Teasing out elevational trends in infraspecific Prunus taxa: A vein analysis approach

Using 33 specimens collected from across their range in Turkey, we demonstrate that the subspecies of Prunus microcarpa C.A.Mey react very differently to altitude. We first outline a simplified, flexible protocol for sectioning and removing the epidermis of small, difficult‐to‐image leaves for leaf vein studies. We then used venation analysis software to evaluate the two subspecies of this wild cherry in relation to altitude. We also found key differences in venation features between short‐shoot and long‐shoot leaves for each taxon. Differences include statistically significant negative correlation between vein density, and positive correlation between areole area and altitude in long‐shoot leaves of Prunus microcarpa subsp. microcarpa, while its short‐shoot leaves had a positive relationship between maximum areole area, and negative relationship between vein density, numbers of veins and endpoints. Meanwhile, P. microcarpa subsp. tortuosa (Boiss. & Hausskn.) Browicz recorded trends that were largely opposite of this, but beside vein thickness and areole area, were not statistically significant. This difference may be part of each taxon's overarching syndrome of anatomical and morphological adaptations to its external environment.


| INTRODUCTION
Leaf venation analysis is a staple in fields such as physiology and ecology and is growing in importance in the realm of systematics, where it has been used in many plant groups, including Rosaceae (Kolivand et al., 2019;Rieger et al., 2003;Ufimov & Dickinson, 2020).The development of advanced computational methodology and software for network analysis has created a more robust mathematical foundation for venation studies, which previously focused on qualitative features (Bühler et al., 2015;Dirnberger et al., 2015;Pagano et al., 2016;Price et al., 2011).Examples include the venation-based separation of soybean (Glycine max (L) Merr) and two Phaseolus vulgaris L. cultivars through machine learning algorithms, demonstrating the differences in venation of closely related species (Larese et al., 2014).Changes in vein features are also known to help plants adapt to different climatic conditions (Okie & Rieger, 2002).
Prunus microcarpa C.A.Mey is a small-leaved, drought-adapted member of the Rose family which may take the form of a shrub or small tree.In Turkey, which lies at the western limit of its range, it is represented by two subspecies, P. microcarpa subsp.microcarpa and P. microcarpa subsp.tortuosa (Boiss.& Hausskn.)Browicz (POWO, 2022).Its leaves, which are often folded along the midrib and undulating or curled, are covered in a thick cuticle and dense trichome cover.Our previous work on Prunus L. revealed increasing difficulty in separating these and other infraspecific taxa of this genus, largely due to variability and overlap in features previously considered diagnostic (Bostanci Ordu et al., 2021;Çiftçi et al., 2022).The characteristics usually used to separate P. microcarpa taxa are leaf shape and indumentum (Boissier, 1867;Browicz, 1972), but these are insignificant when a large number of populations are taken into account (Bostanci Ordu et al., 2021).These features show a certain amount of plasticity based on ecological conditions, which gives even nearly indistinguishable infraspecific taxa the potential to show diverging adaptations to similar abiotic factors.
The current study aimed to assess whether these taxa vary significantly in venation features, which, alongside leaf shape and indumentum, may have adaptive significance (Brodribb et al., 2010).This builds on our previous work on P. microcarpa, in which we found that the differences in leaf morphology between long-shoot and short-shoot leaves were greater than those between leaves of these two infraspecific taxa (Mollman et al., 2022).Heteroblastic development affecting shoot and leaf anatomy and morphology is well documented in many Rosaceae taxa (Costes et al., 2014;Robertson et al., 1992;Rouffa & Gunckel, 1951), but studies on many aspects of this phenomenon are lacking, especially for Prunus.
As such, in order to compare similar groups of leaves, we assessed these two leaf types separately as we explored the differences in vein features of these two taxa and how they change in response to elevation throughout their range in Turkey.The folded, occluded leaves of P. microcarpa taxa present difficulties that must be overcome in undertaking such a study, as does the great variability in size, especially between the two different leaf types.We present strategies for dealing with these types of challenges as we explore the dynamic relationships between these taxa and leaf types.

| Sampling and preparation
A total of 121 leaf samples from 33 P. microcarpa specimens (17 of P. microcarpa subsp.microcarpa, 16 of P. microcarpa subsp.tortuosa) were used in this study (Table 1).Specimens were collected in summer June-July 2021, pressed and deposited as herbarium material at Istanbul University Faculty of Science Herbarium (ISTF).We also obtained leaf samples from herbarium specimens of ISTF and Gazi University Faculty of Science Herbarium (GAZI), with permission of the curators.
Specimens were identified according to Browicz (1972) and Boissier (1867) and sample leaves were sorted into four groups according to subspecies and leaf type: P. microcarpa long shoot (LS) and short shoot (SS) leaves, and P. microcarpa subsp.tortuosa LS and SS leaves.
Leaves were softened by boiling in tap water for 30 min and drying them flat in a standard plant press overnight.
To obtain a standardized section, each pressed leaf was placed abaxial surface up and divided into three equal segments lengthwise with the aid of millimetric graph paper.The middle segment to the right of the midrib was selected for analysis and the margin was removed (Figure 1).This secures a section of leaf with more standard leaf and areole features, as these tend to skew toward the margins and tips, and remain more stable in the center of the blade (Nardini et al., 2008;Zwieniecki et al., 2004).
The standardized samples were then soaked in 0.6% sodium hypochlorite solution until bleached and bubbles appeared under the epidermis.The epidermis and its dense layer of indumentum was then removed to provide an unimpeded view of the veins.The bleached and peeled samples were photographed before and after staining with an aqueous solution of 0.01% Safranin O. Pictures were taken under an Olympus SZX7 Stereomicroscope equipped with a Canon camera and Argenit Kameram Version 3.1.0.0 software.The photographs were saved in tagged image format (.tif) with a 1 millimeter scale bar.Photographs with a clear view of the veins were preprocessed using Adobe Photoshop 2017.0.0 to clean up the edges of the sample, erase the background and adjust contrast.After testing both bleached and stained images, we found that bleached, unstained images with the epidermis removed gave the clearest results in phenoVein (Bühler et al., 2015).

| Venation analysis using phenoVein
Preprocessed images were loaded into phenoVein (workflow outlined in Figure 2) and scale bars were set to 1 mm.Color channel was selected for each sample to give the best contrast between veins and areoles (Figure 2a).Lower threshold values were set to between 50,000 and 65,000 and upper values between 55,000 and 120,000, according to the image (Figure 2b).
The following settings were used for vesselness filtering: lower sigma 4-10 (depending on the image), upper sigma 25-35 (depending on the image), number of steps 6, gap closing kernel 9, and boundary erosion 5 (Figure 2c).Default values were used during skeletonization using threshold values of 4000-10,000, depending on the image T A B L E 1 Material used in this study.

| Statistical analysis
Total sample area, skeleton length, average vein density, number of skeleton pieces, endpoints, branching points and areoles, as well as average, maximum and minimum values for areole area, vein length and vein width were included in statistical analyses.
Nested t test was performed with GraphPad Prism 9.0 software to detect significant values using subspecies as the primary group and leaf type as subgroups.Data categories found to be significant (p < .05)were subjected to principal component analysis (PCA) to compare the leaf groups.Multivariate linear regression (MLR) analysis was used to assess the relationship between venation characteristics (dependent variables) and altitude (independent) data.PCA and MLR were performed in PAST version 4.08 (Hammer et al., 2001).

| RESULTS
Leaves showed important and complex differences in venation features based on altitude, taxon and leaf type during our statistical tures at the high and low edge of their elevational distributions for both P. microcarpa (Figure 3) and P. microcarpa subsp.tortuosa (Figure 4).
Nested t test results found some important differences between venation characteristics of SS and LS leaves in these Prunus microcarpa taxa, but no significant differences between the two infraspecific taxa.Results are provided in Table 2.
SS leaves had the shortest total skeleton length and LS leaves had the longest total length, regardless of taxon.The number of skeleton pieces is analogous to vein number and is greatest in the larger LS leaves.End and branching point counts also grow in tandem with the total number of vein pieces, meaning that LS leaf samples had the highest values.However, at the lower end of the range, both LS and SS samples had equally low minimum values for these characteristics.LS leaves also had the greatest number of areoles.The values of these characteristics tend to increase with leaf size and therefore, the larger and more numerous areoles of LS leaves can be seen as a natural consequence of their larger average areas.All data are provided in Supplementary material.
Although vein density, minimum areole area and maximum vein length are statistically significant among leaf groups, they are less significant (larger p value) than other parameters.This indicates a general trend in which dense venation is seen more often in SS leaves than in LS leaves.The smallest areoles and shortest veins tend to be found in SS leaves.
According to PCA using the significant nested t-test data, LS leaves show greater variation than SS leaves.No significant difference was detected between subspecies across populations when not evaluated for altitude (Figure 5).
Potential correlation between altitude and venation characteristics was tested using multiple linear regression (MLR) analysis, the results of which indicate that the two subspecies react differently to altitude in terms of vein characteristics.P. microcarpa subsp.microcarpa shows an overall more pronounced reaction to changes in altitude than does P. microcarpa subsp.tortuosa (see  P. microcarpa subsp.tortuosa combined data found negative relationships between average areole area and thickest vein data and rising altitude, and both barely fell under the p < .05threshold for significance.The average areole area of LS leaves was also barely significantly correlated with altitude, and no parameters changed significantly for SS leaves of this taxon.Running an MLR analysis without division between taxon or leaf type resulted in no significant relationships between venation characteristics and altitude.
In many cases, leaf venation parameters show opposing relationships with altitude between the two infraspecific taxa.Plots showing these trends are provided for parameters that show a significant relationship with altitude for at least one test group in Figure 6.
Both taxa show a negative correlation between altitude and thickest vein diameter, but the correlation is significant only for P. microcarpa subsp.tortuosa, and is notably the only characteristic which is significant for this taxon and not the other.Average areole area is the only category in which both subspecies are significantly affected by a rise in altitude, and in this case P. microcarpa subsp.tortuosa is negatively correlated, while P. microcarpa's average areole area grows with increasing altitude.In other parameters found significant in at least one group, only P. microcarpa shows a strong reaction to change in altitude.

| DISCUSSION
Because much of the discussion on leaf venation analysis has centered around computational methods of network extraction, most quantitative studies on leaf vein parameters have focused on leaves that are more straightforward in terms of vein visibility and leaf size, such as those of Diego and De Bacco (2021), Price et al. (2011), and Zheng and Wang (2010).Often samples used are either whole leaves or one-size-fits-all sections for all leaves, often with random sampling from various parts of the leaves (Larese et al., 2014;Zhu et al., 2020).
The latter is problematic for comparative studies because venation features change across the surface of a leaf (Nardini et al., 2008;Zwieniecki et al., 2004).Leaves in previous work have been imaged with a scanner (Larese et al., 2014) camera or microscope (Newsome et al., 2020).In the case of P. microcarpa leaves, extreme variability in terms of leaf size meant that a standard section taken from the  F I G U R E 5 PCA using venation data found significant by Nested t test.No significant differences were found between Prunus microcarpa subspecies, but some venation characteristics were significantly different between SS and LS leaves.Percentage of variance: 98.76 for PC1 and 1.09 for PC2.
smallest SS leaves does not cover enough area to be representative of the larger LS leaves, especially given the differences in vein width and length typical for veins of different orders (Roth-Nebelsick et al., 2001).The small size of our sections and their veins necessitated imaging under a stereomicroscope, which has been found to give superior results in some studies (Newsome et al., 2020).A robust foundation to compare many types of leaves even when not exploring dimorphism may also help to prevent biased sampling based on ease of analysis, rather than representing the full range of leaves in a population.
The indumentum of many of our leaves was so dense that veins were completely occluded, necessitating the removal of the epidermis.
The indumentum and thick cuticle, coupled with a very thin and delicate mesophyll, rendered the chemical regime of some previously published protocols, such as that of Dizeo de Strittmatter (1973), too damaging for our samples.We found that boiling leaves prior to clearing them both helped soften the tissues, and also helped to break down the cuticle so that a simple weak bleach solution both cleared the veins and lifted the epidermis enough that it could be removed completely under a dissecting microscope.Taking sections prior to labwork also eased both the chemical load and processing time of each sample.Further tests found that staining with Safranin O was not more effective in achieving the necessary contrast for vein analysis than simply clearing the sections, and in some cases it further damaged tissues already weakened by exposure to bleach solution.We suggest these strategies be used with dried material from difficult plants, such as other drought-adapted species with soft mesophyll, dense indumentum or thick cuticles.
Our previous work on the anatomical (Bostanci Ordu et al., 2021), morphological and molecular (Çiftçi et al., 2022)   tortuosa and we have suggested their treatment as synonyms.The current study shows differences between the two in reaction of their venation features to differences in altitude.While P. microcarpa showed a tendency to alter vein density, areole size and vein length, as well as number of veins and vein endpoints, the subspecies tortuosa showed only weak correlations between areole area and vein thickness and altitude.Overall, P. microcarpa's vein features demonstrated a much greater tendency to react to changes in altitude than did P. microcarpa subsp.tortuosa.Wang et al. (2020) found that while there were general trends among trees and shrubs, plants employed wide-ranging tactics beyond venation to adapt to altitude changes, and phylogeny played a major role in these tactics.Our results suggest that even close infraspecific taxa that show little to no difference in other ways, may employ different venation-based changes in response to altitude.
Venation properties may change significantly according to myriad climatic factors, including water availability, heat and even windspeed, as well as orientation in the canopy and taxonomic placement of the plant (Uhl & Mosbrugger, 1999).These interactions are complex, with variable adaptation strategies adopted by various plant groups (Schneider et al., 2018).In a broad review of venation literature, Sack and Scoffoni (2013), compiled the published data on vein density, leaf area and free-ending vein numbers, as well as biome, climate factors such as precipitation and temperature, and individual leaf data for nearly 800 species.In that review, trees tended to have a higher vein density than shrubs, which had denser veins than herbaceous plants, while tolerance of intense sun and aridity was another predictor of high vein density.It is posited that one adaptation to periodic drought is high water conductance to support photosynthesis when water is available, and high vein density supports this need (Maximov, 1931).
Lower altitudes tend to mean longer, hotter summers than similar habitats at higher altitudes.It follows that many plants have significantly thicker veins or greater vein density and number of veins as we see in P. microcarpa to mitigate the resulting water stress of lower altitudes.Areole size and vein endpoint numbers naturally fall with decreasing vein number and density in P. microcarpa, a trend seen in other Prunus s.l.species investigated by Kawai and Okada (2020).
However, P. microcarpa subsp.tortuosa seems to demonstrate less venation-based adaptation and may thus employ different adaptive strategies.For instance, as indicated by its subspecies epithet, P. microcarpa subsp.tortuosa has a notably tortuous branching pattern (Boissier, 1867).Branching habit and its effects on architecture and resource allocation have been shown to be adaptive in both trees and shrubs (Götmark et al., 2016;Pickett & Kempf, 1980;Sun et al., 2010).Another notable feature is the overall denser indumentum of this taxon, which is a known adaptation to dry conditions.The basis for these different strategies may be purely genetic or informed by other climatic and abiotic factors.In our field work, we observed that P. microcarpa was sometimes found on roadside forest edges, but P. microcarpa subsp.tortuosa was found only in the open.The competitive presence of taller vegetation or intensity of sunlight may also affects these plants' strategies.However, our reliance on herbarium material in which this information was not recorded precluded the use of these data in our analyses.
Leaf characteristics vary along with abiotic conditions such as temperature, light, humidity and available nutrients at the population or individual level (Ashby, 1948;McDonald et al., 2003;Nicotra et al., 2011).In a morphometric study using wild P. avium leaves from Bosnia and Herzegovina, Miljkovi c et al. ( 2019) found significant differences between the three populations growing at different altitudes.
Other studies have observed smaller leaves in individuals living at high altitudes across various plant groups (Liu et al., 2020).On the contrary, the current study found that sample area increased with altitude in both subspecies of P. microcarpa.This corroborates our previous study on leaf morphology of these two taxa in which we found altitude to have a minimal effect on leaf shape, but affected leaf size positively (Mollman et al., 2022).
We also found that plants with large leaves living at higher altitudes generally have lower vein density.Similar trends have been observed in previous studies but these features are known to vary alongside other factors, such as phylogeny or plant form.For instance, in a study of 93 different woody plant species, Wang et al. (2020) found that, for trees, vein density showed a decreasing trend and vein width increased at higher altitudes, while shrubs tended toward denser and finer venation at higher elevations.Our results show that, like the trees of that study, P. microcarpa subsp.microcarpa's vein density drops with increasing elevation, but the veins also become thinner, a characteristic most often seen in the shrubs of Wang et al. (2020).This is not particularly surprising, as the aforementioned study notes that phylogeny overrides climate as a factor because different plant groups employ widely differing strategies to deal with the climatic stressors associated with altitude (Wang et al., 2020).Strategies beyond leaf size and venation features include cuticle conductance, water storage and dehydration tolerance as well as leaf drop, which confer tolerance to drought (Scoffoni et al., 2011).
Principal component analysis results between subspecies and leaf type groups show that LS leaves exhibit wider variation than SS leaves.This may be due to the larger leaves of this test group, which provide more flexibility, but it may also be that LS leaves are necessarily more sensitive to short-term climatic changes because they develop in the same season, as opposed to short-shoot leaves, which develop in the previous year and spend the winter as leaf primordia in buds (Costes et al., 2014;Robertson et al., 1992).
Due to the nature of our study, in which it was necessary to take sections from the center of the leaf, from one side of the midrib, the data presented is necessarily heavily skewed toward minor veins, and major vein data is removed from the data during trimming.However, because minor vein density was directly correlated with total vein density for all Prunus s.l.species and was a general trend throughout the Rosaceae in the data compiled by Kawai and Okada (2020), we do not believe this poses a major weakness in the current study.
Free-ending veins (Number of endpoints) may boost leaf hydraulic conductivity and high numbers are associated with moist habitats (Scoffoni et al., 2011).Free-ending vein numbers were relatively low in both P. microcarpa taxa, which may mitigate risks associated with unstable water flow or damage to veins (Kawai & Okada, 2020).
P. microcarpa showed a statistically significant tendency to lose freeending veins at higher altitudes, while P. microcarpa subsp.tortuosa showed a slight increase.Because the range of both taxa overlaps geographically and elevationally, differences in their free-ending vein numbers may indicate significant differences in overarching adaptive syndromes.
Leaf features may show similar adaptative characteristics for a number of different abiotic factors, even within a single individual (Sack & Scoffoni, 2013).Higher leaf hydraulic conductivity, which is associated with greater vein density, thickness and free-ending veins, is frequently seen in leaves exposed to higher levels of heat, radiation, nutrients and water stress, and helps them adapt quickly to varying levels of water availability through dormancy and stomatal closure (Maximov, 1931;Sack & Scoffoni, 2013).As this study has shown, shoot-dependent leaf dimorphism also has the potential to affect venation features within individuals.And even among closely related taxa, these features may show surprisingly different trends in their reactions to abiotic factors.

(
Figure 2d).After manual correction to the vein skeleton, minimum ending vein size was set to between 2 and 25 pixels to clean up the image (Figure 2e,f).Following venation analysis, images were saved and data output as .csvfile (Figure 2g,h).Data output in .csvformat was compiled, trimming the upper and lower 5% of areole area, vein length and vein width data to remove extreme values resulting from poorly detected veins.The mean, standard deviation, maximum and minimum values of these data were calculated for each specimen.Compilation, trimming, and calculations were performed in Python.

F
I G U R E 2 Venation analysis was performed for four groups of leaves in phenoVein: (a) Load and scale image: color channel is selected to give good contrast with light veins on a darker background and scale bar is set (1 mm) (b) Leaf segmentation: Masking to indicate leaf sample (cyan) and background (red), using lower and upper threshold values (c) Vesselness filtering: The program estimates venation as a 3D network using parameters defined by the user, (d) Skeletonization: Vein structure is extracted according to the 3D network drawn in the previous step based on default parameters.The user specifies the lower limit accepted as a vessel, (e) Manual correction: The user can make corrections to the generated skeleton, adding undetected veins and deleting any falsely detected veins.(f) Re-skeletonization: Skeletonization accounting for user corrections provides an opportunity to clean up false ends and final check of the skeleton.(g) Vein analysis: The program calculates the width, length, and other data of individual vessels, (h) Write results: Images are saved after setting color scale for areole size and data is output.

F
I G U R E 3 Leaf analyses showing stereomicroscope-generated photograph, skeletonized vein network and areole size analysis for representative SS (a, b) and LS leaves (c, d) of P. microcarpa populations from altitudes at the low (a-c, 900 m) and high (b-d, 1911 m) edge of the sampling range for that taxon.ISTF41454 SS (a) and LS (c) leaf; ISTF41329 SS (b) and LS (d) leaf.Scale bars = 1 mm.Photos taken with stereomicroscope and analysis images generated using phenoVein software(Bühler et al., 2015).

F
I G U R E 4 Leaf analyses showing stereomicroscope-generated photograph, skeletonized vein network and areole size analysis for representative (a, b) SS and (c, d) LS leaves of P. microcarpa subsp.tortuosa populations from altitudes at the low (a-c, 437 m) and high (b, 1699 m; d, 1388 m) edge of the sampling range for that taxon.ISTF41373 SS (a) and LS (c) leaf; ISTF41335 SS (b) and ISTF41266 LS (d) leaf.Scale bars = 1 mm.

F
I G U R E 6 Plot graphs showing correlations between altitude and Prunus microcarpa venation data found to be significant ( p < .05) in at least one test group: (a) Average vein density, (b) Number of skeleton pieces, (c) Number of skeleton endpoints, (d) Average areole area, (e) Largest areole area, (f) Longest vein, (g) Thickest vein.Circular data points: P. microcarpa subsp.microcarpa.triangle points: P. microcarpa subsp.tortuosa.Filled points indicate SS leaves.unfilled points are LS leaves.Lines show correlation trend for P. microcarpa subsp.microcarpa (solid lines) and P. microcarpa subsp.tortuosa (dashed lines) leaves combined.

Table 3
area, but a negative relationship between mean vein density, numbers of veins and endpoints.Combined data for both LS and SS leaves for this taxon found a strong positive correlation between longest vein and areole area and a negative relationship for vein density and rising altitude.
T A B L E 2 Nested t-test results for phenoVein-generated venation data of SS and LS leaves of Prunus microcarpa subsp.microcarpa and P. microcarpa subsp.tortuosa.Taxon was used for the primary group and leaf type for the subgroups.
MLR results of venation data and altitude.p values are given for each characteristic for SS and LS leaves of each taxon as well as combined results for both types of leaf.
features of Prunus subgenus Cerasus taxa in Turkey suggest that there is no quantifiable difference between Prunus microcarpa and P. microcarpa subsp.T A B L E 3Note: Signs below each value indicate positive (+) or negative (À) correlation with rising altitude.*p values <.05 are considered significant.