Leaf Traits of Drought Tolerance for 37 Shrub Species Originating from a Moisture Gradient

Identifying the drought-tolerance traits of plant species originating from a moisture gradient will increase our understanding of the differences and similarities in plant drought tolerance. However, which traits can be used to evaluate drought tolerance remain an open question. Here, we conducted a common-garden experiment on 37 shrub species originating from desert to humid regions. The correlations between plant traits and the native environmental conditions were studied. Leaf sizes and Huber values were significantly correlated with most climate variables of the shrubs’ origins. The osmotic potentials at full turgor (π100), turgor loss point (ΨTLP), and midday leaf water potential (Ψmid) were significantly correlated with most climate variables of their origins. We proposed using leaf sizes, Huber values, and ΨTLP as predictors of drought tolerance across shrub species and shrub biomes. Statistically significant correlations were found between π100, ΨTLP, and specific leaf area (SLA). However, owing to the weak correlations between SLA and the climate variables of the shrubs origins and between Huber values and leaf size and turgor loss traits, it was difficult to integrate leaf morphological traits with physiological traits to find a simple way to accurately quantify drought-tolerance-related differences among these shrub species.


Introduction
All ecosystems worldwide will be influenced to certain degree by climate change, with forecast for more frequent and severe drought occurrence in the future [1]. Forest dieback caused by drought has been reported in every terrestrial ecosystem in the past few decades [2][3][4], which may ultimately lead to the collapse of eco-services [5]. The effective protection of existing forests to provide services under these circumstances must involve the cautious consideration of species' capacities to withstand future climatic interferences and acclimation to future climate regimes [6]. Thus, it is crucial to comprehend the function of leaf morpho-physiological traits in facilitating accliantion to extreme heat and drought under predicted climatic changes [6,7]. Identifying the leaf drought-tolerance traits that are positioned on a moisture or aridity gradient may help us further understand the differences in drought tolerance and will assist in characterizing the impact associated with exposure to future climates. This understanding may help us to reveal the species' different vulnerabilities to climate-induced drought and facilitate the development of effective measures to protect forests.
Many morphological and physiological traits are used as indicators of drought tolerance. Leaf morphology or structural traits are linked to mechanisms of desiccation tolerance and thus

Climate Variables
We used the ALA to assess the climatic conditions of the observed species' distributions. The species occurrence records come from a scientific collection of specimens and observations by an individual or member of an organization. Only the validated records of species occurrences layers were overlaid with the selected environmental layers and then downloaded from the ALA using the "mapping and analysis" portal. The climatic layers (based on a gridded dataset,~1 km × 1 km) extracted from the environmental layer portal on the ALA. We downloaded mean annual precipitation (MAT), mean annual temperature (MAP), and mean annual aridity index (AI) data for all sampling points across each species distribution within Australia from the ALA [34]. AI values ranged from 0.175 to 1.381 (Table 1). Previous work showed percentile values of climate variables were key limits for plant survival under drought conditions across Australian vegetation types [38]. Thus we used percentile values to characterize relatively "extreme" edges of the species climatic niche. To characterize the dry ends of the MAP, we used a 0.02 percentile value; while to characterize the warm ends of the MAT data, we used a 0.98 percentile value. In order to compare trait differences across the biome with contrasting habitat preferences, the shrub species were divided into subgroups in accordance with the site MAP. Namely, plants receiving annual rainfalls of less than 400 mm, between 400 mm and 800 mm, and greater than 800 mm were mainly arid shrubs, woody open shrubs, and wet shrubs under forest trees, respectively.

Ψ TLP and Pressure-Volume Traits
Pressure-volume (P-V) curves were generated for a minimum of four leaves or shoots of each shrub species using the bench-drying method in accordance with Tyree and Hammel [39]. For each species, shoots of six individual trees were cut in the morning and rehydrated in deionized water for 1 h, which proved to be adequate time for the full rehydration of most species [29]. One leaf of each shoot was cut, and its turgid weight was measured. Then, the corresponding water potential was measured in the pressure chamber. If this measurement was less than −0.1 MPa, the leaf was discarded because it had not fully rehydrated. The leaf was set on the bench to slowly dehydrate at the indoor temperature of 22 • C. Leaf weight were measured approximately every 30-60 min until the correlation between 1/Ψ and fresh weight formed a straight line consisting of at least four measurements. The leaf was dried in an oven at 70 • C for at least 72h to get a constant weigh. The sample was then weighed to determine the relative water content (RWC) as follows: where FW, DW, and TW represent the leaf fresh, dry and turgid weights, respectively. The Ψ TLP , osmotic potential at full turgor (π100), and elastic modulus at full turgor (ε max ) were derived from the curve of 1/Ψ against RWC using a P-V curve-fitting routine available online (http://landflux.org) that is based on the work of Schulte and Hinckley [40]. Leaf water potential for P-V curves and midday leaf water potential (Ψ mid ) was determined using a Scholander-type pressure chamber (Soil moisture Equipment Corp., Santa Barbara, CA, USA).

Morphological Measurements
All the leaves of three small shoots were removed, and the leaf numbers per shoot were recorded. Leaf size was determined by scanning the leaf area with an LI3100 area meter (Li-Cor, Lincoln, Nebraska, USA). Leaf size was calculated as the average of the three repetitions.
Dry weights (DWs) were measured after drying the leaves at 70 • C for 72 h in an oven. Specific leaf area (SLA) was calculated as fresh leaf area divided by DW. The SLAs of all the leaves used for leaf size measurement were calculated.
For the Huber values (sapwood to leaf area ratio), the above three small shoots were used. Such small shoots lack heartwood; therefore, the sapwood area was approximated using the total cross-sectional area. The Huber value was then computed using the sapwood to leaf area ratios of all the leaves on the small shoots.

Statistical Analyses
To calculate the means and standard errors of each data set, descriptive statistics were used. Pearson's moment correlation tests were used to analyzed the correlation between trait data. Differences among leaf size, SLA, Huber values, π 100 , Ψ TLP , and ε max of the different subgroups of shrubs were tested using one-way analyses of variance, followed by the post-hoc Tukey's HSD multiple comparison tests. Figures were plotted using Origin 8.0 (Origin Lab Corp., Northampton, MA, USA). All the data analyses were carried out using IBM SPSS Statistics (Ver.22, Armonk, NY, USA).

Mean Climatic Parameters for Each Species
Climatic parameters for each shrub species ranged between the two extremes of arid and subalpine rainforest environments (Table 1). Among the natural distribution sites of the 37 shrub species, mean annual rainfalls range from 313 mm to 1334 mm. The AI ranged from 1.381 to 0.175. The subalpine climate is characterized by high rainfall and low temperature conditions, resulting in a high AI value. For example, Grevillea victoriae distribute in the south-eastern part of New South Wales and the mountain areas of Victoria, Australia, while Correa lawrenceana is found in rainforests and sclerophyll forests in Tasmania, Victoria, New South Wales, and Queensland. Along the moisture gradient, the climate for the provenances of shrub species became hotter and drier, resulting in decreasing AI values. For example, Rhagodia spinescens and Halgania cyanea are found in the arid and semiarid parts of Western Australia ( Figure 1).

Dependencies of Leaf Size, SLA, and Huber Value on Provenance Climate
Of the three climatic indices examined, the leaf sizes were significantly greater as MAP and AI values increased and were smaller as MAT values increased (Figure 2a-c). MAT and AI were the best single predictive variables of leaf size. Huber values significantly decreased as MAP and AI values increased (Figure 2h,i). MAP and AI were the best single predictive variable of the Huber value. Huber values partly reflect the water supply and demand balance among shrubs as a result of long-term adaptation to their natural distribution site's water availability. There were weakly positive correlations between Huber values and MAT ( Figure 2g). Correlations between SLA and the three climatic indices were not significant (Figure 2d-f). Thus, SLAs were independent of the three climatic variables. The addition of climatic indicators founded on the dry ends of the species distribution (e.g., MAP, AI, and MAT) did not reinforce the correlations between climatic indicators and morphological traits (data not shown), compared with climatic indicators derived from the means ( Figure 2). Summarily, across the whole data set, leaf size and Huber values were closely correlated with the natural distribution's climatic conditions.

Dependence of Turgor Loss Traits on the Native Climate
We observed strong correlations between the site's climatic variables and most water-related parameters. The correlations between π 100 and the site's MAT and AI values were significant (Figure 3a,c). The π 100 decreased as MAT increased and increased along with AI. There were significantly correlations between Ψ TLP and the three climatic indices (Figure 3d-f). MAT and AI were the best single predictors of Ψ TLP . Ψ TLP decreased as MAT increased and increased as AI decreased. Ψ TLP increased along with MAP. The midday leaf water potential was also closely correlated with the natural distribution site's MAP and AI (Figure 3k,l). The addition of climatic indicators founded on the dry ends of the species distribution (e.g., MAP, AI, and MAT) did not reinforce the correlations between climatic indicators and physiological traits (data not shown) compared with climatic indicators derived from the means (Figure 3).     correlations between ΨTLP and the three climatic indices (Figure 3d-f). MAT and AI were the best single predictors of ΨTLP. ΨTLP decreased as MAT increased and increased as AI decreased. ΨTLP increased along with MAP. The midday leaf water potential was also closely correlated with the natural distribution site's MAP and AI (Figure 3k and l). The addition of climatic indicators founded on the dry ends of the species distribution (e.g., MAP, AI, and MAT) did not reinforce the correlations between climatic indicators and physiological traits (data not shown) compared with climatic indicators derived from the means (Figure 3).

Correlations between Turgor Loss and Other Leaf Traits
We discovered a strong positive correlation between osmotic potential at π100 and ΨTLP ( Figure   4a). Variation in π100 values among the 37 shrub species was a significant element in interpreting the variation in ΨTLP, and there was a positive correlation in wet winters. There were significantly negative correlations among π100, ΨTLP and ε max, demonstrating that osmotic and elastic adjustments coordinated among the leaf water-related traits (Figure 4c,d). The positive correlation between Ψmid and ΨTLP (Figure 4b) indicated that habitat water availability (midday leaf water potential) was also a significant predictor of ΨTLP. We observed weak correlations between leaf morphological traits and most water-related parameters, except SLA, which was strongly correlated with π100, ΨTLP, and ε max ( Figure 5).

Correlations between Turgor Loss and Other Leaf Traits
We discovered a strong positive correlation between osmotic potential at π 100 and Ψ TLP (Figure 4a). Variation in π 100 values among the 37 shrub species was a significant element in interpreting the variation in Ψ TLP , and there was a positive correlation in wet winters. There were significantly negative correlations among π 100 , Ψ TLP and ε max , demonstrating that osmotic and elastic adjustments coordinated among the leaf water-related traits (Figure 4c,d). The positive correlation between Ψ mid and Ψ TLP (Figure 4b) indicated that habitat water availability (midday leaf water potential) was also a significant predictor of Ψ TLP . We observed weak correlations between leaf morphological traits and most water-related parameters, except SLA, which was strongly correlated with π 100 , Ψ TLP , and ε max ( Figure 5).

Leaf Water-Related Traits Correlated with Native Climates across Biomes
We compared species among the following biome categories: arid/desert shrub (n = 4), woody open shrub (n = 13), and wet shrub under forest (n = 20) in a high-altitude mountainous region. Although all the traits varied significantly among biomes (Figure 6), leaf size, Huber value, π 100 , and Ψ TLP values could be used to separate moist from dry biomes.

Figure 5.
Correlations between each of three traits, osmotic potential at full turgor (π100), turgor loss point (Ψ TLP) and modulus of elasticity (ε max), and leaf size, specific leaf area (SLA) and Huber value. Solid lines represent the linear regressions, and the associated r 2 values are provided (*P < 0.05, ** P < 0.01, *** P < 0.001). Data are means ± 1 SE.    We compared species among the following biome categories: arid/desert shrub (n = 4), woody open shrub (n = 13), and wet shrub under forest (n = 20) in a high-altitude mountainous region. Although all the traits varied significantly among biomes (Figure 6), leaf size, Huber value, π100, and ΨTLP values could be used to separate moist from dry biomes.

Discussion
Increasing drought is a critical challenge for species survive and ecosystems stability; therefore, advanced theory and practices are desired for the quantification of species drought tolerances [8]. An increasing quantity of literature indicated within-species correlations between native climate and leaf morphology and physiology were weak in the common garden [12,41], compare with the persistently more robust correlations among species growing at different sites along a wide moisture gradient [6,21,23,34,42] and even in common closely related garden species [43]. In the current study, 37 shrub species with different climatic origins were planted in a common garden for at least five years before they were used in our experiment. However, the significant correlations between their natural distribution sites' climatic conditions and their morphological and physiological traits demonstrated the strong selective pressure of climatic variables in shaping plant ecophysiological traits among Australian shrub species.

The Relevance of Native Climate to Morphological Traits
The correlations between several leaf attributes and environmental factors, especially water or nutrient availability are generally consistent [44]. The reduction in leaf size as rainfall decreases is often observed [45,46], and this correlation is typically associated with an increase in leaf thickness and a decrease in SLA (leaf area per unit dry mass) [46,47]. However, in our study, SLA showed no synchronous decrease as leaf size decreased (Figure 2a-f). Small leaf size decrease the boundary

Discussion
Increasing drought is a critical challenge for species survive and ecosystems stability; therefore, advanced theory and practices are desired for the quantification of species drought tolerances [8]. An increasing quantity of literature indicated within-species correlations between native climate and leaf morphology and physiology were weak in the common garden [12,41], compare with the persistently more robust correlations among species growing at different sites along a wide moisture gradient [6,21,23,34,42] and even in common closely related garden species [43]. In the current study, 37 shrub species with different climatic origins were planted in a common garden for at least five years before they were used in our experiment. However, the significant correlations between their natural distribution sites' climatic conditions and their morphological and physiological traits demonstrated the strong selective pressure of climatic variables in shaping plant ecophysiological traits among Australian shrub species.

The Relevance of Native Climate to Morphological Traits
The correlations between several leaf attributes and environmental factors, especially water or nutrient availability are generally consistent [44]. The reduction in leaf size as rainfall decreases is often observed [45,46], and this correlation is typically associated with an increase in leaf thickness and a decrease in SLA (leaf area per unit dry mass) [46,47]. However, in our study, SLA showed no synchronous decrease as leaf size decreased (Figure 2a-f). Small leaf size decrease the boundary layer's resistances and facilitate maintaining favorable leaf temperatures and higher photosynthetic water-use efficiencies under high solar radiation and low water availability conditions [48]. Thus, a narrow leaf form or small leaf size is considered as an adaptation to xeric environment [12]. Leaf size across biomes differed significantly (Figure 6d). Therefore, in our current study, using leaf size as a predictor of the drought tolerance of shrub species was still practical, as supported by Wright et al. [49].
The SLA usually declines with increasing aridity and nutrient scarcity in native habitats [46,50] because a low SLA is accompanied with drought resistance and also enhances nutrient residence time [51]. The weak correlations between SLA and site climatic variables across plant species were consistent with a study by Niinemets [21] in its inverse measurements (leaf dry mass per unit area). After the rainfall data was log10-transformed, their study showed a significant linear relationship between SLA and the site's mean monthly precipitation during the three driest months. In our study, owing to the variability in the monthly distributions of rainfall at the native climatic sites of the shrub species [52], we were unable to perform such an analysis. The lack of correlation between SLA and the climate of origin, and no significant difference in SLA across the biome, in our study (Figure 2d-f; Figure 6e) strongly suggested that the wide and untimely use of SLA to forecast species' drought tolerances and distributions is likely to yield misleading results [7,16].
The Huber values of the same or closely related species have been interrelated to climatic gradients, such as precipitation or vapor pressure deficit [53][54][55]. However, other studies found no correlation between the Huber value and climate of origin or water availability [26,56,57]. These contradictory conclusions may be attributed to the plasticity of Huber values. This value can enhance as a result of a decrease of leaf area, which can appear under drought stress. For instance, the Huber value may vary seasonally, increasing during the drought season (water stress or high vapor pressure deficit) [58][59][60]. We measured the Huber value during the wet winters when the soil water deficit was slight and the vapor pressure deficit was also low, which partly relieved the influence of the plasticity of the Huber value on their relationship. The significantly close correlations between Huber values and MAP and AI in our study strongly supported the first viewpoint in a more broad scope across the different shrub species with the moisture gradient in the common garden (Figure 2h,i). The Huber values across the biome also showed significant differences (Figure 6f). Thus, we proposed using the Huber value as a predictor of drought tolerance across shrub species and shrub biomes. However, we suggest caution in using Huber values to assess drought-tolerance levels across closely related plant species belonging to the same genus.
Although coherent and universal relationships between most leaf structural traits and native climate variables were found in this case, the explained variances were quite low in some circumstances. Besides the error variance related to climatic data and the plasticity of morphological traits when planting in a common garden, the low universality in climate vs. leaf structure relationships indicates that other plant adaptive traits related to phenology, depth of root systems, and biomass allocation patterns also have important roles in determining the competitive abilities of species with resemble leaf structures along moisture gradients [21].

The Importance of Physiological Traits to Ecological Drought Tolerance
Leaf water potential at wilting (Ψ TLP ), is a deciding factor of the leaves tolerance to water stress and devotes to plant-level physiological drought tolerance [16]. Leaf turgor loss plays a functional role in driving stomatal closure and hydraulic dysfunction [61,62]. Meanwhile, the leaf water potential at Ψ TLP is increasingly being used as a robust functional trait to serve as a proxy for ecological drought tolerance across species and vegetation having gradients of site water availability [6,13,17,28,33,43].
The close positive correlations between site climatic variables and Ψ TLP (Figure 3d-f) indicated that Ψ TLP is a reliable indicator of a species drought tolerance and could be used to anticipate drought responses. Our study also showed that biomes from dry areas had more negative Ψ TLP values than those from moisture areas (Figure 6b), and it reinforced the conclusion that Ψ TLP may imply drought tolerance at the biomes scale [13,42]. Hence, the introduction Ψ TLP into process-based vegetation models may advance the accuracy of forecast of climate change' influence on shrub species in Australia [7,63]. Currently, the use of vapor pressure osmometers, which measure leaf Ψ TLP rapidly, has overcome the inefficiency and time-consumption of the traditional pressure-volume curve approach [27]. We expect the new convenient measuring method to have broad applications in forecasting species reaction to climate fluctuation.

Coordination among Morphological and Physiological Traits and Their Correlations with Native Climates
The close positive correlation between osmotic potential at π 100 and Ψ TLP (Figure 4a) strongly supported the recent analyses that saturated turgor (π 100 ) is the crucial variable driving Ψ TLP across shrub species and, thus, that π 100 and Ψ TLP are robust attributes for forecasting drought tolerance and distributions relating to moisture gradient [13].
There is currently no consensus on whether or not the osmotic or elastic adjustments play a key role in drought acclimation, but increases in ε max and decreases in tissue osmotic potential in response to water stress often co-occurred [21]. The negative relationship between osmotic potential at π 100 and ε max (Figure 4d) indicates species that mount up solutes (i.e., more negative π 100 ) have rigidity cell walls contrasted with species that have less negative π 100 [18]. The interdependence of π 100 and ε max has been verified formerly for several woody species, including Santalum acuminatum (R.Br.) A. DC. (quandong, a shrub), Nuytsia floribunda (Labill.) Fenzi (Western Australian Christmas tree), heath and mallee communities [18], Eucalyptus species [18,64,65], Ziziphus mauritiana Gola [66], and Douglas fir [67]. The larger cell-wall rigidity (higher ε max ) in species with higher concentrations of internal cellular solute (lower π 100 ) may be indispensable during fast re-hydration to maintain tissue and cell-wall integrity. Additionally, if the cell walls succumb in response to turgor pressure generated by osmotic accumulation, then the advantages of osmotic adjustment are lost. Similarly, plants that cannot osmotically adjust may profit from elastic walls to sustain positive turgor pressures at lower cell volumes and water contents [18].
Previous studies on SLA and Ψ TLP under drought stress conditions, involving research on Mediterranean species [68,69], showed no coherent relationship between these attributes. However, these researches were limited to relative few species, and the discrepancy in SLA among those species was minor. In current study, we sampled a various scopes of leaf forms, having a relatively large scope of SLA values (12.22-308.89 cm 2 g −1 ), and found significant correlations between π 100 , Ψ TLP , and SLA. This result was consistent with the research of Niinemets [21] and Mitchell et al. [18]. Shrub species with lower Ψ TLP values preferred to have smaller SLAs, which partly ascertains that increased leaf turgor maintenance is combination with more leaf carbon input [42,69]. Thus, Ψ TLP may be used to forecast a species' position along the "fast-slow" plant economic spectrum [42,62]. However, owing to the weak correlation between SLA and the climate variables of the shrub origins, and the weak correlations between Huber values and leaf size and turgor loss traits, it was still difficult to integrated leaf morphological traits with physiological traits to represent ecological drought-tolerance differences among these shrub species. Therefore, our present study did not support the viewpoint that the integration of comparatively unrelated traits (leaf physiological traits vs leaf morphological traits) attains a more precise characterization of their environmental restrictions, as reported by Costa-Saura [23].

Conclusions
Our study showed that leaf size and Huber values were significant correlated with the climate of the natural distribution site and markedly different between the dry and humid origin shrub biomes. Thus, we proposed the combination of leaf size and Huber value as a predictor of ecological drought tolerance across shrub species and shrub biomes. The close correlations between turgor loss traits and the comprehensive AI suggested that, to a large degree, leaf turgor loss traits were important and have broad ecological significance with regard to climatic zones. Ψ TLP can be used as an index to characterize the drought-tolerance abilities of shrub species and shrub biomes; thereby, defining the scope of their natural distribution along a moisture gradient. It was still difficult to integrate leaf morphological traits with physiological traits to accurately represent drought-tolerance differences among these shrub species. The morphological and physiological traits may be used to quantify drought tolerance, depending on the target concern and the ease of measurement. In spite of the universality of plasticity, Ψ TLP measured in the wet season in a common garden reliably characterized most shrub species' essential drought tolerances and distributions relation to water availability.