Five major shifts of diversification through the long evolutionary history of Magnoliidae (angiosperms)

Background With 10,000 species, Magnoliidae are the largest clade of flowering plants outside monocots and eudicots. Despite an ancient and rich fossil history, the tempo and mode of diversification of Magnoliidae remain poorly known. Using a molecular data set of 12 markers and 220 species (representing >75% of genera in Magnoliidae) and six robust, internal fossil age constraints, we estimate divergence times and significant shifts of diversification across the clade. In addition, we test the sensitivity of magnoliid divergence times to the choice of relaxed clock model and various maximum age constraints for the angiosperms. Results Compared with previous work, our study tends to push back in time the age of the crown node of Magnoliidae (178.78-126.82 million years, Myr), and of the four orders, Canellales (143.18-125.90 Myr), Piperales (158.11-88.15 Myr), Laurales (165.62-112.05 Myr), and Magnoliales (164.09-114.75 Myr). Although families vary in crown ages, Magnoliidae appear to have diversified into most extant families by the end of the Cretaceous. The strongly imbalanced distribution of extant diversity within Magnoliidae appears to be best explained by models of diversification with 6 to 13 shifts in net diversification rates. Significant increases are inferred within Piperaceae and Annonaceae, while the low species richness of Calycanthaceae, Degeneriaceae, and Himantandraceae appears to be the result of decreases in both speciation and extinction rates. Conclusions This study provides a new time scale for the evolutionary history of an important, but underexplored, part of the tree of angiosperms. The ages of the main clades of Magnoliidae (above the family level) are older than previously thought, and in several lineages, there were significant increases and decreases in net diversification rates. This study is a new robust framework for future investigations of trait evolution and of factors influencing diversification in this group as well as angiosperms as a whole. Electronic supplementary material The online version of this article (doi:10.1186/s12862-015-0320-6) contains supplementary material, which is available to authorized users.

Within tribe Bocageeae, genera Bocagea, Cardiopetalum, and Froesiodendron have never been included in a phylogenetic analysis. However, synapomorphies of tribe Bocageeae occur in these three genera, leading Chatrou et al. [1] to argue that they could be securely placed in this monophyletic group. We followed their point of view by incorporating these three genera within one compartment corresponding to the entire tribe Bocageeae. We did not exclude this diversity because it represents 12 % of the species diversity of the defined compartment.
In tribe Duguetieae, Duckeanthus has been only included in a morphological cladistic analysis [20]. In their results, the relationships among the remaining genera of the tribe were not in accordance with those supported in recent molecular analyses [1,3]. Because this genus incorporates only one species, more investigations are needed to confirm its position within the tribe, and a conservative compartment would incorporate more than 99 species (the number of species in Duguetieae), we decided to ignore this species in our analyses.
In tribe Annoneae, the monophyly of Neostenanthera and Asimina has not been tested before.
Consequently, we treated Annoneae as a single compartment in our analyses. This supra-generic compartment has been well supported in the literature [1,3,8] and in our BEAST analyses [see Because there is only one species in this genus, we preferred to exclude it. The genus Diclinanona, also placed in Annoneae, was not sampled in our molecular dating analyses. Its phylogenetic position has been debated in the literature [8,21]. According to Chatrou et al. [ Afroguatteria have never been included in a molecular phylogenetic analysis [1]. Doyle and Le Thomas [22] placed this genus as the sister group to Uvaria in a morphological cladistic analysis in which relationships among genera of Annonaceae were not compatible with the current phylogenetic knowledge of the group. However, because a secure placement of the two species of this genus requires more investigation and the genus Uvaria contains 187 species, we ignored Afroguatteria from our analyses. Cleistochlamys and Gilbertiellia have never been included in a phylogenetic analysis and their placement within Annonoideae was based on an intuitive approach [1]. Because both of them are monotypic genus and represent less than three percent of any secure compartments in which we could incorporate their diversity, we ignored them from our analyses. Because the monotypic genus Schefferomitra has not been placed in a published phylogeny, we preferred to exclude this genus from our analyses. Exellia, not included in our molecular dating analyses, has been placed in an unresolved position in Uvariaeae [1]. Because this is a monospecific genus we decided to exclude it. Pyramidanthe has been placed with a phylogenetic approach in a clade with Dasymaschalon, Desmos, Dielsiothalamnus, Fissistigma, Friesodielsia, Mitrella, Monanthotaxis, Sphaerocoryne, Toussaintia and Uvaria, but the relationships within among these genera remained poorly resolved [23]. Because the incorporation of this monospecific genus would require a conservative compartment consisting of all Uvarieae (at least 415 species), we ignored it.
In tribe Piptostigmateae, Piptostigma was shown to be paraphyletic with respect to Polyceratocarpus [7], even though Chatrou et al. [1] provisionally maintained the two genera.
We defined a compartment including the diversity of the two genera (PP=100% in the present study, see Additional file 2).
In tribe Malmeeae the three genera Bocageopsis, Onychopetalum, and Unonopsis form a clade (PP=100 % in the present study, see Additional file 2), but the relationships among them remain unclear [1,3,8,11]. Because the monophyly of Onychopetalum has never been tested before, we defined a compartment including these three genera (BOU, see Additional file 2). In the same tribe, the monophyly of Pseudomalmea has never been tested either and its relationships are not well established in the present study. In addition, Oxandra has been shown to be polyphyletic even though the support values associated with this polyphyly are low [1,14]. All the genera of Aristolochiaceae (incl. Lactoris) were sampled in our study. However, Hydnoraceae were excluded (see Molecular dataset section in the materials and methods).
Previous studies have placed this parasitic family within Aristolochiaceae [3,33] and, more recently, Naumann et al. [34] refined its position as sister to subfamily Aristolochioideae. In order to incorporate the nine species of Hydnora [31] and the three species of Prosopanche [32] in our analyses, we should define a broader compartment to include Hydnoraceae and All genera of the family were sampled in our chronograms. In the phylogenetic analysis of Salazar and Nixon [38], Cinnamodendron was paraphyletic and the monophyly of Cinnasmoma Degeneriaceae include a single genus, Degeneria, with two species [42]. The monophyly of this taxon has been tested with matK and ndhF markers with the separate molecular datasets of Massoni et al. [3]. Because the genus appears to be monophyletic (J. Massoni, unpubl [46] Hernandiaceae incorporate 62 species in five genera [46]. The monophyly of the four nonmonospecific genera has been well supported in the literature [46]. Lauraceae is the largest family within Magnoliidae, uncluding 2500 to 3500 species [57]. The relationships within this clade are still debated in the literature [52,64,83]. However, several clades are well supported. This is the case for the Perseeae-Laureeae clade [64], named core Lauraceae by Rohwer and Rudolph [65]. Within the core Lauraceae, Li et al. [52] focused on the phylogeny of the Persea group as defined by Rohwer et al. [73]. In addition to the paraphyly of Alseodaphne and Dehaasia, they found Persea to be paraphyletic, with Apollonias nested in. In our chronograms, these two genera were sampled and we find them to be nested in a wellsupported core Lauraceae clade. However, there is a well-supported conflict among the present study and the literature requiring a larger compartment than core Lauraceae. Nothaphoebe, which was nested in this clade in previous studies [52,73], is found in a different position (outside core Lauraceae) in the present study (see also Massoni et al. [3]). This genus, consisting in about 40 species [52], has never been included in a phylogenetic analysis of Lauraceae as a whole, and a maximum of two species (Nothaphoebe umbellifora and N. semecarpifolia) have been simultaneously included in the same phylogenetic analysis [52,73]. In the present study and in Massoni et al. [3], we sampled a different species in order to represent the genus (N. konishii).
Because this is the first introduction of this species in a phylogenetic study, and because more than 95 % of the diversity of this genus has never been included in a phylogenetic approach, it is difficult to affirm or disconfirm a misidentification of the taxa used to generate the sequence.
Because more investigations are needed to elucidate this question, we preferred to define a conservative compartment including all Lauraceae. Indeed, because an infinitesimal part of the species diversity of Lauraceae has been sampled in previous phylogenetic studies, its distribution in smaller compartments will be problematic. The taxonomy within the family varies among authors. The number of genera recognized ranges from eight to five. In the present study, we followed the taxonomic revisions of Vink [109,110], in which five genera of Winteraceae are recognized. The monophyly of these taxa have been tested and confirmed in Marquínez et al. [105] and Pratt [106]. Zygogynum was not sampled in our chronograms. Because in the literature this genus is sister to Pseudowintera [105, 106, 111, 112], we incorporated the two genera in a single compartment (PZ, see Additional file 2).