Seedling response to environmental variability: The relationship between phenotypic plasticity and evolutionary history in closely related Eucalyptus species
Abstract
PREMISE OF THE STUDY:
Phenotypic plasticity is an important means through which organisms cope with environmental variability. We investigated seedling plasticity in the green ash eucalypts within a phylogenetic framework to examine the relationship between plasticity and evolutionary history. The green ashes are a diverse group, which include the tallest flowering plant in the world (Eucalyptus regnans) and a rare mallee less than 1 m tall (E. cunninghamii).
METHODS:
Seedlings of 12 species were exposed to high and low nutrient and water availability in a factorial experiment. Leaf trait and total plant plasticity were evaluated using the phenotypic plasticity index. A phylogeny of the species was estimated using genome-wide scans.
KEY RESULTS:
We found significant differences in functional traits across species, growth forms, and substrates in response to changes in resource availability. Many traits (e.g., leaf width) were highly plastic for most species. Interspecific differences in leaf-level plasticity was significant, however plasticity was not correlated with phylogeny. Species with broader environmental niches had higher leaf-level plasticity than species with narrower environmental ranges.
CONCLUSIONS:
Plastic responses to environmental variability can differ widely among closely related species, and plasticity is therefore likely to be associated with many factors, including habitat and range size, as well as evolutionary history. Our results provided insights for species delimitation in Eucalyptus, which have management implications. Because of the high number of rare species and that other species are commercially important, a more comprehensive understanding of plasticity is essential for predicting their response to changing climates.
Understanding how species adapt and respond to new and heterogeneous environments is a central goal in evolutionary biology. Phenotypic plasticity is the ability of an organism (or genotype) to express a range of phenotypes as a function of the environment (Bradshaw, 1965; Schlichting and Levin, 1986; Scheiner, 1993) and is an important mechanism through which many species cope with environmental variability (Gratani, 2014). The expression of phenotypic plasticity can allow individuals to have a wider tolerance to environmental conditions, thereby enabling higher fitness in a range of habitats, and buffering populations against novel stresses (e.g., Bradshaw, 1965; Ghalambor et al., 2007; Nicotra et al., 2010). Phenotypic plasticity may promote adaptive responses and can influence lineage diversification (Sultan, 2000; Sultan and Spencer, 2002). For example, plasticity may allow individuals that are subjected to new environmental conditions to persist, until natural selection further enhances the fitness of a population (Price et al., 2003; Pigliucci, 2005). Furthermore, plasticity may rapidly change the phenotypes in a population that is targeted by natural selection (Ghalambor et al., 2007; Lande, 2009). The ability of species to respond to environmental variability is likely to be constrained to some extent by their evolutionary history (Kellermann et al., 2012). Because phenotypic plasticity is a quantitative trait that is itself under selection and which has evolutionary consequences, research of plasticity in a phylogenetic framework is important for predicting the response of species to changing climates (Pigliucci, 2005; Matesanz et al., 2010).
A number of species within the plant genus Eucalyptus display a high degree of phenotypic plasticity (Eldridge et al., 1993; Slee et al., 2006). Eucalyptus is a highly diverse genus (comprising more than 700 species) and is found in all climatic regions of Australia (Eldridge et al., 1993; Potts and Wiltshire, 1997). We focus on a group of eucalypts commonly known as the green ashes (subgenus Eucalyptus section Eucalyptus, sensu Brooker, 2000). The green ashes are found in a range of habitats in southeastern Australia, with some species occurring as trees in tall forests on fertile soils and others as smaller trees, or mallees, on sandstone substrates (Ladiges et al., 2010). They include the tallest flowering plant in the world (Eucalyptus regnans, up to 100 m tall), widespread and commercially important trees (e.g., E. obliqua), and a rare mallee that is often less than 1 m in height (E. cunninghamii). Mallees are long-lived multistemmed trees with an underground lignotuber from which they are able to resprout after disturbances, such as fire, frost, and herbivory (Mullette, 1978). While much effort has gone into understanding the evolutionary history of the green ashes, there has been considerable disagreement regarding the divergence (and number) of species (e.g., Ladiges et al., 1989; Brooker, 2000; Hill, 2002). Some green ash species (e.g., E. obliqua and E. dendromorpha) naturally occur as both trees and mallees (Hill, 2002), while other species are mallees in the wild (e.g., E. burgessiana and E. apiculata), but grow as trees when cultivated. Species boundaries within the group are further complicated by interspecific hybridization (Benson and McDougall, 1998). Because of this, many species are difficult to distinguish either chemically (Lassak and Southwell, 1982) or morphologically (Ladiges et al., 1989) and are identified by narrow differences in morphological characters (e.g., leaf length, leaf width, and capsule size), geographic location, or a combination of these. Because the green ashes are a group of closely related species representing a range of growth forms and habitats, they are an appropriate system for investigating the relationship between phenotypic plasticity and lineage diversification.
We examined phenotypic plasticity in response to nutrient and water variability in seedlings of representative species from the green ash group. Comparisons of eucalypt seedlings from xeric or oligotrophic habitats with those from mesic environments have been useful in identifying physiological and morphological traits that have evolved in response to stressful environments (e.g., Anderson et al., 1996; Merchant et al., 2007). Seedling response to environmental variability is important in the establishment of these relatively long-lived species. The seedling stage comprises rapid developmental changes in morphology and biomass allocation patterns (Kitajima, 1996), and seedling morphological and physiological traits are associated with plant establishment and regeneration (Zheng et al., 2009; Bonito et al., 2011). The degree of plasticity in seedling functional traits (e.g., plant biomass, plant height, leaf size, leaf thickness) is therefore likely to have implications for the ability of plants to acclimatize to changes in habitat and environmental conditions. For example, a high degree of plasticity in such traits could facilitate the survival and persistence of seedlings when dispersing to, and establishing in, new environments (Robinson and Dukas, 1999).
Comparative ecophysiological studies that incorporate phylogenetic data have the potential to address important evolutionary questions, including those concerning phenotypic plasticity (Ackerly et al., 2000; Donoghue, 2008). Traits of closely related species are expected to resemble each other more closely than expected by chance (Davies et al., 2013). If plasticity is a phylogenetically conserved trait, then differences in plasticity across species should reflect their relatedness (Valladares et al., 2000), and the capacity to respond to heterogeneous environments may be limited (or promoted) by evolutionary history. However, previous studies on the phylogenetic constraints of plasticity have yielded inconsistent results (Herben et al., 2014). For example, root phenotypic plasticity was found to be phylogenetically conserved in a range of plant species (Kembel and Cahill, 2005). In contrast, phenotypic plasticity in other plant species was not correlated with either phylogenetic relationships or growth form (Godoy et al., 2011). Rather, plasticity may be associated with habitat affiliation (Valladares et al., 2000). Phenotypic plasticity is predicted to be favored in heterogeneous environments (Van Buskirk, 2002; Murren et al., 2015) and to have evolved in response to variable selection pressures that occur in such environments (Bradshaw, 1965). Plasticity mechanisms may be related to habitat stability and productivity (Grime et al., 1986). For example, morphological plasticity is predicted to be pre-eminent in plants from productive environments, and to be not sustainable in slow-growing plants from resource-poor environments (Grime and Mackey, 2002).
In the case of Eucalyptus, nutrient and water availability are considered to be major factors in the evolution of the genus (Wardell-Johnson et al., 1997). Distribution patterns of eucalypts have been correlated with changes in (and interactions between) soil fertility and moisture (e.g., Parsons and Specht, 1967; Parsons, 1968, 1969; Kirkpatrick et al., 1987), while the distribution of specific mallee species has been attributed to only slight differences in water and nutrient availability (Parsons, 1969). However, global rainfall patterns are likely to alter under climate change, with projections of declining rainfall across many parts of the Australian continent over the next couple of decades (Hobday and Lough, 2011). Climate change is also predicted to affect biogeochemical cycling in forest ecosystems, which will likely affect the physical, chemical, and biological processes of soils (Campbell et al., 2009). Given the importance of nutrient and water availability to the distribution patterns of Eucalyptus, the degree of plasticity of species in response to changes in these factors is likely to be crucial to their survival and persistence under changing climates.
In a previous study (Rutherford et al., 2016), we used genomic analyses based on Diversity Arrays Technology (DArT) to estimate the phylogeny of the green ashes. We found that ‘species’ concepts were not always consistent with previous classifications, which were based primarily on morphology (particularly for species from the Sydney region and the Greater Blue Mountains World Heritage Area). Our results indicated that some of the parameters used to identify species in this group (e.g., leaf width) were problematic and not as useful in species delimitation as previously thought. Consequently, a follow-up question emerged from that study: are these characters poor indicators because we are dealing with broader plastic taxonomic units? Furthermore, what role does phenotypic plasticity have in the ability of the green ashes to respond to environmental variability?
In the current study, we investigated the morphological and physiological plasticity of green ash species in response to changes in nutrient and water availability in a common garden experiment. Common garden experiments can allow genetic and environmental factors to be differentiated, thereby enhancing our understanding of evolutionary patterns and adaptation (Warren et al., 2005; Lewis et al., 2011). We used a comparative approach incorporating phylogenetic data to address the following questions: (1) Are plant performance and leaf functional traits plastic across nutrient and water treatments? (2) Is the degree of phenotypic plasticity correlated with phylogenetic relationships, growth form, and substrate (soil type)? (3) Has plasticity affected species delimitation in the green ashes, and what are the implications of this for the conservation and management of species within the group?
MATERIALS AND METHODS
Selection of species and experimental design
Twelve species were selected from subgenus Eucalyptus section Eucalyptus (Table 1). To ensure that all recognized species in the Sydney region and the Greater Blue Mountains World Heritage Area (GBMWHA) were included in the study, we followed the species concepts of Hill (2002). Location and habitat details of each species were obtained from the National Herbarium of New South Wales database (Royal Botanic Garden Sydney) and Benson and McDougall (1998), and are summarized herein. With the exception of E. regnans (which occurs on deep fertile soils in Victoria and Tasmania), the species that were selected are found across a range of altitudes and latitudes in the Sydney region and GBMWHA. This included E. obliqua and E. fastigata, which occur on high fertility soils; as well as E. triflora and E. dendromorpha, both of which have a scattered but locally frequent distribution on low fertility slopes and escarpments. The remainder of species included were mallees on low nutrient ridges, escarpments, plateaus, and hillsides. Eucalyptus cunninghamii and E. laophila are found at elevations of 650–1100 m. Eucalyptus burgessiana and E. apiculata occur at altitudes of 300–750 m, while E. obstans and E. langleyi are found only in coastal areas. Eucalyptus stricta is widespread being found in both coastal and inland habitats (0–1200 m).
| Species | Provenance | Latitude (S) | Longitude (E) | Growth form | Substrate | Altitude (m) | Distribution and conservation |
|---|---|---|---|---|---|---|---|
| E. apiculata | Woodford | 33°45′19.3″ | 150°29′8.3″ | Mallee | Skeletal sandy soil | 550-600 | Rocky outcrops, GBMWHA to Berrima, NSW. Rare and localized |
| E. burgessiana | Springwood | 33°39′20.6” | 150°33′25.2″ | Mallee | Sandy soil | 400-450 | Ridges with rocky outcrops, GBMWHA, NSW. Locally frequent but restricted |
| E. cunninghamii | Kings Tableland | 33°46′15.1″ | 150°22′33.7″ | Mallee | Skeletal sandy soil | 785-805 | Slopes and escarpments, GBMWHA, NSW. Locally frequent but restricted |
| E. dendromorpha | Mount Wilson | 33°31′15.7″ | 150°22′14.8″ | Mallee | Sandy soil | 990-1010 | Escarpment and hillsides above cliffs from GBMWHA to Monga, NSW. Locally abundant but restricted |
| Fitzroy Falls | 34°38′54.1″ | 150°28′47.1″ | Medium tree | Sandy soil | 290-310 | ||
| E. fastigata | Mount Tomah | 33°32′57.9″ | 150°25′26.4″ | Tall tree | Deep, fertile soils | 990-1000 | Wet sclerophyll forest in cold wet areas on fertile soils in NSW and Vic. Widespread |
| E. langleyi | South Nowra | 34°58′25.3″ | 150°29′40.2″ | Mallee | Sandy soil | 195-265 | Confined to Nowra district, South Coast, NSW. |
| Restricted and localized | |||||||
| E. laophila | Wollemi National Park | 33°15′23.7″ | 150°13′7.3″ | Mallee | Skeletal sandy soil | 900-935 | Pagoda rocky outcrops, from Coricudgy to Newnes Plateau, NSW. Locally frequent but restricted |
| E. obliqua | Mount Murray | 34°33′33.5″ | 150°38′22″ | Tall tree | Deep, fertile soil | 600-650 | Wet sclerophyll or grassy forest in Qld, NSW, Vic., SA and Tas. Widespread and locally dominant |
| E. obstans | Royal National Park | 34°7′16.4″ | 151°4′31.7″ | Mallee | Shallow sandy soil | 110-125 | Plateaus and upper slopes in coastal areas from Ku-ring-gai Chase National Park to Jervis Bay, NSW. Locally abundant |
| E. regnans | Rubicon, Vic. | 37°20′ | 145°56′ | Tall tree | Deep fertile soil | 550-600 | Wet sclerophyll forests, mostly in mountainous regions in Vic. and Tas. Widespread and locally abundant |
| E. stricta | Newnes Plateau | 33°27′10.4″ | 150°13′53.1″ | Mallee | Shallow sandy soil | 1170-1180 | Ridges and plateaus from Newnes Plateau to Budawang Ranges, NSW. Widespread and locally abundant |
| E. triflora | Nerriga | 35°5′20.4″ | 150°8′16.8″ | Medium tree | Skeletal sandy soil | 700-800 | Steep sandstone slopes, confined to Budawang Ranges, NSW. Scattered but locally frequent |
We collected fruits from all species found in the Sydney region and GBMWHA. One population was sampled from each of 10 selected species. Two populations were sampled of E. dendromorpha, one from Mount Wilson in the GBMWHA and the other from Fitzroy Falls, 100 km south of Sydney (because the mallee form in the GBMWHA and the tree form from south of Sydney have been previously considered separate taxa; Klaphake, 2012). At least 10 randomly selected plants per population (approximately 10 m apart) were sampled. The growth form of each species and the substrate observed at each site was recorded at the time of collection. Substrates were categorized as either deep, fertile soils (generally derived from basalt), or skeletal and sandy soils (predominantly derived from sandstone). Fruits were returned to UNSW and air-dried in a cool, low-humidity environment until seeds were released. Because of the logistical constraints of field collections, seed for E. regnans was ordered from the Australian Tree Seed Centre (CSIRO, Canberra, Australia) and represented 12 parents from a wild population at Rubicon, Victoria. Habitat details (e.g., altitude, soil type) for this population were provided by the Australian Tree Seed Centre.
In December (early summer) 2013, seeds from all species were germinated on filter paper moistened with deionized water. Twenty-eight pots (15 cm deep and 20 cm diameter) per species were set up in the glasshouse at UNSW, 14 of which contained a low nutrient soil treatment and 14 containing a high nutrient soil treatment. The soil mix used in the high nutrient treatment was an Australian Native potting mix designed for optimal growth of native plants, which comprised 33% Organic Garden Mix (Australian Native Landscapes, Sydney, Australia), 33% washed river sand, 33% coco peat, and 200 mL of a slow release low phosphate fertilizer (Osmocote, Sydney, Australia). In the low-nutrient treatment, the fertilizer was diluted to 20% of that applied to the high nutrient treatment. Once germinated (the emergence of the cotyledons), seedlings for each species were transplanted into pots and assigned to high- and low-nutrient treatments. Over the following few months, a total of three seedlings were planted per pot. All plants were well-watered (to maximum soil water holding capacity) weekly, for three months until established, then thinned until a single plant per pot remained. After this period, pots were randomly assigned to high- and low-water treatments. Plants in the high-water treatment received as ample water as needed (enough to ensure the potting mix remained moist throughout the experiment). The low-water treatment was established by gradually reducing water over the course of 14 d until they received a water allocation of 30–40% of the high water treatments. Therefore, there were four treatments per species (seven replicates per treatment) in a full factorial design: low nutrient and low water (LNLW), low nutrient and high water (LNHW), high nutrient and low water (HNLW), and high nutrient and high water (HNHW). Pot positions were randomized once per month within blocks. The temperature of the glasshouse ranged from 18–30°C during the period of growth, with daily high temperatures generally ranging between 25 and 27°C.
Plant performance, leaf morphology, and leaf functional traits
Once per month, we measured leaf length (including the petiole when present) and leaf width (at the widest point) from 5–10 randomly selected true leaves using a digital Vernier caliper (Kincrome, Scoresby, Victoria, Australia). The number of leaves selected depended on the number present at the time of growth. The position of leaves on the stem was also recorded (node and leaf number) for as long as possible (until the cotyledons fell off the plant) to note any ontogenetic changes.
After eight months of growth, 14 variables representing whole plant and leaf-level traits were measured (Appendix S1, see Supplemental Data with this article). We measured the shoot height of each plant (from the soil surface to the top of the plant) to the nearest millimeter. The stem diameter to the nearest 0.01 mm was measured 2 cm above the soil surface with a digital Vernier caliper, and the total number of leaves per plant was recorded. Plants were then harvested and immediately placed into a sealed plastic bag and stored in a cooler box. Within an hour after harvesting, four leaves per plant were randomly selected and their fresh mass was determined. The length (including the petiole), width (at the widest point), petiole length, and thickness (at four randomly selected points on the leaf) of each leaf were measured with a digital Vernier caliper (following the method outlined in Pérez-Harguindeguy et al., 2013). Each leaf was scanned, and the leaf area (LA, to the nearest square millimeter) was determined using the Leaf Area Measurement version 1.3 software (University of Sheffield, Sheffield, United Kingdom). Each leaf and all other aboveground parts were oven dried for 24 h at 70°C and their dry mass was determined. Fresh mass vs. dry mass (FW/DW) of each plant and each leaf was determined and the specific leaf area (SLA) per leaf was calculated as the total leaf area (in square millimeters) divided by the dry mass of the leaf (in milligrams) using the formula of Wright and Westoby (1999).
Phenotypic plasticity index
Leaf trait and total plant plasticity were evaluated for the two extreme treatments: low nutrient and low water (LNLW), and high nutrient and high water (HNHW). These treatments were chosen because they represented the poorest and most favorable environments. Phenotypic plasticity was calculated with the plasticity index (PI) using the formula created by Valladares et al. (2000):
where Mean(env1) and Mean(env2) are the mean values of a specific trait for each species in environment 1 and 2; and Max(Mean(env1),Mean(env2)) is the maximum mean value (from environment 1 or 2). The PI values range from a value of zero to one (no plasticity to maximum plasticity; Valladares et al., 2000). In the current study, PI was calculated for each trait and species. The mean plant-level PI, mean leaf-level PI, and overall mean PI were calculated.
Statistical analysis
We used ANOVA to assess the main effects of water, nutrients, and species, and interactions between these main effects, on variability in seedling growth. A mixed model ANOVA was used to investigate the effect of nutrients, water, and species on changes in plant-level and leaf-level traits. Nutrient and water treatments were fixed effects and species was a random effect in this analysis. Species was used as a random factor in this analysis because we were only using a random selection of green ash species and we wanted to make inferences about all possible groups from our sample selection (Quinn and Keough, 2002). We used a three-factor ANOVA to assess the effect of water, nutrients, and growth form on plant-level and leaf-level traits. A three-factor ANOVA was also used to examine the effect of water, nutrients, and substrate on plant-level and leaf-level traits. To investigate the effect of the block design of the experiment on our analyses, we performed ANOVAs with and without ‘block’ (i.e., ‘replicate number’) as a fixed effect. We found that the blocking effect did not change the significance of the other effects or interactions (apart from two factors out of all analyses performed), and therefore the ‘block’ effect was removed from the analysis. For each species, differences in all plant-level and all leaf-level traits across treatments were examined using a multivariate analysis of variance (MANOVA). Because we had multiple response variables, many of which are likely to be highly correlated, we considered a MANOVA to be appropriate for investigating treatment differences on all response variables (MANOVA allows all response variables in an experiment to be considered simultaneously, Quinn and Keough, 2002). All these analyses were performed in SPSS version 24.0 (IBM, Armonk, New York, USA).
Individual traits were further analyzed in R version 3.1.3 (R Development Core Team, 2015). We assessed the effect of water and nutrient treatments on variation in each trait using a two-factor ANOVA, focusing on differences between treatments for each growth form and substrate. To examine the effect of water and nutrient treatments on leaf length and leaf width over the period of growth for each species, a two-factor ANOVA was used focusing on differences between treatments at each monthly measurement. We used a single factor ANOVA to assess differences in plant-level, leaf-level, and overall mean plasticity indices between species. Significant differences (at P < 0.05) between pairwise mean values within each ANOVA were assessed using Tukey's honest significant difference (HSD) test. Prior to each analysis, homogeneity of variances among groups was examined with Levene's test, and traits were log10 transformed where necessary to meet the variance assumptions of ANOVA.
While phylogenetic comparative methods (e.g., phylogenetic independent contrasts) are considered the most appropriate way to analyze trait data in closely related species, these methods are most effective using a completely resolved phylogeny (this is because incompletely resolved phylogenies can inflate estimates of phylogenetic conservatism, Davies et al., 2012). In cases where a completely resolved phylogeny is not available, a mixed model ANOVA (with species included as a random factor) has been found to perform well (Funk et al., 2015). Although we estimated the phylogeny of our study species (see below for full details of analysis), it should be noted that some clades had a posterior probability (PP) of less than 0.95 (and therefore, all clades were not entirely resolved). However, because most of the clades in our phylogeny had a PP of 1, we were able to use our phylogenetic analysis to address our questions concerning the relationship between phenotypic plasticity and evolutionary relationships (across species, growth forms, and substrates).
Phylogenetic analysis of plasticity across species, growth forms and substrate
To investigate phylogenetic patterns in phenotypic plasticity (plant-level, leaf-level, and overall mean PI) across the green ash group, we constructed a phylogeny of the study species. In Rutherford et al. (2016), phylogenetic trees were constructed from DArT presence/absence markers, with the Bayesian phylogeny showing a strong correlation with the current taxonomy of the green ashes and closely related eucalypts. Using the molecular data set from Rutherford et al. (2016), a Bayesian phylogenetic tree was constructed with data from the species in the current study, with Eucalyptus piperita selected as the outgroup. Bayesian analyses were conducted in MrBayes 3.2.4 (Huelsenbeck and Ronquist, 2001; Ronquist and Huelsenbeck, 2003; Ronquist et al., 2012) using a restriction site (binary) model of evolution and default priors. The final analysis was run for 40 million generations sampling every 500 generations with two parallel runs each with four chains (three hot and one cold). Convergence was considered reached based on the standard deviation of split frequencies (< 0.01) and the first 25% of trees discarded as burn-in. Ancestral reconstructions of growth form and substrate were traced onto the Bayesian tree using maximum parsimony reconstructions in the Mesquite software package version 3.03 (Maddison and Maddison, 2015).
We tested the effect of phylogenetic history on the evolution of phenotypic plasticity in the green ashes. A range of approaches are available that quantify phylogenetic signal in comparative data sets (Diniz-Filho et al., 2012). These include approaches based on statistical methods (i.e., spatial autocorrelation indices), which quantify the level of phylogenetic autocorrelation for a specific trait (e.g., Moran's I and Abouheif's Cmean, Moran, 1950; Gittleman and Kot, 1990; Abouheif, 1999). However, autocorrelation indices, such as Moran's I, were not originally designed to provide a quantitative interpretation, and the earlier literature suggests that such interpretations are inappropriate (Ord, 1975; Li et al., 2007). Other approaches are based on a theoretical model of trait evolution (e.g., Brownian motion) where phylogenetic signal is estimated using indices such as Pagel's lambda (λ) and Blomberg's K (Pagel, 1997, 1999; Blomberg et al., 2003). Under Brownian motion, trait evolution follows a random walk along the branches of a phylogeny, with phenotypic variance being proportional with branch lengths (Felsenstein, 1985; Martins, 1996). In a comparison of approaches and indices, it was found that Pagel's λ performed better than Blomberg's K on simulated data sets when the underlying evolutionary process was based on a Brownian motion model of trait evolution (Münkemüller et al., 2012). Although computationally more intensive, Pagel's λ provided a reliable effect size measure and performed better at discriminating between complex models of trait evolution (Münkemüller et al., 2012). Therefore, in the current study, Pagel's λ was chosen as the most appropriate parameter to quantify phylogenetic signal.
We estimated phylogenetic signal of plant-level, leaf-level, and overall mean PI using the ‘Continuous’ module in BayesTraits V2 (Pagel, 1997; Pagel and Meade, 2014). The Continuous option implements the generalized least-squares model, which accounts for nonindependence among species (Martins and Hansen, 1997; Pagel and Meade, 2014) and is analytically equivalent to independent contrasts (Garland et al., 2005). We selected 1000 trees (from the MrBayes analysis) to analyze in BayesTraits. We estimated Pagel's λ in Continuous using a random-walk model and Markov chain Monte Carlo (MCMC) analysis. Pagel's λ models the variance of the trait attributable to phylogenetic conservatism or ‘heritability’ (Lee et al., 2013). We compared the estimated value of λ to models where λ is forced to be either 0 or 1. A λ value of 0 suggests that trait evolution is independent of the phylogeny, whereas a λ value of 1 indicates that the pattern of trait evolution is consistent with phylogenetic relationships (Kang et al., 2014). Analyses were run for 140 million generations, with a sampling frequency of 100 generations and a burn-in of 35 million generations. To assess convergence, each analysis was repeated three times. Bayes factors (BF) were used to infer whether the λ parameter improved model fit (using the criterion of twice the marginal log-likelihood differences, Newton and Raftery, 1994; Kass and Raftery, 1995). We calculated BF following the procedure described in the BayesTraits manual (Pagel and Meade, 2014). The BF compares the marginal log likelihoods of two models to assess which model better fits the data. A log BF value of less than 2 suggests that there is weak support for one model over another, whereas, a log BF greater than 2 indicates that there is strong evidence for support of one model over the other (Pagel and Meade, 2014).
RESULTS
Plasticity in plant performance and plant-level traits
We found that all plant-level traits varied significantly in response to changes in nutrient treatments across species, growth forms, and substrates (P < 0.05, Table 2). With the exception of stem diameter, all plant-level traits were significantly different across growth forms (P < 0.05, Table 2). Only stem diameter, total leaf number, and total leaf dry mass were significantly different among species; while total leaf number, total aboveground dry mass, total leaf dry mass, and total FW/DW were significantly different across substrates (P < 0.05, Table 2).
| Height | Stem diameter | Total leaf number | Total aboveground dry mass | Stem dry mass | Total leaf dry mass | Total FW/DW | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| d.f. | F | P | F | P | F | P | F | P | F | P | F | P | F | P | |
| Species | 12 | 2.934 | 0.051 | 4.424 | 0.023 | 2.809 | 0.042 | 2.718 | 0.054 | 1.833 | 0.151 | 3.087 | 0.028 | 7.053 | 0.055 |
| N | 1 | 46.845 | <0.001 | 122.598 | <0.001 | 28.557 | <0.001 | 58.455 | <0.001 | 78.128 | <0.001 | 56.69 | <0.001 | 100.967 | <0.001 |
| W | 1 | 0.022 | 0.885 | 15.288 | 0.002 | 5.508 | 0.036 | 3.873 | 0.071 | 3.832 | 0.073 | 5.557 | 0.035 | 5.15 | 0.042 |
| N × W | 1 | 6.297 | 0.027 | 14.343 | 0.002 | 16.377 | 0.001 | 4.542 | 0.053 | 6.492 | 0.025 | 9.061 | 0.01 | 5.426 | 0.037 |
| Species × N | 12 | 3.404 | 0.022 | 2.328 | 0.079 | 11.512 | <0.001 | 5.385 | 0.003 | 3.991 | 0.012 | 9.195 | <0.001 | 0.865 | 0.597 |
| Species × W | 12 | 0.884 | 0.583 | 0.812 | 0.638 | 1.154 | 0.404 | 0.962 | 0.527 | 1.415 | 0.278 | 1.405 | 0.282 | 1.012 | 0.492 |
| Species × N × W | 12 | 0.99 | 0.458 | 0.781 | 0.67 | 0.471 | 0.931 | 0.611 | 0.832 | 0.691 | 0.76 | 0.359 | 0.976 | 0.877 | 0.571 |
| Growth form | 2 | 14.006 | <0.001 | 1.83 | 0.162 | 24.173 | <0.001 | 6.047 | 0.003 | 4.885 | 0.008 | 6.002 | 0.003 | 3.937 | 0.02 |
| N | 1 | 109.842 | <0.001 | 136.72 | <0.001 | 92.307 | <0.001 | 128.932 | <0.001 | 170.664 | <0.001 | 117.299 | <0.001 | 68.735 | <0.001 |
| W | 1 | 0.066 | 0.798 | 3.803 | 0.052 | 1.109 | 0.293 | 0.564 | 0.453 | 1.771 | 0.184 | 0.824 | 0.365 | 2.586 | 0.109 |
| N × W | 1 | 3.074 | 0.08 | 2.752 | 0.098 | 2.276 | 0.132 | 0.624 | 0.43 | 1.944 | 0.164 | 0.875 | 0.35 | 5.244 | 0.023 |
| Growth form × N | 2 | 1.513 | 0.222 | 0.358 | 0.699 | 4.875 | 0.008 | 3.099 | 0.046 | 3.095 | 0.047 | 2.426 | 0.09 | 2.661 | 0.071 |
| Growth form × W | 2 | 0.019 | 0.982 | 0.285 | 0.752 | 0.111 | 0.895 | 0.443 | 0.642 | 0.37 | 0.691 | 0.155 | 0.856 | 1.069 | 0.344 |
| Growth form × N × W | 2 | 0.371 | 0.69 | 1.125 | 0.326 | 0.378 | 0.685 | 0.321 | 0.726 | 0.199 | 0.82 | 0.222 | 0.801 | 0.117 | 0.889 |
| Substrate | 1 | 2.566 | 0.11 | 3.173 | 0.076 | 46.709 | <0.001 | 5.988 | 0.015 | 0.091 | 0.764 | 7.437 | 0.007 | 4.849 | 0.028 |
| N | 1 | 82.427 | <0.001 | 129.164 | <0.001 | 115.4 | <0.001 | 94.848 | <0.001 | 131.076 | <0.001 | 88.899 | <0.001 | 56.329 | <0.001 |
| W | 1 | 0.088 | 0.767 | 4.64 | 0.032 | 1.533 | 0.217 | 0.374 | 0.541 | 1.028 | 0.311 | 1.095 | 0.296 | 1.269 | 0.261 |
| N × W | 1 | 4.217 | 0.041 | 4.955 | 0.027 | 3.478 | 0.063 | 0.689 | 0.407 | 1.634 | 0.202 | 1.538 | 0.216 | 4.151 | 0.042 |
| Substrate × N | 1 | 0.424 | 0.515 | 0.688 | 0.407 | 9.578 | 0.002 | 2.852 | 0.092 | 0.169 | 0.681 | 2.557 | 0.111 | 0.904 | 0.342 |
| Substrate × W | 1 | 0.035 | 0.852 | 0.08 | 0.778 | 0.003 | 0.958 | 0.751 | 0.387 | 0.721 | 0.396 | 0.055 | 0.815 | 2.122 | 0.146 |
| Substrate × N × W | 1 | 0.628 | 0.429 | 0.075 | 0.784 | 0.004 | 0.949 | 0.344 | 0.558 | 0.331 | 0.565 | 0.002 | 0.966 | 0 | 0.992 |
- d.f. = degrees of freedom, FW/DW = fresh mass vs. dry mass, P = significance.
We found that plants generally grew significantly larger under higher nutrients (differences in growth of E. dendromorpha from Mount Wilson between high and low nutrient treatments are demonstrated in Fig. 1). For the majority of species, most plant-level traits varied significantly in response to nutrients (P < 0.05, Appendix S2). Aboveground dry mass, plant height, and stem diameter were all significantly higher in the high-nutrient treatments compared with the low-nutrient treatments for each growth form and each substrate (P < 0.05, Fig. 2, Appendix S3). In terms of magnitude, the total aboveground dry mass for most species (apart from E. regnans, E. obliqua, E. laophila, and E. apiculata) was more than twice as large in the high-nutrient and low-water (HNLW) and high-nutrient and high-water (HNHW) treatments, compared with the low-nutrient and low-water (LNLW) and low-nutrient and high-water (LNHW) treatments. In E. fastigata, the total aboveground dry mass from the high-nutrient treatments (HNLW and HNHW) was 6–11 times larger than the low-nutrient treatments (LNLW and LNHW).
Changes in growth and vegetative morphology of Eucalyptus dendromorpha from Mount Wilson over a 6 mo period in response to the following treatments: low-nutrient high-water (LNHW) and high-nutrient high-water (HNHW) treatments. (Photography: A. Orme, L. Mou, and S. Rutherford.)
Total aboveground dry mass of green ash (A) species, (B) growth forms, and (C) substrates after 8 mo of the following treatments: low nutrients and low water (LNLW), low nutrients and high water (LNHW), high nutrients and low water (HNLW), and high nutrients and high water (HNHW). Bars show mean ± SE (n = 4–66). Significant differences between treatments for each growth form and substrate are denoted with different letters above bars (P < 0.05, Tukey's HSD post-hoc comparison). See Table 2 for ANOVA of each plant-level trait across species, growth form, and substrate, and Appendix S2 for MANOVA of plant-level traits for each species.
Many plant-level traits did not vary significantly in response to changes in water. Although stem diameter, total leaf number, total leaf dry mass, and total FW/DW varied significantly in response to water across species, only stem diameter varied significantly in response to water across substrates (P < 0.05, Table 2). For individual species, there was no significant effect of water on most plant-level traits (Appendix S2). There were no significant differences in total aboveground dry mass between the LNLW and LNHW treatments or between the HNLW and HNHW treatments for any growth form or substrate (Fig. 2). Similar results were found for height and stem diameter (Appendix S3).
The interaction of water treatment–nutrient treatment was significant for most plant-level traits (with the exception of total aboveground dry mass) for species (P < 0.05, Table 2). However, the interaction of water treatment–nutrient treatment was only significant for total FW/DW for growth form, and was only significant for height, stem diameter, and total FW/DW for substrates (P < 0.05, Table 2). For individual species, interactions of water treatment–nutrient treatment of most plant-level traits were not significant (Appendix S2). Most of the other interactions were not significant for most plant-level traits across species, growth form, or substrate (P > 0.05, Table 2).
Plasticity in leaf morphology and functional traits
We found significant differences in leaf length and leaf width between treatments at each monthly measurement for E. fastigata, E. regnans, E. dendromorpha (Fitzroy Falls), E. dendromorpha (Mount Wilson), E. burgessiana, E. obstans, E. triflora, and E. stricta (P < 0.05, Fig. 3). For these species, leaf length and width was significantly higher in the high nutrient treatments than the low nutrient treatments. In the case of E. cunninghamii, while differences in leaf length among treatments were significant at each monthly measurement, no significant difference in leaf width was found between treatments at 11, 20, and 25 wk of age.
Leaf length and leaf width over a 6 mo period of Eucalyptus species in response to the following treatments: low nutrient and low water (LNLW), low nutrient and high water (LNHW), high nutrient and low water (HNLW), and high nutrient and high water (HNHW). Symbols for leaf length under the different treatments are: LNLW (■), LNHW (
), HNLW (
) and HNHW (□); and for leaf width are: LNLW (▲), LNHW (
), HNLW (
) and HNHW (∆).Values are the mean ± SE (n = 4–7). Differences in leaf length and leaf width across treatments at each time period were significant for E. fastigata, E. regnans, E. burgessiana, E. obstans, and E. stricta (P < 0.05, two-way ANOVA).
As with plant-level traits, all leaf-level traits varied significantly in response to nutrient treatments for species, growth form, and substrate at the end of the study period (P < 0.05, Table 3). With the exception of SLA and leaf FW/DW, leaf-level traits were significantly different among species (P < 0.05, Table 3). All leaf-level traits varied significantly among growth forms, while the majority of leaf-level traits (except for leaf width) varied significantly between substrates (P < 0.05, Table 3). For the majority of species, most leaf-level traits varied significantly in response to nutrients (P < 0.05, Appendix S4). In E. fastigata, E. dendromorpha (Fitzroy Falls), E. langleyi, E. burgessiana, E. stricta, and E. obstans, leaf width at the end of the eight-month study period was 1.4–2 times higher in the HNHW treatment than in the LNHW treatment (Appendix S5). We found significant differences in SLA and leaf thickness across nutrient treatments for most species (P < 0.05, Appendix S4). The SLA was significantly higher in the LNLW treatment than in the high nutrients (HNLW and HNHW) for each growth form and for plants from deep, fertile soils (Fig. 4). For plants from skeletal and sandy soils, SLA was significantly higher in the low nutrients (LNLW and LNHW) than in the high nutrients (HNLW and HNHW) (Fig. 4). Leaf thickness was significantly lower in the LNLW treatment than in the high nutrients (HNLW and/or HNHW) for each growth form and for plants from deep, fertile soils (Appendix S6). For plants from skeletal and sandy soils, leaf thickness was significantly lower in the LNLW and LNHW treatments than in the HNLW and HNHW treatments (Appendix S6).
| Leaf length | Leaf width | Petiole length | Leaf thickness | Leaf area | SLA | Leaf FW/DW | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| d.f. | F | P | F | P | F | P | F | P | F | P | F | P | F | P | |
| Species | 12 | 8.304 | 0.004 | 17.763 | <0.001 | 7.115 | 0.003 | 96.454 | 0.029 | 8.225 | <0.001 | 61.455 | 0.223 | 4.175 | 0.059 |
| N | 1 | 32.084 | <0.001 | 23.041 | <0.001 | 29.268 | <0.001 | 88.581 | <0.001 | 15.391 | 0.002 | 48.583 | <0.001 | 49.329 | <0.001 |
| W | 1 | 0.19 | 0.67 | 0.002 | 0.967 | 1.07 | 0.321 | 0.266 | 0.615 | 0.008 | 0.931 | 1.184 | 0.296 | 3.66 | 0.079 |
| N × W | 1 | 4.626 | 0.052 | 5.273 | 0.04 | 2.916 | 0.113 | 0.548 | 0.473 | 3.89 | 0.071 | 2.237 | 0.16 | 4.278 | 0.06 |
| Species × N | 12 | 1.797 | 0.162 | 3.059 | 0.032 | 3.456 | 0.021 | 0.542 | 0.849 | 5.52 | 0.003 | 0.972 | 0.519 | 0.985 | 0.51 |
| Species × W | 12 | 0.964 | 0.525 | 1.167 | 0.396 | 0.809 | 0.64 | 0.986 | 0.51 | 1.488 | 0.251 | 0.348 | 0.96 | 1.253 | 0.351 |
| Species × N × W | 12 | 1.016 | 0.434 | 1.086 | 0.372 | 0.842 | 0.607 | 1.359 | 0.185 | 0.839 | 0.611 | 1.164 | 0.309 | 1.075 | 0.381 |
| Growth form | 2 | 16.678 | <0.001 | 8.894 | <0.001 | 4.213 | 0.016 | 67.616 | <0.001 | 12.401 | <0.001 | 89.76 | <0.001 | 6.215 | 0.002 |
| N | 1 | 34.525 | <0.001 | 19.5 | <0.001 | 38.98 | <0.001 | 18.349 | <0.001 | 21.497 | <0.001 | 51.648 | <0.001 | 54.145 | <0.001 |
| W | 1 | 0.917 | 0.339 | 0.112 | 0.739 | 0.272 | 0.603 | 0.008 | 0.929 | 0.414 | 0.521 | 0.679 | 0.41 | 2.678 | 0.103 |
| N × W | 1 | 2.35 | 0.126 | 1.74 | 0.188 | 0.704 | 0.402 | 1.212 | 0.272 | 1.25 | 0.264 | 4.559 | 0.033 | 3.991 | 0.047 |
| Growth form × N | 2 | 0.352 | 0.704 | 0.179 | 0.836 | 0.62 | 0.539 | 0.284 | 0.753 | 0.873 | 0.419 | 3.613 | 0.028 | 3.943 | 0.02 |
| Growth form × W | 2 | 0.451 | 0.637 | 0.102 | 0.903 | 0.178 | 0.837 | 0.201 | 0.818 | 0.384 | 0.682 | 0.267 | 0.766 | 0.105 | 0.9 |
| Growth form × N × W | 2 | 0.405 | 0.667 | 0.382 | 0.683 | 0.019 | 0.981 | 0.463 | 0.63 | 0.183 | 0.833 | 0.474 | 0.623 | 0.007 | 0.993 |
| Substrate | 1 | 22.917 | <0.001 | 0.253 | 0.615 | 4.276 | 0.039 | 136.537 | <0.001 | 10.862 | 0.001 | 164.602 | <0.001 | 8.843 | 0.003 |
| N | 1 | 28.206 | <0.001 | 15.548 | <0.001 | 31.522 | <0.001 | 16.047 | <0.001 | 15.434 | <0.001 | 48.998 | <0.001 | 41.04 | <0.001 |
| W | 1 | 0.53 | 0.467 | 0.013 | 0.908 | 0.674 | 0.412 | 0.194 | 0.66 | 0.082 | 0.774 | 0.163 | 0.687 | 2.912 | 0.089 |
| N × W | 1 | 3.258 | 0.072 | 2.226 | 0.137 | 0.637 | 0.425 | 1.011 | 0.315 | 1.423 | 0.234 | 3.814 | 0.052 | 4.093 | 0.044 |
| Substrate × N | 1 | 0.499 | 0.48 | 0.14 | 0.708 | 1.03 | 0.311 | 0.574 | 0.449 | 1.563 | 0.212 | 4.592 | 0.033 | 1.251 | 0.264 |
| Substrate × W | 1 | 0.232 | 0.63 | 0 | 0.986 | 0.014 | 0.905 | 0.228 | 0.633 | 0.022 | 0.883 | 0.262 | 0.609 | 0.057 | 0.811 |
| Substrate × N × W | 1 | 0.785 | 0.376 | 0.698 | 0.404 | 0.036 | 0.85 | 0.464 | 0.496 | 0.311 | 0.578 | 0.372 | 0.542 | 0.002 | 0.963 |
- d.f. = degrees of freedom, FW/DW = fresh mass vs. dry mass, SLA = specific leaf area, P = significance.
Specific leaf area (SLA) of (A) species, (B) growth forms and (C) substrates after 8 mo of the following treatments: low nutrients and low water (LNLW), low nutrients and high water (LNHW), high nutrients and low water (HNLW), and high nutrients and high water (HNHW). Bars show mean ± SE (n = 4–66). Significant differences between treatments for each growth form and substrate are denoted with different letters above bars (P < 0.05, Tukey's HSD post-hoc comparison). See Table 3 for ANOVA of each leaf-level trait across species, growth form, and substrate, and Appendix S4 for MANOVA of leaf-level traits for each species.
We found that leaf-level traits did not vary significantly in response to changes in water for species, growth forms, and substrates (Table 3). For individual species, there was only a significant effect of water on leaf FW/DW in E. regnans and E. obliqua, leaf length in E. stricta, and leaf thickness in E. obstans and E. apiculata (P < 0.05, Appendix S4). Differences in all other leaf-level traits across water treatments were not significant for any species. No significant differences were found between LNLW and LNHW or between HNLW and HNHW for leaf length, leaf width, SLA, and leaf thickness in any growth form or substrate (Fig. 4, Appendices S5 and S6).
Nutrient–water treatment interactions across species, growth forms, and substrate were not significant for most leaf-level traits (Table 3). Similarly, all other interactions were not significant for most leaf-level traits across species, growth form, and substrate (Table 3). For individual species, nutrient–water treatment interaction had a significant effect on leaf length in E. regnans, and leaf width in E. regnans and E. fastigata; leaf thickness in E. apiculata, E. cunninghamii, and E. dendromorpha (Fitzroy Falls); leaf FW/DW in E. dendromorpha (Fitzroy Falls); leaf area in E. regnans and E. fastigata; and SLA in E. stricta (P < 0.05, Appendix S4).
Phylogenetic analysis of plasticity
Values for whole plant phenotypic plasticity (plant-level PI) were greater than that of leaf functional traits (leaf-level PI) for the study species (Fig. 5). Eucalyptus fastigata had the highest overall mean PI, followed by E. burgessiana, E. dendromorpha (Fitzroy Falls), and E. stricta. Species with lower overall mean PI values were E. regnans, E. laophila, E. apiculata, E. triflora, and E. cunninghamii. Differences were only significant for leaf-level PI values (P < 0.001), with E. fastigata and E. burgessiana having significantly higher leaf-level PI than E. regnans and E. laophila. Eucalyptus dendromorpha (Fitzroy Falls) and E. langleyi also had a significantly higher leaf-level PI than E. regnans.
Comparison between the Bayesian phylogeny and the overall mean, plant-level, and leaf-level phenotypic PI of the study species. Character states were traced onto the phylogeny using Mesquite version 3.03 and show: trees on deep, fertile soils (■); medium trees on skeletal and sandy soils (
); and mallees on skeletal and sandy soils (□). The PI values are the mean ± SE (n = 7–14). Differences between species are significant for leaf-level PI (F12, 78 = 3.29, P < 0.001, one-way ANOVA), but not for plant-level or overall mean PI.
Bayesian analyses produced a phylogeny with 12 nodes, eight of which had a Bayesian Posterior Probability greater than 0.95 (Fig. 5). The phylogeny produced here was consistent with Rutherford et al. (2016) in terms of the same overall topology and similar groupings of species. In the current study, the tall trees on deep, fertile soils formed a clade that was sister to the medium trees and mallees on skeletal and sandy soils. However, there was no apparent pattern between plant-level, leaf-level, or overall mean plasticity and phylogenetic relationships. For example, although the tall trees, E. regnans and E. fastigata, were in the same clade, they had the lowest and highest overall mean PI, respectively. Similarly, a mallee from skeletal and sandy soils, E. stricta, had the second highest plant-level PI, yet was in the same clade as another mallee on skeletal and sandy soils, E. apiculata, which had the lowest plant-level PI. This was supported by the BayesTraits analysis, with λ values for plant-level PI, leaf level PI, and overall mean PI being less than 0.5, indicating that plasticity was not correlated with phylogeny (Table 4). The BF indicated that the estimated λ values for plant-level, leaf-level, and overall mean PI better described the data than when λ was fixed at either 1 (where trait evolution is correlated with phylogenetic relationships) or 0 (where trait evolution is independent of the phylogeny). The BF values were all less than 2 indicating that the estimated λ model was more appropriate that when λ was fixed at either 1 or 0 (Table 4).
| λ estimated | λ fixed at 0 | λ fixed at 1 | Log BF for λ= 0 | Log BF for λ= 1 | |
|---|---|---|---|---|---|
| Plant-level PI | |||||
| Harmonic mean | 7.981 | 7.131 | 7.001 | −1.7 | −1.96 |
| Mean likelihood | 9.364 | 9.564 | 9.112 | ||
| Mean λ (95% HPD) | 0.497 (0.00009, 0.947) | 0 | 1 | ||
| Leaf-level PI | |||||
| Harmonic Mean | 13.280 | 12.874 | 14.031 | −0.812 | 1.502 |
| Mean likelihood | 14.725 | 15.075 | 14.095 | ||
| Mean λ (95% HPD) | 0.458 (0.000006, 0.928) | 0 | 1 | ||
| Overall mean PI | |||||
| Harmonic mean | 12.249 | 12.399 | 11.593 | 0.3 | −1.312 |
| Mean likelihood | 13.835 | 14.214 | 13.171 | ||
| Mean λ (95% HPD) | 0.453 (0.0000006, 0.926) | 0 | 1 |
- BF = Bayes Factor, HPD = highest posterior density.
DISCUSSION
Plasticity in seedling functional traits of the green ash eucalypts
High levels of phenotypic plasticity were found in many plant- and leaf-level traits across nutrient treatments for most species. Similarly, most plant- and leaf-level traits varied significantly across nutrient treatments for each growth form and for plants from each substrate. This is consistent with our predictions, and the findings of previous studies demonstrating phenotypic plasticity in plant morphology and physiology (e.g., Cordell et al., 1998; Hovenden and Vander Schoor, 2004; Herben et al., 2014). The high degree of plasticity in leaf morphology (e.g., leaf length and leaf width) found here in response to varying nutrients (for most species, growth forms, and plants from each substrate) agreed with the results of Hovenden and Vander Schoor (2004) who found that leaf morphology in Nothofagus cunninghamii was significantly affected by environmental conditions. However, it should be noted that although the sample size for the medium tree growth form in our study was 55 replicates, this growth form was only represented by two species (E. triflora and E. dendromorpha from Fitzroy Falls). This is in contrast to the other two growth forms, which were represented by more species (the tall trees comprised three species, while the mallees comprised eight species). Therefore, some caution is required when interpreting our analysis for the medium tree growth form.
Unexpectedly, many plant- and leaf-level traits did not vary significantly across water treatments. In fact, when water availability was lowered below a certain threshold during the period of growth (less than 30% of the high-water treatment) plants would start to die (i.e., within 12 h of the treatment being lowered, some individuals were lost). Reducing the water treatment below 30% was tried multiple times, but always resulted in the loss of some individuals. So as not to risk losing more plants, we kept the low-water treatment at 30–40% of the high-water treatment. These findings may be related to the current distribution of the study species, which all occur in higher rainfall environments (700–1800 mm annual rainfall; Benson and McDougall, 1998). A number of studies have demonstrated that some eucalypts may be sensitive to water stress. For example, Stoneman et al. (1994) found that rates of leaf growth and photosynthesis in seedlings of E. marginata were sensitive to water deficits. Similarly, eucalypt species from drier habitats were found to maintain lower osmotic potentials under both well-watered and water-deficit conditions than eucalypt species from higher rainfall environments (including E. obliqua, Merchant et al., 2007). Leaf traits in many Eucalyptus species may show low overall variation to water availability within and between populations. For example, leaf traits were not correlated with water availability between populations across rainfall gradients in E. sideroxylon (Warren et al., 2005). Furthermore, in a study of 28 eucalypt species, differences in hydraulic traits were mainly genotypic in origin rather than environmentally plastic (with no indication of plasticity in these traits across an aridity gradient; Pfautsch et al., 2016). However, in growth experiments, it can be difficult to subject plants to the appropriate levels of water stress without causing plant death. In this study, plants in the low-water treatment were wilting, however it is difficult to assess whether plants are truly water stressed without measuring the soil water potential. Future research focusing on soil water potential, water relations, and hydraulic traits of the study species would enable us to better understand their response to water stress.
Role of plasticity in the evolution and diversification of the green ashes
Phenotypic plasticity (plant-level, leaf-level, and overall mean PI) was not correlated with phylogeny, nor was it associated with growth form. The estimated values of λ (<0.5, Table 4) suggest that there is a weak phylogenetic signal in plasticity and that the degree of plasticity in section Eucalyptus perhaps reflects more of a convergent evolutionary strategy (e.g., Valladares et al., 2000; Godoy et al., 2011; Herben et al., 2014). Unexpectedly, phenotypic plasticity (plant-level, leaf-level, and overall mean PI) was not associated with substrate. Although species in section Eucalyptus occupy high- and low-fertility soils, they occur along an altitudinal gradient (0–1200 m; Benson and McDougall, 1998) with many other environmental factors (e.g., temperature and light) that are highly variable between habitats. In the current study, SLA was higher in plants from deep, fertile soils than those from skeletal and sandy soils. This finding was consistent with that found in a range of species. For example, higher SLA was found for plants from higher fertility soils in a study of 79 perennial species from high and low nutrient habitats (Wright et al., 2001). A similar pattern in SLA was found for temperate rainforest tree species from high- and low-fertility sites (Lusk et al., 1997). However, SLA of each growth form and of plants from each substrate in the current study was significantly higher in the low-nutrient treatments than in the high-nutrient treatments. Once again, this was somewhat surprising because high SLA is often correlated with higher nutrient availability (e.g., Wright et al., 2001, 2004). The SLA is also positively correlated with relative growth rate (Lusk et al., 1997; Wright and Westoby, 1999, 2000). Negative relationships between relative growth rate and leaf nitrogen concentration have been found for other species and it has been suggested this may be because under high nutrient supply, slower growing species store essential nutrients to support growth after soil reserves are exhausted (‘luxury consumption’, Chapin, 1980; Tateno and Chapin, 1997). This type of mechanism may explain the higher SLA in the lower nutrient treatments found for many of the species here.
Phenotypic plasticity, which was once considered to be ‘noise’ masking the true appearance of organisms, is increasingly being recognized to be genetically controlled, heritable, and of significance to the evolution of species (Bradshaw, 2006; Nicotra et al., 2010). For example, phenotypic plasticity can enhance the ability of a species to occur across diverse and variable habitats (Sultan, 2000) and more-widespread species are thought to be better able to capitalize on increased availability of resources (Dawson et al., 2012). In the current study, the rare species, E. cunninghamii, was not as plastic as other species in leaf-level PI and this may explain why this species is restricted to higher altitude escarpments and cliff edges, generally in the upper Blue Mountains. Eucalyptus obliqua and E. fastigata had higher leaf-level PI and this could account for why they are more widespread (E. obliqua has ecotypes, which are able to cope with a wide range of environmental conditions; Eldridge et al., 1993). Eucalyptus regnans, which is also a widespread species, had the lowest leaf-level and overall mean PI. However, although E. regnans is widespread, it has a very narrow environmental range, occurring on high fertility, well-drained soils with high and reliable rainfall (Tng et al., 2012), where fire is the primary form of natural disturbance (Burns et al., 2015). Forests dominated by E. regnans can have high fire intensity (McCarthy et al., 1999), but E. regnans is a fire-sensitive obligate seeder eucalypt, which does not resprout following fire (Nicolle, 2006) despite having epicormic bud-forming structures like other species (Waters et al., 2010).
Implications for species delimitation and consequences for conservation and management
Much of the work on species delimitation in Eucalyptus has focused on seedling or juvenile morphology (e.g., Ladiges et al., 1989, 2010; Brooker, 2000; Slee et al., 2006). The size and shape of juvenile leaves are important taxonomic characters in eucalypts, with juvenile leaf traits frequently being used to discriminate between species with similar adult foliage (King, 1999). Because species are one of the fundamental units of analysis in biology, improving our knowledge of species boundaries is essential when attempting to address many evolutionary, biogeographic, and ecological questions (Rissler and Apodaca, 2007; Wiens, 2007). The success of conservation efforts also requires species to be clearly defined for their effective management (Allendorf and Luikart, 2007; Flot et al., 2010).
In the current study, there were significant differences in leaf width for many species in response to changes in nutrients. Many species that had lower leaf-level PI, such as E. cunninghamii and E. triflora, were relatively easy to identify in the field (E. cunninghamii is distinguished on the basis of its thin, silvery-green leaves, while E. triflora usually has a three-flowered inflorescence). The high degree of phenotypic plasticity in juvenile leaves for many taxa in the current study is consistent with Rutherford et al. (2016), where adult leaf morphology for many green ash species was found to be highly variable. These findings suggest that phenotypic plasticity in leaf morphology may have confounded species delimitations in section Eucalyptus and that some traits, which are regarded as key diagnostic features (e.g., leaf width), are not so useful in understanding evolutionary relationships and patterns of divergence. In the current study, petioles were present on the first true (opposite) leaves of E. regnans, E. fastigata, E. obliqua, and E. cunninghamii. In contrast, the first true opposite leaves were sessile in all the other study species. In the Bayesian phylogeny, the green ash tall trees (E. regnans, E. fastigata, and E. obliqua) were monophyletic and occupied a position as sister to the remainder of the group, while E. cunninghamii was sister to a clade comprising all other green ash mallees and medium trees. Therefore, petiole presence or absence on the first true opposite leaves is associated with phylogenetic relationships, and should be investigated as a potentially genetically fixed trait in section Eucalyptus.
In groups like the eucalypts, where reproductive isolation is weak and hybridization occurs between closely related species, deciding how species and other entities should be delineated for management can be difficult and problematic (Shepherd and Raymond, 2010). There has been considerable disagreement regarding the number of recognized species in the green ash group. While Brooker (2000) considers there to be only 13 species in the group, other authorities recognize more (e.g., Hill, 2002). However, where species boundaries are drawn in the green ashes has significant implications for their management and conservation. For example, many populations of E. burgessiana occur in protected areas (e.g., in the GBMWHA). However, E. obstans, which was once widespread in the Sydney region is much less common (Benson and McDougall, 1998), with some populations occurring outside protected areas (e.g., the Beacon Hill population). If E. burgessiana and E. obstans are considered the “one species” (as defined by Brooker, 2000), then they may be regarded as sufficiently conserved under the current protected areas. However, in Rutherford et al. (2016), E. obstans and E. burgessiana did not form a clade. Therefore, if considered to be two separate taxa (as defined by Hill, 2002), then E. obstans may not be sufficiently protected.
The ecological impacts of climate change will be largely dependent on the ability of species to respond to changing conditions (McLean et al., 2014). The survival and persistence of plant species under climate change will be influenced by the extent of plasticity in functional traits, as well as a number of other factors, such as reproduction, recruitment, dispersal, population size, and competition (Lavergne et al., 2010). It has been suggested that species and populations that display higher plasticity will be able to adjust more rapidly and therefore, may be in a better position to respond to changing climates (West-Eberhard, 2005; Lande, 2009). The high degree of phenotypic plasticity of some species observed here (e.g., E. fastigata and E. stricta) could therefore be advantageous under climate change because it could allow those species to respond to new and variable environments. Conversely, species that are less plastic and are restricted in distribution (e.g., E. cunninghamii and E. triflora) could be particularly vulnerable. For example, leaf size has been found to be highly plastic in some other species of Eucalyptus (e.g., E. tricarpa, McLean et al., 2014). It has been suggested that high plasticity in leaf size may enable leaf hydraulic, stomatal, and boundary layer conductance to be adjusted in response to changes in evaporative demand and light availability (Yates et al., 2010). Other traits where high plasticity may be advantageous in response to altered environmental conditions are seedling height and plant biomass (which are both associated with plant performance, Poorter et al., 2008). Given the large number of rare taxa in this group—and that it includes many commercially important species—future investigations into plastic responses to changing environments is warranted. For example, investigations into both seedling and adult plant responses to changes in temperature and atmospheric carbon would further enhance our understanding of the ability of rare and common green ash species to cope with climate change.
CONCLUSIONS
We found significant plasticity in plant and leaf functional traits in response to changes in resource treatments (particularly nutrients). Plant- and leaf-level plasticity in the green ashes were not phylogenetically constrained. Our results suggest that plasticity in leaf morphology (especially leaf width) of many of the study species may have confounded species delimitations, and this has significant implications for their management because of the many rare and restricted taxa, as well as the commercially important species within the group. We found more widespread species to have higher leaf-level plasticity than species with more restricted distributions. A better understanding of plasticity of the green ashes is likely to be important in predicting their ability to respond to future climate change.
ACKNOWLEDGEMENTS
The authors thank Geoff McDonnell, Josh Griffiths, Clara Pang, Justin Wan, Richard Johnstone, Andrew Orme, Joel Cohen, Aaron Smith, and Lawrence Mou for technical support. The authors thank Maureen Phelan for advice on experimental design, Chris Allen for advice on substrates, Fatih Fazlioglu for advice on using the plasticity index, Peter D. Wilson and Jason Bragg for advice on data analysis, and Miguel Garcia for facilitating access to resources in the Daniel Solander Library (Royal Botanic Garden, Sydney). SR was in receipt of an Australian Postgraduate Award at the time this research was undertaken. The authors thank Jean-François Flot for helpful comments on the manuscript. We also thank two anonymous reviewers for comments that significantly improved the manuscript.
