Introduction 

The tropical rainforest of the Amazon basin covers about 5.1 × 10⁶ km² (out of 6.915 × 10⁶ km² of the total basin area, which extends across several countries) and stores about 86 Pg of carbon in above and below ground biomass ([38], [21]), i.e., 21.8% of total carbon (393.4 Pg) stored in above and below ground biomass of land vegetation ([35]). The Amazon rainforest is also important for recycling through transpiration about 50% of total precipitation of the Amazon region ([39]) and for its outstanding biodiversity. It has been anticipated that atmospheric CO₂ concentration can reach 800 ppm by the end of the 21^st century ([4]) and that changes in rainfall distributions can lead to an increase in frequency and severity of droughts, particularly in Eastern Amazon ([15]). Although a mild dry season has little effect on growth rates of Amazonian trees in the central Amazon ([14], [9]), it has been observed a decline in the amount of carbon stored in above ground biomass per unit land area and an increase in tree mortality under severe drought ([21]).

The most common response observed in free-air CO₂-enrichment is an increase in photosynthetic rates ([2], [43]) and in several vascular plants there is also an improvement in net primary productivity ([34]), but the effect of long-term exposure to elevated CO₂ concentration (eCO₂) on tropical trees still remains to be elucidated. Stomatal conductance (g_s) often declines at eCO₂ ([2]), which contributes to improve water-use efficiency (WUE) and biomass gain of plants under CO₂-enrichment ([29], [30]). Plant WUE (WUE_P), the ratio between total biomass gain and total evaporation of water from the plant ([27]) is a parameter often used to assess plant response to water stress. At the leaf level WUE describes the ratio between net photosynthetic rate (A) and leaf transpiration (E), and both photosynthesis and transpiration are often closely related to stomatal conductance ([32], [27]). Besides g_s, other factors such as biomass allocation to plant organs, whole-plant respiration and variation in boundary layer conductance affect WUE_P, and hence it is not uncommon to find discrepancy between WUE_P and A/E trends ([27]).

In the Brazilian Legal Amazon (5,016,136 km² - [21]) large areas has been deforested over the past decades and hence there are extensive areas available for rehabilitation. It has been estimated that about 120,000 km² could be restored or reforested by 2030 ([6]). Several species, including andiroba (Carapa, a genus that includes 11 species in the Neotropics - [24]) have been tested to evaluate their performance in reforestation projects or forestry systems ([7], [41]). Andiroba is a promising species because it produces wood of good quality and a multipurpose oil (carapa oil) can be extracted from their seeds ([24]). This is particularly important in (agro)forestry systems, as non-timber forest products can generate additional income while the tree is not large enough for timber harvesting ([26]). In the Neotropics, the Carapa tree (andiroba in Brazil) occurs in lowland forests from Central America and the Caribbean to South America ([24]). At sapling stage it can grow about 1-2 cm yr^-1 in diameter and 1-1.5 m yr^-1 in height ([16], [8]). Canopy trees of this species can reach up to 2.0 m in diameter and 60 m in height ([20]). Carapa is more tolerant to Hypsipyla grandella than mahogany - Swietenia macrophylla ([41]), thus it can be used in plantations to replace the highly prized mahogany, which highlights the importance of andiroba in forest management project. The aim of this study was to assess the effect of eCO₂ and drought on photosynthesis, water-use efficiency, and biomass gain in saplings of Carapa surinamensis. In this experiment we hypothesized that subjecting the plants to eCO₂ increases biomass gain and photosynthetic rates, and decreases stomatal conductance, which leads to an increase in water-use efficiency.

  Materials and methods 

Experimental setup and plant material

The experiment was conducted at the National Institute for Research in the Amazon - INPA (03° 05′ 29″ S, 59° 59′ 35″ W), Manaus, Brazil under greenhouse and growth chamber conditions. Seeds of andiroba (Carapa surinamensis Miq, Meliaceae) were germinated in vermiculite and 15 days after emergence, the seedlings were transplanted to large pots (20 cm diameter and 26 cm deep, about 8 L capacity) containing 7 kg of substrate, a mixture of soil forest of the first 20 cm, amended with mulched material and 5 g kg^-1 (soil) of 10-10-10 (NPK) fertilizer. We used large pots to avoid root restriction during the experimental period. Once the plants had reached 60 cm in height (about six months after emergence) the saplings were randomly sorted into four treatments: two CO₂ levels, ambient CO₂ (400 ppm) and eCO₂ (700 ppm) and two water regimes, soil kept at 50% field capacity - FC (hereafter referred to as the drought condition) and 100% FC. The experiment was run for 163 days, from July 14 to December 23, 2015.

Before initiating the experiment and for 30 days, we measured the light intensity (photosynthetically active radiation - PAR) inside the greenhouse, which turned out to be a mean of about 200 µmol m^-2 s^-1 over a 12-h period. Furthermore, because of the latitudinal location (-3.091°), incoming irradiance remains rather constant over the year. Then we set up the growth chamber to emulate the mean light intensity in the greenhouse over a 12-h photoperiod.

Plants subjected to the ambient CO₂ treatment were kept under greenhouse conditions, whereas those subjected to eCO₂ were kept in a growth chamber (TPC-19^®, Biochambers Inc., Winnipeg, Canada; working area of 1.72 m² and 1.52 m height). In the growth chamber CO₂ was kept constant (700 ppm); we used a 12-h photoperiod with PAR of 200 μmol m^-2 s^-1, the light was turned off at night; day/night temperature and relative humidity were 27/25 °C and 80-90%, respectively. We also monitored the light and temperature conditions inside the greenhouse with specific sensors (Li-190 SA^®, Li-Cor, Lincoln, USA and Humitter 50Y^®, Vaisala, Oyj, Finland) connected to a datalogger (Li-1400^®, Li-Cor, Lincoln, USA), which was set to record data at 15-min intervals. On randomly selected days we monitored the CO₂ concentration in the greenhouse with an infrared gas analyzer (LI-6400^®, Lincoln, USA) with empty chamber.

The water volume the soil could hold was recorded and used to calculate soil water content (SWC) at 50% and 100% FC. At these water contents, soil moisture percentages, measured with a probe (MPM160B^®, ICT International, Armidale, Australia), were 21% (v/v, soil at 50% FC) and 31% (v/v, soil at 100% FC). For further information we also measured the predawn leaf water potential of the plants with a Sholander pressure chamber (1505D^®, PMS Instrument Company, Albany, USA). Once a week and during the experimental period the plants in the greenhouse and growth chamber were randomly moved to avoid position effects. At the end of the experimental period we measured gas exchange (data were collected just once), and determined dry matter (DM) of stems (W_S), roots (W_R) and leaves (W_L), total plant DM (W_T), shoot/ root ratio (SRR), leaf area (A_L), specific leaf area (SLA, leaf area to leaf mass ratio), consumptive use of water (CUW - total amount of water consumed by a plant), whole-plant water-use efficiency (WUE_p), and leaf total nonstructural carbohydrates (TNC).

Gas exchange parameters

At the end of the experimental period, gas exchange parameters were measured with a portable gas exchange system (LI-6400XT^®, Li-Cor, Lincoln, USA). The measurements were carried out between 08:00 and 14:00 in two fully expanded leaves per plant, after a stabilization period of about 10 min at ambient and eCO₂ and the light intensity used for measurements. We measured stomatal conductance to CO₂ (g_sCO2), light saturated photosynthesis (A_sat), C_i/C_a (intercellular CO₂ to ambient CO₂ ratio), carboxylation efficiency (CE, the initial slope of the A/C_i regression line - [17]), leaf respiration in the light (R_L), and dark respiration (R_D). A_sat and g_sCO2 were measured at 1000 μmol m^-2 s^-1, ambient temperature (27 °C), relative humidity of 70 ± 5%, and CO₂ concentration of 400 ppm (ambient CO₂ treatment) and 700 ppm (eCO₂ treatment). We also determined light and CO₂-saturated photosynthesis (A_max) at a [CO₂] of 2000 ppm (C_a), as previously described ([31]). To obtain A_sat-mass, A_max-mass, CE_mass and g_sCO2-mass, photosynthetic rates, CE and g_sCO2 per unit area were converted to a mass basis by multiplying them by SLA. For CE, we used the conventional unit (i.e., µbar CO₂ instead of µmol CO₂); thus CE_mass was expressed in mmol g^-1 s^-1 bar^-1. R_L was determined at low light intensity (10-75 μmol m^-2 s^-1) as the intercept of the A/PAR regression line, whereas R_D was determined as the absolute value of A in the dark (PAR = 0) and at the same ambient conditions as those used for measuring A_sat-mass; both R_L and R_D were expressed on a mass basis (i.e., R_L-mass and R_D-mass).

Biomass allocation

At the end of the experimental period (163 days = t₁₆₃), we determined the dry matter of leaves, stems and roots by oven-drying the plant material at 72 °C to constant mass. We also determined the specific leaf area (SLA). Leaf area (A_L) was measured with an area meter (LI-3000^®, Li-Cor, Lincoln, USA). At the beginning (t₀) and at the end of the experiment (t₁₆₃), we measured plant height, H (from ground level to the apical bud). We also measured plant diameter (D) at 6.0 cm from the ground with digital calipers.

The biomass gain (ΔB) of plant organs (ΔW_L, ΔW_S and ΔW_R) and leaf area gain (ΔA_L) during the experimental period was calculated as the difference between the biomass of plant organs recorded at the end of the experimental period (i.e., at t = 163 by harvesting the plants) and that estimated at the beginning of the experiment (initial biomass at t₀). To estimate the biomass at t₀, additional plants (n = 30 plants) were produced and harvested to generate allometric equations and estimate: (i) the total biomass, W_T (g) = 10.157·exp 0.1692D (r² = 0.95), where D (in mm) denotes the plant diameter at 6.0 cm from the base of the plant; (ii) the stem biomass, W_S (g) = 2.4298·exp 0.1909D (r²= 0.94); and (iii) root biomass, W_R (g) = 1.3277·exp 0.1988D (r² = 0.88). Leaf biomass (W_L) was calculated as the aggregate biomass of individual leaflets: W_LL (g) = -0.2052 + 0.04L_L (r² = 0.85), where L_L (in cm) stands for leaflet length, n = 100 leaflets from 30 plants. Leaf area (A_L) at the beginning of the experiment was calculated as the aggregate area of individual leaflets: A_LL (cm²) = -47.117 + 8.5452L_L (r²= 0.89). These equations allowed us to estimate A_L and the biomass gain (ΔB) of plant organs throughout the experimental period (163 days).

The consumptive use of water (CUW)

The CUW during the experimental period was obtained by daily recording the amount of water consumed by the plant. During the whole experimental period (163 days) each potted plant was weighed (at 07:00-08:00, accuracy of 1 g) and the amount of water consumed recorded and restored to keep the soil at its target water content (50% FC or 100% FC). For rewatering, we did not take into account the mass change due to daily plant growth because in comparison with the amount of water lost by total evaporation from the plant, the biomass gain was very low (< 1%). To avoid evaporation from the soil surface the pot was covered with a plastic bag sealed to the base of the plant. Thus, all water lost from each pot was assumed to come only from plant evaporation.

Whole-plant water-use efficiency (WUE_P)

WUE_P was calculated as the ratio of the whole-plant biomass gain (ΔW_T) to CUW. That is, WUE_P = ΔW_T / CUW, where ΔW_T denotes the difference between W_T at t₁₆₃ and W_T at t₀ (i.e., ΔW_T = W_Tt163 - W_Tt0). The relative growth rates (RGR) was calculated as follows: RGR (g g^-1 day^-1) = [ln W_Tt163 - ln W_Tt0]/(t₁₆₃ - t₀), where W_Tt163 and W_Tt0 denote the plant biomass (in grams) at the indicated times, and (t₁₆₃ - t₀) denotes the time interval (t₁₆₃ - t₀ = 163 days).

Total nonstructural carbohydrates (TNC)

Leaf total nonstructural carbohydrate content was measured at the end of the experimental period. Nonstructural carbohydrates were extracted in boiling ethanol (80% v/v), and purified with chloroform. Starch in the ethanol-soluble residue was hydrolyzed in 0.5 M NaOH and the precipitate removed by centrifugation (15 min at 1000×g). Glucose content in the sample was measured spectrophotometrically at 490 nm after reaction with phenol-sulfuric acid, and the sugar content determined using a glucose standard.

Statistical analysis

We used a split-plot experimental design, with CO₂ levels (400 and 700 ppm) as whole plots and water regimes (soil at 50 and 100% FC) as subplots, with five replications (plants). Data were subjected to analysis of variance (ANOVA) and the post-hoc Fisher’s LSD test (p = 0.05) was used for mean separation. Biomass, CUW, CE_mass and TNC data were log₁₀ or log₍₁₀₊₁₎-transformed before conducting the ANOVA. Statistical analyzes were performed using the software package STATISTICA^® ver. 7.0 (StatSoft Inc., Tulsa, OK, USA).

  Results 

In the greenhouse relative humidity was 70-80% and mean temperature 27.5 °C (ranging from 26 °C at night to 29 °C at midday); mean PAR ranged from 185 to 215 µmol m^-2 s^-1 (over a 12-h period) and mean ambient CO₂ concentration was 410 ± 17 ppm (day/night of 400/420 ppm). Over the experimental period, mean predawn leaf water potentials were -0.20 MPa in well-watered plants and -0.34 MPa in those subjected water stress.

Gas exchange

Taking well-watered plants grown at ambient CO₂ as a baseline, A_sat-mass increased 28% (124 to 159 nmol g^-1 s^-1) under eCO₂ in well-watered plants, and 93% (86-166 nmol g^-1 s^-1) in plants submitted to water stress (Fig. 1a). Across water regimes, A_max-mass was 22% lower at eCO₂ (259 against 202 nmol g^-1 s^-1, p = 0.002 - Fig. 1b, Tab. 1), and over CO₂ levels, A_max-mass declined 11% under drought (244 against 217 nmol g^-1 s^-1, p = 0.039 - Fig. 1b, Tab. 1). Values of g_sCO2-mass declined 43% under drought at ambient CO₂ (1.01 to 0.58 mmol g^-1 s^-1), and 39% in well-watered plants at eCO₂ (1.01 to 0.62 mmol g^-1 s^-1, p = 0.001 - Fig. 1c). However, there was no effect of drought on g_sCO2-mass at eCO₂ (p = 0.22- Fig. 1c). There was no difference in C_i/C_a values between water regimes at eCO₂ (mean of 0.59, p = 0.57 - Fig. 1d). At ambient CO₂, however, C_i/C_a slightly decreased (11%) under drought (p = 0.006 - Fig. 1c), but this decline was not strong enough to affect C_i, which only varied in response to the CO₂ treatment. Mean C_i values were 399 ppm at eCO₂ and 240 ppm at ambient CO₂ (Tab. 2). In both CO₂ environments, we found no effect of water stress on CE_mass (p = 0.20 - Tab. 1), but over water regimes it was lower at eCO₂ (p = 0.003 - Fig. 2a). R_L-mass was not affected by drought or eCO₂ (p > 0.05 - Tab. 1), but on average R_D-mass was lower at eCO₂ (Tab. 2).

Fig. 1 - Light saturated photosynthesis on a mass basis (A_sat-mass, panel A), light and CO₂-saturated photosynthesis on a mass basis (A_max-mass, panel B); stomatal conductance to CO₂ on a mass basis (g_sCO2-mass, panel C) and intercellular to ambient CO₂ ratio (C_i/C_a, panel D) in Carapa surinamensis grown at 400 and 700 ppm of CO₂ and two water regimes, soil at 50% FC and 100% FC. Significant differences between boxes are indicated with different small letters within a CO₂ level and different capital letters within a water regime (Fisher LSD test at p ≤ 0.05). The boundaries of the box indicate the 25th and 75th percentile, and the solid line within the box denotes the median. The mean value for each box is available in the appendix table (Tab. 1A). The p values of the comparisons 1 × 4 and 2 × 3 (meaning of numerals is shown in the x-axis of the panel) are given when the interaction (CO₂ level × water regime) is significant (p ≤ 0.05, Tab. 1).

  Enlarge/Shrink   Download   Full Width  Open in Viewer

Fig. 2 - Carboxylation efficiency of Rubisco on a mass basis (CE_mass, panel a) and total nonstructural carbohydrates (TNC, panel b) in Carapa surinamensis grown at 400 and 700 ppm of CO₂ and two water regimes, soil at 50% FC and 100% FC. Significant differences between boxes are indicated with different small letters within a CO₂ level and different capital letters within a water regime (Fisher’s LSD test, p ≤ 0.05). Further information is described in Fig. 1.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

Biomass allocation, water-use efficiency and TNC

Growth in diameter was greater under eCO₂ than in ambient CO₂ conditions (p = 0.001), whereas height growth was enhanced at ambient CO₂ (p = 0.012 - Tab. 1, Tab. 2). Irrespective of the CO₂ treatment, growth rates in diameter and height were greater in well-watered plants than under drought (p ≤ 0.009 - Tab. 1, Tab. 2). CO₂-enrichment did not affect ΔW_L (p = 0.18 - Tab. 1) and, over CO₂ levels, ΔW_L was reduced by 59% under drought stress (46.8 vs. 19.0 g per plant, p < 0.001 - Fig. 3a). The ΔW_S was greater (83%) at eCO₂ (63.2 vs. 34.5 g per plant over water regimes), but it was significantly reduced (52%) by drought over CO₂ levels (66.1 vs. 31.6 g per plant, p = 0.004 - Fig. 3b, Tab. 1). Over water regimes, ΔW_R more than doubled under eCO₂ (30.9 vs. 10.3 g per plant, p < 0.001 - Tab. 1), and across CO₂ levels it was reduced (51%) by drought (27.6 vs. 13.6 g per plant, p = 0.06 - Fig. 3c, Tab. 1). Mean SRR declined under eCO₂ from 6.9 to 3.5 (p = 0.001), with no effect of water regimes (p = 0.24 - Fig. 3d, Tab. 1).

Fig. 3 - Gain of leaf dry matter (ΔW_L, panel a), stem dry matter (ΔW_S, panel b), root dry matter (ΔW_R, panel c), and shoot/root ratio (SRR, panel d) in Carapa surinamensis grown at 400 and 700 ppm of CO₂ and two water regimes, soil at 50% FC and 100% FC. Significant differences between boxes are indicated with different small letters within a CO₂ level and different capital letters within a water regime (Fisher’s LSD test, p ≤ 0.05). Further information is described in Fig. 1.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

Over water regimes RGR was improved (51%) at eCO₂ (7.55 vs. 5.01 g kg^-1 day^-1), and across CO₂ levels, it declined (32%) under drought (7.50 vs. 5.07 g kg^-1 day^-1, p = 0.001 - Fig. 4a, Tab. 1). Total leaf area gain (ΔA_L) was not affected by CO₂-enrichment (p = 0.70), but across CO₂ levels, it was lower under drought (0.71 vs. 0.23 m² per plant, p < 0.001 - Tab. 1, Tab. 2). Although the plants were a little taller under ambient conditions, across water regimes ΔW_T was 73% greater at eCO₂ (129.8 vs. 75.0 g per plant, p = 0.002 - Fig. 4b, Tab. 1) and, over CO₂ levels, it was reduced 54% by drought (140.5 vs. 64.2 g per plant, p = 0.003 - Fig. 4b). One can see in this figure that CO₂ enrichment greatly mitigates the effect of drought, as ΔW_T doubled under drought at eCO₂ (85.5 vs. 43.0 g per plant).

Fig. 4 - Relative growth rate (RGR, panel a), gain of total dry matter (ΔW_T, panel b), consumptive use of water (CUW, panel c) and whole-plant water-use efficiency (WUE_P, panel d) in Carapa surinamensis grown at 400 and 700 ppm of CO₂ and two water regimes, soil at 50% FC and 100% FC. Significant differences between boxes are indicated with different small letters within a CO₂ level and different capital letters within a water regime (Fisher’s LSD test, p ≤ 0.05). Further information is described in Fig. 1.

  Enlarge/Shrink   Download   Full Width  Open in Viewer

Across water regimes, CUW was reduced (43%) under eCO₂ (27.2 vs. 15.5 kg [water] per plant over the whole study period) and across CO₂ levels it was 63% lower under drought (31.2 vs. 11.5 kg [water] per plant, p < 0.001 - Fig. 4c, Tab. 1). Therefore, the WUE_p was substantially improved by CO₂-enrichment, particularly under drought (Fig. 4d). At eCO₂, it was improved by 188% in well-watered plants and by 262% under drought. However, WUE_p was not improved by drought at ambient CO₂ (p = 0.66, Fig. 4d). Across water regimes, SLA declined (23%) at eCO₂ (20.4 vs. 15.7 m² kg^-1, p < 0.001 - Tab. 1, Tab. 2) with no effect of soil water content on SLA (p = 0.29). Likewise, the TNC content was only affected by CO₂ enrichment, with an increase of 27% under eCO₂ (179 vs. 227 mg g^-1, p = 0.002 - Fig. 2b, Tab. 1).

  Discussion 

Irrespective of the water regime, A_sat-mass was greater in plants subjected to eCO₂ than in those kept at ambient CO₂ conditions, which indicates that the new leaves flushed during the experimental period did not experience down-regulation of photosynthetic at sustained eCO₂. Because the C_i values were higher at eCO₂ (240 ppm at ambient CO₂ against 399 ppm at eCO₂), this suggests that at ambient CO₂ the Rubisco carboxylation rate was limited by the diffusion of CO₂ to the carboxylation sites ([27], [30]). Improvement of carbon assimilation under CO₂ enrichment is in accordance with the findings reported by others ([13], [34], [1], [2]). In well-watered plants we found a rise of 28% in A_sat-mass under eCO₂, but that improvement was greater (93%) in plants submitted to drought. The decline of A_sat-mass under drought at ambient CO₂ can be the result of the combined effect of diffusive (g_s and mesophyll conductance) and non-diffusive (biochemical) limitations of photosynthesis ([12], [36], [19]). One can see in Fig. 1a,c that the decrease in A_sat-mass at ambient CO₂ under water stress was associated with a drop in g_sCO2-mass. It has been proposed that the decline in photosynthetic rate under mild drought (leaf water potential above -0.9 MPa) is most of time associated with a decline in stomatal conductance ([12]). The increase of A_sat-mass at eCO₂ concurs with the results reported by Ainsworth & Long ([1]) and Nowak et al. ([34]) who found that photosynthesis increases by about 30-50% under eCO₂. Kelly et al. ([23]) also reported an improvement in photosynthetic rates and a decline in the C_i/C_a ratio in response to CO₂-enrichment, which is in agreement with the results we found in this study (Fig. 1d).

There was a slight decline (17%) of CE_mass under elevated eCO₂ (Fig. 2a), which can be ascribed to a decrease in SLA (Tab. 2), as leaf thickness inversely increased with SLA ([27]). However, it is apparent that this small decline in CE_mass did not impair A_sat-mass which increased at eCO₂. This concurs with the result of a meta-analysis carried out by Ainsworth & Long ([1]) who found just a slight decline (6%) in the maximum carboxylation rate of Rubisco. Both R_L-mass and R_D-mass tended to decrease at eCO₂ (11-20%) but the effect of eCO₂ was only significant for dark respiration (Tab. 1). This is in agreement with the result reported by Curtis & Wang ([13]), who found a small decrease (18%) in R_D-mass under eCO₂. The decrease in leaf respiration can be ascribed, at least in part, to a decrease in SLA under eCO₂. Much of the discrepancy on the effect of eCO₂ on leaf respiration can be attributed to the difficulty in measuring this parameter with gas exchange techniques ([29]). We found an increase (27%) in TNC under eCO₂, and it has been suggested that the accumulation of leaf carbohydrates at eCO₂ can reduce the expression of genes coding for photosynthetic enzymes ([11]), but in this study we did not found evidence of photosynthetic acclimation.

Subjecting the plants to eCO₂ mitigates the negative effect induced by low water availability and, on average, A_sat-mass increased 54% at eCO₂. However, contrary to expectation A_max-mass declined at eCO₂ in well-watered plants and under drought at ambient CO₂. The decline of A_max-mass under ambient CO₂ can be attributed, at least partially, to a reduction in g_sCO2-mass (Fig. 1c), but an effect of mesophyll conductance cannot be ruled out ([19]). In fact, in several species there is a close correlation between mesophyll conductance and photosynthetic rate ([28], [40], [3]). It has been reported that an increase in TNC under eCO₂ can lead to down-regulation of photosynthesis ([30], [18]). Although A_sat-mass was not reduced by eCO₂ (Fig. 1a), down-regulation of A_max-mass in response to an increase in TNC cannot be entirely disregarded. It is important to note, however, that A_max-mass was measured at a C_a value of 2000 ppm (mean C_i of about 1100 ppm) against a C_a of 700 ppm (C_i of 399 ppm) used for measuring A_sat at eCO₂. Kitao et al. ([25]) found that mesophyll conductance exponentially declines with increasing C_i. Furthermore, Leakey et al. ([30]) suggested that the apparent down-regulation of carboxylation rates at eCO₂ can also be caused by a decrease in mesophyll conductance. Thus, it seems plausible to suggest that the decline of A_max-mass at eCO₂ was caused by a decline of mesophyll conductance and, to a lesser extent, by the increase in TNC at eCO₂.

In well-watered plants, g_sCO2-mass was lower under eCO₂ which is not unexpected, as the most common response is a decrease of stomatal conductance under eCO₂ ([13], [1], [29], [30]). One can see in Fig. 1c that g_sCO2-mass did not decline under drought at eCO₂, which leads to similar C_i/C_a values under CO₂ enrichment (Fig. 1d). This shows that stomata were sensitive to drought at ambient CO₂, but rather insensitive at eCO₂, which suggests some stomatal acclimation (a physiological change triggered by a new environmental condition) to eCO₂ ([33]). Yan et al. ([44]) found that under drought, stomatal conductance was dependent on xylem-abscisic acid concentration at ambient CO₂, whereas it was insensitive to abscisic acid and predominantly regulated by leaf turgor at eCO₂. We show that subjecting the plants to eCO₂ negates the effect of drought on stomatal conductance, and consequently water stress did not lead to a reduction in A_sat-mass under eCO₂. Even when predawn leaf water potential only slightly declined under drought (-0.20 to -0.34 MPa), there was a sharp decline of g_sCO2-mass at ambient CO₂ under drought, which suggests that andiroba is rather sensitive to small changes in leaf water potential ([8]).

Besides the effect of eCO₂ on g_sCO2-mass, the pronounced reduction in total leaf area under water stress also contributed to reduce CUW under drought (Fig. 4c). This occurs because of the negative effect of water stress on leaf production, as cell division and leaf expansion are greatly reduced under drought ([5], [42]). At eCO₂ the drop in CUW essentially mirrored the drop in ΔW_L, as g_sCO2-mass did no decline under drought at eCO₂. At ambient CO₂, however, the drop in CUW reflected both the decline in ΔW_L and g_sCO2-mass. Besides, in the greenhouse the air was slightly drier (10% lower RH) and a little (1.5 °C) warmer than in the growth chamber, which could also have had a small effect on CUW and thereby on WUE_P at ambient CO_2.

WUE_P was greatly improved by submitting the plants to eCO₂ (Fig. 4d), because of the combined effect of a decline in CUW and an increase in ΔW_T. The enhancement of WUE_P at eCO₂ is agreement with the results reported by others ([22], [23]). Cernusak et al. ([10]) found that WUE_P of seedlings of Swietenia macrophylla and Ormosia macrocalyx (tropical trees) can improve from 54% (S. macrophylla) to 91% (O. macrocalyx) in response to eCO₂, which shows that WUE_P can exhibit large variation between species even at the same experimental conditions. It is worth noting, that g_sCO2-mass and WUE_P followed different trends. For example, g_sCO2-mass declined under drought at ambient CO₂, whereas WUE_P remained unchanged in that condition (Fig. 4d). This can be explained if we take into account that, besides stomatal conductance, WUE_P also depend on other factors including biomass allocation to plant organs, leaf morphology, plant respiration and canopy boundary layer conductance, which ultimately leads to difference between A/E and WUE_P ([27], [30]).

The ΔW_T increased 73% at eCO₂ and decreased by 54% under drought over CO₂ levels. This is in tandem with the increase of A_sat-mass recorded at eCO₂ (28% and 93%, depending on water regime), which resulted in greater amount of carbon allocated to stems and roots. Our ΔW_T values are greater than the above ground biomass gain (20-30%) reported by others ([13], [1]) in plants subjected to eCO₂. This discrepancy can be explained by considering that Carapa trees grow at high rates during the juvenile stage ([16], [8]). On the other hand, our results are consistent with those reported by Cernusak et al. ([10]), who found that the biomass of well-watered tropical seedlings can increase from about 70% (Swietenia macrophylla) to 150% (Ormosia macrocalyx) at eCO₂. Large biomass allocation to stems and roots suggests that these plant organs can be the predominant sinks of Carapa saplings under CO₂ enrichment, as ΔW_L remained unaltered at eCO₂. In comparison with ambient CO₂, more carbon was allocated to roots at eCO₂ (Fig. 3c), and due to the absence of an effect of eCO₂ on ΔW_L, this ultimately led to a lower SRR. The decline of SRR at eCO₂ is rather unexpected, as in several species it remains unchanged under eCO₂ ([37]). The drastic reduction in ΔW_T at ambient CO₂ under water stress is consistent with the long-term decline in total biomass (250 to 150 Mg(C) ha^-1, in 10 yrs) observed under drought (i.e., a 50% reduction in water supply - [21]). RGR increased under eCO₂, which concurs with the finding reported by Poorter & Nagel ([37]). We found that the amount of carbon stored as TNC increased only 27% at eCO₂, i.e., less than the starch increase of 60-80% reported by Nowak et al. ([34]), which is consistent with the large fraction of biomass accumulated in stems and roots. SLA was greatly decreased under CO₂-enrichment. This is in accordance with the rise in TNC often found at eCO₂ ([13], [1]).

  Conclusions 

We postulated that plants under CO₂-enrichment enhance A_sat-mass, biomass accumulation and WUE_P, and that eCO₂ leads to a reduction of stomatal conductance, which was supported by data. However, surpassing our expectations, submitting the plants to eCO₂ entirely nullifies the negative effect of drought on A_sat-mass and g_sCO2-mass, and improves total biomass accumulation, but unexpectedly CO₂-enrichment leads to a decline in A_max-mass. Total leaf area was greatly reduced by drought, but the absence of a positive effect of eCO₂ on leaf area production was unexpected. These findings widen our understanding of the effect of eCO₂ and water stress on the physiology of Carapa, an important species of multipurpose use, particularly in the Amazon region. It should be acknowledged, however, that several factors can affect the performance of mature trees in their natural environment (e.g., light, nutrient and water availability, and the capacity of the root system to explore the soil for water and nutrients), which suggests caution in extrapolating results from green-house experiments to forest ecosystems.

  List of abbreviations 

(A): net photosynthetic rate; (A_L): total leaf area; (A_sat-mass): light saturated photosynthesis on a mass basis; (A_max-mass): light and CO₂-saturated photosynthesis on a mass basis; (B): biomass; (ΔB): biomass gain over the experimental period; (C_a): ambient CO₂ concentration; (CE_mass): carboxylation efficiency on a mass basis; (C_i): intercellular CO₂ concentration; (CUW): consumptive use of water (amount of water consumed by a plant); (DM): dry matter; (eCO₂): elevated CO₂ concentration; (g_sCO2-mass): stomatal conductance to CO₂ on a mass basis; (Log): logarithm; (PAR): photosynthetically active radiation; (R_L-mass): leaf respiration in the light on a mass basis; (R_D-mass): dark respiration on a mass basis; (RGR): relative growth rate; (SLA): specific leaf area; (SRR): shoot/root ratio; (TNC): total nonstructural carbohydrate; (W_L): leaf DM; (W_S): stem DM; (W_R): root DM; (W_T): total DM; (WUE_P): whole-plant water-use efficiency.

  Acknowledgments 

The authors acknowledge the Brazilian Ministry of Science, Technology, Innovations and Communications (MCTIC/INPA PRJ 15.120), and the National Council for Scientific and Technological Development (CNPq, 302041/2015-0) for grant (to RAM) and scholarship to MFO. We thank the anonymous reviewers for their useful comments and suggestions.

  Author contributions 

MFO collected data and conducted statistical analysis and RAM supervised the experimental work and wrote the article with contributions of the first author.

  References

  Supplementary Material

Tab. S1 - Means of data used in Figs. 1-4.