Phylogenetic relationships of Zieria (Rutaceae) inferred from chloroplast, nuclear, and morphological data

Abstract Zieria Sm. (Rutaceae, Boronieae) is predominantly native to eastern Australia except for one species, which is endemic to New Caledonia. For this study, sequence data of two non-coding chloroplast regions (trnL-trnF, and rpl32-trnL), one nuclear region (ITS region) and various morphological characters, based on Armstrong’s (2002) taxonomic revision of Zieria, from 32 of the 42 described species of Zieria were selected to study the phylogenetic relationships within this genus. Zieria was supported as a monophyletic group in both independent and combined analyses herein (vs. Armstrong). On the basis of Armstrong’s (2002) non-molecular phylogenetic study, six major taxon groups were defined for Zieria. The Maximum-parsimony and the Bayesian analyses of the combined morphological and molecular datasets indicate a lack of support for any of these six major taxon groups. On the basis of the combined Bayesian analysis consisting of molecular and morphological characters, eight major taxon groups are described for Zieria: 1. Zieria cytisoides group, 2. Zieria granulata group, 3. Zieria laevigata group, 4. Zieria smithii group, 5. Zieria aspalathoides group, 6. Zieria furfuracea group, 7. Zieria montana group, and 8. Zieria robusta group. These informal groups, except for of the groups Zieria robusta and Zieria cytisoides, correspond to the clades with posterior probability values of 100.

Zieria consists of prostrate shrubs to small trees, with opposite and trifoliolate, or rarely unifoliolate leaves. Infl orescences are axillary, with four-merous, white or pink fl owers. Th e fruits are comprised of one to four basally connate cocci, which dehisce explosively along the adaxial and apical margins. Th e seeds are usually one (often by abortion of one ovule) per fruit, with a thin brittle testa that is irregularly sculptured. In general, Zieria is distinguished from other genera of the Australian Rutaceae by the combination of opposite leaves, the conspicuous and 4-merous fl owers, free petals, four stamens, free fi laments, a deeply four-lobed disc, and dry, dehiscent fruits. Th is genus is predominantly native to eastern Australia, with the exception of the one species, Z. chevalieri Virot., which is endemic to New Caledonia. Th e distribution in eastern Australia extends from northeastern Queensland to Tasmania and as far west as Kangaroo Island in South Australia.
Sir James E. Smith fi rst described the genus in 1798, in memory of Jan Zier, a Polish botanist. In 1810, H.C. Andrews described the fi rst species, Zieria smithii Andrews, in H.C. Andrew's Botanist's Repository. In 1815, Bonpland published the descriptions of four species and soon after, in 1818, J.E. Smith described fi ve more species. Bentham in his Flora Australiense (1863) described 11 new taxa and provided the fi rst comprehensive key, with descriptions, synonyms and distribution data. For almost 136 years very little taxonomy was completed apart from C.T. White's descriptions of fi ve new taxa in 1942, and Virot's (1953) circumscription of the endemic species from New Caledonia. It was not until 2002 that Armstrong reassessed and revised the classifi cation, including defi ning six major taxonomic groups within Zieria. Accordingly, the nomenclature used in this paper is that of Armstrong (2002) and incorporates the morphological phylogenetic characters from that study (cf. Table 1). Th is study will be the fi rst to test the monophyly of Zieria and its six major taxonomic groups using molecular data.
A subfamilial phylogenetic analysis was completed for Rutaceae by Chase et al. (1999), Groppo et al. (2008Groppo et al. ( , 2012, Poon et al. (2007), Bayly et al. (2013), and Morton and Telmer (2014), using evidence from rbcL and atpB, rps16 and trnL-trnF and trnL-F, xdh, and ITS sequence variation. All of the above authors, except for Bayly et al. (2013), did not include taxa from either Zieria or Neobyrnesia (sister genus to Zieria). Bayly et al. (2013) only included three Zieria species and Neorbyrnesia, and therefore, their relationships to each other and to other taxa of Rutaceae based on molecular techniques need to be examined for the degree of congruence with morphological characters. Of the 32 species used in this study, 21 are considered endangered or vulnerable according to the Environment Protection and Biodiversity Conservation (EPBC) Act (http://www.environment.gov.au/cgi-bin/sprat/public/spratlookupspecies.pl?name=zieria&searchtype=Wildcard).
Molecular studies can produce eff ective and practical solutions for conservation biology to taxonomic uncertainties with respect to rare and threatened taxa and, in light of the high proportion of endangered taxa and overlying distribution patterns for a number of these taxa, examinations should be conducted on Zieria.
Th e goals of this study are (1) to test the monophyly of the genus Zieria and to identify its closest relatives; (2) to evaluate the six taxonomic groups within Zieria as recognized in the most recent revision (Armstrong 2002);and (3) to examine the relationship based on distribution patterns and molecular change of the endangered or vulnerable taxa of Zieria.

Methods
For this study, two non-coding chloroplast regions (trnL-trnF, and the rpl32-trnL) were selected, as well as the Internal Transcribed Spacer (ITS) of the nuclear region and various morphological characters. Th e trnL-trnF region consists of the trnL intron and the trnL-trnF intergenic spacer (Taberlet et al. 1991). Th e rpL32-trnL intergenic spacer is in the SSC (small single copy) region of the chloroplast genome. Th e rpl32-trnL was fi rst used for phylogenetic studies by Shaw et al. (2005). Various workers have found that both of these sequences provided good resolution at the generic and species level (e.g. Wallander and Albert 2000;Baker et al. 2000). Th e ITS region of the 18S-26S nuclear ribosomal DNA (nrDNA) consists of three genes that code for the 18S, 5.8S and 26S ribosomal subunits. Th e three genes are separated by two internal transcribed spacers, ITS1 between 18S and 5.8S and ITS2 between 5.8S and 26S. Morphological characters were taken from information in Armstrong's (2002) taxonomic revision of Zieria (Table 1).

Taxon sampling & DNA extraction
Vouchers for the 33 species used in this study along with the GenBank accession numbers are listed in the Appendix 1. Th e total genomic DNA was extracted from (0.5-1.0 g) fresh or dried leaf material. Leaves were ground with a mortar and pestle and subsequently treated with the DNEasy plant DNA extraction kit from Qiagen (Qiagen, Valencia, California, USA) following the manufacturer's protocol. Alignments were made using the Sequencher software program (Gene Codes Corporation, Ann Arbor, MI), for each marker for 32 Zieria and 1 Neobyrnesia species and also a broader trnL-F alignment with sampling across all Rutaceae subfamilies including Meliaceae and Simaroubaceae as outgroups. All GenBank accessions numbers for the additional sequences can be found in Morton and Telmer (2014) with the exception of Boronia (EU853780), and Medicosma (EU853806) and Euodia (EU493243).

rpl32-trnL
Th e rpl32-trnL gene in 33 species was amplifi ed using the primer pair rpl32F/trnL (Shaw et al. 2005) to acquire the entire region. Th e fi nal PCR cocktail of 50 μl contained the following: 38 μl water, 5 μl of 10% Mg free buff er solution, 3 μl of 25 mM MgCl 2 , 1 μl of 10 mM dNTPs, 0.25 μl Taq polymerase, and 0.5 μl of each primer. Th e amplifying reactions were run for 25 cycles of denaturing for 30 s at 95 °C, primer annealing for 50 s at 57 °C, and elongation for 2 min at 72 °C.

trnL-trnF
Th e trnL intron and the trnL-trnF intergenic spacer for 33 species were PCR-amplifi ed using the universal primers trn-c, trn-d, trn-e, and trn-f as described by Taberlet et al. (1991). For some samples the entire trnL intron/trnL-trnF spacer region was amplifi ed with trn-c and trn-f. In others, two separate amplifi cations were performed, one to amplify the trnL intron with trn-c and trn-d and the other to amplify the trnL-trnF spacer with trn-e and trn-f. In general each 50 μl amplifi cation reaction contained the same proportions as in the rp16 reactions. PCR amplifi cation used a 7-min denaturing step at 94 °C followed by 30 cycles of denaturing for 1 min at 94 °C, primer annealing for 1 min at 45 °C, and elongation for 1 min at 72 °C, with a fi nal 7-min elongation step at 72 °C.

ITS
Th e amplifi cation of the ITS was performed successfully on 33 species using oligonucleotide primers ITS1/ITS4 (White et al. 1990) to acquire the entire region. Th e DNA fragment amplifi ed using these two primers is approximately 800 bp long and includes ITS1, ITS2 and the 5.8S ribosomal gene. Th e basic mix contained the following: 38 μl of water, 5 μl of 10% Mg free buff er solution, 3-6 μl of 25 mM MgCl 2 , 1 μl of 10 mM dNTPs, 0.5 μl of each primer (10 nM), and 0.25 μl Taq DNA along with 1.5 μl of DNA template for each reaction. Th e thermal cycler was programmed to perform an initial 1 cycle of denaturation at 95 °C for 2 min, followed by 24 cycles of 30 seconds at 55 °C, 72 °C for 90 seconds and 95 °C for 30 seconds. Th is was followed by a 10 min. extension at 72 °C to allow completion of unfi nished DNA strands.

Cycle sequencing
Th e PCR products were cleaned using the QIAGEN QIAquick PCR purifi cation kit (QIAGEN Inc., Chatsworth, California, USA) following the protocols provided by the manufacturer. Cleaned products were then directly sequenced using the ABI PRISM Dye Terminator Cycle Sequencing Ready Kit with AmpliTaq DNA Polymerase (Applied Biosystems Inc., Foster City, California, USA). Unincorporated dye terminators were removed using the QIAGEN DyeEx dye-terminator removal system (QIAGEN Inc.) following the manufacturer's recommendations. Samples were then loaded into an ABI 3100 DNA Sequencer. Th e sequencing data was analyzed and edited using the Sequencher software program (Gene Codes Corporation, Ann Arbor, Michigan, USA).

Morphological characters
A morphological dataset of 48 characters was constructed. Twenty-eight characters were coded as unordered binary and 20 as multistate. All but two characters (4-types of pubescence on young branches and 12-presence or absence of revolute lamina margins) were variable within Zieria. Th e invariant characters were included because they were thought to be important in testing the monophyly of the genus. All analyses were conducted as stated in the analysis section. Character states of taxa were taken from Armstrong (2002: 291-294).

Phylogenetic analysis
Boundaries of the trnL intron, rpl32-trnL, and the ITS nuclear gene were determined by comparison with sequences in GenBank. Th e following two alignment criteria and methodology were used: (1) when two or more gaps were not identical but overlapping, they were scored as two separate events and (2) phylogenetically informative indels (variable in two or more taxa) were scored as one event at the end of the data set. All DNA sequences reported in the analyses have been deposited in GenBank (Appendix 1). Maximum-parsimony (MP) analyses of all single markers as well as the combined datasets were performed in PAUP* 4.0b8 (Swoff ord 2002) using the heuristic search option and with uninformative characters excluded. Searches were conducted with 100 random-taxon-addition replicates with TBR branch swapping, steepest descent, and MulTrees selected with all characters and states weighted equally and unordered. All trees from the replicates were then swapped onto completion, all shortest trees were saved, and a strict consensus or majority rule tree was computed. Relative support for individual clades was estimated with the bootstrap method (Felsenstein 1985). One thousand pseudoreplicates were performed with uninformative characters excluded. Ten random-taxon-addition heuristic searches for each pseudoreplicate were performed and all minimum-length trees were saved for each search. To reduce bootstrap search times, branches were collapsed if their minimum length was zero ("amb-").
Th e Bayesian analysis of the combined molecular and morphological analysis used a mixed-model approach (Mr Bayes 3.1.2 Ronquist et al. 2005). MrModelTest v2.3 Crandall 1998, 2001;Nylander 2004) was used to choose the best evolutionary model, as selected by the Akaike Information Criterion. Four independent analyses were run, each performing 10 million generations, sampling every 1000 th generation and using 3 heated and 1 cold chain, and other default settings. Tracer v1.4.1. (Rambaut and Drummond 2007) was used to assess convergence of the runs and to discard the initial 20% of the trees as a burn-in. Branch lengths are averaged from the distribution of trees and the posterior probability values (BPP) for the branches reported (Nylander et al. 2004). Morphological state changes were examined on the combined tree by using MacClade 4.0 (Maddison and Maddison 2000).
To determine the combinability of the data sets, their data structure was compared using methods outlined by Mason-Gamer and Kellogg (1996), who discussed various ways to assess confl ict between data sets. In one method the combination of independent data sets is possible if the trees do not confl ict or if confl ict receives low bootstrap support. Th erefore, each node on the independent trees is tested for congruence against the other. If the nodes do not contain confl icting information, they are congruent and the data sets are combinable. Where there are incongruent nodes, the bootstrap values for each node are examined. If the support is less than 70%, there is no hard confl ict and the incongruence is interpreted as being due to chance. In this study the diff erent data sets were analyzed in combination to see how each data set changed or confi rmed the tree topologies of each other and to adopt a hypothesis of phylogenetic relationships for the genus. (2014) found that species with low genetic diversity are less able to respond to environmental change; therefore this information can be informative and has been considered.

Morton and Schlesinger
Th is study examined the following 15 of the 21 endangered or vulnerable species (Z. adenophora, Z. baeuerlenii, Z. buxijugum, Z. citriodora, Z. collina, Z. convenyi, Z. formosa, Z. granulata, Z. ingramii, Z. murphyi, Z. obcordata, Z. parrisiae, Z. prostrata, Z. verrucosa, and Z. tuberculata). An examination for similarity was made using the distribution patterns and the number of bp changes within all three genes for the taxa in clades that had strong posterior probabilities.

Results
Th e inclusion of gap coding in all data sets containing molecular data resulted in more homoplasy and lack of resolution; therefore, gap coding was not used in the following results. GenBank sequences EU281855-EU281953 were specifi cally generated for this study.

Larger trnL-trnF Family Analysis
Multiple sequence alignment of Zieria and Neobyrnesia with 44 other Rutaceae and closely related taxa resulted in a data matrix of 1038 characters. No regions were excluded. Of the 1038 positions constituting the aligned trnL-trnF sequences, 357 (34%) were variable and 408 (39%) were parsimony-informative. Th e analysis recovered 4,383 equally optimal trees of 1037 steps (CI = 0.57, RI = 0.72; Fig. 1).

trnL-trnF
Multiple sequence alignment of Zieria and Neobyrnesia resulted in a matrix of 1035 characters. A total of 10 gaps were required for proper alignment of the trnL-trnF sequences. Th ese gaps ranged from one to 15 bps. No regions were excluded. Mean percentage G + C content was 56%. Of the 1035 positions, 127 (12.3%) were variable and 33 (3.2%) were parsimony-informative. Th e analysis recovered 35,458 equally optimal trees of 71 steps (CI = 0.59, RI = 0.69).
Zieria was supported as monophyletic in the strict consensus trees (BS 100). Most of Zieria consists of an unsupported grade or small polytomies except for one minor clade with bootstrap support of 75% (Z. furfuracea R.Br. ex Benth. and Z. laxifl ora Domin).

Rpl32-trnL
Multiple sequence alignment of Zieria and Neobyrnesia resulted in a matrix of 1180 characters. Approximately 14 gaps were required for proper alignment of the rpL32-trnL sequences. Th ese gaps ranged from one to 49 bps. No regions were excluded. Mean percentage G + C content was 30%. Of the 1180, 236 (20%) were variable and 46 (3.9%) were parsimony-informative. Th e analysis recovered 87,213 equally optimal trees of 77 steps (CI = 0.69, RI = 0.90).
Zieria was supported as monophyletic in the strict consensus trees (BS 100). Th e tree mainly consists of a polytomy except for one minor clade with bootstrap support greater than 75% (Z. furfuracea and Z. laxifl ora (BS 95%)).

ITS
Multiple sequence alignment of Zieria and Neobyrnesia resulted in a data matrix of 714 characters. Approximately fi ve gaps were required for proper alignment of the ITS sequences. Th ese gaps ranged from one to 16 bps. No regions were excluded. Mean percentage G + C content was 36%. Of the 714, 207 (29%) were variable and 82 (11.5%) were parsimony-informative. Th e analysis recovered 7,259 equally optimal trees of 169 steps (CI = 0.72, RI = 0.84). Zieria is supported as a monophyletic clade in the strict consensus tree (BS 100%). Basal within this clade is Z. citriodora J.A. Armstr., which is sister to Z. aspalathoides A. Cunn. Ex Benth. and Z. ingramii J.A. Armstr. (BS 88%). Th e backbone phylogeny of the genus remained unresolved, however a number of minor clades were inferred. Clades that contain bootstrap support greater than 75% starting from the base of the tree include: 1) a clade containing Z. arborescens Sims sister to a polytomy of Z. covenyi J.A. Armstr., Z. murphyi Blakely and Z. odorifera J. Phylogenetic utility of the three genes (trnL-trnF, rpl32-trnL, and ITS) in Zieria Th e respective numbers of variable and potentially phylogenetically informative characters in each dataset, the consistency indices and the numbers of branches with bootstrap support above 75% can be found in Table 2. Th e ITS sequences produced the most parsimony-informative characters for similar taxon sampling when compared with the other regions: trnL-trnF (33), rpl32-trnL (46), and ITS (82). Th e trnL-trnF gene produced the fewest parsimony-informative characters. Th e ITS gene also had the highest number of resolved nodes at or above 75% bootstrap support when compared with all other genes: trnL-trnF (2), rpl32-trnL (2), and ITS (9). Th e combined parsimony analysis had 7 nodes at or above 75% bootstrap support whereas in the Bayesian analyses 13 branches had posterior probability values higher than 93%. Th ere was no correlation between the increase of the CI and RI values and the increase in the number of informative characters.

Combined molecular MP analysis
Following the methods outlined by Mason-Gamer and Kellogg (1996) and applied by Eldenäs and Linder (2000), the data sets were considered combinable. Within each gene analysis, trnL-trnF, Rpl32-trnL and ITS, the genus was monophyletic with 100% bootstrap support. Among the molecular trees there were no confl icting nodes with bootstrap support greater than 75%; therefore congruence exists between the data sets and a combined molecular analysis was completed.
Multiple sequence alignment of Zieria and Neobyrnesia resulted in a matrix of 2929 characters, of which (32.7%) include at least one accession with a gap. Mean percentage G + C content is 40%. Of the 2929, 570 (19.5%) were variable and 161 (5.5%) were parsimony informative. Th e analysis recovered 2,301 equally optimal trees of 378 steps (CI = 0.57, RI = 0.74; Fig. 2

Morphological-based MP analysis
Of the 48 characters constituting the non-molecular dataset, 48 were variable and 45 (93.8%) were parsimony-informative. Th e analysis recovered 591 equally optimal trees of 278 steps (CI = 0.30, RI = 0.57). Zieria was monophyletic in the strict consensus of these trees (BS 100%). Th e in-group topology consisted of a large grade with only one clade that contained bootstrap support greater than 75% (Z. laxifl ora and Z. laevigata (BS 75%)).

Combined molecular and morphological data
Following the methods outlined by Mason-Gamer and Kellogg (1996), the molecular and morphological data sets contained only one potential hard confl ict between a clade containing Z. fraseri and Z. laevigata (BS 100%) in the molecular data set and a clade containing Z. laxifl ora and Z. laevigata (BS 75%) sister to Z. fraseri in the morphology data set. Th e positions of these three taxa have interchanged among the three separate molecular data sets and this is refl ected in the morphology matrix having all three grouped together. Th e confl ict appears to be due to a lack of resolution within the independent molecular dataset or that some of the morphological characters are homoplasious; therefore congruence exists between the data sets and a combined analysis was completed. Multiple sequence alignment of Zieria and Neobyrnesia resulted in a matrix of 2977 characters, of which 28% include at least one accession with a gap. Of the 2977 positions constituting the aligned sequences, 618 (%) were variable and 209 (%) were parsimony informative. Th e analysis recovered 555 equally optimal trees of 1177 steps (CI = 0.62, RI = 0.59; Fig. 3

majority rule tree).
Zieria was supported as monophyletic in the strict consensus trees (BS 100). Zieria consists mainly of grades except for several minor clades with bootstrap support greater than 75%. Clades that contain bootstrap support greater than 75% starting from the base of the clade include: 1) a clade containing Z. furfuracea and Z. laxifl ora (BS 76%); 2) a clade containing Z. fraseri and Z. laevigata (BS 100%); 3) a clade containing Z. prostrata and Z. smithii (BS 95%); 4) a clade containing a polytomy of (Z.

Conservation
Th is study examined 15 of the 21 endangered or vulnerable species found in Zieria for similarity in their distribution patterns and for the number of bp changes within all three genes inside clades that had strong posterior probabilities.
Th e fi rst clade containing Z. adenodonta and Z. collina (BPP 100 and BS 92%) have similar distribution patterns, however two of the three genes indicated had numerous bp changes (over 10 bps), indicating the taxa are distinct species.
Th e second clade contains eight species, one species being Z. buxijugum (BPP 100 and BS 93%), and although the species all occurred mostly in the southeastern territory (New South Wales, Victoria and Tasmania), they had numerous bp changes between taxa.
Within the clade consisting of Z. aspalathoides, and Z. ingramii (BPP 100), there is distributional overlap, however there are over 30 bp changes among the taxa.
Although Z. adenophora has a non-overlapping distribution pattern from Z. furfuracea, and Z. laxifl ora, the latter two taxa are very similar in distribution pattern. All three taxa have numerous bp diff erences, however Z. furfuracea and Z. laxifl ora only had 2 solid bp diff erences.
Th e third clade consisted of Z. prostrata and Z. smithii (BPP 100 and BS 76%) these taxa have non-overlapping distribution patterns and two of the three genes had numerous bp changes (over 10 bps).
Z. covenyi and Z. murphyi (BPP 83), are from the same area and only had 3 bp changes among all three genes.

Monophyly of Zieria and its closest relatives
We assembled a trnL-F dataset including 44 taxa of Rutaceae to determine the outgroup relationship of Zieria (Fig. 1). Based on this analysis six species of Zieria form a strongly supported clade with Neobrynesia (BS 94%). Th e monophyly of Zieria is also suggest by Bayly et al. (2013) and Appelhans et al. (2014). Bayly et al. (2013) using only rbcL and atpB also included Z. chevalieriiii from New Caledonian, the only disjunct species within Zieria to support not only the monophyly of Zieria but also the outgroup relationship with Neobrynesia.
Sister to this grouping is a clade containing the following taxa: Medicosma, Euodia, Boronia, Sarcomelicope and Melicope (see results for BS values and clade arrangements). We therefore used Neobrynesia as the outgroup for this study. Armstrong (2002), using morphological features, found that Zieria, together with Boronia s. l., Brombya, Medicosma, Neobyrnesia, and Euodia s. s., formed a distinct clade that is characterized by the presence of foliar sclereids. Although we did not include a species of Brombya the remaining members of the above group, plus Sarcomelicope and Melicope, are represented in the clade.

Circumscription of Zieria
Both independent and combined analyses of the molecular and morphological data supported the monophyly of Zieria (Figs 2, 3 and 4), as previously postulated by Armstrong (2002). Th e present study examined forty-eight morphological characters, including vegetative, fl oral, and fruit features (Armstrong 2002). Only one character, leaves palmately trifoliolate, provided a synapomorphy for Zieria (excluding Z. mur-phyi). Other morphological characters that had been used to defi ne the genus were examined (e.g. opposite leaves, 4-merous fl owers, free petals, four stamens, free fi laments, four-lobed disc and dehiscent fruits). Many of these morphological characters (e.g. opposite leaves, 4-merous fl owers, four stamens, free fi laments, and dehiscent fruits) that were used to defi ne the genus are also found in the outgroup Neobyrnesia and in other genera of Australasian Rutaceae, and therefore, are not generic synapomorphies of Zieria (Armstrong and Powell 1980). Th e only other potential synapomorphy of Zieria is the intrafl oral disc with "distinct antesepalous lobes", which in Neobyrnesia is entire. Th is study confi rms the need to identify additional morphological characters that provide synapomorphies for classifi cation at the generic level.

Circumscriptions of the six major groups of Zieria
On the basis of Armstrong's (2002) non-molecular phylogenetic study, six major taxon groups were defi ned for Zieria. Th e MP and the Bayesian analyses of the combined non-molecular and molecular datasets indicate a lack of support for any of these six groups (see Table 1 and Figs 2, 3 and 4).
Th e MP trees (strict-consensus trees from the independent, the combined molecular, and the non-molecular datasets) are poorly resolved and thus do not allow conclusive evaluation of the classifi cation of Armstrong's (2002) six taxon groups. Th e Bayesian tree from the combined molecular and morphological datasets provides groupings with high support; therefore this dataset is used to discuss these relationships (Fig. 4).
Characters that support the six major taxon groups defi ned by Armstrong (2002) are as follows: Group A contains 14 species and is characterized by having distinctly tuberculate younger branches, peduncles, petioles, midveins, and fruits. Group B contains fi ve species. Th e characteristics include younger branches slightly ridged or terete, primary infl orescence bracts boat-shaped and deciduous leaving a scar, and the abaxial surface of the calyx lobes with stellate hairs. Group C consists of four species defi ned by having younger branches distinctly ridged with prominent glabrous leaf decurrencies, lower lamina surface velvet like, midveins glabrous with pellucid glands, infl orescences equal to or longer than the leaves, apex of calyx lobes curved inward adaxially, anthers prominently sharply pointed, and fruits with pellucid glands. Group D comprises four species with the following characteristics: younger branches distinctly ridged with prominent glabrous leaf decurrencies; lower lamina surface glabrous and with pellucid glands that turn black on drying and become sunken; petiole either with pellucid glands or tuberculate; midvein glabrous with pellucid glands; and fruit with pellucid glands. Group E is composed of eight species with younger branches densely pubescent, upper lamina surface with simple hairs, lamina lower surface and midvein hirsute, fi la-ments warty towards the apex, anthers prominently sharply pointed, ovary pubescent, cocci sharply pointed, and fruits glabrous or pubescent. Group F, the fi nal group, consists of seven species. Th e characteristics include upper lamina surfaces that are velvet like, infl orescences equal to or longer than the leaves, primary bracts that are boat-shaped and fruits that are pubescent. In examining the Bayesian clade the following three mixed clades indicate that none of Armstrong's (2002) groups are monophyletic (Fig. 4). 1) Z. montana from Group D forms a sister grouping with Z. southwelli from Group B (BPP 100). 2). Z. furfuracea from Group A forms a sister grouping with Z. laxifl ora from Group C (BPP 100). 3). Z. minutifl ora (F. Muell.) Domin from Group E forms a well-supported polytomy with taxa from Groups A, B, and F (BPP 100).

Tentative new groups for Zieria
On the basis of the combined Bayesian analysis based on three genes (two-cholorplast and one-nuclear) and a morphological matrix (48 features), eight major taxon groups are distinguishable within Zieria. All of these informal groups, except for Groups 1 and 8, correspond to the clades with posterior probability values of 100 (Fig. 4). Th e make-up of these Groups are as follows: Th e examination of the 48 morphological characters within the Bayesian tree revealed no unambiguous synapomorphies. However, sets of morphological synapomorphies in combination provide unique groups of characters to defi ne a clade. Maiden & Betche. Th is group had several synapomorphies including the lamina surface containing pellucid oil glands, and the few fl owered infl orescences being equal to or longer than the leaves.

Z. cytisoides
Because of the lack of resolution, fi ve taxa, Z. citriodora, Z. arborescens, Z. minutifl ora, Z. obcordata and Z. pilosa, will remain unplaced until additional studies are completed. DNA for the following species were not examined and therefore these taxa will not be placed into groups until sequencing and analysis is completed: Z. chevalieriiii, Z. fl oydii, Z. hindii, Z. involucrata, Z. lasiocaulis, Z. obovata, Z. oreocena, Z. rimulosa, Z. robertsiorum, and Z. veronicea. Although six of the eight groups have strong posterior probabilities the relationships between these clades remain uncertain. Th e monophyly of the genus and of six of these groups appears unamibiguous; however, additional molecular and morphological studies are needed to further defi ne the groupings and internal relationships.
Z. covenyi and Z. murphyi (BPP 83), are from the same area and only had 3 bp changes among all three genes. Both taxa have several solid morphological diff erences such as leaves pubescent or glabrous, infl orescence numerous or few and fi laments dilated or not dilated respectively. Because of these solid morphological diff erences these species appear distinct.
Z. furfuracea, and Z. laxifl ora, (BPP 100) were very similar in pattern and had only 2 bp diff erences. Once again an examination of the non-molecular features revealed a number of diff erences such as the leaves having stellate-pubescence vs. being glabrous; fl owers ranging from 21-125 vs. commonly 9; petals valvate vs. imbricate; and fl owering from spring to early summer vs. late winter to spring, to name a few.
Taxa in clades with strong posterior probabilities, with similar distribution patterns and low genetic variation, need to be closely examined before conservation management decisions are made to assure that they are unique species.

Conclusion
Zieria as currently circumscribed (Armstrong 2002) is monophyletic. Th is is supported by the molecular phylogenetic analysis and by one morphological synapomorphy: distinct antesepalous lobes of the gynoecium. Th is study found that the previous six species groups considered by Armstrong (2002) are not monophyletic, and confi rmed that Neobyrnesia is the closest relative to Zieria. Th e analyses identifi ed eight groups within Zieria and six of the eight groups have strong posterior probabilities.
Based on the number of informative characters and the number of branches with supported, ITS is an excellent candidate for higher-level analysis. In addition, ITS produced very few alignment diffi culties within the ingroup and outgroup, and its tree topology remained consistent with that of the other genes.
Of the 32 taxa used in this study, 21 are considered endangered or vulnerable according to the EPBC. Several taxa grouped together and formed clades with strong posterior probabilities. Further examination revealed that two of these groups had similar distribution patterns and low genetic variation but solid diff erences in nonmolecular characters. Th e taxonomic relationships of these taxa should be closely examined as further conservation management decisions are made.
Th e phylogenetic analysis presented here provides the fi rst study within Zieria using both chloroplast and nuclear datasets, as well as a morphological dataset. Topics to be addressed in a future study include the determination of tribal and subtribal groupings and the use of additional taxa and genes to elucidate the biogeographic history of the genus.