Variation in leaf morphological, stomatal, and anatomical traits and their relationships in temperate and subtropical forests

Leaf functional traits have attracted the attention of ecologists for several decades, but few studies have systematically assessed leaf morphological traits (termed “economic traits”), stomatal (termed “hydraulic”), and anatomical traits of entire forest communities, thus it is unclear whether their relationships are consistent among trees, shrubs, and herbs, and which anatomical traits should be assigned to economical or hydraulic traits. In this study, we collected leaf samples of 106 plant species in temperate forests and 164 plant species in subtropical forests and determined nine key functional traits. We found that functional traits differed between temperate and subtropical forests. Leaf traits also differed between different plant functional groups, irrespective of forest type; dry matter content, stomatal density, and cell tense ratio followed the order trees > shrubs > herbs, whereas specific leaf area and sponginess ratio showed the opposite pattern. The correlations of leaf traits were not consistent among trees, shrubs, and herbs, which may reflect different adaptive strategies. Principal component analysis indicated that leaf economics and hydraulic traits were uncoupled in temperate and subtropical forests, and correlations of anatomical traits and economic and hydraulic traits were weak, indicating anatomical traits should be emphasized in future studies.

ranging from 4.78 to 47.32 µm (Supplementary Fig. S2); in subtropical forests, average SL was 10.91 µm, ranging from 2.79 to 43.40 µm. SD ranged from 14.88 to 961.31 pores mm −2 with a mean of 184.35 pores mm −2 in temperate forests ( Supplementary Fig. S3), and SD of plants in subtropical forests ranged from 11.16 to 1403.27 pores mm −2 with an average of 255.12 pores mm −2 . The average SPI was 3.35% and 2.31% in temperate and subtropical forests, respectively. SL and SPI were higher in temperate forests than in subtropical forests, whereas SD was higher in subtropical forests. Furthermore, SL was the highest in herbs, followed by shrubs and trees, and the opposite pattern was observed regarding SD (Fig. 1). CTR  www.nature.com/scientificreports www.nature.com/scientificreports/ forests ranged from 4.37 to 24.15 µm, with an average of 10.70 µm, and was significantly higher than in subtropical forests, ranging from 2.14 to 47.77 µm with an average of 8.60 µm. Regarding plant functional groups, the order of CTR was trees > shrubs > herbs, that of SR was trees < shrubs < herbs, and AB was highest in herbs, irrespective of the forest type (Fig. 1).

Differences in leaf anatomical traits.
Correlation of leaf functional traits and principal component analysis results in temperate and subtropical forests. Weak but significant correlations were observed among leaf economic, hydraulic, and anatomical traits ( Table 2). The relationships between leaf traits were not always observed in trees, shrubs, and herbs, and some relationships only existed in specific plant functional groups and in one of the two forest types (Supplementary Figs S4-11. Principal component analysis (PCA) was employed to test the association of leaf economic, hydraulic, and anatomical traits (Fig. 2). PCA axis 1 explained 29.3% of the total variation and showed strong loadings from leaf morphological traits. PCA axis 2 explained 24.3% of the total variation and showed strong loadings from leaf anatomical traits. PCA axis 3 explained 18.0% of the total variation and showed strong loadings from leaf stomatal traits (Table 3 and Fig. 2).

Discussion
Economic, hydraulic, and anatomical leaf traits varied greatly between temperate and subtropical forests. Regarding economic traits, plants in subtropical forests had higher DMC and LT and lower SLA, compared with plants in temperate forests. A high DMC indicates little intercellular space and high mesophyll resistance to gas diffusion 18 ; therefore, diffusion resistance in subtropical forest plants with a high DMC may be increased to decrease leaf transpiration. Higher LT can increase photosynthetic capacity under strong irradiance by increasing the nitrogen content and the volume of the photosynthetic machinery per unit leaf area, and thick leaves can prevent sunlight damage 19 , which may explain the higher LT in plants in subtropical forests. Lower SLA indicates higher construction cost per unit area, and SLA can reflect the potential of leaves to capture light, therefore higher SLA in plants in temperate forests may be an adaption to low light intensity.   www.nature.com/scientificreports www.nature.com/scientificreports/ Regarding hydraulic traits, subtropical forests had higher SD and SPI and smaller SL than plants in temperate forests. Smaller stomata have a higher surface area to volume ratio, so that they can respond quickly to environmental changes by opening and closing rapidly 20,21 ; higher SD can help reduce CO 2 diffusion resistance caused by a large mesophyll surface under strong irradiance 22 . This may explain small and dense stomata in plants growing in subtropical forests. Furthermore, the rates of cell division are severely constrained by low temperatures, which is reflected in larger and fewer stomata in plants in temperate forests 23 . In line with the results of Sack, et al. 24 , the SPI correlated with the maximum stomatal conductance. CO 2 diffusion is reduced at low temperatures, therefore reduced stomatal conductance likely led to a larger SPI in temperate forests. Regarding leaf anatomical traits, temperate forest had thicker AB and larger CTR than those of subtropical forest. Thicker AB can reduce reflection of scattered light 25 and larger CTR can enhance the cold resistance of plants 26 ; therefore, temperate forest had a higher AB and larger CTR, which is likely an adaption to the cold climate and low light intensity.
The differences in leaf functional traits regarding morphology, stomata, and anatomy were significant between plant functional groups in temperate and subtropical forests. Regarding morphological traits, DMC produced the order trees > shrubs > herbs in both forest types. Higher DMC increases moisture diffusion resistance in leaves, and DMC is an indicator of leaf water content. Hydraulic effects are typically stronger in larger plants 27 , and DMC was indeed highest in trees, followed by shrubs and herbs. SLA showed the opposite pattern, which has previously been observed other in forest types [28][29][30][31][32][33] , and shaded leaves may increase the efficiency of light capture by increasing SLA 34 . This may be a result of the adaptation to declining light intensity form trees to shrubs to herbs. Regarding leaf stomatal traits, SL was highest in herbs, and SD showed the opposite pattern in both forest types, which was in line with the results of Wang, et al. 35 and Liu, et al. 36 , probably because large stomata are critical for herbs to optimize light capture in light-limited environments 20 , and a higher SD increases the ability to regulate leaf transpiration 22 . Regarding leaf anatomical traits, CTR showed the order trees > shrubs > herbs, and SR produced the opposite pattern. Previous studies have shown that CTR decreased and SR increased with a decrease in light intensity 8,37 . Furthermore, photosynthesis occurs predominantly in fully developed palisade tissue 38 and spongy tissue reduces the loss of photons and thus improve photosynthetic efficiency in low-light conditions 39 , which supports our observations of the highest CTR in trees and lowest CTR in herbs. Bone, et al. 40 suggested that higher AB is an adaptation to low-light conditions as it can reduce the reflection of scattered light and increase irradiation intensity within the leaf, which indicates that plants with higher AB may be better adapted to light-limited environments.
Significant correlations of leaf functional traits were not entirely consistently in trees, shrubs, and herbs, and most correlations of leaf functional traits only occurred in specific plant functional groups; however, the correlation of SL and SD was observed in all plant functional groups, which may be explained by physical and energetic constraints 15,20 . Trait correlations may vary as a response to adaption to changing environments, and in a stable environment, multiple combinations of traits that are costly would not be required 17 . For instance, SLA was negatively correlated with DMC and LT, which indicated water retention ability because higher SLA may increase water loss but higher DMC and LT may increase moisture diffusion resistance and distance; however, these negative correlations were not observed in herbs in temperate forests, possibly due to low water loss in low-temperature and light-limited environments. Climate can shape and shift functional biodiversity in forests, and our results indicate that trait correlations may differ between environments, which may be an aspect to consider regarding global climate change.
Correlations of leaf economic and hydraulic traits were very weak, which was consistent with the results of Li, et al. 16 , but contradicted those of Yin, et al. 17 , which may be because water availability was not a limiting factor for plant growth at our study sites. Interestingly, we found that also the correlations of anatomical traits with economic and hydraulic traits were weak. Leaf anatomical traits and hydraulic traits were also uncoupled on the arid Loess Plateau (Supplementary Table S2), and numerous studies found that leaf anatomical traits closely correlated with temperature 41,42 , therefore temperature may be another factor to consider regarding anatomical adaptation strategies and resource usage by plants. As Li, et al. 16 suggested, multi-dimensional adaptation strategies would be expected as an adaption to multifactorial changes in the environment.

Conclusion
There are few systematic studies on the functional traits on entire forest communities. This is the first study to investigate leaf functional traits regarding morphology, stomata, and anatomy of an entire forest community. Our study demonstrated that trait correlations are not consistent, and leaf anatomical traits were uncoupled from leaf economic and hydraulic traits. Anatomical traits and trait co-variation should be further investigated in the future.
Material and Methods study site. The experiments were conducted in temperate and subtropical forests, as these account for 88.3% of the forest area of China. To avoid the effect of anthropogenic disturbance, sampling plots were established within well-protected national nature reserves in China where the vegetation was relatively homogenous and representative of the respective forest type 43 : Temperate forests. Temperate forest plots were established in Changbai Mountain Natural Reserve (42°24′N,  128°05′E, 758 m.a.s.l.), China. The mean annual temperature was approximately 3.6 °C, and mean annual precipitation was 691 mm 44 . The soil was categorized as dark brown forest soil. The vegetation was typical of temperate forests with dominant tree species such as Pinus koraiensis, Tilia amurensis, and Quercus mongolica, among others, dominant shrubs such as Corylus mandshurica, Acanthopanax senticosus, and Ribes mandshuricum, and dominant herbs such as Brachybotrys paridiformis and Phryma leptostachya 45,46 . www.nature.com/scientificreports www.nature.com/scientificreports/ Subtropical forests. Subtropical forest plots were established in the Dinghu Mountain Nature Reserve (N,  23°17′N, 112°54′E, 240 m.a.s.l.), China. The mean annual temperature was approximately 20.9 °C, and mean annual precipitation was 1955 mm 44 . The soil was characterized as a latosol. The vegetation type was considered a southern subtropical evergreen broad-leaved forest, with dominant trees Castanopsis chinensis, Schima superba, and Cryptocarya chinensis, dominant shrubs Psychotria rubra and Ardisia punctata, and dominant herbs Chloranthus spicatus and Alpinia japonica 46 . sample collection. Samples were collected from July to August 2013. First, we established four plots (30 × 40 m) in each forest type. Geographic information (latitude, longitude, and altitude), plant species composition, and ecosystem structure were recorded for each plot. The number, height, diameter at breast height (≥2 cm) of all trees, basal stem diameter of shrubs, and coverage of herbs (and other data) were recorded. Leaf samples of trees, shrubs, and herbs were collected in and around the plots. Briefly, four healthy trees of each species were selected for collecting leaves from the middle and the canopy using long-handle shears or by climbing. From each plant species in each plot, 20 healthy mature leaves were collected from four individuals, representing one replicate. Leaf samples were placed in sealable plastic bags and immediately stored on ice in a cooling box. Measurements or pre-treatments were completed as soon as possible after collection (within 4-8 h).
In total, 106 plant species of temperate forests and 164 plant species of subtropical forests were collected, comprising 95 families and 216 genera ( Table S1).

Measurement of leaf traits.
Leaf traits were assessed regarding morphology, stomata, and anatomy ( Supplementary Fig. S12); units of measurement are shown in Table 1.
Morphological traits. After sampling, LT was measured using a Vernier caliper at an accuracy of 0.02 mm. Leaf area (LA) was measured in 16 leaves (four groups) using a scanner (Cano Scan LIDE 100, Japan) and Photoshop CS software (Adobe, USA). Leaf fresh weight (LFW) was measured using an electronic balance at an accuracy of 0.0001 g; the leaves were subsequently dried to constant weight in an oven 44 to measure leaf dry weight (LDW). DMC (g kg −1 ) and SLA (mm 2 mg −1 ) were calculated according to the following equations: Stomatal traits. After field sampling, eight to ten clean leaves were cut into small pieces (1.0 × 0.5 cm) along the main vein. The pieces were then placed in a formalin acetic alcohol solution (50%; alcohol:formalin:glacial acetic acid:glycerin = 90:5:5:5) as soon as possible (3-6 h after collection) for analyzing leaf stomatal and anatomical traits.
We first examined the pre-treatment leaf samples in the laboratory after air-drying and scraping off surface hair using a razor blade. Stomatal traits were observed by scanning electron microscopy (Hitachi SN-3400, Hitachi, Tokyo, Japan). Three replicates of each species were examined, and two photographs of each replicate were taken in different positions. In each photograph, five stomata were randomly selected to measure SL using an electronic image analysis equipment (MIPS software, Optical Instrument Co., Ltd., Chongqing, China) and the number of stomata (N) in each photo was recorded 36 . SD and SPI were calculated as follows: where 1.12 × 10 −2 is the area of observed photo (mm 2 ).
Anatomical traits. Samples fixed in a formalin acetic alcohol solution were serially dehydrated in ethanol (50-100%) and were then infiltrated using warm paraffin (56-58 °C). Leaf samples of 8-12 µm in size were produced using a rotary microtome (Leica RM 2255, Leica Instruments, Nussloch, Germany). The slides were stained using safranin and fast green (1% aqueous safranin and 0.5% fast green in 95% ethanol). Then, the sections were photographed to measure LT (in µm), palisade thickness (PT; in µm), spongy tissue thickness (ST; in µm), and AB (in µm) producing five measurements per photograph 41 . CTR and SR were calculated as follows: Data analyses. The respective average of all leaf traits was calculated at the species level based on all replicates. The plant species were divided into three functional groups, namely trees, shrubs, and herbs. Then, leaf www.nature.com/scientificreports www.nature.com/scientificreports/ traits at the plant functional group and community level were calculated as the pooled mean of leaf traits per plant functional groups and community.
To compare the differences in leaf traits among different plant functional groups between forest types, a one-way analysis of variance with a least-significant-difference test was performed. A t-test for independent samples was used to test differences in leaf traits between forest types. Correlations of leaf traits were tested using Pearson's correlation after log 10 -transformation of the data to meet the assumption of normal distribution. To test whether the relationships between leaf traits were consistent among trees, shrubs, and herbs, linear and non-linear regressions were fitted, and models with a lower Akaike's information criterion values were chosen as the best-fitting models. Multivariate associations of leaf traits were analyzed with a PCA using R software (version 2.15.2, R Development Core Team 2012). Data analyses and visualization were performed using SPSS 13.0 (SPSS Inc., Chicago, IL, USA, 2004) and SigmaPlot 10.0 software (Systat, USA). Statistical significance is reported at P < 0.05.

Data Availability
The datasets generated and analyzed in the current study are included in its Supplementary Information files.