Desiccation of the leaf mesophyll and its implications for CO2 diffusion and light processing

Abstract Leaves balance CO2 and radiative absorption while maintaining water transport to maximise photosynthesis. Related species with contrasting leaf anatomy can provide insights into inherent and stress‐induced links between structure and function for commonly measured leaf traits for important crops. We used two walnut species with contrasting mesophyll anatomy to evaluate these integrated exchange processes under non‐stressed and drought conditions using a combination of light microscopy, X‐ray microCT, gas exchange, hydraulic conductance, and chlorophyll distribution profiles through leaves. Juglans regia had thicker palisade mesophyll, higher fluorescence in the palisade, and greater low‐mesophyll porosity that were associated with greater gas‐phase diffusion (g IAS), stomatal and mesophyll (g m) conductances and carboxylation capacity. More and highly‐packed mesophyll cells and bundle sheath extensions (BSEs) in Juglans microcarpa led to higher fluorescence in the spongy and in proximity to the BSEs. Both species exhibited drought‐induced reductions in mesophyll cell volume, yet the associated increases in porosity and g IAS were obscured by declines in biochemical activity that decreased g m. Inherent differences in leaf anatomy between the species were linked to differences in gas exchange, light absorption and photosynthetic capacity, and drought‐induced changes in leaf structure impacted performance via imposing species‐specific limitations to light absorption, gas exchange and hydraulics.


| INTRODUCTION
Photosynthesis supports plant growth, development, and reproduction, and to optimise this process, leaves must balance light absorption, carbon capture, and water loss under ever changing conditions. Anatomical and physiological leaf traits play key roles in determining the exchange of light, CO 2 , and water with the environment. How the structural and physiological components of the leaf are affected by drought is of increasing importance given the increasing frequency and duration of drought globally (Brodribb et al., 2020;Choat et al., 2018). Stomata regulate the diffusion of gases across the leaf surface, where water vapour is lost in exchange for CO 2 after crossing the leaf boundary layer, and respond strongly to changes in vapour pressure and soil moisture in many species to minimise water loss (Cowan & Troughton, 1971;Farquhar & Sharkey, 1982;Mott & Peak, 2013;Oren et al., 1999;Turner et al., 1984). After reaching the substomatal cavity, CO 2 molecules are subject to a series of gas and liquid phase resistances along the diffusion pathway through the intercellular airspace, cell walls, membranes, cytosol, and other cellular components to reach carboxylation sites inside chloroplasts. The inverse of the sum of these resistances is used to calculate mesophyll conductance (g m , see Table 1 for symbol definitions), (Flexas et al., 2008;Flexas et al., 2018;Tosens & Laanisto, 2018), and these resistances to the movement of CO 2 should be sensitive to changes in leaf water status. Theoretical predictions and experimental observations have found that both the physical properties of the mesophyll (e.g., cell wall thickness, palisade and spongy mesophyll cell density, mesophyll surface area exposed to the intercellular airspace -IAS) and the underlying physiology (i.e., chloroplast positioning, aquaporins, and carbonic anhydrase activity) strongly influence CO 2 diffusion within a leaf and its concentration at the sites of carboxylation (Flexas et al., 2012;Momayyezi & Guy, 2017a, 2017bMuir et al., 2014;Tholen & Zhu, 2011). The products of photosynthesis are then either consumed locally or exported to the vascular tissue.
Similarly, but in an opposing flow direction, water exits the vascular tissue and travels through the mesophyll, ultimately evaporating into the IAS and lost to the atmosphere via the stomata or across the epidermis. A primary role of the leaf vasculature is therefore to replace the water lost while the stomata are open to sustain the uptake of CO 2 for photosynthesis. During drought, insufficient soil moisture or declines in the hydraulic conductance of the vascular system fail to meet the evaporative demands of leaves, leading to loss of turgor in the mesophyll. Recent work has shown that turgor loss directly influences cell shape and leaf porosity even while the xylem remains functional , and the resulting structural and physiological changes associated with leaf desiccation can significantly alter leaf hydraulic conductance (Buckley et al., 2017;Scoffoni et al., 2014), highlighting the complex sequence of events that take place inside the leaf during drought.
An additional layer of complexity can be observed in the overall structure of the leaf mesophyll and the embedded vasculature, which should not only be organised to facilitate the movement of both carbon and water, but also optimised for the opposing gradients of light and CO 2 within the leaf Evans, 1999;Evans, 2021;Smith et al., 1997;Xiao et al., 2016). A general assumption is that the absorptive, optical, and hydraulic properties of leaves are optimised under well hydrated conditions with the mesophyll cells under full turgor. What then are the effects of turgor loss and the associated changes in cell shape and volume on the processing of light, CO 2 and water as leaves dehydrate? As mesophyll cells lose turgor, there should be consequences for the different physiological roles that those cells contribute to, and perhaps differently in the palisade and spongy mesophyll based on cell size and shape. For example, loss of turgor in the mesophyll should lead to changes in the physical shape of the cells (Canny et al., 2012), which has implications for the surface area exposed to the IAS, the tortuosity of the diffusion pathway for both H 2 O and CO 2 (i.e., decline in g m , Cano et al., 2014) by bringing the epidermis closer to the sites of evaporation within the leaf (Buckley et al., 2017). However, it should also directly affect the optical properties of those cells for light propagation and scattering, leading to sub-optimal light absorption with negative impacts on biochemical activity and light use during photosynthesis. The distribution of mesophyll cells and presence of bundle sheath extensions (BSEs) can influence light distribution with depth into a leaf (Evans & Vogelmann, 2003;Holloway-Phillips, 2019;Smith et al., 1997). Numerous studies have shown significant relationships between optical properties (e.g., absorptance and reflectance) and leaf chlorophyll concentration under water stress (e.g., Carter, 1993;Carter & Knapp, 2001;Gitelson et al., 2003), however, the functional relationship between mesophyll and light absorption with depth into a leaf and under dehydration is not known. Recent studies evaluating mesophyll anatomy at finer scales have shown links between biophysical properties of mesophyll cells and IAS conductance (g IAS ). These linkages are associated with variation in airspace tortuosity (i.e., the ratio of the diffusive path length to the straight path length; τ leaf ), porosity (i.e., IAS volume fraction of the mesophyll; θ IAS ), and path lengthening as a consequence of CO 2 diffusion through each distinct stomate to IAS (λ leaf ) (Earles et al., 2019;Gommes et al., 2009;Harwood et al., 2021;Tosens et al., 2016). Little is known about how water stress influences these relationships, as suggested in CO 2 and water flux models considering mesophyll and vascular geometry (Rockwell et al., 2014c(Rockwell et al., , 2017. Declines in the net assimilation rate (A n ) under water stress are well documented, and arise due to both stomatal and non-stomatal limitations. Loss of turgor in the guard cell complex creates a physical barrier for the diffusion of CO 2 into the leaf, and leads to a depletion of the internal CO 2 supply to carboxylation sites, but also negatively influences photochemistry due to increases in leaf temperature (Brodribb & Holbrook, 2003;Buckley, 2019;Buckley et al., 2017;Galle et al., 2009). Excessive leaf temperatures and desiccation can also lead to permanent damage to photosynthetic machinery (Cano et al., 2013;Chaves et al., 2009;Galmés et al., 2007;Hsiao, 1973;Nadal & Flexas, 2018;Trueba et al., 2019;Urban et al., 2017). A negative response of g m to dehydration occurs under mild water stress, and this response is exacerbated by high light intensity (Flexas et al., 2008;Galle et al., 2009;Zhou et al., 2007), illustrating, the need to understand the coordination of multiple exchange processes since excess light can be detrimental to the photosynthetic machinery when rates of carbon fixation decrease with water stress.
The goal of this study is to unfold the complex links between the leaf anatomical traits and functional diversity in CO 2 , water and light absorption. Here, we explore inherent differences in leaf structure for two walnut species with leaf anatomy contrasting in the fraction of  (Pivovaroff et al., 2014. Therefore, we expected J. regia leaves with less structural and functional support by BSEs (mainly known as parenchyma cells connecting veins to epidermis) to exhibit more volumetric changes through mesophyll cells, porosity, and g IAS under dehydration. In contrast, J. microcarpa with higher cell packing and BSEs was expected to more reflect small changes in cell geometry through light absorption profile, as suggested in species with dense spongy mesophyll through more lower-mesophyll scattering impact (Ren et al., 2019;Smith et al., 2004). To evaluate these hypotheses, we used X-ray micro-computed tomography (microCT) imaging to observe in-depth variation in leaf and cell morphology with dehydration coupled with gas exchange measurements.

| Plant materials and growth conditions
Juglans regia cv. Chandler is the most common hybrid scion from natural populations of J. regia L., and J. microcarpa, is used in J. microcarpa × J. regia crosses to produce rootstocks with resistance to crown gall and root rot diseases (Browne et al., 2015;Hasey, 2016;McGranahan & Leslie, 2009). Juglans microcarpa is reported to be more tolerant to water deficit (Knipfer et al., 2020).
Two-year-old saplings of clonal and non-grafted J. regia and J.

| Photosynthesis measurements
Net assimilation rate (A n ), stomatal conductance (g s ) and the intercellular airspace CO 2 concentration (C i ) were measured on the 4th or 5th leaflet of the most recent fully expanded leaf using LI-COR 6400 XT and LI-COR 6800 systems fitted with 6400-40 and 6800-01A fluorometers, respectively (see Supporting Information Method for A-C i and A-I curves). All measurements were done under PPFD = 1500 (10% blue vs. 90% red) (µmol m −2 s −1 ), chamber temperature at 25°C, ambient chamber CO 2 concentration (C a ) at 400 (µmol mol −1 ), flow rate at 150 (µmol air s −1 ), and vapour pressure deficit between 1.5 and 2.0 kPa. All leaflets were dark adapted for 20 min before all other measurements to obtain the maximum quantum yield of photosystem II. The quantum yield of photosystem II (Φ PSII ) under actinic light was obtained by application of saturating multiphase flashes (>8000 µmol m −2 s −1 ) as per Genty et al. (1989).

| Stable carbon isotope discrimination method
Pre-evacuated 10 ml gas tight vials (Exetainer, Labco, UK) were used to collect air exiting the LI-COR chamber through a tube connected to the cuvette exhaust, either with (CO 2 P = plant CO 2 ) or without (CO 2 R = reference CO 2 ) leaf material inside the chamber. The air exiting the LI-COR cuvette was collected as described by  and analysed for stable carbon isotope composition. A three-way valve was added to the LI-COR 6800 chamber through the exhaust tube. A~2 m sampling tube was connected to the third port, and the valve was opened towards it. After 5 min, the valve was returned to its primary position along the chamber exhaust tube, and 15 ml air was collected from the tube into a gas-tight glass syringe through a brass luer-lock fitting. A needle was connected to the syringe, the syringe's valve was opened, and 3 ml of air sample was flushed through the needle before purging 12 ml of the air into a vial. Sampling started with CO 2 R samples, followed by CO 2 P and then alternating CO 2 R with one CO 2 P sample.
After taking the first CO 2 R sample, a leaf was placed inside the chamber and light adapted for 20 min before taking the first CO 2 P sample. The same protocol was followed for every plant sample,

| Calculation of g m from carbon isotope discrimination
Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) discriminates against 13 CO 2 relative to 12 CO 2 during carboxylation (Guy et al., 1993). The amount of discrimination expressed in vivo depends on the diffusion gradient for CO 2 from the bulk atmosphere. By comparing the observed discrimination (∆ o ) with the predicted discrimination (Δ i ) based only on the diffusion gradient through the stomata (i.e., C a to C i ), the gradient associated with the remaining portion of the diffusion pathway (i.e., C i to C c ) can be estimated and used to calculate g m (Evans et al., 1986). Smaller contributions to total discrimination, associated with respiratory (Δ e ) and photorespiratory carbon flux (Δ f ), must also be accounted for. The effect of g m on overall isotope discrimination (Δ gm ) is then given by: Observed discrimination was calculated according to Evans et al. (1986): Desiccation of the leaf mesophyll and its implications (2) where, δ C δ C and 13 e 13 a are the isotopic ratios of reference CO 2 and unconsumed CO 2 , respectively. ζ is the ratio of the reference CO 2 concentration (C e ) entering the cuvette, as determined by the LI-COR 6800, and the net amount consumed in photosynthesis (i.e., C e -C a ).
Predicted discrimination was calculated from gas exchange data with corrections for ternary effects as per Farquhar and Cernusak (2012): where b is the fractionation in carboxylation of ribulose bisphosphate catalysed by Rubisco (−29‰; Guy et al., 1993). The ternary correction factor, t, is: where E is the transpiration rate and g t ac is the combination of boundary layer and stomatal conductance to CO 2 . The combined factor for diffusional fractionation through stomata and the boundary layer, a′, is: where a and a b are the fractionations occurring during diffusion across the stomata (4.4‰) and through the boundary layer (2.9‰), respectively, and C s is the CO 2 concentration at the leaf surface (Evans et al., 1986).
Discriminations associated with respiration (Δ e ) and with photorespiration (Δ f ) were calculated from Equations (9) and (10) (Farquhar & Cernusak, 2012): where e and ƒ are the fractionations associated with respiration and photorespiration, respectively. We assumed ƒ to be −11.6‰ (Lanigan et al., 2008) and that there is no significant fractionation associated with dark respiration during the day (Wingate et al., 2007). However, because respired carbon was likely fixed during prior photosynthesis in the greenhouse, we took e to equal the difference between δ 13 C e (−32 to −37‰) and the isotopic composition for atmospheric CO 2 (δ 13 C atm ) in the greenhouse (assumed to be −8‰; Alonso-Cantabrana & von Caemmerer , 2015): Discrimination associated with g m is described by Farquhar and Cernusak (2012): where a i is the fractionation factor associated with hydration and diffusion in water (1.8‰ at 25°C). Substituting Equation (3) into Equation (12) and rearranging, g m was then calculated as:

| Calculation of C c
Having obtained g m by the chlorophyll fluorescence method, the CO 2 concentration in the chloroplast (C c ) was estimated: g m obtained from the stable isotope discrimination method was strongly correlated with that estimated using the chlorophyll fluorescence method (see Supplementary Methods) (g m values between 0.03 and 0.19; R 2 = 0.8016, p < 0.0001; Figure S1). Given the potential uncertainties with g m estimates obtained from the variable J method, and the increased sensitivity of certain methods for leaves experiencing water stress, we chose to present g m from data carbon isotopic discrimination technique.

| Leaflet water potential measurements
The two leaflets opposite the one used for gas exchange measurements were used to measure water potentials. The first leaflet was cut at petiolule base and bagged (in a clear bag) for 10 min to allow equilibration within the leaflet. Then, using a razor blade~1 cm of leaflet lamina was cut from either side of the middle vein to fit the short petiolule inside the pressure chamber gasket. Chamber pressure was increased slowly until the balancing pressure was reached.
The second leaflet was covered in a dark bag for 20 min before removal to obtain the water potential of the rachis for the remainder of the leaf.

| Leaflet water potential and leaflet hydraulic conductance
Leaflet water potential (Ψ leaflet ) was measured using a pressure chamber (PMS Instrument Company, Model 1505D) immediately after gas exchange measurements between 10 a.m. to 3 p.m. (Williams & Araujo, 2002) (see Supporting Information Method).
Leaflet hydraulic conductance (K leaflet ) was calculated using in situ evaporative flux method according to Brodribb and Holbrook (2003) and Simonin et al. (2015): E is the transpiration rate (mmol m −2 s −1 ) measured using gas exchange system, and ΔΨ bagged leaflet-unbagged leaflet is the difference between bagged leaflet and unbagged leaflet water potential The Ψ leaflet showed 3%-6% variability between the leaflets and the average Ψ leaflet was 3%-5% more negative than Ψ leaf in each species. The average transpiration rate (E) within and between leaflets on the same leaf were compared for these measurements.
Minimal or no significant difference in E existed across individual leaflets for scaling to the total leaflet area.
To quantify the stomatal aperture under well-watered and dehydrated conditions in each species, both hypostomatous, abaxial epidermis imprints using transparent nail polish (water-based) were obtained from the same leaflets used for the gas exchange measurements. Using the imprints, stomata images were taken on a light microscope at 20x (Nikon C2 + , Nikon Instruments Inc.) and used to measure the stomatal pore dimensions. The inner pore width (μm) was divided by the inner pore length (μm) to calculate the stomatal aperture ratio (Rui & Anderson, 2016). Maximum g s (g smax ) was calculated using the stomata pore dimensions (Franks & Beerling, 2009) and used to interpret changes in stomata opening and g s in the two species and further, test the precision of the stomatal aperture quantification method. To calculate stomata size, guard cells length was multiplied by total width, for closed guard cells (Franks & Beerling, 2009).
Five hundred consecutive slices from the grid and phase stacks were selected for segmentation. The resulting image stack was segmented using the methods presented in Théroux-Rancourt et al.
(2020) ( Figure S3). Six slices were labelled manually per scan and were used to train a random-forest model for automated segmentation of the whole scan image stack. The final segmented stacks had individual labels for the adaxial epidermis, abaxial epidermis, mesophyll cells, intercellular airspace, BSEs, veins, and background outside of the scanned leaf. This final stack was used to extract leaf anatomical traits, that is surface areas, volumes, and lengths.

| Mesophyll surface area and porosity
As described by Théroux-Rancourt et al. (2017), mesophyll porosity, θ IAS (m 3 m −3 ) was calculated as the IAS volume as a fraction of the total mesophyll volume. The IAS volume (V IAS ) to mesophyll cell volume (V mes-cell ) ratio and the mesophyll surface area exposed to the IAS (SA mes ) per mesophyll volume (V mes ) were calculated as V IAS /V mescell (m 3 m −3 ) and SA mes /V mes (μm 2 μm −3 ), respectively ( Figure 1).

| Tortuosity and lateral path lengthening
The tortuosity factor, τ(m 2 m −2 ), was defined as the ratio of the diffusive path length within the IAS to the straightest path length in the absence of any physical obstacles to diffusion between a stomate and the cell surface: where geodesic distance (L geo ) is the distance from the stoma to a cell surface, and Euclidean distance, (L Euc ) (Earles et al., 2018). The L geo and L Euc were mapped and quantified for all voxels along the mesophyll Desiccation of the leaf mesophyll and its implications | 1367 surface and τ was calculated for the whole 3D image array as in Earles et al. (2018). Then, leaf-level tortuosity (τ leaf ) was calculated as the mean of τ values at the edge of mesophyll cells. The lateral path lengthening, λ (m m −1 ) was calculated using L Euc , and a second distance map as described by Earles et al. (2018) to measure the shortest unobstructed distance in a straight line between the abaxial epidermis and all points along the mesophyll surface, L epi (Legland et al., 2016): Similarly, leaf-level lateral path lengthening (λ leaf ), was then calculated as the mean of λ values at the edge of mesophyll cells.

| IAS conductance
The τ leaf , λ leaf and θ IAS were used to calculate leaf-level IAS conductance (g IAS ), where D m is the diffusivity of CO 2 in air (m 2 s −1 ).

| Porosity profiles
MicroCT scans for each species under well-watered and dehydrated conditions ( Figure 2) were used to determine porosity profiles from IAS distribution with leaf depth using a plot profile of grey value distribution across leaf excluding adaxial and abaxial epidermis tissue.
The grey values were used to calculate air volume for a known mesophyll area (4 μm 2 ) and based on mesophyll thickness (μm) per individual within each depth after converting pixel to distance (pixel resolution of 0.65 μm).

| Palisade mesophyll cell diameter at paradermal section
The grid reconstructions of microCT images were used to compare the palisade mesophyll cell diameter through paradermal sections at three depths (20%, 40% and 60%) from the adaxial surface in wellwatered versus dehydrated leaves within 0.02 mm 2 of the mesophyll area.
Chlorophyll distributions were obtained by calculating the relative F I G U R E 1 Leaf three-dimensional projection for Juglans regia and Juglans microcarpa under well-watered and dehydrated conditions F I G U R E 2 MicroCT images, representative slices from scans of Juglans regia and Juglans microcarpa leaves under well-watered and dehydrated conditions. Bar equals 250 μm chlorophyll fluorescence (proportional to chlorophyll concentration) at each relative depth. Light absorption gradients, representing relative chlorophyll distribution patterns were measured using chlorophyll fluorescence imaging of leaf cross sections under direct illumination (Vogelmann & Evans, 2002;Vogelmann & Han, 2000).
Fresh samples in a subset of three were cut into~1 cm 2 from the same leaflets and placed on top of a wet paper to protect the specimen from desiccation in a glass holder on the microscope stage

| Bundle sheath extensions area
MicroCT scans for each species were used to calculate the ratio of the BSEs area where parenchyma cells connecting vascular tissue to both epidermises, as a percentage from the mesophyll cross sectional area (i.e., area of mesophyll cells and airspace) using Image J (Griffiths et al., 2013).
Green light penetrates deeper into leaf and is absorbed more equally throughout the leaf profile (Brodersen & Vogelmann, 2010). Therefore, adaxial fluorescence images from green laser illumination, as described above, were used to determine the relative fluorescence for mesophyll tissue adjacent to BSEs and plot their distribution along the mesophyll in Image J. Fluorescence percentage near BSEs were normalised using maximum values per sample and plotted using epi-and adaxial illumination imaging data for the two species.

| Statistics
Linear regression lines were used to describe relationships between A n and C i and paired t-tests were used to compare differences in estimated g m from the isotope discrimination and chlorophyll fluorescence methods using GraphPad prism 8 (GraphPad Software, Inc).
Mixed linear models were used to compare treatments effects on the following physiological variables: A n , g m , g s , C i , C c , Φ PSII , Ψ leaflet , τ leaf , θ IAS , L leaf , L mes , L epi-adaxial , L epi-abaxial , SA mes /V mes , V IAS /V mes-cell , λ leaf and g IAS in the two species under well-watered and dehydrated treatments using SAS 9.4 (SAS Institute Inc).

| Mesophyll traits and IAS parameters
Total leaf thickness (L leaf ; p = 0.0183), mesophyll thickness (L mes ; p = 0.0203), θ IAS (p < 0.0001), V IAS /V mes-cell , g IAS (p < 0.0001) and λ leaf (p = 0.0023) were greater in J. regia compared to J. microcarpa under well-watered conditions (Figure 3). Lower θ IAS in J. microcarpa aligned with significantly greater SA mes /V mes (p < 0.0001, Figure 3). Tortuosity (τ leaf ) ( Figure 3) and adaxial and abaxial epidermis thicknesses (data not shown) were not statistically different between the species. Water stress reduced L leaf (by 8% vs. 9%) and L mes (by 10% vs. 13%) similarly in J. regia and J. microcarpa, respectively. Although the abaxial epidermis showed some shrinkage under dehydration, the abaxial and adaxial epidermis thickness (L epi-abaxial , L epi-adaxial ) were not significantly reduced in either species. Dehydration increased V IAS /V mes-cell in both species by 20% (p < 0.0001) through reducing both V IAS and V mes-cell but in different rates in each species (p < 0.0001, Figure 3). The reductions were in line with an increase in porosity (θ IAS ) in both species under dehydration, but this effect was greater for J. regia than J. microcarpa (p = 0.0065) and significantly higher SA mes /V mes under drought, in J. regia only (p = 0.010) (Figure 3). g IAS increased equally for J. regia (by 23%) and J. microcarpa (by 21%) (p > 0.05) under Desiccation of the leaf mesophyll and its implications dehydration compared to the well-watered condition. Dehydration reduced g IAS contribution to g m (calculated as described by Niinemets & Reichstein, 2003) from 22% to 9% in J. regia, and 23% to 8% in J. microcarpa (p < 0.05). Although there was a significant increase in τ leaf (by 23%) in J. microcarpa under dehydration (p = 0.010) (Figure 3), path lengthening (λ leaf ) did not change in either species.

| CO 2 and light response curves
Despite species-dependent differences in photosynthetic capacity and greater A n at ambient CO 2 (400 µmol mol −1 ) and higher maximum carboxylation rate (V cmax ) and maximum electron transport rate (J max ) in J. regia as expected (Figures 4 and 5), maximum photosynthesis (A max at F I G U R E 3 Intercellular airspace parameters under well-watered (solid bar) and dehydrated (empty bar) treatments for Juglans regia and Juglans microcarpa. a, leaf tortuosity factor (τ leaf , m 2 m −2 ); b, intercellular airspace porosity (θ IAS , m 3 m −3 ); c, intercellular airspace to mesophyll cell volume ratio (V IAS /V mes-cell ); d, leaf thickness (L leaf , μm); e, mesophyll thickness (L mes , μm); f, mesophyll surface area exposed to the intercellular airspace per mesophyll volume basis (SA mes /V mes , μm 2 μm −3 ); g, lateral path lengthening within intercellular airspace (λ leaf , m m −1 ); h, intercellular airspace conductance (g IAS , mol m −2 s −1 bar −1 ). Data points are means of four biological replicates (ramets) per species under well-watered or dehydrated conditions (±SE) C i greater than 750 µmol mol −1 ) was statistically similar for the two species under well-watered conditions (Figure 4). Juglans microcarpa maintained its photosynthetic capacity (i.e., greater J max ) under dehydration to a greater extent compared to J. regia (17% vs. 52% decrease in A max , respectively; p < 0.0001). Dehydration reduced A n significantly in both species (p = 0.0003) with a greater percent decrease in J. regia (by 47%) compared to a 42% for J. microcarpa (p = 0.0023) under ambient CO 2 and saturating light (1500 µmol m −2 s −1 ) (Figure 4). At lower PPFD (50 to 500 µmol m −2 s −1 ) from adaxial illumination, the percent and absolute reductions in A n were similar between the species ( Figure S2). In general, A n was lower with abaxial illumination, however, absolute and percent reductions in A n were similar to those from adaxial illumination in the two species (p = 0.0014; Figure S2).

| Mesophyll conductance and photosynthesis at ambient CO 2
A n and photosystem efficiency (Φ PSII ) at ambient CO 2 were greater in J. regia than J. microcarpa under control conditions (Figures 5a and 4b, p < 0.0083), in agreement with higher g s (p = 0.0080; Figure 5c), g m (p < 0.0001, stable isotope method, Figure 5d), C i (p = 0.0001; Figure 5e), and C c (p < 0.0001, Figure 5f). Reduced A n under dehydration aligned with reductions in g s and g m ( Figure 5).
On the other hand, Φ PSII and C i decreased significantly in J. regia only and C c showed greater reductions with dehydration in J. regia compared to J. microcarpa ( Figure 5, p < 0.0001). The stomatal aperture ratio (inner pore width/inner pore length) was greater for J. regia under well-watered condition (J. regia 0.47 ± 0.04, J. microcarpa 0.37 ± 0.03), and dehydration induced stomatal closure and increased the ratio by 38% in J. regia versus 61% in to J. microcarpa. The relative changes in the pore dimensions were proportional to g s reduction in J. microcarpa (by 68%), but less than reduction in J. regia (by 58%).
Under well-watered conditions, Ψ leaflet (Figure 5h) and K leaflet ( Figure 5g) were similar for the two species, however, they were correlated negatively in both species (R 2 = 0.9985, p = 0.0008). g s responded negatively to decreasing Ψ leaflet (R 2 = 0.9091, p = 0.0465) with a greater reduction in J. microcarpa compared to J. regia (p < 0.0001) (Figure 6), however, the reductions in K leaflet induced by water stress were not linked with significant changes in the number of embolized conduits for either species (Figure 6). The percent ratio of embolized conduits (C emb ) per number of conduits (C)  in representative microCT images (800 μm of each cross; n = 6) in secondary veins was similar between under well-watered and dehydrated conditions in J. regia (11.3% vs. 12.4%) and J. microcarpa (20.2% vs. 14.2%) (p > 0.05). Similarly, no significant effect on emboli formation was found in tertiary veins under well-watered versus dehydrated conditions in J. regia (7.4% vs. 10.9%) and J. microcarpa, (3.6% vs. 4.6%) (p > 0.05).

| Chlorophyll distribution, light absorption and porosity profiles
Relative chlorophyll distribution was estimated from fluorescence profiles in leaf cross sections using epi-illumination (Figures 7 and S5).
The patterns showed species-specific differences; J. regia exhibited a F I G U R E 4 Photosynthetic CO 2 response curves, relationship between mean A n (net assimilation rate) and C i (intercellular airspace CO 2 concentration) at 1500 µmol m −2 s −1 photosynthetic photon flux density were constructed using FvCB model (Sharkey, 2016), averaged over four replications in Juglans regia (panels a and c) and Juglans microcarpa (panels b and d) under well-watered (solid symbols and top row) and dehydrated (empty symbols and bottom row) treatments (±SE; n = 4). Assimilation rate at saturating CO 2 (A max ), Rubisco and RuBP regeneration limitations are indicated for each species [Color figure can be viewed at wileyonlinelibrary.com] Desiccation of the leaf mesophyll and its implications | 1371 single peak in relative fluorescence around palisade mesophyll, within 0%-20% depth from the adaxial epidermis (Figure 8a), whereas J. microcarpa had double peaks at 10%-40% and 80%-100% depth ( Figure 8c). A rapid attenuation after 20% and leveling off at 60% of the depth in J. regia was different than the pattern for J. microcarpa, where there was a depression between 40% and 80% of leaf depth.
As expected, porosity increased with depth from adaxial surface in two species, but the porosity profile complemented the fluorescence profile better in J. regia with an increase in porosity after 40% of depth, around spongy mesophyll, and a maximum between 90% adaxial depth (Figure 8b). In J. microcarpa, porosity changed less compared to relative fluorescence suggesting that components other than cell packing are involved in fluorescence gradients across the leaf. However, the porosity increased smoothly after 20% depth and reached the maximum between 80% and 100% depth from adaxial surface (Figure 8d). F I G U R E 5 Photosynthetic traits under wellwatered (solid bar) and dehydrated (empty bar) treatments for Juglans regia (J. reg) and Juglans microcarpa (J. mic). a, net assimilation rate (A n , µmol CO 2 m −2 s −1 ); b, photosystem II efficiency (Φ PSII ); c, stomatal conductance (g s , mol m −2 s −1 ); d, mesophyll conductance (g m , mol CO 2 m −2 s −1 ); e, intercellular airspace CO 2 (C i , µmol mol −1 ); f, chloroplast CO 2 (C c , µmol mol −1 ); g, leaflet hydraulic conductance (K leaflet , mmol m −2 s −1 MPa −1 ); h, leaflet water potential (Ψ leaf , MPa). Data points are means of four biological replicates (ramets) per species under wellwatered or dehydrated conditions (±SE). The measurements were taken at 400 µmol mol −1 and 1500 µmol m −2 s −1 photosynthetic photon flux density Under dehydration, J. microcarpa showed an increase in relative fluorescence within the first 20% of the leaf depth and a consistent reduction between 30% and 100% depths (Figure 8c).
In contrast, J. regia did not show a significant difference between watering conditions. Porosity increased across the leaf profile under dehydration in both species (p < 0.002) with a greater increase after 60% of adaxial depth in both species (Figure 8).
Between 0% and 20% depth from the adaxial epidermis and under adaxial illumination, absorption of red light was greater in J. microcarpa under dehydration, but it decreased significantly after 30% depth compared to the well-watered condition in all wavelengths ( Figure 9). Only within 20% depth from the adaxial surface, J. regia showed higher absorption under dehydration at the blue wavelength (Figure 9). Illumination direction had a significant impact on absorption depth in red and blue wavelengths; maximum absorption at adaxial irradiance occurred at first 30% of depth from adaxial surface, whereas it was at first 60% of depth from abaxial epidermis under abaxial illumination. There was no consistent difference in light absorption profiles between wellwatered and dehydrated conditions under abaxial illumination in F I G U R E 6 Leaflet water potential (Ψ leaf , MPa) and leaflet hydraulic conductance (K leaflet , mmol m −2 s −1 MPa −1 ) (a) and stomatal conductance (g s , mol m −2 s −1 ) (b) relationship under well-watered (solid) and dehydrated (empty) treatments in Juglans regia (circle) and Juglans microcarpa (square). MicroCT images on the left are representative slices comparing conduit embolism from scans of J. regia and J. microcarpa leaves under well-watered and dehydrated conditions. Six slices from each microCT scan were used to count number of emboli per secondary and tertiary veins under well-watered and dehydrated conditions. Number of emboli divided by number of veins within the red circles region did not show a significant change in either secondary or tertiary veins under dehydration from well-watered condition. Bar equals 250 μm [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 7 Top images compare microscopic cross sections for Juglans regia and Juglans microcarpa under well-watered conditions. Bottom images compare spatial chlorophyll distribution from original epi-illumination imaging and false-coloured associated pairs, representative of the two species leaves under well-watered and dehydrated conditions [Color figure can be viewed at wileyonlinelibrary.com] Desiccation of the leaf mesophyll and its implications | 1373 J. regia, while the green wavelength showed a significantly higher absorption between 60% and 100% of depth in J. microcarpa.

| Paradermal cells and bundle sheath extensions
The paradermal section images at 20%, 40% and 60% depths from the adaxial epidermal surface showed a significant decrease in the palisade cell diameter (μm) under dehydration in both species; diameters decreased by 14%, 9%, 15% in J. microcarpa and by 14%, 17%, 19% in J. regia (p < 0.05; see example images in Figure 10) at each increasing depth, respectively. BSEs were more prominent in J. microcarpa occupying 15% of the mesophyll volume compared to 8% in J. regia (p = 0.001). That was related to higher vein density with narrow BSEs width ( Figure 11) for J. microcarpa. Using epiillumination data, J. microcarpa had higher fluorescence near the BSEs cells under well-watered than dehydrated condition, compared to J. regia. (Figure 11, p < 0.0001). Under adaxial green wavelength illumination, differences in fluorescence near BSEs between the two species were not significant (normalised data shown in Figure S6). The exchange of water and CO 2 and light absorption gradients are connected via mesophyll geometry and stress-related changes in leaf anatomical characteristics induce responses in g s , g m , and A n . Using microCT imaging, we showed that θ IAS and g IAS in J. regia and J. microcarpa corresponded with species-specific differences in A n , g s and g m obtained from gas exchange, chlorophyll fluorescence, and stable carbon isotope methods (Figures 3, 5 and S1). Well-watered J. regia, with thicker leaves and denser mesophyll cells in the upper palisade, had higher θ IAS , V IAS /V mes-cell and g IAS aligned with greater A n , g s and g m and higher chlorophyll concentration near the adaxial surface. Juglans regia mesophyll structure with higher θ IAS and greater IAS distribution between mesophyll cells increase the g IAS thorough an effective lateral diffusion (Figure 3). J. regia with less BSEs, forms less physical barrier to gas diffusion, thus the resistance to diffusion from gas to liquid phase decreases due to a greater lateral conductivity in this mesophyll type (e.g. homobaric) (Evans & von Caemmerer, 1996;Pieruschka et al., 2005). This inner mesophyll structure couples with a greater need for g s to keep up with the higher demand for C i concentration and higher g smax ( Figure 5). According to lateral CO 2 diffusion modelling by Pieruschka et al. (2007), larger interconnected airspace can improve CO 2 diffusivity through IAS, and enhance C i in coordination with g s as seen in J. regia (Figures 3 and 5). Juglans regia has larger stomata (average size, 122 vs. 72 μm 2 ) but fewer stomata (62 per mm 2 ) than J. microcarpa (79 per mm 2 ), and higher λ leaf in J. regia is linked with its lower stomatal density, which would increase CO 2 diffusion length and higher θ IAS near spongy mesophyll in hypostomatous species (similar to patterns reported by Earles et al., 2018 andHarwood et al., 2021). Therefore, J. regia benefits from a higher CO 2 diffusion capacity, and exhibits improved performance through increasing maximum carboxylation rate (V cmax ) and A n under lower CO 2 concentrations (Figure 4). At ambient CO 2 , where RUBP-regeneration is limiting, J. regia mesophyll with lower diffusion resistance (e.g., more porous leaves) and higher enzymatic activity during CO 2 fixation and carbohydrate formation (i.e., Calvin cycle) shows greater maximum electron transport rate (J max ), g m , and A n under well-watered condition.
Smaller mesophyll cell size potentially improves CO 2 accessibility to Rubisco by enhancing SA mes /V mes and chloroplast surface area (Ren et al., 2019;Terashima et al., 2006;Tholen et al., 2012), and consequently increases conductance within the liquid phase (g liq ), while greater θ IAS is associated with higher g IAS . g IAS contributed 29% and 25% to g m in J. regia and J. microcarpa, respectively, consistent with previous findings in tree F I G U R E 10 Paradermal sections at 20% depth from the adaxial surface in Juglans regia and Juglans microcarpa under well-watered and dehydrated treatments. Similar responses were seen at 40% and 60% depths (images not shown). Bar equals 50 μm F I G U R E 11 Fluorescence percentage normalised using maximum values per sample using epi-illumination. It indicates the spatial distribution of fluorescence in equal horizontal distance from bundle sheath extensions (BSEs) in Juglans regia (circles-left data panel) and Juglans microcarpa (squares-right data panel) under well-watered (solid) and dehydrated (empty) treatments (±SE; n = 4). Positive and negative values present variation in grayscale values at right or left side of BSEs. BSEs location is designated with white arrows; mean J. regia BSE width was~4.7-5.0 μm and mean J. microcarpa BSE width was~3.0-3.5 μm. Zero on the x-axis of the data plots represents the centre of the measured BSE Desiccation of the leaf mesophyll and its implications | 1375 species; Populus tremula (23%-25%; Tosens et al., 2012), Quercus ilex L. (23%; Niinemets & Reichstein, 2003), and four Eucalyptus species (8%-21%; Harwood et al., 2021) under well-watered condition and within the estimated limitation range by g IAS (3%-37%) for A n in hypostomatous species (Parkhurst & Mott, 1990). Higher SA mes /V mes in well-watered J. microcarpa was associated with lower θ IAS and V IAS /V mes-cell , and less variations in airspace distribution and relative fluorescence across leaf profiles (Figures 3 and 8). Juglans microcarpa mesophyll geometry increases resistances for CO 2 diffusion in the gas phase (lower g IAS ) through disconnecting mesophyll tissues and dividing into compartments and to the liquid phase, through potentially higher cell wall thickness (Pieruschka et al., 2008;Ren et al., 2019;Tomás et al., 2013), in line with less diffusion and carboxylation capacity, exhibited as g m , V cmax and J max responses for J. microcarpa (Figures 4 and 5). However, more biomass allocations toward cell packing and extensive bundle sheaths extensions, in species like J. microcarpa, improves the structural tolerance under the environmental stresses (e.g., low water) (Hikosaka & Shigeno, 2009;Niinemets et al., 2007).  (Cui et al., 1991;Gotoh et al., 2018), also reported for intermediate shade-tolerant species (Hanba et al., 2002).
This mesophyll structure appears to be an adaptation to maximize light absorption under varying light conditions (Leegood, 2008;Terashima, 1992;Tholen et al., 2012); as described by Vogelmann et al. (1996), larger IAS acts as "hall of mirrors" and improves absorption by multiple reflections between airspace, mesophyll and epidermal cells. J. microcarpa, on the other hand, had a narrower range of airspace distribution across leaf profiles with a smaller range of relative fluorescence that was distributed more evenly throughout the mesophyll, in line with mesophyll architecture for species from a high light environment (Hanba et al., 2002). The discrepancy between porosity and relative fluorescence profiles in J. microcarpa (Figure 8) could be attributed to light scattering inside the leaf due to other cell types (e.g., more BSEs) (Vogelmann & Evans, 2002), exhibited as significant difference in relative fluorescence near BSEs in two species under epi-illumination ( Figure 11). Mesophyll partitioning due to BSE presence (i.e., heterobaric leaves) is predicted to increase light penetration and overall absorption in various directions, mostly through spongy cells (Vogelmann & Martin, 1993;Xiao et al., 2016).
More BSE volume in J. microcarpa compared to J. regia (15% vs. 8%, respectively) and greater fluorescence from the cells near the BSEs ( Figure 11) under both well-watered and dehydrated conditions indicate that J. microcarpa can utilise light deeper into the leaf (absolute data not shown). Species with BSEs that contain transparent cells with few or no chloroplasts are proposed to acclimate more effectively to drought since the light transmitted through BSEs is elevated at red and blue wavelengths and it may modify available internal light for photosynthetic tissues (Karabourniotis et al., 2000).
BSE-containing species like J. microcarpa may rely more on the structural support by BSEs than turgor associated with water supply to sustain their leaf stiffness (Read & Stokes, 2006) as an ecological response to water shortage in their growth habitat even if it costs a reduction in number of photosynthesising cells and eventually the carbon fixation.
4.2 | Drought-induced changes in photosynthetic capacity related to structure and function changes Dehydration had a negative impact on both species by reducing A n , g s and g m . Juglans regia was shown to be more susceptible to stress with a decline in Φ PSII and a greater imbalance in energy distribution between PSII and PSI by more reduced distribution to PSII, also suggesting photodamage-related decreases in light use efficiency under dehydration ( Figure S2). Larger IAS volume increases evaporation surface for mesophyll cells, resulting in irreversible mesophyll cell shrinkage and potential permanent damage to the photosystems (Buckley et al., 2017;Rockwell et al., 2014c;Sack & Frole, 2006), as reflected by lower A max for J. regia under saturating CO 2 (Figure 4). A further reduction in C i at low g s , concurrent with a decline in Φ PSII suggests an increase in photorespiration (Medrano et al., 2002) as seen in J. regia. According to a sequential baseline presented by Trueba et al. (2019), g s could decrease by 50% before TLP, whereas passing the TLP under severe stress can lead to permanent damage to the chlorophyll fluorescence as occurred in J. regia. This species exhibited a limited range of resilience to drought, as the widely cultivated walnut species in commercial nut production; this is consistent with irrigation management practices aimed at avoiding water stress in this species. In contrast, J. microcarpa with a more conservative water use strategy and a higher intrinsic water use efficiency (WUE i , 90 μmol CO 2 mol −1 H 2 O vs. 76 μmol CO 2 mol −1 H 2 O for J. regia) under the well-watered condition, functions at lower water potentials under drought. The inherently lower g s under well-watered conditions and a greater reduction in Ψ leaflet under dehydration for J. microcarpa suggests that this species may maintain low g s at lower Ψ leaflet closer to its turgor loss point (TLP). Species with prominent heterobaric leaves are also expected to show more nonuniform stomatal closure in response to environment stress (Kamakura et al., 2011). Juglans microcarpa is reported to be less vulnerable than J. regia to the xylem cavitation; it shows 12% decrease in stem xylem hydraulic conductivity at lower Ψ xylem (e.g.,~−1.6 vs. −1.3 MPa in J. regia) (Jinagool et al., 2018). Tyree et al. (1993) also measured about 50% reduction in petiole hydraulic conductivity for J. regia at Ψ xylem− 1.4 MPa. Although TLP was not measured in our study, a greater increase in porosity and decrease in palisade cell diameter at less negative water potential suggest that J. regia functioned closer to its potential TLP under dehydration. These are commensurate with previous studies reporting reductions in mesophyll cell thickness and changes in IAS thickness near TLP to be species-specific (Sancho-Knapik et al., 2011;Scoffoni et al., 2014). In leaves with a structure like J. microcarpa, where the epidermis is more hydraulically integrated due to the presence of BSEs, water can bypass parts of the mesophyll reaching evaporation sites near epidermis as proposed by Zwieniecki et al. (2007). In this system, stomatal function is more closely linked to changes in xylem hydraulic conductance and stomata may show delays in the closure. Therefore, the relative presence of BSEs might play a role in allowing species to operate at more negative Ψ xylem .
Dehydration reduced g s equally in both species, but decreased K leaflet more severely in J. microcarpa, in relative terms ( Figure 5). It has been shown that mutants lacking BSEs (e.g., Solanum lycopersicum) have lower g s , A n and K leaf than wild-type plants (Zsögön et al., 2015). In addition, BSEs are proposed to slow down stomatal closure under stress-induced conditions by enhancing hydraulic conductance through extravascular pathway (Barbosa et al., 2019;Buckley et al., 2011). Despite the higher presence of BSEs, the lower minimum g s (0.02 mol H 2 O m −2 s −1 ) in J. microcarpa compared to J. regia (0.05 mol H 2 O m −2 s −1 ) may represent greater response through stomatal closure via highly reduced K leaflet (driven by changes outside the xylem) at the expense of significantly lower A n (Figures 5 and 6) and higher WUE i at ambient CO 2 under drought. Further, changes in g s associated with stomatal aperture were supported by epidermal imprints, where stomatal aperture ratio decreased more for J. microcarpa (by 61%) compared to J. regia (by 38%). Dehydration reduced g s to 27% and 11% of the g smax for J. regia and J. microcarpa, respectively, further supporting the inherent difference in sensitivity of g s to dehydration. Still, finding no significant emboli formation within xylem veins in either species (Figure 6) was in agreement with recent studies suggesting that declines in leaf hydraulic conductance are mostly due to declines in outside-xylem tissue hydraulic conductance under mild to moderate dehydration and even beyond the TLP (Albuquerque et al., 2020;Scoffoni et al., 2017). Furthermore, the contradicting results for J. microcarpa may highlight the importance of aquaporins activity in changing K leaf under induced conditions, such as a positive link between aquaporins abundance and K leaf under high (vs. low) light was reported in J. regia (Cochard et al., 2007).
Drought-induced shrinkage in mesophyll cells opens up more IAS volume within mesophyll, however nonuniform changes in cell shape can increase resistance to CO 2 diffusion by reducing chloroplast surface area facing cell walls (Cano et al., 2013;Tosens et al., 2012;Xiao & Zhu, 2017). Dehydration increased porosity, leading to an increase in g IAS , however, the g IAS contribution to g m was reduced to 6-8% in both species under dehydration (p < 0.05). Therefore, limitation imposed by g liq to g m may increase under drought via chloroplast re-positioning and activity of CAs and aquaporins (Evans et al., 2009;Miyazawa et al., 2008;Momayyezi et al., 2020;Tholen et al., 2008;Tomás et al., 2013) more so than changes in resistance via anatomical components such as the cell wall thickness and cell wall composition (e.g., lignin deposition) (Evans, 2021;Roig-Oliver et al., 2020). We must note, however, that the porous media approach of Earles et al. (2018) to compute g IAS , a step forward in the representation of the inherent 3D nature of the leaf mesophyll, does not fully account for the specificities of the diffusion within the leaf. As discussed by Harwood et al. (2021), path lengthening is a step forward to account for the discrete nature of stomata along the epidermis, but path shortening within the mesophyll would also occurs because of the gradient of carbon assimilation within the leaf profile, and would theoretically increase g IAS . In the present case, as water stress decreases photosynthesis, the path shortening effect could be smaller and might cancel out the g IAS increases caused by higher porosity and lower tortuosity. Thus, the present result must be seen as a potential increase in g IAS caused by anatomical changes.   in light penetration and overall absorption efficiency through the leaf under stress.

| CONCLUSIONS
Mesophyll structure has a substantial role in both CO 2 diffusion and light absorption. Juglans regia mesophyll with thick palisade layers and higher IAS volume between mesophyll cells and mostly near the spongy layer, has higher g m in line with more carboxylation capacity and greater light absorption under well-watered condition. A more porous mesophyll with less BSEs has less anatomical leverage to tolerate dehydration and maintain the gas exchange in association with hydraulic components, and increases risk of damage to photosynthetic machinery. While more mesophyll cell density with less IAS distribution and greater BSEs (e.g., heterobaric leaf) can increase resistance to CO 2 diffusion and lower overall light absorption and photosynthesis, it performs better in light absorption under drought.
Greater BSEs in J. microcarpa leaves provide physical and hydraulic support leading to less mesophyll cell shrinkage with minimum damage to the carboxylation activity.
Desiccation of the leaf mesophyll and its implications | 1377