Unravelling the Phylogenomic Relationships of the Most Diverse African Palm Genus Raphia (Calamoideae, Arecaceae)

Palms are conspicuous floristic elements across the tropics. In continental Africa, even though there are less than 70 documented species, they are omnipresent across the tropical landscape. The genus Raphia has 20 accepted species in Africa and one species endemic to the Neotropics. It is the most economically important genus of African palms with most of its species producing food and construction material. Raphia is divided into five sections based on inflorescence morphology. Nevertheless, the taxonomy of Raphia is problematic with no intra-generic phylogenetic study available. We present a phylogenetic study of the genus using a targeted exon capture approach sequencing of 56 individuals representing 18 out of the 21 species. Our results recovered five well supported clades within the genus. Three sections correspond to those based on inflorescence morphology. R. regalis is strongly supported as sister to all other Raphia species and is placed into a newly described section: Erectae. Overall, morphological based identifications agreed well with our phylogenetic analyses, with 12 species recovered as monophyletic based on our sampling. Species delimitation analyses recovered 17 or 23 species depending on the confidence level used. Species delimitation is especially problematic in the Raphiate and Temulentae sections. In addition, our clustering analysis using SNP data suggested that individual clusters matched geographic distribution. The Neotropical species R. taedigera is supported as a distinct species, rejecting the hypothesis of a recent introduction into South America. Our analyses support the hypothesis that the Raphia individuals from Madagascar are potentially a distinct species different from the widely distributed R. farinifera. In conclusion, our results support the infra generic classification of Raphia based on inflorescence morphology, which is shown to be phylogenetically useful. Classification and species delimitation within sections remains problematic even with our phylogenomic approach. Certain widely distributed species could potentially contain cryptic species. More in-depth studies should be undertaken using morphometrics, increased sampling and more variable markers. Our study provides a robust phylogenomic framework that enables further investigation on the biogeographic history, morphological evolution and other eco-evolutionary aspects of this charismatic, socially and economically important palm genus.


126
Our phylogenetic analyses recovered five well supported clades. Three clades matched the 127 sections as defined by Otedoh [21]. Raphia regalis was always inferred with strong support as sister to 128 the rest of the genus independent of the inference method (Figures 1, A1). When comparing the two 129 phylogenetic approaches we identified a topological difference in the phylogenetic placement of the 130 section Temulentae, the species R. matombe and the Moniliformes and Flabellatae sections ( Figure 2). 131 In the IQ-TREE analyses we recovered weak support for the Temulentae as sister to all Raphia (except 132 R. regalis) (Figure 2a) while the ASTRAL analysis indicates with higher support that Temulentae is 133 sister to a clade containing R. matombe, Moniliformes & Flabellatae (Figure 2b). 134 The relationships between species in the Raphiate section are weakly to moderately supported 135 in both analyses (Figures 1, A1). Nevertheless, we do recover monophyletic groups in some species 136 consistent with prior morphological identifications. This is the case for individuals of R. laurentii 137 and R. monbuttorum, which despite low support are monophyletic. Furthermore, both these species 138 are recovered as sister, with moderate to high support. However, our species delimitation analysis 139 suggested that individuals identified under both species are conspecific (Figure 1 A). Support is 140 generally higher in the ASTRAL tree, even when taking into account different gene histories, so we 141 suggest that the ASTRAL tree represents a more accurate reconstruction of the phylogeny of Raphia 142 ( Fig. 2b) so we will principally refer to the relationships in this tree from now on.  Figure  A2. Individuals are colour coded based on the hypothesis of species delimitation inferred using SODA with α = 0.01. For a single clade, the Temulentae section marked with a black star and referred to as the "hookeri" complex in the main text, varied between the two values of α ued here. The orange boxes represent the species limits using SODA with a more stringent value of α = 0.005. The black square shows the "zamiana" complex clade. Tip names contain the species name as well as the sequencing ID. B: R. regalis partial inflorescence representing the Erectae section (described here, see discussion). C: R. sudanica inflorescence, representing the Obclavatae section. D: R. palma-pinus inflorescence, representing the Raphiate section. E: R. hookeri inflorescence, representing the Temulentae section. F: R. farinifera inflorescence, representing the Moniliformes (and ex Flabellatae) section. G: R. regalis, note the acaulescent habitat and inflorescences subtended by the leaves (Couvreur 398,Cameroon). H: R. zamiana (Mogue Kamga 17, Gabon). I: R. monbuttorum (Couvreur 1212, Cameroon). J: detail of R. monbuttorum rachillae (Couvreur 1212, Cameroon

144
Our species delimitation approach yielded between 17 (α = 0.005) and 23 (α = 0.01) species ( Figure   145 1). Higher values of α split a clade of closely related individuals (marked with a star in Figure 1), 146 predominantly belonging to R. hookeri, into seven different species. Generally, our species delimitation 147 results corresponded well with our field identifications and using available floras (e.g. [19,20]). In 148 some cases we found that SODA split individuals belonging a priori to a single species into multiple 149 species, for example R. farinifera and R. sudanica (Fig. 1). Conversely, individuals assigned to different 150 species such as R. laurentii and R. monbuttorum were classified as the same species according to SODA 151 delimitation independent of α values. In general, the support among different species as delimited by To further explore genetic structure among our two main species complex, namely the "zamiana" 155 and "hookeri complexes (marked with a black square and a black star in Figure 1, respectively), we 156 used SNPs extracted from the sequence data to look at the variation among individuals. The "hookeri"

Discussion
3.1. Synthesizing morphology and molecules: the sections of Otedoh reevaluated.

174
Our phylogenomic analyses of Raphia provide a novel and overall well supported phylogenetic 175 framework for this important African genus (Figure 1 A). Although some of the morphology based 176 sections of Otedoh [21] were recovered, we also recovered some topological differences (Figure 1 A). Angola (including Cabinda) was different between the two types of analyses ( Figure 2).

182
In all analyses, the acaulescent central African species Raphia regalis is recovered with strong 183 support as sister to the rest of the genus (Figures 1 A, A1). This species, together with R. australis, 184 were placed within the subsection "erectae" within the Moniliformes section [20,21] because the 185 inflorescences were suggested to be "erect", in contrast to the rest of the Raphia species whose 186 inflorescences are hanging or semi-erect (except for R. palma-pinus which also has an erect inflorescence).

187
Our results do not support this classification, as R. australis is recovered as sister to R. farinifera (of the

192
The close relationships between the Moniliformes and Flabellatae section is not surprising.

193
The inflorescences, although different in some aspects such as the clearly racket-shaped partial

209
Our results, however, suggest that certain sections erected by Otedoh [21] are not monophyletic 210 and need to be re-evaluated. Differences in phylogenetic relationships between the concatenated and 211 coalescent approaches have been increasingly reported in this genomic era [28][29][30]. Our results were 212 similar to those in Couvreur et al. [19] where higher bootstrap support were obtained when using the 213 concatenation approach, despite the coalescent approach highlighting considerable gene tree conflict.

214
Here, we favour the phylogenetic hypothesis recovered when using the coalescent approach ( and Flabellatae. 226 We thus recognize five main sections within Raphia based on the phylogenetic results presented 227 here. We then discuss these results in more detail below.