We selected five families of vascular plants, which consisted of one terrestrial species and its palustrine counterpart: (1) Acanthus mollis L. and Hygrophila stricta (Vahl) L. in Acanthaceae (angiosperm); (2) Arum italicum Mill. and Anubias heterophylla Engl. in Araceae (angiosperm); (3) Trachelium caeruleum L. and Lobelia cardinalis L. in Campanulaceae (angiosperm); (4) Polypodium cambricum L. and Leptochilus pteropus (Blume) Fraser-Jenk, in Polypodiaceae (fern); and (5) Pteris vittata L. and Ceratopteris thalictroides L. (Brongn) in Pteridaceae (fern) (Table 1). In the middle of autumn, terrestrial plant species were collected in the field with their underlying substrate (soil) at various coordinates in Mallorca (Spain; Table 1), and placed in plastic bags to be immediately transported to the University of Balearic Islands (Mallorca) where they were transplanted into plastic pots, using a sterile soil–peat mixture (3: 1 v/v). Then, the pots were maintained in a growth chamber under controlled conditions of 25°C, moderate light intensity of 350 μmol m^–2 s^–1 of photosynthetic photon flux density (PPFD), relative humidity above 40%, 12 h photoperiod, and watered to full soil capacity every 3–4 days. At the same time, commercial amphibious plants were distributed inside the same growth chamber as the terrestrial plants in different 34 × 45 cm water-tanks containing 20 ± 5 cm water-level, rooted in gravel/substrate for aquarium plants, and maintained under a moderate irradiance of 100 μmol m^–2 s^–1, according to the low light demand required for growing aquarium species as described in previous studies (Mommer et al., 2005; Koga et al., 2020). Four to six plants per terrestrial and palustrine species were maintained under different availability of light energy and water in each habitat. By doing this, we generated contrasting redox environments according to their different predominance in biomes with contrasting canopy openness and water availability as outlined in next subsection. All plants developed aerial leaves under growth chamber conditions until the beginning of experiments in the middle of winter. The upper-most fully expanded aerial leaves of all species were used for gas exchange, in vivo respiration, and metabolic profiling analyses.

In order to assess the abundance of both terrestrial and palustrine plant species in locations and biomes with different environmental conditions, we studied the spatial distribution of these species considering data of MAT and mean annual precipitation (MAP) from the years 1980 to 2010. Different numbers of records among species were obtained from GBIF (Global Biodiversity Information Facility¹): A. italicum (32875), P. cambricum (17980), L. cardinalis (5375), P. vittata (3906), A. mollis (2628), T. caeruleum (2559), C. thalictroides (2037), L. pteropus (196), A. heterophylla (55), and H. stricta (7). For greater accuracy, we increased the number of records of palustrine plants in Araceae and Acanthaceae, by substituting Higrophylla stricta (7) for Higrophylla ringens (1264) and Anubias heterophylla (55) for Anubias spp. Schott. (617) because of their similar distribution records (Supplementary Figure 1). Then, a random selection of records equalized the number of samples in each family and habitat; 2000 in Campanulaceae; 1500 in Pteridaceae; 1000 in Acanthaceae; 600 in Araceae, and 150 in Polypodiaceae. Finally, the spatial distribution of records randomly selected was studied with QGIS, a GIS software that combines species occurrences from GBIF with climate layers from WorldClim². QGIS rasterized species occurrences and extracted MAT and MAP data across all grid cells of the species occurrence region, at a spatial resolution of 30 arc-seconds (∼1 km). Then, species classification into biomes was performed from a Whittaker diagram of MAT and MAP (Wright et al., 2004).

Leaf gas exchange with Chla fluorescence measurements were recorded every day from 10 am to 2 pm during the last 2 weeks of the experiment with an open infrared gas-exchange analyzer system (Li-6400; Li-Cor Inc., Lincoln, NE, United States) equipped with a leaf chamber fluorometer (Li-6400-40, Li-Cor Inc.) using aerial leaves of terrestrial and amphibious plants under light-saturating photosynthetic photon flux density (PPFD) of 1000 and 400 μmol m^–2 s^–1, respectively (to avoid photodamage as a consequence of a high PPFD), with 10% blue light, a vapor pressure deficit (VPD) of 1.35 ± 0.32 kPa, a CO₂ concentration (C_a) of 400 μmol CO₂ mol^–1, and 25°C air temperature. Net photosynthesis (A_N) and stomatal conductance (g_s) were determined after a steady state was reached (after c. 20 min). Once the gas exchange stabilized, five readings were taken in four to six plants per species, and averaged to be considered as the mean of the measured plant. Intrinsic WUEi was calculated as the ratio between A_N and g_s. After a minimum 30 min under dark conditions, leaf dark respiration (R_dark) was measured in three to five plants per species with at least five readings per plant, and estimations of leaf carbon balance were obtained from the ratio of R_dark to A_N.

The quantum efficiency of the photosystem II (PSII)-driven electron transport was determined using the equation ΦPSII = (F_m′ - F_s)/F_m′, where F_s is the steady-state fluorescence in the light (PPFD = 1000 and 400 μmol quanta m^–2 s^–1 for terrestrial and palustrine plants, respectively) and F_m′ is the maximum fluorescence obtained with a light-saturating pulse (8000 μmol quanta m^–2 s^–1). The electron transport rate (ETR) was calculated as ETR = ΦPSII × PPFD × αβ, where α is the leaf absorptance, assumed to be 0.84, and β is the distribution of absorbed energy between the two photosystems, assumed to be 0.5 (Gallé and Flexas, 2010). At least five readings in two to four plants per species were taken and averaged to be considered as ETR values of the measured plant. The average ETR value for each species was used for estimations of the ratio of ETR to A_N.

For respiratory measurements, the aerial leaves of terrestrial and palustrine plants were harvested and cut into pieces after 30 min in darkness to be placed in a 3 ml stainless-steel closed cuvette maintained at a constant temperature of 25°C. Air samples were sequentially removed from the cuvette and fed into the mass spectrometer (Delta XPlus; Thermo LCC, Bremen, Germany). Changes in the ¹⁸O/¹⁶O ratios and O₂ concentration were obtained to calculate the oxygen-isotope fractionation and the electron partitioning to the AOP (τ_a), allowing calculations of the in vivo activities of AOP and cytochrome oxidase pathway (COP) as described in Del-Saz et al. (2017a). Both end point fractionation values of the AOP (Δ_a) and the capacity of the alternative pathway (V_alt) were determined in leaves of terrestrial and palustrine plants treated with a solution of 10 mM potassium cyanide (KCN) for 30 min. For land plants, Δ_a values (n = 3) of 29.9 ± 0.2‰, 30.0 ± 0.2‰, 30.2 ± 0.5‰, 30.6 ± 0.2‰ and 30.3 ± 0.4‰ were obtained for P. cambricum, P. vittata, A. italicum, A. mollis, and T. caeruleum, respectively. For palustrine plants, Δ_a values of 32.5 ± 0.3‰, 30.8 ± 0.3‰, 31.2 ± 0.8‰, 31.4 ± 0.1‰, and 29.6 ± 0.2‰ were obtained for A. heterophylla, C. thalictroides, H. stricta, L. cardinalis, and L. pteropus, respectively. On the other hand, an assumed value of 20.0‰ for the end point fractionation values of the COP (Δ_c) was used for the electron partitioning calculations as this has been shown to be fairly constant in most of the leaves and species examined (Ribas-Carbó et al., 2005). Total mitochondrial ATP production (ATP_total) together with ATP production via COP (ATP_cop) and AOP (ATP_aop) were modeled from the activities of the COP and AOP of each measurement, assuming that electron flow through the AOP drives the synthesis of 11 ATP for each 6 O₂ consumed whilst 29 ATP are formed for each 6 O₂ consumed via COP (Del-Saz et al., 2017b). Values presented are the mean of six to eight measurements performed in four to six plants per species that were performed from 9 am to 6 pm on the same days as gas exchange measurements were performed during the last 2 weeks of the experiment. In addition, the engagement of AOP (ρ) was calculated as a percentage of the ratio of the in vivo activity of AOP (v_alt) to V_alt.

Terrestrial leaves of palustrine and terrestrial plants were simultaneously sampled after 30 min in darkness on the last day of the experimental period, immediately frozen in liquid nitrogen, and stored at –80°C until further analysis. Metabolite extractions, derivatization and gas chromatography time of flight-mass spectrometry (GC-TOF-MS) analyses were carried out as previously described (Lisec et al., 2006). The GC-TOF-MS system was composed of a CTC CombiPAL autosampler, an Agilent 6890N gas chromatograph, and a LECO Pegasus III time-of-flight mass spectrometer running in EI + mode. Metabolites were identified by comparison with database entries of standards (Kopka et al., 2005; Schauer et al., 2005). The data of each terrestrial species were normalized to the mean of its respective palustrine counterpart (i.e., the value of all metabolites for each palustrine species was set to 1). The data represent averages of three to six measurements corresponding to material harvested from three to six individual plants per species.

Data of A_N, WUEi, total respiration (V_t), in vivo activity of COP (v_cyt), ATP_cop, and ATP_total, were log-transformed to meet homoscedasticity. A two-way analysis of variance (p < 0.05) was performed with habitat level (terrestrial, palustrine) and plant family (Acanthaceae, Araceae, Campanulaceae, Polypodiaceae, and Pteridaceae) as fixed factors (Table 2), and Tukey’s post hoc test (p < 0.05) was used to determine differences in each respiratory and photosynthetic parameter between species (Figures 2, 3, Tables 3, 4, and Supplementary Tables 2, 3). Student’s t-tests were used for statistical analyses in Table 5 in order to compare data from terrestrial species with data from the respective palustrine counterpart in each family. To generate individual fold change data from the physiological parameters, we normalized each measurement of the terrestrial counterpart to the mean of the respective palustrine species, as for the GC-MS metabolite analyses, and Pearson coefficients were obtained with JMP^®, Version 12.1.0 (SAS Institute Inc., Cary, NC, United States, 1989–2007; Table 6). Associations between the respiratory parameters and the metabolite profile were explored by applying the Partial Least Square (PLS) sparse regression as defined previously (Saccenti et al., 2014). Missing data in the metabolome dataset were imputed by employing a random forest imputation method before PLS analysis (Gromski et al., 2014). The “pls” package in R software was used to develop the PLS regression analysis. Also, this package includes a function to implement the variable importance for the projection (VIP) for single-response orthogonal score plsr models (Wehrens and Mevik, 2007).

A species classification into biomes was obtained from a Whittaker diagram of MAT and MAP (Figure 1 and Supplementary Table 1; Wright et al., 2004). We observed species records in all biomes, especially in shrubland, temperate forest, tropical seasonal forest, woodland, and desert (25.6, 24.0, 22.2, 12.6, and 9.89% total records). A low register was found in tropical rainforest, grassland, temperate rainforest, boreal forest, and tundra (4.06, 1.11, 0.48, 0.03, and 0.02% total records). In general, palustrine species were more abundant than terrestrial species in biomes with values of MAP ≥ 1000 mm, such as temperate forest (33.0% palustrine vs. 15.1% terrestrial), tropical seasonal forest (32.3% palustrine vs. 12.2% terrestrial), and tropical rainforest (6.40% palustrine vs. 1.72% terrestrial). In biomes with values of MAP ≤ 1000 mm, palustrine species were more abundant only in woodland (19.3% palustrine vs. 5.96% terrestrial), whilst terrestrial species were more abundant than palustrine species in arid biomes such as shrubland (46.4% terrestrial vs. 4.72% palustrine) and desert (17.6% terrestrial vs. 2.16% palustrine). Specific abundances in each type of biome can be found in Supplementary Table 1.

Regarding net photosynthesis (A_N), comparisons between groups showed no differences between angiosperms (Acanthaceae, Araceae, Campanulaceae) and ferns (Polypodiaceae, Pteridaceae) in terrestrial habitats; however among palustrine species, A_N was significantly lower in the two ferns species compared to the angiosperm L. cardinalis (Campanulaceae; Figure 2A). When comparing between counterparts in each family, A_N was significantly higher (by 2.5-fold) in terrestrial species of Acanthaceae, Araceae, Polypodiaceae, and Pteridaceae. Regarding g_s among terrestrial species, this parameter was significantly lower in the fern P. cambricum (Polypodiaceae) only when compared with the angiosperm T. caeruleum (Campanulaceae). Contrary to what was observed for A_N, no differences were found in g_s when comparing between counterparts in each family (Figure 2B).

With regard to WUEi, no major differences were observed between ferns and angiosperms in terrestrial habitats; whilst among palustrine species, the two ferns species showed a significantly lower WUEi when compared to the angiosperm A. heterophylla (Araceae; Figure 2C). Very similar to the trends observed for A_N, WUEi was significantly higher (by 3.7-fold) in terrestrial counterparts of Acanthaceae, Polypodiaceae, and Pteridaceae, with the terrestrial fern P. cambricum (Polypodiaceae) showing the highest values of WUEi, and both the palustrine angiosperm H. stricta (Acanthaceae) and fern L. pteropus (Polypodiaceae) displaying the lowest values of WUEi (Figure 2C). On the other hand, palustrine plants showed higher averaged values of ETR/A_N (9.25) and R_dark/A_N (0.173) than terrestrial plants (ETR/A_N = 8.10, R_dark/A_N = 0.087) mainly because their small A_N, and secondary, because the lack of major variations in R_dark and ETR (Tables 2, 3 and Supplementary Table 2).

A high heterogeneity was found in V_t, v_cyt, and v_alt among all species. Considering that most of V_t takes place via COX activity, a similar heterogeneity was found in v_cyt and V_t, with both varying significantly by 3.3 and 2.7-fold, across species in the terrestrial and palustrine environments, respectively. Both v_alt and τ_a showed less variability than v_cyt and V_t across terrestrial species (2.0 and 1.6-fold, respectively). In palustrine environments, higher variability was found in v_alt, differing significantly 5.4-fold across species, whilst τ_a showed similar variability to v_cyt and V_t (2.6-fold). When comparing between counterparts in each family, V_t was significantly higher in terrestrial counterparts of Araceae (by 1.7-fold), and in palustrine counterparts from both fern families, Polypodiaceae and Pteridaceae (by 1.6-fold and 2.4-fold respectively; Table 2), differing slightly from v_cyt, which was no different in terrestrial counterparts of Araceae (Table 4). A different pattern was observed for v_alt, which was significantly higher in the terrestrial counterpart of Araceae (4.0-fold) and in the palustrine counterpart of Pteridaceae (2.0-fold). A similar behavior was observed for ATP production modeled from v_cyt and v_alt (Supplementary Table 3). Regarding τ_a, the terrestrial counterparts of Acanthaceae, Araceae, and Polypodiaceae showed significantly higher values than their palustrine counterparts, 1.4, 2.3, and 1.4-fold, respectively. It is worth mentioning that in Polypodiaceae, the two ferns showed the highest values of τ_a in each habitat (Figure 3). On the other hand, leaves of H. stricta showed the highest engagement of AOP (ρ) (57%) mainly because the low V_alt, followed by leaves of plants in Polypodiaceae and Pteridaceae (25.5%) that showed variability in V_alt and v_alt, and by leaves of plants in Campanulaceae and of terrestrial plants in Araceae and Acanthaceae (14%) that displayed large V_alt. The palustrine A. heterophylla showed the lowest ρ (9%) because the low v_alt (Tables 3, 4 and Supplementary Table 2).

In order to better understand the changes in photosynthetic parameters driving the species-specific response of the respiratory parameters, fold changes of A_N, g_s and WUEi values were correlated with fold changes of V_t, τ_a, v_cyt, and v_alt as described in the statistical analyses section. The only significant correlation (r = 0.75) can be found between A_N and τ_a. Similarly, to study whether AOP contributes significantly to ATP synthesis, fold changes of τ_a and ATP_total values were correlated with fold changes of τ_a, ATP_cop and ATP_aop. Significant correlations can be found between ATP_total and energy synthesis by each pathway (ATP_cop and ATP_aop; r = 0.98 and 0.87), and between τ_a and ATP_aop (r = 0.98; Table 6).

By using GC-MS-based metabolite profiling from the aerial leaves of palustrine and terrestrial plants, we annotated 40 metabolites (Supplementary Table 5), including sugars, amino acids, organic acids, antioxidants and secondary metabolite precursors, as well as sugar-alcohols (Table 5). Although the identification of 17 metabolites (glycine, asparagine, tryptophan, phosphoric acid, pyruvic acid, citric acid, malic acid, fumaric acid, 2-oxoglutaric acid, quinic acid caffeoyl, maltose, rhamnose, xylose, raffinose, melibiose, erythritol, and galactinol) were only partly detected (n = 2) or not detected at all (nd) in certain species, they were considered for a general interpretation of the results. Significant changes (Student’s t test, p < 0.05) in metabolite levels were observed for each metabolite, in the comparison between terrestrial and palustrine counterparts in each family, with the exception of threonine, pyruvic acid, fumaric acid, and caffeic acid.

Focusing on photosynthetic routes, we observed that Campanulaceae, the only family which showed no significant differences in A_N between palustrine and terrestrial counterparts, showed the largest number of metabolites (19), mainly sugars and organic acids, with reduced levels in the terrestrial species when compared to the palustrine counterpart (Table 5). In contrast, terrestrial species of Acanthaceae, Araceae, Polypodiaceae, and Pteridaceae, with higher values of A_N than their palustrine counterparts, showed higher levels of sugars such as sucrose, fructose or glucose (Table 5), suggesting a higher energy status. We also observed that Araceae, with significantly higher g_s in the terrestrial counterpart, was the only family also showing higher levels of metabolites such as malate and maltose, which are considered of interest due to their roles in determining stomatal movement (Fernie and Martinoia, 2009; Araújo et al., 2011; Gago et al., 2016).

Regarding respiratory routes, in Araceae, the only family showing higher V_t in the terrestrial counterpart, the lack of change and decrease in citrate and 2-oxoglutarate levels, respectively, together with increases in downstream intermediates (succinate and malate) suggests a high TCA cycle activity (Table 5). This pattern was significantly different (increased citrate levels with no changes in 2-oxoglutarate and malate) in the two terrestrial fern species that displayed lower V_t and v_cyt, when compared to their palustrine counterparts, presumably due to lower TCA cycle decarboxylation activity. In this comparison, pronounced differences in γ-aminobutyric acid (GABA) levels – which are intimately connected to TCA cycle activity – between ferns and angiosperms suggest a different role for the GABA-shunt. In addition, the large accumulation of sugars such as sucrose, glucose, and fructose in ferns (Table 4) coincided with an accumulation of antioxidant and secondary metabolism precursors such as quinic acid and dehydroascorbic acid, likely indicative of a reduction in sugar oxidation by glycolysis and the TCA cycle while also promoting the accumulation of antioxidant and secondary metabolism precursors (Table 5). Notably, in Araceae, the only family showing higher values of v_alt in the terrestrial counterpart, we observed higher levels of metabolites such as valine, isoleucine, and malate, which are considered of interest due to their positive correlation with v_alt in previous studies (Florez-Sarasa et al., 2012; Del-Saz et al., 2016).

Given the observed general tendency of several physiological parameters to correlate with several metabolites (Figures 2A, 3 and Table 5), we further investigated the observed respiratory patterns for each habitat group employing PLS statistical modeling combined with variable importance for projection (VIP) as a criterion to elucidate metabolite relevance from the generated models (Gago et al., 2016). This modeling helps to highlight putative metabolic networks that differentially drive the respiratory processes in the terrestrial as compared to the palustrine species studied. We used V_t, v_cyt, v_alt, and τ_a as response variables and, after cross-validation (CV) of the generated models by the PLS, only models for τ_a can be considered robust due to the display of a R² higher than 0.6, for both terrestrial (R² = 0.62) and palustrine (R² = 0.7) habitats. For palustrine species, significant associations with phosphoric acid, proline, glucose, malic acid, glyceric acid, quinic acid, quinic acid caffeoyl, fructose, GABA, and threonine were observed (Figure 4 and Supplementary Table 4). For terrestrial species, associations with τ_a were observed for trehalose, sucrose, glucose, threonic acid and glycerol (Supplementary Table 4). Interestingly, sugar metabolism was importantly related to τ_a for both lifestyle strategies, glucose being the only metabolite significantly associated in both; despite sugar metabolism in each family differing in the other metabolite associations. Terrestrial species associated mostly with levels of trehalose and sucrose, while palustrine species were mainly associated with phosphoric acid and proline.

In order to characterize terrestrial and palustrine species under the contrasting redox conditions that broadly differentiate both habitats, we decided to maintain plants under different light intensities to fall close to an optimum for each lifestyle. This is because palustrine plants are more often covered by dense canopy trees in humid forests than terrestrial plants in semi-arid Mediterranean forests, according to spatial distribution of plant records and sample collection coordinates of terrestrial plants (Figure 1 and Table 1). Besides, in humid forest, ground layer plant species may display shade adaptations like low light saturation and light compensation points (Chazdon and Pearcy, 1991; Meng et al., 2014), which led us to photosynthetically characterize these species at different PPFD. We did not expose plants to changing light intensities because it is well known that changes in growth light intensity does not affect oxygen isotope discrimination or τ_a as observed in leaves of Arabidopsis thaliana (Florez-Sarasa et al., 2011) and of sun and shade species (Noguchi et al., 2001). However, we ensured that experimental conditions were non-stressful, and enough to allow ETR/A_N values typical of irrigated plants, positive leaf carbon balance and low AOP engagement (and enough overcapacity) in all species (Table 3).

As leaves of terrestrial plants have large energy input because in air the light level is high, the terrestrial species A. mollis, A. italicum, P. cambricum, and P. vittata showed higher A_N than their palustrine counterparts H. stricta, A. heterophylla, L. pteropus, and C. thalictroides in Acanthaceae, Araceae, Polypodiaceae, and Pteridaceae, respectively (Figure 2A). This coincided with higher levels of sugars (e.g., sucrose, fructose, and glucose; Table 5), which were considered as markers of high photosynthetic activity (Gago et al., 2016). In contrast, no differences in A_N were found between T. caeruleum and L. cardinalis in Campanulaceae, which coincides with important reductions in sugars and organic acids in T. caeruleum with respect to L. cardinalis (Table 5). Because the higher A_N, WUEi, the ratio between A_N and g_s, was found to be larger in Acanthaceae, Polypodiaceae, and Pteridaceae (Figure 2C), which could be in line with previous studies describing a differential regulation of ecosystem (WUE) among biomes. In arid ecosystems, WUE is primarily controlled by evaporation; whilst in sub-humid regions, WUE is mostly regulated by assimilation (Yang R. et al., 2016), which could be partly due to a different predominance of palustrine and terrestrial records displaying contrasting values of WUEi (Figure 1 and Figure 2C) agreeing with the idea of water losses acting as a driving force for the evolution in land plants of gas exchange regulation system (Raven, 2002; Berry et al., 2010; Assouline and Or, 2013).

Contrary to A_N, total respiration (V_t) was not higher in the terrestrial species of Acanthaceae, Araceae, Polypodiaceae, and Pteridaceae than in their palustrine counterparts. Differences in V_t were found among families in each habitat and between ferns and angiosperms (Table 4), similar to previous studies (Choy-Sin and Suan, 1974; Boyce and Mohamed, 1987; Davey et al., 2004; Hilman and Angert, 2016; Zhu et al., 2021). Variability was also found regarding v_alt and v_cyt (Table 2). Respiration in leaves is highly variable among species as it depends on leaf characteristics such as leaf lifespan, nitrogen content, growth forms, and differential nutritional requirements, regardless of lifestyle or biome (Grime and Hunt, 1975; Reich et al., 1998; Lusk and Reich, 2000; Millenaar et al., 2001; Wright et al., 2004; Atkin et al., 2015). Moreover, the carbon cost for leaf growth and maintenance may differ among species (Lambers et al., 2008). This is why τ_a, which represents the contribution of AOX to V_t, represents a better proxy to evaluate the importance of AOX activity for plant respiration when comparing among different plant species. In vivo AOX activity accounted for 10-36% of V_t in both palustrine and terrestrial species considered here, which is within the range of values observed under both stressful and non-stressful conditions in terrestrial species (10–50%; Del-Saz et al., 2018a), and here, it was strongly influenced by habitat (Table 2). The contribution of AOX to V_t was significantly higher in terrestrial species from Acanthaceae, Araceae, and Polypodiaceae (Figure 3). In model terrestrial plants, previous studies reported τ_a increases under abiotic stressors mainly due to reductions in v_cyt because the COX pathway is more sensitive to stressors than the AOX pathway (Del-Saz et al., 2018a), which helps to explain the different effect of habitat on both v_cyt and v_alt (Table 2). Considering the highest values of A_N and τ_a observed among terrestrial species (Figures 2A, 3) and the significant Pearson coefficient between these parameters (Table 6), the AOP is likely more important for the dissipation of excess energy in terrestrial plants than in palustrine plants, which is in line with previous studies describing higher oxygen isotope discrimination in sun leaves than in shade leaves (Noguchi et al., 2001). Moreover, this coincided with metabolic increases in the levels of several sugars and A_N (Figure 2A and Table 5). Interestingly, τ_a was variable among terrestrial and palustrine species (Figure 3), suggesting that v_alt is coupled to fundamental metabolic processes under non-stress conditions that may differ among species (Florez-Sarasa et al., 2016). Regarding the differences observed between groups, previous studies suggested that the post-translational regulation of AOXs in ferns may differ from those of angiosperms because of the presence of a SerI residue instead of a CysI residue in the majority of the AOX protein sequences analyzed, which could presumably affect v_alt (Neimanis et al., 2013).

A PLS approach through multivariate regression modeling identified significant relationships only between τ_a and several metabolites in each habitat (Figure 4 and Supplementary Table 4). In terrestrial plants, significant relationships were identified only between τ_a and metabolites related to sugar metabolism (sucrose, glucose, and trehalose). All of these carbohydrates are closely linked to glycolytic activity or sucrose synthesis that are highly dependent on leaf ATP synthesis or requirements (Lunn et al., 2006; Dimroth and von Ballmoos, 2008; Lim et al., 2020). In addition, the accumulation of these sugars likely confers osmotolerance and redox homeostasis in both ecosystems (Robe and Griffiths, 2000). Sucrose is a metabolic precursor of trehalose, via trehalose-6-phosphate, which acts as a signal for high carbon availability in the form of sucrose (Schluepmann et al., 2004; Lunn et al., 2006; Paul et al., 2010; Fichtner and Lunn, 2021), which is in line with the high rates of A_N observed in terrestrial plants (Figure 2A). Trehalose is hydrolyzed by trehalase into glucose, and together with fructose (a product of the reactions catalyzed by both invertase and sucrose synthase) are metabolic precursors of ascorbic acid (AA), one of the most abundant antioxidants in plants (Smirnoff and Wheeler, 2000; Hossain et al., 2017). AA can be metabolized to compounds like threonate (Hancock and Viola, 2005; DeBolt et al., 2006; Smirnoff, 2018) which showed a significant relationship with τ_a in terrestrial plants. Notably, previous studies under salinity conditions highlighted a relationship between the AOP and erythronic acid (Del-Saz et al., 2016), a degradation product of AA (Green and Fry, 2005), reinforcing the role of the AOP in mitochondrial AA synthesis (Millar et al., 2003; Bartoli et al., 2006; Del-Saz et al., 2016). In addition, threonate is also a precursor of osmoprotectants (Guerrier et al., 2000; Jouve et al., 2004; Muscolo et al., 2015). On the other hand, τ_a in terrestrial plants also showed a significant relationship with glycerol, which is a lipid precursor, that similar to trehalose, is thought to be produced as a consequence of an enhanced CO₂ assimilation in the Calvin-Benson cycle and/or from starch degradation (Liska et al., 2004), which corresponds to the highest values of photosynthesis, foliar carbon balance and oxygen isotope discrimination observed in terrestrial plants (Figures 1A, 3 and Table 3).

Palustrine plants displayed a higher energy efficiency of respiration bearing in mind their lower τ_a, the significant Pearson coefficient between ATP_aop and ATP_total (Table 6), and the highest VIP value obtained from the relationship between τ_a and phosphate (Supplementary Table 4), perhaps indicative of a tendency to save phosphorus during oxidative phosphorylation for the benefit of ATP synthesis via COX. Besides, we identified relationships between τ_a and primary metabolites related to sugar metabolism, photorespiration, secondary metabolism, the TCA cycle and ammonium assimilation. Precisely, we found a significant relationship between τ_a and glycerate, corresponding to the described role of AOP in dissipating reducing equivalents from photorespiration (Watanabe et al., 2016; Timm and Hagemann, 2020), and suggesting a role of photorespiration in palustrine plants as previously described (Maberly and Spence, 1989). The relationships between τ_a and acyl-quinic acids (Qui, CQA; Figure 4) in palustrine plants suggest a participation of the AOP in modulating carbon supply for these chlorogenic acids, whose accumulation is associated with enhanced tolerance to oxidative stress (Tamagnone et al., 1998; Niggeweg et al., 2004), and competes with the accumulation of shikimate and derived metabolites (Marsh et al., 2009), such as phenylalanine and tryptophan. The reversible esterification of caffeoyl-CoA (whose metabolic precursor is CA) with Qui produces CQA. By the conversion of Qui to shikimate (Clifford et al., 2017), the shikimate pathway provides precursors for the synthesis of tryptophan that in turn is a metabolic precursor for the biosynthesis of auxins. In heterophyllous amphibious plants, auxin synthesis may be enhanced due to alterations in the perception of blue light in submerged leaves. This is part of a mechanism to coordinate, together with other plant hormones, phenotypic plasticity in leaf form or heterophylly (Nakayama et al., 2012, 2014, 2017; Li et al., 2019, 2021). On the other hand, the significant relationships between τ_a and malate, GABA, and proline suggest that the AOP could also be related to the carbon supply for both the TCA cycle and ammonium assimilation. Through the mitochondrial 2-OG/malate transporter, malate can facilitate GABA transport (Ramesh et al., 2018; Bown and Shelp, 2020), whose synthesis mainly occurs from glutamate by the cytosolic glutamate decarboxylase, alternatively through polyamine degradation (Yang Y. et al., 2016), or by the oxidation of proline to glutamate in mitochondria (Fait et al., 2008; Shelp et al., 2012). Moreover, both GABA and proline may act as osmoprotectants and their catabolism in mitochondria can provide reducing equivalents as substrates for the AOP (Studart-Guimarães et al., 2007; Michaeli et al., 2011; Florez-Sarasa et al., 2021), which is in agreement with the relationships identified between τ_a and these metabolites in palustrine plants (Figure 4 and Supplementary Table 4). On top of this, GABA can act as a transducer of environmental stress signals leading to the activation of genes for ethylene and abscisic acid biosynthesis (Kinnersley and Turano, 2000; Forde and Lea, 2007). Overall, the relationships between τ_a and metabolites related to hormone biosynthesis and signaling in palustrine environments could be especially relevant for heterophyllous amphibious plants. All these signaling metabolites, together with gibberellins, mediate perception and responses to fluctuations of water levels, and control the synthesis of new developing aerial leaves in the transition from a submerged to an aerial habit (Cox et al., 2004; Jackson, 2008; Chater et al., 2014; Kim et al., 2018). Whilst some evidence has suggested that plant hormones such as abscisic acid, ethylene, gibberellins, and auxins are part of signaling networks controlling AOX expression (Ivanova et al., 2014; Berkowitz et al., 2016), their control of in vivo AOX activity remains, even in model terrestrial plants, to be tested.