Ramification has little impact on shoot hydraulic efficiency in the sexually dimorphic genus Leucadendron (Proteaceae)

Plant material

Males and females of both species grow naturally near Du Toitskloof Pass, Western Cape, South Africa. All individuals were co-occurring within ∼20 m of each other on a slope extending westward toward the city of Paarl. The two species differ in the degree of serotiny, with L. rubrum holding its cones for an average of 2.8 years and L. daphnoides for less than one year (Williams, 1972; Harris & Pannell, 2010). Plants were collected and measurements made during November–December 2012 and during April–May 2013. In the field, mature plants were chosen, ignoring any unusually small individuals. Shoots were cut at the plant base with garden shears and immediately recut under water one node above the previous cut. Individual shoots were placed in dark, plastic bags, and their bases submerged in water during transport back to the lab. Shoots were stored in a 4 °C refrigerator until the day of measurement, and any shoots not measured within three days of field collection were discarded.

Measurement of hydraulic conductance

Immediately prior to hydraulic measurements, shoots were defoliated underwater. When necessary, a sharp blade was used to cut leaves at the petiole base. Stems were recut underwater with a new razor blade. Hydraulic conductance was measured on whole, branching shoots of various sizes using a low pressure flow meter (Kolb, Sperry & Lamont, 1996). With this method, the cut stem base was inserted into a compression fitting (Omnifit) that was connected to hard-sided tubing, the opposite end of which was submerged in a vial of filtered 0.01 M KCl that sat on a balance (Mettler-Toledo MS205DU with a resolution of 0.01 mg). The branching shoot was then placed inside a chamber connected to a vacuum pump. Flow rates from the balance were recorded automatically every 5–20 s (the frequency depended on the absolute flow rate) at each of a series of pressures below ambient: 10, 30, 50, 65, 40, 20 kPa. The stable flow rate at each pressure was determined when the coefficient of variation of the previous ten instantaneous flow rates was less than 5%. Hydraulic conductance, K, was obtained by a linear regression of flow rate as a function of pressure (kg s⁻¹ MPa⁻¹). To compare hydraulic efficiency between groups, K was normalized by either entire shoot leaf area (K_LA) or by cross-sectional area of the stem base (K_CSA).

Quantifying ramification and other traits

At the end of each measurement, shoots were assessed for other morphological traits. The cross-sectional area of the stem base was calculated from the average of two perpendicular measurements of stem diameter at the base using manual calipers. Shoot ramification was quantified in two ways. First, we quantified the rate of diameter change down the length the shoot, using a previously described method Harris & Pannell (2010) and described briefly here. Starting at the highest branch, we measured stem cross-sectional area at the midpoint of each consecutive internode down the shoot as well as the relative position of this midpoint along the length of the shoot. The slope of the relationship between the logarithm of stem cross-sectional area and the relative distance from the crown provides an index of ramification. In the previous application of this method, the tallest shoot on the plant was used and all nodes from the top t the base of the plant were included. However, because we sought to directly link shoot reamification to hydraulic conductance and wanted to capture a wide range of shoot sizes, the shoots we measured were not as large as those measured previously by Harris & Pannell (2010). Nonetheless, because their method for quantifying ramification depends on a linear regression, removing the basalmost nodes should not unduly impact estimates of ramification so long as the assumptions of linear regression are met. Furthermore, the method of quantifying ramification used by Harris & Pannell (2010) assumes that all branches arise immediately below terminal inflorescences and at nodes. For some species of Leucadendron, including L. daphnoides, this assumption is valid. However, in other species, particularly males of L. rubrum, most branches occur along the internodes. To account for these many small branches we measured the total number of branch tips on a shoot (regardless of their size and position on the shoot) normalized by the cross-sectional area of the stem base. The number of branch tips per stem cross-sectional area (BTSA) can easily be applied to shoots of any size, including shoot segments from the most recent year of growth that do not have properly defined internodes, as well as to other species with different branching patterns.

Leaves were placed on a flatbed scanner and their total area and number determined per shoot using ImageJ (Schneider, Rasband & Eliceiri, 2012). The average leaf size per shoot (termed here simply ‘leaf size’) was determined by dividing the total leaf area by the total number of leaves on the shoot. For consistency with Corner’s rules, we quantified the average leaf area per branch by dividing the total shoot leaf area by the total number of terminal branches on each shoot.

Data analysis

All data were analyzed using R v. 3.0.2 (Wickham, 2017; R Core Team, 2018). Scaling relationships were determined using standard major axis regression (SMA) as implemented in the ‘smatr’ package (Warton et al., 2012). For each pair of variables, relationships were determined for each species and sex combination, and differences between groups in scaling relationships determined using a likelihood ratio test as implemented in the sma function. For certain scaling relationships, slope tests were used to test whether slopes significantly differed from unity. If the likelihood ratio test (LRT) revealed no significant differences in scaling slopes between groups, data were pooled and the relevant slopes and statistics shown and reported. We also compared K_LA and K_CSA between groups using ANOVA with species, sex, and the interaction between them as factors. If either of these grouping variables were significant, post-hoc Tukey HSD tests were used to determine which groups were significantly different. For visualization purposes, 95% confidence intervals around scaling slopes were estimated by bootstrap sampling 1,000 iterations.

Stem and leaf architectural relationships

Stems with larger basal cross-sectional areas held more leaf area (Fig. 2). There was no significant difference in slopes between sex or species groups (LRT = 4.146, df = 3, P = 0.246), and the global slope was significantly greater than unity (slope = 1.11, R² = 0.83, P < 0.001).

The two metrics of branch ramification did not covary linearly, even in log–log space (Fig. S1.). Subsequent analyses focused solely on BTSA because it can be readily applied to any branching architecture and it conformed to Corner’s rules (Fig. 3). More highly ramified branches held significantly less leaf area per branch tip (slope = − 0.97, R² = 0.93, P < 0.001; slope test: r =  − 0.10, df = 65, P = 0.41) with no significant difference between groups (LRT = 1.50, df = 3, P = 0.68; Fig. 3A). Similarly, more highly ramified shoots had smaller leaves on average (Fig. 3B). Although species exhibited different elevations in their relationships, there was no significant difference in the slopes among groups (LRT = 2.74, df = 3, P = 0.43; L. rubrum: slope = − 0.39, R² = 0.87, P < 0.001; L. daphnoides: slope = − 0.45, R² = 0.46, P < 0.001).

To determine whether larger branches were more hydraulically efficient, we tested whether there were different scaling slopes between K and two metrics of shoot size: total leaf area and basal cross-sectional area (Fig. 4). There was no significant difference in slopes between groups for the scaling of K and leaf area (LRT = 6.19, df = 3, P = 0.10), and the slope test revealed that the scaling slope was not significantly different from unity (r = 0.0097, df = 66, P = 0.94; Fig. 4A). Thus, total shoot leaf area was a strong predictor of whole-shoot hydraulic conductance across species and sexes (R² = 0.83, P < 0.001). Similarly, there was no significant difference between groups in the scaling slopes between K and stem cross-sectional area (LRT = 5.93, df = 3, P = 0.12), and there was no significant difference between the observed slope and unity (r = 0.212, df = 58, P = 0.11; Fig. 4B). Thus, basal cross-sectional area was a strong predictor of whole shoot hydraulic conductance across species and sexes (R² = 0.69, P < 0.001).

To determine whether ramification impacted hydraulic efficiency, we tested whether there was a significant relationship between BTSA and leaf area-specific hydraulic conductance (K_LA) and between BTSA and hydraulic conductance (K) normalized by stem basal cross-sectional (K_CSA). There was no universal effect of BTSA on either K_CSA (P = 0.79; Fig. 5A) or K_LA (P = 0.13, Fig. 5B). Within certain species and sex groups, however, there were some significant effects of BTSA. Only in L. daphnoides females did more highly ramified individuals have lower K_CSA (R² = 0.30, P = 0.02; Fig. 5A). In contrast, more ramified L. rubrum males had higher K_CSA (R² = 0.42, P < 0.02; Fig. 5A). There was no significant effect of species, sex, or the interaction between species and sex on K_CSA in an ANOVA (all P > 0.05). Similarly, only for L. daphnoides females was K_LA lower for more ramified individuals (R² = 0.36, P = 0.01; Fig. 5B), and when L. rubrum males and females were pooled together there was a weak, but significant negative relationship between BTSA and K_LA (R² = 0.16, P = 0.02; Fig. 5B). ANOVA revealed a similar pattern, with a significant effect of the interaction of species and sex (F = 4.59, df = 1, P = 0.04) on K_LA. However, a Tukey post-hoc HSD test found no single pairwise comparison significant, even before Bonferroni adjustment, at the P = 0.05 level. The largest pairwise difference was between L. rubrum males and females, although this difference was not significant in either the Tukey post-hoc comparison (P = 0.14) or in a Welch’s two-sample t-test (t = 1.997, df = 18.55, P = 0.06).

Allometry of leaves and stems obeys Corner’s rules

Corner’s rules predict that “the stouter the stem, the bigger the leaves and the more complicated their form” (Corner, 1949). Corner’s first prediction between stem size and leaf size has been well-characterized for many species including Leucadendron (Bond & Midgley, 1988; Ackerly & Donoghue, 1998; Olson, Aguirre-Hernández & Rosell, 2009). Across species and sexes a single slope explained the relationship between total shoot leaf area and stem cross-sectional area with larger stems holding proportionally more leaf area (i.e., SMA slope not significantly different from unity; Fig. 2). This relationship further suggests that despite large variation in stem diameter both within and among species, leaf area is allocated proportional to stem diameter.

More apropos to the present study, Corner (1949) also predicted that “the greater the ramification, the smaller become the branches and their appendages”. This prediction emerges out of an argument of allocation: given a constant amount of biomass, producing more branches means that each of them must be smaller. In most species of Leucadendron, branches emerge just below terminal inflorescences and seed cones. This branching pattern results in internodal segments bracketed by nodes from which a variable number of branches emerge. In dimorphic Leucadendron, males have more branches at each node than their conspecific females, which is driven by pollinator selection for more inflorescences because inflorescences are borne only on terminal branches (Bond & Maze, 1999). Selection on inflorescence number, therefore, can result in more highly ramified branches and, because of Corner’s rules, smaller leaves (Midgley & Bond, 1989). Using BTSA as a simple metric of ramification, we found that more highly ramified shoots both within and among species and sexes bore less leaf area per branch, driven predominantly by each leaf being smaller (Fig. 3). These results were consistent with Corner’s second prediction, with isometric scaling between BTSA and leaf area per branch tip. Interestingly, while leaves were smaller on more ramified shoots, there was a species-specific effect such that the two species had different intercepts (Fig. 3B). In contrast, another metric of ramification recently used in a broad survey of Leucadendron species (Harris & Pannell, 2010) exhibited no similar scaling relationships with leaf size or with leaf area per tip (Fig. S2).

Little effect of shoot morphology on shoot hydraulics

We sought to test whether these large differences in leaves and stems among species and sexes in Leucadendron led to differences in stem hydraulic efficiency, as has been suggested (Harris & Pannell, 2010). Two main characteristics of woody shoots were thought to be important: the cross-sectional area of a branch and the number of branching nodes. These two traits are, to some extent, linked in Leucadendron because more highly ramified shoots are more likely to have more nodes and smaller terminal branches. As a rough approximation, larger diameter branches should have higher conductance because they have more xylem conduits (assuming no change in conduit size with branch size). However, wood is complex, being composed of vessels, fibers, parenchyma, and pith in various proportions (Zanne et al., 2010), such that hydraulic conductance does not necessarily scale isometrically with cross-sectional area. The complexity of wood structure-function relationships means that simplistic models that treat the stem as a pipe cannot accurately describe shoot water transport capacity (Lehnebach et al., 2018). For example, because stems continually grow, their theoretical hydraulic capacity always increases, yet as xylem ages it becomes progressively non-functional, such that most of the water transpired by leaves is delivered via the current year’s xylem (Melcher, Zwieniecki & Holbrook, 2003; Brodersen et al., 2019). As a result, cross-sectional area of the stem cannot necessarily predict the hydraulic capacity of the stem.

Despite numerous measurements on the hydraulic conductivity of stems, surprisingly few studies have examined the hydraulic conductance or conductivity of branch junctions (nodes). The available data suggest that branch junctions increase hydraulic resistance, but the magnitude of this effect is highly species specific (Ewers & Zimmermann, 1984b; Ewers & Zimmermann, 1984a; Ewers, Fisher & Chiu, 1989; Tyree & Ewers, 1991; Tyree & Alexander, 1993). In Leucadendron, most species produce new branches each year almost exclusively just below terminal inflorescences and, in the case of females, seed cones. Leaves are borne on newer branches, and so transpired water must traverse multiple nodal junctions before it reaches the leaves. Yet, because the effect of junctions on hydraulic resistance is species-specific, simply having more nodes does not necessarily translate into higher resistance.

We found no compelling evidence that hydraulic efficiency differs between conspecific males and females or even between the co-occurring species we studied here, which differ dramatically in their degrees of sexual dimorphism. At the entire shoot level, ramification had little impact on hydraulic efficiency, whether determined on a leaf area basis or on a stem cross-sectional area basis (Fig. 5). Among L. daphnoides females more highly ramified individuals had lower K_CSA, but L. rubrum males–which were the most highly ramified shoots we measured–exhibited the opposite relationship. When measured on a leaf area basis, hydraulic efficiency was lower in more highly ramified L. daphnoides females and among L. rubrum, but only when pooling males and females together. Overall, these results suggest that ramification per se has little impact on hydraulic efficiency at the level of the entire shoot. While branch junctions may add resistance to the shoot, the amount of resistance is likely small and compensated for by other adjustments, such as changes in leaf area or additional xylem area produced in basal segments. Furthermore, assuming a given length of stem growth, adding this stem as an additional branch rather than by growing the length of an existing branch would actually be beneficial because resistances are additive in series and conductances are additive in parallel. Thus, adding a stem length as a branch in parallel would add to the conductance of the entire shoot rather than subtract from it, as adding that stem length in series would. This feature of resistance networks could compensate for the resistance of adding a branch node, and other studies have suggested that trees may become more ramified as they grow taller (i.e., they add parallel conductors) in order to compensate for the added resistance a longer path length (Mencuccini & Grace, 1996). The similar water use efficiencies of Leucadendron males and females supports the equivalence of their hydraulic efficiencies (Midgley, 2010). Thus, larger stems, whether males or females, provide no more water to their transpiring leaves per unit leaf area or stem cross-sectional area than do smaller stems, in contrast to some arguments about how differences in hydraulic efficiency between males and females may be linked to sexual dimorphism in the genus (Harris & Pannell, 2010).

Males and females of each species must survive to the next fire event in order to be represented in the most recent year’s seed crop and, thus, maximize fitness (Midgley, 2000). Conspecific males and females should, therefore, have similar longevities and likely also similar rates of metabolism. This is supported by the similarity in hydraulic efficiencies among conspecifics we studied and the strong correspondence between shoot hydraulic efficiency and photosynthetic capacity (Brodribb & Feild, 2000; Brodribb, Holbrook & Gutiérrez, 2002). If the hydraulic architecture of females were more efficient than that of males, then it is not clear why males would not also have evolved a similar branching and leaf structure given that males and females co-occur and may compete with each other. Also, if the hydraulic architecture of males were less efficient, then males may incur costs associated with reproduction equal to or higher than those of females, contrary to most theory and data regarding the costs of reproduction (Bond & Maze, 1999). Instead, our data and those of Midgley (2010) are consistent with there being similar hydraulic efficiencies between species and sexes, regardless of the degrees of sexual dimorphism and branch ramification.