Regulation of photosynthesis and stomatal and mesophyll conductance under water stress and recovery in olive trees: correlation with gene expression of carbonic anhydrase and aquaporins

1Water stress is considered the main environmental factor limiting photosynthesis, plant growth, and yield worldwide, especially in semi-arid areas, where Olea europaea is well adapted. Under water-stress conditions, which are related to water depletion and/or high atmospheric vapour pressure deficit (VPD), photosynthesis decreases through several mechanisms including stomata closure, reduced mesophyll conductance to CO2 (g m), and feedback regulation by end-product accumulation. As soil water deficit and VPD result in reduced stomatal conductance to CO2 (g sc) (reviewed by; ) and g m in many species, several authors have suggested a possible co-regulation of g sc and g m. However, such co-regulation depends on the species and the prevailing conditions, such as the combination of drought with VPD or with radiation.

2Under natural conditions, water stress normally develops gradually over periods of weeks or months, and hence it is possible that some acclimation occurs in addition to day-to-day variations in response to variable environmental conditions. Acclimation to water stress may comprise responses involving modification of gene expression and plant physiology and morphology, taking place over days to weeks, which lead to a homeostatic compensation for the initial negative effects of water stress on photosynthesis. Nonetheless, little is known about the acclimation of photosynthesis to water-stress conditions in the short term, so studies that specifically address this issue are needed. The recovery phase after relief of stress (i.e. rainfall or irrigation) becomes another important part of the overall plant physiological response to a water-stress period. The capability for photosynthetic recovery from a water-stress period determines the future growth and survival of plants in their habitat, and depends on the degree and velocity of photosynthesis decline during water depletion. Although there is much information on the regulatory mechanisms of the response of g s to water stress and recovery, little is known about the regulation of g m. The main limiting factor for photosynthesis during water stress or recovery can vary depending on species, the intensity of previous stress, light and temperature, plant age, and the application of successive drought and recovery cycles.

3Therefore, besides photosynthetic biochemistry, the regulation of g s and g m is crucial to understand both processes of acclimation to stress and recovery after stress. Despite being one of the most studied physiological variables, the regulation of stomata is not yet fully understood. There is, however, a large consensus on the fact that two main components operate to produce the adequate response of g s to environmental stimuli: a hydropassive loop, related to the hydraulic capacity of the plant, and a hydroactive loop, linked to both the chemical signalling and the photosynthetic capacity of the leaf. Regarding the regulation of g m, recent evidence suggests that anatomical traits, such as cell-wall thickness and chloroplast distribution, are among its stronger determinants. Nevertheless, the most likely candidates proposed for the rapid regulation of g m are aquaporins (AQPs) and carbonic anhydrase (CA). There are several pieces of evidence of AQP involvement in g m regulation. First, inhibiting the activity of some aquaporins by HgCl2, found a decreased g m in Vicia faba. Secondly, the tobacco aquaporin NtAQP1 facilitates CO2 membrane transport when inserted in Xenopus oocytes. Finally, the strongest evidence comes from the observed altered g m in transformed and mutants plants.

4On the other hand, as aquaporins accumulate in cells around stomatal cavities and in guard cells themselves, they may also be involved in the regulation of g s. In fact, transgenic and mutant plants with altered AQPs present differences not only in g m but also in g s. Recently, found a strong correlation between g s and the gene expression of some particular AQPs during water stress and recovery in grapevines. Despite all these findings, the potential relationship between the gene expression of AQPs and the variations in g s and g m is still unclear, as patterns of AQP expression are complex and different AQP forms may induce distinct responses

5Concerning the involvement of CA in the regulation of g m, less evidence has been found. CA might have a role in the regulation of g m through the establishment of the dynamic equilibrium between CO2 and HCO3–. Likewise, the presence of isoforms of CA in all the compartments composing the diffusion way of the liquid phase of CO2 (except on the cellular wall) suggests that they may be involved somehow in the regulation of g m. The elimination of the stroma CA might result in a 50% reduction in g m but, in contrast, studies with mutant plants with reduced CA have shown little reduction in g m. Some authors have claimed that the role of CA could be species dependent and become more important in species with a high CO2 resistance in their cell walls and low g m, as is the case of many drought-adapted species.

6Olea europaea is an excellent plant model for studying physiological and molecular responses to water stress and recovery under realistic water-stress conditions because of its high reputation as drought-tolerant species, as well as its importance in the Mediterranean landscape and agronomical impact. As no previous study has related AQP and CA gene expression with all the potential photosynthesis-limiting factors (g s, g m and biochemistry) during water stress and recovery under field conditions, we studied the expression of two AQP genes characterized in olive, OePIP1.1 and OePIP2.1, and a stromal CA. The experiment was carried out outdoors, under natural conditions of radiation and atmospheric demand, in contrast to most up-to-date studies of the relationships between AQPs and CA, which have been performed in plants grown in growth cabinets. The specific objectives of the present work were: (i) to quantify the photosynthesis limitations imposed by g s, g m and photosynthetic capacity during the water-stress and recovery period, and (ii) to relate these changes to changes in the gene expression of AQPs and CA in order to expand current knowledge on the possible role of the AQP and CA expression in the regulation of g s and g m.

7Eighteen 5-year-old O. europaea L. var. Manzanilla plants were grown outdoors in 50 l pots (see Supplementary Fig. S1 at JXB online) at ‘La Hampa’, an experimental farm of the Consejo Superior de Investigaciones Científicas (37°17′ N, 6°3′ W, altitude 30 m), near Seville (south-western Spain). Pots contained sandy-loam soil, typical of this farm and whose hydraulic properties have been described previously by. The experiment was performed during the summer of 2009. The climate was typically Mediterranean, with a mild, wet season from October to April and hot and dry conditions from May to September. Six trees were pruned by one-third of their leaf area to reduce leaf water demand and to produce a slower evolution of water stress. Branches were cut at dawn after being sprayed with water and were covered with plastic bags to minimize embolism risk. Pruning cuts were covered with wound dressing. Thus, two different water-stress intensities were imposed by withholding irrigation in six non-pruned trees (S) and six pruned trees (SP) on 26 July. Another group of six olive trees was maintained as well watered and was used as control treatment (C). Finally, after 13 d of total water withholding, S and SP plants were rewatered daily for the following 6 weeks and allowed to recover. All the physiological measurements detailed below were taken in leaves of a similar age, including those carried out during the recovery experiment. This ruled out the possibility that new, recently grown leaves were measured.

8Meteorological variables were measured with an automatic weather station (Campbell Scientific Ltd, Shepshed, UK) located 50 m away from the experimental trees. Average values taken over 30min of net radiation, global radiation, photosynthetically active radiation, wind speed, air temperature, relative humidity, and VPD were recorded.

9The volumetric soil water content (θ, mm) of the substrate in the pots was determined by the time–domain–reflectrometry technique using a Tektronix cable tester (Model 1502C; Beaverton, OR, USA). Measurements were made at 7.00 Greenwich Mean Time (GMT) and averaged between 0.05 and 0.20 m in depth because there were no differences between the values recorded at either depth within each treatment. θ was transformed into soil relative extractable water (REW) according to:

10where θ is the actual soil water content, θ min is the minimum soil water content measured during the experiment (mm), and θ max is the soil water content at field capacity (mm). Field capacity was determined to be 0.25 m3 m–3 and the soil water content at permanent wilting point was 0.04 m3 m–3. A REW value of >1 could be observed for several hours after water application, especially in the C treatment, as during that period θ exceeded field capacity.

11Leaf water potential was assumed to be equal to xylem pressure potential at the petiole in transpiring leaves and was measured in six leaves per treatment at pre-dawn (Ψ pd) and 11.00 GMT (Ψ) with a pressure chamber (Soilmoisture Equipment, Santa Barbara, CA, USA). The relative water content of leaves (RWC) was determined at 11.00 GMT in opposite leaves to those sampled to measure Ψ as:

12Olive leaves were weighed immediately on harvest to determine fresh weight. To determine the turgid weight of the samples, these were kept in distilled water in darkness at 4 °C to minimize respiration losses until they reached a constant weight (full turgor, typically after 24h). Their dry weight was obtained after 48h at 70 °C in an oven.

13Leaf gas exchange in leaves similar to those used for plant water relationships was determined simultaneously with chlorophyll fluorescence at 11.00 GMT using the open gas-exchange system Li-6400 (Li-Cor, Lincoln, NE, USA) equipped with an integrated fluorescence chamber head (Li-6400–40; Li-Cor). Photosynthesis was induced with saturating light (1800 µmol m–2 s–1) and an ambient concentration of CO2 of 400 µmol mol–1.

14From the fluorescence measurements, the actual photochemical efficiency of photosystem II (ΦPSII) was determined according to as:

15where F s is the steady-state fluorescence in the light [here photosynthetic photon flux density (PPFD) of 1800 µmol m–2 s–1] and F m′ is the maximum fluorescence obtained with a light-saturating pulse (~8000 µmol m–2 s–1). As ΦPSII represents the number of electrons transferred per photon absorbed by PSII, the rate of electron transport (J) can be calculated as:

16where PPFD is the PPFD incident on the leaf, 0.5 is a factor that assumes equal distribution of energy between the two photosystems, and 0.93 is the leaf absorptance determined in five leaves per treatment using an integrating sphere with a portable spectroradiometer (LI-1800; Li-Cor) and calculating absorptance as 1 – reflectance – transmittance. The relationship between ΦPSII and the quantum efficiency of gross CO2 fixation (ΦCO2) was obtained by varying the light intensity under non-photorespiratory conditions in an atmosphere containing <1% O2. The correlation between ΦPSII and ΦCO2 was done under different conditions of soil water deficit and VPD (see Supplementary Fig. S2 at JXB online) to check for potential errors in the estimation of g m due to non-linearity of electron transport rate.

17Estimation of g m was performed with the ‘variable J method’ of:

18where A N and internal CO2 concentration (C i) were taken from gas-exchange measurements at saturating light, J was estimated from fluorescence, the rate of mitochondrial respiration in the light (R d) was assumed to be the same as measured dark respiration,and Γ* was estimated at the measuring temperature by using the kinetic constants proposed for this parameter by. To estimate R d, we covered the leaves (n=6) for 30min and measured the net assimilation of CO2 rates using a modified 2×3cm broadleaf chamber and an integrated light source (LI-6400-02B; Li-Cor) at a flow rate of 250 µmol air s–1. The relationships between ΦPSII and ΦCO2, and R d were measured in three stages of the experiment: one at the beginning of the experiment, the second 7 d after withholding irrigation, and the third 2 d after resuming irrigation. No significant differences in leaf absorptance were found throughout the experiment, despite changes in the RWC of leaves.

19To assess the limitations imposed by water stress and recovery on photosynthesis, a quantitative limitation analysis of photosynthesis was conducted according to, with modifications. This approach requires measurements of A N, g s, g m, and V cmax to calculate the partition of photosynthesis limitations into components related to stomatal conductance (S L), mesophyll conductance (MC L), and leaf biochemical characteristics (B L), assuming a reference treatment where maximum assimilation rate, g s, g m, and V cmax can be defined (see Supplementary Information in JXB Online for details of equations used). Total limitations (T L) were defined as the sum of S L, MC L, and B L. As actual electron transport rate (ETR, i.e. fluorescence-derived J) is tightly coupled with V cmax and should indeed reflect gross photosynthesis, B L was calculated using ETR instead of V cmax as a surrogate for leaf biochemistry. There were three reasons for the use of ETR as a surrogate of V cmax in the limitation analysis: (i) under our experimental approach, it was unfeasible to build a number of A N–C i curves enough to estimate V cmax for every sampling date; in addition to this, when g s is low, the estimation of V cmax is highly prone to errors due to the low rates of leaf gas exchange; (ii) as actual electron transport rate is tightly coupled with V cmax and should indeed reflect gross photosynthesis, ETR can be used instead of V cmax as a surrogate for leaf biochemistry; the uncertainties in the determination of V cmax reported by several authors can be avoided by using ETR, as has been verified and confirmed by, ); and (iii) we checked that, under our conditions and during the progress of water stress and recovery, there was a good correlation between ETR and V cmax for the range measured; this correlation is shown in Supplementary Fig. S3 at JXB online. In the current study, the maximum assimilation rate, concomitantly with g s, g m, and ETR, was reached under well-watered conditions, and therefore the C plants were used as a reference. However, as A N of the C plants increased during the experiment, presumably due to environmental changing conditions, the values of the C plants for each day were considered as the reference for the S and SP treatments determined on the same day.

20The same leaves used throughout the experiment for gas exchange and chlorophyll fluorescence measurements were harvested at 12.00 GMT for gene expression analysis. They were harvested at the same time of the day to avoid diurnal fluctuations in the transcripts. Three replicates per treatment and day were sampled and were frozen immediately in liquid nitrogen. Leaf tissues of each sample were ground to a fine powder with liquid nitrogen using sterile mortars and pestles. RNA extraction was performed on 100mg of plant material using an RNeasy Plant Mini kit (Quiagen, Venlo, The Netherlands) and treated with RNase-Free DNase (Qiagen) to avoid genomic DNA contamination during RNA purification. The concentration of total RNA fraction in each sample was estimated by measuring the sample absorbance (A 260/A 280) with a spectrophotometer (Lambda 6 UV-VIS; Perkin-Elmer, Buckinghamshire, UK). RNA reverse transcription to cDNA was performed on 1 μg of RNA using a QuantiTect Reverse Transcription kit (Qiagen). The samples were then stored at –80 °C until real-time PCR analysis.

21The three genes whose expression was chosen to be studied as influenced by water stress and recovery were: OePIP1.1 (; GenBank accession no. DQ202708) and OePIP2.1 (; GenBank accession no. DQ202709), encoding functional water-channel proteins in olive; and the gene encoding an olive CA enzyme located in the chloroplast stroma (GenBank accession no. FN814304). We constructed a multiple alignment of the whole Arabidopsis family reported by, including the olive CA. The isoform of CA of olive was confirmed to be aligned with Arabidopsis stroma isoforms βCA (see Supplementary Fig. S4 at JXB online) described by to be targeted to the chloroplasts. The CA full DNA coding sequence of Arabidopsis thaliana was retrieved from phytozome (http://www.phytozome.net) for phylogenetic analysis with the identified O. europaea mRNA (CA, GenBank accession no. FN814304). Alignment and phylogenetic analysis were performed with the seaview toolkit. Multiple alignment of mRNA sequences was performed using the MUSCLE algorithm. As phylogenetic inference could be biased by partial sequences in the alignment, as with the O. europaea partial mRNA in our dataset, prior to phylogenetic analysis we extracted the evolutionary and conserved aligned blocks from the alignment using the Gblock algorithm. A phylogenetic tree was created by the maximum-likelihood approach using the PhyML algorithm, with the K80 nucleotide substitution model, 100 random tree starts, and an optimized ratio between nucleotide transition and transversion (T s/T v ratio). Branch support was estimated by an approximate likelihood ratio test based on the Shimodaira–Hasegawa-like procedure.

22Despite the fact that both OePIPs are more expressed in roots and twigs, we studied them because they are also expressed in leaves and are actually the only PIPs that have been studied and characterized in olive. We included the β-actin gene (GenBank accession no. AF545569) as a housekeeping control. Specific primers were designed by Applied Biosystems (Foster City, CA, USA). The gene expression assays consisted of a 20× mix of unlabelled PCR primers and TaqMan MGB probes (labelled with carboxyfluorescein dye).

23Real-time PCR (7300 Real-Time PCR System; Applied Biosystems) was performed on the samples, and relative gene expression was determined using the relative standard curve method (Applied Biosystems). As a calibrator, a sample from the experiment (C plant) was chosen, the calibration curves being performed with a 1, 10, 50, and 100ng dilution of the cDNA of the same sample, for each of the genes chosen to be studied plus the housekeeping gene. Fast PCR cycles were performed running 96-well Fast reaction plates with 20 µl of the reaction mix in every well and using TaqMan Universal PCR Master Mix (2×; Applied Biosystems). The PCR thermal cycling conditions used were the default ones, as indicated by the manufacturer’s instructions.

24Statistical analysis was carried out using Statistica for Windows v.6.0 software package (StatSoft). Significant differences between means were assessed using one-way analysis of variance applying Tukey’s test (P<0.05). Previously, data were tested for normality and homoscedasticity.

25A d-separation method of path analysis was used to test several conceptual models that could explain the covariance among the measured physiological variables, especially g m, and the genetic expression of AQPs and CA. The proposed models were based on previous knowledge and the hypothesis of interactions between variables. Causal relationships between variables were combined to form directed graphs (the path models). These directed graphs implied a series of independence relationships between pairs of variables. Pearson correlation coefficients between pairs of variables were determined. The relationships among variables were then translated into a structure of variances and covariances that could be tested against the observed data. Path analysis was performed using the Causal Toolbox packages.

26REW dropped rapidly after water withholding in the S and SP treatments (Fig. 1a). Both stress treatments showed REW of <0.3 as soon as 2 d after withholding irrigation (a.w.i.). After that, a gentle decrease in REW down to 0.08 was recorded, suggesting very little water use by the plants. On average, the atmospheric demand for water was high, with maximum VPD values usually >4 kPa, and even reaching 6.8 kPa (Fig. 1b). During the initial drop in REW, both S and SP plants maintained both Ψ and RWC values similar to the C plants (Fig. 2), although with slight larger values for SP. However, differences with C were observed on day 4 a.w.i. and afterwards, with higher values in SP (with a smaller leaf area) than in S. At the end of the stress period, both stress treatments reached similar minimum values of Ψ and RWC of approximately –5.5MPa and 55%, respectively. Leaf gas-exchange variables were more responsive to the decline in soil REW, and a reduction in all three variables, g s, g m, and A N, was observed before changes in leaf water status were noticeable (Fig. 3). In parallel, ETR showed a continuous decrease in both S and SP plants from the beginning (Fig. 4).

27OePIP1.1 expression in S and SP plants peaked on day 4 a.w.i. and returned to C values at the end of the drought period (Fig. 5a). Expression of OePIP2.1 in stressed plants was similar to that in C plants at the beginning of the drought period but diminished at the end (Fig. 5b). For CA, relative expression in stressed treatments was similar to that in C plants during the first days of drought but diminished at the end of this period (Fig. 5c). However, compared with the first day, relative expression in stressed treatments showed a continuous decrease from the beginning, similar to that described for ETR (Fig. 4).

28After resuming irrigation (a.r.i.), leaf water status recovered before gas-exchange variables did. At just 8h a.r.i., once REW was replenished (Fig. 1a), Ψ experienced a large recovery up to values around –2.0MPa, with higher values in SP than in S plants (Fig. 2a). On day 2 a.r.i., RWC recovered fully in S and SP plants, while Ψ did on days 3 and 5 a.r.i. in the case of SP and S plants, respectively. However, g s, g m, A N, and ETR had not reached C plant values in S and SP plants by day 10 a.r.i. In view of these results, another experimental day was programmed on day 47 a.r.i. (Figs 3 and 4), which finally confirmed the recovery in g s. In all variables, SP plants showed a larger capacity for recovery than S plants insofar as, on day 4 a.r.i., g s in S and SP plants had recovered by 6 and 35%, g m had recovered by 35 and 53%, and A N had recovered by 20 and 53%, respectively. OePIP1.1 expression did not change significantly after the recovery irrigation, except for a peak observed on day 2 a.r.i. in S plants, similar to that found during the first stage of the stress period. After this, all three treatments behaved similarly. In contrast, OePIP2.1 expression presented a clear enhanced expression after irrigation, the response being stronger in S than in SP plants. CA expression recovered on day 3 a.r.i. (Fig. 5c).

29During the drought period, the diffusional limitations (D L=S L+MC L) prevailed over B L (Fig. 6). Two main stages in the time course of photosynthesis limitations could be differentiated in both S and SP treatments. In the initial one, during the first 2–4 d the drop in g s was accompanied by a predominant role of S L accounting for up to 50% of the T L in S and 35% in SP plants. In the second one, during the next 9 d without irrigation where B L, and especially MC L, took the dominant role, S L lost importance. During this stage of severe water stress, B L was around 20% of T L. At the end of the drought period, all limitations in S plants were imposed by MC L and B L, accounting for 35 and 65%, respectively. Similar trends were observed in SP plants, with a minimal role of S L (17%).

30During the recovery period, D L also prevailed over B L. The dynamics of the limitations during this period were different between the stressed treatments. In S plants, on 1 d a.r.i. nearly 100% of T L was S L, and it then accounted for about 40% until the end of the experiment. The main difference observed between S and SP plants was that, in the latter, SL was not so exclusive on the first day of recovery, with MC L and B L playing a significant role. On the following days, TL was between 20 and 30% lower in SP than in S plants, with a significantly lower MC L. All limitations had disappeared by d 47 a.r.i.

31Normalizing the gene expression, g s, and g m by the first date (26 July) allowed us to eliminate the factor of different absolute values and concentrate on the evolution along the experiment. There were linear positive relationships of g m (P=0.0001) and g s (P=0.0001) with the expression of OePIP2.1 (Fig. 7b, e) and a hyperbolic one between g m and the expression of CA (P<0.0001) (Fig. 7c). Although OePIP1.1 did not show a significant relationship with g m and g s, this was largely due to the inclusion of the peak points measured just after the imposition of water stress or just after irrigation was resumed. If we discarded these values (on the bottom right corner of Fig. 7a, b), we obtained a significant correlation between variables in both cases (P=0.003, r 2=0.43, and P<0.0001, r 2=0.64, for g m and g s, respectively).

32Several alternative models were performed and studied with a path analysis to reveal the causal relationships that linked gene expression with g m, g s, and A N. Of all the possible combinations among variables, only those that were physiologically sound were tested. We stated a priori the relationships among variables with a strong mechanistic or well-established and accepted empirical basis only. The main underlying hypotheses were: (i) A N is determined mainly by S L+MC L; (ii) AQPs can affect H2O fluxes (g s); (iii) AQPs can affect CO2 fluxes (g m); (iv) PIP1.1 and PIP2.1 interact to modify the traffic to the membranes and to build a tetramer that confers the basis of the CO2 transport function; and (v) CA can affect g m in sclerophyll plants. Fig. 8 shows three of the most representative models tested. These three were consistent with hypotheses 1 and 2. Model 1 included the role of OePIP1.1 in g m, reported for other PIP1s in other species, but, even though CA was included affecting g m directly, this model was rejected on the grounds that P was <0.05 (i.e. it had little prediction capacity). Nevertheless, when in model 2, we changed the role of OePIP1.1 in g m for that of OePIP2.1 and included the interaction between OePIP1.1 and OePIP2.1, simulating the tetramer structure, we obtained a satisfactory explanation for a high proportion of the variance found (P=0.83). However, the best fit to the data was reached with model 3, which completed model 2 by incorporating the role of CA, as a function of A N, in g m (P=0.98). The inclusion of Ψ (as a surrogate of plant water stress) affecting OePIP1.1 and OePIP2.1 reduced the proportion of the variance found (χ 2=12.92; P=0.93), although the resulting model continued to be useful.

33Photosynthetic variables were the first to respond to irrigation withholding, being lower in stressed treatments than in C plants on d 2 a.w.i. (Figs 3 and 4). Leaf water status responded 2 d later (Fig. 2). The initial reduction in g s, larger in S than in SP plants, was enough to keep Ψ at constant values, demonstrating the fine control of plant water status in olive by stomata at the initial stages of water stress (Figs 2a, and 3a). Stomatal closure has been interpreted as a plant mechanism to maintain xylem Ψ well above a critical value avoiding catastrophic hydraulic failure. On d 4 a.w.i., Ψ was around –3.0MPa, corresponding to a xylem Ψ around –5MPa, which is not enough to provoke important losses of hydraulic conductivity in olive (around 15–30% loss of hydraulic conductivity), as some works have reported. However, at this stage, g s had nearly reached its minimum value, and there was no further capacity to control Ψ, which later dropped to very low values of –6MPa in the case of S treatment. The progress of stress in the experiment was quicker than we expected initially. Although we used large pots (50 l), the leaf area and the high atmospheric demand in our experimental site meant that plants consumed the available water in a few days. The reduction in leaf area by pruning was not enough to make large differences between both stressed treatments, but some significant differences were still found between them as shown in Figs 3 and 4, 2 d a.w.i. and especially a.r.i. The time of the day at which Ψ and leaf gas exchange measurements were made were chosen carefully to try to determine the most significant moments: minimum Ψ and maximum g s. The lack of difference in Ψ between treatments on many days, as a consequence of its isohydric behaviour, did not affect the difficulty of unravelling causal relationships, although certainly it would be desirable to repeat the experiment under a more gradual drought-stress imposition in the future.

34After rewatering, leaf water status recovered earlier than photosynthetic variables from drought, matching the C plant values on d 5 a.r.i., while A N and ETR did so on d 10 a.r.i. (Figs 2–4). Olive has a high capacity for recovery from water stress after rewatering, as previous studies have shown. A similar pattern has been reported in tobacco. The value of g s showed the slowest recovery of all the variables studied. The reasons for this commonly slow recovery of g s after restoration of leaf water status is not yet clear, although it has been related to both hydraulic limitations and chemical limitations. The above-mentioned low values of Ψ reached at the end of the drought period (approx. –6MPa) suggested that a large loss of stem hydraulic conductivity could be a hydraulic factor conditioning the slow and partial recovery of g s when water was applied during the recovery stage. However, even at these values of Ψ, olive is not expected to lose much of its hydraulic conductivity: between 35 and 40% according to, explaining the ability of this species to tolerate severe drought. Other components of the hydraulic system could have not recovered after resuming irrigation and would have been limiting, like leaves or roots. However, this does not look to be the case, as reported recently by ). Concerning chemical limitations, several authors have pointed out the accumulation of abscisic acid (ABA) during stress acclimation as a circumstance that might prevent g s from fully recovering once water is available again and plant water status has been restored. explained this role played by ABA in the recovery stage as a way of controlling transpiration rate during the embolism-repairing time. However, the conclusions of that work, carried out in grapevine, are not necessarily applicable to olive, especially as the former has been reported to have a great refilling capacity of embolized vessels. Moreover, even in grapevine, contradictory results have been obtained, with no recovery of g s once ABA accumulation was fully reversed after a few days. Finally, several authors have reported that recovery from water stress depends on the level and velocity of the stress imposition. In this sense, our experiment showed a faster recovery of all physiological variables in SP than in S plants.

35Our results confirm the hypothesis that D L prevails under most water-stress situations in numerous species and that B L usually appears only under severe water stress (Fig. 6). Although during the first 2–4 d of drought S L was the only prevailing limitation, as drought intensified to a severe level both MC L and B L increased. MC L in particular accounted for up to 60% of the T L, reflecting the important and determinant role of g m in the photosynthesis rates. This agrees with previous reports in similar experiments where, under the most severe conditions of water stress MC L proved to be the maximum limitation. Once water was available again after irrigation, the most responsive variable determining the recovery of A N was g m. MC L was reduced to a residual 15–20% in the S treatment, while it was nearly eliminated in the SP treatment (Fig. 6). The faster recovery of photosynthetic variables in SP compared to S plants translated into lower values of T L in SP than in S plants. A good correlation between g s and g m was found for most of the studied period with a g m:g s ratio (both on CO2 basis) of 2.1. Considerable evidence of the close relationship between g s and g m has been reported, but this relationship has been regarded as the result of a tight covariance of these two resistances in the CO2 pathway to the leaf rather than as a cause-effect relationship. In fact, several works have indicated that the relationship between g s and g m can be modified according to changing environmental conditions.

36The expression patterns during stress treatments were different between the two AQPs studied and were more irregular than for CA (Fig. 5). This impression was shown by the peak responses measured at the beginning of both stress and recovery periods in OePIP1.1. The peaks at the onset of drought have been reported in similar studies, where the upregulation of AQP expression, including PIP1 and PIP2, was interpreted as a mechanism to promote water movement inside leaves via symplast by increasing membrane permeability to water when this is less available for the plant. The peak of expression in OePIP1.1 at the start of the re-irrigation period could explain the fast recovery of leaf water status by the possible role of AQPs in the xylem refilling of parenchyma cells. This peak would occur in a moment of high water availability in soil, which would reduce risks of massive losses of water from leaves. On the other hand, the downregulation in OePIP2.1 expression at the end of drought period may encourage cellular water conservation during periods of water stress by reducing membrane water permeability and limiting loss of cellular water. Probably, in order to maintain a suitable water status under abiotic stress, both increased water transport via AQPs in some tissues and reduced in others are required because one of the main role of AQPs is to maintain homeostasis and water balance under water-stress conditions.

37In addition to their role in water transport, there is published evidence of the putative role of AQPs in CO2 transport and g m. Accordingly, we found a linear relationship between OePIP2.1 expression and g m, but no relationship between OePIP1.1 expression and g m (Fig. 7). Although it is found that most members of PIP2 act as water channels while PIP1 members are involved in a CO2 transport function, found differences in g m induced by a PIP2. On the other hand, it has been reported that PIP1 and PIP2 AQPs form heterotetramers, which modify their function as H2O or CO2 membrane transport facilitators depending on their composition; hence, both PIP1 and PIP2 may be inextricably linked with each other and with g m. On the other hand, OePIP2.1 showed a similar degree of correlation with g s, as already observed for PIP2.1 of grapevines. In summary, while it is unclear from the data whether the expression of specific AQPs is related to regulation of both g s and g m, or whether they just operate by affecting one of the two conductances and a tight co-regulation between g s and g m provokes an apparent correlation of the other with PIP expression, the results strongly suggest that AQP expression is involved in setting diffusional limitations to photosynthesis in olives under water stress and recovery. This same link provokes an additional tight correlation between the two apparently less related parameters of g m and the leaf hydraulic conductance.

38The CA expression patterns in stressed treatments throughout the experiment were smoother than AQP patterns and were similar in S and SP plants (Fig. 5c), closely tracking those observed for ETR (Fig. 4). The role of this stroma CA in g m regulation was supported by the relationship found between CA expression and g m (Fig. 7c). The hyperbolic shape of this relationship agrees with the consideration of CA as not being limiting to photosynthesis due to the large amount of CA present in leaves, although this can be species dependent. Although the role of CA in regulating g m is more controversial than the evidence found for AQPs, some authors have reported that this role can be species dependent, gaining importance when g m is low, as happens in sclerophyll species. Therefore, this would also be the case for olive, a sclerophyllous species with a thick cell wall. Following the results of, olive would have to counterbalance the low conductance at the cell-wall sites with a larger conductance at the chloroplast sites, which undoubtedly would make g m more dependent on CA.

39To analyse the causal nature and the structure of the relationships found here (Fig. 7b, c), we considered several causal models with physiological sense that fitted the data. The most robust structure with the highest significant levels in many of the combinations tested was that obtained in model 3 (Fig. 8). This model showed that there could be a direct role of OePIP2.1 and CA expression in g m regulation, as well as an indirect role of OePIP1.1. The interaction between both AQPs is justified in accordance with the conclusions by, who suggested that PIP1 and PIP2 interact to form a heterotetramer. Additionally, the model considers that both PIPs are directly involved in the regulation of g s, that both S L and MC L determine A N, and that CA is in turn regulated by A N. OePIP1.1 and OePIP2.1 in olive have been described to regulate membrane water permeability in leaves, a fact that explains their effect on g s. CA was tested in an independent model (not shown) to regulate g m directly without the role of AQPs, this model being able to explain only a small amount of the data variance (χ 2=22.53; degrees of freedom=16; P=0.126). This supports the hypothesis that the role played by CA in the regulation of g m is not central. In any case, in addition to the implications in sclerophyllus leaves mentioned above, the inclusion of CA as a function of A N is justified because the regulation of CA gene expression is related to the leaf photosynthetic activity by the CO2 inside leaves. CA has been reported to be closely related to Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) activity, and therefore to A N. Furthermore, the time course of CA expression and A N were quite similar throughout the experiment (Figs 5c and 3c) and recent studies have demonstrated the role of CA as an upstream regulator of CO2-controlled stomatal movements in guard cells. While the validation of a causal model does not imply that the hypothesis is true, it does imply that the hypothesis is plausible given the empirical data. C i has long been proposed as a potential signal used by the plant to maintain the equilibrium among g s, g m, and A N. Under water stress, downregulation of g s, g m, V cmax, and ETR could occur to keep C i under relatively constant values during the acclimation process. If this hypothesis is correct, CA would be in an optimal position for regulation of gm, as it uses as C i as a substrate and influences its diffusion inside the cell. Although gene expression may not necessarily reflect protein function because of the post-transcriptional regulation, the good relations of g m with expression of OePIP2.1 and CA and the good fitting of the causal model proposed allow us to go beyond already published literature and to infer a direct role of AQPs and CA in g m regulation, although certainly the mechanistic basis for such a role remains to be elucidated.

40The main objective of this study was to provide a step forward in the putative role played by AQPs and CA in the regulation of g m under water stress. Despite the use of the path analysis to infer some potential causal effects, a novel aspect of this work is the execution of the experiment outdoors under natural conditions. Photosynthetic variables in olive trees responded faster than water status under drought but recovered much later. Despite the severe stress, D L prevailed over B L throughout the experiment. The role of MC L varied throughout the experiment, prevailing over the rest of the limitation components as stress intensified and being present even during the recovery phase. This study supports the hypothesis that the regulation of g m is regulated mainly by AQPs and that both OePIP1.1 and OePIP2.1 are likely to interact to exert a significant effect on g m. The data shown in this study reveals novel evidence on the putative role of CA in the regulation of mesophyll conductance to CO2, which, although small if compared with that of AQPs, is justified by the sclerophyllous nature of olive leaves. Due to the tight correlation between AQPs and both g s and g m, evidence of co-regulation between the two conductances and the fact that H2O and CO2 fluxes in leaves share a part of their pathways, facilitated by AQPs, it is very difficult to relate genes with functions in an unequivocal manner. As far as we know, there are few than expected studies showing this sort of correlation and, of course, no reverse genetics approach is possible to date with woody species like olive trees. Hence we believe that our approach, although certainly limited, is useful for a better understanding of plant physiological response to water stress, and is perhaps the only one currently available for olives. New biochemical probes and techniques are therefore required in order to quantify the activity of AQPs and their genetic expression changes, as well as to follow their spatial distribution in cells.

41Supplementary data are available at JXB online.

42Supplementary Fig. S1. Picture showing the aspect of the experimental set-up: 5-year-old trees in 50 l pots.

43Supplementary Fig. S2. Relationship of the quantum yield of photosystem II (ΦPSII) and ΦCO2 [(A N+R d)/PPFD].

44Supplementary Fig. S3. Relationship between V cmax and ETR used in the quantitative analysis of photosynthesis limitations throughout the experiment.

45Supplementary Fig. S4. Maximum-likelihood tree of Olea europaea carbonic anhydrases (CA) with Arabidopsis thaliana CA.

46Supplementary method. Assessment of limitations to photosynthesis.