Three-dimensional xylem networks and phyllode properties of co-occurring Acacia

Authors


G. F. M. Page. Fax: +61 8 6488 7925; e-mail: page@graduate.uwa.edu.au

ABSTRACT

Reduced leaf size is often correlated to increased aridity, where smaller leaves demand less water via xylem conduits. However, it is unknown if differences in three-dimensional (3D) xylem connectivity reflect leaf-level adaptations. We used X-ray microtomography (micro-CT) to quantify 3D xylem connectivity in ∼5 mm diameter branch sections of co-occurring semi-arid Acacia species of varied phyllode size. We compared 3D connectivity to minimum branch water potential and two-dimensional (2D) vessel attributes derived from sections produced by micro-CT. 2D attributes included vessel area, density, vessel size to number ratio (S) and vessel lumen fraction (F). Trees with terete phyllodes had less negative water potentials than broad phyllode variants. 3D xylem connectivity was conserved across all trees regardless of phyllode type or minimum water potential. We also found that xylem connectivity was sensitive to vessel lumen fraction (F) and not the size to number ratio (S) even though F was consistent among species and phyllode variants. Our results demonstrate that differences in phyllode anatomy, and not xylem connectivity, likely explain diversity of drought tolerance among closely related Acacia species. Further analysis using our approach across a broader range of species will improve understanding of adaptations in the xylem networks of arid zone species.

INTRODUCTION

Drought tolerance of plants in semi-arid and arid environments can be attributed to a suite of adaptations, which include modifications to both leaf and branch hydraulic architecture. Reduced leaf size is one of a suite of functionally coordinated traits correlated with increasing aridity (Givnish 1987; Wright et al. 2004; Ordoñez et al. 2009). However, as leaf size increases, so does the demand for water transported via the xylem pathway (Sperry et al. 2002; Brodribb 2009). An increase in mean vessel area may improve water transport efficiency but be at greater risk of cavitation, which may explain why many arid species have narrow xylem vessels (Zimmermann 1983; Hacke, Jacobsen & Pratt 2009). However, while the relationships between leaf and branch traits is well demonstrated among species and across large environmental gradients (McDonald et al. 2003; Wright et al. 2004; Schenk et al. 2008; Ordoñez et al. 2009), it remains unclear how xylem architecture may relate to intraspecific variation in leaf-level properties even though significant variability has been noted within species and at localized scales (Sack & Holbrook 2006; Choat, Sack & Holbrook 2007; Beaumont & Burns 2009).

The three-dimensional (3D) arrangement, or degree of overlap among xylem vessels, may be a significant determinant of the drought tolerance of woody species, as the connectivity among conduits will influence the continuity of the water column while under stress (Sperry et al. 2002). Pits on the sides and end walls of vessels facilitate the flow of water between vessels and also dictate the air-seeding pressure required to induce cavitation, as smaller pores generally require a greater pressure differential before gas enters a conduit (Zimmermann 1983; Christman, Sperry & Adler 2009). Consequently, the largest pore in a vessel will determine its vulnerability to cavitation. The ‘rare pit hypothesis’ states that as pit area per vessel increases, so does the probability of the occurrence of a rare larger pore (Choat et al. 2005; Hacke et al. 2006; Christman et al. 2009). While total pit area per vessel and vulnerability to cavitation are correlated (Wheeler et al. 2005; Hacke et al. 2006), pit area must vary independently of vessel surface area; otherwise, a shift towards smaller vessels in more drought-tolerant species would be confounded by an increase in vessel surface area. A decrease in pit area is thus more likely to be a function of less vessel overlap rather than of reduced vessel surface area (Wheeler et al. 2005). Consequently, accurate and quantitative determination of vessel overlap will significantly improve understanding of the mechanisms of drought tolerance in any woody species.

While the 3D arrangement of xylem networks is a significant factor in the propagation or containment of hydraulic failure, most investigations of hydraulic function have been restricted to analysis of two-dimensional (2D) xylem structure. One reason for this is that physical ‘pipe-based’ models for understanding xylem function are primarily concerned with cross-sectional vessel attributes. According to the Hagen–Poiseuille equation, the conductivity of an individual vessel scales to the fourth power of its radius. Therefore, differences in the number and diameter of vessels strongly regulate the water transport capacity of xylem (Zimmermann 1983; Zanne et al. 2010). However, while it is often assumed that the trade-off for having more efficient (larger) vessels is a greater risk of cavitation, the 3D arrangement of xylem is not often considered. Larger vessels may not necessarily be at greater risk of cavitation from air seeding when isolated from other vessels (Ellmore, Zanne & Orians 2006; Zanne et al. 2006). A recent study across 3005 angiosperm species found that the size to number ratio of xylem vessels (S) varied independently of the proportion of branch cross-sectional area comprised of vessel lumen (F) (Zanne et al. 2010). Furthermore, mean vessel area was also negatively correlated with vessel density (Zanne et al. 2010). Consequently, an increase in mean vessel area will increase hydraulic efficiency, but not necessarily at the expense of safety if the vessel lumen fraction (F) remains constant or decreases. The degree of connectivity within a 3D xylem network may thus have a large influence on both efficiency and safety (Loepfe et al. 2007), although safety may not be directly related to lumen area (Zanne et al. 2006).

X-ray computed microtomography (micro-CT, or high-resolution computed tomography) is a rapidly developing method for investigating the 3D structure of xylem vessel networks. Only one study so far has used micro-CT to quantitatively analyse the connectivity of xylem vessel networks (Brodersen et al. 2011), improving upon earlier studies that tested the technology as a visualization tool (Steppe et al. 2004). Micro-CT has the advantage of non-destructive and rapid capture and analysis of large volumes of 3D data, which was either impossible to attain through conventional 2D methods or was too time consuming to be practical. However, while there have been rapid advances in the ability to capture and analyse 3D xylem network data (e.g. Brodersen et al. 2011), there has been no quantitative application of this technology to studies into plant adaptation, including drought tolerance.

Here, we used micro-CT to quantify the 3D xylem connectivity of intact branch segments at 3.40 µm resolution from co-occurring forms of Acacia aneura F. Muel. ex Benth and Acacia ayersiana Maconochie. A. aneura (Mulga) and A. ayersiana are the dominant canopy species of mulga woodlands and shrublands that occur across much of semi-arid Australia (Johnson & Burrows 1994). These species exhibit large variation in their tree architecture, and in their phyllode (leaf analog) shape and size (specific leaf area, SLA: 1.5–3 mm2 mg−1) (Pedley 1973; Randell 1992; Miller, Andrew & Maslin 2002). Mulga is also renowned for its drought tolerance (Slatyer 1961, 1965; O'Grady et al. 2009) and its capacity to persist across a wide variety of habitats in arid and semi-arid Australia (Johnson & Burrows 1994; Miller et al. 2002). We recently determined that different phyllode shapes are not uniformly distributed across a semi-arid landscape sequence (Page et al. 2011). Terete phyllodes were dominant in landscape positions that received the least run-off, which we predicted was a consequence of greater drought tolerance and lower rates of water use. In this study, we compared phyllode SLA, microstructural features of xylem connectivity and estimates of hydraulic capacity with pre-dawn and midday branch water potential. We hypothesized that (1) species with smaller and more terete phyllodes would have more negative seasonal leaf water potentials than broader phyllode types; (2) trees with more negative water potentials would exhibit less connectivity within their vessel networks; (3) measures of connectivity within vessel networks would not be correlated with differences in 2D mean vessel lumen area (A), but (4) connectivity would be sensitive to differences in the vessel lumen fraction (F).

MATERIALS AND METHODS

Sample site and selection of branches

We sampled phyllodes and branches from A. ayersiana and two phyllode variants of A. aneura in a low open woodland in a broad valley of the Hamersley Ranges in the Pilbara region of Western Australia (−23.045932S, 118.827095E). The two phyllode variants of A. aneura targeted in this study are common in the mulga woodlands of the Pilbara. Both phyllode types are typically less than 10 cm long but differ in cross section. ‘Terete’A. aneura has almost cylindrical phyllodes, roughly 1 × 1 mm (SLA ∼1.5 mm2 mg−1), whereas ‘broad’A. aneura are roughly 1 × 3 mm (SLA ∼2 mm2 mg−1). A. ayersiana typically have much broader phyllodes and are 0.5 × 12 mm in cross section (SLA ∼3 mm2 mg−1). We selected nine trees that included three replicates each of A. aneura (terete), A. aneura (broad) and A. ayersiana. All trees were approximately the same height (4–5 m) and selected for consistency in growth habit to minimize any morphological differences other than SLA (Fig. 1). Three sub-plots spread over 100 m were selected so that they contained one of each species and phyllode variant within 25 m of one another. One branch was sampled at 2 m height from all nine trees, and selected to be within a range of 70–100 cm long and 5–8 mm in diameter at the cut end. Phyllodes were collected from each branch for measurement with a leaf area meter (LI-3100, Li-Cor Inc., Lincoln, NE, USA), and 10 cm from the terminal end of each branch was cut and transported back to the lab for scanning using micro-CT. All samples were collected from the field in May 2008, following 2 months without any effective rainfall.

Figure 1.

Trees and corresponding phyllode shapes selected for this study. (a) terete Acacia aneura, (b) broad A. aneura and (c) Acacia ayersiana. Each of these three phyllode groups had three replicate trees, one of which is pictured here. Phyllode sizes in picture inserts are relative, and all three trees are of similar height (4–5 m).

Branch water potential

Water potentials were measured on the same day that samples were collected for micro-CT. Measurements were made using a Scholander pressure chamber fitted with a 10 MPa analog gauge (PMS Instrument Co., Albany, OR, USA). Five branches were cut from equidistant positions around the canopy 2 m from the ground at 1 h before sunrise and at 12 pm. Small terminal branches ∼10 cm long were then cut from each branch with a single-sided razor blade and measured. All measurements were made within 5 min of the branch being cut from the tree.

Data acquisition using X-ray micro-CT

One 10–15 mm segment from the terminal end of each branch was scanned at a resolution of between 2.90 and 4.11 µm with a SkyScan 1172 microtomograph (SkyScan, Belgium). Each branch was scanned in 0.30° steps of a 180° rotation with 16 frame averaging. Data were output as transverse slices in a stack of 900 greyscale bitmap images with a resolution of 2000 × 2000 pixels (Fig. 2a). 3D data were reconstructed using Avizo 6 software (Visualization Sciences Group, Burlington, MA, USA), re-sampled to a resolution of 3.40 µm and thresholded to select the xylem vessels with a minimum and maximum greyscale value of 0 and 10 (Fig. 2b,c). While the original model output from the microtomograph was 900 voxels (a volumetric pixel, or 3D pixel) in length, one branch sample had split while drying for half of its length, and so all models were shortened to 450 voxels. As xylem vessels are not uniformly distributed across a transverse branch section, we expected the porosity and permeability of the branch models to vary in sections between the central axis (heartwood) and the outer layers (vascular cambium and bark). Rectangular sections were required for the model analysis, which made it difficult to ensure that different axial segments were evenly represented. Therefore, rectangular blocks of equal size (500 × 500 × 450 voxels) were extracted from each branch model, which bordered the heartwood and outer layers for one half of the branch (Fig. 2a). Thresholded 3D images were then labelled in a binary format where voxels inside the threshold limits (vessel lumen) were labelled ‘1’ and everything else labelled ‘0’, and then exported for computational analysis following the methods of Liu et al. (2009) and Liu & Regenauer-Lieb (2011).

Figure 2.

Creating three-dimensional (3D) models of xylem networks from microtomography (micro-CT) data. (a) Singular transverse slices in greyscale bitmap format from micro-CT output data. 500 × 500 × 450 voxel subsamples (1.75 × 1.75 × 1.575 mm) are outlined by white squares on the raw image. (b) Transverse view of subsampled blocks thresholded at 0–10. Voxels thresholded as vessel lumen are coloured red. White scale bars represent 435 µm. (c) Tangential view of subsampled blocks thresholded at 0–10, with elongated vessels clearly visible and selected in red. White scale bars represent 435 µm. (d) Visualisation of 3D vessel network model, which was then exported in binary format for computational analysis. For scale, the cube is 1.575 mm tall, and 1.75 mm along the top edge.

Connectivity of 3D xylem networks

Voxels labelled as vessel lumen were assigned unique cluster labels when they shared a boundary. In this way, all vessels that were ‘touching’ were assigned the same cluster label; vessels that were not connected at some point in the 3D vessel network were thus readily distinguished (Fig. 3; different colours distinguish between clusters). When a cluster was present on opposite sides of the microstructural model, it was determined to be ‘percolating’, or capable of acting as a pathway for water movement through the branch. It was then possible to determine the size and number of percolating clusters of interconnected vessels present in each branch model, as well as the overall ‘porosity’ of the branch (the ratio of vessel lumen to total model size).

Figure 3.

Surface models of xylem vessel clusters reconstructed from micro-CT images to demonstrate the quantification of the scale of connectivity among xylem vessels. (a) Oblique and transverse views of xylem vessel clusters over four shrinkage operations in a small sub-volume (4.05 × 107 µm3) extracted from a branch model. Separate clusters of interconnected vessels are identified by colour, and vessels not connected to a percolating cluster have been filtered out of the image. Voxel size of the sub-volume is 3.4 µm3. (b) Magnified transverse sections of the sub-volume before and after the first shrinkage operation. The large green cluster of interconnected vessels was split into three separate vessels after shrinking all sides of the cluster inwards by 3.40 µm, and non-percolating vessels filtered out of the image.

To compare the scale of connectivity, or sectorality among branches, we conducted a ‘shrinkage analysis’ by moving the boundaries of every cluster progressively inwards in steps of 1 to 5 voxels (3.40–17.0 µm) and then re-analysing for porosity and percolation (Liu & Regenauer-Lieb 2011). In this way, marginal or very small connections among vessels would be disconnected by small shrinkage operations (i.e. 6.8 µm at the first shrinkage), but more robust connections would be broken only by larger shrinkages (Fig. 4). If sections of a cluster were separated by a shrinkage, the cluster would be differentiated in the re-analysis resulting in more clusters of smaller sizes (Fig. 3). For example, we expected that branches with more isolated vessels, or a higher degree of sectorality, would exhibit a smaller change in the number and size of percolating clusters than branches with more interconnected vessel networks. We calculated a ‘connectivity index’ by dividing the number of clusters before a shrinkage by the number of clusters after. An index <1 indicated an increase in the number of clusters (i.e. greater connectivity), while an index of >1 indicated a decrease (less connectivity). We also calculated the minimum size of the largest cluster in each microstructural model that contained 50% of the voxels labelled as vessel lumen. We expected that this measure would provide another intuitive proxy of connectivity. Greater connectivity within a vessel network would result in less overall clusters, and so fewer but larger clusters comprising 50% of the voxels labelled as vessel lumen, and thus a larger ‘minimum size’.

Figure 4.

Visualizing disconnection of vessel clusters by breaking connections among vessels over four shrinkage steps. The original model is green, and progressive shrinkage steps are indicated by arrows numbered according to each step. Two connections broken by the first shrinkage step are identified by red squares, and the resulting distinct vessel clusters are delineated by colour. A third, more robust connection is only broken by the second shrinkage step (circled). Further shrinkage steps do not break apart any more clusters.

2D analysis of xylem and estimation of theoretical specific leaf conductivity

The total number of vessels, and the area of each vessel lumen, was determined using particle size analysis on three 2D transverse slices from each branch taken from the output of the X-ray microtomograph using ImageJ version 1.42q (Abramoff, Megalhaes & Ram 2004). Each image was thresholded with greyscale values of 0–10 to match the 3D model analysis. The number and size of each particle in the resulting binary image was then exported into R version 2.9.2 (R Development Core Team 2010). The radius of each vessel was calculated using Eqn 1, assuming that each particle was the equivalent of a round xylem vessel. The theoretical conductivity (Φ) for each vessel was calculated using the Hagen–Poiseuille equation (Eqn 2), where η is the dynamic viscosity of water (8.90 × 10−4 Pa·s). The branch theoretical conductivity (Kt) was the sum of all xylem vessels and had the units m4 Pa·s. This measure is also sometimes referred to as ‘hydraulic capacity’ (Reid et al. 2005).

image(1)
image(2)

Specific leaf conductivity (SLC) was calculated for each branch by dividing Kt by the one-sided leaf area. The Huber value (Hv) was calculated by dividing the branch cross-sectional area by the one-sided leaf area. We also calculated the mean vessel area (A), vessel density (N), vessel lumen fraction (F = AN) and the size to number ratio (S = A/N) for comparison with the global metrics of Zanne et al. (2010).

Statistical analysis

We compared the response of 3D xylem network attributes of each Acacia species and phyllode variant to shrinking each cluster over five steps using repeated measures analysis. Linear mixed models (LMM) were fitted using the ‘lmer’ function in the lme4 package (version 0.999375) in R version 2.9.2 (R Development Core Team 2010), with a fixed effect for ‘type’ (Acacia species and phyllode variant), ‘shrink step’ as a fixed ordered time covariate and a random intercept for each tree. A null model and a full model including an interaction between ‘shrink step’ and ‘type’ were compared using both a pair-wise chi-squared likelihood ratio test and a comparison of Akaike information criterion (AIC). When the chi-squared P-value was ≤0.05 and AIC values for the models differed by >10, the model with the smallest AIC value was chosen as the best fit, otherwise the most parsimonious model was preferred (Bolker et al. 2009). A P-value >0.05 indicated that there were no differences among the two species and phyllode variants of Acacia. Comparisons of all 2D branch attributes and vessel metrics were performed using one-way analysis of variance and post hoc comparisons with Tukey's honestly significant difference using the multcomp package (Hothorn, Bretz & Westfall 2008). Correlations among 2D vessel attributes and the connectivity index were tested using linear models fitted in R version 2.9.2.

RESULTS

Branch water potential

Branch water potential was extremely negative in all three phyllode variants. Terete A. aneura was significantly less negative both at pre-dawn and midday than the broad phyllode variant or A. ayersiana, a result that is contradictory to our first hypothesis. Broad A. aneura (−8.93 MPa) and A. ayersiana (−9.72 MPa) also had the most negative pre-dawn water potentials. At midday, branch water potential of both broad A. aneura and A. ayersiana was more negative than could be measured on the Scholander pressure chamber; consequently, for all subsequent data analyses, these samples were designated values of −11 MPa, which are consistent with maximum data reported for mulga elsewhere (Slatyer 1961). In contrast to the broad phyllode variants, branch water potential of terete A. aneura decreased from −6.91 MPa at pre-dawn to −8.13 MPa at midday (Table 1).

Table 1.  2D branch parameters and leaf attributes differ among phyllode variants of Acacia aneura and Acacia ayersiana
ParameterA. aneura (terete)A. aneura (broad)A. ayersianaP-value
  1. Standard errors are listed in parentheses below mean values, where n = 3 for SLA and Hv, n = 9 for water potential measurements and n = 9 for all other parameters which were three replicate transverse branch sections from each of three trees of each type. When parameters were significantly different among tree types in a one-way analysis of variance (P ≤ 0.05), Tukey's honestly significant difference was computed, and significantly different tree types are denoted with different superscript letters in boldface type.

  2. Hv, Huber value; SLA, specific leaf area; SLC, specific leaf conductivity.

Pre-dawn Ψ (MPa)−6.91a−8.93b−9.42b<0.001
(0.51)(0.14)(0.4)
Midday Ψ (MPa)−8.13a<−11b<−11b<0.001
(0.41)(na)(na)
SLA (mm2 mg−1)1.45a1.81a2.86b0.06
(0.06)(0.21)(0.08)
Hv (m2 m−2 × 10−4)27.7a12.8a,b7.33b<0.001
(8.36)(1.77)(0.36)
Mean vessel area (A, µm2)2602452100.31
(56.4)(33.5)(21.2)
Vessel density (N, mm−2)687a587a822b<0.001
(42.2)(52.0)(57.8)
Vessel lumen fraction (F = AN)0.180.140.170.21
(0.04)(0.01)(0.01)
Size to number ratio (S = A/N)3.77 e−7a,b4.35 e−7a2.63 e−7b0.02
(7.32 e−8)(8.31 e−8)(4.54 e−8) 
Kt (m4 Pa·s)1.71 e−131.25 e−131.06 e−130.12
(5.91 e−14)(2.70 e−14)(1.17 e−14) 
SLC (m2 Pa·s)1.14 e−11a3.59 e−12b3.04 e−12b<0.001
(3.35 e−12)(2.77 e−13)(3.99 e−13) 

3D analysis of xylem connectivity

We expected that trees with more negative water potentials (i.e. A. ayersiana) would exhibit less xylem connectivity. However, all three phyllode variants had consistent xylem connectivity within the vessel network over all five cluster shrinkages (LMM P > 0.05 for all 3D model parameters, full model compared to null model, Table 2 and Fig. 5). Shrinking each side of the cluster by 3.40 µm resulted in an increase in the number of clusters in all branches except for one sample of the broad A. aneura that remained the same and one sample of terete A. aneura that decreased slightly (data not shown). Further shrinkage resulted in a decrease in the number of percolating clusters in all branches (Fig. 5). The connectivity index was strongly related to the porosity of the sample, where higher porosity resulted in greater connectivity among clusters. The most porous branches also exhibited the largest increase in cluster number when shrunk by 3.40 µm on each edge (Table 2). The minimum size of the largest clusters in each microstructural model that contained 50% of the voxels labelled as vessel lumen decreased in all branches for each progressive shrinkage step (Table 2). All branch segments were strongly anisotropic and were only percolating in the z-axis, as would be expected for xylem vessels.

Table 2.  3D attributes of xylem networks reconstructed from micro-CT were consistent among the three closely related Acacia species
  Shrink stepP-value
Original12345
(3.40 µm)(6.80 µm)(10.2 µm)(13.6 µm)(17.0 µm)
  • All values listed are means where n = 3. Standard errors are given inside parentheses. P-values were produced from pair-wise chi-squared likelihood ratio tests between the null Linear mixed models (LMM) and the interaction LMM.

  • a

    A larger number indicates that fewer clusters compose 50% of the voxels in the microstructural model labelled as vessel lumen.

  • b

    Minimum cluster size data were log transformed before LMM analysis.

  • micro-CT, microtomography.

Porosity (%) A. aneura17.57.523.081.230.430.100.99
(terete)(3.74)(1.97)(1.02)(0.60)(0.29)(0.08)
A. aneura15.17.063.271.410.490.12
(broad)(1.19)(1.02)(0.60)(0.34)(0.17)(0.05)
A. ayersiana17.77.623.191.300.410.08
 (1.00)(0.51)(0.37)(0.22)(0.10)(0.04)
Number of active clustersA. aneura86.013589.740.02.3300.99
(terete)(71.0)(42.0)(43.2)(39.5)(2.33)(0)
A. aneura48.313488.329.73.00
(broad)(8.83)(39.0)(33.6)(16.8)(2.52)(0)
A. ayersiana36.316410127.72.000
 (16.4)(18.8)(14.8)(10.9)(1.00)(0)
Minimum size of clusters containing 50% of voxelsaA. aneura8.82 e43.18 e49.21 e35.20 e31.09 e31.88 e20.71b
(terete)(7.55 e4)(2.12 e4)(5.38 e3)(4.43 e3)(9.51 e2)(1.54 e2)
A. aneura2.13 e61.38 e47.83 e33.73 e31.20 e32.93 e2
(broad)(1.12 e6)(3.98 e3)(2.21 e3)(1.76 e3)(6.71 e2)(1.63 e2)
A. ayersiana2.40 e61.22 e46.83 e33.21 e31.12 e33.07 e2
 (1.51 e6)(1.63 e3)(6.45 e2)(9.50 e2)(4.78 e2)(1.29 e2)
Figure 5.

The ratio of the number of clusters before to that after a shrinkage step was consistent among Acacia ayersiana (▴), broad Acacia aneura (○) and terete A. aneura (inline image). The number of clusters increased after the first shrinkage (0/1), and then decreased with further shrinkage. The fifth shrinkage step resulted in zero percolating clusters in all branches, and is not included in this figure. The dashed horizontal line indicates the 1/1 line, below which indicates an increase in clusters following a shrinkage step, and above indicates a decrease. Points are group means with n = 3, and vertical bars represent 1 standard error of the mean.

2D analysis of xylem vessel attributes

Mean vessel area (A) was 238 µm2, and distributions of vessel diameters in branch wood were consistent among all three phyllode variants (P > 0.05, Table 1). Vessel density (N) of branch wood was highest in A. ayersiana (822 mm2), while the two phyllode variants of A. aneura had similar vessel densities of ∼637 mm2 (Table 1). However, theoretical hydraulic conductivity (Kt) and the fraction of branch cross-sectional area occupied by vessel lumen (F) were consistent among the branches of all three phyllode variants (Table 1). The size to number ratio (S) is a metric of the relationship commonly observed between mean vessel area (A) and density (N), where larger values are indicative of fewer but larger vessels, and small values represent many vessels with small lumen area (Zanne et al. 2010). We found that A. ayersiana had the lowest vessel size to number ratio (S) (2.63 × 10−7), while broad A. aneura was significantly higher (4.53 × 10−7), and terete A. aneura was not significantly different from either (4.35 × 10−7) (Table 1).

Comparison of 2D vessel characteristics with 3D connectivity

The first shrinkage step was compared against 2D vessel characteristics as it was the only step that resulted in an increase in the number of clusters and thus measured connectivity. As expected (hypothesis 3), neither mean vessel area (A) nor vessel size to number ratio (S) was correlated with the connectivity index (P > 0.05). However, the vessel lumen fraction (F) was negatively correlated with connectivity (Fig. 6, P < 0.001). Thus, as expected (hypothesis 4), there was greater connectivity in branches that had a higher proportion of the stem cross-sectional area allocated to vessel lumen.

Figure 6.

Logarithmic correlation between the vessel lumen fraction (F) and the connectivity index at the first shrinkage for all trees. Connectivity was greatest in branches with more branch cross-sectional area composed of vessel lumen. The fitted line is for all trees, but species and phyllode variants are identifiable: Acacia ayersiana (▴), broad Acacia aneura (○) and terete A. aneura (inline image).

Linking 2D vessel characteristics and 3D connectivity to physiological attributes

Contrary to hypothesis 2, pre-dawn and midday branch water potentials were not correlated with the connectivity index (P > 0.05). However, differences in both pre-dawn and midday water potentials between terete and broad phyllode variants were consistent with differences in SLC. SLC was more than threefold greater in terete A. aneura (1.14 × 10−11) than either broad A. aneura (3.59 × 10−12) or A. ayersiana (3.04 × 10−12, Table 1), but there were no differences in theoretical hydraulic conductivity (Kt) when not scaled for leaf (phyllode) area. While the Hv is also scaled by leaf (phyllode) area, it was consistent both between species and among phyllode variants (Table 1) and, thus, did not correlate with water potential.

DISCUSSION

Our results demonstrate that 3D connectivity is conserved within the xylem networks of two closely related and co-occurring Acacia species, regardless of phyllode type (broad or terete) and minimum branch water potential (Table 1 & 2, Fig. 5). Whole-plant co-ordination of drought adaptation has been suggested as an explanation for non-convergence in branch-level attributes in arid environments (Meinzer et al. 2010). Clearly, our investigation of the role of micrometre-scale xylem connectivity in drought tolerance of terminal branches ∼5 mm in diameter must be considered in the context of the whole transport pathway and, indeed, the whole plant, as the selection for traits that confer fitness under the prevailing environmental conditions occurs at the organismal scale (Meinzer et al. 2010). Furthermore, xylem efficiency and safety may not necessarily always be optimized in arid environments, particularly when plants are regularly prone to conditions that induce extensive levels of cavitation (Hacke et al. 2009). Instead, selection in favour of traits for repair or regrowth of xylem may be more important adaptations for plants in arid and semi-arid environments. While the tolerance of hydraulic systems to stress or maintaining function under increasing drought stress is likely a key attribute of plant fitness in arid environments, it may be equally as important to have the capacity to re-establish function after prolonged periods of hydraulic dysfunction. These closely related Acacia species from semi-arid Australia experience some of the most negative water potentials measured in woody vegetation and can spend large parts of the year with branch water potentials <−10 MPa (Slatyer 1961). Whole-plant adaptation to extreme desiccation is a necessity under these conditions, as is the ability to capture very limited rainfall that only occurs unpredictably throughout the year. Therefore, features of the xylem network that promote either prolonged function under extreme drought or, alternatively, rapid re-activation of the water transport pathway would be essential in these species.

While xylem connectivity was conserved at the species and phyllode variant level, our hypothesis of greater sensitivity of xylem connectivity to the vessel lumen fraction (F) was supported by a significant logarithmic correlation across all trees (P < 0.001, Fig. 6). The lack of correlation of xylem connectivity with either mean vessel area (A) or the size to number ratio (S) also supports the idea that these traits may not always be closely related to xylem network connectivity (hypothesis 3). Our results provide further evidence that vessel lumen fraction (F) and vessel size to number ratio (S) are independent axes for adaptation at the branch level (Zanne et al. 2010). Our data also indicate that even though only 4.8% of variation in mean vessel area (A) and density (N) at a global scale was accounted for by vessel lumen fraction (F) (Zanne et al. 2010), it may be a key trait in determining xylem connectivity and of particular importance at the intraspecific level and in congeneric species.

Smaller leaf size has previously been related to more negative seasonal water potentials and greater drought tolerance (Ackerly 2004; Scoffoni et al. 2011). However, we found the opposite; A. aneura with the smallest (terete) phyllodes maintained less negative branch water potentials. Terete A. aneura also had significantly greater SLC, a trait that may reflect lower rates of water use, which may reduce strain on the water column. In many species, stomatal control of transpiration limits xylem tension to below damaging levels under non-extreme conditions (Sperry et al. 2002; Meinzer et al. 2008), which when combined with lower rates of water use may conserve resources and result in higher minimum seasonal water potentials. Alternatively, our results may be partly explained by the fact that smaller leaf sizes are also correlated with higher temperatures and light exposure owing to a thinner boundary layer that enhances convective cooling (Givnish 1987; Ordoñez et al. 2009). Differences in water potential among these Acacia species and phyllode variants may thus also reflect differences in site characteristics that are not limited to water availability.

Theoretical modelling of xylem networks has concluded that higher connectivity increases efficiency but also results in a greater risk of embolisms being spread via the same pathway (Loepfe et al. 2007). Similarly, vessel overlap or pit area in 2D cross sections has also been correlated with cavitation resistance (Wheeler et al. 2005; Hacke et al. 2006). Consistency among the vessel networks of the three Acacia studied here suggests that at the branch level, they may share a similar level of resistance to the propagation of embolisms. Analysis across a large number of species indicates that the area of overlap between adjacent vessel surfaces must be small to achieve maximal cavitation resistance, yet merely having small pit areas does not confer similar benefits (Hacke et al. 2009). The lack of correlation between pit area and cavitation resistance relates to the ‘rare pit hypothesis’ whereby different species have different frequencies of unusually large pit membrane pores that can be a ‘weak link’ that reduces air-seeding thresholds (Christman et al. 2009). Despite interspecific differences in pit membrane properties which weaken the relationship between pit area and cavitation resistance, both pit area and vessel overlap are often used as a proxy for cavitation risk since, in general, the chances of encountering a ‘leakier’ pit increase with pit area (Wheeler et al. 2005; Christman et al. 2009). Our 3D model of xylem connectivity does not account for differences in membrane porosity among vessels or branches; although in such closely related species, we would not expect porosity to differ greatly. However, 3D models as presented here could be further improved by incorporating stochastic simulations of the ‘rare pit hypothesis’ to further test the influence of connectivity on vulnerability to cavitation and xylem dysfunction (Loepfe et al. 2007; Christman et al. 2009). Indeed, a recent study combined analysis of 3D xylem networks using micro-CT with visualization of pore size and location with electron microscopy (Brodersen et al. 2011). Such approaches will likely underpin improved understanding of the anatomical adaptations in the xylem networks of species adapted to the most extremely droughted environments.

ACKNOWLEDGMENTS

We thank Peter Austin at the Australian Minerals Research Centre for the image capture using micro-CT and Klaus Regenauer-Lieb for advice on image analysis. Funding and field support from Rio Tinto Iron Ore Pty Ltd is gratefully acknowledged, along with field assistance from Kaitlyn Height. We also thank Derek Eamus, Andrew Merchant and Stephen Burgess, who provided valuable comment on an early version of the manuscript. We also thank two anonymous reviewers and Prof Holbrook, who helped to improve the quality of the manuscript.

Ancillary