Functional trade-offs are driven by coordinated changes among cell types in the wood of angiosperm trees from different climates
Summary
- Wood performs several functions to ensure tree survival and carbon allocation to a finite stem volume leads to trade-offs among cell types. It is not known to what extent these trade-offs modify functional trade-offs and if they are consistent across climates and evolutionary lineages.
- Twelve wood traits were measured in stems and coarse roots across 60 adult angiosperm tree species from temperate, Mediterranean and tropical climates.
- Regardless of climate, clear trade-offs occurred among cellular fractions, but did not translate into specific functional trade-offs. Wood density was negatively related to hydraulic conductivity (Kth) in stems and roots, but was not linked to nonstructural carbohydrates (NSC), implying a functional trade-off between mechanical integrity and transport but not with storage. NSC storage capacity was positively associated with Kth in stems and negatively in roots, reflecting a potential role for NSC in the maintenance of hydraulic integrity in stems but not in roots. Results of phylogenetic analyses suggest that evolutionary histories cannot explain covariations among traits.
- Trade-offs occur among cellular fractions, without necessarily modifying trade-offs in function. However, functional trade-offs are driven by coordinated changes among xylem cell types depending on the dominant role of each cell type in stems and roots.
Introduction
Tree wood is multifunctional, simultaneously providing hydraulic transport, mechanical integrity, storage and defence against pest/pathogen attack (Chave et al., 2009). The secondary wood of angiosperms comprises three main cell types (parenchyma, vessels and fibres) that contribute to these distinct functions. The metabolically active radial and axial parenchyma (RAP) cells play a fundamental role in defence and wound repair (Morris et al., 2020), respiration (Rodríguez-Calcerrada et al., 2015), carbohydrate, nutrient and water storage (Plavcová et al., 2016; Secchi et al., 2017); vessels mainly provide long-distance water and nutrient transport (Tyree & Ewers, 1991; Pratt & Jacobsen, 2017); and fibres, as well as lignified parenchyma, provide mechanical strength (Burgert & Eckstein, 2001; Rana et al., 2009; Ziemińska et al., 2013; Fortunel et al., 2014). Trees have evolved in response to a multitude of diverse environmental factors that have led to trade-offs in functions within the xylem space (Baas et al., 2004; Lachenbruch & McCulloh, 2014; Grubb, 2016; Pratt & Jacobsen, 2017; Pratt et al., 2021a). The resulting structure of the xylem space actively influences physiological processes occurring within trees, potentially driving survival or failure when local environmental conditions change (Jacobsen et al., 2007; Choat et al., 2018).
Wood cellular composition is widely divergent among woody angiosperms. Due to the finite volume or given cross-sectional area, Pratt and Jacobsen (2017) described that proportional trade-offs occurred among the major cell types, with strong and negative correlations between fibre and RAP fractions and between fibre and vessel fractions (VFs). However, Fortunel et al. (2014) found a negative relationship between fibre and parenchyma fractions, but no relationship with VFs across 113 Amazonian rainforest tree species and Pratt and Jacobsen (2017) also found that vessel and parenchyma fractions were not correlated with each other across 36 shrub species. This divergence in wood cell fractions raises important questions regarding trade-offs between functions. Whether the allocation trade-off among cellular types really reflects the trade-off in function is not clear, as cell types and functions are entwined and connected in complex ways.
As the largest metabolically active wood fraction, parenchyma cells and the nonstructural carbohydrates (NSC) stored within them play key roles, for example, for resprouting after disturbance (Smith et al., 2018). Axial parenchyma (AP) is also associated with the maintenance of hydraulic conductivity, as angiosperm tree species with higher fractions of AP tend to have larger vessels (Morris et al., 2018), greater hydraulic capacitance (Ziemińska et al., 2020; Aritsara et al., 2021) and lower vulnerability to xylem embolism (Kiorapostolou et al., 2019). Radial parenchyma (RP) is not only a major NSC storage compartment but can also contain a large amount of lignin between individual parenchyma cells and surrounding fibres, thus potentially increasing overall wood density (WD) and improving radial strength (Burgert & Eckstein, 2001; Zhang et al., 2022b), but decreasing the available wood space for other cell types. The presence of large or abundant vessels can result in decreased WD, but a large lumen fraction may be important for increased hydraulic conductivity (Zanne et al., 2010), potentially creating a trade-off between hydraulics and xylem mechanical integrity (Christensen-Dalsgaard et al., 2007a,b; Chen et al., 2020), although studies provide conflicting results. For example, Woodrum et al. (2003) found no relationship between hydraulic conductivity and xylem modulus of elasticity (i.e. resistance to being deformed elastically) in several diffuse-porous Acer species. Differences in function among species may be due to the spatial organization of cell types, but the link between wood function and its spatial structure is still poorly understood. Although there have been several studies on the trade-offs in functions performed by different cell types (Supporting Information Table S1), few have done so across a broad range of tree species from different climates.
Testing hypotheses about the links between wood structures and functions can be facilitated by studying patterns within trees, and in particular, comparing stem and root wood. Regardless of the porosity type in angiosperm stems (i.e. ring- or diffuse-porous), wood becomes more diffuse-porous with distance from the stem in coarse roots (Patel, 1965). Roots usually have fewer vessels and smaller fibre fraction, but greater RAP fraction than stems (see Stokes & Guitard, 1997), thus increasing NSC storage capacity. This gradient in porosity in roots is advantageous for an improved hydraulic efficiency, especially in horizontal lateral roots that take up water from the most superficial soil layers (Fahn, 1967; Pratt et al., 2007). Also, when responding to mechanical stress, such as wind loading, root system anchorage is better improved through morphological changes, rather than modifications in xylem anatomy (Stokes & Guitard, 1997).
Due to the different plant strategies of temperate, Mediterranean and tropical climates as well as wet vs dry conditions, there is likely to be a divergence in wood structure. Tropical trees typically possess wider vessels and more parenchyma than temperate tree species (McCulloh et al., 2010; Morris et al., 2016, 2018), resulting in different wood densities. Despite the numerous studies focussing on variation and functional significance of xylem properties, it remains unclear how patterns of xylem traits and function are altered across different climates and whether functional trade-offs vary due to climate differences.
Allocation to different wood cell types and resulting functional strategies may be driven by coordinated evolutionary constraints. Exploring phylogenetically independent relationships among wood functional traits should provide us with a more genuine insight for correlated evolutionary shifts and their impact on trait variation (Felsenstein, 1985). For example, phylogeny explained more than half of the variation in individual functional traits in the stem xylem of 700 angiosperm tree species in China (Zheng et al., 2019). However, when examining the relationships among traits (e.g. WD and fibre fraction, hydraulic conductivity and embolism resistance, or NSC content and parenchyma fractions), studies have shown that these relationships are not necessarily affected by evolutionary history (Fortunel et al., 2014; Levionnois et al., 2021; Pratt et al., 2021b; Zhang et al., 2022b). Nonetheless, in order to identify trade-offs between anatomical structure and xylem function, phylogenetic analysis is still needed, especially across a broad range of species and climates.
We investigated the covariation in 12 wood traits in stems and coarse roots of 60 angiosperm tree species from temperate, Mediterranean and tropical climates. We asked whether the design of the xylem space captures trade-offs among the three main wood functions in stems and roots and whether this coordination is linked to climate or constrained by evolutionary histories. We hypothesize that (H1) coarse root wood will have more parenchyma and NSC, but lower WD than stem wood and that this pattern will be more pronounced in tropical species. (H2) Due to spatial constraints in a finite stem volume, we expect an allocation-driven trade-off among parenchyma, fibre and VFs. (H3) Due to the multiplicity of traits involved in a given function, functional trade-offs will be less clear than allocation-driven trade-offs between cell type fractions and will also differ between stems and coarse roots, as the relative importance change (e.g. the mechanical integrity of xylem tissue is less important in roots than stems). (H4) Due to the divergence of wood traits across climates, we expect that relationships among functions will also differ. For example, as the seasonal dormancy of temperate and Mediterranean species is more pronounced than that of trees in the tropics, we suppose that high levels of NSC are needed to assure the production of new and large earlywood vessels, especially in ring-porous species.
Materials and Methods
Study sites, species and sampling
This study was conducted on trees from forests growing in three different climates: a temperate forest (Luz-Saint-Sauveur, France; 42°52′25″N, 0°0′26″E), a Mediterranean forest (Montpellier, France; 43°38′46″N, 4°0′3″E) and a tropical forest (Paracou, French Guiana; 5°16′32″N, 52°55′28″W). The Luz-Saint-Sauveur site has a temperate mountain climate, with the heavy rainfall in first half of a year. The mean annual precipitation is 1217 mm, and the mean annual temperature is 11.8°C. The Montpellier site has a typical Mediterranean climate with hot dry summers and cool wet winters. The mean annual precipitation is 588 mm with c. 80% of the annual rainfall occurring between September and April. The mean annual temperature is 15.4°C. The Paracou site has a typical tropical rainforest climate with a distinct dry season from mid-August to mid-November and long rainy season often interrupted by a short drier period between March and April. The mean annual precipitation is 3035 mm with a minimum in September and maximum in May. The mean annual temperature is 26.9°C. The temperature and precipitation data were calculated using mean monthly data from the past 31 yr (Fig. S1: data are from Meteo-France; https://meteofrance.com/).
A total of 60 angiosperm species were selected, with 20 species in each climate, spanning 43 genera and 32 families (Table S2; Zhang et al., 2022a). The selected tree species are commonly found in these forests and represent their dominance to a certain extent. For each species, three healthy, adult, single-stemmed trees were identified, with a distance of at least 20 m between them. Selected stems were between 0.1 and 0.5 m wide at a height of 1.3 m (however, in some adult, small-stature species, diameter was as small as 0.05 m; Table S2). For Mediterranean and tropical trees, samples were collected at the end of August and September 2019, and for temperate trees, samples were collected in September 2020. Although phenological stages differ between tree species, we assumed that NSC would be greatest in temperate and Mediterranean species at the end of the summer season, but that little difference occurs in tropical species throughout the year (Martínez-Vilalta et al., 2016). We sampled all trees between 7:00 h and midday, to reduce variability in NSC content linked to photosynthate production. At a height of 1.3 m, three 0.05-m-long cores were extracted from tree stems with a 4.3-mm-diameter increment borer (n = 540). To collect samples of roots, we excavated a single lateral root (0.02–0.05 m in diameter) and at a distance of 0.3–0.5 m from the base of the tree, we extracted three increment cores or sampled three 0.02-m-long segments of root, depending on accessibility (n = 540). All samples were immediately placed in moist paper, put in a cooler box and taken to the laboratory before midday, where they were kept in a fridge at 4°C.
Measurement of wood traits
Xylem anatomical traits
Within 24 h, one of the three replicate increment cores was trimmed and a 0.02-m-long section was taken from the outer sapwood and placed in a 50% solution of alcohol and water. A 0.02-m-long section of root was also placed in the solution of alcohol and water. These sections were kept for the analysis of anatomical traits. Increment cores were embedded in paraffin individually after dehydration by immersing in a sequence of alcohol solution, whereas sections of root were clamped in a microtome. We used a sliding microtome to cut 15- to 20-μm-thick cross-sections. Cross-sections were stained with a mixture of safranin and Alcian blue (0.35 g safranin in 35 ml 50% alcohol with 0.65 g Alcian blue in 65 ml deionized water) and dehydrated using ethanol series (50%, 75%, 95% and 100%). Finally, sections were mounted on glass slides and observed under a light microscope (Olympus BX 60F; Olympus Co. Ltd, Tokyo, Japan). Minus a few exceptions, three microphotographs of transversal sections were taken for each stem and root sample with an APO (×5 or ×10) lens using a digital camera (Canon EOS 500D; Canon Inc., Tokyo, Japan) and some examples are shown in Fig. 1.
Radial parenchyma and AP were manually coloured using Photoshop (Adobe Systems Incorporated, San Jose, CA, USA; see examples shown in Fig. S2). The proportions of RP (radial parenchyma area divided by xylem cross-sectional area, %) and AP (axial parenchyma area divided by xylem cross-sectional area, %) were estimated using ImageJ 1.52 software (https://imagej.nih.gov/ij/). The proportion of total RAP (i.e. the sum of RP and AP, %), was then calculated. Using the same image, we also determined VF (i.e. the total vessel area divided by xylem cross-sectional area). Some tree species can have fibre-tracheids and tracheids (FT; Table S2), but distinguishing these cell types from fibres in a transversal section is methodologically challenging (Baas & Carlquist, 1985; Wheeler et al., 1989). In addition, FT are usually sparse with thick cell walls like fibres (Wheeler et al., 1989; Ziemińska et al., 2013). Therefore, we analysed fibres with FT together (FFT) in this study as also carried out by Kawai et al. (2021). Thus, the remaining tissue fraction was calculated as 1-RAP-VF. Vessel density (VD) is counted as the average number of conduits per mm2 of xylem cross-section.
For all anatomical traits, we used the mean value calculated from the three images for subsequent statistical analyses.
Hydraulic conductivity
The Kth calculated according to Hagen–Poiseuille's law is thought to be an overestimate relative to the actual hydraulic conductivity due to the uncertain variation of the pit membrane and vessel length inside the tree (Loepfe et al., 2007). However, there is empirical evidence showing that the relationship between these anatomical-based estimates and actual hydraulic measurements tends to be conserved across species (Sperry et al., 2005).
Wood density
The outer bark was removed on one stem and one root sample from each tree. Wood density was then measured by soaking samples in water for 24 h. The saturated volume was determined using the water displacement method (Pérez-Harguindeguy et al., 2016). After determining the saturated volume, samples were dried in a well-ventilated oven at 105°C for up to 72 h and then the dry mass (DM) was weighed. Wood density used here is defined as oven-dry mass divided by saturated volume (Williamson & Wiemann, 2010).
NSC content
Immediately on returning to the laboratory (on the same day the samples were collected), samples were heated in a microwave oven (90 s at 700 W) to stop enzymatic activity (Popp et al., 1996) and then oven-dried at 40°C until a constant weight was obtained. Dried samples were ground into powder using a ball mill (Retsch MM400; Retsch, Haan, Germany), and the ground samples were stored in sealed plastic tubes until analysis. Due to the large number of samples (n = 360), we used the near-infrared spectroscopy (NIRS; Batten et al., 1993) method to develop a model for predicting NSC contents in different organs (Zhang et al., 2022a). The NIRS method has been used successfully in plant ecology for predicting NSC content in different tissue types of a broad range of tree species (Ramirez et al., 2015; Rosado et al., 2019).
As this study is part of a larger study, where NIRS was used to develop a model for predicting NSC in organs of 90 tree species (Zhang et al., 2022a), we were able to create robust models using this larger dataset. First, 113 samples of stem and 140 samples of root were selected, for which values represented a broad range of data. Soluble sugars (SS) were extracted from 10 to 15 mg powder mixed in with 0.5 ml of 80% ethanol and incubated for 20 min at 80°C. The extraction was repeated twice and the three supernatants were collected in a tube and dried under vacuum (Refrigerated CentriVap Vacuum Concentrators, Labconco, Kansas City, MO, USA). The resulting soluble sugar extract was solubilized in 1.5 ml ultrapure water by sonication and agitation and then stored at −20°C. Total SS content was determined by a spectrophotometer at 620 nm (UV–visible DU 640 B; Beckman Coulter, Brea, CA, USA) using anthrone reagent and glucose as standard (Van Handel, 1965). Starch was extracted from the dried pellets in 1.5 ml of 0.2 M KOH solution incubated for 20 min at 80°C. Then, hydrolysis of starch into glucose was carried out in 0.1 M sodium acetate buffer solution at pH 4.75 with amyloglucosidase (EC 3.2.1.3. from Aspergillus niger Sigma A1602). The obtained glucose was determined colorimetrically using an enzymatic glucose oxidase (EC 1.1.3.4; Sigma G6125)/peroxidase (EC 1.11.1.7; Sigma P8125) and O-dianisidine dihydrochloride (Sigma D3252) reagent. After 10 min, 6N hydrochloric acid was added and absorbance was measured at 530 nm, using glucose as a standard (Chow & Landhäusser, 2004). SS and starch contents were expressed as mg g−1 DM and their sum is referred to as the total NSC content.
Then, near-infrared spectra were obtained for all ground samples previously stabilized in a conditioning room at a temperature of 20 ± 2°C and air humidity of 65 ± 5%. A Bruker Vector 22/N spectrometer (Bruker, Burladingen, Germany) was used in reflectance mode. Data were measured for wavelengths between 1000 and 2500 nm with a resolution of 3 nm. The spectra set was thus composed by 501 wavelengths of reflectance values. NIRS spectra were first transformed (Naes et al., 2004) with a standard normal variate (SNV) correction to reduce the effect of irregularities of surface and the intraspectrum variability (correction of the light dispersion). The second derivative was then computed using the algorithm of Savitzky and Golay (1964). The use of this derivative allowed for separating peaks that overlap and for correcting the baseline deviation of spectra and the range of NIRS values was examined.
Partial least squares regression (PLSR) was used to develop calibrations for the predictions of SS and starch content for stem and root samples separately (Zhang et al., 2022a). The calibration R2, standard error and root mean square error of cross-validation values are shown in Table S3.
Statistical analysis
In this study, the minimum, maximum, mean, standard deviation and quartile coefficient of dispersion (QCD) of each trait were calculated. The data were transformed using natural log for all traits to homogenized variance. We used linear mixed models to test the effects of organs (stem and root) and climates (temperate, Mediterranean and tropical) on wood traits. We included organs, climates and their interactions as fixed effects. Species were introduced in models as random effects to account for pseudo-replicates. Tukey honestly significant differences (HSD) post hoc tests were also performed to identify variation between organs and among climates.
We used the standardized major axis (SMA) regression method (Warton et al., 2006) to determine the scaling relationships among the wood traits across all species and among climates in (i) stems and (ii) roots. We also performed phylogenetic generalized least squares (PGLS) to investigate the influence of phylogenetic trends on wood trait associations with all species pooled together. We constructed a phylogenetic tree (Fig. S3) with all 60 species using the R package V.PhyloMaker (Jin & Qian, 2019), with the GBOTB phylogeny as the backbone (Smith and Brown, 2018). All species names were confirmed by the World Flora Online (http://www.worldfloraonline.org/). A Mantel test was performed to compare the pairwise wood trait correlation matrices between Pearson's test with and without phylogenetically independent contrasts (PICs). We also calculated Blomberg's K-value to evaluate the strength of phylogenetic signals for each trait (Blomberg et al., 2003). In order to infer the potential connections between all wood traits, we also conducted a simple structural equation model (SEM) based on our hypotheses. The overall model fit was assessed with several tests, including a nonsignificant chi-squared (χ2) test (P > 0.05), a comparative fit index (CFI > 0.95) and a standardized root mean square residual (SRMR < 0.08; Malaeb et al., 2000; Ali et al., 2016).
All statistical analyses were performed in R v.4.0.2 (R Development Core Team, 2020), using the packages vegan (Dixon, 2003), ape (Paradis et al., 2004), picante (Kembel et al., 2010), lavaan (Rosseel, 2012) and smatr (Warton et al., 2012).
Results
Patterns of wood traits in stem and root across 60 tree species
Across the 60 tree species, all wood traits showed marked interspecific variation (Figs S4, S5; Table 1). VD was the most variable trait in stems, ranging from 1.3 to 451.77 n mm−2 with a QCD of 0.77, and Kth was the most variable trait in roots, ranging from 0.45 to 459.18 kg m−1 MPa−1 s−1 with a QCD of 0.72. The proportion of FFT and WD showed relatively small interspecific variation. In stem, FFT ranged from 0.47 to 0.88 with a QCD of 0.07 and for roots, FFT ranged from 0.27 to 0.88 with a QCD of 0.10 (Table 1). In stems, WD ranged from 0.24 to 0.85 g cm−3 with a QCD of 0.14 and in roots, from 0.23 to 0.88 g cm−3 with a QCD of 0.16 (Table 1). Generally, organ type had a greater effect on traits than the effect of climate: WD, FFT, VF, VD and Kth were all significantly greater in stems than in roots, while AP, RP, RAP, SS, starch and NSC were all significantly lower in stems than in roots (Tables 1, 2).
| Trait | Abbrev. | Unit | Stem | Root | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Min | Max | Mean | SD | QCD | Min | Max | Mean | SD | QCD | |||
| Fibre with fibre-tracheids and tracheids fraction | FFT | – | 0.47 | 0.88 | 0.69 | 0.08 | 0.07 | 0.27 | 0.88 | 0.67 | 0.11 | 0.10 |
| Vessel fraction | VF | – | 0.02 | 0.39 | 0.13 | 0.07 | 0.41 | 0.02 | 0.33 | 0.11 | 0.07 | 0.42 |
| Vessel mean diameter | D m | μm | 16.43 | 216.36 | 75.43 | 41.72 | 0.37 | 20.00 | 218.32 | 73.74 | 34.02 | 0.32 |
| Vessel density | VD | n mm−2 | 1.30 | 451.77 | 57.96 | 71.01 | 0.77 | 2.30 | 270.51 | 33.71 | 33.10 | 0.61 |
| Radial parenchyma fraction | RP | – | 0.04 | 0.29 | 0.13 | 0.05 | 0.30 | 0.05 | 0.37 | 0.14 | 0.06 | 0.28 |
| Axial parenchyma fraction | AP | – | 0.01 | 0.21 | 0.05 | 0.04 | 0.41 | 0.01 | 0.46 | 0.08 | 0.08 | 0.55 |
| Total parenchyma fraction | RAP | – | 0.05 | 0.39 | 0.18 | 0.06 | 0.24 | 0.07 | 0.65 | 0.22 | 0.10 | 0.28 |
| Wood density | WD | g cm−3 | 0.24 | 0.85 | 0.56 | 0.12 | 0.14 | 0.23 | 0.88 | 0.51 | 0.13 | 0.16 |
| Specific xylem hydraulic conductivity | K th | kg m−1 MPa−1 s−1 | 0.81 | 460.05 | 48.47 | 73.53 | 0.69 | 0.45 | 459.18 | 37.11 | 52.77 | 0.72 |
| Soluble sugars content | SS | mg g−1 | 1.22 | 43.35 | 16.38 | 7.98 | 0.35 | 2.84 | 106.37 | 29.91 | 16.99 | 0.35 |
| Starch content | ST | mg g−1 | 0.38 | 110.93 | 32.37 | 20.25 | 0.42 | 7.15 | 360.26 | 105.96 | 67.60 | 0.48 |
| Total NSC content | NSC | mg g−1 | 7.58 | 120.91 | 49.06 | 22.61 | 0.32 | 21.34 | 376.39 | 135.12 | 74.25 | 0.40 |
- Abbrev., abbreviation; Min, minimum; Max, maximum; NSC, nonstructural carbohydrates; QCD, quartile coefficient of dispersion.
| Traits | Organ | Climate | Organ × Climate | ||||||
|---|---|---|---|---|---|---|---|---|---|
| df | F | P | df | F | P | df | F | P | |
| FFT | 1 | 12.52 | < 0.001*** | 2 | 0.51 | > 0.05 | 2 | 16.96 | < 0.001*** |
| VF | 1 | 15.75 | < 0.001*** | 2 | 9.70 | < 0.001*** | 2 | 7.99 | < 0.001*** |
| D m | 1 | 1.24 | > 0.05 | 2 | 25.40 | < 0.001*** | 2 | 27.79 | < 0.001*** |
| VD | 1 | 24.11 | < 0.001*** | 2 | 30.55 | < 0.001*** | 2 | 48.14 | < 0.001*** |
| RP | 1 | 9.19 | < 0.01** | 2 | 1.54 | > 0.05 | 2 | 0.05 | > 0.05 |
| AP | 1 | 49.77 | < 0.001*** | 2 | 5.96 | < 0.01** | 2 | 15.03 | < 0.001*** |
| RAP | 1 | 53.31 | < 0.001*** | 2 | 1.86 | > 0.05 | 2 | 9.25 | < 0.001*** |
| WD | 1 | 48.87 | < 0.001*** | 2 | 4.57 | < 0.05* | 2 | 5.35 | < 0.01** |
| K th | 1 | 11.59 | < 0.001*** | 2 | 4.80 | < 0.05* | 2 | 1.96 | > 0.05 |
| SS | 1 | 217.59 | < 0.001*** | 2 | 21.78 | < 0.001*** | 2 | 7.10 | < 0.001*** |
| ST | 1 | 320.96 | < 0.001*** | 2 | 2.28 | > 0.05 | 2 | 28.13 | < 0.001*** |
| NSC | 1 | 524.27 | < 0.001*** | 2 | 5.55 | < 0.01** | 2 | 23.08 | < 0.001*** |
- AP, axial parenchyma; df, degree of freedom; Dm, vessel mean diameter; F, Fisher value; FFT, fibre with fibre-tracheids and tracheids; Kth, theoretical specific xylem hydraulic conductivity; NSC, non-structural carbohydrates; RAP, radial and axial parenchyma; RP, radial parenchyma; SS, soluble sugars; ST, starch; VD, vessel density; VF, vessel fraction; WD, wood density. P indicates the significance level. *, P < 0.05; **, P < 0.01; ***, P < 0.001.
The stem fibre fraction was lowest in temperate species and highest in tropical species, while no significant difference was found among climates for FFT in roots (Fig. 2a). The VF in both stems and roots was greatest in temperate species and lowest in tropical species (Fig. 2b). RAP and AP in both stems and roots were lowest in temperate species, whereas RP was the greatest in the stem of Mediterranean species, but did not vary in roots among the three climates (Figs 2c, S6a,b). Temperate species had lower WD in both stems and roots compared with Mediterranean and tropical species (Fig. 2d). The Kth of both stems and roots was lower in Mediterranean species than in temperate and tropical species (Fig. 2e). Mediterranean species had more stem SS but less stem starch compared with temperate and tropical species, while stem NSC were not significantly different. However, tropical species possessed lower SS, starch and NSC contents in roots compared with temperate and Mediterranean species (Figs 2f, S6e,f).
Independent trade-offs among wood cell types and functions in stem and root
Clear and strong trade-offs among cellular fractions were found in both stems and roots. VFs and RAP were significantly and negatively related to FFT, and RAP was also negatively correlated with VF in both stems and roots across all species (Figs 3a–c, 4a–c). However, the trade-offs among functions were complex. Kth was negatively associated with WD, while no overall significant relationships between NSC and WD were found in either stem or root (Figs 3d,e, 4d,e). Interestingly, NSC was significantly and positively associated with Kth in stem, whereas in roots, a negative relationship was found between NSC and Kth (Figs 3f, 4f). Wood density was not significantly associated with FFT and Kth was positively associated with VF in both stems and roots (Figs 3g,h, 4g,h). There was a positive relationship between NSC and RAP across the stems of all species, but not in roots (Figs 3i, 4i).
Relationships between wood traits varied between climates. Basically, trade-offs among cellular fractions of stem and root wood were generally observed within climates (Figs 3a–c, 4a–c; Tables S4, S5). However, the significant relationships disappeared in some climates when referring to the traits that acted as proxies for functions. For example, there were no significant relationships between stem WD and Kth within a given climate and no significant relationship was found between root WD and Kth in tropical species (Figs 3d, 4d; Tables S4, S5). Positive relationships between WD and NSC for both stems and roots were only found in Mediterranean species (Figs 3e, 4e; Tables S4, S5). Similarly, stem Kth was only associated with stem NSC in temperate species and no significant correlation was found between root Kth and NSC in any climate (Figs 3f, 4f; Tables S4, S5).
The phylogenetic analysis showed similar results to the cross-species analysis regarding trait relationships (Tables 3, S6). However, many traits had significant phylogenetic signals, including WD, RP, RAP, FFT, VF, Dm and VD in stems and AP, RP, RAP, VD, SS, starch and NSC in roots (Table S7).
| Models | Stem | Root | ||
|---|---|---|---|---|
| R 2 | P | R 2 | P | |
| VF ~ FFT | 0.32 | < 0.001*** | 0.09 | < 0.05* |
| RAP ~ FFT | 0.29 | < 0.001*** | 0.62 | < 0.001*** |
| RAP ~ VF | 0.12 | < 0.01** | 0.06 | > 0.05 |
| Kth ~ WD | 0.07 | < 0.05* | 0.14 | < 0.001*** |
| NSC ~ WD | 0.01 | > 0.05 | 0.01 | > 0.05 |
| NSC ~ Kth | 0.04 | > 0.05 | 0.07 | < 0.05* |
| WD ~ FFT | 0.01 | > 0.05 | 0.01 | > 0.05 |
| Kth ~ VF | 0.09 | < 0.05* | 0.54 | < 0.001*** |
| NSC ~ RAP | 0.25 | < 0.001*** | 0.08 | < 0.05* |
- FFT, fibre with fibre-tracheids and tracheids; Kth, theoretical specific xylem hydraulic conductivity; NSC, non-structural carbohydrates; RAP, radial and axial parenchyma; VF, vessel fraction; WD, wood density.
- P indicates the significance level. *, P < 0.05; **, P < 0.01; ***, P < 0.001.
The result of the SEM confirmed that there were allocation trade-offs among cell fractions in both stems and roots and, as in the previous analyses, the relationships between Kth and NSC were contrasting (Fig. 5). The framework constructed by the SEM showed that WD was not associated with NSC in either stems or roots and negative relationships between WD and Kth were found in both stems and roots (Fig. 5). Additionally, RAP and FFT were strongly and positively associated with WD in both stems and roots (Fig. 5). There were significant relationships among cell types in stem and root wood for species in most climates, but the trade-off among functions was complex and differed across climates (Fig. S7).
Discussion
Variations in wood traits in stems and roots across climates
Corroborating our H1, we showed that the total parenchyma and AP fractions were more abundant in roots than stems when all species pooled together, providing roots with a greater capacity to store NSC. In agreement with Morris et al. (2016), we also found that the total parenchyma fraction was greater in tropical than in temperate species, which was mainly due to the increase in AP rather than RP. Tropical species had fewer but larger vessels in both stems and roots, with greater hydraulic conductivity. By contrast, Mediterranean species had the lowest hydraulic conductivity in both stem and root wood, due mainly to narrower vessels, even though vessel density was greater than in trees from other climates. Species from water-limited environments, such as the Mediterranean area, are characterized by high embolism resistance and low hydraulic transport efficiency (Maherali et al., 2004; Liu et al., 2019). The climate in tropical French Guiana is relatively moist and warm; therefore, the risk of drought-induced embolism is negligible (Ziegler et al., 2019), but there may be exceptions in the case of an extreme drought (Rowland et al., 2015; Fontes et al., 2020), and as vessel size is not constrained, it can be increased to enhance water uptake. Similarly, freeze–thaw-induced xylem embolism is less likely to occur in warmer tropical forests than in temperate tree species or some Mediterranean tree species that are more susceptible to freeze–thaw stress (Sperry & Sullivan, 1992; Hacke et al., 2017). Therefore, wider vessels associated with higher hydraulic conductivity are again more likely to occur in tropical systems.
Wood density in both stems and roots of temperate species was lower than that in species from other climates. However, the variability in data from tropical species was high and the range of density values was also large, as also found by Wiemann and Williamson (2002). Root NSC content was always higher than in the stem, as also found by Martínez-Vilalta et al. (2016), Smith et al. (2018) and Furze et al. (2020). A high NSC content in roots allows for fast root growth in the spring, improving the supply of water and nutrients necessary for budburst or leafing out (Abramoff & Finzi, 2015). Mediterranean species had more NSC in roots, which could aid resprouting and the recovery of hydraulic conductivity if embolism occurs (Yoshimura et al., 2016; Kono et al., 2019), especially during long, dry periods. Tropical species had the least NSC in roots, possibly because they allocate more carbon for growth and respiration throughout the longer growing season, resulting in a lower overall NSC (Körner, 2003; Martínez-Vilalta et al., 2016). In agreement with Martínez-Vilalta et al. (2016), we did not find a significant difference among the three climates with regard to total NSC in stems.
Trade-offs among cell types do not reflect functional trade-offs in stems and roots
Clear allocation trade-offs among cell types were found in both stems and roots across all species, reflecting spatial constraints (Pratt & Jacobsen, 2017) and corroborating our H2. Fibres occupied the most space, at the expense of vessels or parenchyma. However, functions did not exhibit the same patterns and were complex and diverse, which is in agreement with our H3. We found negative relationships between WD and hydraulic conductivity in both stems and roots across all species pooled together, implying that there is a trade-off between mechanical integrity and hydraulic conductivity, as also found by several authors (Smith & Ennos, 2003; Christensen-Dalsgaard et al., 2007a,b; McCulloh et al., 2011; Janssen et al., 2020). Wood density has been used as a proxy for the hydraulic properties of stems (Markesteijn et al., 2011; Hoeber et al., 2014), as hydraulic conductivity is proportional to vessel diameter to the fourth power, suggesting that species with wide vessels have high water transport efficiency and reduced WD due to a lower fibre fraction. Similarly, dense wood has greater cell packing, with narrow vessel lumina (Zanne et al., 2010; Markesteijn et al., 2011). It is assumed that dense tissues reduce hydraulic efficiency because of narrower and potentially shorter vessels, as well as thicker walls that will increase the path length through the pits where sap flows between narrow vessels (Lens et al., 2011; Pratt et al., 2021a). However, this negative relationship between WD and hydraulic conductivity does not seem to be universally valid as other studies failed to find it (Pratt et al., 2007; Martínez-Cabrera et al., 2009; Poorter et al., 2010; Schuldt et al., 2013; Fortunel et al., 2014). A possible explanation may be that other properties could drive WD, such as fibre wall thickness, but do not directly affect hydraulic conductivity (Ziemińska et al., 2013; Hoeber et al., 2014; Badel et al., 2015). Previous studies have also shown that although WD was not related to total fibre fraction, it was positively correlated with fibre wall fraction and negatively with fibre lumen fraction (Jacobsen et al., 2005; Ziemińska et al., 2013; Janssen et al., 2020), properties that were not measured in our study.
We found that WD was positively correlated with total parenchyma fraction, suggesting that WD could be a good predictor of NSC storage capacity. However, though RAP fraction was positively correlated with NSC content, we found no correlation between WD and NSC in either stem or root. The positive relationship between WD and parenchyma content is well documented, as wood with a high proportion of AP usually has fibres with narrow lumina (Woodrum et al., 2003; Martínez-Cabrera et al., 2009; Rana et al., 2009; Zhang et al., 2022b). Additionally, RP can be strongly lignified and high lignin content increases WD (Harada & Wardrop, 1960; Zheng et al., 2016). Also, we expressed NSC content per unit of mass, not sapwood volume. Comparisons of NSC concentrations should also account for changes in the density of structural compounds, which can obscure differences in effective compound concentration per cell unit (Hoch et al., 2002). Overall, our results indicate that WD is a poor predictor of the NSC content in trees.
In agreement with our H3, we found that NSC content was positively associated with hydraulic conductivity in stems, but negatively in roots. Stem wood needs high levels of NSC to allow for the production of new and large earlywood vessels, especially in ring-porous species. The most common species that have ring-porous stems rarely refill winter-embolized vessels and mostly rely on new sapwood growth by cambial activity in the earlywood to carry water during the growing season (Hacke & Sperry, 2001; Barbaroux & Bréda, 2002), which requires the use of local or imported NSCs and represents a medium to long-term strategy for hydraulic recovery. Lateral roots are usually diffuse-porous with more numerous vessels throughout the xylem. Previous studies demonstrated that roots are more vulnerable to xylem cavitation than stems (Sperry & Ikeda, 1997; Maherali et al., 2006; Hacke & Jansen, 2009; Domec et al., 2010) and the embolism repair or vessel refilling may be dependent on the role of SS and parenchyma cells surrounding vessels (Secchi et al., 2017; Tomasella et al., 2020). However, recent studies have shown that coarse roots are not highly vulnerable to xylem embolism as commonly believed (Peters et al., 2020; Wu et al., 2020; Lübbe et al., 2022). In addition, even if root embolism occurs (especially for freeze–thaw-induced xylem embolism), the repair of embolism may probably depend on root pressure (the positive pressure resulting from osmolyte accumulation in root tissues; Wegner, 2014) rather than osmotic regulation achieved by SS (Wang et al., 2022). Thus, the recovery of hydraulic conductivity is less dependent on stored carbon. This opposing strategy may also be related to the different functions performed by RAP in roots and stems (Zheng & Martínez-Cabrera, 2013; Secchi et al., 2017; Aritsara et al., 2021). In agreement with Morris et al. (2018), a large AP fraction was associated with wider vessels, consistent with a potential role in water transport. By contrast, RP was negatively correlated with hydraulic conductivity, demonstrating the functional differentiation of AP and RP in hydraulic efficiency (Zheng & Martínez-Cabrera, 2013). Also, AP may perform different functions in roots compared to stems, where it serves a role for hydraulic functioning, which could lead to different strategies between stems and roots.
The impact of phylogeny and climate on structure–function relationships
Certain wood anatomical traits exhibited strong phylogenetic signals as also observed by Zheng et al. (2019), which supports the notion that wood anatomy is relatively conservative and largely determined by evolutionary history. However, our phylogenetic analysis showed the correlations between PGLS trait values were similar as those with raw trait values, suggesting that evolutionary histories cannot explain the associations among the traits we sampled, as also found by Pratt et al. (2021a). In addition, we found that the relationships among cell types exhibited strong trade-offs across species within each climate, although some relationships were only weakly correlated. This result means that the covariation among cell fractions is conserved and does not experience large fluctuations within a given climatic zone (Morris et al., 2016; Pratt & Jacobsen, 2017). On the contrary, relationships among functions across species are complex and tend to vary among climates, which corroborates our H4. The reason may be that in different climates, WD can be influenced by other properties, such as fibre wall thickness, but hydraulic conductivity might not be directly affected by these properties (Ziemińska et al., 2013; Hoeber et al., 2014; Badel et al., 2015). Additionally, NSC can show seasonal variation (Smith et al., 2018; Furze et al., 2020) and could be involved in the reversal of embolism, especially for trees growing in temperate regions that are susceptible to freeze–thaw-induced xylem embolism (Nardini et al., 2011; Sala et al., 2012).
Caveats
Although cell types and functions in xylem are entwined and interconnected in a complex way, different cell types still have different functions. Tracheids, fibre-tracheids and fibres typically play different roles in the xylem, but are difficult to distinguish in transversal sections (Baas & Carlquist, 1985; Wheeler et al., 1989). Analysing these three cells together in our study may ignore the role played by each cell type, thereby affecting the relationship between allocation and functional trade-offs. Therefore, more effort is required in future studies, especially with regard to distinguishing cell types, to improve our understanding of the relationships between cell type fractions and wood functions in trees. Living wood fibres that retain live protoplast for months or years are often overlooked cell types in structure and function studies, even though they are analogous in their function to AP strands (Fahn & Leshem, 1963; Yamada et al., 2011). Living fibres are much more common than supposed, but they remain poorly understood because they are difficult to detect in fresh wood samples (Gregory, 1978). Further experimental work with improved techniques and methods on how or to what extent living fibres play a major role in xylem function will be important towards reaching this goal.
Conclusion
We evaluated the coordination in wood structure and functions across 60 angiosperm tree species in three distinct climates. By examining the changes in wood design in stems and roots of the same individuals, we found that there were clear and strong trade-offs in carbon allocation among cell type fractions. However, the covariations among functions were more complex. We thus underline that trade-offs in carbon allocation do not necessarily reflect functional trade-offs in stems and roots. We further point out that although WD was negatively correlated with Kth, suggesting a trade-off between mechanical integrity and hydraulic conductivity, no relationship was observed between WD and NSC. Additionally, NSC storage capacity in stems was positively associated with hydraulic conductivity, but negatively in roots, demonstrating that there is also a difference in this functional relationship between stems and roots. Our study provides new insight into the structure and function of xylem within trees from diverse locations and helps us to better understand the behaviour of trees under global climate change and how they can cope with increasing drought events in the future.
Acknowledgements
This study was supported by a grant from Investissement d'Avenir grants of the ANR, France (CEBA: ANR-10-LABX-25-01) and China Scholarship Council (CSC, no. 201806300029, for GZ). We are grateful to J. Cazal (INRAE), S. Nourissier-Mountou (CIRAD), N. Boutahar (CIRAD), M. Ramel (INRAE) and S. Fourtier (INRAE), for their help with field and laboratory work. We thank G. Derroire, A. Dourdain, C. Stahl and S. Coste for their aid with logistical arrangements at Paracou field station and at ECOFOG laboratory in French Guiana. The collecting of genetic resources in French Guiana was authorized by the French Government (APA no. 1371201). Thanks are due to managers of state-owned land for their collaboration and to Mr. and Mrs. R. Lecoustre and B. Dupin (Société Eco-Altitude) for permission to sample trees on private property. The authors would like to thank Prof. L. Poorter (Wageningen University and Research) for his help in the early discussions in the paper. The authors would like to thank SILVATECH (doi: 10.15454/1.5572400113627854E12) from UMR 1434 SILVA, 1136 IAM, 1138 BEF and 4370 EA LERMAB from the research centre INRAE Grand-Est Nancy for its contribution to NSC analyses. The SILVATECH facility is supported by the French National Research Agency through the Laboratory of Excellence ARBRE (ANR-11-LABX-0002-01).
Competing interests
None declared.
Author contributions
GZ and AS planned and designed the research. GZ, ZM, PM, JE, CF, PH, J-LM, JM-V and AS conducted fieldwork. GZ, LB, BG, AS and CF performed laboratory analyses. GZ, ZM and AS analysed the data. All the authors contributed critically to the drafts and gave final approval for publication.
Open Research
Data availability
The data described in this paper are freely and openly accessed at Data INRAE at https://doi.org/10.57745/X6WW8A.
