Introduction

Madagascar is known for its high biodiversity with a flora showing extraordinary levels of species diversity and endemism; around 82% of the vascular plant species are endemic (Callmander & al. 2011). The daisy family (Asteraceae) is one of the most species-rich families in Madagascar, with no fewer than 540 species, of which about 88% are endemic. One representative of the Malagasy Asteraceae is Apodocephala Baker, a small genus of nine species of shrubs or trees characterized by its coriaceous leaves drying brown (S. Razafimandimbison, pers. obs.), discoid capitula and florets with white or whitish corollas (Fig. 1A, B) (Humbert 1955, 1960, 1962). Members of the genus are mainly found in rainforests and rocky habitats at high altitudes (Humbert 1955, 1960, 1962).

Apodocephala is currently placed in the tribe Astereae, where it has been associated with Vernoniopsis Humbert, another genus of small trees with discoid capitula and white florets, but also with Madagaster G. L. Nesom and Rochonia DC., all endemic to Madagascar (Humbert 1960; Bremer 1994; Nesom 2020). The Astereae are one of the largest tribes of the family and typically consist of annual or perennial herbs (rarely shrubs) with heterogamous capitula, containing numerous radiate or filiform marginal florets, and with distributions ranging from the Arctic to the tropics. Astereae members are characterized by having ecaudate and ecalcarate anther bases, disc floret styles with two distinct, non-confluent marginal stigmatic lines and often deltate to triangular or lanceolate style appendages that are adaxially glabrous and with sweeping hairs abaxially (Bremer 1994; Brouillet & al. 2009). Apodocephala is, however, noted to have shortly caudate anthers (Humbert 1960; Bremer 1994) and also differs from the more typical Astereae genera by being shrubs or trees and in the lack of filiform or radiate marginal female florets. The current placement of Apodocephala in the tribe Astereae has not previously been tested by analysis of molecular data.

Fig. 1.

Representatives of the Malagasy endemic Athroismeae – A, B: Apodocephala sp.; C, D: Lowryanthus rubens; E: Anisopappus longipes (Cass.) Wild; F: Athroisma proteiforme (Humbert) Mattf. – Photographs by P. Antilahimena (A, B; MBG-Madagascar, CC-BY-NC-ND © 2021), K. Kainulainen (C), P. P. Lowry II (D; MBG, CC-BY-NC-ND © 2021), F. Ratovoson (E; MBG-Madagascar, CC-BY-NC-ND © 2021) and L. Ramon (F; CC-BY-NC-ND © 2021).

Two of the nine species of Apodocephala, A. oliganthoides Humbert and A. pauciflora Baker (the type of the generic name), were collected in 2020 by one of the authors (SGR) in north-eastern Madagascar. Leaf samples of these species preserved in silica gel (Chase & Hills 1991) were available for the first time for DNA sequencing and analysis. The aim of this study is to test the phylogenetic position and tribal affiliation of the genus Apodocephala within Asteraceae. The taxonomic and biogeographic implications of our findings are discussed.

Material and methods

Phylogenetic analyses

Sequences were aligned using MUSCLE v.3.8.425 (Edgar 2004) as implemented in AliView v.1.24 (Larsson 2014) and manually edited using BioEdit v.7.0.5.3 (Hall 1999). Alignments are available as  supplementary Appendices 2 (wi.51.51205_Appendix_2_ndhF_matrix.nex) and  3 (wi.51.51205_Appendix_3_combined_matrix.nex). Two different datasets were compiled and analysed. We initially performed analyses of a ndhF matrix with 130 taxa, representing a wide coverage of the Asteraceae tribes, in order to determine the tribal position of Apodocephala within the Asteraceae. Boopis anthemoides Juss. (Calyceraceae) was used as outgroup, following Bengtson & al. (2017). The results of the ndhF analyses allowed us to narrow down the sampling to a focus on the tribe Athroismeae. A second dataset of ETS, ITS, ndhF and trnL–trnF sequence data from 45 taxa of the Athroismeae was analysed. Callilepis salicifolia Oliv. was used as outgroup to root trees. All the analyses were conducted with the Bayesian and parsimony methods. Prior to analyses of the combined dataset, each region was analysed separately to check for incongruence by simply comparing topologies.

Bayesian inference analyses were conducted using MrBayes v.3.2.7a (Ronquist & al. 2012) using the online XSEDE platform on the CIPRES Science Gateway (Miller & al. 2010). For the larger ndhF dataset, nucleotide substitution models were set to GTR+I+G; for the combined dataset they were set to GTR+G for ETS and to GTR+I+G for ITS and the plastid markers (ndhF and trnL–trnF), chosen using the Akaike information criterion (AIC) as implemented in jModeltest v.2.1.10v20160303 (Guindon & Gascuel 2003; Darriba & al. 2012). Analyses consisted of two independent runs, with eight chains each and the temperature parameter set to 0.1. The Markov Chain Monte Carlo (MCMC) was run for 75 million generations for the larger ndhF dataset, with a sampling frequency of 7500, and for 30 million generations, with a sampling frequency of 3000, for the combined dataset. Convergence of Markov chains was examined using Tracer v.1.7.1 (Rambaut & al. 2018), as well as by checking average standard deviation values of split frequencies. The first 25% of the trees were excluded as a burn-in phase.

Maximum parsimony analyses were conducted using PAUP v.4.0a169 (Swofford 2002). The most parsimonious trees were searched for using a heuristic search strategy, with 10,000 random addition replicates, TBR branch-swapping and saving only the most parsimonious tree each replicate. Bootstrap support (BS) was estimated using 10,000 replicates, each with 10 random addition replicates and saving only a single most parsimonious tree each replicate. All analyses were run multiple times.

Fig. 2.

Bayesian fifty-percent majority-rule consensus tree from an analysis of the Asteraceae ndhF dataset. Numbers above branches indicate posterior probability (PP) and bootstrap support (BS) values, bootstrap support values < 50 are indicated by a dash.

Results

The ndhF dataset consisted of 2298 aligned characters, 556 of which were parsimony informative, and the combined Athroismeae dataset of 4470 characters, 720 of which were parsimony informative. The parsimony analysis of the ndhF dataset yielded 6113 most parsimonious trees, 2882 steps long (consistency index, CI = 0.38; retention index, RI = 0.67, excluding uninformative characters), and the parsimony analysis of the Athroismeae dataset yielded six most parsimonious trees, 2417 steps long (CI = 0.54, RI = 0.81). Analyses of separate regions produced partly unresolved trees (results not shown), but showed no signs of incongruence. Bayesian and parsimony analyses of the two datasets produced trees with similar topologies, differing only in little-supported nodes. Parsimony analyses resulted in somewhat less-resolved trees.

A majority-rule consensus tree from a Bayesian analysis of the ndhF dataset including representatives from the entire Asteraceae is shown in Fig. 2. The analysis resolved the two sampled Apodocephala species as monophyletic and with strong support as sister to the Malagasy Lowryanthus rubens Pruski (posterior probability, PP = 1.0, bootstrap support, BS = 99, Fig. 2), and placed them in a poorly supported tribe Athroismeae (PP = 0.85, BS = -). A majority-rule consensus tree from a Bayesian analysis of the combined dataset, consisting of both nuclear (ETS, ITS) and plastid (ndhF, trnL–trnF) data, and with a focus on the tribe Athroismeae, is shown in Fig. 3. The analysis confirmed the position of Apodocephala within Athroismeae (PP = 1.0, BS = 95, Fig. 3), and the sister-group relationship between Apodocephala and Lowryanthus (PP = 1.0, BS = 100, Fig. 3). The Apodocephala–Lowryanthus clade was resolved as sister to the subtribe Athroisminae (PP = 1.0, BS = -, Fig. 3).

Discussion

Our results strongly support the phylogenetic placement of Apodocephala in the tribe Athroismeae as sister to Lowryanthus, and therefore we reject its current position in the tribe Astereae (Grau 1977; Bremer 1994; Nesom 2020) (Fig. 2). Apodocephala and the Malagasy genus Vernoniopsis (Astereae) have been suggested to have a close affinity on the basis of their discoid capitula with white florets and presence of shortly caudate anthers (Bremer 1994; Nesom 2020). On the other hand, Vernoniopsis is distinct by having a pappus of scabrid bristles and epaleate receptacles, whereas Apodocephala lacks a true pappus and has paleate receptacles (Humbert 1960). Nesom (2020) placed Apodocephala and Vernoniopsis in his Astereae subtribe Madagasterinae together with Madagaster and Rochonia, all of which are endemic to Madagascar. Madagaster and Rochonia differ from Apodocephala and Vernoniopsis in having radiate heads and campanulate involucres, and from Apodocephala in having epaleate receptacles and a pappus of barbel-late bristles (Humbert 1960; Nesom 2020). Molecular phylogenetic studies place Madagaster within Astereae (Brouillet & al. 2009). Rochonia and Vernoniopsis have yet to be included in a molecular phylogenetic analysis, meaning that their phylogenetic positions within the Asteraceae remain untested. Given the present results, it is possible that these two genera are also misplaced in the Astereae.

The tribe Athroismeae was originally described by Panero & Funk (2002) to accommodate the genera Athroisma DC., Blepharispermum Wight ex DC. and Leucoblepharis Arn., previously members of the Inuleae, but had since then been classified in a broadly delimited tribe Heliantheae, as the Athroisma group (Kim & Jansen 1995). Over the years, the limits of the tribe have greatly changed, as more genera have been found to belong there (Wagstaff & Breitwieser 2002; Panero 2005; Anderberg 2009; Pruski 2014; Bentley & al. 2015; Bengtson & al. 2017). Bengtson & al. (2017) presented the first molecular phylogenetic study of the tribe, which resulted in an amended circumscription. Three monospecific genera (Anisochaeta DC., Artemisiopsis S. Moore and Symphyllocarpus Maxim.), all earlier placed in other tribes, were also shown to belong in the Athroismeae. This broadly delimited Athroismeae are a morphologically diverse lineage of nine genera, which are classified in four subtribes (Athroisminae, Anisopappinae, Lowryanthinae and Symphyllocarpinae). The tribe is difficult to characterize morphologically and there seem to be no obvious apomorphic morphological characters that unite its members. Bengtson & al. (2017) noted that, considering the history of the tribe, it would not be unlikely that new additions to the tribe would be discovered as more of the odd genera of the family Asteraceae were sequenced.

Our analyses resolve the two sampled species of Apodocephala, A. oliganthoides and A. pauciflora, as a monophyletic group, which is sister to the Malagasy Lowryanthus rubens (Fig. 3). The Apodocephala–Lowryanthus clade is in turn sister to the subtribe Athroisminae (Fig. 3). Lowryanthus is a monospecific genus described by Pruski (2014) and consists of shrubs or small trees with bright red, corymbiform-paniculiform synflorescences bearing coral red, discoid capitula with reddish to pink florets and red involucres (Fig. 1C, D). The genus has carbonized, obcompressed, geniculate-rostrate cypselas, which led Pruski (2014) to correctly postulate its close affinities with the Blepharispermum group (= subtribe Athroisminae) of the tribe Athroismeae. Eriksson (1990, 1992) described the Blepharispermum group to have cypselas with a neck, which in some species is abaxially curved; similar features are also found in Lowryanthus. The molecular phylogenetic study of Athroismeae by Bengtson & al. (2017) confirmed the placement of Lowryanthus in the Athroismeae, and showed that the genus is sister to the subtribe Athroisminae. Lowryanthus, however, differs from the members of Athroisminae in having free, solitary capitula rather than capitula assembled in secondary heads, and was therefore placed in its own subtribe, Lowryanthinae (Bengtson & al. 2017).

Fig. 3.

Bayesian fifty-percent majority-rule consensus tree from an analysis of the combined Athroismeae dataset (ETS, ITS, ndhF and trnL–trnF). The different subtribes are marked in the figure. Numbers above branches indicate posterior probability (PP) and bootstrap support (BS) values, bootstrap support values < 50 are indicated by a dash. Scale bar shows number of nucleotide substitutions per site.

The sister-group relationship of Apodocephala and Lowryanthus is supported by both genera having discoid capitula with bisexual florets arranged in corymbiform-paniculiform synflorescences, paleate receptacles and cypselas lacking a true pappus (Humbert 1960; Pruski 2014). Morphological studies of cypsela characters reveal further similarities. Lowryanthus has geniculate-rostrate cypselas with a tightly curved rostrum. This character is also found in Apodocephala, where several species have cypselas elongated to a closely curved, s-shaped neck (Humbert 1960, 1962). Apodocephala coursii Humbert and A. radula Humbert in particular have cypselas resembling those of Lowryanthus. Apodocephala oliganthoides and A. pauciflora, the two species included in the molecular phylogenetic analyses, do not share this character, and it has evidently been lost. Further, both Lowryanthus and members of the Athroisminae have cypselas with long marginal and apical hairs (Eriksson 1990, 1992, 1995; Pruski 2014), which are also found in A. coursii (Humbert 1962). The morphological similarities in cypselas support the position of Apodocephala as sister to Lowryanthus and as part of the Lowryanthinae subtribe of the Athroismeae. The different Apodocephala species, however, show much variation in capitula and cypsela characters, and it is worth noting that the species included in the molecular study are not the ones most resembling Lowryanthus.

Apodocephala and Lowryanthus differ greatly in general appearance, in the colour of the peduncles, capitula and florets, and in the shape of the capitula and involucral bracts (Fig. 1). They also have distinct geographic ranges, meaning that they do not grow sympatrically. Apodocephala and Lowryanthus are not the only examples of morphologically distinct sister genera. Another example from Madagascar is Landiopsis Capuron ex Bosser and Bremeria Razafim. & Alejandro in Rubiaceae, which have been shown to be sister genera by molecular studies but share no known synapomorphies (Alejandro & al. 2005). The former genus is monospecific and is restricted to the dry deciduous forests in northern Madagascar. In contrast, the latter genus is commonly found throughout the lowland and montane rainforests. The Central African genus Colletoecema E. M. A. Petit and its sister genus, the Seychellan Seychellea Razafim. & al. (Rubiaceae), constitute another example, however both genera are restricted to lowland rainforests (Razafimandimbison & al. 2020).

Biogeographic remarks

Madagascar is a biodiversity hotspot, with remarkable levels of diversity and endemism. Biogeographical studies show that the majority of the lineages ancestral to the Malagasy endemics have African origins. Numerous endemic clades result from dispersals from the African mainland, followed by subsequent diversification (Yoder & Nowak 2006). The Malagasy Athroismeae also seem to fit this biogeographical pattern, although a biogeographical analysis would be needed to confirm this. The tribe Athroismeae has a mainly African distribution, but with several species in the Arabian Peninsula, Asia, and even Australia and New Zealand (Centipeda Lour.). All four subtribes of the Athroismeae have representatives in Madagascar, and six out of the ten genera in Athroismeae (Anisopappus Hook. & Arn., Apodocephala, Athroisma, Blepharispermum, Centipeda and Lowryanthus) occur in Madagascar. Comparisons of the phylogeny and distributions show that the Malagasy Athroismeae result from multiple independent colonization events, mostly from the African mainland (Bengtson & al. 2017, 2021). Two genera (Apodocephala and Lowryanthus), and about 12 species (three Athroisma and about nine Anisopappus; the Malagasy Anisopappus under revision, Bengtson & al. in prep.) are endemic to Madagascar and two additional species (Blepharispermum arcuatum T. Erikss. and Centipeda minima (L.) A. Braun & Asch.) are present on the island but are not endemic there.

Anisopappus (Fig. 1E) is represented by nine species in Madagascar, and shown by Bengtson & al. (2021) to be the result of two separate colonization events from mainland Africa followed by subsequent diversification. Athroisma (Fig. 1F) contains three species endemic to Madagascar (Eriksson 1995), although only one of the species is included in this study and in Bengtson & al. (2017); whether these three species result from a single dispersal event or not remains to be seen. The analyses place the Malagasy Athroisma pinnatifidum T. Erikss. within an African clade and as sister to a specimen of Athroisma laciniatum DC. from India. The sister-group relationship indicates that they could be the result of a single dispersal event from Africa to Madagascar followed by secondary dispersal to India or vice versa (a single dispersal event to India from Africa followed by a second dispersal to Madagascar from India). The subtribe Lowryanthinae, consisting of Lowryanthus and Apodocephala, is endemic and appears to result from a single colonization event, followed by subsequent diversification in Madagascar.

Lowryanthus and most Apodocephala are restricted to rainforest habitats. In contrast, Athroisma seems to be confined to dry forest or scrub habitats in western and southern Madagascar (Eriksson 1995), while Anisopappus mainly thrives in open habitats (e.g. woodland savannas or forest gaps) or rocky habitats (Humbert 1960). Lowryanthus rubens is known only from the Bemangidy forest in the northern part of the Tsitongambarika protected area in southeastern Madagascar, where it is found in low-elevation humid evergreen forests between 100 and 300 m elevation (Pruski 2014). In contrast, Apodocephala has a wider distribution ranging from the southeast, the central plateau, the north and northeast of Madagascar and mostly occurs in higher elevations from 700 m to 1500–2000 m, where the species are found in montane rainforests. The only exception is Apodocephala minor Scott Elliot, which occurs at low altitudes in siliceous, rocky habitats in southeastern Madagascar (Humbert 1960). Apodocephala minor and A. radula are the Apodocephala species found in the closest geographical proximity of the known Lowryanthus localities; however, the two Apodocephala species and Lowryanthus differ in habitat and A. radula occurs at higher elevations.

It is worth noting that the subtribe Lowryanthinae joins the growing list of the Malagasy endemic plant lineages. Beside the six Malagasy endemic plant families (Aphloiaceae, Asteropeaceae, Barbeuiaceae, Physena-ceae, Sarcolaenaceae and Sphaerosepalaceae), other examples are: tribe Coleeae (Bignoniaceae), subfamily Didiereoideae (Didiereaceae), subfamily Diegodendroi-deae (Bixaceae), tribe Humbertieae (Convolvulaceae) and tribe Tseboneae (Sapotaceae, Gautier & al. 2013).

Conclusions

Apodocephala is here revealed to be part of the tribe Athroismeae rather than Astereae, and therefore forms another addition to this tribe. Lowryanthus and Apodocephala are sisters that constitute the Malagasy endemic subtribe Lowryanthinae. A more detailed morphological study including all Apodocephala would be rewarding, as it could shed light on the potential morphological synapomorphy of the Apodocephala–Lowryanthus clade (in prep.).

Taxonomic treatment

Lowryanthinae Pruski & Anderb. in Taxon 66: 417. 2017, emend. Bengtson, Anderb. & Razafim. – Type: Lowryanthus Pruski.

Shrubs or trees, 1–4(–30) m tall. Leaves alternate, petiolate; blade oblanceolate to narrowly obovate, coriaceous, glabrous to hirsute, pinnately veined, margin entire to denticulate. Capitulescence terminal, corymbiform-paniculate. Capitula discoid, paleate; florets (1–)3 or 4(–20); corolla white, whitish or red. Cypselas oblong and sub-prismatic to obovate and obcompressed, in some species rostrate, brown or black, epappose.

Genera: Apodocephala Baker, Lowryanthus Pruski.