Leaf Venation Variation and Phenotypic Plasticity in Response to Environmental Heterogeneity in Parrotia subaequalis ( H . T . Chang

Plant leaf vein traits are partially the result of adaptation to environmental factors during long-term evolution. For terrestrial plants, leaf veins greatly vary in size and numbers. Parrotia subaequalis (H. T. Chang) R. M. Hao et H. T. Wei, an endangered tree species endemic to China, has a limited distribution, and inhabits both hillsides and valleys. The variations in P. subaequalis leaf venation and vein density in response to environmental changes were examined by collecting samples from all 14 extant populations and analyzing the association between leaf vein density and environmental factors. The results revealed that leaf characteristics were strongly associated with different habitats. A set of vein traits, namely base angle, intercostal tertiary areole development and shape, and free ending veinlet branching, were related to habitat. Significant relationships between vein density and environmental variables (mean annual temperature, mean annual precipitation, and elevation) were doubtless confirmed by this study. These findings indicate that phenotypic plasticity in leaf vein traits is an important ecophysiological characteristic that enables P. subaequalis to adapt to spatiotemporally fluctuating environments. Furthermore, these results also provide important reference data for in-depth studies on the protection strategies used by the tree.


Introduction
Environmental conditions in combination with resource availability appear to be key factors involved in determining the distribution and functional characteristics of a species inhabiting a particular region [1].Higher plants typically cope with varying environmental conditions by altering their tissues and organs [2].The leaf morphology of angiosperm plants varies across different taxa.The leaves of plants are highly diverse and have various leaf venation patterns.Leaf venation is an important morphological structure, which characterizes the distribution and arrangement of the vein system in the leaves.Leaf veins provide mechanical support to display the leaf towards light, and contain the xylem that transports water and nutrients for photosynthesis and transpiration [3].Leaf vein characteristics and functional traits jointly reflect the adaptation of a given plant to local environments [4][5][6].Vein traits, such as vein density (VD), vary considerably across species.VD strongly influences leaf hydraulic conductance (K leaf ), stomatal density, stomatal conductance, and rates of gas exchange per leaf area [7,8].The leaf hydraulic conductance (K leaf ) represents the capacity of the transport system to deliver water, Forests 2018, 9, 247 2 of 17 which affects stomatal opening and photosynthesis [9].There are individual differences in vein density within the same species when the species is subjected to different environmental conditions.These variations are closely related to plant growth environment, precipitation, and temperature [10,11].Vein traits are thought to reflect the gas and water exchange conditions at the interface between the leaves and the atmosphere, which are greatly influenced by the climatic factors affecting the leaf [11,12].Studies focusing on interspecific patterns between plant traits and climatic factors have identified a correlation between leaf vein density and mean annual precipitation (MAP), and dryness of soil water availability can lead to higher leaf venation densities [13].
Phenotypic plasticity is one of two complementary ecological strategies and it refers to the ability of a specific genotype to shape various phenotypes in response to environmental variation [14].Venation properties have been recognized as a type of phenotypic plasticity that is caused by ontogeny.There have been numerous reports on the correlations between leaf vein density and leaf area [15], epidermal cell size [16], stomatal density [17][18][19], stomatal conductance [19], hydraulic conductance [19,20], and photosynthesis [13].However, variations in leaf venation and leaf vein density under different environmental conditions have not been widely reported until recently [21], especially for endangered species.
Parrotia subaequalis (H.T. Chang) R. M. Hao et H. T. Wei [22], a rare and endangered deciduous tree endemic to China, belongs to the family Hamamelidaceae [23].There are two species in this genus (P.subaequalis and Parrotia persica (DC.)C.A. Mey.), and P. subaequalis is an important living fossil angiosperm species that appeared 67 million years ago [22].Due to the limited geographic range, alternate-year fruit production, serious habitat destruction, and increasing anthropogenic disturbance in the form of timber harvesting have occurred [24].This species has been listed as a rare and endangered plant in China [25] and as a "plant species with extremely small populations" (PSESP) [23].P. subaequalis is a tertiary relic plant and occurs in eastern China where there is a disjunct distribution in Anhui, Jiangsu, Henan, and Zhejiang Provinces (Table 1).In eastern China, the annual precipitation is usually over 1000 mm in humid regions.However, P. subaequalis can only grow and thrive within a small range of temperatures and environmental habitats that have a low heterogeneity [5].Characteristic traits, such as leaf size, leaf length, color, and propagation ability, are highly diverse in P. subaequalis [26].To date, there have only been a few studies on the P. subaequalis leaf trait responses to climatic stress, and the sensitivity of P. subaequalis leaf morphology to climate change is poorly understood.Therefore, an analysis of leaf morphology under different climatic conditions can increase the understanding of its adaptive strategies, and this knowledge can be used to improve the conservation of P. subaequalis populations.
The aim of this study was to test the ecological adaptive strategy of P. subaequalis under different habitats with leaf venation and vein density.The VDs of materials with venation from leaves of 14 extant populations along an elevational gradient from 192-988 m was investigated.The objectives were (1) to determine how the P. subaequalis leaf trait varied under different climatic conditions; (2) to quantify the relationships between leaf trait and climatic factors, such as elevation, mean annual temperature (MAT), and mean annual precipitation (MAP); and (3) to set up a scientific foundation for designing conservation strategies.

Leaf Sample Collection
Fresh leaves of P. subaequalis were collected from all 14 known extant populations (Table 1) located across this historic range in eastern China (Figure 1) in July 2016.Five leaves per tree and 10 mature trees per population were selected, and from the four geographical sides of the crown, four outermost branches were randomly harvested.The distance between individual trees was approximately 50 m, which was far enough apart to minimize the possibility of sampling related individual effects [27,28].Voucher specimens were deposited at Nanjing Forest University (NJFU).Long-term climate data collected over 29 years (1981-2010), including MAT ( • C) and MAP (mm), were obtained for each sampling site from the China Meteorological Data Sharing Service System (CMDSSS; www.cdc.cma.gov.cn[29], Table 1).

Leaf Sample Collection
Fresh leaves of P. subaequalis were collected from all 14 known extant populations (Table 1) located across this historic range in eastern China (Figure 1) in July 2016.Five leaves per tree and 10 mature trees per population were selected, and from the four geographical sides of the crown, four outermost branches were randomly harvested.The distance between individual trees was approximately 50 m, which was far enough apart to minimize the possibility of sampling related individual effects [27,28].Voucher specimens were deposited at Nanjing Forest University (NJFU).Long-term climate data collected over 29 years (1981-2010), including MAT (°C) and MAP (mm), were obtained for each sampling site from the China Meteorological Data Sharing Service System (CMDSSS; www.cdc.cma.gov.cn[29], Table 1).

Leaf Venation
Observations of the leaf venation pattern were carried out according to Zhao et al. (2016) [18], with 15 leaf venation characteristics (Table 2) selected for analysis.Fresh or herbarium leaves were cleared according to Dizeo de Strittmatter (1973) [30] and stained in safranin/80% ethanol.The samples were mounted in DPX (Aldrich Chemical Company, Gillingham, UK) and viewed with the Zeiss Stereo Microscopea Zeiss Axiolab microscope (Carl Zeiss, Oberkochen, Germany), photographed using a color digital imaging camera.

Vein Density Measurements
To determine the vein density, five leaves per location were selected.The vein density (mm•mm −2 ) measurement was performed on cleared leaves [31].An approximately 1-cm 2 adaxial section from the right side of the leaf midrib was excised to determine vein density.In total, 70 images were measured and analyzed using ImageJ image analysis software (National Instituted of Health, NIH, Bethesda, MD, USA) (public software; https://imagej.en.softonic.com/)[32].Vein density (mm mm −2 ) was calculated as the sum of the lengths of all vein segments (mm) per unit area (mm 2 ).

Leaf Venation
Observations of the leaf venation pattern were carried out according to Zhao et al. (2016) [18], with 15 leaf venation characteristics (Table 2) selected for analysis.Fresh or herbarium leaves were cleared according to Dizeo de Strittmatter (1973) [30] and stained in safranin/80% ethanol.The samples were mounted in DPX (Aldrich Chemical Company, Gillingham, UK) and viewed with the Zeiss Stereo Microscopea Zeiss Axiolab microscope (Carl Zeiss, Oberkochen, Germany), photographed using a color digital imaging camera.

Vein Density Measurements
To determine the vein density, five leaves per location were selected.The vein density (mm•mm −2 ) measurement was performed on cleared leaves [31].An approximately 1-cm 2 adaxial section from the right side of the leaf midrib was excised to determine vein density.In total, 70 images were measured and analyzed using ImageJ image analysis software (National Instituted of Health, NIH, Bethesda, MD, USA) (public software; https://imagej.en.softonic.com/)[32].Vein density (mm mm −2 ) was calculated as the sum of the lengths of all vein segments (mm) per unit area (mm 2 ).

Data Analysis
Leaf venation characters were checked between populations using the unweighted pair-group method with the arithmetic means (UPGMA) procedure in PAST 3.14 software (public software, http: //folk.uio.no/ohammer/past).Data analysis between vein density and environmental variables was determined using the linear regression analysis procedure in SPSS 22.0 software (SPSS Inc., Chicago, IL, USA).According to the variation in VD due to elevation, MAT, or MAP, we also obtained a more controlled measurement of the effect of habitat on VD through an analysis of covariance (ANCOVA).For statistical analysis, SigmaPlot 12.5 (Systat Software, Richmond, CA, USA) was employed.

Leaf Venation Clustering Analysis of P. Subaequalis
All the taxa analyzed had a basal pinnate primary vein framework, with no naked basal veins, 11-14 basal veins per lobe, and simple agrophic veins (Figures 2A and 3A).The UPGMA clustering analysis of the 15 characteristics showed that the 14 P. subaequalis populations could be divided into two main groups, namely hillside and valley habitat types (Figure 4A,B; Table A1).The hillside habitat type group consisted of eleven populations, such as Ningbo, Shucheng, Yuexi3 (YX3), and Xinyang populations (Table 3, Figures 2A and 5A).This group displayed different freely ending veinlets (FEV) that were similar to those observed in the Yuexi2 (YX2), Xinyang, Huoshan, and remaining populations (Figure 3D-F,H).These populations have an obtuse base angle, quadrangular and pentagonal but rarely triangular areoles, and moderate areole development.The Changhua, Yixing, and Yuexi3 (YX3) populations can be further recognized by two branched FEVs (Table 3).The Jingde and Huoshan populations have three branched FEVs and their areola shapes are more variable (Table 3; Figure 3B,E,I).The Xinyang population has looped marginal ultimate venation (Figure 3C,G).In contrast, the valley habitat type group contained three populations (Jixi, Jinzhou, and Tingeing) (Figures 2B and 5B).They show moderate areole development and quintenary vein fabric is present (Figure 4F).The Jixi, Tongcheng, and Jinzhai populations have almost the same pattern, with two branching FEVs (Figure 4B,F).Jixi can be distinguished by its more variable areola shape, while the other two have quadrangular and pentagonal, but rarely triangular, areoles (Figure 4F).The Jixi and Tongcheng populations can also be recognized by the presence of looped, marginal, ultimate venation (Figure 4C-E).The base angle in the valley group is a right angle and the areola shape is more variable than in the hillside group.However, this group has almost the same vein pattern, with the same laminar shape, areole development, and intercostal tertiary vein fabric.It can be distinguished by its FEVs.The first group mostly displays two or three branched FEVs (Figure 3H), whereas the second group (Figure 4B) generally has three branched FEVs and their density is also different.

Effects of Elevation and Climatic Factors on VD
The vein density significantly varied from 3.2-13.3mm mm −2 across the 14 sites.Elevation across the 14 sites varied markedly, with an elevation range from 192 m to 988 m (Table 1).The VD values of the 14 populations significantly rose as the elevation increased (R 2 = 0.6270, p = 0.0007; Figure 6).Climate conditions, especially MAT and MAP, clearly varied among sites.The VD had a negative In contrast, the valley habitat type group contained three populations (Jixi, Jinzhou, and Tingeing) (Figures 2B and 5B).They show moderate areole development and quintenary vein fabric is present (Figure 4F).The Jixi, Tongcheng, and Jinzhai populations have almost the same pattern, with two branching FEVs (Figure 4B,F).Jixi can be distinguished by its more variable areola shape, while the other two have quadrangular and pentagonal, but rarely triangular, areoles (Figure 4F).The Jixi and Tongcheng populations can also be recognized by the presence of looped, marginal, ultimate venation (Figure 4C-E).The base angle in the valley group is a right angle and the areola shape is more variable than in the hillside group.However, this group has almost the same vein pattern, with the same laminar shape, areole development, and intercostal tertiary vein fabric.It can be distinguished by its FEVs.The first group mostly displays two or three branched FEVs (Figure 3H), whereas the second group (Figure 4B) generally has three branched FEVs and their density is also different.

Effects of Elevation and Climatic Factors on VD
The vein density significantly varied from 3.2-13.3mm mm −2 across the 14 sites.Elevation across the 14 sites varied markedly, with an elevation range from 192 m to 988 m (Table 1).The VD values of the 14 populations significantly rose as the elevation increased (R 2 = 0.6270, p = 0.0007; Figure 6).Climate conditions, especially MAT and MAP, clearly varied among sites.The VD had a negative Forests 2018, 9, 247 9 of 17 linear relationship with both MAP (R 2 = 0.5188, p = 0.01; Figure 7) and MAT (R 2 = 0.093, p = 0.145; Figure 8).The MAP values for the Anji, Changhua, and Ningbo sites were lower than for the other sites.The MATs for Anji and Changhua were also lower than the other populations (Table 4).Anji, Changhua (820 m), and Ningbo (988 m) populations had a higher VD, but the Jinzhai (450 m), Tongcheng (270 m), and Jixi (684 m) populations had a relatively low VD.Therefore, it can be concluded that P. subaequalis growing in a hillside habitat with a lower MAP had a higher VD, and those growing in a valley habitat with higher MAPs had lower VDs.
Forests 2018, 9, x FOR PEER REVIEW 9 of 17 linear relationship with both MAP (R 2 = 0.5188, p = 0.01; Figure 7) and MAT (R 2 = 0.093, p = 0.145; Figure 8).The MAP values for the Anji, Changhua, and Ningbo sites were lower than for the other sites.The MATs for Anji and Changhua were also lower than the other populations (Table 4).Anji, Changhua (820 m), and Ningbo (988 m) populations had a higher VD, but the Jinzhai (450 m), Tongcheng (270 m), and Jixi (684 m) populations had a relatively low VD.Therefore, it can be concluded that P. subaequalis growing in a hillside habitat with a lower MAP had a higher VD, and those growing in a valley habitat with higher MAPs had lower VDs.Forests 2018, 9, x FOR PEER REVIEW 9 of 17 linear relationship with both MAP (R 2 = 0.5188, p = 0.01; Figure 7) and MAT (R 2 = 0.093, p = 0.145; Figure 8).The MAP values for the Anji, Changhua, and Ningbo sites were lower than for the other sites.The MATs for Anji and Changhua were also lower than the other populations (Table 4).Anji, Changhua (820 m), and Ningbo (988 m) populations had a higher VD, but the Jinzhai (450 m), Tongcheng (270 m), and Jixi (684 m) populations had a relatively low VD.Therefore, it can be concluded that P. subaequalis growing in a hillside habitat with a lower MAP had a higher VD, and those growing in a valley habitat with higher MAPs had lower VDs.Forests 2018, 9, x FOR PEER REVIEW 9 of 17 linear relationship with both MAP (R 2 = 0.5188, p = 0.01; Figure 7) and MAT (R 2 = 0.093, p = 0.145; Figure 8).The MAP values for the Anji, Changhua, and Ningbo sites were lower than for the other sites.The MATs for Anji and Changhua were also lower than the other populations (Table 4).Anji, Changhua (820 m), and Ningbo (988 m) populations had a higher VD, but the Jinzhai (450 m), Tongcheng (270 m), and Jixi (684 m) populations had a relatively low VD.Therefore, it can be concluded that P. subaequalis growing in a hillside habitat with a lower MAP had a higher VD, and those growing in a valley habitat with higher MAPs had lower VDs.According to an analysis of covariance (ANCOVA), the result yields a result that populations from hillside habitats have a stronger response to elevation, MAT, and MAP than valley populations.Elevation as a covariate has a significant effect (p < 0.001) on VD, yet habitat remains a significant predictor for VD (p = 0.003) (Table A2).Of course, plants do not respond to elevation, but rather some other environmental variable that is associated with elevation.However, note that the slope for hillside habitats is substantially more positive than for valley habitats (Figure 9).MAT as a covariate was not a significant predictor of variation in VD (p = 0.218), yet habitat was a significant predictor for VD (p = 0.026) (Table A3).This suggests that MAT is not very important, while habitat is important.However, note that the slope for hillside habitats is substantially more negative than for valley habitats, which is rather flat (Figure 10).According to an analysis of covariance (ANCOVA), the result yields a result that populations from hillside habitats have a stronger response to elevation, MAT, and MAP than valley populations.Elevation as a covariate has a significant effect (p < 0.001) on VD, yet habitat remains a significant predictor for VD (p = 0.003) (Table A2).Of course, plants do not respond to elevation, but rather some other environmental variable that is associated with elevation.However, note that the slope for hillside habitats is substantially more positive than for valley habitats (Figure 9).MAT as a covariate was not a significant predictor of variation in VD (p = 0.218), yet habitat was a significant predictor for VD (p = 0.026) (Table A3).This suggests that MAT is not very important, while habitat is important.However, note that the slope for hillside habitats is substantially more negative than for valley habitats, which is rather flat (Figure 10).MAP as a covariate was a significant predictor of variation in VD (p = 0.022), while habitat was not a significant predictor for VD (p = 0.159) (Table A4).This suggests that MAT is important in driving VD, with no additional significant variation in VD explained by habitat (p = 0.159).Note that a MAP with habitat term in the model was not significant (p = 0.113).However, the slope for hillside habitats is very negative, while the slope for valley habitats is positive (Figure 10).MAP and MAT are highly correlated (R = 0.715, one-tailed p = 0.002), with a similar relationship found in both habitats (Figure 11).Additionally, MAP and altitude are negatively correlated (R = − 0.669, two-tailed p = 0.009), but the association is opposite for the two habitats (Figure 11).MAP as a covariate was a significant predictor of variation in VD (p = 0.022), while habitat was not a significant predictor for VD (p = 0.159) (Table A4).This suggests that MAT is important in driving VD, with no additional significant variation in VD explained by habitat (p = 0.159).Note that a MAP with habitat term in the model was not significant (p = 0.113).However, the slope for hillside habitats is very negative, while the slope for valley habitats is positive (Figure 10).MAP and MAT are highly correlated (R = 0.715, one-tailed p = 0.002), with a similar relationship found in both habitats (Figure 11).Additionally, MAP and altitude are negatively correlated (R = − 0.669, two-tailed p = 0.009), but the association is opposite for the two habitats (Figure 11).MAP as a covariate was a significant predictor of variation in VD (p = 0.022), while habitat was not a significant predictor for VD (p = 0.159) (Table A4).This suggests that MAT is important in driving VD, with no additional significant variation in VD explained by habitat (p = 0.159).Note that a MAP with habitat term in the model was not significant (p = 0.113).However, the slope for hillside habitats is very negative, while the slope for valley habitats is positive (Figure 10).MAP and MAT are highly correlated (R = 0.715, one-tailed p = 0.002), with a similar relationship found in both habitats (Figure 11).Additionally, MAP and altitude are negatively correlated (R = − 0.669, two-tailed p = 0.009), but the association is opposite for the two habitats (Figure 11).

Correlation between Vein Trait and Climatic Conditions
Plants are subjected to multiple environment types and their survival occurs under many different environment conditions [28,33,34].Leaf venation networks provide an integrated link between plant form, function, and climate niche because leaf water transport underlies the variations in plant performance [35,36].This has key implications for the distribution and productivity of ecosystems, as well as applications in paleobiology, agriculture, and plant technology.For example, we can propose leaf venation as a new tool to interpret fossil angiosperm life forms.Vein density values of a standing forest are reflected in its leaf litter, suggesting that fossil leaf assemblages are representative of past forest ecosystems [34].Our results imply that based on leaf venation, 14 P. subaequalis populations could be divided into two groups growing in different habitats.These results also show that the differences in venation between the populations were mainly affected by water availability, and this finding was in agreement with previous study [5,37,38].
As leaf morphology evolves and adapts in response to environmental change, there is the potential for functionally interdependent tissues to independently change [35][36][37][38].Our results show that base angle, FEV, and areole shape changed during adaptive variation in leaf venation.The leaf of P. subaequalis showed varied changes in habitats (hillside and valley).In hillside habitats, leaves expanded their base angle, but the FEVs are lower than growing in valley habitats.Furthermore, FEV development is better in valley habitats, and the areole shape is more optimal compared to the leaves of those growing in hillside habitats.Leaves grown under well-watered conditions in the field show a decrease in leaf hydraulic conductance, whereas leaf hydraulic conductance increases under less-watered conditions [39][40][41].
Vein traits are thought to reflect the gas and water exchange characteristics between the atmosphere and the leaves, which are greatly influenced by climatic factors at the leaf, tree, and even regional scales [15].The vein density was closely correlated to the long-time mean temperature (MAT) and water conditions (MAP).A higher vein density means shorter mesophyll distances, which leads to higher leaf hydraulic conductivity [11].A lower evaporative demand should result in less investment in leaf venation because of the lower leaf hydraulic conductivity requirement.Our study shows that the Ningbo, Changhua, and Anji populations had a higher VD than Tongcheng, Jixi, and Jinzhai populations, which grew in valleys and had the lowest VD.Vein density was negatively and linearly related to MAP and MAT (Figures 7 and 8), which agreed with previous reports by other studies [42,43].
However, the ANCOVA results also suggest that VD is responding more to MAP than anything else, including habitat, but especially in presumably drier hillside habitats.Given the strong positive correlation between MAT and MAP (in both habitats), it makes sense that VD is strongly predicted by MAT.Some studies in south China have suggested that vein density has a strong positive correlation with MAP.This is probably due to the higher temperature and precipitation levels in southern China compared to northern or eastern China, thus leading to higher plant transpiration rates [34].Higher leaf vein densities increased transpiration rates.Furthermore, previous studies indicated that there was a strong correlation between vein density and assimilation rates [44].Leaf vein density also has a significantly positive effect on transport and photosynthetic capacities [16,45,46], and stomata density has a very strong linear correlation with vein density [37,45].Some studies have shown that species with a low stomata density (44 stomata mm −2 ) have a low vein density (2.3 mm mm −2 ), while species with a high stomata density (521 stomata mm −2 ) have a high vein density (12.8 mm mm −2 ) [37,45,47].Cell size has been significantly linked to key function traits, vein density, stomata density, and leaf thickness [16,46].A high vein density has also been strongly associated with small stomata, as well as epidermis, palisade, and xylem cell sizes across a number of species [18].Plants can effectively adapt to environmental changes through different adaptive strategies.These strategies coordinate the density of veins and stomata to maintain an efficient photosynthetic rate, which varies depending on the climatic conditions.For example, an increase in VD when moisture availability decreases leads to increased plant photosynthetic rates during pulses of high water availability in arid regions [42,48].
P. subaequalis is an endemic and endangered tree species from China.Currently, the extant populations are protected [46,49].Nevertheless, a more proactive conservation strategy may be needed.A few populations (Ningbo, Jixi) have less than 20 individuals, and only some of these trees are of a reproductive size [50].The results show that different P. subaequalis populations have different adaptive strategies and physiological reactions.Therefore, management measures that reflect these differences need to be introduced so that habitat quality can be improved and the long-term viability of P. subaequalis can be guaranteed.For instance, we can preserve all the extant populations and their habitats or the establishment of ex situ collections in gardens.

Conclusions
Leaf venation patterns and vein density are key anatomic traits closely related to climatic change.P. subaequalis is one of the endangered species of angiosperms distributed across eastern Asia.To understand the change of leaf morphology under different habitats, we measured 15 leaf venation characteristics of P. subaequalis.We have demonstrated that leaf venation is closely related to different habitats (hillside and valley).In hillsides, leaves expand the base angle, with lower FEV trees growing in valley habitats.FEV development is better in water-abundant valley habitats, and the areole shape is more optimal compared to the leaves growing in hillside habitats.
We further used elevation, MAT, and MAP to investigate the relationship with vein density (VD).The results revealed that a strong positive linear relationship was observed between vein density and elevation (R 2 = 0.6270), and had a slightly negative linear relationship with both MAP (R 2 = 0.5188) and MAT (R 2 = 0.093).We also used ANCOVA to analyse the slopes of different habitats, and the results showed that the slope for hillside habitats is substantially more positive than for valley habitats when comparing altitude and VD.Additionally, for MAT and VD, the slope for hillside habitats is substantially more negative than for valley habitats, which is rather flat.For VD and MAP, the slope for hillside habitats is very negative, while the slope for valley habitats is positive.MAP and MAT are MAP and MAT are highly correlated (R = 0.715, one-tailed p = 0.002); however, MAP and altitude are negatively correlated (R = − 0.669, two-tailed p = 0.009).Populations from hillside habitats have a stronger response to elevation, MAT, and MAP than valley populations.
These results imply that P. subaequalis uses different strategies to adapt to different habitats.According to the analysis of leaf morphological traits, different P. subaequalis populations have gradually adapted to the local climatic condition.Moreover, leaf morphological traits might be used as a reference for the conservation of P. subaequalis populations.

Figure 1 .
Figure 1.Distribution map of the sample Parrotia subaequalis (H.T. Chang) R. M. Hao et H. T. Wei populations across eastern China in 2016.The blue circles represent the extant population sites.

Figure 1 .
Figure 1.Distribution map of the sample Parrotia subaequalis (H.T. Chang) R. M. Hao et H. T. Wei populations across eastern China in 2016.The blue circles represent the extant population sites.

Figure 5 .
Figure 5. Unweighted pair-group method with the arithmetic means (UPGMA) analysis using variables of 15 leaf venation characteristics of P. subaequalis.(A) Hillside habitat type; (B)Valley habitat type.

Figure 5 .
Figure 5. Unweighted pair-group method with the arithmetic means (UPGMA) analysis using variables of 15 leaf venation characteristics of P. subaequalis.(A) Hillside habitat type; (B)Valley habitat type.

Figure 6 .
Figure 6.Regression analysis of the VD (mean ± SD) with elevation.The VD is significantly related to the elevation (p = 0.0007).Each dot represents a sampling site in China.

Figure 7 .
Figure 7. Regression analysis between P. subaequalis of vein density (VD) and the climatic factors of mean annual precipitation (MAP) in 14 sampling sites across eastern China.Each dot represents a sampling site in China.(p = 0.001).

Figure 8 .
Figure 8. Regression analysis between P. subaequalis of vein density (VD) and the climatic factors of mean annual temperature (MAT) in 14 sampling sites across eastern China.Each dot represents a sampling site in China.(p = 0.145).

Figure 6 .
Figure 6.Regression analysis of the VD (mean ± SD) with elevation.The VD is significantly related to the elevation (p = 0.0007).Each dot represents a sampling site in China.

Figure 6 .
Figure 6.Regression analysis of the VD (mean ± SD) with elevation.The VD is significantly related to the elevation (p = 0.0007).Each dot represents a sampling site in China.

Figure 7 .
Figure 7. Regression analysis between P. subaequalis of vein density (VD) and the climatic factors of mean annual precipitation (MAP) in 14 sampling sites across eastern China.Each dot represents a sampling site in China.(p = 0.001).

Figure 8 .
Figure 8. Regression analysis between P. subaequalis of vein density (VD) and the climatic factors of mean annual temperature (MAT) in 14 sampling sites across eastern China.Each dot represents a sampling site in China.(p = 0.145).

Figure 7 .
Figure 7. Regression analysis between P. subaequalis of vein density (VD) and the climatic factors of mean annual precipitation (MAP) in 14 sampling sites across eastern China.Each dot represents a sampling site in China.(p = 0.001).

Figure 6 .
Figure 6.Regression analysis of the VD (mean ± SD) with elevation.The VD is significantly related to the elevation (p = 0.0007).Each dot represents a sampling site in China.

Figure 7 .
Figure 7. Regression analysis between P. subaequalis of vein density (VD) and the climatic factors of mean annual precipitation (MAP) in 14 sampling sites across eastern China.Each dot represents a sampling site in China.(p = 0.001).

Figure 8 .
Figure 8. Regression analysis between P. subaequalis of vein density (VD) and the climatic factors of mean annual temperature (MAT) in 14 sampling sites across eastern China.Each dot represents a sampling site in China.(p = 0.145).

Figure 8 .
Figure 8. Regression analysis between P. subaequalis of vein density (VD) and the climatic factors of mean annual temperature (MAT) in 14 sampling sites across eastern China.Each dot represents a sampling site in China.(p = 0.145).

Figure 9 .
Figure 9. Analysis of covariance (ANCOVA) between vein density (VD) and altitude in hillside and valley habitats.Each dot represents a sampling site in China.E represent ten to the power of N.

Figure 9 .
Figure 9. Analysis of covariance (ANCOVA) between vein density (VD) and altitude in hillside and valley habitats.Each dot represents a sampling site in China.E represent ten to the power of N.

ForestsFigure 10 .
Figure 10.(A) ANCOVA between vein density (VD) and MAT in hillside vs. valley habitats; (B) ANCOVA between vein density (VD) and MAP in hillside vs. valley habitats.Each dot represents a sampling site in China.

Figure 11 .
Figure 11.(A) ANCOVA between MAP and MAT in different habitats (hillside and valley) (p = 0.002); (B) ANCOVA between altitude and MAP in different habitats (hillside and valley) (p = 0.009).Each dot represents a sampling site in China.E represent ten to the power of N.

Figure 10 .
Figure 10.(A) ANCOVA between vein density (VD) and MAT in hillside vs. valley habitats; (B) ANCOVA between vein density (VD) and MAP in hillside vs. valley habitats.Each dot represents a sampling site in China.

Forests 2018, 9 ,Figure 10 .
Figure 10.(A) ANCOVA between vein density (VD) and MAT in hillside vs. valley habitats; (B) ANCOVA between vein density (VD) and MAP in hillside vs. valley habitats.Each dot represents a sampling site in China.

Figure 11 .
Figure 11.(A) ANCOVA between MAP and MAT in different habitats (hillside and valley) (p = 0.002); (B) ANCOVA between altitude and MAP in different habitats (hillside and valley) (p = 0.009).Each dot represents a sampling site in China.E represent ten to the power of N.

Figure 11 .
Figure 11.(A) ANCOVA between MAP and MAT in different habitats (hillside and valley) (p = 0.002); (B) ANCOVA between altitude and MAP in different habitats (hillside and valley) (p = 0.009).Each dot represents a sampling site in China.E represent ten to the power of N.

Table 1 .
Details of sample locations and sizes of 14 populations of Parrotia subaequalis (H.T. Chang) R. M. Hao et H. T. Wei from eastern China.

Table 4 .
Leaf vein density (mean ± Standard Deviation(SD)) of P. subaequalis by site examined in this study.

Table 4 .
Leaf vein density (mean ± Standard Deviation(SD)) of P. subaequalis by site examined in this study.

Table A2 .
The result of ANCOVA between vein density (VD) and altitude in hillside and valley habitats.R 2 = 0.837(Adjusted R 2 = 0.808). a

Table A3 .
The result of ANCOVA between vein density (VD) and MAT in hillside and valley habitats.

Table A4 .
The result of ANCOVA between vein density (VD) and MAP in hillside and valley habitats.

Tests of Between-Subjects Effects Dependent Variable: VD Source Type III Sum of Squares df Mean Square F Sig. Partial Eta Squaed
R 2 = 0.602 (Adjusted R 2 = 0.529). a