Evidence for phylogenetic signal and correlated evolution in plant–water relation traits
Summary
Evolutionary relationships are likely to play a significant role in shaping plant physiological and structural traits observed in contemporary taxa. We review research on phylogenetic signal and correlated evolution in plant–water relation traits, which play important roles in allowing plants to acquire, use, and conserve water. We found more evidence for a phylogenetic signal in structural traits (e.g. stomatal length and stomatal density) than in physiological traits (e.g. stomatal conductance and water potential at turgor loss). Although water potential at turgor loss is the most-studied plant–water relation trait in an evolutionary context, it is the only trait consistently found to not have a phylogenetic signal. Correlated evolution was common among traits related to water movement efficiency and hydraulic safety in both leaves and stems. We conclude that evidence for phylogenetic signal varies depending on: the methodology used for its determination, that is, model-based approaches to determine phylogenetic signal such as Blomberg's K or Pagel's λ vs statistical approaches such as ANOVAs with taxonomic classification as a factor; on the number of taxa studied (size of the phylogeny); and the setting in which plants grow (field vs common garden). More explicitly and consistently considering the role of evolutionary relationships in shaping plant ecophysiology could improve our understanding of how traits compare among species, how traits are coordinated with one another, and how traits vary with the environment.
| Contents | ||
|---|---|---|
| Summary | 392 | |
| I. | Introduction | 392 |
| II. | Phylogenetics: the study of evolutionary relationships among taxa | 393 |
| III. | How shared evolutionary history informs plant–water relation traits | 396 |
| IV. | Phylogenetic signal in structural vs physiological traits | 402 |
| V. | Trait–environment relationships | 402 |
| VI. | The size of the phylogeny | 403 |
| VII. | Conclusions | 403 |
| Acknowledgements | 404 | |
| References | 404 |
I. Introduction
Plants possess a number of different functional traits that allow them to cope with water-deficit conditions and thus affect performance and ultimately fitness (Violle et al., 2007). These traits, often referred to as plant–water relation traits, may be either physiological or structural. Physiological traits include stomatal conductance (gs), water potential at turgor loss point (Ψtlp), and stem hydraulic conductivity (Kh), whereas structural traits include stomatal length (SL), vein density (VD), and Huber value (HV). These traits impact water acquisition, use, and conservation. Plant–water relation traits have been studied to provide insights into how species compare with each other (Cavender-Bares & Holbrook, 2001; Maherali et al., 2004), how these traits are coordinated to form plant strategies (Cavender-Bares et al., 2004; Jacobsen et al., 2007; Zhang et al., 2014), and how these traits vary as a function of the environment (Mitchell et al., 2018; Ramírez-Valiente et al., 2020).
The use of comparative methods has allowed biologists to compare traits among populations, species, and larger taxonomic groups and find the causes of repeated patterns (Olson & Arroyo-Santos, 2015; Olson, 2021). Studying the adaptive origin of traits allows us to ask and answer questions on the possible causes of trait diversity and trait similarity and whether the latter is the product of convergent evolution or shared evolutionary history. In plant physiological ecology, for example, using explicit information about shared evolutionary history (i.e. by examining phylogenetic trees in conjunction with traits) can allow us to understand whether independent evolutionary change or similarity, due to common descent, shaped patterns of trait distributions among species, and how these trait distributions affect the life of plants. How does a trait evolve and why does it evolve that way?
While there is considerable research on plant–water relations, it is not clear to what extent closely related species tend to resemble one another more than distantly related ones. In some subdisciplines of ecology, there has been appreciation for how evolution influences the processes and patterns we observe today and this has led to significant new insights into behavioral ecology, community ecology, and conservation biology (Webb et al., 2002; Blomberg et al., 2003; Wiens & Graham, 2005; Kraft et al., 2007). In community ecology, the use of phylogenies has helped to answer questions on niche evolution and community assembly. Ignoring the contributions of evolutionary history to plant–water relation traits – and to plant physiological ecology more broadly – leaves us vulnerable to misinterpretation of the causes and consequences of the patterns and processes we wish to unravel.
We did a search in Google Scholar using the words ‘phylogenetic signal’ + ‘trait’ + ‘water’ + ‘plant’. We inspected the articles and chose those in which phylogenetic signal in plant–water relation traits (see a list in Table 1) was assessed using either Pagel's λ or Blomberg's K. Some of these articles also evaluated trait–trait correlations in a phylogenetic context and were used to discuss correlated evolution. We reviewed 39 articles for phylogenetic signal and 17 for correlated evolution. We aimed to synthesize the current evidence for (1) the presence of phylogenetic signal in plant–water relation traits, and (2) correlated evolution among those traits. We then leveraged this information to better understand how shared evolutionary history informs plant–water relation traits and possible differences between physiological and structural traits, how incorporating phylogenetic information into our research can improve our understanding of trait–environment relationships, what is the effect of size and type of phylogeny on being able to detect phylogenetic signal, and what research we may prioritize moving forward. But before delving into these questions, we start by defining phylogenetic signal and describing the phylogenetic methods currently used.
| Trait | Trait abbreviation |
|---|---|
| Stomatal length | SL |
| Stomatal area | SA |
| Stomatal density | SD |
| Stomatal pore index | SPI |
| Maximum stomatal conductance | gs,max |
| Transpiration rate | E |
| Leaf hydraulic conductance | Kleaf |
| Vein density | VD |
| Leaf cavitation resistance | P50leaf |
| Leaf area | LA |
| Leaf thickness | LT |
| Turgor loss point | Ψtlp |
| Relative water content at turgor loss point | RWCtlp |
| Stem hydraulic conductivity | Kh |
| Sapwood-specific stem hydraulic conductivity | KS |
| Leaf-specific stem hydraulic conductivity | KL |
| Stem cavitation resistance | P50stem |
| Wood density | WD |
| Huber value | HV |
II. Phylogenetics: the study of evolutionary relationships among taxa
The fact that species are not independent of each other indicates that there is a need for analyses that assess the role of shared evolutionary history in informing observed traits, as well as account for this nonindependence when comparing and contrasting species with one another and with their environment (Felsenstein, 1985; Revell et al., 2008). Furthermore, a phylogenetic perspective can allow us to comprehend how traits evolve over time. For example, understanding trait correlations, associations with environment, rates of evolution, gain, and loss of traits requires data to be placed in a phylogenetic framework. In the following paragraphs, we review the methods used to assess phylogenetic signal and trait–trait covariation in a phylogenetic context (correlated evolution). We provide working definitions for key terms in Box 1.
Box 1. Key phylogenetic terms
Comparative methods: Methods that involve the comparison of two traits across a group of taxa, or the comparison of one trait along an environmental variable (Felsenstein, 1985). Usually, the methods involve correlation and regression analyses to investigate what traits of organisms change with other traits, or with aspects of their environment, potentially providing evidence for the temporal order of changes in traits, and suggesting probable causal pathways (Pagel, 1999).
Correlated evolution: Two or more traits evolve together along lineages, either in a coordinated way (positive association) or trading-off (negative association). Correlated evolution is detected by analyzing the correlation of continuous traits in modern taxa using phylogenetically independent contrasts (PICs) (Martins & Garland, 1991; Price, 1997), or by ancestral reconstructions of discrete traits (Pagel, 1999). Phylogenetically independent contrasts in this framework are used to evaluate whether the divergence of a trait is consistently correlated with the divergence of another trait during the evolution of these traits (Santiago & Kim, 2009).
Blomberg's K: A measure of phylogenetic signal calculated as the ratio of two ratios: the numerator is the mean squared error of the tip trait data measured from the phylogenetically corrected mean (where the mean trait value of the root is subtracted from the tip trait data), divided by the mean squared error of the trait data calculated using the variance–covariance matrix derived from the phylogeny; the denominator is the expected ratio using data from a model under the assumption of Brownian motion of trait evolution (Blomberg et al., 2003).
Brownian motion model of trait evolution: A random walk model in which trait values change randomly along the length of the phylogenies' branches.
Pagel's λ: A phylogeny scaling factor that ‘detects whether the shared evolutionary histories as specified by the phylogeny produce the patterns of similarity observed in the data’ (Pagel, 1999). Values of λ < 1 correspond to traits being less similar among species than expected from their evolutionary relationships, and λ > 1 suggests the opposite.
Phylogeny: A branching tree diagram that shows how organisms are related to each other relative to the position of their most recent common ancestor (Felsenstein, 1983).
Phylogenetically independent contrasts (PICs): An independent comparison among pairs of species and higher nodes in a phylogeny (Felsenstein, 1985; Pagel, 1992). It is calculated by taking the difference between the mean trait value of the two taxa and dividing it by the branch length of the two taxa (Garland, 2005). Phylogenetically independent contrasts meet the assumption that samples are independent of one another, but they assume that trait evolution occurs under Brownian motion.
Phylogenetic generalized least squares (PGLS): A modification of generalized least squares, in which a linear regression is performed using a correlated structure based on a variance–covariance (vcv) matrix of the evolutionary relationships among species. The vcv matrix is obtained from a candidate phylogeny. When trait evolution along the phylogeny is assumed to follow Brownian motion, then PGLS and PIC results are the same, that is, PIC is a special case of PGLS (Garland, 2005).
Phylogenetic conservatism: The tendency (pattern) of closely related species to be more ecologically similar than expected based on their evolutionary relationships (Losos, 2008). High values of phylogenetic signal (the definition of which follows) may indicate phylogenetic niche conservatism, but this alone is not sufficient evidence.
Phylogenetic signal: The tendency (pattern) of closely related species to resemble each other more than they resemble species randomly drawn from the phylogenetic tree (Blomberg & Garland, 2002; Blomberg et al., 2003). Another definition is the degree to which variation in species trait values is predicted by the relatedness of the species (Losos, 2008; Swenson, 2014). The presence of phylogenetic signal is the expected outcome of Brownian motion trait evolution, and it can ensue from genetic drift or natural selection that fluctuates randomly in direction and magnitude through time (Losos, 2008).
Phylogenetics: The field of biology that seeks to understand the evolutionary relationships among species.
Ornstein–Uhlenbeck evolutionary process: A process that models the effects of stabilizing selection around an optimum trait value (Lavin et al., 2008). Accordingly, an Ornstein–Uhlenbeck transformation is the change in branch lengths of a candidate phylogeny (d = 1) to make it more hierarchical (d > 1) or less hierarchical (0 < d < 1) and test for fit of the trait data.
Phylogenetic signal is defined as the tendency of closely related species to resemble each other more than expected by chance (Blomberg & Garland, 2002; Blomberg et al., 2003). As the definition implies, phylogenetic signal is a pattern, with multiple evolutionary processes leading to the presence or lack of it. Processes include natural selection that fluctuates in time, magnitude and direction, and random drift (Losos, 2008). It is important to note that the terms phylogenetic signal and phylogenetic conservatism are interchangeably used in the literature, although they are not exactly the same. Phylogenetic conservatism indicates that closely related species are more similar ecologically than expected by their shared evolutionary history, and even though high phylogenetic signal may indicate phylogenetic conservatism, phylogenetic signal alone is not sufficient to demonstrate conservatism (Losos, 2008). The terms phylogenetic effect, phylogenetic constraint, and phylogenetic inertia have also been used to refer to the role of shared evolutionary history in explaining ecological similarity (Blomberg et al., 2003; Losos, 2008). Following Losos (2008), we adopt the term phylogenetic signal throughout this review.
Multiple indices have been developed to assess the degree of phylogenetic signal. The four most common indices are Moran's I, Abouheif's Cmean, Pagel's λ, and Blomberg's K. The first two indices are based on autocorrelation and are considered statistical approaches that contrast with model-based methods such as Pagel's λ and Blomberg's K, which involve a specification of a mode of evolution (Diniz-Filho et al., 2012). We focus on the latter two, which are the most commonly used in plant physiological ecology.
One of the first methods proposed to assess phylogenetic signal is Pagel's λ (Pagel, 1999). Pagel's λ is the transformation of a phylogenetic tree that best describes the trait data under a Brownian motion (BM) model of trait evolution. When λ is equal to 1, no transformation has been performed, indicating that the phylogeny alone is enough to explain the trait data. As λ decreases, the phylogeny becomes less hierarchical, approaching a star phylogeny, indicating that the phylogeny becomes less important in explaining the data. Values of λ > 1 are not defined (Cadotte & Davies, 2016). A simpler explanation for this concept is often provided by λ = 0 indicating no phylogenetic signal and λ = 1 indicating phylogenetic signal under BM.
A second, more recent method, Blomberg's K, ‘seeks to quantify the degree to which variation in a trait is explained by the structure of a given phylogenetic tree’ (Swenson, 2014). Blomberg's K is a ratio of two other ratios: the numerator is the mean squared error of the tip trait data measured from the phylogenetically corrected mean (where the mean trait value of the root is subtracted from the tip trait data), divided by the mean squared error of the trait data calculated using the variance–covariance matrix derived from the phylogeny (a phylogenetic distances matrix); the denominator is the expected ratio using data from a model under the assumption of BM of trait evolution (Blomberg et al., 2003). Similar to Pagel's λ, K = 0 indicates no phylogenetic signal, 0 < K < 1 indicates that closely related species resemble each other less than expected under the BM model of trait evolution, K = 1 indicates phylogenetic signal as expected by BM evolution, and K > 1 indicates high phylogenetic signal, with closely related species resembling each other more than expected under BM (Blomberg et al., 2003). When K < 1 or > 1, other models of trait evolution may better explain the data. Even though the use of Blomberg's K appears to have increased in recent years, recent reviews demonstrated that Pagel's λ is a more appropriate measure to test phylogenetic signal in ecologically important traits when phylogenetic information is incomplete, especially when there are polytomies in the phylogenetic tree, or when information on branch lengths is not known with certainty (Münkemüller et al., 2012; Molina-Venegas & Rodríguez, 2017).
Both Blomberg's K and Pagel's λ evaluate the strength of the phylogenetic signal relative to the BM model (Fig. 1a,b). Both methods are bound at 0 (indicating no phylogenetic signal) with values of 1 equal to expectations from BM. Whenever testing for phylogenetic signal, we recommend including the null hypothesis being tested to avoid ambiguity: K = 0 or K = 1, or λ = 0 or λ = 1.
It is common to find that closely related species resemble one another more than distantly related ones (Fig. 1c; Ackerly & Reich, 1999; Maherali et al., 2004). For example, correlations among traits may arise due to deep evolutionary divergences, such as the one between angiosperms and gymnosperms (Ackerly & Reich, 1999), or because some processes maintain low trait differences among closely related species. Importantly, Felsenstein (1985) pointed out that because of these evolutionary relationships, species cannot be regarded as independent data points in traditional statistical analyses and proposed a method for testing trait correlations among extant taxa by taking into account evolutionary relationships. This method, known as phylogenetically independent contrasts (PICs), is carried out by computing independent contrasts between pairs of sister taxa: taking the difference between the mean trait value of the two taxa and dividing it by the square root of the sum of the branch lengths linking the two taxa. Phylogenetically independent contrasts meet the assumption that samples are independent of one another, but they assume that trait evolution occurs under BM.
The use of PICs improved the testing of comparative biology hypotheses (Miles & Dunham, 1993; Ricklefs & Starck, 1996; Garland & Díaz-Uriarte, 1999), mainly those related to trait–trait covariation in an evolutionary context. However, PICs have seen a decline in use as phylogenetic generalized least squares (PGLS) and phylogenetic regressions have increased in use because of their flexibility in allowing for multiple dependent variables in PGLS and the incorporation of models of trait evolution other than BM in phylogenetic regressions (Garland, 2005; Blomberg et al., 2012; Symonds & Blomberg, 2014). Results from PGLS under BM are equivalent to PICs.
Accounting for phylogeny in trait–environment relationships is generally done using PGLS, an extension of generalized least squares that includes a correlated structure based on evolutionary relationships among taxa. This correlated structure is a variance–covariance matrix estimated from the phylogenetic tree, where closely related species sharing a relatively recent common ancestor (short branch lengths) have high values of expected trait covariance (Symonds & Blomberg, 2014).
III. How shared evolutionary history informs plant–water relation traits
In the following sections, we review published research on how shared evolutionary history of species informs plant–water relation traits related to stomata, leaf, stem, and whole-plant functioning. We first synthesize the evidence for phylogenetic signal and correlated evolution in those traits (Fig. 2; Tables 2: Supporting Information Table S1).
| Trait | Evidence for phylogenetic signal | Evidence for correlated evolution | Key citations |
|---|---|---|---|
| SL | Yes | LT | Liu et al. (2012); Zhang et al. (2014); Li et al. (2015) |
| SA | Yes | Zhang et al. (2012) | |
| SD | Yes | VD, LT | Li et al. (2015); Liu et al. (2015, 2019); Zhang et al. (2018) |
| SPI | Mixed | Savage & Cavender-Bares (2012); Liu & Osborne (2015); Liang et al. (2019); Liu et al. (2019) | |
| gs,max | No | Kleaf, VD, KL | Hao et al. (2011); Fu et al. (2012); Li et al. (2015); Liu et al. (2015); Liu & Osborne (2015); Scoffoni et al. (2016) |
| E | No | Liu et al. (2015); Liu & Osborne (2015) | |
| Kleaf | No | gs,max | Liu & Osborne (2015) |
| VD | Yes | LA, gs,max | Zhang et al. (2012, 2014); Blonder & Enquist (2014); Blonder et al. (2017) |
| P50leaf | Insufficient data | Kleaf | Yan et al. (2020) |
| P80leaf | Insufficient data | ||
| LA | Mixed | VD | Zhang et al. (2012); Lasky et al. (2014); Gallaher et al. (2019); de Paula et al. (2020) |
| LT | Yes | SL, SD, VD | Kembel & Cahill (2011); Yang et al. (2014); Zhang et al. (2014); Li et al. (2015) |
| Ψtlp | No | P50stem | Fu et al. (2012); Savage & Cavender-Bares (2012); Liu et al. (2015); Liu & Osborne (2015); Bartlett et al. (2016) |
| RWCtlp | Insufficient data | ||
| KS | Yes | WD, P50stem | Fu et al. (2012); De Guzman et al. (2017); Sanchez-Martinez et al. (2020) |
| KL | Mixed | gs,max | Fu et al. (2012); Sedio et al. (2012); Sanchez-Martinez et al. (2020) |
| P50stem | Yes | Ψtlp, KS | Fu et al. (2012); Sanchez-Martinez et al. (2020) |
| WD | No | KS | Hacke et al. (2009); Fu et al. (2012); Zhu et al. (2013) |
| HV | Yes | LA, KS | Bhaskar et al. (2007); Fu et al. (2012); Liu et al. (2015); Bruy et al. (2018); Sanchez-Martinez et al. (2020) |
- Some authors have indirectly addressed the role of evolutionary relationships on these traits, but they did so by using other statistical methods (e.g. QVI or linear mixed models using taxonomy as a factor). These data are not included in this table. Trait abbreviations: stomatal length (SL), stomatal area (SA), stomatal density (SD), stomatal pore index (SPI), maximum stomatal conductance (gs,max), transpiration rate (E), leaf hydraulic conductance (Kleaf), vein density (VD), leaf cavitation resistance (P50leaf and P80leaf), leaf area (LA), leaf thickness (LT), turgor loss point (Ψtlp), relative water content at turgor loss point (RWCtlp), sapwood-specific stem hydraulic conductivity (KS), leaf-specific stem hydraulic conductivity (KL), stem cavitation resistance (P50stem), wood density (WD), and Huber value (HV).
1. Stomatal characteristics and conductance to water vapor
Stomatal conductance (gs) refers to the conductance to water vapor transfer from substomatal cavities to outside air and is calculated as the rate of transpiration divided by the leaf-air vapor-pressure difference. It is a fundamental measure of the physiological response of plants to water availability and demand. Notably, the maximal rate of diffusion is mediated by structural characteristics, namely the maximum size of the stomatal aperture and the number of stomata per unit of leaf area (LA).
Studies analyzing stomatal characteristics have consistently found support for phylogenetic signal in size and density of stomata, but not necessarily for maximum gs (gs,max) or transpiration rate (E), although it is not always clear whether the maximum possible values of gs or E have been obtained. Field studies, such as the study of 85 dicotyledonous species from tropical and subtropical forests by Li et al. (2015), have revealed evidence for a phylogenetic signal in SL and stomatal density (SD), but not for gs,max as calculated according to Franks & Beerling (2009). Examining six deciduous and six evergreen angiosperm trees from a tropical dry karst forest, Fu et al. (2012) also found no evidence for phylogenetic signal in gs,max. Common garden and glasshouse studies have demonstrated no evidence for phylogenetic signal in either gs,max or E across 27 species in Magnoliaceae and 33 species of grasses (Liu et al., 2015; Liu & Osborne, 2015). However, in a common garden experiment with 30 tropical fern species, Zhang et al. (2014) reported a phylogenetic signal for both gs,max and E, as well as SL and SD. Similarly, evidence for a significant phylogenetic signal has been observed in nighttime stomatal conductance in a common garden (Yu et al., 2019). Overall, there is more consistent evidence for phylogenetic signal in various clades for structural stomatal traits such as length, area, and density than for gs,max (Liu et al., 2012; Zhang et al., 2012, 2014; Li et al., 2015; Liu & Osborne, 2015).
The absence of evidence to date for phylogenetic signal in gs,max in natural settings could be because of environment effects on trait expression, despite the fact that gs,max is usually measured during the wet season and early morning hours. For example, even if a phylogenetic signal in gs,max does exist among certain taxa, high phenotypic plasticity in response to environmental conditions may obscure the results (E. Ávila-Lovera et al., unpublished data). Only one common garden study showed a significant phylogenetic signal in gs,max (Zhang et al., 2014), and compared with the greater number of studies, field and common garden, that did not find phylogenetic signal, we suggest that more common garden studies would be useful. Nonetheless, low-to-no phylogenetic signal of gs,max indicates that it is a labile trait (Scheiner, 1993) that can vary considerably with habitat conditions. Another explanation could be that variation in other traits counterbalances variation in SL and SD in such a way that similar values of gs,max can be achieved with a different combination of SL and SD and that gs,max is indeed under selection due to climate.
Despite limited evidence for phylogenetic signal in the magnitude of maximum stomatal conductance, evidence from comparative methods supports the hypothesis that gs,max has evolved in a correlated fashion with leaf hydraulic conductance (Kleaf) in multiple taxa. For instance, species with high gs,max also have high maximum Kleaf (Kleaf,max) across 30 species of Viburnum grown in a common garden (Scoffoni et al., 2016). Stomatal conductance and density have also been found to have evolved in correlation with VD, the ratio of total vein length to LA (mm−1), whereby species with high gs,max or stomatal density also had high VD (Brodribb et al., 2007; Sun et al., 2014; Zhang et al., 2014, 2018; Schneider et al., 2017). Taken together, having high VD and SD are both conducive to high gs,max, and high gs,max is only advantageous if Kleaf,max is also high, as the water flow through the leaf is restricted by the lowest conductance of the water path. High maximum stomatal and hydraulic conductance is expected to be favored in wet habitat species where species can capitalize on high water availability. Examining the supplemental data (table S2) in Scoffoni et al. (2016), we found this to be true. Environmental control of gs,max is then especially important in species from seasonally dry ecosystems, where gs,max should match water availability and demand to avoid drought stress and mortality.
2. Leaf hydraulic conductance and venation
Leaf hydraulic conductance, the rate at which water moves through the leaf per unit area per unit of water potential gradient, is a measure of the hydraulic efficiency of leaves, which affects both leaf water potential and gas exchange (Scoffoni et al., 2016, 2018). Maximum values of Kleaf typically range from 0.76 to 49 mmol m−2 s−1 MPa−1 (65-fold) depending on growth form (crop herbs, tropical and temperate woody angiosperms, and temperate woody gymnosperms) and leaf habit (Sack & Holbrook, 2006). This large variation is typically related to differences in leaf venation and leaf thickness (LT).
Despite a large interest in understanding the factors that influence Kleaf,max (Sack & Holbrook, 2006), we are aware of only one study that has tested possible phylogenetic signal on Kleaf,max. A survey of Kleaf,max in 33 grass species growing in a common garden revealed no evidence for phylogenetic signal (Liu & Osborne, 2015). It is nonetheless possible to infer phylogenetic patterns in Kleaf,max by studying its underlying traits.
Among the traits directly related to Kleaf,max is VD, as greater space destined for xylary water movement influences the total conductance of a leaf. There is evidence for phylogenetic signal in VD in many studies that span from species of a single genus (Paphiopedilum, Orchidaceae, Zhang et al., 2012), temperate angiosperms (Blonder & Enquist, 2014), to tropical ferns and angiosperms (Zhang et al., 2014; Blonder et al., 2017). Only one study found no evidence for phylogenetic signal in VD among 85 dicot species from five tropical and subtropical forests (Li et al., 2015). The latter study concluded that high variation in VD relative to other traits may be the reason for the absence of a phylogenetic signal (Li et al., 2015). The evidence for phylogenetic signal in VD in most studies indicates that shared evolutionary history plays an important role in determining VD values. Interestingly, finding phylogenetic signal in VD may vary depending on the taxonomic scale of interest (e.g. among vs within families), a nuance that is not often explored (Blonder et al., 2017).
Most of the relationships between maximum Kleaf, VD, and leaf cavitation resistance (P50leaf and P80leaf) across a diverse group of species are strong (Sack & Frole, 2006; Brodribb et al., 2007; Brodribb & Jordan, 2008; Johnson et al., 2009; Blackman & Brodribb, 2011; Xiong et al., 2018), and although one may assume those relationships are not affected by evolutionary relationships (Sack & Frole, 2006; Brodribb et al., 2007), tests that formally address the role of shared evolutionary history of the species have the potential to improve our understanding of correlated evolution. For example, the strength of relationships between traits can decrease or become nonexistent when phylogenetic information is incorporated into linear models. For example, VD and leaf size are strongly negatively related to one another using Bayesian PGLS (Schneider et al., 2017), but weakly so when using species data (analyses not shown); this suggests correlated evolution between VD and leaf size in the family Ochnaceae, but a weak association between traits in contemporary species. Kleaf,max is positively related to P50leaf (Yan et al., 2020), with high values of Kleaf,max being accompanied by high values of P50leaf or low hydraulic embolism resistance. This suggests a trade-off between leaf hydraulic efficiency and safety, just as the one widely known for stems. Given the limited studies on this topic, the extent to which these relationships demonstrate correlated evolution with one another remains unresolved.
More studies on the evolution of Kleaf,max and associated traits across a large phylogeny are necessary to draw robust conclusions on the relative role of phylogeny and climate in explaining those traits. One way this can be accomplished is by analyzing large datasets of Kleaf,max values extracted from the TRY Plant Trait Dataset (Kattge et al., 2020) and by examining possible correlates to explore the patterns of trait associations. Leaf hydraulic conductance is the limiting factor for the rate of leaf water loss (Zimmermann, 1978; Sack & Holbrook, 2006), with leaves acting as safety valves (Pivovaroff et al., 2014). Kleaf,max is thus a key trait that determines leaf water relations, as well as leaf photosynthetic capacity (Scoffoni et al., 2016), and is therefore crucial for a better understanding of plant water use in a drying climate. Although measurements of Kleaf,max are time-consuming and difficult, more comprehensive studies would be of interest.
3. Leaf size, shape, and thickness
The size, shape, and thickness (e.g. succulence) of leaves affect water loss by mediating leaf energy balance (i.e. leaf temperature), and succulence also affects capacitance and osmotic pressure (Gotsch et al., 2015). Large and thick leaves require greater water loss to cool down than small and thin leaves (Gates, 1968; Schuepp, 1993). Leaf shape, which can be measured as the leaf length-to-width ratio (LWR), or as a dissection index (DI; how lobed a leaf is), also affects water loss through leaf energy balance (Nicotra et al., 2011).
The evidence for phylogenetic signal of leaf size is mixed. On the one hand, in a study of 25 Atractocarpus species, LA was found to have a significant phylogenetic signal as measured by both Blomberg's K and Pagel's λ (Bruy et al., 2018). Similar results were found in a study of 206 species sampled from herbaria (Gallaher et al., 2019) and in a study of 83 species from two dry neotropical inselbergs (de Paula et al., 2020). Leaf area was ‘conserved’ (as described in Ackerly, 2004) in three California chaparral families (Ericaceae, Rhamnaceae, and Rosaceae), as measured by low values of a quantitative convergence index, which indicates that closely related species were phenotypically more similar than distantly related ones (Ackerly, 2004). Similarly, the LA of species from different forest communities along an elevation gradient was strongly determined by family, regardless of environmental conditions experienced by the individual species (Costa et al., 2018).
On the other hand, in contrast with the evidence for phylogenetic signal of LA among families, there is less evidence for phylogenetic signal in LA when species within single plant families are evaluated (Zhang et al., 2012; Lasky et al., 2014; Liu & Osborne, 2015), except for the Atractocarpus study above (Bruy et al., 2018). This may be true because plant families are a taxonomic classification that groups taxa with more similar characteristics, including leaf size and shape. It seems that the level of taxonomic hierarchy analyzed affects the ability to detect phylogenetic signal, if it exists (see Section VI). Future research should evaluate whether phylogenetic signal is detected in both a large phylogeny and smaller clades within the larger phylogeny, as proposed in recent reviews (Losos, 2008; Münkemüller et al., 2012; Molina-Venegas & Rodríguez, 2017).
There is only limited research on phylogenetic signal of leaf shape, although there have been studies on traits such as leaf DI (Santiago & Kim, 2009) and leaf LWR (Carlson et al., 2011; Mitchell et al., 2015; Gallaher et al., 2019). Leaf DI showed no phylogenetic signal in 15 species within the woody Sonchus alliance when grown in a common garden (Santiago & Kim, 2009); closely related species had divergent values of DI, which was associated with habitat specialization. However, LWR demonstrated a significant phylogenetic signal among 206 species of Poaceae, despite consistent differences between the habitats from which they were sampled (i.e. open, forest margin, and forest understory) (Gallaher et al., 2019). These later results indicate that both phylogeny and light availability have important roles in explaining leaf shape in Poaceae.
Evidence for a significant phylogenetic signal in LT comes from studies in 76 species of Poaceae (Kembel & Cahill, 2011), 17 species of Paphiopedilum (Zhang et al., 2012), 265 species of tropical trees (Yang et al., 2014), 30 tropical ferns (Zhang et al., 2014), and 85 species of dicots (Li et al., 2015). In contrast, Liu & Osborne (2015) found no evidence for phylogenetic signal in LT of 33 species of C4 grasses. Additional support for a phylogenetic signal in LT comes from studying leaf succulence, which is typically determined as the amount of water in a leaf divided by its area. Phylogenetic signal in succulence has been observed in 83 species from two dry Neotropical inselbergs (de Paula et al., 2020). In spite of extensive research on the origin of the cactus-like form, leaf venation patterns in succulent forms, and the evolution of leaf succulence and photosynthetic pathways (Edwards & Donoghue, 2006; Ogburn & Edwards, 2013; Edwards, 2019), formal tests of phylogenetic signal in leaf succulence are surprisingly limited. However, we would expect to find evidence for phylogenetic signal in leaf succulence because it has evolved in correlation with Crassulacean acid metabolism (CAM) (Edwards, 2019).
Leaf area has been measured in a large number of studies, mainly as a correlate for traits of the leaf economic spectrum (leaf life span, specific leaf area (SLA), nitrogen concentration, and mass-based photosynthetic rate) (Wright et al., 2004), but it has rarely been found to evolve in correlation with plant–water relation traits besides the relationships with VD described earlier. In a study of the California chaparral, Ackerly (2004) found weak correlations between traits of the leaf economic spectrum and LA, indicating that leaf economic traits co-evolved irrespective of leaf size. More recent research indicates that hydraulic traits may have evolved independent of leaf economic traits (Sack et al., 2013), and therefore we would not expect correlated evolution of leaf size and shape with other water relations traits.
In summary, there is strong support for phylogenetic signal in leaf size, shape, and thickness, indicating that such traits are more similar among close relatives and could be only minimally responding to the environment, with some exceptions such as the example of plant genera that have greater leaf size in forests than shrublands (Power et al., 2019). These structural traits show low variation or respond more slowly to changes in climate and hence can be considered less labile than physiological traits.
4. Leaf pressure–volume curve characteristics
Pressure–volume (PV) curves describe the relationship between water potential (Ψ) and relative water content (RWC) and are most commonly studied in leaves, but also in stems and roots. The relationship between Ψ and RWC is then used to determine a number of traits including the water potential at turgor loss (Ψtlp), the relative water content at turgor loss (RWCtlp), the modulus of elasticity (ε), and capacitance (C). All of these traits reflect different functional strategies used by plants frequently experiencing water-deficit stress. For instance, species from deserts and Mediterranean-type climate ecosystems usually have more negative values of Ψtlp than species from more mesic environments, indicating that they can maintain turgor at higher water deficits than mesic species (Bartlett et al., 2012).
The water potential at turgor loss is the best-studied PV curve trait in a phylogenetic context. Most studies have found no evidence for phylogenetic signal in Ψtlp (and occasionally ε), as measured by a combination of Blomberg's K and Pagel's λ, across a number of different taxa in different environments (Fu et al., 2012; Savage & Cavender-Bares, 2012; Liu et al., 2015; Liu & Osborne, 2015; Bartlett et al., 2016). However, there was evidence for phylogenetic signal in Ψtlp of 19 Psychotria species, whereby Blomberg's K was significantly different from zero (Sedio et al., 2012). Notably, all of these studies were relatively restricted in the number of species investigated (c. < 50). Similarly, there is no evidence for phylogenetic signal in Ψtlp among populations of Castanopsis fargesii (Liang et al., 2019). Given the importance of Ψtlp in explaining drought tolerance, we performed an analysis of phylogenetic signal in Ψtlp for 527 species in the TRY database (Kattge et al., 2020) and found no evidence for phylogenetic signal using Blomberg's K (Table 3; Fig. 3). Taken together, these results indicate that Ψtlp may be a labile trait that changes in response to the environment. We expect that it would be more likely to find phylogenetic signal in traits that reflect structure (e.g. ε) than in traits that reflect physiology (e.g. osmotic potential), as structural traits may take longer to change in evolutionary time than physiological traits; however, further studies are necessary.
| Trait | Number of species | Blomberg's K | P | Pagel's λ | P |
|---|---|---|---|---|---|
| Structural traits | |||||
| SD | 806 | 0.060 | 0.001 | 0.846 | 1.57 × 10−80 |
| VD | 452 | 0.080 | 0.001 | 0.885 | 2.19 × 10−13 |
| LA | 2082 | 0.053 | 0.004 | 0.998 | 3.08 × 10−271 |
| LT | 5597 | 0.012 | 0.071 | 0.649 | 2.06 × 10−225 |
| HV | 377 | 0.048 | 0.203 | 0.340 | 8.92 × 10−6 |
| Physiological traits | |||||
| gs,mean | 2793 | 0.026 | 0.001 | 0.719 | 1.04 × 10−64 |
| gs,max | 2743 | 0.004 | 0.150 | 0.504 | 8.99 × 10−61 |
| Kleaf | 24 | 0.135 | 0.869 | 0.000 | 1.00 |
| Ψtlp | 527 | 0.011 | 0.142 | 0.690 | 7.04 × 10−13 |
| RWCtlp | 19 | 0.501 | 0.006 | 0.894 | 5.27 × 10−3 |
| KS | 694 | 0.054 | 0.001 | 0.488 | 9.12 × 10−15 |
| KL | 524 | 0.028 | 0.550 | 0.000 | 1.00 |
| P50stem | 740 | 0.055 | 0.001 | 0.857 | 2.84 × 10−93 |
- Traits such as SA, SPI, and P50leaf are not in TRY, and SL had only four observations; hence, those traits are not included in this list. P values come from statistical tests of the null hypotheses that K and λ are different from zero. Bold values indicate significant phylogenetic signal at P < 0.05. Trait abbreviations: stomatal density (SD), vein density (VD), leaf area (LA), leaf thickness (LT), Huber value (HV), mean stomatal conductance (gs, mean), maximum stomatal conductance (gs,max), leaf hydraulic conductance (Kleaf), turgor loss point (Ψtlp), relative water content at turgor loss point (RWCtlp), sapwood-specific stem hydraulic conductivity (KS), leaf-specific stem hydraulic conductivity (KL), and stem cavitation resistance (P50stem).
In some studies, an ANOVA was used with a taxonomic rank (e.g. species or family) as a predictor variable in lieu of tests for phylogenetic signal. For example, a nested ANOVA with predictors ‘genus’ and ‘distribution nested within genus’ was used in a study of 24 tree species from forest habitats of contrasting rainfall seasonality and found that both genus and distribution (habitat) had a strong effect on Ψtlp and osmotic pressure at full turgor (πft) (Baltzer et al., 2008).
Despite an apparent lack of phylogenetic signal in Ψtlp, some other leaf traits have evolved in a correlated fashion with Ψtlp. Using PICs, Ψtlp was found to be negatively related to the leaf maximal tensile strength and the tensile modulus of elasticity of 17 Mediterranean shrub species; these relationships were driven by leaf density (Méndez-Alonzo et al., 2019). Similarly, leaf dry matter content has been found to be negatively correlated with Ψtlp in 27 species in the Magnoliaceae using a PGLS model (Liu et al., 2015). These results indicate that tougher leaves have lower (more negative) Ψtlp, which suggests a conservative water strategy. These relationships between structural traits and Ψtlp indicate coordination between tissue density/dry matter content and the ability for leaves to sustain low water potentials before losing turgor. Similar coordination occurs between gs,max and Ψtlp, whereby stomatal sensitivity to water deficit (Ψgs50, the water potential at which stomatal conductance decreases to 50% of the maximum value) is negatively related to Ψtlp, indicating a trade-off between water status and safety (Henry et al., 2019). These results are consistent with structural and physiological traits involved in controlling leaf responses to dehydration evolving together over time, due to shared selection, evolutionary constraint, or both. Such correlated evolution is advantageous, as coordinated traits allow for a coordinated whole-plant response to different water availability regimes.
5. Stem characteristics: hydraulic conductance and cavitation resistance
The efficiency of water transport through stems is important for maintaining the metabolic functioning of leaves. Stem hydraulic conductivity (Kh) and the sapwood and LA standardized versions (KS and KL, respectively) are measures of such efficiency of water transport. However, stems with high KS are usually more vulnerable to cavitation (Gleason et al., 2016; De Guzman et al., 2017, 2021). Hence, a trade-off exists between hydraulic efficiency and safety; safety is usually described by the water potential at which 50% of hydraulic conductivity is lost (P50stem). Although this trade-off appears to be weak at a global scale (Gleason et al., 2016), it is important in determining plant–water strategies at local and regional scales (Maherali et al., 2006; Jacobsen et al., 2007; De Guzman et al., 2017; Skelton et al., 2021).
There is indirect evidence for a phylogenetic signal in P50stem inferred from the mapping of this trait onto phylogenetic trees. For example, P50stem shows phylogenetic signal in a study of 23 conifer species from Australia, with species from clades in drier climates consistently having more negative P50stem values (Larter et al., 2017). Direct evidence of phylogenetic signal in P50stem comes from a study of 24 liana and 27 tree species from the wet tropics (van der Sande et al., 2019). There was less evidence for phylogenetic signal in P50stem among a more diverse group of plants from Mediterranean, desert, and tropical rain forest environments; instead, there was evidence for convergent evolution, as measured by quantitative convergence indices (Maherali et al., 2004). In species typically found in chaparral communities, P50stem had highly negative values, despite the fact that they belonged to distantly related genera and families (Maherali et al., 2004).
Studies directly testing for phylogenetic signal in KS are limited. While KS showed no significant phylogenetic signal in one study (van der Sande et al., 2019), a recent meta-analysis encompassing 2027 species found a significant phylogenetic signal in KS and other related hydraulic traits such as KL, P50stem, and HV (Sanchez-Martinez et al., 2020). These results indicate that traits related to wood anatomy are structured by both shared evolutionary history and mean annual precipitation (MAP) in angiosperms (Gleason et al., 2016).
Given the coordination among multiple stem traits, including KS, cavitation resistance (e.g. P50stem), and wood density (WD), there has also been interest in testing for correlated evolution among these traits. Phylogenetically independent contrasts of species mean KS and P50stem led to no evidence for evolutionary correlations between the two (Maherali et al., 2004, 2006; Bhaskar et al., 2007; Jacobsen et al., 2007; Zhu et al., 2013). Thus, either the trade-off between hydraulic efficiency and safety is weak, or these two traits have evolved independently of one another (Maherali et al., 2004; Hacke et al., 2009). However, in a study evaluating 12 species of trees and lianas from 11 different families, the relationship between KS and P50stem held after accounting for evolutionary relationships using PICs (De Guzman et al., 2017); in this case, the observed hydraulic safety–efficiency trade-off seems to have evolved independently in multiple families. Similarly, even after accounting for evolutionary relationships, WD was negatively related to KS in species from both arid lands (desert, chaparral, and coastal sage scrub) and subtropical forests, suggesting possible correlated evolution (Hacke et al., 2009; Zhu et al., 2013). Detecting correlated evolution among these hydraulic traits provides evidence for the coordination of these plant–water relation traits.
6. Huber value
The HV is the ratio of sapwood area (or sometimes stem cross-sectional area) to LA distal to the stem and is a measure of the ability of stems to support leaf water demand (Huber, 1928; Tyree & Ewers, 1991). More recently, HV has also been represented as its inverse: the ratio of LA to sapwood area (LA : SA), but we use HV consistently herein.
The HV is predominantly a structural trait that links stem and leaf hydraulics and may therefore be expected to show phylogenetic signal. Indeed, phylogenetic signal in HV has been found in multiple studies (Bhaskar et al., 2007; Hao et al., 2008; Liu et al., 2015; Bruy et al., 2018; Sanchez-Martinez et al., 2020), which indicates similar hydraulic architecture among close relatives. We are aware of only one study that did not find evidence for phylogenetic signal in HV: a meta-analysis of tropical montane cloud forests (Eller et al., 2020); one reason for this may be incomplete phylogenetic sampling as the study is a compilation of different studies.
With respect to correlated evolution, HV has been found to be negatively correlated with both LA and KS (Edwards, 2006; Bhaskar et al., 2007). Hence, other traits in which there is phylogenetic signal, such as LA and KS, are more likely to have evolved in a correlated fashion with HV, which has implications for whole-plant functioning, especially in processes related to water transport and balance. For example, in habitats with low water availability, having low LA and KS coupled with high HV can ensure appropriate water delivery to leaves and sustain leaf stomatal opening and photosynthesis.
IV. Phylogenetic signal in structural vs physiological traits
Among all the traits analyzed in this review, it seems that structural traits were more frequently found to have a significant phylogenetic signal than physiological traits. For example, traits related to the size and number of stomata such as SD, SL, and SA, which influence leaf water loss, had higher values of Blomberg's K and/or Pagel's λ than gs,max (Fu et al., 2012; Zhang et al., 2012, 2014; Li et al., 2015; Liu et al., 2015). Similarly, traits related to the size and shape of leaves (Kembel & Cahill, 2011; Yang et al., 2014; Bruy et al., 2018; Costa et al., 2018; Gallaher et al., 2019) and stem hydraulics (Fu et al., 2012; Liu et al., 2015; Rungwattana et al., 2018; Sanchez-Martinez et al., 2020) showed significant phylogenetic signal in contrast to Ψtlp. As previously noted, phylogenetic signal is a pattern and multiple evolutionary processes can lead to a similar phylogenetic signal. However, the observation that Ψtlp is a trait that is strongly linked to the environment suggests that environmental selection may be driving trait variation in such a way that close relatives resemble each other less than expected, that is, they have low or no phylogenetic signal. Furthermore, the observation that there is more evidence for phylogenetic signal in structural traits than in physiological traits may be explained by: (1) changes in structure evolving more slowly than changes in function; (2) structural traits having lower variation than physiological traits; and (3) structural traits being more heritable than physiological traits. Another possible explanation is that structural traits are often measured in more standardized ways (Pérez-Harguindeguy et al., 2013) than physiological traits. The extent to which artifacts affect the likelihood of detecting phylogenetic signal in hydraulic traits would be of interest in future studies.
Despite the general pattern we found by reviewing each article (Table S1), analyses of phylogenetic signal using data extracted from TRY (Kattge et al., 2020) yield slightly different results (Table 3). First, there are differences in the ability to detect signal and the magnitude of the signal when estimated by Blomberg's K vs Pagel's λ for structural traits such as LT and HV. Second, most structural traits were indeed found to have a significant signal using one or both indices, but other physiological traits such as gs,max and KS also showed significant signal using both indices. Third, Ψtlp showed no phylogenetic signal using Blomberg's K, but it had a significant signal using Pagel's λ. Despite these seemingly contradictory results, it seems that both indices of Blomberg's K and Pagel's λ are greater in structural than in physiological traits, except for RWCtlp. This partly supports the results found when analyzing each article individually and highlights that sampling from bigger phylogenies may have an effect on our ability to detect phylogenetic signal (see Section VI).
Studies on insects, fish, birds, and mammals also show higher phylogenetic signal in morphological than in physiological traits (Blomberg et al., 2003). Other studies on animals have also demonstrated that heritability is low in behavioral and physiological traits compared with morphological and life-history traits (Mousseau & Roff, 1987; Weigensberg & Roff, 1996; Stirling et al., 2002), suggesting that high heritability in structural/morphological traits is key in determining low lability.
A few studies of plants have examined the link between heritability and phylogenetic signal. These studies point toward heritability as a likely cause for the differential phylogenetic signal in structural vs physiological traits. For example, one study of narrow-sense heritability, the ratio of additive genetic variance to the total phenotypic variance, which examined photosynthetic traits (maximum quantum yield of PSII, maximum photosynthetic rate, gs,max, and water-use efficiency (WUE)), rosette size, and SLA of Lobelia cardinalis, found that heritability was higher in rosette size and SLA than in photosynthetic traits (Caruso et al., 2005). Similarly, broad-sense heritability, the ratio of total genetic variance to the total phenotypic variance, was higher in petal size and height than in flowering time in Clarkia xantiana (Gould et al., 2014). However, one study did find higher realized heritability in WUE than in height in Pinus halepensis (Suárez-Vidal et al., 2021). These results provide some evidence for higher heritability in structural than physiological traits in plants, but additional research is needed to further support this hypothesis.
V. Trait–environment relationships
How plant structure and function vary in relation to the environment is a central question of plant ecophysiology. Classical statistical models frequently used to detect trait–environment relationships assume that data points are independent of each other (among other things); however, our review reveals that many plant–water relation traits demonstrate phylogenetic signal (i.e. trait similarity among closely related species). This indicates that using models that assume independence is not statistically appropriate to study these type of data, and formally incorporating the shared evolutionary history of species into these linear models (phylogenetic comparative methods) is necessary to fully understand the evolution of traits in response to the environment. Much of the current interest in studying traits in relation to the environment is to predict plant responses to ongoing anthropogenic climate change. The expectation that plant traits will shift with climate change is well-founded, but the degree to which traits can vary with climate depends on trait lability (acclimation potential) and the evolutionary history of the species in question. For example, that low P50stem values occur in species from temperate mesic regions suggests that phylogeny constrains, at least partly, the adaptive evolution of the trait (Maherali et al., 2004).
The study of the effects of environment on plant–water relation traits that we observed in contemporary taxa while accounting for shared evolutionary history has become increasingly common. For example, terrestrial bromeliads exhibit high variation in leaf shape and area, and this variation is usually associated with the environment they live in: species with high LA and low VD are restricted to moist, aseasonal environments, despite them belonging to the same family, and hence being closely related to each other (Males, 2017). Comparing closely related species pairs of multiple families from contrasting habitats (forest vs shrubland), Power et al. (2019) concluded that environment was a stronger predictor of LA and LT than shared evolutionary history, that is, LA and LT were always higher in forest than in shrubland regardless of species identity. In both these instances, environment played a stronger role in structuring LA than phylogeny did, despite the evidence for phylogenetic signal in LA we identified in this review.
Similarly, KS varies predictably with MAP (Gleason et al., 2016), indicating that species with high hydraulic conductivity perform better in sites with high water availability and often low precipitation seasonality. This may be a strategy by which species can capitalize on high water availability by assimilating more carbon, as leaf photosynthesis usually scales with stem hydraulic conductivity (Brodribb & Feild, 2000; Brodribb et al., 2002; Santiago et al., 2004). Similarly, values of P50stem are significantly related to water scarcity, with species in clades occupying Mediterranean forests, woodland, and scrubs having more negative P50stem values than clades occupying more mesic habitats (Larter et al., 2017). These results point toward the importance of hydraulic traits in determining current species distributions (Costa-Saura et al., 2016; Cosme et al., 2017; Li et al., 2018; Fontes et al., 2020).
Along with evidence for low lability (high signal) in HV, environment seems to contribute to currently observed values. For example, high HVs are usually found in short shrubs with small leaves, low SLA, and low KS in arid environments, whereas low HV values are usually found in tall trees with large leaves, high SLA, and high KS in moist environments (Mencuccini et al., 2019). Similarly, HV has been found to be negatively related to soil moisture in three different lineages of oaks (Cavender-Bares & Holbrook, 2001) and in 51 California coast range angiosperms (Preston et al., 2006). Such observations would indicate that, among different species, HV is strongly influenced by the physical environment.
While there is some evidence for phylogenetic signal in many of the structural plant–water relation traits highlighted in this review, it is also important to note that these traits, although less labile, can often vary as a function of climate. The contribution of climate to plant trait variation may be site- and clade-specific, as some environments (especially those with seasonal water deficits) may impose strong selection on traits regardless of shared evolutionary history. Additional insights into how evolutionary relationships affect trait–environment relationships will further enhance our foundational understanding of plant–water relations, as well as our understanding of how those relationships may change in the future.
VI. The size of the phylogeny
It appears that the size of the phylogenetic tree affects the ability to detect phylogenetic signal in plant–water relation traits, as has been highlighted in relevant reviews (Münkemüller et al., 2012; Kamilar & Cooper, 2013). Some studies may have insufficient taxa to unravel a phylogenetic signal, as larger clades or multiple clades that include more diverse taxa are more likely to yield phylogenetic signal where one exists (Freckleton et al., 2002; Münkemüller et al., 2012). One way to test this hypothesis is to model the probability of detecting phylogenetic signal as a function of the number of species included in the analysis. Using 39 studies, we ran a generalized linear model assuming binomial distribution and found a significant effect of the number of study species on the probability of detecting phylogenetic signal (deviance = 16.648, P < 0.0001). We also found that the mean number of species was greater in studies that found phylogenetic signal than in studies that did not find phylogenetic signal (two-tailed independent t-test; t71 = 4.116, P < 0.001). This suggests that results from studies where small phylogenetic trees were examined and no phylogenetic signal has been found should be interpreted with caution as absence of evidence is not evidence of absence. In addition to the number of species, the taxonomic scale of the study (e.g. within family vs among families) may also determine the ability to detect a phylogenetic signal and deserves further attention moving forward. One study found that better intrafamilial tree resolution affects Blomberg's K values (Seger et al., 2013), with K values decreasing with greater intrafamilial resolution. However, running a generalized linear model in which type of phylogeny (community, family, genus, species, and intraspecies) was included as a covariate of species number showed that the type of phylogeny does not have an effect on the likelihood of detecting phylogenetic signal (deviance = 7.726, P > 0.05).
Even though large studies of plant structure and physiology are logistically demanding and time-consuming, they are necessary to increase the probability of finding a phylogenetic signal where one exists. Databases such as TRY (Kattge et al., 2020) make large data sets available on existing key traits and help test hypotheses of phylogenetic signal, correlated evolution, and trait–environment relationships. With respect to trait–trait correlations, sampling traits in species randomly selected from large phylogenies, rather than focusing on a single life form or interesting study group, can recover the patterns of correlated evolution found in the larger phylogeny (Ackerly, 2000).
VII. Conclusions
We reviewed the phylogenetic signal of plant–water relation traits, as these traits are important in determining plant ecological strategies, especially under drought conditions, and therefore affect plant species' survival. We found that many traits showed significant phylogenetic signal; however, this was predominantly true for structural rather than physiological traits. We speculate that changes in structure evolve more slowly than physiological traits or that structural traits are more constrained in their possible variation than physiological traits, leading to more lability in physiological vs structural traits.
Given that structural traits, such as leaf stomatal density, VD, LA, and LT, typically have high phylogenetic signal, we can use them to estimate mean trait values of species that are difficult to sample (e.g. rare species and/or species in highly diverse ecosystems). While directly assessing species' traits is necessary to advance our understanding of plant functioning, especially in the context of anthropogenic climate change, being able to infer species trait values from phylogeny alone can be fruitful in multiple contexts.
In order to expand our understanding of the phylogenetic signal and correlated evolution of plant–water relation traits in response to climate drivers, more studies are needed that explicitly incorporate phylogenetics in the testing of hypotheses about trait–trait and trait–environment coordination. Additionally, models that seek to understand the responses of plants in large, highly diverse ecosystems (e.g. tropical wet forests) to changes in climate should include phylogenetic information, as some traits may be related to current climate conditions (e.g. P50stem of deciduous angiosperms and MAP) but had not evolved in a correlated fashion (Maherali et al., 2004).
To advance our understanding on the role of shared evolutionary history on trait values, we suggest that future studies: (1) use the well-established methods described in this review (e.g. Blomberg's K and Pagel's λ) rather than linear models (e.g. taxonomic ANOVAs) to estimate phylogenetic signal; (2) explore the possible effect of the size of the phylogeny on detecting phylogenetic signal by randomly removing taxa from taxa-rich phylogenies to artificially create smaller ones; and (3) in the case of species grown in common garden settings, also include trait data of the species in their natural habitat. We believe that incorporating the shared evolutionary histories of species in ecophysiological studies can help us answer questions related to the origin of variation in plant–water relation traits and how shared evolutionary history informs correlated evolution and trait–environment relationships.
Acknowledgements
We thank Ted Garland and Ralph Albuquerque for helpful discussions on methods of comparative biology. We are very grateful to Jennifer Funk, Marjorie Webber, and Finn Piatscheck for their helpful comments on an earlier version of this manuscript. We thank Mark Olson, David Ackerly, and Robert Skelton and the handling editor for constructive reviews. This review was supported by NSF Dimensions grant DEB-1737878 to GRG, and the Smithsonian Tropical Research Institute Earl S. Tupper Postdoctoral Fellowship to EA-L.
Author contributions
GRG and E-L planned and designed the review. EA-L synthesized the literature and analyzed the data. EA-L wrote the manuscript with significant inputs from KW and GRG.
Open Research
Data availability
All the data used in this work are publicly available at TRY Plant Trait Database at https://www.try-db.org/.
