Recircumscription of the Canary Island endemics Argyranthemum broussonetii and A. callichrysum (Asteraceae: Anthemideae) based on evolutionary relationships and morphology

Abstract: Phylogenetic analysis has revealed that Argyranthemum broussonetii is polyphyletic, with subsp. broussonetii (Tenerife) and subsp. gomerensis (La Gomera) resolved in separate clades within the Macaronesian endemic genus Argyranthemum. We show that A. broussonetii subsp. broussonetii and subsp. gomerensis are similar in leaf traits, likely a consequence of adaptation to similar habitats, but that the two can readily be distinguished based on capitula and cypselae characteristics. Indeed, A. broussonetii subsp. gomerensis was found to share greater affinity with A. callichrysum, also from La Gomera based on capitula and cypselae characters, in agreement with its phylogenetic placement. Therefore, we propose that A. broussonetii subsp. gomerensis should be recognized as a subspecies of A. callichrysum, specifically A. callichrysum subsp. gomerensis. A key to differentiate A. broussonetii, A. callichrysum subsp. callichrysum and subsp. gomerensis is provided. Citation: White O. W., Reyes-Betancort J. A., Chapman M. A. & Carine M. A. 2021: Recircumscription of the Canary Island endemics Argyranthemum broussonetii and A. callichrysum (Asteraceae: Anthemideae) based on evolutionary relationships and morphology. – Willdenowia 51: 129–139. Version of record first published online on 29 April 2021 ahead of inclusion in April 2021 issue.


Introduction
Argyranthemum Webb is the largest endemic flowering plant genus of the Macaronesian region. It comprises 24 species (39 terminal taxa when subspecies are considered; Humphries 1976) and is distributed across the archipelagos of Madeira, the Selvagens and the Canary Islands.
Argyranthemum broussonetii (Pers.) Humphries is endemic to the Canary Islands where it is restricted to laurel forest clearings on Tenerife and La Gomera. It is distinguished from other members of the genus by its large bipinnatifid leaves, its often wingless ray cypselae, typically two-winged disc cypselae, and large capitula and ray florets. Humphries (1976) recognized two subspecies, namely A. broussonetii subsp. broussonetii endemic to Tenerife and subsp. gomerensis Humphries endemic to La Gomera ( Fig. 1; Fig. 2). He differentiated the two largely by size, with subsp. broussonetii larger than subsp. gomerensis in stature, leaf size, involucre width and ray cypselae dimensions.
A phylogenetic analysis of Argyranthemum based on chloroplast restriction site markers found that A. brous sonetii is not monophyletic, because the two subspecies were resolved in different clades (Francisco-Ortega & al. 130 White & al.: Recircumscription of Argyranthemum broussonetii and A. callichrysum 1996). However, the sister relationships of A. brous sonetii remained unresolved. More recently, a phylogenetic study of Argyranthemum that employed a Next Generation Sequencing (NGS) approach (White & al. 2020) provided improved resolution of species relationships and confirmed that the two subspecies of A. broussonetii are distinct evolutionary lineages. Specifically, A. brous sonetii subsp. broussonetii is sister to the multi-island endemic A. frutescens (L.) Sch. Bip., whereas A. broussonetii subsp. gomerensis is sister to A. callichrysum (Svent.) Humphries, which is also endemic to La Gomera ( Fig. 2;  Fig. 3). D-statistics computed to test for evidence of hybridization provided no evidence of introgression between A. broussonetii subsp. broussonetii and A. callichrysum, ruling out the possibility that A. broussonetii subsp. gome rensis is the result of historical hybridization between A. broussonetii and A. callichrysum. Francisco-Ortega & al. (1996) suggested that the subspecies of Argyranthemum broussonetii are morphologically distinct and that each should be considered a unique species; however, no subsequent taxonomic work on A. broussonetii has been published. In this study, we reassess the taxonomy of A. broussonetii in the light of recent phylogenetic analyses. The morphological distinctiveness of the two subspecies of A. broussonetii is demonstrated, with subsp. gomerensis found to be more similar to A. callichrysum. We propose that A. brousso netii subsp. gomerensis should be recognized as a subspecies of A. callichrysum and provide a key with which to differentiate A. broussonetii and the subspecies of A. callichrysum.

Material and methods
We identified ten characters that differed between the two subspecies of Argyranthemum broussonetii and A. callichrysum based on Humphries (1976). These were: leaf attachment (petiolate, shortly petiolate or sessile); leaf dissection (bipinnatisect or bipinnatifid); primary lobe length (cm); primary lobe width (cm); primary lobe shape (linear-lanceolate or obovate); capitulum width (cm); ray cypselae colouration (yellow-brown, chestnutbrown, brown-purple or black); ray cypselae arrangement (solitary, coalesced in groups of 2 -6 or both identi-fied in the same capitulum); ray cypselae wings (present or absent); and disc cypselae wing number (zero, one or two wings). Leaf dissection was estimated using a ratio of leaf width and lamina width (Fig. 4). For an estimate of lobe shape, a ratio of lobe length and width was calculated. Descriptions of the characters and definitions are provided in Table 1 and Fig. 4. A total of 27 recent collections of Argyranthemum made between July and August 2015 were examined and scored for these traits together with ten further collections accessioned at the Natural History Museum, London (BM). The 37 voucher specimens examined for the morphometric analysis comprised 16 A. broussonetii subsp. broussonetii, 14 A. broussonetii subsp. gomerensis and seven A. callichrysum.
For continuous characters, we checked for normality using histograms, Q -Q norm plots and a Shapiro-Wilk test. Characters with a non-normal distribution were transformed and we tested for significant differences between taxa using ANOVA followed by post hoc Tukey's honest significant difference test. For discrete variables, we used chi-square tests of independence to identify significant associations between character frequencies and taxa. To further investigate the morphological relationships, we employed the R package PCAmixdata (Chavent & al. 2012(Chavent & al. , 2017 which implements principal components analysis (PCA) using both continuous and discrete variables. For the PCA analysis, we excluded the ratio of lobe length and width to avoid the inclusion of strongly correlated variables. The number of clusters identified by the PCA was determined using the R package mclust (Scrucca & al. 2016; R Core team 2020). Further specimens at BM, ORT and RNG were examined to further test the taxon circumscriptions proposed in light of the morphometric analysis.

Discussion
From a morphological perspective, we found that Argy ranthemum broussonetii subsp. broussonetii and subsp. gomerensis are similar in leaf traits including primary lobe length and leaf dissection (leaf:lamina width ratio), which supports their current taxonomic treatment. However, A. broussonetii subsp. broussonetii and subsp. gomerensis can be distinguished based on leaf attachment, capitula width and cypselae characteristics. Indeed, A. broussonetii subsp. gomerensis shows greater similarity to A. callichrysum based on these characters, which is in agreement with their phylogenetic rela-tionships (White & al. 2020;Fig. 3) and the hypothesis proposed by Francisco-Ortega & al. (1996). While our ordination analysis shows some overlap between A. broussonetii subsp. gomerensis and A. callichrysum (Fig. 7), the two can be differentiated using primary lobe shape (length:width ratio; Fig. 5C) and leaf dissection (leaf:lamina width ratio; Fig. 5D). The differences in leaf morphology between A. broussonetii subsp. gome rensis and A. callichrysum are also apparent in plants grown under common glasshouse conditions (Fig. 2B), ruling out the possibility that the observed differences in wild-collected material were simply due to environmentally induced plasticity. In light of the morphological findings that are congruent with the molecular phylogenetic data, we propose that A. broussonetii subsp. gomerensis should be recognized as a subspecies of A. callichrysum. While A. broussonetii subsp. gomerensis could arguably be recognized as a distinct species, we propose recognition as a subspecies of A. callichrysum given the sister group relationships and overlap in morphology in some instances (see below). The necessary new combination is provided below together with a key to the three taxa considered here.
Morphological traits that are advantageous in a particular habitat may be susceptible to morphological convergence. The two independently derived lineages that comprised Argyranthemum broussonetii as circumscribed by Humphries (1976) occupy humid laurel forests of Tenerife and La Gomera, where they can be found in forest clearings. The similarity in leaf shape exhibited by these two lineages would appear to be a convergence in response to similarities in habitat. Based on our collections and georeferenced samples, the distributions of A. broussonetii subsp. gomerensis and A. callichrysum are largely non-overlapping (Fig. 1). They do co-occur at "El Cano" (M. Fernández Galván, ORT26020; Supplementary Table S2) and "Laderas del Roque Cano" (M. Fernández Galván, ORT26022), although here A. brous sonetii subsp. gomerensis is typically distributed on Nfacing slopes, whereas A. callichrysum is generally found on SW-facing slopes. Two specimens of A. callichrysum from "Taguluche Norte" (E. R. Sventenius, ORT4843 and ORT4846; Supplementary Table S2) are also outliers for this species. In Argyranthemum, hybrids may occur where taxa come into close proximity, given weak barriers to gene flow (Borgen 1976;Brochmann 1984;Brochmann & al. 2000

Taxonomy
Note -in the list of specimens seen, those used in the morphometric analysis are indicated in boldface. Distribution -La Gomera: scattered populations on steep slopes of La Gomera between Las Rosas, La Palmita and Agulo on the NW coast (Fig. 1). Distribution -La Gomera: distributed from Barranco de Argaga on the SW coast to the central mountains of Igualero, Agando and Tagamiche, where it is found on S-facing slopes (Fig. 1).

Key to taxa
Habitat -Associated with xerophytic scrub on rocky slopes or disturbed vegetation. Notes -Chrysanthemum callichrysum, the supposed basionym of Argyranthemum callichrysum, was published by Sventenius (1960). The validity of names published in this work has been called into question, but Bramwell (2019) has argued convincingly that they are validly published. In the case of the Argyranthemum names described by Sventenius, the types were examined by Humphries, who visited ORT in 1971 and cited them in his monograph (Humphries 1976).
Unfortunately, in the case of Chrysanthemum cal lichrysum, the name was not validly published because Sventenius cited two gatherings, "Legit cum flore die 18 Maio 1945 et cum fructu die 12 Augusto 1952", without indicating either one of them as the type. If he had cited only one gathering, the name would have been validly published, even though he did not explicitly designate a type (Turland & al. 2018: Art. 40.2, 40.3). Humphries (1976) explicitly cited a single specimen (or gathering) as the holotype and provided a full and direct reference to the Latin description previously published by Sventenius, thereby validly publishing Argyranthemum callichrysum as the name of a new taxon, not a new combination.
During the course of this study, we were unable to find a specimen with a locality that matched exactly that given in the original description of Chrysanthemum callichrysum by Sventenius ("circa pagum 'Igualero' dictum, versus 1.000 m. supra mare" [near to the village called Igualero, toward 1000 m above the sea]) or in the protologue by Humphries ("Iguelero [sic], 1000 m, 18 May 1946", presumably in error for 1945). Nevertheless, there is a specimen at ORT collected by Sventenius (ORT4828) on the same date (18 May 1945) as that cited in his original publication, from "Fuente de la Yegua", a locality less than 1 km from Igualero. We therefore propose that this specimen is one of those to which Sventenius was referring as well as the holotype cited by Humphries.  Fig. 7. PCA based on continuous and discrete characters. Point colour indicates the taxon recognized in this study, whereas shape indicates the cluster defined using mclust. The proportion of variation explained on each dimension is shown in parentheses. -Taxon abbreviations: bro = Argyranthemum broussonetii; gom = A. callichrysum subsp. gomerensis; cal = A. callichrysum subsp. callichrysum (names as accepted in this study).