Molecular phylogeny of the ‘Cape snow’ genus Syncarpha (Asteraceae: Gnaphalieae) reveals a need for generic re-delimitation

Abstract A phylogenetic hypothesis is presented for the charismatic but taxonomically poorly-known Cape daisy genus Syncarpha , based on near-complete ingroup sampling and good coverage of outgroup taxa. A combination of nuclear ribosomal and chloroplast spacer DNA sequence data gives a well-resolved phyogenetic hypothesis, the robustness of which is assessed via both parsimony bootstrap and Bayesian posterior probabilities based on the uncorrelated lognormal relaxed clock model. Syncarpha species fall into two well-supported and distantly-related clades that last shared a common ancestor in the mid-Miocene. The larger Syncarpha 1 grouping contains the type species and corresponds to African ‘ Helipterum ’; this clade is sister to Edmondia and belongs in a larger clade which also includes the Australian Gnaphalieae. The Syncarpha 2 clade contains the taxa associated with Syncarpha paniculata (formerly Helichrysum paniculatum ) and is more closely related to Plecostachys and some species of Gnaphalium . Formal assessment of monophyly lays the groundwork for future revisionary taxonomic work.


Introduction
The genus Syncarpha DC. (Asteraceae: Gnaphalieae) comprises 28 species as currently circumscribed (Nordenstam, 1989(Nordenstam, , 2003Manning and Goldblatt, 2012). All species are sub-shrubs with homogamous flowerheads surrounded by conspicuous coloured involucral bracts, and almost all Syncarpha species are endemic to the Cape Floristic Region of South Africa (CFR; sensu Goldblatt, 1978) and confined to fynbos vegetation. Despite being a charismatic and conspicuous component of the Cape flora on account of its showy capitula, which are typically large and surrounded by several whorls of elongate yellow, pink or white papery bracts (Fig. 1), the systematics of the genus remains poorly explored. Consequently, Syncarpha has recently been identified as requiring revision (Von Staden et al., 2013), the last comprehensive treatment being that of Harvey in Flora Capensis (Harvey and Sonder, 1894). Originally described in the Linnaean genera Xeranthemum, Staehelina, Helichrysum or Gnaphalium (Linnaeus, 1760, 1763, 1767, many Syncarpha species were transferred by De Candolle (1838) to the genus Helipterum DC., which comprised both South African and Australian taxa and was distinguished from Helichrysum by its plumose pappus bristles, fused at the base into a smooth ring (Hilliard and Burtt, 1981). The name Helipterum is illegitimate (Nordenstam, 1989) and so all the South African members of the genus were transferred by Nordenstam (1989Nordenstam ( , 2003 to the genus Syncarpha DC. Nordenstam (1989Nordenstam ( , 2003 also transferred some Helichrysum and Gnaphalium species to Syncarpha (based on characters presented in Hilliard and Burtt, 1981), but did not provide a description of the genus. The morphological characters defining the genus have nowhere been systematically investigated, although Hilliard and Burtt (1981) discussed character variation in Syncarpha and related genera. Syncarpha species are all subshrubs with felty leaves possessing flat margins; the involucral bracts comprise a basal, undivided stereome (although S. argyropsis is exceptional in having a fenestrated stereome) and a lanceolate, papery lamina; the capitula are homogamous and the receptacles fimbrilliferous. Cypsela hairs in Syncarpha are duplex and either globose or very depressedglobose, while the pappus consists of bristles that vary from smooth through barbellate to plumose and are fused basally into a smooth ring (Hilliard and Burtt, 1981). Apart from the brief description in Anderberg (1991), there is to date no examination of species concepts in light of all available specimens, no key to the species, and no summary of geographic distributions. The need for a comprehensive and modern revision of the genus is underlined by the recent description of several new species (Nordenstam, 1989(Nordenstam, , 2003, and by the red-listing of several species, including S. aurea (threatened), S. chlorochrysum (near threatened), S. dykei (rare), S. lepidopodium (vulnerable), S. marlothii (rare), S. montana (rare), S. recurvata (endangered), S. sordescens (vulnerable) and S. zeyheri (rare; http://redlist.sanbi.org/ index.php; accessed January 2015).
Although phylogenetic research and revisionary taxonomy are sometimes perceived as distinct and competing (for funding) branches of systematics (e.g. see Wortley et al., 2002), these disciplines are, in truth, highly complementary. Both seek to resolve evolutionarily meaningful entities (species or monophyletic higher taxa) via a process of hypothesis testing, and both have important roles to play in the description of biodiversity (Wheeler, 2004). In fact, it is widely accepted in the scientific community that phylogenetic results are indispensible in any modern taxonomic treatment, and in a phylogenetic system of classification, a formal assessment of monophyly constitutes an essential foundation for revisionary work on a particular taxon.
As the basis of a revision currently in preparation, therefore, we generate a phylogenetic hypothesis for Syncarpha with the principal aim of evaluating its monophyly. Although Syncarpha has been included in a number of studies exploring phylogenetic relationships within the tribe Gnaphalieae (Bayer et al., 2000;Bergh and Linder, 2009;Galbany-Casals et al., 2010Bergh et al., 2011) these studies have consistently each sampled only a single species of Syncarpha. Thus, while it is now evident that Syncarpha is nested within the "crown radiation" of Gnaphalieae (Ward et al., 2009), it remains unknown whether the genus is monophyletic and, therefore, best treated in a single revision. We address this gap, by presenting a wellsampled phylogenetic hypothesis including all but two species of Syncarpha, along with representatives of a large range of related genera. Although our main phylogenetic output is a time-calibrated tree generated in BEAST (Drummond and Rambaut, 2007) demonstrating the times of divergence of clades of interest, it is not our goal to here discuss the evolutionary history of the group in any depth. Also, while it is our intention to provide a solid foundation for revisionary work, it is not our intention at this stage to implement formal taxonomic or nomenclatural changes.

Taxon sampling
Our analyses included most species currently accepted in Syncarpha and outgroup sequences from a range of genera representing all major lineages of Gnaphalieae, but especially the "crown radiation" clade of Ward et al. (2009) (Table 1). The bulk of the sequences used were generated de novo, either from field-sampled leaf material (vouchers deposited in the Compton Herbarium, NBG or the National Herbarium, PRE) or from herbarium specimen material, though some outgroup sequences were taken from GenBank. Except for the widespread taxa S. canescens (two accessions), S. paniculata (four accessions) and S. staehelina (two accessions), species were represented by single accessions.

DNA isolation, amplification and sequencing
Silica-dried leaf samples (about 15 mg) were ground using a mixer mill (MM 400, Retsch, Haan, Germany) and total genomic DNA isolated using a modified version of the CTAB protocol described by Doyle and Doyle (1987). Extracted DNA was suspended in 100 μl Tris-EDTA buffer. For herbarium material and problematic field-collected samples, DNA was isolated using the DNeasy Plant Mini Kit (Qiagen GmBH, Hilden, Germany).
For phylogenetic analysis, we sampled two nuclear and two chloroplast regions. The 3′ end of the external transcribed spacer (ETS) of nuclear ribosomal DNA was amplified using the primers ETS-1F and 18S-ETS (Baldwin and Markos, 1998;Markos and Baldwin, 2001), while the associated internal transcribed spacer (ITS) region (comprising the ITS1 and ITS2 spacers and the intervening 5.8S ribosomal gene) was amplified using the ITS4 and ITS5 primers of White et al. (1990). The chloroplast trnL intron and the trnL-trnF intergenic spacer were amplified together using the 'c' and 'f' primers of Taberlet et al. (1991), while the chloroplast trnT-trnL spacer was amplified using the 'a' and 'b' primers developed by the same authors.
Target regions were amplified using PCR, in reaction volumes of 25 μl. Each sample consisted of 2.5 μl of reaction buffer (Kapa Biosystems Inc., Woburn, MA, USA), 0.5 μl DMSO, 1.25 μl of each primer at 10 μM, 0.5 bovine serum albumin (BSA), 6.0 μl (trnL-trnF and trnT-trnL) or 1.5 μl (ETS and ITS) of 25 μM MgCl 2 , 0.3 μl (trnL-trnF and trnT-trnL) or 0.2 μl (ETS and ITS) Taq polymerase (Kapa Biosystems Inc., Woburn MA, USA), 1.2 μl (trnL-trnF and trnT-trnL) or 1.0 μl (ETS and ITS) dNTP at 10 μM, and 9.5 μl (trnL-trnF and trnT-trnL) or 8.3 μl (ETS) or 10.3 μl (ITS) nuclease-free water. Two microlitres (trnL-trnF and trnT-trnL) or 7 μl (ETS) or 6 μl (ITS) of template DNA (at a dilution of either 10 −2 or 10 −3 relative to the raw isolate) was added to each reaction mixture. Amplification was performed using an Applied Biosystems 2720 thermal cycler (Foster City, CA, USA). The same thermal profile was applied to the two chloroplast loci, but different profiles were used for each nuclear region. All profiles started with an initial denaturation step of 2 min at 94°C. This was followed by 30 cycles (trnL-trnF, trnT-trnL and ETS) or 35 cycles (ITS) consisting of: 30 s (trnL-trnF and trnT- Table 1 List of accessions sampled in this study, with an indication of species identity and relevant GenBank accession numbers. Voucher specimen numbers (all deposited at NBG unless otherwise indicated) and localities are provided only for those taxa for which new sequences were generated as part of this study. For species represented by multiple accessions, these are provided with numbers (in parentheses, after the species name) which correspond to the numbers assigned to these accessions in Figs trnL) or 1 min (ETS and ITS) at 94°C; 2 min at 52°C (trnL-trnF and trnT-trnL), 1 min at 55°C (ETS) or 1 min at 45°C (ITS); and 2 min (trnL-trnF, trnT-trnL and ETS) or 1 min at 72°C (ITS). The final extension step comprised 7 min (trnL-trnF, trnT-trnL and ETS) or 10 min (ITS) at 72°C. Successfully amplified products were sequenced in both directions using the original PCR primers at the Central Analytical Facility (CAF), University of Stellenbosch, South Africa using an ABI capillary sequencer (Applied Biosystems, Inc.). Chromatograms were checked and the bi-directional sequences assembled using Geneious Pro v. 5.4.3 (Biomatters Ltd., Auckland, New Zealand). The consensus sequence was aligned manually using BioEdit Sequence Alignment Editor (v. 7.0.9.0, Hall, 1999). Stretches of sequence that could not be unambiguously aligned were excluded from phylogenetic analyses.

Phylogenetic analysis
Trees were rooted on R. pungens (representing the Relhania clade of Bergh and Linder, 2009, this clade being recovered as sister to all remaining Gnaphalieae taxa by several earlier workers: Bayer et al., 2000;Bergh and Linder, 2009;Ward et al., 2009;Montes-Moreno et al., 2010). To identify incongruence among loci, parsimony bootstrap analyses were conducted on each locus (plastid DNA, ETS, ITS), treatment of the two chloroplast regions as a single locus being justified by a preliminary comparison. These analyses were conducted in PAUP version 4.0b10 (Swofford, 2003) using 300 non-parametric bootstrap (BS; Felsenstein, 1985) replicates, heuristic tree searches starting with a simple taxon addition tree and TBR swapping, under a maxtrees setting of 500. Separate bootstrap consensus trees for each gene region or locus were compared visually, with conflict considered to be wellsupported when both competing nodes had BS ≥ 70%.
Combined analysis of the four DNA regions was performed using BEAST v1.5.4 (Drummond and Rambaut, 2007), with nodal support values also being assessed using the parsimony bootstrap as described above. The BEAST input file was generated using BEAUti v1.5.4 (Drummond and Rambaut, 2007), implementing a separate model structure for each region, the optimal model in each case being determined under the AIC in MrModeltest v.2.3 (Nylander, 2004). The GTR + I + Γ model was selected for ITS, the GTR + Γ model for ETS and trnL-trnF and the HKY + Γ model for trnT-trnL. Nodal divergence times were estimated in the context of a Yule tree prior, and a lognormal relaxed clock model (Drummond et al., 2006). Calibration was achieved by applying age priors to two nodes, the Gnaphalieae crown node (Fig. 3, node A) and the 'rest of Gnaphalieae' clade crown node (Fig. 3, node B). We used Bergh and Linder's (2009) posterior age estimates (median ± 95% HPD) to date these nodes, imposing these as normally-distributed age priors in such a way that the prior means and 95% confidence intervals matched the corresponding calibration posteriors as closely as possible. Three BEAST runs were performed, each comprising 20,000,000 iterations. Tracer v1.5 (Drummond and Rambaut, 2007) was used to assess the outputs of these runs for convergence and to determine the proportion of samples to be discarded as burn-in. LogCombiner v1.5.4. (Drummond and Rambaut, 2007) was then used to combine the post burn-in trees and parameter estimates from the three runs. Finally, a maximum clade credibility (MCC) tree, with median node heights, was generated in TreeAnnotator v1.5.4 (available for download from the BEAST website: http://beast.bio.ed. ac.uk/).

Sequence characteristics and parsimony analyses
The alignments and resulting trees are available from TreeBase at the following address: http://purl.org/phylo/treebase/phylows/study/ TB2:S16847?x-access-code=178d776a86900d6b1fea8dfbe99f973a& format=html. The ETS, ITS and combined chloroplast alignments used in analyses were 1311, 669 and 1470 base pairs long, containing 287, 173 and 97 parsimony informative characters, respectively. All three loci yielded 70% majority rule bootstrap consensus trees that were poorly resolved, containing large polytomies (Fig. 2). The available resolution, however, indicated no supported incongruence amongst the three gene trees (except for one localised example discussed below) and several congruent groupings. The plastid data, for example, recovers a clade (S1) containing 18 Syncarpha species, including the type S. gnaphaloides. This clade is also recovered in the ETS gene tree, although here three of the species fall outside of the clade, being instead unplaced on a backbone polytomy (Fig. 2). Another clade, comprising S. sordescens, S. striata, S. recurvata, S. chlorochrysum, S. mucronata and S. paniculata (S2) was consistently retrieved, as was a monophyletic Edmondia. Also, several sister pairs (e.g. S. flava + S. ferruginea, S. gnaphaloides + S. argyropsis, and S. speciossisima + S. vestita) were retrieved in more than one tree. One instance of supported incongruence was noted, but this was confined to relationships within the S1 clade: where the chloroplast data retrieved a well-supported subclade (BS = 89%) comprising S. aurea, S. dreageana, S. ferruginea, S. flava and S. variegata, the ETS data resolved S. dregeana as sister to S. dykei (BS = 71%) and S. variegata as sister to S. loganiana + S. montana (BS = 97%), while S. aurea was placed in a well-supported polytomy with S. affinis, S. staehelina and S. virgata (BS = 96%).
Divergence age estimates indicate a substantial interval between the time at which the two Syncarpha clades last shared a common ancestor (13.2 [7.9, 19.5] Ma) and the times at which they differentiated from their sister clades (Syncarpha 1: 7.0 [3.8, 10.7] Ma; Syncarpha 2: 8.3 [4.3, 12.8] Ma). The radiation of both Syncarpha clades took place after the Miocene-Pliocene boundary, the crown nodes of these two clades being dated to 5.6 [3.0, 8.5] Ma (Syncarpha 1) and 3.7 [1.8, 6.2] Ma (Syncarpha 2). Notwithstanding the recent timing of their diversification, species relationships within both Syncarpha clades are fairly well resolved. Within Syncarpha 1, in both cases in which species are represented by two accessions (S. canescens and S. staehelina), the species are recovered as monophyletic, although with low support for the latter. Within Syncarpha 1, relationships are resolved according to the ETS version of the alternative topologies shown in Fig. 2, and this is probably due to the greater phylogenetic information content of the ETS relative to the chloroplast data. Nevertheless, the ETS topology corresponds better with morphology than the plastid one, for example grouping the three species with distinctive white-and-brown involucral bracts and high-altitude distributions (the 'chocolate clade': S. variegata, S. loganiana and S. montana; PP = 0.98; BS = 98%), a clade that is not recovered in the plastid gene tree. The ETS and combined trees also recover a clade of species with yellow bracts (the 'citrine clade': S. ferruginea, S. flava, S. affinis, S. aurea and S. staehelina; PP = 1.00). Two morphologically distinctive species characterised by very large leaves, elongate, sparsely-branched stems and large, highly-clustered flowerheads (S. milleflora and S. eximia, the 'strawberry everlasting') are recovered as sister taxa in the concatenated tree, albeit with low support. Within Syncarpha 2, S. sordescens and S. striata are sister to each other (PP = 1.00; BS = 100%) and sister to a clade comprising the remaining species (PP = 1.00; BS = 96%). Within this clade, Syncarpha recurvata is sister to a subclade (PP = 1.00; BS = 100%) comprising S. paniculata, S. mucronata and S. chlorochrysum. Relationships amongst these three species are not recovered with any support, and the four accessions of the widespread and morphologically variable species S. paniculata are not recovered as monophyletic. This clade (S. paniculata, S. mucronata and S. chlorochrysum) is probably best considered as a species complex, herewith dubbed the 'S. paniculatacomplex'.

Phylogenetic relationships within tribe Gnaphalieae
Despite the low levels of resolution offered by the individual markers used in this study (Fig. 2), data combination yielded a tree that is both well resolved and generally well supported (Fig. 3). This tree corroborates earlier phylogenetic studies on Gnaphalieae (Bergh and Linder, 2009;Ward et al., 2009) in retrieving a monophyletic 'rest of Gnaphalieae' clade (sensu Bergh and Linder, 2009) with good support. The 'rest of Gnaphalieae' corresponds with the more elegantly named "Gnaphalieae crown radiation" of Ward et al. (2009), except that the latter authors included the Stoebe clade. Most earlier studies, including Ward et al. (2009) identified a topology where the Stoebe clade is sister to the 'rest of Gnaphalieae' clade, and the Metalasia and then the Ifloga clades are successively sister to these (Bayer et al., 2000;Bergh and Linder, 2009; Montes-Moreno et al., 2010 [although they did not include any members of the Ifloga clade]; Bengston et al., 2014). However, in all these studies the branch lengths subtending these nodes are short and in some cases support values are low, indicating that speciation events occurred in rapid succession and that the true branching order may be difficult to discern. Our analysis supports an alternative arrangement, also retrieved by Galbany-Casals et al. (2010; although they did not include any members of the Stoebe clade) and Bengston et al. (2011), in which the Stoebe, Ifloga, and Metalasia clades form a monophyletic group (the SIM clade) that is recovered (though not supported) as sister to the 'rest of Gnaphalieae', hereafter re-named the 'crown radiation' (node B, Fig. 2). Our crown radiation is thus the same as the clade so named by Galbany-Casals et al., 2010, and comprises all of Gnaphalieae, excluding the Relhania clade and also excluding the SIM lineages. There are several morphological synapomorphies that support the hypothesis that the SIM clade shares a common ancestor. Chief amongst these is an unusual leaf morphology, in which the leaf margins are involute, curling over the adaxial leaf surface, which is also usually densely white-felted. Frequently, the leaves are twisted so that parts of this white-felted adaxial surface face downwards. Leaves are also frequently borne on fascicles, most often subtended by a longer leaf. Although a detailed study has not been performed on all species in the SIM clade, several members of the Stoebe and Metalasia clades also share an anomalous form of secondary thickening (Lachnospermum and Phaenocoma from the Metalasia clade; Elytropappus, Disparago, and Stoebe from the Stoebe clade; Adamson, 1934) that may represent an additional, and complex, synapomorphy.
Within the Gnaphalieae crown radiation we recover both the HAP and FLAG clades identified by previous authors (Smissen et al., 2011;Galbany-Casals et al., 2010). Despite sparse sampling of the HAP clade, our data corroborates that of Bergh et al. (2011)

in indicating that
Galeomma is a member of this well-defined and very large clade (Nie et al., 2012;Smissen et al., 2011;Galbany-Casals et al., 2014), a finding supported by the fact that Galeomma, like all members of the HAP clade, has a divided stereome on the involucral bracts (Hilliard and Burtt, 1981;Hilliard, 1983;Anderberg, 1991). All other major groupings of Gnaphalieae lack the divided stereome.
The southern African genus Lasiopogon was demonstrated to be non-monophyletic by Bergh et al. (2011), who found L. debilis to be supported as sister to Gnaphalium declinatum, nested within the crown radiation, while L. glomerulatus, L. muscoides and L. micropoides were placed outside of the crown radiation, findings corroborated in the placement of single species by Bayer et al. (2000) and Galbany-Casals et al. (2010). Full species sampling of Lasiopogon, and phylogenetic analysis incorporating a wide range of crown radiation taxa, is required to determine the generic affinities of species currently assigned to Lasiopogon.

Polyphyly of Syncarpha
Strong evidence for the polyphyly of Syncarpha indicates a need for a realignment of generic boundaries. Our data segregate Syncarpha into two distinct, well supported clades, Syncarpha 1 and Syncarpha 2, each of which is embedded within a broader clade containing representatives of genera previously thought to be distantly-related (Australasian taxa for Syncarpha 1; Gnaphalium and Plecostachys for Syncarpha 2). The two clades of Syncarpha were previously united (Hilliard and Burtt, 1981;Nordenstam, 1989) because they both contained species that are sub-shrubby with grey-felted leaves, have homogamous capitula borne singly or in paniculate synflorescences, and with the capitula surrounded by many series of shiny, white, yellow or pink involucral bracts. The involucral bracts in most species have stereomes which are undivided or only partially fenestrated, a character which was influential in segregating these taxa from Helichrysum (where most were originally described). A stronger uniting feature was the presence of a unique type of cypsela hair (Hilliard and Burtt, 1981) that is very large, remarkably myxogenic, two-celled (lacking a swelling cushion) and depressed-globose in shape.
The two clades were historically recognised under different genera, with Syncarpha 1 species corresponding to the South African taxa of De Candolle's (1838) Helipterum. Helipterum was segregated from Helichrysum on the basis of plumose pappus bristles (De Candolle, 1838). Hilliard and Burtt (1981) examined the morphology of several Helichrysum and 'Helipterum' species and concluded that the plumose pappus character does not hold for all Helipterum species. Instead, they considered Helipterum to be separated from Helichrysum due to the former having an undivided or only slightly fenestrated stereome (with one exception, H. argyropsis). The finding by Hilliard and Burtt (1981) that the pappus bristles in South African Helipterum vary from being smooth with apically clavate cells, to barbellate becoming apically plumose, to plumose throughout, resulted in an acknowledgement that Helipterum cannot be defined by plumose pappus bristles. All species in Syncarpha 1 and Syncarpha 2, however, share the feature of pappus elements being fused basally into a smooth ring (Hilliard and Burtt, 1981).
Syncarpha 2 taxa (the former 'Helichrysum paniculatum group') were all housed in Helichrysum at the time that Helipterum was erected, having non-plumose pappus elements. However, their pappus elements are fused at the base, they have non-fenestrated stereomes in the involucral bracts, and their cypsela hairs are flattened-globose without a swelling cushion. This led Hilliard and Burtt (1981) to (rightly) suggest that they are erroneously housed in Helichrysum, and to suggest that they were instead more closely related to the South African species of Helipterum. When Nordenstam (1989) transferred all the South African species housed in the illegitimate Helipterum to Syncarpha, he agreed with Hilliard and Burtt (1981) and combined the 'Helichrysum paniculatum' group under Syncarpha.
Differences between the two groups Syncarpha 1 (the South African species formerly in Helipterum) and Syncarpha 2 (the species termed the 'Helichrysum paniculatum' group by Hilliard and Burtt, 1981) are evident in the leaves (which are erect with flat margins and obtuse tips in Syncarpha 1, but often apically hooked, weakly to strongly involute, and with acute tips in Syncarpha 2); the pappus bristles (which are scabrid throughout in Syncarpha 2) and the anthers, which are apically caudicled or mucronate in Syncarpha 2.
Although there are morphological differences between the two clades of Syncarpha, each difference has an apparent exception and S. argyropsis from Syncarpha 1, for example, has a very similar appearance to a typical Syncarpha 2 species (Fig. 1). Thus, homoplasy within each of the two clades, and across Gnaphalieae in general, serves to obscure true relationships. High levels of morphological homoplasy are well known across Gnaphalieae (e.g. Anderberg, 1991;Bayer et al., 2000), making generic circumscription extremely difficult on the basis of morphology alone.

Syncarpha 1 clade
The Syncarpha 1 clade + Edmondia is recovered as sister to two Australasian taxa, corroborating the study of Bergh and Linder (2009) which found, on the basis of a more sparsely-sampled Syncarpha 1 + Edmondia (only one species of each) but a much denser sampling of Australasian taxa (24 species), the same sister relationship. This is an interesting result considering that the Syncarpha 1 clade corresponds in its membership to the former genus Helipterum which originally comprised both Cape and Australian members. Helipterum was once one of the largest genera of Australian Gnaphalieae, but after its illegitimacy and polyphyly were recognised, almost all Australasian Helipterum species were transferred to other genera (Wilson, 1989a(Wilson, , 1989b(Wilson, , 1992a(Wilson, , 1992b(Wilson, , 1992c. Most were transferred to Rhodanthe Lindl. but the genera Leucochrysum (A.Cunn. ex DC.) Paul G.Wilson, Hyalosperma Steetz., Gilberta Turcz. and Erymophyllum Paul G.Wilson are also involved. The monophyly of the Australasian members of Gnaphalieae has never been examined with exhaustive sampling, and although several broadly-sampled analyses of the tribe appeared to recover monophyletic or near-monophyletic Australasian radiations (Bayer et al., 2002;Bergh and Linder, 2009;Ward et al., 2009;Smissen et al., 2011), all are based on sparse sampling and none has a majority of well-supported branches. The Australasian gnaphalioid flora is characterised by a large number of small or monotypic genera (Bayer et al., 2002), and by many genera of doubtful monophyly (Schmidt-Lebuhn and Constable, 2013) making interpretation of broadlysampled phylogenetic studies more difficult. Nevertheless, a close relationship between members of Syncarpha 1 and at least some Australasian taxa is clearly indicated by the results of the present, and other studies (e.g. Bergh and Linder, 2009). The Australian gnaphalioid flora is extremely diverse, harbouring ca. 475 species (Bayer et al., 2002), but preliminary phylogenetic analyses (N. Bergh, unpubl.) indicate that the clade Syncarpha 1 + Edmondia remains monophyletic with increased sampling of Australasian taxa (representatives of the genera Anaphalioides, Anemocarpa, Chrysocephalum, Craspedia, Ewartia, Gilberta, Hyalosperma, Leucochrysum, Ozothamnus, Parantennaria, Pterygopappus, Pycnosorus, Raoulia, Rhodanthe and Stuartina). The morphological characters previously thought to unite the South African and Australian species of Helipterum (De Candolle, 1838) are clearly homoplasious.

Syncarpha 2 clade
Gnaphalium appears to be non-monophyletic, since Galbany-Casals et al. (2010) recovered G. supinum as a member of the FLAG clade, while Smissen et al. (2011) recovered the clade ((G. austroafricanum + G. uliginosum) Syncarpha mucronata) as a lineage separate from the FLAG clade. Although we did not include G. uliginosum in our analysis, the above studies indicate that this species is likely to group with G. austroafricanum and Syncarpha 1 in the clade marked 'C' in Fig. 3, distant from the FLAG clade. Gnaphalium uliginosum is the type of the genus, and so the findings of Smissen et al. (2011) support the conclusion that our Syncarpha 2 clade and Plecostachys serpyllifolia are members of a clade ('C' in Fig. 3) that also includes, according to taxa sampled so far, the type species of the genus Gnaphalium as well as mainly southern African members of the genus. Both Plecostachys and the 13 southern African members of Gnaphalium are characterised by grey-felted leaves and involucral bracts with undivided stereomes and white-tipped involucral bracts. The heads in these taxa are heterogamous (comprising both female and hermaphrodite florets) and the Syncarpha 2 clade is thus a lineage of taxa with homogamous capitula contained within this 'true Gnaphalium' clade. The southern African Gnaphalium species vary in the degree of basal fusion of the pappus bristles, but nearly half of the species have the bristles fused basally into a smooth ring, supporting their affinity with Syncarpha 2 species. This character, which was influential in the decision of Hilliard and Burtt (1981) and Nordenstam (1989) to sink the Helichrysum paniculatum group into Syncarpha, is here identified as being homoplasious.
At present, there is still considerable uncertainty surrounding the relationships of the major lineages within Gnaphalieae, especially with regards to the larger genera such as Gnaphalium, and this is likely be resolved only with additional sampling of both taxa and markers.

Morphological characters differentiating Syncarpha 1 and Syncarpha 2
Full identification of the characters that distinguish the two clades of Syncarpha will require detailed examination of morphological features, and good candidates include the leaf margins, pappus bristles, style apex, anther apical appendage and cypsela hairs (Hilliard and Burtt, 1981). This work will form part of a revision of the two clades, but several differences can be enumerated here, despite that fact that most characters exhibit some overlap between the two groups. Leaves in the Syncarpha 1 clade are generally held erect and are usually apically rounded or with a weakly mucronate tip; leaf margins are flat and the hairs on the apical margins are usually distinctly tinged a rust-brown colour. In contrast, leaves in the Syncarpha 2 clade are generally smaller and more rigid, often recurved, with the tips acute to acuminate and often hooked. The leaves in this clade are never characterised by rustcoloured hairs on the margins. In Syncarpha 1 taxa, only the midvein is apparent on the lower (abaxial) leaf surface, if any, while the leaves in Syncarpha 2 are characterised abaxially by three or more very distinct parallel veins originating from the base of the leaf (a feature shared with Plecostachys).
The heads in Syncarpha 1 clade species are in many cases distinctly cylindrical in shape and usually at least 2.0 cm in diameter, although some taxa have heads only about 1.5-2.0 cm wide (S. gnaphaloides, S. argyropsis) The heads in Syncarpha 2 taxa are always fairly small (up to 1.5 but usually 1.0 cm or less in diameter) and frequently globose in outline, at least as wide as they are long, with a rounded apex. Syncarpha argyropsis, in Syncarpha 1, has rounded heads but they are generally larger (to about 2 cm) than in Syncarpha 2 taxa.
The pappus setae in both Syncarpha clades are fused at the base into a smooth ring. The two groups differ in that the pappus bristle shaft in Syncarpha 1 taxa is smooth, barbellate or plumose, with the setae towards the apex of the pappus bristle always becoming longer and ending in clavate cells. In contrast, the pappus bristles in the Syncarpha 2 clade are uniformly scabrid or barbellate.
Two species of Syncarpha for which we were unable to obtain DNA sequence data, S. argentea and S. virgata, can confidently be placed into one of the two clades based on morphological features and affinities with sampled species. Syncarpha argentea with small heads, a southern Cape coastal distribution (see Section 4.7., below) and recurved leaves is most probably a member of Syncarpha 2, while S. virgata most likely belongs in the Syncarpha 1 clade, where its yellow involucral bracts place it in the 'citrine' clade, probably close to S. staehelina with which it has been previously synonymised.

Edmondia
Edmondia comprises three species, all included in our analysis, that form the sister clade to the Syncarpha 1 clade. Edmondia was erected by Cassini (1818), with type species Helichrysum sesamoides (L.) Willd. De Candolle (1838) included Edmondia as a section within the genus Helipterum, but it was later returned to Helichrysum by Harvey (Harvey and Sonder, 1894) where it remained until Hilliard and Burtt (1981) reinstated Edmondia on the basis of (mainly) its distinctive foliage. Our analysis supports the separation of Edmondia and the Syncarpha 2 clade, but the placement of Edmondia sister to Syncarpha 1 raises the question of merging the two genera. Hilliard and Burtt (1981) and Hilliard (1983) used morphological grounds to keep Edmondia separate from Helipterum; since the Syncarpha 1 clade corresponds closely with the concept of Helipterum used by these authors, we feel that the morphological arguments retain their weight and we support maintaining a separate Edmondia and Syncarpha 1. Edmondia species are distinguished from those in Syncarpha 1 by having leaves that are adaxially glabrous but abaxially involute and white-tomentose; by scaly peduncles; by stereomes that are at least partially fenestrated; and by heterogamous heads (although the last character is not present in all species).

Origin and diversification
Based on the calibration employed here, the radiation of the two Syncarpha clades are very recent compared with most other dated fynbos clades, but they are nested within a greater Cape-centred radiation: that of tribe Gnaphalieae (Verboom et al., 2014). Gnaphalieae is an unusual lineage of the Asteraceae in that it houses several radiations of fynbos-vegetation taxa, i.e. those endemic to the oligotrophic, quartzitic environments of the CFR. The recent ages of the Syncarpha radiations closely match the radiations of two other dated fynbos gnaphalioid lineages (Metalasia clade: Bengston et al., 2014;Stoebe clade: Bergh et al. in prep.) in being very recent. The parallels with Metalasia are particularly striking. Like the Metalasia A clade, which is also centred in the eastern CFR, the radiation of the Syncarpha 2 clade appears to have been initiated in the Late Pliocene (Syncarpha 2: 3.7 [1.8, 6.2] Ma; Metalasia A: 3.3 [1.3, 6.7] Ma). By contrast, the Syncarpha 1 clade more closely resembles the Metalasia B clade both in being more widespread, and in having its diversification initiated at the Miocene-Pliocene boundary (Syncarpha 1: 5.6 [3.0, 8.5] Ma; Metalasia B: 6.4 [2.4, 12.7] Ma). The Stoebe clade resembles the first pair of clades more closely in that the radiation of its core clade is more recent, dating to 3.21 [1.5, 5.04] Ma. The reasons for the recentness of these Cape gnaphalioid radiations, which contrasts with a general trend of older radiations in other fynbos vegetation lineages (Linder, 2005(Linder, , 2008Verboom et al., 2014), remains something of a mystery, but a possible explanation is parallel radiation into more recently-exposed and environmentally diverse lowland habitats, that were to a large extent produced by late-Miocene uplift and the resulting rejuvenation of erosion in the Cape lowlands (Cowling et al., 2009).

Conclusions
The two Syncarpha clades last shared a common ancestor 13.2 [7.9, 19.5] Ma, so their evolutionary independence is well-established. Under a phylogenetic system of classification, therefore, the case for treating these Syncarpha lineages as distinct genera is persuasive. Ecological and morphological data provide further support, with Syncarpha 1 species being associated mostly with mid-to high-elevation habitats (and the S. dregeana-S. staehelina subclade is generally associated with high-elevation habitats), typically underlain by Table Mountain Group quartzites, while Syncarpha 2 species occur predominantly at low elevations, often along the southern Cape coast, where they inhabit a variety of geological substrates (e.g. limestones, coastal dunes and alluvial gravels). Also, in contrast to Syncarpha 1 which is widespread across both the western and eastern CFR, Syncarpha 2 has its diversity concentrated in the eastern half of the region. Syncarpha 1 species are characterised by plumose to barbellate pappus bristles, flat leaf margins and rounded leaf tips, whereas Syncarpha 2 species have generally scabrid pappus bristles, involute leaf margins and acute or acuminate, hooked leaf tips. Further morphological investigations are required to determine differences in characters that apparently unite the two clades (such as depressed-globose duplex achenial hairs).