•The lack of extant lianescent vessel-less seed plants supports a hypothesis that liana evolution requires large-diameter xylem conduits. Here, we demonstrate an unusual example of a lianoid vessel-less angiosperm, Tasmannia cordata (Winteraceae), from New Guinea.
•Wood mechanical, hydraulic and structural measurements were used to determine how T. cordata climbs and to test for ecophysiological shifts related to liana evolution vs 13 free-standing congeners.
•The tracheid-based wood of T. cordata furnished low hydraulic capacity compared with that of vessel-bearing lianas. In comparison with most nonclimbing relatives, T. cordata possessed lower photosynthetic rates and leaf and stem hydraulic capacities. However, T. cordata exhibited a two- to five-fold greater wood elastic modulus than its relatives.
•Tasmannia cordata provides an unusual example of angiosperm liana evolution uncoupled from xylem conduit gigantism, as well as high plasticity and cell type diversity in vascular development. Because T. cordata lacks vessels, our results suggest that a key limitation for a vessel-less liana is that strong and low hydraulically conductive wood is required to meet the mechanical demands of lianescence.
One outstanding vascular specialization shared across fern, lycophyte and seed plant climbers is the evolution of considerably larger water-conducting conduits when compared with plants with free-standing habits. The enlarged xylem conduits of lianas can be vessels or tracheids. For example, vessels of extant angiosperm and gnetalean lianas are two to eight times wider (up to 700 μm) and several meters longer than the largest vessels found in free-standing angiosperms (Ewers et al., 1990; Carlquist, 1991; Caballé, 1993; Fisher & Ewers, 1995; Fisher et al., 2002; Gutierrez et al., 2009). Not surprisingly, angiosperm liana stems represent the largest xylem hydraulic capacities known (Gartner et al., 1990; Feild & Balun, 2008; Zhu & Cao, 2009). Tracheids of extinct seed plant and extant fern climbers are massive in their own right, becoming up to 200 μm wide and several centimeters long (Chican, 1986; Veres, 1990; Wilson et al., 2008). The conductances of these fern and gymnosperm vine tracheids represent the hydraulic zenith of their times and approximate those of some angiosperm vessels (Wilson et al., 2008).
Why are lianoid habits and xylem conduit gigantism intertwined? One hypothesis is that large xylem conduits enable the evolution of the unusually slender stems of lianas (Ewers & Fisher, 1991). The reasoning is that the widening and lengthening of xylem conduits represents the most effective mechanism to drive down wood investment, whilst maintaining or increasing hydraulic capacity across the long hydraulic path lengths of water transport by lianas in the forest understory or in large treefall gaps and clearings, as well as supplying hydraulically the high leaf gas exchange capacities found in most lianas (Putz, 1984; Ewers & Fisher, 1991; Zhu & Cao, 2009). The hypothesized mechanism is that conduit hydraulic flow increases disproportionately with xylem conduit size. Thus, with more flow afforded by a shift to a few large conduits, the amount of xylem needed to hydraulically supply a given area of transpiring leaves is decreased – translating into stem slenderness (Ewers & Fisher, 1991).
However, liana stems are more than specialized hydraulic hoses. Although lianas enjoy freedom from many of the compressive forces of gravity by mechanical parasitism, lianas must manage the unique and considerable biomechanical demands of climbing (Rowe et al., 2006; Isnard & Silk, 2009). Indeed, the vascular systems of lianas are among the most heteroxylous vascular systems known, exhibiting a broad array of compliant tissues juxtaposed with stiff fiber systems that cushion megaporous vascular conduits, prevent stem breakage or help heal damaged stems (Carlquist, 1991; Putz & Holbrook, 1991; Caballé, 1993; Isnard & Silk, 2009). Scandent ferns and extinct seed plants with giant tracheids illustrate vividly that liana-linked heteroxylly can be decoupled from vessels. In these taxa, sclerencymatous tissues mechanically support the thin-walled, highly pitted, elongate and wide-diameter tracheids (Veres, 1990; Wilson et al., 2008; Pittermann et al., 2011). Thus, in both vessel-less and vesselled vines, large xylem conduits apparently are essential to keep the requisite amount of xylem hydraulic volume low in the stem. This allows other functions and/or cell types related to storage, carbon transport and mechanical function to be efficiently packaged in the stem vasculature to meet the functional demands of climbing without compromising stem slenderness (Carlquist, 1991; Ewers & Fisher, 1991; Putz & Holbrook, 1991; Caballé, 1993; Rowe et al., 2006; Feild et al., 2009; Hudson et al., 2010).
A phylogenetic pattern supporting the hypothesis that the evolution of lianescence requires conduits with unusually high hydraulic throughput is that no extant cycad, conifer or vessel-less angiosperm climbers are thought to occur (Bailey, 1944; Carlquist, 1975, Carlquist 2009; Wilson & Knoll, 2010). These clades possess vessel-less, homoxylous wood – referring to the dominance by tracheids with small- and uniform-diameter lumens (Bailey, 1944; Carlquist, 1975). As such, these taxa are hypothesized to be evolutionarily obviated from climbing growth forms because of xylem constraints. First, vessel-less homoxylous vasculature composed of relatively small-diameter tracheids would seem to be unable to furnish the sufficiently high xylem hydraulic capacity necessary to generate a narrow stem cross-section with high hydraulic supply to transpiring leaves.
Another hypothesized reason why vessel-less homoxylous taxa are evolutionarily obviated from climbing growth forms is that they lack sufficient developmental potential in the vascular cambium to evolve a specialized vascular stem system required for a lianoid habit – slender but functionally specialized (Carlquist, 1975, 2009; Pace et al., 2009; Wilson & Knoll, 2010). Such is the case because extant seed plant tracheid-based vasculatures, unlike vessel-fiber systems, multitask biomechanical and hydraulic functions (Carlquist, 1975; Pittermann et al., 2006, 2011). The specialization of hydraulics and mechanics pulls optimal tracheid anatomy in opposite directions that conceivably result in evolutionary blind alleys for a theoretically vessel-less, homoxylous vine. For example, a vessel-less liana with wood maximizing hydraulic capacity would exhibit low stem mechanical strength or mechanical compliancy, and therefore function with low ability to survive forest disturbances that demand torsional compliancy (Putz & Holbrook, 1991; Rowe et al., 2006). Alternatively, a vessel-less liana could be strong or compliant by increasing wood density or increasing ray volume, respectively, but both shifts could be argued to entail a large sacrifice in xylem hydraulic conductivity. Thus, because of a hydraulic limit imposed by small conduits, a putative vessel-less homoxylous liana would be hypothesized to be unable to produce the low-cost stem extension growth that enables rapid growth in disturbed habitats, or the long hydraulic path lengths that allow extensive exploratory growth (Putz, 1984; Zhu & Cao, 2009).
However, there are anecdotal reports of a possible vessel-less angiosperm liana, Tasmannia piperita entity cordata Vink (referred to as ‘Tasmannia cordata’ Winteraceae, a magnoliid basal angiosperm lineage; see Supporting Information Methods S1) from subalpine cloud forests of eastern Papua New Guinea. Some reports have described T. cordata as a liana or scandent shrub from the subalpine cloud forests of eastern Papua New Guinea, whereas others have only mentioned free-standing habits (Wade & McVean, 1969; Vink, 1970; Venables, 1984). How T. cordata may climb on other plants and how it produces scandent growth in spite of the theoretical vascular design constraints facing vessel-less homoxylous wooded lianas remain unknown.
Here, we show that T. cordata occurs as a liana with tracheid-based homoxylous wood. Based on previous studies of how most vesselled lianas function in comparison with their free-standing relatives, we used field-based ecophysiological measurements to test a suite of hypotheses on T. cordata’s ecophysiology. As a liana, we hypothesized that, as T. cordata plants climbed onto their hosts, wood mechanical strength would decrease through development and that T. cordata’s wood mechanical strength would be lower than that of self-supporting Tasmannia shrubs and trees (Speck & Rowe, 1999; Rowe et al., 2006; Isnard & Silk, 2009). Because hydraulics and mechanical capacities are traded off to a greater extent in homoxylous tracheid-based woods, we hypothesized that T. cordata would possess larger tracheids that furnish greater stem xylem hydraulic capacities (at the sapwood and shoot scales) and increased stem slenderness (quantified as the xylem area for a given leaf area; Huber value), low wood density and weak mechanical strength. Our final expectation was that T. cordata would exhibit specializations towards greater leaf gas exchange capacity, better tolerance of sun-exposed, disturbed environments and the development of longer hydraulic path lengths than its free-standing congeners.
We tested these hypotheses by first determining how T. cordata functioned biomechanically and hydraulically through ontogeny. Anatomical measurements were used to determine the structural basis for possible changes in stem mechanical and hydraulic properties in T. cordata with ontogeny. We tested for functional shifts related to the evolution of climbing by comparing stem wood biomechanical and hydraulic functions, as well as leaf gas exchange and leaf hydraulic performance, of T. cordata with those of a diverse set of nonclimbing congeners. Because Winteraceae are characterized by homoxylous wood with relatively little structural variation (Carlquist, 1975, 1989; Feild et al., 2002), we applied fine-scale measurements using X-ray analyses to test whether chemical–structural changes in tracheid walls influenced the mechanical function of T. cordata wood relative to that of self-supporting Tasmannia (Evans, 2006). These measurements allowed the teasing apart of how small changes in wood density vs changes in the geometry of load-bearing cell walls of tracheids (i.e. average of the microfibril angle of the S2 tracheid layer, MFA) influence wood mechanical strength. We discuss the impact of our results on the evolution of lianas, in relation to xylem evolution as well as the ecological evolution of Winteraceae.
Understory–subcanopy shrub/small tree, to 7 m tall, in 15–17-m-tall montane cloud forest
Mt Gumi, S07°11.028, E146°28.404, 2187 m, Morobe Province, Papua New Guinea
Tasmannia cordata Vink
Understory-establishing liana climbing to the subcanopy of subalpine shrubberies to upper montane cloud forests
Summit trail, 3485 m, Mt Wilhelm, Chimbu Province, Papua New Guinea
Tasmannia crassipes Vink
Shrub, to 3.5 m tall, in forest edges in subalpine coniferous heath
Imbuka Ridge, above Lake Pindi, S05°47.271, E145°03.126, 3730 m, Mt Wilhelm, Chimbu Province, Papua New Guinea
Tasmannia glaucifolia J. B. Williams
Shrub, to 2 m tall, in open vegetation along small creeks and drainages near subalpine zones
Point Lookout, 1200 m, New South Wales, Australia
Tasmannia heteromera Vink
Understory–subcanopy shrub/small tree, to 3 m tall, leaning over late in ontogeny, in 6–8-m-tall forest and subalpine shrubberies
Lake Pindi, S05°47.620, E145°03.489, 3606 m, Mt Wilhelm, Chimbu Province, Papua New Guinea
Tasmannia insipida R.Br. ex DC
Understory to subcanopy shrub, to 3 m tall, in wet sclerophyll eucalypt woodland
Barrington Tops, 1000 m, New South Wales, Australia
Tasmannia lanceolata (Poir.) A. C. Smith
Alpine shrub to small tree, to 3 m tall
Mt Wellington, 1250 m, Tasmania, Australia
Tasmannia membranea (F.v.M.) A. C. Smith
Understory–subcanopy shrub/tree, to 7 m tall, in 10-m-tall montane cloud forest
Mt Bartle Frere, 850 m, Queensland, Australia
Tasmannia montis-wilhelmii (Hoogl.) A. C. Smith
Large shrub, to 4 m tall, edges of subalpine shrubberies
Lake Pindi, S05°47.609, E145°03.494, 3500 m, Chimbu Province, Papua New Guinea
Tasmannia nettioti Vink
Epiphyte, 0.35 m tall, on moss-laden trees in upper montane cloud forests
Above Bakaia, 2450 m Garasa, near Garaina, Morobe Province, Papua New Guinea
Tasmannia purpurascens (Vickery) A. C. Smith
Understory to subcanopy shrub, to 3 m tall, in wet sclerophyll eucalypt woodland
Gloucester Tops Road, 980 m, New South Wales, Australia
Tasmannia stipitata (Vickery) A. C. Smith
Understory to subcanopy shrub, to 3 m tall, in wet sclerophyll eucalypt woodland
Raspberry Road, Styx River Forest, New South Wales, Australia
Tasmannia subalpina Vink
Shrub, to 2 m tall, alpine tussock grasslands
Bomber crash site, S05°47.604′, E145°03.006, 3781 m Chimbu Province, Papua New Guinea
Tasmannia vickeriana (A. C. Smith) A. C. Smith
Alpine shrub, to 1 m tall
Mt Bawbaw, Victoria, Australia
Stem mechanical tests and X-ray analyses of micromechanical function
Stem bending tests were performed using a three-point bending method (Rowe et al., 2006). The bending apparatus consisted of a carbon tripod (Gitzo G126, Rungis, France) with a bolt-on stainless steel bar with hooks that could be adjusted to set distances. A published approach was used to establish the span length (the distance between two supports during bending) for a range of stem diameters to avoid shear stress during measurements (Rowe et al., 2006; Ménard et al., 2009). Bending forces perpendicular to the stem central axis were generated using scaled spring-gauges (Pesola AG, Baar, Switzerland). A ruler (resolution, 0.5 mm) was affixed to the tripod to measure the stem deflection distance. Flexural stiffness (EI) and structural Young’s modulus of elasticity (ES, MPa) were calculated from the linear relationships of deflection (mm) vs force (N; Rowe et al., 2006).
We sampled stems from 4 to 18 mm in diameter to reconstruct developmental changes in wood mechanical properties in T. cordata. Straight undamaged portions of stem axes with minimal taper and no branch nodes were measured (Rowe et al., 2006). Stems were pooled into four stem diameter classes for climbing plants (< 5.0, 5.1–10, 10.1–15.0 and > 15.1 mm) of T. cordata, and into three diameter classes (< 5, 5.1–10 and 10.1–15 mm) for nonclimbing T. cordata individuals. The selected stem diameter categories corresponded broadly to the ontogenetic stages in T. cordata. The first class included young searcher branches and side branches with juvenile wood, the second class included main stem axes of juvenile plants, the third class included major side branches and main stems dominated by mature wood, and the fourth class included only the largest scandent stems of climbers. We measured 10 stems from five individual plants for each stem diameter class. For the other erect Tasmannia taxa, mean interspecific ES values were based on 30 stems that varied from 8.5 to 15 mm in diameter.
Cellular scale analyses of Tasmannia wood properties were determined using Silviscan2 (Evans, 2006). Using published procedures, the system combines three nondestructive analytical techniques (X-ray densitometry, X-ray diffractometry and image analyses) to determine the wood density (g cm−3), MFA (degrees) and Young’s modulus of elasticity (EX, MPa; Evans, 2006; Keunecke et al., 2009). MFA measurements allow the testing of whether tracheid wall quantity or quality (in terms of cell wall structure) variables, which can vary independently, determine wood mechanical strength. These variables were determined every c. 100 μm from pith to bark in three samples for each species. We measured six taxa, including T. cordata, T. crassipes, T. heteromera, T. lanceolata, T. montis-wilhelmii and T. subalpina. EX was calculated on the basis of the model published previously using wood density and the X-ray diffraction pattern (Evans, 2006).
Stem hydraulic measurements
Stem hydraulic conductance (K; kg MPa−1 s−1) was measured using a portable flow meter. Details on the flow meter, field collection methods used to avoid the measurement of embolized stems and the standardization of K measurements are described elsewhere (Feild et al., 2011). Briefly, we sampled two branches from five individual plants for hydraulic measurements (n =10 branches for each species) during a week of continuous rain to avoid xylem embolism. We checked for the presence of emboli by flushing each measured stem with pressurized water (from a syringe at c. 0.1 MPa pressure) and observing the cut end underwater for air bubbles with a hand lens (Feild et al., 2011). No evidence for native embolism was observed in any of the species measured. All of the shoots for each species sampled possessed fully expanded leaves. In addition, the shoots measured came from the most sunlight-exposed sites for a species to measure maximal stem hydraulic capacity (Brodribb & Feild, 2000). Branches were cut with clippers, double bagged in plastic bags and moved to a field camp laboratory for measurement within 1 h of collection. All sampled branches for each species were 10 cm long with diameters ranging from 3 to 5 mm.
From K data, the hydraulic conductivity (KH, kg MPa−1 m− s−1) was calculated as K multiplied by the stem segment length. Using KH, the sapwood area specific hydraulic conductivity (KS; kg MPa−1 m−1 s−1), which quantified the stem hydraulic efficiency in relation to xylem investment, was calculated as KH divided by the sapwood cross-sectional area without the pith. The sapwood cross-sectional area was measured using ImageJ freeware (National Institutes of Health, Bethesda, MD, USA) on photographed transverse sections of measured stems using a compound microscope (CH2; Olympus, New York, USA) and camera (Nikon D300S, New York, USA; Feild et al., 2011). We determined the leaf area specific hydraulic conductivity (KL, kg MPa−1 m−1 s−1) as KH divided by the surface area of the leaves distal to the cut end of the segment. KL expresses the ability of the shoot xylem to supply transpiring leaves with water (Brodribb & Feild, 2000). The distal leaf area was determined using a digital camera and a thick sheet of glass to press flat leaves next to a ruler. The shoot Huber value (HV), a measure of stem xylem allocational cost relative to leaf area, was calculated as the sapwood area divided by the leaf area.
Xylem anatomy and wood density
Ontogenetic changes in T. cordata wood structure were determined by comparing the xylem characteristics of juvenile (diameter, 2.5–5 mm) and mature (diameter, 15–17 mm) stems. The following characteristics of tracheids were measured: length, lumen diameter, hydraulically weighted lumen diameter (DH), wall thickness and pit membrane surface area (percentage of tracheid volume approximated as a cylinder; Hacke et al., 2007; Hudson et al., 2010). The transverse stem surface area allocated to tracheids vs living tissues (rays and parenchyma) was measured on safranin-stained sections (Hudson et al., 2010). The wood density (ρwood, g cm−3) from stems of all species was measured using volumetric displacement (Hudson et al., 2010).
Leaf gas exchange and leaf hydraulic conductance
Leaf photosynthetic rates (A, μmol CO2 m−2 s−1) under saturating light (1250–1400 μmol m−2 s−1) and optimal conditions of water availability (i.e. leaf water potentials > − 0.9 MPa) were measured using an open gas exchange analyzer (Li-6400XT; LiCor Biosciences, Lincoln, NE, USA). Leaf temperatures were 18–20°C, with CO2 at 380 μmol mol−1 and a vapor pressure deficit of 0.9–1.3 kPa. Measurements under these conditions were taken as Amax for each species. Mean interspecific A values were calculated from three leaves from three to five individual plants measured between 09:00 and 11:00 h.
Leaf hydraulic conductance (Kleaf, mmol m−2 s−1 MPa−1) was determined from the relaxation kinetics of water uptake by shoots attached to a flow meter (Brodribb et al., 2007). We cut shoots with a distal leaf area of 80–200 cm2 from plants at 09:30–11:00 h. These samples were double bagged in plastic and moved to a field camp. After 5 min, the water potentials of two to four leaves (Ψleaf) from the shoot were determined. We measured Ψleaf with a pressure chamber (PMS-1000 chamber; PMS Instruments Inc., Corvallis, OR, USA), a digital pressure gauge (± 0.001 MPa; Ashcroft, Stratford, CT, USA), a ×10 hand lens and compressed N2 gas. Ψleaf ranged from − 0.35 to − 1.1 MPa. The shoot was then attached to the flow meter after re-cutting the end of the stem underwater by at least 2 cm from the first cut in air to remove embolized tracheids. The shoot was attached to a silicone tube filled with filtered, degassed water. From the rehydration kinetics of the water uptake by shoots as the negative Ψleaf relaxed, we calculated the maximum Kleaf, as described previously (Brodribb et al., 2007). Because of logistical constraints, Amax and Kleaf data were available for some of the New Guinean taxa (T. cordata, T.coriacea, T. crassipes, T. heteromera, T. montis-wilhelmii and T. subalpina) and one species from Australia (T. lanceolata).
Lianoid habit of T. cordata
The seedlings and saplings of T. cordata occurred as unbranched treelets, producing leaves in pairs and/or rosettes of up to four leaves (Fig. 1a,b). The internodes between rosettes were up to 40 cm long. Beginning as small as 0.5 m tall, saplings buckled pendantly into the surrounding shrubby understory (Fig. 1a). By leaning, the wiry stems and/or stiff leaf bases of T. cordata grappled or became cradled in the branch junctions of neighboring plants (Fig. 1b,c). No stems of T. cordata twined around neighboring plants. Following anchorage to a host trellis, T. cordata branched with several upward-growing shoots. The lean–lodge–branch pattern resulted in a sparsely leaved network of wiry stems climbing up to 5.5 m above the forest floor in 7–8-m-tall forest. Mature plants consisted of a main lianoid stem with a mean diameter ±SD of 17 ± 2.5 mm (n =10 individuals; Fig. 1d,e), and lianas formed small canopies in the forest subcanopy layer (Fig. 1e). Lianoid stems of T. cordata attached loosely to their hosts because plants could be pulled down without breaking the branches of the liana or the host (Fig. 1e). Compared with self-supporting Tasmannia species, mature lianas showed extensive horizontal exploration of the forest understory and subcanopy zones. Searcher branches spread as far as 8 m away from the main stem. Only plants that climbed flowered, with flowers produced on pendant side branches that dangled over small subcanopy gaps. Tasmannia cordata leaves were cordate shaped throughout ontogeny, and the leaves possessed a thickened petiole base (Fig. 2a).
Developmental changes in xylem structure, mechanical properties, wood density and microfibril angle of T. cordata stems
The stems of T. cordata were predominantly circular in outline, with a small pith (< 2 mm in diameter) and thin bark at < 15% of the total stem diameter in mature stems. Tracheids dominated the stem cross-section (Table 2; Fig. 2b,c). Juvenile (diameter, 4.0–6.5 mm) and oldest mature stem wood (diameter, 15.0–17.5 mm) stems differed in the following ways: tracheids in mature wood were 29% wider; tracheid density was 48% lower in mature wood; tracheids of mature wood were 45% longer; and tracheid walls were, on average, 25% thicker in mature wood (Table 2). Other wood structural characteristics did not differ with stem age (Table 2).
Table 2. Comparative quantitative anatomical characteristics of stem secondary xylem of Tasmannia cordata plants
Tracheid diameter (TD, μm); maximum tracheid diameter (TDmax, μm); hydraulic mean diameter (DH, μm); tracheid density (TDensity, number mm−2); lumen area (LArea, mm2); tracheid length (TL, mm); tracheid wall thickness (TW, μm); tracheid pit area per tracheid (Tpitarea, mm2); percentage area of stem cross-section occupied by tracheids (%Atracheid); percentage area of stem cross-section occupied by rays (%Arays). For juvenile and mature stems, five stems were measured with the following sample sizes: TD, TDmax, DH and TW (n =250); TL (n =150); Tpitarea (n =25). For Atracheid, Arays, TDensity and LArea, five stems of juvenile and mature wood were measured. Significant differences between trait value means, based on Mann–Whitney U-tests, are denoted as: ns, not significantly different; *, P <0.5; **, P <0.1; ***, P <0.01.
11.6 ± 2.1
17.1 ± 0.8
13.2 ± 0.9
2941 ± 505
0.299 ± 0.015
1414 ± 341
3.69 ± 0.79
0.057 ± 0.0055
75.5 ± 2.5
24.6 ± 4.3
16.24 ± 3.5
18.3 ± 1.8
15.3 ± 1.4
1552 ± 128
0.315 ± 0.075
2190 ± 396
3.96 ± 0.76
0.049 ± 0.0077
76.3 ± 2.6
23.7 ± 4.4
Nonclimbing (leaning) and lianoid stems did not differ in ES among the < 5.0-, 5.1–10.0- and 10.1–15.0-mm-diameter classes (Fig. 3a). In addition, ES of lianoid stems of > 15.1 mm in diameter did not differ significantly from the means found across all other climbing and nonclimbing stem diameter classes (Fig. 3a). X-Ray analyses of wood (secondary xylem) structure–function for T. cordata revealed small changes in ρwood, MFA and EX from pith to bark (Fig. 3b–d). First, there was a slight (12%) decrease in ρwood from youngest (0.69 g cm−3) to oldest (0.60 g cm−3) secondary xylem (Fig. 3b). Second, MFA of the secondary xylem decreased overall from 8° to 5°. EX varied from 17 000 to 19 000 MPa with no developmental trend (Fig. 3d).
Stem xylem hydraulic–biomechanical properties and leaf functional performance of T. cordata vs nonclimbing congeners
Across all species, we found three relations in stem hydraulic performance: (1) KS increased with DH (r2 = 0.63, y = 0.053x − 0.344, P =0.07; Fig. 4a); (2) HV decreased as KS increased (Fig. 4b); and (3) KL increased linearly with KS (Fig. 4c; r2 = 0.44, y = 0.813x + 0.454, P =0.08). KS varied nearly six-fold from 0.12 ± 0.065 kg m−2 s−1 MPa−1 in T. lanceolata to 0.69 ± 0.067 kg m−2 s−1 MPa−1 for T. membranea. KL varied three-fold across the sampled taxa from (0.46 ± 0.072) × 104 kg m−2 s−1 MPa−1 in T. lanceolata to (1.24 ± 0.045) × 104 kg m−2 s−1 MPa−1 in T. coriacea (Fig. 4c). Compared with its congeners, mean values of KS, HV and tracheid DH of T. cordata nested in the middle of the other taxa trait ranges (Fig. 4a,b). KL of T. cordata fell below the middle portion of the nonclimber taxa range (Fig. 4c).
ES of T. cordata stems (7249 ± 1222 MPa; n =70; for all diameter classes) was two to eight times greater than that of nonclimbing Tasmannia (Fig. 5a). ES across nonlianoid Tasmannia ranged from 980 ± 250 MPa in T. vickeriana to 3272 ± 890 MPa in T. heteromera. Across all Tasmannia taxa, no relation between ES and ρwood was found (Fig. 5a). However, there was a negative correlation between ρwood and KS (r2 = 0.61, y = − 0.28x + 0.68; P =0.075 Fig. 5b). For six Tasmannia taxa, ES was related to MFA (r2 = 0.70; y = − 269.8x + 7929; P =0.06 Fig. 5c). Tasmannia cordata exhibited the lowest average MFA at 7.5°, whereas other taxa means varied from 15° to 28° (Fig. 5c). Finally, T. cordata functioned with the lowest mean Kleaf and Amax relative to five New Guinean and one temperate Australian nonclimbing Tasmannia (Fig. 6).
We found that T. cordata occurred as a liana with tracheid-based homoxylous wood. Tasmannia cordata climbed as a branch climber (Rowe et al., 2006), meandering up to the forest subcanopy by grappling onto the stems of other plants with stiff cordate leaves and wiry branches (Fig. 1). Two morphological specializations related to T. cordata’s climbing habit included long stem internode lengths and cordate leaves arranged in a whorled, flattened pattern (Fig. 1). The relatively long internode lengths of T. cordata, for example, may increase trellis discovery by searching branches (Putz, 1984). Instead of being related to the capture of bright light (cf. Givnish & Vermeij, 1976), we believe that the cordate leaves function in how T. cordata climbs. For example, leaves of T. cordata were stiff to tip deflection, particularly at the cordate base. Hooking onto neighboring plants in several directions by leaves appears to be aided by the high stiffness of leaves, as well as the whorled architecture of the leaf arrangement in T. cordata. In addition, T. cordata apparently is not a high light-tolerant liana, instead showing a preference for climbing underneath the forest canopy. The cordate leaves of T. cordata lacked the high gas exchange capacities that are widespread across most studied vessel-bearing angiosperm lianas (Fig. 6; Brodribb & Feild, 2000; Brodribb et al., 2007; Feild & Balun, 2008; Zhu & Cao, 2009). To our knowledge, no other Tasmannia or Canellales species possesses cordate leaves and internodes as long as those of T. cordata (Wilson, 1966; Vink, 1970; T. S. Feild, personal observations, 2009–2010). Thus, these features most probably represent newly acquired traits with the origin of lianescence.
Lack of increased xylem hydraulic function with the evolution of vessel-less angiosperm lianescence
Our results also did not support our expected hypothesis that lianescent T. cordata would possess greater leaf or stem hydraulic capacities than its nonclimbing vessel-less relatives. The hydraulic properties of T. cordata stems, including DH, KS and KL, nested in the low to middle regions of the trait spaces populated by nonscandent Tasmannia (Fig. 4). Other tropical free-standing Winteraceae, such as Takhtajania perrieri and Zygogynum species, also functioned with nearly three times greater KS and KL than T. cordata (Feild et al., 2002; Hacke et al., 2007; Hudson et al., 2010). Taxa with xylem hydraulic capacities lower than that of T. cordata included temperate alpine shrubs, an epiphyte and two temperate shrub species (Table 1). Lower hydraulic capacities in these taxa were associated with smaller plant sizes (as in the epiphyte) or climates characterized by frequent freeze–thaw events and lower annual precipitation (Vink, 1970; Feild et al., 2002). Surprisingly, T. cordata did not possess the most slender stems for a given amount of leaf area. HV of T. cordata stems of 2–5 mm in diameter fell in the middle portion of the nonclimber trait range. Tasmannia taxa occurring as small trees, as well as several tall shrub taxa in temperate forests and tropical alpine timberlines, functioned with up to 50% lower HV than T. cordata (Table 1; Fig. 4b).
A shift to strong wood with Winteraceae liana evolution
Unexpectedly, scandent stems of T. cordata consisted of very strong wood, based on a three- to eight-fold greater ES than its free-standing congeners. Relative to other lianas, ES of T. cordata ranked in the upper 10% of moduli reported in a survey of 43 species (Rowe et al., 2006). Also unlike most lianas, T. cordata maintained high ES throughout development (Fig. 3a; Rowe et al., 2006). Thus, the lack of changes in wood mechanical properties found over T. cordata’s development paralleled those seen in the mechanics of semi-self-supporting shrubby scramblers and juvenile growth in some angiosperm lianas (Speck & Rowe, 1999; Rowe et al., 2006; Ménard et al., 2009). However, unlike these angiosperm lianas, T. cordata exhibited a more protracted phase of much greater mechanical stiffness, and high stiffness was associated with extensive horizontal exploration of the understory and subcanopy zones.
Recent evidence suggests that an increase in ρwood is primarily how vessel-less homoxylous conifers increase mechanical strength, albeit with decreased hydraulic flow (Pittermann et al., 2006, 2011). Compared with most congeners, T. cordata possessed dense wood. Compared with vesselled lianas, T. cordata’s wood (0.62 g cm−3) was unusually dense (ρwood of angiosperm and gnetalean lianas ranges from 0.35 to 0.59 g cm−3; n =12 species; Putz, 1990; Putz & Holbrook, 1991; T. S. Feild, unpublished data, 2008). However, high ρwood cannot account for the mechanical performance of T. cordata wood. Such is the case because T. cordata wood did not exhibit the lowest DH and stem hydraulic capacity values, and other taxa with denser woods had smaller diameter conduits and were less conductive (Fig. 5b); moreover, we did not find a relation between ES and ρwood across all species sampled (Fig. 5a). Two groups displaced from the ρwood by ES relation included T. cordata with high ρwood and disproportionately high ES, and T. lanceolata and T. vickeriana with high ρwood and disproportionately low ES. Over the remaining nonclimbing taxa, ρwood and ES were linearly related (Fig. 5a). We found no evidence that T. cordata possessed structurally unusual tracheids, growth rings or specialized ray tissues relative to other Tasmannia that could increase mechanical stiffness (Carlquist, 1975, 1989). Therefore, structural features independent of the suite of tracheid shape traits incorporated into ρwood must have a significant influence on the variation in stem mechanical performance.
We found that the interplay between cell well composition and ρwood explained T. cordata’s high ES. The xylem wall MFA of T. cordata wood was lower than that of other Tasmannia taxa, and there is abundant evidence that the low MFA values measured would increase significantly cell wall resistance to bending (Burgert, 2006; Evans, 2006; Donaldson, 2008; Keunecke et al., 2009). At the opposite end of the mechanical range, T. lanceolata, one of the temperate alpine shrubs with low ES despite high ρwood, possessed the highest MFA (Fig. 5c). The populations of T. lanceolata, as well as the other outlier T. vickeriana, experience deep winter snow loading that presses branches to the ground. Low ES by large MFA may avoid stem snapping by snow loading despite dense wood that would otherwise favor high stiffness (Keunecke et al., 2009). Other cell wall properties influencing ES include lignin content, lignin type and S1–S3 cell wall and matrix composition (Burgert, 2006; Jagels & Visscher, 2006; Keunecke et al., 2009). However, the strong relation between ES and MFA in Tasmannia argues that MFA is the key variable.
Why is T. cordata wood so strong? Indeed, the magnitude by which ES in T. cordata overshadows that of other Tasmannia suggests that the demands for stem rigidity are greater for a liana versus a diverse range of self-supported growth forms. To climb, we hypothesize that the functional constraints of vessel-less homoxylous wood enforce the high ES of T. cordata. For example, the structural simplicity of vessel-less homoxylous wood means that T. cordata lacks specialized tissues and cell types, such as variant cambia and associated parenchymatous fiber tissues that jacket conducting tissues (Putz & Holbrook, 1991). Such tissues offer novel possibilities of dissipating torsional stress, controlling stem fracture and maintaining xylem–phloem transport in response to stem twisting (Putz & Holbrook, 1991). Instead, T. cordata scandent stems, in functioning with high durability, are unlikely to break or become twisted and lose hydraulic conductance after inevitable fallout from their host and during other disturbances that lianas endure (Putz, 1984). Consistent with such a hypothesis, T. cordata stems appeared to be unbreakable in the field. Of the 20 lianas we observed over 2 yr, none exhibited snapped stems, even though 90% of plants possessed several scandent branches that fell from the canopy onto the forest floor or became tangled in piles of woody forest gap debris with windthrow of the host (T. S. Feild and L. Balun, field observations, 2009–2010). Interestingly, even the bark of T. cordata was not structurally modified or exceptionally thick, to add in some compliancy to stem bending as in other lianas (Rowe et al., 2006; T. S. Feild, unpublished observations, 2010).
The mechanism by which T. cordata achieves strength offers a key advantage. The interplay between MFA and ρwood found in T. cordata represents a simple mechanism to afford high ES that can meet the mechanical demands of climbing, whilst avoiding a large sacrifice in stem hydraulic capacity. High ES underpinned primarily by low MFA affords mechanical efficiency by reducing the wood investment necessary to generate high flexural stiffness (Evans, 2006). In addition, the mechanism of relying on MFA to achieve high ES is hydraulically efficient because a decrease in tracheid size and increase in tracheid wall thickness, which reduce xylem water transport capacity and increase construction costs, are sidestepped to develop high ES (Burgert, 2006; Pittermann et al., 2006). This mechanism is probably particularly effective in vessel-less angiosperm wood because of the large contribution made by tracheid walls to the stem bending plane.
Evolutionary significance of Winteraceae lianescence
The mechanism by which T. cordata’s wood functions supports a hypothesis on the key role played by vessels in angiosperm liana evolution. Future studies are necessary, but our results are consistent with the hypothesis that large vessels unburdened the spatial limitations of producing wood with high hydraulic throughput based on tracheids, and may have permitted the evolution of other cell types and tissues to meet the mechanical demands of lianescence in novel ways (Sperry et al., 2007). Thus, vessels are developmental enablers (Donoghue, 2005) for lianescence. Because angiosperms apparently acquired vessels early during their evolution, vessel evolution may explain how lianescence is such a widespread early evolved shift in angiosperm evolution (Feild et al., 2009).
By contrast, T. cordata represents a different way to make a liana, but, without vessels, it is an evolutionary dead end for several reasons. For example, the montane cloud forest habitat of T. cordata is hydraulically permissive for a vessel-less angiosperm liana because evaporative demand is low throughout the year as a result of cool temperatures, high rainfall (4000–6000 mm yr−1) and frequent wind-driven mist (therefore the maximization of water transport is of negligible competitive significance) and competition from larger conduit-bearing lianas is probably low because of intermittent freeze–thaw conditions (Wade & McVean, 1969; Hope, 1976; Schnitzer, 2005). It seems counterintuitive to hypothesize a lianoid origin favored by a hydraulically nondynamic and frosty environment (Schnitzer, 2005). However, such conditions are probably critical for T. cordata in reducing competition by other lianas with large vessels. Such is the case because T. cordata’s small-diameter tracheids conduct little water but are in the size range that is impervious to freeze–thaw embolism (Feild et al., 2002). Finally, another permissive aspect to T. cordata’s habitat is that the exploration of the arboreal zone is fairly straightforward even for a short, vessel-less, dense-wooded and probably slow-growing liana. This is because the branchy aspect of short upper montane cloud forests furnishes an abundance of branch trellises that are reached at less than a meter tall (Wade & McVean, 1969). In the short forests at Mt Wilhelm, a possible advantage of the lianoid habit in T. cordata is that, like another basal angiosperm liana with low gas exchange capacity (Austrobaileya scandens), plants can explore more horizontal space across the forest floor and take advantage of small openings in the subcanopy and sunflecks (Feild et al., 2003; Rowe et al., 2006). Indeed, mature T. cordata lianas are longer in length than other self-supporting Tasmannia taxa are in height (Vink, 1970).
Origin of Winteraceae lianescence
What developmental processes influenced the evolution of lianescence in Winteraceae? A clue shedding light on this question is the growth habit of T. heteromera at Mt Wilhelm. We have observed, as others, that the main stems of T. heteromera lean over and continue to grow horizontally for up to 3 m from the main base in the understory (Venables, 1984). Biomechanically, T. heteromera approaches the performance of T. cordata, with the second largest ES and moderately high ρwood. However, T. heteromera lacks the suite of traits linked to lianescence in T. cordata, including long internodes, stiff cordate leaves and low MFA (Vink, 1970). Tasmannia heteromera is not in the same lineage as T. cordata (Supporting Information Methods S1, Fig S1), but its growth form supports the hypothesis that leaning is an early step towards lianescence (Ménard et al., 2009). In Winteraceae, we would add that increased ES and ρwood, as well as the occurrence in the understory and the expression of leaning earlier in development (Fig. 1a), are additional prerequisites. Because leaning is associated with increased stiffness and denser wood in both Tasmannia taxa, changes in leaf area and the placements of canopy mass drive pendant buckling.
Intriguingly, low MFA appears not to be a novelty of T. cordata, unlike its expression in the stem. Root systems do not permit the same bending tests made on stems. However, preliminary measurements on roots have revealed low MFA (5°–10°) and ρwood ranging from 0.59 to 0.65 g cm−3– the hallmarks of high ES (T. S. Feild, unpublished data on roots of T. coriacea, T. heteromera, Zygogynum. sp. nov. Mt Gumi PNG and T. montis-wilhelmii, 2011). Thus, roots of some nonclimbing Winteraceae apparently function mechanically like the lianoid wood of T. cordata. Future research is necessary; however, an implication is that the transfer of root biomechanical function into the shoot developmental plan occurred during the evolution of lianescence in T. cordata. In some ways, arboreal lianoid stems and roots of Tasmannia occur in mechanically comparable environments. Roots experience minimal compressive forces from gravity because of support by soil. However, tensile stresses may be large because the taxa with low-MFA roots occur in loose watery soils or on steep slopes (Matsumura & Butterfield, 2001).
Relicts that radiated recently in the New Guinea highlands
The Winteraceae is an ancient lineage, stemming from the Early Cretaceous diversification of angiosperms (Feild et al., 2002; Marquínez et al., 2009). Our results on the function and phylogeny of T. cordata offer evidence that functional diversification has been rapid and recent in New Guinea. Growth forms in New Guinean Tasmannia include epiphytes, a semi-scandent leaner, trees, a host of microphyllous alpine shrubs and now a liana (Smith, 1943; Vink, 1970). We have some preliminary evidence for phylogenetic structure among several of the New Guinean taxa (Methods S1). By combining such evidence with the diverse vegetative morphologies found in Tasmannia (Vink, 1970), we believe that the vast majority of the ‘Drimys piperita complex’ entities are species. The reconstructed timeframes for radiation within New Guinean Tasmannia (Oligocene–Miocene) suggest that there have been bursts of diversification within the clade (Fig. S1, Methods S1). However, at present, the phylogeny of New Guinean Tasmannia is poorly resolved and requires more markers and greater taxon sampling (Methods S1). Understanding the causes of these diversification events requires further research, given that there may be > 50 Tasmannia species spread across a diverse range of environments in New Guinea (Methods S1). However, on the basis of the functioning of the xylem of Tasmannia, it is likely that ecological radiation tracked the waves of montane uplift and the variations in cloud base and treeline altitudes – the geological and climatic processes that changed the geographic extent of the very wet hydrological envelope and subalpine zone habitats in which Tasmannia have flourished over the past 34–11 million yr (Feild et al., 2002; Hoorn et al., 2010).
Tasmannia cordata is an example of angiosperm lianescence uncoupled from xylem conduit gigantism, as well as the high plasticity and high cell type diversity in cambial development, which are features emblematic of all other known lianas. However, without highly conductive vessels, strong wood with low hydraulic conductivity is necessary to meet the mechanical demands of climbing. As a liana with low hydraulic capacity, T. cordata is a far cry in ecophysiological performance from the types of vesselled angiosperm lianas that add new dimensions of ecosystem function to high-productivity forests. Future studies are necessary, but our results argue that, rather than a passive consequence of mechanical parasitism, vessel evolution may have enabled lianescence by minimizing the hydraulic allocational constraints on cambial development.
We thank Vincent Don and Kegesugl landowners for their warm hospitality, Samson Takai and other porters for hauling the gas cylinder and Li-COR 6400 to the treeline at Mt Wilhelm, and Papua New Guinea Forest Products for access to the Mt Gumi camp. We appreciate helpful discussions from Tim Brodribb, John Sperry and Jack Putz. We thank Brian O’Meara for advice on molecular clock analyses. Our research was supported by a National Science Foundation research grant (IOB-0714156) and was permitted by the National Research Institute in Papua New Guinea.