Eriostemon, , Wilson, 1998

Orel, Harvey K., McLay, Todd G. B., Neal, Will C., Forster, Paul I. & Bayly, Michael J., 2023, Plastid phylogenomics of the Eriostemon group (Rutaceae; Zanthoxyloideae): support for major clades and investigation of a backbone polytomy, Australian Systematic Botany 36 (5), pp. 355-385 : 379-380

Interpreting the backbone polytomy in the Eriostemon group

Is evolution multifurcating or are our data limited?

The amount of sequence data employed in our analyses represents a ~14-fold increase in dataset size compared with the most recent other studies on the group (106 885 bp in our phylogenomic alignment v. 7579 bp in Appelhans et al. 2021 and 5145 bp in Duretto et al. 2023). Even with this large dataset, we were unable to resolve backbone relationships between four supported major lineages of genera ( Fig. 3). Across all analyses, these lineages were arranged on very short, unsupported branches that we have treated as a

polytomy. Short divergences in molecular phylogenies may be explained by both biological and experimental causes. Biologically, polytomies can be the result of true multifurcations (so-called hard polytomies) caused by rapid radiation and simultaneous divergence events. However, polytomies may arise as a result of reaching the limit of resolution in the estimated tree (so-called soft polytomies) caused by using an insufficient amount of sequence data or sequence variation, conflicting data, or inappropriate methods of analysis ( Slowinski 2001; Whitfield and Lockhart 2007; Lin et al. 2011; Sayyari and Mirarab 2018).

In ptDNA, conflicting phylogenetic signal may be attributed to topological incongruence among gene trees, noise introduced by site saturation, or applying substitution models uniformly across genes with differing rates of evolution.

The first of these causes belongs to the species tree paradigm ( Doyle 2021), in which systematists infer species trees from gene trees under the multispecies coalescent ( MSC) model that assumes free recombination between unlinked genes ( Edwards 2009). Because of this assumption, coalescence-based approaches have generally been applied to studies utilising unlinked nuclear genes. However, several studies have reported conflicting phylogenetic signals among plastid genes (e.g. Zeng et al. 2014; Foster et al. 2018; Gonçalves et al. 2019), and the topic of whether these methods are suitable for analysing plastid loci (which have historically been treated as linked; Doyle 1997) has been subject to recent debate ( Gonçalves et al. 2020; Doyle 2021). Although our analysis of tree space identified three different gene-tree clusters, estimation of the cluster phylogenies produced congruent topologies that differed mainly in their levels of branch support; thus, we found no evidence for conflicting signals among gene trees. This may reflect a failure of Robinson-Fould distances to adequately identify meaningful phylogenetic similarity (see Smith 2022), or simply show that there is no strong conflict among gene trees in each cluster. Instead of differences in tree topology among clusters, the identification of three gene-tree clusters may be the result of differing levels of phylogenetic informativeness across gene trees, with Cluster 1 containing, on average, the most informative genes with generally better-resolved topologies than for Clusters 2 and 3. We attribute the generally lower branch-support values displayed by our ASTRAL phylogeny to the lower resolving power of individual genes v. concatenated genes ( Doyle 2021), rather than conflicting gene-tree topologies (this is also supported by gene discordance factors for the tree in Fig. 3; mean gCF: 41.9, mean gDF1: 5.8, mean gDF2: 5.6 across all branches; see Data availability for file).

Second, we are able to dismiss noise caused by site saturation as a potential source of conflict on the basis of the persistence of the polytomy in the analyses of the likelihood mapping subsets (with noisy loci removed), the DNA alignment partitioned by codon position, and the translated CDS alignment (Supplementary Fig. S6). Similarly, if rate variation among loci were a contributor of conflict, then our IQ-TREE analysis partitioned by locus would have produced a result different from that from the unpartitioned analysis. Hence, we consider it unlikely that conflicting phylogenetic signal is a significant cause of the short, unsupported branches of the polytomy.

Previous molecular phylogenies of Rutaceae have resolved deeper relationships than that of our polytomy with far less ptDNA sequence data and comparable sequence variation (e.g. Groppo et al. 2008, 2012; Bayly et al. 2013; Appelhans et al. 2021). Several studies have also shown that lower-level relationships can be resolved using less data (e.g. Barrett et al. 2014; Bayly et al. 2016; Duretto et al. 2020, 2023). On the basis of these precedents, our dataset should be appropriate for resolving relationships at the taxonomic level of the polytomy. However, some biological characteristics of the plastome may render it inappropriate for resolving divergences that occurred over a short period of time. In plants, the plastome evolves at a relatively slow rate, at least half that of the nuclear genome ( Wolfe et al. 1987). Genomes that evolve at slower rates are less capable of accumulating phylogenetically informative evolutionary changes during rapid radiations. In addition, lineage sorting is expected to progress more rapidly in ptDNA than in nDNA because of the smaller effective population size of the plastome (generally ¼ that of nuclear genes, with the exception of regions such as nuclear ribosomal DNA that are subjected to concerted evolution; Buckler and Holtsford 1996; Palumbi et al. 2001). Because of these features, the plastome may not be as useful as the nuclear genome for investigating lineages that have rapidly radiated, because there is less time for phylogenetically informative mutations to accumulate before lineages are sorted. If rapid radiation has occurred in the distant past, as is potentially the case in the Eriostemon group, this problem is likely to be compounded by subsequent lineage-specific mutations that can mask phylogenetically informative changes accumulated during the radiation ( Whitfield and Lockhart 2007).

Our ASTRAL polytomy test was unable to reject that the polytomy is not a true multifurcation ( P > 0.5 for all polytomy branches). However, as this test relies on calculating gene-tree topology likelihoods under the MSC model, the application of the test to our dataset of plastome loci is probably inappropriate (owing to divergence from the MSC model), and hence this result should be interpreted cautiously ( Sayyari and Mirarab 2018; Doyle 2021).

Results from our tree-topology tests are likely to be more reliable. These showed a trend of higher support for topologies with Clade 1 as sister to the rest of the Eriostemon group, and low support for a multifurcating topology, suggesting that relationships between these lineages are perhaps more likely to be bifurcating rather than multifurcating.

The true nature of relationships around the backbone polytomy in the Eriostemon group remains difficult to answer with the current dataset, and we suggest that further investigation using more appropriate data, in particular, multi-locus nuclear DNA markers, is required before any conclusions are made regarding the hard or soft status of the polytomy. Clarification of the polytomy, and its hard or soft status, will also allow for interpretation of the higher-level branching order and evolution of the group.

The significance of understanding the polytomy

Together, the clades that make up the backbone polytomy in the Eriostemon group (i.e. Clades 1–4) contain 16 genera (13 Australian) and ~204 species (~194 Australian) that include a large proportion of Australia’s Rutaceae diversity (total ~42 genera, ~490 spp.). Taxa in the Eriostemon group account for nearly all of the Rutaceae that occur on low-nutrient soils in dry sclerophyll communities in the south-east and south-west of Australia ( Hartley 1995); the only genera outside this group that occur in similar habitat are Boronia (~125 spp. in Australia), Zieria (~63 spp. in eastern Australia), Neobyrnesia (1 sp. in N Northern Territory) and Cyanothamnus (23 spp. in Australia). Because of this, they are an important component of the Australian flora.

The phylogenetic depth at which the polytomy occurs means that it is largely irrelevant to the taxonomic delimitation of genera but may be pertinent to higher-level classification. Currently, revisions of tribal and subtribal classification are needed for Rutaceae ( Appelhans et al. 2021) . A new tribal classification would probably apply above the level of the polytomy and, hence, not require its resolution, but a robust future subtribal classification would certainly benefit from the resolution of the branching order of Clades 1–4. Subtribes in the Eriostemon group have been largely neglected since Engler (1931), who placed the contemporary genera ( Muiriantha was not yet described) in the Eriostemon group across the following four subtribes: Eriostemoninae (including Asterolasia , Crowea , Drummondita (as a section of Philotheca ), Eriostemon , Geleznowia , Leionema (as a section of Phebalium ), Phebalium and Philotheca ), Nematolepidinae ( Chorilaena and Nematolepis ), Correinae (monotypic, including Correa ) and Diplolaeninae (monotypic, including Diplolaena ). Under this scheme, Eriostemoninae is polyphyletic. On the basis of our phylogeny, one could propose a new system that recognises each of the major lineages of the group (i.e. Clades 1–4) as separate subtribes. Alternatively, subtribe Eriostemoninae could be split into several different subtribes so as to retain Correinae, Diplolaeninae and Nematolepidinae. In any case, a newly proposed classification should be constructed with consideration of relationships across the whole of the Rutaceae to ensure that taxonomic ranks are applied in a consistent manner, and should also consider the morphological diagnosability of subtribes, something that the current study does not thoroughly do.

Beyond taxonomic classification, resolving the branching order of the polytomy would prove useful to studies focussed on macroevolution and biogeography. In terms of morphology, the Eriostemon group displays multiple characters of biological and ecological interest that have evidently experienced state transitions (e.g. in floral features and phyllotaxy); studies investigating the evolutionary histories of characters by using methods that require bifurcating phylogenetic frameworks, such as ancestral-state reconstructions, would be enabled by resolution of the polytomy. Biogeographically, the Eriostemon group is noteworthy for having multiple genera that are disjunct between south-western and south-eastern Australia (e.g. Phebalium , 16 spp. in south-west, 22 spp. in south-east; Nematolepis , 1 sp. in south-west, 6 spp. in south-east; Philotheca section Erionema , 1 sp. in south-west, 14 spp. in south-east; Asterolasia , 5 spp. in south-west, 14 spp. in south-east). This distribution pattern occurs in many plant taxa and has been attributed to vicariance associated with marine inundations of south-central Australia and subsequent edaphic and climatic barriers during the mid-Miocene (~16–14 million years ago; Crisp and Cook 2007; Ladiges et al. 2010, 2012). Compared with other species-rich families in Australia, the Rutaceae is unconventional in having a higher net diversification rate of genera in the south-east than in the south-west from the Eocene–Oligocene (~34 million years ago) extinction pulse until the mid-Miocene ( Nge et al. 2020). Resolution of the backbone polytomy would enable the further testing of hypotheses relating to the timing of this vicariance event in the Australian Rutaceae and offer insight into the drivers of generic diversification, and, specifically, whether the diversification of major lineages (i.e. our Clades 1–4) was influenced by such an event. It could also present an opportunity to provide a more robust timing for the split of Myrtopsis in New Caledonia.