usefulness of morphological characters for infrageneric classification of Elatostema ( Urticaceae )

You are free to share to copy, distribute and transmit the work, under the following conditions: Attribution: You must attribute the work in the manner specified by the author or licensor (but not in any way that suggests that they endorse you or your use of the work). Non-commercial: You may not use this work for commercial purposes. No derivative works: You may not alter, transform, or build upon this work. For any reuse or distribution, you must make clear to others the license terms of this work, which can be found at http://creativecommons.org/licenses/by-nc-nd/3.0/legalcode. Any of the above conditions can be waived if you get permission from the copyright holder. Nothing in this license impairs or restricts the author’s moral rights. Blumea 54, 2009: 181–191 www.ingentaconnect.com/content/nhn/blumea doi:10.3767/000651909X476139 RESEARCH ARTICLE


InTRoduCTIon
Elatostema J.R.Forst.& G.Forst.(Urticaceae) is a taxonomically problematic genus of c. 300 species that is widespread throughout the tropical, subtropical and subtemperate regions of Africa through SE Asia, Australasia to Polynesia.The identification and infrageneric classification of Elatostema has traditionally been based on morphological characteristics (for example, Robinson 1910, Winkler 1922, Schröter & Winkler 1935, Wang 1980a, b, Weddell 1854, 1856, 1857, 1869).The characters used in this phylogenetic analysis have, in part, been based on the morphological features used by these researchers.However, several features have been re-interpreted so that they can be scored more consistently than is possible for many of the more subjectively defined character states used by previous researchers (for example, venation patterns).A few additional characters, not used by previous workers, have also been included (for example, the form of nanophylls).Multimetric overlapping characters (that is, characters defined by dimensions rather than non-overlapping codes) have been included by conversion to non-overlapping or minimal overlapping characters states using the divergence coding technique of Almeida & Bisby (1984).The following multi-metric characters that were converted include: plant height, petiole length, lamina length, lamina width, lamina length to width ratio, lamina asymmetry, leaf vein pairs and male tepal length.However, stipule length was not used because species with non-persistent stipules usually loose their stipules while they are young and these are frequently smaller than, and hence, not comparable with the older, persistent stipules of other species.Up to five herbarium specimens of each species included in this study were analysed to cover the morphological variation within each species.Robinson (1910) In species of Elatostema (s.s.), the involucral bracts are present and enclose the staminate and pistillate flowers, whereas involucral bracts are absent in Pellionia and Procris.The pistillate flowers of Pellionia are arranged in condensed inflorescences, whereas those of Procris are arranged on a fleshy receptacle.The inflorescence of staminate flowers of Pellionia and Procris are always openly paniculate (Qi et al. 2003, Ariyanti 2004, respectively).Robinson (1910) emphasised the usefulness of the inflorescence and pistillate flowers for differentiating species of Elatostema, Elatostematoides, Procris and Pellionia.Robinson (1911) concluded that merosity of the perianth of pistillate flowers was taxonomically useful for distinguishing Elatostema, Elatostematoides, Pellionia and Procris, with Elatostema characterised by 3-partite pistillate flowers, although 2-and 4-partite perianth flowers also appear to be present in otherwise 3-partite flowered inflorescences.Elatostematoides was characterised as having a 5-partite pistillate perianth.However, Robinson mistook the staminodes for the perianth.The perianth is much reduced or absent in all species of Elatostema s.str.(Schröter & Winkler 1935).The pistillate perianth of Pellionia is 4-or 5partite and as long as the ovary.The pistillate perianth of Procris is 3-or 4-partite, deeply divided, widely spreading, and shorter than or as long as the ovary.Thus, Pellionia is indistinguishable from some species of Procris and Elatostematoides.

Morphological characters used by
The presence or absence of staminodes in pistillate flowers is thus not useful for distinguishing the genera discussed by Robinson.However, this character is taxonomically useful, at a tribal level, for distinguishing the Elatostemeae from all other tribes in the family.The pistillate flowers of all genera of Elatostemeae have staminodes, whereas the pistillate flowers of all other tribes lack staminodes (Friis 1993).We included perianth merosity in our dataset to test its usefulness in defining taxon relationships.Schröter & Winkler (1935) provided a comprehensive discussion of morphological characters that they regarded as taxonomically useful for understanding Elatostema and for distinguishing the four subgenera recognised by them, namely subg.Elatostema, subg.Elatostematoides, subg.Pellionia and subg.Weddellia.The morphological characters used by them included features of habit, anisophylly, phyllotaxy, leaf and venation characteristics, cystoliths, stipules and reproductive features (including inflorescence structure, floral and fruiting characteristics).Although promoting the usefulness of these characters, they recognised that the taxonomic circumscription of these four subgenera was often unclear.

Vegetative features
Herbs, subshrubs and shrubs -Members of Elatostema are mostly herbs or subshrubs, rarely shrubs.The majority of taxa of subg.Weddellia and subg.Elatostema are fleshy herbs, whereas many of the species of subg.Pellionia and subg.Elatostematoides have more sclerenchymatous cells in the stem during later growth stages.Therefore, some of these species appear slightly 'woody', here referred to as subshrubs.According to Schröter & Winkler (1935) there are many subshrubs or shrubs in subg.Pellionia, but they did not provide specific examples.Species of subg.Elatostematoides are all subshrubs.

Leaf characteristics
Anisophylly -The leaf-pairs at each node are characteristically very unequal in size in Elatostema such that the foliage appears strongly anisophyllous.This anisophylly is also present in other Urticaceous genera, including Pilea.Schröter & Winkler (1935) considered the presence or absence of the nanophyll (small leaves) and its shape as useful taxonomic characters in distinguishing subgeneric groupings within Elatostema.In subg.Weddellia, all species have shortly petiolate nanophylls, whereas many species in the remaining subgenera lack nanophylls.In subg.Elatostema, only three species have nanophylls (namely, E. ambiguum, E. bulbiferum and E. burmanicum).Sessile, minute, more or less linear and readily caducous nanophylls are found in a few species of subg.Elatostematoides (for example, a few species of the Elatostema rostratum group).Finally, in subg.Pellionia, some species have relatively large, green leaf-like nanophylls (for example, E. filicinum, E. raapii and E. sinuatum, and), some others have minute nanophylls (for example, E. hallieri and E. variolaminosum).
The form of the nanophylls is relatively consistent throughout the genus but has not been used as a taxonomic character by previous researchers.When nanophylls are absent, the normal larger leaves (macrophylls) are spirally arranged.The macrophylls vary in shape from obovate, elliptic, narrowly ovate to ovate or rarely roundish.All normal larger leaves that occur in Elatostema and other Urticacean taxa are here referred to as macrophylls to distinguish them from the much smaller nanophylls inserted at the same nodes.Goebel (1928) observed 'rudimentary leaves' (nanophylls) on newly germinated plants of E. sessile (type species of genus), that became fused with the stipule during later stages of growth.
Phyllotaxy - Schröter & Winkler (1935) described the phyllotaxy of Elatostema as opposite, decussate, similar to other members of the family.However, since one leaf of the leafpair at each node is much reduced compared to the other leaf, this gives the appearance of an alternate/spiral arrangement.
Leaf shape symmetry -The width of the lamina of the macrophyll on each side of the central axis (from midvein to margin) is unequal.
Leaf base symmetry -The base of the broader side of the macrophyll's lamina usually exceeds the length of the narrower side.Occasionally the base is slightly cordate (for example, E. hastatum and E. reticulatum) to more fully cordate (rounded on both sides) (for example, E. griffithianum).
Leaf venation -Robinson (1910) concluded that it was not possible to categorize the venation patterns found in Elatostema because they were too variable.In contrast, Schröter & Winkler (1935) considered venation patterns taxonomically Only one vein extends to the apex, the other vein is much shorter, directed more towards the margin, and has more lateral secondary veins on this same side.This type III venation (another modification of type I) is found in subg.Elatostematoides and in some species of Pellionia.Type IV semi-pinnately to subpinnately nerved, with five subtypes which appear to be minor variants.Type V pinnately nerved on both sides of the central.This type is found in a few species of Pilea, in many species of Pellionia and all Procris species, plus a few species of Elatostema s.str.(for example, E. macrophyllum and E. rugosum).
In his work on the systematics of Chinese species of Elatostema, Wang (1980a) simplified the Schröter & Winkler (1935) classification of venation patterns (reproduced here as Fig. 2).
Cystoliths -Cystoliths are present in all species of the Urticaceae (Bigalke 1933) and their form and distribution have been found to be taxonomically useful for distinguishing infrafamilial groups, even if only as secondary characters.Within the large genus Elatostema, they are, at least, superficially relatively homogeneous.A detailed study of the taxonomic usefulness of cystoliths within Elatostema has not been included in this study; however, additional research is recommended.
Stipules -In Elatostema, the stipules are usually membranous, occasionally somewhat fleshy, and usually glabrous.They mostly vary from narrowly ovate (lanceolate) to almost linear (often referred to as linear-lanceolate), initially often shorter than mature stipules.Some large-leaved species (for example, E. fagifolium, E. nemorosum and E. smilacinum) have large stipules of 30-50 mm long.Elatostema sinuatum and related species possess short to minute stipules of 1-2 mm long (on macrophylls).

Reproductive features Inflorescence
The basic inflorescence structure of Urticaceae is cymose, either dichasial or monochasial (Golenkin 1894, Bernbeck 1932).Schröter & Winkler (1935) placed considerable taxonomic importance on the various types of inflorescence exhibited by the members of the tribe Elatostemeae.
Structure of inflorescence -The inflorescences vary from relatively loosely arranged clusters of flowers (for example, as found in subg.Elatostematoides and subg.Pellionia) to more crowded inflorescences (for example, E. velutinum and E. undulatum, both subg.Pellionia).The inflorescence of the former group lacks an involucre, whereas the latter group has a slightly developed involucre of bracts.Schröter & Winkler (1935) regarded the arrangement of flowers as a very important character for delimiting subgenera of Elatostema.
The male and female inflorescences are 'divided' into a series of compartments with the underlying cymose arrangement maintained (Weddell 1856, Schröter & Winkler 1935: f. 17, 29, 30).Although the latter regarded the female inflorescence of Elatostema vittatum (subg.Elatostematoides) as compartmentalised ('ordered') (their f. 17), we scored it as 'unordered' because the indistinct compartmentalisation meant that the scoring of this character was problematic.
Schröter & Winkler (1935: f. 18-28) discussed and illustrated the extent of fusion between involucral bracts.We did not use this character because it proved difficult to measure consistently.
Receptacle of inflorescence -The inflorescences of Elatostema subg.Elatostematoides, subg.Pellionia and the genus Pilea lack a receptacle (Schröter & Winkler 1935), whereas the receptacle is present for members of subg.Elatostema, subg.Weddellia and the female inflorescences of Procris (absent in male inflorescences of the latter genus).
As for several other morphological characters that Schröter & Winkler (1935) regarded as taxonomically useful, none of their inflorescence features uniquely describe any of the taxa listed in Table 1.There is considerable overlap between the characterstates of most characters used to distinguish these taxa.

Taxon sampling
The taxa of Elatostema used in this study represented each of the four subgenera of the genus recognised by Robinson (1910) and Schröter & Winkler (1935), namely, subgenera Elatostema, Elatostematoides, Pellionia and Weddellia.Species of Pilea and Procris (both Elatostemeae) were included, along with representatives of the Boehmerieae, Parietarieae and Urticeae.Boehmeria calophleba and B. macrophylla were used to root the trees.

Management of morphological data
Measurement of morphological characters were recorded and managed in DeltaAccess software, v1.9 (Hagedorn 2005), with data exported in DELTA-format.These DELTA files were then converted into NEXUS formatted files using the NDE software (Page 2001).These data were then loaded directly into PAUP* v4.0b10 (Swofford 2002) for phylogenetic analyses.

Choice and definition of characters
We use morphological characters to evaluate the morphological support for the monophyly of the infrageneric groupings in Elatostema proposed by Schröter & Winkler (1935, 1936).The choice of characters was based, in part, on the results of Robinson (1910) and Schröter & Winkler (1935, 1936), and, to a lesser extent, on research on other members of the Urticaceae.A list of the morphological characters used in this study is given in Table 2.All characters were treated as ordered and polarised to minimise parallel gain (Kitching in Forey et al. 1992).

Non-overlapping characters converted from overlapping characters
Even though overlapping morphometric characters may be useful in defining taxa, the PAUP* v4.0b10 software (Swofford 2002), excludes overlapping characters.There is an implicit assumption in phylogenetic analysis that character states must be discrete (Pimentel & Riggins 1987, Scotland in Forey et al. 1992) but the recognition of distinct character states in overlapping data is frequently difficult.Characters derived from overlapping morphometric data can be used in cladistic analysis (Ariyanti & Conn 2005).However, the transforming of such data into discrete, non-overlapping states has been controversial.One concern with coding overlapping data into discrete, nonoverlapping states is that this may result in an unacceptable level of data distortion (Chappill 1989).However, Thiele & Ladiges (1988), Thiele (1993) and Ariyanti & Conn (2005) maintain that morphometric data can provide useful information in phylogenetic analyses.
Divergence coding (Almeida & Bisby 1984) uses box plots (or 'box-and-whisker' plot sensu Tukey 1977) to represent the median, 1st quartile and 3rd quartile values to separate overlapping data into non-overlapping character states.A box plot provides a simple graphical summary of a set of data.It shows a measure of central location (the median), two measures of dispersion (the range and inter-quartile range), the skewness (from the orientation of the median relative to the quartiles) and potential outliers (marked individually).The lines extending from the box ('whiskers') indicate the range of values within 1.5× inter-quartile range (Wilkinson et al. 1992).We converted quantitative character states to discrete values to eliminate or, at least, minimise the degree of overlap between the character states in the following way: each character was divided into non-overlapping states by placing the inter-quartile range into a single discrete state.Occasionally, characters of some taxa could not be assigned to a single discrete character state.These were recorded as having more than one character state.Five collections of each taxon were used to generate the distributions for the box plots based on minimum and maximum values of each over-lapping character.The following characters were recoded using the divergence coding technique: plant height; petiole length; lamina length; lamina width; lamina length to width ratio; veins number; length of tepals in male flowers.The codified values for each of these characters are listed in Table 2.

Phylogenetic analyses
Ten uninformative characters were deleted prior to analysis.Heuristic searches were conducted in PAUP* using tree bisection reconnection branch-swapping restricted to 100 trees per replicate and 1 000 replicates of random taxon addition to search for multiple islands of most-parsimonious (MP) trees.Support for clades was estimated by decay analysis.These decay values were conducted in PAUP* with 100 replicates of random taxon addition on each constraint tree, using command files created in MacClade.The level of homoplasy was summarised by the Consistency Index excluding uninformative characters (CI-u), Retention Index (RI) and Rescaled Consistency Index (RC).
All characters were initially equally weighted so that the greatest degree of congruence between as many characters as possible would distinguish the useful characters from the misleading ones.Once the parsimony analysis was completed, the characters were a posteriori re-weighted according to their RC and, for comparison, their CI-u values, such that those characters with a higher 'parsimony fit' to a tree received a higher weighting than those with lower values.Subsequent analyses using CI-u and RC-weighted characters were conducted with branch-swapping limited to a maximum of 200 trees per replicate and 100 random taxon addition to search for multiple islands of most-parsimonious (MP) trees.

Phenetic analysis
Construction of a distance tree using clustering was calculated using the Unweighted Pair Group Method with Arithmetic Mean (UPGMA), producing a tree reflecting the phenotypic similarities between taxa.
The strict consensus tree (Fig. 3) is characterised by a large polytomy containing all members but one of the tribe Elatostemeae, with considerable character support (decay +3).Pilea nummulariifolia is placed as sister to the remaining taxa in the tribe (decay +2).This refutes the monophyly of Pilea, since P. microphylla is closer to members of Elatostema and Procris than it is to P. nummulariifolia.Neither of the other two genera appears monophyletic, although the analysis does not refute monophyly in either case.However, apart from P. reticulatovenosa, the remainder of Procris spp.are placed within a clade (P1; decay +1) indicating they are more closely related to each other than to Elatostema.There are two other clades of Elatostema species: i) clade E1 contains seven species from the Mt Kinabalu area of Sabah and receives some support (decay +2); ii) clade E2 that includes E. latifolium and E. tsoongii (decays +1).
Relationships between these three clades and the remaining members of the tribe are unresolved.Hence there is no support from the morphological database for any of the generic groupings within Elatostemeae.Pilea appears polyphyletic, but this estimate of relationships does not refute the monophyly of either Elatostema or Procris.
Within the non-Elatostemeae taxa, Dendrocnide sinuata and D. stimulans form a strongly supported clade (decay +7) with Urtica dioica sister to these two species (decay +1).Myriocarpa longipes (Boehmerieae) is sister to this latter clade, rather shows a close similarity to clade P3 (Fig. 4), containing all the taxa in the latter, plus clade E5 (Fig. 4) and two additional species, E. kinabaluense and E. lineolatum.
Pilea nummulariifolia clusters with species of the Boehmerieae, Urtica and Parietaria judaica.Species of Urtica are clustered with Parietaria judaica rather than with species of Dendrocnide.Myriocarpa longipes is clustered with species of Boehmeria rather than with Dendrocnide -Urtica dioica (Urticeae) as suggested by the parsimony analysis (Fig. 3).Hence the UPGMA analysis supports the current circumscription of Boehmerieae (based on species included in this study), but does not support the Urticeae.
Analysis of CI-u weighted data found 16 trees (RI = 0.69, RC = 0.22).The strict consensus tree of these (Fig. 6) shows much better resolution of relationships between species than the equally weighted one.It contains a number of clades that are similar to groups in the UPGMA analysis.Within the Elatostemeae there are two main clades labelled A and B (Fig. 6).Within clade A there are two terminal subclades that are similar to group A of the UPGMA analysis (Fig. 5), namely E1* and E4*.The topology and composition of the upper subclade (E4*) is similar to group D1 (Fig. 5) and also to clade E4 (Fig. 4) (containing all members of the latter plus three additional species, E. pedunculosum, E. reticulatum and E. stipitatum).
The lower subclade (E1*) comprises clade E1 (as resolved in Fig. 3) plus E. macrophyllum, and also with some species common with that of group D2 (Fig. 5).Likewise, the species composition of clade B (Fig. 6) is similar to clade B (Fig. 5), as well as to some elements in Fig. 3 and 4. All species of Procris are grouped within the terminal polytomy (P1*) of clade B (Fig. 6), with Procris reticulatovenosa nested among members of Procris pro parte majore clade (P1 of Fig. 3), and E. curtisii-E.sinuatum clade sister to this (also refer Fig. 4, 5).All the remaining eight taxa that attach to the base of the Procris  Since the strict consensus tree was poorly resolved, the majority rule tree (Fig. 4) was considered because it may provide additional information on potential relationships The UPGMA dendrogram (Fig. 5) recognises two major morphological groups (A and B) and several minor ones.Group D1 comprises all the species grouped in clade E3 (Fig. 4) plus five other species that are unresolved in Fig. 4. Group D2 consists of the clade E1 (Fig. 3, 4), E3 (Fig. 4), and E. papillosum.Group B clade in Fig. 4, and are also placed within clade B (Fig. 5), are attached at the base of clade B in Fig. 6.Hence, there is considerable similarity in the estimates of relationships and degree of similarity within the ingroup taxa from all three methods of analysis.Tribe Urticeae is polyphyletic in the CI-u weighted analysis, with Urtica urens placed outside the Urticeae clade and closer to Elatostemeae, as also shown in the parsimony analysis of the equally weighted dataset (Fig. 3, 4).However, Myriocarpa longipes (Boehmerieae) is nested within the main Urticeae clade (Fig. 6), or is placed sister to the main Urticeae clade (Fig. 3, 4).Hence neither analysis provides support for the inclusion of Myriocarpa longipes within Boehmerieae.Analysis of RC-weighted data (not presented here) found 71 trees of 60.80 steps (RI = 0.75, RC = 0.27).The topology of the strict consensus tree of the RC-weighted data was topologically similar to the CI-u weighted strict consensus tree.

dISCuSSIon
The low level of decay support for the clades within the equally weighted strict consensus tree indicates the weakness of the traditionally used morphological characters for indicating evolutionary affinities (Fig. 3).Even though there are several robustly supported clades in the strict consensus trees from the equally and CI-u weighted characters (Fig. 3 and 6, respectively), the RC-weighted tree (not presented here) and in the majority rule tree from the equally weighted analysis (Fig. 4), most of these branches are only supported by one or two character changes as shown in the phylogram of one of the MP trees from the equally weighted analysis (Fig. 7).

Evaluation of the subgeneric classification of Schröter & Winkler
The previous subgeneric classification of Schröter & Winkler (1935, 1936) is mapped onto the strict consensus trees of equally weighted (Fig. 8) and the CI-u weighted morphological analyses (Fig. 9).The following species, E. auriculatifolium, E. bullatum, E. dallasense, E. flavovirens, E. maraiparaiense, E. pinna-tivenium, E. purpurascens and E. serpentinicola have been here assigned to subgenera based on an examination of descriptions (Beaman & Cellinese 2004) and photographs of herbarium specimens (held at L) determined by R.S. Beaman.However, the subgeneric position of E. auriculatifolium is unclear because although the description suggests that this species may be a member of subg.Elatostema, based on the photograph of J. & M.S. Clemens 50984, (L565832 -possibly with male flowers) this species could belong to either subg.Elatostema or subg.Pellionia.Procris, recognised as a distinct genus by Schröter & Winkler (1935, 1936), has been included here for comparison.The lack of resolution in the equally weighted analysis means that there is little information relevant to the subgeneric concepts.together, and these appear to be more closely related to species of subg.Elatostema than to the other species of subg.
Weddellia.All members of subg.Elatostema are placed within clade A, except E. velutinicaule, which is placed at the base of the Elatostemeae-clade.
The inclusion of Pilea microphylla within the representatives of Elatostema in all analyses of the morphological data (Fig. 3 -9) does not support the monophyly of Pilea.However, a larger sample of species of Pilea and an expanded characterset are required before the phylogeny of this genus could be resolved.

ConCLuSIon
Parsimony analysis of the equally weighted morphological dataset provided very little resolution of relationships within Elatostema-Procris.The fit of the data to the resultant MP trees was very low (RC = 0.14), indicating a very high level of homoplasy in many characters.Parsimony analysis with reweighted characters greatly improved the resolution (Fig. 6).
We conclude that the groupings obtained from the CI-u and RCweighted parsimony analyses indicate the existence of some phylogenetic signal in these data, although this is weak and mostly obscured in the equally weighted parsimony analysis.
Based on the analyses of these morphological data, there is no support for the subgeneric classification of Elatostema by Schröter & Winkler (1935).The more recent infrageneric classification of Elatostema by Wang (1980a) into sections and series, has not been considered here.However, since Wang (1980a) more or less maintained the same taxonomic concepts as Schröter & Winkler's infrageneric classification, but   regarded their subgenera as sections, there is no support for his sect.Elatostema or sect.Weddellia.Since the monophyly of Elatostema subg.Pellionia is not supported in our analyses, the recognition of this taxon, but as the distinct genus Pellionia (Wang 1980b), is also not supported by the morphological data used here.Therefore, at least in part, there is no support for the sectional and generic classification of these above taxa as defined by Wang (1980a, b).Furthermore, based on the morphological characters analysed here, the taxonomic status of Pilea and Procris requires further review.A re-assessment of morphological characters and consideration of additional morphological features would be useful.It is recommended that the usefulness of morphological characters in evaluating the phylogeny of infra-generic taxa of Elatostema, together with those of other Elatostemeae, would greatly benefit from being considered within a molecular framework based on analyses of molecular data.Since greater homoplasy is often found in morphological data (Givnish & Sytsma 1997a), morphological data may be misleading compared to the large statistical advantage of molecular data (Givnish & Sytsma 1997b).Studies based on molecular data are more likely than morphological studies to provide accurate insights into phylogenetic relationships.

Fig. 4
Fig.4Majority rule tree, equally weighted morphological dataset; percentage of MP trees supporting each node is given above the line; grey lines = new clades formed on this cladogram that were not resolved on strict consensus tree (Fig.3); other coloured lines refer to clades recognised in strict consensus tree (refer Fig.3for explanation).
All but one representative of Procris are grouped in clade P1, but the representatives of subg.Elatostema, Elatostematoides, Pellionia and Weddellia are placed within the large polytomy.Hence the analysis neither supports nor refutes Schröter & Winkler's groupings.Comparison of the subgeneric classification with the re-weighted analysis is more fruitful (Fig. 9).All representatives of Procris are placed within clade P1*, and all representatives of subg.Pellionia attach as a grade below P1* within clade B, along with E. manillense (subg.Elatostematoides).Therefore, these data support the strong affinities of subg.Pellionia and Procris.Furthermore, these morphological data support the reduction of Procris to an infra-generic level within Elatostema as suggested by Hallier (1896).The two representatives of subg.Elatostematoides are widely separated within both clade A (E. rostratum) and clade B (E. manillense); there is no evidence of them being close relatives.The three species of subg.Weddellia (E.backeri, E. parvum and E. papillosum) are placed within clade A, but only two of these species (E.backeri and E. parvum) are grouped

Fig. 5
Fig.5UPGMA tree of equally weighted morphological dataset (refer Fig.3for further details); coloured lines and text are used to make it easier to distinguish the tribal groupings sensu Gaudichaud (1830) andFriis (1993).A, B, d1 and d2 = clades as referred to in text.

Fig. 6
Fig. 6 Strict consensus of 16 MP trees obtained from heuristic searching of the CI weighted morphological dataset; tree length = 89; RI = 0.69; RC = 0.22; coloured lines indicate clades that are also formed on the strict consensus of the equally weighted data (Fig. 3); labelled clades as referred to in the text.

Fig. 9
Fig. 9 Subgeneric classification (sensu Schröter & Winkler 1935, 1936) mapped on the strict consensus tree of CI weighted morphological dataset.Details of coloured lines are given in Fig. 8.

Fig. 7
Fig. 7 Phylogram (branch lengths proportional to amount of change) of one of the MP trees found from the parsimony analysis of the equally weighted morphological dataset.

Table 1
(Dengler 1999)onomically useful inflorescence characters based on usage bySchröter & Winkler (1935: 17 p.p.).The plant height measurements were based on personal field observations, herbarium notes provided by botanical collectors or, less frequently, the measurement of the herbarium specimen when the whole plant was represented by the collection.More commonly, a combination of these data sources was used.This character compares the width of the lamina on each side of central axis (often midvein).The comparison is made in the central 1/3 of the lamina.Leaves are regarded as having an asymmetric lamina (hence, unequal) if the width of broadest side is at least 1.34 times the width of the narrowest side.Leaves with equally symmetric laminas (equal) have a 'width' ratio of less than 1.34.Aniso-phylly and leaf lamina asymmetry are frequently associated(Dengler 1999).Leaf asymmetry is often present in the Urticaceae; in some species it is subtle, whereas in others the leaf lamina is distinctly asymmetric.All species of Elatostema are characterised by having strongly asymmetric (unequal) leaf laminas.The number of veins in the wider part of the lamina is used when the number of veins on each side of the central axis is unequal.Therefore, the number of veins is not always in pairs.List of morphological characters used in this study, together with the character states recorded for each character.Characters 2, 14, 16 -18, 26 and 52, are overlapping characters that have been converted to non-overlapping character states prior to phylogenetic analyses.
a. absent (cystoliths not present on interstices of adaxial surface) b. present (cystoliths present on interstices of adaxial surface) 36.leafabaxial indumentum venation a. absent (hairs absent from veins of abaxial surface) b. on primary, secondary and tertiary veins 37. leaf abaxial indumentum interstices a. absent (hairs absent from interstices of abaxial surface) b. on intersticesTable 2

Table 3
Morphological data matrix of 55 species and 76 characters (as listed in Table 2); '?' represents inapplicable characters or characters not available from specimens used; character state values in parentheses represent multiple character state scores.

.
Procris reticulatovenosa is placed sister to all other Procris spp.(clade P1) in 55 % of trees, suggesting that Procris may be monophyletic.Elatostema curtisii and E. sinuatum are grouped together with all Procris species (in 94 % of the trees) forming a clade P2.There is a suggestion (in 80 % of the trees) that clade P2 and several other members of Elatostema, namely members of subg.Pellionia, E. manillense (subg.Elatostematoides) and E. auriculatifolium (either a member of subg.Elatostema or subg.