Threshold response of mesophyll CO2 conductance to leaf hydraulics in highly transpiring hybrid poplar clones exposed to soil drying

Mesophyll conductance (g m) has been shown to impose significant limitations to net CO2 assimilation (A) in various species during water stress. Net CO2 assimilation is also limited by stomatal conductance to water (g sw), both having been shown to co-vary with leaf hydraulic conductance (K leaf). Lately, several studies have suggested a close functional link between K leaf, g sw, and g m. However, such relationships could only be circumstantial since a recent study has shown that the response of g m to drought could merely be an artefactual consequence of a reduced intercellular CO2 mole fraction (C i). Experiments were conducted on 8-week-old hybrid poplar cuttings to determine the relationship between K leaf, g sw, and g m in clones of contrasting drought tolerance. It was hypothesized that changes in g sw and K leaf in response to drought would not impact on g m over most of its range. The results show that K leaf decreased in concert with g sw as drought proceeded, whereas g m measured at a normalized C i remained relatively constant up to a g sw threshold of ~0.15mol m–2 s–1. This delayed g m response prevented a substantial decline in A at the early stage of the drought, thereby enhancing water use efficiency. Reducing the stomatal limitation of droughted plants by diminishing the ambient CO2 concentration of the air did not modify g m or K leaf. The relationship between gas exchange and leaf hydraulics was similar in both drought-tolerant and drought-sensitive clones despite their contrasting vulnerability to stem cavitation and stomatal response to soil drying. The results support the hypothesis of a partial hydraulic isolation of the mesophyll from the main transpiration pathway.

Introduction important source of resistance to water transport in highly transpiring plants, whereas the resistance occurring outside the xylem represents about a third of the total resistance to leaf water movement (Sack et al., 2004(Sack et al., , 2005. Nonetheless, a decline in water transport outside the xylem generally occurs during drought stress due to a shift from the symplastic-transcellular to the apoplastic water pathways, which also produces a decrease in K leaf and g sw (Pou et al., 2013). Thus, in highly vascularized plants such as hybrid poplar, it may be expected that during drought, leaf xylem cavitation would precede an eventual shift of water flow to the outer xylary water pathways. This in turn would lead to a decrease in K leaf and gas exchange as a way to protect the leaf against excessive cavitation.
When considering how environmental conditions may affect CO 2 transport inside leaves, one must take into account the mesophyll conductance (g m ) pertaining to the CO 2 diffusion path from the substomatal cavities to the chloroplast stroma (Ethier and Livingston, 2004;Evans et al., 2009;Tholen et al., 2012). Although an apparent dynamic regulation of g m in changing microclimates or growth conditions has frequently been reported in recent years (reviewed in Flexas et al., 2008), the physiological basis of such g m responses and how these changes orchestrate themselves with parallel changes in K leaf and g sw remain poorly understood. For example, g m and g sw appear to decline in concert as a result of drought stress (Warren, 2008;Centritto et al., 2009;Barbour et al., 2010). However, some studies dealing mainly with short-to mid-term drought stress show little to no relationship between g m and g sw (Galle et al., 2009;Resco et al., 2009), thereby suggesting a possible threshold drought response of g m in some species (see, for example, Duan et al., 2011).
Because K leaf is a measure of the ease with which liquid water perfuses through the leaf (Sack and Holbrook, 2006), and since part of this pathway (the mesophyll apoplasticsymplastic-transcellular pathway) is shared with CO 2 diffusing towards the chloroplast as bicarbonate ion and dissolved gas , one would expect a functional linkage between g m and K leaf , be it in the form of the hydration status of the cell, or even perhaps through convergence of aquaporin-mediated membrane transport (Maurel et al., 2008;Uehlein et al., 2012). Aasamaa et al. (2005) showed that K leaf increases with mesophyll density, a trait known to influence g m Fleck et al., 2010). Moreover, they suggested that membrane-associated traits such as membrane permeability, or simply the frequency of plasmalemmas and tonoplasts in the hydraulic flow path of the mesophyll, were most influential on K leaf . The possibility that membranes also exert strong control over g m is currently a much debated subject (reviewed in Kaldenhoff, 2012). Recently, Ferrio et al. (2012) suggested that the relationship between g m and K leaf would only be noticeable below a certain threshold. Presumably, the decrease of K leaf and g m as a result of water stress would be due to an eventual reduction of plasma membrane H 2 O/CO 2 permeability, which in turn would divert symplastic water transport through the more tortuous apoplastic pathway (see also Ferrio et al., 2009).
Despite the plausibility of the hypothesis of Ferrio et al., recent modelling results by Tholen et al. (2012) offer perhaps a simpler alternative explanation. Indeed, they have theoretically demonstrated that a threshold-like decrease of g m with respect to g sw in response to drought or salinity stress is expected merely from the relative increase in photorespiratory CO 2 release ensuing from the eventual reduction of the intercellular CO 2 mole fraction (C i ) at low enough g sw . Thus, the apparent threshold-like response of g m to drought could merely be an artefactual consequence of a reduction of C inot a reduction of, say, cell membrane permeability as Ferrio et al. (2012) suggested.
To examine possible functional links between K leaf , g sw , and g m , as well as to verify the possible occurrence of a thresholdlike response of g m with respect to the other two leaf variables, the time course of leaf gas exchange was determined in parallel to K leaf and g m in hybrid poplars of contrasting drought tolerance during a short-term drought. It was also determined whether a reduction of stomatal limitation via a low CO 2 treatment affects the short-term response of g m after 1 week of reduced irrigation. To avoid any artefactual CO 2 effects on g m during measurements, care was taken to conduct the latter under constant C i . It was hypothesized that g m , thus evaluated at a normalized C i , would remain relatively constant and not display a threshold-like response in relation to g sw and K leaf , and that low CO 2 treatment at the end of the drought period would increase g sw in the short term, but not g m and possibly not K leaf .

Plant material and growing conditions
Frozen cuttings of Assiniboine [(Populus×'Walker':Populus deltoides L.×P.×petrowskiana R.I. Schrod. ex Regel)×male parent unknown] and Okanese [(P.×'Walker')×P.×petrowskiana], with low and high putative drought tolerance, respectively (Silim et al., 2009), were thawed in water for 24 h and sprouted in a mist chamber for ~3 weeks in a 2:1 perlite:Pro-mix growing medium (Premier Horticulture, Rivière-du-Loup, QC, Canada). Plants were transferred to a 2:1 coarse sand:Pro-mix soil in 7 l pots and moved to a greenhouse for ~3 weeks (23/21 °C day/night temperature, 55% relative humidity, 16 h photoperiod). When plants reached ~60 cm in height, they were transferred to a growth cabinet and acclimated for at least 7 d [22/17 °C day/night temperature, 60% relative humidity, 16 h photoperiod, 800 μmol m -2 s -1 photosynthetic photon flux density (PPFD)]. A 20-20-20 complete nutrient solution was applied twice a week for a total of 200 ppm N per week per plant. Experiments described later were carried out in the same growth cabinet under the same growing conditions.

Gas exchange and fluorescence measurement
Gas exchange measurements were carried out when plants had reached a leaf plastochron index (LPI) of at least 12 (Larson and Isebrands, 1971). Plants were measured between 9:00 h and 12:00 h with a LI-6400XT equipped with a 6400-40 Leaf Chamber Fluorometer (LCF, LI-COR Biosciences, Lincoln, NE, USA). Leaves were initially acclimated under 800 μmol m -2 s -1 PPFD, 25.0 ± 0.2 °C leaf temperature, 380 μmol mol -1 leaf chamber air CO 2 (C a ), 17.5 ± 0.5 mmol mol -1 water vapour mole fraction, and 250 μmol s -1 air flow rate. Net CO 2 assimilation rate at ambient conditions (A Ca=380 ), stomatal conductance to water vapour (g sw ), and intercellular CO 2 mole fraction (C i ) were recorded once gas exchange variables reached steady state. Photochemical efficiency of photosystem II (Φ PSII ) was estimated by recording steady-state fluorescence (F s ) and maximal fluorescence under a saturating flash (F m ′) of ~10 000 μmol m -2 s -1 using: (1)

Mesophyll conductance measurements and calibration
The variable J method  was used to estimate g m using: where J f is the photochemical electron transport rate estimated from chlorophyll fluorescence, Γ* is the chloroplastic CO 2 photocompensation point, and R d is the non-photorespiratory mitochondrial respiration in the light, which was estimated from the intersection point of two detailed A-C i curves performed at 500 and 150 μmol m -2 s -1 PPFD. For these 'Laisk method' measurements of R d , a larger 2 × 3 cm leaf chamber and an air flow rate of 500 μmol s -1 were used to minimize chamber leaks (LI-COR, 2012). Since the intersection point of the A-C i curves of the Laisk method can only give an apparent intercellular CO 2 photocompensation point (C i *), this value is usually converted to its chloroplastic Γ* equivalent assuming (von Caemmerer et al., 1994;Ethier and Livingston, 2004): However, this relationship assumes that the great majority of the gross intracellular CO 2 flux crossing the chloroplast envelope also experiences the diffusional resistance due to the cell wall and plasmalemma (r wp ). As explained in detail in Tholen et al. (2012), this is very unlikely, especially around Γ* where the net outgoing flux R d crossing the wall/plasmalemma is typically only ~20% of the intracellular flux entering the chloroplast. Under such conditions, Equation 3 may be re-written as: where r ch is the diffusional resistance due to the chloroplast envelope and stroma. Given that current estimates put r ch ≥ r wp in nonsclerophylls Tholen and Zhu, 2011;Tholen et al., 2012), it is clear that C i * is probably greater than Γ*, rather than smaller as Equation 3 has it (see Tholen et al., 2012 for further discussion). Given such uncertainties, it was chosen here to estimate Γ* in hybrid poplar from previous in vitro determinations of the specificity factor (S c/o ) of Rubisco purified from other woody deciduous species (S c/o =100-104; see Balaguer et al., 1996;Bota et al., 2002), which on average corresponded to a Γ* of ~38 μmol mol -1 at 21% O 2 . This Γ* value appears in accordance with Equation 4 as it is below the average C i * value found here for the poplar clones (40.3 μmol mol -1 ); it also compares very well with the popular tobacco Γ* values commonly used to model photosynthesis and estimate g m (see von Caemmerer et al., 1994;Bernacchi et al., 2002). Furthermore, Galmès et al. (2006) have shown the unreliability of in vivo Γ* estimates under water stress conditions, whereas leaf respiration-in particular dark leaf respiration (R n )-appeared less affected by drought. Consequently, the Laisk method was only used to estimate R d and establish a relationship with R n under wellwatered conditions, and then this relationship was used to estimate R d from R n measurements in all subsequent experiments (see next section).
Calibrations of the relationship between the photochemical electron transport rate estimated from fluorescence measurements (J f ) and gas exchange (J CO2 ) were carried out as described by Hassiotou et al. (2009). Briefly, CO 2 response curves were carried out under non-photorespiratory conditions (1% O 2 ) at a PPFD of 1000 μmol m -2 s -1 by progressively increasing C a from 300 μmol mol -1 up to saturating CO 2 (between 800 and 1000 μmol mol -1 ), which were followed by photosynthetic light responses curves, decreasing PPFD from 1500 to 500 μmol m -2 s -1 (Fig. 1). This allowed the calculation of J f as: where s is the slope and c the intercept of the relationship of Φ PSII and Φ CO2 at 1% O 2 , Φ CO2 being calculated as: Calibrations carried out on water-stressed plants resulted in similar regressions; hence, data from both well-watered and stressed plants were pooled in the final (Fig. 1).
Nevertheless, such calibration of J f under non-photorespiratory conditions did not always produce acceptable g m values (see Gilbert et al., 2012 for a detailed discussion of this issue), as many were negative or too high, particularly when the plants were under water stress. Consequently, an alternative method, in the authors' view more likely to produce more reliable estimates of s under 21% O 2 , was evaluated. It is based on the premise that at 21% O 2 , g m estimates should in theory vary little over a range of C i values falling around the RuBP-limited A-C i curve region where Φ PSII is found to be constant (Tholen and Zhu, 2011). It was confirmed that this Fig. 1. Relationship between the photochemical electron transport rate estimated from fluorescence measurements (J f ) and gas exchange (J CO2 ) under 1% O 2 . Photosynthetic response curves to light (A/Q) and CO 2 (A/C i ) were carried out on three plants per clone, and the slope (s) and intercept of the relationship between Φ CO2 and Φ PSII were used to calibrate J f (see Equation 5). A/C i data were pooled together for clarity. Water-stressed plants were also measured and were in a similar range to wellwatered plants. The dotted lines indicate the range of Φ PSII values observed during the short-term soil drying experiments (see Fig. 3 theoretically based assumption held for poplar using g m measurements obtained with the standard isotopic method of Evans et al. (1986) (GE and SP, unpublished;see also Supplementary Fig. S1 available at JXB online). Thus, for this alternative J f calibration method, detailed A-C i curve analysis around the aforementioned C i region of interest combined with chlorophyll fluorescence measurements under 21% O 2 was used. s and g m were then fitted over the C i range of interest using the RuBP-limited photosynthesis equation given in Ethier et al. (2006) (see their equation 3) in order to minimize the sum of squares of errors. This method hence solves s (and so J f ) and g m simultaneously using the measured A-C i , Φ PSII , Γ*, and R d as inputs.
Net CO 2 assimilation and leaf transpiration rates measured with the LI-6400XT over the aforementioned C a range were corrected for leaks following the method of Flexas et al. (2007a). Briefly, this consisted of subtracting the apparent CO 2 assimilation rate of a dead leaf (immersed in boiling water for 5 min) estimated at the corresponding leaf chamber CO 2 concentration and air flow rate, taking care to allow the leaf to become fully equilibrated with the surrounding water vapour concentration of the chamber (set to 17.5 mmol mol -1 ). For H 2 O leak corrections, the apparent transpiration rate of the empty leaf chamber recorded at 17.5 ± 0.5 mmol mol -1 water vapour concentration and appropriate air flow rate was subtracted from measured leaf transpiration rates.

Post-calibration estimation of R d from dark respiration measurements
Because during the short-term drought experiments the water potential of each leaf used for gas exchange needed to be subsequently rapidly measured with a pressure chamber to determine leaf hydraulic conductance (see 'Leaf hydraulic conductance' below), it was necessary to use a rapid proxy method to estimate R d immediately following the leaf water potential (Ψ leaf ) determination. For this, the leaf petiole was placed in a water-filled beaker and recut under water before clamping the leaf to a LI-6400 equipped with a 2 × 3 cm leaf chamber, then covering it with a dark cloth to measure its R n continuously for ~20 min. To convert from R n to R d , the following relationship previously established on the plants used for J f calibrations was used: Dark respiration was measured at a leaf chamber C a set equal to the room C a to avoid CO 2 leakage effects. It was established that pressurization of leaves prior to the determination of dark respiration rates across a range of Ψ leaf (-0.4 to -1 MPa) had no impact on R n compared with that of unpressurized leaves (data not shown).

Short-term drought
Gas exchange of three plants per clone was repeatedly measured over a 12 d period of soil drying, for a total of nine measuring days. Plants were first measured under well-watered conditions, and the irrigation was then reduced on the following days to achieve a wide range of g sw values. Measurements were carried on the same newly mature leaf (LPI=5 on day 1) throughout the experiment. Soil water potential (Ψ soil ) was monitored on each measurement day using custom-made tensiometers and a digital reader (Tensimeter, Soil Moisture Equipment, Santa Barbara, CA, USA).
In addition, to get a better picture of the time course of A, g sw , g m , and K leaf , continuous measurements of gas exchange and plant and soil water potentials were performed on one plant per clone during 5 d of soil drying. Gas exchange was monitored during daytime with a LI-6400XT equipped with a 2 × 3 cm adaptor chamber for a PAM-2000 chlorophyll fluorometer probe (Heinz Walz GmbH, Effeltrich, Germany), set at a flash intensity above 6000 μmol m -2 s -1 . Through Equation 7, R d was assumed to follow the observed decreasing trend of R n as drought proceeded, dropping by ~40-50% after 5 d (from an estimated average initial R d of ~1.1 μmol m -2 s -1 ). Soil water potential was measured using one tensiometer connected to a datalogger (CR7, Campbell Scientific, Logan, UT, USA). A leaf psychrometer (L-51, Wescor, Logan, UT, USA) was attached on an adjacent leaf to the one used for gas exchange to measure Ψ leaf , while another was attached to an adjacent dark covered leaf to measure Ψ stem . Psychrometer chambers were lightly lined with petroleum jelly to establish a good contact with the leaves (Savage et al., 1983;Campbell and McInnes, 1999). Destructive Ψ leaf and Ψ stem measurements using a pressure chamber were carried out during the time course to correct the psychrometer readings according to a relationship between spot scale (psychrometer) and leaf scale (pressure chamber) water potential measurements established a priori (Ψ leaf =1.39 Ψ psychrometer , n=96, R 2 =0.78; Ψ stem =0.45 Ψ psychrometer , n=113, R 2 =0.54). To maintain a good contact between the psychrometers and the leaves, the psychrometers were removed daily, inspected, and cleaned if necessary, before resetting them to the leaves. Furthermore, the dark cover for Ψ stem was removed daily after the final gas exchange measurements, ~4-5 h before lights off.
Reduction of drought-induced stomatal limitation using low C a Reducing air [CO 2 ] to very low concentrations around drought-or salt-stressed leaves leads to the opening of stomata to levels that result in a significant reduction of diffusional limitations (Centritto et al., 2003;Bunce, 2007). This technique was thus used on waterstressed leaves to decrease stomatal limitation and monitor changes in A, g m , and leaf hydraulics.
Prior to soil drying and subsequent low C a treatment, gas exchange of one leaf (LPI 5-7) was measured under well-watered conditions. Every other day, gas exchange parameters of the same leaf were measured to monitor changes in g sw and to adjust irrigation based on a g sw reduction of ~66% from the well-watered conditions. Soil water content (15 cm TDR probe and Cable tester 1502C, Tektronix, Beaverton, OR, USA) and Ψ soil were monitored daily to adjust irrigation between gas exchange measurements. Well-watered plants were also monitored as comparison.
After 7 d of reduced irrigation, the low C a treatment was initiated. Gas exchange was measured first under ambient C a . The plant was then inserted in an ~13 l cylindrical cuvette designed to control the CO 2 and H 2 O mole fractions around the plant. Between four and seven leaves, including the previously measured leaf, could be inserted in this large plant cuvette, while the upper and lower parts of the plant (~2-4 leaves each) remained under ambient growth chamber atmosphere (see Supplementary Fig. S2 at JXB online for a full description of the cuvette). The target leaf was clamped with the LI-6400XT LCF chamber inside the 13 l plant cuvette and both were sealed. The large plant cuvette operated as an open system, with an inflow of ~12 l min -1 . The plant cuvette CO 2 mole fraction was decreased to ~75 μmol mol -1 and the H 2 O mole fraction was maintained at ~17 mmol mol -1 by circulating the cuvette air through a H 2 O scrubbing Drierite™ column in series with a CO 2 scrubbing column. A LI-840 CO 2 /H 2 O Analyzer (LI-COR Biosciences, Lincoln, NE, USA) was used to monitor the plant cuvette gas concentrations. Leaf temperature was maintained at 25 ± 0.2 °C in the leaf chamber. Air temperature in the plant cuvette was ~26 °C, ~1 °C higher than the ambient air temperature outside the cuvette.
After g sw had peaked and stabilized, the CO 2 mole fraction was increased in the leaf chamber to reach a C i of ~275 μmol mol -1 (see chlorophyll fluorescence calibration in the Results section) and C a in the large plant cuvette was increased in parallel to match the C a established in the leaf chamber. Once leaf gas exchange had reached steady state, three consecutive measurements were taken, each at least 2 min apart.

Leaf hydraulic conductance
For both the short-term drought and the low C a experiment, K leaf was estimated using the evaporative flux method: where E is leaf transpiration measured using the LI-6400XT, and Ψ stem (stem water potential; measured on a dark-adapted leaf) and Ψ leaf are measured using a Scholander-type pressure chamber (Model 610, PMS instruments, Albany, OR, USA) or the leaf psychrometers.

Vulnerability curves
Vulnerability curves were carried out using a low-pressure flow meter (LPFM; Sperry et al., 1988). Shoots were cut from plants in the growth chamber, put in a plastic bag, and brought to the lab. After an equilibration period of at least 5 min, water potential was measured on a leaf. Stems were then cut under water, and stem segments of ~7 cm long were prepared based on previous work by Harvey and van den Driessche (1997). Initial stem hydraulic conductivity (K i ) was determined under a pressure head of 4.7 kPa by collecting water flowing out of the segment into a 10 ml pipette and measuring the increasing water column in the pipette with a pressure transducer connected to a datalogger (CR23X, Campbell Scientific). When steady state was reached, stem segments were flushed at a pressure of 120 kPa for 7 min to remove emboli before measuring maximum hydraulic conductivity (K max ). Both flushes and conductivity measurements were carried out using a KCl (0.1 M) solution filtered at 0.2 μm (Cai and Tyree, 2010). To reach more negative Ψ stem values, shoots were left to dry in the lab and stem segments were measured upon reaching a desired value. Percentage loss conductivity (PLC) was calculated as: The vulnerability curves were constructed by plotting PLC against Ψ stem , and then fitting a Weibull function: where b and c are fitting constants. Fitting was carried out using the nls function in R 3.0.0 (R Core Team, 2013).

Statistical analysis
Statistical analyses were carried out with R 3.0.0. Mixed models were used to determine the hierarchical structure (i.e. plants within a measurement cycle) using the nlme procedure (Pinheiro et al., 2013). Initial rates of leaf gas exchange were compared between clones to examine clonal differences. For the low C a experiment, final measurements were analysed using a two-level factorial design (clone×low C a treatment).

Calibration of the chlorophyll fluorescence method for estimating the photochemical electron transport rates
Estimates of g m based on the calibration of J f under 1% O 2 exhibited a positively skewed shape when plotted against C i , with a narrow maximum at ~200 μmol mol -1 (Fig. 2). However, using the proposed alternative calibration method at 21% O 2 , g m estimates increased slightly and were more stable for several hundreds of ppm above a C i of 200 μmol mol -1 (Fig. 2). Thus, in order to improve the estimates of g m as C i decreases under soil drying conditions, C a was adjusted in the leaf chamber to reach a normalized C i of 275 μmol mol -1 (hereafter named C i =275), a value consistently identified as the minimal C i required to reach a stable and maximal g m .
Using a normalized C i value resulted in <5% of g m estimates being discarded (i.e. negative or too high values) compared with 15% when measuring at ambient C a (380 μmol mol -1 ).
Overall, there was a slight difference in s values between clones (see Fig. 1) and the proposed calibration method yielded s estimates that were ~5% lower than those found under 1% O 2 (Fig. 2).

Gas exchange responses to short-term drought
Drought-sensitive Assiniboine and drought-tolerant Okanese had, under well-watered conditions, a similar photosynthesis rate under ambient C a (A Ca=380 ; F 1,17 =2.11, P=0.16), and a similar normalized g m_Ci=275 (F 1,16 =2.93, P=0.11), although g sw was slightly higher in Assiniboine (F 1,17 =3.62, P=0.07; Fig. 3). Twelve days of reduced irrigation gradually decreased photosynthesis (A Ca=380 ), g sw , transpiration (E), K leaf , and leaf and soil water potential (Ψ leaf and Ψ soil ) to very low values (e.g. a 90% decrease in g sw ; see Figs 3 and 4). The decline in Ψ leaf , E, and K leaf was concomitant with g sw , while the decline in A lagged behind K leaf . Normalized g m_Ci=275 remained constant over most of the drought-induced g sw range, and decreased only when g sw fell below a threshold of ~0.15 mol m -2 s -1 (i.e. 68% and 56% of the maximum g sw for Assiniboine and Okanese, respectively). However, because too low or near zero g sw led to difficult if not impossible g m fits, g m_Ci=275 values below 0.05 mol m -2 s -1 were not included in the analysis as scatter increased  tremendously below that threshold, even using the proposed calibration method.
Although the drought response of A Ca=380 , K leaf , and g m_Ci=275 appeared similar for both clones when plotted against g sw , for any given Ψ soil , g sw of Okanese was always lower than that of Assiniboine. Hence, for Okanese, the minimal g sw was observed at a Ψ soil of -0.03 MPa, compared with -0.06 MPa for Assiniboine (Fig. 4). Assiniboine reached higher g sw values under well-watered conditions, but this did not translate into an increase in A Ca=380 , and both clones exhibited a similar decline in A Ca=380 at common g sw . Similar gas exchange versus leaf hydraulic relationships to the 12 d reduced irrigation experiment were obtained over 5 d of water depletion during which A Ca=380 , g sw , K leaf , and g m (C i =225-275 μmol mol -1 ) were recorded continuously. Once again, g sw and K leaf decreased in concert, the decline in A Ca=380 lagged behind the two previous variables, and g m remained constant up to a g sw threshold of ~0.15 mol m -2 s -1 (Fig. 3; see also Supplementary Fig. S3 at JXB online).

Lowering C a to decrease drought-induced stomatal limitation
After 7 d of reduced irrigation, g sw was decreased by 75% on average in Assiniboine and 69% in Okanese. In both clones, lowering C a around leaves resulted in a 3-to 4-fold increase in g sw and E, which translated into a significantly more negative Ψ leaf (Table 1). When C a was reduced, g sw reached a new steady state within 20-40 min for water-stressed plants, and after 50-75 min in well-watered plants (Fig. 5). Furthermore, g sw rapidly declined after a return to a higher C a in severely stressed plants (white dots in Fig. 5), while well-watered plants showed a more progressive return towards their initial g sw level. In contrast to g sw and Ψ leaf , no significant changes in K leaf , A Ci=275 , and g m_Ci=275 were observed under low C a . There were no clonal differences in gas exchange responses  to low CO 2 treatment under reduced irrigation, although Assiniboine exhibited higher K leaf than Okanese (Table 1).

Vulnerability curves
Since each clone exhibited a similar A versus g sw response to drought, as well as a similar response to low C a exposure, vulnerability to cavitation was measured to determine further the clonal differences in water stress response. Droughttolerant Okanese exhibited a higher reduction in K stem over the observed Ψ stem range compared with Assiniboine (Fig. 6). For instance, the PLC was up to 50% in Okanese upon reaching a Ψ stem of -1 MPa, while Assiniboine had ~15% PLC at the same potential. However, PLC increased rapidly after -1.2 MPa Ψ stem in Assiniboine to reach similar PLC values to Okanese near -1.5 MPa (Fig. 6). Nevertheless, values of Ψ leaf observed throughout the experiment were on average between -0.6 MPa and -0.9 MPa (first and third quartiles) for both clones, and Ψ stem ranged between -0.3 MPa and -0.6 MPa. Thus, both clones usually did not reach a PLC of 50% but suffered only ~25% loss of conductivity under the moderate drought conditions of this study.

Methodological considerations on the estimation of g m
Potential methodological errors and the underlying assumptions necessary in the estimation of g m using the variable   5. Time courses of stomatal conductance (g sw ) when the upper foliage of well-watered and water-stressed Assiniboine plants was exposed to a reduction in ambient CO 2 concentration (C a ~75 μmol mol -1 ). Different levels of water stress were examined, expressed here as the percentage of g sw relative to their initial well-watered state (g sw_ini ) evaluated 7 d prior to the low C a treatment (black circles, well-watered, where g sw increased compared with g sw_ini ; dark grey circles, 44% of g sw_ini ; light grey circles, 15% of g sw_ini ; and white circles, 6% of g sw_ini ). The short arrows indicate when C a was increased from ~75 μmol mol -1 to a CO 2 concentration resulting in a C i of ~275 μmol mol -1 . The data points to the right of the arrows were measured at that C i , showing the high sensitivity of severely stressed plants to the return to near ambient C a .
J method were discussed initially by Harley et al. (1992), and subsequently elaborated upon by several other authors (e.g. Warren 2008;Pons et al., 2009;Gilbert et al., 2012). Theoretically, positively skewed g m versus C i curves like those observed in this study when following the standard 1% O 2 J f calibration protocol (see also Flexas et al., 2007b;Hassiotou et al., 2009;Gilbert et al., 2012) are predicted to result from small overestimations of J f , and/or possibly from underestimations of Γ*, or of R d (see Harley et al., 1992;Gilbert et al., 2012). However, Tholen et al. (2012) demonstrated that even if all relevant parameters in the variable J method equation were exact, g m would still be expected to decrease apparently sharply when the photorespiratory CO 2 flux rises at low C i . One way to alleviate this artefactual CO 2 sensitivity problem is by measuring g m at a normalized C i , preferably around the g m versus C i curve region where g m is expected to peak (i.e. the narrow C i range where estimates of g m are expected to be least affected by small errors of J f , etc.; see Fig. 2). However, even after taking this precaution, it was noted that ~5% of the initial g m estimates for leaves subjected to water stress were deemed unreliable. One possibility that could explain this poor model performance is that the 1% O 2 calibration method incompletely accounts for alternative electron sinks under 21% O 2 ; especially under water stress (Makino et al., 2002;Kitao et al., 2003;Laisk et al., 2006). In an attempt to address this possibility, we tested an alternative J f calibration method performed under 21% O 2 , and which solves s and g m concurrently from the CO 2 response of photosynthesis and Φ PSII in the RuBP-limited A-C i curve region where photorespiration is expected to be low enough to exert little influence on g m (see Fig. 2, and further details in the Materials and methods). From the present assessment, this alternative J f calibration procedure yielded slightly lower photochemical electron transport rates than the 1% O 2 calibration method, which resulted in ~20% higher g m estimates, but did not improve the rejection of data under water stress conditions. Nevertheless, it is believed that calibrating Φ CO2 under 21% O 2 should better account for potential alternative electron sinks under normal atmospheric conditions while greatly simplifying the calibration procedure.
To evaluate possible g m inaccuracies introduced by errors in the estimation of Γ* and R d , an analysis of the sensitivity of g m to changes in Γ* (±10%) and R d (±50%) was performed. As already pointed out by others (e.g. Harley et al., 1992;Pons et al., 2009), g m was significantly more sensitive to potential errors in Γ* in comparison with R d (see Fig.7)-in the present simulations, a 120% increase in R d was necessary to obtain a g m value corresponding to a +10% change in Γ*. However, because the estimation of g m was restricted to a narrow C i range neighbouring 275 μmol mol -1 , such errors in Γ* and/or R d are not expected to change the conclusions regarding the delayed threshold response of g m with respect to g sw (see Fig. 7).

Threshold response of g m to drought
Drought stress usually decreases A, g sw , and K leaf concomitantly (Brodribb and Holbrook, 2006;Brodribb et al., 2007;Warren, 2008;Galle et al., 2009;Barbour et al., 2010). Drought stress has also been shown to decrease g m monotonically with respect to A and g sw (Flexas et al., 2007b(Flexas et al., , 2008Warren, 2008;Perez-Martin et al., 2009), although Flexas et al. (2002) reported a lower rate of g m decrease at g sw >0.15 mol H 2 O m -2 s -1 in grapevine. However, these declines in g m may only be circumstantial since a recent study has shown that g m is expected to decrease apparently sharply in the absence of any change in intraleaf diffusional resistance component, once the photorespiratory rate of a waterstressed leaf rises significantly due to stomatal closure and ensuing reduction in C i (Tholen et al., 2012). Measuring g m at  a normalized C i alleviates this effect, which allowed detection of a threshold response of g m with respect to g sw (~0.15 mol H 2 O m -2 s -1 ), thereby confirming the observations of Ferrio et al. (2012). This recent study demonstrated a strong linear relationship between g m and K lamina (hydraulic conductance of the leaf lamina), but only when K lamina fell below 8 mmol H 2 O m -2 s -1 MPa -1 (~70% of maximum value) as a result of water stress. Above this threshold point, corresponding to g m values ≥0.2 mol m -2 s -1 , the relationship did not hold, as the effective water path length influencing K lamina reached a constant baseline value. Hence, the g m -K lamina sensitivity region corresponded to the point where the water path length between the xylem and the evaporative surface increased significantly above this baseline value.
The small difference in g m threshold between the study of Ferrio et al. and the present study could arise from the different species and their respective hydraulic connectivity, but also because a constant C i was used for g m measurement with the variable J method . Using a normalized C i improved g m estimates, as they reached similar values between clones. Nonetheless, in both studies, g sw responded rapidly to early signs of water stress, and g m followed afterwards. Concomitantly, Φ PSII , although not an ideal proxy for g m , has shown a delay in decrease as Ψ leaf declined, beginning its descent often past a g sw reduction of 50% or even 80% (Brodribb and Holbrook, 2003), as observed in the present study (Fig. 3). Hydraulic compartmentalization of the mesophyll might have contributed to this delayed response of g m . Two factors are important when considering compartmentalization: the linkage between the xylem and the epidermis, and the degree of uncoupling of the mesophyll cell to the hydraulic apparatus (Zwieniecki et al., 2007). Thus, stomata, sensing early signs of water stress through small changes in Ψ leaf , could have started to close while more negative Ψ leaf values (i.e. more severe stress) may have been needed to initiate a response in mesophyll cells or chloroplasts, leading to a decrease in g m . Zwieniecki et al. (2007) explained that such hydraulic compartmentalization would allow mesophyll cells to be buffered against short-term changes in leaf water status, a beneficial characteristic to C-assimilating cells. Also, the bulk of mesophyll cells have been shown to be rather unresponsive to exogeneously applied abscisic acid (ABA) (Shatil-Cohen et al., 2011), which reinforces the idea that mesophyll cells are somewhat isolated from the main hydraulic path linking the xylem conduits and their bundle sheath extensions to the epidermis and stomata. Such structural considerations may act in conjunction with membrane-associated mechanisms such as gating of aquaporin channels and ensuing reduction of the symplastic water transport (Cochard et al., 2007;Almeida-Rodriguez et al., 2010;Pou et al., 2013), which might eventually divert more water to the more tortuous apoplastic pathway (Ferrio et al., 2012). The above model appears to fit well the hydraulic response of the hybrid poplar leaves despite the fact that it makes little mention of xylem cavitation. Certainly, the present results show that Ψ stem did not reach low enough values to induce substantial stem cavitation (see also Secchi and Zwieniecki, 2010), but this does not necessarily mean that significant leaf cavitation could not have occurred. Indeed, Sack et al. (2004Sack et al. ( , 2005 reported that the xylem-related component of total leaf resistance dominates in highly transpiring plants such as hybrid poplar, and recently Johnson et al. (2012) demonstrated that in some species leaf xylem cavitation may account for most of the decrease in K leaf under drought.
Exposure to low C a decouples g s and g m Centritto et al. (2003) proposed a method to separate diffusional limitations from non-diffusional limitations in saltstressed plants: they lowered C a to force stomata to open, thus reducing stomatal limitation upon returning to ambient C a . Surprisingly, the low C a treatment appeared to reduce the mesophyll limitation just as rapidly as the stomatal limitation. However, the C i values at which Centritto et al. (2003) evaluated g m after pre-conditioning the stressed plants to low C a was appreciably higher than before the low C a treatment. Thus, because g m was not evaluated at a normalized C i in this study, it is likely that the conclusions of Centritto et al. (2003) about the rapid response of g m to salt stress were confounded by the low C i photorespiratory artefacts described in Tholen et al. (2012). In the present case, C i was controlled close to 275 μmol mol -1 throughout and no significant change in g m in response to low C a pre-conditioning was detected.
Apart from its obvious consequences on gas exchange, low C a treatment has major consequences for the leaf water status. As in the present study, Bunce (2007) reported a decrease (more negative) in Ψ leaf using low C a on a variety of water-stressed herbaceous and woody plants. Interestingly, the hybrid poplar plants used here reached a common Ψ leaf value (approximately -0.92 MPa to -1.01 MPa) under low C a regardless of drought intensity, which suggests that the degree of stomatal opening could have been modulated to keep Ψ leaf above a catastrophic cavitation level (see also the results of Tardieu and Simonneau, 1998). If leaf xylem cavitation and concomitant decrease in Ψ leaf were responsible for the decline in K leaf under drought, K leaf would have been expected to decrease more as a result of the low C a treatment since Ψ leaf was further reduced (see Table 1 compared with Supplementary Fig. S4 at JXB online). However, it did not, and similar observations by Bunce (2007) led him to question the dominant role of leaf cavitation in drought-induced K leaf decline. One recent alternative explanation is a possible dual regulation of g sw and K leaf by ABA under soil drying conditions (Pantin et al., 2013). Indeed, in the present study, water-stressed plants exhibited greater stomatal sensitivity to changes in C a than well-watered plants (Fig. 5), a trait normally associated with the presence of higher levels of ABA (Raschke, 1975;Bunce, 2007). Cochard et al. (2007) reported that aquaporin expression in the leaf vein bundle sheaths could be a major control point on leaf water status, observing a substantial drop in Ψ leaf when aquaporin activity was inhibited. Shatil-Cohen et al. (2011) have also shown that either drought or exogenous application of ABA triggered a reduction of the permeability coefficient (and thus K leaf ) of bundle sheath cells, but not that of the mesophyll cells. As previously mentioned, this suggests a partial hydraulic isolation of mesophyll cells from the xylem-bundle sheath-stomata pathway, thereby possibly delaying ABA delivery. Since certain aquaporins have been reported to have a CO 2 channel role (Otto et al., 2010;Uehlein et al., 2012), it may be that g m is less sensitive to drought stress than g sw or K leaf because regulation of mesophyll aquaporins is expected to take place at a more advanced drought stress due to hydraulic compartmentalization.
Delayed g m response to drought favours water use efficiency in hybrid poplar One aim of the present study was to compare two clones of contrasting drought tolerance in order to determine if they differ in their g sw , K leaf , and g m relationships. Both droughttolerant Okanese and drought-sensitive Assiniboine exhibited a similar g m threshold response to drought (Fig. 3), potentially suggesting a partial hydraulic isolation of the mesophyll cells. This in turn allowed a delay in the decline of A, thereby favouring a higher water use efficiency (WUE) during the early stages of drought. However, those clones differed in their vulnerability to stem cavitation and in the Ψ soil at which their respective minimal g sw was reached (Fig. 4), as observed in other hybrid poplar studies (Silim et al., 2009;Arango-Velez et al., 2011;Larcheveque et al., 2011). Although the vulnerability curves in the present study showed that Okanese allowed more cavitation to occur at similar Ψ stem values than Assiniboine (Fig. 6), both clones modulated g sw to maintain Ψ stem at levels preventing conductivity to drop below ~25%, a feature previously observed for those same clones under severe drought stress (Arango-Velez et al., 2011). While Okanese maintained similar A Ca=380 and g m to Assiniboine at a given g sw , it adopted a more conservative water use strategy by lowering g sw more than Assiniboine under increasing soil water depletion (see also Arango-Velez et al., 2011). Hence, Okanese exhibited a higher intrinsic water use efficiency (WUE i ) than Assiniboine at comparable Ψ soil (it also displayed a higher g m /g sw ratio; data not shown). Such maintenance of high g m /g sw ratios has recently been identified as an important trait to improve WUE in cultivated plants (Barbour et al., 2010;Flexas et al., 2013).

Conclusion
When subjecting young hybrid poplar clones of contrasting drought tolerance to a short period of water stress, K leaf and g sw decreased monotonically in concert, while g m remained constant over most of its range, causing a delay in the decline of A. Only when g sw fell below an approximate threshold of 0.15 mol m -2 s -1 , ~33% (Assiniboine) to 40% (Okanese) of maximum g sw , was a decrease in g m observed (and this even when measured at a normalized C i to remove the photorespiratory bias modelled by Tholen et al., 2012), in accordance with the recent results of Ferrio et al. (2012). The present results thus support the recent suggestion that the bulk mesophyll is partially isolated (i.e. buffered) from the major transpiration pathway delivering ABA to the stomata (Shatil-Cohen et al., 2011;Pantin et al., 2013). The delayed photosynthetic decline mediated by such a g m threshold response resulted in enhancing WUE in the early stage of drought.

Supplementary data
Supplementary data are available at JXB online. Figure S1. Apparent sensitivity of mesophyll conductance (g m ) to intercellular CO 2 mole fraction (C i ): comparison of the responses observed under three different g m estimation methods. Figure S2. Schematic of the large plant cuvette used in the low C a experiment. Figure S3. Continuous measurements of photosynthesis (A), mesophyll conductance (g m ), stomatal conductance to water (g sw ), and leaf hydraulic conductance (K leaf ) during 5 d of water depletion. Figure S4. Relationship between leaf hydraulic conductance (K leaf ) and leaf water potential (Ψ leaf ) during 5 d of water depletion.