The experiment was carried out in a greenhouse at the Center for Nuclear Energy in Agriculture (CENA-USP) in Piracicaba, São Paulo State, Brazil, from July to November 2018. Plants were grown at temperatures between 18 and 32°C (mean of 25°C) with an average relative humidity of 65%. Three Eucalyptus species with different levels of drought tolerance (Eucalyptus saligna Sm., drought sensitive; Eucalyptus urophylla S.T. Blake, moderate tolerance; and Eucalyptus camaldulensis Dehn., drought tolerant) (Gonçalves et al., 2017) of approximately 90 days old and 30 cm in height, germinated from seeds obtained from the Institute of Forest Research and Studies (IPEF, Brazil), were transplanted into individual plastic pots (5 kg) containing Oxisol soil (16% clay, 5% silt, and 79% sand), collected from the top layer at the Itatinga Experimental Station, Itatinga, São Paulo State, Brazil. The physiochemical characteristics were: pH = 4.2; organic matter = 25 g dm^–3; P = 4 mg dm^–3; K = 0.3 mmol_c dm^–3; Ca = 1 mmol_c dm^–3; Mg = 1 mmol_c dm^–3; H + Al = 25 mmol_c dm^–3; Al = 3 mmol_c dm^–3; B = 0.14 mg dm^–3; Cu = 0.6 mg dm^–3; Fe = 33 mg dm^–3; Mn = 0.8 mg dm^–3; and Zn = 0.8 mg dm^–3.

Based on the soil K critical level (<1.20 mmol_c dm^–3 of K) for Eucalyptus to respond to potassium fertilization, K was replaced by Na (as NaCl), representing substitutions of 0/100, 25/75, 50/50, 75/25, and 100/0 (percentage of Na/percentage of K) for 120 days. Thus, the treatments consisted of five combinations of Na and K application rates (0/0.90, 0.22/0.67, 0.44/0.44, 0.67/0.22, and 0.90/0 mmol_c dm^–3 of Na/millimoles of charge per cubic decimeter of K), which, when added to the soil K content, reached the soil K critical level (1.20 mmol_c dm^–3 of K). The rates 0 and 0.90 mmol_c dm^–3 of Na represented the control (solely K supplied plants) and the K deficiency treatments, respectively. We would like to emphasize that the Na rates employed in studies regarding salinity-induced stress in plants are higher than those used hereby. For instance, Zhang et al. (2016) investigated salinity-induced stress on wheat seedlings’ growth in nutrient solution by adding a NaCl rate of 150 mM, while Quais et al. (2020) characterized the mechanisms of the underlying interactions between rice plants and brown planthopper under salinity stress (100 mM salinity level). According to Madsen and Mulligan (2006), the emergence of Eucalyptus citriodora, E. camaldulensis, Eucalyptus populnea, and Acacia salicina was substantially reduced by adding 100 mM of NaCl, while the survival of established plants was reduced only at 300 and 400 mM of NaCl.

In addition to the treatments with K and Na, all samples received the following complementary fertilization: 135 mg dm^–3 of N plus 20 mg dm^–3 of N in coverage at 2 months after the onset of the treatments, 300 mg kg^–1 of P, 92 mg kg^–1 of Ca, 7.2 mg kg^–1 of Mg (reaching 7 mmol_c dm^–3 in a Ca⁺²/Mg⁺² ratio of 4:1), 45 mg kg^–1 of S, 0.82 mg kg^–1 of B, 4.0 mg kg^–1 of Zn, 3.66 mg kg^–1 of Mn, 1.55 mg kg^–1 of Fe, 1.39 mg kg^–1 Cu, and 0.20 mg kg^–1 of Mo, facilitating adequate plant development (Novais et al., 1991). The sources of the elements used were as follows: (NH₄)H₂PO₄, CaCO₃, MgCO₃, elementary S, CuSO₄.5H₂O, ZnSO₄.7H₂O, FeSO₄.7H₂O, H₃BO₃, MnSO₄.H₂O, and MoO₃.H₂O.

The plants were exposed to two water conditions starting 60 days after the onset of the treatments: well-watered (WW) and water stress (WS) conditions, simulating adequate water availability and drought conditions, respectively. The soil relative water content (SRWC) of both water conditions was controlled daily by the gravimetric method using irrigation with deionized water up to 80 and 35% of the field capacity under WW and WS, respectively. Weighing and watering were conducted on a daily basis at dusk until the pots reached their corresponding target-adapted SRWC (Equation 1) (Xu et al., 2009); every 15 days, two plants were harvested and weighed to maintain the desired SRWC in the pots.

where W_total is the current soil total weight (pot + soil + plant + water), W_pot is the weight of the empty pot, DW_soil is the dry soil weight, SFW is the fresh weight of two plants harvested every 15 days, and WFC is the soil weight at field capacity (soil + pot + water).

The experiment was performed in randomized blocks, with four replications in a 5 × 2 factorial design (five rates of K replacement by Na and two water conditions) for each Eucalyptus species, totaling 120 experimental units with one plant each.

The youngest fully expanded leaf of each plant was used to evaluate gas exchange in the morning (from 9 a.m. to 11 a.m.) using an infrared gas analyzer (IRGA, Li-6400XT, LICOR Inc., Lincoln, NE, United States) at environmental humidity and temperature. The external CO₂ concentration (C_a) was fixed at 400 μmol and the photosynthetically active radiation (PAR) flux density at 1,200 μmol m^–2 s^–1. Photosynthesis (A), stomatal conductance (g_s), and transpiration (E) were measured (Mateus et al., 2019). Mean leaf temperature during the measurement was 30°C. WUE (in grams dry matter per kilogram H₂O^–1) was calculated by dividing the total dry matter (TDM) value (belowground plus aboveground) by water consumption (WC) throughout the experiment (Martin and Thorstenson, 1988), which was obtained by calculating the daily weight reduction due to transpiration (Equation 2).

where S_evap is the soil evaporation from the mean weight loss of four plantless pots.

In the same leaves, the predawn (3 a.m.) and noon (12 p.m.) leaf water potentials (Ψw_PD and Ψw_N, respectively) were also measured using a Scholander pressure chamber (Turner, 1981). All evaluations were realized prior to harvesting.

Stomatal density (Std, stomates per square millimeter) was calculated using the two youngest fully expanded leaves per plant (Mateus et al., 2019) on abaxial and adaxial surfaces, applying the software package ImageJ¹. Complementary micrograph material of Std was obtained by scanning electron microscopy (JEOL JSM-IT300 LV, Tokyo-Japan) at 20 kV, and digital images were recorded (Lavres et al., 2019).

Plants were harvested 120 days after the onset of the treatments, and their leaves, stems, branches, and roots were separated. Leaf area (LA) was obtained by passing all leaves through a leaf area integrator (LI-3100).

After drying in a forced air ventilation oven at 60°C for 72 h, each plant part was weighed to determine dry matter. Subsequently, the plant material was ground in a Wiley-type mill and forwarded to nitric–perchloric digestion (Malavolta et al., 1997) to quantify K and Na by inductively coupled plasma optical emission spectrometry (ICP-OES; iCAP 7000 Series, Thermo Fisher Scientific, Waltham, United States). Based on the leaf K and Na concentrations, we calculated the K/Na ratio, which was correlated with the estimated rate of maximum dry matter production (critical level of 90% maximum yield) of each genotype, obtained by equaling the equation to zero. The accumulations of K and Na were obtained by multiplying the concentration of each element in the tissue by the dry matter production of the respective tissue (root, stems, and leaves) and used to determine the use efficiency (UE, in grams per milligram) (Siddiqi and Glass, 1981) according to Equation 3.

where nutrient refers to K or Na accumulation.

The same samples used for leaf dry matter determination were also used to assess the carbon isotope composition, determined using a mass spectrometer (ANCA-GSL Hydra 20-20 model, SERCON Co., Crewe, GBR) coupled to a C automatic analyzer (Barrie and Prosser, 1996), and the isotope values (in per mill) were calculated via Equation 4 (Farquhar and Sharkey, 1982).

where R is the ratio of ¹³C/¹²C. The reference material is the Vienna Pee Dee Belemnite (PDB).

Data were analyzed by the F test (p < 0.05), and significant differences among means were determined via Tukey’s post-hoc test (p < 0.05) to compare the WW and WS conditions. The significant effects of Na application were described by linear, quadratic, and square root regression models, in which the significant model (p < 0.05) with the highest determination coefficient (R²) was selected. The original data were standardized to be analyzed by principal component analysis (PCA) and cluster analysis, correlating the measured variables in each genotype and water condition. In cluster analysis, the treatments were grouped into functional units by their similarity; for the PCA, we used the treatments with Na supply for the first two main components (PC1 and PC2).

Statistical analyses were performed using the software packages SAS version 9.1 (SAS Institute Inc, 2004) and R version 3.5.1 (R Development Core Team, 2018). Data variability was indicated with standard error and shown graphically using SigmaPlot 11.0 (Systat Software Inc., San Jose, CA, United States).

The leaves of E. saligna and E. urophylla were hypostomatic, occurring mainly on the abaxial surface, with lower than 25 stomates per square millimeter. In E. camaldulensis, however, the leaves were amphistomatous, with stomates occurring on both surfaces. Adaxial stomatal density (Std_AD) was influenced by Na, WS, and Na^∗WS in E. saligna, E. urophylla, and E. camaldulensis (Figures 1A–C). Abaxial stomatal density (Std_AB) was influenced by Na and WS in E. saligna and E. urophylla, whereas in E. camaldulensis, it was affected by Na application (Figures 1D–F). E. saligna and E. camaldulensis seedlings grown under WW and WS conditions showed higher Std_AD and Std_AB levels with partial K replacement by Na. Additionally, the highest Na rate (0.9 mmol_c dm^–3 of Na) led to decreased Std_AB levels by 50, 30, and 20% in E. saligna, E. urophylla, and E. camaldulensis, respectively. Water stress also decreased the mean Std_AB by 15 and 10% in E. saligna and E. urophylla, respectively, irrespective of the Na rate.

Parameters A (assimilation rate; Figures 2A–C), E (transpiration rate; Figures 2D–F), and g_s (Figures 2G–I) were influenced by Na and WS in all species. Partial K replacement by Na (up to 0.44 mmol_c dm^–3) increased A up to 55, 50, and 20% in E. saligna, E. urophylla, and E. camaldulensis, respectively, when compared to the control (0 mmol_c dm^–3 of Na). Meanwhile, the K-deficient plants of all genotypes had lower A. Compared to the control, the intermediary rates of Na also increased E up to 200 and 50% in E. saligna and E. urophylla, respectively, under both water conditions, and 40% in E. camaldulensis under WW. Against the other genotypes, E decreased until the intermediary Na rates for E. camaldulensis under WS conditions. The K-deficient plants had higher E, except for E. urophylla under WW. The g_s increased with partial K replacement by Na up to 250% in E. saligna and to 50% in E. urophylla and E. camaldulensis under both water conditions compared to the control. K-deficient plants had significantly lower g_s in E. urophylla under WW and E. camaldulensis under WS. Considering the mean of all rates, drought increased A by 20% in E. saligna and decreased its values by 15% in E. urophylla and E. camaldulensis; it also reduced the E values by 45, 35, and 75% and the g_s values by 50, 30, and 55%, in E. saligna, E. urophylla, and E. camaldulensis, respectively, compared to those under WW.

Factors Na, WS, and Na^∗WS significantly influenced the leaf carbon isotopic compositions (δ¹³C‰) of E. saligna and E. urophylla, whereas for E. camaldulensis, it was affected by Na and WS (Figures 3A–C). Under WW, for E. saligna and E. camaldulensis, the δ¹³C‰ increased with lower Na application rates (0.22 mmol_c dm^–3 of Na) and decreased with higher Na rates. In contrast, the values of δ¹³C‰ for E. urophylla were reduced at lower rates (0.22 mmol_c dm^–3 of Na). The lowest δ¹³C‰ values were observed in K-deficient plants (0.9 mmol_c dm^–3 of Na) of E. urophylla and E. camaldulensis under both water conditions and in the intermediate Na rates (0.44 and 0.67 mmol_c dm^–3 of Na) of E. saligna under WS. Drought stress increased the δ¹³C‰ values of all species when compared to WW.

In E. saligna, both predawn (Ψw_PD) and noon (Ψw_N) leaf water potentials were affected by Na, WS, and Na^∗WS (Figures 4A,D), while in E. urophylla (Figures 4B,E) and E. camaldulensis (Figures 4C,F), these were affected by Na and WS. The lowest Ψw_PD and Ψw_N values were found at low to intermediate Na rates (0.22 and 0.44 mmol_c dm^–3) in E. saligna as well as for E. urophylla under both water conditions, except the Ψw_N of E. saligna under WW, which increased linearly with Na application. Otherwise, the Ψw_PD of E. camaldulensis increased up to 0.67 mmol_c dm^–3 in both conditions, while the Ψw_N decreased with increasing Na application rates. The Ψw_PD values of all genotypes were lower under WS than under WW, while the opposite was found in Ψw_N values since the WS conditions decreased up to 25, 10, and 55% for E. saligna, E. urophylla, and E. camaldulensis, respectively, considering the mean of all Na application rates.

The TDM of all genotypes was affected by Na, WS, and Na^∗WS (Table 1). Partial K replacement by Na increased the TDM of all genotypes under both conditions, except in E. saligna under WS, which had a higher TDM than the control (0 mmol_c dm^–3 of Na), decreasing linearly with higher Na rates. The K-deficient plants (0.9 mmol_c dm^–3 of Na) had lower TDM levels by 35% under both water conditions compared to the control. Under WW, the maximum TDM (critical level) was reached with 0.048 mmol_c dm^–3 Na and 0.852 mmol_c dm^–3 K, corresponding to 5.3% substitution and 48.9 g per plant.

In E. urophylla, a higher TDM level was found in plants with low Na rates (0.22 mmol_c dm^–3) at both water conditions, whereas K-deficient plants had decreased TDM by 50 and 30% than the control treatment in WW and WS, respectively. The estimated Na rate to give the maximum TDM was 0.06 mmol_c dm^–3 of Na and 0.84 mmol_c dm^–3 of K, which corresponded to 6.7% of substitution reaching 69.9 g per plant. Under WS, the estimated Na rate was 0.085 mmol_c dm^–3 and 0.815 mmol_c dm^–3 of K, corresponding to 9.3% substitution and 55.13 g per plant.

In contrast to the other genotypes, the higher TDM values of E. camaldulensis were observed up to an Na rate of 0.44 mmol_c dm^–3 under both water conditions, which means a substitution of K by Na around 50%. The K-deficient plants had lower TDM by 55 and 35% than the control treatment under the WW and WS conditions, respectively. The rates estimated to obtain the maximum TDM under WW was 0.27 mmol dm^–3 of Na and 0.63 mmol_c dm^–3 of K, corresponding to a substitution level of 30% and 69.81 g per plant. Under WS, the estimated rates were 0.09 mmol_c dm^–3 of Na and 0.81 mmol_c dm^–3 of K, corresponding to 10% K replacement by Na and 46.4 g per plant.

Leaf area (LA) was influenced by Na, WS, and Na^∗WS in E. saligna (Table 1). Under WW, LA decreased up to 30% with higher Na application, while under WS, an increase around 10% occurred with intermediary rates of Na (0.44 and 0.67 mmol_c dm^–3) when compared to the control (0 mmol_c dm^–3 of Na). The LA of E. urophylla and E. camaldulensis was affected by Na and WS. Low to intermediary Na rates (0.22 and 0.44 mmol_c dm^–3) increased the LA of E. urophylla by up to 60 and 25% compared to the control under WW and WS, respectively. In E. camaldulensis, these Na rates increased the LA by 6% compared to the control under WW. Otherwise, K-deficient plants (0.9 mmol_c dm^–3 of Na) significantly decreased the LA up to 40% in all genotypes compared to the control. Water stress decreased the LA by 15, 35, and 36% in E. saligna, E. urophylla, and E. camaldulensis, respectively, according to the mean of all Na rates.

Water use efficiency (WUE) was influenced by Na, WS, and Na^∗WS in E. saligna and E. camaldulensis, whereas in E. urophylla, it was only affected by Na and WS (Table 1). In E. saligna and E. urophylla under WW and in E. camaldulensis under both conditions, low K replacement by Na increased the WUE and decreased it in higher Na rates. Otherwise, plants of E. saligna and E. urophylla under WS decreased the WUE due to Na supply. Drought stress increased the WUE by 33, 35, and 17% in E. saligna, E. urophylla, and E. camaldulensis, respectively, irrespective of the Na rate. In addition, the mean WUE was higher in E. camaldulensis (drought tolerant), followed by E. urophylla (moderate tolerance) and E. saligna (drought sensitive) in both water conditions.

The Na leaf concentration [Na] was influenced by Na and WS in E. saligna (Figure 5A), E. urophylla (Figure 5B), and E. camaldulensis (Figure 5C), While the K leaf concentration [K] was affected by Na, WS, and Na^∗WS (Figure 5D) in E. saligna, E. urophylla (Figure 5E), and E. camaldulensis (Figure 5F). This affected only the Na rates and WS. Overall, K decreased and Na increased with increasing Na rates. In addition, the concentration levels were higher in plants under WW than in WS. In E. saligna under WW, the replacement of 5.3%, corresponding to 90% of TDM and reaching rates of 0.041 and 0.85 mmol_c dm^–3 of Na and K, respectively, decreased the K by 0.06 g kg^–1, while Na increased by 0.12 g kg^–1 compared to the application of only K, increasing plant growth. Furthermore, K of 2.9 g kg^–1 was still above the critical level of K, without symptoms of deficiency. E. urophylla under WW with 6.7% of K replacement by Na reached the rates 0.06 and 0.83 mmol_c dm^–3 of Na and K, respectively; K decreased by 0.13 g kg^–1, while Na increased by 0.018 g kg^–1 compared to the application of only K. Under WS, 9.3% of K replacement by Na reached the rates 0.085 and 0.815 mmol_c dm^–3 of Na and K, respectively; K decreased by 0.07 g kg^–1, while Na increased by 0.052 g kg^–1 compared to the application of only K. Conversely, E. camaldulensis under WW, with 30% of K replacement by Na, reached the rates 0.27 and 0.63 mmol_c dm^–3 Na and K, respectively; K decreased by 0.46 g kg^–1, while Na increased by 0.44 g kg^–1 compared to the application of only K.

Additionally, the leaf K/Na ratios decreased with increasing Na levels. According to the estimated rate of Na to achieve the maximum TDM of E. saligna, E. urophylla, and E. camaldulensis, the optimal leaf K/Na ratios were 1.7, 2.9, and 2.2, respectively, under WW and 2.6, 2.2, and 3.2, respectively, under WS (Table 2). UE_K was affected by Na and WS, increasing up to 30% mainly with low K replacement by Na, while UE_Na was affected by Na, WS, and Na^∗WS (Table 2), decreasing up to 70% with higher K replacement by Na. Water stress decreased the UE of both K and Na compared to WW.

In E. saligna (Figure 6A), the total variance was explained by 64% (PC1 + PC2), with PC1 being explained by Ψw_N, UE_K, [K], and δ¹³C‰, while PC2 was explained by TDM, Std_AB, and [Na]. The parameters A, WUE, LA, E, and g_s contributed with average weights to explain the data variance in PC1 and PC2. Under WW, low to intermediate Na rates (up to 0.44 mmol_c dm^–3) were characterized by higher values of TDM and Std_AB and lower values of Na and δ¹³C‰. Plants under lower Na rates and WS showed higher WUE and δ¹³C‰ values. The Na rate of 0.9 mmol_c dm^–3 resulted in higher Na levels and lower K, TDM, and Std_AB values.

In E. urophylla (Figure 6C), the total variance was explained by 66% (PC1 + PC2), with PC1 being explained by g_s and E and PC2 by WUE, δ¹³C‰, [K], TDM, and Na. The parameters Std_AB, UE_K, A, Ψw_N, and LA contributed with average weights to explain the data variance in PC1 and PC2. Under WW, low to intermediate Na rates were characterized by higher values of TDM, Std_AB, UE_K, A, LA, g_s, and E. Higher Na rates were identified by higher Na and lower TDM, [K], WUE, and δ¹³C‰ levels. Under WS, Na rates of 0 and 0.22 mmol_c dm^–3 resulted in higher WUE and δ¹³C‰ and lower LA, E, and g_s levels.

In E. camaldulensis (Figure 6E), the total variance was explained by 77% (PC1 + PC2), with PC1 being explained by TDM, g_s, LA, Ψw_N, and [Na] and PC2 by WUE, δ¹³C‰, and E. The parameters [K], Std_AB, A, and UE_K contributed with average weights to explain the data variance in PC1 and PC2. Under WW, low to intermediate Na rates were characterized by higher values of TDM, g_s, LA, Ψw_N, and EU_K and by lower values of [Na]. Moreover, the Na rate of 0.9 mmol_c dm^–3 resulted in higher E and [Na] and lower δ¹³C‰, WUE, and [K] values. Plants under WS with low to intermediate Na supply were characterized by higher values of [K], WUE, and δ¹³C‰ and by lower E values.

Cluster analysis showed the formation of three main groups among the treatments in all genotypes. In E. saligna (Figure 6B), group 1 was represented by Na rates of 0, 0.22, and 0.44 mmol_c dm^–3 under WW, which represented the control and low to intermediate rates; group 2 by the rate of 0.67 mmol_c dm^–3 under WW and the rates of 0, 0.22, 0.44, and 0.67 mmol_c dm^–3 under WS; and group 3 contained K-deficient plants (0.9 mmol_c dm^–3 of Na) under both conditions. In E. urophylla (Figure 6D), group 1 comprised plants receiving no Na under WW and 0.22, 0.44, and 0.9 mmol_c dm^–3 of Na under WS, while group 2 contained plants receiving 0.67 and 0.9 mmol_c dm^–3 of Na under WS. In group 3, the plants received 0.22, 0.44, and 0.67 mmol_c dm^–3 of Na under WW and no Na under WS. In E. camaldulensis (Figure 6F), group 1 was composed of plants receiving 0, 0.22, 0.44, and 0.67 mmol_c dm^–3 of Na under WW, while in group 2, plants received 0, 0.22, 0.44, and 0.67 mmol_c dm^–3 of Na under WS. In group 3, plants received only 0.9 mmol_c dm^–3 of Na under both water regimes.