American Journal of Botany 89(5): 865±874. 2002.

PATTERNS OF DIVERSIFICATION IN NEW ZEALAND STYLIDIACEAE1

STEVEN J. WAGSTAFF2,4 AND JULIET WEGE3

2Landcare Research, P.O. Box 69, Lincoln, New Zealand 8152; and 3Herbarium, Royal Botanic Gardens Kew, Richmond, Surrey TW9 3AB, UK

Phylogenetic analysis of ITS and rbcL sequences show that New Zealand fall into two distinct lineages differing in species richness. Each lineage represents a unique dispersal event to New Zealand occurring at different times during the evolutionary history of the family. One lineage comprises seven species of and , while the other consists solely of subulatum. The origin of the Forstera/Phyllachne lineage in New Zealand is equivocal; either a South American or a Tasmanian origin is equally parsimonious. Possible sister groups are F. bellidifolia in and P. uliginosa in South America. Oreostylidium subulatum has an Australian origin. In our analyses O. subulatum is nested in a clade composed entirely of species of , almost all of which are endemic to . Species of Phyllachne share a cushion habit with the outgroup (Donatiaceae) that may have preadapted them to alpine environments in New Zealand. The New Zealand Stylidiaceae have small, white, actino- morphic ¯owers that are well adapted to the unspecialized pollinator fauna. Forstera and Phyllachne share this trait with Donatia; however, the small, white ¯owers of Oreostylidium are a dramatic departure from the colorful, highly specialized ¯owers of Stylidium.

Key words: diversi®cation; DNA; ITS; New Zealand; phylogeny; rbcL; Stylidiaceae.

New Zealand has been isolated by the southern oceans for species in the family; the majority (92%) are placed in the at least the last 80 million years, hence colonization by small Stylidium, while the remainder belong to founder populations must have played an important role in the (Table 1). The southwest of Western Australia is a center of evolution of the ¯ora (Carlquist, 1974; Pole, 1994). Dispersal diversity for both Stylidium and Levenhookia. across an oceanic barrier is an infrequent event that geograph- The Stylidiaceae are characterized by a central ¯oral ically isolates new founders from their mainland progenitors. (or gynostemium) that bears the stigma and anthers at the Once established the new immigrants must overcome the del- apex. The column functions ®rstly to shed pollen on an insect eterious effects of an initial genetic bottleneck. With the var- and secondly, when the stigma becomes receptive, to pick up iability of their mainland relatives no longer accessible, ran- pollen that an insect may have been carrying. In Levenhookia, dom mutation and recombination are the only sources of new the ¯owers possess a modi®ed ®fth corolla lobe (the labellum) variation to rebuild a depauperate gene pool. According to that is hooded over the column, which is released when it is Lloyd (1985), with contrasting traits need not be equally stimulated by an insect visitor or by the growth of the column successful in colonizing islands. Many of the unique features (Erickson, 1958). In Stylidium (the trigger plants), the column of the New Zealand ¯ora probably re¯ect the nonrandom suc- is held under tension and ``triggers'' in response to an insect cess of different individuals in reaching, becoming established, probing for nectar, subsequently depositing pollen on the in- and diversifying there. sect before gradually resetting. In order to accommodate this The trigger family (Stylidiaceae) is comprised of ®ve movement the labellum is re¯exed and reduced in size. Trigger genera and over 240 species that are native to Australia, South- plant ¯owers are morphologically diverse, with variation in east Asia, New Zealand, and South America (Fig. 1). The sub- the size and shape of ¯owers, presence and form of corolla alpine and alpine zones in New Zealand are a center of di- appendages, in¯orescence structure and corolla color combin- versity for Phyllachne and Forstera, two of the ®ve genera ing to produce an array of ¯owers of contrasting appearance. that comprise the family. Seven of the nine species belonging In contrast to Stylidium, and to a lesser degree Levenhookia, to these genera are found there, with all but P. colensoi en- Phyllachne, Forstera and Oreostylidium possess simple, white, demic. The monotypic genus Oreostylidium is also restricted actinomorphic ¯owers with immobile columns. to montane and subalpine zones of New Zealand. These three The debate surrounding the origins of the subalpine and genera, however, comprise only 4% of the total number of alpine ¯ora of New Zealand has centered on the relative im- portance of long-distance dispersal. While explaining the dis- 1 Manuscript received 2 August 2001; revision accepted 11 December 2001. tribution of some elements of the ¯ora in terms of dispersal, The authors thank the curator of the CHR herbarium for permission to Wardle (1968, 1978) proposed that New Zealand's subalpine isolate DNA from specimens; Andy Glazier and Anita Thorne for assistance and alpine ¯ora arose largely through diversi®cation during with the DNA sequencing; KaÊre Bremer for sharing DNA samples of Styli- the late Tertiary. Wardle (1968) suggested that Phyllachne, diaceae and his careful review; Adrienne Markey and David Glenny for ®eld assistance; Christine Bezar, Rob Smissen, Peter Wardle, and an anonymous Forstera, Oreostylidium, and Donatia (the hypothesized sister reviewer for insightful comments on earlier drafts; and Kirsty Cullen and group to Stylidiaceae) are examples of seemingly ancient gen- Rebecca Wagstaff for preparing the distribution map and illustrations. The era that existed in Antarctica and dispersed to New Zealand molecular research was funded by the Foundation for Research, Science and via Tasmania and the subantarctic islands. The ancestors of Technology contract C09618. Additional research was conducted as part of these groups were suggested to have survived in peneplained an Anglo-Australian Postdoctoral Research Fellowship, supported by the Of- ®ce of Science and Technology and administered by the Royal Academy of uplands with soils too leached to support continuous forest, Engineering. before evolving into the subalpine and alpine habitats ®rst cre- 4 Author for reprint requests (e-mail: [email protected]). ated in the late Pliocene. In contrast, Raven (1973) argued that 865 866 AMERICAN JOURNAL OF BOTANY [Vol. 89

and were included within the family in some classi®cations (Mildbraed, 1908). However, on the basis of a number of morphological, anatomical, and em- bryological features (including the absence of the ¯oral column) the Donati- aceae are now considered a separate family (Rapson, 1953; Philipson and Philipson, 1973; Gustafsson and Bremer, 1995) and sister to Stylidiaceae (Gustafsson and Bremer, 1995; Laurent, Bremer, and Bremer, 1999). Voucher information and GenBank accession numbers are listed in the Appendix stored at (http://ajbsupp.botany.org/v89/wagstaff.pdf). The complete data sets are available upon request from the ®rst author and were deposited in TreeBASE (http://www.herbaria.harvard.edu/treebase).

Molecular dataÐOur molecular sampling strategy capitalized on the unique characteristics of rbcL and internal transcribed spacer (ITS) sequences. The gene rbcL is evolving at a relatively slow rate allowing sequence com- parisons among distantly related outgroups, and a large number of published sequences are available for comparison (see Chase et al., 1993; KaÈllersjoÈet al., 1998; and references therein). Finally, Albert et al. (1994) and Bremer and Gustafsson (1997) suggest that rbcL approaches clock-like behavior in its evolution, so that the amount of sequence divergence may be used to estimate the timing of evolution. By comparison, the ITS region is evolving Fig. 1. Generalized austral distribution of Stylidiaceae. Fossil evidence more rapidly than rbcL and provides more informative characters to resolve suggests Antarctica was vegetated until the Pliocene and may have been a corridor for migration between Austalia, New Zealand, and South America. relationships at lower taxonomic levels (Baldwin et al., 1995). Total DNA was extracted from either fresh leaves or leaf fragments dried with silica gel using a modi®cation of the hot CTAB (hexadecyltrimethyl- much of the New Zealand ¯ora arrived via Australia. The col- ammonium bromide) method of Doyle and Doyle (1987). The gene rbcL and lision of the Australian plate with the Asian plate resulted in the ITS region (the 3Ј end of the 18S rDNA gene; ITS-1; the 5.8S rDNA the uplift of mountains in Malaysia, Australia, and New Zea- gene; ITS-2; and the 5Ј end of the 28S rDNA gene) were ampli®ed by poly- land during the late Pliocene and Pleistocene. According to merase chain reaction (PCR). Primer sequences and our ampli®cation and Raven, the uplift of the mountain ranges created suitable hab- sequencing techniques follow Olmstead et al. (1992) for rbcL and Wagstaff itats for the migration of subalpine and alpine plants between and Garnock-Jones (1998) for the ITS region. Excess primers and unincor- porated nucleotides were removed from the PCR products by spin column Asia and Australia. In view of their obscure phyletic relation- centrifugation (QIAquick PCR puri®cation kit; QIAGEN, Clifton Hill, Vic- ships, Raven also considered Phyllachne, Forstera, and Don- toria, Australia). The puri®ed DNA samples were then labeled with Big Dye atia to be examples of genera of great antiquity. Recent mo- terminators (PE Applied Biosystems, Perkin-Elmer, Sydney, New South lecular studies on the New Zealand subalpine and alpine ¯ora Wales, Australia). After this, unincorporated dye terminators were removed con®rm that intensi®ed speciation occurred during times of by alcohol precipitation in the presence of 3.0 mol/L sodium acetate pH 5.2. marked climatic and geologic perturbation during the Pliocene The sequencing was performed at the Waikato University DNA Sequencing (Winkworth et al., 1999). Our aim was to reconstruct phylo- Facility. Both the forward and reverse DNA strands were sequenced to min- genetic relationships among the New Zealand lineages of Styl- imize errors and to con®rm our results. idiaceae, to estimate the timing of diversi®cation, and to ex- The sequence alignment for the ITS region was facilitated by ClustalX plore ecological and adaptive morphological features that (Thompson et al., 1997), and gaps were inserted to ensure positional homol- could account for the differences in species richness. ogy. A gap penalty setting of 75 and a gap extension penalty of 6.6 were initially used to identify and position large gaps in the sequence data. Then MATERIALS AND METHODS low-scoring segments were realigned using a gap penalty setting of 15 and a gap extension penalty of 6.6 with the removing new gaps option turned on. Study groupÐData were obtained for 16 representatives of Stylidiaceae. These settings opened and positioned small gaps. These included members of all ®ve genera and all of the described species from New Zealand. Data were also obtained for the two species that comprise Data analysisÐThe phylogenetic analyses were accomplished using Donatiaceae: D. novae-zelandiae in New Zealand and D. fascicularis in South PAUP* version 4.0d65 (Swofford, 1998), and MacClade version 3.04 (Mad- America. The Donatiaceae were generally considered close to Stylidiaceae dison and Maddison, 1992) was used to explore character evolution. The

TABLE 1. Present-day distributions and species numbers for the family Stylidiaceae. The number in the parentheses represents the number of species present in the region, while the percentage represents the level of endemicity.

Genus No. of species Distribution and levels of endemicity Literature citations Forstera 5 New Zealand (4; 100%), Tasmania (1; 100%) Allan, 1961; Curtis, 1963 Levenhookia 10 Southwest Australia (8; 75%), northern Australia (1; Curtis, 1963; Erickson, 1958; Erickson and 100%), mainland southeast Australia (3; 33%) Tasmania Willis, 1966; Western Australian Herbari- (1; 0%) um, 1998 Oreostylidium 1 New Zealand (1; 100%) Allan, 1961 Phyllachne 4 New Zealand (3; 66%), Tasmania (1; 0%), South America Erickson, 1958; Allan, 1961; Curtis, 1963 (1; 100%) Stylidium 221 Southwest Australia (151; 98%) northern Australia (57; Bean, 1999, 2000; Curtis, 1963; Erickson, 96%) mainland southeast Australia (13; 69%), Tasmania 1958; Western Australian Herbarium, 1998; (5; 0%), Malesia (5; 40%), mainland southeast Asia (3; Raulings, 1999; Raulings and Ladiges, 2001 33%), Sri Lanka (1; 0%) May 2002] WAGSTAFF AND WEGEÐDIVERSIFICATION OF NEW ZEALAND STYLIDIACEAE 867

Fig. 2. Comparison of the maximum parsimony trees obtained from rbcL and ITS sequences. Bootstrap values are given above each node and jackknife values below. analyses were conducted using the PAUP* settings random addition with ten to evolve in a clock-like manner because they are more likely replicates, tree-bisection-reconnection (TBR) branch swapping, MULPARS in to be selectively neutral. effect, and steepest descent. The characters were all unordered and weighted A heuristic search of the rbcL sequences recovered only one equally, and gaps were treated as missing data. The rbcL and ITS data sets maximum parsimony tree in a single island of 259 steps (con- were analyzed separately to explore possible con¯ict. In the absence of con- sistency index ϭ 0.757; retention index ϭ 0.866; rescaled con- ¯ict, the data sets were then combined. The topological constraint option in sistency index ϭ 0.656; excluding uninformative characters) PAUP* and the branch moving option in MacClade were used to assess the (Figs. 2, 3). The rbcL tree branches fairly symmetrically with degree of congruence among competing hypotheses of relationships. Support for the inferred clades was estimated by jackknife (Farris et al., two notable clades of Stylidiaceae being supported. The 1996) and bootstrap (Felsenstein, 1985) analyses. The searches were per- branch leading to Levenhookia diverges at the base, but this formed with 1000 replications excluding uninformative sites; the starting trees relationship is only weakly supported (bootstrap 50%; jack- were obtained by random addition with one replication for each replication, knife 56%). The ®rst clade consists of Stylidium and Oreos- TBR branch swapping, and MULPARS in effect. tylidium and is only weakly supported (bootstrap 67%; jack- The relationship between sequence divergence and time for the gene rbcL knife 77%). Stylidium calcaratum diverges at the base of the has been discussed by Albert et al. (1994) and Bremer and Gustafsson (1997) ®rst clade, and the remaining members form a well-supported and was calculated using the equation: substitution rate ϭ patristic distance clade (bootstrap 96%; jackknife 99%) with O. subulatum nest- (Dp)/number of nucleotides/inferred time since cladogenesis. ed well within this group. The second clade is well supported (bootstrap 99%; jackknife 100%) and consists of a heteroge- RESULTS nous mix of species of Forstera and Phyllachne though neither The rbcL sequences included in our analysis were 1402 nu- of these genera emerge as clades. Forstera bellidifolia is weak- cleotides long (corresponding to positions 27±1427 in tobac- ly supported (bootstrap 49%; jackknife 55%) as sister to the co). Missing data accounted for 4.9% of the matrix. Of the remaining members. Nested within the second clade is a well- 1402 total sites in the rbcL matrix, 1214 sites were constant, supported group (bootstrap 90%; jackknife 93%) that includes 89 variable sites were parsimony-uninformative, and 99 sites P. colensoi and F. tenella. Phyllachne rubra and P. clavigera were parsimony-informative characters. Transitions (128) also are well-supported sisters (bootstrap 87%; jackknife 90%). occurred more frequently than transversions (67); the ratio was The ITS sequences were 782 nucleotides long with gaps about 1.9. Most changes (calculated across the maximum par- inserted to assure positional homology. Missing data and gaps simony tree shown in Fig. 2) occurred in the third codon po- accounted for 16.4% of the matrix (0.5 and 15.8%, respec- sition (202), while progressively fewer occurred in the ®rst tively). The ITS-1 spacer varied in length from 164 nucleo- (38) and second (19) positions. Changes in the third codon tides in Stylidium calcaratum to 251 in Donatia, whereas the position are generally synonymous and hence are more likely ITS-2 spacer varied from 209 in Levenhookia leptantha to 281 868 AMERICAN JOURNAL OF BOTANY [Vol. 89

Fig. 3. Maximum parsimony tree of 259 steps obtained from the rbcL sequences (consistency index ϭ 0.757; retention index ϭ 0.866; rescaled consistency index ϭ 0.656; excluding uninformative characters). The New Zealand members of the Stylidiaceae are encompassed with brackets. Approximate divergence times given as million years before present (MYBP) are based upon the rate calculated for rbcL by Bremer and Gustafsson (1997). Branch lengths are proportional to the number of nucleotide substitutions with a scale bar below. in S. emarginatum. Of the 782 total sites in the ITS matrix, minals was 57.9 Ϯ 14.26, whereas the mean number of sub- 458 sites were constant, 85 variable sites were parsimony-un- stitutions in ITS was 163.3 Ϯ 18.9, which is nearly a threefold informative, and 239 sites were parsimony-informative char- difference in the substitution rate. The rbcL data also exhibited acters. substantially more rate variation across lineages than the ITS A heuristic search of the ITS sequences similarly recovered data (Figs. 3, 4); the standard deviation around the mean was only one maximum parsimony tree in a single island of 670 proportionally higher in the rbcL data. The rate difference be- steps (consistency index ϭ 0.685; retention index ϭ 0.809; tween the Forstera/Phyllachne and the Stylidium/Oreostyli- rescaled consistency index ϭ 0.554; excluding uninformative dium clades was particularly striking; however, the sequences characters) (Figs. 2, 4). The slightly lower consistency, reten- within each of these clades appeared to be evolving in a rel- tion, and rescaled consistency indices indicate more homopla- atively clock-like manner. sy in the ITS sequence data. The ITS tree also branches sym- While the rate of evolution differed between rbcL and ITS, metrically with two notable clades being resolved, the ®rst the tree topology obtained by each data set was largely con- clade consisting of Levenhookia, Oreostylidium, and Stylidium gruent, differing mainly in the con¯icting placement of Lev- (bootstrap 97%; jackknife 100%) and the second clade con- enhookia, which was not well supported by either analysis sisting of Forstera and Phyllachne (bootstrap 83%; jackknife (Figs. 2±4). When the rbcL data were constrained to the to- 91%). As in the rbcL tree S. calcaratum emerges at the base pology of the ITS tree (Figs. 2, 4), one tree of 283 steps was of the ®rst clade, whereas Levenhookia is now nested within obtained; this tree was 24 steps longer than the maximum par- the ®rst clade emerging as sister to S. emarginatum (bootstrap simony tree of 259 steps. Similarly, when the ITS data were 51%; jackknife 71%). Oreostylidium again is nested well with- constrained to the topology of the rbcL tree (Figs. 2, 3), one in the ®rst clade emerging as sister to S. graminifolium (boot- tree of 696 steps was obtained; 26 steps longer than the max- strap 99%; jackknife 100%). The topology of the second clade imum parsimony tree of 670 steps without the constraint. Only is also similar to that obtained by rbcL with F. bellidifolia three branch moves using MacClade were required to account again diverging at the base of the second clade (bootstrap for the differences between the rbcL and the ITS trees. 100%; jackknife 100%) and with neither Phyllachne nor For- Because the results from the rbcL and ITS sequences were stera being monophyletic. However, the New Zealand species largely congruent, the two data sets were combined. The com- form two uniform groups that are well supported. bined analysis also recovered one maximum parsimony tree Absolute genetic distance values indicated that the gene of 943 steps (consistency index ϭ 0.688; retention index ϭ rbcL evolves at a slower rate than the ITS region; the mean 0.812; rescaled consistency index ϭ 0.559; excluding unin- (Ϯ1 SD) number of nucleotide substitutions in rbcL that dis- formative characters) (Fig. 5) that shared aspects of the rbcL tinguish Donatia novae-zelandiae from the Stylidiaceae ter- and ITS trees, but the relationships were more highly resolved May 2002] WAGSTAFF AND WEGEÐDIVERSIFICATION OF NEW ZEALAND STYLIDIACEAE 869

Fig. 4. Maximum parsimony tree of 670 steps obtained from the ITS sequences (consistency index ϭ 0.685; retention index ϭ 0.809; rescaled consistency index ϭ 0.554; excluding uninformative characters). Branch lengths are proportional to the number of nucleotide substitutions with a scale bar below.

and better supported. The tree again branched symmetrically with the same two clades resolved; the ®rst clade comprised Levenhookia, Oreostylidium, and Stylidium (bootstrap 86%; jackknife 95%) and the second was composed of Forstera and Phyllachne (bootstrap 98%; jackknife 100%). As in the rbcL tree, L. leptantha diverged at the base of the ®rst clade (boot- strap 69%; jackknife 84%). The two accessions of O. subu- latum form a well-supported clade (bootstrap 99%; jackknife 100%) nested well within the ®rst clade with S. graminifolium well supported as their sister (bootstrap 100%; jackknife 100%). The branching order of the second clade was identical to that obtained from the ITS sequences. Forstera bellidifolia diverged at the base of the second clade (bootstrap 100%; jackknife 100%). As in the previous analyses, neither Forstera nor Phyllachne is monophyletic, but two smaller clades are monophyletic. The ®rst of these consists of P. rubra, P. col- ensoi, and P. clavigera (bootstrap 86%; jackknife 95%) and the second of F. sedifolia, F. tenella, and F. bidwillii (boot- strap 58%; jackknife 74%). The combined data set was subjected to maximum likeli- hood analysis using the general-time-reversible model (e.g., Yang, 1994) with the following parameters estimated from the combined parsimony tree shown in Fig. 5: (1) the assumed nucleotide frequencies were A ϭ 0.25459, C ϭ 0.23492, G ϭ 0.26019, and T ϭ 0.25030; (2) the estimated proportion of invariable sites ϭ 0.478127 (observed proportion of constant sites ϭ 0.761447); and the estimated value of the gamma shape parameter ϭ 0.541289. The maximum likelihood anal- Fig. 5. Maximum parsimony tree of 943 steps obtained from the com- ysis of the combined data yielded a single tree (Ϫlog ϭ bined rbcL and ITS sequence data (consistency index ϭ 0.688; retention index ϭ 0.812; rescaled consistency index ϭ 0.559; excluding uninformative char- 7789.31019) (Fig. 6). The branching order was identical to the acters). Bootstrap values are given above each node and jackknife values combined parsimony analysis, but the branch lengths differed. below. The maximum likelihood result may provide a more realistic 870 AMERICAN JOURNAL OF BOTANY [Vol. 89

Fig. 6. Maximum likelihood tree (Ϫlog ϭ 7789.31019). Branch lengths are proportional to the number of nucleotide substitutions and have been corrected for multiple hits using the general-time-reversible model (e.g., Yang, 1994) with parameters estimated from the combined parsimony tree shown in Fig. 5. A scale bar is provided below. Divergence times were estimated from the fossil record calibrated using the ®rst appearance of Forstera±type pollen 39 million years ago. Floral illustrations are provided for the taxa highlighted in bold (see Fig. 7). estimate of divergence because corrections are made for mul- result. This feature was used by Mildbraed (1908) to separate tiple substitutions along long branches. These observations the two groups into the tribes Phyllachneae and Stylidieae. also indicate that our phylogenetic inferences are relatively robust to the different assumptions of these two approaches to Origin and dispersalÐColonization of New Zealand by phylogeny reconstruction. members of the Stylidiaceae involved at least two instances of long-distance dispersal. The distribution of Donatia, the out- DISCUSSION group in our study, is very similar to that of Phyllachne. Don- Differences in the origin and timing of evolution and in their atia fascicularis is found in southern South America to latitude ability to adapt to the unique environmental conditions char- 40ЊS, and D. novae-zelandiae is found in both New Zealand acteristic of New Zealand may in part account for the differ- and Tasmania. The Tasmanian species F. bellidifolia diverges ences in species richness among lineages of New Zealand Styl- at the base of the Forstera/Phyllachne clade and the South idiaceae. According to Carlquist (1974) the evolution of ¯oral American species P. uliginosa is sister to the New Zealand morphology, breeding systems, ecological preferences, growth species, hence the origin of the New Zealand clade is equiv- forms, and dispersal mechanisms are all closely interrelated to ocal; with a South American or Tasmanian origin being equal- enhance reproductive success in island plants, and we discuss ly parsimonious. Conspeci®c populations of P. colensoi in some of these interrelationships below. New Zealand and Tasmania probably re¯ect more recent dis- Our results suggest that the New Zealand Stylidiaceae fall persal from New Zealand to Tasmania. The New Zealand an- into two distinct lineages that differ substantially from one cestor of O. subulatum arrived by long-distance dispersal from another in species richness (Figs. 2±6), a result consistent with Australia. Oreostylidium subulatum is nested in the largely that obtained by Laurent, Bremer, and Bremer (1999). One Australian genus Stylidium; S. graminifolium, its sister in our lineage is composed of seven species placed in Forstera and analysis, is widely distributed in eastern Australia and Tas- Phyllachne, while the other New Zealand group consists solely mania. of Oreostylidium subulatum, which is nested within a clade The present distribution of Stylidiaceae suggests that Ant- consisting of Levenhookia and Stylidium. The occurrence of arctica may have played an important role as a corridor for curved, monothecous anthers in both Phyllachne and Forstera, the migration of Stylidiaceae between Australia, New Zealand, as opposed to the dithecous anthers found in Levenhookia, and South America during the Tertiary (Fig. 1). Isolation of Stylidium, and Oreostylidium, lends additional support to this Antarctica appears to have been initiated south of Tasmania at May 2002] WAGSTAFF AND WEGEÐDIVERSIFICATION OF NEW ZEALAND STYLIDIACEAE 871 the boundary between the Eocene and Oligocene some 35 mil- (Raven, 1973). Although montane and cool-temperate envi- lion years ago, separating Antarctica from the Australian areas ronments existed earlier, subalpine and alpine environments (Coleman, 1980; Johnson and Veevers, 1984; Veevers, Powell, probably did not. The ®rst major lowering of temperatures in and Roots, 1991). New Zealand at this time had already been New Zealand took place in the Late Pliocene about 2.4 million isolated for more than 45 million years, but it is possible that years ago, and there appear to have been at least three glacial the ancestors of New Zealand Stylidiaceae reached New Zea- cycles subsequently. During the last, or Otiran, cycle, there land across now submerged islands and/or now uninhabitable was heavy glaciation in the South Island along and west of areas along the Antarctic coast long after the isolation of New the Main Divide. Stewart Island and the subantarctic islands Zealand. Up to the Oligocene±Miocene boundary 23 million were glaciated, and all plants seem to have been eliminated years ago, the Antarctic Peninsula and southern South America from some of the smaller islands. In Australia, the major effect were linked by a land bridge, the Scotia Arc, which ®nally of the pluvial cycles during the Pleistocene was to promote broke open to form the Drake Passage and establish the cir- moist fertile corridors across now arid regions facilitating con- cum-Antarctic current (Barker and Burrell, 1977; Coleman, tact between ¯oras of the eastern and western temperate zones 1980; Dalziel, 1983). There is abundant fossil evidence that of Australia (Raven, 1973). suggests the Transantarctic Mountains were clothed in beech forests until as recently as the Pliocene (Webb and Harwood, Patterns of diversi®cationÐThe New Zealand Stylidiaceae 1993). can be readily distinguished by their growth habit, ¯oral mor- Divergence estimates based upon the gene rbcL (Fig. 3) phology, and ecological requirements, and these factors may suggest that dispersal to New Zealand occurred at two distinct have in¯uenced their ability to diversify in New Zealand (Fig. times during the late Tertiary. The mean (Ϯ1 SD) number of 7). Species of Phyllachne have short stem innovations that are nucleotide substitutions that distinguish Phyllachne uliginosa densely branched, forming compact cushions (Fig. 7A), a fea- from the Forstera/Phyllachne terminals is 8.14 Ϯ 0.89, where- ture that is also characteristic of Donatia. In contrast, the stems as the accessions of Oreostylidium subulatum are distinguished in Forstera are elongate and simply branched to form more from by 4.00 Ϯ 1.41 substitutions. delicate perennial herbs (Fig. 7D). Oreostylidium is distinct Based upon the substitution rate of 0.74 substitutions per mil- from the other New Zealand species and shares a basal-rosette lion years calculated for the gene rbcL by Bremer and Gus- habit (Fig. 7H) with its sister S. graminifolium (Fig. 7E). tafsson (1997), the Forstera/Phyllachne lineage in New Zea- Cushion plants are common elements of the subalpine and land shared a common ancestor with the South American P. alpine vegetation of Tasmania, New Zealand, the subantarctic uliginosa about 6 million years ago and O. subulatum shared islands, and South America and have evolved independently a common ancestor with the Australian S. graminifolium about in several plant groups. This life form is a response to a num- 3 million years ago. ber of variables, including intense radiation, temperature, These divergence times are consistent with the substitution physical and physiological drought, low soil fertility, and wind rate estimated from the fossil record (Fig. 6). According to (Gibson and Kirkpatrick, 1985). According to Godley (1960) Macphail (1997) Forstera-type pollen (Tricolpites stylidioides) the ecological conditions and taxonomic af®nities of cushion ®rst appears in southeastern Australia possibly as early as the bogs in New Zealand and southern Chile are remarkably sim- Oligocene some 39 million years ago, whereas Stylidiaceae ilar where D. fascicularis and P. uliginosa are characteristic pollen is known only from the Quaternary in New Zealand components of the Magellanic moorland region of southern (Mildenhall, 1980). The mean distance from the terminals to Chile. In New Zealand D. novae-zelandiae occurs in mainly the base of the Forstera/Phyllachne clade is 0.074 Ϯ 0.005 subalpine habitats, but descends to low altitudes in Westland substitions per site, and dividing this by 39 million years yields and beside Foveaux Strait (Wardle, 1991). Along with P. col- a substitution rate of approximately 0.0019 nucleotide substi- ensoi it is often dominant in upland bogs. tutions per site per million years (Fig. 6). The mean number is widely distributed in high mountain habitats including bogs of substitutions per site from the New Zealand Forstera/Phyl- throughout mainland New Zealand, with the exception of Mt. lachne terminals to the base of the New Zealand clade is 0.012 Taranaki. In mainland New Zealand the range of P. colensoi Ϯ 0.002, which also equates the ®rst appearance in New Zea- overlaps with P. clavigera, which is found from latitude 42Њ S land to about 6 million years ago (0.012 nucleotide substitu- and in the subantarctic islands and with P. rubra, which has tions per site/0.0019 nucleotide substitutions per site per mil- a more restricted distribution on the Central Otago plateau. lion years). The mean number of substitutions per site from Conspeci®c populations of Phyllachne colensoi in New the Oreostylidium terminals to the base of the Oreostylidium Zealand and Tasmania probably re¯ect recent dispersal from clade is 0.0035, which equates to about 2 million years New Zealand. This species is widespread in the alpine zone (0.0035/0.0019), a ®rst appearance in New Zealand that is of New Zealand, but in Tasmania it is recorded from only 12 somewhat more recent than the estimate obtained from rbcL locations and is generally restricted to areas above 1250 m on sequences alone, but is more consistent with the Quaternary well-drained sites or rocky slopes. The restricted distribution fossil record. The difference may re¯ect a more realistic esti- of P. colensoi in Tasmania may re¯ect its recent establishment mate of branch lengths obtained by maximum likelihood anal- and the lower habitat diversity in the subalpine and alpine ysis. zones (Gibson and Kirkpatrick, 1985). The estimated divergences times (Figs. 3, 6) suggest diver- The spreading, sparsely branched growth habit of Forstera si®cation of Stylidiaceae was likely in¯uenced by major geo- (Fig. 7D) probably evolved as this species expanded into less logical and climatical changes and ¯uctuations in sea level that stressful subalpine and montane herb®elds. The more robust occurred during the late Tertiary and Quaternary. Uplift of the species F. sedifolia and F. mackayii are found in high montane mountains in Australia and New Zealand to their present ele- to subalpine regions, whereas the somewhat more slender spe- vation and subsequent episodes of glaciation created a diver- cies F. bidwillii and F. tenella are found in the subalpine to sity of habitats largely during the Pliocene and Pleistocene low montane areas. 872 AMERICAN JOURNAL OF BOTANY [Vol. 89

Fig. 7. (A) Habit of Phyllachne colensoi (1.5ϫ). (B) Flower (6ϫ). (C) Flower of Forstera bidwillii (1.5ϫ). (D) Habit (6ϫ). (E) Habit of Stylidium graminifolium (0.75ϫ). (F) Flower (3ϫ). (G) Flower of Oreostylidium subulatum (6ϫ). (H) Habit (3ϫ). The illustrations were adapted from Mildbraed (1908).

Although Oreostylidium subulatum shares a similar growth serve as pollinators (Lloyd, 1985). The ¯owers of Phyllachne habit to Stylidium graminifolium (Fig. 7E, H), it has not di- are sessile and solitary at the apices of the stems, scarcely versi®ed in New Zealand. It is restricted to damp grasslands emerging from the surface of the foliage (Fig. 7B). However, and herb®elds in montane and lower subalpine areas (Allan, the total ¯oral display can be quite striking given the densely 1961), whereas S. graminifolium is widespread in eastern branched nature of the stems. A broad spectrum of insect vis- mainland Australia and Tasmania, occurring on a diversity of itors has been documented for P. colensoi in both New Zea- soil types throughout coastal and montane regions (Raulings land (Primack, 1983) and Tasmania (Corbett, 1995). In For- and Ladiges, 2001). Lloyd (1985) postulated that many Aus- stera, the ¯owers form a less conspicuous ¯oral display, borne tralian plant groups are absent or poorly represented in New singularly or in pairs on elongate scapes (Fig. 7C). Forstera Zealand because they possess characters that are poorly adapt- mackyii and F. bidwillii have colored pollinator guides at the ed to New Zealand habitats. throat of the ¯ower. There is only one pollination record for Like many of the ¯owering plants in mountainous regions Forstera (Primack, 1983). of New Zealand, all members of the New Zealand Stylidiaceae In contrast to the other New Zealand Stylidiaceae, Oreos- (including Donatia novae-zelandiae, Donatiaceae) possess tylidium subulatum has undergone a dramatic evolutionary small, unspecialized, white ¯owers (Fig. 7B, C, G). There is transformation in its ¯oral morphology (Fig. 7G). It possesses a general lack of specialist insect pollinators in New Zealand actinomorphic, white, solitary ¯owers with an insensitive col- and hence a disproportionate reliance on unspecialized polli- umn (Fig. 7G, H), whereas its sister, Stylidium graminifolium nators that promiscuously visit a wide range of plants and (along with most of the other species in the genus Stylidium), forage in an imprecise manner. A generalized ¯oral structure possesses zygomorphic, often colorful ¯owers, typically borne may confer a selective advantage by permitting more species on multi¯owered in¯orescences and with sensitive columns of insects and particularly smaller, less specialized insects to (Fig. 7E, F). Laurent, Bremer, and Bremer (1999) suggested May 2002] WAGSTAFF AND WEGEÐDIVERSIFICATION OF NEW ZEALAND STYLIDIACEAE 873 that the small, white ¯owers of O. subulatum developed by ment, the availability of suitable pollinators, and their breeding paedomorphosis (accelerated sexual development in a mor- systems partly account for contrasting evolutionary success in phologically immature plant). The ¯owers of O. subulatum the New Zealand Stylidiaceae. Finally our results have im- also lack the sensitive column that enables precise pollen portant implications for the classi®cation of New Zealand Styl- placement, so presumably they are visited by a range of in- idiaceae. Based upon our study, we would propose two taxo- sects. This is in contrast to most species of Stylidium, in which nomic changes, ®rst to include Oreostylidium in the genus pollinator constancy has been observed (Erickson, 1958). Stylidium, and second, to include Phyllachne in the genus For- It is plausible that Oreostylidium subulatum underwent this stera, Forstera being the older name. There are precedents for transformation prior to dispersal to New Zealand, but we pro- both of these changes. Valid combinations have been pub- pose that this evolutionary change occurred relatively rapidly lished for Stylidium subulatum and Forstera clavigera. These and after it became established in New Zealand. The initial O. changes would render Stylidium and Forstera monophyletic. subulatum founder population was presumably small and may have grown from a single seed. The lack of specialist polli- LITERATURE CITED nators may have also constrained its establishment and diver- si®cation in New Zealand. ALBERT, V. A., A. BACKLUND,K.BREMER,M.W.CHASE,J.R.MANHART, A well-documented system of balanced lethal mutations has B. D. MISHLER, AND K. C. NIXON. 1994. Functional constraints and been shown to minimize the products of self-pollination in rbcL evidence for land plant phylogeny. Annals of the Missouri Botanical southwest species of Stylidium (Coates and James, 1979; Garden 81: 534±567. ALLAN, H. H. 1961. Flora of New Zealand. R. E. Owen Government Printer, James, 1979; Burbidge and James, 1991). Willis and Ash Wellington, New Zealand. (1990) later demonstrated that a self-incompatibility system BALDWIN, B. G., M. J. SANDERSON,J.M.PORTER,M.F.WOJCIECHOWSKI, also operates in the eastern Australian species S. productum C. S. CAMPBELL, AND M. J. DONOGHUE. 1995. The ITS region of nu- and S. graminifolium. It is unknown whether an equivalent clear ribosomal DNA: a valuable source of evidence on angiosperm phy- system operates in the closely allied species Oreostylidium logeny. Annals of the Missouri Botanical Garden 82: 247±277. subulatum. This is a particularly interesting area for further BARKER,P.F.,AND J. BURRELL. 1977. The opening of Drake Passage. Marine Geology 25: 15±34. research given that Godley (1979) has illustrated that self-in- BEAN, A. R. 1999. A revision of Stylidium sect. Debilia Mildbr, S. sect. compatibility in the native New Zealand ¯ora is a relatively Floodia Mildbr. and S. sect. Lanata A.R. Bean (Stylidiaceae). Austro- infrequent phenomenon. baileya 5: 427±455. Although Stylidiaceae ¯owers appear designed to promote BEAN, A. R. 2000. A revision of Stylidium subg. Andersonia (R.Br. ex cross-pollination, high levels of inbreeding can be generated G.Don.) Mildbr. (Stylidiaceae). Austrobaileya 5: 589±649. by geitonogamous self-pollination, which is facilitated by BREMER, K., AND M. H. G. GUSTAFSSON. 1997. East Gondwana ancestry of the sun¯ower alliance of families. Proceedings of the National Academy many-¯owered in¯orescences. The elongate scapes in Styli- of Sciences, USA 94: 9188±9190. dium graminifolium (Fig. 7E) and allied trigger plants support BURBIDGE,A.H.,AND S. H. JAMES. 1991. Postzygotic seed abortion in the between 10 and 110 ¯owers, for example (Raulings and La- genetic system of Stylidium (Angiospermae: Stylidiaceae). Journal of diges, 2001). In contrast, the ¯owers of Oreostylidium subu- Heredity 82: 219±228. latum are borne individually on short scapes (Fig. 7H). The CARLQUIST, S. 1969. Studies in Stylidiaceae: new taxa, ®eld observations, inconspicuous ¯owers combined with the loss of column evolutionary tendencies. Aliso 7: 13±64. CARLQUIST, S. 1974. Island biology. Columbia University Press, New York, movement may indicate a switch toward obligate autogamy, New York, USA. as reported in some species of Parahebe (Garnock-Jones, CARLQUIST, S. 1978. New species of Stylidium, with comments on evolu- 1976; Wagstaff and Garnock-Jones, 2000). Carlquist (1978) tionary patterns in tropical Stylidiaceae. Aliso 9: 308±322. suggested the reduction in in¯orescence size, loss of column CHASE,M.W.,ET AL. 1993. Phylogenetics of seed plants: an analysis of sensitivity, and recurved column in the northern Australian nucleotide sequences from the plastid gene rbcL. Annals of the Missouri species S. reductum were also indicative of a switch to self- Botanical Garden 80: 528±580. COATES, D., AND S. H. JAMES. 1979. Chromosome variation in Stylidium pollination. In an Oreostylidium population examined at crossocephalum (Angiospermae: Stylidiaceae) and the dynamic co-ad- Maungatua the stigmas developed in the bud suggesting that aptation of its lethal system. Chromosoma 72: 357±376. autogamy is extremely likely, at least in this population. Or- COLEMAN, P. J. 1980. Plate tectonics background to biogeographic develop- eostylidium is recorded as white-¯owered (Allan, 1961), al- ment in the southwest Paci®c over the last 100 million years. Palaeo- though the ¯owers on the Maungatua population had a pink- geography, Palaeoclimatology, Palaeoecology 31: 105±121. ish-red abaxial ¯ush and the column was similarly colored at CORBETT, C. 1995. Pollination ecology in a Tasmanian alpine environment. Honors thesis, Department of Geography and Environmental Science, the base. Garnock-Jones (1976) noted that the corollas of taxa University of Tasmania, Hobart, Australia. adapted to autogamy are usually uniformly white. The evo- CURTIS, W. M. 1963. The student's Flora of Tasmania. L. G. Shea, Govern- lution of autogamy would have facilitated the establishment ment Printer, Tasmania, Australia. of an initial Oreostylidium founder population, but would limit DALZIEL, I. W. D. 1983. The evolution of the Scotia Arc: a review. In R. L. the amount of genetic variation within and among populations, Oliver, P. R. James, and J. B. Jago [eds.], Antarctic earth science, 283± thus restricting its ability to adapt to environmental changes 288. Cambridge University Press, London, UK. DOYLE,J.J.,AND J. L. DOYLE. 1987. A rapid DNA isolation procedure from in the longer term. small quantities of fresh leaf tissue. Phytochemical Bulletin 19: 11±15. In summary the sequence data support two distinct clades ERICKSON, R. 1958. Triggerplants. Paterson Brokensha, Perth, Australia. of New Zealand Stylidiaceae that differ in species richness. ERICKSON, R., AND J. H. WILLIS. 1966. Some additions to Australian Styli- The origin of the Forstera/Phyllachne clade is equivocal; how- diaceae. Victorian Naturalist 83: 107±112. ever, our results clearly demonstrate an Australian origin for FARRIS, J. S., V. A. ALBERT,M.KAÈ LLERSJOÈ ,D.LIPSCOMB, AND A. G. KLUGE. the New Zealand endemic Oreostylidium subulatum. Forstera 1996. Parsimony jackkni®ng outperforms neighbor-joining. Cladistics 11: 99±124. and Phyllachne are conspicuous members of the subalpine and FELSENSTEIN, J. 1985. Con®dence limits on phylogenies: an approach using alpine ¯ora of New Zealand, while O. subulatum has a more the bootstrap. Evolution 39: 783±791. restricted distribution. Differences in the time of establish- GARNOCK-JONES, P. J. 1976. Breeding systems and pollination in New Zea- 874 AMERICAN JOURNAL OF BOTANY [Vol. 89

land Parahebe (Scrophulariaceae). New Zealand Journal of Botany 14: RAULINGS, E. J. 1999. Stylidiaceae. In N. B. Walsh and T. J. Entwisle [eds.], 291±298. Flora of , vol. 4, 579±587. Inkata Press, Melbourne, Australia. GIBSON,N.J.,AND B. KIRKPATRICK. 1985. A comparison of the cushion RAULINGS,E.J.,AND P. Y. L ADIGES. 2001. Morphological variation and spe- plant communities of New Zealand and Tasmania. New Zealand Journal ciation in Stylidium graminifolium (Stylidiaceae), description of S. mon- of Botany 23: 549±566. tanum and reinstatement of S. armeria. Australian Systematic Botany,in GODLEY, E. J. 1960. The botany of southern Chile in relation to New Zealand press. and the subantarctic. Proceedings of the Royal Society, B 152: 457±475. RAVEN, P. H. 1973. Evolution of the subalpine and alpine plant groups in GODLEY, E. J. 1979. Flower biology in New Zealand. New Zealand Journal New Zealand. New Zealand Journal of Botany 11: 177±200. of Botany 17: 441±466. SWOFFORD, D. L. 1998. PAUP*: phylogenetic analysis using parsimony (*and GUSTAFSSON,M.H.G.,AND K. BREMER. 1995. Morphology and phyloge- other methods), version 4. Sinauer, Sunderland, Massachusetts, USA. netic interrelationships of the Asteraceae, Calyceraceae, Campanulaceae, THOMPSON, J. D., T. J. GIBSON,F.PLEWNIAK,F.JEANMOUGIN, AND D. G. Goodeniaceae, and related families (). American Journal of Bot- HIGGINS. 1997. The CLUSTAL࿞X windows interface: ¯exible strategies any 82: 250±265. for multiple sequence alignment aided by quality analysis tools. Nucleic JAMES, S. H. 1979. Chromosome numbers and genetic systems in the trig- Acids Research 25: 4876±4882. gerplants of Western Australia (Stylidium; Stylidiaceae). Australian Jour- VEEVERS, J. J., C. POWELL, AND S. R. ROOTS. 1991. Review of sea¯oor nal of Botany 27: 17±25. spreading around Australia. 1. Synthesis of the patterns of spreading. JOHNSON, B. D., AND J. J. VEEVERS. 1984. Oceanic palaeomagnetisum. In J. Australian Journal of Earth Sciences 38: 373±389. J. Veever [ed.], Panerozoic earth history of Australia, 17±38. Clarendon WAGSTAFF,S.J.,AND P. J. GARNOCK-JONES. 1998. Evolution and biogeog- Press, Oxford, UK. raphy of the Hebe complex (Scrophulariaceae) inferred from ITS se- KAÈ LLERSJOÈ , M., J. S. FARRIS,M.W.CHASE,B.BREMER,M.F.FAY,C.J. quences. New Zealand Journal of Botany 36: 425±437. HUMPHRIES,G.PETERSEN,O.SEBERG, AND K. BREMER. 1998. Simul- WAGSTAFF,S.J.,AND P. J. GARNOCK-JONES. 2000. Patterns of diversi®cation taneous parsimony jackknife analysis of 2538 rbcL DNA sequences re- in Chionohebe and Parahebe (Scrophulariaceae) inferred from ITS se- veals support for major clades of green plants, land plants, seed plants quences. New Zealand Journal of Botany 38: 389±407. and ¯owering plants. Plant Systematics and Evolution 213: 259±287. WARDLE, P. 1968. Evidence for an indigenous pre-Quaternary element in the LAURENT, N., B. BREMER, AND K. BREMER. 1999. Phylogeny and generic mountain ¯ora of New Zealand. New Zealand Journal of Botany 6: 120± interrelationships of the Stylidiaceae (Asterales), with a possible extreme 125. case of ¯oral paedomorphosis. Systematic Botany 23: 289±304. WARDLE, P. 1978. Origins of New Zealand mountain ¯ora, with special ref- LLOYD, D. G. 1985. Progress in understanding the natural history of New erence to trans-Tasman relationships. New Zealand Journal of Botany Zealand plants. New Zealand Journal of Botany 23: 707±722. 16: 535±550. MACPHAIL, M. K. 1997. Comment on M. Pole (1994): `The New Zealand WARDLE, P. 1991. Vegetation of New Zealand. Cambridge University Press, FloraÐentirely long-distance dispersal?' Journal of Biogeography 24: Cambridge, UK. 113±117. WARDLE, P., C. EZCURRA,C.RAMIREZ, AND S. J. WAGSTAFF. 2001. Com- MADDISON,W.P.,AND D. R. MADDISON. 1992. MacClade, version. 3.1, Anal- parison of the ¯ora and vegetation of the southern Andes and New Zea- ysis of phylogeny and character evolution. Sinauer, Sunderland, Massa- land. New Zealand Journal of Botany 39: 69±108. chusetts, USA. WEBB,P.N.,AND D. M. HARWOOD. 1993. Pliocene fossil Nothofagus (south- MILDBRAED, J. 1908. Stylidiaceae. In A. Engler [ed.], Das P¯anzenreich IV, ern beech) from Antarctica: phytogeography, dispersal strategies, and 278. Wilhelm Engelmann, Leipzig, Germany. survival in high latitude glacial-deglacial environments. In J. Alden, J. MILDENHALL, D. C. 1980. New Zealand Late Cretaceous and Cenozoic plant L. Mastrantonio, and S. Odum [eds.], Forest development in cold cli- biogeography: a contribution. Palaeogeography, Palaeoclimatology, Pa- mates, 135±166. Plenum Press, New York, New York, USA. laeoecology 31: 197±233. WEGE, J. A. 1999. Morphological and anatomical variation in Stylidium (Styl- OLMSTEAD, R. G., H. J. MICHAELS,K.M.SCOTT, AND J. D. PALMER. 1992. idiaceae)Ða systematic perspective. Ph.D. dissertation, Department of Monophyly of the Asteridae and identi®cation of their major lineages Botany, The University of Western Australia, Nedlands, Australia. inferred from DNA sequences of rbcL. Annals of the Missouri Botanical WESTERN AUSTRALIAN HERBARIUM. 1998. FloraBaseÐinformation on the Garden 79: 249±265. Western Australian ¯ora. Department of Conservation and Land Man- PHILIPSON,W.R.,AND M. N. PHILIPSON. 1973. A comparison of the embry- agement. http://www.calm.gov.au/science/¯orabase.html. ology of Forstera and Donatia. New Zealand Journal of Botany 11: 449± WILLIS,A.J.,AND J. E. ASH. 1990. The breeding systems of Stylidium gra- 460. minifolium and S. productum (Stylidiaceae). Australian Journal of Bot- POLE, M. 1994. The New Zealand ¯oraÐentirely long-distance dispersal? any 38: 217±227. Journal of Biogeography 21: 625±635. WINKWORTH, R. C., A. W. ROBERTSON,F.EHRENDORFER, AND P. J. LOCK- PRIMACK, R. B. 1983. Insect pollination in the New Zealand mountain ¯ora. HART. 1999. The importance of dispersal and recent speciation in the New Zealand Journal of Botany 21: 317±333. ¯ora of New Zealand. Journal of Biogeography 26: 1323±1325. RAPSON, L. J. 1953. Vegetative anatomy in Donatia, Phyllachne, Forstera, YANG, Z. 1994. Maximum likelihood phylogenetic estimation from DNA and Oreostylidium and its taxonomic signi®cance. Transactions of the sequences with variable rates over sites: approximate methods. Journal Royal Society of New Zealand 80: 399±402. of Molecular Evolution 39: 306±331.