We sampled at least one specimen for every species of the Monodoreae tribe, several specimens from the Uvarieae tribe representing different genera, as well as several other Annonaceae outgroups (Chatrou et al. 2012; Suppl. material 3). Within Monodoreae, we sampled, when possible, more than one specimen per species following these cases: one specimen per subspecies or variety; one specimen per major geographic region for widespread species (between East Africa, West Africa, Central Africa); several specimens for morphologically variable species representing that variability (Suppl. material 3).

For all specimens, DNA was extracted from either silicagel dried leaf samples or leaf material taken from herbarium specimens. Details about specimens and vouchers are available in Suppl. material 3.

For the downstream analyses, every sample was assigned a unique identifier as follow: RUN_##_I##_T##_Genus_species-COL_NUM. “RUN_##_I##_T##” is the identifier used for the sequencing process and bioinformatics analyses. “Genus_species” is the most up to date taxonomic identification of the specimen (i.e. after applying the taxonomic changes treated in this paper). To ease the reading of updated names, some of the specimens have their old identification in parentheses like “Genus_species_(ex_old name)”. “COL_NUM” is the three first letters of the name of the collector and the number of collection of the specimen (e.g. “VAL_2540” stands for the specimen n°2540 collected by J.L.C.H van Valkenburg). The details about every sample can be found in the Suppl. material 3.

DNA extraction and NGS library preparations for nuclear exon capture followed Couvreur et al. (2019). Libraries were sequenced using an Illumina HiSeq 2000 plateform using paired-end sequencing of length 150 bp by Novogene Co. Ltd.

Raw read sequences were demultiplexed using the demultadapt script (https://github.com/Maillol/demultadapt). Adapters were removed from the reads using cutadapt 1.18 (Martin 2011) with the “-b” option and the “-O 7 -m 35 -q 20 -e 0.1” parameters. Low quality reads were then excluded (mean quality phred score below 30 and read less than 35 bp long) using a script modified from https://github.com/SouthGreenPlatform/arcad-hts/blob/master/scripts/arcad_hts_2_Filter_Fastq_On_Mean_Quality.pl. Reverse and forward reads were paired according to their names in the fastq files using a custom script (https://github.com/SouthGreenPlatform/arcad-hts/blob/master/scripts/arcad_hts_3_synchronized_paired_fastq.pl). A final step of 6 bp trimming was performed on the reverse reads using FASTX-Toolkit (https://github.com/agordon/fastx_toolkit).

To process the clean reads, we used the HybPiper v1.3.1 pipeline (Johnson et al. 2016) following Couvreur et al. (2019). This pipeline mapped the reads on to the targeted nuclear exons using BWA v0.7.12 (Li and Durbin 2009). The mapped reads were assembled into contigs using SPAdes v3.11.1 (Bankevich et al. 2012). Overlapping contigs were then assembled into ‘supercontigs’ containing both targeted sequence (i.e. exon) and off-target sequence data (i.e. partial intron). The pipeline ends up with one file per gene (supercontig) containing the sequence for every individual. The sequences were then aligned for each gene using MAFFT v7.305 (Katoh and Standley 2013) using the automatic selection of the alignment algorithm (--auto). Alignments were trimmed with Gblocks 0.91b (Talavera and Castresana 2007), to remove poorly aligned regions.

We used two different approaches for the phylogenetic reconstruction: the concatenation approach and the gene tree approach. For each approach, we first removed the supercontigs that were putatively identified as paralogs within the HybPiper pipeline, and then filtered our dataset to select only the supercontigs that are present in 75% of the individuals and for which the sequence contains at least 75% of the length of the target sequence (75/75 filter, following Couvreur et al. 2019).

For each supercontig, the alignments were filled with gaps and missing individuals so that every sequence in the alignment has the same length and every alignment contains all the individuals. The alignments were then concatenated into a single “supermatrix” using the pxcat function in the phyx program (Brown et al. 2017). A phylogenetic tree was inferred from the supermatrix using an optimized maximum likelihood (ML) tree search implemented in RAxML 8.2.9 with the GTRGAMMA model after a rapid bootstrap analysis (Stamatakis 2014). We specified Anaxagorea crassipetala as an outgroup (“-o” option) and constrained the ML search on a backbone topology (“-g” option), in order to avoid long-branch attraction due to the scarce sampling of the outgroups. The backbone topology only includes the outgroup species and follows the results of Couvreur et al. (2019): (Anaxagorea_crassipetala-MAA_9408, (((((Uvariodendron_kirkii-DAG_09, Uvaria_grandiflora-COU_838), Annona_glabra-CHA_467), Duguetia_staudtii-COU_1014), Guatteria_jefensis-MAA_9553), Greenwayodendron_suaveolens-COU_746)).

We inferred a phylogenetic tree for every supercontig using RAxML 8.2.9 (Stamatakis 2014) under the GTRGAMMA model and 100 bootstrap replicates. For every tree obtained, we removed the branches with a bootstrap support less than 10% using the Newick Utilities program (https://github.com/tjunier/newick_utils), in order to improve the accuracy of the inferred species tree (Zhang et al. 2018). We then inferred the summary species tree from all the (unrooted) genes trees using ASTRAL-III 5.7.5 under the multi-species coalescent model (Zhang et al. 2018, 2018). The support at the branches was estimated both with the quartet support (QS) values (“-t 1” option) and with the local posterior probabilities (LPP) (default parameters). The species tree was then rooted on Anaxagorea crassipetala (Anaxagorea_crassipetala-MAA_9408). The phylogenetic trees obtained were then plotted and annotated using the ggtree, ggplot and treeio packages (Wickham 2016; Yu et al. 2017; Wang et al. 2020).

In 1913, Baker described the monotypic genus Dennettia Baker f., with a single species Dennettia tripetala Baker f. having bisexual flowers with “2 rarely 3 sepals” and three petals in a single whorl (Baker 1913). In 2003, Kenfack et al. combined the genus Dennettia into the genus Uvariopsis under the name Uvariopsis tripetala (Baker f.) G.E. Schatz, based on four main arguments (Kenfack et al. 2003). First, the morphological circumscription of Uvariopsis had been extended to also include bisexual flowers by Verdcourt (1986, Up. bisexualis); second, Kenfack et al. (2003), observed specimens of D. tripetala exhibiting flowers with 2 sepals and 4 petals, thus reminding of the typical Uvariopsis flower; third, Kenfack et al. (2003) considered that because Up. congolana has 3 petals, the number of perianth segments is unreliable as a generic distinction for Uvariopsis; and fourth, the pollen microscopic features of D. tripetala are very close to those of Uvariopsis (Kenfack et al. 2003; Doyle et al. 2004). This was initially confirmed based on molecular phylogenetic analyses using few plastid regions, recovering Uvariopsis tripetala as nested within Uvariopsis but based on just three species sampled within Uvariopsis (Couvreur et al. 2008a; Chatrou et al. 2012).

After careful examination of 16 flowering specimens identified as Uvariopsis tripetala (including the type Dennett 44), we found that this species invariably exhibits 3 sepals and 3 petals. The sepals are fused at the base up to more than half of their length, forming a ring curved downward and persisting on the fruiting pedicel. In some rare cases we observed that the lobes of the calyx (i.e. the free parts at the apices of the basally fused sepals) are sometimes almost inconspicuous on the fruiting pedicel, allowing for the confusion with a two sepal-like calyx. To date, and based on the specimens we observed, we haven’t seen any flowers of Up. tripetala with four petals, contrary to the affirmation of Kenfack et al. (2003). If such flower arrangements exist, they must be extremely rare. It is also possible that the four petal flowers observed by Kenfack et al. (2003) were actually flower buds of the new species, Uvariopsis oligocarpa Dagallier & Couvreur (see the Taxonomic Treatment section), as their geographic ranges overlap and leaves are quite similar (but Up. oligocarpa is unisexual). In contrast to the previous phylogenetic analyses, we did not recover Dennettia (Uvariopsis) tripetala as nested within Uvariopsis, but as sister to the monotypic West African genus Monocyclanthus vignei Keay, with strong support in both phylogenetic approaches (Fig. 1, Suppl. materials 1, 2). Dennettia (Uvariopsis) tripetala differs from Monocyclanthus vignei (Keay 1953) by having three petals in one whorl (vs. six petals in one whorl), the lobes of the fused sepals generally distinct (vs. generally inconspicuous), and a receptacle height:width ratio less than 1, i.e. wider than tall (vs. a receptacle height:width ratio greater than 1, i.e. taller than wide) (see Table 1). The single whorl of six petals of Monocyclanthus supposedly results from the compression of separate ancestral whorls (Saunders 2010). For Dennettia, the single whorl of three petals might result from a similar compression of separate ancestral whorls followed by a loss of petals, or from the loss of a whorl of petals.

All these morphological and phylogenetic elements suggest that Dennettia (Uvariopsis) tripetala is part of neither the genus Uvariopsis nor the genus Monocyclanthus. We thus reinstate the monotypic genus Dennettia, with its single species Dennettia tripetala Baker f. (see Taxonomic treatment). This was also suggested previously by Cheek et al. (2022) based on morphology.

Given the above, the morphological circumscription of Uvariopsis must be reevaluated. Uvariopsis was generally circumscribed as having 2 sepals and 4 petals (Kenfack et al. 2003). However, Uvariopsis congolana differs from all other Uvariopsis species by having flowers with 3 petals instead of 4. Uvariopsis congolana was successfully sequenced and nested within Uvariopsis with strong support (sister to Up. dioica, Fig. 1, Suppl. materials 1, 2). Thus, the number of petals (4 versus 3) is not a good diagnostic character for Uvariopsis. However, Up. congolana has 2 sepals, similarly to all the other Uvariopsis species, making this character a good synapomorphy for the genus. The synoptic characters of the genus Uvariopsis Engl. are given in the Taxonomic Treatment section below.

Based on our dense taxon sampling, we identified a group of closely related species for which the phylogenetic limits remain uncertain. Indeed, the four species Up. bakeriana, Up. citrata, Up. korupensis and Up. submontana clustered as monophyletic with strong support (Fig. 1, Suppl. materials 1, 2), but support for the monophyly of each individual species and the relationships between them remains ambiguous (Fig. 1, Suppl. material 2) with some gene tree conflict (Suppl. material 1). They share some morphological characters in particular the large size and shape of their leaves, hence referred to as the “large leaves” group here. They are all trees with large obovate leaves (up to 30–60 cm long) having rounded to cordate bases, and display flowers borne along the trunk with varying degrees on density, from densely packed in Up. submontana to sparsely distributed in Up. citrata (Kenfack et al. 2003; Couvreur and Niangadouma 2016). In addition, flowering pedicels are shorter than 7 cm and petals generally 3 to 10 times longer than wide. Despite these similarities, they differ by several marked characters. Uvariopsis bakeriana is characterized by its unique petals that are narrowly ovate to linear and bright pinkish to deep red in color (Hutchinson and Dalziel 1927b). Uvariopsis citrata is characterized by its leaves emitting a strong lemon scent when crushed and sessile flowers (Couvreur and Niangadouma 2016). Uvariopsis korupensis and Up. submontana have similar flowers with pedicels generally 20–70 mm long, ovate to narrowly ovate shortly fused at base, but are differentiated by their ecology (Up. korupensis occurs below 200 m, Up. submontana above 900 m), the degree of flowers density on the trunk (less densely clustered in Up. korupensis) and the sepals being smaller in Up. korupensis (1–5(7.5) mm long) than in Up. submontana (5–11 mm long) (Gereau and Kenfack 2000; Kenfack et al. 2003).

Based solely on the phylogeny, this group could be considered as a single species. However, given the morphological differences of each species detailed above we choose to keep the taxonomy of this group as it is. Phylogenetic conflict is known to be coincident with morphological innovations (Parins-Fukuchi et al. 2021), as observed in recent and fast diverging groups (Meyer et al. 2017). Here we retrieve such a pattern, with substantial morphological differentiation between Up. citrata, Up. bakeriana and Up. korupensis, but a lack of support at subtending nodes. Uvariopsis korupensis is retrieved as paraphyletic and the two specimens sampled here (Couvreur 1039 and Louis 1863) come from different localities. They might belong to different isolated relictual populations failing to coalesce due to the persistence of ancestral polymorphisms (Pennington and Lavin 2016). The species Up. bakeriana and Up. citrata likely emerged from two populations that diverged from the ancestral Up. korupensis population. This remains to be clarified with a denser sampling (in particular of Up. submontana) or with a phylogeographic and adapted population genomics approaches (Hardy et al. 2013; Migliore et al. 2019; Helmstetter et al. 2020a).

In addition to the results discussed above, our sampling and analysis confirmed the presence of new and/or under-collected species, which we detail here. Thanks to our dense sampling in the genus Uvariopsis, we were also able to test species limits, and concluded that several names are actually synonyms, which we detail under each concerned species in the Taxonomic treatment section below.

Morphologically, Uvariopsis sessiliflora stands out from the other Uvariopsis species by having sessile flowers and leaves less than 20 cm long (Up. citrata also has sessile flowers but leaves more than 30 cm long). The only known specimen representing this species is its type specimen (Mildbraed 5239), collected in 1911 in Lomié, a relatively well collected locality of Cameroon (Sonké and Couvreur 2014). This specimen is relatively poor in information as it has only has a small branch, one flower and one monocarp. Here we were able to sequence this specimen and it turns out that it is phylogenetically embedded within Up. dioica (Fig. 1, Suppl. materials 1, 2), indicating that Up. sessiliflora and Up. dioica are the same phylogenetic species. It seems that the specimen Mildbraed 5239 represents a rare extreme in the morphological variation of Ud. dioica, with severe reduction of the flower pedicel. This also explains why no similar specimen was collected after all this time in a such relatively well explored locality. We thus make it synonym under the name Up. dioica.

The recently described species Uvariopsis etugeana Dagallier & Couvreur (Couvreur et al. 2022), is a rare tree species only known from two localities in Cameroon. Our results confirm it is phylogenetically different from other Uvariopsis species (Fig. 1, Suppl. materials 1, 2).

Uvariopsis noldeae Exell & Mendonça is a species endemic to Angola (Exell and Mendonça 1951; Paiva 1966) and only known from the type specimen (Nolde 576). It is morphologically similar to Uvariopsis solheidii (De Wild.) Robyns & Ghesq., except for the young branches that are sparsely pubescent to glabrous whereas young branches of Up. solheidii are tomentose to shortly tomentose suggesting the names might be synonymous. Our phylogenetic analyses, however, revealed with strong support that Up. noldeae is genetically different from Up. solheidii, being sister to Up. dioica and Up. congolana, with Up. solheidii forming a strongly supported monophyletic group (Fig. 1, Suppl. materials 1, 2).

Uvariopsis congensis differs from the other Uvariopsis species by its relatively small to medium sized leaves (less than 180 mm long), globose flower buds (vs. conical), and short flowering pedicels (less than 12 mm long). It is a widespread species distributed from Central Africa (Cameroon, Gabon) to East Africa (Uganda, Kenya) (Robyns and Ghesquière 1933a; Verdcourt 1971a). Interestingly, some of the specimens we identified as Up. congensis were from West Africa, (Sierra Leone to Ghana), thus outside of its suggested range. A thorough examination of these West African specimens revealed they differed by the morphology of their fruits. Moreover, our molecular phylogenetic analyses retrieved these West African specimens (referred as Uvariopsis oligocarpa on the Fig. 1 and Suppl. materials 1, 2) as monophyletic with strong support, and clearly separated from the Central and East African specimens (Fig. 1, Suppl. materials 1, 2). Sterile specimens of Up. oligocarpa are similar to Up. congensis and Dennettia tripetala, but are easily differentiated when female flowers or fruits are available. We thus describe these specimens under the new species Uvariopsis oligocarpa Dagallier & Couvreur.

Finally, the specimens of Uvariopsis congensis from East Africa (Verdcourt 1971a) are also morphologically different from the specimens distributed in Central Africa. The East African specimens represent trees from 7 to 15 meters tall (vs. shrubs to trees from 2 to 7 meters tall for the Central African specimens) and they have narrowly elliptic laminas with a length:width ratio between 3 and 4 (vs. elliptic to obovate with a length:width ratio comprised between 2.1 and 3.1). Interestingly, the morphological differences between the two morphotypes are not supported by our phylogenetic analyses (Fig. 1, Suppl. materials 1, 2). Thus given our sampling we shall refrain from splitting Up. congensis in two different species. However, to account for the morphological differences between the two morphotypes, we describe a new variety Uvariopsis congensis var. angustifolia Dagallier & Couvreur corresponding to the East African morphotype (see Taxonomic Treatment below).

Collection ATBP 666 (African Tropical Biodiversity Programme, MO04937010) collected from the Budongo Forest Reserve in Uganda (as “Uvariopsis sp. nov. Uganda” in Fig. 1 and Suppl. materials 1, 2), is a sterile specimen tentatively identified as Up. congensis. Indeed, the leaves of this specimen resemble Up. congensis except for its longer petioles when compared to typical Up. congensis (ca. 6 mm vs. 2.5–5 mm in Up. congensis). Interestingly, our results recovered this specimen as sister to the clade formed by Up. lovettiana and Up. congensis, with strong support (Fig. 1, Suppl. materials 1, 2) and thus genetically very different from Up. congensis. This specimen could thus be an undescribed species. Unfortunately, the available material is too poor for a proper species description. This species might be rare occurring only in the Budongo Forest Reserve. Plant surveys of this locality are relatively old (Synnott 1985) and it has recently been assessed as a key biodiversity area (Plumptre et al. 2019). Similarly, a specimen collected from Kigoma area in western Tanzania (Kyoto University Expedition (KUE) 1039, EA000009849), and referred to as “Uvariopsis sp. nov. Tanzania” (Fig. 1, Suppl. materials 1, 2), appears to be a new taxon. Morphologically, it is close to Up. lovettiana but differs in having lamina with rounded base (vs. acute to decurrent), and subsessile pubescent monocarps (vs. stipitate and glabrous). The sequenced specimen is recovered as sister to either Up. noldeae with strong support (Suppl. material 2) or to the group formed by Up. noldeae, Up. dioica and Up. congolana (Fig. 1, Suppl. material 1) with weak support. This uncertainty probably comes from the fact that the specimen has low overall sequencing statistics (22% of the total reference covered at at least 10×; mean depth 6.1×). This would also explain its long sustaining branch. Nevertheless, the specimen passed our quality thresholds and is included. Our results confirm that it is a new species, although its phylogenetic placement remains uncertain. These new species emphasize once again the great diversity of the East African region, with locations (mainly forest remnants) containing numerous endemic species or genera (e.g. Scharff 1992; Couvreur et al. 2009; Gosline et al. 2019; Dagallier et al. 2021).

East Africa is a region that harbours a great plant diversity, with several species being endemic or narrow endemics (Scharff 1992; Burgess et al. 1998, 2007; Couvreur et al. 2009; Gosline et al. 2019; Dagallier et al. 2021; Cheek et al. 2022). This pattern is known to be the result of the topographic heterogeneity of this region (Droissart et al. 2018; Guillocheau et al. 2018) that could promote both recent divergence and persistence of more anciently lineages through time (Fjeldså and Lovett 1997; Rahbek et al. 2019; Dagallier et al. 2020).

Recently, we described three new Uvariodendron species from the East African coastal forests based on their morphology: Uvariodendron dzomboense Dagallier, W.R.Q. Luke & Couvreur, Uvariodendron mbagoi Dagallier & Couvreur, and Uvariodendron schmidtii W.R.Q. Luke, Dagallier & Couvreur (Dagallier et al. 2021). Here, our phylogenetic analyses retrieved these species as monophyletic, confirming their species status (Fig. 1, Suppl. materials 1, 2).

In addition to these recently described species, our results reveal another undescribed species. It is represented by four specimens sampled here that were previously identified as Uvariodendron kirkii. These specimens all come from a restricted area in the Kimboza Forest Reserve (Morogoro, Tanzania), and this species is thus described as Uvariodendron kimbozaense Dagallier & Couvreur. Phylogenetically, it is sister to Ud. kirkii although clearly divergent at the molecular level (Fig. 1, Suppl. materials 1, 2). Morphologically, it also resembles Ud. kirkii, but differs by having larger flowers and different leaf dimensions (see Fig. 3, Table 2 and Taxonomic Treatment). Although the Kimboza Forest Reserve covers just 4.05 km² of lowland rain forest, the description of yet another new tree species underlines once again the major biological importance of this forest for biodiversity conservation (Rodgers et al. 1983). To date, an unpublished check list (see http://www.mikepalmer.co.uk/woodyplantecology/tropical/Kimboza.html) reports that Kimboza harbors 18 strict endemic plant species, including the monotypic Annonaceae genus Mwasumbia (Couvreur et al. 2009) as well as many other trees such as Cola kimbozensis Cheek and C. quentinii Cheek (Cheek et al. 2007), Turraea kimbozensis Cheek (Cheek 1989) or Vitex morogoroensis Walsingham & S.Atkins. (Cheek et al. 2019). This suggests that Kimboza is “probably the richest site for forest limestone point endemic plant species in tropical Africa” (Cheek et al. 2007, 2019).

Figure 3. 

Morphological characters differentiating Uvariodendron kimbozaense Dagallier (A, C, E, G) and Uvariodendron kirkii Verdc. (B, D, F, H) A flowering pedicel with bracts and sepals, side view, note the imbricate sepals B flowering pedicel with bract scar and sepals C flower, with one outer petal and two inner petals gnawed, top view, note the slight transversal curvature of the inner petals D flower, top view, note the ‘boat-shaped’ inner petals E, F simplified representations of the flowers, from above, yellow marking highlights transversal curvature of the petals G, H Simplified representations of the transversal cut of the leaf, by the midrib. A, C Dagallier 49 (type) B, D Dagallier 23. Photos and drawings Léo-Paul Dagallier. Abbreviations: B, bract; brs bract scar; ca, carpel; ip, inner petal; lam, lamina; mlo, lower (abaxial) side of the midrib; mup, upper (adaxial) side of the midrib; op, outer petal; sep, sepal.

In the Flora Zambesiaca (Robson 1960) a potential new species of Uvariodendron was documented without being named because of the scarcity of the material. It is only known from a single specimen (Mendonça 2558A) collected in 1944 in Mozambique. This specimen was sequenced here and it appears genetically distinct from the other Uvariodendron species, being sister to the group of several other East African species (Ud. gorgonis, Ud. schmidtii, Ud. dzomboense and Ud. pycnophyllum) with strong support (see Uvariodendron mossambicense in Fig. 1 and Suppl. materials 1, 2). Our results thus confirm that this specimen represents a new species. Morphologically, it is very similar to Ud. dzomboense by the short leaves (less than 130 mm long), narrowly elliptic and acute to slightly decurrent laminas, and the globose, ca. 5 mm in diameter and sessile flower buds. It differs by the number of carpels (ca. 5 vs. 50–75 in Ud. dzomboense) (Table 3). The description from the Flora Zambesiaca was based on a sheet from LISC. Our examination of this species is based on a sheet from WAG that contains a single monocarp, thus slightly enhancing the initial description of the species (Robson 1960). The monocarp has a short stipe (ca. 12 mm long) which also differentiates it from Ud. dzomboense that has sessile monocarps. Even if the material remains scarce for a complete description, we provide evidence from both phylogeny and morphology that this species is distinct from the other known Uvariodendron species. We thus name this species Uvariodendron mossambicense Robson ex Dagallier & Couvreur and provide a description as complete as possible (see the Taxonomic Treatment section).

East Africa has been relatively well explored botanically, as shown by the completion of the Flora of Tropical East Africa (Beentje 2015) and the publication of several parts of the Flora Zambesiaca (Exell and Wild 1960). However, our descriptions of several new species suggest that other botanical discoveries might arise from further exploration (Cheek et al. 2022). This stresses the importance of local botanical initiatives (e.g. Binggeli 2019; Bingham et al. 2021; Hyde et al. 2021).

In the Flora of Cameroon (Couvreur et al. 2022), the name Uvariodendron mirabile R.E.Fr. was synonymized under Uvariodendron fuscum (Benth.) R.E.Fr. Here we sequenced both the type (Preuss 1378) and paratype (Lehmbach 178) of the name Ud. mirabile, both specimens clustering with specimens identified as Ud. fuscum (e.g. Mildbraed 6428, a paratype of the name Ud. fuscum) (Fig. 1, Suppl. materials 1, 2), confirming they are conspecific. Note that when described, Ud. occidentale was distinguished from Ud. fuscum (as Ud. mirabile) based on morphology (Le Thomas 1967). This holds up from a phylogenetic point of view as Ud. occidentale is clearly separated from Ud. fuscum (Fig. 1, Suppl. materials 1, 2).

Morphologically, Ud. fuscum (Benth.) R.E.Fr., Ud. giganteum (Engl.) R.E.Fr. and Ud. magnificum Verdc. share morphological similarities and differences. They mainly diverge along a morphological continuum by the size of their leaves (lamina length) and the size of their flowers (sepals and petals dimensions, number of carpels), with Ud. magnificum having larger leaves and flowers than Ud. giganteum, with the latter having larger leaves and flowers than Ud. fuscum (Fig. 4, Table 4). In addition, Ud. magnificum flowers have apically imbricate inner petals while the other Uvariodendron species have apically connate inner petals (Verdcourt 1969). In terms of similarities, they all present elliptic to obovate leaves. Uvariodendron fuscum and Ud. giganteum have similar sepals (fused over up to half of their length), inner petals (free, obovate) and outer petals (free, ovate). Uvariodendron giganteum and Ud. magnificum have young branches covered with long and soft hairs producing a whitish appearance quickly falling off with age, whereas the young branches of Ud. fuscum vary from being sparsely pubescent to glabrous. Some of the characters that supposedly discriminate the species are overlapping (e.g. lamina length), particularly concerning Ud. fuscum and Ud. giganteum.

Figure 4. 

Morphological continuum in Uvariodendron fuscum: traits comparison for var. fuscum (fusc.), var. giganteum (gig.) and var. magnificum (mag.). Each point represents a measurement. Vertical bars span the minimum and maximum values of the focal traits. Photos, left: leaf of Couvreur 1046; top right: flower, below view, of Couvreur 1046; below right: flower, top view, of Couvreur 990 (CC BY-NC 4.0 Thomas Couvreur).

Phylogenetically, the specimens of the three species clustered together, with strong support at the crown node and a weak to moderate support at the internal nodes (Fig. 1, Suppl. materials 1, 2). The phylogenetic branches are short (Fig. 1, Suppl. material 2) indicating a low divergence between the specimens and both Ud. fuscum and Ud. giganteum are retrieved as paraphyletic. This suggests that gene flow is still occurring between populations. We thus propose that the three described species are actually three different morphotypes of the same species. Uvariodendron fuscum represents the smaller morphotype (small leaves and flowers relatively to the others), Ud. magnificum represents the greater morphotype (large leaves and flowers relatively to the others), and Ud. giganteum represents the intermediate morphotype. The imbrication of inner petals of Ud. magnificum that Verdcourt (1969) described as a synapomorphy for the species should be considered as a developmental constraint imposed by the size of the petals. Indeed, in bud, we can imagine that the only way for such long (50–53 mm long) and wide (36–38 mm wide) petals to arrange with the other floral elements is to overlap.

The name Ud. giganteum was already reduced into synonymy with Ud. fuscum in the Flora of Cameroon (Couvreur et al. 2022). Here we thus synonymize the name Ud. magnificum with Ud. fuscum. The first mention of Ud. fuscum dates back to 1862, as Uvaria fusca (Bentham 1862), whereas the name Ud. magnificum dates back to 1969 (Verdcourt 1969). Given the priority rule (Turland et al. 2018), the name Uvariodendron fuscum prevails. The description of the species is given in the Taxonomic Treatment section below.

The species Uvariodendron molundense (Engler & Diels) R.E.Fr. is distributed in Cameroon and Gabon. In 1969, Le Thomas synonymized the names Uvariodendron mayumbense and Uvariodendron letestui with Ud. molundense. Indeed, the position of the flowers along the trunk and branches, previously used to discriminate these species (Fries 1930), was no longer valid as some specimens are both cauliflorous and ramiflorous (Le Thomas 1969). Le Thomas (1969) also described the variety Uvariodendron molundense var. citrata form some specimens from Gabon exhibiting a strong lemon scent. The examination of 106 specimens identified as Ud. molundense revealed that this species shows a great morphological variability, especially for the leaf shape. For example, some specimens show young leaves with decurrent bases, whereas older leaves have a rounded base. The specimens identified as Ud. molundense form three phylogenetically supported groups separated by relatively high molecular divergence (Fig. 1, Suppl. materials 1, 2), at least as much molecular divergence as between other clearly morphologically divergent species (e.g. between Ud. dzomboense and Ud. pycnophyllum). Although the three groups have morphological similarities, we were able to distinguish them and we thus propose to separate them in three distinct species: Ud. molundense, Ud. citriodorum (Le Thomas) Dagallier & Couvreur, and Ud. pilosicarpum Dagallier & Couvreur. The full and valid descriptions of Ud. citriodorum and Ud. pilosicarpum are given in the Taxonomic Treatment section.

The genus Uvariopsis Engler was erected in 1899 with the description of Uvariopsis zenkeri Engler (Engler and Diels 1899). As circumscribed at the time, the genus was distinguished by its mainly unisexual flowers, two sepals and four fused petals. A few years later, Diels erected the genus Tetrastemma Diels with the description of Tetrastemma dioicum Diels (Diels 1907). Tetrastemma was also characterised by unisexual flowers with two sepals and four petals, but differed from Uvariopsis by being dioecious and cauliflorous (vs. monoecious and ramiflorous), and by having free petals (vs. fused at base). Based on the free petals character, four other Tetrastemma species were described in the following years: Tetrastemma solheidii De Wild. in 1909 (De Wildeman 1909), Tetrastemma pedunculosum Diels and Tetrastemma sessiliflorum Mildbr. & Diels in 1915 (Diels 1915), and Tetrastemma bakerianum Hutch. & Dalziel in 1927 (Hutchinson and Dalziel 1927a).

In parallel, two other genera, morphologically closed to Uvariopsis and Tetrastemma were described: Thonnera De Wild. and Denettia Baker f. The genus Thonnera included a single species Thonnera congolana De Wild. differing from Tetrastemma by flowers having three petals (De Wildeman 1909). In 1953, however, Fries combined this name into Uvariopsis as Uvariopsis congolana (De Wild.) R.E.Fr. without further justification (Fries 1953). The genus Denettia Baker f. was characterised by having bisexual flowers with three sepals and three petals, which differed from Tetrastemma, Thonnera, and Uvariopsis. It included a single species Denettia tripetala Baker f. (Baker 1913).

In 1933, Robyns & Ghesquière described Uvariopsis vanderystii Robyns & Ghesq., and suggested it was a kind of intermediate species showing characters of Tetrastemma, e.g. cauliflory, and of Uvariopsis, e.g. fused petals (Robyns and Ghesquière 1933b). The same year, they revised the two genera with their ‘Essai de révision des genres Uvariopsis Engl. & Diels et Tetrastemma Diels (Annonacées)’ (Robyns and Ghesquière 1933a). In this publication, they described several new species: Uvariopsis batesii Robyns & Ghesq., Uvariopsis chevalieri Robyns & Ghesq. and Uvariopsis congensis Robyns & Ghesq. The later species is ramiflorous and has free petals, which again showed intermediate characters between Uvariopsis and Tetrastemma. Based on these intermediate species, Robyns & Ghesquière (1933a) extended the morphological concept of Uvariopsis and combined the name Tetrastemma with Uvariopsis. The new species combinations were: Uvariopsis bakeriana (Hutch. & Daltz.) Robyns & Ghesq., Uvariopsis dioica (Diels) Robyns & Ghesq., Uvariopsis pedunculosa (Diels) Robyns & Ghesq., Uvariopsis sessiliflora (Mildb. & Diels) Robyns & Ghesq., Uvariopsis solheidii (De Wild.) Robyns & Ghesq. In 1948, Pellegrin described Uvariopsis letestui Pellegr., a species from Gabon (Pellegrin 1948), and three years later Exell and Mendonça described a species from Angola: Uvariopsis noldeae Exell & Mendonça (Exell and Mendonça 1951). In 1952, Keay published the species Uvariopsis globiflora Keay (Keay 1952). He also combined a species previously described by Chevalier under the name Uvaria spectabilis Chev. (Chevalier 1920) and cited in the Flora of Tropical West Africa as Uvaria spectabilis Chev. ex Hutch. & Dalziel (Hutchinson and Dalziel 1927a) into Uvariopsis guineensis (Chev. ex Hutch. & Dalziel) Keay (Keay 1952). More than 30 years later, Verdcourt described Uvariopsis bisexualis Verdc. from Tanzania. Although this species has bisexual flowers, Verdcourt placed it in Uvariopsis based on pollen morphology closely ressembling species of Uvariopsis (Verdcourt 1986). The concept of Uvariopsis was thus enlarged to include bisexual species.

Gereau and Kenfack (2000) described the species Uvariopsis korupensis Gereau & Kenfack followed by the publication of Uvariopsis submontana Kenfack, Gosline & Gereau (Kenfack et al. 2003). In that later publication, the name Denettia tripetala Baker f. was combined into Uvariopsis tripetala (Baker f.) G.E. Schatz. based on the concept of Uvariopsis now also included species with bisexual flowers (Verdcourt 1986; Kenfack et al. 2003). More recently, the Tanzanian species Uvariopsis lovettiana Couvreur & W.R.Q. Luke (Couvreur and Luke 2010) and the Gabonese species Uvariopsis citrata Couvreur & Niangadouma (Couvreur and Niangadouma 2016) were described. In the Flora of Cameroon Annonaceae (Couvreur et al. 2022), the species Uvariopsis etugeana Dagallier & Couvreur was described and the name Uvariopsis vanderystii Robyns & Ghesq. was synonymized with Uvariopsis pedunculosa (Diels) Robyns & Ghesq. Finally, the species Uvariopsis dicaprio Cheek & Gosline was described from Cameroon (Gosline et al. 2022). In the present work, based on molecular and morphological analyses, we retain the name Dennettia tripetala and thus the genus Dennettia (see discussion above). We also describe a new species, Up. oligocarpa sp. nov., and synonymize the names Up. sessiliflora with Up. dioica, Up. globiflora with Up. guineensis and Up. letestui with Up. solheidii. For the two laters, we combine the names into varieties as Up. guineensis var. globiflora comb. et stat. nov. and Up. solheidii var. letestui comb. et stat. nov., respectively. We also describe a new variety of Up. congensis, Up. congensis var. angustifolia var. nov. and provide a preliminary description of potentially three new species which lack sufficient material to be formally described. This brings the total number of Uvariopsis species to 17 (and three that have yet to be formally described).

The genera Dennettia, Uvariodendron and Uvariopsis are trees or shrubs with plagiotropic branches on an orthotropic axis, with indumentum of simple hairs and with no latex or exudate. A distichous phyllotaxis of the primary axis (the trunk) was reported for Ud. pycnophyllum (Johnson 2003) but no information was reported for the other species nor for Dennettia and Uvariopsis. The trunks are generally cylindrical and straight, with no buttresses. The base of the trunk of some Uvariopsis species (Up. dioica, Up. submonana, Up. korupensis) can be thickened by the growth of numerous inflorescences clumped on the trunk between the ground to generally ca. 50 cm (up to 4 m).

The branches can be sparsely pubescent when young, but the hairs generally fall off with age. The dense brown tomentum on the young branches persisting on older branches distinguishes Ud. calophyllum from the other species of Uvariodendron. The species Ud. fuscum (var. giganteum and var. magnificum) also have a characteristic indumentum of long soft hairs producing a whitish appearance on the younger branches, but that quickly disappear in older branches.

In the three genera, the leaves are distichous, simple, entire, exstipulate, with the midrib impressed above, raised below, the secondary veins weakly brochidodromous to brochidodromous and the tertiary veins reticulate. Dennettia has relatively small (less than 160 mm long) elliptic leaves, while Uvariodendron and Uvariopsis have a range of leaf shapes, from elliptic to obovate to oblong, ranging from 65 to 750 mm long. Uvariodendron fuscum var. giganteum has one of the longest leaves of Annonaceae in Africa, up to 60 cm long (Couvreur et al. 2022). In contrast to other genera (e.g. Uvariastrum; Couvreur 2014) the leaf base shape is not a very useful taxonomic character to distinguish species because several different shapes (e.g. from acute to rounded) can occur within species. However, subcordate leaf bases are generally found in Up. bakeriana, Up. citrata, Up. korupensis, and Up. submontana, and decurrent leaf bases in Ud. pilosicarpum, Up. bisexualis, Up. congensis, Up. lovettiana, Up. oligocarpa and Up. zenkeri. Some species have apparent complex leaf bases, like rounded in shape but minutely decurrent when looking at the very base (Ud. citriodorum, Ud. gorgonis, Ud. molundense). In Uvariodendron kimbozaense, the midrib is slightly raised above with a central groove all along the length of the midrib. This is unique in the genus, but raised midribs occur in other genera such as Isolona, Monodora and Ophrypetalum (Couvreur 2009), and raised and grooved midribs occur in Cremastosperma (Pirie et al. 2018). The petiole in Uvariopsis is generally short and thick, less than 6 mm long with a length:width ratio less than 2, whereas in Uvariodendron the petiole is generally longer than 4 mm with a length:width ratio between 2 and 5.

In Uvariodendron, several species (Ud. dzomboense, Ud. gorgonis, Ud. mossambicense and Ud. schmidtii) show ‘eragrostiform’ leaf buds, that are terminal or axillary buds composed of several (generally 5 to 12) distichous and densely pubescent scales (Figs 25A, 36C, 44B). The term ‘eragrostiform’ relates to the genus Eragrostis (Poaceae) and its typical form of flattened spikelet composed of compact and clustered florets. Even though this feature is striking, it should be noted that not all the specimens of a same species present this kind of buds, so it is hard to use it as a diagnostic character. The phenology or the environment may influence the establishment of eragrostiform buds, which might protect the meristem against drought or herbivores.

The scent of crushed leaves and/or bark can be used as a distinctive character. Uvariodendron citriodorum and Up. citrata emit a strong lemon scent (Le Thomas 1969; Couvreur and Niangadouma 2016), a feature also shared by some specimens of Ud. gorgonis. Crushed leaves and bark of Ud. mbagoi have a strong bergamot scent (between orange and lemon) (Dagallier et al. 2021), and all parts of Ud. anisatum smell of anise. Ud. angustifolium is also reported as having a strong aromatic scent.

In contrast to sterile parts, Dennettia, Uvariodendron and Uvariopsis have very different fertile parts. In Annonaceae, the inflorescence is a rhipidium, i.e. determinate (sympodial) inflorescence with a terminal flower and lateral cymose and monochasial partial inflorescences (Weberling and Hoppe 1996). Inflorescences of Dennettia, Uvariodendron and Uvariopsis are axillary and borne on leafy twigs or borne on old branches (ramiflory) or on the trunk (cauliflory), depending on the species. The position of the flower was regarded as an important character to characterize some species like Ud. molundense and Up. guineensis (Keay 1952; Le Thomas 1967, 1969) but it appears to be unreliable in Uvariodendron and Uvariopsis (see also Le Thomas 1969). In the three genera, the flower pedicels are not articulate and display from one to six bracts at their base and from one to four bracts on the lower half of the pedicel. As flowers mature, most of the bracts fall off, except the uppermost bract. Usually, fallen bracts leave a scar on the pedicel.

In Dennettia, the inflorescences are axillary and borne on leafy twigs or borne on old branches, held by a tiny peduncle. They can be composed of up to six flowers, but one or two-flowered inflorescence are the most common. The flowers of Dennettia are bisexual. They are composed of one whorl of three sepals and one single whorl of three petals, which is characteristic of the genus in the Monodoreae. In Annonaceae, single whorls of three petals can be found in Annickia, Annona, and Dasymaschalon (van Heusden 1992).

In Uvariodendron, the inflorescences are axillary and borne on leafy twigs, borne on old branches or borne on the trunk. The axis of the inflorescences are contracted, rendering their interpretation difficult without a detailed morpho-anatomical study. They appear to be rhipidia composed of one or two (in most species), three (Ud. calophyllum, Ud. gorgonis, Ud. mbagoi) or rarely up to 11 (Ud. kimbozaense) flowers. Generally, in multi-flowered rhipidia, the flower in the terminal position is fully developed while the lower flowers on the axis are buds.

Flowers in Uvariodendron are bisexual with one whorl of three sepals and two whorls of three petals, as usually found in Annonaceae (van Heusden 1992). The fusion and arrangement of the sepals is used to distinguish between species: always fused (Ud. angustifolium, Ud. dzomboense, Ud. fuscum, Ud. mossambicense, Ud. occidentale, Ud. pycnophyllum, Ud. schmidtii), always imbricate (Ud. citriodorum, Ud. kimbozaense, Ud. mbagoi, Ud. pilosicarpum), or from free to fused to imbricate (Ud. anisatum, Ud. calophyllum, Ud. connivens, Ud. gorgonis, Ud. kirkii, Ud. molundense). The sepals are smaller and morphologically different to the petals. The outer and inner petals are subequal in length, ranging from 10 to 70 mm at anthesis. In bud, the outer petals are valvate (adpressed against each other all along the margin) whereas inner petals are valvate only at the apex. In mature flowers, the outer and inner petals are free, but the inner petals can remain adpressed at apex. In most species, the petals are cream to yellow in color, with a red/purple marking at the base of the inner side. Uvariodendron connivens has petals that are pink to red all over and on both sides at anthesis, while the outer petals of Ud. calophyllum are covered with a dense tomentum leading to a brown flower. The stamens are numerous (more than 200), with linear anthers and truncate apical prolongation of the connective. The carpels vary from five to 160, they are free, linear, with a coiled stigma. The monocarps are sessile or subsessile and cylindrical.

Uvariopsis has inflorescences borne on trunk in most of the species, but can be axilary to leaves on young branches (Up. congensis, Up. guineensis, Up. lovettiana, Up. oligocarpa, Up. zenkeri). They are generally composed of one to three flowers. Like in Uvariodendron, the peduncle and partial peduncle are contracted, rendering their interpretation as rhipidium difficult. In particular, Up. dioica, Up. submontana and Up. korupensis have inflorescences in clumps of up to 50 flowers on the trunk.

The flower buds of Uvariopsis have different shapes important to discriminate between species (Kenfack et al. 2003). They can be globose (Up. congensis, Up. guineensis, Up. lovettiana, Up. oligocarpa, Up. pedunculosa, Up. zenkeri), ovoid to conical (Up. bisexualis, Up. citrata, Up. dioica, Up. noldeaea Up. solheidii, Up. submontana), pyramidal (Up. congolana, Up. dicaprio, Up. etugeana), or narrowly and long conical (Up. bakeriana, Up. korupensis). Flowering pedicels in Uvariopsis varies greatly from less than 11 mm (Up. bakeriana, Up. citrata, Up. congensis, Up. etugeana, Up. oligocarpa, and Up. zenkeri) to more than 100 mm up to 400 mm (Up. congolana, Up. pedunculosa). The flowers of Uvariopsis have one whorl of two sepals and one whorl of four petals (except Up. congolana which has three petals). In Annonaceae, flowers with two sepals and four petals are rare and can be found in only a few other genera such as Disepalum or Tridimeris (van Heusden 1992). The petals are usually free, but can be fused in some species (Up. congolana, Up. guineensis, Up. korupensis, Up. pedunculosa, Up. submontana, and Up. zenkeri). Another rare character in Annonaceae exhibited by Uvariopsis is its exclusively unisexual flowers (except for Up. bisexualis that has bisexual flowers). In Annonaceae, unisexual flowers are found in several androdioecious species, e.g. in Polyceratocarpus (Couvreur et al. 2009), Diclinanona (Erkens et al. 2014), Ephedranthus (Lopes et al. 2014) , Pseudephedranthus (Erkens et al. 2017), Klarobelia or Pseudomalmea (Lopes et al. 2018). Species with exclusively unisexual male and female flowers are rarer and found only in a few species within genera such as Anonidium, Monanthotaxis, Pseuduvaria, Stelechocarpus and Winitia (van Heusden 1992; Chaowasku et al. 2013; Hoekstra et al. 2021; Couvreur et al. 2022). Unisexual Uvariopsis species are monoecious, i.e. male and female flowers on the same individual, even in the species Up. dioica that has a misleading specific epithet. The number of stamens varies from 100 to 900, with anthers linear and connective prolongation truncate or absent. The number of carpels varies from three to 280; they are free, linear, with a coiled, truncate or globose stigma.

The fruits of Uvariopsis are composed of several sessile to subsessile monocarps generally cylindrical in shape. Their texture and pubescence can be informative: either smooth and not ridged (Up. congensis, Up. dioica, Up. guineensis, Up. korpensis, Up. lovettiana, Up. oligocarpa) or wrinkled to verrucose (Up. bakeriana, Up. congolana, Up. pedunculosa, Up. solheidii), and either pubescent to tomentose (Up. pedunculosa, Up. oligocarpa, Up. zenkeri) or sparsely pubescent to glabrous.