Dynamics of leaf hydraulic conductance with water status: quantification and analysis of species differences under steady state

Abstract

Leaf hydraulic conductance (K_leaf) is a major determinant of photosynthetic rate in well-watered and drought-stressed plants. Previous work assessed the decline of K_leaf with decreasing leaf water potential (Ψ_leaf), most typically using rehydration kinetics methods, and found that species varied in the shape of their vulnerability curve, and that hydraulic vulnerability correlated with other leaf functional traits and with drought sensitivity. These findings were tested and extended, using a new steady-state evaporative flux method under high irradiance, and the function for the vulnerability curve of each species was determined individually using maximum likelihood for 10 species varying strongly in drought tolerance. Additionally, the ability of excised leaves to recover in K_leaf with rehydration was assessed, and a new theoretical framework was developed to estimate how rehydration of measured leaves may affect estimation of hydraulic parameters. As hypothesized, species differed in their vulnerability function. Drought-tolerant species showed shallow linear declines and more negative Ψ_leaf at 80% loss of K_leaf (P₈₀), whereas drought-sensitive species showed steeper, non-linear declines, and less negative P₈₀. Across species, the maximum K_leaf was independent of hydraulic vulnerability. Recovery of K_leaf after 1 h rehydration of leaves dehydrated below their turgor loss point occurred only for four of 10 species. Across species without recovery, a more negative P₈₀ correlated with the ability to maintain K_leaf through both dehydration and rehydration. These findings indicate that resistance to K_leaf decline is important not only in maintaining open stomata during the onset of drought, but also in enabling sustained function during drought recovery.

Introduction

In dicotyledons, the leaf hydraulic conductance strongly constrains gas exchange and growth (Sack et al., 2003; Sack and Holbrook, 2006). The resistance of open stomata to vapour diffusion out of the leaf is typically far greater than the hydraulic resistance to bulk flow of the liquid through the plant, and transpiration rates are thus dictated by this diffusion process, which in turn depends on the stomatal and boundary layer conductances and the difference in vapour pressure between the intercellular air spaces of the leaf and the atmosphere (Cowan, 1972; Sack and Tyree, 2005; Sack and Holbrook, 2006). However, the maintenance of open stomata depends on the leaf being well hydrated; that is, having a high leaf water potential (Ψ_leaf), which, in turn, depends on the plant hydraulic conductance being sufficiently high. Because in dicotyledons, the leaf accounts for on average 30% of the plant hydraulic resistance (Sack et al., 2003; Sack and Holbrook, 2006), the leaf hydraulic conductance (K_leaf=flow rate/water potential driving force, i.e. 1/leaf hydraulic resistance) is thus a critical variable. Water enters the petiole, moves through several vein orders of diminishing size, then exits into the bundle sheath and moves through or around cells before evaporating into the intercellular airspace and being transpired from the stomata. The K_leaf declines with ψ_leaf during drought, due to losses of conductance resulting from cavitation and/or collapse of xylem conduits, and/or to decline in the permeability of extra-xylem tissues, and this response drives stomatal closure to prevent leaf desiccation (e.g. Salleo et al., 2000; Brodribb and Holbrook, 2004a; Sack and Holbrook, 2006; Scoffoni et al., 2008; Brodribb and Cochard, 2009; Brodribb et al., 2010; Scoffoni et al., 2011). Understanding species variation in hydraulic vulnerability is thus critical, and several techniques have been applied, especially the rehydration kinetics method (RKM; Supplementary Table S1 available at JXB online; Brodribb and Holbrook, 2003a). The aim of this study was to quantify this response using an independent, steady-state method, for species varying strongly in drought tolerance, and to determine the ability of dehydrated leaves to recover in K_leaf after rehydration.

Previous studies using the RKM found species to vary strongly in leaf hydraulic vulnerability, quantified as the Ψ_leaf at 50% loss of K_leaf (P₅₀; e.g. Hao et al., 2008; Blackman et al., 2009; Chen et al., 2009; Johnson et al., 2009a; Saha et al., 2009). Additionally, species with a low P₅₀ also had low osmotic potential at the turgor loss point (Blackman et al., 2010), and could thus maintain stomata open as leaves dehydrate. Further, these studies tested the classic trade-off between hydraulic efficiency and safety, previously found for stems, and showed this to be absent in leaves: the maximum K_leaf for hydrated leaves (K_max) was independent of P₅₀ (Sack and Holbrook, 2006; Blackman et al., 2010).

Notably, the various methods for measuring K_leaf all have value but can raise potential concerns (Sack and Tyree, 2005). There was thus a need to test leaf hydraulic vulnerability with a method independent of the RKM. The typically used RKM measures K_leaf from water uptake into the mesophyll of a dehydrated leaf for a known time, and involves some uncertainty because uptake to leaf cells continues even after leaf collection for Ψ_leaf determination, though a recently modified version of the RKM (‘dynamic RKM’) has overcome this limitation (Brodribb and Cochard, 2009; Blackman and Brodribb, 2011; Brodribb, Blackman, and PrometheusWiki contributors, 2011). Additionally, in the RKM, water uptake into mesophyll cells might not always mimic the complete pathways of natural transpiration (Scoffoni et al., 2008). Furthermore, the RKM may give low resolution of K_leaf declines in the well-hydrated range of the vulnerability curve if such leaves rehydrate completely during measurement. The evaporative flux method (EFM) has the advantage of allowing K_leaf measurement during steady-state transpiration and, further, using the EFM, leaves can be acclimated to high irradiance, which influences K_leaf for many species (Sack et al., 2002; Nardini et al., 2005; Tyree et al., 2005; Cochard et al., 2007; Sellin and Kupper, 2007; Scoffoni et al., 2008; Sellin et al., 2008). One previous study applied a variant of the EFM to generate vulnerability curves (the heat-flux method, ‘Heat-FM’; Brodribb and Holbrook, 2006) which involved some complexity. A heat gun was used on the leaf to drive a transiently high transpiration rate, after which the stomata closed, establishing a lower flow rate. The leaf was removed, and K_leaf was determined as the steady-state flow rate divided by the final Ψ_leaf (Ψ_final), and the vulnerability curve was determined as K_leaf plotted against Ψ_final. However, that method could not determine the lowest Ψ_leaf induced in the leaf during the high transpiration rates driven by the hot air (Ψ_lowest), which may have triggered the K_leaf decline. In this study, the EFM was modified to allow measurement of both Ψ_lowest and Ψ_final, such that K_leaf could be plotted against both.

A second aim of this study was to refine the statistical analysis of the K_leaf decline with dehydration for improved accuracy and mechanistic insight. Typically, studies have fitted the same function for all species, chosen for approximate fit to the data; polynomial (including linear), sigmoidal, and logistic functions have all been used (Supplementary Table S1 at JXB online). However, species may differ in the shape of their vulnerability curve, and choosing the appropriate function is important both for accuracy and also to allow interpretation of the underlying processes (Brodribb and Holbrook, 2006, 2007). Notably few studies have directly discussed the underlying basis for different shapes of vulnerability curves, probably due to the lack of an approach to select the appropriate function objectively, but the literature has pointed to several potential mechanisms for differently shaped curves (reviewed in Table 1). As a next step, a rigorous analysis is needed to resolve species differences in the shape of the function. Thus, for 10 diverse species, the maximum likelihood function was selected for each species. Drought-tolerant species were hypothesized to show shallower, linear declines, whereas drought-sensitive species were expected to show stronger initial K_leaf declines due to greater sensitivity in one or more components of the water transport system. Tests were made of the impact on estimated hydraulic vulnerability parameters of using different functions as in previous studies (Supplementary Table S1), and the degree to which it matters how vulnerability curves are plotted, namely whether unbinned data for K_leaf are plotted against Ψ_lowest or Ψ_final, or whether data are binned by Ψ_leaf intervals.

Table 1.

Mechanisms that would theoretically influence the shape of the response of leaf hydraulic conductance (K_leaf) to dehydration (i.e. decreasing leaf water potential, Ψ_leaf) and thus the function that best fitted to the data. A linear decline implies no threshold Ψ_leaf before which K_leaf declines (i.e. K_leaf declines immediately as Ψ_leaf declines), and also a proportional decline of K_leaf with Ψ_leaf. A non-linear decline of K_leaf with Ψ_leaf can include a threshold Ψ_leaf before the decline begins and/or a disproportionate decline of K_leaf as Ψ_leaf declines. For these possibilities three types of mechanisms were included—those relating to air seeding causing cavitation in the xylem conduits (and analogous effects would occur given collapse of xylem conduit walls), those arising from venation architecture, and those arising in the pathways outside the xylem. References are provided to studies of these potential mechanisms per se and/or on their influence on the shape of stem or leaf vulnerability curves.

Shape of Kleaf decline	Air seeding	Venation architecture	Pathways outside the xylem
Linear decline:
No threshold before decline	If air seeding begins at high Ψleaf because of large pit membrane pore size (Neufeld et al., 1992; Pammenter and Vander Willigen, 1998)		If a loss of membrane permeability (e.g. due to aquaporin activity or loss of cell turgor) begins immediately as Ψleaf declines (Brodribb and Holbrook, 2006)
Proportional decline of Kleaf with declining Ψleaf	If conduits of different sizes all have a wide range in maximum pit membrane pore size such that cavitation occurs equally across conduit sizes (Pammenter and Vander Willigen, 1998; Choat et al., 2005)	If higher major vein length/area (=vein density) confers hydraulic redundancy, such that first embolisms of the vein xylem conduits do not cause a dramatic decline (Scoffoni et al., 2011)	If membrane permeability declines linearly as the average cell tugor declines with Ψleaf (Kubiske and Abrams, 1990; Brodribb and Holbrook, 2006). If the Kleaf declines due to loss of water-filled pathways through cell walls (Pieruschka et al., 2010)
Non-linear decline (logistic, sigmoidal, exponential):
Threshold before decline	If a threshold for air seeding determined by the largest pit membrane pore size leads to a retention of Kleaf until a Ψleaf threshold (Neufeld et al., 1992; Pammenter and Vander Willigen, 1998; Domec et al., 2006)		If there is a threshold Ψleaf below which aquaporins are deactivated and membrane permability declines (North and Nobel, 2000; Miyazawa et al., 2008). If the Kleaf is insensitive to turgor or turgor is maintained by osmotic adjustment until a cavitation threshold is reached (Brodribb and Holbrook, 2006)
Disproportionate decline of Kleaf with declining Ψleaf	If larger conduits conferring the bulk of the vein xylem conductivity have larger pit membrane pores or greater pore numbers, and cavitate first, followed by smaller conduits that have decreasing impact on Kleaf (Neufeld et al., 1992; Pammenter and Vander Willigen, 1998; Tyree and Zimmermann, 2002)	If leaves with lower major vein density suffer strong decline in Kleaf with first embolism of xylem conduits in the low-order veins (Scoffoni et al., 2011)	If strong declines due to aquaporin deactivation occur at high Ψleaf (Johansson et al., 1998; Kim and Steudle, 2007; Scoffoni et al., 2008). If a greater loss of turgor in cells with relatively weak solute potential (e.g. bundle sheath cells) during leaf dehydration lead to especially rapid decline in Kleaf (Nonami and Schulze 1989; Koroleva et al., 1997)

A third aim in this study was to quantify the recovery of K_leaf with rehydration, a related, essential process that has received little attention. One previous study found that excised and dehydrated sunflower leaves recovered rapidly in K_leaf when rehydrated with petioles under water (Trifilo et al., 2003a). Species differences in this ability were tested for. Species with the greatest hydraulic vulnerability were hypothesized to show the greatest recovery, as they would derive most benefit. Further, all studies of vulnerability have involved leaf rehydration during measurement, but none has accounted for this in interpretation; tests were developed to determine how the measurements might be affected. The main benefit of a low hydraulic vulnerability has typically been framed as the ability to keep stomata open without dehydrating the mesophyll. It was hypothesized that a low hydraulic vulnerability would also confer the ability to maintain K_leaf through both dehydration and rehydration.

Materials and methods

Plant material

This study was conducted alongside a study of the importance of venation architecture and leaf size in determining species variation in hydraulic vulnerability (Scoffoni et al., 2011). Ten species were selected across nine families and spanning a wide range of drought sensitivity; five species were native to dry habitats (mainly California chaparral) and five species to moist habitats (Table 2). Study species included mature trees and shrubs in and around the campus of University of California, Los Angeles and Will Rogers State Park, Los Angeles, California, and sunflower Helianthus annuus var. Sunspot grown from seeds (Botanical Interests; Broomfield, Colorado, USA) in 3.6 l pots in a greenhouse (average minimum, mean, and maximum values for temperature. 21.1, 23.2, and 26.0 °C; for humidity, 44, 51, and 59%). Sunflowers were irrigated every 2 d, with 200–250 ppm of 20:20:20 N:P:K; the irradiance measured at mid-day on a sunny day was up to 550 μmol photon m⁻² s⁻¹, and on average 300 μmol photon m⁻² s⁻¹ (LI-250 light meter; LI-COR Biosciences, Lincoln, NE, USA).

Table 2.

Study species, family, native range, and mean values ±SE for pressure–volume curve parameters and leaf hydraulic vulnerability parameters, i.e. leaf hydraulic conductance at full hydration (K_max), leaf water potential at 50% and 80% decline of leaf hydraulic conductance (P₅₀ and P₈₀), calculated from the maximum likelihood function for the ‘Ψ_lowest unbinned’ plot, and results of t-tests on species’ means (for hydraulics parameters) or of analyses of variance for the difference between moist and dry area species, and among species nested within those categories (for pressure–volume parameters). Data are from Scoffoni et al. (2011).

Species	Family	Native rangea	Kmax (mmol m−2 s−1 MPa−1)	P50 (–MPa)	P80 (–MPa)	Turgor loss point (–MPa)	Osmotic potential (–MPa)	Modulus of elasticity (MPa)	Saturated water content (g g−1)
Dry habitat species
Cercocarpus betuloides	Rosaceae	California. Mexico	4.36	2.76	5.25	2.59±0.03	1.64±0.04	10.1±0.701	0.79±0.02
Comarostaphylis diversifolia	Ericaceae	California. Mexico	2.96	2.85	4.56	3.45±0.34	2.51±0.34	17.3±2.23	0.70±0.01
Hedera canariensis	Araliacaeae	Canary Islands	5.73	0.64	1.18	1.98±0.09	1.49±0.07	17.9±1.28	2.81±0.09
Heteromeles arbutifolia	Rosaceae	California. Mexico	20.7	2.57	4.12	2.53±0.10	2.08±0.10	16.4±0.486	1.38±0.07
Quercus agrifolia	Fagaceae	California. Mexico	3.96	2.40	3.83	3.00±0.12	2.31±0.12	12.8±0.787	0.93±0.01
Moist habitat species
Camellia sasanqua	Theaceae	Japan	5.99	1.78	2.84	2.12±0.18	1.61±0.04	7.98±1.11	1.74±0.03
Helianthus annuus	Asteraceae	Across N. America	6.45	0.83	1.16	1.09±0.12	0.875±0.10	13.3±1.31	11.2±0.79
Lantana camara	Verbenaceae	Pantropical	11.4	0.80	1.41	1.37±0.04	1.10±0.04	9.14±0.525	2.73±0.15
Magnolia grandiflora	Magnoliaceae	Southern USA	5.24	0.42	2.06	2.06±0.05	1.43±0.34	5.49±0.792	1.50±0.07
Platanus racemosa	Platanaceae	California, Mexico	34.1	0.09	0.35	2.03±0.06	1.54±0.12	4.85±0.331	1.34±0.03
Average ±SE	Dry habitat species		7.55±3.32	2.24±0.41	3.79±0.69	2.71±0.14	2.01±0.19	14.9±1.49	1.32 ±0.04
	Moist habitat species		12.6±5.48	0.78±0.28	1.56±0.42	1.74±0.09	1.31±0.14	8.16±1.51	3.71 ±0.21
ANOVA or t-test	Dry/moist species		NS	*	*	******	******	******	******

^aCroat (1978); Kitamura and Murata (1979); eFloras (2008).

NS, P >0.05; *P <0.025;***P <0.001.

Experiments were conducted in May–September 2008. On the day prior to measurements, for 3–10 plants per species, exposed branches with mature, healthy leaves were collected into plastic bags with moist paper towel; for sunflowers, whole shoots were collected. Each shoot was re-cut by at least two nodes in the laboratory under ultrapure water (MilliPore, 0.22 μm Thornton 200CR, Molshem, France) and rehydrated overnight at laboratory temperature (20–25 °C), covered with dark plastic bags.

Measuring the dehydration response of K_leaf with the evaporative flux method

Using the EFM, K_leaf is determined as the ratio of steady-state transpirational flow rate (E, mmol m⁻² s⁻¹) to the water potential driving force (ΔΨ_leaf, MPa; Sack et al., 2002). Notably, in this system, the overall driving force for flow through the whole leaf is the water potential gradient between the outside air and the water entering the petiole, but the important component of that driving force is the vapour pressure gradient between the outside air and leaf air spaces; this vapour pressure driving force, and stomatal conductance, determine the transpiration rate (see Introduction). However, for the liquid phase part of flow (i.e. the hydraulic system), the driving force at steady state is the water potential gradient between the leaf mesophyll where water evaporates (estimated as the Ψ_leaf measured at the end of the measurement, i.e. the Ψ_final) and the water entering the petiole at atmospheric pressure (i.e. 0 MPa relative pressure).

In this study, the focus was on the dehydration response of the whole-leaf hydraulic system, including the petiole. The leaf was cut from the shoot with a fresh razor blade under ultrapure water that was used as flow solution (0.22 μm Thornton 200 CR; MilliPore), degassed for at least 8 h with a vacuum pump (Gast, Benton Harbor, MI, USA), and refiltered (0.2 μm; Syringe filter, Cole-Parmer, Vernon Hills, IL.USA). The petiole was then rapidly connected to silicon tubing (Cole-Parmer) under ultrapure water to prevent air entering the system. The tubing connected the leaf to a cylinder on a balance (models XS205 and AB265, ±10 μg sensitivity; Mettler Toledo, Columbus, OH, USA) that logged data every 30 s to a computer for the calculation of flow rate through the leaf (E). Leaves were held adaxial surface upwards in wooden frames strung with fishing line above a large box fan (Lakewood Engineering & Manufacturing Company, Chicago, IL, USA). Leaves were illuminated with >1000 mmol m⁻² s⁻¹ photosynthetically active radiation at the leaf surface by floodlights (model 73828 1000 W, ‘UV filter’; Sears, Roebuck, Hoffman Estates, IL, USA) suspended above a Pyrex container (Corning Incorporated, Corning, NY, USA) filled with water to absorb the heat of the lamp. Leaf temperature was determined using a thermocouple (Cole-Parmer) and maintained between 23 °C and 28 °C.

Leaves were allowed to transpire on the apparatus for at least 30 min and until the flow rate stabilized, with no upward or downward trend, and with a coefficient of variation <5% for at least five measurements made at 30 s flow intervals. When the flow rate was very low (<8 μg s⁻¹), stability was determined with the same criterion, but using the running averages of the last five 30 s intervals. Previous studies found these criteria to be sufficient for stabilization of E, Ψ_leaf, and K_leaf; tests with longer measurement periods after stable flow was established showed no relationship of K_leaf to measurement time for seven species of a wide range of leaf capacitance (Scoffoni et al., 2008; Pasquet-Kok et al., 2010). The minimum 30 min flow period was chosen to ensure that leaves had sufficient time to acclimate to high irradiance, which has been found to enhance K_leaf by up to 8-fold depending on species, apparently due to the expression and/or activation of aquaporins (Sack et al., 2002; Nardini et al., 2005; Tyree et al., 2005; Cochard et al., 2007; Scoffoni et al., 2008; Voicu et al., 2008). Measurements were discarded if the flow suddenly changed, due either to apparent leakage from the seal or to blockage in the system by particles or air bubbles. Following the stabilization of the flow rate, leaf temperature was recorded with a thermocouple and the final five flow rate measurements were averaged. The leaf was quickly removed from the tubing, the petiole was dabbed dry, and the leaf was placed into a sealable bag (Whirl-Pak; Nasco, Fort Atkinson, WI, USA), which had been previously exhaled in, to halt transpiration. Following at least 30 min equilibration, the final leaf water potential (Ψ_final) was measured with a pressure chamber (Plant Moisture Stress, Model 1000, Albany, OR, USA). K_leaf was calculated as E/–ΔΨ_leaf (where ΔΨ_leaf=Ψ_final–0 MPa) and further normalized by leaf area measured with a LI-COR 3100 leaf area meter. To correct for changes in K_leaf induced by the temperature dependence of water viscosity, K_leaf values were standardized to 25 °C (Weast, 1974; Yang and Tyree, 1993; Sack et al., 2002).

To determine the stomatal conductance of leaves measured with the EFM, the final E was divided by the mole fraction vapour pressure deficit (VPD), derived from temperature and relative humidity (RH) measurements in the lab from a weather station that logged measurements every 5 min (HOBO Micro Station with Smart Sensors; Onset, Bourne, MA, USA), where mole fraction VPD=[1–(RH×VP_sat)]/101.3 kPa, and VP_sat is the saturation vapour pressure determined using the Arden Buck equation (Buck, 1981).

The EFM was modified to allow determination of K_leaf for dehydrated leaves. Shoots were cut into segments with at least three leaves under ultrapure water and then dehydrated with a fan for different periods of time to a range of Ψ_leaf values. The bench drying of shoots to achieve a leaf vulnerability curve has been used in studies using the RKM (e.g. Brodribb and Holbrook, 2003a; Blackman et al., 2009), and previous studies found similar vulnerability curves when constructed from bench-drying shoots as from leaves on plants progressively droughted (Brodribb and Holbrook, 2004a; Blackman et al., 2009; Pasquet-Kok et al., 2010). In the present study, shoots were allowed to equilibrate for at least 30 min before two leaves were excised and measured for initial Ψ_leaf (Ψ_o) using a pressure chamber. If the difference in the Ψ_leaf of those two leaves was >0.1 MPa, the shoot was discarded; for very dehydrated shoots, this range was extended to 0.3 MPa. The third leaf (typically the middle leaf) was used to determine K_leaf with the EFM. When dehydrated leaves are measured with the EFM, the stomata open (see Results); before steady-state flow is achieved, the leaf may rehydrate such that Ψ_final is less negative than Ψ_o, or, alternatively, the leaf may further dehydrate such that Ψ_final is more negative than Ψ_o. For each species, at least six K_leaf values were obtained for each 0.5 MPa interval from full hydration to strong dehydration. Outlier tests were conducted for each 0.5 MPa interval (Dixon test; Sokal and Rohlf, 1995); 0–4 outliers were removed over the whole curve for given species (representing 0–8% of the 26–74 data points per curve).

To test the importance of the method for constructing vulnerability curves, these were determined in three ways previously applied (Supplementary Table S1 at JXB online). First, K_leaf was plotted against whichever was lowest, Ψ_o or Ψ_final (=‘Ψ_lowest’); that is, the Ψ_leaf associated with the strongest dehydration experienced during the experiment, and each leaf was considered as a data point (‘unbinned Ψ_lowest’). Additionally, K_leaf was plotted against Ψ_lowest with data averaged in 0.5 MPa bins (‘binned Ψ_lowest’), with the exception of H. annuus averaged in 0.2 MPa bins because of its distinctively narrower K_leaf response, with negligible values below –1.5 MPa. Finally, K_leaf was plotted against Ψ_final rather than Ψ_lowest (‘Ψ_final’), with each leaf considered as a data point. Determination of these alternative versions of the vulnerability curve also allowed interpretation of the recovery of K_leaf during the measurement (see section below).

In the above-described methods, as in previous studies of K_leaf, the pressure chamber balance pressure was taken as Ψ_leaf. In actuality, the balance pressure for an equilibrated leaf gives the xylem pressure potential (P_x), and –P_x is less negative than the bulk Ψ_leaf by the amount of the vein xylem solute potential (π_x; Tyree and Zimmermann, 2002). Notably, previous studies on a range of species have measured π_x values of approximately –0.05 MPa, a difference that would not affect the present findings significantly (Boyer, 1967). Tests were carried out to verify such low π_x for C. sasanqua, H. arbutifolia, and L. camara. Shoots of four leaves were rehydrated overnight and dehydrated to a range of Ψ_leaf (–0.04 MPa to –1.5 MPa). Two leaves were excised for initial Ψ_leaf measurement, a third was bagged for determination of initial π_x, and the fourth was placed in the EFM apparatus until a steady-state flow rate was achieved. Leaf vein π_x was determined using vapour pressure osmometry (Vapro 5520, Wescor Inc., Logan, UT, USA). The leaf margin was excised to open the tips of the midrib and second-order veins, and the leaf was pressurized in the pressure chamber and xylem sap exuded from the petiole was collected onto a filter paper, while moist paper towels surrounded the chamber and petiole to minimize evaporation. The filter paper was transported to the osmometer in a weighing bottle filled with moist paper towel. All π_x values were less negative than the least negative measurable value with this instrument, –0.05 MPa, and thus indistinguishable from pure water in the instrument, indicating that the present findings would not be significantly impacted by π_x.

Model testing and estimation of parameters for the decline of K_leaf with dehydration

Maximum likelihood was used to select the function for each species’ K_leaf vulnerability response (Burnham and Anderson, 2002), using the optim function in R 2.9.2 (http://www.r-project.org; Burnham and Anderson, 2004; Sack et al., 2006; the scripts are available on request). A linear function (K_leaf=aΨ_leaf+y_o), was tested, in addition to sigmoidal and logistic functions , as used previously in the literature on leaf vulnerability (Supplementary Table S1 at JXB online), and an exponential function ($K_{\text{leaf}}=y_0+a{e}^{-b{\Psi }_{\text{leaf}}}$), as previously used for whole-plant vulnerability (Iovi et al., 2009). The maximum likelihood parameters were determined by the Simulated Annealing procedure for global optimization, followed by the Nelder–Mead simplex procedure for local optimization; standard errors for parameters were generated from the Hessian matrix. For each data set, functions were compared using the Akaike Information Criterion (AIC), corrected for low n. The function with the lowest AIC value was chosen as the best fit function for that data set, with differences >2 considered as meaningful (Burnham and Anderson, 2002, 2004).

To compare species in their hydraulic parameters, and to determine correlations between hydraulic parameters and other leaf traits, values for the maximum K_leaf at full hydration (K_max) and the Ψ_leaf at which K_leaf had declined by 50% and 80% (P₅₀ and P₈₀) were determined from the vulnerability curves. For these parameters, each species’ maximum likelihood function was used [i.e. that with lowest AIC and highest r² determined from the unbinned data plots (‘unbinned Ψ_lowest’ and ‘Ψ_final’)]. The steepness of the vulnerability curve was also determined, as the first derivative of the maximum likelihood function at Ψ_leaf= –0.5 MPa, where the steepest declines were observed. As an additional method for determining K_max, the average K_leaf for points above –0.5 MPa was calculated for each species; this was the method used in most previous leaf hydraulics studies that measured only K_leaf for hydrated leaves, and not its vulnerability to dehydration (e.g. Sack et al., 2002; Brodribb and Holbrook, 2003b; Nardini et al., 2005).

To determine the degree to which the choice of function and data set matters, tests were made of the sensitivity of vulnerability curve parameters (K_max, P₅₀, and P₈₀) to the choice of function, and, for each function, of plotting K_leaf against ‘unbinned Ψ_lowest’, ‘binned Ψ_lowest’, or ‘Ψ_final’.

Hydraulic safety margins were calculated as the difference between the Ψ_leaf at which the leaves of a given species lose turgor (π_TLP; data from Scoffoni et al., 2011) and those at which hydraulic function was substantially lost (P₅₀ or P₈₀). Positive numbers indicate a safety margin, whereas negative numbers indicate a loss of hydraulic function even above the turgor loss point.

Testing the recovery of leaf hydraulic conductance after dehydration

Experiments were performed to test the recovery of K_leaf for leaves rehydrated after dehydration (method after Trifilo et al., 2003a). For the 10 species, shoots were dehydrated with a fan to a known Ψ_leaf below their respective turgor loss points (determined as described in the following section). Leaves from each shoot were excised in air using a fresh razor blade and measured for Ψ_leaf (Ψ_dehydration), and other leaves were excised under ultrapure water, and rehydrated for 1 h with their petioles under water in a beaker, covered with a dark plastic bag. Following rehydration, leaves were equilibrated in a plastic bag for at least 10 min and either had their petioles cut in air and were measured for Ψ_leaf (Ψ_rehydration), or had their petioles re-cut under ultrapure water and were immediately connected to the EFM to determine K_leaf (n= 4–12 per species). The percentage recovery of K_leaf was determined as the K_leaf after rehydration divided by the K_leaf at Ψ_dehydration, which was estimated from the species’ maximum likelihood vulnerability curve ×100%. The recovery was considered significant if the K_leaf after 1 h rehydration was greater than the K_leaf at Ψ_dehydration (t-test; Minitab Release 15). The recovery was determined as complete if K_leaf after 1 h rehydration was not significantly lower than the K_leaf at Ψ_rehydration which was estimated from the species’ maximum likelihood vulnerability curve.

Testing for the recovery of leaf hydraulic conductance during EFM measurement

As in other methods for determining leaf hydraulic vulnerability (i.e. RKM and Heat-FM; see Introduction), the EFM partially rehydrates the dehydrated leaf, as the petiole is connected to water at atmospheric pressure. Two analyses were developed to test for the potential recovery of K_leaf during the EFM measurement. The first analysis was a test of residual variation. If K_leaf recovered completely during the EFM measurement, one would expect no influence of the dehydration treatment prior to measurement on the final K_leaf value; rather, the measured K_leaf would simply relate to Ψ_final (i.e. the leaf water potential during the final steady-state flow). Thus, for each species, from the maximum likelihood vulnerability curve for the ‘Ψ_final’ plot, the residuals of K_leaf against Ψ_final were calculated. These residuals represented the variation in K_leaf unrelated to Ψ_final. A test was made for correlation of these residuals with Ψ_lowest values (Minitab Release 15). If the residual K_leaf variation was negatively correlated with Ψ_lowest, there was a persistent impact of Ψ_lowest on K_leaf, independently of Ψ_final. In other words, the effect of the dehydration treatment persisted even at the end of the EFM measurement, and, thus, the K_leaf had not recovered completely during the measurement. The second analysis was the calculation of an index of the recoverability of K_leaf during the EFM. For each species, a sample of the vulnerability data was selected that was analogous to the 1 h rehydration experiment (see previous section). Data were selected for leaves that had been dehydrated to a Ψ_leaf below the turgor loss point but that had rehydrated during the EFM measurement to Ψ_leaf values similar to those for leaves measured by the EFM after the 1 h rehydration experiment (n=4–7 for each species). The percentage recovery of K_leaf during EFM was determined as the average measured K_leaf for this leaf sample divided by the K_leaf at Ψ_dehydration, which was estimated from the species’ maximum likelihood vulnerability curve ×100%. The significance of the recovery of K_leaf was tested as for leaves in the 1 h rehydration experiment.

Given that some species showed a partial recovery of K_leaf with rehydration during the EFM (see Results), a theoretical consideration was made of how K_leaf recovery during measurement should influence the calculation of vulnerability parameters. Based on the diversity of tissues in the leaf hydraulic pathway, the vulnerability of K_leaf is expected to involve several components, some of which might be recoverable on a short time scale, while others might be reversible after a longer time scale under low tension (cf. Brodribb and Holbrook, 2006; Scoffoni et al., 2008). The most appropriate vulnerability plot would depend on the degree to which leaves are recoverable in the short term (Fig. 1). Bounding cases were considered in which (a) leaves were non-recoverable in K_leaf during the measurement; (b) leaves were totally recoverable; and (c) leaves were partially recoverable. In case (a) in which K_leaf is non-recoverable, an accurate vulnerability curve would be obtained by plotting K_leaf against Ψ_lowest, as only the minimum Ψ_leaf during the whole experiment is important for influencing K_leaf. In case (a), plotting K_leaf against Ψ_final would overestimate the leaf’s vulnerability. In contrast, in case (b) in which K_leaf recovers completely during measurement, an accurate vulnerability curve would be determined by plotting K_leaf against Ψ_final, because only the Ψ_leaf during steady state at the end of the measurement is important for influencing K_leaf. In case (b), plotting K_leaf against Ψ_lowest would underestimate the leaf’s vulnerability. Finally, in case (c), in which K_leaf is partially recoverable, the accurate vulnerability curve would be intermediate between the plots of K_leaf against Ψ_final and against Ψ_lowest. Additional scenarios were not considered, for example if leaves recover in K_leaf differently depending on their degree of dehydration; notably, such scenarios should fall within the bounding cases considered. Tests were conducted to determine whether the estimation of vulnerability parameters K_max, P₅₀, and P₈₀ was improved by using for each species the plot appropriate to its K_leaf recovery. Thus, for the species that showed no K_leaf recovery during EFM measurement, parameters were re-calculated from the maximum likelihood function for the ‘Ψ_lowest unbinned’ plot and, for the species with partial recovery, parameters were averaged from those determined from the ‘Ψ_final’ and ‘Ψ_lowest unbinned’ plots. These re-calculated parameters were compared with those determined using the ‘Ψ_lowest unbinned’ for all species, as has been the most typical procedure in previous studies (Supplementary Table S1 at JXB online).

Fig. 1.

A theoretical framework for the construction of vulnerability curves according to the degree that leaves recover in leaf hydraulic conductance (K_leaf) with rehydration. The black line is the ‘true’ vulnerability curve, the grey line is the vulnerability curve plotting K_leaf against ‘Ψ_lowest’, and the grey dotted line is the vulnerability curve plotting K_leaf against ‘Ψ_final’. Bounding cases were considered: (a) leaves were non-recoverable in their K_leaf during the measurement; (b) leaves were totally recoverable in their K_leaf; and (c) leaves were partially recoverable in their K_leaf (see the Materials and methods).

Statistical analysis of differences among species and trait correlations across species

Trait differences between moist and dry habitat species were tested using analyses of variance (ANOVAs) with species nested within habitat type, or by using t-tests on species means (Table 1; Minitab Release 15). All data were log-transformed to improve normality and heteroscedasticity (Sokal and Rohlf, 1995). Correlations among traits were considered significant only if P <0.05 for both Spearman and Pearson coefficients (r_s and r_p, respectively); when relationships were non-linear, correlations for log-transformed data were determined. Standard major axes were fitted when determining slopes of relationships between traits, to account for error in both x and y variables (using SMATR; Sokal and Rohlf, 1995; Warton et al., 2006).

Results

Vulnerability curves: species differences in the response of K_leaf to dehydration

The EFM was effective for determining vulnerability curves for leaves dehydrated from near full turgor to beyond the turgor loss point (Fig. 2). Leaves that had been previously dehydrated opened their stomata and established steady-state transpiration during the EFM measurement, as indicated by even the lowest transpiration rates observed representing stomatal conductance values 2.2- to 7.3-fold higher than cuticular conductance for these species (Supplementary Table S2 at JXB online).

Fig. 2.

Vulnerability curves for leaf hydraulic conductance (K_leaf) for 10 species varying widely in drought tolerance, determined using the evaporative flux method using three different plots (‘Ψ_lowest unbinned’, ‘Ψ_lowest binned’, and ‘Ψ_final’). For the ‘Ψ_lowest unbinned’ and ‘Ψ_final’ panels, each point represents a different leaf measured. Standard errors are represented for each bin point in the ‘Ψ_lowest binned’ plot. The lines plotted are the maximum likelihood functions using each plot for each species (Supplementary Table S3 at JXB online).

Species differed significantly in the shape of the leaf hydraulic vulnerability curves. For four species the linear function was selected by maximum likelihood for K_leaf plotted against ‘Ψ_lowest unbinned’, and for six species a non-linear function was selected (Fig. 2; Supplementary Table S3 at JXB online). The logistic function was selected for five species and the sigmoidal for C. betuloides. Species from dry habitats had a greater tendency to show a linear decline in K_leaf as one of their selected functions, that is, within an AIC of 2 of the maximum likelihood function (4/5 species versus 1/5 for moist habitats; P=0.018; proportion test). The slope of the vulnerability curve at Ψ_leaf= –0.5 MPa varied from –10 mmol m⁻² s⁻¹ MPa⁻² to –0.5 mmol m⁻² s⁻¹ MPa⁻², and drought-sensitive species had on average 3-fold steeper slopes than drought-tolerant species (–6.5 mmol m⁻² s⁻¹ MPa⁻² versus –1.6 mmol m⁻² s⁻¹ MPa⁻², respectively; t-test; P=0.009, n=5).

Vulnerability curves: sensitivity of derived parameters to the choice of function and plot

The use of maximum likelihood to select the vulnerability function for each species based on plots of K_leaf against ‘Ψ_lowest unbinned’ was considered to be the most appropriate practice, and was the one used for interpretation and comparison among species. However, because many previous studies have applied a single function and plot to all species’ data, the sensitivity of the derived vulnerability parameters to the choice of function and plot and whether such choices affected the resolution of species ranking in vulnerability was tested. Notably, the functions selected by maximum likelihood with AIC values within 2 of the minimum depended on the choice of plot, and multiple functions were often selected for given species (Fig. 2, Supplementary Table S3 at JXB online). Thus, when using the ‘Ψ_lowest binned’ plot, the linear function was selected for 8/10 species, the logistic for two, and the exponential for one species. In contrast, when using the ‘Ψ_lowest unbinned’ plot, the logistic function was selected for eight species, the sigmoidal for six, the linear for five, and the exponential for four. When using the ‘Ψ_final’ plot, the logistic was selected for nine species, the exponential for eight, the sigmoidal for five, and the linear for two. The best fit function selected using the ‘Ψ_lowest unbinned’ plot was one of those selected when using the ‘Ψ_lowest binned’ data set for 5/10 species, and when using the ‘Ψ_final’ plot for only 3/10 species.

The estimation of vulnerability parameters K_max, P₅₀, and P₈₀, was sensitive to the function and the plot used, but typically the values determined in different ways were correlated across species (Fig. 2, data in Supplementary Table S4 at JXB online). When using the ‘Ψ_lowest unbinned’ plot, the K_max, P₅₀, and P₈₀ values generated by the four different functions, averaged across species, varied by 12–27%, 0.21–0.76 MPa, and 0.12–0.74 MPa, respectively, and correlated across species in 15/18 comparisons (r_p=0.81–0.99; P <0.05). The use of the three plots produced K_max values from the four given functions that varied on average by 3–40%, and correlated across species in 11/12 comparisons (r_p=0.64–0.99; P <0.05; Fig. 2). Notably, for a species such as Platanus, with a steep initial hydraulic decline, determining K_max from a ‘Ψ_lowest unbinned’ plot was critical to resolve its high K_max. For P₅₀ and P₈₀, the use of the ‘Ψ_lowest unbinned’ and ‘Ψ_lowest binned’ plots produced values for given functions that differed on average by 0.08–0.6 MPa, and correlated across species in 7/8 comparisons (r_p=0.56–0.99; P <0.05). In contrast, the use of the ‘Ψ_final’ plot produced P₅₀ and P₈₀ values 0.8–2 MPa less negative than when using the other plots (Fig. 2; paired t-test; P <0.05), and values were not correlated across species (r_p= –0.40 to 0.48; P=0.11–0.81). The values of K_max determined using the function selected using the ‘Ψ_lowest unbinned’ or ‘Ψ_lowest binned’ plots did not differ on average across species from those determined by taking the mean of K_leaf values at Ψ_leaf of 0 to –0.5 MPa (data in Supplementray Table S5 at JXB online; P=0.10–0.15; paired t-test). However, K_max determined using the ‘Ψ_final’ plot was on average 44% higher than K_max determined by taking the mean of K_leaf values at Ψ_leaf of 0 to –0.5 MPa (P=0.02), but again the species’ values with the two methods were correlated (r_p=0.74; P <0.01).

Species variation in maximum K_leaf and vulnerability, and lack of an efficiency–safety trade-off

Species were compared in the parameters determined from their maximum likelihood functions using the ‘Ψ_lowest unbinned’ plot (Table 2). Species differed by >11-fold in K_max, with no average differences between species from moist and dry habitats, though species differences were significant (considering K_max as the mean of K_leaf values at Ψ_leaf of 0 to –0.5 MPa; ANOVA; P <0.001). Species also differed greatly in their vulnerabilities, varying 32-fold in P₅₀ and 15-fold in P₈₀, from the most vulnerable species (H. annuus and P. racemosa) with values less than –1 MPa to the least vulnerable species, C. diversifolia, with P₅₀ and P₈₀ values of –3.54 MPa and –5.25 MPa, respectively (Fig. 2). Species’ P₅₀ and P₈₀ values were strongly correlated (r_p and r_s=0.88–0.96, P <0.01). Species with greater vulnerability (i.e. with less negative P₅₀ and P₈₀ values) had steeper vulnerability curve slopes (r_p and r_s= –0.72 to –0.83, P <0.01; data in Supplementary Table S5 at JXB online). On average, species from dry habitats had 2.4- to 2.9-fold more negative P₅₀ and P₈₀ than species from moist habitats.

Species with lower vulnerability had greater hydraulic safety margins. Thus, safety margins based on P₅₀ were negatively correlated with P₅₀, and safety margins based on P₈₀ were negatively correlated with both P₅₀ and P₈₀ (r_p and r_s= –0.70 to –0.95; P <0.05; data in Supplementary Table S5 at JXB online). Safety margins based on P₅₀ ranged from –1.9 MPa to 0.17 MPa and were positive for two species (C. diversifolia and H. arbutifolia); thus, most species lost leaf turgor at lower Ψ_leaf than P₅₀ as determined using the steady-state method. However, safety margins calculated from P₈₀ ranged from –1.7 MPa to 2.7 MPa, and seven species had positive safety margins. Safety margins did not differ between habitat types (t-test, P <0.05).

Both P₅₀ and P₈₀ were independent of K_max across species (|r_p| and |r_s|=0.37–0.62, P >0.05).

Recovery of K_leaf with leaf rehydration and a new importance for leaf hydraulic vulnerability

Species varied strongly in the ability to recover in K_leaf after dehydration below their turgor loss point (such that K_leaf declined by 57–97% depending on species) followed by 1 h rehydration with their petioles under water (Fig. 3). For four species (C. diversifolia, H. annuus, L. camara, and M. grandiflora), K_leaf increased 2.2- to 2.8-fold (Fig. 3; P <0.05); C. diversifolia and M. grandiflora recovered fully in K_leaf to their expected values. Three of these species were moist habitat species (L. camara, H. annuus, and M. grandiflora) and one was a dry habitat species (C. diversifolia). The six other species showed no significant recovery. The percentage recovery of K_leaf after rehydration did not correlate with K_max, P₅₀, or P₈₀ (P >0.05; data in Supplementary Table S5 at JXB online).

Fig. 3.

Recovery of leaf hydraulic conductance (K_leaf) after 1 h rehydration with their petioles under water, for 10 species varying widely in drought tolerance. The grey curves are the best-fit functions of the species’ response to dehydration from Fig. 2; open and filled symbols represent the predicted K_leaf at the dehydrated leaf water potential, and K_leaf after 1h rehydration respectively; stars on the x-axis represent the turgor loss point. Species depicted in the upper four panels showed significant recovery in K_leaf (*P=0.04; **P=0.001; ***P <0.001); only C. diversifolia and M. grandiflora showed total recovery. Species depicted in the lower panels showed no significant recovery in K_leaf.

For the six species that did not recover in K_leaf with 1 h rehydration, a nearly perfect correlation was found of the ability to maintain K_leaf after dehydration and rehydration episodes and low P₅₀ and P₈₀ (r_s and r_p= –0.94 to –0.98; P <0.005; Fig. 4). Thus, among the species that did not recover in K_leaf with rehydration, a low vulnerability predicted the ability to retain hydraulic capacity despite strong, short-term dynamics in water status.

Fig. 4.

The ability of hydraulic vulnerability to predict the degree that leaf hydraulic conductance (K_leaf) was maintained after strong dehydration and rehydration for 1 h with petiole in water, calculated as K_leaf after rehydration divided by maximum K_leaf (K_max). Filled circles represent species without recovery of K_leaf and open circles species that did show recovery of K_leaf. The line was fitted only for species without recovery of K_leaf (**P=0.005; ***P <0.001).

Testing K_leaf recovery during EFM, and its impact on the estimation of hydraulic parameters

No species showed full recovery in K_leaf of dehydrated leaves during EFM measurements; for all species there was a persistent impact of dehydration. When the residuals of K_leaf against Ψ_final were plotted against Ψ_lowest (see the Materials and methods), this correlation was significant for seven species (r_p= –0.49 to –0.79; n=25–74; P <0.05; Table 3). For the other three species (C. sasanqua, H. annuus, and H. canariensis) the lack of significant correlation of residuals with Ψ_lowest did not imply a complete recovery of K_leaf during EFM measurement. In the case of H. annuus and H. canariensis, the Ψ_lowest values were typically the Ψ_final values because the leaves dehydrated further during measurement, rather than recovering in Ψ_leaf, and, in the case of C. sasanqua, because the Ψ_lowest correlated with Ψ_final (r_p=0.53; P <0.001) there may not have been sufficient residual variation for a powerful test. There were broadly consistent results in the second analysis of the recovery of K_leaf during the EFM measurement (i.e. the calculation of the percentage recovery of K_leaf for leaves that rehydrated over the same Ψ_leaf interval as the 1 h rehydration experiment). Again there was no evidence for total recovery of K_leaf. There was a significant partial recovery of K_leaf in 3/10 species (P <0.007; Table 3), with K_leaf increasing by 158–178%. Across species, the recovery of K_leaf during the EFM was positively correlated with that observed after 1 h rehydration (r_p and r_s=0.83–0.84; P <0.05). The percentage recovery of K_leaf during the EFM was 13% lower on average than that after 1 h rehydration, consistent with the leaf rehydrating for a shorter period of time, under subatmospheric pressure (paired t-test; P=0.04).

Table 3.

Results from the tests of the recovery of leaf hydraulic conductance (K_leaf) during the evaporative flux method (EFM), and during 1 h rehydration in the dark. In the residual test for recovery during the EFM, significance indicates that K_leaf did not fully recover. For the indices of K_leaf recovery during the EFM, and during the 1 h rehydration experiments, significance before the comma indicates some degree of significant recovery, and significance after the comma indicates that K_leaf did not recover fully (see the Materials and methods).

Species	Residual test for recovery during EFM, R2 (n)	Index of recovery in Kleaf during EFM (% increase)	Index of recovery in Kleaf after 1 h rehydration (% increase)
Camellia sasanqua	0.029NS (41)	114NS, **	58.9NS, ***
Cercocarpus betuloides	0.48*** (70)	119NS, **	119NS, **
Comarostaphylos diversifolia	0.33*** (57)	178**, *	259***, NS
Hedera canariensis	0.036NS (41)	159**, **	150NS, ***
Helianthus annuus	0.017NS (36)	124NS, *	230**, *
Heteromeles arbutifolia	0.62*** (58)	66.4NS, *	79.3NS, *
Lantana camara	0.61*** (25)	161NS, **	284**, **
Magnolia grandiflora	0.24* (74)	158**, ***	218*, NS
Platanus racemosa	0.35*** (38)	104NS, *	130NS, *
Quercus agrifolia	0.38*** (46)	72.2NS, **	113NS, ***

^*P <0.05; **P <0.01; ***P <0.001. NS, non-significant

Given that three species indicated partial K_leaf recovery during EFM measurement, an analysis was made of its potential influence on derived vulnerability parameters (see the Materials and methods). Re-calculating these species’ K_max, P₅₀, and P₈₀ values while considering the partial K_leaf recovery produced values that were correlated with those determined using both the ‘Ψ_lowest unbinned’ and ‘Ψ_lowest binned’ plots (r_s and r_p=0.57–0.99; P <0.001–0.09), indicating that species comparisons using those vulnerability plots are robust even despite partial K_leaf recovery. However, the re-calculated parameters accounting for partial recovery did not correlate with those determined using the ‘Ψ_final’ plot (r_s and r_p=0.08–0.30; P=0.16–0.83).

Discussion

The new steady-state EFM developed for determining the hydraulic vulnerability of leaves acclimated to high irradiance allowed an independent confirmation and extension of key relationships first shown using rehydration methods. Additionally, refined statistical methods for analysing vulnerability data allowed fitting of the appropriate function for each species and considering the effect of recovery during the measurement. These approaches showed novel variation among species in leaf vulnerability, and relationships with species’ habitat. Further, rehydration experiments quantifying the rapid recovery of K_leaf after dehydration indicated novel species variation, and a new role for leaf vulnerability in determining function after episodes of dehydration and rehydration. This work provided new insights into the vulnerability response, and will additionally enable higher resolution in future work investigating the underlying mechanisms for leaf hydraulic vulnerability.

Species’ differences in K_leaf decline and potential mechanisms

Species differed strikingly in their vulnerability parameters P₅₀ and P₈₀, and in the shape of their vulnerability curves. Notably, because species varied strongly in initial K_leaf values (K_max) and in the steepness of their decline in K_leaf, P₈₀ was useful to allow comparison of species’ vulnerabilities at a similar stage of their trajectory, namely after the steepest decline phase (Fig. 2), whereas P₅₀ values often occurred in the middle of the steepest decline, which for some species occurred at very high Ψ_leaf. For such species the P₅₀ may not be an effective index of drought resistance. Further, it is noted that several species (e.g. P. racemosa) had very high K_max, with substantial K_leaf decline before Ψ_leaf reached –0.5 MPa. Though part of the true range of leaf hydraulic behaviour in such species, such very high K_leaf values are outside of the range found in nature, as they would not occur for leaves transpiring in vivo, in which the soil and plant hydraulic resistance would cause a further Ψ_leaf drop not experienced by leaves in the EFM. Species with such steep, non-linear decline were typical of moist habitat species, whereas species with shallow, linear declines were associated with dry habitats.

The K_leaf decline during dehydration arises due to loss of hydraulic conductance in the petiole and/or vein xylem, and/or the extra-xylem pathways (Table 1). The importance of (i) cavitation due to air seeding in major veins leading to subsequent embolism was supported by studies showing ultra-acoustic emissions that may reflect cavitation events (Kikuta et al., 1997; Salleo et al., 2000; Johnson et al., 2009a), as well as dye and cryo-scanning electron microscope studies showing embolism in vein xylem (Salleo et al., 2001; Nardini et al., 2003, 2008; Johnson et al., 2009a), measurement of relatively low air seeding pressures in the leaf petiole and midrib (Choat et al., 2005), and a correlation across species of hydraulic vulnerability with low major vein length per leaf area, as such leaves have less xylem redundancy to protect from the impact of embolism (Scoffoni et al., 2011). Another mechanism may be (ii) the collapse of xylem conduits in the leaf veins; indeed xylem cell collapse has been found for tracheids in the vein of pine needles and in the transfusion tissue of a tropical conifer, at Ψ_leaf values as high as –1.5 MPa, in advance of cavitation (Cochard et al., 2004; Brodribb and Holbrook, 2005). Indeed, xylem cell collapse has been hypothesized to occur in the minor vein xylem in angiosperms too, but has not yet been visualized directly (Blackman et al., 2010). Additionally, K_leaf decline might relate to (iii) the loss of turgor in living cells in the extra-xylem flow pathways (Brodribb and Holbrook, 2006), in particular the cells of the bundle sheath, mesophyll, and epidermis, which may shrink with walls retracting, and/or may undergo plasmolysis. Tissues with low solute potential, such as bundle sheath, might lose turgor in advance of the mesophyll (Giles et al., 1974; Palta and Leestadelmann, 1983; Nonami and Schulze, 1989; Canny and Huang, 2006). Such changes in cell volume and turgor may alter the flow pathways, and additionally reduce membrane permeability, for example via deactivation of aquaporins (Kim and Steudle, 2007). A final mechanism for the K_leaf decline especially in well-hydrated leaves is (iv) the evaporation of liquid water in the cells walls during transpiration, leaving walls moist but with empty pores and thus lower permeability (Kim and Steudle, 2007; Lee et al., 2009; Voicu et al., 2009).

The shapes of functions fitted to K_leaf data from the present study using maximum likelihood provide several key insights and hypotheses for the action of these mechanisms and point to a diversity in specific impacts across species. Given that embolism or collapse of vein xylem conduits is a principal driver of the K_leaf decline, the linear decline observed for four species implies that air seeding or collapse begins at high Ψ_leaf for these species (see references in Table 1). The linear decline also implies that conduits of different sizes tend to have approximately equal distributions of air seeding pressures and tendencies to collapse, and/or that a high major vein density provides redundancy that protects the leaf from a disproportionate effect of cavitation of the major vein xylem. A linear decline of K_leaf would also be consistent with a direct role for loss in mesophyll, epidermis, or bundle sheath cell volume or turgor, or the number of water pathways through cell walls declining approximately linearly with Ψ_leaf above the turgor loss point (Table 1). The logistic decline observed in five species and sigmoidal decline in C. betuloides indicate a qualitative difference. Given that xylem cavitation and/or collapse play a principal role, for these species the steep decline at high Ψ_leaf that slows with ongoing dehydration is consistent with an unequal distribution of air seeding pressures, for example the larger vessels that confer the bulk of vein xylem conductivity cavitating and/or collapsing first, and smaller vessels having lower air seeding pressures or wall strength and losing function at lower Ψ_leaf (Table 1). A disproportionate decline at high Ψ_leaf could also relate to species having low major vein densities; and thus, embolism occurring early in these veins, would result in substantial declines in K_leaf (Table 1). If losses in cell permeability are important, the disproportionate decline at high Ψ_leaf could relate to a strong sensitivity of K_leaf to losses in volume in particular cells, with low solute potential, for example bundle sheath cells, that may shrink at high Ψ_leaf and/or undergo aquaporin deactivation (Table 1). If losses of cell wall pathways contribute to the loss of K_leaf, a disproportionate decline at high Ψ_leaf would be consistent with the cell walls behaving as observed for other porous media that show non-linear declines in conductivity with declining water potential, for example soil (Laio et al., 2001). The species variation in vulnerability curves points to the critical importance of research to disentangle the specific mechanisms of K_leaf decline for given species. Notably, previous work has shown species variation in partitioning of hydraulic resistance between petiole and lamina, and among vein orders, and between the vein xylem and extra-xylem pathways (Trifilo et al., 2003b; Sack et al., 2004, 2005). These species differences would also result in variation in the important mechanisms underlying sensitivity to hydraulic decline because K_leaf would be most sensitive to declines in conductance in the component that accounted for the greatest part of the leaf resistance (Scoffoni et al., 2011).

In this study the focus was on on the response of K_leaf to dehydration under high irradiance. It is noted that many species show an increase of K_leaf under high irradiance, and this response may interact with the response to dehydration (Kim and Steudle, 2007; Lee et al., 2009; Voicu et al., 2009). The decline in conductance under low irradiance occurs in the extra-xylem tissues (Nardini et al., 2005); thus, under low irradiance, the extra-xylem tissues would account for a greater proportion of leaf resistance, and cavitation or collapse of vein xylem would have a lesser impact on K_leaf, and any reduction in the permeability of extra-xylem tissues due to dehydration would have a stronger impact (Nardini et al., 2005; Scoffoni et al., 2008; Voicu et al., 2008). The interaction of the light and dehydration responses of K_leaf is an important area for future investigation.

Quantifying the vulnerability of K_leaf: importance and limitations of the steady-state method

Since it is not yet possible to determine K_leaf directly across a full range of Ψ_leaf in vivo, hydraulic methods have been applied to excised leaves. The EFM is the latest of several approaches to measuring K_leaf vulnerability on excised leaves. These methods have advantages over indirect methods, such as the audio method, which registers amplified ultrasonic acoustic emissions (UAEs) within drying plant tissue, hypothesized to arise from cavitation (Milburn and Johnson, 1966; Tyree and Dixon, 1983, 1986; Kikuta et al., 1997; Johnson et al., 2009a), or visual methods using dye or cryo-scanning electron microscopy that directly demonstrate embolism in dehydrated leaves (Salleo et al., 2000, 2001), and collapse of conduits in dehydrated conifer leaves (Cochard et al., 2004; Brodribb and Holbrook, 2005), because these methods do not provide information of possible extra-xylem decline, or directly measure hydraulic vulnerability. Hydraulic methods applied to excised leaves include, in addition to the EFM, the high pressure flowmeter (Nardini et al., 2001), the vacuum pump method (Lo Gullo et al., 2003), and the RKM, most frequently used for determining leaf hydraulic vulnerability, which estimates K_leaf from the uptake of water during rehydration by analogy to the charging of a capacitor in series with a resistor (Brodribb and Holbrook, 2003a, 2004a, b, 2006; Woodruff et al., 2007, 2008; Hao et al., 2008; Blackman et al., 2009; Brodribb and Cochard, 2009; Johnson et al., 2009a, b; Saha et al., 2009; Blackman and Brodribb, 2011). As described in the Introduction, these methods all have merits and disadvantages. The steady-state EFM is independent of the RKM, and here it confirmed and extended key findings.

Several limitations of the EFM applied to excised leaves apply equally to the other methods for leaf vulnerability. These methods cannot assess the decline of K_leaf and Ψ_leaf that occurs in vivo, when xylem water is under tension, and leaf cells are equilibrated at very low water potentials; the xylem cells may be collapsed, leaf cells shrunken, and aquaporins inactivated (Cochard et al., 2002, 2004; Brodribb and Holbrook, 2005, 2006; Canny and Huang, 2006). Excising the leaf under water relieves the tension, and some of these effects might be reversed rapidly. Discovery of such effects would require new in vivo methods for measuring K_leaf decline. In the meantime, vulnerability measured on excised leaves must be considered as conservative, because these methods measure only the K_leaf decline that is not instantly recoverable, for example embolism in veins, which may require many minutes to hours of low tension and active processes to recover (Tyree and Zimmermann, 2002; Bucci et al., 2003; Trifilo et al., 2003a), and persistent effects on living tissues. Further, all the methods may be affected by recovery of K_leaf with rehydration during the measurement itself, but the analysis in this study showed that comparative estimates of hydraulic vulnerability remained robust despite such recovery.

Linkage of vulnerability with drought sensitivity

Species of dry habitats had lower vulnerability (i.e. lower P₅₀ and P₈₀) than species of moist habitat. This finding was consistent with that of a study of Australian species using the RKM (Blackman et al., 2010), here extended with the steady-state method to a set of species very diverse in drought tolerance. This study also confirmed no trade-off across species between K_max and hydraulic vulnerability, as previously reported using RKM (Blackman et al., 2010) and in a meta-analysis combining data collected with different methods (Sack and Holbrook, 2006), a relationship frequently found for stems (Tyree and Zimmermann, 2002; Maherali et al., 2004; Meinzer et al., 2010). Notably, species from dry and moist habitats did not differ on average in their K_max. This finding is consistent with multiple types of adaptation to drought. Some drought-tolerant species use water sparingly via low maximum rates of gas exchange, consistent with low K_max, while others conduct rapid gas exchange when water is available, consistent with high K_max, and then ‘gear-down’ during shortage, (Maximov, 1931; Grubb, 1998), as illustrated by species such as H. arbutifolia (maximum photosynthetic rate of 14 μmol CO₂ m⁻² s⁻¹; Valladares and Pearcy, 1997).

It was also found that across species P₅₀ and P₈₀ were strongly correlated with the bulk leaf turgor loss point (π_TLP) and osmotic potential (Scoffoni et al., 2011; Fig. 5). This finding confirmed and extended the correlation previously reported between P₅₀ and π_TLP for 19 species using the RKM (Blackman et al., 2010). A low π_TLP might confer resistance to K_leaf decline directly, if it allows cells to preserve their structural integrity at lower bulk Ψ_leaf (Blackman et al., 2010). Previous work has demonstrated the heterogeneity of solute potential and across lamina locations and tissues (Slavik, 1959; Nonami and Schulze, 1989; Koroleva et al., 1997, 2002), and the correlation with vulnerability might be even stronger with the turgor loss point of individual tissues important in the water flow pathways, for example the bundle sheath, rather than for the bulk leaf.

Fig. 5.

Correlation of the leaf water potential at 80% loss of leaf hydraulic conductance (P₈₀) with osmotic potentials (a) at full turgor (π_o) and (b) at turgor loss point (π_TLP), for 10 species of a wide range of drought tolerance. Fitted standard major axes: (a) π_o=0.30×P₈₀+0.85; (b) π_TLP=0.42×P₈₀+1.1. Data for π_o and π_TLP are from Scoffoni et al. (2011).

One consequence of the correlation of vulnerability and π_TLP is a mechanism for inducing protective stomatal closure in drought-sensitive species (Brodribb and Cochard, 2009; Hao et al., 2010). The narrow safety margins found in this study were consistent with past studies showing angiosperms often operating at close to cavitation thresholds (Lo Gullo et al., 2003; Brodribb and Holbrook, 2004a, b) in contrast to conifers and ferns which can have wide safety margins (Brodribb and Holbrook, 2004b). Declines in K_leaf accelerate further declines in Ψ_leaf at a given transpiration rate, and guard cells lose turgor against the background of epidermal cell pressure (Franks and Farquhar, 1999; Damour et al., 2010). After that point, cuticular water loss would lead to slower declines of Ψ_leaf and of K_leaf (Pasquet-Kok et al., 2010). In contrast, in species with low hydraulic vulnerability, the maintenance of K_leaf would allow stomata to remain open without desiccating the mesophyll during diurnal water stress or soil drought (Brodribb and Holbrook, 2003a). This contribution of K_leaf sensitivity to stomatal control is important in whole-plant drought tolerance (Brodribb and Cochard, 2009; Blackman et al., 2010).

Species differences in K_leaf recovery and a new importance of resistance to K_leaf decline

A strong, novel variation across species was found in the ability of dehydrated leaves to recover rapidly in K_leaf with 1 h of rehydration. Six species showed no recovery and four increased in K_leaf by 2.5- to 2.8-fold. This study thus partially confirmed one previous report of a complete recovery for sunflower (Trifilo et al., 2003a). Typically K_leaf did not fully recover after 1 h rehydration, indicating a partial irreversibility consistent with embolisms that require refilling, or losses of cell permeability that might require energy transduction for recovery (Bucci et al., 2003).

The present data on vulnerability and recovery highlighted a new importance for leaf hydraulics in determining performance with changing plant water status. A recent meta-analysis of data for 31 species found that at minimum daily Ψ_leaf, species varied greatly in their K_leaf decline, with roughly half the species being below P₅₀ (Johnson et al., 2009b). Our study showed that among species that did not recover rapidly in K_leaf with rehydration, a low hydraulic vulnerability conferred the ability to maintain K_leaf at a high value through both dehydration and rehydration. Species resistant to hydraulic decline could thus maintain K_leaf at high levels despite transient but severe dynamics in Ψ_leaf, and gain a benefit in maintaining performance during diurnal water stress or soil drought. These findings are consistent with the correlation of low leaf hydraulic vulnerability and the ability of severly droughted plants to recover in transpiration after rewatering (Blackman et al., 2009; Brodribb and Cochard, 2009). Tests are needed of the degree that rapid leaf hydraulic recovery, as shown in this study, contributes to whole-plant hydraulic recovery and tolerance of dynamic water regimes.

Supplementary data

Supplementary data are available at JXB online.

Table S1. A summary of previous studies of leaf hydraulic vulnerability on whole leaves, indicating the various methods used, the different functions fitted to the data, and whether the data were binned or not before line fitting.

Table S2. Minimum and maximum transpirational flow rates (E) for each species measured with the evaporative flux method and corresponding estimated stomatal conductances (g), and cuticular conductances for these species.

Table S3. Parameters for the decline of leaf hydraulic conductance (K_leaf) with declining leaf water potential for 10 species, fitted with four different functions, R² for observed values plotted against predicted values from the fitted function, and values for the Akaike Information Criterion (AIC). For each function, three plots were tested for K_leaf against leaf water potential: (1) ‘Ψ_lowest unbinned’; (2) ‘Ψ_lowest binned’; (3) and ‘Ψ_final’ (see the Materials and methods for additional information).

Table S4. Parameters of leaf hydraulic vulnerability curves (K_max, P₅₀, and P₈₀) determined by fitting four functions to the data for each species (linear, sigmoidal, logistic, and exponential) and using three kinds of plots ‘Ψ_owest unbinned’, ‘Ψ_lowest binned’, and’ Ψ_final’).

Table S5. Species means ±SEs for leaf hydraulic vulnerability parameters and pressure–volume parameters for 10 species ranging widely in drought tolerance.