Resolving basal lamiid phylogeny and the circumscription of Icacinaceae with a plastome-scale data set

Lamiidae are a major clade of asterid angiosperms, including ∼40000 species, or ∼15% of angiosperm species richness (Refulio-Rodriguez and Olmstead, 2014). Earlier studies have referred to this clade informally as asterids I (Chase et al., 1993), euasterids I (APG, 1998; Soltis et al., 1999a; APG II, 2003), or lamiids (Bremer et al., 2002; Judd and Olmstead, 2004; APG III, 2009). Although Cantino et al. (2007) applied the name Garryidae to this clade, it has more commonly been called Lamiidae (Refulio-Rodriguez and Olmstead, 2014), and this treatment will be followed in the Companion Volume of the PhyloCode (R. G. Olmstead, University of Washington, personal communication; note that all clade names are in italics following Cantino et al., 2007).

Lamiids are generally characterized by superior ovaries and corollas with late sympetaly (Erbar and Leins, 1996; Judd et al., 2008), but considerable variation in morphology occurs across the clade, and nonmolecular synapomorphies are unclear (Stevens, 2001 onward; Judd and Olmstead, 2004). Phylogenetic studies over the past 20 years (e.g., Olmstead et al., 1992, 1993, 2000; Chase et al., 1993; Soltis et al., 1999a, 2000, 2011; Savolainen, 2000a, b; Albach et al., 2001; Bremer et al., 2002; Refulio-Rodriguez and Olmstead, 2014) have clarified the composition of Lamiidae. However, relationships among the major lineages have largely eluded resolution. Most lamiid diversity falls into a clade informally known as the “core lamiids” (Refulio-Rodriguez and Olmstead, 2014; the formal name Lamianae will be applied to this clade in the upcoming Companion Volume to the PhyloCode, R. G. Olmstead, University of Washington, personal communication), which includes four species-rich groups—Boraginales, Gentianales, Lamiales, and Solanales—as well as the small, phylogenetically isolated genus Vahlia Thunb. Relationships among these five groups have varied across studies and rarely received strong bootstrap support (Olmstead et al., 1992, 1993, 2000; Soltis et al., 1999a, 2000, 2011; Savolainen, 2000a, b; Albach et al., 2001; Bremer et al., 2002), although a recent study by Refulio-Rodriguez and Olmstead (2014) found moderate support for (Gentianales, ((Solanales + Vahliaceae), (Boraginales + Lamiales))).

Subtending the core lamiids are multiple phylogenetically isolated lineages of uncertain placement, which we informally call the “basal lamiids” because they possibly comprise a grade at the base of the lamiid tree. These include Garryales (Eucommiaceae and Garryaceae), Icacinaceae, and the monogeneric families Metteniusaceae and Oncothecaceae. Relationships among the basal lamiids are poorly known largely because previous studies (e.g., Kårehed, 2001; González et al., 2007; Lens et al., 2008; Soltis et al., 2011; Byng et al., 2014; Refulio-Rodriguez and Olmstead, 2014) have included an insufficient sampling of characters and/or taxa to investigate this phylogenetic problem comprehensively.

The Icacinaceae, with ∼34 genera and 200 species (Kårehed, 2001; Byng et al., 2014), constitute the most diverse basal lamiid family. They are of particular importance for understanding basal lamiid phylogeny because the family, as currently recognized (Kårehed, 2001), is likely not monophyletic, with some members considered probably more closely related to other basal lamiid groups (Lens et al., 2008; Byng et al., 2014). The traditional circumscription of the family (referred to henceforth as Icacinaceae s.l.) has included ∼54 genera and 400 species, united largely by the presence of superior, unilocular ovaries with two pendant ovules, only one of which matures (Engler, 1893; Howard, 1940; Sleumer, 1942, 1969, 1971; Kårehed, 2001). The family was previously considered a rosid group and was placed with Celastraceae and Aquifoliaceae (Miers, 1852, 1864), although earlier authors (de Candolle, 1824; Bentham, 1841, 1862) had associated members of Icacinaceae s.l. with Olacaceae of Santalales (a superasterid; Soltis et al., 2011).

The most widely adopted infrafamilial classification of Icacinaceae s.l. was that of Engler (1893) and Sleumer (1942), who divided the family into four tribes—Icacineae (including >30 genera), Iodeae (Hosiea Hemsley & E. H. Wilson, Iodes Blume, Mappianthus Hand.-Mazz., Natsiatopsis Kurz, Natsiatum Buch.-Ham. ex Arn., and Polyporandra Becc.), Sarcostigmateae (Sarcostigma Wight & Arn.), and Phytocreneae (Chlamydocarya Baill., Miquelia Meisn, Phytocrene Wall., Polycephalium Engl., Pyrenacantha Wight, and Stachyanthus Engl.)—based largely on wood anatomical characters. However, various lines of morphological evidence—e.g., pollen (Dahl, 1952; Lobreau-Callen, 1972, 1973), leaf epidermal characters (van Staveren and Baas, 1973), nodal anatomy (Bailey and Howard, 1941a), and wood anatomy (Bailey and Howard, 1941b–1941d)—suggested that the aforementioned tribes, and especially Icacineae, might not be monophyletic. Bailey and Howard (1941b) instead organized the family into three informal groups based on vessel characters of the primary and secondary xylem and nodal anatomy (i.e., the presence of uni- vs. trilacunar nodes). Group I included ∼13 genera from the Icacineae characterized by trilacunar nodes and vessels of both the primary and secondary xylem with scalariform performations. Group II included ∼10 genera from the Icacineae characterized by trilacunar nodes and vessels of the secondary xylem with scalariform-porous perforations. Group III included ∼23 genera, representing all four tribes, characterized by unilacunar nodes and vessels of the secondary xylem with simple perforations (Bailey and Howard, 1941a, b).

Multiple phylogenetic studies have shown Icacinaceae s.l. to be highly polyphyletic (Soltis et al., 1999a, 2000; Savolainen et al., 2000b; Kårehed, 2001), with members falling near the base of either the lamiids or the campanulids (=euasterids II). Kårehed (2001) conducted the first family-wide phylogenetic investigation of Icacinaceae, based primarily on ndhF sequences—although a sparse sampling of several other loci (rbcL, atpB, and 18S rDNA) was also included—and ∼70 morphological characters across 45 of the ∼54 traditional genera. As a result, Kårehed (2001) transferred ∼18 genera to the campanulid families Cardiopteridaceae (Aquifoliales), Pennantiaceae (Apiales), and Stemonuraceae (Aquifoliales). The remaining 34 genera (only 16 of which were sampled for molecular characters) were provisionally retained in Icacinaceae by Kårehed (2001), although it was evident that these genera might not constitute a monophyletic group. The 34 genera appeared to comprise four clades—which Kårehed (2001) informally called the Apodytes, Cassinopsis, Emmotum, and Icacina groups—but the relationships among these groups and the other basal lamiid lineages were unclear.

More recent studies of Icacinaceae have included greater sampling of morphological/anatomical characters (Lens et al., 2008) or additional genera not included in previous studies (Angulo et al., 2013; Byng et al., 2014). However, these studies were still unable to clarify the circumscription of the family and relationships among basal lamiids. Nevertheless, Byng et al. (2014) confirmed the placement of Dendrobangia Rusby among the basal lamiids—as opposed to in Cardiopteridaceae (Aquifoliales), where it had been placed previously (Kårehed, 2001)—and resolved some relationships within the Icacina group.

Resolving lamiid relationships, particularly toward the base of the tree, is critical for establishing an improved classification system for this clade. It is also essential for interpreting patterns of character evolution across not only the lamiids but also the whole of the core asterids (Stevens, 2001 onward; Endress and Rapini, 2014), i.e., the clade comprising the lamiids + campanulids (∼80000 species). The core asterids have been referred to informally as euasterids (e.g., APG I, 1998) and formally as the Gentianidae (Cantino et al., 2007); we will use this latter name throughout the paper. In many respects, the basal lamiids differ morphologically from the core lamiid clade. For example, the core lamiids are variable in habitat and habit and characterized by showy, sympetalous flowers with epipetalous stamens and distinctly two-carpellate/loculate gynoecia; the fruits are variable but usually contain multiple relatively small seeds (Stevens, 2001 onward; Judd et al., 2008). In contrast, the basal lamiids are strictly woody (trees, shrubs, or lianas) and generally occur in tropical rainforest (Sleumer, 1971; Carpenter and Dickison, 1976; González et al., 2007; Hua and Howard, 2008). The flowers are small, with petal apices often inflexed in bud, varying degrees of perianth connation, and unilocular gynoecia (Sleumer, 1942; Howard, 1942b, d; González and Rudall, 2010). The fruits are large and fleshy, generally drupaceous, usually containing one large seed (Sleumer, 1942; Howard, 1942b, d; González and Rudall, 2010). If the basal lamiids do indeed form a grade leading to the core clade, this topology would parsimoniously suggest that the aforementioned characters are ancestral for the lamiids. Furthermore, many of these characters are also shared by members of Aquifoliales (Howard, 1942c, d, 1943a–1943c; Stevens, 2001 onward), which are sister to the rest of the campanulids (Soltis et al., 2011), suggesting that these morphological features might represent ancestral states for Gentianidae as a whole (see also Endress and Rapini, 2014).

To resolve basal lamiid phylogeny, we sequenced 50 plastid genomes across the core asterids, focusing on the basal lamiid genera, and combined these with publically available asterid plastome data for comprehensive phylogenetic analyses. The resulting data matrix comprised 112 accessions, including all families and 36 of the 38 currently recognized genera of basal lamiids. On the basis of our results, we present a phylogenetic classification for Icacinaceae and the basal lamiids, providing formal definitions for 10 clade names following the PhyloCode version 4c (Cantino and de Queiroz, 2010; http://www.ohio.edu/phylocode/toc.html). This treatment includes the conversion of six names already recognized under the International Code of Nomenclature for Algae, Fungi, and Plants (ICN; McNeill et al., 2012) (e.g., Icacinaceae Miers) and four new clade names. We also offer suggestions for the application of these clade names under a rank-based system (namely, the Angiosperm Phylogeny Group). Finally, we discuss the general implications of our results for understanding patterns of character evolution across Gentianidae.

MATERIALS AND METHODS

RESULTS

Of the 59113 DNA characters analyzed, 31051 were constant and 28062 were variable. We recovered nearly identical relationships across all analyses (both ML and Bayesian), with differences in topology restricted to areas of poor support—there were no strongly supported conflicting relationships among the major lineages. Of the four partitioning schemes in the ML analyses, the gene-codon scheme (219 partitions) had the highest likelihood score (−lnL = 598491.98) and is depicted in Figs. 1 and 2 and discussed in the text. The –lnL scores of the other ML analyses were 613007.34 (no partition), 609229.67 (codon partition), and 607903.56 (gene partition). The best –lnL score of the gene-codon partitioned Bayesian analysis was 596805.40; the other partitioning schemes did not reach stationarity. The posterior probabilities of the gene-codon partitioned Bayesian analysis are mapped onto the best ML tree in Figs. 1 and 2 and discussed in the text. All trees resulting from these analyses (except those already shown in the text) are available in the online supplemental files (online Appendices S5–S11).

The overall topology recovered is as follows (Figs. 1 and 2). Lamiidae are strongly supported as monophyletic. The core lamiids form a well-supported clade (BS 100/PP 1.0), but relationships among the major core lineages (i.e., Boraginales, Gentianales, Lamiales, Solanales, and Vahlia) are more poorly supported given their modest/low ML bootstrap values. However, the Bayesian posterior probability values are generally strong. Boraginales were recovered as sister to Gentianales (BS 50/PP 0.99), with these together being sister to a clade of Lamiales, Solanales, and Vahlia (BS 66/PP 1.0). Vahlia was recovered as sister to Lamiales but with very weak support (BS <50/PP 0.74).

The 33 genera comprising Icacinaceae sensu Kårehed (noted with asterisks in Fig. 1) form two distinct clades: one includes 21 genera and corresponds to the Icacina and Cassinopsis groups of Kårehed (2001); the other clade comprises 11 genera, including 10 icacinaceous genera, primarily from the Apodytes and Emmotum groups of Kårehed (2001), as well as Metteniusa embedded well within the clade, sister to Ottoschulzia Urb. The first Icacinaceae clade (including 21 genera) and Oncothecaceae form a well-supported clade (BS 100/PP 1.0) sister to the rest of the lamiids. The second clade of Icacinaceae (including Metteniusa) is placed with strong support (BS 90/PP 1.0) as sister to the remaining lamiids, with Garryales highly supported as sister to the core lamiids (BS 98/PP 1.0).

Within the first Icacinaceae clade, Cassinopsis Sond. is sister to a clade comprising most of the genera of the Icacina group (BS 78/PP 0.99). The Icacina group, in turn, includes four well-supported clades (Fig. 3), which are discussed in more detail below; in general, relationships are well resolved and well supported across the entire clade. Within the second major clade of Icacinaceae sensu Kårehed, Platea Blume + Calatola Standl. are strongly supported (BS 100/PP 1.0) as sister to a clade of genera from the Emmotum and Apodytes groups. However, Pittosporopsis Craib (which, until now, had never been included in a molecular phylogenetic analysis) and Metteniusa (Metteniusaceae) are nested in the Emmotum group, and Dendrobangia is nested in the Apodytes group, in all cases with strong support (Fig. 1).

DISCUSSION

PHYLOGENETIC CLASSIFICATION

Our results provide a well-resolved and strongly supported hypothesis of basal lamiid relationships. This improved phylogenetic framework offers an excellent opportunity to revise the classification of basal lamiids. Here we present phylogenetic definitions following the PhyloCode version 4c (Cantino and de Queiroz, 2010; http://www.ohio.edu/phylocode/toc.html), including the recognition of four new clade names, as well as the conversion of six names already recognized under the ICN. The clades named here are highlighted in boldface in Fig. 4, and the definitions for the clades are presented below. Following the definitions, we offer suggestions for the application of these clade names within the Angiosperm Phylogeny Group system, including a scheme of families and orders building on previous studies (e.g., Kårehed, 2001; Byng et al., 2014).

Metteniusidae G. W. Stull, D. E. Soltis, and P. S. Soltis, new clade name.

Definition: The least-inclusive clade containing Metteniusa edulis H. Karst. 1860 (Metteniusaceae), Garrya elliptica Douglas ex Lindl. 1834 (Garryales), and Gentiana acaulis L. 1753 (Lamianae). This is a node-based definition in which all the specifiers are extant. Abbreviated definition: < Metteniusa edulis H. Karst. 1860 & Garrya elliptica Douglas ex Lindl. 1834 & Gentiana acaulis L. 1753.

Etymology: Derived from Metteniusa (name of an included genus), established by Hermann Karsten in honor of German botanist Georg Heinrich Mettenius.

Reference phylogeny: This paper is the primary reference; see Figs. 1, 2, and 4. See also Burge (2011).

Composition: Metteniusaceae, Garryales, and Lamianae (core lamiids: Boraginales, Gentianales, Lamiales, Solanales, and Vahlia).

Diagnostic apomorphies: No non-DNA synapomorphies are currently known.

Synonyms: None.

Comments: This is a newly discovered clade, lacking a pre-existing name. The composition of Metteniusaceae (discussed more below) and the positions of Metteniusaceae and Garryales in relation to the Lamianae were poorly supported or unresolved in previous studies (e.g., Soltis et al., 2011; Byng et al., 2014; Refulio-Rodriguez and Olmstead, 2014). Our results provide strong support that Metteniusaceae (as here circumscribed) and Garryales are successively sister to Lamianae. However, if upon further study Metteniusaceae is found to be sister to Icacinales, the name Metteniusidae would become a synonym with the prior name Lamiidae. Morphological synapomorphies for Metteniusidae are currently unknown.

We chose Metteniusa edulis as an internal specifier because it is the type of Metteniusa, which is the basis of the name Metteniusidae. Although we did not include this species in our analyses—instead, we included Metteniusa tessmanniana Sleum. (Sleum.)—the monophyly of Metteniusa is supported by numerous morphological features (Lozano-Contreras and de Lozano, 1988). We chose Garrya elliptica as an internal specifier because it is the type of the genus. Although we did not include G. elliptica in this study, a recent study (Burge, 2011) demonstrated that Garrya is monophyletic and that G. elliptica is closely related to G. flavescens, which we included in our analyses. The last internal specifier, Gentiana acaulis, was included in our analyses and occupies a highly nested position in the phylogeny of Metteniusidae.

Garryidae R. G. Olmstead, W. S. Judd, and P. D. Cantino 2007: 836 [G. W. Stull, D. E. Soltis, and P. S. Soltis], converted clade name.

Definition: The least-inclusive clade containing Garrya elliptica Douglas ex Lindl. 1834 (Garryales) and Gentiana acaulis L. 1753 (Lamianae). This is a node-based definition in which all the specifiers are extant. Abbreviated definition: < Garrya elliptica Douglas ex Lindl. 1834 & Gentiana acaulis L. 1753.

Etymology: Derived from the included genus Garrya Douglas ex Lindl.

Reference phylogeny: This paper is the primary reference; see Figs. 1, 2, and 4. See also Bremer et al. (2002) and Refulio-Rodriguez and Olmstead (2014).

Composition: Garryales and Lamianae (core lamiids: Boraginales, Gentianales, Lamiales, Solanales, and Vahlia).

Diagnostic apomorphies: No non-DNA synapomorphies are currently known.

Synonyms: None.

Comments: The name Garryidae was originally applied to the lamiid/euasterid I clade as a whole (Cantino et al., 2007). However, Refulio-Rodriguez and Olmstead (2014) instead applied the name Lamiidae to the lamiid/euasterid I clade, and this procedure will be followed in the Companion Volume to the PhyloCode (R. G. Olmstead, University of Washington, personal communication). We therefore apply the name Garryidae to a less inclusive clade of lamiids, i.e., Garryales + Lamianae. Several previous studies recovered Garryales sister to the core lamiids, albeit generally with weak support (Bremer et al., 2002; Refulio-Rodriguez and Olmstead, 2014). Our analyses place Garryales sister to the core lamiids with strong ML and Bayesian support. Morphological synapomorphies for this clade, however, are currently unknown.

Garrya elliptica, one of the internal specifiers, represents the type of the genus Garrya, and it is closely related to the species of Garrya (G. flavescens) that we included in our phylogenetic analyses (Burge, 2011). The other internal specifier, Gentiana acaulis, was included in our phylogenetic analyses.

Garryales Lindl. 1835: 16 [G. W. Stull, D. E. Soltis, and P. S. Soltis], converted clade name.

Definition: The least-inclusive clade containing Garrya elliptica Douglas ex Lindl. 1834 (Garryaceae) and Eucommia ulmoides Oliv. 1890 (Eucommiaceae). This is a node-based definition in which all the specifiers are extant. Abbreviated definition: < Garrya elliptica Douglas ex Lindl. 1834 & Eucommia ulmoides Oliv. 1890.

Etymology: Derived from the included genus Garrya Douglas ex Lindl.

Reference phylogeny: Soltis et al. (1999a: supplemental fig. 11A) is the primary reference phylogeny. See also Soltis et al. (2000, 2011), Bremer et al. (2002), Savolainen et al. (2000a), Refulio-Rodriguez and Olmstead (2014), and this paper.

Composition: Garryaceae, Eucommiaceae.

Diagnostic apomorphies: Production of the latex gutta percha and dioecy.

Synonyms: None.

Comments: Numerous phylogenetic studies show that Aucuba (Garryaceae or sometimes treated in its own family, Aucubaceae Bercht. & J. Presl), Garrya (Garryaceae), and Eucommia (Eucommiaceae) form a well-supported clade (e.g., Soltis et al., 1999a, 2000, 2011; Refulio-Rodriguez and Olmstead, 2014; this paper). Although APG I (1998) recognized Garryales as including Aucubaceae, Eucommiaceae, Garryaceae, and Oncothecaceae, in APG II (2003) Oncothecaceae were excluded from the order and Aucubaceae were included in Garryaceae. This procedure was followed in APG III (2009). We apply the converted clade name Garryales to this same circumscription of taxa: Eucommiaceae (Eucommia) + Garryaceae (Garrya + Aucuba). The internal specifers selected—Garrya elliptica and Eucommia ulmoides—are the type species of their respective genera. The production of the latex gutta percha appears to constitute a synapomorphy of Garryales. Another possible synapomorphy of this clade is dioecy.

Icacinales Tiegh. ex Reveal 1993: 175 [G. W. Stull], converted clade name.

Definition: The least-inclusive crown clade containing Icacina oliviformis (Poir.) J. Raynal 1975 (Icacinaceae) and Oncotheca balansae Baill. 1891 (Oncothecaceae). This is a node-based definition in which all the specifiers are extant. Abbreviated definition: < Icacina oliviformis (Poir.) J. Raynal 1975 & Oncotheca balansae Baill. 1891.

Etymology: Derived from Icacina (name of an included genus), which refers to the resemblance of the type species, Icacina oliviformis (Poir.) J. Raynal (=Icacina senegalensis A. Juss.), to Chrysobalanus icaco L. of Chrysobalanaceae (Jussieu, 1823).

Reference phylogeny: This paper is the primary reference; see Figs. 1, 2, and 4.

Composition: Icacinaceae and Oncothecaceae.

Diagnostic apomorphies: No non-DNA synapomorphies are currently known.

Synonyms: None.

Comments: The sister relationship of Icacinaceae, as circumscribed here, and Oncothecaceae is a novel result, not found in previous studies of lamiid phylogeny (e.g., Bremer et al., 2002; González et al., 2007; Lens et al., 2008; Soltis et al., 2011; Refulio-Rodriguez and Olmstead, 2014; Byng et al., 2014), which have generally placed Oncotheca with various genera of Metteniusaceae, as circumscribed here, although never with strong support.

The use of Icacinales for the clade comprising Icacinaceae (as circumscribed here) and Oncothecaceae is a novel application of the name. Van Tieghem (1897) first proposed Icacinales as a new order, but this name was not validly published until more recently (Reveal, 1993). Van Tieghem's circumscription of Icacinales more or less corresponded to Icacinaceae s.l., which he divided into multiple families (Emmotaceae, Iodaceae, Icacinaceae, Leptaulaceae, Phytocrenaceae, Pleurisanthaceae, and Sarcostigmataceae). Takhtajan's (1997) circumscription of Icacinales included Icacinaceae s.l. and three other families (Aquifoliaceae, Phellinaceae, and Sphenostemonaceae), which are now recognized as distantly related from Icacinaceae (Soltis et al., 2011). Oncothecaceae had generally been placed in Theales (Cronquist, 1981; Takhtajan, 1997).

Icacina oliviformis, the type species of Icacina, is synonymous with Icacina senegalensis A. Juss. 1823, which is the original name upon which the genus Icacina was based. The epithet “oliviformis” (originally treated as Hirtella olivaeformis Poir. 1813) takes priority over “senegalensis” as it is the older name. Although Icacina oliviformis, one of the internal specifiers, was not included in our analyses, we did include Icacina mannii, which Byng et al. (2014) found to be sister to Icacina oliviformis with strong support. Note, however, that in Byng et al. (2014), I. oliviformis is listed under the synonym I. senegalensis. Previous studies (Kårehed, 2001; Bremer et al., 2002) including Icacina oliviformis (listed under the synonym Icacina senegalensis) also found this species to be related to other members of Icacinaceae as here circumscribed. Oncotheca balansae, which was selected as the other internal specifier for Icacinales, is the type of Oncotheca and was included in our phylogenetic analyses.

Icacinaceae Miers 1851: 174 [G. W. Stull], converted clade name.

Definition: The most-inclusive crown clade containing Icacina oliviformis (Poir.) J. Raynal 1975, Mappia racemosa Jacq. 1797, and Pyrenacantha malvifolia Engl. 1893 but not Oncotheca balansae Baill. 1891 (Oncothecaceae). This is a branch-modified node-based definition. Abbreviated definition: >∇ Icacina oliviformis (Poir.) J. Raynal 1975 & Mappia racemosa Jacq. 1797 & Pyrenacantha malvifolia Engl. 1893 ∼ Oncotheca balansae Baill. 1891.

Etymology: Derived from Icacina (name of an included genus), which refers to the resemblance of the type species, Icacina oliviformis (Poir.) J. Raynal (=Icacina senegalensis A. Juss.), to Chrysobalanus icaco L. of Chrysobalanaceae (Jussieu, 1823).

Reference phylogeny: This paper is the primary reference; see Figs. 1, 2, and 4. See also Kårehed (2001), Bremer et al. (2002), Angulo et al. (2013), Byng et al. (2014).

Composition: Alsodeiopsis, Casimirella, Cassinopsis, Desmostachys, Hosiea, Icacina, Iodes, Lavigeria, Leretia, Mappia, Mappianthus, Merrilliodendron, Miquelia, Natsiatopsis, Natsiatum, Nothapodytes, Phytocrene, Pleurisanthes, Pyrenacantha, Rhyticaryum, Sarcostigma, Sleumeria, and Stachyanthus.

Diagnostic apomorphies: No non-DNA synapomorphies are currently known.

Synonyms: None.

Comments: The name Icacinaceae was first proposed by Miers (1851) for ∼13 genera formerly treated as the Icacineae tribe of Olacaceae. Others (e.g., Engler, 1893; Bailey and Howard, 1941a; Sleumer, 1942) later applied the name Icacinaceae to a much larger assemblage of genera (∼54) and species (∼400). Based on molecular-morphological phylogenetic analyses, Kårehed (2001) recognized a much-reduced circumscription of Icacinaceae including ∼34 genera and 200 species. None of the aforementioned circumscriptions were monophyletic. We apply the name Icacinaceae to a clade of 23 genera. This represents the most-inclusive monophyetic assemblage of genera from Icacinaceae s.l. including Icacina.

Because the position of Cassinopsis sister to the rest of Icacinaceae is not fully supported, we did not choose this species as an internal specifier. Instead we adopted a branch-modified node-based definition with three other species as internal specifiers, affording flexibility to include/exclude Cassinopsis upon further analyses. If it becomes sister to Oncotheca, for example, Icacinaceae (as defined above) would still exist; its composition would simply change slightly, in that Cassinopsis would be excluded. This situation would render Icacinaceae and Icacinoideae synonyms, with Icacinaceae taking priority. The internal specifiers Mappia racemosa and Icacina oliviformis (=the type of Icacina and thus ultimately Icacinaceae) represent the two clades successively sister to the rest of the family (after Cassinopsis). The third internal specifier, Pyrenacantha malvifolia, occupies a highly nested position in the phylogeny.

Icacinoideae Engl. 1893: 242 [G. W. Stull], converted clade name.

Definition: The least-inclusive clade containing Icacina oliviformis (Poir.) J. Raynal 1975, Mappia racemosa Jacq. 1797, and Pyrenacantha malvifolia Engl. 1893. This is a node-based definition in which all of the specifiers are extant. Abbreviated definition: < Icacina oliviformis (Poir.) J. Raynal 1975 & Mappia racemosa Jacq. 1797 & Pyrenacantha malvifolia Engl. 1893.

Etymology: Derived from Icacina (name of an included genus), which refers to the resemblance of the type species, Icacina oliviformis (Poir.) J. Raynal (=Icacina senegalensis A. Juss.), to Chrysobalanus icaco L. of Chrysobalanaceae (Jussieu, 1823).

Reference phylogeny: This paper is the primary reference; see Figs. 1–4. See also Kårehed (2001), Lens et al. (2008), Angulo et al. (2013), and Byng et al. (2014).

Composition: All genera of Icacinaceae except Cassinopsis.

Diagnostic apomorphies: Unilacunar nodes and simple perforation plates.

Synonyms: No formal synonyms exist, although several informally named groups constitute close approximations to Icacinoideae: the Icacina group of Kårehed (2001) and group III of Bailey and Howard (1941b).

Comments: Engler (1893) used the name Icacinoideae for one of three subfamilies of Icacinaceae, the other subfamilies being the monogeneric Cardiopteridoideae and Lophopyxidoideae. Engler (1893) recognized 36 genera in Icacinoideae, which he divided among four tribes: Icacineae, Iodeae, Phytocreneae, and Sarcostigmateae. Subsequently, the circumscription of Icacinoideae was expanded to include >50 genera, while Cardiopteris Wall. ex Royle (Cardiopteridoideae) and Lophopyxis Hook.f. (Lophopyxidoideae) were recognized as dubious members of Icacinaceae (Sleumer, 1942; Bailey and Howard, 1941a). The above circumscriptions of Icacinoideae were shown to be polyphyletic by numerous phylogenetic studies (Savolainen et al., 2000a, b; Soltis et al., 2000; Kårehed, 2001).

We apply the name Icacinoideae to a clade comprising 22 genera. No formal name with a closer correspondence to this clade exists in the literature. Icacinoideae as here circumscribed includes 11 genera from the traditional Icacineae tribe and all genera of the traditional Iodeae, Phytocreneae, and Sarcostigmateae tribes, plus the newly described genus Sleumeria (Utteridge et al., 2005). All the genera comprising Icacinoideae were included in our analyses except Sleumeria and Natsiatopsis. However, Byng et al. (2014) confirmed the placement of Sleumeria in this clade based on phylogenetic analyses of the plastid loci matK, ndhF, and rbcL (although Sleumeria was only represented by ndhF in the analyses), and multiple lines of morphological evidence (mentioned in this paper) suggest that Natsiatopsis falls within this clade close to Natsiatum. Icacinoideae corresponds closely to several informal groupings suggested in earlier studies—i.e., the Icacina group of Kårehed (2001) and group III of Bailey and Howard (1941b). The internal specifiers Mappia racemosa and Icacina oliviformis represent the two clades successively sister to the rest of Icacinoideae. The third internal specifier, Pyrenacantha malvifolia, occupies a highly nested position in the phylogeny. This clade is diagnosed by unilacunar nodes and vessels with simple perforation plates. Cassinopsis, which is sister to Icacinoideae, is distinguished by having trilacunar nodes and vessels with sclariform perforation plates.

Metteniusaceae H. Karst. ex Schnizl. 1860: 142 [G. W. Stull], converted clade name.

Definition: The least-inclusive clade containing Apodytes dimidiata E. Mey. ex Arn. 1840, Metteniusa edulis H. Karst. 1860, and Platea parviflora A. O. Dahl 1952. This is a node-based definition in which all of the specifiers are extant. Abbreviated definition: < Apodytes dimidiata E. Mey. ex Arn. 1840 & Metteniusa edulis H. Karst. 1860 & Platea parviflora A. O. Dahl 1952.

Etymology: Derived from Metteniusa (name of an included genus), established by Hermann Karsten in honor of German botanist Georg Heinrich Mettenius.

Reference phylogeny: This paper is the primary reference; see Figs. 1, 2, and 4.

Composition: Apodytes, Calatola, Dendrobangia, Emmotum, Metteniusa, Oecopetalum, Ottoschulzia, Pittosporopsis, Platea, Poraqueiba, and Rhaphiostylis.

Diagnostic apomorphies: No non-DNA synapomorphies are currently known.

Synonyms: None.

Comments: Metteniusa has generally been treated as the sole constituent of its own family, Metteniusaceae. González et al. (2007) showed that Metteniusa represents a basal lamiid lineage, but their sampling was too skeletal to place the genus precisely among the basal lamiids. Phylogenetic analyses by Byng et al. (2014) suggested that Metteniusa is closely related to several genera of the Emmotum group of Icacinaceae sensu Kårehed (2001): Emmotum, Ottoschulzia, Oecopetalum, and Poraqueiba. Our results are consistent with Byng et al. (2014) in that Metteniusa is nested in the Emmotum group and further show that this clade is related to multiple additional genera of Icacinaceae s.l.: Apodytes, Calatola, Dendrobangia, Platea, and Rhaphiostylis. The name Metteniusaceae was selected for this clade (comprising 11 genera) because it is the oldest family name associated with this clade's constituent genera. Morphological synapomorphies of this clade are currently unknown. The internal specifers used in this definition were all included in our phylogenetic analyses, except Metteniusa edulis, the type of Metteniusa. Since the clade name Metteniusaceae is based on the genus name Metteniusa, the type of the genus must be included in the definition (Cantino et al., 2007). The monophyly of Metteniusa is supported by numerous morphological features (Lozano-Contreras and de Lozano, 1988), suggesting that Metteniusa edulis is likely closely related to Metteniusa tessmanniana, which was included in our analyses.

Plateoideae G. W. Stull, new clade name.

Definition: The least-inclusive clade including Calatola mollis Standl. 1923 and Platea latifolia Blume 1826. This is a node-based definition in which all of the specifiers are extant. Abbreviated definition: < Calatola mollis Standl. 1923 & Platea latifolia Blume 1826.

Etymology: Derived from the included genus Platea Blume.

Reference phylogeny: This paper is the primary reference; see Figs. 1, 2, and 4. See also Kårehed (2001) and Byng et al. (2014).

Composition: Calatola and Platea.

Diagnostic apomorphies: Possible synapomorphies for this clade include unisexual flowers and indeterminate inflorescences.

Synonymy: None.

Comments: Kårehed (2001) recovered a sister relationship between Calatola and Platea based on morphological phylogenetic analyses. Based on three plastid loci, Byng et al. (2014) also recovered this relationship with moderate support. The results in this paper corroborate these earlier analyses with strong support. Calatola mollis and Platea latifolia were selected as internal specifiers because the former was included in our phylogenetic analyses and the latter represents the type of Platea, the basis of the name Plateoideae.

Apodytoideae G. W. Stull, new clade name.

Definition: The least-inclusive clade including Apodytes dimidiata E. Mey. ex Arn. 1840, Dendrobangia boliviana Rusby 1896, and Rhaphiostylis ferruginea Engl. 1909. This is a node-based definition in which all of the specifiers are extant. Abbreviated definition: < Apodytes dimidiata E. Mey. ex Arn. 1840 & Dendrobangia boliviana Rusby 1896 & Rhaphiostylis ferruginea Engl. 1909.

Etymology: Derived from the included genus Apodytes E. Mey. ex Arn.

Reference phylogeny: This paper is the primary reference; see Figs. 1, 2, and 4. See also Kårehed (2001) and Byng et al. (2014).

Composition: Apodytes, Dendrobangia, and Rhaphiostylis.

Diagnostic apomorphies: No non-DNA synapomorphies are currently known.

Synonymy: None.

Comments: Several studies (Kårehed, 2001; Lens et al., 2008) have recovered a sister relationship between Apodytes and Rhaphiostylis, but these analyses did not include the Neotropical genus Dendrobangia. Byng et al. (2014) recovered Dendrobangia as sister to Apodytes + Rhaphiostylis with weak Bayesian support. Our analyses recovered Apodytes sister to Dendrobangia + Rhaphiostylis with strong ML and Bayesian support. Although our results differ from those of Byng et al. (2014) in the placement of Dendrobangia, both indicate that these three genera form a clade. Morphological synapomorphies for this clade are not currently known. The internal specifiers selected were included in our phylogenetic analyses. Apodytes dimidiata and Dendrobangia boliviana also constitute the type species of their respective genera.

Metteniusoideae G. W. Stull, new clade name.

Definition: The least-inclusive clade including Emmotum fagifolium Desv. ex Ham. 1825 and Metteniusa edulis H. Karst. 1860. Abbreviated definition: < Emmotum fagifolium Desv. ex Ham. 1825 & Metteniusa edulis H. Karst. 1860.

Etymology: Derived from Metteniusa (name of an included genus), established by Hermann Karsten in honor of German botanist Georg Heinrich Mettenius.

Reference phylogeny: This paper is the primary reference; see Figs. 1, 2, and 4. See also Duno de Stefano and Fernández-Concha (2011) and Byng et al. (2014).

Composition: Emmotum, Metteniusa, Oecopetalum, Ottoschulzia, Pittosporopsis, and Poraqueiba.

Diagnostic apomorphies: Possible synapomorphies for this clade include stamen connectives extending beyond the anther sacs, fleshy petals, and fruits with persistent styles.

Synonymy: None.

Comments: In his combined molecular-morphological analyses, Kårehed (2001) recovered relationships among Emmotum, Oecopetalum, Ottoschulzia, and Poraqueiba, which he informally called the Emmotum group. Kårehed's (2001) analyses, however, did not include Pittosporopsis or Metteniusa. A recent paper by Byng et al. (2014) found moderate support for a clade including the Emmotum group plus Metteniusa, but Pittosporopsis was not included in their analyses. Our analyses corroborate this general result and further show, with strong support, that Pittosporopsis belongs in this clade.

The internal specifier Emmotum fagifolium represents the type of the genus Emmotum. Although we did not include this species in our phylogenetic analyses—instead, we included Emmotum nitens (Benth.) Miers—a morphology-based phylogeny of the genus (Duno de Stefano and Fernández-Concha, 2011) found numerous morphological synapomorphies supporting its monophyly. The other internal specifier, Metteniusa edulis, represents the type of Metteniusa, the monophyly of which is supported by numerous morphological features (Lozano-Contreras and de Lozano, 1988).

ACKNOWLEDGEMENTS

G.W.S. thanks S. R. Manchester for his invaluable support and assistance throughout the execution of this project. The authors thank A. Liston, R. G. Olmstead, P. F. Stevens, and an anonymous reviewer for helpful suggestions that improved the manuscript and W. S. Judd for assistance with PhyloCode names. K. Perkins, J. Solomon, J. J. Wieringa, and G. Thijsse kindly facilitated sampling of herbarium material for DNA sequencing from FLAS, MO, WAG, and L, respectively. The authors thank M. Deyholos and J. Leebens-Mack for contributing multiple 1KP samples used in this paper, Z. K. Zhou and his students for help with fieldwork in China, M. J. Moore for providing unpublished data and advice on plastome assembly, M. A. Gitzendanner for assistance with library construction and data processing, and A. A. Crowl for help with analyses. This work was supported by NSF grant DEB-1310805 (Doctoral Dissertation Improvement Grant to S. R. Manchester, P.S.S., and G.W.S.), NSF grant DEB-0743457 (to P. Jørgensen), the American Society of Plant Taxonomists, the Society of Systematic Biologists, the Botanical Society of America, the Explorers Club, and the Torrey Botanical Society.