Expanded generic concepts for Kermadecia and Persoonia, Proteaceae of New Caledonia and neighboring islands

ABSTRACT Proteaceae, the Macadamia nut family, are diverse in New Caledonia with c. 50 species distributed in nine genera in tribes Embothrieae, Macadamieae, Persoonieae, and Roupaleae, where they represent an important component of the rainforest and particularly the maquis. The family is also represented by a few species in the neighboring archipelagos of Vanuatu and Fiji, which marks the eastern limit of its distribution in the Pacific. Here, we address some issues regarding generic limits within Pacific Proteaceae, using new molecular data and a review of morphology. The generic limits within the tribe Macadamieae have long been problematic, particularly amongst Kermadecia, Sleumerodendron, and Turrillia, which are part of the subtribe Gevuininae. Molecular phylogenetics shows that they form a monophyletic group, while morphological data indicate that they are only weakly differentiated. We conclude that they are better treated as a single genus, Kermadecia, as has been done previously, with eight species. Kermadecia elliptica is placed in synonymy under K. rotundifolia for the first time, and several names in the genus are lectotypified. As previously observed, we found that the New Caledonian endemic Garnieria spathulifolia is nested in Persoonia (tribe Persoonieae) from which it differs only by the number of ovules and is therefore transferred to it and a lectotype is designated.


Introduction
Proteaceae are represented in New Caledonia by a rich and unique assemblage of nine genera and c. 50 species that are all endemic to the archipelago (Jaffré et al. 2001;Morat et al. 2012). Until recently, few alterations to the list of species had been made since their last complete revision published as volume two of the series Flore de la Nouvelle-Calédonie (Virot 1968). Subsequent taxonomic work included the transfer of the endemic species of Macadamia F.Muell. to the new endemic genus Virotia L.A.S. Johnson & B.G.Briggs (Johnson and Briggs 1975;Mast et al. 2008). Work has recently been initiated to refresh the taxonomy of the family to prepare its evaluation for the IUCN red list. This work has included the description of a new species in Kermadecia Brongn. & Gris (Hopkins and Pillon 2019), another in Virotia (Hopkins and Pillon 2020), and a reappraisal of the New Caledonian Grevillea R.Br., with 10 species recognised instead of three (Majourau and Pillon 2020). In addition, two New Caledonian species of Knightia R.Br. have been transferred to a newly reinstated endemic genus, Eucarpha (R.Br.) Spach (Mabberley and Moore 2022). In Fiji and Vanuatu, the family is represented by two species in each archipelago, which have had varying generic placements, sometimes associated with New Caledonian taxa.
The three south-western Pacific genera, Kermadecia (five spp., New Caledonia), Sleumerodendron Virot (one sp., New Caledonia), and Turrillia A.C.Sm. (three spp., Vanuatu and Fiji) are closely related members of the Proteaceae, tribe Macadamieae, subtribe Gevuininae (Weston and Barker 2006;Weston 2007). The Gevuininae contains eight genera: Bleasdalea F.Muell. ex Domin, Cardwellia F.Muell, Euplassa Salisb., Gevuina Molina, Hicksbeachia F.Muell, as well as Kermadecia, Sleumerodendron and Turrillia and according to Weston and Barker (2006) and Sauquet et al. (2009) it is strongly supported as monophyletic. It can be defined within the Macadamieae by the following combination of characters: Plants bisexual with adult leaves alternate. Flower pairs are subtended by a scalelike bract, a common peduncle of each flower-pair present or absent, and flower pedicellate or sessile, floral bract absent. Perianth zygomorphic (actinomorphic in Hicksbeachia), staminal filaments adnate to tepals, hypogynous disc variable. Fruit is drupaceous with a hard inner mesocarp and a succulent outer mesocarp with radially oriented fibres (dry, follicular in Cardwellia). Seed are globose to compressedellipsoid, not winged (flat, elliptical, winged in Cardwellia) (Weston and Barker 2006;Weston 2007).
In the past, the limits of some genera in the Gevuininae have been quite fluid; for example, the Australian species now known as Bleasdalea bleasdalei (F.Muell.) A.C.Sm. & J.E.Haas has been placed in no fewer than six other genera at various times, including both Kermadecia and Turrillia. Because of their similarities, Kermadecia, Sleumerodendron, and Turrillia have quite a complex and intertwined taxonomic history, which is outlined below and summarised in Table 1, with a synopsis of names, authors, types, and synonyms given in the final section. The main part of this paper presents novel molecular data and a review of morphology based on both our own observations and the existing literature for these three genera in order to determine whether the current taxonomy is satisfactory and to clarify generic limits. In the results and discussion, Kermadecia s.s. refers to the five species regarded as belonging to this genus at the start of our study, i.e.  (Turrill) A.C.Sm. (Fiji)). Occasionally the term Kermadecia p.p. is used to refer to Kermadecia s. s. minus K. pronyensis.
Another debate regarding generic concepts in Pacific Proteaceae involves Garnieria Brongn. & Gris in the tribe Persoonieae and whose taxonomic history is also outlined below. According to Weston and Barker (2006) and Weston (2007), members of Persoonieae are characterised as follows: Bisexual trees or shrubs. The cotyledons are elliptic to linear, sessile, semi-circular to triangular in cross-section. Leaves entire. Inflorescence a raceme or spike. Stamens are usually monomorphic. Carpel shortly stipitate, ovules 1-7. Fruit a drupe, the stony endocarp penetrating between the seeds. Seeds are ovoid, not winged. A second aim of the current paper is to present molecular data for Persoonia and the related genera Garnieria, Acidonia, and Toronia, with a much shorter discussion of morphology (see Discussion), leading to the formal synonymising of Garnieria under Persoonia in the Taxonomic section.

Taxonomic history of Kermadecia, Turrillia, and Sleumerodendron
Kermadecia was established by Brongniart and Gris (1863) to accommodate their three new species, K. elliptica, K. rotundifolia, and K. sinuata. In the same publication, they also described Adenostephanus austrocaledonicus Brongn. & Gris, which was subsequently transferred to Kermadecia by Jackson (1894), based on Bentham andHooker (1880, see Turner 2016), a move supported by Guillaumin (1935), though he noted that the flowers were still unknown, implying some uncertainty about the generic placement. One reason why Brongniart and Gris (1863) considered that the species they described in Adenostephanus and their new genus Kermadecia did not fit into genera already known from New Caledonia was the form of the ovules: in each ovary two orthotropous (rather than anatropous) ovules were pendulous from the summit of the locule, with the micropyle oriented downward.
Kermadecia pronyensis Guillaumin was described in Guillaumin (1922). Initially, he placed it in Grevillea, transferring it to Kermadecia when he recognised that the ovules were pendulous rather than attached to the side of the ovary wall (Guillaumin 1935). In his synoptic flora for New Caledonia, he recognised seven species within Kermadecia (Guillaumin 1948), including the four listed above, K. austrocaledonica (Brongn. & Gris) Benth. & Hook.f. ex B.D.Jacks. and two taxa (K. leptophylla Guillaumin and K. neurophylla Guillaumin) that Virot (1968) subsequently placed in Macadamia and that are now in Virotia (Johnson and Briggs 1975;Mast et al. 2008).
In his treatment for New Caledonia, Virot (1968) regarded Kermadecia as consisting of seven species: four from that archipelago (K. elliptica, K. pronyensis, K. rotundifolia, K. sinuata) plus the three from Vanuatu and Fiji. These three were originally described in Kermadecia, as K. vitiensis (Turrill 1915), K. lutea (Guillaumin 1932) and K. ferruginea (Smith 1936), before being transferred first to Bleasdalea by Smith and Haas (1975) and then to Turrillia (Smith 1985). Virot (1968) also established the monotypic genus Sleumerodendron to accommodate the plant known previously as either Adenostephanus austrocaledonicus or K. austrocaledonica. He remarked on the similarities of S. austrocaledonicum to Euplassa and Gevuina but did not discuss differences from Kermadecia except in the keys to the genera. Adenostephanus Klotzsch is now a synonym of the South American genus Euplassa (Plana and Prance 2004).
As mentioned above, Smith and Haas (1975) took the three species from Vanuatu and Fiji out of Kermadecia and placed them in Bleasdalea, alongside B. papuana (Diels) Domin from New Guinea, and B. bleasdalei from north-east Australia. Kermadecia now contains only four species, all from New Caledonia, and Sleumerodendron continues to comprise just one.
In his account for Fiji, Smith (1985) established the genus Turrillia to replace the name Bleasdalea, which he considered was invalidly published. He transferred Table 1. Generic placement of epithets associated with the name Kermadecia according to various authors. 1 Some names are placed in these genera by implication. 2 This species was originally described as Grevillea pronyensis Guillaumin (Guillaumin 1922). 3 This epithet has also been placed in Adenostephanus, Euplassa,Grevillea,Gevuina,Kermadecia,and Roupala Aubl. by various authors. 4 This epithet has also been placed in Euplassa and Gevuina by some authors (IPNI 2022) Guillaumin (1935Guillaumin ( , 1948 Virot (1968) Smith and Haas (1975) Smith (1985) Weston and Barker (2006); Weston (2007)  the five species that he had previously placed in Bleasdalea to Turrillia but was aware of the lack of clarity at the generic level in this part of the family and discussed several scenarios, including whether the taxa from Australia and New Guinea (T. bleasdalei (F. Muell.) A.C.Sm. and T. papuana (Diels) A.C.Sm.) might belong to the otherwise South American genus Gevuina. In their new classification for the family, Weston and Barker (2006) and Weston (2007)
The molecular phylogenetic analysis by Sauquet et al. (2009) recovered a monophyletic group in which Placospermum C.T.White & W.D.Francis (one sp., Australia), in subfamily Persoonioideae, tribe Placospermeae (Weston 2007) is sister to Toronia (Persoonia (Acidonia + Garnieria)) (=tribe Persoonieae). However, in a well-sampled molecular phylogenetic analysis of Persoonia (93 out of 101 species), Holmes et al. (2018) found that it was paraphyletic, with Toronia, as well as Acidonia and Garnieria nested within it. They nevertheless refrained from subsuming these three monotypic genera into Persoonia, although acknowledging Persoonia would be a "highly corroborated, easily identified clade" if these three were included.
For Garnieria spathulifolia, we amplified two molecular markers: the nuclear DNA internal transcribed spacer (ITS) region including ITS-1, 5.8S and ITS2 and the chloroplast trnL-F region including the trnL intron, trnL (UAA) 3' exon and the trnL-trnF intergenic spacer using the primers P1L/ P2R designed by Barker et al. (2002) and the primers c/d and e/f designed by Taberlet et al. (1991) respectively.
The PCR amplification program consisted of 5 min of initial denaturation at 80°C, followed by 30 cycles of 1 min denaturation at 95°C, 1 min of annealing at 50°C, 1 min 30 elongation at 65°C; and a final hold of 4 min at 65°C. We separated the waxy PCR product by electrophoresis on 2% (w/v) agarose gel and excised the resulting band. We cleaned it up with QIAquick ®Gel Extraction Kit (Qiagen) and we cloned the purified fragment in pGEM®-T Easy vector (Promega Corp.) according to the manufacturer's instructions.
Purification and bidirectional sequencing were carried out by an external service (Genoscreen, Lille, France) on an automated sequencer using BigDye Terminator V3.1 reagents. In ITS PCR product of Garnieria, we coded as per the IUPAC nucleotide ambiguity scheme (Cornish-Bowden 1985) for the sites that had double peaks. The sequences were edited in MEGA X (Kumar et al. 2018), aligned using MUSCLE (Edgar 2004) and cleaned from problematic alignment blocks using Gblocks 0.91 (Castresana 2000) using the less stringent options set.
A list of accessions with voucher information and GenBank accession numbers for the Gevuininae dataset is provided in supplemental data online. A collection (Weston 1639), with a sequence previously published under the name K. elliptica, is here re-identified as K. cf. brinoniae. These two species were previously confused and K. brinoniae was described several years after the Weston collection was first identified (Hopkins and Pillon 2019). The four sheets of Weston 1639 from Mont Koghis in NSW are all juvenile foliage and do not allow a firm identification at species-level, but K. brinoniae is the only species of Kermadecia s.s. that has previously been collected in this area. Our sampling of Gevuininae contains all the species we currently recognise (see taxonomic section) in New Caledonia and Vanuatu, but neither of the two Fijian species.

Phylogenetic analysis
We conducted phylogenetic analyses using Bayesian inference. The best substitution models for each DNA region were determined using jModelTest version 2.1.6 (Darriba et al. 2012) setting Akaike's information criterion (AIC). Then, we combined these models in Mr Bayes v. 3.2.7 (Ronquist and Huelsenbeck 2003) linking the branch lengths but not the sampling of character-state frequencies, substitution rates, the gamma shape parameter, and the proportion of invariant sites.
We performed a Markov chain Monte Carlo (MCMC) analysis for 5 million generations with eight chains, sampling every 1000th generation. We assessed the convergence of runs by examining the ESS of parameters under Tracer v.1.7 (Rambaut et al. 2018), using the recommended threshold of 200 (Drummond et al. 2006). We generated majority-rule consensus trees discarding a burn-in fraction of 25%.

Morphological observations
All material of Kermadecia s.l. at P (Le Bras et al. 2017) and K was seen by us plus some at NOU (Bruy et al. 2022) and BM. Images of types in other herbaria were studied using JSTOR Global Plants 2022 (continuously updated) and databases for some individual herbaria, located using Index Herbariorum (Thiers et al. continuously updated). In the morphological section, characters for K. elliptica are included in K. rotundifolia and the reasons for this are discussed in the Taxonomic section. In the synopsis of names, authors, synonyms, and types for Kermadecia s.s., Sleumerodendron and Turrillia, publication details for species names were taken initially from the International Plant Names Index (IPNI 2022) and the Australian Plant Names Index (APNI 2022) and verified by us.

Typification
Several names are lectotypified below, in accordance with the International Code of Nomenclature (article 9, Turland et al. 2018) if the original author(s) of a name did not specify a holotype in the protologue. In these cases, the details of the type as cited in the protologue are quoted although this is not done for names already lectotypified or with a holotype. In three cases where the original author mentioned a single collection in the protologue, or mentioned a single gathering as the "typus," Smith and Haas (1975) stated that the holotype of the name was at a particular herbarium and referred to the duplicates at other herbaria as isotypes. Under ICN Art. 9.10, on the correction of a misused term, we have changed the status of their "holotypes" to lectotypes. In one instance, a second step of lectotypification is required because the material designated as the holotype by Smith and Haas consists of two sheets.
Five of the names discussed in the taxonomic section were published by Brongniart and Gris based on material collected by Eugène Vieillard in New Caledonia between 1855 and 1860. The numbering and labelling of Vieillard's collections are notoriously complex as each of his numbers generally refers to what he thought was a species rather than to an individual gathering, and so material he collected from different localities and on different dates can all have the same number. Furthermore, when duplicates were distributed, material from different gatherings may have sometimes become muddled and handwritten data were not always copied accurately or completely. This applies particularly to Vieillard's later collections (1861-1867) rather than to those dated 1855-1860. While it is usually clear which material at P was used by Brongniart and Gris for their descriptions of the species discussed here, comparison between specimens housed in different herbaria is not straightforward and many sheets outside P that were labelled as isotypes during the 20 th century are unlikely to be part of the original material. Barcodes are given here for sheets that are known or thought likely to be part of the type and most but not all remaining sheets with the same number are mentioned by herbarium abbreviation only, although a few may eventually prove to be isolectotypes.

Molecular phylogenetics
No major conflicts were observed between tree topologies (i.e. no strongly supported conflicting nodes) recovered in individual gene analyses. We obtained a combined matrix of 14 terminals and 9,942 sites for the Gevuininae (with 18% of gaps) and a combined matrix of 85 terminals and 1,405 sites for the Persoonieae. In our Bayesian phylogenetic analyses, convergence of runs is obtained with all parameters ESS values ≥200. In the majority-rule consensus tree of Gevuininae (Figure 1) inferred from atpB, ITS, matK, ndhF, PHYA, rbcL and waxy, Kermadecia s.s. is a strongly supported (PP = 1) monophyletic clade, with Turrillia, Sleumerodendron and Euplassa each recovered as successive sister taxon with strong support.
In the majority-rule consensus tree of Persoonieae ( Figure 2) inferred from DNA ITS and trnL-F sequences, this tribe is recovered as a strongly supported monophyletic clade (PP = 1). The genus Persoonia is paraphyletic with the monotypic Acidonia, Garnieria, and Toronia nested in it. Garnieria is unresolved at a basal node along with Persoonia species (PP = 0.86).

Morphology in Kermadecia, Turrillia, and Sleumerodendron
The results of the literature review and our own observations are summarised in Table 2, and a selection of specimens that demonstrate these characters given in Appendix 2. Some of the characters described below were illustrated in Smith and Haas (1975, Figs. 1-12 & 36-51) and Figure 3.

Leaves and indumentum.
Leaves of mature plants are simple in all species except T. vitiensis and sometimes T. lutea, in which they are imparipinnate. In Kermadecia s.s. and T. ferruginea, the blades are relatively broad and the margins are not or rarely toothed (T. vitiensis) though frequently slightly sinuated, misshapen, or damaged. The leaves of Sleumerodendron have comparatively narrow blades and entire margins, and those of T. lutea are narrowly ovated and entirely or rarely coarsely toothed. The juvenile leaves of Sleumerodendron and Kermadecia are apparently difficult to distinguish (Virot 1968).
Ferruginous indumentum is characteristic in several species, and white or grey indumentum in others.
The trichomes vary from minute, curled structures to longer, straight or curved ones. Indumentum is commonly but not universally present on the leaves and young vegetative axes, on the inflorescence axes and outer surface of the perianth, and on the ovary.
2. Inflorescences and flowers. Inflorescences are raminascent (i.e. borne on the branches and proximal to the leaves), axillary or in pseudo-terminal groups. Although many appear axillary when mature, they probably, at least sometimes, originate from supra-axillary buds. In Turrillia ferruginea, occasionally several axes arise in series in a single leaf axil. When the inflorescence is a pseudo-terminal group, a number of lateral axes are borne towards the end of a shoot, distal to the leaves, and  the shoot's apical bud is dormant; each axis has a scar at the base from a fugacious bract, except for the most proximal axes that may be in leaf axils. In all species, the flowers are arranged in pairs (or very exceptionally singly) along almost the full length of the main inflorescence axis; flower-pairs are organised irregularly or are sometimes opposite. Each main axis is usually unbranched except for the flower-pairs, and racemose (most species) or forming a panicle of racemes in K. rotundifolia and occasionally in K. sinuata and T. lutea. Weston (2007) used the term conflorescence (i.e. a compound inflorescence consisting of two or more unit inflorescences, in which the main axis does not end in a flower but the axes of the branches do so, McCusker 2007). In Table 2, the main axis of each inflorescence is classified as either robust or slender according to species.
Floral characters were outlined in the keys and descriptions of Virot (1968), Haas (1975, 1985) and Weston (2007). In the species of Kermadecia s.s. except K. pronyensis, (i.e. Kermadecia p.p.) each flower is borne on a comparatively long, slender, free pedicel (rarely a few flowers are single or the pedicels are fused in pairs). These pedicels are arranged in collateral pairs (Figure 3a), each subtended by a small free bract, or the bracts are fused into a single structure that may have a bifurcate tip. The apex of each pedicel is unequal so the base of the flower is oblique. The hypogynous disc at the base of the ovary is semi-circular to crescent-shaped, well developed on the anterior side where the pedicel is shorter and absent on the posterior side (Figure 3b d). All four tepals are either curved back or rolled up at, or soon after, anthesis. In K. sinuata, the stiffness of the perianth appears to prevent the tepals from rolling up.
In Turrillia, Sleumerodendron and K. pronynensis, the flowers are sessile in pairs on a comparatively short, broad peduncle (especially short in K. pronyensis) (Figure 3g, l, q), which is sometimes grooved down the centre (very rarely a few flowers per inflorescence are borne on free pedicels). Each peduncle has a small, fugacious bract or its scar at the base. The apex of the peduncle is acute, with one flower inserted on either side, attached obliquely to its asymmetric base. Plotting whether the flowers are sessile or pedicellate onto the phylogeny for the clade formed by Kermadecia s.s., Sleumerodendron, Turrillia, and Euplassa, suggests that sessile flowers on short peduncles are plesiomorphic, with the longer, free pedicels in Kermadecia p.p. derived from this arrangement. In Turrillia and K. pronyensis, the disc is also a shieldlike or crescent-shaped structure on the anterior side of the flower, extending ½ (¾) way round the base of the ovary (Figure 3h -j, m -o). In contrast, in Sleumerodendron, the disc consists of 4 minutes, free lobes ( Figure 3r). Rao (1967) described the disc of "Kermadecia sp." as having the two anterior (sic) lobes suppressed and the two posterior ones connate to form a two-lobed, crescent-like gland, with no vestigial remnant of the suppressed parts. However, the anterior and posterior are reversed in his description compared with their use here. According to Smith and Haas (1975), the disc of Kermadecia s.s. is formed of three fused glands and that of Turrillia (as Bleasdalea) by only two, although they included additional taxa in Bleasdalea. At × 10, the discs of Kermadecia s.s. and Turrillia appear quite similar in our observations. The number of lobes forming the disc in Kermadecia p.p. needs confirmation from developmental studies.
In Turrillia, Sleumerodendron, and K. pronyensis, two adjacent tepals, one from each flower of a pair (the dorsal perianth segment according to Smith and Haas 1975), have a tendency to remain erect at or immediately post anthesis, while the remaining three tepals per flower are recurved or rolled back, though this is more apparent in some specimens than others.
In all three genera the ovary is relatively narrow; according to Virot (1968), it is quadrangular-trapezoid in section in Kermadecia p.p. but not angular in K. pronyensis and Sleumerodendron. In all species, it narrows gradually but only slightly into a long style, which is either straight and erect or curved away from the midline of the flower-pair. The style is stiff and straight in K. sinuata, which has the largest flowers, and most curved in Sleumerodendron, in which the flowers are relatively small and delicate, but whereas some variation is species specific, some may be due to the age of the flower or the speed of drying. Smith and Haas (1975) reported the orientation of the ovary as diagonal in Kermadecia s.s. and antero-posterior in Sleumerodendron and Turrillia (as Bleasdalea), but this is difficult to confirm and again anatomical studies are required.
The distal 1-2 mm of the style forms a pollen presenter that is usually papillose, often expanded and either barrel-shaped or slightly splayed on the downward-facing surface (not expanded in K. sinuata and scarcely expanded in T. vitiensis, though this may depend on the floral stage). According to Ladd (1994), the pollen presenter is of the same type in Turrillia and Kermadecia s.s. In T. lutea at least, the stigma appears to be a minute pore at the very tip of the style though some species have lateral or ventral stigmas (Weston 2007). The structure and functioning of the pollen presenters and the position of the stigmas in the related genera Gevuina and Euplassa were described in Prance et al. (2007). Flowers of several species are brown or brownish on the outer surface, from the dense indumentum, and commonly yellowish, cream-coloured or sometimes pale green on the inner surface, with the style often green. Turrillia lutea and T. vitiensis both have bright yellow flowers. S. austrocaledonicum is the only species in this group in which the tepals are dark purple or violet on the outer surface, with yellow-green tips and purple on the inner surface, though again style is green; this colouration suggests a different pollinator from that of Kermadecia s.s. In Fiji, T. lutea is reported to have been visited by birds (field notes of Wheatley 767).
3. Fruits. In all three genera, the fruits are woody, indehiscent and drupaceous, and scarcely to moderately laterally compressed. Weston and Crisp (1996) suggested they were likely to be dispersed by bats. Although the ovaries of several species have dense indumentum, mature fruits are glabrous. Probably, all fruits contain a single seed: see Virot (1968) for Kermadecia s.s., Smith and Haas (1975) for Turrillia (as Bleasdalea). Small differences in fruit shape and degree of fissuring of the fruit-wall occur. The species of Kermadecia p.p. have similar fruits that are inequilateral in lateral view, somewhat ellipsoidal, obovoid, or gibbous, with the dorsal side more developed than the ventral one (Figure 3e -f). The dorsal margin is either slightly angled distally or rounded (K. sinuata), and the base is cuneate with the attachment scar central or somewhat lateral. Towards the apex, the fruit is blunt or obtuse and the style base forms a small mucro positioned slightly towards the ventral margin. In ventral view, the fruits are elliptic and in crosssection, somewhat trullate and slightly angled, especially along the dorsal line (see Fig . 4 in Smith and Haas (1975), K. brinoniae, as K. elliptica, based on MacKee 16,339). In dry fruits of this species, the fruitwall has a tendency to be corrugated and occasional fissures can develop; in the other species of Kermadecia p.p., the fruit-wall is almost smooth when mature.
In Sleumerodendron, the fruits are sub-circular lateral view, slightly inequilateral, and better developed on the dorsal side (Figure 3 s -t). Young fruits can be ± globose, but most mature fruits are slightly flattened laterally so that in ventral view they are elliptic to ± broadly elliptic, and in cross-section, oblong-elliptic to almost circular and not angled along the dorsal line. The basal attachment scar is at the lowest point and slightly asymmetrically placed. The apex is ± round with the mucro small or absent. In dry fruits, the fruitwall has numerous deep fissures, revealing that the outer layer of the mesocarp is composed of short fibres perpendicular to the surface (Smith and Haas 1975, fig. 8).
The fruits of Turrillia are more variable. In lateral view, they are ± rhomboidal, almost circular but slightly gibbous, obovate/obtrullate, or turbinate (see illustration for T. vitiensis in Gillespie 1932) and often inequilateral (Figure 3p). Distally, many have an umbo positioned either centrally or towards the dorsal margin, with the style-base at its apex. The base is cuneate to broadly cuneate and the attachment scar ± oblique or not. In ventral view, they are ovate, broadly ovate, or obovate, and in transverse section elliptic or broadly elliptic, and not especially angled along the dorsal line. In dry fruit, the fruit-wall is stony-woody and its surface is ± smooth, usually without fissures, though occasionally old fruits have a few cracks that may be the result of the drying process (e.g. perhaps in F.D. 1031, T. vitiensis). Exceptionally, the fruit-wall starts to develop cracks that resemble those of Sleumerodendron, as in Raynal RSNH 16,338 (T. lutea).

Kermadecia, Turrillia, and Sleumerodendronhow many genera should be recognised?
Comparisons of Kermadecia s.s., Sleumerodendron and Turrillia (sometimes as Bleasdalea) based on the generic descriptions of Virot (1968), Smith and Haas (1975) and Weston (2007), on characters from the keys in these publications, and on our observations show that Kermadecia s.s. overlaps with the other two in significant characters, especially of the flowers and their arrangement. Because of the heterogeneity in floral arrangements in Kermadecia s.s., the genus keys out twice in Weston (2007)  flowers in pairs on short peduncles places K. pronyensis with Turrillia and Sleumerodendron. Although little fruiting material of K. pronyensis is available (none at K and only three sheets at P), as noted above, the fruits of this species are markedly similar to those of T. ferruginea and differ somewhat from those of the remaining species of Kermadecia s.s. If K. pronyensis remains in Kermadecia s.s., then Turrillia must also be included because no characters would strongly differentiate between them (Tables 2  & 3). The phylogeny also supports this conclusion because K. pronyensis is sister to the other species of Kermadecia s.s.
The position of Sleumerodendron, as sister to the subclade formed by Kermadecia s.s. + Turrillia lutea (Figure 1), is somewhat different because it has several distinctive characters, including its floral disc (four small, free lobes), the deep fissuring of the fruit wall, and its purple, more delicate flowers. A case can be made either for uniting it with Kermadecia s.s. + Turrillia, which would make the resulting genus more heteromorphic, or for maintaining it as a monotypic genus. One factor to consider in deciding the fate of Sleumerodendron must be how narrowly or broadly other genera in the Gevuininae are circumscribed and whether Kermadecia s.l., including both Turrillia and Sleumerodendron, would be markedly more heteromorphic than the remaining genera in the subtribe. A second factor is what morphological characters would define an expanded Kermadecia s.s. + Turrillia if Sleumerodendron was also included.
The two Australian endemics, Cardwellia and Hicksbeachia, are both narrowly defined and distinctive. The first differs from all other members of the subtribe by its numerous ovules, winged seeds, and follicular fruits. Members of the second are small unbranched trees or multi-stemmed from the ground, with actinomorphic flowers and inflorescences borne on the stems proximal to the leaves (caulinascent). In Figure 1, Hicksbeachia pinnatifolia F.Muell. is sister to Bleasdalea bleasdalei, also from Australia, and the two differ in floral symmetry and the structure of the floral disc. The two South American genera, Gevuina (imparipinnate leaves with relatively small, toothed leaflets) and Euplassa (paripinnate leaves with toothed or entire margins), also differ in their floral discs and leaf rachises (see key in Prance et al. 2007). With 20 species, Euplassa shows a wider range of variations than any other genus, notably in the form of the floral disc (see Plana 2007, fig. 14).
As noted above, Sleumerodendron shares with most species in Kermadecia s.s. + Turrillia its simple leaves with entire margins and comparatively large globular fruits with fibrous fruit walls. Although its inclusion in a broader Kermadecia makes the resulting genus more heteromorphic, on balance, we consider it preferable to include it in Kermadecia s.l., and the resulting genus would not encompass a wider range of variations than does Euplassa. Kermadecia s.l. can be distinguished from other genera of the Gevuininae by the adult foliage consisting largely of simple, entire leaves (or imparipinnate and/or leaflets sparsely serrate in T. vitiensis and occasionally in T. lutea). Kermadecia is the oldest of the three generic names and all the necessary combinations within it already exist.

The circumscription of Persoonia
All combinations of Persoonieae species already exist in the genus Persoonia, except for Garnieria spathulifolia. This genus has never been critically compared to Persoonia, and the only character that appears to distinguish them is the number of ovules (see key in Weston 2007). Virot (1968) wrote that Garnieria has 3-7 ovules and up to seven seeds per fruit, and Weston (1995c) indicated two ovules in Acidonia, and 1-2 in Persoonia. Otherwise, G. spathulifolia fits within the variation described for the genus Persoonia in Australia (Weston 1995c). A broader concept of Persoonia, which would be strongly supported as monophyletic ( Figure 2) and that equates to the tribe Persoonieae seems preferable. A new combination in Persoonia is therefore provided below for the species currently in Garnieria.  Table 3. Comparison of genera in the subtribe Gevuininae (Proteaceae). Data from pers. obs. at K and P plus Hyland (1995 -Cardwellia), Pennington (2007 -Gevuina), Plana (2007) and Plana & Prance (2004 -Euplassa), Smith (1985 -Turrillia), Smith & Haas (1975 -Bleasdalea, Turrillia [as Bleasdalea]), Virot (1968 -Kermadecia s.s., Sleumerodendron), (Weston 1995b -Bleasdalea [as Gevuina]), Weston (1995b -Hicksbeachia), Weston (2007 -all The sheet stated here to be an isolectotype at P (P00645323!) resembles the lectotype in both its leaves and phenological state (flowers). Sheets at B (B10 0295559!, Vieillard s.n., 1861-1867) and K (K00073959!, Vieillard 1109, s.loc., s.dat.) both have mature fruits and lack flowers. They are potentially from the same gathering as each other, but they do not match the protologue and so are excluded from being isolectotypes. The only character that distinguishes between K. elliptica and K. rotundifolia is their inflorescence structure (a simple raceme of flower-pairs vs. a trident of flower-pairs). However, field observations and photographs show that individuals of K. rotundifolia can bear both unbranched and branched inflorescences. Branched inflorescences can also be found occasionally in K. sinuata. Branching of the main inflorescence axis is therefore not a reliable character on its own to distinguish between Kermadecia species. Furthermore, we were unable to find any characters of the foliage that differ consistently between K. elliptica and K. rotundifolia and so we put the former into the synonymy of the latter here.
It is unlikely that all of the specimens we have seen with the number Vieillard 1105 belong to the same gathering. The lectotype, with the locality Balade and the date 1855-1860, would have been part of the material available to Brongniart and Gris in their publication of 1863 and two of the sheets have Brongniart's writing on them. Amongst the material of Vieillard 1105 seen at K and online (via JSTOR Global Plants), sheets A00035492 and HBG-508308 have the same label data as the lectotype and resemble it morphologically; K000736970 and K000736971 lack original label data but may also be part of this gathering. However, the localities given on sheets in various other herbaria include "prope Wagap" or "Wagap -Balade," or they have no locality, and the dates given are 1864 or 1861-1865, or they have no date. While some that lack a locality or date may prove to be isolectotypes, most of these specimens, at B, E, K, G, GH, KFTA, L, MEL, MPU, and NY, are unlikely to be so, although close comparison with the sheets at P is required to be certain.

Kermadecia sinuata
Three sheets at P have labels with the locality of Diaoué and are dated 1855-1860 (lectotype and isolectotypes). Most of the sheets labelled Vieillard 1103 at A, B, HBG, K, L, MEL, MPU, and NY have the date 1861-1867 and the locality Wagap; a few lack a locality, or a date, or both; it is doubtful that any of them are part of the original material. Turrill, Hooker's Icon. Pl. 31: t. 3022. 1915;J. Linn. Soc., Bot. 43: 36. 1915 Lectotype (first step lectotypification designated by Smith and Haas (1975, p. 146