American Journal of Botany 96(12): 2214-2223. 2009. COORDINATION OF FOLIAR AND WOOD ANATOMICAL TRAITS CONTRIBUTES TO TROPICAL TREE DISTRIBUTIONS AND PRODUCTIVITY ALONG THE MALAY-THAI PENINSULA1 JENNIFER L. BALTZER,2'3'7 DORTHEA M. GREGOIRE,2 SARAYUDH BUNYAVEJCHEWIN,4 N. SUPARDI M. NOOR,5 AND STUART J. DAVIES3'6 2Biology Department, 63B York Street, Mount Allison University, Sackville, New Brunswick, E4L 1G7 Canada; 3Center for Tropical Forest Science-Arnold Arboretum Asia Program, Harvard University Herbaria, Harvard University, Cambridge, Massachusetts 02138 USA; 4Royal Forest Department, Chatuchak, Bangkok 10900 Thailand; Torest Research Institute Malaysia, Kepong 52109 Selangor, Malaysia; ^Center for Tropical Forest Science, Smithsonian Tropical Research Institute, P.O. Box 0843-03092, Balboa, Panama, Republic of Panama Drought is a critical factor in plant species distributions. Much research points to its relevance even in moist tropical regions. Recent studies have begun to elucidate mechanisms underlying the distributions of tropical tree species with respect to drought; however, how such desiccation tolerance mechanisms correspond with the coordination of hydraulic and photosynthetic traits in determining species distributions with respect to rainfall seasonality deserves attention. In the present study, we used a common garden approach to quantify inherent differences in wood anatomical and foliar physiological traits in 21 tropical tree species with either widespread (occupying both seasonal and aseasonal climates) or southern (restricted to aseasonal forests) distributions with respect to rainfall seasonality. Use of congeneric species pairs and phylogenetically independent contrast analyses allowed exami- nation of this question in a phylogenetic framework. Widespread species opted for wood traits that provide biomechanical support and prevent xylem cavitation and showed associated reductions in canopy productivity and consequently growth rates compared with southern species. These data support the hypothesis that species having broader distributions with respect to climatic vari- ability will be characterized by traits conducive to abiotic stress tolerance. This study highlights the importance of the well-estab- lished performance vs. stress tolerance trade-off as a contributor to species distributions at larger scales. Key words: abiotic stress; climatic variability; gas exchange; geographic range size; Malaysia; Thailand; tropical forest dynam- ics; wood anatomy. The connection between plant traits and environmental conditions has long been recognized (e.g., Cowles, 1899) and continues to be an active area of research contributing sub- stantially to our understanding of species distributional pat- terns at local, regional, and global scales (Wright et al., 2006; Ackerly and Cornwell, 2007; Cornwell and Ackerly, 2009). Much recent work has emphasized trade-offs between perfor- mance and stress tolerance with respect to foliar and wood traits (Reich et al., 2003; Wright et al., 2004; Hacke et al., 2006; Chave et al., 2009) as well as whole-plant measures such as growth and mortality (Russo et al., 2005, 2008). Fur- thermore, there is functional coordination among these traits with limits as to the trait combinations and values possible 1 Manuscript received 8 December 2008; revision accepted 13 August 2009. Many thanks to the Forest Research Institute of Malaysia and the Thai Royal Forest Department for permission to conduct this research. S. Phillips was integral in the experimental establishment, and F Abidin assisted in nursery care. S. Nishimura (NIES) provided nursery space at Pasoh and assistance in various forms. Thanks to S. Thomas for use of gas- exchange equipment and taxonomic assistance. Wood anatomical trait measurement was greatly facilitated by M. Schneider through generous use of equipment and methodological direction. Many thanks to B. Pratt for his thorough and helpful editorial contributions and three anonymous reviewers. Research was supported by the Center for Tropical Forest Science-Arnold Arboretum Asia Program, a CTFS research grant, and Natural Science and Engineering Research Council of Canada Postdoctoral and Discovery funding to J.L.B. 7 Author for correspondence (e-mail: jbaltzer@mta.ca) doi:10.3732/ajb.0800414 (Wright et al., 2004; Chave et al., 2009). Generally speaking, we expect that a species will be characterized by a suite of functional traits that provide some adaptive advantage in terms of growth and/or survival in its native habitat but may have negative consequences for competitive ability or survival in another habitat (Ackerly, 2003). Species range sizes vary by orders of magnitude, and with increasing range, species encounter greater variability with re- spect to environmental conditions. It is predicted that species having the capacity to span broader environmental gradients will necessarily be more tolerant of abiotic stress (Morin and Chuine, 2006). Recent work examining the distribution of tropical tree species with respect to a rainfall seasonality gra- dient provided evidence of this trade-off in that widespread species occupying both aseasonal and seasonal forests had lower growth rates and sensitivity to local habitat variability when compared with species restricted to the aseasonal forests (Baltzer et al., 2007). The basis of such differences presum- ably lies in trade-offs occurring among the functional traits that are best suited for the resource availability encountered by each distributional grouping. If seasonal drought is contribut- ing to these distributional limits then we would expect to see more conservative functional traits that enhance drought toler- ance in the widespread species. Lethal water potential (i.e., the lowest water potential at which a plant can maintain living tis- sue) and the corresponding minimum relative water content have been shown to correspond well with tree species' distri- butions along rainfall and seasonality gradients in the equato- rial tropics (Engelbrecht and Kursar, 2003; Baltzer et al., 2008). However, the mechanistic basis of differential lethal 2214 December 2009] BALTZER ET AL.?TRAIT COORDINATION AND TREE SPECIES DISTRIBUTION 2215 water potentials presumably lies in the ability to prevent or minimize cavitation and resulting xylem dysfunction during drought, thereby maintaining water flow through the plant and supporting living tissues. There should thus be an important contribution of wood anatomical traits (either via contribution to cavitation avoidance or tolerance) with implications for po- tential canopy productivity due to the inherent coordination of these traits. In the present study, we attempt to quantify such functional coordination of foliar and wood anatomical traits and its role in tropical tree distributions with respect to rainfall seasonality. It is well known that canopy photosynthesis is largely under the control of hydraulic architecture as stomatal response is critical to avoidance of hydraulic limitations (Meinzer et al., 1995; Nardini and Salleo, 2000; Sperry, 2000). Furthermore, hydraulic and xylem anatomical traits form the basis of species' resistance to loss of xylem function via cavitation (the sponta- neous transition of water from liquid to gas when under exces- sive tension). It is well accepted that an important mechanism affording drought tolerance is a "safe" hydraulic system but that the traits providing such safety may compromise the hy- draulic efficiency via increased resistance to water transport (Baas et al., 2004; Hacke et al., 2006; Pratt et al., 2007). In this way, safety from embolism limits canopy productivity and should result in lower growth rates. Thus, one would anticipate that the leaf and wood economics spectra will also be corre- lated, with stress tolerant species having both conservative fo- liar and safe wood traits. Indeed, there is a growing body of theoretical and empirical evidence of the coordination between the plant hydraulic system and foliar photosynthetic traits (Bro- dribb and Feild, 2000; Brodribb et al., 2002; Katul et al., 2003; Bucci et al., 2004; Santiago et al., 2004; Ishida et al., 2008). Poleward from the equator, a strong gradient in rainfall sea- sonality impacts forest structure, diversity, and productivity (Gentry, 1988; Clark et al., 2001; Engelbrecht et al., 2006). Wa- ter availability exerts a strong influence over wood anatomical structure and clear patterns of differentiation in the conduit structure have long been recognized between dry and wet cli- mates (e.g., Carlquist, 1977). Furthermore, global examinations of hydraulic traits as a function of climate have demonstrated a significant correlation between resistance to cavitation and mean annual precipitation in evergreen angiosperms (Maherali et al., 2004). In the present study, we quantified a range of wood anatomical features and foliar physiological traits in tree spe- cies occurring along the Malay-Thai Peninsula having wide- spread (occurring in both aseasonal and seasonally dry forests) or southern (restricted to aseasonal forests) distributions. We predicted that wood anatomical and leaf trait coordination should be evident such that "safer" wood anatomical features that contribute to cavitation resistance during drought will come at the cost of lower canopy productivity. Therefore, we would expect more conservative leaf traits in those species capable of occupying seasonally dry forests. Baltzer et al. (2007) previ- ously demonstrated performance differences along this same gradient, and we anticipate that coordination of the hydraulic system and photosynthetic traits underlies these performance- related differences. MATERIALS AND METHODS Study area?The Kangar-Pattani Line (hereafter the KPL) is a key phyto- geographical transition that bisects the Thai-Malay Peninsula near the political border (Van Steenis, 1950). At this transition, over 500 genera of vascular plants meet their distributional limits. Because there is presently no geophysical barrier at the KPL, two mechanisms have been proposed as potential explana- tions for this floristic transition. The first is historical in nature and hypothesizes that during the Miocene and Pliocene eras seaways bisected the Thai-Malay Peninsula at both the KPL and the Isthmus of Kra (Fig. 1) (Woodruff, 2003). A second, environmentally derived mechanism has also been invoked for explain- ing distributional patterns at the KPL. Specifically, the KPL coincides with a major climatic transition that occurs along the Peninsula; forests south of the KPL are perhumid (i.e., relative humidity typically 100% or greater) and asea- sonal, while those to the north are seasonally dry with a 2-3 mo drought with little or no change in total annual rainfall; such a climatic transition could pre- sumably play a critical role in determining the distributions of plant species (Whitmore, 1984; Ashton, 1997). Forests to the north of the KPL are evergreen, seasonal forests with low incidence of drought deciduousness. Plant material and growth conditions?The mature seeds of 21 tree species (Table 1) were collected between July and October 2005 in the Pasoh Forest Reserve, Malaysia (2?58'N, 102?18'E) and the Khao Chong Peninsular Botani- cal Garden, Thailand (7?34'N, 99?47'E) (hereafter Pasoh and Khao Chong; Fig. 1). Average rainfall in Pasoh is 1950 mm-y-1 (0 mo of drought) and 2700 mm-y-1 (2-3 mo of drought) in Khao Chong. Minimum/maximum tempera- tures daily temperatures in Pasoh and Khao Chong are 23/33?C and 22/34?C, respectively. Species were classified as southern or widespread based upon flo- ristic records for the region. Primary sources were the Tree Flora of Malaya (Whitmore, 1972), Flora Malesiana (Van Steenis, 1950), and Flora of Thai- land (Smitinand and Larsen, 1970), all of which list the states/provinces and other countries in which the species occurs. Additional range data were col- lected from other reliable sources (e.g., Symington, 2004; Van Welzen and Chayamarit, 2005). Species with no floristic records in the seasonally dry for- ests north of the KPL were classified as southern, while those whose distribu- tions traversed the rainfall gradient were classified as widespread. For all but two species, the seeds of two or more individuals were obtained (Table 1). All seeds were germinated and seedlings grown in polybags (15.2 cm diameter x 22.9 cm height) filled with clay-rich soil collected near Pasoh. A wooden struc- ture covered with neutral density shade cloth provided 15% full sunlight during the study. Minimum/maximum temperature, humidity, and vapor pressure defi- cit (VPD) within the structure were 23/31?C, 62/100%, and 0/0.23 psi, respec- Khao Chong, Thailand ^x^. Kangar-Pattani Line Pasoh, Malaysia \? K 4 ^i? ^J\ Fig. 1. Map of the locations of the Kangar-Pattani Line (KPL; dashed line), Pasoh Forest Reserve (Pasoh; Peninsular Malaysia, 2?58'N, 102? 18'E) and Khao Chong Peninsular Botanical Garden (Khao Chong; Peninsular Thailand, 7?34'N, 99?47'E). North of the KPL the climate is largely sea- sonal while to the south rainfall is primarily aseasonal. Modified from Bal- tzer et al. (2008). 2216 AMERICAN JOURNAL OF BOTANY [Vol. 96 TABLE 1. List of study species, family and distributions (Dist; S = southern, W = widespread) in relation to the Kangar-Pattani Line. Habit indicates the location in the canopy: E, emergent; M, main canopy; U, understory. Abbreviations (Abb.) correspond to those in Figs. 2-4. Location indicates the seed collection site (P = Pasoh, KC = Khao Chong); number of seed sources is indicated in parentheses. In the Dipterocarpaceae, mast fruiting often did not allow accurate counts of seed sources; thus the numbers represent the seed collection locations. Perforation plate type and vestures are defined in Table 2. Family Species Perforation plate Dipterocarpaceae Parashorea stellata Kurz. Parashorea densiflora (Y.S1.) ex Sym. Shorea guiso (Blanco) Blume Shorea lepidota (Korth) Blume Shorea macroptera Dyer Shorea parvifolia Dyer Euphorbiaceae Mallotus penangensis Mull. Arg. Neoscortechinia kingii Hk.f. Fabaceae Millettia atropurpurea (Wall.) Benth. Sindora coriacea (Baker) Prain Sindora wallichii Graham ex Benth Fagaceae Lithocarpus wrayi (King) A. Lamus Quercus semiserrata Roxb. Phyllanthaceae Aporosa globifera Hook f. Aporosa microstachya Hook f. Aporosa symplocoides (Hook f.) paqx Polygalaceae Xanthophyllum affine Korth. Sapotaceae Palaquium maingayi K&G Palaquium sumatrana Burck Payena lucida (Don) DC Violaceae Rinorea anguifera (Lour.) OK E pst W KC(5) Scalariform No ME pcle s P(l) Simple No E sgu w PC) Simple No E sle s P(>3) Simple Yes E sma s P(>3) Simple Yes E spa s P(>3) Simple Yes M mpe s PC) Simple Yes M nki s PC) Simple No M mat w P (2); KC (5) Simple Yes E SCO w P (3); KC (2) Simple Yes M swa s PC) Simple Yes M lwr s PC) Scalariform No M qse w KC(2) Simple No U agl s PC) Scalariform No U ami w PC) Simple No U asv s PC) Scalariform No M xaf w P(l) Simple Yes M pma s PC) Scalariform No M psu w KC(2) Simple No M plu w P(>3) Simple No U ran w PC) Simple No lively. Seedling locations within the nursery were randomly assigned and rotated regularly to avoid confounding effects of light and temperature gradi- ents. Seedlings were watered daily if diurnal rainfall had been insufficient. Af- ter 1 year of growth, stems and leaves were harvested from the saplings. The stems were stored in 70% ethanol for processing at a later date. Foliar trait measurements?Using a LI-6400 gas-exchange system (Licor, Lincoln, Nebraska, USA) we measured gas exchange on recent, fully expanded leaves of four individuals per species (and provenance where seeds of multiple provenances were collected). Measurements were made from 9 to 16 July 2006 before noon with cuvette conditions maintained at 370 ppm C02 and 60-80% relative humidity (RH) and leaf temperatures between 25 and 30?C. Gas-ex- change measurements were made at light levels of 0 and 1500 umol-m~2-s_1. Stability of gas-exchange values was determined visually using graphics avail- able in the LI-6400 programming. Maximum stomatal conductance to water (gs) corresponding to Amax was used to characterize stomatal response across species and distribution. Gas-exchange leaves were harvested, measured for fresh leaf area, dried at 60?C, and weighed for calculation of leaf mass per area (LMA). Wood anatomical trait measurements?Stem sections from three speci- mens of each species were used to characterize anatomical traits. Stem cross and tangential sections were made for each sample using a sliding microtome at 20 um and 5 um thickness, respectively. All sections were wet mounted and digital images obtained using a Carl Zeiss Photomicroscope I (Oberkochen, Germany) combined with a PixeLink PL-A662 Camera Kit (Prague, Czech Republic). Photographic scale was calibrated using a standard ocular micrometer slide. All vessels in each image (-100 vessels) were measured. The diameter and total area of every vessel in each cross-sectional image was measured, as was total xylem area. From these measurements, maximum vessel diameter and vessel fraction (vessel area/xylem area) were determined. Vessel diameter contributes directly to the upper limit on rate of flow in the vessel and correlates positively with the pore area in the perforation plates that connect adjacent xylem vessels (Chave et al., 2009). Larger vessels mean more efficient flow but correspond with larger pit area and higher probability of large pit pores. Vessels with large pores are more prone to cavitation by air seeding (Choat et al., 2003; Hacke et al., 2006). Similarly, a larger vessel fraction accommodates greater water flow to the canopy. Hydraulically weighted mean diameter was calculated as: 2(f.rVf/) following Sperry et al. (1994). This weights the proportional importance of radii to the estimated hydraulic conductance of the conduits and corresponds closely with hydraulic conductance (Sperry et al., 1994; Pratt et al., 2007). The double wall thickness of adjacent conduits (i) was quantified from the tangen- tial sections and the maximum conduit span (b) estimated following Hacke et al. (2001); these two metrics allow for the calculation of conduit reinforce- ment as (t/b)2. Wood density was calculated as the oven-dried mass of the wood sample divided by the green volume measured following removal of bark and pith (Hacke et al., 2000). Wood density correlates with cavitation resistance and stress tolerance because high wood density is generally associated with greater structural reinforcement of the xylem (Hacke et al., 2001). Both wood density and conduit reinforcement have been shown to be excellent predictors of the water potential at which 50% loss of hydraulic conductivity by cavitation is observed (Hacke et al., 2001; Pratt et al., 2007). Perforation plate type and vesture presence were also determined from the tangential sections. The type of perforation plates present influence the rate at which water can move up and down the vessels (hydraulic efficiency). Simple perforation plates having a single opening create less friction as water moves through them than do scalari- form perforation plates (multiple elongated openings separated by ladder-like bars) (Baas, 1986; Carlquist, 2001; but see Schulte et al., 1989); however, the relative importance of this is a function of the pore size and number in the sca- lariform perforation plate (Schulte, 1999). Statistical analysis?In two cases (Millettia atropupurea and Sindora cori- acea), seeds were collected from both the Pasoh and Khao Chong provenances. These samples are included in all analyses as separate data points. Sample sizes for gas-exchange and wood anatomy traits were 4 and 3, respectively. To test for the effect of distribution on wood and leaf traits, we conducted a random- ized block ANOVA with distribution as the fixed variable and genus as the random variable. Dependent variables included four foliar traits (area-based photosynthetic and respiration rates [Amax and Rd respectively], maximum sto- matal conductance [gj, and leaf mass per area [LMA]) and five wood traits (maximum vessel diameter, mean hydraulic diameter, wood density, (t/b)2, and vessel fraction). Trait names, abbreviations, and units of measure can be found in Table 2. Species mean trait values can be found in Appendix SI (see Supple- mental Data with the online version of this article). To meet assumptions of normality maximum vessel diameters, vessel fraction, and wood density were log-transformed prior to analysis. To assess how measured traits were associ- ated and where the two distributional groups occurred in multivariate trait space, we conducted a principal component analysis (PCA). Due to unusually high maximum vessel diameter values in the two provenances of Milletia atro- purpurea, this species was excluded from the PC A. LMA was not correlated with any of the significant components, so it was removed from the analysis. Perforation plate type and presence/absence of vesturing were coded with December 2009] BALTZER ET AL.?TRAIT COORDINATION AND TREE SPECIES DISTRIBUTION 2217 dummy variables (0 = simple, 1 = scalariform; 0 = vestures absent, 1 = vestures present). Differences in loadings along the first two axes as a function of distri- bution were tested using ANOVA. These analyses were conducted using the program R (v. 2.1; R Foundation for Statistical Computing, Vienna, Austria). To determine coordination of and potential shifts in hydraulic and photosyn- thetic traits as a function of distribution, we employed standardized major axis regression analysis using the SMATR program (Falster et al., 2003). For these analyses, species' mean values were used. None of the significant linear relation- ships showed elevational shifts in the intercept or shifts along a common slope; therefore, only pooled cross-species bivariate trait correlations are presented. To account for phylogenetic relatedness in the bivariate relationships, we conducted phylogenetically independent contrasts (PICs) (Felsenstein, 1985). We used a maximally resolved tree created in the program Phylomatic 2 (Webb and Dono- ghue, 2007), which employs the angiosperm phylogeny (Stevens, 2001) as the hypothesis for the phylogenetic relationships among study taxa. Phylogenetic re- lationships to the genus level of the broad range of taxa included in our study are well resolved. The tree was then run through the program bladj (Webb and Dono- ghue, 2007), which uses the angiosperm node ages of Wikstrom et al. (2001) to assign conservative branch lengths by placing the nodes evenly between dated nodes, and between dated nodes and terminals. This minimizes variance in branch length, within the constraints of dated nodes and can be a marked improvement on using the number of intervening nodes as a phylogenetic distance (Webb, 2000). Where polytomies occurred that could not be resolved by genus-specific phytogenies, we randomly split the species to force the bifurcating structure nec- essary for PIC. There were only two instances where this had to be done: the light-red meranti section of Shorea (S. parvifolia, S. macroptera and S. lepidota) and the genus Aporosa. Because of the low number of polytomies, altering the order of the branching pattern in these had little impact (data not shown). Correla- tion coefficients were calculated for resulting PICs on a pairwise basis. There were four significant components in the PCA explain- ing 84% of the variation across species; the first two compo- nents explained 55 % of the total variation. The two distributional categories separated significantly along the first axis (Fll9 = 8.12, P = 0.0102); widespread species tended to be located in the right quadrats, while southern species were located in the left quadrats (Fig. 4). All traits with the exception of perfora- tion plate type and gs loaded quite strongly onto the first axis (Fig. 4). This axis essentially corresponds with the safety-effi- ciency trade-off with widespread species located largely to the right of the axis with high wood density and conduit reinforce- ment and southern species to the left with high physiological rates, large maximum vessel diameters and hydraulic means, and high vessel fractions. Along the second axis, there seems to be a bit of a breakdown in terms of the trait coordination with physiological traits and perforation plate type (0 = simple, 1 = scalariform) loading positively and wood anatomical traits, with the exception of the vessel fraction, loading negatively. Distributional grouping (S vs. W) did not separate significantly along the second axis (Fll9 = 1.26, P = 0.2752); however, it should be noted with the exception of perforation plate type (axis 1) and Rd (axis 2), all traits loaded significantly onto both axes (Fig. 4A). In general, species having widespread distribu- tions were characterized as having conservative, stress tolerant traits, while southern species tended to have traits associated with faster growth. RESULTS Does distribution with respect to seasonality predict leaf and wood traits??Distribution with respect to rainfall season- ality affected all wood traits examined with the exception of maximum vessel diameter and mean hydraulic diameter (Fig. 2). Vessel fraction was significantly lower in widespread spe- cies while both (tlb)2 and wood density were significantly higher in species having widespread distributions with respect to rain- fall seasonality (Fig. 2). Foliar traits similarly showed system- atic shifts with respect to species' distributions with the exception of leaf mass per area (Fig. 3). Area-based photosyn- thetic and respiration rates and maximum stomatal conductance were all significantly higher in southern-distributed species though the difference in stomatal conductance was only mar- ginally significant (Fig. 3). Coordination of anatomical and photosynthetic traits? There were several hydraulic-photosynthetic trait pairs that showed coordination within our cross-species analysis (Table 3 A). Both Amax and gs showed a significant, positive relationship with LMA and correlated positively with one another (Table 3A). Rd showed significant positive correlations with maximum vessel di- ameter and mean hydraulic diameter (Table 3A). LMA did not correlate with any of the wood anatomical traits measured. Vessel fraction was positively correlated with both maximum vessel diameter and hydraulic mean, which showed the strongest corre- lation of all pairs, and correlated negatively with wood density. There was a moderate positive correlation between wood density and {tlb)2 (Table 3A). Most of the same relationships existed when phylogeneti- cally independent contrasts were used for the correlation analy- sis, and for the most part, these relationships were strengthened TABLE 2. Symbols, units of measurement, and definitions for all traits measured in the study. Trait Symbol Units Definition Photosynthetic capacity Hmol-nr--s Dark respiration rate ?d Hmol-nr--s Maximum stomatal conductance g, mol-m~2-s~ Leaf mass per area LMA g-cm~2 Maximum vessel diameter Max vessel |_im Vessel fraction Vessel fraction ? Wood density Density g-cm-3 Conduit reinforcement (*/6y ? Mean hydraulic diameter Hydraulic mean |_im Perforation plate type Perforation ? Vestures Vestures Area-based carbon assimilation at saturating light conditions Area-based carbon loss in the dark Resistance for diffusion of water vapor through the stomata Leaf mass expressed per unit leaf area Upper fifth percentile of all cross-sectional vessel diameters Ratio of cross-sectional area occupied by vessels to that occupied by all xylem tissue Mass of oven-dried wood per green volume Double wall thickness (t) relative to the maximum span of the vessel (b) (see Hackeetal., 2001) Hydraulically weighted mean diameter that weights the importance of radii in proportion to estimated hydraulic conductance of the conduits (Sperry et al., 1994) Openings in end wall of vessel elements where primary wall and middle lamella have been hydrolyzed allowing free water flow between vessels Bordered pits where pit chamber or outer pit aperture is wholly or partially lined with projections from secondary cell wall (see Jansen et al., 2004) 2218 AMERICAN JOURNAL OF BOTANY [Vol. 96 E CD in in a > x to V) c cu "D T3 O O E to