Sap fluxes from different parts of the rootzone modulate xylem ABA concentration during partial rootzone drying and re-wetting

Highlight Following drying and re-wetting events during partial rootzone drying irrigation, xylem ABA concentration was best explained by accounting for sap fluxes from both parts of the rootzone.


Introduction
Soil moisture varies considerably both temporally (due to rainfall events or irrigation of crop plants) and spatially (roots usually dry the surface soil layers while considerable moisture may be available at depth), and both plant fitness and crop productivity depend on the root system capturing sufficient water to sustain growth. There has been considerable agronomic interest in different irrigation techniques such as partial rootzone drying (PRD; Dry et al., 1996;Kang and Zhang, 2004) that explicitly aim to vary soil moisture within part or all of the rootzone. Compared to crops grown with conventional deficit irrigation (DI, where water is applied to the entire rootzone), crops grown with PRD (the alternate irrigation and drying of only part of the root system) had significantly higher yield in six (out of 15) experiments (Dodd, 2009), but the physiological mechanisms underpinning these responses remain elusive. PRD was conceived as an irrigation technique that aimed to alter root-to-shoot chemical signalling by drying part of the rootzone, thereby stimulating root synthesis of ABA and its subsequent transport to the shoot in order to partially close the stomata, thereby increasing leaf water use efficiency (Zhang and Davies, 1991;Stoll et al., 2000;Kang and Zhang, 2004). Thus, the increased crop yields of PRD plants compared with DI plants may be partially attributed to the impacts of the drying and rewetting cycles (that characterize PRD) on root-to-shoot ABA signalling (Dodd et al., 2006). However, relatively few studies have actually measured ABA concentrations in xylem sap or leaves of field-grown plants (but see Topcu et al., 2007;Rodrigues et al., 2008;Hutton and Loveys, 2011;Pérez-Pérez et al., 2012;Romero et al., 2012). Even under controlled environment conditions, PRD increased (Dodd, 2007), decreased (Dodd, 2007;Wang et al., 2012) or had no consistent effect (Wang et al., 2010(Wang et al., , 2012 on xylem ABA concentration compared with DI plants, perhaps due to the timing of measurements during the drying/ rewetting cycles. Taken together, these results suggest that the agronomic promise of PRD is unlikely to be consistently translated into improved crop water use efficiency in the field, unless irrigation managers can better predict its physiological effects. Consequently, laboratory studies with 'two root-one shoot' grafted plants (Dodd, 2007) determined the contributions of different parts of the root system to total sap flow and leaf xylem ABA concentration ([X-ABA] leaf ) during PRD, which better explained [X-ABA] leaf than assuming it was determined by total soil water availability (Dodd et al., 2008a(Dodd et al., , b, 2010. While these studies only exposed plants to a single soil drying cycle where the wet and dry parts of the rootzone remained the same (PRD-Fixed), in the field PRD usually alternates the wet and dry parts of the rootzone (Stoll et al., 2000;Romero et al., 2012). Soil drying and re-wetting cycles stimulated root growth (Mingo et al., 2004), enhanced soil nutrient availability (Wang et al., 2010), and altered root-sourced chemical signalling to the shoots by transiently increasing [X-ABA] leaf (Dodd et al., 2006;Romero et al., 2012). However, it was not clear whether this increase was due to the remobilization of ABA that had accumulated in the previously dried rootzone, or due to the drying of previously irrigated roots.
Measuring and modelling xylem ABA concentrations (Dodd, 2008;Dodd et al., 2008aDodd et al., , b, 2010Liu et al., 2008;Plauborg et al., 2010) is challenging, due to the methodological difficulties of collecting an authentic xylem sap sample. In field-grown crops, root xylem sap can only be sampled by de-topping the plant and collecting sap at relatively low flow rates compared with whole plant transpiration rate, which artificially increases root xylem sap ABA concentration, [X-ABA] root (Else et al., 1994). Consequently, many studies have collected xylem sap from detached leaves or stems, by measuring their water potential and then applying an overpressure (Jachetta et al., 1986;Dodd, 2007). While varying the overpressure applied to detached tomato leaves had minimal effects on [X-ABA] leaf , actual concentrations can be higher than [X-ABA] root , depending on both the accuracy with which root xylem sap flow rate is matched with transpirational flow rate (Dodd et al., 2008a;Netting et al., 2012) and/or a dilution of leaf apoplastic sap with symplastic contents during sap collection (Jachetta et al., 1986;Borel and Simonneau, 2002). For this reason, modelling xylem ABA concentrations of plants exposed to heterogeneous soil moisture may be more informative when a single xylem sap sampling methodology and/or site of xylem sap sampling (either root or leaf) is adopted.
This study aimed to predict the [X-ABA] leaf of plants exposed to temporal and spatial differences in soil moisture imposed by fixed or alternate PRD. Whereas previous models of [X-ABA] leaf (Dodd et al., 2008a(Dodd et al., , b, 2010 relied on direct measurements of root xylem ABA concentrations from different parts of the root system, this study developed a model that collected xylem sap only from leaves (and thus could be applied to field-grown plants). Initial experiments with 'two root-one shoot' grafted sunflower plants determined whether sap flow gauges and soil moisture sensors gave similar relationships between the fraction of water uptake by roots in drying soil and soil water content (θ). Then water was withheld from the entire rootzone of own-rooted tomato plants to determine the relationship between [X-ABA] leaf and θ, which was used to predict [X-ABA] root from measurements of θ by assuming that [X-ABA] remained constant in transit from roots to shoots. Finally, different models were used to predict [X-ABA] leaf of split-root tomato plants at different times in soil drying and re-wetting cycles during PRD. Whereas xylem ABA concentration during fixed PRD was related strictly to the soil water content of the irrigated rootzone, both drying and re-irrigated parts of the rootzone contributed following PRD alternation.

Materials and methods
Determining the contributions of different parts of the root system to total sap flow in 'two root-one shoot' plants exposed to a single PRD drying cycle The initial experiments used 'two root-one shoot' sunflower (Helianthus annuus L. cv. Tall Single Yellow) plants since the grafting procedure gave higher success rates in this species and because the cylindrical stems of sunflower were ideal for measuring sap flow from different root systems. Seeds (Moles Seeds, Essex, UK) were placed on two layers of filter paper (Whatman No. 1) moistened with distilled water in a covered Petri dish and germinated in the dark for 48 h. Five seedlings (typical radical length 20 mm) were placed each side of a vertical, watertight plastic partition in a 3.0 l pot (17 cm diameter, 13 cm high) filled with an organic loam (John Innes No. 2, J Arthur Bowers, Lincoln, UK) substrate with a gravimetric water content (θ) at a drained capacity of 0.63 g g -1 . The substrate was watered to drained capacity prior to seedling placement, then pots were placed in a plastic container, the top of the container covered with aluminium foil (to exclude light and promote hypocotyl extension), and the container placed in a walk-in controlled environment room. After one week, the aluminium foil was removed and the plants grown for a further two weeks before 'two root-one shoot' grafting was implemented with uniform seedlings, as described previously (Dodd, 2007). A plastic bag was secured around the pot base with a rubber band, and the grafted plants, which resembled an inverted 'Y', were allowed to establish for 2 weeks. When the plastic bags were removed, only one grafted plant was allowed to grow in each pot. Plants were watered daily and allowed to grow for a further two weeks prior to experiments.
Prior to sap flow measurements, the entire pot was watered to drained capacity. Sap flow through each hypocotyl (below the graft union) was measured using the heat balance technique with commercially available sensors (Model SGA-5, Dynagage ® , Dynamax Inc, Houston, TX, USA) suitable for stems of 5-7 mm diameter, which was directly measured above the sensor after installation. Sensor installation and operation were according to the manufacturer's instructions (Dynagage, 2005). Foam (15 mm thick) and aluminium foil shielded the sensors from direct radiation. Power input to the heater was constant for all measurements. Sap flow was recorded every 10 s and averaged over 5 min using a datalogger (Model DL2e, Delta-T Devices, Cambridge, UK). For each hypocotyl, Qf (heat loss by convection by the sap; Fig. 1A) was expressed as a fraction of total Qf (Fig. 1C). Previous measurements established that total Qf (the sum of both hypocotyls) was temporally correlated with whole plant transpiration measured gravimetrically by placing the plant on a balance (Dodd et al., 2008a).
After fitting the sap flow gauges, two theta probes (Model ML2X, Delta-T Devices) with 6.5 cm pins were placed vertically into the top of each soil compartment to measure θ, and measurements recorded every 5 min using a datalogger (Model DL2e, Delta-T Devices). During the experiments, water was supplied to one soil compartment (Figs 1B, 3B) to ensure that θ exceeded 0.3 g g -1 , corresponding to a soil matric potential of -0.1 MPa (Dodd et al., 2006).

Modelling leaf xylem ABA concentration of own-rooted plants following PRD alternation
Subsequent experiments used tomato (Solanum lycopersicum Mill. cv. Ailsa Craig) as it was easier to collect sufficient xylem sap from individual detached tomato leaves. Seeds (Moles Seeds, Essex, UK) were individually sown in a well-watered peat-based substrate (Levingtons M3, Levington Horticulture Ltd., Ipswich, UK) in seedling trays, with a single seed in each separate compartment (30 mm deep×20 mm×20 mm). After 14 d, when the first true leaf had emerged, seedlings were transferred to 'net pots' (Teku, Pöppelmann Plastiques, Pöppelmann, France) of 50 mm diameter×50 mm deep with 5 mm×7 mm pores in the sides. After one more week, plants in the 'net pots' were transplanted into custom-made 3.0 L split pots (17 cm diameter, 13 cm high). A vertical, watertight plastic partition separated two halves of the 3.0 l pot, with a gap (50 mm deep×50 mm wide) in the centre of the partition to allow each 'net pot' to be inserted into the substrate, minimizing seedling disturbance. Both 'net pots' and split pots were filled with an organic loam (John Innes No. 2, J Arthur Bowers, Lincoln, UK) and watered daily until each experiment commenced.
Several batches of tomato plants were produced as described above, with an initial experiment aiming to define relationships between [X-ABA] leaf and whole pot θ and leaf water potential (Ψ leaf ). Eight weeks after the seeds were planted, different irrigation treatments were applied by withholding irrigation from the entire pot (homogeneous irrigation, DI) or half of it (heterogeneous irrigation, PRD). To generate a range of whole-pot soil water contents, water was withheld for 24-48 h, but all plants (15 for PRD and 7 for DI) were sampled on the same day (between 10 00 h and 17 00 h) by sequentially excising three fully expanded leaves (Leaves 5-7 numbering from the base of the plant) from each plant. Ψ leaf was measured using a Scholander type pressure chamber (Plant Moisture Systems, Santa Barbara, CA, USA), then an overpressure (0.4 MPa) was applied to the leaf to express xylem sap which was collected in a pre-weighed Eppendorf tube and frozen in liquid nitrogen for later determination of ABA concentration via radioimmunoassay (Quarrie et al., 1988). Two measurements of soil water status of each compartment were made by inserting a theta probe (Model ML2X, Delta-T Devices) into the top of the pot, after which the soil (including roots and 'net pot') was carefully removed from the pot, weighed, and then oven-dried to determine θ. To quantify the fractional soil water uptake from each compartment in subsequent experiments (each comprising four plants per week), two theta probes (Model ML2X, Delta-T Devices) were placed vertically into the top of each soil compartment to measure θ. Then water was supplied to only one soil compartment to implement partial rootzone drying (PRD-Fixed). After 3 d of soil drying and before the start of the photoperiod on the fourth day, the wet and dry soil compartments were alternated (PRD-Alternated) to compare the relationships between soil and plant variables with those plants where the wet and dry soil compartments were fixed. Fully expanded leaves (Leaves 5-7 numbering from the base of the plant) were detached to measure Ψ leaf and to collect xylem sap as described above at the end of the third day of soil drying (PRD-Fixed, Leaf 5) and 2 h (Leaf 6) and 6 h (Leaf 7) after the wet and dry soil compartments were alternated. Preliminary experiments revealed that no more than three leaves could be harvested from the one plant without substantially affecting the relationship between [X-ABA] leaf and soil water content (data not shown). An additional group of well-watered (θ >0.35 g g -1 ) plants were sampled to compare the effects of the PRD-Fixed and PRD-Alternated treatments on whole pot soil water content (θ pot ), Ψ leaf , and [X-ABA] leaf .
In an attempt to understand the variation in [X-ABA] leaf generated in response to fixed and alternate PRD, measured [X-ABA] leaf was compared with the [X-ABA] leaf predicted from three models where: (i) [X-ABA] leaf depended only on whole pot soil water content (as in Fig. 2A) using the relationship: where θ is the mean soil water content, determined by theta probe measurements, derived by averaging both sides of the pot from DI plants.
(ii) [X-ABA]leaf depended only on leaf water potential (as in Fig. 2E) using the relationship: where Ψ leaf is the leaf water potential of an individual leaf from DI plants.
(iii) [X-ABA] leaf depended on θ of each compartment of the split-pot, which affected both [X-ABA] root emanating from, and soil water uptake by, roots in those compartments, according to a simple model (Dodd et al., 2008a): where F dry and F wet represent the fractions of sap flow, and [X-ABA] root-wet and [X-ABA] root-dry represent root xylem ABA concentrations from wet and dry parts of the root system, respectively. Since xylem sap was collected only from leaves in this study, [X-ABA] root-wet and [X-ABA] root-dry were simulated using equation 1, by assuming no change in [X-ABA] in transit between roots and shoots and considering only soil water content of wet or dry sides of the pot, respectively.

Statistical analysis
Within an irrigation treatment (PRD-Fixed or PRD-Alternated) or combined across irrigation treatments, regression analysis determined the significance of relationships between soil and plant variables. Two-way analysis of variance (ANOVA) determined whether the wet or dry root system of PRD plants, or irrigation treatment, altered relationships between soil and plant variables. A change in the sensitivity of the y-variable to the x-variable is given by a significant interaction term (x-variable by treatment or part of the root system).

Results
Before imposing a PRD treatment on a typical 'two root-one shoot' sunflower plant, sap flow through the two hypocotyls ( Fig. 1A) and soil moisture of the two soil compartments (Fig. 1B) were similar. Even though soil moisture declined from ~0.51 to ~0.34 g g -1 in both compartments during the second photoperiod, sap flow through both hypocotyls was maintained. At the beginning of the third photoperiod, one half of the pot was watered to impose PRD, thereby raising soil water content back to ~0.54 g g -1 (Fig. 1B) which maintained sap flow through this part of the root system (Fig. 1A). Soil water content continued to decline in the dry side of the pot (albeit at a reduced rate) and sap flow from that root system started to decline, thus the fraction of total sap flow through the dry and wet root systems decreased and increased, respectively (Fig. 1C). A similar pattern was noted by calculating (from soil moisture readings) the fractions of water uptake from each side of the root system (Fig. 1C). Thus, proportional water uptake from different parts of the rootzone could be determined with sap flow gauges (only in grafted plants) or soil moisture sensors.
To determine the xylem ABA responses of own-rooted, ungrafted tomato plants at different times within soil drying and re-wetting cycles during PRD, it was necessary to collect xylem sap from different leaves of the same plant. Since the sensitivity of leaf xylem ABA concentration ([X-ABA] leaf ) to soil water content (θ) and leaf water potential (Ψ leaf ) was not significantly altered by the sequential removal of three leaves (Fig. 2), each plant could be sampled three times during PRD cycles.
Withholding irrigation from the entire rootzone, or half of it, produced an exponential relationship between [X-ABA] leaf and whole-pot soil water content (θ pot ) ( Fig. 2A, B). Homogeneous or heterogeneous irrigation induced a similar response, as there was no difference in the slope of the relationship according to whether DI or PRD was applied (P=0.71). However, changes in θ pot produced by DI or PRD were not correlated with Ψ leaf (Fig. 2C, D). By contrast, the spatial distribution of irrigation altered the relationship between Ψ leaf and [X-ABA] leaf (a difference in the slope of relationship: P=0.028) since [X-ABA] leaf correlated with Ψ leaf only in homogeneously irrigated plants (Fig. 2E). The ability of these relationships between [X-ABA] leaf and θ pot ( Fig. 2A; equation 1) and Ψ leaf (Fig. 2E; equation 2) to predict [X-ABA] leaf was tested in subsequent experiments that exposed own-rooted, ungrafted tomato plants to PRD for 3 d and sampled leaves on three occasions (before alternation as PRD-Fixed plants, and 2 h and 6 h after alternating the wet and dry parts of the rootzone as PRD-Alternated plants).
In a typical plant, frequent watering maintained θ of the wet (left) side of the pot greater than 0.3 g g -1 , but θ of the dry (right) side of the pot rapidly decreased below 0.2 g g -1 once irrigation was withheld (Fig. 3A), such that water uptake from this compartment virtually ceased on the second day (Fig. 3B). At the end of the third day of drying, Ψ leaf and [X-ABA] leaf of this plant were -0.58 MPa and 160 nM, respectively (Fig. 3D), while θ of the wet and dry compartments were 0.35 g g -1 and 0.16 g g -1 (Fig. 3A) and the fractions of soil water uptake from these compartments were 0.98 and 0.02, respectively (Fig. 3C).
Similar measurements were made at this time in the PRD cycle in 11 other plants, when the average θ of the wet and dry sides were 0.30 g g -1 and 0.14 g g -1 , respectively, while the fractions of soil water uptake from these compartments in these PRD-Fixed plants were 0.98 and 0.02, respectively (Table 1). Across all plants in this study, average (mean of 12 plants) Ψ leaf and [X-ABA] leaf of PRD-Fixed plants were -0.60 MPa and 273 nM, respectively (Table 2). At this time, Ψ leaf was negatively and [X-ABA] leaf positively correlated with θ of the irrigated compartment (Fig. 4A, C). Predicting [X-ABA] leaf based on either whole pot θ (equation 1) or Ψ leaf (Equation 2) overestimated its value by 75% and 54%, respectively (Table 3). Multiplying the fraction of soil water uptake from each compartment by the predicted [X-ABA] root based on its θ, and summing these terms (Equation 3), underestimated [X-ABA] leaf by only 10% (Table 3).
Returning to the typical plant, the dry (right) side of the pot was re-watered at the beginning of the fourth day and irrigation withheld from the previously irrigated (left) side (Fig. 3A). After alternating irrigation, soil water uptake increased progressively in the re-watered (right) side, while it decreased in the newly-drying (left) side (Fig. 3B). Two hours after alternation, Ψ leaf and [X-ABA] leaf of this plant were -0.74 MPa and 209 nM, respectively (Fig. 3D), while θ of the wet and dry sides were 0.40 g g -1 and 0.29 g g -1 (Fig. 3A) and the fractions of soil water uptake from these sides were 0.61 and 0.39, respectively (Fig. 3C). Relationships between leaf xylem ABA concentration ([X-ABA] leaf ) and whole-pot soil-water content (A, B), leaf water potential and wholepot soil water content (C, D), and [X-ABA] leaf and leaf water potential (E, F) in tomato plants exposed to homogeneous (deficit irrigation; A, C, E) and heterogeneous (fixed PRD; B, D, F) irrigation. Each point represents an individual leaf (three leaves were sampled sequentially from each plant), with regression lines fitted where significant (P <0.05).
Similar measurements were made at this time in the PRD cycle in 11 other plants, when the average θ of the newly irrigated (right) and now drying (left) sides were 0.32 g g -1 and 0.22 g g -1 , respectively, while the fractions of soil water uptake from these compartments in these PRD-Alternated plants were 0.72 and 0.28, respectively (Table 1). Across all plants in this study, average Ψ leaf and [X-ABA] leaf of PRD-Alternated   plants were -0.59 MPa and 270 nM, respectively (Table 2). At this time, variations in Ψ leaf and [X-ABA] leaf were correlated with θ of both sides of the pot (Fig. 4). Predicting [X-ABA] leaf based on either whole pot θ (equation 1) or Ψ leaf (equation 2) overestimated its value by 12% and 50%, respectively (Table 3). Multiplying the fraction of soil water uptake from each compartment by the predicted [X-ABA] root based on its θ, and summing these terms (equation 3), underestimated [X-ABA] leaf by only 3% (Table 3). Returning again to the typical plant 6 h after alternation, soil water uptake from the drying (left) side of the pot continued to decrease (Fig. 3B). At this time, Ψ leaf and [X-ABA] leaf of this plant were -0.64 MPa and 180 nM, respectively (Fig. 3D), while θ of the wet and dry sides were 0.40 g g -1 and 0.22 g g -1 (Fig. 3A) and the fractions of soil water uptake from these sides were 0.80 and 0.20, respectively (Fig. 3C).
Similar measurements were made at this time in the PRD cycle in 11 other plants, when the average θ of the newly irrigated (right) and now drying (left) sides were 0.31 g g -1 and 0.18 g g -1 , respectively, while the fractions of soil water uptake from these compartments in these PRD-Alternated plants were 0.90 and 0.10, respectively (Table 1). Across all plants in this study, average Ψ leaf and [X-ABA] leaf were -0.63 MPa and 225 nM, respectively (Table 2). At this time, variations of [X-ABA] leaf and Ψ leaf were only correlated with changes of θ from the newly-irrigated (right) side (Fig. 4B, D). Predicting [X-ABA] leaf based on either whole pot θ (equation 1) or Ψ leaf (equation 2) overestimated its value by 1.6-fold (~60%) and Fig. 4. Relationships between leaf xylem ABA concentration ([X-ABA] leaf ) (A, B) and leaf water potential (C, D) and soil water content from both sides (left and right) of the root system of tomato plants. In PRD-Fixed plants, left and right sides corresponded to wet and dry parts of the root system while, in PRD-Alternated plants, they corresponded to the newly-drying and just re-watered parts of the root system. In PRD-Alternated plants measurements were taken 2 h and 6 h after irrigation was alternated. Each point represents a single leaf and regression lines were fitted where significant (P <0.05). NC means not correlated.  2.4-fold, respectively (Table 3). Multiplying the fraction of soil water uptake from each compartment by the predicted [X-ABA] root based on its θ, and summing these terms (equation 3), overestimated [X-ABA] leaf by only 31% (Table 3). Therefore at all stages of the PRD cycle, equation 3 best estimated [X-ABA] leaf.

Table 3. The ability of different models to predict leaf xylem ABA concentration ([X-ABA] leaf ) of fixed and alternated PRD tomato plants
Applying PRD-Fixed and PRD-Alternated irrigation decreased θ pot compared with well-watered plants, with θ pot lower when heterogeneous soil moisture conditions were clearly established (Table 2). In both PRD-Fixed and PRD-Alternated plants, Ψ leaf decreased and [X-ABA] leaf increased similarly compared with well-watered plants (Table 2). It was also of interest to determine whether PRD alternation affected the relationships between [X-ABA] leaf , θ pot , and Ψ leaf . When θ pot decreased below 0.26 g g -1 , Ψ leaf declined similarly in PRD-Fixed and PRD-Alternated plants (Fig. 5B). When θ pot exceeded 0.26 g g -1 , PRD-Alternated plants had a lower Ψ leaf 6 h after alternation of the wet and dry sides than after 2 h. [X-ABA] leaf increased as Ψ leaf declined (Fig. 5C) or as θ pot (Fig. 5A) increased, but the timing (2 h and 6 h after alternation) or occurrence (PRD-Fixed versus PRD-Alternated) of PRD alternation did not affect the sensitivity of ABA signalling. Generally, all PRD plants had similar relationships between [X-ABA] leaf , θ pot , and Ψ leaf independently of alternating the wet and dry sides of the pot.

Discussion
The irrigation technique of partial rootzone drying was conceived as a field adaptation of laboratory split-root soil drying experiments, to enhance root-to-shoot chemical signalling to improve crop water use efficiency by causing partial stomatal closure and decreasing excessive vegetative vigour (Dry et al., 1996;Kang and Zhang, 2004). While PRD can outyield conventional deficit irrigation (DI) where the entire rootzone is irrigated (Dodd, 2009), understanding the physiological mechanisms underpinning this technique has received comparatively little attention. Although many authors have postulated a role for root-sourced ABA in causing these physiological responses, there is considerable variability in the relative response of [X-ABA] leaf when comparing PRD and DI plants (Dodd, 2007;Wang et al., 2012), perhaps related to the timing of measurements during drying/re-wetting cycles (Dodd et al., 2006) and/or total soil water availability (Romero et al., 2012). Although models have demonstrated the importance of sap flow from roots in drying soil in predicting [X-ABA] leaf of PRD plants (Dodd et al., 2008a(Dodd et al., , b, 2010, these require [X-ABA] root as an input variable, making them difficult to apply to field situations. By collecting xylem sap only from detached leaves and measuring water uptake by different parts of the root system, this study developed a model, suitable for field application, that better predicted [X-ABA] leaf of PRD plants (throughout drying and rewetting cycles) than assuming that [X-ABA] leaf was simply related either to total soil water availability or to leaf water potential.
In previous studies, [X-ABA] leaf was predicted by quantifying both the fractions of sap flow, and root xylem ABA concentrations, from different parts of the root system using specially constructed 'two root-one shoot' grafted plants (Dodd et al., 2008a(Dodd et al., , b, 2010, and assuming [X-ABA] was not altered in transit from roots to shoot. However, a major limitation of applying this model to field-grown, own-rooted plants exposed to PRD is the difficulty of determining the fraction of sap flow from each side of the root system. Continuous soil moisture monitoring can infer plant water uptake (Puertolas et al., 2013), which was correlated with sap flow from different parts of the root system in 'two root-one shoot' grafted plants (Fig. 1). After withholding water from part of the rootzone, root water uptake and sap flow declined similarly as the soil dried. Thus soil-moisture sensors can determine when sap flow from the drying side of the root system ceases, although vertical gradients in soil moisture (and root water uptake) may complicate interpretation (Puertolas et al., 2013).
Another complexity of modelling [X-ABA] leaf during PRD is the periodic alternation of wet and dry sides, requiring repeated measurements from single plants to test the adequacy of models at different stages of the drying cycles. To monitor the effects of soil drying/re-wetting cycles on the [X-ABA] leaf of individual plants, three leaves were sampled (Fig. 3D). Preliminary experiments established that relationships between [X-ABA] leaf and both plant (leaf water potential) and soil water status (Fig. 2) did not vary with the number of leaves (per plant) sampled when plants were allowed to dry the soil in a split pot uniformly. Nevertheless, heterogeneous soil moisture increased the variation of [X-ABA] leaf (from 60-1150 nM in PRD plants and from 215-645 nM in DI plants across a similar θ and Ψ leaf range; Fig. 2) as observed previously in tomato (Dodd, 2007), probably due to spatial differences in root water uptake when θ of the dry side of the pot decreased below 0.3 g g -1 (Fig. 3). An additional contributing factor may be xylem sectoriality in this species (Zanne et al., 2006), with specific roots supplying water (and possibly chemical signals) to specific leaves in the shoot, but the leaf sampled did not affect the relationship between [X-ABA] leaf and θ in PRD plants (Fig. 2B), probably because roots were stochastically distributed between the two soil compartments of the split pot. Since roots of PRD plants are exposed to a greater range of soil moisture at a given θ pot than DI plants, different water uptake fractions from each part of the root system can affect [X-ABA] leaf .
Based on previous studies with PRD plants, the optimal moment to alternate irrigation is when sap flow from the dry rootzone significantly decreases; thus limiting ABA export from roots to shoots (Dodd et al., 2008a). When soil moisture was clearly heterogeneous (6 h after alternation and during fixed PRD), water uptake from the dry rootzone practically ceased ( Fig. 3C; Table 1). Under these conditions, variations of Ψ leaf and [X-ABA] leaf were mostly related to changes in the θ of the irrigated rootzone (Fig. 4), as previously modelled (Dodd et al., 2008a, b). Furthermore, soil water status of this compartment is important to maintain high plant water status during PRD (Wang et al., 2012). At these times of the PRD cycle, [X-ABA] leaf was best predicted with a model that included the water-uptake fractions from each part of the root system (Table 3). As previously observed (Dodd et al., 2008b), predicting [X-ABA] leaf based on the predetermined relationship between [X-ABA] leaf and whole pot θ (average of dry and wet sides of the pot, equation 1) significantly overestimated [X-ABA] leaf (Table 3). Similarly, predicting [X-ABA] leaf based on Ψ leaf (equation 2) also substantially overestimated its concentration (Table 3). Thus accounting for soil water uptake from different parts of the root system best predicted [X-ABA] leaf once water uptake from the dry rootzone had declined to 10% or less (Table 1), but these conditions may occur for limited periods of time in field-grown plants due to the availability of soil moisture at depth. Two hours after re-watering the previously dry rootzone, water uptake from both soil compartments significantly contributed to the total sap flow ( Fig. 3; Table 1) and thus [X-ABA] leaf . In this case, whole pot θ alone (equation 1) and accounting for water uptake from different parts of the rootzone (equation 3) showed a statistically similar ability to predict [X-ABA] leaf (Table 3). By contrast, in an experiment where the fraction of soil water uptake from the drying compartment remained greater than 25% for four days of the five days of a PRD cycle, relating [X-ABA] leaf to total soil water availability (and ignoring relative water uptake from the two compartments) was the better performing model (Liu et al., 2008). However, the reliability of this conclusion depended on the range of soil water availability considered (Dodd et al., 2008b), demonstrating that preliminary experiments to parameterize any model must occur over a similar range of soil water contents as tested experimentally (cf. Fig. 2A, C and Fig. 5A, B). An alternative view, that [X-ABA] leaf can be predicted from Ψ leaf, has received comparatively little attention since the relationship between [X-ABA] leaf and Ψ leaf varied according to whether plants received PRD or DI (Dodd, 2007;Dodd et al., 2008a;cf. Fig. 2E and 2F here) and whether PRD was alternated or fixed (Dodd et al., 2006) and sometimes there was no significant relationship between these variables. Accordingly, predicting [X-ABA] leaf from Ψ leaf (equation 2) systematically overestimated [X-ABA] leaf at all stages of the PRD cycle (Table 3), suggesting again that accounting for soil water uptake from different parts of the root system best predicted [X-ABA] leaf , when soil moisture heterogeneity existed.
After alternating the wet and dry soil compartments of the pot, [X-ABA] leaf was similar to fixed PRD plants (Table 2), contrary to previous observations. Greater [X-ABA] leaf of alternated plants compared with fixed plants (Dodd et al., 2006-tomato grown in the same substrate as in this study) was suggested to result from mobilizing root-sourced ABA (that had accumulated during soil drying) to the transpiration stream following re-watering the originally dry column. Two hours after irrigation alternation, variations of [X-ABA] leaf were related to changes in soil water status of both sides of the pot (Fig. 4A, B). At this moment, water uptake from the previously dry side of the pot was re-established (Table 1), yet [X-ABA] leaf did not show any unexpected increase due to 'extra ABA' transported from the previously dried roots. Reconciling these apparently contradictory observations requires more detailed information on the sensitivity of root ABA accumulation in response to soil drying. However, when there were more pronounced vertical gradients in soil moisture [as probably occurred in Dodd et al. (2006) where plants were grown in 30 cm high soil columns], there was less pronounced root ABA accumulation when soil moisture was 0.13-0.25 g g -1 compared with a more homogeneous soilmoisture distribution (Puertolas et al., 2013) which characterizes the soil environment in the 13 cm high pots used here. Instead, differences in the pot surface area-to-volume ratio in the soil compartments in the two studies (cf. Dodd et al., 2006 versus this study) will expose different numbers of roots to different environmental (and soil-moisture) conditions at the edge of the pot. Clearly, more attention must be given to measuring root ABA accumulation in the field in attempting to explain why PRD alternation only sometimes stimulates xylem ABA concentration (cf. Pérez-Pérez et al., 2012versus Romero et al., 2012. In summary, soil moisture sensors accurately estimated water-uptake fractions from different parts of the rootzone in plants grown in split pots. This information improved the prediction of [X-ABA] leaf in plants exposed to fixed and alternate PRD, compared with prediction based on total soilwater availability or leaf water potential alone. Further work is required to establish why PRD alternation did not enhance [X-ABA] leaf (Fig. 5), contrary to previous work, and whether any changes in root-to-shoot ABA signalling are related to crop yields.