Contrasting Vegetation Gradient Effects Explain the Differences in Leaf Traits Among Woody Plant Communities in the Amazonia-Cerrado Transition

The link between functional traits and changes in the environment are challenging, especially in systems of high diversity, such as transitional regions between savannas and tropical forests that are considered unstable and highly vulnerable to climate change. Here we evaluate a series of anatomical traits, key morphological traits, and nutrient concentration for the most representative woody species of four vegetation types (i.e. semideciduous seasonal forest, transition forest, typical cerrado and rocky cerrado) in the Amazonia-Cerrado transition. Our aim was to understand how trees respond to the distinct and unique environmental conditions in each area. Overall, species growing in closed-canopy environments exhibited acquisitive strategies (i.e., higher leaf water content and higher specic leaf area). On the other hand, species from open-canopy environments have adopted conservative strategies (i.e., higher leaf thickness and higher trichomes density). These differences represent a divergence in the adaptive strategies of woody species of the Amazonia-Cerrado transition and are important indicators of how vegetation can respond to future climate change and water availability. Despite these differences, stomatal density and leaf nutrient concentration at a mass basis was surprisingly similar across trees from all vegetation types, suggesting some common environmental drivers, such as likely similar nutrient limitation and climate seasonality.


Introduction
Trees display different ecological strategies in response to contrasting habitats under distinct environmental stresses (Borges et al. 2019). Ecological strategies are expressed as a balance between functional traits and the environment, driven by environmental constraints, which may favour different sets of traits (Wright et al. 2006). Functional traits are morpho-physiological-phenological characteristics that directly or indirectly affect plant performance through their effects on growth, survival, and reproduction (Violle et al. 2007). Such adaptations are essential for the survival and expansion of species ecological limits (Ghalambor et al. 2007), and are governed by modi cations in leaf traits, which enable species to overcome different environmental lters (Borges et al. 2019), particularly associated with resource availability, such as water, nutrients, and light (Wellstein et al. 2017). The coordinated relationship between functional traits can be expressed in a gradient of strategies that ranges from fast acquisition to e cient resource conservation (Wright et al. 2004;Reich 2014). Functional traits linked to faster growth rate and higher productivity (acquisitive strategy) are favourable in environments with high water and light availability enhancing photosynthetic e ciency. In contrast, traits representing higher resource conservation and drought tolerance (i.e., conservative strategy), are widely reported in environments under water and nutritional de cits In the Amazonia-Cerrado transition, we can nd contrasting vegetations side by side as is the case in our study, two are open-canopy (typical cerrado and rocky cerrado) and two are closed-canopy (semideciduous forest and transitional forest) that represent an open-air laboratory (Marimon- Leaves are the most sensitive and environmentally adaptable organs (Li and Bao 2014), being directly exposed to sunlight and high temperatures (Dickison 2000). In an environment under water limitation and high light intensities, plants tend to develop thicker leaves to avoid leaf tissue damage (Cornelissen et al. 2003;Rozendaal et al. 2006). Higher leaf thickness also allows species to increase water conservation and resistance to physical damage (Yates et  American biomes, the Brazilian Savanna, known as the Cerrado, and the Amazonia, there are an extensive transitional area, the Cerrado-Amazonia transition, with unique characteristics, such as a shared oristic composition, edaphic and climatic parameters, and occurrence of res, mainly in savanna formations (Furley and Ratter 1988;Ratter 1993;Marimon et al. 2006Marimon et al. , 2014. These transitional regions are considered relatively unstable and may be highly susceptible to global climate change (Marimon et al. 2014; Morandi et al. 2018;Araújo et al. 2021a). The prediction regarding the climate is that a progressive increase of the temperature in the next decades (Gatti et al. 2014) will continue to occur, causing more frequent and more intense dry seasons (Coe et al. 2016). Furthermore, in this region, the photosynthetic functions of trees are affected by high temperatures, making photosystem II more vulnerable to ongoing and future climate changes, putting at risk the persistence of tree species over time (Tiwari et al. 2020; Araújo et al. 2021a). Therefore, understanding leaf morphological and anatomical patterns at the community level are fundamental to clarify how selective processes act on different species and to describe functional patterns (Pearson and Dawson 2003).
Our study aims to compare morpho-anatomical traits and leaf nutrient concentrations in distinct plant communities along a gradient with multiple environmental factors in the Cerrado-Amazonia transition. Here we introduce a set of ecologically important functional traits that promote new insights into the combination of leaf traits (e.g., stomata density, stomata size, maximum opening of stomata pore and trichomes density) in addition to morphological traits and concentration nutrients in the leaves that allow to understand the dimensions of the ecological strategies adopted by woody plants in different vegetation types.
The plant species occur along a gradient with multiple combined environmental factors (Marimon-Junior and Haridasan 2005; Araújo et al. 2021b). We wait for a gradient response in leaf traits from the vegetation more exposed to light, high temperature, and under stronger water limitation (rocky cerrado and typical cerrado) to the closed-canopy semideciduous seasonal forest. We hope that the cerradão (i.e., . Speci cally, we hypothesize that community-level traits will display a gradient response going from functional traits linked to a faster growth rate and typical of closed-canopy environments (e.g. higher speci c leaf area, higher leaf water mass content, higher leaf nutrient concentration and larger stomata) to functional traits linked to a slower growth rate and typical of opencanopy environments (e.g. higher trichomes density, higher leaf thickness, lower speci c leaf area and lower leaf nutrient concentrations and smaller stomata) (

Study area
The four study areas are in Nova Xavantina, eastern Mato Grosso state, Central Brazil, in the Amazonia-Cerrado transition zone ( Figure 1). The climate is seasonal with two well-distinct periods, the rainy

Data collection
We selected seven species that represented (> 80% IVI) in the plant communities in each vegetation type, according to the importance value index (IVI, Table 1), which considers species relative density, frequency, and dominance (Martins 1991). For each species, we sampled the ve largest individuals and, for each one, we selected terminal branches completely exposed to the sun to avoid light variations throughout the day and we selected eight fully expanded tree canopy leaves or pathogen-free lea ets in the peak rainy period (February). We use ve leaves to measure morphological traits and with the same leaves, we determine the concentration of foliar nutrients and we collect more three leaves to measure anatomical traits.

Leaf morphological traits and nutrient analysis
Functional traits assessed in this study are listed in Table 2. We measured leaf thickness (mm) using an electronic digital micrometre (± 0.001 mm). Fresh leaves were scanned, and leaf area measured with ImageJ software (Abràmoff et al. 2004). We estimated the wet weight of the leaves with a precision balance (± 0.001 g), then placed them in paper bags in an oven at 60 °C, and after 72 h determined the dry weight. The leaf water content (g g -1 ) was calculated as the ratio between wet and dry weights. The speci c leaf area (cm 2 g -1 ) was calculated as the ratio between leaf area and leaf dry mass (Pérez-Harguindeguy et al. 2013). Dried samples were sent to the laboratory of Soil Department of the Federal University of Viçosa (UFV, MG, Brazil), for nutrient concentration analyses. N and P (g kg -1 ) were determined by Kjedahl digestion and UV-Vis spectroscopy, respectively, and the other nutrients (Ca, Mg and K, g kg -1 ) were determined by atomic absorption spectrometry.

Leaf anatomical traits
We sampled three leaves of each individual and, on abaxial and adaxial leaf surfaces, we applied the foliar surface imprinting method (Weyers and Johansen 1985) using high moulding technology silicone.
After drying, we obtained reverse imprints from the moulds, using colourless nail polish and photographed the imprints on a light optical microscope (Zeiss Primo Star), with a coupled camera taken under microscopic magni cation of 40x. We randomly selected 10 elds from each leaf and processed the images with ImageJ software (Abràmoff et al. 2004). The measured parameters were: guard cell length ("L" in μm), guard cell pair width ("W", in μm), stomata size ("S", estimated as S = L * W, according to Franks et al. 2009a;b) and maximum area of the stomatal pore ("amax", in μm 2 ). The maximum area of the stomatal pore was calculated as amax = α * S, where α = 0.12, according to Franks and Beerling (2009). Stomata density (mm -2 ) was calculated for each individual as the average number of stomata counted in a given area. The same area was used to calculate trichomes density (mm -2 ), when present. A total of 25 stomata per individual were measured, and stomata densities, lengths, and widths were averaged per species.

Statistical analysis
All analyses were performed in R software (2019). Before performing all analyses, we tested the normality and homocedastity of the data using the Shapiro-Wilk and Levene tests (Levene 1961;Shapiro-Wilk 1965). All variables were log10-transformed when necessary. We created a variance partition to understand which organizational levels (species and individuals) best explain the variation of each functional traits and, for that, we used different groupings of linear models and adjusted the separate models for each functional traits (Rosas et al. 2019). We introduce species and individuals as nested random factors and vegetation type as ( xed factor) to analyze how the variation of functional traits are distributed among these organizational levels, and we use functional traits as the response variable for each of the models. To test for differences in functional traits between vegetation types, we compared the functional traits using Generalized Linear Mixed Models (GLMM) in the lme4 package (Bates et al. 2014), with the vegetation type as predictors ( xed effect) and species as a random effect. We carried out a principal component analysis (correlation PCA), to investigate how leaf functional traits were distributed among tree communities. In order to estimate allometric lines of best t for the bivariate relationships, we tted standardized major axis and tested for differences in inclination (slopes) between the four vegetation types using SMATR package (Warton et al. 2012).

Results
For most of the leaf traits we measured, the largest variations were attributed to differences between species, ranging from > 65% of the variation explained by species differences for leaf nutrient concentration, 72% for morphological traits, and 70% for anatomical traits (Figure 2; Table S1). Individuallevel explained between 36 to 47% of the variation in K and the density and size of stomata ( Figure 2).
Species that grow in closed-canopy environment generally showed different leaf traits compared to species that grow in open-canopy environments, except for leaf nutrient concentration at a mass basis (Table 3; Table S2). Species in the semideciduous forest showed higher speci c leaf area and higher leaf water content (Table 3). Unexpectedly, species in the semideciduous forest showed the maximum area of the stomatal pore and stomatal size similar to species that grow in rocky cerrado (Table 3). On the other hand, species in open-canopy environments showed smaller SLA and thick leaves, combined with higher trichomes density (Table 3). At a mass basis (mg g -1 ), there was no diference between species from distinct vegetation types for leaf macronutrient concentrations (Table S2; Table S3). At an area basis, however, species growing in the semideciduous forest showed lower leaf concentration of N, P and Mg compared to species in other vegetation types (Table 3). On the other hand, the species of the transitional forest, typical cerrado and rocky cerrado showed similar leaf nutrient concentrations (Table 3). There were no differences in leaf K and Ca concentration between vegetation types (Table 3; Table S4). Indeed, the major source of variability for all nutrients were within species (Figure 2).
The PCA explained 51% of the total data variation in the rst two axes and the variation explained by the ve axes contributed with > 80% (Table S5), and differences between the distinct plant communities re ected in clear discrimination between closed and open-canopy vegetation, revealed by PC2 (Figure 3 Figure 4a). Stomata density was also positively related with leaf N concentration, but only for transition forest and rocky cerrado (transition forest: R 2 = 0.19, P = 0.042; rocky cerrado: R 2 = 0.24, P = 0.023; Figure  4b). There was also a positive relationship between leaf P concentration and stomata size for the two closed-canopy communities (semideciduous forest: R 2 = 0.15, P = 0.048; transition forest: R 2 = 0.33, P = 0.006; Figure 4c). We did not nd any relationship between anatomical traits, and there was no relationship between stomata density and stomata size for all data together or for each vegetation separately.

Discussion
Together, our ndings support that the trees from distinct vegetation types exhibit strategies related to invest in water-saving (species open-canopy environments) and optimize carbon acquisition (species closed-canopy environments) ( dimensions are strongly associated with seasonal environments and re ect climate adaptations. In the rocky cerrado, where access to water and soil nutrients is more limited (Marimon and Haridasan 2005) we expected that the stomatal dimensions would be smaller than the values recorded in the semideciduous forest (Franco 2002;Pearce et al. 2006). However, in this case, it is possible that other morpho-anatomical strategies, such as the greater thickness leaves (Table 3)  Overall, species growing in closed-canopy environments exhibited acquisitive strategies (i.e., higher leaf water content and higher speci c leaf area). On the other hand, species from open-canopy environments have adopted conservative strategies (i.e., higher leaf thickness and higher trichomes density), validating the economic spectrum leaf and re ecting the environmental differences between these vegetation types We also found that species in the semideciduous forest invest in higher speci c leaf area at a given leaf N concentration probably to overcome light competition by increasing the leaf area, light interception, and photosynthetic rates (Grime 1983 , from an evolutionary point of view, the investment in leaf area may be more advantageous, even considering that thinner leaves are more susceptible to herbivores or prone to water loss (Westoby et al. 2002). Unexpectedly, in our study semideciduous forest species showed the lowest N and P leaf concentration on an area basis, which may re ect a strong nutrient limitation in these forest soils (Marimon et al. 2014). Furthermore, species from open-canopy environments invested in higher leaf thickness that helps reduce leaf damage caused by herbivores and increase leaf lifespan (Grime 1983;Cornelissen et al. 2003;Bündchen et al. 2015). Interestingly, species from the transitional forest showed markedly similar traits compared to species from rocky cerrado and typical cerrado, such as low SLA, high TRD and LET, which indicate mixed strategies responding to mixed environmental drivers. These ndings reinforce the idea that transitional forest is typical contact areas between savannas and forests (Ratter 1993;Ivanauskas et al. 2008), especially in the Amazonia-Cerrado transition (Marimon et al. 2014;Marques et al. 2020). These different combinations of ecological strategies in transitional forest can be advantageous in dealing with different environmental pressures, an essential condition for the persistence of species over time.
The positive relationship between N leaf tissue concentrations and stomata density for rocky cerrado and for transitional forest species may suggest a stronger pressure to optimize resources in these vegetations, particularly water and nitrogen, probably to maximize photosynthesis (Wright et al. 2003). Moreover, P and stomata size scaled positively, but only for forests formations, where plants should invest in primary growth to reach the canopy and successfully compete for light. Whereas P limitation is associated with lower wood density and greater hydraulic conductivity (Resco de Dios 2003), the coordination between P availability and stomata size (re ecting the anatomical adjustment to stomata opening control) in low-P soil and light-limited vegetation might be a key adaptation. Thus, the interaction between nutrients and water availability may have critical implications for the future distribution of plants and their responses to increasing drought severity and length (Cramer et al. 2009).
In addition, the relatively high trichomes density observed in rocky cerrado and typical cerrado species, may also re ect adaptation to control water de cit due increase water vapour concentration around the leaf boundary layer (Fahn and Cutler 1992;Larcher 2000). It is noteworthy that the combination of high temperature, high light incidence and low humidity of the air are determining factors in uencing opencanopy environments species (Franco 2002;Araújo et al. 2021b). Trichomes are important adaptive strategies that help decrease light incidence and leaf temperature (Klich et al. 1997), which also promote water saving and concurrently avoid photoinhibition damage.
There was an overlap in leaf traits between all vegetation types analysed, which may be expected in a transition zone characterized by ecological tensions (Furley and Ratter 1988;Ratter 1993;Marimon et al. 2006Marimon et al. , 2014. Notably, the similar leaf nutrient concentration at a mass basis and similar stomata density between trees of all vegetation types suggest some common environmental drivers, such as likely similar nutrient limitation and climate seasonality. However, morphological and anatomical adjustments are associated with a strong connection between plant structures and functioning and we could nd two major clusters with different ecological strategies, where trees from closed-canopy habitats optimize carbon acquisition, while trees in open-canopy environments invest in water-saving strategies (Bedetti et al. 2011). In this context, our hypothesis was partially supported.
Our ndings suggest that trees from closed-canopy environments may be more vulnerable to drought events as they present a combination of functional traits that are less safe to deal with prolonged droughts, especially semideciduous forest species that share no traits to manage larger water balance during droughts and may suffer more from water stress. severe consequences, such as changes in plant communities composition, structure and functioning. Therefore, processes such as water use e ciency will be critical for the survival of species that inhabit hot and dry regions (Hulme 2005 Leaf phosphorus P g cm -2 Leaf potassium K g cm -2 Leaf calcium Ca g cm -2 Leaf magnesium Mg g cm -2 Speci c leaf area SLA cm 2 g -1 Leaf thickness LET mm Leaf water content LWC g g -1 Trichomes density TRD mm -2 Stomata density STD mm -2