2.1 Plant material and ozone exposure

Cuttings of two Populus × canadensis Moench clones (Populus deltoides × P. nigra) Robusta (RB) and Soligo (SL) were grown for 5 weeks in 10-L pots filled with commercial potting soil (Terreau Uni Alsa LFc, Gramoflor SPI Universel) mixed with a slow-release 13:13:13 N:P:K fertiliser (Nutricote T 100 Fertil) stacked on a 2 cm layer of clay pebbles. Photoperiod, relative humidity and temperature in eight environment-controlled growth chambers were set at 14 h (300 ± 30 µmol m⁻² s⁻¹, photosynthetic photon flux density [PPFD] at leaf height), 65 ± 5% and 23/18°C (day/night), respectively. In each chamber, growth conditions were recorded individually and verified to be similar throughout the study period. A total of 12 replicates of both clones were evenly distributed in the eight climate chambers. After a 7-day acclimation period, four chambers were exposed to charcoal-filtered air, whereas the other four were treated daily with O₃ at 120 ± 5 ppb for 13 h per day for 21 days. An O₃ generator produced O₃ from pure O₂ (OZ500; Fischer, Bonn, Germany and CMG3-3; Innovatec II) and an O₃ analyser (O341M, Environment S.A.) continuously monitored the O₃ concentration in the climate chambers. Plants were irrigated daily with tap water and rotated regularly to avoid any effect of chamber position. All measurements were carried out with five to six plants per treatment and per clone.

2.2 Gas exchange and chlorophyll fluorescence measurements

2.3 Estimation of leaf gas exchange characteristics

2.4 Quantitative limitation analysis of photosynthesis

2.5 Light and electron microscopy and anatomical measurements

Sample preparation for light microscopy and transmission electron microscopy (TEM) followed Tosens et al. (2012b). After the gas exchange measurements, two leaf discs with a diameter of 10.6 mm were taken from that same leaf from each replicate, avoiding large veins. The discs were then infiltrated in a fixation buffer (2% glutaric aldehyde in 0.1 M phosphate buffer, pH = 7.1) in a syringe under vacuum and post-fixed in a 2% osmium tetroxide solution for 2 h. Subsequently, stripes of leaves of ~5 × 2 mm were cut from the leaf discs and dehydrated in a series of increasingly stronger ethanol solution and embedded in LR white resin (Electron Microscopy Sciences) according to standard procedure (Tosens et al., 2012b). An ultramicrotome (Leica EM UC7, Leica) was used to prepare semi-thin leaf cross-sections of 1 μm for light microscopy, and ultra-thin cross sections of 70 nm for TEM. The sections for light microscopy were stained with toluidine blue (Sigma Aldrich), while the sections for TEM were mounted on carbon-covered copper meshes and stained with uranyl acetate and lead citrate (Electron Microscopy Sciences). The semi-thin sections were viewed in bright field and photographed at ×100 magnification with an EVOS cell imaging system (Thermo Fisher Scientific). The ultra-thin sections were viewed and photographed with a Philips Tecnai 10 TEM (FEI) using an accelerating voltage of 80 kV. Images were analysed with ImageJ software (US National Institutes of Health, Bethesda, MD, USA).

2.6 Mesophyll conductance modelled based on anatomical characteristics

2.7 Quantitative analysis of partial limitations of g_m

2.8 Statistical analyses

All statistical analyses were performed in R (R Core Team, 2018). A two-way analysis of variance (ANOVA) was run using the emmeans (Lenth et al., 2018) and rstatix (Alboukadel Kassambara, 2019) packages, with clone and treatment as factors to test their effects on different characteristics. The ANOVA assumptions of homoscedasticity and normality were checked by performing a Levene's test and Shapiro test, respectively. The data were log-transformed when requirements were not fulfilled. If a significant effect was observed, a post hoc analysis was performed using Tukey's multiple pairwise comparison test using the glht function of the multcomp package (Hothorn et al., 2009). The statistical relationships between clones and treatments were explored by linear and non-linear regression analyses, depending on which model provided a better fit. For each relationship, the coefficient of determination, the Pearson correlation coefficient and the respective p values were obtained using the ggscatter function of the ggpubr package (Alboukadel Kassambara, 2016) and Microsoft Office Excel, 2019. A principal component analysis (PCA) was used to identify the major axes of variation among the major leaf integrative and anatomical traits associated with g_m, A_net and g_sc. The R functions prcomp (stats package) and ggbiplot (ggplot2 package) were used to perform and visualise the PCA. The data were grouped by clone and treatment and the different measurement units of the several variables considered were scaled using the ‘R scale = TRUE’ setting within the prcomp function. The first two dimensions were retained for further analysis (for Eigenvalues and component loadings see Table S2).

3.1 Strong decrease of mesophyll conductance and assimilation rate and while stomatal conductance and V_cmax remained unaffected under elevated O₃

Net assimilation rate (A_net) was 18% higher in the hybrid poplar clone Robusta (RB) than in the clone Soligo (SL) regardless of O₃ treatment (Figure 1a). Stomatal (g_sc, Figure 1b) and mesophyll (g_m Figure 1c) conductances to CO₂ were similar among clones, but the CO₂ concentration at the carboxylation sites in the chloroplast (C_c, Figure 1d), maximum carboxylase activity of Rubisco (V_cmax(C_c), Figure 1e) and the capacity for photosynthetic electron transport rate (J₁₀₀₀(C_c), Figure 1f) were greater in RB than those in SL (18% for C_c, 26% for V_cmax(C_c) and 41% for J₁₀₀₀(C_c)).

Upon O₃ exposure, A_net showed a similar decrease in both clones, 21% in RB and 22% in SL (Figure 1a). g_sc was not affected by O₃ (Figure 1b), but O₃ exposure caused a substantial decrease in g_m in both clones, by 57% in RB and 62% in SL (Figure 1c). C_c decreased in both clones, by 20% in RB, and even more, by 33% in SL (Figure 1d). Neither V_cmax (C_c) nor J₁₀₀₀(C_c) were affected by O₃ in both clones (Figure 1e,f). While intercellular CO₂ partial pressure (C_i) was unaffected by O₃, the CO₂ drawdown, C_i -C_c increased by 78% and 99% in RB and SL, respectively (Figures S2 and S3).

3.2 Mesophyll conductance was the most limiting factor for net assimilation rate under elevated O₃

A quantitative A_net limitation analysis found that stomatal limitation of photosynthesis (l_s) was the most important limitation in control plants with 43 ± 2.9%, while l_m and l_b accounted for 26 ± 2.4% and 31 ± 1.1% on average across both clones (Figure 2). In control plants, l_b was higher in SL than RB, whereas l_s and l_m did not differ between SL and RB (Figure 2). Under elevated O₃, l_s and l_b decreased to the same extent in both clones (20% and 27%, respectively), while l_m increased by 66% on average in both clones, being the most important limitation of A_net in O₃-stressed plants (Figure 2).

3.3 Leaf integrative traits showed pronounced intraspecific differences, but were mostly unaffected by O₃

LMA was 16% higher in RB than in SL (Table 2). Leaf density (D_leaf) was in the same range for both clones (Table 2). Leaf and mesophyll thickness (T_leaf and T_mes) 14% and 15% higher, respectively, in RB compared to SL (Table 2 for quantitative differences, Figure 3a–d for microscopic images). Similarly, palisade and spongy thickness (T_pal and T_spo) was 16% and 13% thicker, respectively, in RB. However, the T_pal/T_spo ratio was significantly decreased under O₃ in both genotypes, by 14% in RB and 12% in SL. The fraction of intercellular airspace (f_ias) was 24% higher in RB compared to SL, regardless of the O₃ treatment. O₃ affected D_leaf only in SL (+18%). LMA, T_leaf & T_mes, T_pal & T_spo and f_ias were not affected by O₃.

3.4 Cell wall and cytoplasm thickness exhibited a clone-dependent increase, while S_c/S remained unaffected

Cell wall thickness (T_cw) was on average 24% thicker in RB while cytoplasm thickness (T_cyt) was almost identical for both clones (Table 3 and Figure 3i–l). Under O₃, T_cw increased by a similar magnitude under elevated O₃ in both clones, by 44% in RB and 41% in SL (Table 3 for quantitative differences, Figure 3e–h for microscopic images). In contrast, T_cyt showed a very different magnitude of increase in the two clones with a 77% increase in RB versus 44% in SL. Chloroplast length (L_chl) decreased by 17% in RB and 13% in SL, while chloroplast thickness (T_chl) was not affected by O₃ (Table 3 and Figure 3e–h). This translated into an increase of the T_chl/L_chl ratio by 19% in RB and 14% in SL. The interchloroplast distance (ΔL_cyt2) remained unaffected in SL, but it increased by 214% in RB. The mean number of chloroplasts per cell was not affected in none of the clones. Neither S_c/S nor S_mes/S were affected by O_3, whereas S_c/S_mes decreased by 30% in RB, but did not change in SL (Table 3).

3.5 A_net and g_m depended more strongly on leaf anatomical traits in the clone Robusta

A PCA was performed to characterise the relationship of gas exchange characteristics with leaf integrative and anatomical traits in both clones and treatments. The first two principal components explained 54.9% of variation in data (Figure 4). Axis 1 explained 36.3% and was positively associated with T_cw, T_cyt and especially with ΔL_cyt2, while a negative association was found for S_c/S_m and notably for T_pal/T_spo (Figure 4). Axis 2 explained 18.6% of the variation and was negatively associated with S_c/S, A_net, LMA and T_leaf (Figure 4). The traits T_chl/L_chl, g_sc, and D_leaf did not contribute to any dimension while S_mes/S and f_ias were associated with both dimensions. Thus, g_m was positively associated to A_net, S_c/S and S_c/S_m, but negatively correlated to T_cw, T_cyt and ΔL_cyt2 (Figure 4). No correlation was found between g_m and S_mes/S, T_leaf, LMA as well as f_ias. Axis 1 separated the two treatment groups, although those of SL were less clearly distinguished (Figure 4). Axis 2, in turn, tended to separate the two clones (Figure 4).

3.6 Importance of S_c/S_m, T_cw and T_cyt as drivers of g_m response under elevated O₃

T_cw and T_cyt were strongly negatively related to g_m in RB, while in SL these trends were also present yet non-significant (Figure 5a,b). Moreover, a large range of T_cw under O₃ was observed in both clones (Figure 5a), whereas for T_cyt this was only the case in SL (Figure 5b). Interestingly, greater T_cw values in RB did not result in a lower g_m under O₃ exposure compared to SL with lower T_cw. A positive relationship was found between S_c/S_m and g_m (Figure 5c). However, this relationship for all data pooled was significant only for RB (Figure 5c). SL showed a high variation in the control treatment, and a positive relationship was only present for O₃-treated plants (Figure 5c).

3.7 Anatomical model of mesophyll conductance underestimates g_m values under control conditions

An anatomical model was applied to estimate diffusion conductance as driven by leaf anatomical traits (g_{m_Anat}) and quantify the role of each component of the CO₂ mesophyll pathway in limiting g_{m_Anat}. The modelled and measured g_m values were strongly correlated and were close to the 1:1 line (Figure 6a). However, the model underestimated g_m values exceeding 0.2 mol mol⁻² s⁻¹ bar⁻¹ by 51% and 40% in control of RB and SL, respectively (Figure 6a; Figure S4). There was a strong fit between g_m and g_{m_Anat} in the O₃ treatment group of RB. However, estimates for SL were 41% higher for g_{m_Anat} suggesting that potentially important variables remain uncaptured, which is also consistent with the PCA results (Figure 4).

3.8 Increased cell wall and cytoplasm thickness are the main drivers of liquid-phase limitation while gas-phase limitation remained unaffected under O₃

The gas phase conductance (g_ias) was not affected by O₃ nor by clone (Figure 6b). The liquid phase conductance (g_liq) decreased by 26% in RB and 19% in SL in O₃-exposed plants (Figure 6b). Moreover, g_liq was 22% higher across both treatments in SL (Figure 6b). Gas phase limitation (l_ias) decreased by 28% on average while liquid phase limitation (l_liq) increased by 11% on average under O₃ (Figure 6c). Cell wall conductance (g_cw) showed a strong decrease of 61% and 51% in RB and SL, respectively (Figure 6d). In addition, g_cw was twofold higher in SL compared to RB in both treatments. Cytosol conductance (g_cyt) decreased by 47% in RB and 28% in SL (Figure 6d). Stroma conductance (g_str) was not affected by O₃, but it was on average 10% higher in SL (Figure 6d). Cell wall limitation (l_cw) more than doubled by about 108% on average in both clones under O₃, yet being 43% higher in RB than in SL across both treatments (Figure 6e). Cytosol limitation (l_cyt) increased by 49% on average, whereas the share of g_liq limited by the stroma (l_str) showed a mean decrease of 15% in both clones. Plasma membrane and chloroplast envelope limitation (l_pl/env) decreased by 26% in RB but only by 13% in SL, leading to a 24% higher l_pl/env in SL under O₃ compared to RB (Figure 6e).

3.9 Considerable increase of cytosolic limitations under elevated O₃

According to the quantitative anatomical model, the inner cell liquid path diffusion conductance was divided into the partial liquid phase conductance for exposed cell wall portions lined with chloroplasts (g_cel1) and for cell wall portions without chloroplasts (g_cel2). g_cel1 decreased in both clones by 14% on average (Figure 7a). Yet, SL had on average a 15% higher g_cel1 across both treatments. g_cel2 remained unaffected by O₃ but was on average 42% lower across treatments and clones compared to g_cel1 (Figure 7a). The corresponding path limitation l_cel1 increased by 7% on average while l_cel2 remained unaffected (Figure 7b). Also, l_cel2 was 82% higher than l_cel1. Cytosolic conductance from the plasmalemma perpendicular to the cell wall (g_cyt1) showed a decrease of 46% in RB and 30% in SL in O₃-exposed plants (Figure 7c). Stromal conductance on the same path (g_str1) was unaffected by O₃ but on average 16% higher in SL than in RB across both treatments (Figure 7c). Cytosolic conductance parallel and/or diagonal from the plasmalemma to the cell wall (g_cyt2) decreased by 47% in RB and 28% in SL in response to O₃, while the respective stromal conductance (g_str2) increased by 20% on average across both clones (Figure 7c). Cytosolic limitation perpendicular to the cell wall (l_cyt1) increased by 52% on average, while the respective stromal limitation (l_str1) remained unaffected under O₃ (Figure 7d). Cytosolic limitation parallel to the cell wall (l_cyt2) increased by 52% on average under O₃ with RB showing a 29% higher increase than SL. The respective stromal limitation (l_st12) decreased by 23% on average yet being 8% lower in RB than in SL. The plasma membrane and chloroplast envelope limitations (l_pl/env) decreased by 11% in RB and 5% in SL (Figure 7d).

4.1 Mesophyll conductance limits net assimilation rate the most under elevated O₃

We analysed the response of key foliage photosynthetic traits to chronic O₃ exposure and observed a decrease in g_m, C_c and A_net (Figure 1a–f). However, stomatal conductance (g_sc) and foliage biochemical photosynthesis potentials, C_c-based V_cmax and J₁₀₀₀, were not affected by O₃ (Figure 1a–f), which is in accordance with our previous findings (Joffe et al., 2022). Thus, these data indicate that the reduction in A_net by O₃ exposure was primarily driven by reductions in g_m in our study. Estimation of g_m from gas exchange and chlorophyll fluorescence is sensitive to parameters assumed in the model, in particular Г*, J and R_d for which literature reported methodological issues regarding their assessment (Farquhar & Busch, 2017; Gong et al., 2018; Tcherkez et al., 2017). However, it was previously reported that varying these parameters by ±10% did not conflict with our conclusions and confirmed the large impact of O₃ on g_m (Joffe et al., 2022). g_m estimates are also sensitive to light capture properties of the leaf, including the leaf absorbance and (α) and the fraction of photons absorbed by PSII (β). Sensitivity analysis revealed that changing both factors could explain the differences of g_m among clones, but not those observed between treatments (Table S3). This confirms again that the large decrease in g_m under O₃ is real and not due to differential leaf traits. Despite g_sc not being affected by O₃, it was much lower in SL under elevated O₃ compared to RB (Figure 1b) suggesting a different magnitude of O₃ uptake between the two clones (Dumont et al., 2014; Dusart et al., 2019). As PCA suggests, g_sc was neither related to A_net and g_m nor to any anatomical feature (Figure 4), so a direct relationship between O₃ uptake and impaired CO₂ assimilation was unlikely in this study. Nevertheless, a decoupling of photosynthetic activity and g_sc under chronic O₃ exposure has been observed in previous works (Hoshika et al., 2012; Lombardozzi et al., 2012). This can be due to the often observed O₃-caused stomatal sluggishness affecting g_sc response to CO₂ in comparison to A_net or g_m (Hoshika et al., 2018). A strong decrease of A_net and C_c, and g_m, while C_i remained unaffected supports the explanation that reduction in photosynthetic activity was entirely due to non-stomatal factors (Fig. S2). This goes along with a significant shift from l_s and l_b towards l_m under O₃, underlining the relevance of g_m but also suggesting a substantially different role of g_sc and g_m in regulating A_net under stress and non-stress conditions (Figure 2) (Harmens et al., 2017; Shang et al., 2020). Finally, this also suggests that the often-cited co-regulation between g_s and g_m can be altered (Flexas et al., 2008; Ma et al., 2021), especially under stress.

4.2 Chronic O₃ exposure affects anatomical characteristics rather than leaf integrative traits

O₃ had a different impact on the main integrative and anatomical leaf traits (Tables 2 and 3, Figure 4). O₃-induced structural modifications have been shown to be species-, treatment- and leaf age-dependent (Ferdinand et al., 2000; Gao et al., 2016; Oksanen et al., 2001; PÄÄkkÖNEN et al., 1996; Xu et al., 2023). Yet, in our study, none of the leaf integrative traits except for the ratio of palisade to spongy tissue (T_pal/T_spo) and D_leaf were affected by chronically elevated O_3. Moreover, only T_pal/T_spo was correlated to g_m (Figure 4). A similar response has been observed in field-grown Populus tremuloides and Betula papyrifera (Oksanen et al., 2004). It was suggested that O₃ particularly targets the palisade tissue due to the higher surface to volume ratio in comparison to spongy tissue. Thus, O₃ damage might be earlier and stronger visible in the palisade tissue but eventually also in the spongy tissue as the O₃ exposure continues (Giacomo et al., 2010). A higher D_leaf can be due to an increased proportion of cell wall material of non-photosynthetic tissue, the proportion of which can increase under O₃ (Günthardt-Georg et al., 1997; Niinemets, 1999). At the same time, contrasting responses were observed during free-air O₃ exposure experiments with an O₃-induced increase of D_leaf and T_leaf remaining unaffected in Fagus crenata (Hoshika et al., 2020) while both traits decreased in Populus deltoids and Populus euamericana (Xu et al., 2023). Indeed, in our study T_leaf was not affected, while D_leaf increased, albeit only in SL. This emphasises a considerable species and clone-dependence of O₃ effects on leaf structure (Pääkkönen et al., 1995). As for leaf integrative traits in general, the observed modifications of the key leaf structural trait LMA are also strongly treatment- and species-specific (Hoshika et al., 2020; PÄÄKKÖNEN et al., 1998; Poorter et al., 2009). Here we found LMA to be unaffected in both clones, which suggests that only a longer or/and more severe exposure to O₃ or drought would have caused measurable modifications and thus an impact on LMA leading to a decrease of g_m as previously observed (Hoshika et al., 2020; Joffe et al., 2022; Tosens et al. 2012a, 2012b; Xu et al., 2023). Moreover, this might also explain the lack of any robust relationship between integrative leaf traits and g_m (Figure 4), which could have caused an O₃-induced decrease of g_m as observed by Hoshika et al. (2020). Therefore, in our experimental setting, leaf integrative traits LMA, D_leaf and T_leaf failed to explain g_m response to chronic O₃ (see Supplementary Note 1 for the discussion of the anatomical and cellular traits).

4.3 Discrepancy between observed and anatomy-based g_m estimates at higher end of the range

An anatomical model (Equation 10) was used to quantify the anatomical CO₂ diffusion pathway, aiming to determine the relative importance of the main anatomical traits limiting g_m. Across a broad range of species the model has demonstrated strong predictive power within a range of 0.030–0.2 mol m⁻² s⁻¹ bar (Tomás et al., 2013; Tosens & Laanisto, 2018). However, it tends to underestimate g_m values above the upper limit of this range (>0.2 mol m⁻² s⁻¹ bar). Such underestimations have been observed previously, e.g., when comparing multiple species (Peguero-Pina et al. 2012; Veromann Jürgenson, 2020; Xiong, 2023), across clones (Lei et al., 2022; Tomás et al., 2014), or when applying different treatments within a single species (Carriquí et al., 2019). Here we found g_{m_Anat} to yield a good fit within the treatment group (Figure 6a), when considering cell wall porosity as a function of T_cw as suggested by Tosens et al. (2012a, 2012b), while control replicates did not respond anymore to porosity values being higher than 0.35. Regarding membrane permeability, a 1:1 relationship was only achieved when assuming unrealistic high values of g_pl and g_en as well as r_f,i as also previously reported (Han et al., 2018; Tomás et al., 2013, 2014). Therefore, empirically determined values for membrane permeability seem to be indispensable to improve the anatomical model of g_m, as also recently been pointed out (Evans, 2021). In addition, the anatomical model assumes passive diffusion through membranes, probably regulated by biochemical components such as aquaporins (or cooporins) (Evans et al., 2009). To date, conclusions regarding their role in regulating g_m are still inconsistent (Cousins et al., 2020). Indeed, recent studies could not find a clear evidence for the often-referred aquaporin-g_m relationship in this context (Flexas et al., 2018), but did also point out a still existing lack of understanding of the physiological role of aquaporins in CO₂ transport facilitation (Clarke et al., 2022; Kromdijk et al., 2020). A final important point is the assumption of the anatomical model that Rubisco is evenly distributed in the chloroplast stroma, which seems rather unlikely. It is more likely that most of the Rubisco is located near the chloroplast envelope and at the side exposed to the cell wall. This would in turn increase g_{m_anat} and thus yield a better agreement with measured g_m, especially in the control group (Figure 6a).

4.4 An increased subcellular diffusion limitation due to altered chloroplast positioning and thicker cell walls lowers g_m under elevated O₃

S_c/S has been identified as another major determinant of g_m (Evans, 1994; Lu et al., 2016; Tosens et al., 2012b, 2015), although it can be occasionally unrelated to g_m (Hanba et al., 2004). S_c/S was not affected under O₃ (Table 3), which is in accordance with recent studies looking at the response to O₃ in Populus (Xu et al., 2023) and to drought in Helianthus annuus (Roig-Oliver et al., 2020). This is in contrast with previous studies in which different Populus cultivars were exposed to salt, light and water stress resulting in a decrease of S_c/S (Liu et al., 2021; Tosens et al., 2012b). We further observed a decrease in chloroplast length (L_chl) and total chloroplast surface length facing the intercellular airspace (L_c) which then lead to a decrease in S_c/S_m yet not in S_c/S (Figure 3 and Table 3). However, the observed decrease in S_c/S_m, while the number of chloroplasts per cell remained constant (Table 3), suggests that the distance between chloroplasts also increased, as confirmed by a greater ΔL_cyt2 under O₃ (Table 3). Hence, the consideration of the fraction of total exposed mesophyll area covered with chloroplasts or S_c/S_m in addition to S_c/S, might serve as a more comprehensive explanatory variable of g_m. Also, the increase in T_cyt indicates that O₃ not only reduced chloroplast exposition but also increased the distance to cell wall. In fact, low g_m was strongly related to higher T_cyt and lower S_c/S_m (Figure 5b,c). It suggests not only that O₃-induced lateral chloroplast shrinkage had a major negative impact on the cytosolic CO₂ pathway (Figures 3 and 5b,c) but also, a genotype-dependent chloroplast migration towards the cell centre as a possible O₃ protecting mechanism (Lei et al., 2022; Nauš et al., 2010). This is insofar important, as Xu et al. (2023) observed no change in S_c/S either but a decreased ΔL_cyt2 being also accompanied by a reduced number of chloroplasts after long-term O₃ exposure. Collectively, it indicates that observed g_m decrease is due to a change in chloroplast positioning, size and/or numbers (see Supplementary Note S2 for the discussion of the quantitative limitation analysis of gas and liquid-phase g_m components).

4.5 g_m—Anatomy relationship is clone-dependent in poplar

Despite having more favourable leaf integrative (thinner tissues) and anatomical traits (shorter pathway lengths) which might actually have suggested a higher g_m (Figure 8) (Nadal et al., 2018; Niinemets, 1999), g_m values of SL were not different to those of RB, regardless the estimation method used (Figure 1c; Figure S3). These results contrast to those of Xu et al. (2023), who found that the clone with shorter pathways also yielded a higher g_m, though this effect vanished during the applied long-term O₃ treatment. Moreover, a relationship between leaf anatomy and g_m was nearly absent in SL, while except for S_mes/S, all main variables correlated with g_m in RB (Figure 4). A similar (apparent) decoupling of leaf anatomical traits and g_m was observed between several Vitis vinifera cultivars under different levels of drought stress (Tomás et al., 2014) and wild and domesticated cotton (Lei et al., 2022). The latter also reported that thicker cell walls in the wild type did not alter g_m as there might have been possible offsetting effects between the also highly limiting but at the same time lower T_cw and T_cyt in the domesticated clones. In contrast, both T_cyt and T_cw increased in both of our clones but as a response to stress rather than due to breeding, with T_cw being higher in RB under O₃ (Table 3). Yet, it was observed that chloroplasts can move to optimise or reduce exposure to low or high light (Jeong et al., 2002; Loreto et al., 2009; Tholen et al., 2008). Hence, the fact that the above described higher ΔL_cyt2 as a function of a lower S_c/S_m (data not shown) was only significant in RB, suggests partly preserved chloroplast integrity and functioning in this clone, offsetting the otherwise less favourable anatomical traits with regard to g_m compared to SL. A recent O₃ sensitivity study in poplar concluded that a more comprehensive analysis of several variables is necessary when assessing stress response in closely related clones, as the observed differences of leaf traits are not as large as within interspecies comparisons (Shang et al., 2020). In this regard, recent literature has identified cell wall composition as a key variable, which can affect both porosity and tortuosity and thus g_m (Flexas et al., 2021). Therefore, different properties and stress-induced changes of cell wall components, as recently found among different tomato clones (Roig-Oliver et al., 2022), could also be involved in the different g_m-anatomy relationships between clones.