Phylogenomics and biogeography of Cunoniaceae (Oxalidales) with complete generic sampling and taxonomic realignments.

PREMISE: Cunoniaceae are a family of shrubs and trees with 27 genera and ca. 335 species, mostly confined to tropical and wet temperate zones of the southern hemisphere. There are several known issues regarding generic limits, and the family also displays a number of intriguing long-range disjunctions.
METHODS: We performed a phylogenomic study using the universal Angiosperms353 probe set for targeted sequence capture. We sampled 37 species covering all genera in the Cunoniaceae, and those in the three closely related families of the crown Oxalidales (Brunelliaceae, Cephalotaceae, and Elaeocarpaceae). We also performed analyses for molecular dating and ancestral area reconstruction.
RESULTS: We recovered the topology (Cunoniaceae, (Cephalotaceae, (Brunelliaceae, Elaeocarpaceae))) and a well-resolved genus-level phylogeny of Cunoniaceae with strongly supported clades corresponding to all previously recognized tribes. As previously suspected, the genera Ackama and Weinmannia were recovered as paraphyletic. Australasia was inferred as the likely ancestral area for the family.
CONCLUSIONS: The current distribution of Cunoniaceae is best explained by long-distance dispersal with a few possible cases of Australasian-American vicariance events. Extinctions may have been important in determining the mostly Oceanian distribution of this family while some genera in the tribe Cunonieae and in New Caledonia have undergone recent bursts of diversification. New generic diagnoses, 80 new combinations, and one new name are provided for a recircumscribed Ackama (including Spiraeopsis), a much smaller Weinmannia (mostly New World), and a resurrected Pterophylla to accommodate Old World taxa previously in Weinmannia.

Cunoniaceae are a medium‐sized family of flowering plants with 27 genera and ca. 335 species (Bradford et al., 2004) of shrubs and trees mostly confined to tropical (including montane) and wet temperate zones of the southern hemisphere. The greatest diversity of species and genera is found in Oceania, particularly New Guinea, eastern Australia, and New Caledonia. The family has a rich fossil record, especially in Australia (Barnes et al., 2001), but also in other areas where it is no longer extant, such as Antarctica (Poole et al., 2003), Sweden (Schönenberger et al., 2001), and Burma (Chambers et al., 2010). Cunoniaceae are now placed in the order Oxalidales (APG IV, 2016), along with six other families (Fig. 1). Huaceae (2 genera and 3 spp., tropical Africa) appears to be sister to the rest of the order (Soltis et al., 2011), which is composed of two well‐supported clades: Connaraceae + Oxalidaceae, and Brunelliaceae + Cephalotaceae + Cunoniaceae + Elaeocarpaceae (Wang et al., 2009; Soltis et al., 2011). The relationships within the latter clade, which could be referred to as crown Oxalidales or Elaeocarpineae Engler (1898, p. 148), remain unresolved, and therefore the sister group of Cunoniaceae is uncertain.

FIGURE 1

One representative of each of the four crown families of Oxalidales, clockwise from top left: L. (Cunoniaceae), drawn by Matilda Smith and reproduced from Curtis’ Bot. Mag. vol. 139, tab. 8504 (1913); Labill. (Cephalotaceae), drawn by Louis‐Constantin Stroobant and reproduced from L. B. van Hoote, Flore des serres et des jardins de l’Europe, vol. 3, tab. 8 (1847); Cuatrec. (Brunelliaceae), artist unknown. Reproduced from the Project to digitize the drawings of the Royal Botanical Expedition of the New Kingdom of Granada (1783–1816), directed by José Celestino Mutis: www.rjb.csic.es/icones/mutis. Royal Botanic Garden‐CSIC; Sm. (Elaeocarpaceae), drawn by Walter Hood Fitch and reproduced from Curtis’ Bot. Mag. vol. 78, tab. 4680 (1852).

Hufford and Dickison (1992) conducted the first cladistic analysis of the family, using morphological and anatomical characters. A molecular phylogenetic analysis, based on plastid trnL and rbcL sequences, was later published, including 23 of the currently accepted genera in the family (Bradford and Barnes, 2001). This study confirmed the placement of Bauera Banks ex Andrews, Davidsonia F.Muell., and Eucryphia Cav. within Cunoniaceae, three genera that have often been placed in their own monotypic families. This study also enabled the delimitation of several monophyletic tribes: Caldcluvieae, Codieae, Cunonieae, Geissoieae, Schizomerieae, and Spiraeanthemeae, although the position of some other genera (e.g., Eucryphia, Acrophyllum Benth.) remained unresolved. This phylogenetic framework paved the way for subsequent, more narrowly targeted studies. In a phylogenetic analysis of the tribe Cunonieae (Cunonia L., Pancheria Brongn. & Gris, Vesselowskya Pamp., Weinmannia L.), Bradford (2002) found support for the division of Weinmannia, the largest genus of the family (ca. 150 species), into five monophyletic sections, while only recovering weak support for the monophyly of the entire genus. A new, monotypic genus from New Caledonia, Hooglandia McPherson & Lowry, was later discovered and described (McPherson and Lowry, 2004), with molecular phylogenetics indicating that it occupied an isolated position (Sweeney et al., 2004). A phylogenetic analysis of Spiraeanthemeae (Pillon et al., 2009a) suggested that two genera, Acsmithia Hoogland and Spiraeanthemum A.Gray, were not monophyletic and were therefore reunited. A later study of Geissoieae (Hopkins et al., 2013), including Geissois Labill., Pseudoweinmannia Engl., and the first published molecular data for Lamanonia Vell., led to the placement of the two Australian species of Geissois in the newly described Karrabina Rozefelds & H.C.Hopkins. Before the present work was initiated, no molecular data were available for the New Guinean endemics Aistopetalum Schltr. (2 spp.) and Opocunonia Schltr. (1 sp.), and in addition, questions regarding the monophyly and, hence, the limits of Ackama A.Cunn. and Weinmannia remained unclear (Y. Pillon, unpublished data).

Most phylogenetic studies of Cunoniaceae have been conducted with a handful of plastid (Bradford and Barnes, 2001) or nuclear (Pillon et al., 2009b) genes that lacked the power to resolve recalcitrant nodes. High‐throughput DNA sequencing permits data gathering from a large number of loci, thereby increasing significantly the resolution of phylogenetic studies. Plastome phylogenomic studies (e.g., Givnish et al., 2010; Drew et al., 2014; Li et al., 2019) have become common, but they rely on a number of loci that are tightly linked. Cytoplasmic gene flow implies that such approaches could be misleading (Rieseberg and Soltis, 1991), particularly between closely related species, but also even at higher ranks, e.g., between orders (Sun et al., 2015). Therefore, approaches using many unlinked nuclear loci are expected to yield more robust phylogenies. One such approach, targeted sequence capture, has now been made accessible to all angiosperm researchers through the development of a universal probe set, known as Angiosperms353 (Johnson et al., 2019). Also, targeted sequence capture employs short‐read sequencing, which permits the use of degraded genomic DNA, such as that typically found in and extracted from herbarium specimens, i.e., museomics (e.g., Zedane et al., 2016), which can be a good substitute for fresh material for species that are extinct or difficult to procure. The efficacy of target sequence capture using Angiosperms353 on DNA from herbarium specimens has been well demonstrated (Brewer et al., 2019).

The primary aims of this study were to (1) clarify the relationships between Cunoniaceae and its closest relatives in the crown Oxalidales (viz Brunelliaceae, Cephalotaceae, and Elaeocarpaceae), (2) produce a phylogenetic tree of Cunoniaceae based on a sampling of every genus, (3) investigate the limits of two particularly problematic genera: Ackama and Weinmannia, and (4) investigate the biogeographical history of this predominantly southern hemisphere group. Secondary goals were to ensure that Cunoniaceae genera for which we sampled several species were monophyletic and to make the necessary taxonomic and nomenclatural adjustments to genera and species on the basis of a strongly supported phylogeny that was further supported by morphological characters whenever possible.

MATERIALS AND METHODS

Sampling

The sampling included all 27 genera currently recognized within Cunoniaceae (following Bradford et al., 2004; Pillon et al., 2009a; Hopkins et al., 2013), as well as all of the genera of the related families: Brunelliaceae (monogeneric, Kubitzki, 2004), Cephalotaceae (monogeneric, Conran, 2004), and Elaeocarpaceae (12 genera, Coode, 2004), except Sloanea L. in the last family, for which sequencing was not successful (Appendix S1). Within Cunoniaceae, we sampled all five sections within Weinmannia (sect. Weinmannia, Fasciculatae, Inspersae, Leiospermum, Spicatae; sensu Bradford, 1998, 1998). To test generic monophyly and for biogeographic analysis, one species from each geographical area for those taxa with a disjunct distribution was included in the sampling: Cunonia (southern Africa and New Caledonia), Eucryphia (South America and Australia), Ackama (Australia and New Zealand), and Weinmannia sect. Weinmannia (Americas and Mascarenes). Two species of Connaraceae, Manotes expansa Sol. ex Planch. and Rourea calophylla (Gilg ex G. Schellenb.) Jongkind, and one of Oxalidaceae, Sarcotheca macrophylla Blume, were chosen as outgroup taxa. Accessions were sourced from the DNA & Tissue Bank at the Royal Botanic Gardens, Kew, or were selected from herbarium specimens.

DNA extraction, library preparation, hybridization, and sequencing

DNA was extracted using a modified CTAB protocol (Doyle and Doyle, 1987) and purified using Mag‐Bind TotalPure NGS (Omega Bio‐tek, Norcross, GA, USA). The quality and concentration of the DNA extracts were assessed using a 1.5% agarose gel (to evaluate average fragment size) and a Qubit 3.0 fluorometer (Life Technologies, Carlsbad, CA, USA). DNA extracts with fragment sizes above 350 bp were sheared using a Covaris M220 Focused‐ultrasonicatorTM with Covaris microTUBES AFA Fiber Pre‐Slit Snap‐Cap (Covaris, Woburn, MA, USA). Dual‐indexed libraries for Illumina sequencing were prepared using the DNA NEBNext UltraTM II Library Prep Kit using half the recommended volume, with Dual Index Primers Set 1, NEBNext Multiplex Oligos for Illumina (New England BioLabs, Ipswich, MA, USA). The quality of the resulting libraries was evaluated on an Agilent Technologies 4200 TapeStation System using High Sensitivity D1000 ScreenTape (Agilent Technologies, Santa Clara, CA, USA). Libraries were pooled (equimolar 1 μg per pool) and enriched using the Angiosperms353 probe kit (Catalog #308196; Johnson et al., 2019) following the manufacturer’s protocol v4 (4.0; http://www.arborbiosci.com/mybaits‐manual). Hybridizations were performed at 65°C for 28–32 h in a Hybex Microsample Incubator (SciGene, Sunnyvale, CA, USA) and using red Chill‐out Liquid Wax (Bio‐Rad, Hercules, California, USA) to prevent evaporation. Enriched products were amplified with KAPA HiFi 2X HotStart ReadyMix PCR Kit (Roche, Basel, Switzerland) for 10 cycles. PCR products were then cleaned using the QIAquick PCR purification kit (Qiagen, Hilden, Germany). Products were quantified with a Qubit 3.0 fluorometer, in some cases, reamplified a second time between 3 and 8 cycles. Final products were run on an Agilent Technologies 4200 TapeStation System using High Sensitivity D1000 ScreenTape to assess quality and average fragment size. Several pooled libraries were multiplexed, and sequencing was performed at the Royal Botanic Gardens, Kew on an Illumina MiSeq (Illumina San Diego, CA, USA) with v 3 reagent chemistry (2 × 300‐bp paired‐end reads) or at Macrogen (Takeley, UK) on an Illumina HiSeq to produce 2 × 150‐bp paired‐end reads.

Read mapping and sequence alignments

The reads of the sequencing output (.fastq) were trimmed using Trimmomatic (Bolger et al., 2014) to remove reads with a quality score below 30 and reads that had any 4‐bp window below 30, retaining reads with at least 36 bp (LEADING: 30 TRAILING: 230 SLIDING WINDOW:4:30 MINLEN:36). The MINLENGTH setting was also used with length set to 36 to remove shorter reads that might not be uniquely positionable against other sequences. Paired reads and combined unpaired reads were used to recover target sequences using HybPiper version 1.3 (Johnson et al., 2016) using a target file available at https://github.com/mossmatters/Angiosperms353. Reads were mapped to de‐gapped medoid sequences using BLASTX (Camacho et al., 2009), each gene was assembled de novo using SPAdes (Bankevich et al., 2012), and coding sequences were extracted using Exonerate (Slater and Birney, 2005). HybPiper was run using the BLAST option (Altschul et al., 1990) because it has been found to produce longer sequences (Murphy et al., 2020). Noncoding sequences (i.e., introns and untranslated regions [UTRs]) flanking the coding sequences were recovered using the script intronerate.py available with HybPiper. Gene matrices were aligned separately using MAFFT V7 (mafft‐7.419‐gcc_fc6.x86), with accuracy‐oriented methods (‐‐localpair ‐‐maxiterate 1000) and the option to generate reverse complement sequences to align them together with the remaining sequences based on 6‐mer counting (‐‐adjustdirectionaccurately). Matrices were subsequently trimmed using phyutility (https://github.com/blackrim/phyutility) to delete sites that were missing 80% data (‐clean 0.8). Gene trees from trimmed matrices were generated using IQtree V1.6.12 (Minh et al., 2020), using ultrafast bootstrap with partition models (Chernomor et al., 2016). In a first iteration, generated gene trees were evaluated using TreeShrink 1.3.1 (Mai and Mirarab, 2018) to identify and exclude branches that increased the diameter of each gene tree by more than 20% using centroid re‐rooting (‐b 20 ‐c). Each locus was then realigned, trimmed, and analyzed using IQtree with bipartition support assessed with 1000 UFBoot2 (Hoang et al., 2018) bootstrap replicates (‐bb 1000), while collapsing branches with support values below 10% (Mirarab, 2019 [Preprint]) using Newick Utilities 1.6 (Junier and Zdobnov, 2010). We additionally excluded genes that contained data for fewer than 25% of species. A species tree was constructed from the set of trees produced based on the supercontigs (exons + introns) individually produced with IQ‐Tree. Tree and extensive branch annotations were generated using ASTRAL‐II (Mirarab and Warnow, 2015) using alternative quartet topologies (‐t 2): indicating the local posterior probabilities of the percentage of quartets in gene.

Divergence time estimates

To limit the effects of rate and topology heterogeneity between genes on divergence time estimations and the computational times required for such analysis, we first selected a set of genes using Sortadate (Smith et al., 2018). Genes chosen were those that were at least 10% concordant (bipartition >0.1) with the species tree produced with ASTRAL, had a root‐to‐tip variation less than 0.003, and had a tree length exceeding 2.1. The threshold for the agreement with the species tree was based on those used in previous studies (e.g., Shee et al., 2020), while those for the root‐to‐tip variation and tree length were determined using the median of the values obtained with Sortadate for all gene trees.

Divergence times were estimated using the Bayesian approach implemented in BEAST v1.10.4 (Suchard et al., 2018) on the CIPRES Science Gateway V. 3.3 (https://www.phylo.org/). We used the selected genes as independent partitions with their specific DNA substitution models and clock models, both unlinked. The topology of the tree was constrained to the relationships retrieved in the ASTRAL analysis. We used an uncorrelated relaxed clock with a lognormal distribution prior and a Yule process tree prior. Six fossil taxa were used for calibration using a normal distribution prior (Table 1). Five independent runs were conducted with 100 million generations, sampled every 2000 generations. Parameter convergence and appropriate effective sample sizes were verified in Tracer v1.7 (Rambaut et al., 2018). Tree files were combined in LogCombiner with a burnin of 25% (based on results visualized in Tracer), while the maximum credibility tree and associated posterior probabilities were computed in TreeAnnotator (Suchard et al., 2018). Diversification rates were calculated using equation 4 of Magallón and Sanderson (2001).

TABLE 1

List of the six fossil taxa used as calibration points in the molecular dating analysis of Cunoniaceae and relatives. Ages are reported in millions of years. SD, standard deviation.

Fossil taxa	Period	Position	Mean age, Myr (SD)	Reference
1. Tropidogyne	Early Cretaceous (Upper Albian)	Crown node of Oxalidales	100.5 (1.0)	Chambers et al. (2010)
2. Lacinipetalum	Early Paleocene	Stem node of tribe Schizomerieae	66.0 (1.0)	Jud et al. (2021)
3. Eucryphia	Late Paleocene	Stem node of Eucryphia	66.0 (1.0)	Barnes et al. (2001)
4. Codia	Middle Eocene‐Oligocene	Crown node of Codia and Callicoma	47.8 (1.0)	Barnes et al. (2001)
5. Vesselowskya	Early Oligocene	Stem node of Vesselowskya	33.9 (1.0)	Barnes et al. (2001)
6. Elaeocarpus	Early Oligocene	Crown node of Elaeocarpus	33.9 (1.0)	Crayn et al. (2006)

Biogeographical patterns

Ancestral range estimation for Cunoniaceae was performed using the dispersal–extinction cladogenesis (DEC) (Ree and Smith, 2008) and DEC+J models as implemented in the R package BIOGEOBEARS (Matzke, 2013). Geographic areas, as defined by Buerki et al. (2011), were used with the addition of Antarctica as one of the available regions because of its potential role in the past as a dispersal route for plants and animals (de la Estrella et al., 2017, 2019). The adopted area of delimitation approximates that of Buerki et al. (2009) and de la Estrella et al. (2019) with some modifications as follows: A, Africa; B, Madagascar (including Comoro and Mascarene Islands); C, Australia/New Guinea; D, New Caledonia; E, New Zealand; F, Americas; G, West Malesia (including India and Sri Lanka); H, Pacific Islands; I, Antarctica (see Appendix S2 for detailed definition). India and Sri Lanka were included within West Malesia to reduce the number of areas in our analysis, as only one sampled genus occurs in India (Elaeocarpus). Area assignments for each terminal are listed in Appendix S3. The biogeographical models proposed by Buerki et al. (2011) and de la Estrella et al. (2019) with five time slices reflecting the probability of area connectivity through time served as the basis for the development of a model tailored for Cunoniaceae (Appendix S2). Four of the time slices from de la Estrella et al. (2019) were employed here because the oldest one, 160–125 Ma, is older than the time spanned by the current analysis (i.e., 100.2 Ma). We performed these biogeographical analyses including and excluding Antarctica as an area to ascertain its impact on the estimation of the biogeographical patterns. The outgroup taxa were removed before these analyses.

RESULTS

Cunoniaceae were recovered as the sister group of a clade comprising (Brunelliaceae, (Cephalotaceae, Elaeocarpaceae)) (Fig. 2). The sister relationship of Cephalotaceae to the Brunelliaceae + Elaeocarpaceae clade is only moderately supported (local posterior probability, LPP = 0.97; quartet values q1 = 0.45, q2 = 0.31, q3 = 0.24). Within Cunoniaceae, most nodes are strongly supported (i.e., LPP = 1) with a few exceptions. Spiraenthemum, Hooglandia, Aistopetalum, and Bauera were the successive sister groups to the rest of the Cunoniaceae, although the placement of Bauera is only weakly supported (LPP = 0.36; q1 = 0.51, q2 = 0.27, q3 = 0.22).

FIGURE 2

Genus‐level phylogenetic tree of Cunoniaceae and its relationships with the closely related families Brunelliaceae, Cephalotaceae, and Elaeocarpaceae. Tree based on the universal Angiosperms353 probe set for targeted sequence capture. Numbers below branches represent local posterior probability values, and pie charts indicate quartet support.

The tribes Caldcluvieae (Ackama, Caldcluvia D.Don, Opocunonia, Spiraeopsis Miq.), Codieae (Callicoma Andrews, Codia J.R.Forst. & G.Forst., Pullea Schltr.), Cunonieae (Cunonia, Pancheria, Vesselowskya, Weinmannia), Geissoieae (Geissois, Karrabina, Lamanonia, Pseudoweinmannia), Schizomerieae (Anodopetalum A.Cunn. ex Endl., Ceratopetalum Sm., Platylophus D.Don, Schizomeria D.Don) were all recovered as monophyletic with strong support (LPP = 1; except for tribe Codieae, LPP = 0.95). Davidsonia was the sister of tribe Schizomerieae, but this relationship was only moderately supported (LPP = 0.92; q1 = 0.42, q2 = 0.33, q3 = 0.26). Within tribe Schizomerieae, the relationships between the genera differed from the published studies based on two plastid genes. For instance, Anodopetalum was previously recovered as the sister group to Platylophus with strong support (Bradford and Barnes, 2001; Hopkins et al., 2013), while here Ceratopetalum is the sister group to Platylophus with strong support (LPP = 1).

A strongly supported (LPP = 1) “core Cunoniaceae” (sensu Bradford et al., 2004), including Acrophyllum, Eucryphia, Gillbeea F.Muell., and the tribes Caldcluvieae, Codieae, Cunonieae, and Geissoieae, was recovered as in previous studies (Bradford and Barnes, 2001; Hopkins et al., 2013). Relationships resolved within tribe Codieae agree with previous studies (Bradford and Barnes, 2001; Hopkins et al., 2013). The relationships between the four genera of tribe Geissoieae were here only moderately supported (Fig. 2), with Lamanonia and Geissois as sister groups (LPP = 0.86), and Karrabina and Pseudoweinmannia as sister groups (LPP = 0.70). Opocunonia was recovered in the tribe Caldcluvieae as the sister group to Caldcluvia (LPP = 0.93). In this tribe, the genus Ackama was paraphyletic with respect to Spiraeopsis Miq., with the Australian and New Zealand species of Ackama forming separate subclades, based on our sampling.

In Cunonieae, Weinmannia formed two distinct clades (Fig. 2). The species belonging to the four Old World sections (sect. Fasciculatae, Inspersae, Spicatae, and Leiospermum) formed a monophyletic group (LPP = 1) sister to Cunonia + Pancheria (LPP = 1). The fifth section, composed of taxa from the Americas and the Mascarenes (sect. Weinmannia), was sister to this assemblage. The two sections of Weinmannia endemic to Madagascar and the Comoros (Inspersae and Spicatae) formed a monophyletic group sister to section Leiospermum from the Pacific Islands. Several groups showing major geographic disjunctions were also recovered: Cunonia (southern Africa, New Caledonia), Eucryphia (South America, Australia), and Weinmannia sect. Weinmannia (Americas, Mascarenes) were all monophyletic. The other genera endemic to America (Caldcluvia, 1 sp., Lamanonia, 6 spp.) and South Africa (Platylophus, 1 sp.) all have their closest relatives in Oceania.

A set of 41 gene trees were selected based on their concordance, which was further reduced to include only those that comprised at least 48 of the 56 taxa (86%) included in this study. This resulted in the set of 14 genes totalling 51,614 characters that were used in the molecular dating analyses.

All parameters from the combined BEAST analyses reached convergence, except for a few for which the effective sample size was slightly under the generally accepted threshold of 100. Most age estimates have relatively small confidence intervals (Appendix S4), including the estimate for the crown node of Elaeocarpaceae (66.7 Ma) and the crown node of Cunoniaceae (88.6 Ma; Appendix S4). The DEC+J model was favored over the DEC model (p < 0.01), which is indicative of a greater role of vicariance in explaining the biogeographical patterns observed in the study group. The analyses with Antarctica included or not as an available area produced very similar results; only the results including Antarctica are presented hereafter (Fig. 3). The most likely ancestral area for the crown node of Cunoniaceae is a combination of Australia/New Guinea (area C) + New Caledonia (area D); the same ancestral area is assigned to the subsequent node in Cunoniaceae. The remainder of the earliest‐diverging nodes in Cunoniaceae are reconstructed as occurring in Australia/New Guinea with dispersal to other regions later in the history of the group. For Elaeocarpaceae, America is the most likely area assigned to the crown node of the family, with Australia/New Guinea the second most likely reconstruction. In the clade comprising the genera Vallea Mutis ex L.f. and Aristotelia L’Hér., America is the most likely ancestral area, while for the clade comprising the remainder of the family, Australia/New Guinea is favored.

FIGURE 3

Ancestral area reconstruction in Cunoniaceae and related families using the DEC+J model.

DISCUSSION

This study resolved the phylogenetic relationships amongst the largest clade within Oxalidales as follows: (Cunoniaceae, (Cephalotaceae, (Brunelliaceae, Elaeocarpaceae))). It also confirmed that the monospecific Cephalotaceae (Albany pitcher plant) from southwestern Australia is nested in an otherwise entirely woody and mostly tropical clade. No obvious floral characters support this particular arrangement amongst the four families (Matthews and Endress, 2002). All four are predominantly found in the southern hemisphere, although Elaeocarpaceae (Hably et al., 2007; Manchester and Kvaček, 2009), like Cunoniaceae, has credible fossils in several regions of the northern hemisphere where it no longer occurs.

The monophyly of Cunoniaceae with its current limits is confirmed, and our analysis supports the inclusion of the families Baueraceae, Davidsoniaceae, and Eucryphiaceae within it (Angiosperm Phylogeny Group, 1998). We found support for the monophyly of all six of the tribes delineated by Bradford and Barnes (2001): Caldcluvieae, Codieae, Cunonieae, Geissoieae, Schizomerieae, and Spiraeanthemeae, the last being now monogeneric (Pillon, et al., 2009a). Assigning all genera to tribes would require the description of several new tribes (mostly monotypic) and will be dealt with elsewhere. The generic relationships recovered within Elaeocarpaceae agree with those of Crayn et al. (2006).

The Caldcluvieae currently comprises four genera: Ackama, Caldcluvia, Opocunonia, and Spiraeopsis (Bradford et al., 2004). Although this tribe contains only 12 species, seven generic names have been published for them. The last complete species‐level revision was by Hoogland (1979), who treated the entire group as a single genus, Caldcluvia, but this broad generic concept has not been followed in most subsequent works (de Lange et al., 2002; Hopkins and Hoogland, 2002; Bradford et al., 2004; Heslewood and Wilson, 2013). Here we found that Ackama was paraphyletic with Spiraeopsis nested inside it, and so we sink Spiraeaopsis into Ackama, while Opocunonia and Caldcluvia s.s. remain as distinct, monotypic genera (see Appendix 1: Taxonomic treatment). Characters that unite Ackama s.l. and distinguish it from Caldcluvia s.s. and Opocunonia are the size and shape of the inflorescence, the size of the flowers and length of their pedicels, and the number and arrangement of the stipules.

Our data supported the paraphyly of the genus Weinmannia that was suspected by Bradford (2002) and Y. Pillon (unpublished data). The species of Weinmannia fell into two clades. The first, comprising sect. Weinnmannia, was represented in our analysis by one species from the Americas (W. pinnata L., the type of the genus) and one from the Mascarenes (W. tinctoria Sm.). The second clade includes the remaining species in our analysis, all from the Old World, representing sections Fasciculatae (Malesia and Pacific), Inspersae (Madagascar), Leiospermum (Pacific), and Spicatae (Madagascar + Comoros). We therefore propose splitting Weinmannia into two genera, Weinmannia s.s., which equates to sect. Weinmannia, and Pterophylla D.Don, which we re‐establish for the clade including the four other sections (see Appendix 1: Taxonomic treatment). These two groups differ in inflorescence architecture, corolla, and seeds.

This phylogenomic study provides insights into the enigmatic biogeography of Cunoniaceae, traditionally considered a “Gondwanan” family (Raven and Axelrod, 1974). Our ancestral area reconstruction indicates a combination of Australia/New Guinea (area C) + New Caledonia (area D) as the likely ancestral area for the family (Fig. 3). Indeed, the greatest number of extant species and genera are found in Oceania, and particularly in eastern Australia, New Guinea, and New Caledonia (Fig. 4). The three lineages in the basal grade (Spiraeanthemum, Hooglandia, Aistopetalum) are all restricted to Oceania. Few lineages are found outside Oceania and nearby Southeast Asia (Malesia), and they are all phylogenetically distantly related to one another. The divergence of the southern African Cunonia capensis L. (24.5 Ma), the South American Lamanonia (20.8 Ma) and Eucryphia (14.5 Ma), and the Malagasy Pterophylla (25.6 Ma) from their Oceanian relatives postdate land connections between Australia and their current ranges. Their distributions may therefore be explained by long‐distance dispersal. It is worth noting that the only major difference between the reconstructions including and excluding Antarctica concerns the ancestral area of the node subtending Platylophus and Ceratopetalum (27.9 Ma). It is reconstructed as Australia/New Guinea when Antarctica is not accounted for and as a combination of Africa and Australia/New Guinea when Antarctica is included (Fig. 3). This scenario could suggest that the ancestor of Platylophus would have dispersed to Africa from Australia/New Guinea via Antarctica. The older timing of divergence of the mostly American Weinmannia s.s. (32.3 Ma) and the American Caldcluvia (41.2 Ma) from their Oceanian relatives may also be compatible with colonization through Antarctica. Indeed, until the Eocene, Antarctica was either directly connected to, or separated by relatively narrow seas from, both South America and Australia (Scotese, 2004; Müller et al., 2016), increasing the possibility of biotic exchange between these two now distant landmasses. It is not precisely clear when these landmasses became irreversibly separated by sea barriers, but it was probably sometime in the Eocene, ca. 35–41 Ma (Stickley et al., 2004; Scher and Martin, 2006). The same “via‐Antarctica” scenario might also apply to two Elaeocarpaceae genera confined to South America: Crinodendron Molina (58.1 Ma) and Vallea (41.8 Ma). Nevertheless, the ancestral area reconstruction using the models proposed by de la Estrella et al. (2019) did not suggest Antarctica as an ancestral area for Cunoniaceae, although the family has a substantial fossil record there (Cantrill and Poole, 2012).

FIGURE 4

Distribution and diversity of Cunoniaceae according to the new taxonomic framework presented here. *Ceratopetalum is still extant in Australia, New Guinea, and New Britain. †Extinct genera (age of fossil deposit).

The parsimonious scenario of an Oceanian origin of the extant diversity of Cunoniaceae with repeated dispersal events to Africa, the Americas, and Madagascar, may however, be in conflict with the fossil record. Firstly, Cretaceous fossils from Sweden (Schönenberger et al., 2001) and Burma (Chambers et al., 2010; Poinar and Chambers, 2017, 2019) have been attributed to this predominantly southern hemisphere family although some authors have recently considered the fossil evidence for Cunoniaceae in the northern hemisphere to be equivocal (Carpenter and Rozefelds, 2021). However, the attribution of fossils from Greenland and North America (Manchester and Kvaček, 2009) and Italy (Hably et al., 2007) to the extant genus Sloanea (Elaeocarpaceae) is more difficult to challenge, and the two closely related families have very similar present‐day distributions. Secondly, the Southern Cone of South America appears to have been another significant center of diversity for the Cunoniaceae, with fossils described from the Paleocene and Eocene of Argentina. These comprise two extinct genera, Lacinipetalum Jud, Gandolfo, Iglesias & Wilf (Jud et al., 2018) and Cunoniantha Jud & Gandolfo (Jud and Gandolfo, 2021), as well as an extinct species of Ceratopetalum (Gandolfo and Hermsen, 2017), a genus now composed of nine extant species in Australia and New Guinea. In addition, fossils from the Oligocene of Tasmania have been placed in the genera Acsmithia Hoogland (=Spiraeanthemum), Callicoma, Schizomeria, and Vesselowskya, although these genera no longer occur on this island (Carpenter and Buchanan, 1993), and a fossil assigned to Codia was described from the middle Eocene to Oligocene of Western Australia (Barnes and Hill, 1999), although this genus is now confined to New Caledonia. Extinction has clearly been important in the history of the Cunoniaceae and this family has other features of relictual groups. Most genera have few species with a median of only three. Furthermore, the family is similar to several small families of the ANA grade (Amborellales, Nymphaeales, Austrobaileyales), Chloranthales and Magnoliids in being over‐represented in eastern Asia and Oceania, where supposed climatic stability may have played a key role in the survival of these ancient lineages (Morley, 2001; Buerki et al., 2014; Pouteau et al., 2015). Thus, the Australia/New Guinea + New Caledonia origin inferred here for Cunoniaceae may be the result of their greater persistence in these regions, compared to areas such as South America (Kooyman et al., 2014; Barreda et al., 2021).

One tribe, nevertheless, is remarkably successful at the present time: Cunonieae. It comprises two thirds of the species in the family and contains the four largest genera (Weinmannia s.s., Pterophylla, Pancheria, Cunonia). Crown Cunonieae (i.e., excluding Vesselowskya Pamp.) has a diversification rate of 0.316, 2.2 times higher than the rate for the entire family (0.144). The tribe also has the largest distribution, with major radiations in the Andes, Madagascar, and New Caledonia. The genus Pterophylla is widespread in the Pacific and has reached the remote islands of the Marquesas and Rapa Iti. Weinmannia s.s. has itself an enigmatic distribution, divided between the Americas and the Mascarenes. Although the present study included only a single species from each of these two areas, for which divergence is estimated at 13.4 Ma, the previous study by Bradford (2002) included two additional ones in a phylogenetic analysis and numerous others in his morphological study. He found that the temperate W. trichosperma Cav. (Southern Cone of South America) was resolved as sister to a tropical clade (Americas + Mascarenes), suggesting a dispersal from the Americas to the Mascarenes. This may be one of the most intriguing dispersal events known, and the disjunction is similar to one in the palm tribe Chamaedoreeae, which comprises four genera in the neotropics and Hyophorbe in the Mascarenes (Baker and Couvreur, 2013). The disjunction in Cunonia between South Africa and New Caledonia is similar to that in Dietes Salisb. ex Klatt (Iridaceae), which is found in Africa, including southern Africa, and on Lord Howe Island (Goldblatt, 1981), and is also probably due to transoceanic dispersal. It is not clear why Cunonieae diversified and dispersed more than other tribes. It has small winged or hairy seeds, features found in several other genera of the family.

Outside Cunonieae, all genera with more than 10 species occur in New Caledonia. The remarkable success of Cunoniaceae on this archipelago (particularly Geissois: 13 species, Codia: 15 spp., Cunonia: 24 spp., Pancheria 27 spp.) mirrors that of the relictual conifer family Araucariaceae, in which the largest (and recent, <20 Ma) radiations of Agathis Salis. and Araucaria Henkel & W.Hochst. are also in New Caledonia (Setoguchi et al., 1998; Kranitz et al., 2014). Both families have a marked bias toward ultramafic substrates, an important feature of New Caledonia, and one that has probably had major effects on the survival (or not) of immigrant taxa and their diversification on the island (Pillon et al., 2010; Isnard et al., 2016).

CONCLUSIONS

The phylogenomic approach using Angiosperms353 has allowed us to resolve relationships across Cunoniaceae and neighboring families of Oxalidales, as well as issues of generic delimitation relating to Ackama and Weinmannia. It is hoped that the classification obtained in this study will be stable in the future, requiring few changes in generic concepts. The resulting Cunoniaceae have a total of 27 genera and ca. 335 species (Table 2). This phylogenetic framework for Cunoniaceae and the related families Brunelliaceae, Cephalotaceae, and Elaeocarpaceae was used to produce a linear sequence of genera for herbarium arrangement (Appendix S5) following the rules of Trias‐Blais et al. (2015). Cunoniaceae has a mixture of relict features and recent radiations that will be better understood with further research on fossils and through species‐level phylogenies.

TABLE 2

The 27 genera recognized within Cunoniaceae in this study, their tribal placement, number of species, distribution, and selected taxonomic references.

Genus	Tribe	No. of species	Distribution	Selected taxonomic references
Ackama	Caldcluvieae	10	Australia, New Zealand, Malesia, Solomon Islands	Hopkins and Hoogland (2002); de Lange et al. (2002); Schönenberger et al. (2020); APNI (2020)
Acrophyllum	unplaced	1	Australia	Hoogland (1960, 1981)
Aistopetalum	unplaced	2	New Guinea	Hopkins and Hoogland (2002)
Anodopetalum	Schizomerieae	1	Tasmania	Barnes and Rozefelds (2000)
Bauera	unplaced	4	Australia	APNI (2020)
Caldcluvia	Caldcluvieae	1	South America	Hoogland (1979); Rodriguez et al. (2018)
Callicoma	Codieae	1	Australia	APNI (2020)
Ceratopetalum	Schizomerieae	9	Australia, New Guinea	Rozefelds and Barnes (2002)
Codia	Codieae	15	New Caledonia	Hopkins et al. (2014)
Cunonia	Cunonieae	24 + 1	New Caledonia + South Africa	Hopkins et al. (2014); Goldblatt and Manning (2000)
Davidsonia	unplaced	3	Australia	Harden and Williams (2000)
Eucryphia	unplaced	5 + 2	Australia + South America	Taylor and Hill (1996); Rodriguez et al. (2018)
Geissois	Geissoieae	19	New Caledonia, Fiji, Vanuatu, Solomon Islands	Hopkins (2006); Hopkins et al. (2014)
Gillbeea	unplaced	3	Australia, New Guinea	Rozefelds and Pellow (2000); Hopkins and Hoogland (2002)
Hooglandia	unplaced	1	New Caledonia	McPherson and Lowry (2004)
Karrabina	Geissoieae	2	Australia	Schimanski and Rozefelds (2002, as Geissois); Hopkins et al. (2013)
Lamanonia	Geissoieae	6	South America	Zickel and Leitão Filho (1993); Hopkins (2018a)
Opocunonia	Caldcluvieae	1	New Guinea	Hopkins and Hoogland (2002)
Pancheria	Cunonieae	27	New Caledonia	Hopkins et al. (2014)
Platylophus	Schizomerieae	1	South Africa	Goldblatt and Manning (2000)
Pseudoweinmannia	Geissoieae	2	Australia	Rozefelds and Pellow (2011)
Pterophylla	Cunonieae	68	Madagascar, Comoros, Malesia, Pacific Islands	Madagacar and Comoros: Bradford (2001, 2001, 2001); Bradford and Miller (2001); Rogers (2017). Malesia and Pacific; Hopkins (1998, 1998, 1998); Hopkins and Bradford (1998, 1998); Hopkins and Florence (1998); Hopkins et al. (1998)
Pullea	Codieae	3	Australia, Malesia, Fiji	Hoogland (1979); Hopkins and Hoogland (2002)
Schizomeria	Schizomerieae	9	Australia, Malesia, Solomon Islands	Hopkins (2018b)
Spiraeanthemum	Spiraeanthemeae	19	Australia, New Guinea, Moluccas, Pacific Islands	Pillon et al. (2009a)
Vesselowskya	Cunonieae	2	Australia	Rozefelds et al. (2001)
Weinmannia	Cunonieae	90 + 2	Americas, Caribbean + Mascarenes	America: Bernardi (1961, 1963). No recent checklist available for all American taxa, but several regional treatmentsa. Mascarenes: Scott and Bosser (1997)

Regional treatments for American Weinmannia include: Central America and Mexico (Morales, 2010, 2011), Venezuela (Bradford and Berry, 1998), Ecuador (Bradford, 1999; Harling, 1999), Peru (Zarucchi, 1993), Bolivia (Harling and Fuentes, 2014), Chile (Rodriguez et al., 2018), Southern Cone (Hopkins, 2008).

AUTHOR CONTRIBUTIONS

Yohan Pillon: Conceptualization (equal); Supervision (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Helen C.F. Hopkins: Data curation (equal); Writing – original draft (equal); Writing – review & editing (equal). Olivier Maurin: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Niroshini Epitawalage: Data curation (equal). Jason Bradford: Validation (equal); Writing – review & editing (equal). Zachary S Rogers: Investigation (equal); Validation (equal); Writing – original draft (equal); Writing – review & editing (equal). William J Baker: Conceptualization (equal); Funding acquisition (equal); Methodology (equal); Project administration (equal); Supervision (equal); Writing – review & editing (equal). Felix Forest: Conceptualization (equal); Formal analysis (equal); Funding acquisition (equal); Project administration (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal).

Conceptualization: F.F., O.M., Y.P., W.J.B.; data curation: H.C.F.H., O.M., N.E.; formal analysis: F.F., O.M.; funding acquisition: F.F., W.J.B.; investigation: Z.R.; methodology: W.J.B.; project administration: F.F., W.J.B.; supervision: Y.P., W.J.B.; validation: J.B., Z.R.; visualization: F.F., O.M., Y.P.; writing original draft: F.F., O.M., Y.P., H.C.F.H., Z.R.; writing, review, and editing: F.F., H.C.F.H., J.B., O.M., Y.P., W.J.B., Z.R.

DATA AVAILABILTY STATEMENT

DNA sequences produced for this study are available in European Nucleotide Archive under umbrella project PRJEB35285; accessions numbers are provided in Appendix S1.

APPENDIX S1. List of sequenced material with full information on herbarium vouchers and data quality.

Click here for additional data file.

APPENDIX S2. Biogeographical model.

Click here for additional data file.

APPENDIX S3. Geographical area assignments for each terminal included in the biogeographical analysis.

Click here for additional data file.

APPENDIX S4. Dated tree of crown Oxalidales.

Click here for additional data file.

APPENDIX S5. Linear sequences of families and genera in the crown clade of Oxalidales for use in herbarium curation.

Click here for additional data file.

APPENDIX S6. The 12 currently recognized species in the Caldcluvieae, showing their distribution and historical and current generic placements.

Click here for additional data file.

TABLE 3

Comparison of selected morphological features of Ackama, Opocunonia, Spiraeopsis, and Caldcluvia s.s.

Characters	Ackama (New Zealand)	Ackama (Australia)	Spiraeopsis	Opocunonia	Caldcluvia s.s.
Number of species	2	2	5	1	1
Leaf type	imparipinnate	imparipinnate	imparipinnate	imparipinnate or trifoliolate	simple
Venation	semi‐craspedodromous	semi‐craspedodromous	semi‐craspedodromous	semi‐craspedodromous to craspedodromous	craspedodromous
Stipule position, number, margin	interpetiolar, 1 pair per node, toothed in A. rosifolia	interpetiolar, 1 pair per node, not toothed	interpetiolar, 1 pair per node, not toothed	interpetiolar, 1 pair per node, usually stalked and bilobed, sometimes toothed	lateral, 4 per node, sometimes asymmetric, toothed
Trichome type	simple	simple	simple and stellate	simple	simple
Presence of raised glands, peltate scales, or orbicular glands	yes	yes	yes	no (yes?)	no
Inflorescence size, shape a	large, triangular in outline	large, triangular in outline	large, triangular in outline	small, rounded or flattish in outline	small, rounded in outline
Pedicels (length)	± absent	± absent	1–2.5 mm	3–5 mm	4–8 mm
Flower size b (length)	small (1.5 mm)	small (1–1.5 mm)	small (1.5–2 mm)	large (2–4 mm)	large (3–4 mm)
Petals	(4–) 5, oblanceolate‐elliptic	4–6, oblanceolate‐elliptic	4–5, oblanceolate‐elliptic	5–6 (–7), ovate	4 (–5?), oblanceolate‐elliptic
Carpel number	2 (–3)	2–4	2–5	2	2
Replum in fruit	absent	absent	present	absent	present
Sepals in fruit	persistent	persistent	persistent	persistent	caducous
Seed shape, pubescence	ovoid, pubescent all over	ovoid, pubescent all over	spindle‐shaped, glabrous	spindle‐shaped, glabrous	spindle‐shaped, glabrous

Observations based on herbarium material at K and supplemented with data taken from literature: Godley (1979); Webb and Simson (1991); de Lange et al. (2002); Hopkins and Hoogland (2002).

Inflorescence size – large: typically 25–50 cm long though sometimes smaller in S. clemensiae L.M.Perry; small: typically 10–15 cm long.

Flower size – length measured from the base of the calyx to the tips of the lobes.