2.1 Taxon sampling for phylogeny

We examined the same 169 accessions as for Wu et al. ([21]; q.v. for complete list), comprising 122 species from 47 genera of Urticaceae, representing 87% of the recognized genera [19]. The molecular data consisted of DNA sequences from two nuclear (ITS and 18S), four chloroplast (matK, rbcL, rpll4-rps8-infA-rpl36, trnL-trnF) and one mitochondrial (matR) loci.

2.2 Morphological characters and character state coding

Morphological information was mainly obtained from observations of plants in the field, and herbarium material (Studied at BM, E, K, KUN and PE); this was complemented with literature [17, 22, 27, 35, 36, 38–45]. All field work in this study was carried out on the mountains or periurban areas of China, where no specific permissions were required for these locations or activities, and no endangered or protected species were sampled. Therefore, the sampling in this study did not violate any law, rule or regulation in China and all around the world, requiring no ethical or institutional approval.

Ancestral character states were reconstructed for 19 morphological characters: habit, cystolith presence, cystolith form, stigma form, phyllotaxis, stipule presence, stipule form, stipule fusion, stipule position, pistillate perianth presence, pistillate perianth lobes fusion, achene symmetry, external morphology of achene, leaf venation apparentness, leaf venation—pinnate versus palmate, types of palmate venation, number of stamens, stinging hair presence and filament (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. The coding and scoring for the characters in this study.

https://doi.org/10.1371/journal.pone.0141821.t001

Regarding cystoliths, those that occur within the leaf blade or stem can be detected with a light microscope, but the character in two genera in Clade IV (Maoutia and Leucosyke) could not confidently be determined via light microscope due to their dense indumentum. Therefore we referred to spodogram-research by Bigalke [48], whose research using scanning electron microscope and stereo-light microscope determined that the cystoliths of Leucosyke, Maoutia and Gibbsia occur only in hairs or papillate hairs (cystolith-hairs) (Henk C. den Bakker, private communication), where they cannot be detected by light microscope. Therefore, cystoliths were coded as present only if they could be observed on the leaf blade or stem using a light microscope.

Regarding pistillate perianth lobes fusion, accessions were coded either as free (0) or connate (1), with the latter including lobes fused into a tube plus those fused only at the base. Regarding stigma, eleven types of stigma were identified by Chen [36], we detected ten types except for trifid-filiform among the taxa examined. Regarding venation, we referred to the description in the relative monographs [47, 49]. Regarding stipule form, as it is usually considered to be very important in the classification of section within Urticaceae [38], so we included this character analysis, which referred to the leaf description [46].

2.3 Ancestral state reconstruction

Considering that different analysis methods exhibit different advantages and limitations [50–54], we performed our analyses using both maximum parsimony (MP) and maximum likelihood (ML). Both methods were conducted in the program Mesquite v. 2.75 [55]. In the MP reconstruction, the character states were treated as unordered. We employed the Markov k-state 1 (Mk1) parameter model of evolution for the ML reconstruction, with equal probability for any particular character change. The data matrixes used for MP and ML analyses are presented in S2 and S3 Tables respectively.

3.1 Habit

The ancestral state for Urticaceae, and also for clade II+III is herbaceous (Fig 2). There have been at least five reversions to herbaceousness in Archiboehmeria, Chamabainia, subclades 1A5+1A6, 1C+1D, and Boehmeria of 1A1. Hemi-epiphytes independently arose at least two times within Urticaceae, including as clear synapomorphies for Poikilospermum and Coussapoa. Shrub habit has arisen independently at least six times within Urticaceae. Tree habit independently originated at least three times, i. e. in Dendrocnide, Cecropia + Myriocarpa and Gesnounia.

3.2 Cystolith presence

We found that cystoliths are present in all Urticaceae except clade IV (excluding cystoliths concealed within hairs; see methods), making this a probable synapomorphy for Clade IV (Fig 2). However, the ancestral state in Urticaceae is uncertain, because cystoliths are absent in the outgroups examined, hence the ancestral state could also be absent, with presence acquired independently in Clades I and II+III. The ML analysis gave a 56% probability for absent cystoliths to be ancestral within the family (see S6 Fig).

3.3 Cystolith form

Three different cystolith forms occur within Urticaceae (Fig 2), and they appeared nearly stable within each of clades I, II and III. The ancestral cystolith form is punctiform, and a switch to linear is an apomorphy for Clade II that is still present in all but six of the Clade II species examined. The exceptions are four Pilea species, two each with punctiform and virgate cystoliths, plus both Gyrotaenia species also with virgate cystoliths. Outside Clade II, only Urera alceifolia has linear cystoliths. Clade III also has two small clades with virgate cystoliths, plus several species each in Urtica, Urera and Poikilospermum that can have either virgate or punctiform cystoliths. Clade I shows no change from the ancestral punctiform cystoliths except for virgate cystoliths in some specimens of Soleirolia soleirolii.

3.4 Stigma form

We determined that filiform stigmas were ancestral in Clade I (S1 Fig), whereas penicillate stigmas were ancestral in Clades II, III and IV; either of these states might be ancestral for Urticaceae, with 62% and 38% probabilities respectively, according to ML analysis (S8 Fig). Transitions to several other states occur within all four major clades, so filiform stigma is not a good diagnostic trait for Clade I.

Capitate stigmas arose at least once in each of Clades I, II and III, although shifts to other states often seem to follow. Subulate stigmas evolved independently at least 3 times within Clade III, and appeared to be a synapomorphy of two genera: Girardinia and Laportea. The ligulate stigma has occurred at least six times across clades I, III and IV, and appears to be a synapomorphy of each of Archiboehmeria, Dendrocnide, and Poikilospermum. The peltate stigma independently originated at least three times within Urticaceae, and it can be considered a synapomorphy of each of Oreocnide and Cecropia. Both circular and oblong stigmas arose only once, in Sarcochlamys and Obetia respectively. The semilunar stigma was unique to the two Myriocarpa species within subclade (2D), whereas the spatulate stigma was unique to Touchardia.

3.5 Phyllotaxis

Only two states occur in Urticaceae, of which alternate is ancestral, and opposite arose at least seven times, most commonly within Clade I (Fig 2). In addition, there have been at least three reversions to alternate phyllotaxis in Sarcopilea, and some species of Boehmeria. Opposite leaves can be considered a synapomorphy of each of 3A (Urtica + Hesperocnide), Droguetia +Australina, Chamabainia and Gonostegia. This character therefore needs to be used cautiously; however there was only one origin of opposite phyllotaxy each within Clades II (clade 2F) and III (Urtica), so it can be a useful character within those clades.

3.6 Stipule presence

Stipules are present in all taxa except those of subclade 1C (Fig 2). Therefore stipule presence is ancestral in the family, and their loss is a unique and diagnostic synapomorphy for subclade 1C, i.e. a reduced Parietariaeae (Gesnouinia, Soleirolia, and Parietaria).

3.7 Stipule form

All six possible states arose more than once and hence show homoplasy (S2 Fig), but a few genera do have synapomorphic stipule forms without subsequent changes, making it a useful diagnostic character for these if used with others; such genera include Forsskaolea, Gonostegia, Dendrocnide, Lecanthus and Procris.

Lanceolate was inferred as the ancestral state for stipules within the family (S2 Fig), but numerous transitions to other states occur in Clades I and especially II and III, including within several genera. Only a few clades lack stipule form variation, notably clades IV and 1A1+2+3, both of which retain the plesiomorphic lanceolate state throughout.

Broad triangle stipules independently evolved at least seven times, and this state can be considered a synapomorphy of each of Nanocnide, Sarcopilea, Gonostegia, Forsskaolea, and Hemistylus. The narrow triangle was gained at least four times throughout the family: in the common ancestors of Pilea and of clades 1A4+5+6+7+8+9, (both with many subsequent changes and reversions) and twice within Elatostema. Linear stipules arose three times, but only in Clades II and III. It appears to be a synapomorphy of Obetia and Urtica, although with two subsequent changes to oblong stipules in Urtica. It also arose in clade 2A+2B, but with four subsequent changes of state within this clade. Oblong stipules likewise occur only in Clades II and III, where they originated at least five times: once within Pilea and twice within Urtica, and as apparent synapomorphies for Lecanthus and for Girardinia. Ovate stipules arose at least four times, but again only in Clades II and III. This character appears synapomorphic for Procris and Dendrocnide, and otherwise arose within Elatostema and Pilea.

Overall, stipules are far more plastic within clades II+III (21 changes inferred, six possible states) that I+IV (7 changes inferred, three possible states).

3.8 Stipule fusion

The ancestor of Urticaceae is indicated to have had unfused (free) stipules (Fig 2) by MP analysis; this is congruent with the ML analysis (S12 Fig). We detected at least 18 switches within Urticaceae, including the common ancestor of Clades III, and those of many small clades and several genera. These included at least 12 switches to the fused stipule state, plus one clear independent origin in Clade IV, with at least five incidences of revision to free stipules. The fused state appears synapomorhic for Procris, Clade 2E+2F (Lecanthus+ Pilea+Sarcochlarmys), Clade IV, Debregeasia, and Archiboehmeria. Only one genus, Urtica, was variable for this character. Furthermore, the amplexicaul stipule was unique to subclade 4A, (Fig 2), making it a clear synapomorphy for this subclade.

3.9 Stipule position

Stipule position is very closely related to stipule fusion, with almost all taxa that have free stipules also having interpetiolar stipules (including the inferred common ancestor of the family), whereas fused stipules are normally intrapetiolar (Fig 2). The sole exception is Urtica, wherein all stipules are interpetiolar but both fused and free stipules occur, making fused interpetiolar stipules a trait unique to one clade within Urtica. Hence these two characters are tightly linked in their evolution, and should be treated together.

3.10 Pistillate perianth presence

Pistillate perianths are present almost throughout Urticaceae, and were clearly ancestral (Fig 2). Loss of pistillate perianth occurred only twice, and appears therefore to be a synapomorphy for each of Phenax and Forsskaolea.

3.11 Pistillate perianth lobes fusion

Free and connate tepals are ancestral to Clades II+III and I+IV respectively, and are almost diagnostic for these clades (Fig 2). The MP analysis could not determine that ancestral state for the family, but ML analysis gives a 80% likelihood for connate to be the ancestral state (S15 Fig). The lobes are free almost throughout clades II and III; although six switches to connate have occurred they affect few species: the two Myriocarpa species in Clade II, and in Clade III the monotypic Hesperocnide, two Urera spp, and the genus Poikilospermum. Only a single change has occurred in Clade I, wherein Sarcochlamys is unique in having free tepals.

Despite the exceptions noted, this character is the closest we have found to a diagnostic trait distinguishing the two sister clades that comprise Urticaceae.

3.12 Achene symmetry

Straight achenes are ancestral in Urticaceae, and in Clades I, II and IV. These clades contain respectively one (Pilea), two (Sarcochlamys and Soleirolia) and one (Leucosyke) genera in which a change to the oblique achene state appears to be a synapomorphy (Fig 2). Two reversions have also occurred within Pilea.

Oblique achenes are also a synapomorphy for Clade III, within which three reversals to the straight state have occurred, i.e. Touchardia, Poikilospermum, and the common ancestor of Nanocnide and Urtica.

3.13 External morphology of achene

The ancestral state for external achene morphology in Urticaceae was smooth-and-dull or tuberculate (S3 Fig); these two states occurred across all four major clades I-IV. ML analysis gives a higher probability for smooth-and-dull (80%) as the ancestral state (S17 Fig). Throughout the family, at least 40 changes of state (including reversals) were detected, the highest for any character examined. Three other achene states were widespread within the family: verrucose, smooth-and-shiny, and reticulate. Transititions between smooth-and-dull and verrucose were particularly common, with large clades containing a mixture of these states occuring within each of Clades I, II and III. Smooth-and-shiny achenes were gained once in Clade III (Touchardia) and two or three times in Clade I. Reticulate achenes arose once in each of Clades I (Oreocnide), II (Gyrotaenia) and IV (Coussapoa) and are hence diagnostic for those genera within those clades.

The remaining states, ribbed and linolate, occurred only in Clade II. Ribbed achenes appear synapomorphic to Elatostema, although switches to three different states then occurred within this genus (linolate, smooth, tuberculate). Otherwise, linolate achenes occur only in Procris (a likely synapomorohy) and a single species of Pellionia.

Nine genera exhibited within-genus variation for external achene morphology, most notably Elatostema, in which four of the eight different states observed by [27] were detected in our more limited sample.

3.14 Leaf venation apparentness

Inapparent leaf venation evolved only once within the family (Fig 2), so it can be considered as an unique synapomorphy of Sarcopilea.

3.15 Leaf venation—pinnate versus palmate

The ancestral state is palmate venation (Fig 2). Pinnate venation independently originated at least seven times, once in Clade IV and several times each in Clades II and III, often within genera. Pinnate venation is completely absent from Clade I, but can be considered as synapomorphies of Procris, Dendrocnide and Coussapoa. Genera that contain both states are Urera, Elatostema, Myriocarpa and Pellionia.

3.16 Types of palmate venation

Trinerved palmate venation is clearly ancestral within Urticaceae, and the only state present in Clade I (S4 Fig). Semi triplinerved venation arose only once, as an apomorphy for Clade 2A+2B+2C, although reversions and switches within Elatostema limit the diagnostic value of this character for the clade. Three Elatostema species were plastic for palmate venation type, with two containing both trinerved and semi triplinerved venation, whereas Elatostema tenuicaudatum var. tenuicaudatum contained both trinerved and triplinerved venation, the only instance of the latter character detected in the whole family.

Outside of clade 2A+2B+2C, all palmate-nerved species are trinerved except for a handful of transitions to >3 nerves, in a subclade of Urtica, Obetia (a possible synapomorphy), and most of Clade IV. This could be a synapomorphy for Clade IV with reversal, or it could have arisen twice within that clade. Overall, this character exhibits more changes within Clade 2A+2B+2C than among the rest of the family put together.

3.17 Number of stamens

The ancestral stamen number of more than one has been retained in most Urticaceae, the sole exception being subclade 1D (tribe Forsskaoleae), for which a single stamen is hence a unique synapormophy (Fig 2).

3.18 Stinging hairs presence

Stinging hairs were absent in the common ancestor of Urticaeae, and occur only in Clade III, for which they are clearly a synapomorphy (Fig 2). However, there were also two clear losses of this character (reversions), in Poikilospermum and Touchardia.

3.19 Filament

Inflexed filament was reconstructed as ancestral for Urticaceae, and occurred in all but four genera of the family (Fig 2). The exceptions were those formerly placed in Cecropiaceae, i.e. Poikilospermum (Clade III) plus the three genera comprising sublcade 4A of Clade IV. These had straight filaments, which hence evolved at least twice within Urticaceae.

4.1 Patterns of character evolution within Urticaceae

Morphological classification of Urticaceae was traditionally based on a series of key diagnostic characters including habit, cystolith form and presence, stigma, phyllotaxis, stipule presence, form and position, pistillate perianth, achene, leaf venation, stamens, stinging hairs, filament, hook hairs and so on [17, 22, 27, 33–35, 38, 56], A few other characters, like achene symmetry, have been largely overlooked. Our ancestral state reconstructions indicated that most of these traits were not good tools to classify the investigated taxa at higher levels, because most were highly homoplastic; e.g. stigma and stipule form exhibited 29 and 28 transitions respectively (S1 and S2 Figs), whereas achene surface underwent 40 transitions (S3 Fig). Hence it is unsurprising that morphology-based classifications of the genus have disagreed, both with each other and with molecular phylogenies, about the placement of certain genera.

Most characters varied independently of one another, with one exception: stipules were almost always either interpetiolar and free, or intrapetiolar and fused, despite at least nine switches for each state (Fig 2). The only exception to this pattern is a clade of four Urtica species (ardens, fissa, mairei and zayuensis), which uniquely have fused interpetiolar stipules. Perhaps surprisingly, these characters show no obvious correlation with phyllotaxis.

Nine of the 19 characters examined exhibited more than 10 state changes, and of these four were much more plastic within Clades II+III than in I+IV (S1 Table), whereas only habit and phyllotaxis had more transitions in Clades I and IV than II and III (Fig 2, S1 Table). Furthermore, 12 and 15 different characters were variable within Clades II and III respectively, whereas only nine were variable within Clade 1A, which though a subclade of Clade I is larger than II or III (S1 Table). This could indicate that morphological change has been slower in Clade 1A than elsewhere, that it is younger, and/or that speciation within it has been more rapid than elsewhere, in such a way that morphological change has not kept pace.

4.2 Taxonomic implications for tribes and subfamilies within Urticaceae

4.3 Large, paraphyletic or problematic genera within Urticaceae

Looking at the larger genera within Urticaceae, taxonomic difficulty in Clade I might reflect relatively few useful characters, whereas in Clades II and III difficulties might reflect too much homoplasy and reversals. Of those in Clade 1A, only two characters (habit and phyllotaxis) were variable within Boehmeria, whereas external achene morphology was the only variable character within Debregeasia. By comparison, Elatostema (Clade II) has four variable characters, whereas Pilea (Clade II, 2F) and Urtica (Clade III, 3A) have four and five respectively even if derived genera Sarcopilea and Hesperocnide are excluded. Lastly, Urera (Clade III) is highly paraphyletic and the clade containing it (3F) is variable for 12 different characters.

Pilea and Urtica become monophyletic if Sarcopilea and Hesperocnide are included, respectively. Sarcopilea has three apomorphies—alternate phyllotaxy, inapparent venation and straight achene, of which only the last has arisen independently in other Pilea; it also has a dramatically different distribution (Caribbean, as opposed to Old and New world tropics for Pilea; see Table 1 in [21]). However making all genera in Clade 2F monophyletic would divide Pilea into at least four genera if Sarcopilea isn’t included, so we suggest that Sarcopilea domingensis is best treated as an aberrant species of Pilea, whose novel traits likely result from founder effect following long dispersal. By contrast, Hesperocnide differs from Urtica only in connate perianth lobes, which given the other variation within Urtica noted above makes uniting these genera less problematic.

A polyphyletic Boehmeria is spread across subclades 1A1 and 1A3 with a single species in 1A8. Only two of the nineteen characters examined (habit and phyllotaxy) distinguish any of these species from one another, yet both of these vary within subclades and even within species. Conversely, a homoplasy for smooth achenes (S3 Fig) and a reversal to lanceolate stipules (S2 Fig) make B. rugulosa (Clade 1A8) resemble other Boehmeria species. Hence none of the characters we examined are at all useful in separating Boehmeria into segregate monophyletic genera, explaining why this has never been done before. Therefore, Boehmeria calls for much more detailed investigations in the future.

Pouzolzia Gaudich. is even more polyphyletic/paraphyletic: its members form four distinct lineages, two each within Clades 1A5 and 1A7, meaning their common ancestor is that of Clades 1A5+6+7+8+9, shared with eight other genera. Among characters examined, it appears that Pouzolzia is defined by two that are plesiomorphic within this clade, i.e. narrow triangle stipules (S2 Fig), and free interpetiolar stipules (Fig 2); species or clades that have altered states for one of both characters are placed in other genera. That said, the characters examined provide only limited material for a revision of this clade. Clades 1A7 and 1A8 can be separated by stipule fusion and position; stipule form would be similarly useful but for homoplasy. However clade 1A5 completely lacks detected apomorphies, and this makes a taxonomic overhaul of this group difficult. Possibly a further study similar to this one but focused only on this clade, looking at more characters, might be useful.

Relationships among Elatostema, Pellionia and Procris (clade 2A+B+C here) have been controversial for over a century [20, 21, 27, 47, 57–60]. Elatostema and Procris as currently subscribed are monophyletic [21], whereas Pellionia is paraphyletic with respect to them both, becoming monophyletic only if P. repens and P. tsoongii are moved to new, segregate genera. Among these, Procris can be distinguished by fused, intrapetiolar stipules. Four other characters vary within this group, i.e. stipule form, achene exterior and two venation characters (S2, S3, and S4 Figs), but all of these vary at least as much within Elatostema as they do outside it. Hence our results provide little data to resolve the taxonomy of this clade. A further more detailed morphological survey focused on this clade, and including additional characters such as morphology of bract and bracteoles [61], might have more success.

A third highly paraphyletic genus is Urera, which forms three distinct subclades of Clade 3F, each from a different biogeographic region (Africa, the Americas, Hawaii; see Fig 1 and Table 1 in [21]). The other three genera in this clade all have apomorphies unique within this clade: shiny smooth achenes and spatulate stigma for Touchardia, straight filament for Poikilospermum, and circular stigma, >3-nerved leaves and linear, fused interpetiolar stipules for Obetia. However, although six characters are variable within Urera, not one of these distinguishes the three subclades of the genus, whereas all six vary within the clade that is sister to Poikilospermum. Hence Urera appears to be a genus somewhat defined by characters that are plesiomorphic within Clade 3F, but other than biogeography we could not find defining characters that would be useful for breaking up Urera into monophyletic segregates.

One further pair of non-monophyletic genera are Myriocarpa and Gyrotaenia, whose species are mixed together in Clade 2D. However, these genera are separated by five of the traits examined, the two Myriocarpa species examined having apomorphies for virgate cystoliths, fused intrapetiolar stipules, connate female perianth lobes (unique in Clade II), and smooth-and-dull achenes. The only other character that varies within Clade 2D is leaf venation type, which distinguishes Myriocarpa cordata from all others. It is very unlikely that four out of five apomorphies (treating the stipule characters as one) within this clade should have occurred independently in two distinct lineages. Instead, the apparent non-monophyly of Myriocarpa might be a signature of some of reticulate evolution, possibly indicating a role for hybridization in this clade.

4.4 Adaptive evolution of characters in Urticaceae

Many of the characters examined here are likely adaptive, and hence might have responded to shifts in habitat or vegetation type [31]. Within Urticaceae, most species occur either in forests or at their margins [17], with many preferring wet places, along streams or under shade in the tropics. It is well known that convergent evolution may produce similar phenotypes in different geographically isolated regions [62], with similar conditions selecting for the same traits [63], and this could explain some of the homoplasy seen within this family. Competitive interactions might further drive switches to both homoplasious and novel character states [64]. Furthermore, some Urticaceae are pioneer species, and morphological changes could accompany switches between such habitats.

Two apparently adaptive traits that have been lost in certain lineages are stinging hairs and inflexed stamens. Stinging hairs occur in four Eudicot families (Urticaceae, Euphorbiaceae, Loasaceae, and Hydrophyllaceae [65]. Most research on stinging hairs has focused on their toxicology [37, 38, 45], ontogeny and morphology [37], whereas there has been relatively little on their evolution (but see [66]). They are affected by grazing pressure [66], and possibly also biogeographic context and competition [62]. Both instances of stinging hair loss in Urticaceae are associated with long dispersal: into SW China for Poikilospermum and into Hawaii for Urera glabra + Touchardia. For Touchardia, the loss is likely an adaptive response to the absence of native herbivorous mammals in Hawaii, but for Poikilospermum it could reflect a chance loss of key genes due to a founder effect, or period of low herbivore numbers at the time.

Stamens that are inflexed in the bud can explosively expel pollen by bending outwards elastically and suddenly during anthesis [31]. This is likely adaptive for wind-pollination, especially in understorey environments with limited air movement, where most Urticaceae grow. This trait is shared with the majority of Moraceae and apparently also Celtis tetrandra outside of Urticaceae, so the inflexed filament character may be older than the family Urticaceae. We detected independent losses in the two clades comprising former Cecropiaceae: Poikilospermum and Cecropia+Myrianthus+Coussapoa. A possible explanation for this loss is that all these three genera are tall lianas or trees, wherein elevated flowers can get more wind, which might remove the need for explosive pollen expulsion.

Overall, character evolution in Urticaceae must have been affected by a variety of ecological, genetical, and developmental factors as well as biogeography and chance events like dispersal [31, 67]. Research in all these areas will help untangle the complex causes involved, whereas a dated phylogeny might uncover instances of rapid adaptive radiation.