Parameter | Value | ||
---|---|---|---|---|---|
Priors 1 | Priors 2 | ||||
Aperture pattern | Cytokinesis type | BF | 18.9 | 16.3 | |
Z (%) | q12 | 99.8 | 99.3 | ||
q21 | 29.8 | 31.7 | |||
q13 | 0.0 | 0.0 | |||
q31 | 1.3 | 1.7 | |||
q24 | 26.4 | 27.7 | |||
q42 | 82.2 | 82.1 | |||
q34 | 0.0 | 0.0 | |||
q43 | 41.4 | 45.7 | |||
Aperture pattern | Tetrad form | BF | 49.8 | 49.2 | |
Z (%) | q12 | 99.2 | 91.9 | ||
q21 | 13.0 | 12.0 | |||
q13 | 0.4 | 4.9 | |||
q31 | 0.1 | 1.1 | |||
q24 | 21.7 | 18.9 | |||
q42 | 95.2 | 90.6 | |||
q34 | 0.2 | 1.6 | |||
q43 | 3.4 | 2.1 | |||
Aperture pattern | Mode of callose deposition | BF | 12.7 | 16.3 | |
Z (%) | q12 | 84.3 | 40.3 | ||
q21 | 14.7 | 16.5 | |||
q13 | 0.5 | 0.5 | |||
q31 | 2.7 | 15.0 < | |||
q24 | 9.5 | 5.6 | |||
q42 | 6.5 | 18.9 < | |||
q34 | 8.0 | 28.7 < | |||
q43 | 69.0 | 62.8 |
- For each pair of traits X–Y tested, the value of the Bayes factor (BF) and of the percentage of times each transition rate was assigned to zero (Z) in the reversible-jump Markov chain Monte Carlo (RJ MCMC) chain are given. Results are given for the two sets of hyperpriors used: gamma (Priors 1) and exponentials (Priors 2). Bold numbers indicate significant values of BF or transitions rates rarely assigned to zero (Z < 10%). Grey numbers depict improbable transition rates (Z ≥ 90%). The ‘<‘ marks indicate discrepancies between the two sets of hyperpriors used.
In the context of the loss of the triaperturate morphology, we found differences in the amount of variation observed between the three microsporogenesis features: the numbers of species and of clades with atypical microsporogenesis states were different for the three microsporogenesis features. Tetrad form, callose deposition and cytokinesis varied in 31, 25 and 12 species, respectively, and these alterations were distributed among 15, 10 and 6 independent phylogenetic samples, respectively (Figs 5, 6).
Within each microsporogenesis trait, in the context of the loss of the triaperturate morphology, the different states of each trait were not observed in equal proportions (Fig. 9). For the tetrad form, within the 18 groups of species producing atypical pollen morphology, the tetrahedral, rhomboidal and tetragonal states were observed in 16, 14 and 10 groups, respectively, whereas other tetrad forms (linear, irregular and T-shaped) were observed at a lower frequency, in one or two groups only (Fig. 9a). For the mode of callose deposition, we observed a higher number of occurrences for the centripetal modes than for the centrifugal modes (Fig. 9b). In the same way, simultaneous cytokinesis was observed much more frequently than successive cytokinesis (Fig. 9c).
Discussion
The stability of pollen morphology and development in eudicots is a fact that is widely recognized among botanists. The eudicot clade encompasses phenomenal variation in morphological, anatomical, ecological and biochemical characters (Judd & Olmstead, 2004), though microsporogenesis and pollen morphology in this group are relatively conserved: most species produce pollen with three equatorial apertures through a very stereotypic microsporogenesis pattern. According to the criterion of Williams (1992), this pattern may thus be considered as evolutionary stasis. The nature of the evolutionary processes underlying the conservation of pollen aperture pattern and microsporogenesis is fundamental for our understanding of, on the one hand, the evolutionary morphology of plants (Walker, 1974; Furness & Rudall, 2004) and, on the other hand, a common evolutionary pattern (evolutionary stasis) (Hunt, 2007; Kellermann et al., 2009; Piras et al., 2009; Ellegren, 2010; Lavoué et al., 2011; Van Bocxlaer & Hunt, 2013).
In this study we hypothesized that stasis in microsporogenesis was due to a selective pressure related to the determination of a pollen morphological trait (aperture pattern). This was tested by studying the variation of microsporogenesis in a context where the hypothetical selective pressure must be absent. Such conditions are found in species with atypical aperture patterns because in these species microsporogenesis is not the determinant of the aperture pattern. The disconnection of microsporogenesis from aperture pattern determination requires the loss of the equatorial aperture pattern, a transformation that could be constrained. However, Dobritsa and Coerper (2012) have shown that a mutation in a single gene (INP1) was sufficient to lose all three apertures in Arabidopsis thaliana, resulting in plants with an inaperturate pollen morphology lacking any other visible morphological abnormality, and otherwise normal and fertile. Moreover, inaperturate pollen appeared independently in distant groups both in eudicots and in monocots (Furness & Rudall, 1999a; Furness, 2007). This suggests that aperture loss, and thus the disconnection of microsporogenesis from the determination of aperture pattern is not a difficult transformation to achieve.
According to our rationale (see 1 and Fig. 4), if species with atypical aperture patterns present variable microsporogenesis then microsporogenesis is not constrained, and it is likely that the evolutionary stasis in microsporogenesis is due to selection on pollen morphology. A drawback of this rationale is that finding an absence of variation in species with atypical aperture pattern would be in part inconclusive because this result could be explained by constraints or by unidentified selective pressures acting on microsporogenesis. However, we found a remarkable burst of variation affecting all of the three microsporogenesis traits studied in the species producing atypical aperture patterns, clearly showing that there is a relationship between the involvement of microsporogenesis in pollen aperture pattern ontogeny and stasis.
Evolutionary transitions from the triaperturate to an atypical pollen morphology were correlated with the occurrence of intra-individual and inter-specific variation in microsporogenesis. Some slight variation in microsporogenesis was observed among the triaperturate species investigated herein, but it is quite negligible if compared with the huge variation observed in species with atypical pollen. Only a few species producing atypical pollen morphologies had microsporogenesis identical to that of the triaperturate species. Instead, species producing atypical pollen morphology exhibited a wide range of microsporogenesis pathways. For example, in Reinwardtia, we observed, within a single plant, four different microsporogenesis patterns, though pollen morphology was always the same, namely global.
The finding that in most of our species with atypical aperture patterns pollen morphology could not be associated with any microsporogenesis pattern, and that microsporogenesis was variable at the intra-individual level without any change in pollen morphology, is consistent with previous works (Wodehouse, 1935; Blackmore & Crane, 1998; Ressayre et al., 2002a; Furness, 2007) proposing that in these species microsporogenesis is disconnected from pollen aperture pattern determination. The fact that under these conditions microsporogenesis was highly variable also supports the view that microsporogenesis is not strongly constrained by the genotype-phenotype map or by other unidentified selective pressures.
Our evolutionary model (Fig. 8) supports a scenario in which changes in aperture pattern occur first, allowing for a release of selection on microsporogenesis and then, in most cases, this is followed by the occurrence of variation in microsporogenesis. Note that if the variation observed was due to a release of a developmental constraint on microsporogenesis, one would expect that variation in microsporogenesis appears first and is accompanied by variation in aperture pattern but it is the alternative scenario which is supported by the data (Fig. 8). The conservation of microsporogenesis pattern no. 17 is likely due principally to a selective pressure related to its implication in the determination of aperture pattern. It follows that pollen morphologies produced by this microsporogenesis pattern should present some improved performance as compared with the other morphologies shown in Fig. 3.
One possibility is that selection comes from the direct effects of aperture number on pollen reproductive success. Dajoz et al. (1991, 1993) and Till-Bottraud et al. (1999) have shown the existence of a trade-off such that increased aperture number correlates with a faster pollen germination rate upon landing on the stigma, but also with decreased pollen survival during dispersal. Another function of apertures is to contribute to the accommodation of pollen wall deformations further to volume changes of the cell due to water exchanges with the environment (Wodehouse, 1935; Halbritter & Hesse, 2004; Katifori et al., 2010). Moreover, aperture sites are involved in the regulation of pollen hydrodynamics (Heslop-Harrison, 1979; Firon et al., 2012). The aperture pattern is thus linked with many functions, and it could be that aperture patterns (especially the triaperturate one) resulting from eudicot typical microsporogenesis do a relatively good trade-off between all these functions. Future experimental work on the relationships between pollen morphology and pollen performance will have to test this prediction. Another possibility could be that it is the flexibility in pollen morphology provided by the microsporogenesis pattern no. 17 that represents an advantage. There is growing evidence that mechanisms facilitating the generation of variation can be maintained in the long-term by lineage selection, despite no obvious short-term advantage and even strong short-term disadvantages (Goldberg et al., 2010; de Vienne et al., 2013). Under this scenario, the stasis on the microsporogenesis would have been maintained, in the long-term, by selection of lineages having morphological diversity and rapid adaptation in a changing environment. This hypothesis is consistent with the maintenance of a single microsporogenesis pattern despite environmental change over long periods of time. However, this hypothesis lacks an explanation for how microsporogenesis has been maintained in the short-term, and thus it calls for the possibility that other evolutionary forces are involved in the maintenance of the stasis. Accordingly, our observations on the microsporogenesis variation in the absence of selection due to aperture pattern suggest that there might be constraints and/or additional selective pressures acting on the possible variations of microsporogenesis.
We found that the tetrad form and callose deposition change more often and more rapidly in response to the absence of selective pressure than does the cytokinesis type (Figs 5, 6). Similar results have been obtained in inaperturate monocot species, which have been found to be highly labile in the tetrad form but not in the cytokinesis type (Nadot et al., 2006; Pereira Nunes et al., 2009). These differences in the timing of changes between characters may be interpreted in three non-exclusive ways. First, spontaneous variation rates may differ among the three traits, for example, because of differences in the mutation rates between different parts of the genome (Tian et al., 2008) or because of differences in the robustness to mutations of each of these traits (Wagner, 2011). A second explanation is that the morphogenetic system is such that certain changes must necessarily occur if certain others are to become established (Alberch & Gale, 1985). Finally, the possibility that the microsporogenesis traits under study are involved in adaptive functions other than those due to determination of the aperture pattern cannot be ruled out. Interestingly, the character states of the typical eudicot microsporogenesis pathway predominate even in the absence of selective pressure due to aperture pattern (Fig. 9), suggesting that the selective pressure studied here is not the only force accounting for this pattern. It is unlikely that long-term stasis can be maintained by a single selective pressure given the environmental variation encountered by eudicot species since their radiation. Our finding that there might be a combination of factors contributing to the pattern of variation offers a possible solution to the problem of stasis maintenance in a changing environment.
Conclusions
The observed stasis in pollen morphogenesis is due to selective pressures acting on aperture pattern; however, there is also a bias in the generation of variation in microsporogenesis, which can result in a bias in the variation and evolution of aperture pattern. This bias might be a developmental constraint (Maynard Smith et al., 1985). We thus conclude that selection on pollen morphology and, to a lesser extent, constraints and/or other selective pressures, influence the evolution of pollen morphology and development. The data presented in this study reinforce the trend initiated by others (Beldade et al., 2002; Houle et al., 2003; Varela-Lasheras et al., 2011; Davis et al., 2014) indicating that the role of constraints in morphological stasis may have been overestimated. Our study, together with these others, shows that the stasis can be broken when a given selective pressure is manipulated. There is thus empirical evidence from several biological models supporting the idea that selection is an important force in the determination of stasis. Nevertheless, it is unlikely that a single, invariant, selective pressure can maintain morphological constancy for groups that encounter repeated environmental change during long periods of time as is the case for eudicots. We thus advocate the possibility that a combination of factors including selection and developmental constraints are necessary to fully explain patterns of stasis.
Acknowledgements
The authors thank S. Nadot and A. Ressayre for insightful ideas, A. Meade for help with BayesTraits program, and J. Doyle, D. Houle, M. Rausher, L. Venable, J. Schenk and A. Winn for comments on the manuscript. The authors thank H. Hallbritter for sharing pollen micrographs, A. Dubois, C. Raquin, L. Saunois and S. Siljak-Yakovlev for technical help, and the staff of the botanic gardens (Parc Botanique de Launay; Muséum national d'Histoire naturelle; Royal Botanic Gardens, Kew; Jardins Botaniques de la Ville de Paris) for facilitating the access to living material. B.A., P-H.G., C.P. and A.M-V. received financial support from the ‘Action Transversale du Muséum: Formes possibles, Formes realisées'. A.M-V. thanks the Bentham-Moxon Trust for funding.