Functional trade-offs in volume allocation to xylem cell types in 75 species from the Brazilian savanna Cerrado

Abstract

Background and Aims

Xylem is a crucial tissue for plant survival, performing the functions of water transport, mechanical support and storage. Functional trade-offs are a result of the different assemblages of xylem cell types within a certain wood volume. We assessed how the volume allocated to different xylem cell types can be associated with wood functional trade-offs (hydraulics, mechanical and storage) in species from the Cerrado, the Brazilian savanna. We also assessed the xylem anatomical characters linked to wood density across species.

Methods

We analysed cross-sections of branches collected from 75 woody species belonging to 42 angiosperm families from the Cerrado. We estimated the wood volume fraction allocated to different cell types and performed measurements of vessel diameter and wood density.

Key Results

The largest volume of wood is allocated to fibres (0.47), followed by parenchyma (0.33) and vessels (0.20). Wood density is positively correlated to cell wall (fibre and vessel wall), and negatively to the fractions of fibre lumen and gelatinous fibres. We observed a trade-off between hydraulics (vessel diameter) and mechanics (cell wall fraction), and between mechanics and storage (parenchyma fraction). The expected positive functional relationships between hydraulics (vessel diameter) and water and carbohydrate storage (parenchyma and fibre lumen fractions) were not detected, though larger vessels are linked to a larger wood volume allocated to gelatinous fibres.

Conclusions

Woody species from the Cerrado show evidence of functional trade-offs between water transport, mechanical support and storage. Gelatinous fibres might be potentially linked to water storage and release by their positive relationship to increased vessel diameter, thus replacing the functional role of parenchyma and fibre lumen cells. Species can profit from the increased mechanical strength under tension provided by the presence of gelatinous fibres, avoiding expensive investments in high wood density.

INTRODUCTION

A widely accepted theory in land plant ecology states that patterns of species performance within a given community are based upon a fast–slow spectrum that integrates traits of the plant body as a whole, including leaves, stems and roots (Reich, 2014). This spectrum is operated by trade-offs involving resource acquisition, i.e. cheap tissue investment with a faster return, and resource conservation, i.e. expensive investment with slower return, which has been well demonstrated in leaves (Wright et al., 2004). A large proportion of those trade-offs, however, are related to the properties of wood, which sustains multiple biological functions and can have considerable ecological implications since hydraulic functions and carbon balance supply are linked to plant distribution and survival across species and biomes (Chave et al., 2009; Adams et al., 2017).

Wood, or secondary xylem, in angiosperms is a tissue composed of different cell types that performs a number of biological roles, namely water transport, executed primarily by vessels; storage of water carried out primarily by cell lumens (vessels and fibres); storage of nutrients (carbohydrates) carried out mainly by axial and ray parenchyma cells; and also mechanical functions carried out by secondary cell walls (fibre and vessel walls). These biological functions, however, are strongly connected and can partially overlap. For example, parenchyma cells that mainly carry out a storage function might have an influence on hydraulics via increasing hydraulic capacitance (Scholz et al., 2011; Pfautsch et al., 2015), facilitating embolism reversal by releasing water into the transpiration stream during drought (Brodersen et al., 2010; Trifilò et al., 2019) and by having an influence on embolism resistance since vulnerable species have larger amounts of parenchyma (Kiorapostolou et al., 2019; Pratt et al., 2021a, b). The volume and amount of cell (fibre) lumen spaces as well as living cells (parenchyma cells) are assumed to affect the extent and dynamics of water release into the transpiration stream, affecting hydraulic dysfunction (Meinzer et al., 2009; Jupa et al., 2016). There is also evidence that parenchyma plays a role in mechanical functions due to its link with the modulus of elasticity (stiffness), with radial tensile strength (Reiterer et al., 2002; Woodrum et al., 2003) and with wood density (Zheng and Martinez-Cabrera, 2013; Kiorapostolou et al., 2019; Dória et al., 2019a). A relationship between mechanical and hydraulic functions has also been reported by the indicative correlation between embolism resistance and lignified cell wall contents, such as fibre wall area and thickness (Lens et al., 2016; Dória et al., 2018, 2019b), vessel wall reinforcement (Hacke et al., 2001; Jansen et al., 2009) and lignin content (Pereira et al., 2017). The link between mechanical and hydraulic functions has also been hypothesized from the evidence of the fibre bridges detected by the connection between pitted fibre and tracheids, and isolated vessels, which theoretically provides a safer accessory pathway for water transport (Cai et al., 2014).

In this context, the spatial organization and assemblage of the different xylem cell types in a certain wood volume have a direct impact on wood functionality due to the link between structure and function (Bittencourt et al., 2016; Gleason et al., 2016; Pratt et al., 2021a), and thereby result in conflicting functional demands (Baas et al., 2004; Pratt and Jacobsen, 2017; Janssen et al., 2020). For instance, wood density and hydraulic properties, both traits related to plant ecological fitness, are directly determined by the xylem anatomical structure (Jacobsen et al., 2007; Zieminska et al., 2013, 2015), meaning that the different proportion of wood cell fractions in a wood will impact biological functions at different scales. Therefore, a link between both xylem anatomical traits and properties with plant life history performance should appear (Pratt et al., 2007, 2012; Janssen et al., 2020; Percolla et al., 2021).

Wood density has emerged as an integrative trait of diverse wood properties. It is the most commonly measured functional trait, and is a good indicator of where species lie along the fast–slow trade-off continuum: usually, fast-growing species are characterized by low wood density, and slow-growing species have high wood density (ter Steege and Hammond, 2001; Muller-Landau, 2004; Fajardo, 2022), though this pattern might not be sustained within species (Deslauriers et al., 2009; Rossi et al., 2015), showing that wood density might not be a good predictor of competitive ability among individuals of the same species (Fajardo, 2016). Expensive xylem construction costs, theoretically linked to higher wood densities, result in narrower vessels with thicker walls, providing safer conduction against hydraulic failure, but lower hydraulic efficiency (Tyree and Zimmerman, 2002). On the other hand, higher efficiency in conduction is linked to larger vessels, faster plant growth and lower xylem construction costs, i.e. less xylem tissue for a given amount of leaf area (Poorter et al., 2010; Gleason et al., 2012), and consequently have lower wood density. Nevertheless, the functional meaning of wood density is not entirely clear because it is a very diverse trait describing multiple functions (Fichtler and Worbes, 2012). It is positively related to wood mechanical properties (mechanical strength and stiffness) (Onoda et al., 2010) and to plant drought mortality threshold (Liang et al., 2021), and negatively to water storage and capacitance (Bucci et al., 2004; Scholz et al., 2007). Based on the xylem anatomical traits, the variation in wood density across species is largely explained by the variation in fibre wall and lumen fractions (Zieminska et al., 2013, 2015). However, none of the previous studies assessed the particular role of gelatinous fibres (G-fibres) in wood density variation. G-fibres are a special type of fibre with an additional inner cellulosic layer known to increase mechanical resistance against tension (Clair et al., 2008). Also, this additional layer, called the G-layer, has a high percentage of hydrophilic content compared with the other cell wall layers (Bowling and Vaughn, 2009). Other studies point to a role for rays and axial parenchyma in explaining such variation (Zheng and Martinez-Cabrera, 2013; Dória et al., 2019a). Additionally, wood density has long been reported as positively linked to plant embolism resistance by the avoidance of conduit implosion under high negative pressures (Hacke et al., 2001). Nevertheless, an indirect nature of such a relationship has also been acknowledged via the direct correlation between embolism resistance and the thickness of intervessel pit membranes (Li et al., 2016; Dória et al., 2018; Levionnois et al., 2021).

In this study, we collected wood traits in species from the Brazilian Cerrado, the savanna-like ecosystem that dominates the Brazilian central plateau, characterized by a strongly seasonal climate with distinctive and regular wet and dry periods (Oliveira and Marquis, 2002). Our dataset, representing 75 species from 42 angiosperm families, aims to understand how the wood volume is allocated to the different xylem cell types and explain the functional space partition of the xylem components and the resulting wood density of Cerrado species. In particular, we assess the occurrence of potential trade-offs between xylem functional components, including hydraulic, storage and mechanical functions, by testing the following: (1) a trade-off exists between hydraulics and mechanical function (i.e. vessel diameter vs. total cell wall fraction); (2) a trade-off exists between storage of water and carbohydrates and mechanical functions (i.e. ray and axial parenchyma fraction vs. total cell wall fraction); and (3) a positive relationship exists between hydraulic conductivity and carbohydrate storage (i.e vessel diameter vs. parenchyma), and between hydraulic conductivity and water storage (i.e. vessel diameter vs. fibre lumen fraction and/or vessel diameter vs. fibre lumen fraction + axial parenchyma fraction). Also, we expect to find higher wood density in species with a higher wood volume allocated to the cell wall (fibre wall and vessel wall), including higher proportions of G-fibres.

MATERIALS AND METHODS

Plant material and study area

We collected data on 153 individuals, 1–5 individuals sampled per species, of 75 woody species of angiosperms belonging to 42 families (Supplementary data Table S1). We selected the most representative woody species of the Cerrado based on species lists from the literature (Ratter et al., 1997; Durigan et al., 2002) and from the Herbarium Irina Delanova Gemtchujnicov (BOTU) from the São Paulo State University. Cerrado trees are generally short with few and contorted branches, and therefore the samples were collected from the most developed and mature branch of mature individuals in an area of 180 ha located in the mid-western region of São Paulo State, Brazil (22°82ʹ01.8ʹʹS and 48°74ʹ02.6ʹʹW). Annual temperature in the region is 20 °C, with an annual rainfall of 1306 mm. The dry season with monthly precipitations of < 60 mm lasts 4 months (Sonsin et al., 2012).

Wood anatomy

We prepared permanent anatomical slides for light microscopy from the outer sapwood of the branch samples. The samples were cut in transverse, tangential and radial sections of 20 μm in thickness using a sliding microtome. After bleaching with 1 % sodium hypochlorite and rinsing with water, the sections were stained with a mixture of 1 % safranin and 1 % astra blue in water (1:9), dehydrated in an ethanol series (30, 50, 70 and 96 %), treated with a Parasolve clearing agent (butyl acetate) and mounted in Entellan^® resin (Sonsin et al., 2012).

Wood volume fraction, wood density and anatomical measurements

We calculated the wood volume occupied by different cell types by assessing the presence of vessel wall, vessel lumen, fibre wall, fibre lumen, G-fibre, parenchyma-like fibre bands, rays and axial parenchyma. The differences in cell types were recognized in cross-sections, except for the parenchyma-like fibre bands, which include septate and living fibres, that were recognized in tangential sections (Carlquist, 2014; Silva et al., 2015). We identified G-fibre in cross-section by the presence of the inner G-layer, which is dyed blue when using the double staining with safranin and astra blue. All measurements were performed on cross-sections. We took microscopic pictures of the cross-sections using a camera attached to a microscope at ×10 magnification (picture size = 1388 × 1040 pixels). For each individual, we took two pictures, from two different regions along the radial direction from bark to pith, and averaged the data of the two regions. For estimating wood volume fractions occupied by different cell types, we followed the grid method proposed by Smith (1967) and modified by Zieminska et al. (2015). A grid with a total of 269 cross-points, 50 μm apart in a horizontal and vertical direction, was overlaid on the wood anatomical image, and the different cell types were coded with different colours using an image analysis software. Afterwards, the images generated by the coloured points were transferred to ImageJ^® to estimate the fractions of cell types based on the proportion of each colour and on the total number of points. Vessel diameters (μm) were measured with ImageJ from pictures taken from the same slides where the wood volume fraction of cell types was calculated.

Wood density was determined for each individual in wood samples of 1 cm³ in volume, and defined as the ratio between oven-dried mass (at 100 °C until constant weight of the sample) and fresh volume (by the weight of water displacement method) (USDA Forest Service, 1956).

Statistical analyses

Linear mixed effect models (LMMs) (Gelman and Hill, 2006; Pinheiro et al., 2018) were used to test for the relationship between the different tissue fractions to infer functional trade-offs between xylem cell types, and multiple mixed effect linear regression was used for the relationship between tissue fractions and wood density. All LMMs had species as the random variable. For data selection, we used variance inflation factor (VIF) analysis, keeping variables with a VIF value <2 (Zuur et al., 2010). For the LMMs with two or more predictive variables, we calculated the hierarchical partitioning (Chevan and Sutherland, 1991) for the variables retained in the model to assess their relative importance. A log transformation, when necessary, was applied to meet homoscedasticity or normality (Zuur et al., 2007). A principal component analysis (PCA) was conducted to observe the global relationships among the studied traits. A Pearson correlation was performed amongst all variables assessed in the study to observe general correlations. All analyses were performed in R 3.6.3 (R Core Team, 2020) using the packages lme4 and Vegan.

RESULTS

Wood volume allocation to xylem cell types

In our dataset of 75 woody species, the largest fraction of wood volume (0.47) is allocated to fibres (Fig. 1A). It varies 6.5-fold, ranging from 0.12 to 0.78 within individual species, showing a coefficient of variability (CV) of 0.29 (Fig. 1B). The higher fractions of wood volume allocated to fibres (>0.55) are observed in Melastomataceae (genera Pleroma and Miconia), while the higher values for G-fibres are observed in the genus Tapirira (Anacardiaceae) (Supplementary data Table S1).

Fig. 1.

Average volume fractions and variation of wood cell types across 75 species from the Cerrado. Average (asterisk), maximum and minimum values (dots), and CV (coefficients of variability = standard deviation/average) are given.

Parenchyma is the second most abundant tissue, and corresponds to 0.33 of the wood volume (Fig. 1A), showing the largest variability among cell types (CV of 0.43), corresponding to 10-fold, from 0.07 to 0.70 (Fig. 1B). Out of the total parenchyma fraction, 0.21 is due to the ray fraction (from 0.07 to 0.51; a variation of 4-fold), and 0.12 is due to the axial parenchyma (from absent to 0.42; a variation of 42-fold), with a CV of 0.40 and 0.43, respectively (Figs. 1A, B). The higher ray fractions are detected in the genera Ouratea (Ochnaceae), Myrsine (Primulaceae) and Byrsonima (Malpighiaceae), and the lower fractions in Persea (Lauraceae) and Aegiphila (Lamiaceae) (Supplementary data Table S1). For axial parenchyma, the higher fractions are observed in Fabaceae, more specifically in the genera Dimorphandra, Dalbergia and Leptolobium, as well as in Qualea (Vochysiaceae). The lower fractions occur in genera that lack axial parenchyma, such as Connarus (Connaraceae) and Tabernaemontana catharinensis (Apocynaceae) (Supplementary data Table S1).

The vessel fraction corresponds to 0.20 in wood volume (Fig. 1A) and varies 7-fold, from 0.06 to 0.43, corresponding to a CV of 0.37 (Fig. 1B). The higher vessel fractions are observed in Apocynaceae, in the genera Aspidosperma and Tabernaemontana, and in the genus Erythroxylum (Erythroxylaceae), while the lower fractions are observed in Fabaceae (genera Bowdichia and Dalbergia), and in Vochysiaceae (genera Qualea and Vochysia) (Supplementary data Table S1).

Drivers of wood density variation across Cerrado species

On average, wood density is 0.51 g cm^–3, varying >3-fold from 0.24 to 0.77 g cm^–3. Higher wood density occurs in Miconia albicans (0.737 g cm^–3) and lower density in Annona coriacea (0.24 g cm^–3) (Supplementary data Table S1). Wood density is positively correlated to the variation in fibre wall fraction (χ² = 48.42, P < 0.001) and negatively to the variation in fibre lumen fraction (χ²= 40.28, P < 0.001) (Fig. 2). Both variables expain similar percentages: the fibre wall fraction explains 53 % and fibre lumen fraction explains 47 % of the variation in wood density. Accordingly, species with high wood densities, such as Miconia albicans, Erythroxylum pelleterianum and Eugenia rigida, allocate on average 0.49 of the wood volume to the fibre wall and <0.11 to the fibre lumen. Species with low wood densities, such as Annona coriacea and Cecropia pachystachya, allocate on average 0.22 of the wood volume to the fibre wall and 0.04–0.16 to the fibre lumen (Supplementary data Table S1).

Fig. 2.

Wood density variation and its relationship with wood cell types across 75 species from the Cerrado. Each dot represents one individual. The strength of the relationship (R²) is given for each of the regression lines.

Trade-offs in wood volume allocated to the different cell types

We detect a trade-off between vessel diameter and wood volume allocated to the cell wall (fibre wall and vessel wall) (χ² = 5.15, P < 0.05) (Fig. 3A). A strong trade-off emerges between the wood volume allocated to the cell wall and parenchyma storage cells. This relationship is driven by both types of parenchyma cell fractions, more specifically the ray fraction (χ² = 11.24, P < 0.001) (Fig. 3B), and axial storage cells (the sum of axial parenchyma and parenchyma-like fibre bands) (χ² = 20.97, P < 0.001) (Fig. 3C). No relationship is detected either between vessel diameter and wood volume allocated to axial storage cells (χ² = 2.05, P > 0.05) (Fig. 3D) or between vessel diameter and wood volume allocated to the fibre lumen fraction (χ² = 1.2476, P > 0.05) (Fig. 3E); in the same way, no relationship is found between vessel diameter and wood volume allocated to the sum of fibre lumen fraction + axial storage cells fraction (χ² = 3.0518, P > 0.05) (Fig. 3F).

Fig. 3.

Trade-offs in wood volume allocation to wood cell types and vessel properties across 75 species from the Cerrado. Each dot represents one individual. The strength of the relationship (R²) is given for the linear regression lines. Fraction of axial storage cells = the sum of axial parenchyma and parenchyma-like fibre band fractions.

A Pearson correlation matrix showing the relationships among the variables measured is presented in Supplementary data Fig. S1.

Relationships among variables

When analysing the association among all variables measured using a PCA, we find that the first axis (PC1) explains 29 %, the second axis (PC2) explains 20 % and the third (PC3) explains 16 % of the total variance (Fig. 4; Supplementary data Table S2). PC1 is mainly associated with vessel diameter, and wood volume allocated to the fibre wall and to axial parenchyma. PC2 is mainly associated with wood density, and wood volume allocated to fibre lumen and to rays (Fig. 4).

Fig. 4.

Principal component analysis of wood anatomical traits and wood density of 75 Cerrado species. The variables represent the fraction of the wood volume allocated to each xylem tissue, wood density (g cm^–3) and vessel diameter (μm).

Regarding the associations between xylem tissue fractions and wood density, the PCA detects a positive association between wood density and wood volume allocated to the vessel wall and fibre wall (Fig. 4); the latter trait is also detected by the LMM in a positive relationship with wood density (Fig. 2A). The PCA confirms the negative relationship between wood density and wood volume allocated to fibre lumen, detected by the previously described model (Fig. 2B). The PCA reveals that wood density is negatively associated with vessel diameter and wood volume allocated to G-fibres (Fig. 4).

The PCA shows a negative association between wood volume allocated to fibre traits (fibre wall, fibre lumen and parenchyma-like fibre bands) and that allocated to parenchyma cells (rays and axial parenchyma). The analysis also reveals a negative association between vessel diameter and wood density and wood volume allocated to vessel wall (Fig. 4). Also, the PCA confirms the absence of a relationship between vessel diameter and wood volume allocated to the fibre lumen and to axial parenchyma and the parenchyma-like fibre fraction, detected previously by the model (Fig. 3E, F).

DISCUSSION

Wood volume allocation to different cell types in Cerrado species

In this study, we assessed the partition of wood volume to different xylem cell types of Cerrado species to test the functional trade-offs between hydraulics, mechanics and storage functions in trees. The largest fraction of wood volume is allocated to fibres, which show the lowest variability, in contrast to the largest variability presented by the total parenchyma fraction which occupies the second position in the allocation of wood volume fraction (Fig. 1B). The extremes of variability occupied by fibre and parenchyma fractions can be interpreted based on the negative correlation between the volume fraction of fibre and parenchyma, which will consequently result in a functional trade-off between mechanical and storage functions (Morris et al., 2016; Pratt and Jacobsen, 2017), a matter which we discuss in the next paragraphs. In contrast, the variation of vessel fractions appears to be more independent than the coupled variation of fibre and parenchyma fractions. Therefore, the remaining xylem volume occupied by vessels was not as variable as the volume occupied by fibres and parenchyma, as also acknowledged in the literature (Zanne et al., 2010; Percolla et al., 2021).

The lower variability in total fibre fraction might have an ecological explanation based on the fact that the variation in morphological traits is the key to understanding how an organism responds to ecological factors and how such a response, in turn, affects the ecosystem functioning (Díaz and Cabido, 2001; Raffard et al., 2017; Wright et al., 2017). Therefore, the lower variability of fibre fraction coupled to its greater fraction in wood volume might reflect the degree of importance of the biological roles played by fibres in the performance and survival of species, and thereby it is under important selection, which potentially results in low trait variability (Pulido-Rodriguez et al., 2020). In addition to their function in mechanical support, fibres influence plant hydraulics via the hydraulic–mechanical trade-off. This trade-off is operated by an increase in either the ratio between vessel wall and lumen area, or the supportive fibre matrix, which would both enhance wood density and prevent vessel implosion (collapse) (Hacke et al., 2001; Jacobsen et al., 2005; Schweingruber et al., 2006). In accordance, collapse of xylem conduits has been experimentally observed only in cells that lack a robust support of fibre matrix, for instance in leaves (Cochard et al., 2004; Zhang et al., 2016) and in low lignin stems of poplar mutants (Kitin et al., 2010). Also, several studies show that an increase in mechanical strength, i.e. a higher percentage of lignified cell types, is linked to higher embolism resistance (Jacobsen et al., 2007; Pratt et al., 2007; Doria et al., 2018, 2019b). In addition to the effect on hydraulic functions, fibres can also play a role in storage via septate fibres and parenchyma-like fibre bands (Carlquist, 2014; Pratt et al., 2021b).

Wood density is explained by the wood volume allocated to cell wall and void volume

The main sources of wood density variation are the wood volume allocated to the fibre wall and fibre lumen fractions, which supports our previous expectation and confirms the previous findings in the literature for neotropical, Australian and French Guiana species (Jacobsen et al., 2007; Martinez-Cabrera et al., 2009; Zieminska et al., 2013; Fortunel et al., 2014; Janssen et al., 2020). The relative importance and strength of the two relationships explaining wood density variation is similar, but with opposite effects. As larger fibre lumens are associated with lower wood density, this leads to an inverse relationship between water conductivity and storage and the wood volume allocated to cell wall fraction – a matter which we discuss below. In this way, capacitance is found to correlate with wood density, such that species with lower wood density have higher capacitance (Scholz et al., 2007; McCulloh et al., 2012).

Wood density variation also appears to be explained by the variation of wood volume allocated to the vessel wall. These results support the idea that the main drivers of wood density are the cell walls and void wood volume (Lachenbruch and McCulloh, 2014), represented here by the volume fraction of fibre and cell walls, and the fractions of fibre lumen. Additionally, since wood density is linked to the variation of fibre volume fraction, we are tempted to suggest a close relationship between wood density and mechanical properties, modulus of elasticity and rupture, which represent the wood components that bear forces against micro-failures (microscopic breaks) (Onoda et al., 2010; Plavcová et al., 2019).

Our results show a negative relationship between wood density and wood volume allocated to G-fibres. G-fibres are specialized cells characterized by the presence of an inner cell wall layer, named the G-layer, that exhibits gel-like shrinkage during drying (Clair et al., 2008). The walls of G-fibres contain low lignin and high cellulose contents, with the angles of the microfibrils of cellulose in the G-layer arranged in a different orientation from that of the ordinary fibres (Barnett and Jeronimidis, 2003), which provides an important mechanical function reinforcing the structure and stability of mature stems (Fisher, 2008; Abasolo et al., 2009). Despite their importance for the mechanical function, our results show that a high wood volume allocated to G-fibres does not enhance wood density. This strategy may be energetically advantageous because production of cellulose molecules, abundantly present in G-fibres, is less expensive than the production of the complex organic polymers of lignin present in the walls of ordinary (libriform) fibres. Therefore, Cerrado species can profit from the higher mechanical strength under tension provided by the G-fibres, without investing in wood density.

Simultaneously, increased wood density is linked to narrower vessels, and an increase in wood volume allocated to G-fibres is linked to larger vessels. The negative relationship between increased wood density and vessel diameter corroborates the findings that species with lower wood density tend to be more vulnerable to embolism than species with higher wood density (Jacobsen et al., 2007; Christoffersen et al., 2016; Dória et al., 2018; Olson et al., 2020). The functional explanation for such a relationship relies either on the fact that narrow vessels potentially confer embolism resistance (Jacobsen et al., 2019; Olson et al., 2020), or on the fact that high wood densities (thicker cell walls) are associated with greater ability of conduits to resist deformation and air seeding (Hacke et al., 2001; Bittencourt et al., 2016), while thinner cell walls (low wood density) have conduits that are more prone to deform or to suffer micro-ruptures under high negative pressure, which could lead to distension of the intervessel pit membranes, allowing the air bubbles (embolism) to spread from one vessel to another (Li et al., 2016; Olson et al., 2020). Supporting that idea, a recent meta-analysis showed that selection apparently eliminates individuals with vessel walls ‘too thin’ for their diameter (Echeverría et al., 2022). Given that large vessels increase hydraulic conductivity (Tyree and Zimmerman, 2002) and decrease wood density, our study suggests that Cerrado species counteract the trade-off between hydraulic efficiency and mechanics by an increase in mechanical resistance provided by the G-fibres, without necessarily affecting wood density variation.

Functional trade-offs in wood volume allocation to different cell types

The allocation of a larger volume of wood to parenchyma cells (axial parenchyma and rays) results in less wood volume allocated to the cell wall. A higher amount of parenchyma in wood would contribute to a more embolism-vulnerable xylem since it leads to a lower fraction of fibres, which is linked to low wood density and, consequently, a mechanically weaker stem and thus less embolism-resistant xylem (Lens et al., 2016; Pereira et al., 2017; Dória et al., 2018, 2019b). In the same way, a higher amount of parenchyma is linked to a high level of non-structural carbohydrate storage, and consequently to a dehydration resistance (Pratt et al., 2021b). Accordingly, species from tropical humid forests vulnerable to embolisms generally show a higher allocation of wood volume to axial parenchyma than fibre walls (Janssen et al., 2020), which is a strategy for storing non-structural carbohydrates that prevent carbon starvation through the maintenance of a homogeneous activity and distribution of carbohydrates (Plavcová et al., 2016; Dickman et al., 2018; Janssen et al., 2020). On the other hand, the dynamics of non-structural carbohydrates have been recognized as mediating drought-induced tree mortality across tropical species (Signori-Muller et al., 2021), and experimental evidence shows that plants with a higher non-structural carbohydrate content show higher survivorship under drought by avoiding dehydration (Adams et al., 2013; Pratt et al., 2021b). Species belonging to common families in the Cerrado, e.g. Fabaceae and Vochysiaceae, are characterized by large fractions of parenchyma surrounding the vessels (confluent and aliform parenchyma), and therefore have the largest amount of paratracheal parenchyma. Therefore, the selective advantage of a large amount of axial parenchyma in species from a seasonally dry environment such as the Cerrado may rely on the avoidance of dehydration via greater non-structural carbohydrate storage. In addition to the role of non-structural carbohydrates, parenchyma also plays a role in wood capacitance. Water is stored in elastic tissues, i.e. in parenchymatic cells, and as they shrink and swell the water is released into the transpiration stream, as shown by a moisture release curve (Jupa et al., 2016; Pratt and Jacobsen, 2017), though the release of water from living cells generally occurs at more negative water potentials than that from capillary (cell lumen) storage (Jupa et al., 2016).

The absence of a relationship between vessel diameter and axial parenchyma fraction, as well as between vessel diameter and fibre lumen fraction, contradicts our expectation since the positive relationship between conductance and storage functions in wood is recognized across species and biomes (Jupa et al., 2016; Morris et al., 2016; Pratt and Jacobsen, 2017). However, although larger vessels do not lead to a high fraction of parenchyma cells or a high fraction of fibre lumens, it does lead to a high wood volume allocated to gelatinous fibres. Besides the mechanical function under tension stress developed by the G-layer in G-fibres, the hygroscopically active nature of the G-layers due to the high contents of hydrophilic pectins, hemicelluloses and arabinogalactans is largely accepted (Arend, 2008; Bowling and Vaughn, 2009). Such a nature has also been confirmed by in situ Raman spectroscopy (Gierlinger and Schwanninger, 2006). It is an attempt to functionally associate vessel diameter and the fraction of G-fibres based on the known association between axial parenchyma and vessel diameter which ultimately reflects on water transport efficiency (Trifilò et al., 2014; Morris et al., 2018). This association is functionally explained by the increased water released into the transpiration stream mediated by the mobilization of the stored non-structural carbohydrates in parenchyma cells, which would counteract the negative pressures inside large vessels, and reduce the probability of embolism formation (Holbrook, 1995; Borchert and Pockman, 2005; Beikircher and Mayr, 2017; Trifilò et al., 2019). This hypothesis is supported by the larger vessel-facing pits (Plavcová and Hacke, 2011) and high numbers of aquaporins observed in vessel-associated cells, suggesting an exchange of water between these cells and the vessels (Sakr et al., 2003; Almeida-Rodriguez and Hacke, 2012). Additionally, the release of water stored in cell compartments into the transpiration stream is driven by the percentage of fibre lumen area (Pratt et al., 2007), which in the case of the species studied here is likely to be influenced by the percentage of G-fibre. Therefore, our data strongly suggest that G-fibres of Cerrado species may be providing storage for water to flow into the transpiration vessel stream. The mechanism of water flow from the G-layer to the transpiration stream is still unknown and deserves future investigations.

Our expectation of a trade-off between mechanical and hydraulic functions by the negative relationship between wood volume allocated to lignification characters (wood density and vessel wall) and vessel diameter was confirmed. Wood density has been used as a proxy for hydraulic properties (Lachenbruch and McCulloh, 2014; Olson et al., 2020). However, there is no evidence that wood density would directly impact hydraulic conductivity, and therefore it also has been referenced as a poor proxy for conductivity in other studies (Christoffersen et al., 2016; Gleason et al., 2016). The relationships reported in the literature could be due to the correlations between traits that affect lignification, and consequently wood density, also affecting the traits directly linked to hydraulics, such as thicker intervessel pit membranes (Dória et al., 2018, 2019b), or by the link between the void volume that is potentially available for water (affecting conductivity), also affecting wood density (low wood density) (Zanne et al., 2010).

Conclusions

In this study, we measured wood traits in 75 species from the Brazilian Cerrado to assess the functional space partitioning of the xylem components and investigate the presence of trade-offs between hydraulic, storage and mechanical functions. The largest fraction of wood volume is allocated to fibres, followed by parenchyma and vessels. Species show evidence of trade-offs, i.e. conflicting demands, between hydraulics and mechanical functions, and between mechanics and storage. The expected relationships between hydraulics and storage of water and carbohydrates were not confirmed in the studied species. However, the positive relationship between hydraulic conductance (vessel diameter) and wood volume allocated to G-fibres indicates a possible role for G-fibres in water storage and release into the transpiration stream. Wood density is explained by the variation of wood volume allocated to the cell wall (fibre and vessel wall) and void volume (fibre lumen and vessel lumen). An increase in the fraction of G-fibres is not linked to an increase in wood density, suggesting that Cerrado species can profit from the higher mechanical strength under tension without investing in wood density. Evidence of trade-offs among the wood traits of Cerrado species sustains the hypothesis of the wood fast–slow spectrum, i.e. expensive investment (high wood density and large fraction of cell wall) withstanding high hydraulic efficiency (large vessels).

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Table S1: measurements of the volume fractions of wood tissues of the 75 studied species of Cerrado. Table S2: output of the three principal components of the principal component analysis of wood traits of 75 Cerrado species. Figure S1: Pearson correlation matrix of all the traits assessed in the 75 Cerrado species studied.

FUNDING

This work was supported by CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior, Brazil, as part of the programme CAPES PrInt UNESP [grant nos 88887.468867/2019-00 and 88887.508247/2020-00], and by Fonds de Recherche du Québec Nature et Technologies (FRQNET), Canada, in the programne PBEEE - Bourses de court séjour de recherche ou perfectionnement Québec-Brésil [grant no. 304990], both to L.C.D; and FAPESP - São Paulo State Foundation for Science, São Paulo State Government, Brazil [grant nos 2003/13578-9, 2009/17778-9, 2012/50413-7, 2015/14954-1 and 2019/09417-8]; CNPq – Brazilian National Council for Science, Brazil [grant no. 304715/2018-2] and a research productivity fellowship and grant to C.R.M.