Three new tribes in Myrtaceae and reassessment of Kanieae

Keywords: Cloezia, Kania, molecular phylogenetics, Myrtaceae, taxonomy, tribes, Tristaniopsis, Xanthomyrtus.

Kanieae Engler was named as a monogeneric tribe erected to accommodate the genus Kania Schltr., which had been published earlier by Schlechter (1914). When Schlechter described the genus, he was uncertain of its affinities and, after considering placement in Clusiaceae, Myrtaceae and Saxifragaceae, finally described it as an aberrant genus in Saxifragaceae. Engler’s placement of Kanieae within its own subfamily, Kanioideae, within Saxifragaceae, was similarly a reflection of its anomalous position in that family. Morphological investigations (Erdtman and Metcalfe 1963; Weberling 1966) strongly suggested that Kania had Myrtaceous affinities. Van Steenis (1969) noted that vegetative characters, leaf venation type, presence of an intramarginal vein and presence of oil glands clearly indicated that the genus was a member of the family Myrtaceae. However, he took a conservative view of the generic position of Schlechter’s Kania eugenioides Schltr. and transferred the species to the genus Metrosideros Banks ex Gaertn., listing a further six names as synonyms. Wilson (1982) accepted Kania as a genus distinct from Metrosideros and made new combinations for two Philippine species that had originally been described in the genera Cloezia Brongn. & Gris and Tristania R.Br., and Scott (1983) increased the number of accepted species when he published a further two new species from West Papua. Subsequently, Scott (1990) transferred a further West Papuan species from Myrtella F.Muell. to Kania, making a total of six named taxa. However, it is likely that several of the synonyms listed by van Steenis should also be recognised as distinct species (G. P. Guymer, pers. comm.) to bring the total to ~10.

When van Steenis (1969) reduced Kania to synonymy under Metrosideros, he downplayed the value of two distinctive floral features, namely, the elongated anther connectives and the placentas in the basal angles of the loculi, remote from the base of the style. Regarding the latter, van Steenis noted the similarity of this arrangement to that found in the Australian monotypic genus Lysicarpus F.Muell. A third taxon, the New Caledonian genus Cloezia, is also known to have basal placentas remote from the base of the terminal style. The significance of this morphological arrangement in all three genera was first pointed out by Dawson (1972a) in his assessment of Cloezia (as Mooria Montrouz.) in relation to Metrosideros. However, he concluded that Cloezia and Metrosideros were not closely related because the placentas in Metrosideros and allied genera are always adjacent to the style base, even in the South American genus, Tepualia Griseb., which has a basal placenta (see, for example, the description and illustrations in Dawson 1972b). On the basis of the similarity in placentation, Dawson (1972a) suggested that the affinities of Cloezia might lie with Lysicarpus and Kania. Briggs and Johnson (1979) adopted this view and included these three genera in their informal ‘Kania alliance’. However, as noted by Wilson (2011), the ovules of Kania species are scattered on the placentas, whereas those of Lysicarpus and Cloezia are arranged in a more-or-less circular series.

Tristanieae Peter G.Wilson (Wilson et al. 2005) originally comprised three genera, namely, Tristania, Thaleropia Peter G.Wilson and Xanthomyrtus Diels, although the last of these showed significant variation in fruit and seed characters and molecular support for its inclusion in the tribe was modest. Recent phylogenetic analyses by Biffin et al. (2010), Thornhill et al. (2015), and Maurin et al. (2021) recovered a clade that includes these genera but also includes Cloezia. The analysis of Wilson et al. (2005) had not confidently placed Cloezia and they considered it to be incertae sedis. Wilson (2011), on the basis of the ovule arrangement, tentatively included Cloezia in Kanieae sens. lat.

Early evidence from pollen (Pike 1956) found that the pollen of Metrosideros parviflora C.T.White (a synonym of Kania eugenioides sens. lat.) did not conform to that of other Metrosideros species. Pike summarised the differences in pollen morphology as follows: ‘the grains are smaller and the colpi are absent on the polar surfaces’ (p. 40). Erdtman (in Erdtman and Metcalfe 1963) examined pollen from a type specimen of Kania eugenioides and found much the same, but recorded the colpi as ‘narrow, tenuimarginate, about 2.5–3 µm long, with tapering ends’ (p. 249). Gadek and Martin (1981) examined Kania pollen by light microscopy only; they confirmed the differences between Kania and Metrosideros but did not detect colpi. Thornhill et al. (2012a, 2012b) also confirmed the size difference between the genera but the improved resolution of the scanning electron microscope showed more detail such that Kania pollen could now be described as ‘parasyncolpate with arcuate colpi’ (Thornhill et al. 2012a, p. 262); however, in general form, Kania pollen was not dissimilar to pollen of species of Lysicarpus and Tristaniopsis Brongn. & Gris, differing only by being obscurely parasyncolpate with a less ornamented exine. In contrast with this, Thornhill et al. (2012b), in agreement with all previous workers (Pike 1956; McIntyre 1963; Gadek and Martin 1981), noted that all Metrosideros species examined had much larger pollen (~11–17 µm long and wide, compared with ~7 × 11 µm), and almost all taxa they examined had well developed apocolpial islands.

Pollen morphology of Cloezia is neutral on the question of relationships. Thornhill et al. (2012a) found little difference in pollen morphology between Cloezia and the genera of Kanieae sens. lat. and that there was little difference in pollen morphology between Cloezia and Xanthomyrtus, because both have parasyncolpate pollen that is similar in size and exine pattern. In strong contrast with these taxa, the genera of core Tristanieae (Tristania and Thaleropia) share highly derived pollen that is the smallest found in the family so far (~7 μm in diameter) and is triporate and acolpate with a psilate exine (Pike 1956; Gadek and Martin 1981; Patel et al. 1984 for Tristania; Thornhill et al. 2012a for both genera).

Comparative wood anatomy has provided some insights into relationships of these taxa. An apparently informative feature of wood anatomy in some tribes of Myrtaceae is the presence of elongated vessel-ray pitting, which Ingle and Dadswell (1947) found could be used to distinguish Syzygium P.Browne ex Gaertn. and its allies (tribe Syzygieae) from Eugenia L. sens. strict. (tribe Myrteae), confirming that these taxa were not congeneric. Metcalfe (in Erdtman and Metcalfe 1963) noted similar pitting in wood from a type specimen of Kania eugenioides, an observation confirmed by a more recent image in an atlas of woods (Ilic 1991). Similar vessel-ray pitting has been observed in the tribe Metrosidereae. Ingle and Dadswell (1953) described the wood of Tepualia as having vessel-ray pits that appear simple and rounded to elongated, and Meylan and Butterfield (1978), who studied the woods of three New Zealand species of Metrosideros, described the vessel-ray pits as ‘commonly axially elongated and large and form prominent cross fields’ (pp. 94, 96, 98). So, wood anatomy does show more similarity between Kania and Metrosideros than between Kania and many other genera. In contrast to this, the vessel-ray pitting in Cloezia is fine and alternate, similar to intervessel pitting (P. Gasson, pers. comm.), very like that recorded for Xanthomyrtus by Ingle and Dadswell (1953; confirmed by P. Gasson).

The phylogenetic analysis of Wilson et al. (2005), based on sequences of the plastid matK gene, was accompanied by a revised classification of Myrtaceae. In this classification, the tribe Kanieae included Kania, the type of the tribe, and seven other genera, including Barongia Peter G.Wilson & B.Hyland, Basisperma C.T.White, Lysicarpus, Mitrantia Peter G.Wilson & B.Hyland, Ristantia Peter G.Wilson & J.T.Waterh., Sphaerantia Peter G.Wilson & B.Hyland, and Tristaniopsis. The main morphological characters given for the tribe included ‘stamens frequently in bundles’ and ‘style base not adjacent to placentas’ (Wilson et al. 2005, p. 15), but these features are not unique to this tribe. Wilson (2011) tentatively included Cloezia in Kanieae but analyses by Biffin et al. (2010), Thornhill and Crisp (2012), Thornhill et al. (2015) and Maurin et al. (2021) indicated that this genus is instead weakly associated with the tribe Tristanieae.

More recently, Tarran et al. (2016) discussed myrtaceous leaf fossils from an Early Oligocene site in north-western Tasmania. These authors identified several characters on the cuticles of fossil leaves that were found in association with fossil Metrosideros fruits and were potentially of diagnostic value within Kania and associated genera. They were (1) peristomatal rings, (2) distinctive granulate-papillose cuticular texture, (3) striate water stomata and lid cells, and (4) varying degrees of stomatal clumping. The authors noted, from a comparative study of 175 species of extant taxa, that this combination of features was shared with very few of them. An earlier suggestion by Pole (1992) that these fossils might represent a species of Xanthomyrtus was rejected on the basis of differences in the nature of the stomatal clumping that he recorded, plus other cuticular characters that were not found in Xanthomyrtus but were present in Kania and, to a lesser extent, in some Metrosideros species.

Some, but not all, of these lines of evidence suggest that Kania is more closely related to Metrosideros than it is to the other genera that were grouped with it by Wilson et al. (2005) in the tribe Kanieae. Equally, these data provide no support for a possible relationship with Lysicarpus and Cloezia, as suggested by Dawson (1972a). The aim of the present paper is to establish the affinities of Cloezia and Xanthomyrtus, which have both been poorly resolved in previous studies, and to re-examine the relationships of Kania, and other genera presently assigned to the tribe Kanieae, by expanding the phylogenetic analysis of this and related tribes of capsular Myrtaceae. To this end, our primary goal was to generate new sequences of Kania to replace the very poor DNA sequence utilised by Wilson et al. (2005) and, additionally, to broaden the number of regions sequenced for each taxon. The phylogeny will be augmented with more detailed observations on epidermal and floral characters.

We compiled a 61-taxon molecular dataset including limited representation of both subfamilies and all tribes in family Myrtaceae. Where possible, we utilised existing sequences available on GenBank to augment our own data, so that for some taxa, sequences are from different accessions for some loci (all details given in Table 1). For Kania, we sampled three new accessions of K. eugenioides sens. lat., with DNA extracted from leaf or seeds. To cover the groups historically associated with Kania, we sampled eight species in six genera from the remainder of Kanieae and four species of Metrosideros sens. lat. (Metrosidereae). To represent other tribes of Myrtaceae subfamily Myrtoideae, we included one to six samples from each of the remaining tribes sensu Wilson et al. (2005), but with Tristanieae sens.strict. (Tristania + Thaleropia), we added two species of Cloezia and three of Xanthomyrtus, often considered genera of uncertain affinity, in line with the apparent phylogenetic position of these two genera in recent analyses (Biffin et al. 2010; Thornhill and Crisp 2012; Thornhill et al. 2015). We rooted the trees using Heteropyxis Harv. and Psiloxylon Thouars ex Tul. (subfamily Heteropyxidoideae Reveal) as outgroups, on the basis of previous research showing them to represent the sister lineage in the family (Wilson et al. 2001, 2005; Thornhill et al. 2015); more distant outgroups proved difficult to align at some loci. Details of all taxa included in the molecular analyses and associated GenBank numbers are provided in Table 1.

Table 1.  Taxa, vouchers and accession numbers.

New extractions of total genomic DNA were made mostly from frozen silica-dried leaf material, but some were from fresh material, and a few from leaf or seed taken from herbarium specimens. Tissue was disrupted dry with tungsten beads by using the Qiagen Tissue Lyser (Qiagen, Hilden, Germany), and extractions used the Qiagen DNeasy Plant DNA Mini kit following the manufacturer’s protocol.

Where possible, sequences were compiled for a total of six regions, including two from the nuclear-encoded internal transcribed spacer (ITS) and external transcribed spacer (ETS) regions of the rRNA gene, plus four plastid regions, including three contiguous components of the trnK intron, the matK-coding region (matK) and its 5′ and 3′ spacers (preM and postM respectively), and the psbA–trnH intergenic spacer (psbA–trnH). Details of primers used for PCR amplification and sequencing as well as details of PCR reactions were those outlined in Wilson and Heslewood (2016).

Sequence chromatograms were edited in Sequence Navigator (ver 1.0, Applied Biosystems) or GeneStudio Professional (ver. 2.2.0.0, GeneStudio, Inc., see https://genestudio.software.informer.com/) and consensus sequences generated were then aligned manually in PAUP* (ver. 4.0a build 169 for 32-bit Windows, see http://phylosolutions.com/paup-test; Swofford 2003). In aligning sequences, gaps were positioned to maximise conformity to known indel types such as simple and inverted duplications of adjacent sequences (Levinson and Gutman 1987; Golenberg et al. 1993). Overlapping indels of different lengths, and insertions of the same length but bearing different relationships to surrounding sequence, were treated as having independent origins, whereas indels of the same length and position and showing minor differences in nucleotide sequence were scored as the same state (Simmons and Ochoterena 2000). Potentially informative indels were scored as additional presence or absence characters and appended to the database. Gaps were treated as missing data in the phylogenetic analyses. Coding sequences of the matK gene were translated in MacClade (ver. 4.08a, see https://mesquiteproject.github.io/MacClade//macclade; Maddison and Maddison 2000) to check for internal stop codons.

Preliminary analyses using maximum parsimony or Bayesian inference were run using either individual loci, or the concatenated plastid or nuclear loci, each run with or without appended indels. Heuristic searches were conducted in PAUP* using tree bisection reconnection branch-swapping on best trees to recover the most-parsimonious (MP) trees. One thousand replicates of random taxon-addition searching were conducted in which multistate characters were treated as polymorphisms, so as to detect multiple islands of trees. Where preliminary analyses of single plastid loci exhausted computer memory, restricted heuristic searching was conducted, saving only 100 trees per replicate. Relative support for the clades identified by parsimony analysis was estimated using the jackknife rather than bootstrap resampling in PAUP*, following the recommendations of Simmons and Freudenstein (2011). For jackknife analyses, 10 000 replicates of faststep searching were conducted in which each replicate used random-taxon addition, no branch swapping, and the percentage of characters deleted was set at 33%. Jackknife (jk) values >50% were interpreted as weak support for clades, >75–89% as moderate support, 90–99% as strong support and 100% jackknife was considered robust. Sequence statistics for each locus are presented in Table 2.

Table 2.  Sequence statistics for molecular data.

The MP phylogenies generated were compared with those obtained using the Markov-chain Monte Carlo (MCMC) method implemented in MrBayes (ver. 3.2.7a, see https://github.com/NBISweden/MrBayes/; Ronquist et al. 2012) in the CIPRES Science Gateway (ver. 3.3, see https://www.phylo.org; Miller et al. 2010). The most appropriate nucleotide substitution models to apply in likelihood-based analyses were determined using the Akaike information criterion (AIC) in MrModelltest (ver. 2.3, J. A. Nylander, see https://github.com/nylander/MrModeltest2/releases/tag/v2.3), with data partitioned into the six regions indicated above, with each partition assigned a unique substitution model. Under the AIC, five regions fit general time-reversible likelihood (GTR) substitution models (nst = 6), with gamma distribution of rate variation among sites (GTR + Γ model; preM, matK, postM), or also with a proportion of invariant sites (GTR + Γ + I model; ITS, psbA–trnH). The ETS region fit a Hasegawa–Kishino–Yano substitution model (nst = 2, HKY + Γ + I model). Where Bayesian analyses also included indels, these were binary encoded as an extra partition, and we applied a default two-state Markov model with gamma distribution of rates and coding set to variable (because there were no invariant sites). Statefreqpr was set to fixed (empirical) for this partition to reflect only having two states.

Bayesian posterior probabilities (PP) were estimated using two independent runs of 10 million generations by using four chains with tree sampling every 1000 generations. All parameters were set to be unlinked and with rates variable between partitions, with all other priors for the analysis set flat (i.e. as Dirichlet priors). Runs were assessed as sufficient when displaying convergence of effective sample size (ESS) for all statistics in Tracer (ver. 1.7.1, see https://github.com/beast‐dev/tracer/releases/tag/v1.7.1, accessed 5 March 2020), the standard deviation of split frequencies was clearly <0.01 and the PSRF for all parameters neared 1.000. Trees generated before the four Markov chains reaching stationarity (the burn-in ~25%) were discarded. The remaining trees were used to construct a 50% majority-rule consensus tree, with nodes assigned posterior probabilities (PP) of 0.95–1.00 considered as supported.

TreeGraph 2 (ver. 2.15.0–887 β, see http://treegraph.bioinfweb.info/; Stöver and Müller 2010) was used to construct the figures of the phylogenetic trees. The PP (upper) and jk (lower) support values were imported onto the Bayesian consensus trees for each analysis and various annotations made to clades. Clades with strong support (1.00 PP, ≥90% jk) are indicated by heavier lines. Supplementary figures mapping jackknife (jk) values of >50% onto the strict consensus of the most parsimonious trees are also supplied for referencing conflicting areas.

Cuticles were mounted on glass slides for standard light microscopy (LM) or on aluminium stubs for analysis by scanning electron microscope (SEM) following the protocols described in Tarran et al. (2016).

Fruits of Kania sp. and Tristaniopsis collina Peter G.Wilson & J.T.Waterh. were cleared in a solution of 5% potassium hydroxide (KOH) over a medium heat. The fruits were left in the solution until the flesh became translucent and soft enough to be teased away if necessary. The remaining parts were thoroughly rinsed to remove any traces of the KOH, then bleached in a solution of commercial grade bleach until the vascular skeletons became white to translucent. The skeletonised fruits were then placed in a solution of 10% Safranin O, and left to stain, then the bleaching, rinsing and staining were repeated until the lignified vascular structures were darkly stained. Excess stain was rinsed off, the fruits were then stored in deionised water and photographed using a Nikon D5000 digital SLR with a macro lens over a bright light box. Full details of specimens used in these studies are given in Table 3.

Table 3.  Voucher details for specimens examined for morphological characters.

Aligned sequence lengths, variable characters, number of scored informative indels and models applied to each partition for Bayesian analyses are presented in Table 2. Although 21 taxa were missing some data (1–10 taxa lacking sequence at individual loci), our dataset was largely complete. There is some level of saturation of substitutions in the two nuclear regions in this dataset, reducing their utility at resolving deeper levels of relationships across the family, with homoplasy likely confounding the phylogenetic signal. In this family, these nuclear loci will be most useful for within-tribe analyses. Including indels, the nuclear dataset had 51% variable characters, 36% of which were informative under parsimony (compared with 35% variable characters, 20% informative under parsimony for the plastid data). Regardless of differences in the arrangements of some poorly supported branches uniting tribes in separate analyses, all analyses retrieved the same robustly supported major clades.

Inclusion of scored indels in both Bayesian and parsimony analyses resulted in improvements in branch supports. Therefore, indels were included in all analyses presented here. Mostly comprising small sections of sequence that could not be unambiguously aligned, a total of 156 bp, including a 93-bp highly variable portion of the psbA–trnH alignment, were excluded from analyses, leaving a 5008-bp alignment to be used in analyses, inclusive of 114 appended indels. Separate analyses of plastid (3470 bp including 41 indels) and nuclear (1538 bp including 73 indels) data retrieved clades corresponding to most currently recognised tribes, with the major difference being in the composition of the Kanieae clade of Wilson et al. (2005), but there were differences in supported relationships within and among some tribes in these analyses, and between the two types of analysis. For this reason, we have not combined the datasets, but present the results for analyses of the separate genomic regions.

Heuristic searching of the combined plastid dataset yielded 24 equally most parsimonious (MP) trees of 2247 steps in a single island. The MP strict consensus tree (Supplementary Fig. S1) resolved most of the major lineages of the subfamily Myrtoideae congruent with Wilson et al. (2005), and although relationships between many tribes were resolved, most lacked support. The Bayesian analysis of these data showed the same tribal structure but with less resolution between clades. Jackknife supports >50% from the MP analysis are indicated on the Bayesian majority-rule consensus tree (Fig. 1) and the MP strict consensus tree (Supplementary Fig. S1).

Fig. 1.  Bayesian 50% majority-rule consensus tree of combined plastid data. Values shown on tree indicate clade support from Bayesian posterior probabilities (PP, above branches) and jackknife values from maximum parsimony analysis of >50% (jk, below). Thick lines received strong support 1.00 PP and jk ≥ 90%. New or revised tribal assignments are indicated in bold.

Sampling of some groups was limited but the analyses provided continued support for most previously recognised tribes. The core Myrtaceae (subfamily Myrtoideae), tribes Backhousieae, Chamelaucieae, Leptospermeae, Myrteae, Syzygieae and Xanthostemoneae all received robust support (100% jk, 1.00 PP); Eucalypteae, Lophostemoneae and Metrosidereae all have strong support (99% jk, 1.00 PP). By contrast, the tribe Kanieae is not resolved as monophyletic. The type genus, Kania, is moderately supported as sister to Metrosidereae (81% jk, 1.00 PP), but is not at all closely associated with genera formerly placed with it in the tribe Kanieae. Those other genera, the Tristaniopsis group, are weakly monophyletic (68% jk, 1.00 PP), but there is robust internal support (100% jk, 1.00 PP) for the monophyly of a subclade comprising Ristantia, Mitrantia and Sphaerantia. The current tribe Tristanieae is rendered paraphyletic by the placement of Cloezia, and the clade is only weakly supported (52% jk, 1.00 PP). Rather, a weak clade (52% jk, 0.99 PP) places Cloezia (100% jk, 1.00 PP) sister to Xanthomyrtus (97% jk, 1.0 PP), that clade being sister to a robust Tristanieae sens. strict., comprising Tristania + Thaleropia (100% jk, 1.00 PP).

Relationships between some tribes and tribal groupings also receive support in these analyses. As in previous analyses, Xanthostemoneae and Lophostemoneae are resolved as sister (96% jk, 1.00 PP) and form the first diverging lineage in the subfamily, with modest support (68% jk, 0.99 PP); Chamelaucieae and Leptospermeae (99% jk, 1.00 PP) form a strong clade; Melaleuceae (84% jk, 1.00 PP) and Osbornieae are still resolved as sister taxa but with modest support (70% jk, 1.00 PP). Relationships of two other genera that have been unclear previously, Syncarpia Ten. and Lindsayomyrtus B.Hyland & Steenis, remain unresolved.

Heuristic searching of the combined nuclear dataset yielded 36 equally most parsimonious (MP) trees of 2912 steps in a single island. The Bayesian analysis of these data showed a largely similarly resolved structure but with some areas of conflict (Fig. 2). Jackknife supports >50% from the MP analysis are indicated on the Bayesian majority-rule consensus tree (Fig. 2) and the MP strict consensus tree (Supplementary Fig. S2). Again Myrtoideae and most of the existing tribes were retrieved, although supports were somewhat lower with this dataset; Syzygieae and Xanthostemoneae received robust support (100% jk, 1.00 PP), as did clades of Xanthomyrtus and Cloezia; Backhousieae, Eucalypteae, Leptospermeae, Metrosidereae all have strong support (>90% jk, 1.00 PP). Moderate support for Kania + Metrosidereae (84% jk, 1.00 PP) is again found with the nuclear data. The remainder of the present Kanieae, the Tristaniopsis group, is again found to form an unrelated and modestly supported clade (72% jk, 1.00 PP). This group is resolved as a supported sister to a weak Lophostemoneae + Xanthstemoneae (<50% jk, 1.00 PP). Lophostemoneae + Xanthstemoneae is no longer the first diverging lineage in the nuclear analyses, with Myrteae shown as the unsupported first lineage to diverge (<50% jk, 0.86 PP) outside a polytomy containing all remaining tribes. There is very little resolution of the backbone of the tree. A feature of the Leptospermeae clade is that Leptospermum J.R.Forst & G.Forst. is shown to be paraphyletic, a situation first demonstrated by O’Brien et al. (2000). In the plastid analysis, L. grandifolium Sm., a representative of Leptospermum sens. strict., is sister to other members of the tribe (Fig. 1, 69% jk, 1.00 PP) with L. anfractum A.R.Bean nested among the remaining genera as sister to Neofabricia Joy Thomps. Here, in the nuclear analysis, Leptospermum is still found to be paraphyletic, but the topology is rather different, with L. anfractum sister to other members of the tribe (98% jk, 1.00 PP) and L. grandifolium sister to Kunzea Rchb.

Fig. 2.  Bayesian 50% majority rule consensus tree of combined nuclear data. Values shown on tree indicate clade support from Bayesian posterior probabilities (PP, above branches) and jackknife values from maximum parsimony analysis of >50% (jk, below). Thick lines received strong support 1.00 PP and jk ≥ 90%. New or revised tribal assignments are indicated in bold.

A notable difference between the two nuclear analyses is the placement of Cloezia. In the nuclear MP analysis (Supplementary Fig. S2), it is placed in a clade with Chamelaucieae, Leptospermeae and Eucalypteae, rather than as a sister to Xanthomyrtus where it is placed in all other analyses, albeit on a long branch. Although there is strong support from the plastid Bayesian analyses for the sister arrangement with Xanthomyrtus (0.99 PP), the clade is unsupported by the nuclear data (0.83 PP), and there is no jackknife support in either MP analysis for Cloezia’s placement. This is evidence that the genus forms a divergent lineage and confirms that its status needs reassessment.

The major differences between the plastid and nuclear analyses lie in largely unsupported resolution of relationships among tribes. Deep branches separating clades tend to be very short and thus are supported by few characters, so it is not unexpected that resolution is poor at this level. As discussed above, there is some conflict in placement of Cloezia. Although there is modest support for the placement of Lindsayomyrtus sister to Chamelaucieae + Leptospermeae with the plastid data (74% jk, 0.97 PP, Fig. 1, Supplementary Fig. S1), the nuclear MP analysis has it as unsupported sister to Syncarpia (<50% jk, Supplementary Fig. S2) and the nuclear Bayesian analysis as unsupported sister to Melaleuceae + Osbornieae (0.75 PP, Fig. 2). Again, this supports the distinctiveness of the genus and confirms that its recognition as a monotypic tribe is warranted.

Leaf cuticles of Kania show stomatal clumping that is uneven and interrupted, as illustrated in K. eugenioides (Fig. 3a, b). However, the two species of Xanthomyrtus examined show very distinctive stomatal clumping with distinct bands of dense stomata, as can be seen in X. montivaga A.J.Scott (Fig. 3c, d), where the stomatal distribution is clearly independent of underlying venation patterns. Neither of these stomatal arrangement types is typical in the Myrtaceae and most other species of Myrtaceae demonstrate stomatal distribution types more typical of other dicotyledonous angiosperm leaves. Either the stomata are evenly distributed and unaffected by underlying venation, illustrated in Metrosideros laurifolia Brongn. & Gris. (Fig. 4a), or else stomata are restricted in areolae, as on the cuticles of Lophostemon confertus (R.Br.) Peter G.Wilson & J.T.Waterh. (Fig. 4b). In the latter case, the stomata are evenly distributed in the areolae and the gaps occur over leaf veins, which interrupt the underlying spongy mesophyll. The resulting arrangement of stomata does not constitute stomatal clumping.

Fig. 3.  Examples of variation in stomatal distribution. Light microscopy. (a) Cuticle of Kania eugenioides, showing ‘clumping’ of stomata with no clear distribution into vein islets or areolae; scale bar: 200 μm. and (b) A close up of the clumped stomata, showing a disorganised distribution in K. eugenioides. (c) Xanthomyrtus montivaga, showing an alternative form of aggregation of stomata into distinct zones, with large non-stomatal areas between zones, with no clear relation to underlying venation; scale bar: 200 μm. (d) A close up of the aggregated stomata. There are no spaces between any of the subsidiary cells of stomata in Xanthomyrtus species, and stomata are approximately half the size (~5 μm).

Fig. 4.  Examples of the most common forms of stomatal distribution in cuticles from across the Myrtaceae. Light microscopy. (a) Cuticle of Metrosideros (Carpolepis) laurifolia, showing an even distribution of stomata, and (b) Lophostemon confertus, showing separation of stomata by major and minor leaf venation into vein islets or areolae. Scale bars: 500 μm.

Water stomata, with associated cuticular striations, occur in both Kania and Tristaniopsis. Well-developed cuticular striations are found in Kania species (Fig. 5a, c), and also in some other members of the present tribe Kanieae, such as Barongia lophandra Peter G.Wilson & B.Hyland (Fig. 5b). They can also be found in some species of Metrosideros, such as M. robusta A.Cunn. (Fig. 5d), and Tristaniopsis, for example, T. laurina (Sm.) Peter G.Wilson & J.T.Waterh. (Fig. 5e), but are not quite as well developed. By contrast, both water stomata and cuticular striations are absent from Xanthomyrtus species, as observed in X. montivaga (Fig. 5f) and X. flavida (Stapf) Diels.

Fig. 5.  Cuticles of several species from the tribe Kanieae. (a–f) SEM images. (a) Kania eugenioides; scale bar: 50 μm. (b) Barongia lophandra; scale bar: 50 μm. (c) Kania urdanetensis (Elmer) Peter G.Wilson; scale bar: 50 μm. (d) Metrosideros robusta; scale bar: 50 μm. (e) Tristaniopsis laurina; scale bar: 20 μm. Note cuticular striations radiating from the water stomata and the papillose texture in a–e (but not as well developed in Metrosideros and Tristaniopsis). (f) Xanthomyrtus montivaga; scale bar: 20 μm. The cuticles of Xanthomyrtus lack water stomata entirely, as well as any associated cuticle striations.

The cleared fruit of Kania sp. (Fig. 6a) shows five major veins in the hypanthium, leading to each of the five sepals, with weaker secondary branches leading to the sepals. There also appears to be a well developed band of vascular tissue encircling the hypanthial rim. By contrast, the cleared fruit of Tristaniopsis collina (Fig. 6b, c) does not possess five strongly developed major veins in the hypanthium. Several veins of similar size and staining quality are seen running up to the sepals, but also leading to the petals and staminal bundles, which in Tristaniopsis species are located opposite each petal. There is no strong correlation between vein size and perianth and there is no band of vascular tissue encircling the hypanthial rim.

Fig. 6.  Vascularisation in skeletonised fruits. (a) Kania sp., scale bar: 1 mm; and (b, c) Tristaniopsis collina, scale bars: 2 mm.

The present study confirms most of the previous tribal groupings (Wilson et al. 2005; Biffin et al. 2010; Thornhill et al. 2015). However, note that the so-called BKMMST clade (Backhousieae, Kanieae, Metrosidereae, Myrteae, Syzygieae, Tristanieae) of Biffin et al. (2010) was not recovered by our analyses. The chief difference is that we did not find evidence of a robust connection between the Myrteae and the other genera in that grouping. Rather, in our analyses, the Myrteae was associated with an unsupported group comprising many of the remaining tribes (<50% jk, 0.64 PP, plastid), or was unsupported as the earliest diverging lineage in the subfamily (<50% jk, 0.86 PP, nuclear). Thornhill et al. (2015) also failed to find support for the BKMMST clade, with the Myrteae having no (0.78 PP) support as sister to the others.

A recent large-scale study across the order Myrtales (Maurin et al. 2021) targeted a comprehensive suite of more conserved low-copy nuclear genes. That analysis also found little support for the so-called BKMMST grouping of tribes, with Syzygieae consistently falling outside a clade comprising the other tribes. That study included a wider sampling of genera assigned to the Kanieae and concurs with our finding that Kania, which was represented only by a poor, partial sequence in the single plastid locus analysis of Wilson et al. (2005), is now resolved as sister to the Metrosidereae with moderate jackknife support, quite separate from the remainder of tribe Kanieae, the Tristaniopsis group.

The phylogenetic positions of Cloezia and Xanthomyrtus have often been the subject of debate. Wilson (2011) tentatively included Cloezia in Kanieae sens. lat., on the basis of its placentation being similar to that found in Lysicarpus. However, phylogenetic analyses have shown both Cloezia and Xanthomyrtus to form a clade with core members of Tristanieae. In both Biffin et al. (2010) and Thornhill et al. (2015), they were successive sisters to the strongly supported core Tristanieae, but the relationship of Cloezia to the other taxa was not strongly supported, with a PP of ≤0.95 in the former study, and PP of only 0.31 in the latter study, which analysed sequence data from exactly the same regions. In the present analyses, the three taxa form a single clade but there is only nominal support from parsimony for the placement of Xanthomyrtus in a clade that includes the tribe Tristanieae (≤52% jk, 1.00 PP). Rather, Xanthomyrtus is resolved as weakly sister to Cloezia, a degree of relationship also recovered by Maurin et al. (2021) where PP was only 0.54.

Morphological data do not assist with resolution of these relationships. There is little difference in pollen morphology between Cloezia and Xanthomyrtus; both have parasyncolpate pollen that is similar in size and exine pattern (Thornhill et al. 2012a). Wood anatomy is similarly uninformative. The vessel-ray pitting in Cloezia is fine and alternate, similar to intervessel pitting (P. Gasson, pers. comm.), and in Xanthomyrtus it is described as ‘small, half-bordered’ by Ingle and Dadswell (1953, p. 384), so there is little distinction there. In the context of the family, both of these characters (pollen morphology and wood anatomy) would be interpreted as plesiomorphic and, therefore, not be reliable indicators of shared evolutionary history. By contrast, the genera of core Tristanieae, Tristania and Thaleropia, share a highly derived pollen type that is the smallest found in the family so far, and are triporate or acolpate with a psilate exine (Pike 1956, pp. 39, 46; Gadek and Martin 1981, p.179; Patel et al. 1984, p, 939 for Tristania; Thornhill et al. 2012a, p. 267 for both genera).

In the case of Kania, our results point to a closer relationship with Metrosideros than with those genera previously included in Kanieae sens. lat. There is some support for this from wood anatomy. Vessel-ray pits are described as ‘very large, either circular to horizontally elongated or forming almost scalariform series’ (Erdtman and Metcalfe 1963, p. 250) in stems from a type specimen of Kania eugenioides, and as ‘simple and rounded to elongated’ in wood of Metrosideros (Ingle and Dadswell 1953, p. 378; Meylan and Butterfield 1978). However, there is less support from pollen morphology because Metrosideros pollen is much larger than Kania pollen and has distinct apocolpial islands (Pike 1956; McIntyre 1963; Gadek and Martin 1981; Thornhill et al. 2012b).

Leaf epidermal characters identified by Tarran et al. (2016) definitely favour a closer relationship between Kania and at least some species of Metrosideros, although the latter differs significantly in lacking clumped stomata. The findings relating to floral vascularisation are more significant because the reduction in the number of main vascular traces to only five has not been reported elsewhere in the family. Wilson (1993, 2011) was the first to suggest that this feature was a likely synapomorphy for the tribe Metrosidereae, on the basis of the published observations of Dawson (1970a, 1970b, 1972b, 1972c, 1972d, 1975) who provided illustrations of transverse sections of flowers or developing fruits in the Metrosideros group that consistently showed five main veins in the hypanthium. The cleared flower of Kania (Fig. 6a) clearly shows five major vascular traces leading to the sepals with strong secondary branches to the petals. This approaches the pattern of vascularisation observed in Metrosidereae, where the five traces are sometimes heavily thickened in both extant (for example, Dawson 1975, fig. 11) and fossil (Pole et al. 2008, fig. 12; Tarran et al. 2017, fig. 4) taxa. In Kania, there is evidence of a transition to five well developed veins, but the pattern is not as distinctive as it is in many taxa of Metrosidereae. In contrast with this, the vascularisation of the flower of Tristaniopsis collina (Fig. 6b) does not show a particularly strong association of vascular traces with perianth parts.

Morphological characters, particularly cuticle micromorphology and floral vascularisation, indicate a greater affinity between Kania and the tribe Metrosidereae than between it and the genera usually placed in the tribe Kanieae, a relationship also strongly supported by our molecular phylogenetic analysis. Kania is independent of the other genera with which it has been grouped in the tribe Kanieae (sensu Wilson et al. 2005) and shows a robust affinity with the tribe Metrosidereae. However, we also conclude that, because Kania differs from Metrosidereae in anther morphology (prominent connective), placentation (placenta remote from base of style), distinctive cuticular characters, and genetic distance, that retention of Kanieae as a separate, monogeneric tribe is justified. A further consequence of our analysis is that a new tribe is required to accommodate most other genera presently assigned to the Kanieae. This new tribe, Tristaniopsideae, is described below.

The phylogenetic analysis also confirms previous relationships among taxa grouped with Tristanieae. Xanthomyrtus had been referred to Tristanieae, but differs from the two core genera, Tristania and Thaleropia, in having a predominantly four-merous perianth, a compressed-reniform seed with a crustaceous testa, an embryo more like that of Xanthostemon F.Muell. and a fleshy fruit. All three genera do have leafy cotyledons that lie face-to-face, but in Xanthomyrtus the hypocotyl is bent so that it lies along the edges of the cotyledons (accumbent), as noted by Landrum and Stevenson (1986), whereas in the other two it is straight (Dawson 1974; Wilson 1993). Wilson et al. (2005) did not place Cloezia in any tribe, whereas Wilson (2011) tentatively included it in Kanieae sens. lat., on the basis of the arrangement of the ovules. However, as already noted, recent phylogenies that have included Cloezia have placed it in a grade with Xanthomyrtus and core Tristanieae. In our analysis, and that of Maurin et al. (2021), Cloezia was found to be sister to Xanthomyrtus rather than to core Tristanieae and differs significantly from the other genera in having a strongly exserted capsule with basal placentas bearing ovules in a circular series. The Bayesian analyses showed Cloezia to be on a strongly supported long branch, an indicator of early divergence and long isolation. Consequently, our preference is to recognise new tribes to accommodate both Xanthomyrtus and Cloezia to reflect the genetic distance (long branches) and their marked morphological divergence, particularly in placentation and embryo features.

Recognition of these extra tribes, and emending the circumscription of Kanieae, brings to four the number of tribes that do not occur naturally on the Australian mainland today.

Type: Tristania R.Br.

Trees or shrubs; leaves opposite, growth monopodial. Inflorescences thyrsoids or cymes; flowers 5-merous, yellow or orange to red; stamens free or fused into 5 groups opposite petals, usually fewer than 25. Ovary half-inferior, style inserted in the apex of the ovary, style base adjacent to placentas; ovary usually trilocular. Fruit a capsule. Seed linear, embryo straight, cotyledons lying face to face. Pollen grains quite small with a smooth exine.

A small tribe of 2 genera: Tristania, Thaleropia

Kanieae Peter G.Wilson ex Reveal, Phytoneuron 2012–37: 217 (2012), isonym.

Type: Kania Schlr.

Trees or shrubs; leaves opposite. Inflorescence axillary, cymes or panicles; flowers yellow; stamens free, in a single whorl on the hypanthial rim, evenly spaced or, occasionally, grouped opposite the petals; anthers with elongated connectives. Style terminal on the ovary; ovules scattered on basal placentas that are remote from the style. Fruit a capsule, exserted from the hypanthium; seeds linear; embryo straight; cotyledons lying face-to-face.

A monogeneric tribe of ~10 species that occurs only in Malesia (New Guinea and the Philippines). Fossil evidence (Tarran et al. 2016, 2017) indicates that Kania may have been present in Australia in the late Eocene to Oligo-Miocene.

Reveal (2012, p. 217) questioned the validity of the tribal name given in Wilson et al. (2005) and republished the tribe as ‘Kanieae Peter G.Wilson ex Reveal, trib. nov., based on Kanioideae Engl.’, with the presumed implication that the simultaneous publication of Kanioideae and Kanieae by Engler (1930) made the latter name superfluous. However, alternative advice (W. Greuter, pers. comm., 2014) is that the name Kanieae was validly published and that the Reveal name is an isonym.

Type: Xanthomyrtus Diels.

Trees or shrubs; branchlets hairy, often conspicuously glandular. Inflorescence of monads or triads. Flowers yellow, mostly 4-merous, sessile; stamens usually numerous, 1(–2)-seriate, free. Ovary inferior, usually 2- or 3-locular; ovules 10–20, arranged around the margin of the axile placenta; stigma small. Fruit a fleshy berry, reddish to blue-black; seeds many, small, with a crustaceous testa. Embryo with broad cotyledons lying face to face; hypocotyl accumbent.

A monogeneric tribe of 23 species, New Caledonia and Malesia (Philippines, Borneo, Sulawesi, Maluku, New Guinea)

Type: Cloezia Brongn. & Gris.

Shrubs or small trees. Inflorescences usually axillary cymes or monads. Flowers 5-merous, yellow or white; stamens in a single whorl, as long as the petals, anthers dorsifixed, versatile, connective sometimes expanded apically; ovary half inferior, 3-locular; ovules few in a ±circular series on the basal placenta; style terminal, remote from the placenta, stigma small. Fruit a woody loculicidal capsule, exserted from the hypanthium; seeds linear; embryo straight, cotyledons lying face to face.

A monogeneric tribe of five species, endemic to New Caledonia.

Type: Tristaniopsis Brongn. & Gris.

Trees or occasionally shrubs. Inflorescences determinate (panicles, metabotryoids, thyrsoids or cymes). Flowers whitish to yellow. Stamens usually in multiple whorls (not in Mitrantia) and grouped opposite petals, sometimes fused into fascicles. Style-bases not adjacent to placentas, ovules often arranged in circular or semi-circular series. Fruit a capsule, frequently exserted from the fruiting hypanthium (except in Sphaerantia). Seeds various; hypocotyl straight and cotyledons sometimes foliaceous. Hypanthium vascularisation not reduced to 5 main veins.

A tribe comprising seven genera, Tristaniopsis, Lysicarpus, Barongia, Sphaerantia, Ristantia, Mitrantia, and Basisperma. Tristaniopsis is a genus of ~50 species, with a distribution extending from Myanmar and Thailand in the north, through Malesia and extending to eastern Australia and New Caledonia. The remaining genera are small, comprising between one and three species, and are narrow endemics in Papua New Guinea (Basisperma) and Queensland.

The phylogenies show some well supported groupings of genera within the new tribe. The three genera Sphaerantia, Ristantia and Mitrantia form a strong subclade (>97% jk, 1.00 PP), agreeing with previous analyses (Wilson et al. 2005) and strongly correlated with pollen morphology (Thornhill et al. 2012a) and shared presence of oil glands in the pith (P. G. Wilson, pers. obs.). Oil glands in the pith are also a feature of Basisperma (P. G. Wilson, pers. obs.), which was the basis for the comment in Wilson (1982) that Basisperma had no close affinities with the ‘Kania Alliance’ of Briggs and Johnson (1979). The shared occurrence of oil glands in the pith suggested that the genus was very likely to have affinities with these particular taxa, and this has now been confirmed in genomic analyses (Maurin et al. 2021).

New sequence data for this study are available from GenBank https://www.ncbi.nlm.nih.gov/genbank/: OM218672–OM218697 (ITS); OM730292–OM730334 (ETS); OM752313–OM752353 (trnK); OM752354–OM752372 (psbA–trnH). Other data, including molecular alignments and morphological scoring, that support this study will be shared upon reasonable request to the corresponding author.

Peter Wilson is an Associate Editor of Australian Systematic Botany but did not at any stage have editor-level access to this manuscript while in peer review, as is the standard practice when handling manuscripts submitted by an editor to this journal. Australian Systematic Botany encourages its editors to publish in the journal and they are kept totally separate from the decision-making processes for their manuscripts. The authors have no further conflicts of interest to declare.

Myall Tarran’s palaeobotanical research was funded through the University of Adelaide as part of his PhD studies.

Supplementary material is available online.